Home
OpenAI Launches Math and AI Advisory Group; Mysterious Model Solves 100 World-Class Problems in 24 Days

On September 22, OpenAI announced the formation of an "Advisory Group on Mathematics and Artificial Intelligence." Headquartered at the Institute for Advanced Study in Princeton, this independent body comprises nine mathematicians who operate separately from OpenAI.
More notably, OpenAI revealed that a mysterious model, training since August 28, has successfully resolved over 100 longstanding open problems across various mathematical disciplines, including the controversial Navier-Stokes Millennium Prize Problem.
Mathematicians express astonishment at the rapid progress, sparking internal debate on academic disclosure
OpenAI noted that the model's mathematical advancements have surprised its internal team, prompting discussions on the most appropriate method for informing the broader academic community.
Recognizing mathematics as a foundational science, OpenAI acknowledged that AI-driven discoveries could have widespread applications. Consequently, the company is adopting a cautious approach to developing and deploying these capabilities, currently exploring broader implementation strategies.
This follows an open letter by prominent mathematicians Terence Tao, Yu Deng, Peter Scholze, and others, who raised concerns about using the solution of open mathematical problems as a benchmark for new AI systems. OpenAI stated that these perspectives highlight the need for more thorough engagement with the mathematical community.
Advisory group features three Fields Medal winners and operates without compensation from OpenAI
OpenAI clarified that the advisory group functions independently, retaining the right to offer unsolicited advice, publicly comment on OpenAI's impact on mathematics, and independently adjust its membership. Members are unpaid, though the group does not influence the pace of OpenAI's internal mathematical research.
The founding members include François Charles (École Normale Supérieure, ENS-PSL), Camillo De Lellis (Institute for Advanced Study, Gran Sasso Science Institute), Timothy Gowers (Collège de France, University of Cambridge), Martin Hairer (École Polytechnique Fédérale de Lausanne, Imperial College London), Nikhil Srivastava (Berkeley, Simons Institute), Ulrike Tillmann (University of Oxford, INI), Ravi Vakil (Stanford University), Edward Witten (IAS), and Melanie Matchett Wood (Harvard University). Among them, Gowers, Hairer, and Witten are Fields Medalists; Camillo De Lellis is also a signatory of the aforementioned open letter.
The advisory group will advise on the review and dissemination of emerging results, assist OpenAI in assessing the significance of findings, and provide opinions on academic standards in mathematical research. Additionally, it will recommend how OpenAI's tools can better support mathematical research and education.
Related article
Anthropic Confirms Claude Account Hacked After Users Lose Access
Multiple Claude users have recently reported on social media that their account quotas were being depleted rapidly without any active usage. Anthropic’s investigation revealed that hackers gained unauthorized access to user accounts by stealing login
Senspeech X2.5 Twin Stars: First Million-Token Context on the Edge, Fully Trained with Domestic Computing Power
On September 1st, iFLYTEK’s wholly-owned subsidiary, Ciyuan Xinghuo, officially launched and open-sourced two edge-side general large models: Xinghuo X2.5-4B and Xinghuo X2.5-1.7B. These models are the first of their kind to natively support a contex
Meitun LongCat Launches LoHoSearch as BrowseComp Scores Drop Below 30%
Search agent capabilities have been largely defined by BrowseComp over the last year. Yet, this benchmark is losing its relevance — it drove model performance from 30% to 90% in just ten months, and its value is rapidly diminishing. On July 17, Meitu
Related Special Topic Recommendations
Comments (0)
0/500

On September 22, OpenAI announced the formation of an "Advisory Group on Mathematics and Artificial Intelligence." Headquartered at the Institute for Advanced Study in Princeton, this independent body comprises nine mathematicians who operate separately from OpenAI.
More notably, OpenAI revealed that a mysterious model, training since August 28, has successfully resolved over 100 longstanding open problems across various mathematical disciplines, including the controversial Navier-Stokes Millennium Prize Problem.
Mathematicians express astonishment at the rapid progress, sparking internal debate on academic disclosure
OpenAI noted that the model's mathematical advancements have surprised its internal team, prompting discussions on the most appropriate method for informing the broader academic community.
Recognizing mathematics as a foundational science, OpenAI acknowledged that AI-driven discoveries could have widespread applications. Consequently, the company is adopting a cautious approach to developing and deploying these capabilities, currently exploring broader implementation strategies.
This follows an open letter by prominent mathematicians Terence Tao, Yu Deng, Peter Scholze, and others, who raised concerns about using the solution of open mathematical problems as a benchmark for new AI systems. OpenAI stated that these perspectives highlight the need for more thorough engagement with the mathematical community.
Advisory group features three Fields Medal winners and operates without compensation from OpenAI
OpenAI clarified that the advisory group functions independently, retaining the right to offer unsolicited advice, publicly comment on OpenAI's impact on mathematics, and independently adjust its membership. Members are unpaid, though the group does not influence the pace of OpenAI's internal mathematical research.
The founding members include François Charles (École Normale Supérieure, ENS-PSL), Camillo De Lellis (Institute for Advanced Study, Gran Sasso Science Institute), Timothy Gowers (Collège de France, University of Cambridge), Martin Hairer (École Polytechnique Fédérale de Lausanne, Imperial College London), Nikhil Srivastava (Berkeley, Simons Institute), Ulrike Tillmann (University of Oxford, INI), Ravi Vakil (Stanford University), Edward Witten (IAS), and Melanie Matchett Wood (Harvard University). Among them, Gowers, Hairer, and Witten are Fields Medalists; Camillo De Lellis is also a signatory of the aforementioned open letter.
The advisory group will advise on the review and dissemination of emerging results, assist OpenAI in assessing the significance of findings, and provide opinions on academic standards in mathematical research. Additionally, it will recommend how OpenAI's tools can better support mathematical research and education.
Anthropic Confirms Claude Account Hacked After Users Lose Access
Multiple Claude users have recently reported on social media that their account quotas were being depleted rapidly without any active usage. Anthropic’s investigation revealed that hackers gained unauthorized access to user accounts by stealing login
Senspeech X2.5 Twin Stars: First Million-Token Context on the Edge, Fully Trained with Domestic Computing Power
On September 1st, iFLYTEK’s wholly-owned subsidiary, Ciyuan Xinghuo, officially launched and open-sourced two edge-side general large models: Xinghuo X2.5-4B and Xinghuo X2.5-1.7B. These models are the first of their kind to natively support a contex
Meitun LongCat Launches LoHoSearch as BrowseComp Scores Drop Below 30%
Search agent capabilities have been largely defined by BrowseComp over the last year. Yet, this benchmark is losing its relevance — it drove model performance from 30% to 90% in just ten months, and its value is rapidly diminishing. On July 17, Meitu











