Abstract: Compared with chess-playing program ”Deep Blue”, supervised learning of policy networks，rollout policy，reinforcement learning of policy networks and reinforcement learning of policy networks of AlphaGo are studied. A Monte Carlo tree search(MCTS) algorithm guiding by the policy and value networks is analyzed. Based on AlphaGo's technological breakthroughs, potential applications of artificial intelligence(AI) in physics domain, information domain, cognition domain and social domain of war space are forecasted, and AI programs funded by Defense Advanced Research Projects Agency(DARPA) are analyzed. Finally，the revolutionary impacts of AI on military domain are studied based on the Observation, Orientation, Decision, Action(OODA) loop theory.