Google DeepMind has won the first of five games of Go to the champion of this strategy game of Chinese origin, which is a milestone for artificial intelligence.
Google DeepMind’s artificial intelligence program, AlphaGo software has beaten legendary South Korean Go player Lee Sedol. It is the first of five planned games, which constitute the greatest challenge between man and machine so far. In the same way as the duel between the world chess champion Gari Kasparov and the supercomputer Deep Blue, IBM, marked a milestone for the computing power of computers, these Go games are destined to be the equivalent for artificial intelligence.
Go is a board game of Chinese origin, which is among the oldest in the world. The goal is covering an area of the board larger than the opponent’s. For this, pieces are used, which are strategically positioned on the board. The rules are simple, but the possibilities are huge and varied.
The DeepMind program, AlphaGo, has won the first of five games that are being disputed in Seoul, the capital of South Korea. To this country belongs Lee Sedol, who has been world champion of this game for the last decade. Throughout history there has been only one other player who surpasses Sedol in international titles, also South Korean Lee Chang-ho. Over the next few days it is expected that another four games will take place, in which the final result of the duel will be decided.
The fact that an artificial intelligence software has won the best Go player, while waiting for the resolution of the rest of the games, is a qualitative leap compared to what had been achieved in other disciplines where man and machine have faced each other. Despite the simplicity of the Go rules, the game is a challenge of observation and intuition. We are not only talking about calculating possibilities but about a more human factor, such as intuition.
The player has to learn how the opponent acts in order to face him. And this is achieved by deepening in some fields of artificial intelligence, such as machine learning, which allows the algorithm to regulate its behavior based on the information it receives, and the artificial neural networks, which is intended to mimic the functioning of the nervous system of animals.
Reaching the level of professional Go players is difficult because requires advanced machine learning. In October 2015, the AlphaGo program won the European champion of this game and became the first software to win a professional Go player. But Lee Sedol is in another league, because he is at the top of this discipline. The South Korean player withdrew from the game with 28 minutes and 28 seconds left in a game that lasted almost three and a half hours.
The duel is very reminiscent of the one held by the chess player Gari Kasparov and Deep Blue. At that time the publicity that was given to the duel was even greater, although in this case Google has also taken care of spreading the event. The Mountain View company bought DeepMind, a British company specializing in artificial intelligence, in 2014. AlphaGo is one of the programs of the entity, which works with neural networks so that they are able to playing video games in a similar way as people do.
Images: vpickering