the discussion isn’t really about whether language models can become strong chess players though, the point is they seem to struggle to consistently make valid moves. Most humans don’t need to read two books to pick that up, just a couple lines of basic instructions