Scott Aaronson gave some issues with it being cellular automata even with threads connecting them (search "Bell Inequality" https://arxiv.org/pdf/quant-ph/0206089.pdf), so Wolfram has moved on to graph automata things due to, if I am remembering right, someone figuring out efficient approximations of determining graph isomorphisms and it making evocative pictures with some simple transform models, and also it solving the Bell's Inequality no-go problem.