This repository accompanies the v1 preprint “Toward a Physical Theory of Intelligence.” It collects exploratory code, configurations, and artifacts used to probe a candidate physical formulation of intelligence via controlled computational experiments. The theory is developed to advance AI safety research by grounding design and evaluation of intelligent systems in physical principles.
- Foundational v1 research artifact; interfaces, analyses, and conclusions are provisional.
- Some results are conjectural; empirical validation and methodological extensions are ongoing.
- Shared for transparency, discussion, and future development; not a polished library or production-ready framework.
A_Physical_Theory_of_Intelligence.pdf— current preprint.src/experiment_*— experiment scripts for memory capacity, substrate comparisons, edge-of-chaos sweeps, and energy cellular automata (see per-experiment READMEs for execution details).configs/— runnable configuration presets used in the experiments.outputs/— generated artifacts from runs (illustrative, not comprehensive).
License: MIT (see LICENSE).