Visual architecture and cognitive architecture

Traditional architectures have fundamental epistemo-logical problems. Perception is inherently resource limited so controlling perception involves all the same AI-complete problems of reasoning about time and resources as the full-scale planing problem. Allowing a planner to transparently assume that the information it needs will automatically be present and up-to-date in the model thus presupposes a solution to a problem at least as diicult as planning itself. Although one can imagine many possible solutions to this problem, such as allowing the planner to recurse on its own epistemological problems, there have been no convincing attempts at this. In this paper, I compare behavior-based and traditional systems in terms of their representational power and the strengths of their implicit epistemo-logical theories. I argue that both have serious limitations and that those limitations are not addressed simply by joining the two into a hybrid. I discuss my work with using vision to support real-time activity and give an example of an interesting intermediate point between reactive and classical architec-tures that preserves the simplicity and parallelism of behavior-based systems while supporting \symbolic" representations. Traditionally, AI theories have assumed, either implicitly or explicitly, an architecture in which modules of the mind (perception, reasoning, motor control, etc.) are linked by way of some centralized database-like structure, often referred to as a world model ((26] for example). Recently, a number of alternative ar-chitectures have been proposed which, to greater or lesser degrees, claim to do away with world models or with representations entirely 8]]3]. Many of the criticisms of traditional architecture revolve around speed and timescale. Planning, so the story goes, is slow but exible, while feedback loops are fast but stupid. A common approach, both in this special issue and in the literature in general, is to adopt a hybrid which fuses a slow planner running on a long timescale and a set of fast feedback loops running on a short timescale. The problem with this argument is that planning isn't slow, it's combinato-rially explosive. Running an O(2 n) algorithm on a timescale ten times slower is the same as running it on a computer ten times faster: it simply lets you increase n by three. If timescale were the true problem, faster CPUs would make tiered architectures obsolete in a few years. I believe the true issues are not speed in the sense of timescale , but combinatorics and epistemology. The former has been extensively discussed, …