Artificial intelligence may heal computer networks

With that in mind, a University of North Texas professor and a team of graduate students have created an independent software program that can think, work and even reproduce its own workforce when needed.

The new technology uses artificial intelligence to help maintain control over the constant influx of data within a computer or network. The software, now being tested, uses intelligent mobile agents, or IMAs, to watch over computer networks.

IMAs are programs designed to move through a computer system, much like a computer librarian, to gather, control and organize data, and even conduct maintenance and make repairs.

"There are many issues facing technology because of its rapid expansion," explains Dr. Armin Mikler, assistant professor in the UNT computer sciences department. He says constant influx of information makes it difficult to manage.

Dr. Mikler says that the number of computers and users has grown much more rapidly in the past two decades than the number of people managing the data.

As the number of Internet users increases, Dr. Mikler says, computer networks are expanding at an unprecedented rate, creating problems for system managers and business users, who have trouble keeping up with the flow of information.

Dr. Mikler says that, conventionally, system managers have worked within the network, using whatever resources were available, to manage the flow of data.

But as technology increases exponentially, the manpower cost of doing that is too great.

Enter the IMA, which Dr. Mikler says "is just a representative doing a task on behalf of someone." The agents share tasks and communicate with one another, adapting to their environment as needed. But don't think they're going to make the human computer network manager obsolete.

"Just as computers didn't take away jobs – they only increased jobs – these agents will only enhance the technology we have available," Dr. Mikler says. "They're going to help make people's jobs easier and make them more efficient."

What makes Dr. Mikler's research so progressive is that it combines the artificial intelligence of software agents with mobility, which allows IMAs to move about inside the computer or network.

When too much information is traveling through one part of the network and causing a slow-down, IMAs can redirect data without human intervention.

Cliff Cozzolino, a graduate student serving as the lead person on Dr. Mikler's research team, says that the mobility is made possible in much the same way that a Web browser transmits data to a server. When a problem is detected by the program, IMAs are dispatched to fix it, using methods already adopted by the Internet for transmitting information.

"The program is actually constructed through very conventional means," Mr. Cozzolino says. "The basic concept is not novel, but what we are doing is moving those applications into a level that hasn't been used before. We are taking the mobility issue and applying it to artificial intelligence."

"What it means is that you have an entire population [of mobile agents] working on different problems at the same time," Dr. Mikler says. "It is like having a company that can dispatch workers to different places, but if one worker doesn't have the proper tools, he can call on another worker who has those tools.

"In that same way, if the agent doesn't have the ability to resolve a problem, it can reproduce as many of its kind as it needs."

Once the agents have resolved the problem, whether it's rerouting information to fix an overload or repairing software, they "go away" – either killing themselves off or merging with other existing agents, morphing into a single agent.

The biggest issue facing the project is security. The agents travel through a computer system in much the same way as a virus, so Dr. Mikler's research team wants to find a way to keep malicious agents from entering the system.

He says that his team is developing solutions based on existing computer security and that the agents are being tested individually on computers in the school's lab. Once testing is complete, the agents will be placed in a cluster of 16 computers linked to a high-speed network for real-world testing.

After that, the next step is to publish the results and gain attention from the technology industry. Without industry support, Mr. Cozzolino says, the research means little."We'll always move ahead with it, whether the industry supports it or not, because we believe the applications are endless," he says.

"In the future, you could use them to gather marketing information, get stock quotes and do research for you. We're already using that at a certain level on the Internet, such as conducting searches on the World Wide Web, but this takes it into the system level. It's not just applications; it is in an embedded form as system software, so it operates automatically and autonomously."

Dr. Mikler says the school's close proximity to Richardson's Telecom Corridor has encouraged researchers to do more than create an academic prototype. "Being here in the Metroplex, with so many technology-related companies so close to us, I think we have the responsibility to go beyond academic research and seek out applications for it," he says.

"Our challenge is to bridge the research with industry applications. We think because of where we are, we are ideally suited to collaborate with industry and work out the concerns and issues together. If we can do that, we can have a prototype of it out in the industry in two years."

Mike Crawford, president of the high-tech marketing firm M/C/C, says the added aspect of mobility creates some exciting prospects.

"I think that UNT is obviously very forward-thinking," he says. "This kind of technology is just now in its infancy. The ability to multiply is going to create an even more productive and efficient computer network than we're seeing today."

Mr. Crawford predicts that will eventually translate into improved customer service from the industry, which could diagnose and resolve problems before complaints are ever lodged.

"There are so many applications we haven't even dreamed of yet," he says. "By the time we see this on the market, I think technology will have advanced to the point where we can use it in ways that aren't even possible for us to think of right now."

But Kyle Tanouye, a systems engineering manager in Dallas, is skeptical that such technology is the best answer. Sometimes, he says, there's just no substitute for human intuition.

"In some cases, when you have a problem, there might be certain things you need to do before you reset a system that a program wouldn't know," he says. "Maybe you've just done something that needs saving that [the agent] isn't aware of. There's that element of human involvement that I think you need."

Paula Felps is a free-lance writer in Dallas.