There is more to setting up business in Silicon Valley than dispatching a salesperson with a fist-full of business cards

Computers are remarkably stupid things and only look clever because of the software which makes them work. Now a Dublin research group has developed a way to make computers seem even smarter, so much so that they may have a role in the State's burgeoning teleservices industry.

Teleservices and international call centres are a growth area for the Republic, according to a spokesman from the IDA. It has 60 foreign companies on its books, employing 6,500 people and this is expected to reach 10,000 by 2000. "The number of companies is building all of the time. We have become the number one centre for call centres in Europe," the spokesman said. The 10,000 employment target wasn't expected to be reached before 2005, he added.

It has been difficult to keep up with demand for trained staff, however, despite Government support for special diploma and certificate courses which are training people for the teleservices industry. This is where the UCD group might be able to help.

Dr Barry Smyth of UCD's computer science department described his group's work as something of a "breakthrough" in artificial intelligence (AI). A measure of its importance is seen in the research team's recent publication in the leading international journal, Artificial Intelligence.

"The basic research that has fuelled the paper is a definite breakthrough," Dr Smyth said. "How long it will take for an all-singing, all-dancing version is more difficult. We need to translate the research into a marketable product."

The AI technique which drives the system is known as case-based reasoning, he explained. This is based on holding records of previous problems as a resource that can be used to find answers to new problems. Humans do this intuitively, drawing on memory to help us find ways around new hurdles, but computers can't without the right kind of software.

"Problem solving is such a difficult task that the computational cost of solving complex problems has been prohibitive, particularly if solving from scratch," he said. Yet many companies hold onto databases of old problems to assist human operators working on phone-in teleservices for their clients. The UCD work takes this a step further by enabling the software to do its own decision making and problem solving.

"The basic difficulty with developing good case-based reasoning systems is how do you detect the similarity between current problems and previous problems?" he said. "There can be a lot of difficulty detecting similarities."

The UCD system uses what the team called "adaptation knowledge" to help it recognise the similarities, for example the parallels in two very different tasks, sorting numbers in ascending order or descending order. If the system is shown how to do an ascending sort, it should then "know" how to solve the problem in descending order. The key was in detecting "the subtle matches between situations", he said, even when these seemed dissimilar.

Their modified software also improves as it goes along. "The more you learn about what you can and cannot do, the better your similarity detector can do. Your system begins to tune to your environment and the accuracy goes up," he said.

Many teleservices and computer help-desks use simple case-based reasoning, but not at this level. Dr Smyth, who carried out the research in conjunction with Prof Mark Keane, believes that it could change the future of the industry.

"I would predict that this sort of system will be the next generation in these teleservices companies," he said.

Prof Keane also believes that the industry will benefit, but so too will computer research. "This work is as important for what it says about Irish research in computer science, as it is for the teleservices industry." The Republic he added could be an "international player" in computer science research.

Dr Smyth envisages such a system providing a frontline for initial client inquiries, for example on a help-desk.

The most likely interface for such a system, he said, would be via Internet with the client typing in an inquiry and the computer coming up with one or a number of possible solutions. It would not be a replacement for human problem solvers he said, but it could make their lives easier. "It is relieving the load and the burden on the teleservices market." Problems handled by call centres were often just a matter of a "vocabulary mismatch", he said, and the UCD work provided an answer to these simple problems. "This bridges that gap."

The UCD team does not have its research in a marketable form, but are ready to do so. "We don't have a product yet but have the expertise to develop a prototype product quickly. There isn't much that has to go into developing a usable product."