Columns

Agents are abuzz these days

When i/s managers first saw the World Wide Web, most saw the promise of a standard WAN protocol, an instant GUI and easier application distribution. Of course, the Web is more than that. It is now that "great computer in the sky;" a fabulous store of information that knowledge workers scour incessantly; and the world's biggest client/server network. Quickly, the Web is becoming a massive distributed system that shifts the balance of power away from desktops just as desktops in the '80s shifted the balance away from mainframes.

Have you experienced this? Working as a technical journalist, I have discovered that it's become increasingly easier to go out on the Web to find some bit of information than it is to find that same bit on my computer, in my office or (sometimes) on my desk. This said, the Web as a resource leaves much to be desired. It seems a natural fit for agent technology, that longtime frothing adjunct to artificial intelligence (AI).

Software agents are intended, like most technology, to do your work for you. Information overload is the enemy. You program the agent, describing features that interest you. It sifts through a data stream and selects material according to your established profile. Established can mean static. Tastes and interests change. In terms of "what is hip," feature-based filtering has given way to collaborative filtering. In the latter format, a system makes recommendations to one user according to selections made by other like-minded users.

It's as in a honey hive, where individual bees pool their intellects to create a collective calculator. After scouting out food sources, the bees report their findings. In fact, they do a 'waggle' dance or cakewalk. Others are impressed enough to sample the source and report back. Eventually, a quorum is reached: "There's a good spot on the 'back 40.' "

The technology of software agents -- they could just as well be called filters or automatons -- has figured in some recent news. Witness Lycos' purchase of Wise Wire, the Carnegie-Mellon offshoot that, like a lot of AI firms, went through many marketing iterations before finding a conceptual home. Wise Wire started out as Empirical Media. Its early motto: To provide an online filtering service that organizes and delivers a super-personalized stream of information. At another point the motto was: Marketer of a family of products that apply machine learning and artificial intelligence to deliver filtered online content that continually adapts to users' personal preferences.

The lure of agent gold may soon attract Internet-hopeful Vignette Corp. It is reportedly positioning the next version of its Story Server product as able to deliver personalized, interactive content. That's not all. IBM now fields an Intelligent Agents Consulting and Programming Services Group that addresses what it sees as potentially lucrative (and usually Internet-related) knowledge management issues. Look for other agent providers, such as Autonomy, AgentSoft and Net Perceptions -- the latter backed by Bill Gates' old pal, Paul Allen -- to show up with alliances and new product news.

As usual these days, though, the big newsmaker is Gates' Microsoft. The company bought collaborative-filtering star Firefly Network Inc. in April. An MIT offshoot on the banks of the Charles River, Firefly started life as Agents Inc. with a mission to provide intelligent agent technology for 'personalized, unbiased product recommendations in the electronic marketplace.' The company built a popular Web site, but even though co-founder Pattie Maes hit #25 on Time's list of the 'Cyber Elite,' business was not brisk. However, Maes and company were smart enough to see the dark side of agents -- a Web site could garner and then abuse your personal profile -- and set up the means to ensure trustworthy, private information exchange. This, rather than a wish to foster collaborative databases, seems to be behind Microsoft's purchase.

Microsoft said its purchase of Firefly will hasten its move to implement products that meet the World Wide Web Consortium's Platform for Privacy Preferences Protocols (P3P). It certainly buys it an important seat on that protocol committee. Perhaps Microsoft will promote other aspects of agent technology over time, even if it is just to bury the ghost of "Bob," the Microsoft agent software that caused a comic sensation a few summers ago.

And why not? Some of the neatest stuff on the Web uses the filters. Amazon.com, for example, uses NetPerceptions' product to guess which books interest buyers. And Barnes and Noble's site uses Firefly.

While plenty of I/S hands must concern themselves with more mundane filtering technology -- the Web can be a window on some pretty ugly stuff -- it is worth keeping an eye on collaborative filters. As the information workload tilts Webward, such tools will become much more important.

About the Author

Jack Vaughan is former Editor-at-Large at Application Development Trends magazine.