Computer Chronicles Revisited 24 — Edward Feigenbaum

For this episode, I’m going to handle things a bit differently. There was only a single guest–Stanford University computer science professor Edward Feigenbaum–and the subject is one that, quite frankly, does not strike me as all that interesting. So rather than do an extended point-by-point recap of the episode, I’m just going to summarize in broad strokes. Trust me, if you had watched the episode, you’d thank me.

Stanford Professor Discusses AI, Future of Japanese Computing Initiative

You may recall that Feigenbaum appeared in an earlier episode, which I recapped in Part 20. That clip was actually taken from the start of his interview from this episode. And really, this episode covers a lot of the same ground, which focused on knowledge-based or “expert” computer systems. Here, Feigenbaum talked about the state of such systems and the potential threat from Japan’s Fifth Generation computer project.

The Fifth Generation was the term used by Feigenbaum and others to describe to an initiative by the Japanese government to “develop computers with reasoning capabilities, rather than the ability to merely perform calculations,” according to a 1992 article in The New York Times. Japan’s Ministry of International Trade and Industry started the 10-year Fifth Generation program in 1982 to basically spur breakthroughs in artificial intelligence.

Speaking with Stewart Cheifet and Gary Kildall–as well as Herbert Lechner, who was along for the ride–Feigenbaum explained that work in the United States on “expert systems” had been going on since about 1965, the beginning of the second decade in AI research. At that time, the emphasis was on “generality,” or the ability of a program to shift its attention from one area to another easily, rather than speed and performance. But the more current research done by people at Stanford focused on creating expert programs, which required encoding a great deal of knowledge about a field into the machine.

Feigenbaum said now–in late 1983–we were starting to see the industrialization of AI, similar to the transition that took place with computing itself during the late 1940s and 1950s, when the technology moved out of the laboratory and into practical uses. For example, he said the University of Pittsburgh now had a “spectacular program” used for clinical diagnosis that knew about 500 different diseases and 3,500 different signs and symptoms.

After a lengthy demonstration of a program developed by Feigenbaum to help the French national oil company Elf Aquitaine identify drilling locations, the conversation eventually shifted to Fifth Generation computers. Feigenbaum co-authored a book on this subject with Pamela McCorduck called, appropriately, The Fifth Generation–Artificial Intelligence and Japan’s Computer Challenge to the World. He said the Japanese strategy was interesting in that they currently lacked market share in the worldwide computer market. It was difficult to gain market share due to IBM’s dominance. So what you had to do was gain a position in an area where IBM has not been active–and that was artificial intelligence. The Japanese goal was therefore to get there first and fast so that they could get a large chunk of the AI/expert systems market before IBM woke up.

Right now, Feigenbaum said, the United States was clearly ahead of the Japanese when it came to computers. But he thought with this new dedicated effort via the Fifth Generation project, the U.S. computer manufacturers had to move pretty fast if they wanted to stay ahead.

Coleco Launches Adam, Apple Wins Lawsuit Against Clonemaker

Stewart Cheifet presented this week’s “Random Access” segment, which I’m dating at sometime during October 1983.

The Five Generations of Computers

Let’s get the obvious question out of the way–what did “Fifth Generation” mean? Or put another way, what were the first four generations. The simple explanation is as follows:

The term “Fifth Generation” is now used to refer to essentially any computer developed after 1980. In terms of the Japanese Fifth Generation project, it was, as described by the aforementioned New York Times article, a 10-year project that largely failed to yield significant commercial results. Indeed, Edward Feigenbaum told the Times that while the Japanese government’s investment had yielded some interesting computer designs and software, in the end “no one is using the technology.”

And contrary to the fears raised by Feigenbaum and others that the Fifth Generation project represented an existential threat by the Japanese to U.S. computer dominance, the Times noted that “most American computer scientists and executives have long since stopped paying attention to the Fifth Generation project.” The Times added that American computer scientists–presumably including Feigenbaum–had overstated the Japanese threat “[i]n order to coax more support from the United States Government for computer science research.”

Feigenbaum Joined the Air Force (in His Late 50s!)

A decade after his Computer Chronicles appearance, Edward Feigenbaum served a three-year stint as chief scientist of the United States Air Force during the Clinton administration. It was no doubt a highlight in a career that began when he was named one of the original members of Stanford University’s computer science faculty in 1965. At Stanford, Feigenbaum developed (with Joshua Lederberg) one of the first expert systems projects called DENDRAL. He would go on to serve as chairman of Stanford’s computer science department and director of its Computer Center. He also co-founded three AI-related startups, including IntelliCorp, which was featured in a previously reviewed Computer Chronicles episode.

Feigenbaum retired from the Stanford faculty in 2000. He turned 85 this past January.

Notes from the Random Access File