Architectural factors for intelligence in autonomous systems

Abstract

The question of measuring intelligence in artifacts and, in particular, evaluating architectures for it, requires not just an understanding of the very nature of intelligence -quite elusive objective, indeed- but an appropriate stance for evaluation. In this paper we argue that it is not just the case that architectures provide intelligence, but that they really provide a substrate for intelligent behavior in the execution of a particular task. The measuring-intelligence-for-autonomy position becomes maximally relevant in the context of the increased uncertainty levels that the upcoming challenging applications are posing to cognitive architectures. This fits our understanding of intelligence as the capability of maximizing information utility. Copyright © 2007, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.