Artificial intelligence (A.I) has become an important part of game industry and in most cases it determinate the game design. The automation in game it is needed for the game to become more entertained for the player. The evolution of the games is often base on how efficient is the gameplay of the non-player characters (NPCs) that the player must defeat to complete the game. A.I is not used only in game but in military, medical, corporate and advertising applications. Game A.I is the effort of game industry to surpass the scripted gameplay of the games as far as the NPSs interaction how sophisticated is the environment and how complex the system of the game can become and how to make it as interactive as possible for player. System like that learn from user input, user inputs ca be certain action such as how the player fights what button it prefers to use or how the player it interacts with the environment as whole. These inputs help the algorithm to evolve and adapt his own non-prescript behavior and develop new techniques for the NPCs to interact with player. Another use of A.I in games is to create an NPC that learn from other NPC how to play, after some generations of training has the knowledge how to defeat specific NPCs. The goal of our research is the second example how to use artificial intelligent techniques to train an NPC to defeat another NPC in battle. In this paper we present a genetic algorithm (GA) for the generation of evolving artificial neural networks called NEAT and we use this algorithm in a game called street fighter. We will discuss how this algorithm can train an NPC to defeat other NPC in battle.