Essay – A Stanford Reunion – Artificial Intelligence’s Early Years

STANFORD, Calif. — The personal computer and the technologies that led to the Internet were largely invented in the 1960s and ’70s at three computer research laboratories next to the Stanford University campus.

One laboratory, Douglas Engelbart’s Augmentation Research Center, became known for the mouse; a second, Xerox’s Palo Alto Research Center, developed the Alto, the first modern personal computer. But the third, the Stanford Artificial Intelligence Laboratory, or SAIL, run by the computer scientist John McCarthy, gained less recognition.

That may be because SAIL tackled a much harder problem: building a working artificial intelligence system. By the mid-1980s, many scientists both inside and outside of the artificial intelligence community had come to see the effort as a failure. The outlook was more promising in 1963 when Dr. McCarthy began his effort. His initial proposal, to the Advanced Research Projects Agency of the Pentagon, envisioned that building a thinking machine would take about a decade.

Four and a half decades later, much of the original optimism is back, driven by rapid progress in artificial intelligence technologies, and that sense was tangible last month when more than 200 of the original SAIL scientists assembled at the William Gates Computer Science Building here for a two-day reunion.

During their first 10 years, SAIL researchers embarked on an extraordinarily rich set of technical and scientific challenges that are still on the frontiers of computer science, including machine vision and robotic manipulation, as well as language and navigation.

In 1966, the laboratory took up residence in the foothills of the Santa Cruz Mountains behind Stanford in an unfinished corporate research facility that had been intended for a telecommunications firm.

The atmosphere, however, was anything but button-down corporate. The antiwar movement and the counterculture were in full swing, and the lab reflected the widely disparate political views and turmoil of the time. Dr. McCarthy was a committed leftist who would gradually move to the right during the ’60s; Les Earnest, the laboratory’s deputy director, who had worked in government intelligence, would move to the left.

The graduate students soon discovered the building’s attic and took up residence there. Mr. Earnest found a clever way, known in the parlance of the A.I. community as a “hack,” to pay for a sauna in the basement of the building, and because many of the young researchers were devotees of Tolkien’s “Lord of the Rings,” they created a special font in Elvish and used it to identify offices as places from Middle Earth.

The scientists and engineers who worked at the laboratory constitute an extraordinary Who’s Who in the computing world.

Dr. McCarthy coined the term artificial intelligence in the 1950s. Before coming to SAIL he developed the LISP programming language and invented the time-sharing approach to computers. Mr. Earnest designed the first spell-checker and is rightly described as the father of social networking and blogging for his contribution of the finger command that made it possible to tell where the laboratory’s computer users were and what they were doing.

Among others, Raj Reddy and Hans Moravec went on to pioneer speech recognition and robotics at Carnegie Mellon University. Alan Kay brought his Dynabook portable computer concept first to Xerox PARC and later to Apple. Larry Tesler developed the philosophy of simplicity in computer interfaces that would come to define the look and functioning of the screens of modern Apple computers — what is called the graphical user interface, or G.U.I.

Don Knuth wrote the definitive texts on computer programming. Joel Pitts, a Stanford undergraduate, took a version of the Space War computer game and turned it into the first coin-operated video game — which was installed in the university’s student coffee house — months before Nolan Bushnell did the same with Atari. The Nobel Prize-winning geneticist Joshua Lederberg worked with Edward Feigenbaum, a computer scientist, on an early effort to apply artificial intelligence techniques to create software to act as a kind of medical expert.

John Chowning, a musicologist, referred to SAIL as a “Socratean abode.” He was invited to use the mainframe computer at the laboratory late at night when the demand was light, and his group went on to pioneer FM synthesis, a technique for creating sounds that transforms the quality, or timbre, of a simple waveform into a more complex sound. (The technique was discovered by Dr. Chowning at Stanford in 1973 and later licensed to Yamaha.)

The laboratory merged with the computer science department at Stanford in 1980, reopened in 2004, and is now enjoying a renaissance. Its trajectory can be seen in the progress made since 1970, when a graduate researcher programmed a robot to automatically follow a white line under controlled lighting conditions at eight-tenths mile per hour. Thirty-five years later, a team of artificial intelligence researchers at Stanford would equip a Volkswagen Touareg named Stanley with lasers, cameras and a cluster of powerful computers to drive autonomously for 131 miles over mountain roads in California at an average speed of 19.1 miles per hour to win $2 million in the 2005 Darpa Grand Challenge, a robotic vehicle contest.

“We are a first-class citizen right now with some of the strongest recent advances in the field,” said Sebastian Thrun, a roboticist who is the director of SAIL and was one of the leaders of the Stanley team.

The reunion also gave a hint of what is to come. During an afternoon symposium at the reunion, several of the current SAIL researchers showed a startling video called “Chaos” taken from the Stanford Autonomous Helicopter project. An exercise in machine learning, the video shows a model helicopter making a remarkable series of maneuvers that would not be possible by a human pilot. The demonstration is particular striking because the pilot system first learned from a human pilot and then was able to extend those skills.

But an artificial intelligence? It is still an open question. In 1978, Dr. McCarthy wrote, “human-level A.I. might require 1.7 Einsteins, 2 Maxwells, 5 Faradays and .3 Manhattan Projects.”

Recommend Next Article in Science (3 of 23) » A version of this article appeared in print on December 8, 2009, on page D4 of the New York edition.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: