Google’s AI Lab Unveils AlphaGeometry, Solving Geometry Problems Near Human Performance
DeepMind, the UK-based artificial intelligence lab owned by Alphabet’s parent company, has developed an AI system called AlphaGeometry that can solve complex geometry problems close to the performance level of human Olympiad gold medalists.
Breaking Barriers in Mathematical Reasoning for AI
AlphaGeometry, which consists of a neural language model and a rule-bound deduction engine, has revolutionized synthetic data generation by creating one billion random diagrams of geometric objects. This process, known as “symbolic deduction and traceback,” resulted in a final training dataset of 100 million unique examples.
The Future of Artificial General Intelligence
Evan Chen, a maths coach and former Olympiad gold medalist, evaluated a sample of AlphaGeometry’s solutions and praised its output as clean, human-readable proofs using standard geometry techniques. This is in contrast to the messy numerical solutions often produced by AI systems when brute forcing mathematical problems.
AI’s Progress in Mathematics and Logic
While AlphaGeometry currently handles only the geometry portions of Olympiad tests, its skills could have earned a bronze medal on some past exams. DeepMind aims to continue improving its math reasoning abilities to the point where it can pass the entire multi-subject Olympiad.
Achieving Generalized Artificial Intelligence
By mastering Olympiad problems, DeepMind and Google believe they are one step closer to achieving more generalized artificial intelligence that can automatically discover new knowledge.
