Human programmers seem to know a lot about programming. This suggests a way to try to build automatic programming systems: encode this knowledge in some machine-usable form. In order to test the viability of this approach, knowledge about elementary symbolic programming has been codified into a set of about four hundred detailed rules, and a system, called PECOS, has been built for applying these rules to the task of implementing abstract algorithms. The implementation techniques covered by the rules include the representation of mappings as tables, sets of pairs, property list markings, and inverted mappings, as well as several techniques for enumerating the elements of a collection. The generality of the rules is suggested by the variety of domains in which PECOS has successfully implemented abstract algorithms, including simple symbolic programming, sorting, graph theory, and even simple number theory. In each case, PECOS's knowledge of different techniques enabled the construction of several alternative implementations. In addition, the rules can be used to explain such programming tricks as the use of property list markings to perform an intersection of two linked lists in linear time. Extrapolating from PECOS's knowledge-based approach and from three other approaches to automatic programming (deductive, transformational, high level language), the future of automatic programming seems to involve a changing role for deduction and a range of positions on the generality-power spectrum.

I will remain Ion8 and deeply indebted to Herb Simon, H. T. Kung, AI Newell, Jay

How do practicing physicians make clinical decisions? What techniques can we use in the computer to produce programs that exhibit medical expertise?

In A. Hanson and E. Riseman (Eds.), Computer vision systems. New York: Academic Press, 335-352.

Retrospective on lessons learned from the Dendral project."The DENDRAL and Meta-DENDRAL programs are products of a large, interdisciplinary group of Stanford University scientists concerned with many and highly varied aspects of the mechanization of scientific reasoning and the formalization of scientific knowledge for this purpose. An early motivation for our wok was to explore the power of existing Al methods, such as heuristic search, for reasoning in difficult scientific problems. Another concern has been to exploit the AI methodology to understand better some fundamental questions in the philosophy of science, for example the processes by which explanatory hypotheses are discovered or judged adequate. From the start, the project has had an applications dimension. It has sought to develop "expert level" agents to assist in the solution of problems in their discipline that require complex symbolic reasoning. The applications dimension is the focus of this paper."Artificial Intelligence 11 (1-2): 5-24

SU/X and SU/P are knowledge-based programs which employ pattern-invoked inference methods. Both tasks are concerned with the interpretation of large quantities of digitized signal data. The task of SU/X is to understand "continuous signals", that is, signals which persist over time. The task of SU/P is to interpret protein x-ray crystallographic data. Some features of the design are: (1) incremental interpretation of data employing many different pattern-invoked sources of knowledge, (2) production rule representation of knowledge, including high level strategy knowledge, (3) "opportunistic" hypothesis formation using both data-driven and model-driven techniques within a general hypothesize-and-test paradigm; and (4) multilevel representation of the solution hypothesis.

Expert system technology has been successfully applied to many practical problems, but. In this paper we discuss some of the problems confronting computer vision and present an approach to the development of general knoledge-based vision systems. The primary mechanism is a rule-based approach for the generation of initial object hypotheses, which allow focus of attention strategies. The rule set, applied to the attributes of the lines, regions, and surfaces in an intermediate symbolic representation, is constructed interactively with visual feedback to the user. Simples rules are defined as ranges over a feature value which are converted to a vote for an object label; complex rules are constructed via a functional combination of the output from the simple rules.

"The terms adaptation, learning, concept-formation, induction, self-organization, and self-repair have all been used in the context of learning system (LS) research. The research has been conducted within many different scientific communities, however, and these terms have come to have a variety of meanings. It is therefore often difficult to recognize that problems which are described differently may in fact be identical. Learning system models as well are often tuned to the require- ments of a particular discipline and are not suitable for application in related disciplines."In Encyclopedia of Computer Science and Technology, Vol. 11. Dekker

"A CASNET model consists of three main components: observations of a patient, pathophysiological states, and disease classifications. As observations are recorded, they are associated with the appropriate intennediate states. These states, in turn, are typically causally related, thereby forming a network that summarizes the mechanisms of disease. It is these patterns of states in the network that are linked to individual disease classes." Artificial intelligence, August, 1978. Reprinted in Clancey & Shortliffe. Readings in Medical Artificial Intelligence: The First Decade. Ch. 7.