OpenAI assures that one of its artificial intelligence models has solved more than one hundred mathematical problems that had been in study for years. He advertisement It comes while the company tries to rebuild its relationship with researchers after the controversy over its proposed solution to the Navier-Stokes problemreports The Verge.
The work corresponds to an internal model whose training began on August 28 and which not yet available to the public. OpenAI has not revealed its name and describes it as a system significantly more capable than GPT-6 Astra. It is the same AI that was attributed to the Navier-Stokes solution announced on September 8.
The company has not published in its announcement the list of those hundred problems nor the corresponding proofs and it does not detail how many results have been reviewed by external specialists. He claims that they cover most areas of mathematics and that The speed of its advances has surprised its own researchers.
The Navier-Stokes problem seeks to know whether the equations that describe how water or air move can end up giving physically impossible results, such as infinite speed. That would not mean that a liquid could move like that, but rather that those equations would no longer serve to describe what happens. It is one of the Clay Institute’s seven Millennium Problems and a mathematical challenge with nearly ninety years of history.. OpenAI maintains that it has managed to solve the problem after deploying about 10,000 AI agents for 88 hours.
The publication sparked a dispute over authorship and the use of previous research by mathematicians Tristan Buckmaster and Levent Alpöge. Buckmaster reported pressure to exclude Alpöge from a joint publication for working on Anthropic and asked if OpenAI had accessed his private queries in the tool Codex. On September 10, the company assured that an internal investigation had ruled out that Buckmaster’s consultations in Codex during the two months prior to the announcement had influenced the system.
OpenAI’s response to criticism includes collaborate with an independent advisory group made up of nine mathematicians. Is called AGMAI and has the support of Princeton Institute for Advanced Study. Its members include Timothy Gowers and Martin Hairer and its first task will be to guide the publication of the results that the company says it has obtained.
The committee will advise on how to review findings and disseminate them while respecting academic standards. OpenAI wants to give a voice to the mathematics community and receive guidance on tools that help research and learn. Group members will not be paid by the company and will be able to publish their recommendations and criticize their decisions.
However, the creation of the group does not clear up questions about its ability to influence OpenAI’s decisions. The researchers consulted by The Verge They question their representativeness and how much the company will listen to their recommendations. Hairer acknowledges that Their role is advisory and the company can ignore their advice.