Mathematicians Hate AI. They Can’t Quit It
Powerful AI models have created an existential risk to the field, but researchers can’t stop relying on them because they’re too useful.
Mathematicians find themselves caught in the AI revolution, struggling to accept the company models that dominate their work. Even if they disagree, it's difficult to avoid using AI's usefulness, says NYU professor Timothy Buckmaster. Since accusing OpenAI of copying his approach, Buckmaster has continued using the company's coding agent Codex to refine his research papers.
OpenAI's investigation concluded that Buckmaster's Codex prompts couldn't have influenced the system in any way, including through training. However, the company still amended its announcement about solving Navier-Stokes, stating that the result "could not have been influenced" by his prompts.
Other mathematicians have raised similar concerns, such as Andreas Thom, whose techniques in geometric group theory were allegedly used by OpenAI's Astra model to prove a long-standing problem. Despite OpenAI's claim that "no progress" had been made on the issue in the last decade, Thom pointed out a 2019 paper and other mathematicians' work.
The lack of understanding behind AI's processes raises existential questions for mathematicians, who rely on attribution and building on each other's work. Cornell mathematician Alex Townsend has seen colleagues question their purpose in a field increasingly shaped by trillion-dollar companies. Some mathematicians are advocating for more oversight, while others, like Thom, feel uneasy about the possibility of their work being used without proper attribution.
Written by urgent.news from Wired's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
