Computer dilemma as scientist warns of `unfriendly' machines

Two visions of the future of computing were unveiled yesterday - on different continents and offering two different directions for the evolution of machines: increasingly fast, but stupid; or comparatively clever, but slow.

Experts say that eventually such systems could be defined as being "alive", and might even be used by governments to choose the best allocation of taxes, benefits and welfare policy. Within a generation, they could be more intelligent than people - though there could well be hidden drawbacks in that.

The first version of the future appeared in the United States, where the chip-maker Intel showed off a computer able to carry out 1 trillion (a million million) calculations each second. That is almost three times faster than the previous record-holder, built by Hitachi last year.

The $50m (pounds 30m) machine, custom-built for the US Department of Energy, will mathematically simulate nuclear explosions, predict weather and try to forecast natural disasters.

Yet the new machine - built by linking hundreds of Pentium processors, found in household PCs - is still "stupid": it does only what it is programmed to do.

Meanwhile, in the Science Museum, London, Professor Igor Aleksander was showing off an alternative vision: a computer that is conscious of its environment and reacts to it.

Professor Aleksander's system, called "Magnus", is an attempt to replicate consciousness by creating programs which act in the same semi-predictable way as neurons in the brain. Magnus can "explore" its world, which consists of pictures in the laptop computer where it exists, and express "hunger" - which is sated by linking to pictures of fruit. So far it has cost pounds 500,000 to develop the program, which is now effectively programming itself.

But experts are mixed in their opinion of the future.

"There's no point in making a computer that's exactly like a human being - there are plenty of people already," said Peter Cochrane, head of research and development for BT, adding that it would be good to have computers that could assess and pick the best of all options.

Present fiscal systems are hopelessly outdated, he said. Even "The Chancellor [of the Exchequer] is trying to move into the 21st century while using 16th-century management systems ... In 15 to 20 years we will have a new breed of computers with intelligence, that are `alive' - in that they will have births, deaths, and propagation. "

But some of those scenarios might not benefit us, according to Professor Kevin Warwick of the University of Reading, whose new book March of the Machines, published next spring, warns that computers of the future might not be friendly.

"There's an advantage in military, financial and industrial terms to having big, stupid computers," he said. "But in 10 years we'll have machines as complex as a brain. And then it's anybody's guess what will happen."