Amorphous Intelligence Experiments

Author: gadwant

This directory contains code for illustrative experiments demonstrating key phenomena of amorphous intelligence.

Structure

code/
├── experiments/
│   ├── exp1_reasoning_divergence.py    # Experiment 1: Reasoning divergence under isolation
│   ├── exp2_epistemic_drift.py         # Experiment 2: Epistemic drift over time
│   ├── exp3_semantic_skew.py           # Experiment 3: Semantic skew in concept interpretation
│   └── exp4_synchronization.py         # Experiment 4: Opportunistic synchronization effects
├── utils/
│   └── __init__.py                      # Utility functions
└── README.md                            # This file

Requirements

• Python 3.8+
• numpy
• (Optional) Actual LLM APIs for real model deployment

Usage

Each experiment can be run independently:

# Experiment 1: Reasoning Divergence
python experiments/exp1_reasoning_divergence.py

# Experiment 2: Epistemic Drift
python experiments/exp2_epistemic_drift.py

# Experiment 3: Semantic Skew
python experiments/exp3_semantic_skew.py

# Experiment 4: Synchronization
python experiments/exp4_synchronization.py

Note

These experiments use real local LLMs via Ollama. Requirements:

1. Ollama installed and running.
2. Models pulled: mistral, llama3, gemma3.

The scripts will automatically connect to your local Ollama instance (default: http://localhost:11434). Execution times may vary depending on your hardware (GPU vs CPU).

Results

Each experiment generates a JSON results file in experiments/results/:

• exp1_reasoning_divergence.json
• exp2_epistemic_drift.json
• exp3_semantic_skew.json
• exp4_results.json

These results can be used for analysis and visualization to support the paper's claims.