Scientists have created the world's strongest AI, defeating the best AI in chess

DeepMind Technologies, Alphabet's AI subsidiary - who has been working on AI for many years, has created a new artificial intelligence platform called AlphaZero. This AI takes advantage of deep neural networks to quickly learn to play Go, chess and Shogi flags without any human intervention.

DeepMind Technologies, Alphabet's AI subsidiary - who has been working on AI for many years, has created a new artificial intelligence platform called AlphaZero. This AI takes advantage of deep neural networks to quickly learn to play Go, chess and Shogi flags without any human intervention.

AlphaZero was brought out to compete with the world's best AI on chess games, including:

  1. Stockfish: AI world champion chess game.
  2. Elmo: Winner of the 27th World Computer Shogi Championship in 2017.
  3. AlphaGo Zero: AI is said to be the most powerful Go player in history, thanks to Deep Mind himself.

In all three competitions, AlphaZero is only provided with basic knowledge of the rules of the game. After that, he taught himself by playing with himself millions of matches. Initially, AlphaZero only tried many random tactics to win, then through a trial and error process called "reinforcement learning", it gradually became aware of which strategies were most effective.

In the course of training and training, AlphaZero uses 5,000 tensor processing units (TPU - 1 single unit can handle more than 100 million photos on Google Photos every day) for 9 hours with chess, 12 hours with Shogi flags, and 13 days with go flags.

After completing the learning process, AlphaZero is allowed to play against other opponents. And of course, AlphaZero quickly eliminated opponents to become the strongest player in history.

Scientists have created the world's strongest AI, defeating the best AI in chess Picture 1Scientists have created the world's strongest AI, defeating the best AI in chess Picture 1

The DeepMind team has equipped AlphaZero with a system combined by learning algorithms and the "Monte Carlo tree search" (MCTS), a method to help Go players to determine which steps to take next, To help him learn to play chess and Shogi flags. The results show that MCTS can be adapted into other complex games.

An interesting thing is that AlphaZero does not need a person to guide, but does self-study tactics and gives ideas that people have never seen before. AlphaZero has an aggressive playing style and a high degree of flexibility that surprised the chess player Matthew Sadler.

With unique tactics and abilities, AlphaZero becomes a great chess teaching tool, encouraging the creation of unique strategic gameplay unprecedented for those who want to learn chess.

See more:

  1. This is how AI defeats the best Dota 2 players
  2. Virtual broadcasters run on the world's first artificial intelligence, just like real people
  3. This is the first advertising in the world written entirely by artificial intelligence
5 ★ | 1 Vote