By the end of the film, humans' innovations have outlasted them. The earth is populated by what appears a race of super-intelligent androids.
Outrageously far fetched? Maybe not.
Some prominent real-world researchers who work in the field believe the film's robots - played by human actors with special-effects help - are a reasonable approximation of where robotics is headed.
But it's anybody's guess when robots might encompass human intelligence.
Currently, scientists working on humanoid robots - that is, robots designed in the form of a human - are focused on building machines that can understand and obey voice commands, not issue them.
"If the movie showed the level we're working on today, it wouldn't be a good movie," said Maja Mataric, a robotics researcher at the University of Southern California. Directed by Steven Spielberg and based on a project started by the late English filmmaker Stanley Kubrick, the film opens June 29.
Japanese firms Honda and Sony have already built humanoids that can walk, wave and make some rudimentary dance steps. Within a decade, the robots ought to begin handling their design purpose: caring for Japan's burgeoning elderly population.
When Honda unveiled its P-3 humanoid, a plastic-sheathed robot that looks like a slimmed-down Michelin Man, U.S. government agencies began funding humanoid robot researchers like Mataric, whose robot Adonis is learning to dance the Macarena.
"Humanoids are now in vogue," says Mataric. She expects to see humanoid helpmates sold to consumers within five years, and an all-robot soccer team able to defeat a team of humans by 2025.
"It's very exciting," Mataric says.
NASA, as well, is developing "Robonaut," a humanoid designed to handle dangerous space tasks such as space station repairs.
The design of humanoid robots involves myriad sciences, from biology to computer engineering, that progress at different rates. The idea that science can someday build a copy of a human elicits predictions that vary by hundreds of years.
At Brandeis University, artificial intelligence researcher Jordan B. Pollack is a pessimist. Pollack believes technology will prevent "A.I."-style robots from emerging for a century or more.
"We've fundamentally underestimated how good a programmer Mother Nature is," says Pollack.
For a human, the process of opening the eyes, glancing about, and understanding where you are is a small matter. Programming a robot to do the same thing is a monumental task.
"When you build a robot, it occurs to you how extremely complex these things are," said Cynthia Breazeal, a robotics researcher at the Massachusetts Institute of Technology.
At Carnegie Mellon University, robotics researcher Reid Simmons estimates robots will be smart enough in 25 to 50 years to hold intelligent conversations witpeople.
Further afield is the notion that human tinkering with computers and robots might unleash a superior race of immortal beings. A common science fiction theme, it has been pondered seriously by respected scientists such as author-inventor Ray Kurzweil and Bill Joy, Sun Microsystems' chief scientist.
To Kurzweil, author of future-predicting books on machine intelligence, humans are just a few decades from manufacturing robots such as those depicted in "A.I." - machines that will soon demand civil rights.
"Within 30 years we'll have machines that reproduce the full range of human intelligence," says Kurzweil, whose inventions include speech recognition software and a music synthesizer. "People will feel threatened by this technology."
With scientists now scanning and reverse-engineering the human brain, a computer that surpasses human capabilities will soon be possible, says Kurzweil. Place one of these computer brains inside a human-looking robot, and we'll have robots smarter than humans - robots capable of pursuing their own priorities, he believes.
As humans and machines merge, androids will express the full range of human emotion, thus making it irrelevant whether they're made of carbon matter, as are humans, or silicon, says Kurzweil.
Hans Moravec, a government-funded researcher at Carnegie Mellon University, thinks people will eventually be phased out by robots who perform human roles better than humanly possible.
"Rather quickly, they could displace us from existence," Moravec has written. He is not alarmed by the thought of humanity's demise. Rather, he believes helpful robots will offer us a comfortable retirement.
The future imagined by Kurzweil contains more than robots. He describes rich virtual reality environments used for everyday communication, such as business meetings.
Kurzweil predicts that microscopic robots, or "nanobots," could be placed in the human brain to augment human intelligence.
"This is so far in the future that it's science fiction," says Breazeal, whose own robot, Kismet, is the current closest thing to "A.I.'"s main character, a child robot named David.
Kismet is the first humanoid that can respond to human emotion by changing its facial expression.
These days, humanoids represent a fraction of robotic research. Most scientists seek task-oriented robots that, for reasons of efficiency, rarely resemble humans.
"If I want a robot that rolls around a factory floor, do I want a humanoid to drive a machine? Or do I build a machine that drives itself?" asks Simmons. "It's more efficient to build something for the task itself."
Such robots take myriad shapes, from wheeled machines built to explore Mars or underground sewer and gas pipes, to helicopter 'bots that can chase a fleeing vehicle, or welding robots that assemble cars.
"Real robots are kind of workaday, dumb systems," says Pollack.
It is already possible to build single-task robots that outperform humans. Fo instance, IBM's scientists were able to program the supercomputer known as Deep Blue to play chess well enough to defeat the world champion human chess player, Gary Kasparov.
"But Gary Kasparov does all these other things that Deep Blue will never be able to do," says Breazeal.
Not so far-fetched is the idea that humans can develop pet-like bonds with a robot. Some researchers are already working toward "domestic robots," says Breazeal.
"If you could buy a robot that gave you rewarding interaction, like a pet, but also helped your children learn French, you might want it," she says.
Breazeal's own robot, Kismet, with its repertoire of facial expressions, elicits pleased reactions from humans. "They don't treat it like a machine, they treat it like a creature," she says.
Sony's AIBO robotic dogs elicit similar reactions. The company has already sold thousands of the $1,500 robots, which can follow certain voice commands, make simple decisions and express emotions via lights and sounds.
Much of the optimism about advancing artificial intelligence is due to leaps in computational power that permit ever faster and smaller microprocessors.
But computers are just one facet of robotics. Other disciplines required to build a mechanical human - including power source and sensory technologies, natural language recognition, motors and pneumatics - lag behind.
"Cameras are not eyes. Human vision is much more complex," says Mataric. "And we don't know how to do skin. You can touch a robot in most places and it won't even know you exist."
And even the optimistic Kurzweil believes human emotions, especially love, are half a century away from being replicated by machines. Should that day arrive, says Kurzweil, machines will have become human.
In the movie "A.I.," the robot-boy David expresses an unrequited love for his human owner, all the while wishing he were "a real boy."
To Kurzweil, this is evidence of David's humanity.
"He's not biological," says Kurzweil. "But he should be considered a real boy if he can express that deepest, richest emotion we have. He should be considered part of human civilization."
© MMI The Associated Press. All Rights Reserved. This material may not be published, broadcast, rewritten, or redistributed