Skip to Content

On January 28, a day of calm seas off the California coast, computer scientist Jim Gray left San Francisco on his 40-foot yacht Tenacious and made for the Farallon Islands, 43 kilometers beyond the Golden Gate Bridge, where he planned to scatter the ashes of his recently deceased mother. He failed to return that evening.

For the next four days, the U.S. Coast Guard searched the ocean around the Farallons but found no trace of him. Gray’s friends and colleagues, however, refused to give up. A technical fellow at Microsoft and a pioneer in the development of database systems and transaction processing, Gray, 63, was one of the most beloved figures in the computer science community. Executives at Amazon, Sun, Oracle, Google, Microsoft, and other companies organized an intense private search, even enlisting a plane owned by NASA–a close cousin of the U-2 spy plane–and a satellite operated by mapping company DigitalGlobe to collect thousands of new images of the areas to which Tenacious might have drifted.

Despite all this firepower, though, Gray’s friends knew they’d need outside help to analyze the images they accumulated. So engineers at ­Amazon divided the images into tiles, each showing a 300-by-300-meter square of ocean, and on February 2 they uploaded the tiles to Amazon Mechanical Turk, a website where people can earn micropayments in return for completing quick tasks–such as recognizing objects in photographs–that are difficult for computers but easy for humans. More than 12,000 volunteers spent five days scanning 560,000 images for blobs of white pixels that might be Tenacious. They spotted a candidate, but planes dispatched to that area found nothing.

Gray’s family called off the search on February 16, and his disappearance remains a mystery. But the massively distributed, Internet-based search for Tenacious, probably the largest effort of its kind in history, stands as a tribute to Gray and as a powerful example of the emerging technology of “human-assisted search.” Few of the technology’s applications are of the life-and-death variety. But having followed it for some time, I believe that it will soon become pervasive, and that it will dramatically change our assumptions and expectations about the search process and about the types of work that can be accomplished using the Internet. In contrast to the Web’s famous disintermediating effects in commerce, human-assisted search is a form of “reintermediation”–an acknowledgement that software isn’t always king, and that sometimes it helps to have a middleman.

Google’s remarkable success at taming the jungle of text-based Web pages fostered the dream among some researchers that all digital information could be indexed, organized, and comprehended algorithmically–that is, using software alone. If this dream ever comes true, it will be far behind schedule. Meanwhile, entrepreneurs at Web companies such as Amazon have begun to show how the brainpower of thousands of Internet users can be harnessed for specific tasks that remain beyond the capabilities of software.

Mechanical Turk, named after an 18th-century automaton that supposedly played chess but actually concealed a human chess master, is at the forefront of this trend. The idea is simple: somebody with a job to be done, such as transcribing a podcast or proofreading a contract, enters the details into a Web page at Amazon (to do this, however, a person needs technical savvy or the help of a Web developer). Mechanical Turk outsources these so-called human intelligence tasks (HITs) to willing workers across the Internet, who earn a small fee for each one they complete.

HITs can be both boring and touchingly human. (A Pasadena family whose Yorkie had been abducted offered workers $0.10 for every time they posted notes about it on forums, message boards, or MySpace pages.) The search for Jim Gray, however, points toward a significant role for the technology in the future. Imagine armies of PC owners reviewing airport security videos from around the country for the face of a single wanted fugitive, or scrutinizing telescope images for signs of dangerous new near-Earth asteroids. The Web provides tools for interaction that are turning people’s pattern recognition skills into a valuable commodity.

The “people-powered search” company ChaCha is another case in point. Since last fall, the company has recruited 30,000 live “guides”: mostly retirees, college students, and work-at-home moms who labor at their leisure, spending as little or as much time as they like sharing their expertise on the best Web resources in various topic areas. (Guides, who must be invited to join ChaCha by other guides, can earn $5 to $10 per search hour.) The free search service begins with a text box for searching ChaCha’s traditional Web index. If a regular search doesn’t turn up satisfactory results, a user can click a link labeled “Chat Live with a Guide,” which sends the visitor’s query to the appropriate human. Once paired with a guide, the visitor receives instant messages containing greetings and, sometimes, requests for clarification. The guide then selects five or ten promising links on the subject and sends them back to the user’s screen, along with the same keyword-related ads that show up beside instant search results and are the company’s main revenue source. The results are also added to the ChaCha Web index, which consequently grows in quality over time.

A few dot-com-era companies, such as Webhelp.com, tried to market human-assisted search services and failed. But in the era of MySpace, ­YouTube, Skype, and instant messaging via phone and PC, people are more ready for the concept of working with a live person online; in fact, they’re thirsty for a little human interaction and human wisdom on the Web, at least in the eyes of Brad Bostic, ChaCha’s president and chief operating officer. “It’s both technology and culture that are changing,” Bostic told me. “A few short years ago, e-mail was the standard mechanism for interacting over the Internet. Now people turn to instant messaging and other outlets, not only to gain information but to gain social fulfillment.”

I’ve found ChaCha’s guides to be consistently pleasant and straightforward, even if they do sometimes fail to scratch my information itch. To one of my queries, about irrigation for desert gardens, a guide named Navindra responded frankly, “Umm, I am telling you from up front that your info seems a bit hard to find.” No worries: Navindra soon transferred me to Fabrice, who was able to locate two makers of drip-irrigation systems. The entire interaction took 22 minutes. I could probably have found the same information faster on my own, using Google, but it certainly wouldn’t have left me with the same sociable glow.

I expect that ChaCha searches will go faster and produce better results as the guides gain experience and as the company improves the tools it gives them for plumbing the Web. And ­Bostic’s point–that the average netizen is increasingly comfortable turning to fellow users for information–is part and parcel of the explosion in “social computing,” manifested in the great popularity of other user-driven reference sites, such as Wikipedia and Yahoo Answers, and in the proliferation of Internet-based instant messages, which now outnumber voice calls.

Amazon Mechanical Turk, ChaCha, image-tagging startup Polar Rose, collaborative search engine PreFound.com, and other new human-dependent Web ventures are significant because they are the first to exploit a new economic phenomenon: Internet piecework. Humans, it turns out, are even better than computers at completing some big ­information-­intensive tasks such as indexing the Web or searching satellite photos for lost vessels–as long these big jobs are broken into thousands of small ones and distributed to willing workers. That’s what the new online tools do, with great efficiency. And companies are already discovering that it doesn’t take much to recruit workers–just a chance to earn a few extra bucks, in ChaCha’s case, or a humanitarian urge to help in a crisis, in the case of the Jim Gray search.

But whether brokering this type of piecework can become a business big enough to meet the expectations of Wall Street and the venture capital crowd remains to be seen. For now, the companies enabling human-assisted search are putting their faith where dot-com entrepreneurs put theirs–in the Internet’s ability to aggregate millions of users, and in Web software and hardware that can process a continuous flood of transactions swiftly and cheaply. “How do you make many small things add up to a big thing? By making your system amenable to ­handling lots and lots of them,” says Peter Cohen, director of the Mechanical Turk project at Amazon. “We wouldn’t be doing this unless we thought there was going to be a business here for us.”

Computerization and the productivity gains that go with it have plenty of unintended side effects, from longer unemployment lines to the voice-­activated phone menus that make it hard to get assistance from a person at your bank or cable company. But the next time you look for help on the Web, it just might come from a human.

Wade Roush is a Technology Review contributing editor.

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.