DeepMind’s AI has shown that it can solve International Mathematical Olympiad (IMO) problems at the level of a silver medalist. Using the sophisticated AlphaProof and AlphaGeometry 2 systems, the AI was able to solve four of the six challenging IMO problems.
AlphaProof is a reinforcement learning-based system developed for formal mathematical reasoning. It uses the formal language Lean and combines a pre-trained language model with the AlphaZero algorithm, which is known for its success in games such as chess and Go. This method enables the AI to formally verify mathematical proofs and minimize errors.
AlphaGeometry 2 is an improved version of an AI for solving geometric problems. With a neuro-symbolic hybrid system, it uses a large data set and a faster symbolic engine to solve complex geometry problems. Remarkably, it has solved 83% of IMO’s historical geometry problems, significantly outperforming its predecessor.
The performance of AI at IMO is an important milestone in the development of AI and demonstrates its potential in the field of advanced mathematical reasoning. This breakthrough could pave the way for further advances in science and technology and provide new tools for mathematicians and researchers.
Sources:
