Title:
Artificial Intelligence in Computer Games: Not Only Strong, But Also Entertaining

Speaker:
IKEDA, Kokolo (JAIST)

Abstract:
Many researchers have applied Artificial Intelligence (AI) methods to games such as Chess, and have improved them through many games.  "Strong" computer players have been extensively researched, and they are now stronger than almost all human players in many two-player board games.  Then, recently "entertaining" or "educating" AIs in games have started to attract more attention, and two such researches are briefly introduced in this presentation.  Firstly, we show how to generate a computer player that acts not as a "strong opponent" but as a "trustable team-mate" in cooperative multi-player games. For such purpose, it is necessary for the AI to estimate the human player's preferences.  Secondly, we show how to generate a random sequence which "seems to be unbiased" from a human player's point of view.  It is found that some kind of bias can make common players believe that the sequence is unbiased.