AlphaGo and AlphaZero
AlphaGo was the DeepMind program that beat the world's best Go players, a milestone long thought decades away.
Overview
AlphaZero then mastered Go, chess, and shogi entirely through self-play, learning superhuman skill from scratch.
Deep Dive
Go has more possible board positions than atoms in the observable universe, making brute-force search hopeless and intuition essential. In 2016, AlphaGo defeated legendary champion Lee Sedol 4-1, with its famous 'Move 37' stunning experts as creatively non-human. AlphaGo learned from human expert games plus self-play. In 2017, AlphaZero went further: starting with only the rules and no human data, it taught itself by playing millions of games against itself, surpassing the best Go, chess, and shogi programs within hours to days. A later system, MuZero, even learned the rules of games on its own. These milestones showcased how reinforcement learning plus search can discover strategies beyond human knowledge.
Technical Insight
AlphaZero combines a deep neural network with Monte Carlo Tree Search (MCTS). The network outputs a policy (which moves look promising) and a value (who is likely winning), guiding the search to explore only the most relevant lines instead of every branch. Through self-play reinforcement learning, the network's predictions and the search results reinforce each other, steadily improving. No human games or hand-crafted evaluation functions are needed, just the rules and a reward for winning.
Strategic Impact
Vendor strategy
Vendor roadmaps influence what features your team can build next.
Cost and budget
Commercial terms and deployment options affect long-term cost and risk.
Risk and safety
Company incentives shape product defaults, safety posture, and openness.
The Future of AlphaGo and AlphaZero
The AlphaZero recipe, learning by self-play guided by search, now influences robotics, scientific discovery, and large-language-model reasoning, where models 'search' over solution steps. Descendants like MuZero and AlphaProof apply these ideas to planning without known rules and to mathematics. Expect self-play and tree search to keep powering systems that must plan, strategize, and discover novel solutions, increasingly fused with the reasoning techniques now appearing in frontier AI models.
Real-World Implementation
Defeating world Go champions Lee Sedol (2016) and Ke Jie (2017) in landmark matches
AlphaZero teaching itself superhuman chess in hours, revealing fresh opening and sacrifice ideas studied by grandmasters
MuZero mastering Go, chess, shogi, and Atari games without being told the rules
Inspiring self-play and search methods now used in robotics, math (AlphaProof), and LLM reasoning
Risks & Guardrails
Launch announcements may outpace stability in real production workflows.
API pricing or policy shifts can break assumptions overnight.
Single-vendor dependency increases lock-in and migration costs.
Implementation Roadmap
Evaluate providers using your own tasks and datasets.
Review privacy, security, and legal terms before integration.
Maintain a fallback plan across models or vendors.
Monitor release notes so roadmap changes do not surprise teams.
Keep Exploring
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the AlphaGo and AlphaZero quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Next guide
Zhipu GLM Models
Frequently asked questions
What is AlphaGo and AlphaZero?
AlphaGo was the DeepMind program that beat the world's best Go players, a milestone long thought decades away. AlphaZero then mastered Go, chess, and shogi entirely through self-play, learning superhuman skill from scratch.
Why was the game of Go considered especially hard for computers?
Go's enormous branching factor makes brute-force search infeasible, so success required learned intuition.
Who did AlphaGo famously defeat 4-1 in 2016?
AlphaGo beat top Go champion Lee Sedol 4-1 in a widely watched 2016 match.
How did AlphaZero learn to play, unlike AlphaGo?
AlphaZero started from just the rules and learned solely by playing against itself, without human game data.
Which search method does AlphaZero pair with its neural network?
AlphaZero uses Monte Carlo Tree Search guided by a neural network's policy and value predictions.
What two things does AlphaZero's neural network predict to guide its search?
The network outputs which moves look promising (policy) and an estimate of who will win (value), focusing the search.