OpenAI has announced the formation of an independent Advisory Group focused on Mathematics and Artificial Intelligence. This initiative comes amid rapid advancements in the company’s internal AI models, which have demonstrated an unprecedented ability to solve complex mathematical problems.
Rapid AI Breakthroughs in Mathematics
The pace of progress in mathematical fields has surprised internal researchers at OpenAI. Following the training of a new internal model that began on August 28, the system has reportedly solved not only the Navier–Stokes Millennium Prize problem but has also resolved over 100 long-standing open problems across various mathematical disciplines. The company noted that the potential applications stemming from these novel discoveries are substantial, making the responsible development and deployment of such AI capabilities crucial beyond the purely mathematical realm.
Addressing Community Concerns
The move to create an advisory body was partly prompted by external criticism. In a recent open letter titled “A Severe Misalignment of AI in Mathematics,” professional mathematicians voiced concerns regarding the potential negative consequences of using the resolution of open problems merely as a metric for evaluating new AI systems. These critiques emphasized the necessity for AI technology companies to engage thoughtfully with the mathematical community.
To address these concerns and ensure responsible advancement, OpenAI is collaborating with a group of established mathematicians to form the advisory panel. This group is intended to act as a critical link between the broader mathematical community and the public, ensuring that mathematicians have a substantive voice in the company’s future direction.
Scope and Independence of the Advisory Group
The advisory panel will provide guidance on several key areas. Specifically, its function will include helping OpenAI evaluate the significance of emerging research results, advising on methods to coordinate the dissemination of findings, and setting professional standards for mathematical research. Furthermore, the group will advise on how AI tools can support both learning and formal mathematical investigation, aiming to place powerful tools directly in the hands of mathematicians.
Crucially, the group has been structured to operate with complete autonomy from OpenAI. Its members retain the freedom to offer advice that the company has not explicitly requested, critique OpenAI’s impact on the field, and make their recommendations public. The group’s value relies on its members exercising independent judgment and challenging the company’s own assumptions. Financially, the members will not receive payment from OpenAI, and the group retains the authority to adjust its membership as deemed appropriate. The company emphasized that the group will not be responsible for dictating the pace of OpenAI’s internal mathematical development.
Initial Membership Roster
The Advisory Group on Mathematics and Artificial Intelligence, hosted at the Institute for Advanced Study, has announced its initial members, including:
- François Charles (ENS-PSL)
- Camillo De Lellis (IAS, GSSI)
- Timothy Gowers (Collège de France, Cambridge)
- Martin Hairer (EPFL, Imperial College London)
- Nikhil Srivastava (Berkeley, Simons Institue)
- Ulrike Tillmann (Oxford, INI)
- Ravi Vakil (Stanford)
- Edward Witten (IAS)
- Melanie Matchett Wood (Harvard)
While this collaboration marks a significant initial step, OpenAI acknowledged that complex questions remain regarding how AI can truly enhance mathematical comprehension and how the resulting benefits can be distributed across the wider academic community. The company stated its commitment to keeping mathematicians at the core of shaping the answers.
