Reference: Feigenbaum, E.; Engelmore, R.; Gruber, T. R.; & Iwasaki, Y. Large Knowledge Bases for Engineering: The How Things Work Project of the Knowledge Systems Laboratory. Knowledge Systems Laboratory, November, 1990.
Abstract: We view the limitation of highly specialized, narrowly scoped knowledge bases as the single greatest impediment to achieving higher levels of competence in expert systems and other AI programs. Programs must know more than they know today, and be able to use more general forms of knowledge, if they are to become more intelligent. To build programs with more generally useful knowledge will require advances in the representation of knowledge and appropriate reasoning processes. Our long-term goal is to explore the limits of performance/competence achievable by intelligent systems. Considering the principle that intelligent performance is strongly dependent on the knowledge given to systems, we are concentrating on the question of how to represent general-purpose scientific and engineering knowledge that can be used in a variety of important tasks. This paper summarizes the technical issues that motivate the How Things Work project, discusses some of the tangible results expected, and concludes with a section on the scientific and social importance of the research.
Full paper available as ps, hqx.