New “leap” in artificial intelligence from Google’s DeepMind

© naftemporiki.gr New “jump” in artificial intelligence from Google DeepMind

Google’s DeepMind announced that it had another significant leap in the field of artificial intelligence with a computer that learned to play the Chinese game “Go” without any help from human players. It is recalled that the AlphaGo program has already won two of the world’s top “Go” players, and has begun his education learning from thousands of games played by people players.

However, as the BBC says, the new program, AlphaGo Zero, started with a blank “Go” board and zero data except for the rules – and then started playing alone. Within 72 hours he was able to win the original program by 100-0 .

DeepMind Chief Executive, Demis Hassabis, said the system could now have a wider range of applications in the field of scientific research.

“We are fascinated because we believe it is now good enough to make significant progress on real problems, although we are still obviously far from full artificial intelligence”

told to the BBC and other reporters.

The company’s software had defeated South Korea’s Li Stcentol 4-1 last year, and in May he defeated the Chinese Khei. Note that this is a game where it is assumed that the possible moves exceed the number of people in the universe.

However, although AlphaGo took months to reach the level where a professional player could face, AlphaGo Zero did it within just three days, using a fraction of computing power. He also developed techniques that a professional player performing DeepMind’s advisory duties said he had never seen before.

Source: www.naftemporiki.gr