From 59344336473ecfe6075be8abe50b0cc65563d0fa Mon Sep 17 00:00:00 2001 From: Holly Grimm Date: Fri, 17 Jul 2026 14:08:53 -0600 Subject: [PATCH] fix(courses): reconnect lecture videos; drop junk part ids; move transcripts MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Social Sciences: ActInf_Basics_Lecture and Introduction_Lecture had empty parts while their real uploads sat in Other/ — parts filled, transcripts moved in, stubs removed. Chris Fields course gains its 4 Lecture/Discussion videos as parts with transcripts folded, and loses the junk 'transcripts' part id (boost_gs062 likewise dropped from GuestStream_062). 2023 interval transcripts move from their Other/ duplicates to the canonical items. Missing-transcript worklist 22->18; strict-manifest stray ids 7->5. Co-Authored-By: Claude Fable 5 Claude-Session: https://claude.ai/code/session_015NjFiMWkgrVy4foGF1Zbax --- INDEX.json | 88 +- INDEX.md | 26 +- .../First_Interval}/transcript.txt | 0 .../Second_Interval}/transcript.txt | 0 .../ActInf_Basics_Lecture/metadata.json | 167 +- .../ActInf_Basics_Lecture}/transcript.txt | 0 .../Introduction_Lecture/metadata.json | 114 +- .../Introduction_Lecture}/transcript.txt | 0 .../metadata.json | 31 +- .../transcript.txt | 2788 ++++++++++++++++- .../GuestStream/GuestStream_062/metadata.json | 5 - .../Other/BNLnbOFdgc0/README.md | 5 - .../Other/BNLnbOFdgc0/metadata.json | 177 -- .../Other/JkORFoyk8o8/README.md | 5 - .../Other/JkORFoyk8o8/metadata.json | 124 - .../Other/TOZp_XNYijQ/README.md | 5 - .../Other/TOZp_XNYijQ/metadata.json | 125 - .../Other/TOZp_XNYijQ/transcript.txt | 1141 ------- .../Other/WkWIqpxWRM4/README.md | 5 - .../Other/WkWIqpxWRM4/metadata.json | 137 - .../Other/WkWIqpxWRM4/transcript.txt | 453 --- .../Other/hSj7eOYuZe4/README.md | 5 - .../Other/hSj7eOYuZe4/metadata.json | 124 - .../Other/hSj7eOYuZe4/transcript.txt | 639 ---- .../Other/rOoTkgLI1Z8/README.md | 5 - .../Other/rOoTkgLI1Z8/metadata.json | 135 - .../Other/rOoTkgLI1Z8/transcript.txt | 541 ---- 27 files changed, 3120 insertions(+), 3725 deletions(-) rename data/video/activeinferenceinstitute/{Other/rIemcswLfGg => Applied Active Inference Symposium/2023 Ecosystem Symposium/First_Interval}/transcript.txt (100%) rename data/video/activeinferenceinstitute/{Other/PVeyvHSAwmk => Applied Active Inference Symposium/2023 Ecosystem Symposium/Second_Interval}/transcript.txt (100%) rename data/video/activeinferenceinstitute/{Other/BNLnbOFdgc0 => Courses/ActiveInferenceForTheSocialSciences/ActInf_Basics_Lecture}/transcript.txt (100%) rename data/video/activeinferenceinstitute/{Other/JkORFoyk8o8 => Courses/ActiveInferenceForTheSocialSciences/Introduction_Lecture}/transcript.txt (100%) delete mode 100644 data/video/activeinferenceinstitute/Other/BNLnbOFdgc0/README.md delete mode 100644 data/video/activeinferenceinstitute/Other/BNLnbOFdgc0/metadata.json delete mode 100644 data/video/activeinferenceinstitute/Other/JkORFoyk8o8/README.md delete mode 100644 data/video/activeinferenceinstitute/Other/JkORFoyk8o8/metadata.json delete mode 100644 data/video/activeinferenceinstitute/Other/TOZp_XNYijQ/README.md delete mode 100644 data/video/activeinferenceinstitute/Other/TOZp_XNYijQ/metadata.json delete mode 100644 data/video/activeinferenceinstitute/Other/TOZp_XNYijQ/transcript.txt delete mode 100644 data/video/activeinferenceinstitute/Other/WkWIqpxWRM4/README.md delete mode 100644 data/video/activeinferenceinstitute/Other/WkWIqpxWRM4/metadata.json delete mode 100644 data/video/activeinferenceinstitute/Other/WkWIqpxWRM4/transcript.txt delete mode 100644 data/video/activeinferenceinstitute/Other/hSj7eOYuZe4/README.md delete mode 100644 data/video/activeinferenceinstitute/Other/hSj7eOYuZe4/metadata.json delete mode 100644 data/video/activeinferenceinstitute/Other/hSj7eOYuZe4/transcript.txt delete mode 100644 data/video/activeinferenceinstitute/Other/rOoTkgLI1Z8/README.md delete mode 100644 data/video/activeinferenceinstitute/Other/rOoTkgLI1Z8/metadata.json delete mode 100644 data/video/activeinferenceinstitute/Other/rOoTkgLI1Z8/transcript.txt diff --git a/INDEX.json b/INDEX.json index ef603b73f..be3ed4d65 100644 --- a/INDEX.json +++ b/INDEX.json @@ -1,7 +1,7 @@ { - "count": 579, - "videos": 746, - "unique_videos": 742, + "count": 573, + "videos": 744, + "unique_videos": 740, "items": [ { "path": "data/video/activeinferenceinstitute/ActiveInferAntStream/ActiveInferAntStream_001", @@ -190,7 +190,7 @@ "parts": [ "rIemcswLfGg" ], - "has_transcript": false + "has_transcript": true }, { "path": "data/video/activeinferenceinstitute/Applied Active Inference Symposium/2023 Ecosystem Symposium/Second_Interval", @@ -199,7 +199,7 @@ "parts": [ "PVeyvHSAwmk" ], - "has_transcript": false + "has_transcript": true }, { "path": "data/video/activeinferenceinstitute/Applied Active Inference Symposium/2024/part 1", @@ -345,8 +345,10 @@ "path": "data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/ActInf_Basics_Lecture", "series": "Courses", "item": "ActInf_Basics_Lecture", - "parts": [], - "has_transcript": false + "parts": [ + "BNLnbOFdgc0" + ], + "has_transcript": true }, { "path": "data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/CollectiveBehavior_Discussion", @@ -379,8 +381,10 @@ "path": "data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/Introduction_Lecture", "series": "Courses", "item": "Introduction_Lecture", - "parts": [], - "has_transcript": false + "parts": [ + "JkORFoyk8o8" + ], + "has_transcript": true }, { "path": "data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/NormsScripts_Discussion", @@ -439,7 +443,6 @@ "series": "Courses", "item": "PhysicsAsInformationProcessing_ChrisFields", "parts": [ - "transcripts", "hOto96Q8sYQ", "jWEcQVcgleM", "_NfqCpsZPeM", @@ -449,7 +452,11 @@ "Il9sBNAia7g", "bSPQ3BkqTh4", "cxJuA012HvM", - "SV0glS6stuA" + "SV0glS6stuA", + "WkWIqpxWRM4", + "TOZp_XNYijQ", + "rOoTkgLI1Z8", + "hSj7eOYuZe4" ], "has_transcript": true }, @@ -1010,7 +1017,6 @@ "series": "GuestStream", "item": "GuestStream_062", "parts": [ - "boost_gs062", "ciFMH8svne4" ], "has_transcript": true @@ -3010,15 +3016,6 @@ ], "has_transcript": true }, - { - "path": "data/video/activeinferenceinstitute/Other/BNLnbOFdgc0", - "series": "Other", - "item": "BNLnbOFdgc0", - "parts": [ - "BNLnbOFdgc0" - ], - "has_transcript": true - }, { "path": "data/video/activeinferenceinstitute/Other/C1Uqxg7mDCo", "series": "Other", @@ -3073,15 +3070,6 @@ ], "has_transcript": true }, - { - "path": "data/video/activeinferenceinstitute/Other/JkORFoyk8o8", - "series": "Other", - "item": "JkORFoyk8o8", - "parts": [ - "JkORFoyk8o8" - ], - "has_transcript": true - }, { "path": "data/video/activeinferenceinstitute/Other/KR9KCKQ-KJo", "series": "Other", @@ -3134,7 +3122,7 @@ "parts": [ "PVeyvHSAwmk" ], - "has_transcript": true, + "has_transcript": false, "duplicate_of": "Applied Active Inference Symposium/2023 Ecosystem Symposium/Second_Interval" }, { @@ -3155,15 +3143,6 @@ ], "has_transcript": true }, - { - "path": "data/video/activeinferenceinstitute/Other/TOZp_XNYijQ", - "series": "Other", - "item": "TOZp_XNYijQ", - "parts": [ - "TOZp_XNYijQ" - ], - "has_transcript": true - }, { "path": "data/video/activeinferenceinstitute/Other/TaFwI2zr_lE", "series": "Other", @@ -3200,15 +3179,6 @@ ], "has_transcript": true }, - { - "path": "data/video/activeinferenceinstitute/Other/WkWIqpxWRM4", - "series": "Other", - "item": "WkWIqpxWRM4", - "parts": [ - "WkWIqpxWRM4" - ], - "has_transcript": true - }, { "path": "data/video/activeinferenceinstitute/Other/WxTOeCnQNmY", "series": "Other", @@ -3327,15 +3297,6 @@ ], "has_transcript": true }, - { - "path": "data/video/activeinferenceinstitute/Other/hSj7eOYuZe4", - "series": "Other", - "item": "hSj7eOYuZe4", - "parts": [ - "hSj7eOYuZe4" - ], - "has_transcript": true - }, { "path": "data/video/activeinferenceinstitute/Other/hZBkiePQtvc", "series": "Other", @@ -3425,18 +3386,9 @@ "parts": [ "rIemcswLfGg" ], - "has_transcript": true, + "has_transcript": false, "duplicate_of": "Applied Active Inference Symposium/2023 Ecosystem Symposium/First_Interval" }, - { - "path": "data/video/activeinferenceinstitute/Other/rOoTkgLI1Z8", - "series": "Other", - "item": "rOoTkgLI1Z8", - "parts": [ - "rOoTkgLI1Z8" - ], - "has_transcript": true - }, { "path": "data/video/activeinferenceinstitute/Other/rs6EjQM6EdI", "series": "Other", diff --git a/INDEX.md b/INDEX.md index ddf76f258..f597d6f01 100644 --- a/INDEX.md +++ b/INDEX.md @@ -1,6 +1,6 @@ # ActiveInferenceJournal — Index -579 items · 746 video records · 19 series · 742 unique video IDs. Root: `data/video/activeinferenceinstitute/`. Machine index: `INDEX.json`. +573 items · 744 video records · 19 series · 740 unique video IDs. Root: `data/video/activeinferenceinstitute/`. Machine index: `INDEX.json`. ## ActiveInferAntStream (17) - ✓ [`ActiveInferAntStream_001`](data/video/activeinferenceinstitute/ActiveInferAntStream/ActiveInferAntStream_001/) — 2 video record(s) @@ -24,8 +24,8 @@ ## Applied Active Inference Symposium (10) - ✓ [`2021 Symposium with Karl Friston`](data/video/activeinferenceinstitute/Applied%20Active%20Inference%20Symposium/2021%20Symposium%20with%20Karl%20Friston/) — 3 video record(s) - ✓ [`2022 Symposium on Robotics`](data/video/activeinferenceinstitute/Applied%20Active%20Inference%20Symposium/2022%20Symposium%20on%20Robotics/) — 2 video record(s) -- · [`First_Interval`](data/video/activeinferenceinstitute/Applied%20Active%20Inference%20Symposium/2023%20Ecosystem%20Symposium/First_Interval/) — 1 video record(s) -- · [`Second_Interval`](data/video/activeinferenceinstitute/Applied%20Active%20Inference%20Symposium/2023%20Ecosystem%20Symposium/Second_Interval/) — 1 video record(s) +- ✓ [`First_Interval`](data/video/activeinferenceinstitute/Applied%20Active%20Inference%20Symposium/2023%20Ecosystem%20Symposium/First_Interval/) — 1 video record(s) +- ✓ [`Second_Interval`](data/video/activeinferenceinstitute/Applied%20Active%20Inference%20Symposium/2023%20Ecosystem%20Symposium/Second_Interval/) — 1 video record(s) - ✓ [`part 1`](data/video/activeinferenceinstitute/Applied%20Active%20Inference%20Symposium/2024/part%201/) — 1 video record(s) - ✓ [`part 2`](data/video/activeinferenceinstitute/Applied%20Active%20Inference%20Symposium/2024/part%202/) — 1 video record(s) - ✓ [`part 3`](data/video/activeinferenceinstitute/Applied%20Active%20Inference%20Symposium/2024/part%203/) — 1 video record(s) @@ -45,18 +45,18 @@ ## Courses (13) - ✓ [`ActInf_Basics_Discussion`](data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/ActInf_Basics_Discussion/) — 1 video record(s) -- · [`ActInf_Basics_Lecture`](data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/ActInf_Basics_Lecture/) — 0 video record(s) +- ✓ [`ActInf_Basics_Lecture`](data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/ActInf_Basics_Lecture/) — 1 video record(s) - ✓ [`CollectiveBehavior_Discussion`](data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/CollectiveBehavior_Discussion/) — 1 video record(s) - ✓ [`CollectiveBehavior_Lecture`](data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/CollectiveBehavior_Lecture/) — 1 video record(s) - ✓ [`Conclusion_Lecture`](data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/Conclusion_Lecture/) — 1 video record(s) -- · [`Introduction_Lecture`](data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/Introduction_Lecture/) — 0 video record(s) +- ✓ [`Introduction_Lecture`](data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/Introduction_Lecture/) — 1 video record(s) - · [`NormsScripts_Discussion`](data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/NormsScripts_Discussion/) — 0 video record(s) - ✓ [`NormsScripts_Lecture`](data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/NormsScripts_Lecture/) — 1 video record(s) - ✓ [`SemioticsSemantics_Discussion`](data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/SemioticsSemantics_Discussion/) — 1 video record(s) - ✓ [`SemioticsSemantics_Lecture`](data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/SemioticsSemantics_Lecture/) — 1 video record(s) - ✓ [`SocialConstraints_Discussion`](data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/SocialConstraints_Discussion/) — 1 video record(s) - ✓ [`SocialConstraints_Lecture`](data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/SocialConstraints_Lecture/) — 1 video record(s) -- ✓ [`PhysicsAsInformationProcessing_ChrisFields`](data/video/activeinferenceinstitute/Courses/PhysicsAsInformationProcessing_ChrisFields/) — 11 video record(s) +- ✓ [`PhysicsAsInformationProcessing_ChrisFields`](data/video/activeinferenceinstitute/Courses/PhysicsAsInformationProcessing_ChrisFields/) — 14 video record(s) ## GuestStream (127) - ✓ [`GuestStream_001`](data/video/activeinferenceinstitute/GuestStream/GuestStream_001/) — 1 video record(s) @@ -120,7 +120,7 @@ - ✓ [`GuestStream_059`](data/video/activeinferenceinstitute/GuestStream/GuestStream_059/) — 1 video record(s) - ✓ [`GuestStream_060`](data/video/activeinferenceinstitute/GuestStream/GuestStream_060/) — 1 video record(s) - ✓ [`GuestStream_061`](data/video/activeinferenceinstitute/GuestStream/GuestStream_061/) — 1 video record(s) -- ✓ [`GuestStream_062`](data/video/activeinferenceinstitute/GuestStream/GuestStream_062/) — 2 video record(s) +- ✓ [`GuestStream_062`](data/video/activeinferenceinstitute/GuestStream/GuestStream_062/) — 1 video record(s) - ✓ [`GuestStream_063`](data/video/activeinferenceinstitute/GuestStream/GuestStream_063/) — 1 video record(s) - ✓ [`GuestStream_064`](data/video/activeinferenceinstitute/GuestStream/GuestStream_064/) — 1 video record(s) - ✓ [`GuestStream_065`](data/video/activeinferenceinstitute/GuestStream/GuestStream_065/) — 1 video record(s) @@ -336,7 +336,7 @@ - ✓ [`OrgStream_008`](data/video/activeinferenceinstitute/OrgStream/OrgStream_008/) — 1 video record(s) - ✓ [`OrgStream_009`](data/video/activeinferenceinstitute/OrgStream/OrgStream_009/) — 1 video record(s) -## Other (59) +## Other (53) - ✓ [`-39CESDAfLM`](data/video/activeinferenceinstitute/Other/-39CESDAfLM/) — 1 video record(s) - ✓ [`0A36OXIDLhQ`](data/video/activeinferenceinstitute/Other/0A36OXIDLhQ/) — 1 video record(s) - ✓ [`1ZJBJGzOn0g`](data/video/activeinferenceinstitute/Other/1ZJBJGzOn0g/) — 1 video record(s) @@ -345,28 +345,24 @@ - ✓ [`8X9K_gZWhqs`](data/video/activeinferenceinstitute/Other/8X9K_gZWhqs/) — 1 video record(s) - ✓ [`9v22c92LWHQ`](data/video/activeinferenceinstitute/Other/9v22c92LWHQ/) — 1 video record(s) - ✓ [`B64lXpQAOus`](data/video/activeinferenceinstitute/Other/B64lXpQAOus/) — 1 video record(s) -- ✓ [`BNLnbOFdgc0`](data/video/activeinferenceinstitute/Other/BNLnbOFdgc0/) — 1 video record(s) - ✓ [`C1Uqxg7mDCo`](data/video/activeinferenceinstitute/Other/C1Uqxg7mDCo/) — 1 video record(s) - ✓ [`E3xAZAs7YeA`](data/video/activeinferenceinstitute/Other/E3xAZAs7YeA/) — 1 video record(s) - ✓ [`Ei8LHRlmbzI`](data/video/activeinferenceinstitute/Other/Ei8LHRlmbzI/) — 1 video record(s) - ✓ [`EkYlOX-f9sk`](data/video/activeinferenceinstitute/Other/EkYlOX-f9sk/) — 1 video record(s) - ✓ [`FlZa_odQymU`](data/video/activeinferenceinstitute/Other/FlZa_odQymU/) — 1 video record(s) - ✓ [`G_Jksni5o8I`](data/video/activeinferenceinstitute/Other/G_Jksni5o8I/) — 1 video record(s) -- ✓ [`JkORFoyk8o8`](data/video/activeinferenceinstitute/Other/JkORFoyk8o8/) — 1 video record(s) - ✓ [`KR9KCKQ-KJo`](data/video/activeinferenceinstitute/Other/KR9KCKQ-KJo/) — 1 video record(s) - ✓ [`LI_ZIXys1xU`](data/video/activeinferenceinstitute/Other/LI_ZIXys1xU/) — 1 video record(s) - ✓ [`MRTULbP1ZKs`](data/video/activeinferenceinstitute/Other/MRTULbP1ZKs/) — 1 video record(s) - ✓ [`OCCIDeeKOMA`](data/video/activeinferenceinstitute/Other/OCCIDeeKOMA/) — 1 video record(s) - ✓ [`P3h51-J4g8M`](data/video/activeinferenceinstitute/Other/P3h51-J4g8M/) — 1 video record(s) -- ✓ [`PVeyvHSAwmk`](data/video/activeinferenceinstitute/Other/PVeyvHSAwmk/) — 1 video record(s) — duplicate of `Applied Active Inference Symposium/2023 Ecosystem Symposium/Second_Interval` +- · [`PVeyvHSAwmk`](data/video/activeinferenceinstitute/Other/PVeyvHSAwmk/) — 1 video record(s) — duplicate of `Applied Active Inference Symposium/2023 Ecosystem Symposium/Second_Interval` - ✓ [`QmVuId8SY8k`](data/video/activeinferenceinstitute/Other/QmVuId8SY8k/) — 1 video record(s) - ✓ [`RDDfKm4yQIo`](data/video/activeinferenceinstitute/Other/RDDfKm4yQIo/) — 1 video record(s) -- ✓ [`TOZp_XNYijQ`](data/video/activeinferenceinstitute/Other/TOZp_XNYijQ/) — 1 video record(s) - ✓ [`TaFwI2zr_lE`](data/video/activeinferenceinstitute/Other/TaFwI2zr_lE/) — 1 video record(s) - ✓ [`TsQQX3ZdEmA`](data/video/activeinferenceinstitute/Other/TsQQX3ZdEmA/) — 1 video record(s) - ✓ [`U6Edc13EQ_k`](data/video/activeinferenceinstitute/Other/U6Edc13EQ_k/) — 1 video record(s) - ✓ [`UYD8CR_Xorg`](data/video/activeinferenceinstitute/Other/UYD8CR_Xorg/) — 1 video record(s) -- ✓ [`WkWIqpxWRM4`](data/video/activeinferenceinstitute/Other/WkWIqpxWRM4/) — 1 video record(s) - ✓ [`WxTOeCnQNmY`](data/video/activeinferenceinstitute/Other/WxTOeCnQNmY/) — 1 video record(s) - ✓ [`XcoPRGZRO28`](data/video/activeinferenceinstitute/Other/XcoPRGZRO28/) — 1 video record(s) - ✓ [`YxrIbWTjSp4`](data/video/activeinferenceinstitute/Other/YxrIbWTjSp4/) — 1 video record(s) @@ -380,7 +376,6 @@ - ✓ [`gS-qhMNFm84`](data/video/activeinferenceinstitute/Other/gS-qhMNFm84/) — 1 video record(s) - ✓ [`geITY8aJJEA`](data/video/activeinferenceinstitute/Other/geITY8aJJEA/) — 1 video record(s) - ✓ [`hCIxFkdu_-Y`](data/video/activeinferenceinstitute/Other/hCIxFkdu_-Y/) — 1 video record(s) -- ✓ [`hSj7eOYuZe4`](data/video/activeinferenceinstitute/Other/hSj7eOYuZe4/) — 1 video record(s) - ✓ [`hZBkiePQtvc`](data/video/activeinferenceinstitute/Other/hZBkiePQtvc/) — 1 video record(s) - ✓ [`jFqT_ZR0Z8w`](data/video/activeinferenceinstitute/Other/jFqT_ZR0Z8w/) — 1 video record(s) - ✓ [`jnavry7c11A`](data/video/activeinferenceinstitute/Other/jnavry7c11A/) — 1 video record(s) @@ -390,8 +385,7 @@ - ✓ [`p77yZJ0Bf38`](data/video/activeinferenceinstitute/Other/p77yZJ0Bf38/) — 1 video record(s) - ✓ [`q5ZKdxc2tZI`](data/video/activeinferenceinstitute/Other/q5ZKdxc2tZI/) — 1 video record(s) - ✓ [`qUJK1IDxKzg`](data/video/activeinferenceinstitute/Other/qUJK1IDxKzg/) — 1 video record(s) -- ✓ [`rIemcswLfGg`](data/video/activeinferenceinstitute/Other/rIemcswLfGg/) — 1 video record(s) — duplicate of `Applied Active Inference Symposium/2023 Ecosystem Symposium/First_Interval` -- ✓ [`rOoTkgLI1Z8`](data/video/activeinferenceinstitute/Other/rOoTkgLI1Z8/) — 1 video record(s) +- · [`rIemcswLfGg`](data/video/activeinferenceinstitute/Other/rIemcswLfGg/) — 1 video record(s) — duplicate of `Applied Active Inference Symposium/2023 Ecosystem Symposium/First_Interval` - ✓ [`rs6EjQM6EdI`](data/video/activeinferenceinstitute/Other/rs6EjQM6EdI/) — 1 video record(s) - ✓ [`wNCZmPM0V44`](data/video/activeinferenceinstitute/Other/wNCZmPM0V44/) — 1 video record(s) - ✓ [`yRhEbr4bpB0`](data/video/activeinferenceinstitute/Other/yRhEbr4bpB0/) — 1 video record(s) diff --git a/data/video/activeinferenceinstitute/Other/rIemcswLfGg/transcript.txt b/data/video/activeinferenceinstitute/Applied Active Inference Symposium/2023 Ecosystem Symposium/First_Interval/transcript.txt similarity index 100% rename from data/video/activeinferenceinstitute/Other/rIemcswLfGg/transcript.txt rename to data/video/activeinferenceinstitute/Applied Active Inference Symposium/2023 Ecosystem Symposium/First_Interval/transcript.txt diff --git a/data/video/activeinferenceinstitute/Other/PVeyvHSAwmk/transcript.txt b/data/video/activeinferenceinstitute/Applied Active Inference Symposium/2023 Ecosystem Symposium/Second_Interval/transcript.txt similarity index 100% rename from data/video/activeinferenceinstitute/Other/PVeyvHSAwmk/transcript.txt rename to data/video/activeinferenceinstitute/Applied Active Inference Symposium/2023 Ecosystem Symposium/Second_Interval/transcript.txt diff --git a/data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/ActInf_Basics_Lecture/metadata.json b/data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/ActInf_Basics_Lecture/metadata.json index 959ad2c8f..b8d6be78c 100644 --- a/data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/ActInf_Basics_Lecture/metadata.json +++ b/data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/ActInf_Basics_Lecture/metadata.json @@ -5,9 +5,172 @@ "channel": "ActiveInferenceInstitute", "category": null, "episode": null, + "title": "Active Inference for the Social Sciences — Basics of ActInf (Lecture) ~ Basics of ActInf (Lecture) — Course home", + "date": "2023-07-11", + "guests": [ + "Ben White" + ], + "other_participants": [ + "Daniel Friedman" + ], "github": "https://github.com/ActiveInferenceInstitute/ActiveInferenceJournal/tree/main/data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/ActInf_Basics_Lecture", "enriched_from": [ - "generated" + "generated", + "coda", + "curated", + "youtube" + ], + "parts": [ + { + "video_id": "BNLnbOFdgc0", + "url": "https://www.youtube.com/watch?v=BNLnbOFdgc0", + "title": "Basics of Active Inference (Lecture) ~ Ben White ~ Active Inference for the Social Sciences 2023" + } ], - "parts": [] + "sessions": [ + { + "index": 1, + "session_name": "BNLnbOFdgc0_sess01", + "start": "0:00:00", + "title": "Introduction" + }, + { + "index": 2, + "session_name": "BNLnbOFdgc0_sess02", + "start": "0:01:05", + "title": "Aims" + }, + { + "index": 3, + "session_name": "BNLnbOFdgc0_sess03", + "start": "0:05:41", + "title": "Visual Perception" + }, + { + "index": 4, + "session_name": "BNLnbOFdgc0_sess04", + "start": "0:08:08", + "title": "Perceptual Inference" + }, + { + "index": 5, + "session_name": "BNLnbOFdgc0_sess05", + "start": "0:10:17", + "title": "Readings" + }, + { + "index": 6, + "session_name": "BNLnbOFdgc0_sess06", + "start": "0:10:45", + "title": "Guiding Questions" + }, + { + "index": 7, + "session_name": "BNLnbOFdgc0_sess07", + "start": "0:11:20", + "title": "Attunement" + }, + { + "index": 8, + "session_name": "BNLnbOFdgc0_sess08", + "start": "0:12:20", + "title": "Ocean" + }, + { + "index": 9, + "session_name": "BNLnbOFdgc0_sess09", + "start": "0:12:36", + "title": "affordances" + }, + { + "index": 10, + "session_name": "BNLnbOFdgc0_sess10", + "start": "0:14:03", + "title": "inactivism" + }, + { + "index": 11, + "session_name": "BNLnbOFdgc0_sess11", + "start": "0:17:21", + "title": "representations" + }, + { + "index": 12, + "session_name": "BNLnbOFdgc0_sess12", + "start": "0:22:29", + "title": "extended active inference" + }, + { + "index": 13, + "session_name": "BNLnbOFdgc0_sess13", + "start": "0:24:49", + "title": "decoupled agents" + }, + { + "index": 14, + "session_name": "BNLnbOFdgc0_sess14", + "start": "0:26:28", + "title": "dark room problem" + }, + { + "index": 15, + "session_name": "BNLnbOFdgc0_sess15", + "start": "0:29:08", + "title": "Colin Klein 2018" + }, + { + "index": 16, + "session_name": "BNLnbOFdgc0_sess16", + "start": "0:31:48", + "title": "Active Inference Agential Behavior" + }, + { + "index": 17, + "session_name": "BNLnbOFdgc0_sess17", + "start": "0:33:00", + "title": "Active Inference and Folk Psychology" + }, + { + "index": 18, + "session_name": "BNLnbOFdgc0_sess18", + "start": "0:35:17", + "title": "Direct Response Decline" + }, + { + "index": 19, + "session_name": "BNLnbOFdgc0_sess19", + "start": "0:39:54", + "title": "Expectations" + }, + { + "index": 20, + "session_name": "BNLnbOFdgc0_sess20", + "start": "0:43:31", + "title": "Examples" + }, + { + "index": 21, + "session_name": "BNLnbOFdgc0_sess21", + "start": "0:44:15", + "title": "Allostasis" + }, + { + "index": 22, + "session_name": "BNLnbOFdgc0_sess22", + "start": "0:48:29", + "title": "Emotion" + }, + { + "index": 23, + "session_name": "BNLnbOFdgc0_sess23", + "start": "0:54:14", + "title": "Allostatic Control" + }, + { + "index": 24, + "session_name": "BNLnbOFdgc0_sess24", + "start": "0:58:32", + "title": "Interceptive Prediction" + } + ] } diff --git a/data/video/activeinferenceinstitute/Other/BNLnbOFdgc0/transcript.txt b/data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/ActInf_Basics_Lecture/transcript.txt similarity index 100% rename from data/video/activeinferenceinstitute/Other/BNLnbOFdgc0/transcript.txt rename to data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/ActInf_Basics_Lecture/transcript.txt diff --git a/data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/Introduction_Lecture/metadata.json b/data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/Introduction_Lecture/metadata.json index a67a38487..fe7c6c017 100644 --- a/data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/Introduction_Lecture/metadata.json +++ b/data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/Introduction_Lecture/metadata.json @@ -5,9 +5,119 @@ "channel": "ActiveInferenceInstitute", "category": null, "episode": null, + "title": "Active Inference for the Social Sciences — Introduction Lecture ~ Introduction (Lecture) — Course home", + "date": "2023-07-07", + "guests": [ + "Avel Guénin-Carlut" + ], + "other_participants": [ + "Daniel Friedman", + "Ben White" + ], "github": "https://github.com/ActiveInferenceInstitute/ActiveInferenceJournal/tree/main/data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/Introduction_Lecture", "enriched_from": [ - "generated" + "generated", + "coda", + "curated", + "youtube" + ], + "parts": [ + { + "video_id": "JkORFoyk8o8", + "url": "https://www.youtube.com/watch?v=JkORFoyk8o8", + "title": "Introduction (Lecture) ~ Avel Guénin-Carlut ~ Active Inference for the Social Sciences 2023" + } ], - "parts": [] + "sessions": [ + { + "index": 1, + "session_name": "JkORFoyk8o8_sess01", + "start": "0:00:00", + "title": "Introduction" + }, + { + "index": 2, + "session_name": "JkORFoyk8o8_sess02", + "start": "0:02:19", + "title": "Teaching Team" + }, + { + "index": 3, + "session_name": "JkORFoyk8o8_sess03", + "start": "0:04:48", + "title": "Motivation" + }, + { + "index": 4, + "session_name": "JkORFoyk8o8_sess04", + "start": "0:06:15", + "title": "Explanation of Cultural Dynamics" + }, + { + "index": 5, + "session_name": "JkORFoyk8o8_sess05", + "start": "0:09:15", + "title": "Cognitive Laws of Behavior" + }, + { + "index": 6, + "session_name": "JkORFoyk8o8_sess06", + "start": "0:13:24", + "title": "Active Inference" + }, + { + "index": 7, + "session_name": "JkORFoyk8o8_sess07", + "start": "0:17:10", + "title": "Normativity" + }, + { + "index": 8, + "session_name": "JkORFoyk8o8_sess08", + "start": "0:23:52", + "title": "The Case" + }, + { + "index": 9, + "session_name": "JkORFoyk8o8_sess09", + "start": "0:24:54", + "title": "Public Opinion" + }, + { + "index": 10, + "session_name": "JkORFoyk8o8_sess10", + "start": "0:28:49", + "title": "Nationalism" + }, + { + "index": 11, + "session_name": "JkORFoyk8o8_sess11", + "start": "0:32:48", + "title": "Energy Capture" + }, + { + "index": 12, + "session_name": "JkORFoyk8o8_sess12", + "start": "0:36:56", + "title": "Structural Agency" + }, + { + "index": 13, + "session_name": "JkORFoyk8o8_sess13", + "start": "0:40:37", + "title": "Landscape" + }, + { + "index": 14, + "session_name": "JkORFoyk8o8_sess14", + "start": "0:43:32", + "title": "World Work" + }, + { + "index": 15, + "session_name": "JkORFoyk8o8_sess15", + "start": "0:52:17", + "title": "Projects" + } + ] } diff --git a/data/video/activeinferenceinstitute/Other/JkORFoyk8o8/transcript.txt b/data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/Introduction_Lecture/transcript.txt similarity index 100% rename from data/video/activeinferenceinstitute/Other/JkORFoyk8o8/transcript.txt rename to data/video/activeinferenceinstitute/Courses/ActiveInferenceForTheSocialSciences/Introduction_Lecture/transcript.txt diff --git a/data/video/activeinferenceinstitute/Courses/PhysicsAsInformationProcessing_ChrisFields/metadata.json b/data/video/activeinferenceinstitute/Courses/PhysicsAsInformationProcessing_ChrisFields/metadata.json index f396fb77f..1278c47b0 100644 --- a/data/video/activeinferenceinstitute/Courses/PhysicsAsInformationProcessing_ChrisFields/metadata.json +++ b/data/video/activeinferenceinstitute/Courses/PhysicsAsInformationProcessing_ChrisFields/metadata.json @@ -24,11 +24,6 @@ "youtube" ], "parts": [ - { - "video_id": "transcripts", - "url": "https://www.youtube.com/watch?v=transcripts", - "title": "" - }, { "video_id": "hOto96Q8sYQ", "url": "https://www.youtube.com/watch?v=hOto96Q8sYQ", @@ -90,6 +85,32 @@ "video_id": "SV0glS6stuA", "url": "https://www.youtube.com/watch?v=SV0glS6stuA", "title": "“Physics as Information Processing” by Chris Fields ~ Course overview" + }, + { + "video_id": "WkWIqpxWRM4", + "url": "https://www.youtube.com/watch?v=WkWIqpxWRM4", + "title": "Chris Fields ~ Physics As Information Processing ~ 2023 Lecture 2 ~ Lecture 2 — Course information", + "slides_url": "https://docs.google.com/presentation/d/1LL5t6OgkeFTXWdmHoAmUphVc21xEdb758UruGRR_eLw/edit?usp=sharing", + "date": "2023-06-15" + }, + { + "video_id": "TOZp_XNYijQ", + "url": "https://www.youtube.com/watch?v=TOZp_XNYijQ", + "title": "Chris Fields ~ Physics As Information Processing ~ 2023 Lecture 3 ~ Lecture 3 — Course information", + "slides_url": "https://docs.google.com/presentation/d/1YZS8aunpEfoZGvhVk01fulDFOkIKj9PMUeszcKZ195g/edit?usp=sharing", + "date": "2023-07-13" + }, + { + "video_id": "rOoTkgLI1Z8", + "url": "https://www.youtube.com/watch?v=rOoTkgLI1Z8", + "title": "Chris Fields ~ Physics As Information Processing ~ 2023 Discussion 1 ~ Discussion 1 — Course information", + "date": "2023-06-03" + }, + { + "video_id": "hSj7eOYuZe4", + "url": "https://www.youtube.com/watch?v=hSj7eOYuZe4", + "title": "Chris Fields ~ Physics As Information Processing ~ 2023 Discussion 2 ~ Discussion 2 — Course information", + "date": "2023-07-01" } ], "sessions": [ diff --git a/data/video/activeinferenceinstitute/Courses/PhysicsAsInformationProcessing_ChrisFields/transcript.txt b/data/video/activeinferenceinstitute/Courses/PhysicsAsInformationProcessing_ChrisFields/transcript.txt index 24b930efb..943dd8105 100644 --- a/data/video/activeinferenceinstitute/Courses/PhysicsAsInformationProcessing_ChrisFields/transcript.txt +++ b/data/video/activeinferenceinstitute/Courses/PhysicsAsInformationProcessing_ChrisFields/transcript.txt @@ -13263,4 +13263,2790 @@ anonymous way for you to contribute. Lastly, you can go to register to join disc live interactive discussion live streams about two weeks after each lecture, you can complete this form and we'll add you to calendar events. So thanks again for your interest in the course, Physics is Information Processing, and we hope to see you either watching and re-watching the videos, asking questions -and answers, and maybe even registering to join the discussions. Thanks and see you there. \ No newline at end of file +and answers, and maybe even registering to join the discussions. Thanks and see you there. + +## WkWIqpxWRM4 + +## WkWIqpxWRM4 + +Hello everyone, it's June 15th, 2023. We are here in lecture two of Chris Field's course, +Physics as Information Processing. So thank you, Chris. Looking forward to the lecture. +Thank you, Daniel. And yes, welcome to this session. This session is titled Why Quantum Physics? +And if you will recall from the first session, we reviewed the history of physics and some of the +history of mathematics and computer science from the end of the 19th century through to +the beginning of the 21st century and discussed the slow development from classical thermodynamics +of quantum information theory and specifically characterizing quantum information theory as a new +kind of physics that describes systems that are exchanging finite amounts of discreetly encoded +information across some intervening boundary. And so we use this very conventional graphical notation of +two agents or physical systems, Alice and Bob, A and B, that are exchanging energy and information +across some boundary. And I always use a blue ellipse like this to indicate the boundary that separates Alice +from Bob. And this new approach to physics is entirely topological. It's not geometric. So it doesn't assume a +particular background space-time. So it doesn't assume that Alice and Bob are spatially separated. +And so this makes it a very different kind of theory from classical information theory in that the channel +that separates Alice from Bob, this boundary, is a boundary in state space. It's not a boundary in some +geometric embedding space. And in particular, it's not a boundary or a channel that's embedded in a three-dimensional space +that separates Alice from Bob. +So that's what we talked about last time. And today, what I want to discuss is how quantum theory in particular +makes this idea of physics as a theory of communication simple and obvious. And quantum theory, of course, has a +terrible reputation of being abstruse and mathematically incredibly complicated and counterintuitive and +difficult to understand. And this is one quotation among many from leading physicists pointing out that +quantum mechanics is just difficult. And as you probably know, there's an entire philosophical industry of +interpretations of quantum theory that try to make sense of its ontology. And so what I don't want to do today +is try to introduce quantum mechanics. We're not doing any of this. Some of you will recognize this as the table of contents, +the first part of the table of contents of the famous textbook by Landau and Lifshitz. But if you've studied quantum theory +in undergraduate or graduate school, you've probably dealt with a textbook structure much like this one. +And you've probably been introduced to quantum mechanics as it was traditionally conceived as +essentially a mechanical theory, a theory of motion with wave functions and particle representations and on and on and on. +So we're not going to try to do this. What we're going to do instead is take a completely +information-theoretic approach and we're going to characterize information transfer in a quantum-theoretic way +without any assumptions about mechanics or space-time or any of that. +So instead of this, we're going to ask a simple question. By a simple question, I mean a question with +just a yes-no answer or a binary answer. So let's use up or down as our example question. +And to ask this question, we need three things. First, we need an action. We need a way of asking it. +For example, making a sound or writing something down. +Second, we need a thing that we can ask the question of. We need an environment or a friend or the rest of +the world or an experimental apparatus or some system or other that we're going to act on to ask this question. +And the third thing that we need to ask this question is a shared language. +And it's the shared language that's often neglected in models based in physics. +And I think the importance of the shared language became clear in classical information theory. +And it's become clear again, of course, in quantum information theory. But this has only happened over the last few decades. +So what I want to do today is to construct in a step-by-step fashion a formal representation of these three things that we need to ask a simple question. +And what we'll find is that that formal representation is, in fact, quantum theory. +So it's quantum theory that provides us with a formal representation of asking these simple binary questions. +And of course, if you can ask a binary questions, you can ask any finite number of binary questions. +So you can construct an arbitrarily complex finite decision tree that ends up formulating +an arbitrarily complex but still finite question. +So this really is a general model of communicative interactions between agents. +So let's go about this and talk first about the action of asking a question. +And so the first thing I want you to do is actually say out loud, up or down. +Actually ask a question. +And then note two things about what just happened. +The first thing to note is that it took some time to do that. +That action actually required a finite amount of time. +A couple of seconds, not very long, but it's not zero. +It always takes time to ask a question. +And the second thing to note is that it took energy to actually say up or down. +It requires a little bit of exertion to ask a question. +So let's put those two together and define a word. +We're going to define the word action to mean some energy that's spent in some time. +So asking a question is an action. +And it qualifies as an action because it requires spending some energy in some finite amount of time. +And so we can write this as action equals energy times time. +Or I'll sometimes use the abbreviation E times T. +And action is always required when there's some interaction or two-way action between systems. +So whenever we talk about physical interaction, we're always talking about two systems doing something to each other that requires both energy and time. +So let's define another concept, a concept of a minimum action needed to ask a question, and specifically a question with a binary answer. +And we'll just pick a symbol to stand for that minimum action. +And the symbol we'll use because it's conventional is H. +And it turns out that this H is Planck's constant. +The number that Planck came up with when he was trying to solve the black body problem back in 1900 that we talked about last time. +And recall, that was the problem of characterizing the spectrum of radiation emitted by some physical system that was hot. +How much energy comes out as a function of temperature. +And if you assume that the spectrum of energy coming out is continuous, you get predictions that are wrong. +And Planck got predictions that were correct by assuming that that spectrum was discrete. +And that this H represented the minimum amount of energy that could be emitted in a particular time. +So it represented the minimum energy times time or action. +So let's look at an example of that. +And here's an example that may be familiar. +It's the example of your own visual system. +So in the retinal, the cells of your retinas, there are many, many molecules that are rhodopsins of one variety or other that respond to different wavelengths of visible light. +So let's look at that. +And they have some efficiency for capturing light. +And let's just assume that they're optimally efficient, that they're as efficient as they could possibly be. +And that means that they require the minimum amount of energy to change shape, which is what they do when they're impacted by light, +to capture the energy and hence the information that's available in the light. +And the minimum energy to do anything that records one bit of information, so yes or no, up or down, is 0.7. +So log 2 of Boltzmann's constant times the temperature. +And again, we saw this last time. +This is from roughly 1875 and was then built into classical information theory by Shannon and then by Landauer in the mid-20th century. +So what we're doing is assuming that since rhodopsin is optimally efficient, all it requires is this amount of energy to change shape. +And it's known experimentally that rhodopsin takes about 200 femtoseconds to change shape. +So that's 200 times 10 to the minus 15th seconds. +So it's a very, very fast reaction, chemical reaction. +So we can just calculate what the action is. +And at 37 degrees centigrade, so human body temperature, kT is about 4.3 times 10 to the minus 21 joule. +And joule is just the standard unit of energy. +So we can multiply everything together. + +KT times 0.7 times this response time of 200 femtoseconds. +And if you're familiar with quantum mechanics, you'll instantly recognize that number because it's very close to the value of Planck's constant, which is 6.6 times 10 to the minus 34 joule seconds. +So this says that rhodopsin is very near to optimal. +The best it would be able to do would be Planck's constant. +That's the minimum action that can be undertaken. +So rhodopsin is clearly doing very well. +It can't be as efficient as we're assuming it to be, but it's pretty efficient. +So this number Planck's constant is, and this idea of a minimum action is not something we just pulled out of the air. +It's something that one can observe when you're looking at what real biological systems in particular do and other physical systems do. +So let's think about the second component of this project of asking a question, and that's the thing to be asked. +And we want to ask a very simple question. +We want to have a theory of asking very simple questions. +So let's define the simplest thing that we can ask a binary question to. +And let's say this is the thing that we can ask a binary question of using just this minimum action h. +So we're going to ask for the minimal thing that it's possible physically to ask a question of. +And this thing needs a name, so let's call it a qubit. +And that word is a contraction of quantum bit. +And we can draw it like this, a little sphere with an arrow through it. +That's called a block sphere. +And we use a symbol q, and that notation with the bar and the angle bracket is called a bra notation, and it originates with Dirac. +So it's a traditional way of writing inner products, actually, in quantum theory, +to indicate the state of q. +So this thing called a qubit only responds to binary questions. +So it clearly can have only two states, the state that answers yes and the state that answers no. +And conventionally, we draw the arrow going up if the answer is yes and the arrow going down if the answer is no. +And so those are the two states that the qubit can be in when it's answered a question. +So now let's draw a picture for the act of asking a question of q in the state q. +And let's draw it like this. +This curvy arrow is the action of asking the question and getting the answer back. +And we're asking it of this minimum system, the qubit. +And let's call this entire act of h. +And write h acting on q by this expression hq. +So again, we're developing the notation of quantum theory, which +all of you who know the theory already will find familiar. +So the third thing that we have to build into this picture and into the theory +is the shared language. +So we have to know what we mean by up or down if we're going to ask the question up or down. +And we have to have some representation of what it means for the qubit to distinguish up or down. +So let's say that to us, up and down mean this little diagram with an arrow that's labeled up on one end and down on the other end. +So we're defining up and down in a very simple way as just opposites. +We don't know anything else about them except that they're opposites. +So this is a stripped down one bit question. +We could have called them one and minus one. +We could have called them a and b. +We could have called them anything. +We're just indicating that up and down are two opposing concepts. +And so this object that we've used to make this definition of the relationship between our two words, up and down, is called a reference frame. +It's a physical object that encodes a concept or a meaning. +And if you think about using language, using English, for example, you can understand me speaking in English because your brain has the right sorts of neural circuits to allow you to understand English and produce English using your hand and your mouth. +And those neural circuits are reference frames in your brain for understanding and producing English. +And they encode the relationships between different words. +So you all know about chat GPT. +You can think of all of the weights, all of the 10 to the 19th or however many parameters there are in GPT-4 as a reference frame that describes the relationships between words and phrases in English and many other languages. +So this idea of a reference frames. +So this idea of a reference frame is an instance of Landauer's principle, which is often summarized as saying that information is physical. +What that means is that information always has to have some encoding. +And that encoding, that object, can always be thought of as a reference frame. +So let's put all this stuff together. +What we have now is a reference frame that defines the difference between up and down. +We have an action that we've labeled H. +And the action acts on this object, the qubit, which is in some state. +And what we've drawn here now is an experiment. +And you can think of the incoming part of the arrow that goes from the reference frame to the qubit as a preparation of the qubit. +So you can think of it as I'm doing something to the qubit to put it in a particular state, like up. +Because that's the state I want it to be in. +And I can think of this returning part of the arrow as a measurement of the state that the qubit is actually in. +So if the qubit is actually in this state with the arrow pointing up, then I'm going to get back the answer up when I act on the qubit to measure its state. +So now we can recognize this action H as the Hamiltonian operator in quantum theory, which always takes some time, which we represent by an interval, dt, to act on some state, and in particular this binary state q. +So this whole picture depicts the action of a very simple Hamiltonian operator. +And again, I want to emphasize that we've just defined the qubit as the simplest thing that we can ask a binary question of. +So how does this H work? +We can use H to prepare this qubit in one of two states, which we're going to label 1 and minus 1. +The state 1 corresponds to the up direction, and the state minus 1 corresponds to the down direction. +And doing this to q, so preparing q in one of these two states takes energy. +And we know how much. +Boltzmann told us back in 1875 that the minimum energy to reduce the entropy of the universe by one bit is log 2 of kt. +We saw that earlier when we were thinking about rhodopsin. +So we can say that acting with H always requires some number times kt, where that number is bigger than the log of 2. +Remember, the log of 2 is just in there to get units of bits of binary questions, as opposed to units of nats, which are questions that have e answers, where e is the exponential base, the natural exponential base. +But we want to use units of bits, so we have to always have this factor of log 2. +So this lets us write H in a particular way. +We can take out this energetic term that Boltzmann tells us we have to include, beta times kt. +And what's left over is, in a sense, the pure representation of the action stripping out the energy requirement. +And we're going to call that m, because m is the standard symbol to use in quantum theory for a measurement operator. +And m is now dimensionless. +H has units of energy. +Kt is an energy. +And m has no units at all. +So it's H without the energy. +So now we can ask, what is this operator m that we've defined as the simplest possible action on a qubit, but abstracting away the amount of energy that we have to use to act on the qubit? +Well, what m does is just encode a bit or decode a bit. +So we can use this operator m to say, I want q to have the value 1. +I want it to be up. +So I can sort of reach out with my operator m and grab q and turn it so that its arrow is pointing up. +And that's the operation that's called preparation in physics. +It's what it means to prepare an experimental apparatus, for example. +You actually grab hold of the apparatus and you twist some knobs and flick some switches to prepare it to be in some state. +And you can use m to measure a state that isn't known. +So I can use m to grab a hold of q and look at it and determine whether it's in state 1 or minus 1. +And I can represent these as just a cycle that starts with measuring and then prepares, or that starts with preparing and then measuring. +And if I represent it that way, what I see is m having a value, which is either the value I want to encode or the value that I have measured, and q. +And m acts on that value and q to produce a particular state of q, to put q in a particular state or to realize that q is in that state. +And so m is actually an identity acting on some value and q. +So it maps 1 and q to 1 and q, and it maps minus 1 and q to minus 1 and q, if you do the whole cycle of preparation and then measurement. +And that makes sense. +Notice it's actually an axiom in some formulations of quantum theory that if I measure the state of some object, what I'm doing is also preparing it in that state. +So if I immediately measure it again, I'll see the same value. +And that makes sense physically. +If I put my coffee cup on the table and then I look for it, it's likely going to be on the table unless something else intervenes. +Well, what does this mean? +It means that we can do a little bit of algebra. +It means that we can write this original expression of acting on q with our Hamiltonian in this somewhat more explicit form of using the energy beta kt and our operator m on q. +And what I get out of that is the energy beta kt multiplied by q. +So hq equals eq. +That's the Schrodinger equation. +So that's one of the central equations of quantum theory. +What it says is that this operator h is the total energy operator on a system. +Its eigenvalues are the total energies of the system. +And we can now see where that energy is coming from in this very, very simple stripped down kind of representation of an action. +The energy is the thermodynamic energy that has to be spent to determine one binary outcome, up or down. +So, now, if you're familiar with quantum mechanics, you will have seen exponentiated operators like the next line here of algebra. +This operator p, which is time dependent, which is the exponential of h multiplied by it over h-bar, is called the propagator. +It's the unitary operator in quantum theory, the operator that conserves probability or conserves information. +And so, using the representation of h as beta kt, we can see that this operator p is just e to the minus beta m. +If we recall the wick rotation from the first session, this strange equivalence of it over h-bar with 1 over kt that wasn't discovered until 50 years after the invention of quantum theory. +And it's still a very debated thing. +I mean, people worry about what this equation really means. +But one thing that it means in this setting, where h has this particularly simple form, is that the propagator is now just the exponential of a number multiplied by this very simple stripped down operator that encodes a bit or decodes a bit. +But we know something about the exponential of a bit. +But we know something about the exponential propagator just by looking at its form. +If you're, again, familiar with these sorts of exponential operators, you know that they're wave equations. +And we can make sense of that using the diagram up above. +So, what m is doing is just performing a cycle. +It's just mapping 1 and q back to itself or minus 1 and q back to itself. +So, we can think of m as just taking this object and flipping it back around until it gets itself again. +And that kind of cycle can always be turned into a wave equation. +And e to the i over h-bar times ht is a canonical form of a wave equation where the horizontal axis here on the wave is time. +And if you look at the time of the time of the thing spinning around like this, which is exactly the same thing as a wave traveling through time. +So, that tells us something very important. +It tells us that this equivalent expression e to the minus beta m is also a wave equation. +So, it tells us that this operator m has built into it some sense of time. +It's the sense of time of this prepare-measure cycle. +And that's implicit in m, right? +We've used the wick rotation to get rid of the kind of what I'll end up calling the external time t, the objective time t. +And we'll see in July where this internal time comes from in terms of the wick rotation. +But we can say now that it's an agent-specific time. +It's an internal time reference frame that's built into our idea of this system, Alice, that can act on the world. +And it represents the time that's required to ask this question. +Okay. +Chris, can I ask a question at this point? +Yes. +So, on this slide with the operator m, I'm curious, how can the agent deploying the operator m tell the difference or break the symmetry between preparation and measurement? +Or are they just completely dual pictures? +Yeah, they're completely dual. +So, the agent never breaks the symmetry. +The agent can't actually distinguish the act of preparing from the act of measuring. +Every preparation is a measurement and every measurement is a preparation. +Right. +That's in fact what this axiom of… +Unitarity? +…measuring a state leaves the particle in the measured state means. +Right. +Right. +So, yeah, mathematically they're duals. +Mm-hmm. +Thank you. +Okay. +So, let's now use this theory that we've developed to describe the interaction between Alice and Bob. +And let's