Mathematicians grapple with AI's growing role in research and attribution
Leading mathematicians acknowledge AI's utility in solving complex problems but warn that its unchecked use threatens traditional credit and research practices.
Mathematician Tristan Buckmaster, a professor at New York University, says OpenAI employed his earlier work to accelerate a solution to the Navier-Stokes existence and smoothness problem, a $1 million prize problem. Despite the dispute, Buckmaster continues to use OpenAI’s Codex and Anthropic’s Claude to refine his papers and explore the logical steps taken by the AI systems.
OpenAI has investigated the claim and amended its announcement, stating that Buckmaster’s prompts in the two months before the result could not have influenced the system’s training. The company cited an email response to WIRED confirming the clarification.
German mathematician Andreas Thom, whose research in geometric group theory was cited by OpenAI’s Astra model, expressed surprise at the claim and sought clarification from OpenAI researchers. He was told that his interactions with ChatGPT had not been used for training. Thom noted that AI challenges the ability to trace individual contributions in mathematical discoveries.
Other scholars, including Cornell mathematician Alex Townsend, describe a mixed response: AI enables achievements unattainable without it, yet raises questions about the role of human researchers. Townsend reports growing interest among colleagues in accessing advanced models and adjusting privacy settings to prevent their work from being incorporated into training data.
Critics within the mathematical community, such as a group of 25 Fields Medalists, have called for alignment between AI developers and researchers. Over 4,000 signatories of the Leiden Declaration have proposed guidelines to safeguard the discipline. More than 2,000 individuals affiliated with Caltech urged the suspension of an AI-focused math hackathon sponsored by Anthropic and OpenAI, though OpenAI later withdrew.
The consensus among many mathematicians is that while AI offers significant efficiency gains, reliance on the technology may become essential for early-career researchers, prompting calls for broader discussion on its long-term impact on the field.