In the long negotiation between human knowledge and the tools built to extend it, a fracture has appeared. OpenAI withdrew its $10,000 sponsorship from Caltech's Mathathon competition this week after prominent mathematicians accused AI firms of misrepresenting research, generating hollow mathematical work, and potentially appropriating years of human scholarship through training data. The dispute, sharpened by OpenAI's contested claim of solving the Navier-Stokes Problem, raises a question older than any algorithm: who owns the fruit of an idea, and what do we owe the minds that planted it?
OpenAI Pulls Mathathon Sponsorship Amid Mathematician Backlash Over AI 'Slop'
Where mathematicians used 'destructive,' OpenAI chose 'disruptive.'
Mark
So OpenAI just pulled out because mathematicians didn't like the competition?
Mimi
Not exactly. The mathematicians said the competition itself was built on a flawed premise—that AI should be used to solve math problems. They see that as the companies trying to rebrand what they do as research when it's really just pattern-matching on stolen work.
Luke
But wait—did OpenAI actually pull out because of the letters, or did they pull out and then frame it as recognizing disruption? The timeline matters. Did they announce the withdrawal before or after the second letter?
Mimi
The letters came first, then Roberts announced the withdrawal. So yes, it was a response.
Mark
What's the Navier-Stokes thing about? Why does that matter?
Mimi
It's a 150-year-old unsolved problem in mathematics. If OpenAI actually solved it, that would be huge. But the mathematicians are saying OpenAI didn't actually solve it—and worse, that they might have stolen the solution from researchers who were close to cracking it.
Luke
Might have. That's important. Do we know if they actually did? Or is that speculation?
Mimi
It's speculation from Buckmaster and Alpöge. They said OpenAI "somehow appropriated" their work via training data. But there's no proof yet.
Mark
So the real fight is about whether AI companies should be allowed to use mathematicians' published work to train their models?
Mimi
Partly. But it's also about whether using that training data to generate solutions counts as research or as something else entirely—something the mathematicians are calling misconduct.
Luke
And the Mathathon was supposed to be a way for OpenAI to show they could do this responsibly?
Mimi
Yes. The question was literally about responsible use. The irony is not lost on anyone.
Mark
Will the competition still happen?
Mimi
Yes. The organizers say they're finding other sponsors. But it's a smaller pool now, and the whole thing is tainted by the accusation.
Il Polso
- OpenAI's disputed claim of solving the Navier-Stokes Problem lit a fuse — two researchers publicly suggested the company may have absorbed their own unpublished breakthrough through training data.
- Mathematicians responded with coordinated force, publishing open letters signed by dozens of prominent figures accusing AI firms of 'slop mathematics' and outright research misconduct.
- The language of the conflict itself became a battleground: where mathematicians said 'destructive,' OpenAI's research lead carefully chose 'disruptive' — a word that reframes harm as progress.
- OpenAI pulled its $10,000-per-team sponsorship from the Mathathon, a competition explicitly designed to explore responsible AI collaboration with human mathematicians.
- Organizers say alternative funding is already in negotiation and the October 30th first round will proceed — but the institutional trust the event was built on has visibly cracked.
In the long negotiation between human knowledge and the tools built to extend it, a fracture has appeared. OpenAI withdrew its $10,000 sponsorship from Caltech's Mathathon competition this week after prominent mathematicians accused AI firms of misrepresenting research, generating hollow mathematical work, and potentially appropriating years of human scholarship through training data. The dispute, sharpened by OpenAI's contested claim of solving the Navier-Stokes Problem, raises a question older than any algorithm: who owns the fruit of an idea, and what do we owe the minds that planted it?
OpenAI announced Thursday it would withdraw its sponsorship from the Caltech Mathathon, a 40-hour competition in which 100 teams of mathematicians attempt to answer a single question: how can AI responsibly augment human mathematical understanding? The decision followed an open letter from current and former Caltech mathematicians accusing AI companies of running a campaign of scientific misinformation. Within days, 25 additional prominent mathematicians had signed a second letter amplifying the charge.
