OpenAI has assembled an advisory group of nine top mathematicians, including Fields Medalists Timothy Gowers and Edward Witten, to guide the release of AI-generated solutions to over 100 open math problems. The group will assess significance and coordinate dissemination, operating independently from OpenAI.
TL;DR
- OpenAI's new internal model has reportedly solved over 100 open math problems, surprising even its own mathematicians.
- The advisory group will help assess the significance of these results and guide their release, with no direct decision-making power.
- The group's formation follows concerns raised by mathematicians about using open problems as AI benchmarks.
What happened
OpenAI announced the formation of an Advisory Group on Mathematics and Artificial Intelligence on Monday. The group, hosted at Princeton's Institute for Advanced Study and online at agmai.org, includes nine mathematicians: François Charles, Camillo De Lellis, Timothy Gowers, Martin Hairer, Nikhil Srivastava, Ulrike Tillmann, Ravi Vakil, Edward Witten, and Melanie Matchett Wood. OpenAI approached some members to set up an external advisory board, but the group chose to operate independently instead.
The group's remit includes assessing the significance of new math results generated by OpenAI's internal model, advising on their release, and considering academic and professional standards. They can also offer unsolicited advice, comment on OpenAI's impact on mathematics, and make their advice public. OpenAI does not pay group members, and the group can change its own membership. However, the group will not advise on pacing OpenAI's internal progress in mathematics, and has no decision-making power at any AI company.
OpenAI began training a new internal model on August 28. The model has reportedly resolved more than 100 long-standing open problems across most areas of mathematics, in addition to the Navier-Stokes result announced earlier this month. The pace of the model's progress surprised OpenAI's own mathematicians, leading to internal discussions on how to inform the community.
Why it matters
This development highlights the potential of AI to accelerate progress in mathematics, a field often considered resistant to automation. The advisory group's work could set a precedent for how AI-generated results are vetted and released in other scientific disciplines.
For AI developers and startups, this news underscores the importance of interdisciplinary collaboration. As AI systems become more capable, engaging with domain experts will be crucial for validating and contextualizing results. The group's independent status may also serve as a model for other companies seeking to build trust in their AI systems' outputs.
Investors should note the competitive implications. OpenAI's mathematical breakthroughs, if verified, could demonstrate the broad capabilities of its models. However, the formation of the advisory group also reflects the challenges of integrating AI into established fields. The group's ability to guide OpenAI's engagement with the math community may influence the company's reputation and partnerships.
Key facts
- The advisory group includes nine mathematicians: François Charles, Camillo De Lellis, Timothy Gowers, Martin Hairer, Nikhil Srivastava, Ulrike Tillmann, Ravi Vakil, Edward Witten, and Melanie Matchett Wood.
- OpenAI's internal model has reportedly solved over 100 open math problems, in addition to the previously announced Navier-Stokes result.
- The group is hosted at Princeton's Institute for Advanced Study and online at agmai.org.
- OpenAI does not pay group members, and the group can change its own membership.
- The group will not advise on pacing OpenAI's internal progress in mathematics and has no decision-making power at any AI company.
- The group's formation follows an open letter signed by 25 Fields Medal winners, raising concerns about using open problems as AI benchmarks.
- Only one group member, Camillo De Lellis, signed the open letter.
- The group is asking mathematicians for input through a form on its website.
Context
This development comes as AI systems are increasingly being applied to scientific research. While AI has shown promise in fields like drug discovery and materials science, mathematics has remained a challenging domain due to its abstract nature and reliance on rigorous proof.
The formation of the advisory group also reflects growing concerns about the responsible development and deployment of AI. As AI systems become more capable, engaging with domain experts and the broader community will be crucial for building trust and ensuring that AI's benefits are widely shared.
OpenAI's mathematical breakthroughs, if verified, could have implications for the broader AI industry. Demonstrating the ability to solve long-standing problems in mathematics could attract investment and partnerships, while also raising questions about the future of work in AI-driven research.