Google’s AI division DeepMind has unveiled two advanced AI systems, AlphaProof and AlphaGeometry 2, demonstrating significant progress in solving complex mathematical problems, a challenging area for AI, Reuters reports.
These systems solved four out of six questions at the 2024 International Math Olympiad, marking the best performance by an AI system in this competition so far. AlphaProof, which combines Google’s language model Gemini with AlphaZero, solved three problems, including the hardest question, which only five out of over 600 human contestants solved. AlphaGeometry 2 solved one additional problem.
These advances highlight improvements in AI reasoning capabilities, a critical area where previous models have struggled. Meanwhile, Microsoft-backed OpenAI is also developing similar reasoning technology under the project name “Strawberry.”





