Humans should focus on coming up with conjectures and new problems to solve. I don't see LLMs doing that.