Keynote Lecturer: Julian Togelius
Artificial intelligence and games go way back. At least to Turing, who re-invented the Minimax algorithm to play Chess even before he had a computer, and to Samuel, who invented a predecessor of TD-learning in order to build a Checkers-playing program in the 1950s. Games are important for AI because they are designed to challenge and train human cognitive capabilities, and are thus uniquely relevant benchmark problems. They are also uniquely convenient benchmark problems, as they allow unbiased comparison between algorithms and can be executed thousands of times fast than realtime. But one should also not forget the financial clout of the games industry and games' appeal for new students. While research on board games such as Chess and Go has been part of AI research since its inception, the last decade has seen the rise of a research community around AI for video games, and not only for playing them. In this talk I will outline some of the most important trends in recent years. One is General Video Game Playing: developing controllers that can learn to play not just a single game, but a large variety variety of them. Another is Procedural Content Generation, where AI algorithms are used to generate content for games, or even design the games themselves. Yet another trend is AI-assisted design tools, which provide the game designer with instant feedback and suggestions and thus scaffolds the game design process. These research topics inform each other, with general video game playing algorithms being important for procedural content generation and AI-assisted design tools. Finally, I will try to convince you that your own research is important for this endeavor and that you should consider steering your research towards AI for games.