OpenAI Forms Math Panel After Claiming 100-Plus AI Solutions
The company says a new internal model has produced results on more than 100 open problems, prompting outside mathematicians to advise on review and publication.
Topics
News
- Apple Tightens Mac Access Rules as AI Agents Gain Reach
- Trump Puts Intelligence Chief Jay Clayton in Charge of AI Task Force
- IBM Widens AI, Quantum Research With IIT Bombay, IISc
- OpenAI, Synopsys Build AI Model That Could Slash Chip Design Time
- Pentagon Plans Four-Star Command for AI, Drone Warfare
- California Draws the Line on AI Firings, Layoffs and Workplace Surveillance
OpenAI has turned to an independent group of mathematicians for advice on how to review and release a growing number of AI-generated results after saying a new internal model has resolved more than 100 long-standing open problems.
The company said it began training the model on August 28. It is the same system OpenAI says produced its proposed solution to the Navier-Stokes Millennium Prize problem, which it released this month with a Lean formalization. The result is still being scrutinized by the wider mathematical community.
OpenAI said the speed of the model’s progress surprised mathematicians inside the company and triggered discussions over how potentially significant results should be assessed, communicated and released.
The move also follows criticism from prominent mathematicians. An open letter titled “A Severe Misalignment of AI in Mathematics” argued that using unresolved mathematical problems as benchmarks for AI systems could distort research incentives and create costs for the mathematical community.
OpenAI said the concerns demonstrated the need for greater involvement from researchers outside AI companies.
The new Advisory Group on Mathematics and Artificial Intelligence, hosted at the Institute for Advanced Study, will advise on the significance and release of emerging results and on academic standards for AI-assisted mathematical research.
The group says it operates independently of AI companies. Its members are unpaid, can publish recommendations and criticism, and have no decision-making authority over OpenAI.
Its immediate task is to advise OpenAI on the release of what the company describes as a large number of significant mathematical results produced by its internal model.
The panel will not, however, advise OpenAI on whether it should slow the development of its mathematical AI capabilities.
Initial members include François Charles, Camillo De Lellis, Timothy Gowers, Martin Hairer, Nikhil Srivastava, Ulrike Tillmann, Ravi Vakil, Edward Witten and Melanie Matchett Wood.
OpenAI described the arrangement as a first step toward giving mathematicians a larger role in decisions over how increasingly capable AI systems participate in research.
