An AI proving deep theorems is a mathematician; its proofs advance math, while humans still set the questions.
