‘I’m Really Terrified’: A Mathematician Grapples With AI’s Recent Breakthroughs

Mathematicians may no longer be able to compete without AI for breakthrough discoveries.

OpenAI claims its AI has solved a 90-year-old math problem, potentially winning a $1 million prize, though the solution requires independent verification. The announcement has been complicated by claims from mathematician Tristan Buckmaster, who alleges OpenAI preempted his own work, developed with Anthropic researcher Levent Alpöge, and attempted to influence credit attribution. This incident highlights the rapid advancements in AI's mathematical capabilities and the ensuing race among corporate labs for recognition ahead of potential IPOs. Concurrently, Anthropic announced its AI, Claude, proved thousands of theorems and formalized Fermat's Last Theorem. These developments, alongside OpenAI's previous breakthroughs on other long-standing problems, suggest AI is accelerating mathematical discovery but raising concerns among human mathematicians about their future role and the changing landscape of their field.

The rapid integration of AI into pure mathematics, exemplified by OpenAI's and Anthropic's recent achievements, signals a profound shift. It suggests that the ability to perform groundbreaking mathematical research may soon become dependent on AI tools, potentially marginalizing human mathematicians who lack access or proficiency. This democratization, while opening doors for broader participation, also challenges the traditional value placed on years of human expertise and intellectual struggle.

The controversy over credit attribution in the Navier-Stokes problem underscores the growing tension between AI-driven research and human contributions, particularly as companies race to demonstrate AI superiority for market advantage. This dynamic raises questions about intellectual property, the future funding of human mathematical research, and whether human mathematicians will transition from being primary discoverers to interpreters or curators of AI-generated findings.