AlphaProof, a product of DeepMind, demonstrated its aptitude on questions from this year’s Mathematical Olympiad, marking a significant advancement in the battle to create substantial proofs with artificial intelligence.
AlphaProof combines a language model with the technique of reinforcement learning, using the ‘AlphaZero’ engine that the company has used successfully to tackle games such as Go and some mathematical problems.
On July 25, DeepMind declared that four out of the six questions presented to school pupils at the 2024 International Mathematical Olympiad (IMO) had been resolved by its artificial intelligence (AI) systems. The AI generated rigorous, step-by-step proofs that were marked by two leading mathematicians and earned a score of 28 out of 42 — just one point shy of the gold medal range.
DeepMind claims to be on the verge of besting the world’s top school students at solving mathematics problems, having already beaten humans in a variety of simple and sophisticated games.
This is the first time any AI system has been able to achieve medal-level performance in the IMO.
