DeepMind Reaches Key Milestone in Solving Mathematical Problems

DeepMind has marked a significant advancement in AI's capabilities by demonstrating its prowess in solving complex mathematical problems. On July 25, the London-based machine-learning company revealed that its AI system, AlphaProof, successfully tackled four out of six problems from this year’s International Mathematical Olympiad (IMO) held in Bath, UK. The AI provided rigorous, step-by-step proofs for these problems, which were evaluated by top mathematicians Joseph Myers and Fields Medal winner Tim Gowers. With a score of 28 out of 42, AlphaProof nearly reached the gold-medal range.

This achievement follows DeepMind's successes in other domains, such as mastering the game of Go and strategy board games. The company's latest accomplishment signifies a notable leap in AI's ability to handle complex mathematical proofs, an area previously dominated by human expertise. Problems from the IMO, a prestigious competition for young mathematicians, have become a crucial benchmark for measuring progress in AI's theorem-proving capabilities.

Pushmeet Kohli, DeepMind's Vice President of AI for Science, highlighted the significance of this milestone, stating, "This is the first time an AI system has achieved medal-level performance. It represents a key step forward in developing advanced theorem provers." As DeepMind and other organizations race to create AI that can generate proofs for substantial mathematical research questions, this accomplishment underscores AI’s growing potential in tackling some of the field’s most challenging problems.

SOURCE: https://www.nature.com/articles/d41586-024-02441-2

CREDITS: NATURE JOURNALS