From games to biology and beyond: 10 years of AlphaGo’s impact

2026-03-09 · Source: Google DeepMind News · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Data Science & Analytics · Depth: Advanced, extended

Summary

DeepMind's AlphaGo, an AI system, defeated world champion Go player Lee Sae Dol in March 2016, a decade earlier than experts predicted, marking the beginning of the modern AI era. AlphaGo utilized deep neural networks, advanced search, and reinforcement learning, learning from human games and self-play to master Go's 10^170 possible positions. Its "Move 37" demonstrated AI's capacity for novel strategies beyond human intuition. This breakthrough led to AlphaGo Zero, which learned from scratch, and AlphaZero, mastering multiple perfect information games. The underlying techniques have since been applied to scientific challenges, including solving the 50-year protein folding problem with AlphaFold 2, achieving a Nobel Prize in Chemistry for its creators. Further applications include mathematical reasoning with AlphaProof and Gemini's Deep Think mode, algorithm discovery with AlphaEvolve, and AI co-scientists for hypothesis generation, impacting fields from fusion energy to genome understanding.

Key takeaway

For AI scientists developing advanced systems, AlphaGo's legacy underscores the power of combining intuition and calculation to tackle vast search spaces. Focus on building systems that can go beyond human-trained data through self-learning and robust verification mechanisms. Your work should aim to identify and exploit "Move 37" moments in scientific domains, pushing the boundaries of what's currently understood to accelerate fundamental breakthroughs.

Key insights

AlphaGo's victory in Go catalyzed modern AI, demonstrating systems can surpass human intuition and accelerate scientific discovery.

Principles

• Combine deep neural networks with advanced search and reinforcement learning.
• AI can generate novel strategies beyond human expert knowledge.
• Self-play training can lead to superior performance over human-data-only training.

Method

AlphaGo learned plausible moves from human games, then played hundreds of thousands of games against itself, reinforcing winning strategies. It then considered only the most fruitful paths to find optimal moves.

In practice

• Apply AlphaGo's search and reasoning to complex combinatorial problems.
• Use AI to discover more efficient algorithms for fundamental operations like matrix multiplication.
• Integrate AI agents for scientific collaboration and hypothesis generation.

Topics

• AlphaGo
• Reinforcement Learning
• Protein Folding
• Artificial General Intelligence
• Algorithmic Discovery

Best for: AI Scientist, AI Researcher, Research Scientist, Machine Learning Engineer

Related on AIssential

• 10 years of AlphaGo: The turning point for AI | Thore Graepel & Pushmeet Kohli — AlphaGo's victory in Go demonstrated AI's ability to combine intuition and calculation, surpassing human expertise and fostering new scientific discoveries.
• 10 Years of AlphaGo: The Turning Point for AI | Thore Graepel & Pushmeet Kohli — AlphaGo's victory marked AI's transition beyond human-level intelligence, combining intuition and calculation to solve complex problems.
• Partnering with Ineffable Intelligence: A Superlearner for the Era of Experience — Superintelligence may emerge from AI systems learning purely through self-experience, unconstrained by human data.
• AI and the long game — AI's evolution, exemplified by AlphaGo's victory, shifts focus from human-machine competition to understanding intelligence through continuous scientific exploration.
• Move 37 — AlphaGo's Move 37 redefined AI's reasoning capabilities, surprising human experts.
• Google Drops Veo 3.1 Lite — AI systems can surpass human intuition and knowledge, driving scientific discovery in complex domains.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Google DeepMind News.