Robots? No fears

how far science is behind the fiction.

SCIENCE fiction has never been short of robots. Writers have conceived a world where machines not only walk and talk but are pre-programmed with a portfolio of human weaknesses: they're fussy (C3PO in Star Wars), lazy (Bender from Futurama); even manic depressive (Marvin the paranoid android from the Hitchhiker's Guide To The Galaxy).

Now we have the cute/cringe-making robot, depending on your point of view. In Steven Spielberg's futuristic film AI: Artificial Intelligence, child actor Haley Joel Osment plays David, one of a new breed of cyborg aware of its own existence. He is programmed to love - but not to cope with rejection.

The film is set in the middle of this century. Does that mean science is about to bring us the first anthropomorphic androids? After all, we've been looking forward to a future where our every whim is carried out by automatons for 50 years. But we're still stuck with video recorders that cut off the end of the film and washing machines that turn our whites pink.

"Scientists involved in artificial intelligence are divided on the issue," says Nick Pears, lecturer in computer science at York University. "It's fairly well accepted that we can create robots that can behave intelligently to a certain level, and be useful to us in our every day life.

"When you are talking about robots having human emotions, feeling remorse when a friend dies; and intentionality, in the sense of knowing

whether it wants to do something; and consciousness, being aware of its surroundings - whether we can achieve this is debatable.

"My opinion is, in the long term, it is possible in principle."

Nick has seen AI: Artificial Intelligence and, as entertainment, he wouldn't recommend it. As for the science: "I don't think anything I saw in that film is beyond the realms of possibility. But it's not like it's right around the corner: it's not going to happen in my lifetime."

What about that other science fiction nightmare: robots designing ever-more intelligent replacements, until machines take over the planet and enslave us?

Calm yourself, says Nick. "The process of designing something requires an enormous amount of intelligence and cognitive abilities. You are not going to get robots designing other robots until they are close to human intelligence capabilities.

"Because that's so far away it is not something people should be concerned about right now. I have come across people who believe there should be legislation already to stop this sort of thing happening. But we should be more worried about a meteorite hitting the earth or a mega-tsunami after an island collapses into the sea."

A cyborg master-race may be some time off, but science has already made some remarkable breakthroughs in the fields of robotics and artificial intelligence. The two are not the same: an intelligent robot uses AI, but an AI system is not necessarily a robot. It might be a computer program that can "think" for itself.

Back in 1989 the first chess computer defeated a Grand Master. More recently, scientists have developed Pegasus, a computer that can "talk" to airline travellers and provide them with the right tickets for their journey.

In-car system has been developed that can drive your car for you along a motorway. Video cameras linked to a processor can monitor busy road junctions and call the emergency services if there is a crash. Robotically-assisted surgeons can make intricate incisions without hand tremors.

These are all examples of artificial intelligence in action today. But much of the work is at the prototype stage. "It's happening already," says Nick. "It's just this question of reliability that people are working on. The term we use is robustness to the variability of the environment."

For example, the in-car computer can drive you safely along an empty test track, but would struggle in traffic.

Nick has been interested in AI since he embarked on a PhD in robotics in 1985. His particular speciality is visual perception to guide robots.

For three years, he has been developing Arnold, a three-wheeled robot which is the size of a small dog, and has a passing resemblance to one too.

'Arnold' doesn't stand for anything, incidentally: they tried to think of an acronym to fit its name, but couldn't think of anything that didn't sound nerdish.

Arnold can find its way around a strange room without crashing into the furniture or anything else. It has a video camera eye, and several ultrasound sensors which emit pulses of sound and can measure the distance from an object by the echo.

This information is transmitted to a remote computer which sends back a signal to steer Arnold and determine its speed.

All this to make a robot do what humans take for granted: navigate a room without bumping into things.

Nick nods. "There's a saying in AI: the easy things are hard and the hard things are easy. What that means is things that a human being might find quite difficult, like playing chess to a particular level, is comparatively easy to automate in a computer.

"But it's very, very difficult to get a robot to move in a cluttered environment."

This only emphasises how magnificently powerful the human brain is.

As Nick points out, the sheer processing power involved in simply catching a ball is immense.

Not that robots are incapable of ball games: Nick and his colleagues are to enter RoboCup 2002, a five-aside football tournament between competing teams of droids.

Developing artificially intelligent machines in this way could bring all sorts of benefits for future generations. They could assist in everything from medical diagnoses to the housework.

But whether we will ever have robot servants is a moot point, according to Nick. "It's a social issue, whether you want a robot hanging about at home doing your tidying up or looking after your kids. I wouldn't want one of those things getting in my way. But then again, I have only just got a mobile phone."

Updated: 10:25 Friday, September 21, 2001