Technology and Science08 October, 20262 min read
Association for Human Mathematics Calls for Boycott After OpenAI Publishes 377 Math Findings
This article is an automated aggregation of coverage from news outlets. Our editorials offer in-depth analysis curated by the NewsCord team.
OpenAI released hundreds of AI-generated mathematical findings, with reports citing 377 or over 700 manuscripts. Association for Human Mathematics called for a boycott of OpenAI's outputs.
5 outlets told this the same way.
377 findings spark split
OpenAI published 377 mathematical findings spanning algebra, number theory, theoretical computer science, mathematical logic, and topology, and the release reopened debate over the role of artificial intelligence in frontier research. OpenAI’s announcement followed the company’s prior claim that it had deciphered the Navier–Stokes equation, a Millennium Problem for which a reward of one million dollars was offered for every solution.
The Association for Human Mathematics called for a boycott after OpenAI’s online publication of more than 700 AI-generated evidence manuscripts, and the group warned of a future where research is dictated by AI labs. Terence Tao, president of the Association for Human Mathematics, relayed the association’s statement on his blog and described a “Math 2.0” era where solving problems would no longer be the discipline’s main goal.
Standards, verification, and Lean
The Advisory Group on Mathematics and Artificial Intelligence, hosted by Princeton University’s Institute for Advanced Studies, said its recommendations included a first request “to stop testing advanced mathematical problems on proprietary models.” TechCrunch reported that only 10 of the 719 manuscripts included releases of the model’s chain of thought, and it said 42% of the proofs released by OpenAI had not undergone formalization.
Melanie Wood, a Harvard mathematician and board member, said in an email that “We want to create standards and practices so that mathematicians can understand the results published by AI laboratories and advance the field,” and she told TechCrunch that “there is not human understanding of them at the point of release, and now the work begins.” Dan Roberts, OpenAI’s head of research, argued that it was important to test internal models to develop better tools, and that demonstrations were a byproduct of those tests.
Consequences for the field
Tristan Buckmaster, a mathematician at New York University who worked on the Navier–Stokes problem solved by OpenAI, said it remained unclear whether mathematicians using AI models were unknowingly contributing information that allowed the machine to jump ahead to the final answer. Buckmaster added, “With so many results published at once, I don’t think they’ve done any kind of due diligence at all,” and he framed the concern around how people might take someone’s work and drive it to a conclusion.
The “lost in translation” paper raised questions about whether models can be trusted to formalize their own solutions without human involvement, and it concluded that “the NL proof by OpenAI and other autoformalised Lean proofs should not prima facie be trusted.” The Association for Human Mathematics said mathematicians did not request the work and argued that OpenAI ignored the advisory group’s premise that leading AI corporations should not test advanced mathematical problems on internal models.