Wednesday, March 9, 2016

Artificial intelligence, computer Google beats world champion Go – Rai News



A new won AI challenge: in Seoul in South Korea, AlphaGo, a computer developed by Google DeepMind division beat the legendary player of Go, Lee Se-Dol. The prize was $ 1 million.

Share

Seoul Lee Se-Dol, world champion of the ancient Chinese game of Go has said “very surprised” by the AlphaGo capabilities, artificial intelligence developed by Google subsidiary based in London. His comment was made during the press conference which was held in Seoul after what has been called the “game of the century” between human and AI. “I never thought I would lose,” he continued Lee, who trusts in the victory in four matches scheduled.

AlphaGo, a program developed by the unit of Google DeepMind, he beat the legendary player in the first game. Lee retired after three and a half hours from the start of the challenge, whose prize was to be $ 1 million.
Founder DeepMind division of Google said to have a lot of respect for Lee If Dol-and for his incredible ability: “he brought AlphaGo extremes of its capabilities,” said the head of the development team, David Silver.

Most of the atoms in the universe
Go is a board game originated in China over 2500 years ago and very popular in East Asia. Played by over 40 million people worldwide, has simple rules: players take turns must place the white or black stones on a table, trying to capture the opponent’s stones or dominate the blanks to gain territory . But the simplicity of the rules hides a profound complexity of the game: the possible positions are higher than the number of atoms in the universe.

This complexity is what makes it very difficult to play Go to a computer and then challenge irresistible to artificial intelligence. Researchers use the games as a testing ground to invent flexible algorithms to solve problems, sometimes so similar to that of humans.

The computer learns to win
of AlphaGo development team is more than pleased with the results, as reported by the blog DeepMind. “We are thrilled to have won one of the great challenges of artificial intelligence. However, the most significant aspect for us is that AlphaGo is not just an ‘expert’ system built using traditional rules; instead uses general techniques of machine learning to figure out for themselves how to win a Go. Games are the perfect platform to develop and test the artificial intelligence algorithms quickly and efficiently. “

LikeTweet

No comments:

Post a Comment