On Informational Principles of Embodied Cognition
For many decades, Artificial Intelligence adopted a platonic view that intelligent behaviour is produced in the "brain" only and any body is only an incidental translator between thought and action. In the last two decades, in view of the successes of the subsumption architecture and embodied robotics, this perspective has changed to acknowledge the central importance of the body and the perception-action loop as whole in helping an organisms' brain to carry out useful ("intelligent") behaviours. A central keyword for this phenomenon is, of course, "environmental/morphological computation" (Paul 2006; Pfeifer and Bongard 2007).
The question arises, how/why exactly does this work? What are the principles that make environmental computation work so successfully and how can the contribution that the body provides to cognition be characterized objectively?
In the last years, Information Theory has been identified as providing a natural language to characterize cognitive processing, cognitive invariants as well as the contribution of the embodiment to the cognitive process. The talk will present a number of intuitive and less intuitive consequences of these considerations and provide some - sometimes quite surprising - illustrations of the power of the informational view of cognition.
Daniel Polani obtained his PhD in 1996 from the Johannes Gutenberg University in Mainz on the Genetic Optimization of Self-Organizing Maps. Following a research visit to the University of Texas at Austin in 1997 and a postdoctorate in Mainz, he became Research Fellow at the Institute for Neuro- and Bioinformatics at the University of Luebeck in 2000. In 2002, he joined the Adaptive Systems and Algorithms Research Groups at the School of Computer Science at the University of Hertfordshire as Principal Lecturer. Since 2008, he is Reader in Artificial Life at the University of Hertfordshire. He is interested in the evolution of the perception-action loop, and in particular, the foundations of intelligent behaviour in biological as well as artificial agents, especially in terms of optimality principles on the basis of Shannon's theory of information.