Feigenbaum, Edward

(1936–  ) American

Computer Scientist

Edward Feigenbaum is a pioneer artificial intelligence researcher, best known for his development of expert sys-tems (see artificial intelligence). Feigenbaum was born in Weehawken, New Jersey. His father, a Polish immigrant, died before Feigenbaum’s first birthday. His stepfather, an accountant and bakery manager, was fascinated by science and regularly brought young Edward to the Hayden Plane-tarium’s shows and to every department of the vast Museum of Natural History. The electromechanical calculator his father used to keep accounts at the bakery particularly fas-cinated Edward. His interest in science gradually turned to a perhaps more practical interest in electrical engineering.

While at the Carnegie Institute of Technology (now Carnegie Mellon University), Feigenbaum was encour-aged to venture beyond the more mundane curriculum to the emerging field of computation. He became interested in John Von Neumann’s work in game theory and deci-sion making and also met Herbert Simon, who was con-ducting pioneering research into how organizations made decisions (see von Neumann, John). This in turn brought Feigenbaum into the early ferment of artificial intelligence research in the mid-1950s. Simon and Alan Newell had just developed Logic Theorist, a program that simulated the process by which mathematicians proved theorems through the application of heuristics, or strategies for breaking prob-lems down into simpler components from which a chain of assertions could be assembled leading to a proof.

Feigenbaum quickly learned to program IBM main-frames and then began writing AI programs. For his doc-toral thesis, he explored the relation of artificial problem solving to the operation of the human mind. He wrote a computer program that could simulate the human pro-cess of perceiving, memorizing, and organizing data for retrieval. Feigenbaum’s program, the Elementary Perceiver and Memorizer (EPAM), was a seminal contribution to AI. Its “discrimination net,” which attempted to distinguish between different stimuli by retaining key bits of informa-tion, would eventually evolve into the neural network (see neural network). Together with Julian Feldman, Feigen-baum edited the 1962 book Computers and Thought, which summarized both the remarkable progress and perplexing difficulties encountered during the field’s first decade.

During the 1960s, Feigenbaum worked to develop sys-tems that could perform induction (that is, derive general principles based on the accumulation of data about specific cases). Working on a project to develop a mass spectrom-eter for a Mars probe, Feigenbaum and his fellow research-ers became frustrated at the computer’s lack of knowledge about basic rules of chemistry. Feigenbaum then decided that such rules (or knowledge) might be encoded in such a way that the program could apply it to the data being gathered from chemical samples. The result in 1965 was Dendral, the first of what would become a host of success-ful and productive expert systems (see expert system). A further advance came in 1970 with Meta-Dendral, a pro-gram that could not only apply existing rules to determine the structure of a compound, it could also compare known structures with the existing database of rules and infer new rules, thus improving its own performance.

  During the 1980s, Feigenbaum coedited the four-volume Handbook of Artificial Intelligence. He also introduced expert systems to a lay audience in two books, The Fifth Generation (co-authored with Pamela McCorduck) and The Rise of the Expert Company. Feigenbaum combined scientific creativity with entre-preneurship in founding a company called IntelliGenetics and serving as a director of Teknowledge and IntelliCorp. These companies pioneered the commercialization of expert systems. In doing so, Feigenbaum and his colleagues publicized the discipline of “knowledge engineering”—the capturing and encoding of professional knowledge in medi-cine, chemistry, engineering, and other fields so that it can be used by an expert system. In what he calls the “knowl-edge principle” he asserts that the quality of knowledge in a system is more important than the algorithms used for reasoning. Thus, Feigenbaum has tried to develop knowl-edge bases that might be maintained and shared as easily as conventional databases.

Remaining active in the 1990s, Feigenbaum was second president of the American Association for Artificial Intel-ligence and (from 1994 to 1997) chief scientist of the U.S. Air Force. In 1995, Feigenbaum received the Association for Computing Machinery’s prestigious A. M. Turing Award. Founder of the Knowledge Systems Laboratory at Stanford University, Feigenbaum remains a professor emeritus of computer science at that institution.