Luis von Ahn wants to steal your brain.

Not all of it, mind you. Von Ahn just wants a piece of your time so he can tap into your brain's cognitive powers. You'll never get your time back, but he promises the whole process will be painless, even pleasurable.

	Doreena Balestreire/Pittsburgh Post-Gazette
	Luis von Ahn, a graduate student in computer science at Carnegie Mellon University has created a giant special-purpose supercomputer that uses human brains to do the computing.

All he asks is that you play something called the ESP Game.

It's a simple game -- players who visit the www.espgame.org Web site are automatically and anonymously paired up and shown photos or other images. Each player types in words that describe the image until, without seeing the other's list, they have a match. Then they are shown another image and the game repeats. The less time it takes to match words, the higher the score.

If the Internet were cable TV's Game Show Channel, this would be "The Match Game" and Von Ahn would be Gene Rayburn.

The players may not realize it, but the lists of descriptive words that they're generating could eventually be used by search engines such as Google to improve Internet searches for images.

They also are doing something that no computer program has ever managed to accomplish: analyzing an image and accurately describing it in words.

In effect, what von Ahn is creating with his game is a giant, special-purpose supercomputer that uses human brains to do the computing. And the 24-year-old von Ahn, a graduate student in computer science at Carnegie Mellon University, says this approach, which he calls "Stealing Cycles from Humans," could be applied to a wide variety of problems that are too great for any individual but also beyond the capabilities of conventional computers.

As he sees it, the "Matrix" movies have it all wrong. Machines would never tap people as an energy source, because people actually consume energy. But they might well tap people for cognitive abilities that machines can't duplicate.

"There's some meat to his idea," said his mentor, Manuel Blum, a CMU professor and a pioneer in the field of theoretical computer science. Producing word descriptions of images with the ESP Game is nice, of course, but the bigger idea is to entice people to cooperatively solve problems that defy electronic computers.

Adam Smith's markets
More recently, that logic was at the root of the "terrorism futures" market, a controversial experiment set up by the Defense Advanced Research Projects Agency. Part of a program called FutureMAP, the experiment was designed to see if investing in the likelihood of events such as suicide bombings or political assassinations could help predict unrest in the Middle East.

The experiment was of questionable taste, at best, and the agency scrapped it soon after it began to make headlines in late July. But as ghoulish as such an experiment might seem, many computer scientists saw it as a technically valid way to address an unsavory problem -- harnessing expertise about the possibility of future terrorism.

Von Ahn's approach is to use online games, rather than markets. "There are all these problems that you should be able to solve by playing games," he said.

Unlabeled images on the Internet is one. Only about 2 percent of all Internet images are adequately labeled, he said, which creates a problem when people use search engines such as Google to find them, or when blind or visually impaired people use special reading aids to help them navigate the graphics-heavy Web.

The Google engine can't evaluate the images themselves, so it looks for key words in text that appears near an image, he explained. "It's amazing how well that works," he said. Still, a Google image search for "car" returns, among other things, a map of Chicago. A search for "airplane" finds many images of airplanes, but also a snapshot of clouds as seen out an airplane window.

Players of the ESP Game could help fine-tune those selections.

If an average of 5,000 people played the ESP Game 24 hours a day -- and the most popular Internet games often attract many times that number of players -- von Ahn calculates that a one-word label could be generated for all 425 million images searched by Google in just a month. Six-word descriptions could be produced in six months.

For its part, Google is mum about what it's doing to improve image searches, or whether it even considers them a problem.

"We're always working to make the quality of our searches better," said Nathan Tyler, a Google spokesman. "But we don't talk about how we do it."

Tyler was unfamiliar with the ESP Game. "But this seems like an interesting approach," he added.

How popular the ESP Game might be is hard to say. In the past several months, about 400 people have played it and about half have visited more than once. One poor soul, desperately in need of a life, has played it more than 100 times.

Are you hot?

Von Ahn doesn't personally find it a stimulating game, but that doesn't seem a criterion for popularity. One of the most popular Web games of late, Word Whomp, "is an incredibly stupid and simple game," he contended, in which players form words out of a set of letters. Thousands play it every day. And he has seen fellow graduate students spend hours on the Hot or Not site -- www.hotornot.com -- rating the hotness of pictures of people on a scale from 1 to 10.

Determining whether a person is hot or not, it turns out, is something else that computers can't do. Similarly, a game that helped identify whether images are pornographic or not might be another way to "steal human cycles" for someone trying to filter out pornographic images from Web access.

As it happens, Web pornography played a role in the ESP Game.

For several years now, von Ahn has been the leader of the CAPTCHA project, one of Blum's research efforts, which has developed tests for determining whether visitors to Web sites are humans or software programs.

That's a big concern for Web portals such as Yahoo and Hotmail that offer free e-mail accounts. Software programs that automatically fill out registration forms can rapidly obtain thousands of e-mail addresses that can be used for sending spam e-mail. Telling the difference between legitimate human visitors and these software robots is a critical need.

The CAPTCHA tests are simple for humans to pass, but hard for computers. A typical test features a word with fuzzy or distorted letters, or words overlapping each other, or a word superimposed on a complex background; visitors to the site are asked to type a word they see. Yahoo began using the CAPTCHAs on its Web registration form several years ago; other Web sites quickly copied the idea.

But at least one potential spammer managed to crack the CAPTCHA test. Someone designed a software robot that would fill out a registration form and, when confronted with a CAPTCHA test, would post it on a free porn site. Visitors to the porn site would be asked to complete the test before they could view more pornography, and the software robot would use their answer to complete the e-mail registration.

It's not a practice that rapidly or easily overcame the CAPTCHA test, but the tactic of getting humans to unwittingly do cognitive work for a computer program inspired Von Ahn to develop the ESP Game.

The approach could be applied to other problems.

Security cameras, for instance, are cheap to install but expensive to monitor. By designing an online detective game that combined real security camera images with fake images that included interesting activities, players might keep an eye on all the images if they were rewarded for rapidly identifying suspicious activity.

Blum suggested another problem that might be solved with this approach: Internet searches concerning mathematics. Different mathematicians use different symbols to represent the same variable in equations -- what one labels X, another might label T. Humans can recognize that it's the relationships of the variables that matters, not the labels, but computers can't.

Blum doesn't know quite what the solution to that problem might be, but von Ahn could be the person to find it. A native of Guatemala, von Ahn had a mathematics background and an interest in theoretical cryptography when he entered the CMU doctoral program.

Though cryptography remains his primary research interest, his work on nonmathematical computer science projects, such as CAPTCHA and ESP, is taking up a growing amount of his time.

"Far be it from me to tell him he has to do some mathematics," Blum said. "Luis is a terrific programmer. ... He has more ideas than he can shake a stick at."

On the Web: ESP Game: www.espgame.org

The CAPTCHA Project: www.captcha.net.