A question for high-level math people: what is the difference or gap in capability (if any) between AI being able to solve preexisting open questions/probs, vs AI being able to venture forward on its own and identify genuine new math problems nobody ever thought to ask or isolate before?