Google researchers introduced Differentiable Logic Cellular Automata (DiffLogic CA), which applies differentiable logic gates to cellular automata. This method successfully replicates the rules of Conway’s Game of Life and generates patterns through learned discrete dynamics. The approach merges Neural Cellular Automata (NCA), which can learn arbitrary behaviors but lack discrete state constraints, with Differentiable Logic Gate Networks, which enable combinatorial logic discovery but have not been tested in recurrent settings. This integration paves the way for learnable, local, and discrete computing, potentially advancing programmable matter. The study explores whether Differentiable Logic CA can learn and generate complex patterns akin to traditional NCAs……
