Mobile Intelligent Agents
By John Geirland, Wed Dec 11 13:00:00 GMT 2002
Artificial life is poised to invade the wireless world...
In Michael Crichton's latest novel, "Prey," an artificial life computer program called PREDPREY enables countless malicious "nanobots" to run amok in the real world, bent on transforming everything in their path to a milky residue. Meanwhile, a less malevolent form of artificial life is poised to invade the wireless world - fish. Imagine blowfish swimming onto your mobile phone from someone else's wireless device in England or South Korea.
Arriving with a passport detailing its previous ports of call, this scaly cold-blooded visitor is a "mobile intelligent agent" created by Pasadena, California-based DALi, Inc. DALi (Distributed Artificial Life) is among a number of university and corporate research labs and startups actively seeking to extend agent technology to wireless devices.
An intelligent agent is a piece of software that does things on your behalf. Agents can send multiple emails to recipients on a Buddy List, shuffle the appointments on a calendar program, or make you believe you're trading instant messages with Austin Powers. Most agents work quietly behind the scenes and can be found in some surprising places, like the brake systems in new model cars.
Most agents live and execute their code on larger systems like servers and PCs. Handheld devices generally don't have the processing power to host most agents, though agents are getting slimmer and devices more powerful. For now, wireless IM is providing a bridge.
In an IM messaging environment, wireless users employ natural language commands to order agents to do things like get a number from a corporate phone directory or send notifications to network administrators. That saves a lot of headaches, given the interface limitations of most handheld devices.
Agent development companies, like Sunnyvale, California-based ActiveBuddy, are adding wireless connectivity to their development kits. "It's a short leap," notes Forrester Research senior analyst Charles Golvin. ActiveBuddy chief technology officer Tim Kay says the company's new connectivity service, launched last October, "allows our clients to deploy their existing interactive agents [to RIM pagers, WAP and SMS devices] without making any changes at all."
Meanwhile, university and corporate research labs are quietly developing infrastructure for a new generation of wireless agents. The Intelligent Software Agents Group at Carnegie Mellon University in Pittsburgh, Pennsylvania has developed a domain-independent toolkit for agent development called RETSINA (as in the Greek wine). Research professor Katia Sycara and her colleagues are building agents they hope will keep your car safely on the road and your social life securely on track.
Imagine you're driving from Washington D.C. to New York City to see a friend. The personal agent living in your car's telematics system sends an email to your friend's personal agent suggesting dinner. More agents are "spawned" that zero in on your shared dining preferences (sushi) and access Mapquest or some other location-based service to find a conveniently located restaurant. Still other agents oversee the whole process.
Of course, some fear agents could act like pushy concierges. What if an overzealous agent compares a user's calendar with their location (via GPS) and makes the snap judgment that the user won't make it to the opera on time? "I sold your tickets," the agent notifies the hapless users. "But look at the fine restaurant I booked for you with the money I saved on your tickets!" "A lot of these deployments will require an opt-in feature," notes Jonathan Prial, vice president of business development and sales for IBM's pervasive computing division.
Intelligent agents emerged from the field of artificial intelligence. In the early going, AI researchers made the mistake of trying to develop large agents that could handle a myriad of functions. Nowadays the focus is on building smaller, specialized agents that communicate directly with users or with each other. "A single user would need a variety of agents distributed in the infosphere," says Carnegie Mellon's Sycara.
This proliferation of agents in the "infosphere" presents a problem. How do you find the agent that you need - or how do agents find each other? Sycara likens it to trying to locate a plumber in a telephone book that lists only names and numbers. Agents need to somehow "advertise their capabilities," she says. Her group is collaborating with researchers at Nokia to develop semantically meaningful interfaces and descriptions of devices - a "Semantic Web" - that would make it easier for users and other agents to find each other.
Context is Key
Another issue that agent researchers are contending with is "context." "Interactions [with agents on wireless devices] will happen in context," says IBM's Prial. "Context in terms of location, application and transaction." Knowing the context of the transaction can simplify the delivery of information, he says, saving wireless users from having to click through a lot of screens.
Notification is another "context" issue. How does an agent know when and how to interrupt a mobile user? What if the user is in a theater or meeting? "If you use your calendar management system religiously," says Prial, "the agent can figure out where you are" and send a message in the appropriate fashion. (How many people are "religious" about keeping their calendars up to date?) "Agents have a number of ways to reach you, including sending a silent message," he adds.
The most intriguing development in the agent field has been the emergence in recent years of "mobile agents." Most agents are "static" and spend all their lives on a single device, like a server or PC. Mobile agents in contrast are nomadic, capable of traveling through the network in search of a host that has the processing resources or information it needs to fulfill its mission. (Important note: "Mobile" refers to the behavior of the agent, not the type of device it lives on.) DALi, Inc. co-founder Todd Papaioannou believes this quality of mobility offers a solution to the problem of scarce bandwidth.
Despite the promise of 3rd generation wireless networks, bandwidth is likely to remain a scarce resource for many wireless users, Papaioannou says. He likens Internet bandwidth to a vast circulatory system. The large arteries make up the Internet backbone, where mega-corporations and academic institutions operate. Corporate LAN's and home users dwell in the narrower tubing of the veins, while wireless users are left with the tiny capillaries.
"It makes sense [for wireless users] to create an entity [a mobile agent] that can travel out to the center of the network and take advantage of the fat pipes," Papaioannou says. Once a mobile agent is dispatched, the user can disconnect from the network and move on to other tasks. "Users want to minimize data transport back and forth because it's costly," IBM's Prial notes.
If anything is holding back the proliferation of mobile agents, it is security. The security issue cuts two ways. Companies and individuals are understandably squeamish about allowing an un-vetted piece of code to land on their machines and execute. (Isn't that just what a computer virus or worm does?)
Similarly, businesses may fear that a competitor will capture their mobile agents and steal the proprietary information they have collected. Which is why Papaioannou feels that gaming, not enterprise applications, will be the impetus for getting mobile agents on more devices - particularly Java phones and PDA's. (Many mobile agents are built with Java.) "There's less of a trust issue involved in gaming," he says. Which brings us back to the wandering fish.
Papaioannou - who has a PhD from England's Loughborough University and wrote his dissertation on mobile intelligent agents - founded DALi, Inc. with three friends in March 2000. Their goal was to build the largest mobile agent system on the planet. The result was DALiWorld, a "massively distributed aquatic virtual world." These are not like the animated flounder that roam on screen savers. These bottom feeders were spawned with the company's DALi platform. In true Darwinian fashion, Papaioannou and his partners selected the top performing swimmers for breeding and disposed of the rest.
Users extend this virtual ocean to their own systems by downloading the DALiWorld client (an agent environment) from the company's web site. The gaming element involves what people do with the fish - and what the fish do with each other. Different kinds of fish can inter-breed, produce new species and exhibit "emergent behaviors" - a subject of growing interest among the computer scientists who build agent societies and communities.
Whether fish, fowl or nano-beasty, Papaioannou believes mobile intelligent agents will be part of our wireless future. "The concept of sending something slightly more intelligent than a SMS message to someone else's phone is going to catch on," he says.
John Geirland is co-author of "Digital Babylon," a book about the online entertainment business, and writes about mobile wireless developments from Los Angeles.