According to a recent report on the website of New Scientist, the new artificial intelligence "DeepNash" (Deep Nash) developed by the UK based company "Deep Thinking" has learned to use bluff and other deceptive means to defeat human opponents in the game "Strategy". In the 50 ranking matches with professional human players, "Deep Nash" won 84%, ranking among the top three. The relevant research was published in the latest issue of Science. Army chess is an extremely complex game with 10535 possible game scenarios. In contrast, chess and poker are much less. Julian Perrault of Deep Thinking said: "The most surprising thing is that 'Deep Nash' has learned to use weak pieces to bluff in the game and will sacrifice important pieces to win." The research team pointed out that "Deep Nash" overcame chess by fighting himself in 5.5 billion games. It does not try to play by searching all possible game scenarios, nor is it trained to compete with specific players. Instead, it uses an algorithm to continuously guide its behavior to the optimal strategy guided by economic game theory, which can ensure that its winning rate against the perfect opponent is at least 50%. Julian Toglius, an associate professor of New York University, said that under this strategy, although the opponent would hide some information, the game has a large number of possible game scenes, and each round can take many different means, "Deep Nash" can still make decisions to win, "this is something we can't really do before". "Deep Nash" has defeated human and artificial intelligence rivals. In the 50 ranking matches between the online game platform and professional human players, its winning rate was 84%, ranking among the top three. It also won 97% of the games against the top chess game robots of the Western Army, including several robots that won the World Computer Championships. Georges Yannakaski of the University of Malta said: "Good players tend to remember their opponents' pieces and predict their deployment patterns. 'Deep Nash' does well in both aspects: it may have a competitive advantage in memory, and it will make false claims and surprise." (Liu Xinshe)