think of a situation in which Alice prepares a qubit that Bob then measures. +And then Bob prepares the very same qubit. +And then Alice measures it. +So, Alice acts with her Hamiltonian operator, which we'll label HA. +And Bob acts with his Hamiltonian operator that we label HB. +And let's say that Alice prepares the qubit as end state up. +And Bob can then measure the qubit end state up and say, oh, the qubit is now up. +And if Bob then prepares the qubit end state down, Alice can measure the qubit end state down and say, oh, the qubit switched from the up that I prepared to down. +So, Bob must have prepared it as down. +So, what's happened here? +Alice and Bob have actually shared a bit flip. +Right. +Alice said up and Bob said down. +So, and Alice heard him say down. +So, they've shared this idea of change by interacting with each other. +Now, this works as long as they mean the same thing by up and down. +So, we've implicitly assumed in talking about them communicating that they both have this reference frame that distinguishes up from down. +And this is actually a completely general model of what I'll call a noiseless one qubit channel. +Now, what does that mean? +It's noiseless because there's nothing else in the picture. +Right. +There's nothing in the picture that can jiggle the qubit when Alice and Bob aren't looking. +Alice and Bob are the only systems we're talking about. +Alice and Bob are the only systems that are in the picture. +So, there's no kind of third party or environment or heat bath or anything like that that's going to mess with the qubit and disrupt their interaction. +So, that's a noiseless channel. +And it's a one qubit channel because they just have one qubit that they're interacting with. +And they're alternately preparing and measuring its state. +Now, clearly, if they can interact using one qubit, they can interact using any finite number of qubits. +So, here's a picture of an in qubit channel that Alice and Bob, which here I've labeled A and B because they took this figure from a paper, are sharing. +And Alice and Bob both have n-qubit operators, one for each of the n qubits that they share. +And they just do this prepare and measurement cycle on all n qubits. +So, you can think of Alice preparing the n qubits, Bob measuring the n qubits, and then Bob preparing and Alice measuring. +And that allows them to exchange n-qubit messages. +And we can make n as large as we want. +And so, they can exchange messages of arbitrary complexity. +And if they share the meaning of up and down, then they can exchange these messages with no noise. +So, they can communicate. +And we now have our theory of two communicating agents that are sharing arbitrary finite information. +Now, I've labeled q1 through qn in this picture a holographic screen. +And in fact, that's exactly what it is. +Recall from session one, this slide illustrating the holographic principle. +Shannon made this point that a bit of information is an answer to a yes-no question. +And so, we can write down a number of possible states. +We can write down an entropy. +And the limiting case of that entropy is the entropy of a black hole, the densest possible encoding of information in space or on a boundary. +And this holographic principle allows us to see any boundary, as we discussed in session one, as an encoding of bits. +So, the previous boundary of these n-qubits, we can think of as a holographic screen. +Now, why is that important? +It's important because it allows us to use this idea of holography and this idea of a fixed encoding of a message or a fixed encoding of an observable entropy at some density that was developed in cosmology. +So, we can import a whole bunch of theory that was done someplace else into quantum information theory and see that it works perfectly because it describes exactly the same situation. +Now, what does this tell us philosophically? +It tells us that Alice and Bob can learn about each other at most n-bits per communication cycle. +So, the maximal amount of information that they can get from an observation is the n-bits that the other one wrote on the screen. +So, this tells us something very important, that when an agent is looking at their environment, which is this other agent, Bob, the amount of information that can be extracted from the environment is actually proportional to the size in bits of the boundary. +And the holographic principle just makes that size finite in space. +And I said early on that there's no embedding space-time in this theory that Alice and Bob are not thought of as separated in space. +But as soon as we use the holographic principle, we can think of the boundary between them as having a spatial coordinate, even though they don't have a spatial coordinate. +So, we'll get back to this point in September when we talk about theories of emergent space-time. +But all of these theories of emergent space-time come back to this idea that boundaries can have spatial extent, even though the systems that they separate don't have any natural spatial description or natural spatial degrees of freedom. +Okay, so now let's recall another picture from session one. +We talked a little bit about Feynman's theory of scattering and the origin of Feynman diagrams. +And the idea that if Alice prepares some system in some initial state and then measures the state later, if we want to calculate what the new state is, we have to take into account all of the possible paths that join the initial state that Alice prepared to the final state that Alice measured. +So, what does that mean? +So, what does that mean? +I mean, here effectively, if we think of, suppose I'm Alice and the boundary is my screen. +I'm talking to the screen. +And then later on, I'm going to measure the screen. +And between my preparation actions and my measurement actions, a bunch of stuff happens on the other side of the screen. +And if I want to predict what's going to happen on the other side of the screen, I've got to take into account every possible thing that could happen. +And I have to give some weight to those possible things that could happen. +And quantum theory, in fact, tells you how to assign those weights to things that could happen in order to make a measurement given some theory of the mechanism that makes things happen. +But the fact that all of the paths have to be counted is what unitary evolution means. +This evolution that preserves information, that doesn't take any away, it doesn't add any. +It just represents it in different ways. +So, if you look at this screen and imagine that Alice is interacting with her boundary, and then she's interacting with her boundary again, it looks just like the previous screen. +And the wavy red arrows are what's happening inside Bob. +The wavy red arrows are a representation of what's going on inside Bob as he measures his side of the screen and then prepares his side of the screen. +So, what Feynman is representing here is just the action of the environment or the action of the system while Alice isn't looking between Alice's preparation stage and her measurement stage while Bob is doing his measurement and preparation. +So, this diagram is actually completely symmetrical. +We could have drawn an arrow where the wavy red things are and drawn wavy red things where we've talked about Alice and we'd have the same picture just from Bob's point of view. +We can also think of this, of course, as Alice communicating with her future self by writing a memory record someplace in the world and then coming back and reading it. +And if Alice writes a memory on the world and then comes back later on and reads it, she actually has to take into account some theory of all of the things the world might have done to her memory while she wasn't looking. +So, this picture of Feynman's actually tells us a lot about memory. +And when we think about organisms running around in the world doing things like we do all the time, we tend to think of memories as things that are persistent and unchanging and all of that. +But we know that that's not true. +We know that memories degrade in various ways. +We know that adversarial third parties can come and change what we've done so that they fool us into thinking that we've remembered something that we haven't and so on and so forth. +So, that whole picture of memory is actually encoded in this picture that Feynman drew in the late 50s to represent scattering of particles. +So, what we've done here is started with this simple idea of asking questions. +And we've actually constructed a fair amount of quantum theory. +We constructed the Schrodinger equation. +We constructed the unitary propagator. +We constructed the idea of a reference frame, which we'll talk about intensely next time in July. +So, we can sum this up by saying that quantum theory simply is a theory of communicating agents. +And that quantum mechanics of the theory that Feynman was complaining about in that earlier slide, +that quantum mechanics generates this problem that seems intractable of how to understand measurement, +because the theory is in fact about measurement. +It's not about mechanical motion at all. +So, quantum theory actually tells us nothing about ontology. +And this is why the interpretations of quantum theory, the many worlds interpretation, +where you have all these branching universes and so forth, are not really science. +They're philosophy. +They're attempts to understand an ontology that's generated by a theory that doesn't talk about ontology. +It talks about communication between agents. +But we've left something very important out. +We've left out a consideration of what happens when Alice and Bob don't mean the same thing by up and down. +And we've left out the famous phenomenon of entanglement. +Spooky action at a distance, as Einstein called it. +And it turns out these things that we've left out are very, very closely related. +And we're going to see why in July, but I'm going to give you a brief preview now. +So what happens when Alice and Bob actually mean different things by up and down? +What if their reference frames aren't aligned? +Well, the first thing to note is that nothing in the theory says they have to be aligned. +So we have to consider the possibility that they're not. +And in fact, both Alice and Bob have to have what's called free choice of reference frame. +When Alice prepares her instrument, she has to be able to prepare it any way she wants to in order to do an experiment. +And when Bob makes up a sentence to tell Alice, he has to be able to choose any language he wants to use. +He's a free agent in this way. +And if they don't have that freedom, it turns out they're entangled. +So reference frame non-alignment and entanglement are very, very closely coupled. +Now, misalignment of reference frames is what produces superpositions of outcomes. +So if Alice makes her measurements with her up and down arrow arranged vertically, for example, +Alice is not going to have any idea what Bob did. +I mean, she can't see distinctions in the left, right direction. +She can only see distinctions in the up and down direction. +So all she can do is flip a coin. +If she wants to know what Bob prepared, it's 50% likely he prepared up for her, what she calls one. +And it's 50% likely that he prepared what's down for her. +She can't tell. +Now, if Bob's reference frame is pointing at about 45 degrees, +so up for him means 45 degrees off of Alice's up, +then she can sort of tell what he did. +If he's pointing 45 degrees from up, then she's probably going to measure up. +And if he's pointing 45 degrees from down, she's probably going to measure down. +So her outcomes are 75% for 1 and 25% for minus 1 if he prepares at this 45 degree angle. +So we can write this in a particular way. +We can say that the value that Alice gets acting with her M on a cubit that Bob prepared is some number times what she measures is up, +and some other number plus times what she measures is minus 1, +where those numbers, the squares of those numbers add to 1. +So why do they have to add to 1? +Well, they add to 1 because of the Pythagorean theorem. +So 1 is the distance, the total distance between the two outcomes, +and A squared plus B squared are the other two sides of that triangle. +So another way to think of it is that the metric distance in possibility space is 1, +and A squared and B squared are just the Euclidean distance components. +Again, Euclidean distance is based on the Pythagorean theorem. +So this is the Born rule in quantum theory. +It's the rule that tells us what happens when we're trying to measure a superposition, +and it tells us where these superpositions come from. +We have a mismatch in reference frames between two agents that are communicating. +So if we go back to the free energy principle, which is what this whole active inference class is about, +and we formulate it in quantum theory, we get a very simple idea of what the free energy principle is. +The FEP in quantum theory just says that interacting systems are going to behave in a way that aligns their reference frames, +and that makes their communication as good as it can be. +But since aligning reference frames perfectly leads to entanglement, +the free energy principle drives systems asymptotically toward entanglement. +Well, the principle of unitarity says that all systems will approach entanglement asymptotically. +So we can see the FEP is actually a classical statement of the principle of unitarity. +And what we'll see in the next session is why this is the case, +why active inference, when we think of it quantum mechanically, +is a way of talking about the principle of unitarity, which is the deepest axiom of quantum theory. +And we'll also see that this question of alignment becomes more and more complex as the complexity of the messages that the systems are trying to communicate to each other increases. +And that message complexity correlates with system complexity. +So that aligning complicated systems is very hard, although aligning very simple systems is fairly easy. +And if you think about what's the simplest possible theory in science, it's particle physics, because you're dealing with these very simple things that interact in extremely simple ways. +And the really difficult parts of science are things like sociology, where you're dealing with incredibly complicated systems like humans that interact in incredibly complicated ways. +And all we have are very vague kind of folk theories of how that works. +But we know from that example that aligning complex systems is extremely difficult. +That's why humans don't get along very well. +So the second discussion section, which Andrew will be leading, is on Saturday, the 1st of July, a couple of weeks from now. +The third lecture session will be on July 13th. +And I hope you all will contribute to the interactive question and answer session on the web and show up for the discussion on Saturday, the 1st of July. +And join us again in July for the discussion of quantum reference frames and reference frame alignment. +So thank you very much. +Thank you, Chris. +All right. +Thank you all. +Again, please check out the Chorus syllabus. +Ask questions through the site and see you at the discussion. +See you next time. +Thank you. +Bye. +Bye. + +## TOZp_XNYijQ + +## TOZp_XNYijQ + +Hello and welcome everyone. It's July 13th, 2023. We're here in Physics as Information Processing +with the third lecture by Chris Fields. So please check out the course website, +CODA document, where you can submit questions, check out the recordings and descriptions of +all sessions and register if you want to participate in one of the discussion sections. +So thank you again, Chris, to you. Thank you and welcome to Physics as Information Processing +session three, which is about quantum reference frames. So in this course, we've talked about +quantum information theory, which we characterized as a general theory that describes the exchange +of information, exchange of finite information, between two finite agents that are separated +by some boundary. And we've emphasized that this is a topological theory, not a geometric theory. +So it makes no assumptions about an embedding spacetime. And last time, we showed how to characterize +this boundary of script B as an array of N, some finite N quantum bits or qubits, and how +to characterize the interaction of the agents A and B with the boundary as an alternating cycle of preparing +and measuring these qubits in this array. And we showed that their interaction could be represented +as a sum of operators, each of which act on just one of the qubits. And we also talked about the need for each +of the agents, A and B, to choose a reference frame for their interaction with the qubits that defined which way was up. +So what the orientations of the qubits meant. And emphasized that the two observers, A and B, could choose different +reference frames, so different ways of measuring and preparing their qubits that use different meanings for the idea of up for each qubit. +And we ended the last session with a pointer forward toward the quantum formulation of the free energy principle, which can be stated as the claim that interacting systems behave in a way that aligns their reference frames. +And I'll point it out at that time, and I'll point it out at that time, that what this means is that the two agents, A and B, approach entanglement asymptotically, and that therefore the FEP is a classical limit of the principle of unitarity, which is the core principle of quantum theory. +So we'll see in this session why this is true. So what we'll talk about today is what reference frames are in general, how they make observations and actions meaningful, so how they give physics a semantics. +And then we'll talk about the free energy principle. +So what is a reference frame? The idea that physics involves reference frames goes back at least to Galileo, but was probably familiar well before that, just not formalized in the way that Galileo formalized it. +So the key idea can be thought of in terms of motion. +And you're all familiar, for example, with riding a bicycle or driving a car and seeing the world as if it's passing by you at some velocity, with some velocity. +And of course, with respect to the world, you're passing by it with some velocity. +And this idea that velocity is not absolute, but that's relative to some observer who measures it in some reference frame is the foundation for theories of relativity. +So Galileo developed his theory of relativity, which basically was just a way of converting from classical velocities measured by one observer to classical velocities measured by another observer that was moving with respect to the first one. +And this provided the foundation for Einstein's theories of relativity, which modified Galileo's by adding the concept that the observers have to change, exchange information, and can only exchange it at the speed of light, which is finite. +So this idea in physics has a long history. +Of course, the core idea is that measurements are always with respect to something or other. +And the something or other is a reference frame. +And different agents can choose different reference frames. +And this idea is very old. +Here we have a picture of Anubis choosing a feather as the reference frame with which to measure the heart of the pharaoh. +So this goes back to well before the Common Era. +This is about 2000 BCE. +So this is a very old idea that's been elaborated over the centuries as science has developed. +And in the 90s, a fairly big change in how this idea was thought about occurred. +In classical physics, reference frames were always thought of as mathematical abstractions. +And an observer and a reference frame were more or less identified by Galileo. +And that identification is carried forward into Einstein, who thinks about observers, for example, sitting on photons, and observers in spaceships, and observers riding on railroad trains, and this sort of thing. +And always that just means think of a reference frame that's doing something or other. +So a reference frame in classical physics is essentially just a coordinate system. +And these abstract coordinate systems can be employed in thought experiments. +And so they are a way to illustrate classical intuitions about motion. +And in the 90s, as quantum information theory developed, it was realized that a reference frame actually consists of a physical object. +I mean, even in the picture from the Egyptian tomb, the reference frame is a physical feather. +And that means that the reference frame itself is described by quantum theory. +It's some quantum system. +And if you think about reference frames that are familiar, for example, meter sticks for measuring distance, or clocks for measuring time, or gyroscopes for batteries, +or even the Earth's gravitational fields, which typically defines up for us, +those are all physical objects that can be described by quantum theory. +Well, this was an important insight because it led to the realization that reference frames as quantum systems are unique. +And uniqueness in quantum theory is very strong. +There's the no-cloning theorem, which states that you can't reproduce, exactly reproduce, any unknown quantum state. +And that applies to reference frames because they encode what Bartlett et al. in that paper called non-fungible information, +which just means information that can't be completely represented in any finite bit stream. +So unless you know the state of a reference frame, you can't clone it, since these reference frames encode more than any bit strings worth of information in their quantum phases. +So they're unique, and they're unique, and their states are unknown. +And the consequence of this is that if Alice wants Bob to use her reference frame to make a measurement, she can't just send him a description. +She has to actually send the reference frame. +And this has an important consequence for physics because it makes all replicate measurements approximate, +because observers obviously aren't using exactly the same physical objects as their reference frames. +But there's a deeper implication of this, which is that Alice's ultimate rock bottom reference frame is actually implemented by her body, by her brain. +And using any external reference frame requires having this internal reference frame that makes the external reference frame meaningful and usable. +So for example, this thing is a piece of a meter stick, a third of a meter stick. +But for me to use it to make measurements, I have to actually have some concept of distance. +And that concept of distance is implemented by my nervous system. +And that implementation is my reference frame for length. +And it's what gives meaning to any external measurement of length. +It's what makes my measurements of length actionable by me using my effectors or action capabilities, whatever they are. +So my internal reference frames are what make information I get from the world actionable. +And in Bateson's sense of differences that make a difference, actionability is the core of semantics. +Something is meaningful if I can do something about it. +So this idea that we all contain embodied reference frames gives physics a semantics by giving observers a semantics. +And this is new. +This is an idea that never really appeared in physics before, that measurements automatically had to have some kind of semantics. +But reference frames give semantics to measurements. +So here's an example of an embodied reference frame. +This is the reference frame that bacteria use for chemotaxis, i.e. to move toward molecules in the environment that they like, like sugars, and to move away from molecules that they don't like, like acids. +And it's been worked out in great detail over decades of experimental work. +And this is a rough schematic. +The reference frame has three parts. +There's the measurement part, which is conveniently located in the front of the bacterium. +And it consists of a bunch of receptors that probe the environment. +And they can bind molecules, which are called ligands, out in the environment. +And when they bind them, they send a signal into the back of the bacterium. +And it turns out that there's a little miniature reference frame attached directly to these ligands, which is this R and B system, which controls the ligands, not the ligands, the receptor sensitivity. +Then the back of the reference frame is the action component. +And it consists of a set of molecules that control a motor, a molecular motor, that derives the flagella on the bacteria. +And the flagellum allows it to swim forwards or turn in some other direction. +And depending on how it spins, the creature either goes in the direction of something it likes or it stops and turns around. +And the stuff in the middle is a little molecular feedback loop that represents the amount of ligand in the environment that the bacterium expects to see. +So this is the middle of the reference frame, and it defines the default value for what the bacterium is observing in its environment. +And if you're thinking in terms of the free energy principle or Bayesian inference, then you'll immediately see the concentration of this molecule is encoding a prior probability, the bacterium's prior probability for what its environment is like. +And the zero point, the default value of that encoding is the ligand concentration in which the bacterium does nothing. +So a QRF, in general, can be characterized as an expected value, which is here the concentration of this molecule, key Y phosphorylated, plus a bunch of computation, which is all the rest of the things in this diagram. +So this is a very simple QRF that's embodied by a bacterium. +Here's a more complicated QRF that's embodied by eukaryotic cells, so more complicated cells like our cells. +And this is the reference frame that makes decisions about whether cells are going to divide. +So on the top, you see the cells' membrane, and there are all sorts of receptors sticking out of it that detect hormones and neurotransmitters and growth factors and this, that, and the other thing, many of which are produced by other cells. +And those receptors communicate, and those receptors communicate via a lot of chemistry with the nucleus, which is contained in another membrane, and so serves as, in a sense, the outside upon which the reference frame is acting. +And what this reference frame does is control the transcription of large numbers of molecules from DNA. +So it basically controls what proteins the cell is making. +And those proteins end up doing something, like restructuring the cytoplasm and actually driving cell division. +And the stuff in the middle of this reference frame, which I've labeled the integrative center, are three molecules called phylogen. +They're connected by positive, by bidirectional feedback loops. +And they basically negotiate the default value of the environment at which the cell does nothing. +So they're the middle of the reference frame that represents the cell's expectation values about what's going on in its environment. +So you can imagine now what reference frames encoded by our nervous system look like. +They're extremely complicated and exist at multiple scales, from the scale of molecules in cells to the scale of large-scale connections between parts of our brains. +So how do we represent all this stuff? +There are many ways to do this. +But this is a particularly simple way. +So it assumes that whatever's in the middle, which is here called c prime, is the most general thing that the inputs can give information to, and the most general thing that can give information to the outputs. +And most general here has a category theoretic definition. +So it could be stated in a way that's equivalent across many different specific mathematical forms. +But if you're familiar with neural networks, this will look to you a lot like the skeleton of a variational autoencoder. +And that's basically what it is. +And I don't want to deal with this formalism except to say that it exists and is very convenient, except to say that we can attach this kind of formal diagram as a representation of a QRF. +So we can attach this to our representation of the boundary between two interacting agents. +And we can use these formal structures to drive the measurement operators that actually implement physical changes on this boundary. +And I'll represent this whole business with just a triangle labeled Q for some QRF that's attached to the boundary. +And we can use this simple notation then to investigate what sorts of structures one gets when one starts to think about systems as implementing specific QRFs. +And here's the structure of the minimal system that we can really think of as an interesting observer. +And this is a system that has some QRF that I've called E that looks at its environment, the part of its environment that it can actually observe and act on. +And it defines a sector of the boundary, which mathematically is just the image of this QRF E that the system can manipulate. +So this is the part of the boundary. +So this is the part of the environment or the part of the boundary that the system actually sees and can do something about. +And for this observer to be interesting, it needs to be able to detect the difference between what it sees now and what it saw just then so that it can decide whether to take some action or other. +So it needs a memory. +So it needs a memory. +And that requires another QRF that has some allocated piece of the boundary. +Recall, the boundary is where classical information gets written in this purely quantum theoretic formalism. +So it needs a memory sector that that QRF could act on. +And then in order to do this computation and specifically in order to write information to the boundary, in order to actually flip bits on the boundary, it needs free energy. +So the rest of its boundary gets devoted to extracting free energy from the environment and exhausting free energy, waste heat, in other words, into the environment. +And so it has this nice little thermodynamic cycle that drives everything else. +And finally, in order to tell now from just a second ago, the information in memory, it needs a clock that marks the information in memory as not current. +So as past information. +And if it can store multiple memory records, it needs to be able to count and say, this piece of remembered information is actually more recent than that piece over there. +So it has to implement a little internal clock. +And that clock is its internal time QRF. +And we've seen that QRFs act on boundaries. +This clock acts on a boundary. +But it doesn't act on B. +It acts on an internal boundary that's essentially the boundary between it, the clock, and the points at the ends of these QRFs that actually encode information that's either read from the environment or written to the environment. +So already this minimal observer has a little bit of hierarchical structure. +Which is that its clock has to, in some sense, be higher up in the hierarchy than the parts, the QRFs that interact directly with the environment. +So if you have a structure like this, you face a metabolic trade-off. +For a fixed amount of energy that you can extract from the environment, you can allocate it either into learning more about the environment or remembering more about what you learned before. +And so any system has to make this decision about how they're going to use energy. +Are they going to devote it to learning more or remembering more? +And your laptop makes that decision. +Memory is cheap for your laptop. +Your brain makes that decision. +Memory is much more expensive for your brain. +And all cells have to make that decision. +And the universal solution to this trade-off is coarse-graining memory, i.e. remembering things in less detail than you saw them in the beginning. +Ander, you have a comment? +Yes. +I wanted to ask a question now before it gets too late. +So two or three slides ago, when we were looking at the diagram of the cone-cocoin diagrams, I was wondering if the classifiers identified with measurement and preparation correspond to sensory and action degrees of freedom in the earlier picture with the bacteria. +And the classifier in the middle, would that, in that picture, would it correspond to the default chemical concentration? +Yes, exactly. +Okay. +Perfect. +Yeah, that crossing point in the middle is specifically where the default is encoded. +Thank you for pointing that out. +Okay. +So let's follow the consequences of this metabolic issue. +If you have a system that's a little bit smarter than our minimal system that can actually detect and measure the states of objects, it has to do even more work because it has to encode even more QRFs. +So think of measuring the state of some object like a voltmeter in a laboratory. +If I tell you to go into the lab and tell me what the voltmeter is reading, then the first thing you have to do is find the voltmeter. +And your process of finding the voltmeter is independent of your process of reading what it's actually saying on its dial. +So in order to measure the state of an object, you have to have two QRFs. +One that lets you find the object and then the other that lets you measure the state of interest of the object. +And we call these the R sector for reference that lets you find the object and the P sector for pointer because old instruments used to have pointers on dials that read analog numbers. +But what thinking about object identification tells you is that every object that you can identify leads to having to sector both the informative part of your boundary and the memory part of your boundary into two new sectors. +One to store information about object identification. +And one to store information about the states of those objects. +So this requires much more metabolism. +And again, the solution is coarse-grained. +Typically what we coarse-grained is our memory of the object that we've identified. +So what does this do to systems? +Because making measurements costs energy and recording memories costs energy, we have to make decisions about what QRFs to use at what time given our finite energy budgets. +We can't look at the world with all of our measurement capabilities at the same time because it's too expensive. +And now we see the effect of a theorem in quantum information theory, which is that if two QRFs can't be deployed simultaneously for whatever reason, that corresponds to them not commuting as operators. +And if they don't commuting as operators, they have to be implemented by compartments that are sufferable. +So compartments that only communicate classically. +So you end up looking like this diagram where there are a bunch of components that do, um, +the consequence of this theorem is that as soon as one has too many sensory and motor capabilities to run all at the same time given your energy budget, +the amount of memory of the same time given your energy budget. +So you end up being compartmentalized into compartments that only communicate classically. +So you end up looking like this diagram where there are a bunch of components that do, um, sensory motor loops or memory. +one or the other, and they can only communicate classically, and they're controlled by some sort +of attention system that selects which ones to activate, depending on the amount of free energy +that's available. And clearly this clock function becomes part of this control system. +So the physics alone of any system that's acting as a fairly, that executes a fairly complicated +interaction with its environment, gives us this hierarchical compartmentalized structure +without having to add any biological or cognitive science type assumptions at all. +So we can go in three directions from this. We can go deeper into the biology, which we're going to do +in October. We can go deeper into the physics, which we're going to do in August and September. +And we can go deeper into this issue of alignment, which is what we're going to do today. +So there are four possibilities if you think about two agents that have QRFs that are interacting +across this boundary. And the first one is trivial. If the B agent, the agent on the right side, +doesn't actually employ any QRFs, if it just interacts with the boundary thermally, +then there is no alignment. You're in a trivial situation where A can measure something about B, +but not in return. If the B agent deploys a QRF that's much broader in scope than the A agent, +what we show here in panel B, then the A agent can learn something about the B agent, +that misses most of what the B agent is doing. In panel C, we have agents with QRFs of the same +size on the boundary, but they're not lined up correctly. So each one misses part of what the +other is doing. And finally, in D, we have the optimal situation where they have QRFs of the same +size, and they're actually lined up. So each sees what the other is doing, at least on this sector +of the boundary. And since the agents have to allocate some of the boundary to free energy supply +and can't look at that part, then they can't tell what's happening across their entire boundaries. +They can only tell what's happening in this sector, where they can actually deploy a QRF. +But let's now look at this interesting situation, Situation D, where the QRFs are aligned between A and B. +As I said, that's the optimal channel for A and B. +And if we think about this, we can write the variational free energy, or the uncertainty, +or the prediction error, for A, for the QRFx, as the action that it sees, i.e. B's action, +the difference between that and what it expects B to do, so what it can predict about B's action. +And this is just the notion of free energy from the free energy principle, except translated into +the situation where we have two QRFs. And A can measure what B has done, +but it can also use its QRF to predict, because it's predicting the very state that it's going to observe next +of the same sector that B is using. +Now, if we look at this diagram, we can see easily that the way to minimize the variational free energy, +and therefore maximize predictive capability, +is for these two QRFs to actually be the same. +So if A is doing the same thing to the boundary that B is, A can predict B's next action perfectly, +because B is going to do whatever A is doing. +So what the free energy principle tells us in this case is that A and B, +as they minimize their variational free energies, will approach having identical QRFs. +Well, again, we have a theorem that tells us that if these QRFs are shared, A and B are not separable, +but entangled. +And this is also fairly easy to see, because it depends on this notion of non-fungibility. +If X, A, and X, B both encode non-fungible information, which is quantum systems they do, +their states aren't clonable, so they can't be copies of each other, +because one can't clone an unknown system. +So they can only be identical if they're actually identical systems. +This is exactly the same point made earlier about Alice having to actually send her QRF to Bob +if she wants Bob to use her QRF. +So minimizing VFE for quantum systems is becoming entangled. +So the classical free energy principle, which tells us to minimize VFE, +actually drives quantum systems toward entanglement. +And that's why it's a special case of the principle of unitarity, +which just says that any isolated joint system is going to asymptotically approach entanglement. +So this raises an obvious question, which is why all of us aren't entangled with our environments all the time, +and entangled with each other. +And the answer is that entanglement is really hard for complex systems. +And this illustrates the sense in which the FEP is moving systems in two different directions at once. +The FEP is driving systems to maintain their separability, +but it's also driving systems toward entanglement. +And as the system becomes more and more complex, +the difference between those two outcomes becomes larger and larger. +So recall from last time that the condition for separability +is that the two systems have dimensions that are much larger than the dimension of their boundary. +And this is the condition that allows them to have internal states, +to have states that aren't located on their boundaries. +And clearly, to formulate the free energy principle, you have to have internal states. +So if you have a complex QRF that's doing a lot of computation, +then you need a lot of internal degrees of freedom to implement that computation. +So you need a lot of internal states. +So you have to have large dimensions. +So complexity favors separability. +As a system becomes more complex, it has to have more internal states. +So it has to be, in a sense, more separable from whatever it's interacting with. +Now, think about this classically. +If you have a complex system that's doing a lot of computation, +then it requires a lot of free energy, +and it has to dissipate a lot of entropy in the form of heat, typically. +But something, some form of energy transfer that its environment is going to see as noise, +as not informative. +So a complex system is working very hard to predict its environment's next state. +But by working very hard, it's dumping a lot of entropy into its environment. +And putting in that extra entropy makes the environment's behavior even harder to predict. +So the more you work at predicting what the environment is going to do, +the more you yourself make the environment's behavior less predictable. +And this is the paradox of the FEP, which is not really a paradox. +It's just what the physics does. +So we can think of the FEP by driving systems to increase their distinctness from their environments, +or to self-evidence, as Carl Fristin calls it, +as driving this increased complexity that both makes predictions better +and makes predictions harder by transferring entropy into the environment. +So what this tells us is that the FEP by itself +will drive the development of systems with larger and larger complexity. +And this clearly has enormous implications for evolutionary theory, +which we'll talk about in October. +But I want to emphasize here that this is just the physics. +It's just what the FEP does when it's forced to maintain a boundary. +And before stopping here, +I want to consider one aspect of this in particular, +which is language. +For example, human language, +but any kind of language connecting any systems +that are communicating with QRFs that are fairly close. +And language is a complex QRF +that we basically share. +Otherwise, things like this interaction would be impossible. +But we also remain distinct systems. +So the question arises, +how do we manage to do this? +How do we manage to largely share a QRF +while escaping becoming entangled? +And this is a question that's largely been ignored in physics +because physicists have tended to ignore language +and just assume that observers can communicate classically. +So, for example, in talking about a Bell-EPR experiment, +which is a canonical experiment for measuring entanglement, +the observers have to be able to share their results +or they can't observe entanglement, right? +The observed result is a bunch of observational statistics +that violate Bell's inequality. +And those statistics can only be computed +given the results that both of the observers obtain. +So the observers have to share a language +to encode these results. +So how come the observers aren't entangled? +How come they are still separable? +What we end up appealing to +is some sort of common cause in the past +to explain why they share a language. +They both learned it from their culture. +So their culture is some sort of common cause. +But all such assumptions are fine-tuning assumptions. +So they all have basically the status +of the value of something like the fine structure constant +without which we wouldn't have atoms, +and so we wouldn't have us. +This value has to be within very small limits +of the value we actually observe. +And that's a case of fine-tuning. +The universe seems fine-tuned in a way +that allows us to exist. +And if it wasn't fine-tuned in that way, +we wouldn't exist, and so we wouldn't observe anything. +But fine-tuning assumptions are a problem. +They're a pointer to something we don't understand. +And so I just want to leave you with the point +that we have to be very careful +when talking about interacting observers. +So in talking about two components of some system +that jointly interact with some environment, +but also communicate classically. +And if you think back a few slides, +that's exactly the structure +that we've shown that the physics forces +any complicated system to have. +So next time, what we're going to focus on +is this very question +of how we describe observers +that are both interacting with the quantum system +and communicating classically in some language. +So that session will be in August. +At the end of this month, +there's a discussion of this session with Ander. +And I encourage you all to +use the interactive Q&A to ask questions. +You've been asking very good questions +that I hope I've been managing to answer. +So thank you very much. +And Daniel, over to you. +Cool. +Well, we have a few minutes for some questions. +So maybe, Ander, +if you want to bring up any questions first. +Oh, unmute, then go for it. +I was curious if you could say any words +on the slide where you had all the possible alignments +of QRFs and what you meant more by interacting thermally. +So when you have a trivial agent on one side, +what exactly does it mean +to say that you're interacting thermally? +If you think about a system +that is only interacting thermally with its environment, +you end up thinking classically about, +for example, an ideal gas +that's in some sort of container. +And some of the molecules are very near the surface. +And their motions depend on the motions +of all the other molecules that are inside. +So there's a set of internal states +that interact randomly +with the external states that are near the boundary. +And so those external states +interact randomly with the boundary. +But notice in thinking in this way, +you've imposed a space-time embedding. +So you've actually used space +as a way of separating the internal states +from the boundary states. +And a system that, +if you drop this assumption of a spatial boundary, +then a system that's only interacting thermally +with its environment +is going to look like a system +that doesn't have any real internal states. +It's going to look like a system +that just has boundary states. +And so a pure thermal interaction +is going to start to look like entanglement. +So we're actually only really interested +in systems that aren't interacting +only purely thermally with their environments. +We're interested in systems +that have enough internal structure +that they're using that thermal energy +to do something internally. +So to give their internal dynamics +some particular shape +so that their internal dynamics +is actually circling around +some well-defined attractor +that requires some energy +to maintain itself. +Yeah, and continuing with that question, +I saw in the paper, +free energy principle for quantum systems, +that if you're in that situation, +when both Keras are non-trivial, +but perhaps B's is larger than A's, +A's perceives non-local hidden variables. +So I wonder if there is a way +to relate those +by taking the limit of making A small +and recover the thermal interaction. +Presumably, if you let A become small, +you should recover that thermal interaction, +so to speak, right? +Yeah, that's a very interesting comment, actually. +So if you're a system +and all of your environment's variables +are effectively hidden to you, +then you can't do any information processing +because you don't have any access +to any of your system, +your environment's variables. +So yeah, your interaction can only be thermal. +I think that's correct. +Right. +Thank you. +I guess, yeah, go for it. +If I may say, +let's just so we, +since we're in this topic +and we don't have to do context switching that way, +on the other hand, +you have noise, right? +In one of those other situations, +I can't recall which one, +I guess, +I guess in the reversed one, right? +When A square F is larger than that of B, +so it encompasses more of the boundary +degrees of freedom, +those that are not overlapping, +A perceives as noise. +Is that, that's correct, right? +Yeah. +Yeah. +A will be looking at B's thermal sector, +B's free energy sector. +Right. +So A may be interpreting, +in a sense, +what it's doing as informative. +And A may even be seeing +what B is doing as informative. +But, +so, +so noise for B may be signal for A, +depending on, +on what, +how A's QRF is structured. +But, +A is, +is seeing +more information +than B is trying to send it. +So, +it's, +it's essentially the, +the reverse of the hidden variable case. +There are +variables +in the other system +that aren't telling you anything +about what the other system +is actually doing. +That, +that's very interesting +in terms of +reading too much +or too little +between the lines, +connecting too many dots +or connecting not enough dots. +But these challenges +where noise +for one +quantum cognitive agent +might be signal for another +and vice versa, +those are the challenges +that we want to address +because to +have a physics +that doesn't recognize +those situations +is +kind of taking the view +from nowhere +or not really addressing +the complexity +of a cognitive engagement. +Yeah, +it's, +it's describing +interactions +between +effectively systems +that don't share +the same language. +I have a question +about the free energy principle +here. +Only in the last +several years +has the quantum +free energy principle +relationship +come into light. +So maybe +could you just share +a little bit about +what incremental +or quantum +advances +enabled that to happen +or what degrees +of freedom +were pre-existing +or added +to enable +work arising +from perception +and action +in neural systems +to also +so naturally +enter this context +with the quantum +information? +Well, +I think +at least +from +a historical +point of view, +that connection +was made +based on the analogy +between a Markov blanket +and a holographic screen. +so these +are structures +developed in +different parts +of physics +that essentially +serve the same +function +of limiting +the information +that one system +can have +about another system. +So, +it's that +functional +analogy +between +a Markov blanket +defined +classically +and a holographic +screen defined +in quantum theory +that +made it seem +obvious +that we could +translate +the free energy +principle +into a quantum +theoretic principle. +Yes. +In Livestream 40, +we talked a lot +about this, +but I think +that remains +one of the +most remarkable +advances +inside or +outside +the free energy +principle +literature +based upon +this functional +acknowledgement +of boundaries +and the +conditions +that they +entail, +which again +are the +challenges +that we +want to +address +lest we +ignore +boundaries. +Right. +Yeah, +precisely. +I mean, +the free energy +principle +is all about +what happens +when you have +a boundary. +The waste +heat, +is this +the actual +thermal exhaust +from our +CPU, +or how can we +think about the +heat being +released +topologically +rather than +geometrically, +like the +location where +it's released? +What does it +mean to +consider +physical +processes +topologically +that we +might otherwise +just use a +geometric way +of thinking +about? +Well, +you basically +put your finger +on it, +and that you're +not worried +about the +spatial +difference +between +the geometric +spatial +difference +between +where one +kind of +information +flows through +the boundary +and where +another kind +of information +flows through +the boundary. +And you +can think +about this +in Markov +blanket +terms. +And the +Markov +blanket is +just a +network, +and some +of the +causal +arrows +flowing +through +that +blanket +network +are +effectively +transmitting +noise. +And you +can think +it in the +same way +with respect +to a +holographic +screen. +If you're +doing a +certain amount +of information +processing, +and this +brings up +the whole +issue of +how efficient +is a +system +in +its +information +processing, +how close +is it +to the +land hour +limit, +basically, +can it +extract +more +energy +from some +bits +than it +has to +use to +modify the +states of +other bits. +things. +That's +what this, +if