The mathematicians' core accusation was pointed: AI firms were producing what they called 'slop mathematics' — consuming existing research without generating genuine understanding, and effectively scooping human researchers on problems they had spent years pursuing. The criticism arrived alongside a flashpoint. Earlier that week, OpenAI had claimed to have solved the Navier-Stokes Problem, a longstanding puzzle about predicting fluid movement. Mathematicians immediately disputed the validity of the solution. NYU's Tristan Buckmaster and Anthropic's Levent Alpöge went further, speculating that OpenAI may have learned of their own approaching breakthrough and absorbed it through training data — an accusation that remained unresolved.
OpenAI research lead Dan Roberts framed the withdrawal with careful language, describing AI's advance in mathematics as 'disruptive' — a word that carries the glow of innovation rather than the weight of harm that 'destructive,' the mathematicians' chosen term, implies. The Mathathon had been structured as a genuine partnership: OpenAI and Anthropic had jointly pledged $2 million in computational credits. A spokesperson confirmed that talks with replacement funders were already underway and that the October 30th opening round would not be substantially affected.
What began as a sponsorship dispute has become something larger — a proxy conflict over what AI companies owe to the researchers whose work trained their systems, and whether speed and scale in mathematical discovery represent genuine progress or a sophisticated form of plagiarism. OpenAI has stepped back from the line the mathematicians drew. Whether that step is a reckoning or a pause is not yet clear.
OpenAI announced Thursday it would withdraw its sponsorship from the Caltech Mathathon, a competition where 100 teams of mathematicians will spend 40 hours and $20,000 in computational credits each trying to answer a single question: how can AI tools responsibly augment human mathematical understanding? The decision came days after a group of current and former Caltech mathematicians published an open letter accusing AI companies of running what they called a campaign of scientific misinformation, and warning that the Mathathon itself would prove destructive to the mathematical community.
The letter's core complaint was sharp and specific. AI firms, the mathematicians argued, were generating what they termed "slop mathematics"—work that consumed existing research without contributing genuine understanding, effectively scooping human researchers on solutions to problems they had spent years pursuing. This constituted research misconduct, they said. Within days, 25 more prominent mathematicians signed a second open letter amplifying the criticism.
The timing was not accidental. Earlier in the week, OpenAI had announced it had solved the Navier-Stokes Problem, a longstanding mathematical puzzle about predicting fluid movement. The claim immediately drew fire. Mathematicians disputed whether OpenAI's solution was valid, and two researchers—NYU mathematician Tristan Buckmaster and Levent Alpöge from Anthropic—openly speculated that OpenAI might have learned of their own approaching breakthrough and appropriated their work through training data. The accusation hung in the air without resolution.
Dan Roberts, an OpenAI research lead, framed the company's withdrawal in measured language. "We recognize that the rapid progress of AI in mathematics is disruptive," he tweeted. The word choice was telling. Where mathematicians used "destructive," OpenAI chose "disruptive"—a term that carries connotations of innovation rather than harm.
The Mathathon had been structured as a genuine collaboration between AI companies and the mathematical research community. OpenAI and Anthropic had jointly pledged $2 million in computational credits. OpenAI's share was $10,000 per team across the competition. A Mathathon spokesperson confirmed to Gizmodo that the organization was already in talks with other firms to replace that funding. "We do not anticipate that this will affect the event in any substantial way," the spokesperson said. The first round is scheduled to begin October 30th.
What had begun as a sponsorship of a mathematics competition had become a proxy war over what AI companies owe to the researchers whose work trained their systems, and whether speed and scale in mathematical discovery constitute genuine progress or merely a form of sophisticated plagiarism dressed in the language of innovation. The mathematicians had drawn a line. OpenAI had stepped back from it. Whether that retreat signals a broader reckoning or merely a tactical pause remains unclear.
Citazioni salienti
We recognize that the rapid progress of AI in mathematics is disruptive.— Dan Roberts, OpenAI research lead
We do not anticipate that this will affect the event in any substantial way.— Mathathon spokesperson
Vuoi la storia completa?
Leggi l'originale su Gizmodo ↗
Technology