OpenAI's 719 Math Proofs Fall Short, Institute for Advanced Study Group Says
Summary
OpenAI releases 719 manuscripts of claimed solutions to hard math problems. The Institute for Advanced Study's AGMAI says the release falls short of its guidelines, noting only 10 included model chain of thought. Cambridge and King's College London mathematicians report discrepancies between OpenAI's natural language proof and Lean code on a Navier-Stokes-derived problem.
Key Points
- Terence Tao writes on social media that AI prompters solving problems autonomously lack interest in the broader field.
- The AGMAI asks frontier labs to stop testing advanced mathematical problems on proprietary models in its first request.
- Harvard mathematics professor Melanie Wood tells TechCrunch no human understanding of the solutions exists at the point of release.