you will, +metabolic +question is +all about. +In the +case of an +organism, +is it, +can I get +more energy +from some +of the +chemistry +that I +have access +to than +I need +to use +to modify +the chemistry +that I need +to modify? +And if +not, +then life +stops, +right? +Everything +grinds to +a halt. +Now, +can the +heat +exhaust +informationally +be useful +for another +agent somehow? +I mean, +can't we +recover heat +in a +steam engine +and use it +to do +useful work? +not to +the point +of a +perpetual +motion +machine, +but how +can we +recover +exhaust, +and then +what does +the +perpetual +motion +machine +look like? +Well, +the environment +is the +system sitting +over on the +other side +of the +boundary +that is +absorbing +whatever +waste heat +is generated +by the +system, +let's +call it +Alice, +that we're +interested in. +So that +environment +may be, +and Alice +can't tell +this, +but it +may be +divided up +into lots +of different +compartments, +some of +which are +basically +sucking in +Alice's +waste heat +and doing +something with +it. +So this +is really +a question +about how +is the +environment +structured, +and what +operations +does the +environment +execute +with the +input that +it gets? +How does +it interpret +that input? +it? +And what +would +an actual +motion +be? +This gets +back to +this question +about how +do we talk +about multiple +agents that are +communicating with +each other? +Right? +Instead of +talking about +an environment +that we don't +characterize at +all, how +do we talk +about an +environment that +includes some +specified or +preferred systems +that are +treated as or +known to be +other agents? +So let me +get back to +your question +about a +perpetual +motion machine. +Notice that +in classical +thermodynamics, +there's always +the assumption +that there's a +big environment +out there, +you know, +the rest of +the universe +or whatever, +the whole +environment that's +not on Earth, +that is an +arbitrarily large +source of, +or it's not +an arbitrarily +large source of +free energy, +but it's a +large source of +free energy, +and it's a +large sink +for entropy, +a kind of +arbitrarily large +sink for +entropy. +So the +second law +of thermodynamics +says the +entropy of +this entire +system, +counting the +whole surrounding +environment, +doesn't decrease. +So it can +increase and +still satisfy +the second law. +So this gives +you the +classical, +you know, +19th century +heat death of +the universe +idea. +energy, +and so +your proposed +machine is +sitting +somewhere +embedded in +this gigantic +environment that +can provide +some finite +amount of free +energy, but can +exhaust effectively +an arbitrary +amount of +entropy. +entropy. +So any +machine has +to dissipate +entropy as +it runs to +fill up this +kind of +arbitrary +sink of +entropy that's +the classical +rest of the +universe. +Now compare +that to the +picture in +quantum theory. +The picture in +quantum theory is +the universe by +definition is an +isolated system. +So its +total +information +content doesn't +change. +Right? +It's evolving +unitarily. +Unitary +operations preserve +information. +So they're just +moving information +around. +So from +any, +from the +point of view +of any local +system, +the local +system's getting +information from +its environment, +so it sees its +environment losing +information, so +gaining entropy. +What this tells +us is that +entropy is in +fact not a +globally +definable +quantity. +Entropy is +always defined +relative to some +observer, some +division, some +boundary. +And this is +another +kind of big +change in the +last few +decades and +has led to, +for example, +entropic +definitions of +time. +My local +external time +arrow is +whatever direction +I see entropy +increasing the +most. +so again, I +would refer to +people like +Rivelli and +Max Tegmark +who developed +many of these +ideas. +Awesome. +All right. +Thank you, +Chris and +Ander. +Looking forward +to seeing +everyone's +questions +submitted and +their preparation +and participation +and measurement +in the +discussion +section. +So, thank +you. +Until next. +All right. +Thanks again, +Daniel. +Thank you, +Ander. +We'll see you +later. +Bye. +Bye. + + + + + + + + + + + + + + + + + + + + + + + + + + + + +Thank you. + +## rOoTkgLI1Z8 + +## rOoTkgLI1Z8 + +Hello and welcome everyone. It's June 3rd, 2023. We're in the first discussion section for Physics +as Information Processing course with Chris Fields. So welcome to all who are here on and +off camera. Really exciting to see so many people joining for this participatory discussion section. +So in this first interval, we're going to go to anyone who raises their hand and it would be +awesome to hear from everyone who would like to share what excites them about what they heard in +the first lecture. How has their background and journey led to them joining? What are they hoping +to learn and do in this course? So first hand, we'll get to go. Maybe, Ander, you can actually +start first, just speaking as a normal participant. And then meanwhile, anyone is welcome to raise +their hand. That sounds good. So I am a mathematician by training, +professor of physics. But I've always been interested in the more physics side of things. +And in particular, you know, I followed Chris's work pretty closely since probably 2019. +And I don't know, I'm just very interested in the whole ontology of quantum measurement, what it means. +I feel like, you know, very often when it's taught in physics classes, it's hand-waved away. And even +though I myself subscribe to the philosophy of, you know, shut up and compute, sometimes, you know, +not everything. You need to do a little bit of philosophy maybe. And this may be one of the things +where you need to do it. And you need to address it from first principles, right? And I think Chris is +interested in that. So I'm just here to learn more. And yeah, that's basically it. +Awesome. I'll just go from the top left, because I'm not necessarily watching like what order, but +Ross first. +Hi, I'm interested in the kind of overlap between cognitive science and physics, +physics, and seeing if there are any panpsychist implications of the observation that physical +processes seem to follow the same kind of principle that we see applied to brains and cognitive systems. +Awesome. Thank you. Avel? +Hi, I'm Avel Guénin-Carluge. I have mixed backgrounds, mostly physics and cognitive science. Right now, +I could be qualified as a philosopher of physics and cognitive science. And I'm interested in +basically formalizing participatory realism, the idea that the observer participates in creating reality. +I did some work in formalizing that in the case of social norms, and things like language. And I +met a block based on my lack of knowledge of quantum physics and the mathematical tools +that I used for that. So I'm here in a course to get those formal skills. +Awesome. Thank you, Avel. Blue? +I think Alex was first. +Okay. Alexi, go for it. +Oh, sorry. +Hi, I'm Alexi. I'm a clinical psychologist in Maryland, in the States. I work with patients, +do psychotherapy mostly. My interest came through Mark Solms, who is the creator and leader of +Neuropsychoanalysis. And he has been collaborating with Friston on heart problem of consciousness. And +he's now working on AI with emotions. And so he, you know, I blame him, I say it with respect and love +that for introducing mathematics into our field, which was mostly devoid of mathematics. And people +started looking at these partial differential equations and Markov blankets. So I wrote a paper +about a possible extension of their model to include chaos theory. And I just want to understand deeper +concepts involved. And I have some understanding of them, but perhaps not enough. And I'm enjoying +this course a great deal. I mean, I think Chris, among other things, a fantastic teacher, and he is +able to express things very clearly. And it just, I have all, I have goosebumps. So I'm very grateful to +everybody. And thank you so much. Cool. Blue, then Francesco. +Hi, everyone. I'm Blue Knight. I'm with the Active Inference Institute. So I've been +tracking this work of Chris's, you know, connecting quantum information and active inference for, I don't +know, a couple of couple of years now through through that thread. I'm super interested, actually, in +the aspect of space time arising from communication. I think that that's like fascinating to dive into and +how it connects to the like Buddhist and maybe more Eastern religion, like concept of dependent +arising, like how like the interdependence of all things leads to their existence. And so from that +aspect, I think that that that's like the what I'm most interested in exploring. My background is +neuroscience, but I've also been studying and practicing Buddhism for like 25 years. So yeah, +that's it. Thanks. +Awesome. Francesco, then Corby. +Thanks, Daniel. Ciao, everyone. Super happy to be here. I also have a mixed background because I +studied cognitive anthropology and philosophy of science. And currently, a few months ago, I started a PhD in +artificial intelligence in an eighth application in the educational field. And yes, I'm actually super excited +about the all the science implications that Chris Fields mentioned in the first lecture, the connection, I have +no background in physics, but the connection between the first principles of physics and free energy principle +and scientific cognition in general is what is really getting me excited in these days. Thank you. +Cool. Corby, then anyone else who raises their hint? +Hey, yeah, great to be here. My background is in quantum engineering. I've always been motivated by +figuring out what the next big thing in computing was in high school. Morph's law was sort of coming to an +end. And it seemed like a really big opportunity to figure out what the next big computing architecture +might be. So I did academic research and no more computing, quantum computing, and quantum sensors. +And I felt like I was a little bit too early for all those things. So I went into cloud and cloud gaming +and streaming AI services. But I'm still extremely motivated to figure out what the next big thing is +in the holographic principle seems like a good foundation to explore that. +Awesome. I'll raise my own hand, then anyone else is welcome to go. So I'm Daniel. I'm a researcher at the +Institute and have known Ander for some years and learned about all kinds of stochastic matrices and +how the math and the quantum were linked. And we had talked and joked about learning from Chris +fields for a long time. So when the opportunity arose to enact it and extend it and open it, +we were very excited and also don't have much of a formal background in physics outside of working +with Yakup and Ali on Livestream 49. So it's just going to be a great time to learn and see where it all goes. +So anyone else wants to say hello in this kind of opening section? +Okay, well, continuing forward, of course, please just feel free to raise your hand anytime you'd like to +jump into the stack. Ander, for several minutes or however you see, could you reflect on lecture one? +Where did Chris enter? How did he proceed? And where did the lecture take us? +Can you hear me? Yeah. Thank you. Good. Go for it. +All right. Let's take a look at the slides from last time. So I think if we were to summarize this +whole lecture, basically what Chris was trying to say is that there are many hints coming from different +directions pointing at the same thing, right? Pointing at the fact that information will have a central +aspect in physics in the future, but also possibly biology, right? So the earliest hints came from +statistical mechanics, if you wish, but also quantum mechanics, right? And here we need to make a +like an important but subtle distinction between what it is the uncertainty principle, right? Because, +you know, we think of that as the informational part, right? Like a limit on the resolution on information. +And what is the observer effect? Just to make that distinction clear, I think Chris is talking about +the observer effect. The uncertainty principle is an artifact of the wave nature of, you know, quantum mechanics, +right? You can get it from properties of the Fourier transform, Cauchy-Schwarz inequality. But at the end of +the day, it's what I was saying earlier, right? What is the basic picture of measurement, right? And that's +sort of what Chris with Glacebrook and Marciano is trying to address. And the first two hints came from +from quantum mechanics around 1900 and statistical mechanics a few decades earlier. +So then there's, you know, more stuff, right? Like, obviously, physicists resisted to inertia of hundreds +of years of a totally objective, you know, observer independent universe, but the hints continue to pile on. +So then there's some stuff coming from math, but perhaps the most interesting is, again, you know, how +all those paradoxes from statistical mechanics were made sense of with Landauer's proposal. +And then as physics got more fancy past, you know, QFT and stuff, you know, people started thinking about +quantum information more seriously. So we had Bell's theorem and all these proposals by Deutsch and +Wheeler and so on. And I should say that this is not a fringe idea in physics. The paradigm of it from +QBIT is very much mainstream. I mean, Google it, a lot of this is just think about it, you know, +I don't know how many of them, you know, talk about it openly, but you know, it's been a widespread +folk intuition, so to speak, that, you know, information will have this sort of central role. +And the last sort of hints pointing in that direction of information, having a central law came +on the one side from high energy stuff. And that's the holographic principle is, it originally arose in +the context of black hole thermodynamics, but then, you know, it was extended to, to, you know, what's +called the ADS CFT correspondence. And the very most recent stuff was, you know, it's basically +Christians were having essential law came on the one side from high energy stuff. +And so I can speak a little more loudly. It's all good. Continue on. +Yeah, I thought there was a question. Someone was just unmuted. It's all good. +Okay. Yeah. So, so that's basically it, right? So let's, there's hints from many directions. And +most recently, you know, the work of Friston and Levin coming from biology, all pointing in the same +direction, right? That we might have to make sense of some sort of, eventually scale-free +information processing for that, for, for what it means to be an observer, you know, sort of all, +all of these things are pointing in that direction. And I think Chris's work is, is addressing that +head on, right? As I said earlier, you know, sometimes, you know, it's, it's good to do computations, +but sometimes you need to take a step back and think more philosophically about what you're doing +and what the bigger picture is. And then maybe after people talk a little bit, I can, you know, +we can warm up to the second lecture. We can, the other windows that I have open here, +some of them are the papers that are to come. +Yeah. Anyone, please feel free to raise your hand on this. Alexei, and then anyone else? +Alexei So I had one thought about this thing, you know, when, when you've mentioned that there's +various perspectives on the same concept and I'm trying to be cautious on the one hand, I think, +again, this amazing clarity and precision with which Chris expressed thoughts is, is fantastic and +enjoyable. On the other hand, it requires simplification and generalization. And what I +observe in my field is take people take a term like entropy and run with it. And then they get very +quickly from thermodynamic entropy to Shannon to, you know, whatever entropy and they talk about sort of, +and that's not the case. I mean, if we go back to Clausius and all the others, there was a +system of postulates and axioms that must be maintained. And if you step outside of that, so like +us organisms were open systems, we're not closed systems. So we should not apply directly second +world thermodynamics. And, you know, I, I deal a little more with Kolmogorov's and I entropy and, +which is not the same thing. And I guess when we very quickly move from sort of thermodynamics to, +you know, to statistical mechanics with Boltzmann to, you know, all the way around, and we assume it is +the same thing, I think we may need to be careful because we are just ignoring the differences and +assumptions. You know, Shannon talked about telegraph. He talked about a receiver getting a message from +a sender, right? And Clausius talked about, you know, gas. So, but, but I do think it's a balancing act. +And there's no way to not have simplification and generalization. I just wanted to say somewhere, +somehow we need to kind of, you know, go back there and just say, you know, entropy is not entropy, +is not entropy. And all these things are kind of, you know, long papers, right? +Yes, Jakob, then Ali, then anyone else. +Um, I, um, I just wanted to, uh, ask for, I guess, perhaps, uh, slightly longer comment from Alexi on +that, what do you see as the fundamental differences in different descriptions of entropy and whether there is +some, I guess, general concept that could be extracted from all of them, such as level of +information or uncertainty about a particular system, even though these specific, uh, formulations of +entropy, they, um, talk about, they were derived from different specific systems and they have different +equations describing them and they mean slightly different things. Do, um, do you think it would +be wrong to say that there is some commonality between them? +Should I answer Daniel or? +Yeah, please. +Yeah, I don't know. +I just think we need to at some point formalize it and, um, and be careful with these generalizations. +For example, uh, when we talk about active inference and Friston, right. You know, when we say living +organisms must minimize, I'm quoting Friston, by the way, these are not my thoughts, the entropy of +sensory states, right. And then people take that idea sometimes in generalize, and they say that +that's what life is minimization of entropy, right. But if you actually look into human brain and measure +entropy, then it goes all over the place up, down and in between. Uh, and so I, I don't think that's +an accurate description. And the reason for that is where we shifted away from the, uh, assumptions. +And, uh, again, so Kolmogorov's in I entropy talks about the phase space and the divergence of +trajectories in phase space and sort of the volume in phase space. While, when you talk about this, +for instance, uh, entropy of sensory states, these are probability distributions. +And, uh, Shannon talked about the uncertainty of the receiver about the message sent by the sender. +Dr. So if you want to extract the same thing, I suppose you can, but, but it's important not to +lose at least the difference between what kind of object are we applying to? Are we looking at +probability distributions? Are we looking at phase space? Are we looking at ideal gas? Where are we at? +And we're just not to say entropy is entropy is entropy. That's not the case. I mean, it's just a formula, +but depending on the context, assumptions and things, it can be very different. +Thanks, Ollie and Ava. +Uh, yes. In regard to Alex's comment, I just wanted to mention the research of, um, people like +Rudolf Hanel and Stephen Serner, uh, because, uh, for many years, uh, they've been exploring this idea of +different conceptions of entropy and, uh, they've come up with at least a dozen different meanings for +entropy and they formulated it, uh, in a kind of generalized framework, uh, but at the same time, +they've also devised, uh, a way to somehow, uh, classify them into equivalence classes and to see whether, +uh, one or more of those, uh, conceptions of entropy can be theoretically derived from, uh, from some +others or they're totally, uh, distinct and different, um, uh, conceptions of entropy. So, yes, I agree that +in, uh, some situations we need to be more specific about what exactly we mean by entropy and information. +Uh, I mean, based on, uh, the question or the situation of interest. +Thanks. Ava? +So just to put in the entropy discussion, I want to emphasize that entropy, uh, PLNP, so it's a function of +probability discussion, which is, uh, I think, I'm not sure what the term is in English, but it's a measure of how big, uh, something is. +Uh, and to, to get that, you have to define a space. And in many cases, it's pretty arbitrary to decide what +this space you're looking at is. For example, you talk, I don't know if it was Alexei, it was Alexei. +Someone talked about, uh, uh, uh, entropy of, uh, fMRI. Well, those are aggregate measurements of neural +activity. This is not the space of all possible configuration. And even if I add something like the +space of all, let's say functional configuration, the state in your own that reflect activity, this would be +another thing than the space of, uh, I don't know, atomic configuration where the specific atom are. So +yeah, you don't have a warranty that when you define entropy, you define it on a meaningful space. +That's not given. So let's, I kind of think there is one entropy and that's it and it's meaningful constructs. +But also we have to, if we agree to that, then it entails that we have to find it, that it's not just +given out there in the data. +Hmm. +Well, certainly being precise in the formalism and in natural language for what we mean by information +and entropy and all these terms is pretty important. I mean, the course we're in is physics as information +processing, but without an understanding of how information is being meant here or what measurements +and transformations were, were performing, what informational operations, then it, it, um, +is a non sequitur. So how do we, how do we go about, um, clarifying or understanding what kinds of +theoretical results in a given domain or like what is shown on the timeline here might apply to some other +area. +Just one thought. Anyone else can raise their hands with, uh, a thought or a question. Otherwise, +Anders, you can, um, maybe flip over to, to select a few of the submitted questions. +Cool. Could you zoom in, um, with control plus a fair amount? +Thanks. So everyone, um, we really appreciate the submitted questions. We've been able to +get 13 answered by Chris so far. And, um, as with the transcripts of the discussions and lectures, +the questions and answers are going to be part of what is published by the ACDIMP journal. So we really +encourage basic questions, advanced questions, just, there's no question that doesn't make sense to +submit. So under maybe just pick one question that you think's interesting to begin with and summarize, +uh, Chris's response. And then everyone else is welcome to kind of add some other thoughts on this. +Yeah. So I'm going to scroll up and down just to sort of sample. Uh, let me start with the first two. +Can you hear me well? Yep. Uh, since perhaps, I think you submitted this as a trial, right? Is that, +is that right? Yeah. So what is information? What is time? Um, and here's Chris's response. I guess +everyone has a chance to look at it. Now the one, if, if I am to summarize this, the one thing I've heard +that, you know, personally, the one I liked the most, um, answer, at least as, as an adage, +to summarize it is that information, um, are differences that make a difference. This, +I might have heard from Chris. I might have heard from someone else. I don't remember, +but that's one thing. Now onto the question of what is time. And I guess this also addresses, uh, +uh, Alexis question of what is entropy. I, so I'll, I'll make a bigger point now. Uh, what is time? +This is partially addressed. And, and I guess the main thing I'll be doing here is rather than look back, +look forward into what's to come in the course. And in particular, I'm thinking about, um, +um, yeah, this, this, these two papers, uh, and in particular, +particular, and I've had a chat with, always Chris about this, uh, this picture over here. Okay. +Can you zoom in more? Yeah. I'll get back to it in a bit, but basically this, this sort of like +kills two birds with a stone, right? This picture, and especially the text around it, +and the title of this paper is the physical meaning of the holographic principle. Okay. +You guys can see here archive 2020, 22, 10, blah, blah, blah, blah. Anyways, this, uh, addresses the +notion of, you know, the time and entropy, because the, the point that's the claim that's made in this +paper is that these are local observer relative definitions. And we can, uh, look into more detail +around it, but basically the coordinate TA time is an entropic measure. So it, it is non-decreasing +and it goes up with the observer measuring, uh, entropy in the system. Because of the symmetry of +the definition, the second law applies to both. So both observer and environment see entropy increasing +in their environment and vice versa. Uh, anyways, from, from this picture, one gets that you have +this internal circle time QRF. So this just is, is a necessary artifact of, of recording information +into a memory. That's, that's the claim made in the paper. And that essentially induces the passage of +the time, at least from the perspective of the observer. Um, it also, I think, uh, Alexei was talking about, +which I totally, uh, understand and agree that this different notions of, of entropy. Now, I suppose as the, +the, you know, I, I would be the one to, to like definitions and so on. Um, but I, but I do think it's, it's, +it's good to, to have a certain amount of flexibility. Um, so I see the same as energy, +right? Like there are different kinds of energy, you know, there's energy mass equivalences and, +you know, if you push it into supersymmetry, you can think of it as equivalent to force also. +But I, I do think there is some value in being slopy or intuitive about, about the notion. So, so the +fact that, you know, entropy, yes, there are different definitions. I, I don't, I don't see it quite as a +problem. You know, one can explore here around this, this page, there is a definition of, you know, +local entropic time and yeah, one can think of it as, you know, face space or one can think of it as, +you know, coding theory, a la Shannon, but it is at the end of the day, I do think it is true that both +of them are hinting at the same direction, just like, you know, you may have different kinds of energy, +but, you know, they're not unrelated concepts. And, and yeah, I hope one day we get very precise +definitions on everything, but, but for now I think it's, it's good to speculate. +Okay. Thanks. +So we can, there's a question. +Do I see? +Yeah, Corby and then anyone else on this? +Yeah, quick question. So I haven't read that specific paper, but is there any work on the +frequency or the rate at which one could read and write information on the boundary? +Yes. Well, I, I don't, I don't, I can't think of the, the paper exactly, but this is lower bounded. +I think Chris talked about it. If not in the last lecture, I, I think I've seen a talk of him on +YouTube. It's lower bounded by the uncertainty principle, right? So, so for instance, plants, +Planck's constant is in units of action, which is energy times time, right? So Landauer's principle +says you need a certain amount of energy to, to erase information and, and you have time, right? +And then you have an energy time uncertainty principle, putting those things together, right? +You have any minimum amount of action, which is, you know, literally the minimum, um, the action is +measuring what I said earlier, right? Like differences that make a difference, the minimum difference that +you can make. So the minimum difference to, you know, erase a bit or rewrite it on the boundary, +so to speak, right? Uh, so yeah, that would be given by the uncertain relation energy times time. +It's I, I don't know exactly what the figure is, but it's obviously very small compared to everyday life. +Got it. And would, is there work on the upper bound? Like for example, would it be the number of +degrees of freedom on boundary? No, I don't think it's bounded from above, but, but obviously, you know, +then we, we do want to buy it from above for technological reasons, if you wish. But I, I think +that is a, an engineering problem, not, not one of physics. +Um, Andre, few, few questions on this figure. Um, what are Y and E? +Uh, B, the blue oval, as Chris has pointed out is going to consistently be the screen, +the blanket. And so we often associate that with, uh, at least some blankets include +these information gathering and utilizing systems. So what is happening internal to the agent here? +What kind of computation is happening? And is this something that is actually proposed to be occurring? +Or is this a schematic or a map of how we could think about something occurring? +So I think this question is going to be better addressed by Chris himself, probably in the next +lecture. Um, for now, what I can say is I can reiterate what I was trying to say earlier, which is that, +what this addresses, you know, this basic measurement picture, so to speak, is the observer effect, right? +Like this notion that you cannot observe something without, without disturbing the environment. +So if you then try to make sense of this by constraining, you know, from first principles, +Landauer's principle and so on, uh, basic requirements on information processing from the point of view of energy, +then you get that this boundary is constrained in some ways, right? This Y is what's called the memory +sector, meaning that you don't just act on, on, on the immediate information that you perceive. You may +also act on, you know, stuff that's stored that you have perceived before. And then sure, E, the red part, +is what you're perceiving right now, the environment. Uh, but actually, I don't know if I have it in here. +Uh, but this is further broken up into other parts, so not schematically here. Let me try to find it over here. +Now, I don't know. Actually, bear with me for a second. I think you might be here. +That's the same picture. +Anyways, I, I'm not, I don't want to waste time by trying to find it, but +back to the original picture. Yeah. Sure. It is what you're measuring. And again, Chris will talk +about this in more detail. Uh, but from first principle assumptions, +they deduce that you're going to have to, you know, deploy your finite resources for gathering +information in clever ways on the boundary, right? So you're going to have to deploy some resources for +memory, some resources, uh, as a free energy, uh, as a heat sink, right? So you're going to be +processing information. So you're going to be dumping, you know, so-called low quality stuff +into the environment. You're going to be dumping heat. And then older than the memory, you're going +to be focusing on what's going on right now in this environment. And what I wanted to say is that it can +be further broken up into what's called pointer and reference sectors. So you can imagine this +E sector here has to be broken into further sectors because then there's the other first +principle notion. Okay. So what does it mean to observe sense something, the differences that make +a difference, right? So you can only, there's only information with context, right? There is no +information per se, it's all semantics, right? So you only observe things in context. So some things +are going to have to stay the same, namely, you know, reference states here and some things are +going to bear, right? So you can think of this as the gauge in your car, right? Obviously the circle +stays the same, right? And that's your context, so to speak, that's the, with the, you know, numbers in +it, they never go away, they stay the same. But as you speed up around, the gauge is going to change. +And that's how, you know, only in that context, you can tell your own speed when you're driving the +car. Right? So anyways, this is just a teaser for things to come. Now we can go back to more questions +then. What do you think? I'll pick maybe two or three more over here. Yeah, pick another question +or anyone raising their hand or if anyone in the live chat wants to write something, I'll read it. +How about the wick rotation? Could you summarize what the wick rotation is and then state this +question about it? So, yeah, that's a great question. I knew the answer, but I wish I knew +the answer better. But, informally speaking, the wick rotation is just, so it's, but Chris said, +you know, you take a quantity multiplied by I, square root of negative one, that amounts to a 90 +degree rotation in a complex plane. Fine. Now, that appears to be a trick. And what it does effectively +is take you from quantum mechanics to statistical mechanics, roughly speaking. +Now, why it works, that's probably, I mean, I have no clue. I don't know the answer myself. But now, +when things appear to be a trick and they get you something, maybe there's more to it than what's, +you know, apparent at first. But I myself don't know the answer. It's a very good question. +But I do think it may have to do with the fact that, on the one hand, quantum and statistical +mechanics both are talking about constraints on information processing, maybe from different +perspectives. But again, that's just speculation on my end. So, it's maybe a way of relating the two. +Can we just review, just read Chris's full answer to this? Read the question and then Chris's full answer? +So, as the weak rotation tells us that matter and energy are somehow +exists 90 degrees perpendicular to a three-dimensional space. +And then I'll just read Chris's answer. Then anyone, please raise your hand if you want to, +like, add something more on this. He wrote, +In a sense, yes. Time is perpendicular to our three dimensions of space, matter, and energy. We +ourselves exist in time. Wick is pointing to the intimate relationship between how we measure time +and how we determine or judge that something maintains its identity through time, i.e. stays the +same thing through time. To tell time by a clock, I have to be sure I'm looking at the same clock, +set the same way, etc. We'll get to this in July and August. +Chris Cousin, Jr.: Could you connect this, Andr, to some of the semantic information flows +discussed in Livestream 17 or any of the other cone-co-cone diagrams from the papers that you showed +earlier? +Andr, Jr.: Sure. Again, I have to preface it by saying that I am working on understanding it myself, +but just from, you know, small email exchange that I had with Chris Amps, what's said in this paper, +one can sort of see that it's related to the picture I was showing earlier. +I mean, this one, right? But let's scroll down a little bit to page, +page 2324. Here they talk about the reportation again. +Ah, excuse me. +So, +I think it has to do with, at least from what I heard from Chris' email, has to do with this notion of +local time, the definition of local time in this observation picture. +And, you know, the fact that, you know, you rotate 90 degrees twice, you get the opposite sign, +basically, right? So, the local time, effectively, that's going to ensure that the picture is symmetric and that, +you know, the flow is going to be opposite in the direction towards the observer and towards the environment. +Okay, Alexei, then anyone else raising their hands? +Alexei, I am struggling to understand and part of my difficulty bear with me for being so kind of, +but I remember Sean Carroll's lecture on time in different versions of physics, and it took a whole +hour. So, are we talking Newtonian time, which is universally the same? Are we talking relative time? +You know, philosophically, are we talking about people who believe that only present here and now is +real while future and past do not exist? You know, so when we say time is perpendicular to +my sense is that, just like Chris mentioned in the lecture, we're reinventing the time. We're creating +a new definition by calling it the same word we're all familiar with, and that creates some dissonance +where I just think that maybe we need to kind of say that this is something new. And if it's not +something new, then which version of the time are we talking about? Einsteinian, Newtonian, or whatever, +you know? It's a simple question before we say it's perpendicular. Do people agree that future is +real or not? You know? +I'll give a thought on this, Alexei. Again, speculation, neither here nor then. +If we take Chris seriously that communication is an ontological primary, and that rather than +putting communications protocols in some sort of time, Newtonian time, Einsteinian time, these are +these different time concepts, these different scaffolds that then communication gets inscribed +within or stretched within. But if we start with a topological understanding of communication, +rather than a geometric understanding of space-time, so the topology of the blanket and the observer +in their sensorium, then different kinds of cognitive agents, ranging from what we would call non-living to +what we would call living, they, depending on their cognitive sophistication, are able to +create or project different time concepts. Time concepts that are not useful will lead to their +own material dissolution, the failure to maintain that kind of cognitive sophistication, which requires +not just information gathering and utilization, but also energy gathering and utilization. +And so when you have multiple agents who are, whether you conceptualize them as within a message +passing relationship and having no surprise on the clock on the wall, then they're like in a local time +synchrony. But the primary case is one entity engaging with its sensorium and its self-modeling time in terms of progression. +And so these previous time concepts and these different attempts to kind of survey the structure of space-time +reflect different cognitive models that cognitive agents have proposed for time. +And that communication actually like takes ontological primacy over all of these ramified, +potentially overlapping, potentially non-overlapping time concepts. +But they're downstream of agents' cognitive models, not vice versa. +I'm not sure. It's how I interpreted what Chris said about communication being an ontological primary though. +So if we take one instance of a COVID-19 virus, a single kind of instance, right? It's got no memory and no +ability to plan. Is there a local time for that virus versus, you know, humanity observing how the virus +spreads across the globe? And, you know, we have agents with memory and planning, you know? So +is there time relative to a single photon? You know, I just, I mean, it's, it's, my hunch is that there's a +possibility that we're simply inventing something brand new and defining it brand new. And I just, +the drawing parallel back to all these other times, you know, I understand the effort, but it's, it's, it's just, +uh, is it, is that a brand new time? Is it some other kind of time? I don't know. +Can I, Daniel, can I say the words? +So I, I understand Alex's question. And I think it's sort of related to, at least morally speaking, +to your concern about different kinds of entropy, right? And what I can say is that also they have +Glace Brook, Marciano and so on. They have this recent paper on quantum error correction. +So I have yet to read it and take a closer look, but clearly talk about space-time, right? And my punch +to address Alex's question is that if you could sort of recover the more familiar notions of time +in a correspondence principle sort of way, you know, by taking some, you know, Newtonian limit or +H bar goes to zero or speed of light goes to infinity and so on, uh, that would be very satisfactory. Sure. +They may be doing a new thing, but, you know, recovering all stuff in the new thing, uh, is a good +sanity check. And so long as we can see that being worked out, I think I'd be very happy. +And just to the virus point, I would argue that viruses through their physical engagement, they do +have memory, but they don't necessarily have the awareness or a metacognition on that memory, +memory, except as granted by intergenerational evolutionary type pressure, but they do have past events +modifying them within some, um, set of possibilities. +So they're just being them and their internal time is, is on there. There isn't a cognitive observer necessarily +in terms of like a narrative. I, I am watching the clock, but they do change. +But it's certainly very provocative if this is not, or we could also look to Kronos and Kairos +and about the relationship of time and timeliness and agentic time. So I think there's, and then at +this point, are we stretching just this one four letter word too far, just like we brought up with +information entropy. I mean, can we think about these as kind of like continents and then just, +if someone said, I visited this continent, you know, that you need to specify a lot more or what +granularity are these topics, Corby and then anyone else raising their hands. +Yeah. On this topic for a Markov blanket, you have a transition matrix. My question is, how do you, +how are you allowed to increment that transition matrix to calculate a future state? Like, what does +that mean in terms of time? Uh, are you only allowed to increment it by like the lower bound time, +which is H over LN KBT? Does that, does that question make sense? +I'll give a first pass on this. We have discrete time and continuous time models in active inference. +For a discrete time model where you have the B transition matrix, then the choice of delta T +is a modeler's decision. And we talked about this a lot in live stream 52 on accelerated optimization. +There's some delta T that's like going to get you the most accelerated optimization. If you have too +small of delta T, then you're either, um, oversampling or you're maybe even physically unrealistic like you +pointed to, or if you have some long delta T, you're also not making an informative simulation +in the discrete time case where you're explicitly making time steps. Contrast with the continuous time +generative model, where we're doing more of a Taylor series expansion on the generalized coordinates, +at which point delta T is not required. You're merely just taking an extended approximation from a +given snapshot. And so delta T doesn't even come into play. Got it. Okay. That makes sense. +I don't know how it ties into time, but it helps clarify a little bit. +Okay. Anyone else can raise their hand or, um, under kind of maybe one more question or one more topic +from a paper. Just what's one more puzzle piece that you think gives us a first coat of paint from +questions submitted or from one of these papers that you have up? +What's something that, that helps us connect the dots a little bit? +Uh, so I don't, I don't think I have a good answer. Let me look at the questions again. +Uh, +but regarding time, I'll, I'll just reiterate, uh, what I said as I look over these questions, +which is that I, I believe Chris will address it in, in this bit. I, I don't know if you can see what +I'm sharing on the screen, but the paper whose title is communication protocols, blah, blah, blah, +from March of this year, uh, does talk about space time. It's a quantum, uh, error correction code. +And to me personally, I, it'd be nice to see that can be more familiar notions of time, such as +special relativity can be recovered as limits in correspondence principles. +Um, +yeah. Do you, you want me to say anything else? +If, if anyone could even just summarize, like what, what would these, what are these codes referring to? +Or how are, are these codes one in the same as communication protocols? +Or are they related to the blanket and the, the heat that was being output as waste? +What will, what will these codes allow us to do better if we understood them more? +I don't know that I can answer that right now. I don't, I don't think they understand the question. +Um, yeah, I, I just don't know. Sorry. +Terry and then anyone else raising their hands. +Um, I, I'm not a physicist. I don't understand maths at all. +Um, but what attracted me to this, um, idea of physics as information processing is that, um, +um, it feels like at a base level, the, whatever the X unknowable external state is, it could be perceived as +made up of information. You know, we, we, we kind of think of ourselves living in a particulate +universe, but, you know, at, at, at a, at a base level, there are these, +and stop me if I'm just talking 14 year old nonsense here, but you know, that, that, that, +that this, this notion of quantum fields, they feel like informational fields. And, um, that the, +the concept of, of physics is purely information that we then interpret as a physical universe, +which we inhabit because we can't see beyond our Markov blanket. Um, +I find this a really exciting idea that as a, that whenever you start to think of the, um, +surprisal as the, um, variational free energy between, you know, sort of your, your prior and your +posterior, if you start to think of that variational free energy, not as energy, but as an information +gap, so that if you see the energy as actual, you know, cause energy, when you think of energy +and physics terms, it's kind of this magical thing that does, that allows work to happen or, you know, +but, but, but what is it, you know, we've got a, a term for it and we've got, we've got, we've got a, +you know, we, we, is it jewels, you know, or whatever, you know, but, but, but what does it +mean? It has this magical property, but if you think of it as an information, um, then it starts +to make sense. So variational free energy starts to be seen as a difference in information between +what you, uh, what your, your prior and your posterior. +So whenever you start to think about time, time is this conceptual metric that we use +almost because we have memory. Uh, I, and, um, I, I wonder if we are getting, uh, trapped in this +semantics and all I have is English because I don't understand the maths really, but, um, we're, we, we, +we, you, you know, you guys come in here and you do understand the maths and then you convert it into +English, which I think I understand. And then I can think of it like a kid in a Swedish shop or +a physics class at the 14 years of age. I don't really know what I'm saying. I don't know what +I'm saying, but at, at a fundamental level, everything that we're talking about is in for, +I, I, I can understand it as information and I'm excited to see how things develop. +What are your thoughts on this craziness? +So many, uh, great pieces in there. +Physics as information processing. What is it that actually allows the steam engine to run and +the ball to move that kind of physics. And then as the timeline laid out also occurring during a period +of, of immense technology and, and apparatus development and, and digitization and all of +this and unconventional computing as well. It's almost like this, uh, materialist dualism +that instead of with, with mind and body, that there was some kind of dualism between the types +of forces and energies and ways of thinking about what made the steam engine work and then what +makes a computer work. And so you'd understand how the computer could fall off a cliff, but then you +wouldn't necessarily be able to use that exact kind of mechanics and physics and dynamics to study like +how the computer operated, but you could say it weighed 11 pounds. And so I think the, the provocation +and then, which is as you hinted towards, um, first off would be awesome to have a natural language, +math and graphical synthesis, the triple play, but also then to use a common approach to understanding +this and what wasn't working or is not proposed to work is trying to fit information into +that non-cognitive universe. But rather, if we take the cognitive primacy of communication, +then all of these forces as inferred by the sensorium, which is all we're ever going to have, +totally the most valid first principles constraint to make, maybe then we'll be able to have a lot +more synthesis about even how steam engines work when we understand that we're bound in that inference +by our sensorium and all that comes along with that and the variational free energy bounding surprise on +our generative model. Dean? I'm going to, I'm going to come at this from the position of a 14 year old lens, +because I don't think there's anything unique about that. Um, I think on most basic level, +what I took away from Chris's lecture was that there are, that there's information processing of at least +two types. Uh, one is figuring it out as in figure five or any of these other figures that are up on the +screen right now or anything that stabilizes the world around us. I think examples have come up in +this conversation already of top apology or geometry. There's a second type of information processing, +which is outing the figure type of information, which is searching. And I think Alexi used the +expression, which time are we talking about? So these two types of processes are quite different. +If you do a comparative analysis of the two, you have information practicing and you have to have +both of the processes to really get a sense conceptually of what it is that communication is really talking +about. If you are information practicing your wayfinding until you collapse to something we might +describe in a verb tense as way found. It's in the quantizing of that invariant state, that vast ocean, +if you want to think of it metaphorically out to all, all horizons on the manifold into a variant or a +nuanced or a discretized or an orientation type of state where all of a sudden things like currents and +temperature shifts and even things that we've applied labels to like entropy and a transition matrix suddenly +appear out of the seeming nothingness or the invariance. So that's not the math, but that's kind of stepping +back and saying, what is the math trying to say? I think it boils down to the comparative analysis. +I'm hoping that in the coming lectures, Chris starts opening that up and doing kind of the back and +forth between the two types of figuring it out and outing the figures because I think when both are +part of the conversation, that's when we start to get a sense of what's really going on. +Chris, thank you. In this sort of closing sessions, especially if anyone hasn't added anything, +please feel free to give any last reflections or questions arising. So first, as many people as would +like can reflect that way. And then after every hand has been addressed, then, Ander, if you could just give +us a kind of look ahead. So Avil first, then anyone else with just where does this discussion take us +and what do you want to continue to learn more about? Avil. +Avil. So two things first, if I walk out of here without understanding what is the current co-configure, +I will be disappointed. And I don't actually expect it to happen in like 12 hours. That seems like +something. But more seriously, I can agree without difficulty that there is a unity between physical +structure and communication protocols. But communication protocols or quantum reference frame kind of +individual, physical states, individual meaningful measurement outcomes that are reified as physical +possibilities. But the question is then where do communication protocols slash quantum reference frame +come from? Like what are the physical determinants that bring about those things? And that seems like a +question that could be addressed at the conceptual level of this course. And I would like, I saw the +live stream 17. So that did not help. Anyway, I would like to have more grip into that if we can address it in the +form courses. Thank you. Thank you. Anyone else, especially if they haven't added anything yet? +Want to give kind of a closing thought. +Yeah, there's a lot to learn. And Chris brings up a lot in his lecture, even though we've only had one. It's one of the denser hours of epistemic resourcing that I've seen. And there's also all of these accessory and cutting edge works that +that under showing. So I'm really looking forward to how we can continue to curate and improve the questions that we have. +So that the uncertainties in our generative model that we have, how would this be useful? What does this mean? Why is