{"slug": "analysis-ai-is-changing-the-equation-for-mathematicians", "title": "Analysis: AI is changing the equation for mathematicians", "summary": "OpenAI and unaffiliated mathematicians announced in September 2026 that they used AI to solve one of the seven Millennium Prize Problems, each carrying a $1 million prize, according to an analysis of the field. OpenAI CEO Sam Altman then claimed the team had made \"substantial progress on another Millennium Prize Problem,\" and OpenAI said its in-house AI had since solved over 100 previously unsolved math problems that have not been released publicly. The developments drew criticism from 25 Fields medalists, who signed a statement at mathandai.org calling current use of AI in math \"a general threat to intellectual work.", "body_md": "In the past five years, AI has gone from being slightly dumber than an average grade-school math student to helping solve problems that befuddled some of the best mathematical minds in the world.\n\nIn 2021, preeminent AI company OpenAI [bragged](https://openai.com/index/solving-math-word-problems/) that its latest models were only slightly worse than grade-school students at solving grade-school math. By 2022, [it was solving](https://openai.com/index/formal-math/) competitive “Olympiad” problems given to the top high school mathematicians in the world.\n\nBy late 2025-2026, AI started to help solve some of [Erdős’ problems](https://mathweb.ucsd.edu/~erdosproblems/), a famous listing of hundreds of unsolved math problems. Beyond being a neat tool, this showed that AI could help make novel discoveries in math that humans alone had not been able to solve. \n\nStill, many in the math community saw these developments mostly as a fun, potentially useful tool for solving problems that humans just had not had the time to get to.\n\nBut in September 2026, a bomb dropped. OpenAI and unaffiliated mathematicians (who discovered what first or stole from whom is [still being hashed out](https://cybernews.com/ai-news/openai-navier-stokes-theft-accusations/)) announced that they had used AI to solve one of the Millennium Prize Problems — seven of the most important unsolved math problems, each of which had a $1 million prize attached. \n\nOnly one of these problems has been solved before, by a reclusive, eccentric Russian mathematician who [rejected the prize money](https://www.theguardian.com/world/2010/mar/23/grigory-perelman-rejects-1m-dollars) and retired shortly thereafter. This was not an out-of-the-way problem sitting on the shelf that had simply remained unsolved because of lack of attention; as a Millennium problem, it was one of the holy grails of mathematics. \n\nAnd then shortly afterward, OpenAI CEO Sam Altman [claimed that their team](https://www.nytimes.com/2026/09/10/science/tristan-buckmaster-openai-math-navier-stokes.html?unlocked_article_code=1.AFE.TK7P.Y7WvGPkfHaSk&smid=url-share) had made “substantial progress on another Millennium Prize Problem.”\n\n### From grade school to the holy grail\n\nThere is a specter looming here of superhuman mathematicians able to make progress on problems that have stymied humans for centuries.\n\nMichael Dorff, chair of the mathematics department at Brigham Young University, explained, “AI can piece together enormously complicated mathematical solutions by accessing a vast body of mathematical knowledge, synthesizing that knowledge to create potential solutions, and testing their validity to produce a final complete solution that perhaps no human could achieve in a lifetime.”\n\nThis power is making some mathematicians nervous. What is the point of an endeavor for humans if computers can beat them at it? Most prominently, a group of 25 [Fields medalists](https://www.mathunion.org/imu-awards/fields-medal) — the rough equivalent of a Nobel Prize for mathematicians — recently [signed a statement](https://mathandai.org) criticizing the current use of AI in math, arguing that it represents “a general threat to intellectual work.”\n\n[In OpenAI’s response](https://openai.com/index/advisory-group-on-mathematics-and-ai/) to these concerns, they mentioned that the in-house AI used to solve the Millennium problem had since solved *over 100* previously unsolved math problems, which they haven’t yet released to the public (as of this article going to press). \n\nOf course, this flood of results has to be verified by humans. In the past, proof of a solution like this was a peer-reviewed, human-by-human process. Today, more and more theories are being verified using automated theorem checkers that can tell researchers if the proof is correct as long as it is properly typed into the computer program’s language.