Anthropic AI model makes progress on Riemann hypothesis
International Desk
An Anthropic employee and mathematician has claimed to have utilised the company's artificial intelligence chatbot to successfully prove the renowned Riemann hypothesis. Levent Alpöge stated on Friday, one day following the company's US stock market debut, that he spent approximately six weeks using the Claude chatbot to develop a comprehensive proof of the mathematical problem.
Advertisement
Should the findings be verified as accurate, the proof would rank among the most important in historical records. Alpöge characterised the achievement as "one dawn across science of something far greater". This announcement arrives as artificial intelligence organisations, including OpenAI, have recently reported numerous solutions to significant advanced mathematics problems that have remained unsolved by scholars for decades.
The proof regarding the Riemann hypothesis presented on Friday was generated using an internal, non-public iteration of the Claude chatbot. The mathematical challenge is nearly 170 years old and focuses on prime numbers; it is regarded as the most famous problem in the field due to its immense complexity. Named for Bernhard Riemann, the hypothesis suggests that prime numbers possess a specific order along the number line. Because prime numbers serve as the foundation for higher mathematics, any breakthroughs involving them are expected to have wide-ranging consequences.
Alpöge reported that he contacted Anthropic CEO Dario Amodei on Sunday to seek guidance on how to announce the discovery. "The precise timing is horrible," Alpöge remarked, noting the company's public stock listing. "If I could have, I'd have finished this all much earlier." He further added, "I am clearly unilaterally responsible for feeling I just could not live if I waited even longer. There is an obvious cynical view, to do with the timing, but it is false."
The full textual proof, which Alpöge submitted in Lean—a computer language utilised for verifying mathematical proofs—spans more than 250,000 pages. Prominent mathematician Terrance Tao noted on Friday that the verification of the proof remains a distant prospect. "I imagine the digestion process will take some time, though in this case at least we have a human author to communicate with to assist with this task," Tao stated. In contrast, no human was clearly involved in other major mathematical proofs produced by AI earlier this year.
Advertisement
In September, OpenAI alleged that staff using a non-public version of its AI software solved a portion of a difficult problem related to Navier-Stokes equations, which describe fluid motion, in 88 hours. This task involved more than 10,000 AI agents, which are computer-programmed bots that function with a degree of autonomy, as part of a training exercise. The Navier-Stokes proof released by OpenAI has not yet been confirmed as correct, and it sparked criticism from academic and mathematical organisations regarding the application of AI to solve such problems.
Anthropic has also asserted that it has made additional progress and found solutions in fields including mathematics and physics. A collective of mathematics scholars argued that AI companies highlighting significant solutions to demonstrate the sophistication of their commercial products is "detrimental to the science of mathematics". In a group letter, the scholars stated, "We are witnessing a general threat to intellectual work, with misalignment between the outcome of the use of AI and its initial purpose."
A month later, OpenAI announced what it claimed were solutions, proofs, or significant progress concerning 372 additional outstanding higher mathematics problems. It will likely require months or longer for these claims to be evaluated by experts to determine their validity.