The brain strikes again

Human beings are still the most intelligent agents around.

Published May 10, 1997 7:00PM (EDT)

Human sweat isn't supposed to be part of the information future. That's why we invented computers, and the Net, and fancy-pants infobot programs. Our software servants will tackle all the informational heavy lifting, searching and filtering and retrieving, while we of the flesh-and-blood persuasion kick back and click through the latest Microsoft sitcom on WebTV.

Sounds great, doesn't it? But before we get too starry-eyed, maybe we should rein in those cyberpunk dreams. There's a new fad sweeping the online marketplace -- human-based data retrieval. Got a question? Forget about clunky search engines, unwieldy Internet directories and stupid intelligent agents. Instead, go ask a person. Hop on over to answers.com or HumanSearch, where live operators are standing by to take your e-mail call.

Yep, the information revolution is already being counter-revolutionized. The entrepreneurs behind answers.com and HumanSearch are convinced that no matter how good computers get at giving us what we want, they'll never out-perform good old-fashioned people.

Both sites operate on essentially the same structure. After a short registration process, a user inputs a question. Part-time researchers then comb the Internet and other reference works to concoct an answer, which is e-mailed, within 48 hours, to the inquirer. Answers.com charges a fee on a sliding scale based on how detailed an answer you want. HumanSearch offers its services for free, planning to eventually recoup its operating expenses through site advertising or via other software products.

Both Clay Johnson, a University of Rhode Island undergrad who created HumanSearch, and Clement Izzi, the CEO of answers.com, concede that their businesses are staked right on top of a huge contradiction. They are well aware that a crucial component of the Net's seductive lure has been its promise to cut out the middle man, to provide access to the entire sum of human knowledge, to connect each and every one of us to the infonuggets of our desire. But by introducing a new layer between the individual and the raw material of information, answers.com and HumanSearch are essentially declaring that the Net doesn't deliver on that promise. Isn't that going against the digital grain?

"Absolutely," says Clay Johnson. "But the problem is there is just too much information on the Net. It's too hard to find what you want."

Granted. The Net is badly organized, always changing and a huge time sink. So is the world in general, which is why the practice of "information brokering" has been around for probably as long as people have been curious. But the Net is a digital environment, and from the beginning its advocates have claimed that it offers breakthrough potential for the alleviation of information anxiety.

Why throw human bodies at the problem? What about the software solution? What about better tools -- knowbots and searchbots and softbots, smart filters and news agents and autonomous avatars? All over the world thousands of computer scientists, in both academia and the corporate world, are frantically at work crafting software programs that aim to make sense out of information chaos. Answers.com and HumanSearch seem hopelessly retro next to this vast oncoming array of natural language processing artificially intelligent software. People looking up answers? How quaint. How archaic.

Johnson and Izzi shrug at the prospect of competition from artificially intelligent software.

"They'll put us out of business when they're successful," admits Izzi. But he doesn't sound very worried. Nor does Johnson: ''They'll never be as smart as humans."

Both men offer a refreshing counterpoint to the hype that prevails in the intelligent agent research community. Not only do intelligent agents not work very well right now, they argue, but the technology's prospects for being able to meaningfully satisfy critical human requirements in the foreseeable future are slim.

Johnson recounted testing one of the most highly touted agent products currently in the marketplace, AgentWare's AutoNomy search robot. He asked it to look for everything on the Web that related to HumanSearch. After searching, says Johnson, the AutoNomy agent delivered a list of Web pages that included the HumanSearch home page and "15 pages about the rebellion in Zaire."

So yes, there still are some bugs in the system. And it is far from clear whether those bugs will ever be entirely eradicated. It's one thing to design a successful chess playing program. That's the kind of problem that responds well to increased processing power. But to understand the subtleties of human speech? To know what a question means? To search the Net and retrieve only those left-wing critiques of Web-based business models that avoid neo-Marxist cant?

As Johnson observes, the day a machine can answer that kind of question, there probably won't be any need for humans at all. And that would be a sad thing, so maybe we should be glad that answers.com and HumanSearch are declaring that there is still a place for humans in the digital wilderness, that our blood and guts are still good for something.

But the fact that enterprises like these online information brokers are emerging is also a declaration of surrender, an admission that the Net is not the fundamentally empowering instrument we once thought it could be. Information, the cyberpunks enjoy declaiming, wants to be free. But once enough information is free, it starts becoming expensive again.

Clay Johnson and his fellow Rhode Island students believe that they can offer their service for free, that they can scale up as demand rises, and that advertising revenue will pay the bills. But there's a reason traditional information brokering has always been a costly service. There are a lot of people out there with questions, and answering all of them is a labor-intensive business. To keep on top of that demand, Johnson and his cohorts are going to require substantial investment capital. Humans -- at least the kind who can answer questions -- don't come cheap. Any business model that permits answering questions for free will be the true information gold mine. HumanSearch and answers.com are just first stabs at building businesses out of harnessing human brainpower on the Net. And progress is being made, incrementally, in software agent technologies. Perhaps one day, our digital butlers will free us from information toil and trouble. But for now, a wry thing has happened on the way to the cyberpunk nirvana. Look what the digital revolution has wrought: The best intelligent agent money can buy is once again a human being.


By Andrew Leonard

Andrew Leonard is a staff writer at Salon. On Twitter, @koxinga21.

MORE FROM Andrew Leonard


Related Topics ------------------------------------------