Game playing has long been a subject of fascination within the AI field, and Brown CS Professor Amy Greenwald recently hosted a AAAI Panel, “Advancing AI By Playing Games”, with some of the most influential figures in this specialty. The speakers included Dr. Michael Bowling of the University of Alberta (leader of the research group that built the first poker playing AI to beat professional players), Dr. Murray Campbell of IBM (member of the team that developed Deep Blue, the first computer chess program to defeat a world champion), Garry Kasparov (world-renowned chess grandmaster), Dr. Hiroaki Kitano (CEO of Sony Computer Science Laboratories), and Dr. David Silver of University College London (leader of the AlphaGo project, the first computer Go program to defeat a world champion). These individuals are responsible for some of the most pivotal advancements in Game-Playing AI.
“They are incredibly inspiring people,” says Amy when asked about her experience hosting such a star-studded panel. “All of them are really deep thinkers.” This natural curiosity for learning is clearly evident in their work, research that has been crucial in driving the break-neck speed of discovery in AI. “They work on problems for which success is not guaranteed,” explains Amy. “It is a high risk/high reward enterprise.”
Why has there been such a fascination with game-playing AI in particular over the past several decades? “Well, we all love playing games,” laughs Amy, “but beyond that, being good at playing games and solving puzzles has always been a sign of intelligence.” She explains that success at game-playing serves as a benchmark of human intelligence; likewise, it provides a definitive way to measure just how “smart” an AI really is. “The speed of discovery is increasing at an astronomical rate. There are multitudes of people in countries like China and India, and these countries prioritize and emphasize computer technology.”
Even with such rapid development, game-playing AI has encountered a number of challenges that have yet to be overcome. “Most of the successes have been in two-player zero-sum games, where there is a winner and a loser. Games with more than two agents, and general-sum games –that have both competitive and cooperative elements– are very, very challenging,” explains Amy. Even with these obstacles, however, the future of AI is bright. “Dr. Kitano’s Robocup organization has set a goal of robots winning the world cup in soccer by 2050. The kind of teamwork required to play soccer well may translate well, for example, to disaster response, and there are many other real-world applications for game-playing AI.”
For more information, click the link that follows to contact Brown CS Communications Outreach Specialist Jesse Polhemus.