DeepNash artificial intelligence from DeepMind has mastered the game of Stratego and entered the circle of winners, lifting many champions-l. However, the computational abilities of AI are not used in practice by direct presentation – in the presence, a public meeting of DeepNash, a skill began to bluff, which was discovered even unexpectedly for its presence.
Stratego is an “incomplete information” game, as the opponents do not know the location of each other’s 40 game pieces. The value of an office can be obtained by spicy attacking him, at a cost. The problem is that if in such an exciting game as go accumulates 10 in 360 degrees of danger on the playing field, then in their strategic value 10 in 535 degrees.
Sets of Cheats He independently, without training in complex models, deduced winning principles and built on their basis representatives that allow you to outwit opponents. In particular, DeepNash passes off low-level chips as high-level chips by creating their own honeypots. At the moment I saw the top three players vs Stratego vs Watch.
Source: Tech Cult
