Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Project AXIOGEN: The Emergence of Universal Logic Syntax

"Human brains have a mathematical thing inside them as if it was base code of an OS... once a kid learns the syntax of math, they know that 1+1=2 without needing to be taught."

AXIOGEN (Axiom + Genesis) is an experimental AI research project investigating the "Firmware of the Mind." It demonstrates that mathematical intuition, logical reasoning, and curiosity are not "learned behaviors" in the traditional sense, but are the mathematical axioms of survival that emerge inevitably when a neural system interacts with physical constraints.


📂 The Evolutionary Lineage (File Structure)

The project is structured as a biological timeline, where each stage "gives birth" to the next by passing down evolved neural weights (.pkl files).

AXIOGEN/
├── stage1/             # The Wolf (Motor Cortex & Physics)
│   ├── axiogen_evo_stage1.py
│   └── world_alpha.xml
├── stage2/             # The Scientist (Curiosity & Mapping)
│   ├── axiogen_stage2.py
│   └── axiogen_stage1_BEST.pkl (DNA from Stage 1)
├── stage3/             # The Architect (Sequential Logic)
│   ├── stage_3.py
│   └── axiogen_stage2_AUTOSAVE.pkl (DNA from Stage 2)
├── stage4/             # The Sentinel (Grounded Plasticity)
│   ├── stage4.py
│   └── axiogen_stage3_AUTOSAVE.pkl (DNA from Stage 3)
└── final_test/         # The University (Generalization Test)

🔬 Multi-Stage Cognitive Scaffolding

Stage 1: The Wolf (Hardware Drivers)

Objective: Master the physics of a 2D body.

  • Hypothesis: Efficient navigation is a mathematical necessity.
  • Achievement: The agent evolved to handle inertia, friction, and targeting. It reached the "Speed of Light" ceiling (maximum theoretical efficiency allowed by the physics engine) within 43 generations.
  • Key Result: Developed the "Subconscious" motor drivers required for all higher logic.

Stage 2: The Scientist (Intrinsic Motivation)

Objective: Exploration without external reward.

  • Mechanism: Implemented a "Hippocampus" memory grid and a Boredom Timer.
  • Transfer Learning: Injected the "Wolf" brain. Within 10 generations, the agent re-wired its sensors to treat walls as obstacles rather than food.
  • Result: Developed a systematic "Roomba-like" scanning pattern, proving that Curiosity is an optimal algorithm for uncertainty reduction.

Stage 3: The Architect (Conditional Logic)

Objective: Solve a sequential puzzle (Key $\rightarrow$ Gate $\rightarrow$ Goal).

  • Mechanism: Expanded the input layer to 10 dimensions, including an "Internal State Bit."
  • Breakthrough: Achieved Generation 0 Success. Because the "OS Drivers" (navigation) were already stable, the brain mapped the logic of "If Key=1, Then Goal" instantly.
  • Conclusion: Logic is simply 3D navigation applied to abstract states.

Stage 4: The Sentinel (Grounded Cognition)

Objective: Individual learning through "Pain" and "Energy."

  • Innovation: Intra-Life Plasticity. Agents adjust their own synaptic weights in real-time when they feel "Pain" (collision) or "Success" (key found).
  • Achievement: Created a Grounded Intelligence that "feels" the environment. The numbers became "meanings" (Pain = Negative Stimulus).

🎓 The University (Final Generalization Exam)

The "University" test drops the final Stage 4 brain into three "Planets" it has never seen before, with no further training allowed.

Environment Requirement Result
Planet Harvester High-speed targeting GRADUATED
Planet Snake Precision navigation in narrow space FAILED (Overfitting to open space)
Planet Hunter Dynamic tracking of moving goals GRADUATED

Scientific Insight: The failure in Planet B proves that while the "Logical OS" is universal, "Spatial Resolution" is limited by the training environment's geometry.


🛠️ Technical Implementation

Requirements

  • Python 3.9+
  • pygame (Visual engine)
  • neat-python (Evolutionary algorithm)
  • pandas & matplotlib (Analytics)

Installation

pip install pygame neat-python pandas matplotlib

📜 Philosophical Conclusion

Project AXIOGEN proves that the "mathematical thing inside the brain" is the result of Cognitive Scaffolding. We do not teach the AI logic; we provide the physical pressure (Starvation, Pain, Obstacles) and the AI reveals the logic through its neural connections.

The resulting .pkl brain files are not "code for a game"—they are the compiled syntax of survival.


💡 Author's Note

This project explores the bridge between Neuroscience and Artificial General Intelligence (AGI). It argues that AGI cannot be built through data alone, but must be evolved through grounded physical experience.

About

An approach to AGI

Resources

Stars

2 stars

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages