The "AI" that won Go was Monte Carlo tree search on a neural net "memory" of the outcome of millions of previous games; this is a LLM solving open ended problems. The tasks are hardly even comparable.