OpenAI wants to consult elite mathematicians about how to not fumble again — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI formed the Advisory Group on Mathematics and Artificial Intelligence (AGMAI), a nine-member panel drawn from Stanford, Harvard, Oxford, and Cambridge with multiple Fields Medals and MacArthur grants, hosted by Princeton's Institute for Advanced Study.
- AGMAI will operate independently from OpenAI with members unpaid, free to offer unsolicited advice, publicly comment on AI's impact on mathematics, and make its advice public, advising on review, communication, and academic standards for emerging results.
- The panel's first task is advising OpenAI on coordinating release of results from its unreleased model, which has reportedly resolved more than 100 long-standing open problems across most areas of mathematics — the same model that solved the Navier-Stokes Millennium Prize problem and is closing in on another.
- The group's formation follows months of controversy over OpenAI's announcements, including accusations of scooping researchers, failing to credit underlying human work, and stealing from users of its models.
- Martin Hairer (Imperial College London/EPFL), an AGMAI member, published a blog post addressing 'torrent of misinformation' about the group's origins and asserting it is 'genuinely independent of OpenAI and any other of the so-called frontier AI labs.'
- Researchers Simon Machado (ETH Zurich) and Francesco Fournier-Facio (Heriot-Watt) questioned whether the elite cadre represents the broader community, with Fournier-Facio calling it 'ivory tower' and Machado saying their reality differs from most mathematicians' experience.
- Kevin Buzzard (Imperial College London) questioned the need for such an elite body, saying the company simply needed 'better understanding of our subject' — 'it wasn't proofs of hard theorems.'
Why it matters: OpenAI's unreleased model reportedly resolved more than 100 long-standing open problems, a volume that forced the company to seek outside help on releasing results without further inflaming a community already angered by accusations of scooping and uncredited work. Whether a nine-person elite panel can credibly speak for mathematicians broadly — and whether OpenAI will actually heed its advice — will shape how AI labs handle research credit as model outputs increasingly collide with academic priorities.
Ask SkimNews
