“We are currently facing the very specific challenge of advising OpenAI on how to coordinate the release of a large number of significant results in mathematics that they report have been produced by their internal model,” the Advisory Group on Mathematics and Artificial Intelligence wrote in its launch statement.
OpenAI announced the group on Monday. OpenAI began training a new internal model on 28 August. In the same post, it said the model has resolved more than 100 long-standing open problems across most areas of mathematics. That is in addition to the Navier-Stokes result OpenAI announced earlier this month.
According to OpenAI, the pace of the model’s progress surprised its own mathematicians. It led to internal discussions on how to inform the community so the field can prepare and adapt. The announcement does not list the problems.
Nine mathematicians, hosted in Princeton
The Institute for Advanced Study in Princeton hosts the group, which also sits online at agmai.org. Its nine initial members are François Charles (ENS-PSL), Camillo De Lellis (IAS, GSSI), Timothy Gowers (Collège de France, Cambridge) and Martin Hairer (EPFL, Imperial College London). The others are Nikhil Srivastava (Berkeley, Simons Institute), Ulrike Tillmann (Oxford, INI), Ravi Vakil (Stanford), Edward Witten (IAS) and Melanie Matchett Wood (Harvard).
OpenAI approached some of the members about setting up an external advisory board, according to the group. In agreement with OpenAI, they created an independent group instead and invited the others to join.
What it will and won’t do
OpenAI said the group will help it assess the significance of new results. It will also advise on how to coordinate their release. Its remit covers academic and professional standards too, and how OpenAI’s tools can support mathematical research and learning.
The group will operate independently, according to OpenAI. It can offer advice the company has not requested, comment on OpenAI’s impact on mathematics and make its advice public. OpenAI does not pay its members, and the group can change its own membership.
There are limits on both sides. OpenAI wrote that the group “will not be responsible for advising us on how to pace our internal progress on mathematics”. The group said it has no decision-making power at any AI company, and that responsibility for decisions rests with the company. It said it will publish its recommendations, and it is asking mathematicians for input through a form on its website.
The letter that preceded it
OpenAI cited a recent open letter, “A Severe Misalignment of AI in Mathematics”. In it, mathematicians raised concerns about solving open problems as a benchmark for AI systems. TechCrunch reported that 25 Fields Medal winners signed that letter. De Lellis is the only member of the new group who did, according to TechCrunch.
Researchers told The Verge the group is a good first step. Many said they still had basic questions: what it will do, how much influence it will have, and whether OpenAI will listen. Some also questioned whether such a small group of prominent researchers can represent the wider mathematical community.