OpenAI forms independent mathematics advisory panel amid AI proof backlash

OpenAI Creates Independent Mathematics Advisory Panel

OpenAI announced on Monday a new independent Advisory Group on Mathematics and Artificial Intelligence to guide how the company publicly presents and releases AI-generated mathematical results.

The nine-member panel is hosted by the Institute for Advanced Study in Princeton, New Jersey, and includes many prominent mathematicians. Members will not be paid by OpenAI and can freely publish advice, criticize the company, and control their own membership.

OpenAI said the new group operates independently and may publicly offer unsolicited advice, though it will not directly influence the company's internal research pace or decisions.

The panel is meant to advise OpenAI and other AI companies effectively on interactions with the mathematical research community and norms. Its creation follows recent accusations that OpenAI scooped researchers, publicly failed to credit human work, and stole from its users.

Prominent named members also include Dr. Timothy Gowers of the College de France and Cambridge and Dr. Edward Witten of the renowned Institute for Advanced Study.

Backlash Follows AI Math Breakthroughs

OpenAI's unreleased internal AI model reportedly controversially solved the famous sought-after Navier-Stokes Millennium Prize problem in mathematics earlier this month. The same model has reportedly already resolved more than one hundred longstanding open problems across most major areas of mathematics.

Mathematicians say the dense, 166-page AI-generated Navier-Stokes manuscript produced by OpenAI is still hard to read and has yielded very little overall human understanding so far.

Dr. James Maynard at the University of Oxford said extracting human understanding from the proof has been very difficult so far. Dr. Javier Gomez-Serrano at Brown University said the AI-generated paper is currently clearly not written for ordinary humans at all.

OpenAI marshalled about 10,000 AI agents that worked for 88 hours using roughly 130 billion output tokens at a cost of about 6 to 10 million dollars.

Mathematicians around the world had worked toward the Navier-Stokes answer for many years and sensed they were finally getting close. Dr. Martin Hairer at the EPFL said everyone agreed that it would be the next Millennium Prize problem ultimately solved.

Criticism Centers on Elite Representation

Critics questioned whether just nine prominent researchers can still truly represent the wider mathematical community, with one mathematician calling the group an ivory tower in practice.

Dr. Francesco Fournier-Facio, an incoming mathematics professor at Heriot-Watt University in Scotland, said the group feels like an ivory tower. Dr. Simon Machado at ETH Zurich in Switzerland said the members' daily reality differs very greatly from most mathematicians' reality.

Dr. Martin Hairer wrote in a blog post that the group was formed after an approach from OpenAI but remains genuinely independent of all frontier AI labs.

All members signed no restrictive agreements beyond obvious confidentiality requirements, receive no financial compensation, and rely on the Institute for Advanced Study for daily technical support.

The Information reported Thursday that OpenAI employees expect to soon crack the famous Hodge Conjecture, also another Millennium Prize problem. The report said OpenAI reportedly seeks to ultimately dominate the field of mathematics as an important long-term pathway to recursive self-improvement.