This year will finally be the breakthrough for intelligent agents.
January 4, 2001
Digital Valets Might Happen in 2001
by Kevin Maney, USA Today
OK, maybe you've heard that a thousand times. Like, you're going,"Oh, and nuclear fusion hot water heaters are going to be the next big Christmas gift."
But this agent stuff could happen. Microsoft says that in 2001, the company will finally make intelligent agents a mass-market reality. And if Microsoft says something, you just salute and say, "Yes, sir! Anything you say, sir!"
Although, even Microsoft is hedging. The concept of agents is so sullied by eternal disappointment, Microsoft doesn't want to call these things agents anymore. One term they're using is digital valet, which sounds like something Cary Grant might have used to fetch Excel files that keeptrack of his tuxedo trousers.
Intelligent agents - valets, whatever - are pieces of software that are supposed to be able to go out on the Internet and do unpleasant or tedious tasks on your behalf. They might make travel arrangements. They might sort e-mail and voice messages. They might, if they become very good, manage your entire team while you sit poolside and gulp cocktails with little umbrellas in them.
The idea of agents has been around for decades. Since at least the late 1980s, companies and research labs have often talked about them as if they were right around the corner. AT&T made a marketing video in 1993 showing ultra-efficient humanized agents carrying out orders and interactinglike digital versions of Radar O'Reilly from M*A*S*H. General Magic was a well-financed start-up in the early 1990s aimed squarely at commercializing agents. Carnegie Mellon University has agent research projects with names like Joccasta and Pleiades. IBM Research has piles of agent stuff going on.
Yet so far, agents are about as common as public sightings of Salman Rushdie.
We're starting to get some little tastes of agents. There was one over the holiday shopping season. Let's say you were foolish enough to promise your kids they'd get a PlayStation 2 for Christmas. For all the chance you had of coming through, you might as well have promised that Emperor Akihito would drop by with one. But if you were desperate, your best bet would've been to grab a PlayStation 2 autoscanner from Techbargains.com.
Once you had the autoscanner running, it would check Web retailers such as Amazon.com, Kbkids.com and Walmart.com every 15 minutes, looking to see whether a sudden shipment of PS2s landed there. If it did, a notice would pop up on your PC screen and, with a click, you could buy it. Clever and agentlike, but not very sophisticated. As agents go, it's about as brilliant as an amoeba.
The thing is, the notion of truly powerful agents is very appealing, which is why companies and researchers keep trying. If agents were to workas the visionaries see them, they'd be like a team of little assistants waiting inside your computer or out on the Internet somewhere. You could ask them to take care of anything, anytime. You wouldn't have to give them health benefits. They'd never whine.
Yet, alas, that's too much to ask of today's technology. Agents like that would require a ton of computing power, much better networks and agreements between industries and companies that would allow agents to roam freely around every computer network. Then, if all that could occur, agents would be so smart they would whine.
"I don't subscribe to the idea that we're even close to having 'your agent,'" says Craig Mundie, Microsoft executive vice president. "A critical mass of things would have to come together."
So for now, Microsoft is taking aim at a more limited kind of agent- one that sits on your PC and learns how to do small tasks.
One effort centers on research by Microsoft scientist Eric Horvitz.This agent would look at all the electronic messages coming to you. It would also have some idea of where you are and what you're doing. A camera on the PC could let the agent know if you're looking hard at the screen - probably doing work - or out of the room. It would know some things about your priorities (what kinds of messages you read immediately; which ones can wait) based on what you tell it and what it learns .
It sorts messages and decides what to present and when. "It's like having a really great butler," Horvitz says. "It balances the value of being bothered vs. the cost." Microsoft executives say Horvitz's agents will start appearing in products in 2001. A few other firms, including start-upWorldStreet, are working on similar products.
Mundie talks of a news valet, which he says is simple enough to make happen soon. It's more of an evolution from Web crawlers available today.This kind of agent would learn what you like to know about, then go prowling the Web's media sites for appropriate stories, pulling the pieces into a single basket on your computer.
Software pioneer Dan Bricklin likens the concept to a parking valet, "who does what you could do, but does it instead to save you some time - for a small fee or a tip," he says.
Certainly seems more probable than the older concept of the überagent. "We're seeing gradual progress toward agents in small steps," says Chris Meyer, vice president at Cap Gemini Ernst & Young. The valet, especially, "isn't that dramatic."
Anyway, Microsoft has some motivation for pulling it off. Its executives have said that they're not sure how Microsoft will make money on the company's dot-net initiative, which is supposed to be Microsoft's next great phase. Collecting zillions of small fees and tips sounds like a decent start.
So, you see, maybe, after all this time and effort, agents will actually take off this year. Hey, it's a new era. Anything long promised and never delivered now seems within reach, doesn't it? I wonder what's next. Laser guns? Time machines? Anti-gravity pills? Where's Joe Firmage when you need him?
Copyright 2001, USA TODAY. Reprinted with permission.