How this Google's AI has become invulnerable

The Google laboratory, DeepMind, has given life to a new form of artificial intelligence, which is not just a simple evolution of the old versions, but simply the new digital God.

Google's AI  has become invulnerable

We know since 1997 that a machine (Deep Blue) can beat the human reference (Garry Kasparov) in a game of chess. Since then, the artificial intelligence (AI) succeed and become ever faster and more intelligent. So, whether it's a chess game, Go or Shogi - that reference in the field of the complex strategies of our "human" point of view - we only have little chance in front of "Godlike", a powerful neural network used by these new generations AI, who terrace all the human champions who dare to face it.

But since a few days, all hope is vain, we have absolutely no chance. AlphaZero, the great new AI straight from the laboratories of DeepMind (owned by Google) learns a new way, and the results are... scary! Older versions of AI are growing because of how "tutoring". For example, for the game of chess, are given AI rules, and many, many examples of the best parts of chess to the world. AI learns then in a few days to become as good as the best players of chess worldwide. Learning by example (method "tutoring"), AI reached the perfection of the example, the human.

Alphazero  had learned the rules

But not AlphaZero, who learns without a frame, i.e. in a way no tutoring. If we take the example of the game of chess, Alphazero is given the rules, and that's all. Alphazero must then play against itself, to learn alone in his corner. Does this method work? Stockfish, the best ever chess, a monster power program, got trouncing by Alphazero who had learned the rules... just 4 hours earlier. And it only took him only 2 hours to be unbeatable in Shogi, and 8 hours to defeat AlphaGo Zero, his older brother, who was until then the reference in terms of AI.

"We now know who is our new Lord," said the researcher of chess David Kramaley, the CEO of the Chess table site. "It's like an alien civilization who invents his own math," said, in turn, the computer scientist Nick Hynes at MIT. "This will definitely revolutionize the game, but think about how it could be applied outside of chess. This algorithm could manage cities, continents, worlds". Indeed, AlphaZero was not designed to play logic games, but to learn alone to solve any combinatorial problem. Anyone. Is Pandora's box open? Deus Ex Machina.

Did you like this article? Then share it with your friends by clicking on the buttons below:

December 08, 2017