Abstract

Artificial Intelligence researchers have produced models of different components required for intelligence such as knowledge representation, reasoning, natural language processing, learning, neural computation, and robotics. While isolating these components is vital for studying the important issues in each area, nobody believes that these subfields have been completely solved. The field has produced many systems that exhibit interesting behavior or intelligent results, yet we know very little about how to combine their power. In recent years, there has been an increasing interest in the integration of different aspects of intelligence in systems known as integrated architectures and that are capable of producing general intelligent behavior. The design space of integrated architectures is large and not well understood. Researchers are exploring the different issues by exploring points in that space, trying to gain a better understanding in the process. This paper discusses what constitutes an integrated architecture, and what the desired capabilities of such a system are. Next, we introduce some of the architectures that have been proposed. We conclude with a discussion that compares the architectures and their stands on different design issues and a reflection on what is still missing in the current systems.

Date

March 18, 1991

Authors

Yolanda Gil

Journal

Proc. of Approaches to non-conventional Computing: towards Intelligent Systems (TECCOMP 91)

Pages

41-55