this variable this way? What is the implication for that? +Those are all the great questions. And let's just get them inscribed on our classical documents, so that we can have other quantum +cognitive agents reduce their uncertainty by looking at the responses. So with that, I wonder if you could just look ahead to lecture two, and just like point towards where we're going to be going. +Yes. So lecture two will, I think from what I heard from Chris is going to start to address +this basic ontological aspects of the quantum measurement. So the two sources to look at would be +two papers, mainly this one, physical meaning of the holographic principle. This one's by far the most pedagogical one. So if I had to suggest that folks look at one, it would be this one. It's very easy to read, I think. +But maybe for a bit more detail, there's the slightly older one. And they talk at length about COCO and diagrams here. +Free energy principle for generic quantum systems. And so I think that's what's going to come in these two lectures. Basically the main ideas behind these two papers. +Anja, that's all I have to say. +All right. Well, thank you for all who joined this first discussion. Hope that it was fun for you and for the audience. And check out the course site for the readings to get ahead of the second lecture. +We'll all watch the second lecture and come back in about a month for the second discussion. Everyone's welcome to join and the website has the registration information. +And so now that we've seen a little bit of one way it can be, we may do something similar or we may do something different for the second discussion. +So thanks again, everybody. See you later. +Thank you. +Thank you. + +## hSj7eOYuZe4 + +## hSj7eOYuZe4 + +Hello, it's July 1st, 2023. We're in Physics as Information Processing in the second discussion. +So thanks all for joining. We'll see who else comes in. If you're watching live, feel free to +write questions in the live chat and I'll relay them onwards. So Andre, thanks. And it would be +awesome to begin with a little bit of some overview and some points that you wanted to get to. +Yeah, thank you, Daniel. So good morning, everyone. So if you would start by bringing up a bit +cold with Chris's spoke of last time. I have a few slides that I singled out. So now we're getting to +reformulation of quantum theory in terms of reformulation of +opposed to dynamical theory. And as we've talked in previous sessions, we need to +think a little bit philosophically what the ontology of measurement entails. +So Chris, last lecture started with these three basic ingredients. +So an action, that's number one, a thing to ask about. This is the qubit, the physical degrees +of freedom that we need to perturb in some way to ask a question. And then finally, a semantics, +a language that makes sense, a context that makes sense of that information that we gather. +So I should say, number one, I came up in the first session when, I can't exactly remember, +but at some point we talked about what is information. And we came up with, or we brought up +a definition that goes something like this. It's a nice adage. It's differences that make a difference. +This is, Chris told me, I can't remember who first coined this adage, but I think it's quite old, +at least early, early 70s. Anyway, so number one, basically, let's curve that part, right? +If I want to gather some information, I need to perturb a system somehow. So I need to act. +And this is literally in units of actions, right? So units of energy times time. And he provided the example +of rhodopsin, right? Which is apparently near optimal from a quantum perspective. This is mechanically +what goes on in your eyes as light comes in. And this is literally the difference that makes a difference +in your brain. The second part were the physical degrees of freedom. So recall that +we had a boundary, okay? So this theory of quantum mechanics as measurement necessarily breaks down +the system into a bipartite system with a boundary. And you are in a boundary, which are a lot smaller than +than the degrees of freedom of A and B, at least in terms of the dimension of the Hilbert space. +And the other medium change, right? Sorry, can you repeat, Andre? +I repeat what? Sorry? Just last 30 seconds. +Yeah. So the second ingredient is the physical medium in which this information exchange occurs. +So the physical degree of freedom, the qubits are arranged and these are physically perturbed in order to +mediate that difference that makes a difference. So we can think of this duly as either preparing those qubits +or decoding those qubits. And then finally, there was the notion of the reference frame. +Okay. So this theory necessitates a free choice of the agent of a reference frame. And physically, +what this amounts to is a free choice of a measurement axis for those qubits. So we can think of +measuring Z-spin in those qubits on the boundary. So, you know, that requires a free choice of that axis. +And this, you know, amounts to your reference frame, the context in which those measurements are performed. +On one extreme case, we have that if those references are perfectly aligned, +then the exchange of information is noiseless. There was somewhere over here, I think. +I can't quite recall. But generically, these two reference frames of Alice and Bob A and B are not +exactly the same. They get to choose the reference frames. I think it's all the way on the bottom, +actually. Right? So they can be aligned. If they're aligned, they are exchanging noiseless information. +Right? There is no randomness. If they are not +superposition, and that's when you're going to get, you know, probabilities in the outcomes, +trivial probabilities. So Marcia, good? +Yes. +Okay. So I have a few more comments of the main things. Okay. It'll take me a few minutes, +and then we can chat back and forth. Perhaps this slide is +the most interesting, important. I remember this is where I asked the question myself because I was, +I still, you know, it's not easy to wrap your head around what's going on here. +So we said that measurement and preparation are completely dual pictures. +So this is going to be expanded on in the next lecture. But a direct corollary of that is that +if you prepare, Chris said it in similar terms, if you prepare something in a certain state, and then you +measure it right away, you should expect that it's still in that state. Right? So, +intuitively speaking, this induces a cycle, right? And where you have preparation and measurement, +and you're back to square one, so to speak. And in generating that cycle, obviously, there has to +be some notion of time. Right? So that's where we're going to have to make sense of the distinction +between internal and external time. If we take this measurement operator, and here perform some +operations, we can write it as a unitary propagator from quantum mechanics, which again preserves +information. It just leaves the state as it is. But with this little trick of the weak rotation, you can +write it as e to the measurement up to a constant. And that, you know, precisely encodes, at least +morally speaking, what we just said, right, that this preparation measurement thing is a cycle. And more +importantly, it gets rid of this t external time. And presumably, in here, it's going to be replaced by +some notion of internal time, we're going to have an internal time QRF, that is +sort of quote, unquote, the true local time, as the observer proceeds. +I'm not entirely certain of the details. But now looking at it here, the beta, which +presumably there's a, you know, beta and inverse temperature here, I think this is going to be +intimately related to notions of thermal time. This has been explored in papers by Rovelli. +Call, you know, internal time QRFs and thermal time is the way in which the observer observes the +passage of time. Those papers of Rovelli are linked in the Q&A, by the way, so I can also provide links later. +I think there might have been a question. Should I stop? +Yeah, Dean, go for it. +I was curious about this, because what I heard Chris start this session with was that he was going to try to +help us understand when we are looking at something through a mechanical lens. So quantum mechanics. +And to me, this slide was the way of addressing, as he said, his words were characterized information +transfer in a quantum theoretic way, that the that the translation from H to M, took the mechanics aspect +out of the change or timepiece. It was the purification, the most theoretical form of transfer +of information, as opposed to holding up just the one ordered or sequenced or mechanical lens. +Did I I heard you ask the question? And I kind of smiled because I thought to myself, no, I understand +why Anders is asking this question. It's, can something be both gripper in this in the conceptual +meaning sense ontologically, and gripped a physical object. And I think what the I thought, maybe I was +misinterpreting, I'd like your what your perspective on this is, that ability to go from H to M is what +allows us to have both the mechanical and the theoretical, the transference in its purest sense. +Is that being able to have both? Both can be true? Is that what it I think? +I think the answer is, yes, for sure. Yeah. Yeah. Yeah. I +I think much else to say here. It's, it is, it is true that you can think of the information transfer +picture as an enlargement or an addition to the mechanical picture. And that I don't think that +the mechanical picture has to be just, you know, thrown out completely. I think, I think it's interesting +to keep it to see how it can be recovered from the information processing picture. This reminds me, +actually, there is an Italian physicist, I think his name is Cribella. And Dean, if you want, I can, +I can link you the papers later, who is very fond of the information theoretical approach to quantum +mechanics and deducing quantum mechanics from first principles. And he has a paper. I think the title is +physics without physics or something like that, where, you know, he, he treats, um, uh, this whole +thing as a seller, big seller automaton. And by taking the appropriate scaling limits, he derives QFT. +I'm interacting QFT. So, so he says, okay, let me start with some axioms, how information should be +the information processing. I mean, I'm very, very loose here, then get a cellular automaton out of this. +And then take scaling and recover QFT, non interacting QFT, at least. So, you know, start with a few axioms, +a lot of information processing, recovery mechanics from that. Does that add to your question a little bit? +Yeah, so, so all I'm, all I'm trying to do is, is tie this back into some of the live streams that we've +done in the past. And I know I use different language. I, I am not a scientist in the, I've not +studied quantum other than sort of in passing ways. But I do think that what this spoke to me about was +something that I've always tried to include, which is that there's the figure it out piece, which to me is the +conceptual and the meaning part that Chris talked about very clearly here. And then there's the outing of the +figure, there's the sort of physical edge and boundary that if you don't have an awareness of both things happening +concurrently, the preparation and the measure are unitary, have sort of minimum of two state, +then all bets are off in terms of being able to think of these, these propagators and operators and +all of the ways that we're trying to explain the situation, they're off. So we have to have both. +We have to have that sense that there is a transfer, but there are, and there is the mechanics, we have +to keep both juggling all the time. So that's where I, that's where this one particular slide was really +quite both bias confirm confirming for me, but I think it also, as you say, is going to open up a lot +that July. Yeah, I agree. I think this is the most, you know, important or mysterious if you wish a slide. +Yeah. And it turns the pyramid, the conventional approach, here's a pyramid into a top that can be spun +essentially with a rotation and it, and it validates how information transfer probably gets short shrift +in most of this quantum conversation, but it's, it's probably at least an equal to, to the mechanical +part that what, what order are we delivering this information in? So I'm at that point, I'll stop. Thanks. +Yeah. +Thanks. Alexi. +Hi, can you hear me? +Yep. +So, I mean, I, I'm an, you know, my mathematics teacher in high school used to say, this is beautiful, +but I don't quite understand it. So I wanted to, maybe I have two questions, but I'll start with, +with this simpler one. So when Chris talked about the reference frame and said common language, +you know, I think in Shannon common language is the alphabet. There's a receiver and sender and +the alphabet that they share. And then Chris said, this reference frame is a physical thing like +neurons or hands or mouth. And then I got lost because, you know, if, uh, when I encode, I modify +something in the, on the hard drive, then the physical components of the hard drive is the reference, +uh, frame. But, you know, this is not the shared language, you know, so that's my first question. +Maybe I'll stop here to, to really, I understand we're dealing in metaphors, but I want to understand +that, you know, the reference frame up or down, if Bob and Alice understand it similarly, then +is that really physical? +I think the answer is that +Chris would probably give a better answer. +Start from the beginning. +Go. +Is that both are true at the same time. I think Chris would give you a better answer, +but the, the semantics or the, sorry, the shared language or reference frame has to have a physical +embodiment. And, and that's a hardware, a piece of hardware. It's, it's a neuron or something. Um, +in this case, it's the, the, your free choice of the spin axis and that, that's your reference frame. +But at the same time, it's, it's a physical, it's a physical piece of hardware. So reference frames are both +the piece of hardware, but also a piece of semantics. +So if I take it from the language, I don't quite understand to computer science. So if Alice and Bob +communicate through a piece of memory, whether it's random access memory, hard drive, then Alice encodes +something on that hard drive and Bob reads from the hard drive. So in that sense, they understand how +to encode and decode, but they use the same physical piece. That's the reference frame. Would that be correct? +I might, I might just add one again, continuing with the computer science, let's just, and the Shannon +alphabet. Let's just say that these are plain text character encodings. And so they share the same +set of character, a through Z encodings that, that is reflected somewhere physically. +And so if, if a message were, if, if, if a message were sent from another letter, it wouldn't render +properly. It's like outside, they couldn't differentiate Omega from Zeta because they only have the A +through Z software pack. And so there's an informational distinguishments that's facilitated by the +information that's classically inscribed on the screen, the memory. And there is, is an embodied physical +as well as a semantic aspect to that situation. +All right. I think I'm getting closer to understanding. So to move away from computer +science, if Alice is a painter, she took a brush, dipped it in red paint and painted on the canvas red. +And Bob has the same understanding of, of red as Alice. And then he looked at the canvas and read, +read that it is both the physical, you know, a mark on the canvas and, and the symbol of red. Right. +Yes. And if, if, if somebody had red, green color blindness, +that physical apparatus was not able to, to informationally distinguish, then they, they +couldn't make this wick rotation into the quantum cognitive space. They wouldn't be able to, to, um, +detect that there was a red figure painted on a green background. +Yeah. +Yeah. +Okay. +The key component here is that the, the reference frame is inside of the edge. It's, you know, +if you're a color blind, there is something physical inside of your head or your eyes or whatever +that does not allow you to see the difference. So, so your semantics are affected ultimately by +a physical piece of hardware. +Well, yeah, maybe I used a bad example because color actually is computed in the brain and color +blindness is not necessarily physical. So, uh, okay. But I, I think I get it closer, but you know, +if we, if I may ask another question, I get lost with a saying that preparation and measurement are, +they have to happen both. And that when you prepare or when you read, you modify. So I am okay with the +act of writing, the act of encoding. So when I encode something such as I, you know, mark red on the +canvas, or I write to the hard drive, a, I modify the hard drive. That is true. But if I, if I am asking a +question, right, I'm not modifying anything on the canvas. So in what way the act of asking the +question or preparation is modifying the qubit, this is where I'm lost. You know, if I, does that make +sense? +Yeah. I think this is exactly the question I asked, uh, to Chris in, in the stream. Um, so I can't quite +remember his answer, but I think it was basically that they're completely dual pictures. Um, in the +sense of, you know, they're, they're both two ways of looking at the same thing. Uh, it's kind of like, +you know, evolving towards the future or, or, or, or the path you have time symmetry, you know, +you can, you can do either. Um, yeah. So I, I myself, I'm slightly confused by the very, +but I think the answer is that they are anything at the same, the picture of, or am I preparing +or just completely dual? Okay. That's it. Well, do you remember that, Daniel? +Yeah. Okay. I'll give a different angle than the, um, the formal duality relationship. So let's think +about our measuring device is going to be a carpet and we're interested in whether somebody has walked +on the carpet. So to prepare the carpet, we comb it very nicely in one direction. And I'm sure +there's analogies there with the, um, preparation of qubits and a quantum computer, but we're going +to comb the carpet as an act of preparation. And then something is going to happen. Maybe it's a +split second later. And so we're very confident that it should be still nicely combed, or maybe it's +going to be a whole working day later, but now we're going to take that probe, the tip, whatever it +may be, and we're going to run it along the carpet. And we're going to measure the state of, of the +direction of the fibers of the carpet. So if we didn't prepare the measurements, we wouldn't have a +calibrated reference point to know how things had even changed. And so there's, there's a cycle of +preparing the apparatus through something like a calibration, or at least like a emptying of memory +to be able to then make an observation and to have that observation, have the kind of meaning that +we would have expected, like purge the photo detector, make the measurement. Now you have to purge it again. +That's the preparation. What do you think Alexei? That is a beautiful metaphor. And I dig that, except +I think I heard Chris say that when you prepare, you may cube it into up or down. And this is when +I'm lost because when you comb the carpet, you don't do that. You don't put the footstep on it. In fact, +there may not be any footstep, right? And this is, this is where I'm missing, you know, I think in quantum +physics as well, or if you measure there's the cause and effect relationships are broken here for me. +You know, if in my act of preparation, I'm opening my canvas to, to read the message, but I do not know +what the message will be, I'm game. But if he's saying that when I prepare, I modify the carpet +such that it becomes either up or down, actually not blank, but up or down, then I'm lost. +You have to comb the carpet up or down. There's no blank carpet. So you, you, um, you know, you have +a slate, you have an etch a sketch. You could either make all the pixels white, and then you'd be able +to measure a message in the dark, or you can make all the message dark or the slate dark. And then you +could be able to measure something that was light, but you pick a direction of polarization +in your preparation or calibration step so that deviations from that can be the measurements. +So you're saying carpet is binary. Carpet doesn't have multiple states. Carpet can only have two +states. So if I prepare it, I, I effectively put it in one of those states. Well, let's say anything. +Yeah. Let's say we had the carpet at a 45 degree angle. Mathematically, we could describe that as a +mixed state and it may not be the most effective for measuring the movement of people across the +carpet, but yeah, you could, um, define it, um, or comb it that way. It doesn't necessarily, if it, +if it were only a binary, it wouldn't be a qubit, then it would be a classical bit. +I actually, yeah. Andre, go for it. And then Dean. +Yeah. With, with, with, you know, I, I, you know, recall lecture a couple of slides more. +We briefly talked about it like 10 minutes ago by preparing, you are choosing an axis and whatever +we're done in that axis. Now that doesn't mean that the environment gets to do its own thing. Right. +And a superposition of, uh, outcomes, but at least you're measuring that axis. Right. So it's, it's going +to be either up or down in that, in, you can add it to picture. And I have, uh, I just want to say a few +words on graphic screens and define my diagrams and maybe clarify the question a little more. +Dean. Dean. +On this really, really interesting. I don't know how metaphorically you could explain the carpet +as a wave, but I think once you get away from the, it's up or down, that possibility enters the +conversation. And the only reason I, I wonder whether or not metaphorically the wave aspect of that can be +included in how you're describing the carpet is that Chris spent quite a bit of time on the all possible +paths argument. And in particular, and I love this when Alice isn't looking. +Dean. Right. Well, in that, in that wave equation, how could Alice not be looking? +That's what gets real. I mean, Alice could not be looking if it's, if it's binary, if it's or up or down, +but if it's wave, it's on a, it's on a, it's a unitary propagator. I'm not sure how she takes her eyes off of this +preparation measure thing that's going off all the time. So this is, this is actually an, what I consider to be a feature or an opportunity or an opening +in the sense that from a strictly information transfer piece, not a mechanical piece. Now, not a physical thing. +From an information transfer piece, maybe the carpet can be metaphorically understood as a wave. +That the eyes of Alice or whoever's doing the preparing and the measuring, Bob, Alice, I don't care, +peaches. That there is actually a difference then between something that is exchanged and something +that is informationally transferred. And again, I don't think that the four of us here sitting right now +are going to resolve that. But I wonder if part of our, part of the difficulty in this is +finding the physical metaphor that explains the theoretical transfer in such a way that it doesn't, +it, it, it both breaks the symmetry and holds it. Because again, we're back to gripper and grip. We're +back to figuring it out and outing the physical figures in the state that we're in. And we have to do both +at once. So again, Daniel, you're, you're the creative guy. How do you turn your carpet metaphor into a wave +equation? Maybe there's a subwoofer in the room, or maybe somebody is waving the carpet. But what you, +what you said about the, what is exchanged and what is transferred, +maybe the transfer is the classical components. I don't know which will be exchange or transfer, +but one of them is the classical inscription. That's how many gigabytes were written or read to +the hard drive. But what's exchanged, let's just say, are the semantics of meaning. And given an +inappropriate co-preparation, maybe only one bit needs to be transferred for the exchange of a message +that was already triggered to be set off. And then also this entire discussion, it kind of strikes me +as like the, uh, Cartesian duality of our time, but rather than mind and body, it's the physicality +of information syntax and the metaphysicality of information semantics. And that's these kind of +two worlds with their own physics and metaphysics that we're trying to connect. And it's like, +similarly, well, is the information, how does the information world cause something physical? +How does the physical cause something informational? And I think that the entire journey that the course +and everyone is on is, uh, physics as information processing that those as a qualified and distinguished, +but minimum of two set Alexi. Yeah. Thank you so much. Uh, can you hear me? Yeah. I I'm yeah. +Thinking out loud maybe, but I think the way to get closer to understanding in my head is to let go of +the metaphor of asking a question, right? So if Alexi is talking to Daniel through zoom and I pose +even a closed question up or down. So I, you know, uh, formulated the question and, you know, uh, +released that into zoom space and then I'm silent and I opened my listening to your answer, but you, +Daniel have agency to say up or down. I did not, I do not have any effect on whether you will say up +or down, right? That's the difference. So me asking you a question does not modify, uh, the answer you, +you decide Alice talking to Bob, Bob has the agency to answer the question, not Alice. And that that's +where I'm stuck. So I don't know if asking a question continues to be a good metaphor, prepare, +preparing the instrument for measurement is, is maybe better. And, and so, but, but actually +Alice is not asking Bob, Alice is acting on a qubit, right? And qubit is a, is a simple thing that, +you know, can, can literally physically be up or down, I guess. And so that's, that's maybe, +maybe that's the answer that Alice does not ask in Bob. Alice does something to the qubit. +And because qubit can be up or down, then that's no longer asking a question, +asking a question implies your, your correspondent can, can, can tell you anything. +When we're dealing with the very sophisticated process of natural language question asking, +I think we're dealing, we're, we're dealing with something that that's orders, um, of composition +above the kind of, the kind of binary ask. So I, I, I agree with that. Um, it's almost like, um, +Alice prepares the photo receptor. So it's slate is refreshed and then it's the photon or the light +emitting object that has the agency to send the photon or not. I mean, maybe the measure, the +preparation occurs and then no one comes back to check whether or not agency was even exerted. +And I think when we really focus on like the nano probe, like atomic force microscopy, +really tracing with a super fine tip over the surface of something or scattering x-rays, +when we really think about the, the, the, um, moment of information transfer, +we realize like nothing comes for free. And I think that this question asking framework, +other than really conveniently re-deriving quantum mechanics in one hour, incredibly helps us +understand that we don't get to ask it for free in terms of time. We don't get to ask it for free +in terms of energy. And we don't get to ask free floating questions where the question and or +agentic responder are not involved already in a semantics. So I think it's calling attention to, +the components of an experiment or a question or an opening, but agree that it's not always as simple +as a binary question, but in the special, simple basal case, it does start there. +Andre, do you want to look, you want to look at any of the, um, the, uh, participant submitted +questions? If there was any that struck your face. +Let me finish real quick. I have like maybe five minutes to wrap this up. Uh, so, and, and this, +I think is going to at least in part answer one of Dean's concerns. So the next thing we should, uh, +very briefly cover. So let's talk a little bit about, you know, holographic screens. And also, +uh, this is the paper on the, you know, physical meaning of holographic principle. So let's take a +look at also this picture from that paper. This is the holographic boundary and, you know, +Landauer's principle and all these thermodynamic constraints on information processing +necessitate, uh, the breakdown of this boundary in terms of, you know, free energy extraction sectors, +memory sectors, and so on. And by the way, if you, if you read, um, the paragraphs around this picture, +there's more details on how the time QRF is induced, but we'll leave that as it is for now. Uh, +Dean, what I meant by this answering your, your, your concern, your concern was that you don't see how +Alice can take off her eyes from the picture. Yeah. Uh, but that's precisely what's going on here, +right? There is, there is a sector that's being observed, which is a subset of these qubits +that are being observed, but presumably she has the freedom to, you know, say, hey, this sector, +I'm now I, I declare as my observation sector, this I declare as my memory sector, this I declare as my +thermo, uh, free energy extraction sector. So it is because of these fundamental limits +on the amount of information processing. It is not true that she can, she doesn't have an, an idealized, +you know, you know, uh, Laplacian, if you wish, um, access to all the degrees of freedom. She can only +observe finally, uh, a finite amount of stuff. So, so she, she is forced to take her eyes off at least +part of the system, at least for a while. Right. Because otherwise, you know, all those energy and time +constraints that Daniel was just mentioning, you know, they're not, we, you have to abide by those. +So, um, so a few things that Chris mentioned in the last lecture when it came to this holographic +screen is, well, number one, uh, that all the information exchange is mediated by bits, uh, here +at the whole qubits at this holographic screen. This in turn enforces, um, the finiteness in the +information transfer, if you wish, there is not infinitely, uh, an infinite amount of information +in finite space. Even if you were to pack it at Planckian density, which is presumably the limit +that black holes achieve, it's still a finite amount of information in a finite amount of. +So if you have finite icons, you are taking. Uh, the other is the following. Um, and again, uh, Chris +will deal with this at greater detail in perhaps lecture five, I think maybe four, but it's the notion +of emergent space time, which is that A and B separately. So you think of, you know, these are +presumably the internal degrees of freedom, what's going on inside of A, Alice at an abstract level, +there may not be a notion of space, but, you know, the notion of space emerges at the boundary. +And, uh, that's, uh, not easy to explain. I think presumably has to do with and stuff like that. Um, +and I don't know the full details myself, but, but for now, that's, that's, that's one of the key things +that, that, that Chris mentioned, which is that an agent, you know, may not have space, but it may emerge +and the last thing, and, uh, I'm glad that to see that this was mentioned already. So it saves us time. +I don't have to say it again. It'll take us a minute. This five months. And this again answers your, +can you turn off your video under just so you don't have lag? Yeah. Thank you. +One second. +One second. +Yes. +Can you, yeah. +Me? +Go for it. +Okay. So this notion of, uh, Feynman diagrams, right? So what is this whole thing about unitary +evolution or preparation? The, the scattering picture as Alice preparing a vector of her boundary, +right? And writing it into her memory. And then +measuring it at some point in the future, right? Requires that you consider the possible +what happened to that preparation. Okay. So if there's some amount of time passes, +you prepared, you know, your state at time t, you look at it again, at time t plus dt, +you have to take into account all the possible histories. +Clearly, I, myself of the system, or at least that part of the system, right? Dean is just saying, +I prepared it here. Now it's up to B what happens until the next time I look at it, +up to the environment. So I, I, I have to, in, in that unitary evolution, I consider all the +possibilities, right? And that's, that's exactly, you know, the Feynman, you know, the role for. +So let's just look at the conclusion. +I'll leave it there for now. No, we can open up the discussion. So there's two things that are now, +where was this? Yeah. Yes. So, go ahead. +Yeah. So all I was, I understand, or I think I understand, I think I understand, +that if I'm looking at on, or if I'm attending to on, I'm not looking at off. What, what I saw in that +wave thing with Feynman was, I could also be looking at the switch, where the two different +places of on and off exist. And when I'm looking at the entire switch, I'm not looking at one or the +other. I'm looking at the, at that basal, as Daniel described, that basal opportunity of both. And, +and I think that on that, I don't know how that ties in with reference frames, but on that level that says, +I'm now looking at all possible. In that moment, I don't think at least from that perspective, +that reference frame, I'm zooming into a on or off binary choice decision branch, I'm, I'm still at +that place where the preparation allows for, for both. I'm not, I'm literally not taking my eyes off +at that place. So again, I'm not arguing with you, because I what you described, and the image that +you brought up, all that is, is bringing it down to a granular level and saying, if I'm paying attention +to this, I can't be paying attention to that. But if I physically, from a transfer perspective, +back up, so that I can see both, both possibilities at once, I don't know that I'm taking my eye off of +either. That's, that's the only thing that I was bringing up. Now, of course, Chris will probably hear +this and go, Dean, you got to go and rethink this whole thing through, which is great, because I'm +here to learn. But that's, again, if I miss, I miss spoke about what I was trying to say, it wasn't +because I, I don't see those moments when I'm directing my, my preparation over here. And that's +going to give me one kind of measurement versus if I'm taking in a larger timeframe, and now or larger +spatial frame or larger holographic screen, like I don't even know how to describe it, because I just +don't have the words for it yet. But there you go. I just want to be clear that I'm not, I don't think +I'm confused. I just want to know the difference between when I'm looking at on or off. And when +I'm looking at the entire switch, that's all. I'll add one point on this slide. It goes back to some of +the earliest points that Chris made about the ontological primacy of communication. Space and +time arise from communicating agents. Communicating agents are not something that we have to squeeze +into an a priori spatial temporal framework. So quantum mechanics, saying here on the slide, +is about measurement, not mechanical motion. So we don't need to wait around or speculate as to how +from the mechanical or physical hard drive. Measurement arises or happens. We start with the measurements +occurring. And then different kinds of sophisticated agents can conceptualize of movements through spaces +and times. And this also relates to Bayesian mechanics as a mechanics and a physics +of movement on information geometric spaces. And so the physics can be projected by appropriately prepared +agents. But we don't necessarily need to wonder about how that agent fits within a mechanical setting. +And I guess this is where humans are with the I am a strange loop, Gertl Escherbach, +gripper and gripped mystery of being able to see it multiple ways. +Andre, do you want to go to the questions? +Yeah, let me just say, are some remarks here? I think what Chris is the punching line of the paper, +free energy principle for quantum systems, which is, you know, morally speaking, it's similar to +the relaxation to thermodynamical equilibrium, which is that if you have an isolated +bipartite quantum system a and b are going to turn um are going to tend towards reference frame +alignment and therefore entanglement and this these two things are are basically +very intimately related by aligning the reference frames you'll get a and b +will be driven towards maximal entanglement +so uh should i uh take a look at the questions and i can open this yeah +so so uh if you guys have access to the q a +uh there's been more questions in the last you know bringing you guys's attention to a few of them +uh this is where um chris links uh to the papers of robelli which i mentioned earlier in passing +when we were talking about the preparation measurement cycle and the notion of thermal time or +uh you know subjective time if you wish somebody brought up this remark that you know performance +some calculations here you can relate t time to a change in um +uh a change in the volume of phase space or you know more precisely the the amount of available states +in the state space you know with the log and whatever uh with respect to the energy right and you know +you know so so this you know restates the +the idea that we're talking about earlier right like in your intrinsic uh subjective time being +this ds the e derivative a thermal time basically speaking that is uh observer dependent +so that's a a good question to uh take a look at +um there's another one what is what is meant by local free of choice uh this is a very good answer +to local here means that's on boundary right isolated uh systems are idealized by you know +definition there's always going to be an observer interacting with an environment so +a nice a purely isolated system is an idealization so what local free choice means here is exactly what +what we said earlier right that you get to choose the spin axis at the boundary qubits +um +and there were some uh nice questions relating mark of blankets to the scattering picture +uh so this one for instance is also worth a look +uh +so somebody you know brought up this you know very good point about the scattering matrix +the scattering matrix is by the way you know let's not think about it mathematically just um +let's think about the picture right you you imagine two particles that are interacting +and far away they are free right they are spatially separated and interacting free particles +and somehow at some point they come to interact right and then you have to write all the finite diagrams +for all the exchange of bosons or whatever the interaction is and and that's how you would compute +the amplitudes for the scattering process right how those two particles that were you know separated came +to interact and then they go off to infinity again uh chris here brings up the point that in the sense +the mark of blanket is you know in the opposite direction right you see a marker blanket in at least in the graph theory sense you start with a node and the marker blanket is just the set of nodes that +separate in terms of conditional probability that node from everyone else right and then you can have edges which are going to be directly uh causal networks so acyclic +no no no cycles directed edges that's that's how the graph looks +and you you build up so essentially you're taking the opposite um picture right you start with a node +which in this case would be the interaction and then you build up to infinity as opposed to in the +scattering picture which is like you start with particles not interacting at infinity you see how they go off to infinity again +and then the interaction is is in the middle uh so that obviously the answer is is a bit more uh precise but it's worth taking a look at +and this is a an interesting one too of a symmetrical input input what bandwidth so again if you think of it as a graph nothing is constraining you to say that you're going to have +more incoming errors than outcome errors into the boundary right that that's just the way it is you know if you think of it a graph there is no symmetry imposed the only natural sense of symmetry would be energy conservation and that's exactly what you know chris says here right which is that um +and i'm not sure how you related to a conservation of not quite information but you know i wouldn't know how to phrase it myself +i would just take a look at the question but basically no there's a priori nothing that necessitates symmetry in input output in markov blankets +but if you are to talk about any sort of symmetry the natural language would not be in terms of incoming versus +zaka mineros but rather energy conservation which we we saw is is you know a close casting of information transfer right because of +the minimal requirements for action to rewrite a qubit +and the last question that i would like to bring your you guys's attention to i mean all of them are great +uh but we have finite resources and finite times on internet uh would be +would be uh this one okay and i'm not gonna go into too much detail but this is again about +the local wait no one second +um +so it turned on my camera by the way dino go for it +and also people people can see all the questions on the website right and submit their own as well +yes for sure uh so and now i lost track of the question that i wanted to show you guys +i think it's this one over here yes uh yes exactly this one so by the way all of them are good even +this one the holographic principle does it make it easier to compute any physics problems well please take +a look at chris's answer but all i can say is that these ideas of holography are i'm not an expert i'm +not a condensed matter physicist but they're they're used all over physics nowadays not necessarily just in high energy +i know that condensed matter theorists use them all the time so the answer to this is +uh an emphatic yes the holographic at least adacft is used all over but again i'm not a physicist myself +um so this is the last question i wanted to bring your attention to and uh here chris uh this this may +be the one worth taking a look at in terms of uh in order to answer all uh you know the earlier +speculations that we had about time internal clock here somebody asks about um +synchrony and +you know just the perception of time with respect to external clocks i suppose +and chris here answers that at the end of the day any even if you're using an external clock +everything that you perceive outside at the end of the day has to be in terms of an internal clock +any information incoming you're gonna have to you know make causal sense of it you're gonna have to +write it into a memory in terms of an internal clock so so you're never going to be able to +ease so to speak increase your precision of the passage of time by taking an external clock +even if you think it's more precise than anything you have inside i think that's basically what his +answer amounts to and this is morally related to you know all this frame halting problem stuff where because +i'm never as a system you're never going to give uh an infinitely precise you know i'm speaking very loosely +here an infinitely precise description or yourself you're you're never gonna morally speaking again +that's that's what it amounts to you're never going to be able to outdo your internal clock by +by trying to leverage something external like like you know an atomic clock that even if you think it's more +precise at the end of the day your own subjective perception of information will depend on internal clocks +so we can uh open up things any concluding remarks and any other alexi or dean or anyone else who may +have joined i don't have anything else to say at this point i mean where where or when are we we had session one +with a real big historical overview section two we focused on the measurability question which is answered with +the question and we talked about some of the constraints of question asking namely that it takes time it takes +energy you have to ask it to somebody or something on an interface and it has to have a semantic quantum +reference frame and this is dovetailing with a free energy principle of course because the free energy principle +as it says in the syllables to what does the free energy principle apply and chris says everything +measurable that's why it's a theory of every thing that we choose to measure model not ontological +statements about what is and then where are we heading into in session three +have you looked ahead under or do you want me to read what's in the syllabus +uh please do yeah yeah i think basically uh my short answer to that and possibly my full answer also +is that it's just going to take care of the loose ends in this lecture two so this lecture two is a +warm-up to the to the real thing which we need to talk more seriously about reference frames to to in +order to get the precise uh picture of this quantum measurement process yeah yeah it says that session +three will introduce the idea of quantum reference frames and the representation using hierarchies +of binary classifiers these formal structures provide a semantics for measurement and hence provide +the basis for a theory of meaning for interacting agents the language of qrfs allows a particularly +straightforward and intuitive definition of variational free energy and so allows a fully general +quantum formulation of the fep we will see that the fep is a classical limit on the principle of +unitarity the fundamental principle of quantum theory uh by the way uh daniel can i i'm not sure +if you're busy but i just got an email from chris i think he may be watching us and i think he's just +confused about the link so chris i don't know if you're there still but um do you think it's worth +it we send him the link real quick if you can come for a few minutes yeah or i will oh i'll send chris the +um the zoom link now so yeah hold on hold on to your qubit chris we're coming yeah +so let's let's let's give him a few minutes and maybe he's still watching and he'll he'll answer the +questions better than i could well while daniel is sending him sending him a a lifeline can i just from +the from the the rookie perspective on this session one seemed like there was a big time window like +the historical piece over a using an external clock and a timeline this one felt like the parabola +and the upper down and the entanglement was almost instantaneous in terms of that plank length that +we were really holding up as comparative to that timeline a chronological timeline and so i'm guessing +that now that we've established those two as different where does that go but again that's +just the rookies view of things hey chris welcome yeah hi this is a very interesting discussion +um and it it's uh given me some pointers as to what to emphasize at the beginning of session three +uh but i i can say in response to dean's question that the time when alice isn't looking is the time +during which bob is making his measurement and then his next preparation step +so if you if you consider again that slide that shows uh sort of the fineman representation of what's +happening as this unitary uh flow that's what bob is doing and i think andrew said that +oh from alice's point of view bob is just some quantum system and she's banged on that system by asking a question +and now that system is doing something or other that she can't see until it it i.e bob bangs on his side of the screen +and she gets some information by making a measurement uh so she she may be not looking for a very short period of time +uh right i mean according to her internal clock she's not looking for just the time it takes for one of her +internal ticks to take place uh but uh in in general uh we're not saying anything about how long that time is +right if you think of this in a scattering theory way uh then we're talking about the time between two protons +leaving the rings on the uh large hadron collider to the time that all the stuff that comes out of their +collision is measured by the detectors which is a nanosecond or something like that some very short period of time +uh so just just think of that interval as the the measure of the intrinsic +uh time dependence external time dependence of the interaction someone has to do something and then +someone else has to do something and there's a gap uh between what the first person does and what they see +in response oh let's see what's the other question uh the the other central question that came up today +was the question of why observing something requires action and of course in the in the classical picture +observing something doesn't require any action at all uh the the event just happens and this is the the +fundamental if you will distinction between classical and quantum theory is that making an observation +in in in quantum theory is taking an action on the world and there's a lovely uh metaphor that's that's not +correct in detail but it's it's very useful i think conceptually uh which heisenberg came up with to try +to explain his uncertainty principle to his classical colleagues he said look suppose i'm trying to +to to to to see a very small particle and measure its motion oh i have to have a microscope to see this particle +and to measure its motion i need to to light it up i need to shine some light on it and those photons of +light are actually impacting the thing i'm trying to see +so the the the act of observation itself requires banging something into an object +and when we're just looking out at the world it's the sun that's banging the light into the object +not us but we can we can sort of merge those two systems together because it's the sun that's taking the +action that lets us get some information but the sun is perturbing the things that we're looking at +by banging photons into them um and i think that that picture of of heisenberg's is what uh he used to +to actually derive the uncertainty principle in terms of momentum and uh position because banging something +into something is transferring momentum to it photons transfers of momentum to the tiny particle in the +microscope and that perturbs its position on the microscope stage which is precisely what heisenberg +is trying to measure so that's his his point in that analogy observation itself always takes action +and that's why observation and preparation are duals in quantum theory because they each require a minimum +one unit of action and i think andrew you said the the duality can be thought of in terms of time +and variance and that's exactly correct preparation is just measurement run backwards right and vice versa +and if you let me say a few words i think this also answers my own confusion and the one that alexi the +question that alexi posed earlier which is this duality between preparation and measurement in your +picture that you're bringing up chris with heisenberg by shining photons on a particle that i'm trying +to observe um these pictures of preparation and measurement are dual because in in doing so i'm +constraining the system in trying to measure it i'm constraining it so it is so in me trying to gather +information by shining photons on it what i'm really doing is i'm constraining it i'm just saying +you can only take this this you know momenta or or position is would that be correct +uh yeah in fact you're in the microscope example it's particularly clear you're you're bouncing a photon +off of the object and then it's that photon that you're detecting right oh there was one other +question that uh can be answered quickly it was the question about uh what's the reference frame that +uh alice is using when she's uh measuring up or down measuring her spin up or down and of course +there's there there's several reference frames involved here oh so let's work inwards uh in a real system +alice is sitting here she's got a stern gerlock apparatus which is a vertical magnet apparatus +and she's measuring the spin of an electron that's that's traveling through this stern gerlock device +and what the stern gerlock device does is is orient the spin uh with a magnetic field that's that's not +homogeneous so it ends up pointing physically up uh as it comes through the apparatus so how is that +apparatus oriented i mean what defines the up direction for that apparatus well in the practical setting what +defines