OpenAI Says Its AI Has Solved More Than 100 Open Maths Problems
OpenAI has launched a new independent Advisory Group on Mathematics and Artificial Intelligence, hosted at the Institute for Advanced Study, to integrate mathematicians' input into its research. This follows controversy over OpenAI's rapid math solutions and concerns raised by Fields Medalists regarding AI's impact on intellectual work. The group will offer advice on results and releases but cannot influence OpenAI's research pace.
OpenAI says a new internal AI model has resolved more than 100 long-standing open problems in mathematics, prompting the company to create an independent advisory group to help mathematicians engage with the rapid progress.
The model began training on August 28, according to OpenAI. The company says its work has covered most areas of mathematics, including the Navier–Stokes problem, one of the seven Millennium Prize Problems.
The announcement comes as AI's role in mathematical research is becoming harder for the field to ignore. OpenAI has already published a series of AI-generated mathematical results this year, while mathematicians have raised concerns about how quickly such work is being produced and how it should be evaluated.
OpenAI creates a mathematics advisory group
The new Advisory Group on Mathematics and Artificial Intelligence will be hosted by the Institute for Advanced Study in Princeton, New Jersey.
Its nine initial members include mathematicians who will advise on the importance of new results and how they should be communicated to the wider mathematical community. Members will serve without pay and will be able to offer advice publicly and make their own recommendations about membership.
The group will have independence in how it conducts its advisory work. It will not, however, control OpenAI's research programme or decide how quickly the company develops its mathematical systems.
That limit is significant. The group can question, assess and advise on OpenAI's work, but it cannot tell the company to slow down its research.
More than 100 problems, but the proofs still matter
OpenAI's claim covers more than 100 long-standing problems, but the company has not presented all of those results publicly.
That leaves an important part of the story with mathematicians who will have to examine the individual proofs and determine what they establish.
OpenAI has previously released ten mathematical results produced with an internal version of its models, including work in geometry, coding theory, group theory, cryptography and theoretical computer science. The company also released formal Lean certificates for those results.
Its Navier–Stokes work has now been published with both a written proof and a Lean formalisation, giving mathematicians material they can inspect rather than simply taking the company's claim at face value.
Why mathematicians are paying attention
The speed of these developments has created unease among some mathematicians.
Earlier this month, a group of Fields Medal-winning mathematicians signed an open letter raising concerns about AI companies increasingly competing with mathematicians to solve prominent problems.
The disagreement is not simply about whether AI can produce useful mathematics. It also concerns verification, attribution and the role of human researchers when an AI system generates a proof or mathematical idea.
OpenAI's new advisory group is meant to give mathematicians a stronger role in that conversation. But because it cannot control the pace of the company's internal research, its influence will depend largely on the weight its members carry within the wider mathematical community.
For now, OpenAI is making a much larger claim about what its models can do in mathematics. The next step is for mathematicians to examine the results closely and establish which of those solutions genuinely hold up.