Abstract:
After encountering strong protests from nearly 800 mathematicians in the academic community who signed an open letter, the artificial intelligence giant OpenAI announced that it has officially withdrawn its sponsorship and computing support for the California Institute of Technology (Caltech) AI Mathematics Hackathon (Mathathon). This university competition, originally scheduled to be held on October 30, was accused of transforming rigorous pure mathematics research into a promotional show for commercial AI companies and generating a lot of untested academic noise. It triggered a rare storm of backlash in top academic circles, and also put co-sponsor Anthropic into a hot debate about whether to stay or stay.
The hackathon event called "Caltech Mathathon" was originally planned to be supported by two AI cutting-edge laboratories, OpenAI and Anthropic, with a total of approximately US$2 million in computing power points. The organizer aims to organize participating undergraduate students and young researchers to try to solve unsolved open frontier mathematical problems by inputting prompt words into the two companies' most advanced frontier large language models. However, after the activity plan was announced, it was immediately condemned by the academic community. Current and former mathematics scholars at Caltech took the lead in launching a public letter of protest, which quickly gained signatures from as many as 771 mathematicians around the world.
The open letter bluntly criticized that commercial AI companies are using extremely serious pure mathematics research as cheap "advertising and marketing opportunities", and warned that such hackathon activities characterized by violent stacking of computing power are "very likely to cause devastating damage to the mathematics community." The scholars pointed out in the letter that the so-called long proofs generated by large models are often full of "AI mathematical garbage (AI slop)" that appears to be compliant but is actually difficult to verify. This not only forces the academic community to spend a lot of valuable peer review energy to manually check for false results, but also creates a moral hazard for sponsors to use computing power to steal the real wisdom of outstanding undergraduates.
Faced with fierce resistance from the academic community, Dan Roberts, head of research at OpenAI, publicly confirmed the withdrawal of sponsorship and stated on social media that the company has deeply realized that the rapid progress of AI in the field of mathematics has had a severe impact. In the future, it hopes to have more in-depth dialogue and communication with the traditional mathematics community in terms of integrating this technology and conveying its social impact. As a co-sponsor, Anthropic has not yet responded positively to whether it will follow up on its withdrawal. Its silence has triggered further comparison and observation of the two laboratories' academic expansion positions.
It is worth noting that this sponsorship withdrawal comes during a sensitive window period when the credibility of commercial AI giants in the mathematics community is undergoing a severe test. Prior to this, OpenAI was involved in widespread controversy for claiming to have solved the existence and smoothness problems of the Navier-Stokes equation through tens of thousands of AI agents working together for 88 hours. At the same time, it was revealed that the agency was suspected of pressuring a New York University mathematician to remove co-authors with an Anthropic background from another important mathematics paper, triggering a fierce battle between academic ethics and the attribution of predatory results.
Academic observers point out that mathematics has always been the ultimate touchstone for measuring logical reasoning and formal rigor, and it is also the core battlefield for AI laboratories to compete for the right to speak on the capabilities of next-generation models. However, the group resistance encountered by the Caltech Hackathon clearly shows that model output driven by pure engineering computing power and public relations, if separated from the strict verification norms and mutual trust of peers of the traditional academic community, will ultimately be difficult to cross the deep gap of academic legitimacy. OpenAI's forced withdrawal is not only a public reputational setback, but also a wake-up call for the collaborative rules between the entire artificial intelligence industry and hard-core scientific research.
Comments