that apparatus orientation is the earth's gravitational field right we're sitting we're in +we're here in the laboratory and the earth's gravitational field tells us what direction is +down it's toward the center of the earth and so up is away from the center of the earth and that's +what defines the axis for the apparatus which is a great big piece of metal so it's very responsive to the +gravitational field so now we can ask about alice right alice is also embedded in this gravitational field +and she also can detect the gravitational field she can tell which way is down toward the center of the earth +by the the force acting on her feet and that measurement is taking place in alice's brain +so in alice's brain there's a representation of which way is up and that's the internal reference frame of +alice that lets her do the experiment you know if if she was freely floating in space but the apparatus was +anchored to the gravitational field she wouldn't be able to tell that it was anchored to the gravitational field +that's einstein's the point that einstein makes in all of those experiments where he has observers flying +around in spacecraft uh that they have to have their own reference frame uh what what einstein didn't +emphasize was that reference frame is in their head it's encoded by their nervous system uh so +uh i i think it was daniel who said this exactly correctly that uh the physical object uh in her +head the the actual neural network that tells her which way the gravitational field is uh is itself a +semantic object i mean that's that's what gives the word up and down meaning for alice and what is what +is what do i mean when i say the word meaning i mean it lets her take some action um semantics boils down to +actionability if it gives things significance and this this was the point that that bateson was making +when he said differences that make a difference uh for an organism uh something is meaningful if it +actually makes a difference for what i can do uh in this world that i'm trying to survive in +oh so i'll i'll leave it at that but this is this is very very useful for uh leading off the next +session so thank you very much for for all that discussion awesome thank you i guess if i could +say one thing here is that uh let me scroll back up to the three points early in the slides +differences that make a difference would tell you that you require action to to perturb the system and +measure and then the semantics or the contextuality that may be encoded in the in the in the nervous +system would tell you what is the difference that is being made would that be correct yeah right what +what tells you that there's a difference there it's it's it's what what you're able to detect and +represent internally oh it's almost like step one do differently step two you have to ask it to a +different thing and then three you need the semantic frame that makes a and b different +difference differences all the way down um well we will transcribe this and fix the few patchy moments +enter we'll work together on updating the transcript so that everybody who wants to be well prepared +for the measurement of the third lecture will be absolutely in place +in place thank you chris for joining that's a fun fun cap to this discussion so all right thank you see +you all next time bye +thank you + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +Thank you. diff --git a/data/video/activeinferenceinstitute/GuestStream/GuestStream_062/metadata.json b/data/video/activeinferenceinstitute/GuestStream/GuestStream_062/metadata.json index 2dd0639f6..6ba532f0d 100644 --- a/data/video/activeinferenceinstitute/GuestStream/GuestStream_062/metadata.json +++ b/data/video/activeinferenceinstitute/GuestStream/GuestStream_062/metadata.json @@ -21,11 +21,6 @@ "generated" ], "parts": [ - { - "video_id": "boost_gs062", - "url": "https://www.youtube.com/watch?v=boost_gs062", - "title": "" - }, { "video_id": "ciFMH8svne4", "url": "https://www.youtube.com/watch?v=ciFMH8svne4", diff --git a/data/video/activeinferenceinstitute/Other/BNLnbOFdgc0/README.md b/data/video/activeinferenceinstitute/Other/BNLnbOFdgc0/README.md deleted file mode 100644 index 79260b9b4..000000000 --- a/data/video/activeinferenceinstitute/Other/BNLnbOFdgc0/README.md +++ /dev/null @@ -1,5 +0,0 @@ -# BNLnbOFdgc0 - -Series: **Other** - -- [Basics of Active Inference (Lecture) ~ Ben White ~ Active Inference for the Social Sciences 2023](https://www.youtube.com/watch?v=BNLnbOFdgc0) — `BNLnbOFdgc0` diff --git a/data/video/activeinferenceinstitute/Other/BNLnbOFdgc0/metadata.json b/data/video/activeinferenceinstitute/Other/BNLnbOFdgc0/metadata.json deleted file mode 100644 index c43ca9cd7..000000000 --- a/data/video/activeinferenceinstitute/Other/BNLnbOFdgc0/metadata.json +++ /dev/null @@ -1,177 +0,0 @@ -{ - "series": "Other", - "item": "BNLnbOFdgc0", - "source": "youtube", - "channel": "ActiveInferenceInstitute", - "category": "Courses/ActiveInferenceForTheSocialSciences", - "episode": null, - "title": "Active Inference for the Social Sciences — Basics of ActInf (Lecture) ~ Basics of ActInf (Lecture) — Course home", - "date": "2023-07-11", - "guests": [ - "Ben White" - ], - "other_participants": [ - "Daniel Friedman" - ], - "github": "https://github.com/ActiveInferenceInstitute/ActiveInferenceJournal/tree/main/data/video/activeinferenceinstitute/Other/BNLnbOFdgc0", - "enriched_from": [ - "coda", - "generated", - "youtube" - ], - "parts": [ - { - "video_id": "BNLnbOFdgc0", - "url": "https://www.youtube.com/watch?v=BNLnbOFdgc0", - "title": "Basics of Active Inference (Lecture) ~ Ben White ~ Active Inference for the Social Sciences 2023", - "duration": 5575.0, - "upload_date": "" - } - ], - "sessions": [ - { - "index": 1, - "session_name": "BNLnbOFdgc0_sess01", - "start": "0:00:00", - "title": "Introduction" - }, - { - "index": 2, - "session_name": "BNLnbOFdgc0_sess02", - "start": "0:01:05", - "title": "Aims" - }, - { - "index": 3, - "session_name": "BNLnbOFdgc0_sess03", - "start": "0:05:41", - "title": "Visual Perception" - }, - { - "index": 4, - "session_name": "BNLnbOFdgc0_sess04", - "start": "0:08:08", - "title": "Perceptual Inference" - }, - { - "index": 5, - "session_name": "BNLnbOFdgc0_sess05", - "start": "0:10:17", - "title": "Readings" - }, - { - "index": 6, - "session_name": "BNLnbOFdgc0_sess06", - "start": "0:10:45", - "title": "Guiding Questions" - }, - { - "index": 7, - "session_name": "BNLnbOFdgc0_sess07", - "start": "0:11:20", - "title": "Attunement" - }, - { - "index": 8, - "session_name": "BNLnbOFdgc0_sess08", - "start": "0:12:20", - "title": "Ocean" - }, - { - "index": 9, - "session_name": "BNLnbOFdgc0_sess09", - "start": "0:12:36", - "title": "affordances" - }, - { - "index": 10, - "session_name": "BNLnbOFdgc0_sess10", - "start": "0:14:03", - "title": "inactivism" - }, - { - "index": 11, - "session_name": "BNLnbOFdgc0_sess11", - "start": "0:17:21", - "title": "representations" - }, - { - "index": 12, - "session_name": "BNLnbOFdgc0_sess12", - "start": "0:22:29", - "title": "extended active inference" - }, - { - "index": 13, - "session_name": "BNLnbOFdgc0_sess13", - "start": "0:24:49", - "title": "decoupled agents" - }, - { - "index": 14, - "session_name": "BNLnbOFdgc0_sess14", - "start": "0:26:28", - "title": "dark room problem" - }, - { - "index": 15, - "session_name": "BNLnbOFdgc0_sess15", - "start": "0:29:08", - "title": "Colin Klein 2018" - }, - { - "index": 16, - "session_name": "BNLnbOFdgc0_sess16", - "start": "0:31:48", - "title": "Active Inference Agential Behavior" - }, - { - "index": 17, - "session_name": "BNLnbOFdgc0_sess17", - "start": "0:33:00", - "title": "Active Inference and Folk Psychology" - }, - { - "index": 18, - "session_name": "BNLnbOFdgc0_sess18", - "start": "0:35:17", - "title": "Direct Response Decline" - }, - { - "index": 19, - "session_name": "BNLnbOFdgc0_sess19", - "start": "0:39:54", - "title": "Expectations" - }, - { - "index": 20, - "session_name": "BNLnbOFdgc0_sess20", - "start": "0:43:31", - "title": "Examples" - }, - { - "index": 21, - "session_name": "BNLnbOFdgc0_sess21", - "start": "0:44:15", - "title": "Allostasis" - }, - { - "index": 22, - "session_name": "BNLnbOFdgc0_sess22", - "start": "0:48:29", - "title": "Emotion" - }, - { - "index": 23, - "session_name": "BNLnbOFdgc0_sess23", - "start": "0:54:14", - "title": "Allostatic Control" - }, - { - "index": 24, - "session_name": "BNLnbOFdgc0_sess24", - "start": "0:58:32", - "title": "Interceptive Prediction" - } - ] -} diff --git a/data/video/activeinferenceinstitute/Other/JkORFoyk8o8/README.md b/data/video/activeinferenceinstitute/Other/JkORFoyk8o8/README.md deleted file mode 100644 index 1822dcc55..000000000 --- a/data/video/activeinferenceinstitute/Other/JkORFoyk8o8/README.md +++ /dev/null @@ -1,5 +0,0 @@ -# JkORFoyk8o8 - -Series: **Other** - -- [Introduction (Lecture) ~ Avel Guénin-Carlut ~ Active Inference for the Social Sciences 2023](https://www.youtube.com/watch?v=JkORFoyk8o8) — `JkORFoyk8o8` diff --git a/data/video/activeinferenceinstitute/Other/JkORFoyk8o8/metadata.json b/data/video/activeinferenceinstitute/Other/JkORFoyk8o8/metadata.json deleted file mode 100644 index 86479b48f..000000000 --- a/data/video/activeinferenceinstitute/Other/JkORFoyk8o8/metadata.json +++ /dev/null @@ -1,124 +0,0 @@ -{ - "series": "Other", - "item": "JkORFoyk8o8", - "source": "youtube", - "channel": "ActiveInferenceInstitute", - "category": "Courses/ActiveInferenceForTheSocialSciences", - "episode": null, - "title": "Active Inference for the Social Sciences — Introduction Lecture ~ Introduction (Lecture) — Course home", - "date": "2023-07-07", - "guests": [ - "Avel Guénin-Carlut" - ], - "other_participants": [ - "Daniel Friedman", - "Ben White" - ], - "github": "https://github.com/ActiveInferenceInstitute/ActiveInferenceJournal/tree/main/data/video/activeinferenceinstitute/Other/JkORFoyk8o8", - "enriched_from": [ - "coda", - "generated", - "youtube" - ], - "parts": [ - { - "video_id": "JkORFoyk8o8", - "url": "https://www.youtube.com/watch?v=JkORFoyk8o8", - "title": "Introduction (Lecture) ~ Avel Guénin-Carlut ~ Active Inference for the Social Sciences 2023", - "duration": 5614.0, - "upload_date": "" - } - ], - "sessions": [ - { - "index": 1, - "session_name": "JkORFoyk8o8_sess01", - "start": "0:00:00", - "title": "Introduction" - }, - { - "index": 2, - "session_name": "JkORFoyk8o8_sess02", - "start": "0:02:19", - "title": "Teaching Team" - }, - { - "index": 3, - "session_name": "JkORFoyk8o8_sess03", - "start": "0:04:48", - "title": "Motivation" - }, - { - "index": 4, - "session_name": "JkORFoyk8o8_sess04", - "start": "0:06:15", - "title": "Explanation of Cultural Dynamics" - }, - { - "index": 5, - "session_name": "JkORFoyk8o8_sess05", - "start": "0:09:15", - "title": "Cognitive Laws of Behavior" - }, - { - "index": 6, - "session_name": "JkORFoyk8o8_sess06", - "start": "0:13:24", - "title": "Active Inference" - }, - { - "index": 7, - "session_name": "JkORFoyk8o8_sess07", - "start": "0:17:10", - "title": "Normativity" - }, - { - "index": 8, - "session_name": "JkORFoyk8o8_sess08", - "start": "0:23:52", - "title": "The Case" - }, - { - "index": 9, - "session_name": "JkORFoyk8o8_sess09", - "start": "0:24:54", - "title": "Public Opinion" - }, - { - "index": 10, - "session_name": "JkORFoyk8o8_sess10", - "start": "0:28:49", - "title": "Nationalism" - }, - { - "index": 11, - "session_name": "JkORFoyk8o8_sess11", - "start": "0:32:48", - "title": "Energy Capture" - }, - { - "index": 12, - "session_name": "JkORFoyk8o8_sess12", - "start": "0:36:56", - "title": "Structural Agency" - }, - { - "index": 13, - "session_name": "JkORFoyk8o8_sess13", - "start": "0:40:37", - "title": "Landscape" - }, - { - "index": 14, - "session_name": "JkORFoyk8o8_sess14", - "start": "0:43:32", - "title": "World Work" - }, - { - "index": 15, - "session_name": "JkORFoyk8o8_sess15", - "start": "0:52:17", - "title": "Projects" - } - ] -} diff --git a/data/video/activeinferenceinstitute/Other/TOZp_XNYijQ/README.md b/data/video/activeinferenceinstitute/Other/TOZp_XNYijQ/README.md deleted file mode 100644 index b8e335231..000000000 --- a/data/video/activeinferenceinstitute/Other/TOZp_XNYijQ/README.md +++ /dev/null @@ -1,5 +0,0 @@ -# TOZp_XNYijQ - -Series: **Other** - -- ["Physics as Information Processing" ~ Chris Fields ~ Lecture 3](https://www.youtube.com/watch?v=TOZp_XNYijQ) — `TOZp_XNYijQ` diff --git a/data/video/activeinferenceinstitute/Other/TOZp_XNYijQ/metadata.json b/data/video/activeinferenceinstitute/Other/TOZp_XNYijQ/metadata.json deleted file mode 100644 index 4f2bbe2da..000000000 --- a/data/video/activeinferenceinstitute/Other/TOZp_XNYijQ/metadata.json +++ /dev/null @@ -1,125 +0,0 @@ -{ - "series": "Other", - "item": "TOZp_XNYijQ", - "source": "youtube", - "channel": "ActiveInferenceInstitute", - "category": "Courses/PhysicsAsInformationProcessing_ChrisFields", - "episode": null, - "title": "Chris Fields ~ Physics As Information Processing ~ 2023 Lecture 3 ~ Lecture 3 — Course information", - "date": "2023-07-13", - "guests": [ - "Chris Fields" - ], - "other_participants": [ - "Daniel Friedman" - ], - "github": "https://github.com/ActiveInferenceInstitute/ActiveInferenceJournal/tree/main/data/video/activeinferenceinstitute/Other/TOZp_XNYijQ", - "slides_url": "https://docs.google.com/presentation/d/1YZS8aunpEfoZGvhVk01fulDFOkIKj9PMUeszcKZ195g/edit?usp=sharing", - "enriched_from": [ - "coda", - "generated", - "db", - "youtube" - ], - "parts": [ - { - "video_id": "TOZp_XNYijQ", - "url": "https://www.youtube.com/watch?v=TOZp_XNYijQ", - "title": "\"Physics as Information Processing\" ~ Chris Fields ~ Lecture 3", - "duration": 3699.0, - "upload_date": "" - } - ], - "sessions": [ - { - "index": 1, - "session_name": "TOZp_XNYijQ_sess01", - "start": "0:00:00", - "title": "Introduction" - }, - { - "index": 2, - "session_name": "TOZp_XNYijQ_sess02", - "start": "0:00:45", - "title": "Quantum Information Theory" - }, - { - "index": 3, - "session_name": "TOZp_XNYijQ_sess03", - "start": "0:01:18", - "title": "Last Time" - }, - { - "index": 4, - "session_name": "TOZp_XNYijQ_sess04", - "start": "0:03:23", - "title": "Outline" - }, - { - "index": 5, - "session_name": "TOZp_XNYijQ_sess05", - "start": "0:03:46", - "title": "Reference Frames" - }, - { - "index": 6, - "session_name": "TOZp_XNYijQ_sess06", - "start": "0:11:53", - "title": "Embodied Reference Frames" - }, - { - "index": 7, - "session_name": "TOZp_XNYijQ_sess07", - "start": "0:17:17", - "title": "Representation" - }, - { - "index": 8, - "session_name": "TOZp_XNYijQ_sess08", - "start": "0:19:22", - "title": "Observer" - }, - { - "index": 9, - "session_name": "TOZp_XNYijQ_sess09", - "start": "0:22:53", - "title": "Metabolic Tradeoff" - }, - { - "index": 10, - "session_name": "TOZp_XNYijQ_sess10", - "start": "0:24:51", - "title": "Consequences of Metabolic Tradeoff" - }, - { - "index": 11, - "session_name": "TOZp_XNYijQ_sess11", - "start": "0:29:27", - "title": "Three Directions" - }, - { - "index": 12, - "session_name": "TOZp_XNYijQ_sess12", - "start": "0:43:08", - "title": "Interactive QA" - }, - { - "index": 13, - "session_name": "TOZp_XNYijQ_sess13", - "start": "0:49:49", - "title": "Free Energy Principle" - }, - { - "index": 14, - "session_name": "TOZp_XNYijQ_sess14", - "start": "0:52:52", - "title": "Physical Processes" - }, - { - "index": 15, - "session_name": "TOZp_XNYijQ_sess15", - "start": "0:55:26", - "title": "Environment" - } - ] -} diff --git a/data/video/activeinferenceinstitute/Other/TOZp_XNYijQ/transcript.txt b/data/video/activeinferenceinstitute/Other/TOZp_XNYijQ/transcript.txt deleted file mode 100644 index 0336704af..000000000 --- a/data/video/activeinferenceinstitute/Other/TOZp_XNYijQ/transcript.txt +++ /dev/null @@ -1,1141 +0,0 @@ -## TOZp_XNYijQ - -Hello and welcome everyone. It's July 13th, 2023. We're here in Physics as Information Processing -with the third lecture by Chris Fields. So please check out the course website, -CODA document, where you can submit questions, check out the recordings and descriptions of -all sessions and register if you want to participate in one of the discussion sections. -So thank you again, Chris, to you. Thank you and welcome to Physics as Information Processing -session three, which is about quantum reference frames. So in this course, we've talked about -quantum information theory, which we characterized as a general theory that describes the exchange -of information, exchange of finite information, between two finite agents that are separated -by some boundary. And we've emphasized that this is a topological theory, not a geometric theory. -So it makes no assumptions about an embedding spacetime. And last time, we showed how to characterize -this boundary of script B as an array of N, some finite N quantum bits or qubits, and how -to characterize the interaction of the agents A and B with the boundary as an alternating cycle of preparing -and measuring these qubits in this array. And we showed that their interaction could be represented -as a sum of operators, each of which act on just one of the qubits. And we also talked about the need for each -of the agents, A and B, to choose a reference frame for their interaction with the qubits that defined which way was up. -So what the orientations of the qubits meant. And emphasized that the two observers, A and B, could choose different -reference frames, so different ways of measuring and preparing their qubits that use different meanings for the idea of up for each qubit. -And we ended the last session with a pointer forward toward the quantum formulation of the free energy principle, which can be stated as the claim that interacting systems behave in a way that aligns their reference frames. -And I'll point it out at that time, and I'll point it out at that time, that what this means is that the two agents, A and B, approach entanglement asymptotically, and that therefore the FEP is a classical limit of the principle of unitarity, which is the core principle of quantum theory. -So we'll see in this session why this is true. So what we'll talk about today is what reference frames are in general, how they make observations and actions meaningful, so how they give physics a semantics. -And then we'll talk about the free energy principle. -So what is a reference frame? The idea that physics involves reference frames goes back at least to Galileo, but was probably familiar well before that, just not formalized in the way that Galileo formalized it. -So the key idea can be thought of in terms of motion. -And you're all familiar, for example, with riding a bicycle or driving a car and seeing the world as if it's passing by you at some velocity, with some velocity. -And of course, with respect to the world, you're passing by it with some velocity. -And this idea that velocity is not absolute, but that's relative to some observer who measures it in some reference frame is the foundation for theories of relativity. -So Galileo developed his theory of relativity, which basically was just a way of converting from classical velocities measured by one observer to classical velocities measured by another observer that was moving with respect to the first one. -And this provided the foundation for Einstein's theories of relativity, which modified Galileo's by adding the concept that the observers have to change, exchange information, and can only exchange it at the speed of light, which is finite. -So this idea in physics has a long history. -Of course, the core idea is that measurements are always with respect to something or other. -And the something or other is a reference frame. -And different agents can choose different reference frames. -And this idea is very old. -Here we have a picture of Anubis choosing a feather as the reference frame with which to measure the heart of the pharaoh. -So this goes back to well before the Common Era. -This is about 2000 BCE. -So this is a very old idea that's been elaborated over the centuries as science has developed. -And in the 90s, a fairly big change in how this idea was thought about occurred. -In classical physics, reference frames were always thought of as mathematical abstractions. -And an observer and a reference frame were more or less identified by Galileo. -And that identification is carried forward into Einstein, who thinks about observers, for example, sitting on photons, and observers in spaceships, and observers riding on railroad trains, and this sort of thing. -And always that just means think of a reference frame that's doing something or other. -So a reference frame in classical physics is essentially just a coordinate system. -And these abstract coordinate systems can be employed in thought experiments. -And so they are a way to illustrate classical intuitions about motion. -And in the 90s, as quantum information theory developed, it was realized that a reference frame actually consists of a physical object. -I mean, even in the picture from the Egyptian tomb, the reference frame is a physical feather. -And that means that the reference frame itself is described by quantum theory. -It's some quantum system. -And if you think about reference frames that are familiar, for example, meter sticks for measuring distance, or clocks for measuring time, or gyroscopes for batteries, -or even the Earth's gravitational fields, which typically defines up for us, -those are all physical objects that can be described by quantum theory. -Well, this was an important insight because it led to the realization that reference frames as quantum systems are unique. -And uniqueness in quantum theory is very strong. -There's the no-cloning theorem, which states that you can't reproduce, exactly reproduce, any unknown quantum state. -And that applies to reference frames because they encode what Bartlett et al. in that paper called non-fungible information, -which just means information that can't be completely represented in any finite bit stream. -So unless you know the state of a reference frame, you can't clone it, since these reference frames encode more than any bit strings worth of information in their quantum phases. -So they're unique, and they're unique, and their states are unknown. -And the consequence of this is that if Alice wants Bob to use her reference frame to make a measurement, she can't just send him a description. -She has to actually send the reference frame. -And this has an important consequence for physics because it makes all replicate measurements approximate, -because observers obviously aren't using exactly the same physical objects as their reference frames. -But there's a deeper implication of this, which is that Alice's ultimate rock bottom reference frame is actually implemented by her body, by her brain. -And using any external reference frame requires having this internal reference frame that makes the external reference frame meaningful and usable. -So for example, this thing is a piece of a meter stick, a third of a meter stick. -But for me to use it to make measurements, I have to actually have some concept of distance. -And that concept of distance is implemented by my nervous system. -And that implementation is my reference frame for length. -And it's what gives meaning to any external measurement of length. -It's what makes my measurements of length actionable by me using my effectors or action capabilities, whatever they are. -So my internal reference frames are what make information I get from the world actionable. -And in Bateson's sense of differences that make a difference, actionability is the core of semantics. -Something is meaningful if I can do something about it. -So this idea that we all contain embodied reference frames gives physics a semantics by giving observers a semantics. -And this is new. -This is an idea that never really appeared in physics before, that measurements automatically had to have some kind of semantics. -But reference frames give semantics to measurements. -So here's an example of an embodied reference frame. -This is the reference frame that bacteria use for chemotaxis, i.e. to move toward molecules in the environment that they like, like sugars, and to move away from molecules that they don't like, like acids. -And it's been worked out in great detail over decades of experimental work. -And this is a rough schematic. -The reference frame has three parts. -There's the measurement part, which is conveniently located in the front of the bacterium. -And it consists of a bunch of receptors that probe the environment. -And they can bind molecules, which are called ligands, out in the environment. -And when they bind them, they send a signal into the back of the bacterium. -And it turns out that there's a little miniature reference frame attached directly to these ligands, which is this R and B system, which controls the ligands, not the ligands, the receptor sensitivity. -Then the back of the reference frame is the action component. -And it consists of a set of molecules that control a motor, a molecular motor, that derives the flagella on the bacteria. -And the flagellum allows it to swim forwards or turn in some other direction. -And depending on how it spins, the creature either goes in the direction of something it likes or it stops and turns around. -And the stuff in the middle is a little molecular feedback loop that represents the amount of ligand in the environment that the bacterium expects to see. -So this is the middle of the reference frame, and it defines the default value for what the bacterium is observing in its environment. -And if you're thinking in terms of the free energy principle or Bayesian inference, then you'll immediately see the concentration of this molecule is encoding a prior probability, the bacterium's prior probability for what its environment is like. -And the zero point, the default value of that encoding is the ligand concentration in which the bacterium does nothing. -So a QRF, in general, can be characterized as an expected value, which is here the concentration of this molecule, key Y phosphorylated, plus a bunch of computation, which is all the rest of the things in this diagram. -So this is a very simple QRF that's embodied by a bacterium. -Here's a more complicated QRF that's embodied by eukaryotic cells, so more complicated cells like our cells. -And this is the reference frame that makes decisions about whether cells are going to divide. -So on the top, you see the cells' membrane, and there are all sorts of receptors sticking out of it that detect hormones and neurotransmitters and growth factors and this, that, and the other thing, many of which are produced by other cells. -And those receptors communicate, and those receptors communicate via a lot of chemistry with the nucleus, which is contained in another membrane, and so serves as, in a sense, the outside upon which the reference frame is acting. -And what this reference frame does is control the transcription of large numbers of molecules from DNA. -So it basically controls what proteins the cell is making. -And those proteins end up doing something, like restructuring the cytoplasm and actually driving cell division. -And the stuff in the middle of this reference frame, which I've labeled the integrative center, are three molecules called phylogen. -They're connected by positive, by bidirectional feedback loops. -And they basically negotiate the default value of the environment at which the cell does nothing. -So they're the middle of the reference frame that represents the cell's expectation values about what's going on in its environment. -So you can imagine now what reference frames encoded by our nervous system look like. -They're extremely complicated and exist at multiple scales, from the scale of molecules in cells to the scale of large-scale connections between parts of our brains. -So how do we represent all this stuff? -There are many ways to do this. -But this is a particularly simple way. -So it assumes that whatever's in the middle, which is here called c prime, is the most general thing that the inputs can give information to, and the most general thing that can give information to the outputs. -And most general here has a category theoretic definition. -So it could be stated in a way that's equivalent across many different specific mathematical forms. -But if you're familiar with neural networks, this will look to you a lot like the skeleton of a variational autoencoder. -And that's basically what it is. -And I don't want to deal with this formalism except to say that it exists and is very convenient, except to say that we can attach this kind of formal diagram as a representation of a QRF. -So we can attach this to our representation of the boundary between two interacting agents. -And we can use these formal structures to drive the measurement operators that actually implement physical changes on this boundary. -And I'll represent this whole business with just a triangle labeled Q for some QRF that's attached to the boundary. -And we can use this simple notation then to investigate what sorts of structures one gets when one starts to think about systems as implementing specific QRFs. -And here's the structure of the minimal system that we can really think of as an interesting observer. -And this is a system that has some QRF that I've called E that looks at its environment, the part of its environment that it can actually observe and act on. -And it defines a sector of the boundary, which mathematically is just the image of this QRF E that the system can manipulate. -So this is the part of the boundary. -So this is the part of the environment or the part of the boundary that the system actually sees and can do something about. -And for this observer to be interesting, it needs to be able to detect the difference between what it sees now and what it saw just then so that it can decide whether to take some action or other. -So it needs a memory. -So it needs a memory. -And that requires another QRF that has some allocated piece of the boundary. -Recall, the boundary is where classical information gets written in this purely quantum theoretic formalism. -So it needs a memory sector that that QRF could act on. -And then in order to do this computation and specifically in order to write information to the boundary, in order to actually flip bits on the boundary, it needs free energy. -So the rest of its boundary gets devoted to extracting free energy from the environment and exhausting free energy, waste heat, in other words, into the environment. -And so it has this nice little thermodynamic cycle that drives everything else. -And finally, in order to tell now from just a second ago, the information in memory, it needs a clock that marks the information in memory as not current. -So as past information. -And if it can store multiple memory records, it needs to be able to count and say, this piece of remembered information is actually more recent than that piece over there. -So it has to implement a little internal clock. -And that clock is its internal time QRF. -And we've seen that QRFs act on boundaries. -This clock acts on a boundary. -But it doesn't act on B. -It acts on an internal boundary that's essentially the boundary between it, the clock, and the points at the ends of these QRFs that actually encode information that's either read from the environment or written to the environment. -So already this minimal observer has a little bit of hierarchical structure. -Which is that its clock has to, in some sense, be higher up in the hierarchy than the parts, the QRFs that interact directly with the environment. -So if you have a structure like this, you face a metabolic trade-off. -For a fixed amount of energy that you can extract from the environment, you can allocate it either into learning more about the environment or remembering more about what you learned before. -And so any system has to make this decision about how they're going to use energy. -Are they going to devote it to learning more or remembering more? -And your laptop makes that decision. -Memory is cheap for your laptop. -Your brain makes that decision. -Memory is much more expensive for your brain. -And all cells have to make that decision. -And the universal solution to this trade-off is coarse-graining memory, i.e. remembering things in less detail than you saw them in the beginning. -Ander, you have a comment? -Yes. -I wanted to ask a question now before it gets too late. -So two or three slides ago, when we were looking at the diagram of the cone-cocoin diagrams, I was wondering if the classifiers identified with measurement and preparation correspond to sensory and action degrees of freedom in the earlier picture with the bacteria. -And the classifier in the middle, would that, in that picture, would it correspond to the default chemical concentration? -Yes, exactly. -Okay. -Perfect. -Yeah, that crossing point in the middle is specifically where the default is encoded. -Thank you for pointing that out. -Okay. -So let's follow the consequences of this metabolic issue. -If you have a system that's a little bit smarter than our minimal system that can actually detect and measure the states of objects, it has to do even more work because it has to encode even more QRFs. -So think of measuring the state of some object like a voltmeter in a laboratory. -If I tell you to go into the lab and tell me what the voltmeter is reading, then the first thing you have to do is find the voltmeter. -And your process of finding the voltmeter is independent of your process of reading what it's actually saying on its dial. -So in order to measure the state of an object, you have to have two QRFs. -One that lets you find the object and then the other that lets you measure the state of interest of the object. -And we call these the R sector for reference that lets you find the object and the P sector for pointer because old instruments used to have pointers on dials that read analog numbers. -But what thinking about object identification tells you is that every object that you can identify leads to having to sector both the informative part of your boundary and the memory part of your boundary into two new sectors. -One to store information about object identification. -And one to store information about the states of those objects. -So this requires much more metabolism. -And again, the solution is coarse-grained. -Typically what we coarse-grained is our memory of the object that we've identified. -So what does this do to systems? -Because making measurements costs energy and recording memories costs energy, we have to make decisions about what QRFs to use at what time given our finite energy budgets. -We can't look at the world with all of our measurement capabilities at the same time because it's too expensive. -And now we see the effect of a theorem in quantum information theory, which is that if two QRFs can't be deployed simultaneously for whatever reason, that corresponds to them not commuting as operators. -And if they don't commuting as operators, they have to be implemented by compartments that are sufferable. -So compartments that only communicate classically. -So you end up looking like this diagram where there are a bunch of components that do, um, -the consequence of this theorem is that as soon as one has too many sensory and motor capabilities to run all at the same time given your energy budget, -the amount of memory of the same time given your energy budget. -So you end up being compartmentalized into compartments that only communicate classically. -So you end up looking like this diagram where there are a bunch of components that do, um, sensory motor loops or memory. -one or the other, and they can only communicate classically, and they're controlled by some sort -of attention system that selects which ones to activate, depending on the amount of free energy -that's available. And clearly this clock function becomes part of this control system. -So the physics alone of any system that's acting as a fairly, that executes a fairly complicated -interaction with its environment, gives us this hierarchical compartmentalized structure -without having to add any biological or cognitive science type assumptions at all. -So we can go in three directions from this. We can go deeper into the biology, which we're going to do -in October. We can go deeper into the physics, which we're going to do in August and September. -And we can go deeper into this issue of alignment, which is what we're going to do today. -So there are four possibilities if you think about two agents that have QRFs that are interacting -across this boundary. And the first one is trivial. If the B agent, the agent on the right side, -doesn't actually employ any QRFs, if it just interacts with the boundary thermally, -then there is no alignment. You're in a trivial situation where A can measure something about B, -but not in return. If the B agent deploys a QRF that's much broader in scope than the A agent, -what we show here in panel B, then the A agent can learn something about the B agent, -that misses most of what the B agent is doing. In panel C, we have agents with QRFs of the same -size on the boundary, but they're not lined up correctly. So each one misses part of what the -other is doing. And finally, in D, we have the optimal situation where they have QRFs of the same -size, and they're actually lined up. So each sees what the other is doing, at least on this sector -of the boundary. And since the agents have to allocate some of the boundary to free energy supply -and can't look at that part, then they can't tell what's happening across their entire boundaries. -They can only tell what's happening in this sector, where they can actually deploy a QRF. -But let's now look at this interesting situation, Situation D, where the QRFs are aligned between A and B. -As I said, that's the optimal channel for A and B. -And if we think about this, we can write the variational free energy, or the uncertainty, -or the prediction error, for A, for the QRFx, as the action that it sees, i.e. B's action, -the difference between that and what it expects B to do, so what it can predict about B's action. -And this is just the notion of free energy from the free energy principle, except translated into -the situation where we have two QRFs. And A can measure what B has done, -but it can also use its QRF to predict, because it's predicting the very state that it's going to observe next -of the same sector that B is using. -Now, if we look at this diagram, we can see easily that the way to minimize the variational free energy, -and therefore maximize predictive capability, -is for these two QRFs to actually be the same. -So if A is doing the same thing to the boundary that B is, A can predict B's next action perfectly, -because B is going to do whatever A is doing. -So what the free energy principle tells us in this case is that A and B, -as they minimize their variational free energies, will approach having identical QRFs. -Well, again, we have a theorem that tells us that if these QRFs are shared, A and B are not separable, -but entangled. -And this is also fairly easy to see, because it depends on this notion of non-fungibility. -If X, A, and X, B both encode non-fungible information, which is quantum systems they do, -their states aren't clonable, so they can't be copies of each other, -because one can't clone an unknown system. -So they can only be identical if they're actually identical systems. -This is exactly the same point made earlier about Alice having to actually send her QRF to Bob -if she wants Bob to use her QRF. -So minimizing VFE for quantum systems is becoming entangled. -So the classical free energy principle, which tells us to minimize VFE, -actually drives quantum systems toward entanglement. -And that's why it's a special case of the principle of unitarity, -which just says that any isolated joint system is going to asymptotically approach entanglement. -So this raises an obvious question, which is why all of us aren't entangled with our environments all the time, -and entangled with each other. -And the answer is that entanglement is really hard for complex systems. -And this illustrates the sense in which the FEP is moving systems in two different directions at once. -The FEP is driving systems to maintain their separability, -but it's also driving systems toward entanglement. -And as the system becomes more and more complex, -the difference between those two outcomes becomes larger and larger. -So recall from last time that the condition for separability -is that the two systems have dimensions that are much larger than the dimension of their boundary. -And this is the condition that allows them to have internal states, -to have states that aren't located on their boundaries. -And clearly, to formulate the free energy principle, you have to have internal states. -So if you have a complex QRF that's doing a lot of computation, -then you need a lot of internal degrees of freedom to implement that computation. -So you need a lot of internal states. -So you have to have large dimensions. -So complexity favors separability. -As a system becomes more complex, it has to have more internal states. -So it has to be, in a sense, more separable from whatever it's interacting with. -Now, think about this classically. -If you have a complex system that's doing a lot of computation, -then it requires a lot of free energy, -and it has to dissipate a lot of entropy in the form of heat, typically. -But something, some form of energy transfer that its environment is going to see as noise, -as not informative. -So a complex system is working very hard to predict its environment's next state. -But by working very hard, it's dumping a lot of entropy into its environment. -And putting in that extra entropy makes the environment's behavior even harder to predict. -So the more you work at predicting what the environment is going to do, -the more you yourself make the environment's behavior less predictable. -And this is the paradox of the FEP, which is not really a paradox. -It's just what the physics does. -So we can think of the FEP by driving systems to increase their distinctness from their environments, -or to self-evidence, as Carl Fristin calls it, -as driving this increased complexity that both makes predictions better -and makes predictions harder by transferring entropy into the environment. -So what this tells us is that the FEP by itself -will drive the development of systems with larger and larger complexity. -And this clearly has enormous implications for evolutionary theory, -which we'll talk about in October. -But I want to emphasize here that this is just the physics. -It's just what the FEP does when it's forced to maintain a boundary. -And before stopping here, -I want to consider one aspect of this in particular, -which is language. -For example, human language, -but any kind of language connecting any systems -that are communicating with QRFs that are fairly close. -And language is a complex QRF -that we basically share. -Otherwise, things like this interaction would be impossible. -But we also remain distinct systems. -So the question arises, -how do we manage to do this? -How do we manage to largely share a QRF -while escaping becoming entangled? -And this is a question that's largely been ignored in physics -because physicists have tended to ignore language -and just assume that observers can communicate classically. -So, for example, in talking about a Bell-EPR experiment, -which is a canonical experiment for measuring entanglement, -the observers have to be able to share their results -or they can't observe entanglement, right? -The observed result is a bunch of observational statistics -that violate Bell's inequality. -And those statistics can only be computed -given the results that both of the observers obtain. -So the observers have to share a language -to encode these results. -So how come the observers aren't entangled? -How come they are still separable? -What we end up appealing to -is some sort of common cause in the past -to explain why they share a language. -They both learned it from their culture. -So their culture is some sort of common cause. -But all such assumptions are fine-tuning assumptions. -So they all have basically the status -of the value of something like the fine structure constant -without which we wouldn't have atoms, -and so we wouldn't have us. -This value has to be within very small limits -of the value we actually observe. -And that's a case of fine-tuning. -The universe seems fine-tuned in a way -that allows us to exist. -And if it wasn't fine-tuned in that way, -we wouldn't exist, and so we wouldn't observe anything. -But fine-tuning assumptions are a problem. -They're a pointer to something we don't understand. -And so I just want to leave you with the point -that we have to be very careful -when talking about interacting observers. -So in talking about two components of some system -that jointly interact with some environment, -but also communicate classically. -And if you think back a few slides, -that's exactly the structure -that we've shown that the physics forces -any complicated system to have. -So next time, what we're going to focus on -is this very question -of how we describe observers -that are both interacting with the quantum system -and communicating classically in some language. -So that session will be in August. -At the end of this month, -there's a discussion of this session with Ander. -And I encourage you all to -use the interactive Q&A to ask questions. -You've been asking very good questions -that I hope I've been managing to answer. -So thank you very much. -And Daniel, over to you. -Cool. -Well, we have a few minutes for some questions. -So maybe, Ander, -if you want to bring up any questions first. -Oh, unmute, then go for it. -I was curious if you could say any words -on the slide where you had all the possible alignments -of QRFs and what you meant more by interacting thermally. -So when you have a trivial agent on one side, -what exactly does it mean -to say that you're interacting thermally? -If you think about a system -that is only interacting thermally with its environment, -you end up thinking classically about, -for example, an ideal gas -that's in some sort of container. -And some of the molecules are very near the surface. -And their motions depend on the motions -of all the other molecules that are inside. -So there's a set of internal states -that interact randomly -with the external states that are near the boundary. -And so those external states -interact randomly with the boundary. -But notice in thinking in this way, -you've imposed a space-time embedding. -So you've actually used space -as a way of separating the internal states -from the boundary states. -And a system that, -if you drop this assumption of a spatial boundary, -then a system that's only interacting thermally -with its environment -is going to look like a system -that doesn't have any real internal states. -It's going to look like a system -that just has boundary states. -And so a pure thermal interaction -is going to start to look like entanglement. -So we're actually only really interested -in systems that aren't interacting -only purely thermally with their environments. -We're interested in systems -that have enough internal structure -that they're using that thermal energy -to do something internally. -So to give their internal dynamics -some particular shape -so that their internal dynamics -is actually circling around -some well-defined attractor -that requires some energy -to maintain itself. -Yeah, and continuing with that question, -I saw in the paper, -free energy principle for quantum systems, -that if you're in that situation, -when both Keras are non-trivial, -but perhaps B's is larger than A's, -A's perceives non-local hidden variables. -So I wonder if there is a way -to relate those -by taking the limit of making A small -and recover the thermal interaction. -Presumably, if you let A become small, -you should recover that thermal interaction, -so to speak, right? -Yeah, that's a very interesting comment, actually. -So if you're a system -and all of your environment's variables -are effectively hidden to you, -then you can't do any information processing -because you don't have any access -to any of your system, -your environment's variables. -So yeah, your interaction can only be thermal. -I think that's correct. -Right. -Thank you. -I guess, yeah, go for it. -If I may say, -let's just so we, -since we're in this topic -and we don't have to do context switching that way, -on the other hand, -you have noise, right? -In one of those other situations, -I can't recall which one, -I guess, -I guess in the reversed one, right? -When A square F is larger than that of B, -so it encompasses more of the boundary -degrees of freedom, -those that are not overlapping, -A perceives as noise. -Is that, that's correct, right? -Yeah. -Yeah. -A will be looking at B's thermal sector, -B's free energy sector. -Right. -So A may be interpreting, -in a sense, -what it's doing as informative. -And A may even be seeing -what B is doing as informative. -But, -so, -so noise for B may be signal for A, -depending on, -on what, -how A's QRF is structured. -But, -A is, -is seeing -more information -than B is trying to send it. -So, -it's, -it's essentially the, -the reverse of the hidden variable case. -There are -variables -in the other system -that aren't telling you anything -about what the other system -is actually doing. -That, -that's very interesting -in terms of -reading too much -or too little -between the lines, -connecting too many dots -or connecting not enough dots. -But these challenges -where noise -for one -quantum cognitive agent -might be signal for another -and vice versa, -those are the challenges -that we want to address -because to -have a physics -that doesn't recognize -those situations -is -kind of taking the view -from nowhere -or not really addressing -the complexity -of a cognitive engagement. -Yeah, -it's, -it's describing -interactions -between -effectively systems -that don't share -the same language. -I have a question -about the free energy principle -here. -Only in the last -several years -has the quantum -free energy principle -relationship -come into light. -So maybe -could you just share -a little bit about -what incremental -or quantum -advances -enabled that to happen -or what degrees -of freedom -were pre-existing -or added -to enable -work arising -from perception -and action -in neural systems -to also -so naturally -enter this context -with the quantum -information? -Well, -I think -at least -from -a historical -point of view, -that connection -was made -based on the analogy -between a Markov blanket -and a holographic screen. -so these -are structures -developed in -different parts -of physics -that essentially -serve the same -function -of limiting -the information -that one system -can have -about another system. -So, -it's that -functional -analogy -between -a Markov blanket -defined -classically -and a holographic -screen defined -in quantum theory -that -made it seem -obvious -that we could -translate -the free energy -principle -into a quantum -theoretic principle. -Yes. -In Livestream 40, -we talked a lot -about this, -but I think -that remains -one of the -most remarkable -advances -inside or -outside -the free energy -principle -literature -based upon -this functional -acknowledgement -of boundaries -and the -conditions -that they -entail, -which again -are the -challenges -that we -want to -address -lest we -ignore -boundaries. -Right. -Yeah, -precisely. -I mean, -the free energy -principle -is all about -what happens -when you have -a boundary. -The waste -heat, -is this -the actual -thermal exhaust -from our -CPU, -or how can we -think about the -heat being -released -topologically -rather than -geometrically, -like the -location where -it's released? -What does it -mean to -consider -physical -processes -topologically -that we -might otherwise -just use a -geometric way -of thinking -about? -Well, -you basically -put your finger -on it, -and that you're -not worried -about the -spatial -difference -between -the geometric -spatial -difference -between -where one -kind of -information -flows through -the boundary -and where -another kind -of information -flows through -the boundary. -And you -can think -about this -in Markov -blanket -terms. -And the -Markov -blanket is -just a -network, -and some -of the -causal -arrows -flowing -through -that -blanket -network -are -effectively -transmitting -noise. -And you -can think -it in the -same way -with respect -to a -holographic -screen. -If you're -doing a -certain amount -of information -processing, -and this -brings up -the whole -issue of -how efficient -is a -system -in -its -information -processing, -how close -is it -to the -land hour -limit, -basically, -can it -extract -more -energy -from some -bits -than it -has to -use to -modify the -states of -other bits. -things. -That's -what this, -if you will, -metabolic -question is -all about. -In the -case of an -organism, -is it, -can I get -more energy -from some -of the -chemistry -that I -have access -to than -I need -to use -to modify -the chemistry -that I need -to modify? -And if -not, -then life -stops, -right? -Everything -grinds to -a halt. -Now, -can the -heat -exhaust -informationally -be useful -for another -agent somehow? -I mean, -can't we -recover heat -in a -steam engine -and use it -to do -useful work? -not to -the point -of a -perpetual -motion -machine, -but how -can we -recover -exhaust, -and then -what does -the -perpetual -motion -machine -look like? -Well, -the environment -is the -system sitting -over on the -other side -of the -boundary -that is -absorbing -whatever -waste heat -is generated -by the -system, -let's -call it -Alice, -that we're -interested in. -So that -environment -may be, -and Alice -can't tell -this, -but it -may be -divided up -into lots -of different -compartments, -some of -which are -basically -sucking in -Alice's -waste heat -and doing -something with -it. -So this -is really -a question -about how -is the -environment -structured, -and what -operations -does the -environment -execute -with the -input that -it gets? -How does -it interpret -that input? -it? -And what -would -an actual -motion -be? -This gets -back to -this question -about how -do we talk -about multiple -agents that are -communicating with -each other? -Right? -Instead of -talking about -an environment -that we don't -characterize at -all, how -do we talk -about an -environment that -includes some -specified or -preferred systems -that are -treated as or -known to be -other agents? -So let me -get back to -your question -about a -perpetual -motion machine. -Notice that -in classical -thermodynamics, -there's always -the assumption -that there's a -big environment -out there, -you know, -the rest of -the universe -or whatever, -the whole -environment that's -not on Earth, -that is an -arbitrarily large -source of, -or it's not -an arbitrarily -large source of -free energy, -but it's a -large source of -free energy, -and it's a -large sink -for entropy, -a kind of -arbitrarily large -sink for -entropy. -So the -second law -of thermodynamics -says the -entropy of -this entire -system, -counting the -whole surrounding -environment, -doesn't decrease. -So it can -increase and -still satisfy -the second law. -So this gives -you the -classical, -you know, -19th century -heat death of -the universe -idea. -energy, -and so -your proposed -machine is -sitting -somewhere -embedded in -this gigantic -environment that -can provide -some finite -amount of free -energy, but can -exhaust effectively -an arbitrary -amount of -entropy. -entropy. -So any -machine has -to dissipate -entropy as -it runs to -fill up this -kind of -arbitrary -sink of -entropy that's -the classical -rest of the -universe. -Now compare -that to the -picture in -quantum theory. -The picture in -quantum theory is -the universe by -definition is an -isolated system. -So its -total -information -content doesn't -change. -Right? -It's evolving -unitarily. -Unitary -operations preserve -information. -So they're just -moving information -around. -So from -any, -from the -point of view -of any local -system, -the local -system's getting -information from -its environment, -so it sees its -environment losing -information, so -gaining entropy. -What this tells -us is that -entropy is in -fact not a -globally -definable -quantity. -Entropy is -always defined -relative to some -observer, some -division, some -boundary. -And this is -another -kind of big -change in the -last few -decades and -has led to, -for example, -entropic -definitions of -time. -My local -external time -arrow is -whatever direction -I see entropy -increasing the -most. -so again, I -would refer to -people like -Rivelli and -Max Tegmark -who developed -many of these -ideas. -Awesome. -All right. -Thank you, -Chris and -Ander. -Looking forward -to seeing -everyone's -questions -submitted and -their preparation -and participation -and measurement -in the -discussion -section. -So, thank -you. -Until next. -All right. -Thanks again, -Daniel. -Thank you, -Ander. -We'll see you -later. -Bye. -Bye. - - - - - - - - - - - - - - - - - - - - - - - - - - - - -Thank you. \ No newline at end of file diff --git a/data/video/activeinferenceinstitute/Other/WkWIqpxWRM4/README.md b/data/video/activeinferenceinstitute/Other/WkWIqpxWRM4/README.md deleted file mode 100644 index 80b8cec97..000000000 --- a/data/video/activeinferenceinstitute/Other/WkWIqpxWRM4/README.md +++ /dev/null @@ -1,5 +0,0 @@ -# WkWIqpxWRM4 - -Series: **Other** - -- ["Physics as Information Processing" ~ Chris Fields ~ Lecture 2](https://www.youtube.com/watch?v=WkWIqpxWRM4) — `WkWIqpxWRM4` diff --git a/data/video/activeinferenceinstitute/Other/WkWIqpxWRM4/metadata.json b/data/video/activeinferenceinstitute/Other/WkWIqpxWRM4/metadata.json deleted file mode 100644 index bbfc90a14..000000000 --- a/data/video/activeinferenceinstitute/Other/WkWIqpxWRM4/metadata.json +++ /dev/null @@ -1,137 +0,0 @@ -{ - "series": "Other", - "item": "WkWIqpxWRM4", - "source": "youtube", - "channel": "ActiveInferenceInstitute", - "category": "Courses/PhysicsAsInformationProcessing_ChrisFields", - "episode": null, - "title": "Chris Fields ~ Physics As Information Processing ~ 2023 Lecture 2 ~ Lecture 2 — Course information", - "date": "2023-06-15", - "guests": [ - "Chris Fields" - ], - "other_participants": [ - "Daniel Friedman" - ], - "github": "https://github.com/ActiveInferenceInstitute/ActiveInferenceJournal/tree/main/data/video/activeinferenceinstitute/Other/WkWIqpxWRM4", - "slides_url": "https://docs.google.com/presentation/d/1LL5t6OgkeFTXWdmHoAmUphVc21xEdb758UruGRR_eLw/edit?usp=sharing", - "enriched_from": [ - "coda", - "generated", - "db", - "youtube" - ], - "parts": [ - { - "video_id": "WkWIqpxWRM4", - "url": "https://www.youtube.com/watch?v=WkWIqpxWRM4", - "title": "\"Physics as Information Processing\" ~ Chris Fields ~ Lecture 2", - "duration": 3413.0, - "upload_date": "" - } - ], - "sessions": [ - { - "index": 1, - "session_name": "WkWIqpxWRM4_sess01", - "start": "0:00:00", - "title": "Introduction" - }, - { - "index": 2, - "session_name": "WkWIqpxWRM4_sess02", - "start": "0:02:54", - "title": "Quantum Theory" - }, - { - "index": 3, - "session_name": "WkWIqpxWRM4_sess03", - "start": "0:04:09", - "title": "Quantum Mechanics" - }, - { - "index": 4, - "session_name": "WkWIqpxWRM4_sess04", - "start": "0:05:19", - "title": "Simple Questions" - }, - { - "index": 5, - "session_name": "WkWIqpxWRM4_sess05", - "start": "0:07:44", - "title": "Action" - }, - { - "index": 6, - "session_name": "WkWIqpxWRM4_sess06", - "start": "0:11:07", - "title": "Example" - }, - { - "index": 7, - "session_name": "WkWIqpxWRM4_sess07", - "start": "0:14:49", - "title": "What to be asked" - }, - { - "index": 8, - "session_name": "WkWIqpxWRM4_sess08", - "start": "0:17:36", - "title": "Shared language" - }, - { - "index": 9, - "session_name": "WkWIqpxWRM4_sess09", - "start": "0:20:32", - "title": "Hamiltonian operator" - }, - { - "index": 10, - "session_name": "WkWIqpxWRM4_sess10", - "start": "0:25:04", - "title": "Operator M" - }, - { - "index": 11, - "session_name": "WkWIqpxWRM4_sess11", - "start": "0:34:16", - "title": "Alice and Bob" - }, - { - "index": 12, - "session_name": "WkWIqpxWRM4_sess12", - "start": "0:38:26", - "title": "holographic principle" - }, - { - "index": 13, - "session_name": "WkWIqpxWRM4_sess13", - "start": "0:42:01", - "title": "unitary evolution" - }, - { - "index": 14, - "session_name": "WkWIqpxWRM4_sess14", - "start": "0:46:39", - "title": "summary" - }, - { - "index": 15, - "session_name": "WkWIqpxWRM4_sess15", - "start": "0:48:35", - "title": "free choice of reference frames" - }, - { - "index": 16, - "session_name": "WkWIqpxWRM4_sess16", - "start": "0:49:43", - "title": "Superpositions" - }, - { - "index": 17, - "session_name": "WkWIqpxWRM4_sess17", - "start": "0:55:24", - "title": "Conclusion" - } - ] -} diff --git a/data/video/activeinferenceinstitute/Other/WkWIqpxWRM4/transcript.txt b/data/video/activeinferenceinstitute/Other/WkWIqpxWRM4/transcript.txt deleted file mode 100644 index e6b32b5c1..000000000 --- a/data/video/activeinferenceinstitute/Other/WkWIqpxWRM4/transcript.txt +++ /dev/null @@ -1,453 +0,0 @@ -## WkWIqpxWRM4 - -Hello everyone, it's June 15th, 2023. We are here in lecture two of Chris Field's course, -Physics as Information Processing. So thank you, Chris. Looking forward to the lecture. -Thank you, Daniel. And yes, welcome to this session. This session is titled Why Quantum Physics? -And if you will recall from the first session, we reviewed the history of physics and some of the -history of mathematics and computer science from the end of the 19th century through to -the beginning of the 21st century and discussed the slow development from classical thermodynamics -of quantum information theory and specifically characterizing quantum information theory as a new -kind of physics that describes systems that are exchanging finite amounts of discreetly encoded -information across some intervening boundary. And so we use this very conventional graphical notation of -two agents or physical systems, Alice and Bob, A and B, that are exchanging energy and information -across some boundary. And I always use a blue ellipse like this to indicate the boundary that separates Alice -from Bob. And this new approach to physics is entirely topological. It's not geometric. So it doesn't assume a -particular background space-time. So it doesn't assume that Alice and Bob are spatially separated. -And so this makes it a very different kind of theory from classical information theory in that the channel -that separates Alice from Bob, this boundary, is a boundary in state space. It's not a boundary in some -geometric embedding space. And in particular, it's not a boundary or a channel that's embedded in a three-dimensional space -that separates Alice from Bob. -So that's what we talked about last time. And today, what I want to discuss is how quantum theory in particular -makes this idea of physics as a theory of communication simple and obvious. And quantum theory, of course, has a -terrible reputation of being abstruse and mathematically incredibly complicated and counterintuitive and -difficult to understand. And this is one quotation among many from leading physicists pointing out that -quantum mechanics is just difficult. And as you probably know, there's an entire philosophical industry of -interpretations of quantum theory that try to make sense of its ontology. And so what I don't want to do today -is try to introduce quantum mechanics. We're not doing any of this. Some of you will recognize this as the table of contents, -the first part of the table of contents of the famous textbook by Landau and Lifshitz. But if you've studied quantum theory -in undergraduate or graduate school, you've probably dealt with a textbook structure much like this one. -And you've probably been introduced to quantum mechanics as it was traditionally conceived as -essentially a mechanical theory, a theory of motion with wave functions and particle representations and on and on and on. -So we're not going to try to do this. What we're going to do instead is take a completely -information-theoretic approach and we're going to characterize information transfer in a quantum-theoretic way -without any assumptions about mechanics or space-time or any of that. -So instead of this, we're going to ask a simple question. By a simple question, I mean a question with -just a yes-no answer or a binary answer. So let's use up or down as our example question. -And to ask this question, we need three things. First, we need an action. We need a way of asking it. -For example, making a sound or writing something down. -Second, we need a thing that we can ask the question of. We need an environment or a friend or the rest of -the world or an experimental apparatus or some system or other that we're going to act on to ask this question. -And the third thing that we need to ask this question is a shared language. -And it's the shared language that's often neglected in models based in physics. -And I think the importance of the shared language became clear in classical information theory. -And it's become clear again, of course, in quantum information theory. But this has only happened over the last few decades. -So what I want to do today is to construct in a step-by-step fashion a formal representation of these three things that we need to ask a simple question. -And what we'll find is that that formal representation is, in fact, quantum theory. -So it's quantum theory that provides us with a formal representation of asking these simple binary questions. -And of course, if you can ask a binary questions, you can ask any finite number of binary questions. -So you can construct an arbitrarily complex finite decision tree that ends up formulating -an arbitrarily complex but still finite question. -So this really is a general model of communicative interactions between agents. -So let's go about this and talk first about the action of asking a question. -And so the first thing I want you to do is actually say out loud, up or down. -Actually ask a question. -And then note two things about what just happened. -The first thing to note is that it took some time to do that. -That action actually required a finite amount of time. -A couple of seconds, not very long, but it's not zero. -It always takes time to ask a question. -And the second thing to note is that it took energy to actually say up or down. -It requires a little bit of exertion to ask a question. -So let's put those two together and define a word. -We're going to define the word action to mean some energy that's spent in some time. -So asking a question is an action. -And it qualifies as an action because it requires spending some energy in some finite amount of time. -And so we can write this as action equals energy times time. -Or I'll sometimes use the abbreviation E times T. -And action is always required when there's some interaction or two-way action between systems. -So whenever we talk about physical interaction, we're always talking about two systems doing something to each other that requires both energy and time. -So let's define another concept, a concept of a minimum action needed to ask a question, and specifically a question with a binary answer. -And we'll just pick a symbol to stand for that minimum action. -And the symbol we'll use because it's conventional is H. -And it turns out that this H is Planck's constant. -The number that Planck came up with when he was trying to solve the black body problem back in 1900 that we talked about last time. -And recall, that was the problem of characterizing the spectrum of radiation emitted by some physical system that was hot. -How much energy comes out as a function of temperature. -And if you assume that the spectrum of energy coming out is continuous, you get predictions that are wrong. -And Planck got predictions that were correct by assuming that that spectrum was discrete. -And that this H represented the minimum amount of energy that could be emitted in a particular time. -So it represented the minimum energy times time or action. -So let's look at an example of that. -And here's an example that may be familiar. -It's the example of your own visual system. -So in the retinal, the cells of your retinas, there are many, many molecules that are rhodopsins of one variety or other that respond to different wavelengths of visible light. -So let's look at that. -And they have some efficiency for capturing light. -And let's just assume that they're optimally efficient, that they're as efficient as they could possibly be. -And that means that they require the minimum amount of energy to change shape, which is what they do when they're impacted by light, -to capture the energy and hence the information that's available in the light. -And the minimum energy to do anything that records one bit of information, so yes or no, up or down, is 0.7. -So log 2 of Boltzmann's constant times the temperature. -And again, we saw this last time. -This is from roughly 1875 and was then built into classical information theory by Shannon and then by Landauer in the mid-20th century. -So what we're doing is assuming that since rhodopsin is optimally efficient, all it requires is this amount of energy to change shape. -And it's known experimentally that rhodopsin takes about 200 femtoseconds to change shape. -So that's 200 times 10 to the minus 15th seconds. -So it's a very, very fast reaction, chemical reaction. -So we can just calculate what the action is. -And at 37 degrees centigrade, so human body temperature, kT is about 4.3 times 10 to the minus 21 joule. -And joule is just the standard unit of energy. -So we can multiply everything together. - -KT times 0.7 times this response time of 200 femtoseconds. -And if you're familiar with quantum mechanics, you'll instantly recognize that number because it's very close to the value of Planck's constant, which is 6.6 times 10 to the minus 34 joule seconds. -So this says that rhodopsin is very near to optimal. -The best it would be able to do would be Planck's constant. -That's the minimum action that can be undertaken. -So rhodopsin is clearly doing very well. -It can't be as efficient as we're assuming it to be, but it's pretty efficient. -So this number Planck's constant is, and this idea of a minimum action is not something we just pulled out of the air. -It's something that one can observe when you're looking at what real biological systems in particular do and other physical systems do. -So let's think about the second component of this project of asking a question, and that's the thing to be asked. -And we want to ask a very simple question. -We want to have a theory of asking very simple questions. -So let's define the simplest thing that we can ask a binary question to. -And let's say this is the thing that we can ask a binary question of using just this minimum action h. -So we're going to ask for the minimal thing that it's possible physically to ask a question of. -And this thing needs a name, so let's call it a qubit. -And that word is a contraction of quantum bit. -And we can draw it like this, a little sphere with an arrow through it. -That's called a block sphere. -And we use a symbol q, and that notation with the bar and the angle bracket is called a bra notation, and it originates with Dirac. -So it's a traditional way of writing inner products, actually, in quantum theory, -to indicate the state of q. -So this thing called a qubit only responds to binary questions. -So it clearly can have only two states, the state that answers yes and the state that answers no. -And conventionally, we draw the arrow going up if the answer is yes and the arrow going down if the answer is no. -And so those are the two states that the qubit can be in when it's answered a question. -So now let's draw a picture for the act of asking a question of q in the state q. -And let's draw it like this. -This curvy arrow is the action of asking the question and getting the answer back. -And we're asking it of this minimum system, the qubit. -And let's call this entire act of h. -And write h acting on q by this expression hq. -So again, we're developing the notation of quantum theory, which -all of you who know the theory already will find familiar. -So the third thing that we have to build into this picture and into the theory -is the shared language. -So we have to know what we mean by up or down if we're going to ask the question up or down. -And we have to have some representation of what it means for the qubit to distinguish up or down. -So let's say that to us, up and down mean this little diagram with an arrow that's labeled up on one end and down on the other end. -So we're defining up and down in a very simple way as just opposites. -We don't know anything else about them except that they're opposites. -So this is a stripped down one bit question. -We could have called them one and minus one. -We could have called them a and b. -We could have called them anything. -We're just indicating that up and down are two opposing concepts. -And so this object that we've used to make this definition of the relationship between our two words, up and down, is called a reference frame. -It's a physical object that encodes a concept or a meaning. -And if you think about using language, using English, for example, you can understand me speaking in English because your brain has the right sorts of neural circuits to allow you to understand English and produce English using your hand and your mouth. -And those neural circuits are reference frames in your brain for understanding and producing English. -And they encode the relationships between different words. -So you all know about chat GPT. -You can think of all of the weights, all of the 10 to the 19th or however many parameters there are in GPT-4 as a reference frame that describes the relationships between words and phrases in English and many other languages. -So this idea of a reference frames. -So this idea of a reference frame is an instance of Landauer's principle, which is often summarized as saying that information is physical. -What that means is that information always has to have some encoding. -And that encoding, that object, can always be thought of as a reference frame. -So let's put all this stuff together. -What we have now is a reference frame that defines the difference between up and down. -We have an action that we've labeled H. -And the action acts on this object, the qubit, which is in some state. -And what we've drawn here now is an experiment. -And you can think of the incoming part of the arrow that goes from the reference frame to the qubit as a preparation of the qubit. -So you can think of it as I'm doing something to the qubit to put it in a particular state, like up. -Because that's the state I want it to be in. -And I can think of this returning part of the arrow as a measurement of the state that the qubit is actually in. -So if the qubit is actually in this state with the arrow pointing up, then I'm going to get back the answer up when I act on the qubit to measure its state. -So now we can recognize this action H as the Hamiltonian operator in quantum theory, which always takes some time, which we represent by an interval, dt, to act on some state, and in particular this binary state q. -So this whole picture depicts the action of a very simple Hamiltonian operator. -And again, I want to emphasize that we've just defined the qubit as the simplest thing that we can ask a binary question of. -So how does this H work? -We can use H to prepare this qubit in one of two states, which we're going to label 1 and minus 1. -The state 1 corresponds to the up direction, and the state minus 1 corresponds to the down direction. -And doing this to q, so preparing q in one of these two states takes energy. -And we know how much. -Boltzmann told us back in 1875 that the minimum energy to reduce the entropy of the universe by one bit is log 2 of kt. -We saw that earlier when we were thinking about rhodopsin. -So we can say that acting with H always requires some number times kt, where that number is bigger than the log of 2. -Remember, the log of 2 is just in there to get units of bits of binary questions, as opposed to units of nats, which are questions that have e answers, where e is the exponential base, the natural exponential base. -But we want to use units of bits, so we have to always have this factor of log 2. -So this lets us write H in a particular way. -We can take out this energetic term that Boltzmann tells us we have to include, beta times kt. -And what's left over is, in a sense, the pure representation of the action stripping out the energy requirement. -And we're going to call that m, because m is the standard symbol to use in quantum theory for a measurement operator. -And m is now dimensionless. -H has units of energy. -Kt is an energy. -And m has no units at all. -So it's H without the energy. -So now we can ask, what is this operator m that we've defined as the simplest possible action on a qubit, but abstracting away the amount of energy that we have to use to act on the qubit? -Well, what m does is just encode a bit or decode a bit. -So we can use this operator m to say, I want q to have the value 1. -I want it to be up. -So I can sort of reach out with my operator m and grab q and turn it so that its arrow is pointing up. -And that's the operation that's called preparation in physics. -It's what it means to prepare an experimental apparatus, for example. -You actually grab hold of the apparatus and you twist some knobs and flick some switches to prepare it to be in some state. -And you can use m to measure a state that isn't known. -So I can use m to grab a hold of q and look at it and determine whether it's in state 1 or minus 1. -And I can represent these as just a cycle that starts with measuring and then prepares, or that starts with preparing and then measuring. -And if I represent it that way, what I see is m having a value, which is either the value I want to encode or the value that I have measured, and q. -And m acts on that value and q to produce a particular state of q, to put q in a particular state or to realize that q is in that state. -And so m is actually an identity acting on some value and q. -So it maps 1 and q to 1 and q, and it maps minus 1 and q to minus 1 and q, if you do the whole cycle of preparation and then measurement. -And that makes sense. -Notice it's actually an axiom in some formulations of quantum theory that if I measure the state of some object, what I'm doing is also preparing it in that state. -So if I immediately measure it again, I'll see the same value. -And that makes sense physically. -If I put my coffee cup on the table and then I look for it, it's likely going to be on the table unless something else intervenes. -Well, what does this mean? -It means that we can do a little bit of algebra. -It means that we can write this original expression of acting on q with our Hamiltonian in this somewhat more explicit form of using the energy beta kt and our operator m on q. -And what I get out of that is the energy beta kt multiplied by q. -So hq equals eq. -That's the Schrodinger equation. -So that's one of the central equations of quantum theory. -What it says is that this operator h is the total energy operator on a system. -Its eigenvalues are the total energies of the system. -And we can now see where that energy is coming from in this very, very simple stripped down kind of representation of an action. -The energy is the thermodynamic energy that has to be spent to determine one binary outcome, up or down. -So, now, if you're familiar with quantum mechanics, you will have seen exponentiated operators like the next line here of algebra. -This operator p, which is time dependent, which is the exponential of h multiplied by it over h-bar, is called the propagator. -It's the unitary operator in quantum theory, the operator that conserves probability or conserves information. -And so, using the representation of h as beta kt, we can see that this operator p is just e to the minus beta m. -If we recall the wick rotation from the first session, this strange equivalence of it over h-bar with 1 over kt that wasn't discovered until 50 years after the invention of quantum theory. -And it's still a very debated thing. -I mean, people worry about what this equation really means. -But one thing that it means in this setting, where h has this particularly simple form, is that the propagator is now just the exponential of a number multiplied by this very simple stripped down operator that encodes a bit or decodes a bit. -But we know something about the exponential of a bit. -But we know something about the exponential propagator just by looking at its form. -If you're, again, familiar with these sorts of exponential operators, you know that they're wave equations. -And we can make sense of that using the diagram up above. -So, what m is doing is just performing a cycle. -It's just mapping 1 and q back to itself or minus 1 and q back to itself. -So, we can think of m as just taking this object and flipping it back around until it gets itself again. -And that kind of cycle can always be turned into a wave equation. -And e to the i over h-bar times ht is a canonical form of a wave equation where the horizontal axis here on the wave is time. -And if you look at the time of the time of the thing spinning around like this, which is exactly the same thing as a wave traveling through time. -So, that tells us something very important. -It tells us that this equivalent expression e to the minus beta m is also a wave equation. -So, it tells us that this operator m has built into it some sense of time. -It's the sense of time of this prepare-measure cycle. -And that's implicit in m, right? -We've used the wick rotation to get rid of the kind of what I'll end up calling the external time t, the objective time t. -And we'll see in July where this internal time comes from in terms of the wick rotation. -But we can say now that it's an agent-specific time. -It's an internal time reference frame that's built into our idea of this system, Alice, that can act on the world. -And it represents the time that's required to ask this question. -Okay. -Chris, can I ask a question at this point? -Yes. -So, on this slide with the operator m, I'm curious, how can the agent deploying the operator m tell the difference or break the symmetry between preparation and measurement? -Or are they just completely dual pictures? -Yeah, they're completely dual. -So, the agent never breaks the symmetry. -The agent can't actually distinguish the act of preparing from the act of measuring. -Every preparation is a measurement and every measurement is a preparation. -Right. -That's in fact what this axiom of… -Unitarity? -…measuring a state leaves the particle in the measured state means. -Right. -Right. -So, yeah, mathematically they're duals. -Mm-hmm. -Thank you. -Okay. -So, let's now use this theory that we've developed to describe the interaction between Alice and Bob. -And let's think of a situation in which Alice prepares a qubit that Bob then measures. -And then Bob prepares the very same qubit. -And then Alice measures it. -So, Alice acts with her Hamiltonian operator, which we'll label HA. -And Bob acts with his Hamiltonian operator that we label HB. -And let's say that Alice prepares the qubit as end state up. -And Bob can then measure the qubit end state up and say, oh, the qubit is now up. -And if Bob then prepares the qubit end state down, Alice can measure the qubit end state down and say, oh, the qubit switched from the up that I prepared to down. -So, Bob must have prepared it as down. -So, what's happened here? -Alice and Bob have actually shared a bit flip. -Right. -Alice said up and Bob said down. -So, and Alice heard him say down. -So, they've shared this idea of change by interacting with each other. -Now, this works as long as they mean the same thing by up and down. -So, we've implicitly assumed in talking about them communicating that they both have this reference frame that distinguishes up from down. -And this is actually a completely general model of what I'll call a noiseless one qubit channel. -Now, what does that mean? -It's noiseless because there's nothing else in the picture. -Right. -There's nothing in the picture that can jiggle the qubit when Alice and Bob aren't looking. -Alice and Bob are the only systems we're talking about. -Alice and Bob are the only systems that are in the picture. -So, there's no kind of third party or environment or heat bath or anything like that that's going to mess with the qubit and disrupt their interaction. -So, that's a noiseless channel. -And it's a one qubit channel because they just have one qubit that they're interacting with. -And they're alternately preparing and measuring its state. -Now, clearly, if they can interact using one qubit, they can interact using any finite number of qubits. -So, here's a picture of an in qubit channel that Alice and Bob, which here I've labeled A and B because they took this figure from a paper, are sharing. -And Alice and Bob both have n-qubit operators, one for each of the n qubits that they share. -And they just do this prepare and measurement cycle on all n qubits. -So, you can think of Alice preparing the n qubits, Bob measuring the n qubits, and then Bob preparing and Alice measuring. -And that allows them to exchange n-qubit messages. -And we can make n as large as we want. -And so, they can exchange messages of arbitrary complexity. -And if they share the meaning of up and down, then they can exchange these messages with no noise. -So, they can communicate. -And we now have our theory of two communicating agents that are sharing arbitrary finite information. -Now, I've labeled q1 through qn in this picture a holographic screen. -And in fact, that's exactly what it is. -Recall from session one, this slide illustrating the holographic principle. -Shannon made this point that a bit of information is an answer to a yes-no question. -And so, we can write down a number of possible states. -We can write down an entropy. -And the limiting case of that entropy is the entropy of a black hole, the densest possible encoding of information in space or on a boundary. -And this holographic principle allows us to see any boundary, as we discussed in session one, as an encoding of bits. -So, the previous boundary of these n-qubits, we can think of as a holographic screen. -Now, why is that important? -It's important because it allows us to use this idea of holography and this idea of a fixed encoding of a message or a fixed encoding of an observable entropy at some density that was developed in cosmology. -So, we can import a whole bunch of theory that was done someplace else into quantum information theory and see that it works perfectly because it describes exactly the same situation. -Now, what does this tell us philosophically? -It tells us that Alice and Bob can learn about each other at most n-bits per communication cycle. -So, the maximal amount of information that they can get from an observation is the n-bits that the other one wrote on the screen. -So, this tells us something very important, that when an agent is looking at their environment, which is this other agent, Bob, the amount of information that can be extracted from the environment is actually proportional to the size in bits of the boundary. -And the holographic principle just makes that size finite in space. -And I said early on that there's no embedding space-time in this theory that Alice and Bob are not thought of as separated in space. -But as soon as we use the holographic principle, we can think of the boundary between them as having a spatial coordinate, even though they don't have a spatial coordinate. -So, we'll get back to this point in September when we talk about theories of emergent space-time. -But all of these theories of emergent space-time come back to this idea that boundaries can have spatial extent, even though the systems that they separate don't have any natural spatial description or natural spatial degrees of freedom. -Okay, so now let's recall another picture from session one. -We talked a little bit about Feynman's theory of scattering and the origin of Feynman diagrams. -And the idea that if Alice prepares some system in some initial state and then measures the state later, if we want to calculate what the new state is, we have to take into account all of the possible paths that join the initial state that Alice prepared to the final state that Alice measured. -So, what does that mean? -So, what does that mean? -I mean, here effectively, if we think of, suppose I'm Alice and the boundary is my screen. -I'm talking to the screen. -And then later on, I'm going to measure the screen. -And between my preparation actions and my measurement actions, a bunch of stuff happens on the other side of the screen. -And if I want to predict what's going to happen on the other side of the screen, I've got to take into account every possible thing that could happen. -And I have to give some weight to those possible things that could happen. -And quantum theory, in fact, tells you how to assign those weights to things that could happen in order to make a measurement given some theory of the mechanism that makes things happen. -But the fact that all of the paths have to be counted is what unitary evolution means. -This evolution that preserves information, that doesn't take any away, it doesn't add any. -It just represents it in different ways. -So, if you look at this screen and imagine that Alice is interacting with her boundary, and then she's interacting with her boundary again, it looks just like the previous screen. -And the wavy red arrows are what's happening inside Bob. -The wavy red arrows are a representation of what's going on inside Bob as he measures his side of the screen and then prepares his side of the screen. -So, what Feynman is representing here is just the action of the environment or the action of the system while Alice isn't looking between Alice's preparation stage and her measurement stage while Bob is doing his measurement and preparation. -So, this diagram is actually completely symmetrical. -We could have drawn an arrow where the wavy red things are and drawn wavy red things where we've talked about Alice and we'd have the same picture just from Bob's point of view. -We can also think of this, of course, as Alice communicating with her future self by writing a memory record someplace in the world and then coming back and reading it. -And if Alice writes a memory on the world and then comes back later on and reads it, she actually has to take into account some theory of all of the things the world might have done to her memory while she wasn't looking. -So, this picture of Feynman's actually tells us a lot about memory. -And when we think about organisms running around in the world doing things like we do all the time, we tend to think of memories as things that are persistent and unchanging and all of that. -But we know that that's not true. -We know that memories degrade in various ways. -We know that adversarial third parties can come and change what we've done so that they fool us into thinking that we've remembered something that we haven't and so on and so forth. -So, that whole picture of memory is actually encoded in this picture that Feynman drew in the late 50s to represent scattering of particles. -So, what we've done here is started with this simple idea of asking questions. -And we've actually constructed a fair amount of quantum theory. -We constructed the Schrodinger equation. -We constructed the unitary propagator. -We constructed the idea of a reference frame, which we'll talk about intensely next time in July. -So, we can sum this up by saying that quantum theory simply is a theory of communicating agents. -And that quantum mechanics of the theory that Feynman was complaining about in that earlier slide, -that quantum mechanics generates this problem that seems intractable of how to understand measurement, -because the theory is in fact about measurement. -It's not about mechanical motion at all. -So, quantum theory actually tells us nothing about ontology. -And this is why the interpretations of quantum theory, the many worlds interpretation, -where you have all these branching universes and so forth, are not really science. -They're philosophy. -They're attempts to understand an ontology that's generated by a theory that doesn't talk about ontology. -It talks about communication between agents. -But we've left something very important out. -We've left out a consideration of what happens when Alice and Bob don't mean the same thing by up and down. -And we've left out the famous phenomenon of entanglement. -Spooky action at a distance, as Einstein called it. -And it turns out these things that we've left out are very, very closely related. -And we're going to see why in July, but I'm going to give you a brief preview now. -So what happens when Alice and Bob actually mean different things by up and down? -What if their reference frames aren't aligned? -Well, the first thing to note is that nothing in the theory says they have to be aligned. -So we have to consider the possibility that they're not. -And in fact, both Alice and Bob have to have what's called free choice of reference frame. -When Alice prepares her instrument, she has to be able to prepare it any way she wants to in order to do an experiment. -And when Bob makes up a sentence to tell Alice, he has to be able to choose any language he wants to use. -He's a free agent in this way. -And if they don't have that freedom, it turns out they're entangled. -So reference frame non-alignment and entanglement are very, very closely coupled. -Now, misalignment of reference frames is what produces superpositions of outcomes. -So if Alice makes her measurements with her up and down arrow arranged vertically, for example, -Alice is not going to have any idea what Bob did. -I mean, she can't see distinctions in the left, right direction. -She can only see distinctions in the up and down direction. -So all she can do is flip a coin. -If she wants to know what Bob prepared, it's 50% likely he prepared up for her, what she calls one. -And it's 50% likely that he prepared what's down for her. -She can't tell. -Now, if Bob's reference frame is pointing at about 45 degrees, -so up for him means 45 degrees off of Alice's up, -then she can sort of tell what he did. -If he's pointing 45 degrees from up, then she's probably going to measure up. -And if he's pointing 45 degrees from down, she's probably going to measure down. -So her outcomes are 75% for 1 and 25% for minus 1 if he prepares at this 45 degree angle. -So we can write this in a particular way. -We can say that the value that Alice gets acting with her M on a cubit that Bob prepared is some number times what she measures is up, -and some other number plus times what she measures is minus 1, -where those numbers, the squares of those numbers add to 1. -So why do they have to add to 1? -Well, they add to 1 because of the Pythagorean theorem. -So 1 is the distance, the total distance between the two outcomes, -and A squared plus B squared are the other two sides of that triangle. -So another way to think of it is that the metric distance in possibility space is 1, -and A squared and B squared are just the Euclidean distance components. -Again, Euclidean distance is based on the Pythagorean theorem. -So this is the Born rule in quantum theory. -It's the rule that tells us what happens when we're trying to measure a superposition, -and it tells us where these superpositions come from. -We have a mismatch in reference frames between two agents that are communicating. -So if we go back to the free energy principle, which is what this whole active inference class is about, -and we formulate it in quantum theory, we get a very simple idea of what the free energy principle is. -The FEP in quantum theory just says that interacting systems are going to behave in a way that aligns their reference frames, -and that makes their communication as good as it can be. -But since aligning reference frames perfectly leads to entanglement, -the free energy principle drives systems asymptotically toward entanglement. -Well, the principle of unitarity says that all systems will approach entanglement asymptotically. -So we can see the FEP is actually a classical statement of the principle of unitarity. -And what we'll see in the next session is why this is the case, -why active inference, when we think of it quantum mechanically, -is a way of talking about the principle of unitarity, which is the deepest axiom of quantum theory. -And we'll also see that this question of alignment becomes more and more complex as the complexity of the messages that the systems are trying to communicate to each other increases. -And that message complexity correlates with system complexity. -So that aligning complicated systems is very hard, although aligning very simple systems is fairly easy. -And if you think about what's the simplest possible theory in science, it's particle physics, because you're dealing with these very simple things that interact in extremely simple ways. -And the really difficult parts of science are things like sociology, where you're dealing with incredibly complicated systems like humans that interact in incredibly complicated ways. -And all we have are very vague kind of folk theories of how that works. -But we know from that example that aligning complex systems is extremely difficult. -That's why humans don't get along very well. -So the second discussion section, which Andrew will be leading, is on Saturday, the 1st of July, a couple of weeks from now. -The third lecture session will be on July 13th. -And I hope you all will contribute to the interactive question and answer session on the web and show up for the discussion on Saturday, the 1st of July. -And join us again in July for the discussion of quantum reference frames and reference frame alignment. -So thank you very much. -Thank you, Chris. -All right. -Thank you all. -Again, please check out the Chorus syllabus. -Ask questions through the site and see you at the discussion. -See you next time. -Thank you. -Bye. -Bye. \ No newline at end of file diff --git a/data/video/activeinferenceinstitute/Other/hSj7eOYuZe4/README.md b/data/video/activeinferenceinstitute/Other/hSj7eOYuZe4/README.md deleted file mode 100644 index 203891f23..000000000 --- a/data/video/activeinferenceinstitute/Other/hSj7eOYuZe4/README.md +++ /dev/null @@ -1,5 +0,0 @@ -# hSj7eOYuZe4 - -Series: **Other** - -- ["Physics as Information Processing" ~ Ander Aguirre ~ Discussion 2](https://www.youtube.com/watch?v=hSj7eOYuZe4) — `hSj7eOYuZe4` diff --git a/data/video/activeinferenceinstitute/Other/hSj7eOYuZe4/metadata.json b/data/video/activeinferenceinstitute/Other/hSj7eOYuZe4/metadata.json deleted file mode 100644 index df3e0ffe3..000000000 --- a/data/video/activeinferenceinstitute/Other/hSj7eOYuZe4/metadata.json +++ /dev/null @@ -1,124 +0,0 @@ -{ - "series": "Other", - "item": "hSj7eOYuZe4", - "source": "youtube", - "channel": "ActiveInferenceInstitute", - "category": "Courses/PhysicsAsInformationProcessing_ChrisFields", - "episode": null, - "title": "Chris Fields ~ Physics As Information Processing ~ 2023 Discussion 2 ~ Discussion 2 — Course information", - "date": "2023-07-01", - "guests": [ - "Ander Aguirre", - "Chris Fields" - ], - "other_participants": [ - "Daniel Friedman" - ], - "github": "https://github.com/ActiveInferenceInstitute/ActiveInferenceJournal/tree/main/data/video/activeinferenceinstitute/Other/hSj7eOYuZe4", - "enriched_from": [ - "coda", - "generated", - "youtube" - ], - "parts": [ - { - "video_id": "hSj7eOYuZe4", - "url": "https://www.youtube.com/watch?v=hSj7eOYuZe4", - "title": "\"Physics as Information Processing\" ~ Ander Aguirre ~ Discussion 2", - "duration": 4393.0, - "upload_date": "" - } - ], - "sessions": [ - { - "index": 1, - "session_name": "hSj7eOYuZe4_sess01", - "start": "0:00:00", - "title": "Introduction" - }, - { - "index": 2, - "session_name": "hSj7eOYuZe4_sess02", - "start": "0:03:00", - "title": "Physical Degrees of Freedom" - }, - { - "index": 3, - "session_name": "hSj7eOYuZe4_sess03", - "start": "0:06:00", - "title": "Internal and External Time" - }, - { - "index": 4, - "session_name": "hSj7eOYuZe4_sess04", - "start": "0:09:00", - "title": "Mechanical and Theoretical" - }, - { - "index": 5, - "session_name": "hSj7eOYuZe4_sess05", - "start": "0:15:25", - "title": "Reference Frame" - }, - { - "index": 6, - "session_name": "hSj7eOYuZe4_sess06", - "start": "0:16:52", - "title": "Computer Science" - }, - { - "index": 7, - "session_name": "hSj7eOYuZe4_sess07", - "start": "0:18:27", - "title": "Painting" - }, - { - "index": 8, - "session_name": "hSj7eOYuZe4_sess08", - "start": "0:19:41", - "title": "Preparation Measurement" - }, - { - "index": 9, - "session_name": "hSj7eOYuZe4_sess09", - "start": "0:24:28", - "title": "Carpet is Binary" - }, - { - "index": 10, - "session_name": "hSj7eOYuZe4_sess10", - "start": "0:26:13", - "title": "Carpet is a Wave" - }, - { - "index": 11, - "session_name": "hSj7eOYuZe4_sess11", - "start": "0:30:56", - "title": "Asking a Question" - }, - { - "index": 12, - "session_name": "hSj7eOYuZe4_sess12", - "start": "0:32:22", - "title": "Binary Questions" - }, - { - "index": 13, - "session_name": "hSj7eOYuZe4_sess13", - "start": "0:40:38", - "title": "OnOff Comparison" - }, - { - "index": 14, - "session_name": "hSj7eOYuZe4_sess14", - "start": "0:43:27", - "title": "Quantum Mechanics" - }, - { - "index": 15, - "session_name": "hSj7eOYuZe4_sess15", - "start": "0:46:13", - "title": "Questions" - } - ] -} diff --git a/data/video/activeinferenceinstitute/Other/hSj7eOYuZe4/transcript.txt b/data/video/activeinferenceinstitute/Other/hSj7eOYuZe4/transcript.txt deleted file mode 100644 index eee602292..000000000 --- a/data/video/activeinferenceinstitute/Other/hSj7eOYuZe4/transcript.txt +++ /dev/null @@ -1,639 +0,0 @@ -## hSj7eOYuZe4 - -Hello, it's July 1st, 2023. We're in Physics as Information Processing in the second discussion. -So thanks all for joining. We'll see who else comes in. If you're watching live, feel free to -write questions in the live chat and I'll relay them onwards. So Andre, thanks. And it would be -awesome to begin with a little bit of some overview and some points that you wanted to get to. -Yeah, thank you, Daniel. So good morning, everyone. So if you would start by bringing up a bit -cold with Chris's spoke of last time. I have a few slides that I singled out. So now we're getting to -reformulation of quantum theory in terms of reformulation of -opposed to dynamical theory. And as we've talked in previous sessions, we need to -think a little bit philosophically what the ontology of measurement entails. -So Chris, last lecture started with these three basic ingredients. -So an action, that's number one, a thing to ask about. This is the qubit, the physical degrees -of freedom that we need to perturb in some way to ask a question. And then finally, a semantics, -a language that makes sense, a context that makes sense of that information that we gather. -So I should say, number one, I came up in the first session when, I can't exactly remember, -but at some point we talked about what is information. And we came up with, or we brought up -a definition that goes something like this. It's a nice adage. It's differences that make a difference. -This is, Chris told me, I can't remember who first coined this adage, but I think it's quite old, -at least early, early 70s. Anyway, so number one, basically, let's curve that part, right? -If I want to gather some information, I need to perturb a system somehow. So I need to act. -And this is literally in units of actions, right? So units of energy times time. And he provided the example -of rhodopsin, right? Which is apparently near optimal from a quantum perspective. This is mechanically -what goes on in your eyes as light comes in. And this is literally the difference that makes a difference -in your brain. The second part were the physical degrees of freedom. So recall that -we had a boundary, okay? So this theory of quantum mechanics as measurement necessarily breaks down -the system into a bipartite system with a boundary. And you are in a boundary, which are a lot smaller than -than the degrees of freedom of A and B, at least in terms of the dimension of the Hilbert space. -And the other medium change, right? Sorry, can you repeat, Andre? -I repeat what? Sorry? Just last 30 seconds. -Yeah. So the second ingredient is the physical medium in which this information exchange occurs. -So the physical degree of freedom, the qubits are arranged and these are physically perturbed in order to -mediate that difference that makes a difference. So we can think of this duly as either preparing those qubits -or decoding those qubits. And then finally, there was the notion of