Nils Nilsson, a retired Stanford professor is the author of ÏThe Quest for Artificial Intelligence'. Mail Tribune Photo / Jamie Lusch - Jamie Lusch

Artificial intelligence

The victory of IBM mega-brain Watson over human champions in last week's "Jeopardy!" may have sparked some concern over artificial intelligence, but one of its pioneers says there's no need to fear — it's already here.

Sort of.

Many daily functions, such as banking, wouldn't be possible without AI, says Nils Nilsson, author of "The Quest for Artificial Intelligence: A History of Ideas and Achievements" (Cambridge University Press, 2010) and now retired in Medford.

When you verbally tell Siri, a smart phone application, "need reservations at good Italian restaurant in Ashland at 7," you're asking computers in about five different corporations to evaluate your intent, "parse" your words, search ratings, confirm the table and show you a map — and that's AI.

"Siri is not an instance of AI 'taking over,' " Nilsson says. "But, it does have many capabilities that are parts of AI. I don't believe that AI will 'take over.'

"We don't understand enough yet about how the human brain works to be able to duplicate most of what it does in programs," says Nilsson.

Nilsson, 78, is a Ph.D. graduate of Stanford University and also a retired professor and former chairman of its Department of Computer Science.

Working for Stanford Research Institute starting in 1961, Nilsson built neural networks and developed pattern-recognition software that enabled its robot, "Shakey," to reason and make choices for itself, Nilsson says.

"It could perceive its environment with television and it had touch sensors, so when it bumped into something, it could respond," he says. "It could make models of its environment and plan to achieve a goal, such as pushing an obstacle out of the doorway to get to the next room."

Nilsson says Shakey could monitor the execution of its plans, recover from errors and figure out other strategies.

These AI functions are found in computers and such online tools as Google and spam filters, he says.

"Many people are not excited to see machines do things that we think only humans can do," Nilsson says.

When thinking about AI, people have to realize there are many types of intelligence — and AI can do only some of them, he says, noting "the same goes for any human being. We're good at some things and not others."

The IBM supercomputer Watson, which recently won three times more money than the best human contestants on "Jeopardy!," has "the ability to estimate the likelihood that different sources answer a question, using massive amounts of data," but it doesn't have all the kinds of intelligence humans have.

"I don't know when (computers) will achieve that," he says.

Computers can perform face recognition and convert speech to text, but they're not perfect yet, and they lack the human gift of common-sense reasoning — how we "just know" some things, but don't know how we got there, he says.

Will computers fulfill the fears of science-fiction writers, growing as smart as we are and taking over?

"I don't think that's going to happen," he says. "In a sense, computers already have taken over a large part of our lives, just as the automobile before it has done. We couldn't do without computers now. It's built into our lives. We couldn't even do banking without computers."

Nilsson and others delved into AI starting 50 years ago, at a time when it was pure research, the applications weren't known and researchers were following their profession's dictum, "If we build it, we'll understand it."

AI has been researched, first, because of potential commercial applications, including software sales, and secondly, he says, because science learns more about neurodynamics of both human and computer brains.

Nilsson is finishing a book on beliefs, and how we get and evaluate them by looking at (or not looking at) the evidence, consequences and explanations. Some beliefs are reached by skepticism and some because "beliefs are like clothes — we like some because they feel good."

Another researcher in the early days of AI, Fred Tonge of Ashland, helped develop software and applications for assembly lines. He also says he's "not frightened" that AI machines will take over the world.

"In some dimensions, they already do exceed us and will continue to expand," says Tonge, 80. "I have no idea if they will follow Isaac Azimov's three laws of robotics (about not harming humans). I'm always amazed how accurate science fiction turns out to be."

John Darling is a freelance writer living in Ashland. E-mail him at

Share This Story