In what is being heralded as a breakthrough in artificial intelligence Google’s DeepMind computer program, AlphaGo, has mastered the game of Go, or 围棋 in Chinese, first invented in China over 3,000 years ago.
Since the 1990s, computers have been getting better at beating their human counterparts in board games like chess and checkers. 1996 was seen as a breakthrough year for Artificial Intelligence when the IBM computer program “Deep Blue” beat the reigning chess world champion Gary Kasparov.
Computer programs beating the world’s best chess players are now a common occurrence. But up until recently it was thought almost impossible for a computer program to truly master the ancient Asian game of Go and beat a champion human player.
Go was first invented in China over 3,000 years ago, with some evidence that it is even older dating the first game being played in 2,356 BC.
The game can look deceptively simple with two players using black and white stones on a grid to try and “capture” more territory than their opponent. But while a player in a typical chess game has about 20 moves to choose from each turn in Go that increases to 200 per turn.
Underlying the difficulty of building a computer game that can truly master the ancient game, the DeepMind programmers estimated that there are in fact more possible positions in Go than atoms in the universe.
For all these reasons it was previously believed to be impossible to build a computer program that could successfully take on the complexity of the game.
Normally when computers play board games they analyze every possible move, and its consequences, before deciding on which move to make. While this method has proved successful in chess the complexity of Go means this approach is simply impractical.
Instead Google’s team of researchers took the new approach of “teaching” the program how humans played, and then letting its learning software play game after game for practice.
Ultimately the program was able to beat the reigning European champion in a series of games 5-0. It’s the first time a computer has beaten a professional player in a complete game, the computer developers said.
DeepMind now intends to take on an even harder challenge with a match between AlphaGo and Lee Sedol – the world’s top Go player – lined up for March in Seoul.
Story by CCTV News