OpenAI’s math solutions aren’t meeting the field’s standards yet
Mathematicians say OpenAI's hundreds of claimed hard-problem proofs miss standards for human understanding.
OpenAI published 719 claimed solutions to difficult open mathematics problems after consulting mathematicians, but critics say the release misses community standards for human understanding. The Advisory Group on Mathematics and Artificial Intelligence, hosted at Princeton's Institute for Advanced Study, had urged labs to stop testing advanced problems on proprietary models. Only ten manuscripts included chain-of-thought, and 42 percent of the proofs had not been formalized. A University of Cambridge and King's College London paper found at least two discrepancies between OpenAI's natural-language argument and its Lean code for a problem derived from the Navier-Stokes equations.