A Reasonable Expectation: Computers That Can Think
Three decades after a handful of engineers at MIT began the quest for a computer that could think like a human being, “smart” machines have emerged from the realm of fantasy into the American marketplace.
Experts in the field say that emotional computers capable of anger and ambition can no longer be pooh-poohed as the stuff of science fiction.
“There is no reason to think . . . those things are impossible,” said Marvin Minsky, a professor at the Massachusetts Institute of Technology and an expert on artificial intelligence, as the field is known. “It’s just a matter of time.”
Arthur D. Little Inc. estimates that “thinking” computers--that is, machines that imitate the way the human brain reasons--will constitute a $126-billion business by the year 2005 and that machines with electronic brains will be a part of almost every car, home and office.
“Ten years ago, this whole field was out there in the cold, in the woodshed, with researchers begging for grants,” said Philip Cooper, chairman of Palladian Software Inc., a growing artificial intelligence company. “Now, artificial intelligence is being used commercially for the very first time.
“It is still quite primitive. It is not divinity in a box. But it’s pointing at what’s coming down the road.”
That road has taken research from the back-room laboratories, in which it began, into the nation’s most prestigious scientific centers. For example, the University of Illinois at Urbana-Champaign in October announced plans for the $40-million Beckman Institute, which will research the processes of the human brain and how they apply to computers.
At the same time, the Defense Department is spending $600 million to develop “smart” weapons systems, including an unmanned tank that can “see” with a television camera in order to navigate across battlefields.
On a more personal level, Cooper’s Cambridge-based Palladian Software has just introduced a $95,000 computer software program called Financial Advisor, which the company says has the analytic abilities of many high-priced executives holding master of business administration degrees.
Using Financial Advisor requires no special skills, Cooper says. All a user needs to do is supply the machine with information--reams of it, in some cases--about a company and its plans, say, to build a new plant, introduce a product or invest in a risky prospect.
The program, written by eight Ph.Ds from MIT’s Sloan School of Management, is not without a sense of humor. “Do you believe your product enjoys a competitive advantage others do not?” it queries users considering a new market entry. If the answer is yes, the computer screen flashes, “Shall I tell you about Japan and Taiwan?”
After all the information has been fed into the computer, it tabulates it, analyzes it, makes graphs, factors in inflation, taxes and other economic elements and then makes recommendations about a course of action.
Can’t Make Decisions
“What it cannot do is make the decision,” Cooper said. “That’s what the manager must do. This won’t replace people. It’s a tool to free managers to use their talents for more creative tasks.”
The Palladian program, like other artificial intelligence machinery, is different from an ordinary computer because it digests information with an enormously complicated language called LISP.
LISP computes symbols rather than numbers and letters and is punctuated with parentheses, leading industry insiders to joke that LISP stands for “lots of idiotic stupid parentheses.” It actually is a contraction of List Processing.
Financial Advisor’s potential has made believers--and buyers--out of 25 major American companies, including Coopers & Lybrand, Texas Instruments and Ryder Systems Inc.
“It’s like having a team of experts at your disposal,” said Michael Gilboy, director of strategic business planning at Coopers & Lybrand, a large public accounting firm based in New York. “It answers some questions I wouldn’t even have thought of asking.
Equal to an MBA
“When I give it a problem, I don’t have to worry if it handled the tax treatment right or depreciation correctly,” Gilboy added. “I think it could replace an MBA. . . . Why not?”
Although Palladian is selling corporations the first commercial artificial intelligence with widespread applications, about 600 other firms, many of them 10- to 15-person operations, are developing custom-made “thinking” computers for individual clients and companies.
Ford Motor Co., for example, recently signed a $14-million agreement with Carnegie Group Inc. of Pittsburgh and Inference Corp. of Los Angeles for development of computer programs that will be able to approve credit applications and diagnose brake systems.
Westinghouse Electric Corp. uses an artificial intelligence system to monitor steam turbines, alerting managers when something is going wrong and suggesting solutions.
And the Artificial Intelligence Applications Center at Arthur D. Little, an international consulting firm based here, is working on a computer for the National Aeronautics and Space Administration that might be able to predict and analyze tornadoes three hours before they strike.
“Lately, we’ve had a lot of people coming to us and saying, ‘Hey, what’s artificial intelligence? We have to get involved in it,’ ” said Marla E. McDonald, an artificial intelligence specialist at Arthur D. Little.
Solving Everyday Problems
McDonald said that the company has decided to leave behind the “what-if” side of artificial intelligence to create systems that people can actually use to resolve workaday problems.
“We are very application-conscious,” she said. “We won’t research (artificial intelligence) just for (its own) sake.”
The message is similar at Symbolics, at five years old the rich grandfather of the artificial intelligence industry, with sales of $69 million in its 1985 fiscal year.
About 30% of Symbolics’ business is for the Defense Department and NASA, but most of its customers are Fortune 500 industrial companies, such as General Electric Co., Bell Labs and Honeywell Inc.
“There is no question that most of our customers are still in the development stage,” said Nan Boyd DuCharme, head of investor relations for Symbolics. “But this was the year . . . we began to see some practical applications.”
Palladian’s new software, she added, “was a breakthrough . . . . It brought artificial intelligence to the common man.”
Fear of Computers Cited
Experts inside the artificial intelligence industry note with some consternation that the public still fears the notion of computers that have the ability to reason.
“Companies don’t want their clients to know they use artificial intelligence,” McDonald said. “They are afraid somehow the public will think there is a Wizard of Oz pulling the strings.
“People have a vision of ‘Star Wars,’ of a computer running wild, of computers being able to reproduce themselves and rule the world. They have a vision of comic book and cartoon concepts they have built up over the past 30 years.”
McDonald and many others find those “cartoon concepts” absurd, but MIT’s Minsky, considered the founding father of artificial intelligence, thinks the door to the future has opened wide enough to prove anything is possible.
Machines That Love, Hate
He predicts that machines will be able to be programmed for feelings like love, fear, hatred and ambition.
“I recommend fear,” he said. “If the chimps had known about people, they would have feared them, too.”
Cooper’s vision of future machines is less threatening.
Computers, he says, “probably will never be able to be original. They will never make a moral judgment or be artistic or musical or creative. That will always be the difference between people and machines.
“But the more (that) machines free people from the more routine kinds of intelligence, the more it frees them to do the original and creative thinking that pushes mankind forward.”
