AI Tool Solves Hundreds of Unsolved Math Problems, Igniting Debate Over Future of Mathematics

OpenAI has triggered a significant upheaval in the mathematics world by announcing that an unreleased artificial intelligence tool has solved hundreds of previously unsolved problems. The tech giant, known for ChatGPT, published 722 papers containing complete or partial solutions to 372 difficult mathematical challenges. Among these are two of the seven renowned Millennium Prize Problems, each associated with a $1 million reward for a correct solution. This development, already being called the 'mathpocalypse,' has raised concerns among mathematicians about the potential obsolescence of their field.

This advancement follows closely on the heels of OpenAI's recent release of a proposed proof for the Navier-Stokes equation, another Millennium Prize Problem. Experts have expressed astonishment at the rapid progress of AI, which has evolved from systems struggling with basic exams to demonstrating PhD-level expertise in mathematics within a mere two years.

However, the approach taken by OpenAI has drawn criticism. Many mathematicians are expressing outrage, with some warning that the company's actions could "destroy the mathematical community." The academic community is divided, with some mathematicians lauding the release as a pivotal moment in mathematical history, while others condemn OpenAI's methods.

Dr. Levent Alpöge, a mathematician at AI firm Anthropic, described the event as "obviously the most significant moment in mathematical history." Conversely, many experts are angered by what they perceive as an "unsustainable" approach from OpenAI. Instead of adhering to traditional peer review and journal publication, OpenAI uploaded the solutions directly to GitHub, a code repository.

These purported proofs will require months of scrutiny by mathematicians for verification, and initial assessments have already revealed significant errors. Just days after the release, OpenAI had to retract three papers due to basic mistakes and amend others containing flaws that invalidated their results. Dr. Melissa Lee of Monash University noted in The Conversation that this "suggests, at the least, a lack of sufficient vetting before publication." Some mathematicians have described the papers as poorly written and, in some cases, containing unintelligible content.

OpenAI stated in a blog post that the goal was to "push the frontier of human knowledge and enable further progress in mathematics." However, many mathematicians argue that the company has not meaningfully advanced the field. Terence Tao, a professor at UCLA and a highly regarded mathematician, commented on Mastodon that problems are being solved by AI prompters lacking interest in the broader mathematical field, who do not fully comprehend the AI's output. He further stated that "Solutions to open problems are now being harvested at large scale in an unsustainable fashion, leaving entire fields of mathematics much less fertile than when such problems were solved in the traditional 'Math 1.0' fashion."

The Association for Human Mathematics, representing over 800 mathematicians, issued a strong statement rejecting OpenAI's claims of advancing the subject. The group asserted that "Releasing over 700 files at once is not a demonstration of scholarship, but a demonstration of power."

OpenAI claims its release methods were developed in consultation with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study. However, this group has stated that they specifically advised against using proprietary models to solve unproven solutions and sharing the results without thorough explanation, declaring, "we do not endorse this practice, and we ask them to stop testing advanced mathematical problems on proprietary models."

Many mathematicians have shared personal accounts of their life's work being completed rapidly by AI. Professor Hugo Duminil-Copin of Université de Genève, a Fields Medal recipient, was among many who shared their experiences on the Proofs and Prompts forum. He expressed that while he anticipated AI surpassing human capabilities, the announcement's force was unexpected, with dozens of papers addressing his areas of research, leaving him unsure where to focus.

Professor Henry Wilton of the University of Cambridge remarked, "If OpenAI wanted to destroy the mathematical community, this would be a great way to go about it."