NEWS RELEASE

10/28/03

Dawn Levy, News Service: (650) 725-1944, dawnlevy@stanford.edu

EDITORS: This profile was written by Geoff Koch, a Stanford graduate student in journalism. A photo of Nilsson is available at http://newsphotos.stanford.edu.

Relevant Web URLs:
Video of Shakey the robot:
www.ai.sri.com/videos

Computing with a human touch: Stanford researcher reflects on a career spent in artificial intelligence

Artificial intelligence (AI) pioneer Nils Nilsson has spent a career thinking about the difference between computers and human beings. His conclusion?

"There will always be some difference between computers and human beings, but I think the intellectual and even creativity differences ultimately will narrow," he said.

The Kumagai Professor of Engineering, Emeritus, and chair of the Computer Science Department from 1985 to 1990, Nilsson has earned the right to be taken seriously on such matters. On Oct. 21, Nilsson reflected on his career at a Stanford Computer Forum lecture. The talk repeated one he gave in Acapulco in August to the International Joint Conference on Artificial Intelligence when he was presented the organization's Research Excellence Award.

Bart Selman, a computer science professor at Cornell, had introduced Nilsson at the Acapulco conference. "Nils played a crucial role in nurturing new talent in the field of artificial intelligence and shaping the research agenda for the community to pursue," he wrote in an e-mail interview. Selman first met Nilsson in the 1980s as a doctoral student at the University of Toronto; Nilsson was a distinguished visitor. "For a beginning graduate student, meeting Nils, a founder of our field, was very inspirational."

Nilsson enrolled as a freshman at Stanford in 1951, just a few years after famed British mathematician Alan Turing began speculating about the possibility of machine intelligence. Nilsson earned his doctorate here in 1958. In 1961, having served three years in the U.S. Air Force, Nilsson began exploring neural networks as a young scientist at Stanford Research Institute, which changed its name to SRI International in 1977, seven years after it separated from the university.

In some ways, neural networks function like the human brain. They employ statistical techniques so machines can recognize speech and patterns and perform other useful tasks. Nilsson and his colleagues developed the A* algorithm, which helps search graphs for the shortest distance between two points. Developed in the early 1960s, this algorithm is employed today by anyone who uses mapping applications on the Web to obtain point-to-point driving directions.

Having done this work a few decades before the commercial web and companies like Mapquest.com existed, Nilsson and his colleagues never made a dime from this early work. "It would be nice to get a penny for every search," he laughed.

Nilsson's next project at Stanford Research Institute involved a rather famous robot named Shakey. A 1969 video on the SRI website (www.ai.sri.com/videos) shows a boxy wheeled contraption bumping its way around a room as researchers sporting sideburns and turtlenecks take notes.

The dated appearance of researchers and robot aside, the historical video offers a glimpse at one of the world's first autonomous robots. Using information from a TV camera, range finder and sensors, Shakey could plan to achieve goals and follow a route -- an impressive engineering feat at the time.

With the goal of imbuing machines with human-level intelligence, AI has progressed over the last several decades, along the way causing both optimism and concern. In 1997, reigning world chess champion Garry Kasparov suffered a much publicized defeat to IBM's Deep Blue supercomputer.

The head of the IBM team compared the moment to the landing on the moon, and others declared Deep Blue a prototype of the computers that would solve a range of problems in the future.

However, some like Boston Globe columnist Ellen Goodman expressed concern. "It struck a chord with anyone who has ever been downsized or cyberized and come to regard the brave new computer world as a people-unfriendly environment," she wrote after the event.

Nilsson is not persuaded by these fears and believes the laborsaving benefits of artificial intelligence eventually will improve the lives of many people.

Though he has dabbled in business over the years -- he co-founded a publishing company and has served on the boards of several technology companies -- Nilsson has never considered leaving Stanford.

"I enjoy university-style research," he said.

According to Nilsson, in recent years it has become more challenging to find funding support for curiosity-driven research. He says that agencies such as the National Science Foundation and the Defense Advanced Research Projects Agency increasingly are awarding grants to projects with distinct technical or societal goals. Yet with the rise in computing power, research today is dramatically less expensive than it was in Nilsson's early days at SRI.

Shakey needed room-sized, million-dollar computers to process information and send instructions. Last year, a group of Nilsson's undergraduate students used laptops to develop a virtual soccer game where sophisticated and independent programs compete against each other.

Nilsson sees work to achieve something akin to human-level abilities in computers as the most exciting artificial intelligence challenge of the next 50 years.

He points to the work of David Cope, a researcher at the University of California-Santa Cruz, as evidence that some aspects of the goal may be within reach.

Cope has spent years experimenting with software that can mimic the style of various composers including Bach, Chopin and Mozart. The software is so good that, in live performances, even music aficionados can have difficulty distinguishing between the human- and machine-composed scores.

Those worried about the rise of machines might take some solace in Cope's description of his work.

"I'm not sure that using computers as tools makes composing any the less a 'province of human beings,'" Cope said in an e-mail interview. "Most important for me is simply listening to the output, which, interestingly, makes me appreciate the 'masters' even more."

-30-

By Geoff Koch