Robots to get minds of their own

SCIENTISTS HAVE developed an "electronic neuron" that could in a decade lead to robots able to mimic the decision-making processes of the human brain - and so become autonomous machines.

The work was initially part of a project sponsored by the US military, raising the possibility that it might have led to "Terminator" robots like those depicted by Arnold Schwarzen-egger in the eponymous film. Those robots turned on their creators and waged war against them; these ones, it is hoped, will be dispatched into space to be explorers.

The key to the breakthrough, by researchers at Raytheon Systems and the University of Texas at Dallas, is an electronic model of a nerve, in which gallium arsenide processors replace the biological tissue.

The work, reported today in New Scientist magazine, is thought to be an important milestone in the attempt to develop an artificial nervous system. Computer researchers have for years been refining "neural networks", which use software to mimic the way in which a collection of biological nerves can sift different data. But the American work would produce a much faster machine because the "neurons" are built from hardware, meaning their "reactions" are inherent in the design, rather than being run as a software program, where each action is the consequence of following a set of instructions.

By putting many such hardware neurons on to a chip to form a "cortex", and connecting that to a number of external sensors, the artificial system could observe the effects of its decisions - and so be able to learn, anticipate and even become conscious.

Larry Cauller, a neuro- scientist who is the project's principal researcher at the University of Texas in Dallas, told New Scientist: "What we've done is take a biological approach to building such an autonomous machine."

However Dr Cauller is optimistic that he is not creating a race of Terminators. He said: "I can't imagine a brain or automaton that searches and learns from its environment becoming a soldier. I prefer to see it as an explorer. You could even hook one of these brains up to the Internet and let it explore a virtual medium."

An ideal application would be exploring planets with hostile atmospheres, where humans might not be able to bear extremes of temperature or gravity. First designs are already undergoing tests. "We hope that it will start learning from environmental stimuli early next spring," Dr Cauller said.