OpenAI’s mathematical proofs face scrutiny over lack of human understanding

OpenAI recently released hundreds of solutions to complex math problems, but mathematicians argue the results lack sufficient human understanding. The Advisory Group on Mathematics and Artificial Intelligence (AGMAI), based at Princeton’s Institute for Advanced Studies, urged labs to stop testing advanced math on proprietary models. OpenAI claimed to follow some guidelines, such as releasing results quickly and including model reasoning, but only 10 of 719 papers included detailed thought processes. AGMAI also called for formalizing proofs, a step only 42% of OpenAI’s papers achieved. A Cambridge and King’s College paper highlighted discrepancies between natural language explanations and formal code in OpenAI’s solutions to Navier-Stokes problems. Terence Tao criticized the lack of human engagement with AI-generated results. Melanie Wood noted that without human understanding, the work is just beginning.