Research Programs Based on Machine Intelligence Games

Abstract

Games have played a significant role throughout the history of artificial intelligence and robotics. Machine intelligence games are examined here from a methodological perspective, focusing on their role as generators of research programs. These research programs are schematized in terms of framework building, subgoaling, and outcome appraisal processes. The latter process is found to involve a rather intricate system of rewards and penalties, which take into account the double allegiance of participating scientists, trading and sharing interchanges which occur in multidisciplinary research environments, in addition to expected industrial payoffs, and a variety of other research benefits in the way of outreach and dissemination of results, recruitment of junior researchers and students’ enrolment. Examples used here to illustrate these various aspects of the outcome appraisal process include RoboCup and computer chess, Go, Poker and video-games. On the whole, a reflection on research programs that are based on machine game playing opens a window on central features of the complex systems of rewards and penalties that come into play to appraise machine intelligence investigations.