Artificial Intelligence: A Rand Perspective

This article presents a brief history of artificial intelligence research at the Rand Corporation. Rand has long been a leader in the field of AI, beginning with the seminal work of Newell, Shaw, and Simon some thirty years ago, and continues with recent work in expert systems and knowledge-based simulation. This article traces the major accomplishments in AI at Rand with particular emphasis on Rand’s research during the past decade. The references highlight the major Rand documents on AI and related subjects This article is a slightly revised version of the introduction to the book Expert Systems: Techniques, Tools, and Applications, by Philip Klahr and Donald A. Waterman (Reading, Massachusetts: Addison-Wesley, 1986). THE AI MAGAZINE Summer, 1986 55 building one of the first stored-program digital computers, the JOHNNIAC (see Figure 1) (Gruenberger, 1968);l George Dantzig and his associates were inventing linear programming (Dantzig, 1963); Les Ford and Ray Fulkerson were developing techniques for network flow analysis (Ford & Fulkerson, 1962); Richard Bellman was developing his ideas on dynamic programming (Bellman, 1953); Herman Kahn was advancing techniques for Monte Carlo simulation (Kahn, 1955); Lloyd Shapley was revolutionizing game theory (Shapley, 1951-1960); Stephen Kleene was advancing our understanding of finite automata (Kleene, 1951); Alfred Tarski was helping to define a theory of computation (Tarski, 1951); and James Culbertson (Culbertson, 1952, 1953) and Alton Householder (Householder, 1951a, 1951b) were investigating the relationship between neural nets, learning, and automata.2 Within this milieu, Newell, Shaw, and Simon were developing methods and directions for AI research. Perhaps equally important was their development of appropriate computational tools for AI programming. Using the notion of linked-list structures to represent symbolic information, Newell and his associates developed the first symbolmanipulating and list-processing languages, a series of IPL (Information-Processing Language) languages that culminated in IPL-V (Newell, 1963c; Newell & Tonge, 1960). In their 1963 paper (Bobrow & Raphael, 1963), Dan Bobrow and Bert Raphael (both of MIT at the time but also Rand consultants) included IPL-V as one of the earliest and most highly developed list-processing languages. Because of Rand’s unique computing environment and its close ties to the Carnegie Institute of Technology, several Carnegie graduate students were attracted to Rand, and several Ph.D. dissertations emerged including those of Fred Tonge (Tonge, 1959, 1960) and Ed Feigenbaum (Feigenbaum, 1959, 1961). During the early 1960s Feigenbaum, in collaboration with Simon, continued to publish Rand reports (Feigenbaum, 1964; Feigenbaum & Simon, 1961a, 1961b, 1961c, 1962; Simon & Feigenbaum, 1964) describing his experiments with his verbal learning program EPAM. Even after completing his work at Carnegie, Feigenbaum remained a Rand consultant and was highly influential in Rand’s research on expert systems and expert system languages that emerged in the early 1970s. Newell and Simon were also Rand consultants during the 1960s and 1970s. One of their associates, Don Waterman, joined Rand in the mid-1970s and brought much of their influence on the use of production systems to Rand’s first work on expert systems. lThe JOHNNIAC, named after John van Neumann, a Rand consultant in the late 1940s and early 1950s was in operation from 1953 to 1966. It was used extensively by Newell, Shaw, and Simon in their work on information processing 2Much of the credit for creating this intellectually stimulating environment belongs to John Davis Williams, who led Rand’s Mathematics Department in the 1950s and also to the United States Air Force for its generous sponsorship of a broad range of research activities. AI also had its share of controversy, however, at Rand and elsewhere. Given its quick rise to popularity and its ambitious predictions (Simon & Newell, 1958), AI soon had its critics, and one of the most prominent, Hubert Dreyfus, published his famous critique of AI (Dreyfus, 1965) while he was consulting at Rand. In addition, the early promise of automatic machine translation of text from one language to another (the emphasis at Rand was on translation from Russian to English) produced only modest systems, and the goal of fully automated machine translation was abandoned in the early 1960s. The research in machine translation did, however, serve to elucidate the difficult problems of automated language understanding and translation. As a result, work in this area turned toward fundamental and generic issues of linguistic theory, and Rand engaged in over a decade of activity in computational linguistics. By 1967 Rand researchers had produced a wealth of literature (over 140 articles) on linguistic theory and research methods, computational techniques, the English and Russian languages, automatic content analysis, information retrieval, and psycholinguistics (Hays, Henisz-Dostert, & Rapp, 1967). In addition, David Hays produced one of the earliest textbooks on computational linguistics (Hays, 1967). During the 1960s Rand provided a center in which natural language researchers from all over the world could meet, communicate, and collaborate. Special seminar programs and summer symposia (for example, Kochen et al., 1964) provided ample opportunities for researchers to exchange ideas and test theories. Work at Rand during this period included a number of developments: Martin Kay and his associates were working on the MIND system, which focused on research in morphology (Kay & Martins,l970), semantic networks (Kay & Su, 1970; Shapiro, 1971), ‘and parsing (Kaplan, 1970, 1971; Kay, 1967); Jane Robinson was developing new syntactic analyzers (Robinson & Marks, 1965); Roger Levien and Bill Maron were developing the relational data file for information retrieval and question answering (Levien, 1969; Levien & Maron, 1965, 1966); Larry Kuhns was developing a sophisticated query language for database inference (Kuhns, 1967, 1970); and, in a somewhat different area, work was beginning on a new theory of “fuzzy sets” (Bellman, Kalaba, & Zadeh, 1964). Human-Oriented Environments Since its early involvement with the JOHNNIAC in 1953, Rand has continually worked on the development of human-oriented interfaces. Although much of this work has been outside the AI framework, the research has provided interactive computing environments that have made AI systems easier to design, implement, debug, and understand. Today, computational environments appropriate for AI systems comprise a prominent subfield of AI research. 56 THE AI MAGAZINE Summer, 1986 JOHNNIAC, One of the First Stored-Program Digital Computers