- Dhaka
Google DeepMind has developed two AI systems that together solved four out of six problems from the 2024 International Mathematical Olympiad (IMO), matching the performance level of a silver medalist. This is the first time an AI system has achieved such a result in the world’s most prestigious high school math competition. The two systems—AlphaProof, focused on algebra and number theory, and AlphaGeometry 2, designed for geometry—tackled problems known for requiring deep human reasoning. AlphaProof successfully solved three problems, including the toughest one that only five human participants got right, while AlphaGeometry 2 completed a geometry problem in just 19 seconds.
Their combined score was 28 out of 42 points, which is equivalent to this year’s silver medal cutoff. Esteemed mathematician and Fields Medal winner Timothy Gowers praised the AI’s performance, calling it beyond what he thought was currently possible. AlphaProof uses a mix of language models and game-style AI to create and check math proofs, while AlphaGeometry 2 improves on earlier models using symbolic reasoning and large synthetic datasets.
Despite the success, the AI systems still struggled with two combinatorics problems, showing they haven’t yet reached full human-level ability. Google DeepMind views math competitions as a strong benchmark for testing AI reasoning skills and plans to release more details about the project while continuing to improve their math-solving AI tools.
Write your opinion