I'm trying to grok what was done here. Was the chip design itself generated by an LLM to implement a model that was learned by ML from the Connect 4 data? Or is there some other deterministic method for turning a learned Connect 4 model into an ASIC? For that matter, what is the nature of the model implemented in the chip?
(author) The model is learning directly from a dataset of (input bits, output bits). It is directly predicting output bits (index of which column to play next) from input bits (the state of the board right now). It is doing so using logic gates directly as compute primitives.