Cellular automata are the heavyweight champs of emergent pattern generation. Deep learning is the undefeated contender for pattern recognition. What happens when we put them in the ring? We will look at some simple experiments training deep networks on cellular automata, and discuss what each paradigm can bring to the other in terms of testing their ultimate limits, as well as practical implications for research.
Kovas Boguta is a senior engineer building AI infrastructure at Twitter Cortex. Previously he worked on data infrastructure and analytics at Weebly, and on Mathematica and Wolfram Alpha at Wolfram Research. His favorite lisp is Clojure.