DeepMind's latest artificial intelligence, DeepNash, has learned how to beat almost all human players at a war game.
It's an example of how an artificial intelligence learns how to play a game and hones its skills to become a top-rated player.
The 40 game pieces that are hidden from sight are the identity of each player's game piece. It's different from other board games like Go and chess, which previous DeepMind artificial intelligences have learned to master.
Each player needs to balance all possible outcomes when making a decision, as DeepMind explains on its website, which is why it's called a game of imperfect information.
It's more complex than other games because of that. Go has ten to the power of several hundred possible game states, which is more than both poker and chess.
DeepNash has reached an all-time top-three ranking among human experts on the world's biggest online Stratego platform, according to DeepMind.
The artificial intelligence created an unpredictable strategy to make sure it was kept guessing. It was able to bluff its opponent by playing low-ranking pieces as if they were more valuable.
The coauthor of a new paper about the artificial intelligence published in the journal Science said he was surprised by the level of play. I had never heard of an artificial player that could win a match against a human player.
The classic game of imperfect information is called "mastering strago".
DeepMind scientists created artificial intelligence that can generate videos from one frame.
All Rights Reserved © 2024
