Twenty-five Fields Medal winners have signed an open letter warning that the race to solve famous math problems with AI is threatening the integrity of the discipline.
Tristan Buckmaster, a professor at New York University, told the media that OpenAI pressured him to stop crediting a collaborator from Anthropic on a significant proof. He questioned whether the company used Codex to generate a ground-breaking solution over a weekend of inference.
On Thursday, OpenAI removed its sponsorship of a mathematics event at CalTech after researchers there criticised the company.
The letter argues that AI solutions are only useful if the community can understand and communicate them. The signatories say the current approach leaves no time for proper documentation or citing previous work. OpenAI’s proof remains unverified.
They wrote: “As in all creative professions, this raises severe attribution and plagiarism questions.” Without mathematicians willing to integrate these ideas into the canon, the human transmission chain is lost.
Fear is growing that the culture of open research is being threatened. If a frontier lab sees a path to a discovery, it can spend tens of millions of dollars using large language models to beat original researchers to a proof. This dynamic will incentivise secrecy.
This letter follows the Leiden Declaration, released by a working group of mathematicians in June. That document grappled with how LLM proofs will change their work and offered recommendations for institutions and policymakers.
As with software engineering, mathematicians find justification in the work around the work. The value in math is not just the proofs and credit, but the intellectual super-structure that nourishes students and finds new questions.
If you do not care about the cutthroat world of high-stakes mathematical proofs, remember: your field of interest is next.
The letter concludes: “The issues the mathematical community faces now are similar to issues that other scientific and creative professions are facing, and indicate issues that all of humanity might face: how to make sure that, as AI changes the way work is done, we do not lose sight of what that work was meant to achieve in the first place.”
What it means
Researchers face a new pressure to hide their methods. Labs can now use AI to produce results faster than humans can verify them, creating an environment where sharing data and credit becomes a liability.
