Computer with a brain prepares for first deep thought

SCIENTISTS believe they are on the verge of building a computer with common sense to match the science-fiction creations of Arthur C Clarke's 'Hal', in 2001: A Space Odyssey, and Douglas Adams's 'Deep Thought', in the Hitch-Hiker's Guide to the Galaxy.

They are predicting a breakthrough within two years in the attempt to create the philosopher's dream of disembodied rationality - 'artificial intelligence' in a computer's memory.

The nine-year-old project is near to achieving the point when the machine can 'learn how to learn' by browsing through electronic libraries.

Doug Lenat, director of the CYC project, run by the Microelectronics and Computer Technology Corporation in Austin, Texas, said the work is about to enter the 'critical mass' phase when the computer is able to use the common sense it has been given to build up its knowledge.

'The role of humans in the project would change from brain surgeons to tutors answering CYC's questions about difficult sentences and passages. We are just at the verge of it now.'

The CYC project has been likened to creating an electronic encyclopaedia. However, Dr Lenat and his team - nicknamed 'cyclists' - do not instruct the system to store ordinary facts but to understand simple and complex rules that help it interpret the ambiguities of the world.

'We have blasted through all the problems that have stalled artificial intelligence research for three decades . . . CYC will communicate like an intelligent entity,' Dr Lenat said. 'We are codifying common sense.' This enables the computer to clarify lingual ambiguities - for instance it knows a 'red conductor' could be a Communist musician or a red piece of conducting metal.

Research aimed at building an intelligent computer began at the end of the 1950s when 'classical AI' work was born. Attempts to date have failed miserably. The CYC project is seen as the last attempt by the AI 'classicists' to build a thinking machine.

Critics, such as Igor Aleksander, Professor of Electrical Engineering at Imperial College in London, believe that computer circuits based on the nerve connections of the brain - neural networks - stand more chance of creating a thinking machine.

'Those of us working on neural networks see the CYC project as the last throes of doing AI the wrong way. I feel the aim of the project is completely confused although it may have its uses as a sort of computerised encyclopaedia,' Professor Aleksander said.

But if Dr Lenat's prediction comes true, the day of a 'Deep Thought' tackling the question to the mystery of Life, the Universe and Everything may not be so far off - though 'the program may take a little time to run'.