the reference frame. -Okay. So this theory necessitates a free choice of the agent of a reference frame. And physically, -what this amounts to is a free choice of a measurement axis for those qubits. So we can think of -measuring Z-spin in those qubits on the boundary. So, you know, that requires a free choice of that axis. -And this, you know, amounts to your reference frame, the context in which those measurements are performed. -On one extreme case, we have that if those references are perfectly aligned, -then the exchange of information is noiseless. There was somewhere over here, I think. -I can't quite recall. But generically, these two reference frames of Alice and Bob A and B are not -exactly the same. They get to choose the reference frames. I think it's all the way on the bottom, -actually. Right? So they can be aligned. If they're aligned, they are exchanging noiseless information. -Right? There is no randomness. If they are not -superposition, and that's when you're going to get, you know, probabilities in the outcomes, -trivial probabilities. So Marcia, good? -Yes. -Okay. So I have a few more comments of the main things. Okay. It'll take me a few minutes, -and then we can chat back and forth. Perhaps this slide is -the most interesting, important. I remember this is where I asked the question myself because I was, -I still, you know, it's not easy to wrap your head around what's going on here. -So we said that measurement and preparation are completely dual pictures. -So this is going to be expanded on in the next lecture. But a direct corollary of that is that -if you prepare, Chris said it in similar terms, if you prepare something in a certain state, and then you -measure it right away, you should expect that it's still in that state. Right? So, -intuitively speaking, this induces a cycle, right? And where you have preparation and measurement, -and you're back to square one, so to speak. And in generating that cycle, obviously, there has to -be some notion of time. Right? So that's where we're going to have to make sense of the distinction -between internal and external time. If we take this measurement operator, and here perform some -operations, we can write it as a unitary propagator from quantum mechanics, which again preserves -information. It just leaves the state as it is. But with this little trick of the weak rotation, you can -write it as e to the measurement up to a constant. And that, you know, precisely encodes, at least -morally speaking, what we just said, right, that this preparation measurement thing is a cycle. And more -importantly, it gets rid of this t external time. And presumably, in here, it's going to be replaced by -some notion of internal time, we're going to have an internal time QRF, that is -sort of quote, unquote, the true local time, as the observer proceeds. -I'm not entirely certain of the details. But now looking at it here, the beta, which -presumably there's a, you know, beta and inverse temperature here, I think this is going to be -intimately related to notions of thermal time. This has been explored in papers by Rovelli. -Call, you know, internal time QRFs and thermal time is the way in which the observer observes the -passage of time. Those papers of Rovelli are linked in the Q&A, by the way, so I can also provide links later. -I think there might have been a question. Should I stop? -Yeah, Dean, go for it. -I was curious about this, because what I heard Chris start this session with was that he was going to try to -help us understand when we are looking at something through a mechanical lens. So quantum mechanics. -And to me, this slide was the way of addressing, as he said, his words were characterized information -transfer in a quantum theoretic way, that the that the translation from H to M, took the mechanics aspect -out of the change or timepiece. It was the purification, the most theoretical form of transfer -of information, as opposed to holding up just the one ordered or sequenced or mechanical lens. -Did I I heard you ask the question? And I kind of smiled because I thought to myself, no, I understand -why Anders is asking this question. It's, can something be both gripper in this in the conceptual -meaning sense ontologically, and gripped a physical object. And I think what the I thought, maybe I was -misinterpreting, I'd like your what your perspective on this is, that ability to go from H to M is what -allows us to have both the mechanical and the theoretical, the transference in its purest sense. -Is that being able to have both? Both can be true? Is that what it I think? -I think the answer is, yes, for sure. Yeah. Yeah. Yeah. I -I think much else to say here. It's, it is, it is true that you can think of the information transfer -picture as an enlargement or an addition to the mechanical picture. And that I don't think that -the mechanical picture has to be just, you know, thrown out completely. I think, I think it's interesting -to keep it to see how it can be recovered from the information processing picture. This reminds me, -actually, there is an Italian physicist, I think his name is Cribella. And Dean, if you want, I can, -I can link you the papers later, who is very fond of the information theoretical approach to quantum -mechanics and deducing quantum mechanics from first principles. And he has a paper. I think the title is -physics without physics or something like that, where, you know, he, he treats, um, uh, this whole -thing as a seller, big seller automaton. And by taking the appropriate scaling limits, he derives QFT. -I'm interacting QFT. So, so he says, okay, let me start with some axioms, how information should be -the information processing. I mean, I'm very, very loose here, then get a cellular automaton out of this. -And then take scaling and recover QFT, non interacting QFT, at least. So, you know, start with a few axioms, -a lot of information processing, recovery mechanics from that. Does that add to your question a little bit? -Yeah, so, so all I'm, all I'm trying to do is, is tie this back into some of the live streams that we've -done in the past. And I know I use different language. I, I am not a scientist in the, I've not -studied quantum other than sort of in passing ways. But I do think that what this spoke to me about was -something that I've always tried to include, which is that there's the figure it out piece, which to me is the -conceptual and the meaning part that Chris talked about very clearly here. And then there's the outing of the -figure, there's the sort of physical edge and boundary that if you don't have an awareness of both things happening -concurrently, the preparation and the measure are unitary, have sort of minimum of two state, -then all bets are off in terms of being able to think of these, these propagators and operators and -all of the ways that we're trying to explain the situation, they're off. So we have to have both. -We have to have that sense that there is a transfer, but there are, and there is the mechanics, we have -to keep both juggling all the time. So that's where I, that's where this one particular slide was really -quite both bias confirm confirming for me, but I think it also, as you say, is going to open up a lot -that July. Yeah, I agree. I think this is the most, you know, important or mysterious if you wish a slide. -Yeah. And it turns the pyramid, the conventional approach, here's a pyramid into a top that can be spun -essentially with a rotation and it, and it validates how information transfer probably gets short shrift -in most of this quantum conversation, but it's, it's probably at least an equal to, to the mechanical -part that what, what order are we delivering this information in? So I'm at that point, I'll stop. Thanks. -Yeah. -Thanks. Alexi. -Hi, can you hear me? -Yep. -So, I mean, I, I'm an, you know, my mathematics teacher in high school used to say, this is beautiful, -but I don't quite understand it. So I wanted to, maybe I have two questions, but I'll start with, -with this simpler one. So when Chris talked about the reference frame and said common language, -you know, I think in Shannon common language is the alphabet. There's a receiver and sender and -the alphabet that they share. And then Chris said, this reference frame is a physical thing like -neurons or hands or mouth. And then I got lost because, you know, if, uh, when I encode, I modify -something in the, on the hard drive, then the physical components of the hard drive is the reference, -uh, frame. But, you know, this is not the shared language, you know, so that's my first question. -Maybe I'll stop here to, to really, I understand we're dealing in metaphors, but I want to understand -that, you know, the reference frame up or down, if Bob and Alice understand it similarly, then -is that really physical? -I think the answer is that -Chris would probably give a better answer. -Start from the beginning. -Go. -Is that both are true at the same time. I think Chris would give you a better answer, -but the, the semantics or the, sorry, the shared language or reference frame has to have a physical -embodiment. And, and that's a hardware, a piece of hardware. It's, it's a neuron or something. Um, -in this case, it's the, the, your free choice of the spin axis and that, that's your reference frame. -But at the same time, it's, it's a physical, it's a physical piece of hardware. So reference frames are both -the piece of hardware, but also a piece of semantics. -So if I take it from the language, I don't quite understand to computer science. So if Alice and Bob -communicate through a piece of memory, whether it's random access memory, hard drive, then Alice encodes -something on that hard drive and Bob reads from the hard drive. So in that sense, they understand how -to encode and decode, but they use the same physical piece. That's the reference frame. Would that be correct? -I might, I might just add one again, continuing with the computer science, let's just, and the Shannon -alphabet. Let's just say that these are plain text character encodings. And so they share the same -set of character, a through Z encodings that, that is reflected somewhere physically. -And so if, if a message were, if, if, if a message were sent from another letter, it wouldn't render -properly. It's like outside, they couldn't differentiate Omega from Zeta because they only have the A -through Z software pack. And so there's an informational distinguishments that's facilitated by the -information that's classically inscribed on the screen, the memory. And there is, is an embodied physical -as well as a semantic aspect to that situation. -All right. I think I'm getting closer to understanding. So to move away from computer -science, if Alice is a painter, she took a brush, dipped it in red paint and painted on the canvas red. -And Bob has the same understanding of, of red as Alice. And then he looked at the canvas and read, -read that it is both the physical, you know, a mark on the canvas and, and the symbol of red. Right. -Yes. And if, if, if somebody had red, green color blindness, -that physical apparatus was not able to, to informationally distinguish, then they, they -couldn't make this wick rotation into the quantum cognitive space. They wouldn't be able to, to, um, -detect that there was a red figure painted on a green background. -Yeah. -Yeah. -Okay. -The key component here is that the, the reference frame is inside of the edge. It's, you know, -if you're a color blind, there is something physical inside of your head or your eyes or whatever -that does not allow you to see the difference. So, so your semantics are affected ultimately by -a physical piece of hardware. -Well, yeah, maybe I used a bad example because color actually is computed in the brain and color -blindness is not necessarily physical. So, uh, okay. But I, I think I get it closer, but you know, -if we, if I may ask another question, I get lost with a saying that preparation and measurement are, -they have to happen both. And that when you prepare or when you read, you modify. So I am okay with the -act of writing, the act of encoding. So when I encode something such as I, you know, mark red on the -canvas, or I write to the hard drive, a, I modify the hard drive. That is true. But if I, if I am asking a -question, right, I'm not modifying anything on the canvas. So in what way the act of asking the -question or preparation is modifying the qubit, this is where I'm lost. You know, if I, does that make -sense? -Yeah. I think this is exactly the question I asked, uh, to Chris in, in the stream. Um, so I can't quite -remember his answer, but I think it was basically that they're completely dual pictures. Um, in the -sense of, you know, they're, they're both two ways of looking at the same thing. Uh, it's kind of like, -you know, evolving towards the future or, or, or, or the path you have time symmetry, you know, -you can, you can do either. Um, yeah. So I, I myself, I'm slightly confused by the very, -but I think the answer is that they are anything at the same, the picture of, or am I preparing -or just completely dual? Okay. That's it. Well, do you remember that, Daniel? -Yeah. Okay. I'll give a different angle than the, um, the formal duality relationship. So let's think -about our measuring device is going to be a carpet and we're interested in whether somebody has walked -on the carpet. So to prepare the carpet, we comb it very nicely in one direction. And I'm sure -there's analogies there with the, um, preparation of qubits and a quantum computer, but we're going -to comb the carpet as an act of preparation. And then something is going to happen. Maybe it's a -split second later. And so we're very confident that it should be still nicely combed, or maybe it's -going to be a whole working day later, but now we're going to take that probe, the tip, whatever it -may be, and we're going to run it along the carpet. And we're going to measure the state of, of the -direction of the fibers of the carpet. So if we didn't prepare the measurements, we wouldn't have a -calibrated reference point to know how things had even changed. And so there's, there's a cycle of -preparing the apparatus through something like a calibration, or at least like a emptying of memory -to be able to then make an observation and to have that observation, have the kind of meaning that -we would have expected, like purge the photo detector, make the measurement. Now you have to purge it again. -That's the preparation. What do you think Alexei? That is a beautiful metaphor. And I dig that, except -I think I heard Chris say that when you prepare, you may cube it into up or down. And this is when -I'm lost because when you comb the carpet, you don't do that. You don't put the footstep on it. In fact, -there may not be any footstep, right? And this is, this is where I'm missing, you know, I think in quantum -physics as well, or if you measure there's the cause and effect relationships are broken here for me. -You know, if in my act of preparation, I'm opening my canvas to, to read the message, but I do not know -what the message will be, I'm game. But if he's saying that when I prepare, I modify the carpet -such that it becomes either up or down, actually not blank, but up or down, then I'm lost. -You have to comb the carpet up or down. There's no blank carpet. So you, you, um, you know, you have -a slate, you have an etch a sketch. You could either make all the pixels white, and then you'd be able -to measure a message in the dark, or you can make all the message dark or the slate dark. And then you -could be able to measure something that was light, but you pick a direction of polarization -in your preparation or calibration step so that deviations from that can be the measurements. -So you're saying carpet is binary. Carpet doesn't have multiple states. Carpet can only have two -states. So if I prepare it, I, I effectively put it in one of those states. Well, let's say anything. -Yeah. Let's say we had the carpet at a 45 degree angle. Mathematically, we could describe that as a -mixed state and it may not be the most effective for measuring the movement of people across the -carpet, but yeah, you could, um, define it, um, or comb it that way. It doesn't necessarily, if it, -if it were only a binary, it wouldn't be a qubit, then it would be a classical bit. -I actually, yeah. Andre, go for it. And then Dean. -Yeah. With, with, with, you know, I, I, you know, recall lecture a couple of slides more. -We briefly talked about it like 10 minutes ago by preparing, you are choosing an axis and whatever -we're done in that axis. Now that doesn't mean that the environment gets to do its own thing. Right. -And a superposition of, uh, outcomes, but at least you're measuring that axis. Right. So it's, it's going -to be either up or down in that, in, you can add it to picture. And I have, uh, I just want to say a few -words on graphic screens and define my diagrams and maybe clarify the question a little more. -Dean. Dean. -On this really, really interesting. I don't know how metaphorically you could explain the carpet -as a wave, but I think once you get away from the, it's up or down, that possibility enters the -conversation. And the only reason I, I wonder whether or not metaphorically the wave aspect of that can be -included in how you're describing the carpet is that Chris spent quite a bit of time on the all possible -paths argument. And in particular, and I love this when Alice isn't looking. -Dean. Right. Well, in that, in that wave equation, how could Alice not be looking? -That's what gets real. I mean, Alice could not be looking if it's, if it's binary, if it's or up or down, -but if it's wave, it's on a, it's on a, it's a unitary propagator. I'm not sure how she takes her eyes off of this -preparation measure thing that's going off all the time. So this is, this is actually an, what I consider to be a feature or an opportunity or an opening -in the sense that from a strictly information transfer piece, not a mechanical piece. Now, not a physical thing. -From an information transfer piece, maybe the carpet can be metaphorically understood as a wave. -That the eyes of Alice or whoever's doing the preparing and the measuring, Bob, Alice, I don't care, -peaches. That there is actually a difference then between something that is exchanged and something -that is informationally transferred. And again, I don't think that the four of us here sitting right now -are going to resolve that. But I wonder if part of our, part of the difficulty in this is -finding the physical metaphor that explains the theoretical transfer in such a way that it doesn't, -it, it, it both breaks the symmetry and holds it. Because again, we're back to gripper and grip. We're -back to figuring it out and outing the physical figures in the state that we're in. And we have to do both -at once. So again, Daniel, you're, you're the creative guy. How do you turn your carpet metaphor into a wave -equation? Maybe there's a subwoofer in the room, or maybe somebody is waving the carpet. But what you, -what you said about the, what is exchanged and what is transferred, -maybe the transfer is the classical components. I don't know which will be exchange or transfer, -but one of them is the classical inscription. That's how many gigabytes were written or read to -the hard drive. But what's exchanged, let's just say, are the semantics of meaning. And given an -inappropriate co-preparation, maybe only one bit needs to be transferred for the exchange of a message -that was already triggered to be set off. And then also this entire discussion, it kind of strikes me -as like the, uh, Cartesian duality of our time, but rather than mind and body, it's the physicality -of information syntax and the metaphysicality of information semantics. And that's these kind of -two worlds with their own physics and metaphysics that we're trying to connect. And it's like, -similarly, well, is the information, how does the information world cause something physical? -How does the physical cause something informational? And I think that the entire journey that the course -and everyone is on is, uh, physics as information processing that those as a qualified and distinguished, -but minimum of two set Alexi. Yeah. Thank you so much. Uh, can you hear me? Yeah. I I'm yeah. -Thinking out loud maybe, but I think the way to get closer to understanding in my head is to let go of -the metaphor of asking a question, right? So if Alexi is talking to Daniel through zoom and I pose -even a closed question up or down. So I, you know, uh, formulated the question and, you know, uh, -released that into zoom space and then I'm silent and I opened my listening to your answer, but you, -Daniel have agency to say up or down. I did not, I do not have any effect on whether you will say up -or down, right? That's the difference. So me asking you a question does not modify, uh, the answer you, -you decide Alice talking to Bob, Bob has the agency to answer the question, not Alice. And that that's -where I'm stuck. So I don't know if asking a question continues to be a good metaphor, prepare, -preparing the instrument for measurement is, is maybe better. And, and so, but, but actually -Alice is not asking Bob, Alice is acting on a qubit, right? And qubit is a, is a simple thing that, -you know, can, can literally physically be up or down, I guess. And so that's, that's maybe, -maybe that's the answer that Alice does not ask in Bob. Alice does something to the qubit. -And because qubit can be up or down, then that's no longer asking a question, -asking a question implies your, your correspondent can, can, can tell you anything. -When we're dealing with the very sophisticated process of natural language question asking, -I think we're dealing, we're, we're dealing with something that that's orders, um, of composition -above the kind of, the kind of binary ask. So I, I, I agree with that. Um, it's almost like, um, -Alice prepares the photo receptor. So it's slate is refreshed and then it's the photon or the light -emitting object that has the agency to send the photon or not. I mean, maybe the measure, the -preparation occurs and then no one comes back to check whether or not agency was even exerted. -And I think when we really focus on like the nano probe, like atomic force microscopy, -really tracing with a super fine tip over the surface of something or scattering x-rays, -when we really think about the, the, the, um, moment of information transfer, -we realize like nothing comes for free. And I think that this question asking framework, -other than really conveniently re-deriving quantum mechanics in one hour, incredibly helps us -understand that we don't get to ask it for free in terms of time. We don't get to ask it for free -in terms of energy. And we don't get to ask free floating questions where the question and or -agentic responder are not involved already in a semantics. So I think it's calling attention to, -the components of an experiment or a question or an opening, but agree that it's not always as simple -as a binary question, but in the special, simple basal case, it does start there. -Andre, do you want to look, you want to look at any of the, um, the, uh, participant submitted -questions? If there was any that struck your face. -Let me finish real quick. I have like maybe five minutes to wrap this up. Uh, so, and, and this, -I think is going to at least in part answer one of Dean's concerns. So the next thing we should, uh, -very briefly cover. So let's talk a little bit about, you know, holographic screens. And also, -uh, this is the paper on the, you know, physical meaning of holographic principle. So let's take a -look at also this picture from that paper. This is the holographic boundary and, you know, -Landauer's principle and all these thermodynamic constraints on information processing -necessitate, uh, the breakdown of this boundary in terms of, you know, free energy extraction sectors, -memory sectors, and so on. And by the way, if you, if you read, um, the paragraphs around this picture, -there's more details on how the time QRF is induced, but we'll leave that as it is for now. Uh, -Dean, what I meant by this answering your, your, your concern, your concern was that you don't see how -Alice can take off her eyes from the picture. Yeah. Uh, but that's precisely what's going on here, -right? There is, there is a sector that's being observed, which is a subset of these qubits -that are being observed, but presumably she has the freedom to, you know, say, hey, this sector, -I'm now I, I declare as my observation sector, this I declare as my memory sector, this I declare as my -thermo, uh, free energy extraction sector. So it is because of these fundamental limits -on the amount of information processing. It is not true that she can, she doesn't have an, an idealized, -you know, you know, uh, Laplacian, if you wish, um, access to all the degrees of freedom. She can only -observe finally, uh, a finite amount of stuff. So, so she, she is forced to take her eyes off at least -part of the system, at least for a while. Right. Because otherwise, you know, all those energy and time -constraints that Daniel was just mentioning, you know, they're not, we, you have to abide by those. -So, um, so a few things that Chris mentioned in the last lecture when it came to this holographic -screen is, well, number one, uh, that all the information exchange is mediated by bits, uh, here -at the whole qubits at this holographic screen. This in turn enforces, um, the finiteness in the -information transfer, if you wish, there is not infinitely, uh, an infinite amount of information -in finite space. Even if you were to pack it at Planckian density, which is presumably the limit -that black holes achieve, it's still a finite amount of information in a finite amount of. -So if you have finite icons, you are taking. Uh, the other is the following. Um, and again, uh, Chris -will deal with this at greater detail in perhaps lecture five, I think maybe four, but it's the notion -of emergent space time, which is that A and B separately. So you think of, you know, these are -presumably the internal degrees of freedom, what's going on inside of A, Alice at an abstract level, -there may not be a notion of space, but, you know, the notion of space emerges at the boundary. -And, uh, that's, uh, not easy to explain. I think presumably has to do with and stuff like that. Um, -and I don't know the full details myself, but, but for now, that's, that's, that's one of the key things -that, that, that Chris mentioned, which is that an agent, you know, may not have space, but it may emerge -and the last thing, and, uh, I'm glad that to see that this was mentioned already. So it saves us time. -I don't have to say it again. It'll take us a minute. This five months. And this again answers your, -can you turn off your video under just so you don't have lag? Yeah. Thank you. -One second. -One second. -Yes. -Can you, yeah. -Me? -Go for it. -Okay. So this notion of, uh, Feynman diagrams, right? So what is this whole thing about unitary -evolution or preparation? The, the scattering picture as Alice preparing a vector of her boundary, -right? And writing it into her memory. And then -measuring it at some point in the future, right? Requires that you consider the possible -what happened to that preparation. Okay. So if there's some amount of time passes, -you prepared, you know, your state at time t, you look at it again, at time t plus dt, -you have to take into account all the possible histories. -Clearly, I, myself of the system, or at least that part of the system, right? Dean is just saying, -I prepared it here. Now it's up to B what happens until the next time I look at it, -up to the environment. So I, I, I have to, in, in that unitary evolution, I consider all the -possibilities, right? And that's, that's exactly, you know, the Feynman, you know, the role for. -So let's just look at the conclusion. -I'll leave it there for now. No, we can open up the discussion. So there's two things that are now, -where was this? Yeah. Yes. So, go ahead. -Yeah. So all I was, I understand, or I think I understand, I think I understand, -that if I'm looking at on, or if I'm attending to on, I'm not looking at off. What, what I saw in that -wave thing with Feynman was, I could also be looking at the switch, where the two different -places of on and off exist. And when I'm looking at the entire switch, I'm not looking at one or the -other. I'm looking at the, at that basal, as Daniel described, that basal opportunity of both. And, -and I think that on that, I don't know how that ties in with reference frames, but on that level that says, -I'm now looking at all possible. In that moment, I don't think at least from that perspective, -that reference frame, I'm zooming into a on or off binary choice decision branch, I'm, I'm still at -that place where the preparation allows for, for both. I'm not, I'm literally not taking my eyes off -at that place. So again, I'm not arguing with you, because I what you described, and the image that -you brought up, all that is, is bringing it down to a granular level and saying, if I'm paying attention -to this, I can't be paying attention to that. But if I physically, from a transfer perspective, -back up, so that I can see both, both possibilities at once, I don't know that I'm taking my eye off of -either. That's, that's the only thing that I was bringing up. Now, of course, Chris will probably hear -this and go, Dean, you got to go and rethink this whole thing through, which is great, because I'm -here to learn. But that's, again, if I miss, I miss spoke about what I was trying to say, it wasn't -because I, I don't see those moments when I'm directing my, my preparation over here. And that's -going to give me one kind of measurement versus if I'm taking in a larger timeframe, and now or larger -spatial frame or larger holographic screen, like I don't even know how to describe it, because I just -don't have the words for it yet. But there you go. I just want to be clear that I'm not, I don't think -I'm confused. I just want to know the difference between when I'm looking at on or off. And when -I'm looking at the entire switch, that's all. I'll add one point on this slide. It goes back to some of -the earliest points that Chris made about the ontological primacy of communication. Space and -time arise from communicating agents. Communicating agents are not something that we have to squeeze -into an a priori spatial temporal framework. So quantum mechanics, saying here on the slide, -is about measurement, not mechanical motion. So we don't need to wait around or speculate as to how -from the mechanical or physical hard drive. Measurement arises or happens. We start with the measurements -occurring. And then different kinds of sophisticated agents can conceptualize of movements through spaces -and times. And this also relates to Bayesian mechanics as a mechanics and a physics -of movement on information geometric spaces. And so the physics can be projected by appropriately prepared -agents. But we don't necessarily need to wonder about how that agent fits within a mechanical setting. -And I guess this is where humans are with the I am a strange loop, Gertl Escherbach, -gripper and gripped mystery of being able to see it multiple ways. -Andre, do you want to go to the questions? -Yeah, let me just say, are some remarks here? I think what Chris is the punching line of the paper, -free energy principle for quantum systems, which is, you know, morally speaking, it's similar to -the relaxation to thermodynamical equilibrium, which is that if you have an isolated -bipartite quantum system a and b are going to turn um are going to tend towards reference frame -alignment and therefore entanglement and this these two things are are basically -very intimately related by aligning the reference frames you'll get a and b -will be driven towards maximal entanglement -so uh should i uh take a look at the questions and i can open this yeah -so so uh if you guys have access to the q a -uh there's been more questions in the last you know bringing you guys's attention to a few of them -uh this is where um chris links uh to the papers of robelli which i mentioned earlier in passing -when we were talking about the preparation measurement cycle and the notion of thermal time or -uh you know subjective time if you wish somebody brought up this remark that you know performance -some calculations here you can relate t time to a change in um -uh a change in the volume of phase space or you know more precisely the the amount of available states -in the state space you know with the log and whatever uh with respect to the energy right and you know -you know so so this you know restates the -the idea that we're talking about earlier right like in your intrinsic uh subjective time being -this ds the e derivative a thermal time basically speaking that is uh observer dependent -so that's a a good question to uh take a look at -um there's another one what is what is meant by local free of choice uh this is a very good answer -to local here means that's on boundary right isolated uh systems are idealized by you know -definition there's always going to be an observer interacting with an environment so -a nice a purely isolated system is an idealization so what local free choice means here is exactly what -what we said earlier right that you get to choose the spin axis at the boundary qubits -um -and there were some uh nice questions relating mark of blankets to the scattering picture -uh so this one for instance is also worth a look -uh -so somebody you know brought up this you know very good point about the scattering matrix -the scattering matrix is by the way you know let's not think about it mathematically just um -let's think about the picture right you you imagine two particles that are interacting -and far away they are free right they are spatially separated and interacting free particles -and somehow at some point they come to interact right and then you have to write all the finite diagrams -for all the exchange of bosons or whatever the interaction is and and that's how you would compute -the amplitudes for the scattering process right how those two particles that were you know separated came -to interact and then they go off to infinity again uh chris here brings up the point that in the sense -the mark of blanket is you know in the opposite direction right you see a marker blanket in at least in the graph theory sense you start with a node and the marker blanket is just the set of nodes that -separate in terms of conditional probability that node from everyone else right and then you can have edges which are going to be directly uh causal networks so acyclic -no no no cycles directed edges that's that's how the graph looks -and you you build up so essentially you're taking the opposite um picture right you start with a node -which in this case would be the interaction and then you build up to infinity as opposed to in the -scattering picture which is like you start with particles not interacting at infinity you see how they go off to infinity again -and then the interaction is is in the middle uh so that obviously the answer is is a bit more uh precise but it's worth taking a look at -and this is a an interesting one too of a symmetrical input input what bandwidth so again if you think of it as a graph nothing is constraining you to say that you're going to have -more incoming errors than outcome errors into the boundary right that that's just the way it is you know if you think of it a graph there is no symmetry imposed the only natural sense of symmetry would be energy conservation and that's exactly what you know chris says here right which is that um -and i'm not sure how you related to a conservation of not quite information but you know i wouldn't know how to phrase it myself -i would just take a look at the question but basically no there's a priori nothing that necessitates symmetry in input output in markov blankets -but if you are to talk about any sort of symmetry the natural language would not be in terms of incoming versus -zaka mineros but rather energy conservation which we we saw is is you know a close casting of information transfer right because of -the minimal requirements for action to rewrite a qubit -and the last question that i would like to bring your you guys's attention to i mean all of them are great -uh but we have finite resources and finite times on internet uh would be -would be uh this one okay and i'm not gonna go into too much detail but this is again about -the local wait no one second -um -so it turned on my camera by the way dino go for it -and also people people can see all the questions on the website right and submit their own as well -yes for sure uh so and now i lost track of the question that i wanted to show you guys -i think it's this one over here yes uh yes exactly this one so by the way all of them are good even -this one the holographic principle does it make it easier to compute any physics problems well please take -a look at chris's answer but all i can say is that these ideas of holography are i'm not an expert i'm -not a condensed matter physicist but they're they're used all over physics nowadays not necessarily just in high energy -i know that condensed matter theorists use them all the time so the answer to this is -uh an emphatic yes the holographic at least adacft is used all over but again i'm not a physicist myself -um so this is the last question i wanted to bring your attention to and uh here chris uh this this may -be the one worth taking a look at in terms of uh in order to answer all uh you know the earlier -speculations that we had about time internal clock here somebody asks about um -synchrony and -you know just the perception of time with respect to external clocks i suppose -and chris here answers that at the end of the day any even if you're using an external clock -everything that you perceive outside at the end of the day has to be in terms of an internal clock -any information incoming you're gonna have to you know make causal sense of it you're gonna have to -write it into a memory in terms of an internal clock so so you're never going to be able to -ease so to speak increase your precision of the passage of time by taking an external clock -even if you think it's more precise than anything you have inside i think that's basically what his -answer amounts to and this is morally related to you know all this frame halting problem stuff where because -i'm never as a system you're never going to give uh an infinitely precise you know i'm speaking very loosely -here an infinitely precise description or yourself you're you're never gonna morally speaking again -that's that's what it amounts to you're never going to be able to outdo your internal clock by -by trying to leverage something external like like you know an atomic clock that even if you think it's more -precise at the end of the day your own subjective perception of information will depend on internal clocks -so we can uh open up things any concluding remarks and any other alexi or dean or anyone else who may -have joined i don't have anything else to say at this point i mean where where or when are we we had session one -with a real big historical overview section two we focused on the measurability question which is answered with -the question and we talked about some of the constraints of question asking namely that it takes time it takes -energy you have to ask it to somebody or something on an interface and it has to have a semantic quantum -reference frame and this is dovetailing with a free energy principle of course because the free energy principle -as it says in the syllables to what does the free energy principle apply and chris says everything -measurable that's why it's a theory of every thing that we choose to measure model not ontological -statements about what is and then where are we heading into in session three -have you looked ahead under or do you want me to read what's in the syllabus -uh please do yeah yeah i think basically uh my short answer to that and possibly my full answer also -is that it's just going to take care of the loose ends in this lecture two so this lecture two is a -warm-up to the to the real thing which we need to talk more seriously about reference frames to to in -order to get the precise uh picture of this quantum measurement process yeah yeah it says that session -three will introduce the idea of quantum reference frames and the representation using hierarchies -of binary classifiers these formal structures provide a semantics for measurement and hence provide -the basis for a theory of meaning for interacting agents the language of qrfs allows a particularly -straightforward and intuitive definition of variational free energy and so allows a fully general -quantum formulation of the fep we will see that the fep is a classical limit on the principle of -unitarity the fundamental principle of quantum theory uh by the way uh daniel can i i'm not sure -if you're busy but i just got an email from chris i think he may be watching us and i think he's just -confused about the link so chris i don't know if you're there still but um do you think it's worth -it we send him the link real quick if you can come for a few minutes yeah or i will oh i'll send chris the -um the zoom link now so yeah hold on hold on to your qubit chris we're coming yeah -so let's let's let's give him a few minutes and maybe he's still watching and he'll he'll answer the -questions better than i could well while daniel is sending him sending him a a lifeline can i just from -the from the the rookie perspective on this session one seemed like there was a big time window like -the historical piece over a using an external clock and a timeline this one felt like the parabola -and the upper down and the entanglement was almost instantaneous in terms of that plank length that -we were really holding up as comparative to that timeline a chronological timeline and so i'm guessing -that now that we've established those two as different where does that go but again that's -just the rookies view of things hey chris welcome yeah hi this is a very interesting discussion -um and it it's uh given me some pointers as to what to emphasize at the beginning of session three -uh but i i can say in response to dean's question that the time when alice isn't looking is the time -during which bob is making his measurement and then his next preparation step -so if you if you consider again that slide that shows uh sort of the fineman representation of what's -happening as this unitary uh flow that's what bob is doing and i think andrew said that -oh from alice's point of view bob is just some quantum system and she's banged on that system by asking a question -and now that system is doing something or other that she can't see until it it i.e bob bangs on his side of the screen -and she gets some information by making a measurement uh so she she may be not looking for a very short period of time -uh right i mean according to her internal clock she's not looking for just the time it takes for one of her -internal ticks to take place uh but uh in in general uh we're not saying anything about how long that time is -right if you think of this in a scattering theory way uh then we're talking about the time between two protons -leaving the rings on the uh large hadron collider to the time that all the stuff that comes out of their -collision is measured by the detectors which is a nanosecond or something like that some very short period of time -uh so just just think of that interval as the the measure of the intrinsic -uh time dependence external time dependence of the interaction someone has to do something and then -someone else has to do something and there's a gap uh between what the first person does and what they see -in response oh let's see what's the other question uh the the other central question that came up today -was the question of why observing something requires action and of course in the in the classical picture -observing something doesn't require any action at all uh the the event just happens and this is the the -fundamental if you will distinction between classical and quantum theory is that making an observation -in in in quantum theory is taking an action on the world and there's a lovely uh metaphor that's that's not -correct in detail but it's it's very useful i think conceptually uh which heisenberg came up with to try -to explain his uncertainty principle to his classical colleagues he said look suppose i'm trying to -to to to to see a very small particle and measure its motion oh i have to have a microscope to see this particle -and to measure its motion i need to to light it up i need to shine some light on it and those photons of -light are actually impacting the thing i'm trying to see -so the the the act of observation itself requires banging something into an object -and when we're just looking out at the world it's the sun that's banging the light into the object -not us but we can we can sort of merge those two systems together because it's the sun that's taking the -action that lets us get some information but the sun is perturbing the things that we're looking at -by banging photons into them um and i think that that picture of of heisenberg's is what uh he used to -to actually derive the uncertainty principle in terms of momentum and uh position because banging something -into something is transferring momentum to it photons transfers of momentum to the tiny particle in the -microscope and that perturbs its position on the microscope stage which is precisely what heisenberg -is trying to measure so that's his his point in that analogy observation itself always takes action -and that's why observation and preparation are duals in quantum theory because they each require a minimum -one unit of action and i think andrew you said the the duality can be thought of in terms of time -and variance and that's exactly correct preparation is just measurement run backwards right and vice versa -and if you let me say a few words i think this also answers my own confusion and the one that alexi the -question that alexi posed earlier which is this duality between preparation and measurement in your -picture that you're bringing up chris with heisenberg by shining photons on a particle that i'm trying -to observe um these pictures of preparation and measurement are dual because in in doing so i'm -constraining the system in trying to measure it i'm constraining it so it is so in me trying to gather -information by shining photons on it what i'm really doing is i'm constraining it i'm just saying -you can only take this this you know momenta or or position is would that be correct -uh yeah in fact you're in the microscope example it's particularly clear you're you're bouncing a photon -off of the object and then it's that photon that you're detecting right oh there was one other -question that uh can be answered quickly it was the question about uh what's the reference frame that -uh alice is using when she's uh measuring up or down measuring her spin up or down and of course -there's there there's several reference frames involved here oh so let's work inwards uh in a real system -alice is sitting here she's got a stern gerlock apparatus which is a vertical magnet apparatus -and she's measuring the spin of an electron that's that's traveling through this stern gerlock device -and what the stern gerlock device does is is orient the spin uh with a magnetic field that's that's not -homogeneous so it ends up pointing physically up uh as it comes through the apparatus so how is that -apparatus oriented i mean what defines the up direction for that apparatus well in the practical setting what -defines that apparatus orientation is the earth's gravitational field right we're sitting we're in -we're here in the laboratory and the earth's gravitational field tells us what direction is -down it's toward the center of the earth and so up is away from the center of the earth and that's -what defines the axis for the apparatus which is a great big piece of metal so it's very responsive to the -gravitational field so now we can ask about alice right alice is also embedded in this gravitational field -and she also can detect the gravitational field she can tell which way is down toward the center of the earth -by the the force acting on her feet and that measurement is taking place in alice's brain -so in alice's brain there's a representation of which way is up and that's the internal reference frame of -alice that lets her do the experiment you know if if she was freely floating in space but the apparatus was -anchored to the gravitational field she wouldn't be able to tell that it was anchored to the gravitational field -that's einstein's the point that einstein makes in all of those experiments where he has observers flying -around in spacecraft uh that they have to have their own reference frame uh what what einstein didn't -emphasize was that reference frame is in their head it's encoded by their nervous system uh so -uh i i think it was daniel who said this exactly correctly that uh the physical object uh in her -head the the actual neural network that tells her which way the gravitational field is uh is itself a -semantic object i mean that's that's what gives the word up and down meaning for alice and what is what -is what do i mean when i say the word meaning i mean it lets her take some action um semantics boils down to -actionability if it gives things significance and this this was the point that that bateson was making -when he said differences that make a difference uh for an organism uh something is meaningful if it -actually makes a difference for what i can do uh in this world that i'm trying to survive in -oh so i'll i'll leave it at that but this is this is very very useful for uh leading off the next -session so thank you very much for for all that discussion awesome thank you i guess if i could -say one thing here is that uh let me scroll back up to the three points early in the slides -differences that make a difference would tell you that you require action to to perturb the system and -measure and then the semantics or the contextuality that may be encoded in the in the in the nervous -system would tell you what is the difference that is being made would that be correct yeah right what -what tells you that there's a difference there it's it's it's what what you're able to detect and -represent internally oh it's almost like step one do differently step two you have to ask it to a -different thing and then three you need the semantic frame that makes a and b different -difference differences all the way down um well we will transcribe this and fix the few patchy moments -enter we'll work together on updating the transcript so that everybody who wants to be well prepared -for the measurement of the third lecture will be absolutely in place -in place thank you chris for joining that's a fun fun cap to this discussion so all right thank you see -you all next time bye -thank you - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -Thank you. \ No newline at end of file diff --git a/data/video/activeinferenceinstitute/Other/rOoTkgLI1Z8/README.md b/data/video/activeinferenceinstitute/Other/rOoTkgLI1Z8/README.md deleted file mode 100644 index eb65c7732..000000000 --- a/data/video/activeinferenceinstitute/Other/rOoTkgLI1Z8/README.md +++ /dev/null @@ -1,5 +0,0 @@ -# rOoTkgLI1Z8 - -Series: **Other** - -- ["Physics as Information Processing" ~ Ander Aguirre ~ Discussion 1](https://www.youtube.com/watch?v=rOoTkgLI1Z8) — `rOoTkgLI1Z8` diff --git a/data/video/activeinferenceinstitute/Other/rOoTkgLI1Z8/metadata.json b/data/video/activeinferenceinstitute/Other/rOoTkgLI1Z8/metadata.json deleted file mode 100644 index 13013af67..000000000 --- a/data/video/activeinferenceinstitute/Other/rOoTkgLI1Z8/metadata.json +++ /dev/null @@ -1,135 +0,0 @@ -{ - "series": "Other", - "item": "rOoTkgLI1Z8", - "source": "youtube", - "channel": "ActiveInferenceInstitute", - "category": "Courses/PhysicsAsInformationProcessing_ChrisFields", - "episode": null, - "title": "Chris Fields ~ Physics As Information Processing ~ 2023 Discussion 1 ~ Discussion 1 — Course information", - "date": "2023-06-03", - "guests": [ - "Ander Aguirre" - ], - "other_participants": [ - "Daniel