ERIC Number: ED178060
Record Type: Non-Journal
Publication Date: 1978-Oct
Reference Count: N/A
Machine Methods for Acquiring, Learning, and Applying Knowledge.
Hayes-Roth, Frederick; And Others
A research plan for identifying and acting upon constraints that impede the development of knowledge-based intelligent systems is described. The two primary problems identified are knowledge programming, the task of which is to create an intelligent system that does what an expert says it should, and learning, the problem requiring the criticizing and expanding of current knowledge to improve overall system performance. In this view, learning produces new knowledge which must be accommodated to implement an improved system. Motivating problems, technical requirements for supporting system extensibility and improvability, and research proposed to achieve these objectives are described and related to previous research. Numerous examples are drawn for a heuristic program that plays a moderately difficult card game (hearts). Several appendices provide additional details on many technical issues, including the representation of knowledge and the structure of knowledge bases, the design of a knowledge programmer, various control methods, the design of a learning workbench, an illustrative learning scenario, and various learning heuristics. (Author)
Descriptors: Artificial Intelligence, Automation, Computer Oriented Programs, Cybernetics, Game Theory, Information Processing, Programing, Research Proposals
The Rand Corporation, 1700 Main Street, Santa Monica, California 90406 ($5.00 per copy)
Publication Type: Reports - Descriptive
Education Level: N/A
Authoring Institution: Rand Corp., Santa Monica, CA.