OpenAI announced on 21 September that it has set up an independent Advisory Group on Mathematics and Artificial Intelligence — a nine-member body hosted at the Institute for Advanced Study in Princeton. The group will assess how significant the company’s model-generated mathematical results are and coordinate their release, but it has no influence over how fast OpenAI works.
Key takeaways
- Nine members, hosted by the Institute for Advanced Study in Princeton
- The group does not advise on the pace of OpenAI’s internal mathematics work
- OpenAI says an internal model has resolved more than 100 open problems
- Those include a solution to Navier–Stokes, one of the Millennium Prize problems
- 25 Fields Medal winners had earlier signed a letter criticising the AI labs’ race
What the group can and cannot do
The remit is narrow and precisely drawn — as TechCrunch reports, it also covers the right to offer advice nobody asked for.
| The group can | The group cannot |
|---|---|
| Judge the significance of model-generated results | Advise on the pace of internal work |
| Coordinate how results are announced | Make decisions at any AI company |
| Speak publicly without an invitation | — |
[The group] will not be responsible for advising us on how to pace our internal progress on mathematics.
OpenAI, in its announcement of the advisory group.
The Institute for Advanced Study puts it more bluntly: "we do not have decision making power at any AI company".
A hundred problems, none peer-reviewed
The claim the whole structure rests on is a strong one: an internal OpenAI model is said to have resolved more than 100 open problems across most areas of mathematics, including Navier–Stokes, one of the seven Millennium Prize problems?Millennium Prize problems: Seven open mathematical problems announced in 2000 by the Clay Mathematics Institute, each carrying a one-million-dollar prize.. None of those results has been independently reviewed yet, and that gap is precisely what the group exists to address.
Background: the letter from 25 Fields medalists
The body is not being created in a vacuum. Earlier, 25 Fields Medal?Fields Medal: The highest distinction in mathematics, awarded every four years to mathematicians under 40. winners signed an open letter criticising how AI labs compete over famous mathematical problems. Of the group’s nine members, only one — Camillo De Lellis of the IAS — also signed that letter. OpenAI did not invite the loudest critics to the table, but a broader community less invested in the dispute.
Why it matters
A review body created by the party being reviewed always walks a narrow line. Here the line is drawn openly: the group speaks to results, not to pace. That is sensible from OpenAI’s side and awkward for the mathematicians who wanted the opposite. The arrangement’s value therefore rests on one thing — whether nine people will use their right to say uncomfortable things when the results turn out weaker than advertised.
What's next
- The results, including the Navier–Stokes solution, await independent verification — that is what will settle the claim about 100-plus problems
- The group can speak publicly without an invitation from OpenAI, so its stance at the next results announcement is the first real test
