Twenty-five mathematicians who have won the Fields Medal, the discipline's highest honor, have signed an open letter arguing that artificial intelligence laboratories are threatening mathematical research by competing to solve famous unsolved problems without adequate time for verification or attribution. The letter follows accusations this week that OpenAI pressured a NYU professor not to credit a collaborator at Anthropic for work on a mathematical problem, and OpenAI's subsequent withdrawal of sponsorship from a CalTech math event after criticism from researchers there.
Twenty-five Fields Medalists warn AI labs are rushing to solve math problems without proper attribution
Twenty-five mathematicians who have won the Fields Medal, the discipline's highest honor, have signed an open letter arguing that artificial intelligence laboratories are threatening mathematical research by competing to solve famous unsolved problems without adequate time for verification or attribution. The letter follows accusations this week that OpenAI pressured a NYU professor not to credit a collaborator at Anthropic for work on a mathematical problem, and OpenAI's subsequent withdrawal of sponsorship from a CalTech math event after criticism from researchers there.
NYU professor Tristan Buckmaster accused OpenAI this week of pressuring him not to credit a collaborator who works for Anthropic for solving an important math problem. Buckmaster also questioned whether OpenAI had used collaborative work with Codex, an AI system, to produce its own proof of the Navier-Stokes equations, which OpenAI announced as a breakthrough discovery. The proof remains unverified.
The signatories argue that rapid announcements of AI-generated mathematical solutions leave insufficient time for proper documentation, the identification of novel methods, and the citation of prior work. They wrote: "Often these solutions are announced in a rush, leaving no time for a proper writeup, the isolation of new methods and ideas, and citing relevant previous work of others." They added that the practice "raises severe attribution and plagiarism questions."
The mathematicians contend that mathematical advances require human expertise to integrate new findings into established mathematical knowledge and to pass understanding to students and future researchers. Without this integration, they argue, AI-generated proofs would remain incomplete contributions to the discipline.
The letter also reflects concerns among mathematicians that AI companies may incorporate their own published work into proprietary models. Some researchers expressed worry that their use of Codex could have been fed into OpenAI's newer systems, creating an incentive for secrecy that could undermine the traditional culture of open research collaboration.
A related document, the Leiden Declaration, was released by a working group of mathematicians in June and addresses similar concerns about how large language model proofs will alter mathematical work, offering recommendations for mathematicians, institutions, and policymakers.
- The source does not provide the full text or detailed contents of the open letter signed by the twenty-five mathematicians.
- The source does not name the specific math problem that Buckmaster and his Anthropic-affiliated collaborator solved.
- The source does not state OpenAI's response to Buckmaster's accusation or its stated reasons for withdrawing CalTech sponsorship.
- The source does not specify which CalTech researchers criticized OpenAI or the precise nature of their criticism.
- The source does not describe what verification of OpenAI's Navier-Stokes proof would entail or when or whether such verification is expected.
- The source does not explain how researchers' Codex usage could have been incorporated into OpenAI's later models, or provide evidence that this transfer has occurred.
Read the original at TechCrunch