# AlphaZero

> computer program playing chess, shogi, and Go

**Wikidata**: [Q44860007](https://www.wikidata.org/wiki/Q44860007)  
**Wikipedia**: [English](https://en.wikipedia.org/wiki/AlphaZero)  
**Source**: https://4ort.xyz/entity/alphazero

## Summary
AlphaZero is a computer program developed by Google DeepMind that specializes in playing chess, shogi, and Go at a world-class level. It is an artificial intelligence model that learned to master these games purely through self-play, without relying on human data or handcrafted rules. AlphaZero achieved superhuman performance in all three games by using a reinforcement learning algorithm.

## Key Facts
- **Developer**: Created by Google DeepMind, a UK-based artificial intelligence company.
- **Inception**: First introduced in 2018.
- **Games Played**: Demonstrated expertise in chess, shogi, and Go.
- **Aliases**: Known as Alpha Zero, Alpha0, Alpha 0, 阿尔法零 (Chinese), and Alpha零.
- **Technical Basis**: Functions as an artificial intelligence model and computer program.
- **Learning Method**: Trained solely through self-play using reinforcement learning.

## FAQs
### Q: Who developed AlphaZero?
A: AlphaZero was created by Google DeepMind, a leading artificial intelligence research organization.

### Q: What games can AlphaZero play?
A: AlphaZero specializes in chess, shogi (Japanese chess), and Go, achieving superhuman performance in all three.

### Q: How does AlphaZero differ from traditional AI game programs?
A: Unlike earlier AI systems that relied on human expertise or large datasets, AlphaZero learned strategies from scratch through self-play.

## Why It Matters
AlphaZero represents a landmark achievement in artificial intelligence research, demonstrating the power of reinforcement learning and neural networks to master complex games without human intervention. Its ability to surpass world-class engines like Stockfish (chess) and Elmo (shogi) highlighted the potential of self-learning systems. AlphaZero’s approach has influenced broader AI research, emphasizing the importance of generalization and minimal human input. Its success also reshaped strategic understanding in games like chess, as it discovered unconventional moves and tactics. As a result, AlphaZero is widely regarded as a milestone in the development of autonomous, adaptive AI systems.

## Notable For
- **Multi-Game Mastery**: First AI to achieve world-leading performance in chess, shogi, and Go simultaneously.
- **Self-Learning**: Learned strategies purely through self-play, without human data or domain-specific algorithms.
- **Defeated Specialized Engines**: Outperformed Stockfish (chess) and Elmo (shogi) in head-to-head matches.
- **Scientific Recognition**: Published in the journal *Science* in 2018, detailing its methodology and results.

## Body
### Development
AlphaZero was developed by Google DeepMind, a subsidiary of Google focused on artificial intelligence research. The program was first announced in 2018, building on earlier work from the same team, including AlphaGo, which defeated a human world champion in Go in 2016.

### Technical Approach
- **Reinforcement Learning**: AlphaZero used a neural network to evaluate positions and a search algorithm to plan moves. It improved through self-play, adjusting strategies based on game outcomes.
- **Neural Network Architecture**: The system relied on a single neural network to guide both move selection and position evaluation, demonstrating efficiency compared to domain-specific engines.

### Achievements
- **Chess**: Defeated Stockfish, a world-class chess engine, in a 100-game match (64 wins, 36 draws, 0 losses).
- **Shogi**: Outperformed Elmo, a leading shogi engine, winning 90% of games.
- **Go**: Demonstrated superhuman strength without relying on the specialized techniques used by its predecessor, AlphaGo.
- **Publication**: Its breakthroughs were documented in a 2018 *Science* paper titled "A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play."

## References

1. [Source](https://golden.com/wiki/AlphaZero-639PV3P)