Harvard mathematicians are split over whether correct AI proofs need human understanding, after OpenAI’s Navier-Stokes claim.
A proof does 2 things: it shows a statement is true, and it teaches other mathematicians a method they can reuse, and AI proofs often deliver only the first.
Mathematicians value proofs for 2 reasons, certainty and technique, and an AI proof that nobody understands gives them certainty while withholding the technique.