Authors:

 Krueger, Ted.

Title:

 There is No Intelligence

Keywords:

 intelligence, artificial intelligence, behavior, solution space
click here to download the full paper

Abstract:

 A substantial portion of contemporary interactive, autonomous and adaptive work references intelligence. Work, both artistic and technical, that seeks to embody intelligence in networks or artifacts requires an operational definition of the concept. Some notion of the-thing-itself is required in order to make claims about the artificial version that is to be produced and to guide design activities that are involved in its production. Yet definitions of intelligence, either natural of artificial, are typically vague. Definitions of intelligence in these contexts vary widely from overstatement (common techno - hype) to convenience (intelligence is defined as the things that the project does) or avoidance( let's not waste time defining intelligence, but simply get on with the work at hand).

Intelligence is an attribution made on the basis of judgements about the quality of the interaction between an agent and an environment; a capacity for appropriate behaviors within a specified context. It is an opinion. The assumption is that these behaviors are caused by an underlying intelligence. It is the implicit assumption of 'cause and effect'; a linear connection between the observed and an assumed generator of that behavior from which the notion of intelligence is derived. While the simple newtonian universe has been undergoing revision for several centuries, its technological and mechanistic bias remains a substrate for our notions of intelligence and thereby our understanding of ourselves.

This paper argues that there is no need to assume the existence of a thing that can be called intelligence. However, we can not also regress into a kind of neo-behaviorism in which only the external activity is available for examination. An alternative characterization is developed in which the possibility for responses to environmental conditions describe a space of potential solutions.

This space may be continuous or discontinuous and developed over n-dimensions. Trajectories through this space - developing out of both the properties of the context and the properties of the solution space - interact to determine the behavior of the system. This model allows for a specification of the properties and distribution of the solution space and so incorporates it into the design process.
Krueger, Ted. tkrueger@comp.uark.edu School of Architecture University of Arkansas http://comp.uark.edu/~tkrueger Ted Krueger is the E Fay Jones Visiting Professor of Architecture at the University of Arkansas for 1996-99. He is interested in the cultural implications of technical developments.