TECHNOLOGY TRANSFER Rl And Beyond: AI Technology Transfer At DEC

This article describes one person’s experience in coming from an academic environment to work at Digital Equipment Corporation. The author feels his own experience has paralleled the transfer of AI technology from academia to industry, where AI researchers must live up to very different expectations, but, also enjoy very different rewards. This article covers the historical background of DEC’s involvement with AI, the development of Rl-known internally and henceforth in this article as XCONP and DEC’s experiences with it and its consequences. Finally, the article offers advice for other corporations planning to develop their own capabilities in AI. DEC and AI-A Historical Perspective DEC is unlike most other corporations, even most other computer corporations, in that it has a fairly long and involved historical relationship with AI. I won’t give all the details of that history. Rather, I will explain how that history has affected perceptions and attitudes at DEC and other corporations, and how those attitudes have influenced DEC’s decisions in the area of artificial intelligence. I’ve divided this history into two distinct parts. The first part describes DEC’s years as the main hardware vendor for the AI community, and the second part describes DEC’s more recent role as a designer and user of AI software. DEC was the main hardware vendor for the AI community during the 197Os, a period characterized by the birth and development of expert systems in academic research laboratories. Since DEC had a very close relationship with the academic computing laboratories, much of this work was done on the DECsystem-10 and its successor, the DECsystem-20. The two major dialects of LISP, MACLISP and INTERLISP, were developed on these machines; so were most of AI’s early successes and disappointments. This facilitated an exchange of engineering talent between university AI labs and DEC’s engineering labs. Thus, even before the commercialization of AI and DEC’s entry into the AI market, DEC engineers were familiar with AI software projects and techniques. As AI became increasingly commercial, DEC’s history of involvement This article is an edited version of Dr Polit’s presentation at the Technology Transfer Symposium held at the AAAI-83 conference with AI led people both inside and outside DEC to believe that DEC was a natural supplier of machines and tools for the development of AI systems. Consequently, DEC was open to entering the AI software market very quickly I’ll discuss XCON in a bit more detail later on, but right now I’ll describe how its success affected attitudes towards AI within and outside DEC. XCON was originally devclopcd at Carnegie-Mellon IJniversity in 1979 by John McDermott and several associates. It was one of many DECPmiivcrsity collaborations; DEC profited from almost all of these, as did the universities. One benefit of this project was that a closer tie was created between DEC and a major AI lab. Although the system received DEC’s moral and financial support during its development, it, did not immediately represent a large commitment of DEC to artificial intelligence. The successful completion of RI and its incorporation as XCON, however, changed attitudes towards AI within DEC The demonstrated practicality of AI made it a more legitimate subject of industrial interest. Also, somewhat ironically, it suddenly gave DEC the image of being a leading practitioner of AI, a prominent example of an industrial firm successfully using an AI tool. Finally, and importantly, considerable cxpcrience was gained by bringing XCON to use at DEC. Building Expert Systems I’ll now give a brief review of the steps involved in building expert systems as they are described by many researchers. The five steps involved in building an expert system are: Step 1: problem recognition, Step 2: task definition, Step 3: initial design, Step 4: knowledge acquisition, and Step 5: system maintenance. First, during step 1, someone must recognize that there is a problem to be solved and determine whether AI is an appropriate way of solving it. Frequently, the problem is perceived as a bottleneck in a larger process; sometimes it is a scarcity of traiued personnel. Second, during step 2, researchers must define the functions the AI system will perform. These two steps may be the most difficult ones, because researchers must have a comprehensive underst,anding 76 THE AI MAGAZINE Winter. 1985 AI Magazine Volume 5 Number 4 (1984) (© AAAI)