Last Modified: ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?



Homepage / Publications & Opinion / Archive / Articles, Lectures, Preprints & Reprints

The love machine - 'A.I.' stars a robot with feelings. How close is that to reality?
Star-Ledger, 28 June 2001
By Kevin Coughlin

In the 45 years since the term "artificial intelligence" was coined, the quest for thinking machines has given the world automated airline ticketing, a programmable puppy named Aibo, and Deep Blue, the cyber chess champ.

Tomorrow, Steven Spielberg unveils something more profound:

A machine that is all heart.

"A.I." is a "Pinocchio"-meets-"E.T." fable about David, a cyborg "boy" wired to love his adoptive parents forever. While sci-fi films like Stanley Kubrick's "2001: A Space Odyssey" abound with dire warnings of computers outsmarting us, "A.I." worries more about the cruelties we might inflict on them.

As David desperately struggles to become "real" and win his mother's love, Spielberg (with help from the late Kubrick) spins another Frankenstein yarn on the pitfalls of man-as-Creator.

But how close is the science to the fiction in this cautionary tale? When will the first Davids bound off assembly lines and into our arms?

Within decades, say futurists, confident the human brain will yield its secrets and serve as a blueprint.

Never, say skeptics, noting computers still can't think for themselves, let alone feel.

"It's too complex. We are not God," said Nirwan Ansari, a computer science professor at the New Jersey Institute of Technology. Philosopher-turned-computer scientist Jim Waldo agrees.

"We don't understand emotions at all. ...Emotions are much more complex than anything we've tried to model in a computer," said Waldo, who teaches at Harvard University.

Hugo de Garis is striving to build an "artificial brain" in a Brussels laboratory. The computer scientist predicts "loving robots" will follow in 50 to 100 years.

"In the second half of the 21st century, we will know how the brain works and can put the same principles in our artificial brains. Then AI will take off," de Garis said via e-mail.

Ray Kurzweil, inventor of optical scanners for the blind, contends the emerging field of nanotechnology will produce tools for deciphering the human brain. Then, AI can get emotional.

"In my view, our emotional capacity represents the most intelligent thing we do," Kurzweil writes on his Web site. "It's the cutting edge of human intelligence, and as the film portrays, it will be the last exclusive province of biological humanity, one that machines will ultimately master as well."

Kurzweil expects these creations will revere humankind. Bill Joy, on the other hand, finds the notion of cuddly androids a bit "squishy" and fears something darker.

"If we build humanoid robots, the first customer is likely to be the military, not a person in need of a child," said Joy, chief scientist for Sun Microsystems, and author of a Wired magazine essay on the dangers of unbridled technology.

Artificial intelligence, according to the Dictionary of Computing & Digital Media, is "software that makes decisions based on accumulated experience and information" with human-like functions "such as learning, adapting, reasoning, and self-correcting."

Critics put it another way.

"Anything that's impossible but can get funded is artificial intelligence," quipped Waldo, who also works for Sun.

Indeed, ever since the term AI was born, at a Dartmouth University conference in 1956, defining intelligence has proven as tricky as designing it.

The military was a driver back then. Now, corporations hope artificial intelligence will solve problems that defy easy yes/no answers. Scheduling factory workers, say, or squeezing the most from an oil field involves complex, subjective factors. Educated guesses count. Researchers have experimented with software that "learns" from experience, software that draws upon expert human advice, and software that betters itself by constantly winnowing algorithms in head-to-head tests.

But is that thinking? Waldo said it's like asking if submarines can swim. "They move through the water, but is that swimming?"

Even pioneers of AI differ. Marvin Minsky of the Massachusetts Institute of Technology insists artificial intelligence should be patterned after human thought processes. John McCarthy, who left MIT to start Stanford's AI lab in the early 1960s, prefers mathematical approaches.

The most enduring benchmark of machine intelligence remains the Turing Test, proposed in 1950 by a man many consider the father of computer science, the late British mathematician Alan Turing.

Turing, who predicted thinking machines would arrive by 2000, suggested a machine can be deemed intelligent if it fools an interrogator into thinking it is human.

Some say that milestone was reached at MIT in 1966, two years before the movie "2001" made the fictional computer HAL a symbol of digital evil.

Tossing out open-ended questions, a program named Eliza played virtual psychiatrist. Folks began pouring out their troubles, to the dismay of inventor Joseph Weizenbaum. The danger there wasn't any monolithic, all-powerful computer; rather, it was people placing too much faith in a simple machine.

Today, and the bar keeps getting raised. Computer chess epitomized artificial intelligence -- until IBM's Deep Blue clobbered Garry Kasparov. Now, some view the achievement as mere brute force, a number-crunching trick by super-fast computers. Likewise, advances in speech- and visual recognition programs are turning those once-lofty pursuits into basic engineering exercises, said Waldo.

Emotions aren't the biggest challenge for AI programmers, Minsky contends.

"In my view, emotional states are just different ways to think. The really hard part will be getting machines to be able to do commonsense thinking," said Minsky.

A Texas firm, Cycorp Inc., is trying to talk sense to a computer named Cyc. Researchers are programming its encyclopedic memory with kid stuff that confounds most machines: Trees are usually outdoors, people stop buying things when they die, glasses of liquid should be carried right side up, and so forth.

Yet even if Cyc understands humans, it will have a face only a motherboard could love. MIT is perfecting a cuter machine, Kismet, that responds to human cues like a pet. Kismet is no match for Spielberg's Teddy bear -- a "super toy" that vies with Haley Joel Osment's David as the real star of "A.I." -- but at least it's a nod toward tomorrow's touchy-feely robots.

Just be patient, counsels Peter Cochrane, former tech guru for British Telecom.

Like a machine.

"There are no limits," said Cochrane, a consultant for ConceptLabs of California. "We are still at the crystal set stage of AI, and even when we progress to the cell phone level, we will still have only just begun!"

Kevin Coughlin covers technology. He can be reached at kcoughlin@starledger.com or (973) 392-1763.

All materials created by Peter Cochrane and presented within this site are copyright ? Peter Cochrane - but this is an open resource - and you are invited to make as many downloads as you wish provided you use in a reputable manner