OpenAI's math proofs fall short of expert standards

OpenAI released hundreds of claimed solutions to difficult math problems, but fell short of guidelines set by the Advisory Group on Mathematics and Artificial Intelligence. The lab tested proprietary models despite AGMAI's request not to, and only released chain-of-thought reasoning for 10 of 719 manuscripts. A paper from Cambridge and King's College London found at least two discrepancies between OpenAI's natural language proofs and their Lean code translations for a Navier-Stokes problem. Critics including mathematician Terence Tao argue AI-generated proofs lack the human understanding and community engagement that make mathematical results meaningful.