DeepMind Solves Nine Erdős Problems, Passes Peer Review in Science
DeepMind used a system called AlphaProof Nexus to solve nine Erdős problems and published the result in Science. OpenAI, by contrast, posted its own proofs without any review.
Google DeepMind built a system called AlphaProof Nexus. It runs on the company's Gemini 3.1 Pro model. The system writes mathematical proofs in Lean, a formal language that automatically checks every step of a proof. AlphaProof Nexus solved nine of the 353 open Erdős problems, a long-standing list of unresolved math questions. Two of those nine had gone unsolved for 56 years. The results went through peer review and were published in the journal Science.
Around the same time, OpenAI, another AI lab, posted hundreds of its own math proofs on GitHub, a code-hosting site, without any review. Mathematicians led by Terence Tao, through the Association for Human Mathematics, called for a boycott of that approach. DeepMind's proofs, by contrast, went through standard scientific peer review. The gap is a reminder for anyone using AI-generated results. An answer that looks finished still needs a separate check before you can trust it.
Source: the-decoder.com
Free course
Stop reading about AI — start building with it
The free Claude Code course: your first site, tool or game — no coding. No upsells, no cross-sells — nothing to buy here.
Start free →
Author
Evgenii Arsentev
PhD · Founder & CEO, Dad4Kids
Articles · Latest articles