Collins Conference Room
Working Group
 

Abstract.  In recent years, it has been shown that the physical properties of a body can perform functions that are normally attributed to the brain. Two examples are the morphology of an insect’s eye, which contributes to motion detection, and tendon friction in the human hand, which contributes to grasping. These examples illustrate that computations (motion detection, position control), which would normally be assigned to the brain, are somehow (at least partially) outsourced to the body. This outsourcing of computation to the morphology has led to the notion of Morphological Computation, which is now a central concept in the field of embodied artificial intelligence. The problem with this notion is that not all physical processes in the body can be considered as computation in the classical sense. Hence, one important question is to determine which processes should be considered as morphological computation and which should be considered as purely physical processes?

The current trend is to answer this question by breaking down morphological processes into different concepts, such as morphological computation, morphological control, physical processes that contribute to a behavior, and physical processes that are not behavior relevant. We believe that this is the wrong approach. Instead of diversifying morphological contributions to intelligence, we should find a unifying perspective, which is the goal of this Working Group. The long-term goal is to open the pathway to a formal treatment of this fundamental concept of embodied artificial intelligence.

Purpose:
Research Collaboration
SFI Host:
Nihat Ay and Keyan Ghazi-Zahedi

More SFI Events