Classical artificial intelligence tends to analyze a task logically and plan solutions. “This allows systems to be built that always work well when the problem can be precisely translated into an abstract-mathematical form,” reports the team headed by Prof. Martin Butz from the Chair of Cognitive Modeling . On the other hand there are artificial neural networks, which are currently the object of intensive research.

BrainControl is a computer game that is based neither on the one nor the other concept. The goal is rather to explore new variants of artificial intelligence. BrainControl simulates a 2D world as well as virtual figures cooperating and learning within it. The user communicates with them by means of human language. The agents learn how their environment works through interactions, and they act autonomously. They gradually require less and less feeding with set instructions. A two-minute introductory video clearly illustrates how this works.

BrainControl is basic research done in an entertaining manner . It is expected to improve AI in the long term. Games tend to be well suited for this kind of research. Since Google’s software beat its professional opponents in the board game Go a year ago, human dominance in chess has also become a thing of the past: In December, a Google program learned the game of kings on its own within four hours – and went on to defeat the best players in the world.