Abstract
Van Gelder's clear distinction between the quantitative nature of dynamical systems and the nonquantitative nature of computational processes provides a firm basis for distinguishing between processes that happen in time and processes that happen over time. Symbolic reasoning, the presumed basis of intelligent behavior in robots, happens over time. However, the movements and actions that robots must make to behave intelligently, happen in time. Attempting to connect the two, as classical artificial intelligence and robotics have presumed to be necessary, has produced a tension and an arbitrarily moving interface in the construction of robots. Adopting a robotic version of the dynamical hypothesis offers sound theoretical and scientific justification for those robotics researchers who continue to insist that getting the interaction dynamics of intelligent behavior right is a purely dynamical matter, and never a symbolic computational one.
