Organic neural cellular automata

Neural cellular automata (NCAs) are a class of algorithms in which each pixel follows a set of rules based on its neighboring pixels and decides what color it should be in the next frame.

Mitosis patterns in neural cellular automata feature little blob things that seem to be doing exactly what biological cells do—they’re splitting and replicating, making copies of themselves.

These blobs are replicators, and observing them helps appreciate how populations of replicators grow—if population growth follows an exponential curve, even a very bad replicator can create explosive population expansion.

In the mitosis patterns, natural selection is not happening—there’s no genetic difference between these blobs and thus no selective advantages.

Another mitosis pattern features blobs that seem to stick to the grid more and form longer chains of cells that sort of look like bacteria, similar to the bacteria that causes strep.

Organic NCA patterns are typically found fairly close to one another in parameter space—where one interesting pattern is found, tweaking the parameters a bit finds another and another and another.

Worm-like patterns feature structures that seem to grow outwards in long heads almost as if they were searching for something, with other worms budding off of existing heads.

The favorite NCA pattern features heads growing out as if they’re seeking something, sometimes interacting with other heads or little seed things that populate the screen which in turn grow into their own heads.

Observing slime mold patterns really opened eyes as to how living organisms can take advantage of simple emergent systems like this and use them to search and spread, grow and compete.