A group of DeepMind scientists has developed an AI program that teaches you how to defeat humans in the three most complex board games thanks to an algorithm that competes without instructions.
The researchers introduced AlphaZero, a computer program capable of "not only reaching the superhuman domain in some of the most complex board games known, such as chess, shogi and go, but also teaching oneself to play them without knowledge" . above, except for the rules of each game. "
The system was created by experts from DeepMind, a British company owned by Alphabet, the parent company of Google.
The results published in the journal Science represent an important step towards the development of an AI of players who can learn and master any game quickly.
The novelty of AlphaZero is that it avoids the need for information derived from humans and instead is able to learn chess, shogi and go repeatedly playing against itself until dominating each of the games.
From scratch (hence his name) without help or examples of real players. On other occasions, the computer learned because it fed it with all the human knowledge, millions of pieces and real examples, and from there the silicon brain chose the best strategies.
This new system was able to overcome the most advanced AI programs specialized in these three games after a few hours of self-learning.
Researchers say they now have to resort to a new generation of games, such as multiplayer video games, to tackle the next set of challenges for artificial intelligence systems.