Ten years after AlphaGo defeated world Go champion Lee Sae Dol, Demis Hassabis reflects on how the system's core techniques — deep neural networks, reinforcement learning, and tree search — have catalyzed breakthroughs far beyond games. AlphaFold solved the 50-year protein folding problem and is used by 3 million researchers; AlphaProof and Gemini's Deep Think achieved medal-level performance at the International Mathematical Olympiad; AlphaEvolve discovered novel matrix multiplication algorithms. Hassabis argues that combining Gemini's multimodal world models, AlphaGo's search and planning methods, and specialized AI tools is the roadmap toward AGI — a system capable of true creativity and fundamental scientific discovery.
Sort: