Programmable Cellular Automata: readable code rules evolved via genetic programming
Amidos2006 · x · 2026-09-09
Amidos2006 introduces Programmable Cellular Automata (PCA), addressing the pain of authoring cellular automata: classic CA rules are lookup tables, while neural CAs are expressive but black boxes.
The core idea modularizes CA into small programs:
- Local Functions: neighborhood → value
- Global Functions (optional): whole state → value, breaking strict locality but enabling faster completion (e.g., counting)
- Decision Function: combines outputs → next cell state
Since rules are programs, they can be evolved with genetic programming: the chromosome is the list of all functions, crossover swaps programs between parents, and mutation replaces a program with a new one.
Related event: Programmable Cellular Automata: readable code replaces lookup tables(2 posts)→
More from Research
- Autoresearch Loop with Tinker Reproduces Self-Distillation Papers at Predictable Cost — SRSchmidgall · 2026-09-10
- Researcher Lets Codex Run the Experiments, Publishes Recurrent Model Length-Extrapolation Paper — qixing_huang · 2026-09-10
- Author defends paper's narrow scope: precision over breadth, more to come — brwilder · 2026-09-10
- Modeling LLMs as holding latent beliefs yields high-accuracy behavior predictions — brwilder · 2026-09-10
- PyData Amsterdam talk: Recursive Language Models with DSPy, from prompts to programs — lateinteraction · 2026-09-10
- Microsoft's EigenLI: training-free spectral compression of ColBERT-style representations — lateinteraction · 2026-09-10