\n\nStill, a lot of human hours are required to confirm that the AI-generated proofs are, in fact, valid. And the Clay Mathematics Institute, which manages the Millennium Prizes, has not yet declared the Navier-Stokes problem — the problem OpenAI claims to have solved — officially resolved.\n\nSo what are mathematicians to do?\n\n### AI and math have a bright future\n\nMany are more optimistic than others, seeing an opportunity more than a threat. While Pace Nielsen, an expert in computational number theory at Brigham Young University, concedes he’s not sure “how exactly AI will be used in the future,” he says, “One thing is clear: When used properly, these tools remove the need to do those parts that humans find boring or tiresome.”\n\n“Hopefully, we don’t start finding truth boring!” Nielsen told me. “We live in exciting times. AI is revolutionizing how mathematics is being done.”\n\nReferencing what he’s seen discussed in the broader conversation, the professor noted, “The economics of mathematics has been upended. Proofs, the lifeblood of mathematicians, were a scarce commodity; now they are easily produced by machines. If we use these tools wisely, great things can be accomplished.”\n\nDorff agrees. “AI is a tool — a very advanced and useful tool … If mathematicians were solely interested in the final solution to a problem, then maybe AI would rob the mathematician’s craft of its beauty. But there is more to mathematics than just the final solution to a problem.”\n\nHe continued, “The reasoning behind how the solution was developed, the understanding of the ideas needed to solve the problem, and the ways to apply that knowledge to a new problem or situation are essential parts of a mathematician’s role.”\n\nEarly data suggests that most mathematicians agree with Nielsen and Dorff’s perspective of AI as a tool. [One survey](https://www.scirp.net/journal/paperinformation?paperid=151492), fielded before the Millennium Problem was solved, found that 8 in 10 mathematicians anticipate “that AI will evolve into a collaborative partner rather than a replacement for human mathematicians.” This is about the same number admitting to regular AI usage, with 7 in 10 also reporting at least moderate gains in their productivity as mathematicians because of AI. \n\n### Lingering nervousness\n\nStill, Nielsen recognizes the valid nervousness some might have as AI surpasses human ability on some tasks. “In just six months, AI went from giving nonsensical comments to answering some of the deepest questions. This should make everyone a little nervous about what comes next, nonmathematicians included.”\n\nDorff remains optimistic. “Most people think of mathematics as a set of tools and procedures. While those are valuable, mathematics is much more than that. Mathematics is a way of reasoning and thinking. It fosters traits like persistence, creativity and clear thinking, which will always be beneficial and needed in society.”\n\nNervousness about AI’s impact on one’s job and other established elements of a career is obviously not limited to mathematicians. A [recent Gallup poll](https://news.gallup.com/poll/714368/workers-fear-job-losses-technology.aspx) shows that 27% of Americans fear that technology will make their jobs obsolete — about double what it was a decade ago. \n\nThat number rises to a little over 1 in 3 harboring these fears when we look at just the younger workers aged 18-44. Even though the recent gains in AI ability seem especially aimed at intellectual work (robots still have a long way to go before they’ll be doing our dishes and landscaping our yards), noncollege graduates are [even more scared](https://news.gallup.com/poll/714368/workers-fear-losing-jobs-technology.aspx?) about their jobs being outsourced to technology. \n\nOne thing seems clear: Mathematics, and many other fields, will never be the same after AI. The question is whether it will add to the beauty of the mathematician’s craft or detract from it.", "url": "https://wpnews.pro/news/analysis-ai-is-changing-the-equation-for-mathematicians", "canonical_source": "https://www.deseret.com/education/2026/09/22/is-ai-changing-math/", "published_at": "2026-09-22 22:44:22+00:00", "updated_at": "2026-09-22 23:25:14.549648+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-research", "large-language-models"], "entities": ["OpenAI", "Sam Altman", "Michael Dorff", "Brigham Young University", "Fields Medal", "Clay Mathematics Institute", "Millennium Prize Problems", "Erdős problems"], "alternates": {"html": "https://wpnews.pro/news/analysis-ai-is-changing-the-equation-for-mathematicians", "markdown": "https://wpnews.pro/news/analysis-ai-is-changing-the-equation-for-mathematicians.md", "text": "https://wpnews.pro/news/analysis-ai-is-changing-the-equation-for-mathematicians.txt", "jsonld": "https://wpnews.pro/news/analysis-ai-is-changing-the-equation-for-mathematicians.jsonld"}}