Apocalyptic visions of autonomous computers or robots dominating the human race may be a tantalising prospect for film producers and script writers but the technological alchemy needed to produce such a being is only in its infancy.

But satellites that are capable of decision-making may not be too distant a prospect. The US National Aeronautics and Space Administration (NASA) is already developing artificial intelligence which will equip satellites to make their own decisions, as more crowded skies necessitate independent control.

According to Corkman, Mr Paul Morris, who is working with an international team of computer scientists at NASA in California on long-range research in artificial intelligence, control systems have become over-subscribed as space agencies have launched more and more smaller, cheaper and faster probes.

The use of artificial intelligence would obviate the need for much of the communication between ground stations and probes, and improve the rate of success of deep space missions which often suffer from the need for constant instruction on movements and paths from earth.

"We have developed a system called `Remote Agent' which can carry out tasks independently during missions and can adjust to changing conditions in space as well as react to problems which probes or satellites encounter there."

Last April, Remote Agent took over control of the Deep Space One probe for a week without a glitch, and now a modified version of the system is being considered as part of the control system for the next Mars mission, Marie Curie in 2001, which will be similar to the pathfinder mission of last year.

"It's not like Hal out of Space Odyssey: 2001 but it can deal with routine problems, like those encountered by the last Mars polar lander which lost a lot of time because instructions had to be sent across space from Earth to modify the original instructions given to the vehicle."

The mission to investigate Mars was hampered by the inability of the rover to react to problems and the mission lost days as scientists sent new instructions to stop the probe repeating the same flawed programmed movements.

The main advantage offered by conditional commands is that a vehicle or probe will only carry out the tasks if the conditions are right. This will avoid the scenario of the last mission when the vehicle got stuck on a rock repeating the same useless movements over and over again because it was programmed to do so.

"Mission managers are very reluctant to try out the new system on multimillion dollar projects as the technology is as yet unproven but it could be very useful for Earth observing and controlling satellites orbiting Earth."

As well as being able to control the engine to steer the craft into the optimum position, it can change the direction and position of the cameras to get the best angle, and make sure the view is not obstructed by objects such as asteroids or debris in space.

The system is by no means foolproof because it can only deal with problems or scenarios for which it has been programmed. But it can be programmed with conditional commands which allow it to react to different situations in different ways.

The hope is that the system will be used in a new Spacetech 3 mission which will see spacecraft fly in formation in space but the whole project is still on the drawing board. NASA budget cutbacks also mean it is never clear which missions will actually go ahead.

Mr Morris, a computer scientist specialising in artificial intelligence, believes that computers will ultimately match human intelligence. "In 50 to 100 years we will have very flexible capable systems approaching normal intelligence," he says.

"There has been slow but steady progress in the field but when you think that AI is only 50 years old compared to physics and maths which have been developed for hundreds of years, a lot has been achieved."

Mr Morris did a masters degree in maths-physics in University College Cork before going on to do PhD research in maths before switching to computer science at the University of California at Irvine. He worked as a consultant for Lockheed and as a consultant on Y2K before moving to NASA.

He believes that artificial intelligence will never replace humans on space missions as the sense of achievement involved is created by humans exploring a planet or travelling across a galaxy.

However, autonomous systems could prove very useful in a support role to human missions in space. Astronauts currently spend much of their time doing routine tasks. Scientists at NASA are developing more intuitive graphic user interfaces which with automation systems, like Remote Agent, could free scientists to concentrate on the experiments that are central to their mission.

In the context of the massive costs of space missions and taxpayers growing unwillingness to fund such projects, artificial intelligence is cheap to develop and to replicate.

There was significant initial interest from commercial and industrial sectors in developing artificial intelligence.

But says Mr Morris: "The problem was people made overly optimistic predictions when it wasn't clear what AI could actually do, and gave the business community unrealistic expectations."

The field then went through a lull in the 1990s as businesses pulled out as results failed to materialise. Involvement is now mainly confined to larger operations which have more money to invest in research and are more patient.

Computer giants like Microsoft and Bell are involved in research in the area although the results remain secret. Boeing recently used AI systems to design its new airplane the B-777.

Stanley Kubrick amazed the world with Space Odyssey: 2001 as he forecast a vision of a future with virtual people and a computer landscape. But how far away is the day when we could encounter computers that have a capacity to think independently?

"That question is for the far future as we are no where near the level of Hal, but the question of computers becoming self-aware is under the control of humans," Mr Morris says.

"People have a choice as to whether they make computers self-aware and if they do the whole issue of rights and slavery would come into the equation. Overall computers are probably more useful to people as tools rather than as independent thinking entities."

The prospect of your computer refusing to work because you have not given it enough holidays or lunch breaks is still a distant prospect.