Wednesday, March 9, 2016

Google software beat the world champion Chinese game “Go” – Il Sole 24 Ore

History of the article

Close

This article was published on March 9, 2016 at 12:07 hours.
the last change is the March 9, 2016 at 12:11 hours.

the computer may apprestarsi to win his final battle against the intelligence and human imagination. The Google software AlphaGo Wednesday defeated the world champion of the ancient Chinese game Go: This is the South Korean Lee Sedol, 33 years. Go is considered one of the most difficult board games in the world, given the huge amount of possible moves. Is so far believed that the man was favored over the machine, thanks to his imagination.

The AlphaGo software, developed by Google DeepMind on the mechanisms of the human brain model, has also demolished this limit. Google’s computers had already defeated 5 games to 0 the European go champion, but Lee Sedol was another opponent: professional player since he is 12 years old, has won 18 times the world championship. A Lee, who played from Seoul, are 4 other matches, up to Tuesday, March 15, to seek revenge. And to restore the primacy of imagination and human intelligence.

“I am very surprised because I did not think I would have lost the game – said the player in the flesh -. A mistake I made at the beginning lasted until the end. ” Lee added that the initial strategy AlphaGo was “excellent” and was amazed by an unconventional move of the software, that a human would not do. The bag up for grabs is one million dollars. If the computer wins, the prize will go to charity.

But how to play Go?

It is perhaps the oldest strategic board game for two players. It originated in China, where he played for at least 2500 years; It is very popular in East Asia, but has spread around the world in recent years. It has simple rules, but the complexity of the strategy depends on the fact that the game of square grid ( goban ) is composed of 18 boxes per side and the pieces are to be placed at the intersections of the grid (19×19). A platform, therefore, much more extensive classic checkers (8×8), which multiplies exponentially the possible game configurations: there are as many as the number starting with 208, followed by 168 zeros!

two players must alternately placing black and white pieces (called “stones”) on empty intersections of the board. The aim of the game is the control of a goban greater than that controlled by the opponent; for this purpose the players try to have their stones so that they can not be captured, at the same time carving out of territories that the opponent can not invade without being caught.

You can in fact capture a stone or a group of enemy stones by surrounding them completely with their stones, so that they have no adjacent free intersections. Arrange the stones close to each other makes it possible to strengthen each other and avoid capture; on the other hand, arrange them away from each other allows to create influence on the whole goban. Players try to satisfy both the offensive and defensive needs and choose among the priorities tactics and their strategic plans. The game ends when the players “pass” consecutively, indicating that neither of them can increase its territory or reduce his opponent’s.

Force Lee, then. Without forgetting that the software, however, has been created by human intelligence …



Permalink

LikeTweet

No comments:

Post a Comment