Twenty-five Fields Medal winners signed an open letter criticizing AI laboratories for their approach to solving mathematical problems. The mathematicians argue that these companies prioritize speed over proper attribution and verification, potentially harming the tradition of open research.
Tristan Buckmaster, a professor at NYU, recently alleged that OpenAI pressured him to exclude a collaborator from Anthropic when crediting a mathematical solution. He also questioned whether OpenAI utilized his work with the Codex model to develop its own proof.
OpenAI withdrew its sponsorship of a CalTech mathematics event on Thursday following criticism from researchers at the institution. The company's recent proof remains unverified by the broader mathematical community.
The signatories warn that frontier AI labs may spend tens of millions of dollars to use large language models to outpace human researchers. This competitive dynamic threatens to incentivize secrecy and disrupt the established human transmission of knowledge.
This letter follows the Leiden Declaration published in June, which addressed the impact of LLM-generated proofs on academic work. Mathematicians emphasize that the value of their field lies in the intellectual structure that fosters new ideas and education.
