| Title | Multigame Playing – A Challenge for Computational Intelligence |
| Speaker | Prof. Jacek Mańdziuk |
| Chair | Zexuan Zhu |
Abstract
Recent advances in artificial systems regarding many popular board games have posed a spectacular challenge to human supremacy in chess, checkers, Othello, or backgammon, and raise the question – “Quo vadis board game research?”, or alternatively “What may be the reason for further pursuing Artificial/Computational Intelligence (AI/CI) research in this area?”
In the talk, I will emphasize the idea of returning to the roots of AI/CI applicability to board games, which relied on attempting to imitate the human way of playing (or, in a broader perspective, to mimic human intelligence). In particular, the talk will present an argument for the potential benefits of developing cognitively-plausible multigame playing systems, i.e. agents being able to autonomously learn how to play any game within an arbitrarily defined genre at a satisfactory level, as long as the rules of the game are provided. An efficient implementation of a multigame playing paradigm requires designing general-purpose learning and reasoning methods that abstract from the specificity of particular games. Such methods are relatively easily developed and mastered by human players, however, they are still far from being efficiently applied by machines.
The talk will provide an overview of previous AI/CI approaches to multigame playing and subsequently focus on the research framework known as General Game Playing (GGP), recently proposed at Stanford University. GGP encompasses the genre of finite, discrete, synchronous, multi-player, perfect-information games and includes many of the most popular board games, such as chess, checkers, Othello or Go, yet extending far beyond this selection. In GGP, game rules are provided to the players in the form of Prolog-like description. GGP poses several challenges to AI/CI, and addressing them requires the development of innovative solutions in various research fields. The list of core issues includes example-based learning and generalization (with the autonomous building of an evaluation function being one of the subgoals), efficient knowledge transfer between games, opponent modeling and automated reasoning in a multi-agent environment, evolution/coevolution of knowledge, or development of universally-applicable heuristic search methods.
Some of these tasks, especially those related to knowledge discovery and the autonomous construction of the evaluation function, are well suited to Memetic Computing methods on the condition that there is sufficient time allocated to initial game analysis. Another challenging research avenue in which Memetic Computing has potential, is the development of general-purpose adaptive meta-heuristic search methods, which would gear their search policy towards game-specific goals, autonomously inferred from game’s definition.
The above issues will be addressed in the talk from both theoretical and practical perspective.
Biography
Jacek Mańdziuk, Ph.D., D.Sc., is an Associate Professor at the Warsaw University of Technology. He gained an M.Sc. (Honors) and Ph.D. in Applied Mathematics from the Warsaw University of Technology, Poland in 1989 and 1993 respectively., and a D.Sc. degree in Computer Science from the Polish Academy of Sciences in 2000. In 2011 he was awarded the title of Full Professor (Professor Titular) by the President of the Republic of Poland. His research interests include the application of Computational Intelligence methods to game playing, bioinformatics, financial modeling, time series prediction and development of constructive learning methods for artificial agents.
He is the author of two books (including Knowledge-free and Learning-based Methods in Intelligent Game Playing, Springer 2010), and co-author of one textbook and over 80 refereed papers. He has served as a Program Committee Member at over 60 international conferences, was a Program Co-Chair of the International Workshop on Adaptive Systems in Soft Computing and Life Sciences, USA, in 2002 and a panelist on the Computational Intelligence and Games panel at the IEEE World Congress on Computational Intelligence (WCCI’08) in Hong Kong. He acts as a reviewer for several science journals dedicated to Computational and Artificial Intelligence. He is an Associate Editor of the IEEE Transactions on Computational Intelligence and AI in Games, an Editorial Board Member for the International Journal On Advances in Intelligent Systems, a member of IEEE (M’05, SM’10) and a member of the Games Technical Committee of the IEEE Computational Intelligence Society (CIS). He has been a founding chair of the IEEE CIG Task Force on Neural Networks for Games (since 2008) and a founding chair of the IEEE CIS Emergent Technology Technical Committee Task Force on Towards Human-like Intelligence (since 2011). He is a member of the Lifeboat Foundation Robotics/AI Advisory Board.
He has received a First Degree Award for Excellence in Research (6 times) and a Second Degree Award for Excellence in Research (4 times) from the Warsaw University of Technology. He was a recipient of the Senior Fulbright Research Award and has visited the University of California at Berkeley and the International Computer Science Institute in Berkeley (USA). Recently he has been a visiting professor at Yonsei University (South Korea), Nanyang Technological University (Singapore) and the University of Alberta (Canada).
