OpenAI recruits mathematicians as scientists warn about speed of AI progress
AI solves the most complex problems of our time (photo: Getty Images)
OpenAI has created an independent advisory group on mathematics and AI. The new body will operate at the Institute for Advanced Study in Princeton, according to OpenAI.
AI breakthrough: Why are scientists concerned?
Training of the new internal model began on August 28, and in a short period of time, the algorithm not only solved one of the Millennium Prize Problems, Navier–Stokes, but also tackled more than 100 longstanding open problems across most areas of mathematics.
The rapid pace of progress surprised even OpenAI’s own experts, prompting internal discussions about the safe and responsible publication of the results.
At the same time, the scientific community expressed concern.
In an open letter, 25 Fields Medalists emphasized the risks of using complex mathematical problems as ordinary benchmarks for AI.
The scientists warned that laboratories’ race to make high-profile announcements threatens traditional intellectual work and academic standards.
Composition and powers of the independent advisory group
In response to the criticism, OpenAI invited leading scientists to join the advisory group. Its initial members include nine prominent mathematicians: François Charles, Camillo De Lellis, Timothy Gowers, Martin Hairer, Nikhil Srivastava, Ulrike Tillmann, Ravi Vakil, Edward Witten and Melanie Matchett Wood.
The advisory group will operate independently. Its responsibilities will include:
Evaluation and communication — analyzing the scientific value of new AI results and coordinating their publication.
Academic standards — providing recommendations on professional ethics and the integration of AI tools into educational and research processes.
Independence of expression — members will serve without compensation, may publicly criticize AI’s impact on mathematics, and may provide advice without an official request.
Autonomy — the group will independently determine its membership and expand it at its own discretion.
Limits of influence over cutting-edge development
Despite its broad advisory powers, the group will have no direct influence over the dynamics of OpenAI’s internal development processes.
As clarified by the company and the Institute for Advanced Study, the group has not been tasked with regulating the pace of internal progress of AI models, and all responsibility for final decisions and the deployment of technologies remains with the company itself.
