Artificial Intelligence

Anthropic mathematician Levent Alpöge claims Lean‑verified proof of the 167‑year‑old Riemann Hypothesis

· GlobeWire Daily Editorial Desk · Source: techcrunch.com

2 min read

Anthropic mathematician Levent Alpöge claims Lean‑verified proof of the 167‑year‑old Riemann Hypothesis
TechCrunch · resized and padded · Photo source

Summary

Levent Alpöge, a researcher at Anthropic, announced that he has produced a Lean‑compatible proof of the Riemann hypothesis, a problem first posed in 1859 and one of the seven Millennium Prize problems. The claim is still being examined by the mathematics community; early checks by several experts suggest the Lean code compiles, but a full human‑readable verification may take months.

Details

  • What was announced: Alpöge posted an essay describing a proof that can be compiled in the open‑source proof assistant Lean. The proof relies on deep automorphy‑lifting arguments, Galois representations, and several advanced results (Clozel’s purity, Deligne–Ribet, etc.).
  • Community reaction: A Zulip chat among mathematicians (including Chris Birkbeck, Andrew Sutherland, and others) showed rapid interest. About 20 participants reacted with surprise emojis, and several reported that the Lean file compiled without errors.
  • Human readability: Alpöge admits the current version is not yet understandable to most mathematicians. He plans to work with collaborators, notably NYU’s Tristan Buckmaster, to produce a conventional exposition.
  • Parallel human effort: While Alpöge’s AI‑assisted proof is being vetted, Stanford post‑doc Alessandro Masullo released a traditional, non‑AI proof of the General Riemann hypothesis. Its correctness also remains under peer review.
  • Implications: If the Lean proof holds, it would be the first major Millennium problem solved with substantial AI assistance, showcasing how proof assistants can formalize extremely complex arguments. It would not, however, replace the need for human insight or verification.
  • How to verify: Interested readers can download the Lean repository from the link provided in Alpöge’s essay, run the build locally, and compare the formal statements with the published mathematical literature. Independent verification by other proof‑assistant experts is expected in the coming weeks.

Practical takeaways

  • The claim is not yet confirmed as a mathematically accepted proof.
  • Researchers should treat the announcement as a pre‑publication result that may evolve.
  • The AI component (Anthropic’s Claude model) was used to explore argument fragments; the final Lean code was assembled by Alpöge.
  • The broader AI‑math community sees this as a milestone, but also a reminder that human‑level understanding remains essential for acceptance.

--- *All statements are attributed to the original TechCrunch article and to publicly posted comments; no new quotes or unpublished data have been added.*

Share this article

About the author

GlobeWire Daily Editorial Desk

More to read