Friedman" - ], - "github": "https://github.com/ActiveInferenceInstitute/ActiveInferenceJournal/tree/main/data/video/activeinferenceinstitute/Other/rOoTkgLI1Z8", - "enriched_from": [ - "coda", - "generated", - "youtube" - ], - "parts": [ - { - "video_id": "rOoTkgLI1Z8", - "url": "https://www.youtube.com/watch?v=rOoTkgLI1Z8", - "title": "\"Physics as Information Processing\" ~ Ander Aguirre ~ Discussion 1", - "duration": 3988.0, - "upload_date": "" - } - ], - "sessions": [ - { - "index": 1, - "session_name": "rOoTkgLI1Z8_sess01", - "start": "0:00:00", - "title": "Introduction" - }, - { - "index": 2, - "session_name": "rOoTkgLI1Z8_sess02", - "start": "0:01:00", - "title": "Ander Aguirre" - }, - { - "index": 3, - "session_name": "rOoTkgLI1Z8_sess03", - "start": "0:02:08", - "title": "Ross" - }, - { - "index": 4, - "session_name": "rOoTkgLI1Z8_sess04", - "start": "0:02:46", - "title": "Avo" - }, - { - "index": 5, - "session_name": "rOoTkgLI1Z8_sess05", - "start": "0:03:39", - "title": "Alexia" - }, - { - "index": 6, - "session_name": "rOoTkgLI1Z8_sess06", - "start": "0:04:57", - "title": "Francesco" - }, - { - "index": 7, - "session_name": "rOoTkgLI1Z8_sess07", - "start": "0:06:00", - "title": "Corby" - }, - { - "index": 8, - "session_name": "rOoTkgLI1Z8_sess08", - "start": "0:07:37", - "title": "Daniel" - }, - { - "index": 9, - "session_name": "rOoTkgLI1Z8_sess09", - "start": "0:14:14", - "title": "Alex" - }, - { - "index": 10, - "session_name": "rOoTkgLI1Z8_sess10", - "start": "0:22:04", - "title": "Information" - }, - { - "index": 11, - "session_name": "rOoTkgLI1Z8_sess11", - "start": "0:23:04", - "title": "Questions" - }, - { - "index": 12, - "session_name": "rOoTkgLI1Z8_sess12", - "start": "0:35:21", - "title": "Wick Rotation" - }, - { - "index": 13, - "session_name": "rOoTkgLI1Z8_sess13", - "start": "0:38:31", - "title": "Local Time" - }, - { - "index": 14, - "session_name": "rOoTkgLI1Z8_sess14", - "start": "0:44:31", - "title": "Space Time" - }, - { - "index": 15, - "session_name": "rOoTkgLI1Z8_sess15", - "start": "0:51:36", - "title": "Information Processing" - }, - { - "index": 16, - "session_name": "rOoTkgLI1Z8_sess16", - "start": "0:55:04", - "title": "Materialist Dualism" - }, - { - "index": 17, - "session_name": "rOoTkgLI1Z8_sess17", - "start": "0:57:17", - "title": "Two Types of Information Processing" - } - ] -} diff --git a/data/video/activeinferenceinstitute/Other/rOoTkgLI1Z8/transcript.txt b/data/video/activeinferenceinstitute/Other/rOoTkgLI1Z8/transcript.txt deleted file mode 100644 index f55a18cd0..000000000 --- a/data/video/activeinferenceinstitute/Other/rOoTkgLI1Z8/transcript.txt +++ /dev/null @@ -1,541 +0,0 @@ -## rOoTkgLI1Z8 - -Hello and welcome everyone. It's June 3rd, 2023. We're in the first discussion section for Physics -as Information Processing course with Chris Fields. So welcome to all who are here on and -off camera. Really exciting to see so many people joining for this participatory discussion section. -So in this first interval, we're going to go to anyone who raises their hand and it would be -awesome to hear from everyone who would like to share what excites them about what they heard in -the first lecture. How has their background and journey led to them joining? What are they hoping -to learn and do in this course? So first hand, we'll get to go. Maybe, Ander, you can actually -start first, just speaking as a normal participant. And then meanwhile, anyone is welcome to raise -their hand. That sounds good. So I am a mathematician by training, -professor of physics. But I've always been interested in the more physics side of things. -And in particular, you know, I followed Chris's work pretty closely since probably 2019. -And I don't know, I'm just very interested in the whole ontology of quantum measurement, what it means. -I feel like, you know, very often when it's taught in physics classes, it's hand-waved away. And even -though I myself subscribe to the philosophy of, you know, shut up and compute, sometimes, you know, -not everything. You need to do a little bit of philosophy maybe. And this may be one of the things -where you need to do it. And you need to address it from first principles, right? And I think Chris is -interested in that. So I'm just here to learn more. And yeah, that's basically it. -Awesome. I'll just go from the top left, because I'm not necessarily watching like what order, but -Ross first. -Hi, I'm interested in the kind of overlap between cognitive science and physics, -physics, and seeing if there are any panpsychist implications of the observation that physical -processes seem to follow the same kind of principle that we see applied to brains and cognitive systems. -Awesome. Thank you. Avel? -Hi, I'm Avel Guénin-Carluge. I have mixed backgrounds, mostly physics and cognitive science. Right now, -I could be qualified as a philosopher of physics and cognitive science. And I'm interested in -basically formalizing participatory realism, the idea that the observer participates in creating reality. -I did some work in formalizing that in the case of social norms, and things like language. And I -met a block based on my lack of knowledge of quantum physics and the mathematical tools -that I used for that. So I'm here in a course to get those formal skills. -Awesome. Thank you, Avel. Blue? -I think Alex was first. -Okay. Alexi, go for it. -Oh, sorry. -Hi, I'm Alexi. I'm a clinical psychologist in Maryland, in the States. I work with patients, -do psychotherapy mostly. My interest came through Mark Solms, who is the creator and leader of -Neuropsychoanalysis. And he has been collaborating with Friston on heart problem of consciousness. And -he's now working on AI with emotions. And so he, you know, I blame him, I say it with respect and love -that for introducing mathematics into our field, which was mostly devoid of mathematics. And people -started looking at these partial differential equations and Markov blankets. So I wrote a paper -about a possible extension of their model to include chaos theory. And I just want to understand deeper -concepts involved. And I have some understanding of them, but perhaps not enough. And I'm enjoying -this course a great deal. I mean, I think Chris, among other things, a fantastic teacher, and he is -able to express things very clearly. And it just, I have all, I have goosebumps. So I'm very grateful to -everybody. And thank you so much. Cool. Blue, then Francesco. -Hi, everyone. I'm Blue Knight. I'm with the Active Inference Institute. So I've been -tracking this work of Chris's, you know, connecting quantum information and active inference for, I don't -know, a couple of couple of years now through through that thread. I'm super interested, actually, in -the aspect of space time arising from communication. I think that that's like fascinating to dive into and -how it connects to the like Buddhist and maybe more Eastern religion, like concept of dependent -arising, like how like the interdependence of all things leads to their existence. And so from that -aspect, I think that that that's like the what I'm most interested in exploring. My background is -neuroscience, but I've also been studying and practicing Buddhism for like 25 years. So yeah, -that's it. Thanks. -Awesome. Francesco, then Corby. -Thanks, Daniel. Ciao, everyone. Super happy to be here. I also have a mixed background because I -studied cognitive anthropology and philosophy of science. And currently, a few months ago, I started a PhD in -artificial intelligence in an eighth application in the educational field. And yes, I'm actually super excited -about the all the science implications that Chris Fields mentioned in the first lecture, the connection, I have -no background in physics, but the connection between the first principles of physics and free energy principle -and scientific cognition in general is what is really getting me excited in these days. Thank you. -Cool. Corby, then anyone else who raises their hint? -Hey, yeah, great to be here. My background is in quantum engineering. I've always been motivated by -figuring out what the next big thing in computing was in high school. Morph's law was sort of coming to an -end. And it seemed like a really big opportunity to figure out what the next big computing architecture -might be. So I did academic research and no more computing, quantum computing, and quantum sensors. -And I felt like I was a little bit too early for all those things. So I went into cloud and cloud gaming -and streaming AI services. But I'm still extremely motivated to figure out what the next big thing is -in the holographic principle seems like a good foundation to explore that. -Awesome. I'll raise my own hand, then anyone else is welcome to go. So I'm Daniel. I'm a researcher at the -Institute and have known Ander for some years and learned about all kinds of stochastic matrices and -how the math and the quantum were linked. And we had talked and joked about learning from Chris -fields for a long time. So when the opportunity arose to enact it and extend it and open it, -we were very excited and also don't have much of a formal background in physics outside of working -with Yakup and Ali on Livestream 49. So it's just going to be a great time to learn and see where it all goes. -So anyone else wants to say hello in this kind of opening section? -Okay, well, continuing forward, of course, please just feel free to raise your hand anytime you'd like to -jump into the stack. Ander, for several minutes or however you see, could you reflect on lecture one? -Where did Chris enter? How did he proceed? And where did the lecture take us? -Can you hear me? Yeah. Thank you. Good. Go for it. -All right. Let's take a look at the slides from last time. So I think if we were to summarize this -whole lecture, basically what Chris was trying to say is that there are many hints coming from different -directions pointing at the same thing, right? Pointing at the fact that information will have a central -aspect in physics in the future, but also possibly biology, right? So the earliest hints came from -statistical mechanics, if you wish, but also quantum mechanics, right? And here we need to make a -like an important but subtle distinction between what it is the uncertainty principle, right? Because, -you know, we think of that as the informational part, right? Like a limit on the resolution on information. -And what is the observer effect? Just to make that distinction clear, I think Chris is talking about -the observer effect. The uncertainty principle is an artifact of the wave nature of, you know, quantum mechanics, -right? You can get it from properties of the Fourier transform, Cauchy-Schwarz inequality. But at the end of -the day, it's what I was saying earlier, right? What is the basic picture of measurement, right? And that's -sort of what Chris with Glacebrook and Marciano is trying to address. And the first two hints came from -from quantum mechanics around 1900 and statistical mechanics a few decades earlier. -So then there's, you know, more stuff, right? Like, obviously, physicists resisted to inertia of hundreds -of years of a totally objective, you know, observer independent universe, but the hints continue to pile on. -So then there's some stuff coming from math, but perhaps the most interesting is, again, you know, how -all those paradoxes from statistical mechanics were made sense of with Landauer's proposal. -And then as physics got more fancy past, you know, QFT and stuff, you know, people started thinking about -quantum information more seriously. So we had Bell's theorem and all these proposals by Deutsch and -Wheeler and so on. And I should say that this is not a fringe idea in physics. The paradigm of it from -QBIT is very much mainstream. I mean, Google it, a lot of this is just think about it, you know, -I don't know how many of them, you know, talk about it openly, but you know, it's been a widespread -folk intuition, so to speak, that, you know, information will have this sort of central role. -And the last sort of hints pointing in that direction of information, having a central law came -on the one side from high energy stuff. And that's the holographic principle is, it originally arose in -the context of black hole thermodynamics, but then, you know, it was extended to, to, you know, what's -called the ADS CFT correspondence. And the very most recent stuff was, you know, it's basically -Christians were having essential law came on the one side from high energy stuff. -And so I can speak a little more loudly. It's all good. Continue on. -Yeah, I thought there was a question. Someone was just unmuted. It's all good. -Okay. Yeah. So, so that's basically it, right? So let's, there's hints from many directions. And -most recently, you know, the work of Friston and Levin coming from biology, all pointing in the same -direction, right? That we might have to make sense of some sort of, eventually scale-free -information processing for that, for, for what it means to be an observer, you know, sort of all, -all of these things are pointing in that direction. And I think Chris's work is, is addressing that -head on, right? As I said earlier, you know, sometimes, you know, it's, it's good to do computations, -but sometimes you need to take a step back and think more philosophically about what you're doing -and what the bigger picture is. And then maybe after people talk a little bit, I can, you know, -we can warm up to the second lecture. We can, the other windows that I have open here, -some of them are the papers that are to come. -Yeah. Anyone, please feel free to raise your hand on this. Alexei, and then anyone else? -Alexei So I had one thought about this thing, you know, when, when you've mentioned that there's -various perspectives on the same concept and I'm trying to be cautious on the one hand, I think, -again, this amazing clarity and precision with which Chris expressed thoughts is, is fantastic and -enjoyable. On the other hand, it requires simplification and generalization. And what I -observe in my field is take people take a term like entropy and run with it. And then they get very -quickly from thermodynamic entropy to Shannon to, you know, whatever entropy and they talk about sort of, -and that's not the case. I mean, if we go back to Clausius and all the others, there was a -system of postulates and axioms that must be maintained. And if you step outside of that, so like -us organisms were open systems, we're not closed systems. So we should not apply directly second -world thermodynamics. And, you know, I, I deal a little more with Kolmogorov's and I entropy and, -which is not the same thing. And I guess when we very quickly move from sort of thermodynamics to, -you know, to statistical mechanics with Boltzmann to, you know, all the way around, and we assume it is -the same thing, I think we may need to be careful because we are just ignoring the differences and -assumptions. You know, Shannon talked about telegraph. He talked about a receiver getting a message from -a sender, right? And Clausius talked about, you know, gas. So, but, but I do think it's a balancing act. -And there's no way to not have simplification and generalization. I just wanted to say somewhere, -somehow we need to kind of, you know, go back there and just say, you know, entropy is not entropy, -is not entropy. And all these things are kind of, you know, long papers, right? -Yes, Jakob, then Ali, then anyone else. -Um, I, um, I just wanted to, uh, ask for, I guess, perhaps, uh, slightly longer comment from Alexi on -that, what do you see as the fundamental differences in different descriptions of entropy and whether there is -some, I guess, general concept that could be extracted from all of them, such as level of -information or uncertainty about a particular system, even though these specific, uh, formulations of -entropy, they, um, talk about, they were derived from different specific systems and they have different -equations describing them and they mean slightly different things. Do, um, do you think it would -be wrong to say that there is some commonality between them? -Should I answer Daniel or? -Yeah, please. -Yeah, I don't know. -I just think we need to at some point formalize it and, um, and be careful with these generalizations. -For example, uh, when we talk about active inference and Friston, right. You know, when we say living -organisms must minimize, I'm quoting Friston, by the way, these are not my thoughts, the entropy of -sensory states, right. And then people take that idea sometimes in generalize, and they say that -that's what life is minimization of entropy, right. But if you actually look into human brain and measure -entropy, then it goes all over the place up, down and in between. Uh, and so I, I don't think that's -an accurate description. And the reason for that is where we shifted away from the, uh, assumptions. -And, uh, again, so Kolmogorov's in I entropy talks about the phase space and the divergence of -trajectories in phase space and sort of the volume in phase space. While, when you talk about this, -for instance, uh, entropy of sensory states, these are probability distributions. -And, uh, Shannon talked about the uncertainty of the receiver about the message sent by the sender. -Dr. So if you want to extract the same thing, I suppose you can, but, but it's important not to -lose at least the difference between what kind of object are we applying to? Are we looking at -probability distributions? Are we looking at phase space? Are we looking at ideal gas? Where are we at? -And we're just not to say entropy is entropy is entropy. That's not the case. I mean, it's just a formula, -but depending on the context, assumptions and things, it can be very different. -Thanks, Ollie and Ava. -Uh, yes. In regard to Alex's comment, I just wanted to mention the research of, um, people like -Rudolf Hanel and Stephen Serner, uh, because, uh, for many years, uh, they've been exploring this idea of -different conceptions of entropy and, uh, they've come up with at least a dozen different meanings for -entropy and they formulated it, uh, in a kind of generalized framework, uh, but at the same time, -they've also devised, uh, a way to somehow, uh, classify them into equivalence classes and to see whether, -uh, one or more of those, uh, conceptions of entropy can be theoretically derived from, uh, from some -others or they're totally, uh, distinct and different, um, uh, conceptions of entropy. So, yes, I agree that -in, uh, some situations we need to be more specific about what exactly we mean by entropy and information. -Uh, I mean, based on, uh, the question or the situation of interest. -Thanks. Ava? -So just to put in the entropy discussion, I want to emphasize that entropy, uh, PLNP, so it's a function of -probability discussion, which is, uh, I think, I'm not sure what the term is in English, but it's a measure of how big, uh, something is. -Uh, and to, to get that, you have to define a space. And in many cases, it's pretty arbitrary to decide what -this space you're looking at is. For example, you talk, I don't know if it was Alexei, it was Alexei. -Someone talked about, uh, uh, uh, entropy of, uh, fMRI. Well, those are aggregate measurements of neural -activity. This is not the space of all possible configuration. And even if I add something like the -space of all, let's say functional configuration, the state in your own that reflect activity, this would be -another thing than the space of, uh, I don't know, atomic configuration where the specific atom are. So -yeah, you don't have a warranty that when you define entropy, you define it on a meaningful space. -That's not given. So let's, I kind of think there is one entropy and that's it and it's meaningful constructs. -But also we have to, if we agree to that, then it entails that we have to find it, that it's not just -given out there in the data. -Hmm. -Well, certainly being precise in the formalism and in natural language for what we mean by information -and entropy and all these terms is pretty important. I mean, the course we're in is physics as information -processing, but without an understanding of how information is being meant here or what measurements -and transformations were, were performing, what informational operations, then it, it, um, -is a non sequitur. So how do we, how do we go about, um, clarifying or understanding what kinds of -theoretical results in a given domain or like what is shown on the timeline here might apply to some other -area. -Just one thought. Anyone else can raise their hands with, uh, a thought or a question. Otherwise, -Anders, you can, um, maybe flip over to, to select a few of the submitted questions. -Cool. Could you zoom in, um, with control plus a fair amount? -Thanks. So everyone, um, we really appreciate the submitted questions. We've been able to -get 13 answered by Chris so far. And, um, as with the transcripts of the discussions and lectures, -the questions and answers are going to be part of what is published by the ACDIMP journal. So we really -encourage basic questions, advanced questions, just, there's no question that doesn't make sense to -submit. So under maybe just pick one question that you think's interesting to begin with and summarize, -uh, Chris's response. And then everyone else is welcome to kind of add some other thoughts on this. -Yeah. So I'm going to scroll up and down just to sort of sample. Uh, let me start with the first two. -Can you hear me well? Yep. Uh, since perhaps, I think you submitted this as a trial, right? Is that, -is that right? Yeah. So what is information? What is time? Um, and here's Chris's response. I guess -everyone has a chance to look at it. Now the one, if, if I am to summarize this, the one thing I've heard -that, you know, personally, the one I liked the most, um, answer, at least as, as an adage, -to summarize it is that information, um, are differences that make a difference. This, -I might have heard from Chris. I might have heard from someone else. I don't remember, -but that's one thing. Now onto the question of what is time. And I guess this also addresses, uh, -uh, Alexis question of what is entropy. I, so I'll, I'll make a bigger point now. Uh, what is time? -This is partially addressed. And, and I guess the main thing I'll be doing here is rather than look back, -look forward into what's to come in the course. And in particular, I'm thinking about, um, -um, yeah, this, this, these two papers, uh, and in particular, -particular, and I've had a chat with, always Chris about this, uh, this picture over here. Okay. -Can you zoom in more? Yeah. I'll get back to it in a bit, but basically this, this sort of like -kills two birds with a stone, right? This picture, and especially the text around it, -and the title of this paper is the physical meaning of the holographic principle. Okay. -You guys can see here archive 2020, 22, 10, blah, blah, blah, blah. Anyways, this, uh, addresses the -notion of, you know, the time and entropy, because the, the point that's the claim that's made in this -paper is that these are local observer relative definitions. And we can, uh, look into more detail -around it, but basically the coordinate TA time is an entropic measure. So it, it is non-decreasing -and it goes up with the observer measuring, uh, entropy in the system. Because of the symmetry of -the definition, the second law applies to both. So both observer and environment see entropy increasing -in their environment and vice versa. Uh, anyways, from, from this picture, one gets that you have -this internal circle time QRF. So this just is, is a necessary artifact of, of recording information -into a memory. That's, that's the claim made in the paper. And that essentially induces the passage of -the time, at least from the perspective of the observer. Um, it also, I think, uh, Alexei was talking about, -which I totally, uh, understand and agree that this different notions of, of entropy. Now, I suppose as the, -the, you know, I, I would be the one to, to like definitions and so on. Um, but I, but I do think it's, it's, -it's good to, to have a certain amount of flexibility. Um, so I see the same as energy, -right? Like there are different kinds of energy, you know, there's energy mass equivalences and, -you know, if you push it into supersymmetry, you can think of it as equivalent to force also. -But I, I do think there is some value in being slopy or intuitive about, about the notion. So, so the -fact that, you know, entropy, yes, there are different definitions. I, I don't, I don't see it quite as a -problem. You know, one can explore here around this, this page, there is a definition of, you know, -local entropic time and yeah, one can think of it as, you know, face space or one can think of it as, -you know, coding theory, a la Shannon, but it is at the end of the day, I do think it is true that both -of them are hinting at the same direction, just like, you know, you may have different kinds of energy, -but, you know, they're not unrelated concepts. And, and yeah, I hope one day we get very precise -definitions on everything, but, but for now I think it's, it's good to speculate. -Okay. Thanks. -So we can, there's a question. -Do I see? -Yeah, Corby and then anyone else on this? -Yeah, quick question. So I haven't read that specific paper, but is there any work on the -frequency or the rate at which one could read and write information on the boundary? -Yes. Well, I, I don't, I don't, I can't think of the, the paper exactly, but this is lower bounded. -I think Chris talked about it. If not in the last lecture, I, I think I've seen a talk of him on -YouTube. It's lower bounded by the uncertainty principle, right? So, so for instance, plants, -Planck's constant is in units of action, which is energy times time, right? So Landauer's principle -says you need a certain amount of energy to, to erase information and, and you have time, right? -And then you have an energy time uncertainty principle, putting those things together, right? -You have any minimum amount of action, which is, you know, literally the minimum, um, the action is -measuring what I said earlier, right? Like differences that make a difference, the minimum difference that -you can make. So the minimum difference to, you know, erase a bit or rewrite it on the boundary, -so to speak, right? Uh, so yeah, that would be given by the uncertain relation energy times time. -It's I, I don't know exactly what the figure is, but it's obviously very small compared to everyday life. -Got it. And would, is there work on the upper bound? Like for example, would it be the number of -degrees of freedom on boundary? No, I don't think it's bounded from above, but, but obviously, you know, -then we, we do want to buy it from above for technological reasons, if you wish. But I, I think -that is a, an engineering problem, not, not one of physics. -Um, Andre, few, few questions on this figure. Um, what are Y and E? -Uh, B, the blue oval, as Chris has pointed out is going to consistently be the screen, -the blanket. And so we often associate that with, uh, at least some blankets include -these information gathering and utilizing systems. So what is happening internal to the agent here? -What kind of computation is happening? And is this something that is actually proposed to be occurring? -Or is this a schematic or a map of how we could think about something occurring? -So I think this question is going to be better addressed by Chris himself, probably in the next -lecture. Um, for now, what I can say is I can reiterate what I was trying to say earlier, which is that, -what this addresses, you know, this basic measurement picture, so to speak, is the observer effect, right? -Like this notion that you cannot observe something without, without disturbing the environment. -So if you then try to make sense of this by constraining, you know, from first principles, -Landauer's principle and so on, uh, basic requirements on information processing from the point of view of energy, -then you get that this boundary is constrained in some ways, right? This Y is what's called the memory -sector, meaning that you don't just act on, on, on the immediate information that you perceive. You may -also act on, you know, stuff that's stored that you have perceived before. And then sure, E, the red part, -is what you're perceiving right now, the environment. Uh, but actually, I don't know if I have it in here. -Uh, but this is further broken up into other parts, so not schematically here. Let me try to find it over here. -Now, I don't know. Actually, bear with me for a second. I think you might be here. -That's the same picture. -Anyways, I, I'm not, I don't want to waste time by trying to find it, but -back to the original picture. Yeah. Sure. It is what you're measuring. And again, Chris will talk -about this in more detail. Uh, but from first principle assumptions, -they deduce that you're going to have to, you know, deploy your finite resources for gathering -information in clever ways on the boundary, right? So you're going to have to deploy some resources for -memory, some resources, uh, as a free energy, uh, as a heat sink, right? So you're going to be -processing information. So you're going to be dumping, you know, so-called low quality stuff -into the environment. You're going to be dumping heat. And then older than the memory, you're going -to be focusing on what's going on right now in this environment. And what I wanted to say is that it can -be further broken up into what's called pointer and reference sectors. So you can imagine this -E sector here has to be broken into further sectors because then there's the other first -principle notion. Okay. So what does it mean to observe sense something, the differences that make -a difference, right? So you can only, there's only information with context, right? There is no -information per se, it's all semantics, right? So you only observe things in context. So some things -are going to have to stay the same, namely, you know, reference states here and some things are -going to bear, right? So you can think of this as the gauge in your car, right? Obviously the circle -stays the same, right? And that's your context, so to speak, that's the, with the, you know, numbers in -it, they never go away, they stay the same. But as you speed up around, the gauge is going to change. -And that's how, you know, only in that context, you can tell your own speed when you're driving the -car. Right? So anyways, this is just a teaser for things to come. Now we can go back to more questions -then. What do you think? I'll pick maybe two or three more over here. Yeah, pick another question -or anyone raising their hand or if anyone in the live chat wants to write something, I'll read it. -How about the wick rotation? Could you summarize what the wick rotation is and then state this -question about it? So, yeah, that's a great question. I knew the answer, but I wish I knew -the answer better. But, informally speaking, the wick rotation is just, so it's, but Chris said, -you know, you take a quantity multiplied by I, square root of negative one, that amounts to a 90 -degree rotation in a complex plane. Fine. Now, that appears to be a trick. And what it does effectively -is take you from quantum mechanics to statistical mechanics, roughly speaking. -Now, why it works, that's probably, I mean, I have no clue. I don't know the answer myself. But now, -when things appear to be a trick and they get you something, maybe there's more to it than what's, -you know, apparent at first. But I myself don't know the answer. It's a very good question. -But I do think it may have to do with the fact that, on the one hand, quantum and statistical -mechanics both are talking about constraints on information processing, maybe from different -perspectives. But again, that's just speculation on my end. So, it's maybe a way of relating the two. -Can we just review, just read Chris's full answer to this? Read the question and then Chris's full answer? -So, as the weak rotation tells us that matter and energy are somehow -exists 90 degrees perpendicular to a three-dimensional space. -And then I'll just read Chris's answer. Then anyone, please raise your hand if you want to, -like, add something more on this. He wrote, -In a sense, yes. Time is perpendicular to our three dimensions of space, matter, and energy. We -ourselves exist in time. Wick is pointing to the intimate relationship between how we measure time -and how we determine or judge that something maintains its identity through time, i.e. stays the -same thing through time. To tell time by a clock, I have to be sure I'm looking at the same clock, -set the same way, etc. We'll get to this in July and August. -Chris Cousin, Jr.: Could you connect this, Andr, to some of the semantic information flows -discussed in Livestream 17 or any of the other cone-co-cone diagrams from the papers that you showed -earlier? -Andr, Jr.: Sure. Again, I have to preface it by saying that I am working on understanding it myself, -but just from, you know, small email exchange that I had with Chris Amps, what's said in this paper, -one can sort of see that it's related to the picture I was showing earlier. -I mean, this one, right? But let's scroll down a little bit to page, -page 2324. Here they talk about the reportation again. -Ah, excuse me. -So, -I think it has to do with, at least from what I heard from Chris' email, has to do with this notion of -local time, the definition of local time in this observation picture. -And, you know, the fact that, you know, you rotate 90 degrees twice, you get the opposite sign, -basically, right? So, the local time, effectively, that's going to ensure that the picture is symmetric and that, -you know, the flow is going to be opposite in the direction towards the observer and towards the environment. -Okay, Alexei, then anyone else raising their hands? -Alexei, I am struggling to understand and part of my difficulty bear with me for being so kind of, -but I remember Sean Carroll's lecture on time in different versions of physics, and it took a whole -hour. So, are we talking Newtonian time, which is universally the same? Are we talking relative time? -You know, philosophically, are we talking about people who believe that only present here and now is -real while future and past do not exist? You know, so when we say time is perpendicular to -my sense is that, just like Chris mentioned in the lecture, we're reinventing the time. We're creating -a new definition by calling it the same word we're all familiar with, and that creates some dissonance -where I just think that maybe we need to kind of say that this is something new. And if it's not -something new, then which version of the time are we talking about? Einsteinian, Newtonian, or whatever, -you know? It's a simple question before we say it's perpendicular. Do people agree that future is -real or not? You know? -I'll give a thought on this, Alexei. Again, speculation, neither here nor then. -If we take Chris seriously that communication is an ontological primary, and that rather than -putting communications protocols in some sort of time, Newtonian time, Einsteinian time, these are -these different time concepts, these different scaffolds that then communication gets inscribed -within or stretched within. But if we start with a topological understanding of communication, -rather than a geometric understanding of space-time, so the topology of the blanket and the observer -in their sensorium, then different kinds of cognitive agents, ranging from what we would call non-living to -what we would call living, they, depending on their cognitive sophistication, are able to -create or project different time concepts. Time concepts that are not useful will lead to their -own material dissolution, the failure to maintain that kind of cognitive sophistication, which requires -not just information gathering and utilization, but also energy gathering and utilization. -And so when you have multiple agents who are, whether you conceptualize them as within a message -passing relationship and having no surprise on the clock on the wall, then they're like in a local time -synchrony. But the primary case is one entity engaging with its sensorium and its self-modeling time in terms of progression. -And so these previous time concepts and these different attempts to kind of survey the structure of space-time -reflect different cognitive models that cognitive agents have proposed for time. -And that communication actually like takes ontological primacy over all of these ramified, -potentially overlapping, potentially non-overlapping time concepts. -But they're downstream of agents' cognitive models, not vice versa. -I'm not sure. It's how I interpreted what Chris said about communication being an ontological primary though. -So if we take one instance of a COVID-19 virus, a single kind of instance, right? It's got no memory and no -ability to plan. Is there a local time for that virus versus, you know, humanity observing how the virus -spreads across the globe? And, you know, we have agents with memory and planning, you know? So -is there time relative to a single photon? You know, I just, I mean, it's, it's, my hunch is that there's a -possibility that we're simply inventing something brand new and defining it brand new. And I just, -the drawing parallel back to all these other times, you know, I understand the effort, but it's, it's, it's just, -uh, is it, is that a brand new time? Is it some other kind of time? I don't know. -Can I, Daniel, can I say the words? -So I, I understand Alex's question. And I think it's sort of related to, at least morally speaking, -to your concern about different kinds of entropy, right? And what I can say is that also they have -Glace Brook, Marciano and so on. They have this recent paper on quantum error correction. -So I have yet to read it and take a closer look, but clearly talk about space-time, right? And my punch -to address Alex's question is that if you could sort of recover the more familiar notions of time -in a correspondence principle sort of way, you know, by taking some, you know, Newtonian limit or -H bar goes to zero or speed of light goes to infinity and so on, uh, that would be very satisfactory. Sure. -They may be doing a new thing, but, you know, recovering all stuff in the new thing, uh, is a good -sanity check. And so long as we can see that being worked out, I think I'd be very happy. -And just to the virus point, I would argue that viruses through their physical engagement, they do -have memory, but they don't necessarily have the awareness or a metacognition on that memory, -memory, except as granted by intergenerational evolutionary type pressure, but they do have past events -modifying them within some, um, set of possibilities. -So they're just being them and their internal time is, is on there. There isn't a cognitive observer necessarily -in terms of like a narrative. I, I am watching the clock, but they do change. -But it's certainly very provocative if this is not, or we could also look to Kronos and Kairos -and about the relationship of time and timeliness and agentic time. So I think there's, and then at -this point, are we stretching just this one four letter word too far, just like we brought up with -information entropy. I mean, can we think about these as kind of like continents and then just, -if someone said, I visited this continent, you know, that you need to specify a lot more or what -granularity are these topics, Corby and then anyone else raising their hands. -Yeah. On this topic for a Markov blanket, you have a transition matrix. My question is, how do you, -how are you allowed to increment that transition matrix to calculate a future state? Like, what does -that mean in terms of time? Uh, are you only allowed to increment it by like the lower bound time, -which is H over LN KBT? Does that, does that question make sense? -I'll give a first pass on this. We have discrete time and continuous time models in active inference. -For a discrete time model where you have the B transition matrix, then the choice of delta T -is a modeler's decision. And we talked about this a lot in live stream 52 on accelerated optimization. -There's some delta T that's like going to get you the most accelerated optimization. If you have too -small of delta T, then you're either, um, oversampling or you're maybe even physically unrealistic like you -pointed to, or if you have some long delta T, you're also not making an informative simulation -in the discrete time case where you're explicitly making time steps. Contrast with the continuous time -generative model, where we're doing more of a Taylor series expansion on the generalized coordinates, -at which point delta T is not required. You're merely just taking an extended approximation from a -given snapshot. And so delta T doesn't even come into play. Got it. Okay. That makes sense. -I don't know how it ties into time, but it helps clarify a little bit. -Okay. Anyone else can raise their hand or, um, under kind of maybe one more question or one more topic -from a paper. Just what's one more puzzle piece that you think gives us a first coat of paint from -questions submitted or from one of these papers that you have up? -What's something that, that helps us connect the dots a little bit? -Uh, so I don't, I don't think I have a good answer. Let me look at the questions again. -Uh, -but regarding time, I'll, I'll just reiterate, uh, what I said as I look over these questions, -which is that I, I believe Chris will address it in, in this bit. I, I don't know if you can see what -I'm sharing on the screen, but the paper whose title is communication protocols, blah, blah, blah, -from March of this year, uh, does talk about space time. It's a quantum, uh, error correction code. -And to me personally, I, it'd be nice to see that can be more familiar notions of time, such as -special relativity can be recovered as limits in correspondence principles. -Um, -yeah. Do you, you want me to say anything else? -If, if anyone could even just summarize, like what, what would these, what are these codes referring to? -Or how are, are these codes one in the same as communication protocols? -Or are they related to the blanket and the, the heat that was being output as waste? -What will, what will these codes allow us to do better if we understood them more? -I don't know that I can answer that right now. I don't, I don't think they understand the question. -Um, yeah, I, I just don't know. Sorry. -Terry and then anyone else raising their hands. -Um, I, I'm not a physicist. I don't understand maths at all. -Um, but what attracted me to this, um, idea of physics as information processing is that, um, -um, it feels like at a base level, the, whatever the X unknowable external state is, it could be perceived as -made up of information. You know, we, we, we kind of think of ourselves living in a particulate -universe, but, you know, at, at, at a, at a base level, there are these, -and stop me if I'm just talking 14 year old nonsense here, but you know, that, that, that, -that this, this notion of quantum fields, they feel like informational fields. And, um, that the, -the concept of, of physics is purely information that we then interpret as a physical universe, -which we inhabit because we can't see beyond our Markov blanket. Um, -I find this a really exciting idea that as a, that whenever you start to think of the, um, -surprisal as the, um, variational free energy between, you know, sort of your, your prior and your -posterior, if you start to think of that variational free energy, not as energy, but as an information -gap, so that if you see the energy as actual, you know, cause energy, when you think of energy -and physics terms, it's kind of this magical thing that does, that allows work to happen or, you know, -but, but, but what is it, you know, we've got a, a term for it and we've got, we've got, we've got a, -you know, we, we, is it jewels, you know, or whatever, you know, but, but, but what does it -mean? It has this magical property, but if you think of it as an information, um, then it starts -to make sense. So variational free energy starts to be seen as a difference in information between -what you, uh, what your, your prior and your posterior. -So whenever you start to think about time, time is this conceptual metric that we use -almost because we have memory. Uh, I, and, um, I, I wonder if we are getting, uh, trapped in this -semantics and all I have is English because I don't understand the maths really, but, um, we're, we, we, -we, you, you know, you guys come in here and you do understand the maths and then you convert it into -English, which I think I understand. And then I can think of it like a kid in a Swedish shop or -a physics class at the 14 years of age. I don't really know what I'm saying. I don't know what -I'm saying, but at, at a fundamental level, everything that we're talking about is in for, -I, I, I can understand it as information and I'm excited to see how things develop. -What are your thoughts on this craziness? -So many, uh, great pieces in there. -Physics as information processing. What is it that actually allows the steam engine to run and -the ball to move that kind of physics. And then as the timeline laid out also occurring during a period -of, of immense technology and, and apparatus development and, and digitization and all of -this and unconventional computing as well. It's almost like this, uh, materialist dualism -that instead of with, with mind and body, that there was some kind of dualism between the types -of forces and energies and ways of thinking about what made the steam engine work and then what -makes a computer work. And so you'd understand how the computer could fall off a cliff, but then you -wouldn't necessarily be able to use that exact kind of mechanics and physics and dynamics to study like -how the computer operated, but you could say it weighed 11 pounds. And so I think the, the provocation -and then, which is as you hinted towards, um, first off would be awesome to have a natural language, -math and graphical synthesis, the triple play, but also then to use a common approach to understanding -this and what wasn't working or is not proposed to work is trying to fit information into -that non-cognitive universe. But rather, if we take the cognitive primacy of communication, -then all of these forces as inferred by the sensorium, which is all we're ever going to have, -totally the most valid first principles constraint to make, maybe then we'll be able to have a lot -more synthesis about even how steam engines work when we understand that we're bound in that inference -by our sensorium and all that comes along with that and the variational free energy bounding surprise on -our generative model. Dean? I'm going to, I'm going to come at this from the position of a 14 year old lens, -because I don't think there's anything unique about that. Um, I think on most basic level, -what I took away from Chris's lecture was that there are, that there's information processing of at least -two types. Uh, one is figuring it out as in figure five or any of these other figures that are up on the -screen right now or anything that stabilizes the world around us. I think examples have come up in -this conversation already of top apology or geometry. There's a second type of information processing, -which is outing the figure type of information, which is searching. And I think Alexi used the -expression, which time are we talking about? So these two types of processes are quite different. -If you do a comparative analysis of the two, you have information practicing and you have to have -both of the processes to really get a sense conceptually of what it is that communication is really talking -about. If you are information practicing your wayfinding until you collapse to something we might -describe in a verb tense as way found. It's in the quantizing of that invariant state, that vast ocean, -if you want to think of it metaphorically out to all, all horizons on the manifold into a variant or a -nuanced or a discretized or an orientation type of state where all of a sudden things like currents and -temperature shifts and even things that we've applied labels to like entropy and a transition matrix suddenly -appear out of the seeming nothingness or the invariance. So that's not the math, but that's kind of stepping -back and saying, what is the math trying to say? I think it boils down to the comparative analysis. -I'm hoping that in the coming lectures, Chris starts opening that up and doing kind of the back and -forth between the two types of figuring it out and outing the figures because I think when both are -part of the conversation, that's when we start to get a sense of what's really going on. -Chris, thank you. In this sort of closing sessions, especially if anyone hasn't added anything, -please feel free to give any last reflections or questions arising. So first, as many people as would -like can reflect that way. And then after every hand has been addressed, then, Ander, if you could just give -us a kind of look ahead. So Avil first, then anyone else with just where does this discussion take us -and what do you want to continue to learn more about? Avil. -Avil. So two things first, if I walk out of here without understanding what is the current co-configure, -I will be disappointed. And I don't actually expect it to happen in like 12 hours. That seems like -something. But more seriously, I can agree without difficulty that there is a unity between physical -structure and communication protocols. But communication protocols or quantum reference frame kind of -individual, physical states, individual meaningful measurement outcomes that are reified as physical -possibilities. But the question is then where do communication protocols slash quantum reference frame -come from? Like what are the physical determinants that bring about those things? And that seems like a -question that could be addressed at the conceptual level of this course. And I would like, I saw the -live stream 17. So that did not help. Anyway, I would like to have more grip into that if we can address it in the -form courses. Thank you. Thank you. Anyone else, especially if they haven't added anything yet? -Want to give kind of a closing thought. -Yeah, there's a lot to learn. And Chris brings up a lot in his lecture, even though we've only had one. It's one of the denser hours of epistemic resourcing that I've seen. And there's also all of these accessory and cutting edge works that -that under showing. So I'm really looking forward to how we can continue to curate and improve the questions that we have. -So that the uncertainties in our generative model that we have, how would this be useful? What does this mean? Why is this variable this way? What is the implication for that? -Those are all the great questions. And let's just get them inscribed on our classical documents, so that we can have other quantum -cognitive agents reduce their uncertainty by looking at the responses. So with that, I wonder if you could just look ahead to lecture two, and just like point towards where we're going to be going. -Yes. So lecture two will, I think from what I heard from Chris is going to start to address -this basic ontological aspects of the quantum measurement. So the two sources to look at would be -two papers, mainly this one, physical meaning of the holographic principle. This one's by far the most pedagogical one. So if I had to suggest that folks look at one, it would be this one. It's very easy to read, I think. -But maybe for a bit more detail, there's the slightly older one. And they talk at length about COCO and diagrams here. -Free energy principle for generic quantum systems. And so I think that's what's going to come in these two lectures. Basically the main ideas behind these two papers. -Anja, that's all I have to say. -All right. Well, thank you for all who joined this first discussion. Hope that it was fun for you and for the audience. And check out the course site for the readings to get ahead of the second lecture. -We'll all watch the second lecture and come back in about a month for the second discussion. Everyone's welcome to join and the website has the registration information. -And so now that we've seen a little bit of one way it can be, we may do something similar or we may do something different for the second discussion. -So thanks again, everybody. See you later. -Thank you. -Thank you. \ No newline at end of file