Mathematician Tristan Buckmaster continues using OpenAI’s models despite accusing the company of using his work to solve a legendary math problem with a $1 million bounty.
He has been using Codex to tidy up his research papers and understand the logical steps OpenAI’s agents might have taken to reach the final proof.
Buckmaster says he is 'stuck' with AI due to its usefulness and the companies' monopolistic control over the technology.
Buckmaster worked alongside Anthropic researcher Levent Alpöge on the Navier-Stokes existence and smoothness problem, which OpenAI solved using tens of thousands of agents. He claims the company only deployed these agents after learning the equation was close to being solved.
OpenAI later amended its announcement to say Buckmaster’s Codex prompts could not have influenced the system in any way, including through training.
Other mathematicians, like Andreas Thom, have raised similar concerns. Thom says OpenAI claimed its Astra model used his techniques to solve a long-standing problem he had been working on for two decades.
He questioned how OpenAI learned about his work and pointed out the company’s press release overlooked his 2019 paper. OpenAI later amended its statement, and Thom says he does not trust the company’s claims about the data’s origin.
"I set it aside," Thom recalls.
"I'm not so much interested in these political things; I want to work on mathematics." He admits he will never know whether his work actually fed the result, and he acknowledges that AI is making it impossible to trace who contributed what to mathematical breakthroughs.
This poses existential questions for mathematicians who see their field slipping from their understanding.
Cornell mathematician Alex Townsend says many of his colleagues are now asking about AI and how to access higher-powered models. He feels both excited and nervous about the technology’s impact on their work. Townsend and others are calling for ground rules on how to release results and properly attribute human contributions in AI-driven mathematics.
Buckmaster is calling for a detente in AI-driven mathematics while mathematicians and AI laboratories set some ground rules on how to release results, including getting their references right.
He plans to clean up the papers he published prematurely last week, calling them 'AI slop.' He is open to discussing this with Open, but he says they must be careful with collaboration.
Source: wired
