Google DeepMind, a British-American AI research lab, has achieved a “silver-level standard” in advanced mathematics. Its AlphaProof and AlphaGeometry 2 systems solved four out of six problems from this year’s International Mathematical Olympiad (IMO).
The IMO is a prestigious global mathematics competition, known for its challenging problems. Achieving silver-level status in this competition highlights the advanced capabilities of these AI systems in solving complex reasoning and geometry problems.
AlphaProof, a reinforcement learning-based system, and AlphaGeometry 2, an enhanced version of a geometry-solving model, demonstrated remarkable performance.
AlphaProof combines elements from AlphaGo and AlphaZero and is trained using the formal programming language Lean. AlphaGeometry 2 is a neuro-symbolic hybrid system built on Google’s Gemini AI model, improving its geometry problem-solving skills.
The IMO, held annually since 1959, serves as a key benchmark for AI’s mathematical reasoning. Google’s recent achievement highlights its significant progress in AI research amid Alphabet’s $2.3 billion Q2 loss due to heavy investment in AI development.
In related news, OpenAI is advancing AI capabilities with its new project, “Strawberry,” which aims to enhance reasoning and deep internet research, as reported by Reuters on July 15. On July 25, OpenAI launched SearchGPT, an AI-powered search engine prototype.
Google DeepMind’s achievement of a “silver-level standard” in advanced mathematics underscores the impressive capabilities of its AI systems, AlphaProof and AlphaGeometry 2. This milestone demonstrates significant progress in AI’s ability to tackle complex reasoning and geometry challenges.
