Playing Stratego Games, Google AI Becomes One Of The Top Players In The World

DeepNashplays Stratego (photo: dock. Google DeepMind)

JAKARTA - Google DeepMind's latest artificial intelligence, DeepNash has learned how to beat almost all the human players in the war game, Stratego. Making him one of the top players in the world.

Stratego merupakan permainan dengan melibatkan dua pemain yang berusaha merebut bendera musuh. DeepNash dijadijar cara bermain Stratego, yang mencoba mencari tahu masing-masing dari 40 mata permainan lawan mereka yang berada di lokasi terhid di seluruh papan.

Due to Stratego's complexity, DeepNash took a modelless approach to their algorithms. Here, AI doesn't try to model its opponent's behavior exactly, a kind of blank slate to study.

The arrangement was very useful in the early stages of the game, when DeepNash knew very little about his opponent's piece, it made predictions very difficult.

DeepMind's team then used deep reinforcement learning to move DeepNash, with the aim of finding a balance of games. Such as strengthening learning that helps decide on the next best steps in every game step, while DeepNash provides an overall learning strategy.

To evaluate the system, the team also engineered a tutor using knowledge from the game to filter real mistakes that may not make sense in the real world.

According to DeepMind's team, Stratego is an imperfect game of information and requires each player to balance all possible outcomes when making decisions, making it a much more complex game than chess, Go ortens, which DeepMind AI has previously studied.

In terms of numbers, Go has ten powers of 360 possible status of the game, far more thanUCH or chess while Stratego has ten powers of 535.

DeepMind's team stated that DeepNash became very good at Stratego so that it reached the top three of all time among the human players on the world's largest online Stratego platform, Gravon.

"The level of the game DeepNash surprised me," said one author of a new paper on AI published in the journal Science, Vincent de Boer, who is also a former Stratego World Champion.

Launching Futurism, Tuesday, December 6, AI is developing an unpredictable strategy to ensure its human opponents continue to guess, which involves spreading bait to get them off its path.

In fact, AI learns how to bully its opponent by playing low-rank pieces as if they were much more valuable.

"I've never heard of an artificial Stratego player who is close to the level needed to win a game against experienced human players," said Boer.