CMV: AI will relatively soon, solve the Millennium Prize Problems in a single day. I mostly just don't know enough about the topic to have a firm belief. I just have a thoughtful and considered intuition in one direction. What AI seems to be best at is pattern recognition well beyond human ability, abstraction and the ability to holistically parse language systems and produce meaningful sequences of that language from inputs. Mathematics is a language. Presumably you could train a model on all the known work in mathematics, set theory, and symbolic logic. All progress in mathematics is built upon previous results. It is an almost purely theoretical system, you don't need to conduct experiments in the world to produce most mathematical results. I do understand that in serious mathematics you aren't just following algorithmic rules, you are constructing strong mathematical arguments and then evaluating them, but I think, since this is a purely symbolic practice, AI would be quite good at it.…