Advisory Group on Mathematics and AI launches at IAS, nine mathematicians to advise OpenAI on releasing results its internal model produced
- The Advisory Group on Mathematics and Artificial Intelligence, hosted at the Institute for Advanced Study and online at agmai.org, says its current task is advising OpenAI on how to coordinate the release of a large number of significant math results the company reports its internal model produced.
- The group formed after OpenAI approached some of its members about an external advisory board, and the members agreed with OpenAI to build an independent group instead and invite others to join.
- Members are François Charles, Camillo De Lellis, Timothy Gowers, Martin Hairer, Nikhil Srivastava, Ulrike Tillmann, Ravi Vakil, Edward Witten, and Melanie Matchett Wood, and they take no payment for the work.
- The recommendations go public on the group's site, and the members state they have no decision-making power at any AI company, with responsibility for company decisions resting with the company.
- The group is collecting input from mathematicians through a form at agmai.org/input, and Tao notes in a bracketed aside that the guest post itself was converted from a different file format using AI.
Hacker News opinions
I've got a lot more respect for mathematicians after the last few weeks. They looked at what AI is good at and what it isn't, then came back with actual proposals for their field instead of either worshipping the machine or yelling at it to go away.
That's because their culture is built on proof. Every line has to hold up beyond doubt. You see the opposite in software engineering, where code quality signals are mostly subjective.
Careful with the 'deeply human' framing. There's an open letter circulating that warns joining events tied to Anthropic and OpenAI could plausibly hurt participants' reputations. Human doesn't always mean sensible.
As a mathematician I'm a bit disappointed in my colleagues here. It reads like sour grapes. Nobody will even entertain the possibility that they're wrong about math as we know it being gone.
I don't think the community is adapting as well as people claim. The 3D designers at my company handled the GPT-6 Astra shift better, and their situation is far more concrete than this.
Mathematicians reach for thinking before emotion, the same way any community reaches for the tools it has. Whether that produces better outcomes is still an open question.
It's not the whole math community. Plenty are reacting badly, same as any profession getting upended. This particular group just happens to accept the inevitable. Software engineers adapted way faster, and coding agent revenue shows it.
The real purpose here is PR. AI companies use math results to show machines beating humans, and OpenAI announced today they solved 100 open math problems without even naming them. Even a group like this is a band aid.
Chess and Go players already went through this mentally. Lee Sedol quit Go. That's the closest preview we have of what happens to a field when a machine does it better.
Those problems don't exist in a vacuum. Calling a discovery a publicity stunt is a weird way to frame it, and the people downstream don't care how the result was found.
That OpenAI announcement is about exactly this group. It exists to tell OpenAI whether and how to release those 100+ problems, since they don't want another blowback from mathematicians. Once the model ships, anyone could do it anyway.