To Infinitude and Beyond

January 31, 2012
 
Bright ideas
Bright ideas: In his first two weeks at Carnegie Mellon, von Ahn was awarded a MacArthur "genius" grant and named one of Popular Science's Brilliant Young Scientists of 2006.Justin Merriman
by Sally Ann Flecker

>
How is computer scientist Luis von Ahn ’00 taking the Internet to where it’s never gone before? By tapping into its most powerful resource: the millions of people who use it every day.

The machines—those great, solemn machines—moving in persistent rhythm through the rituals of their work. That's what the boy breathed in, whether he knew it then or not. He spent most of his childhood playing here—a candy factory in Guatemala owned by his mother's family. Even the apparatus with the least exciting job—the one that wrapped the cough drops, for instance—was a marvel to him. Sometimes he would look at something—a gummy bear or a marshmallow—and try to work backwards to the system that could produce such a thing. The process was always very clever, beyond any mechanism he could imagine.

You might think the boy who notices such things will grow up to be an engineer or an inventor. And he does, in a way. But what is important here is how he learns to twist and turn his ways of thinking into something every bit as clever as those enigmatic machines. Because that's who the boy becomes— someone who dreams up things beyond what everyone else can imagine. Great minds may think alike. But the greatest minds think in completely unexpected ways. They're the ones who take you where you had no idea you wanted to go. When you get there, you say, of course, how could it have been any other way? And that is exactly what this boy will do.

When Luis von Ahn '00 gives a talk, he'll show on the projector screen the kind of squiggly, hard-to-read word you often have to type to gain access to a site on the Internet. "How many of you have gone on a website where you've been asked to read a distorted sequence of characters like this?" he'll ask his audience. A few hands will clap. Next he asks, "How many of you found it really, really annoying?" There'll be some laughter. Then he'll deliver his punch line: "Okay. Outstanding. So. I invented that." He's got the timing of a standup comedian delivering a monologue, perhaps not what the audience is expecting when they come to hear one of the leading computer scientists of our day.

At thirty-two, he's still ever-so-slightly baby-faced. His Latino inflection is easy on the ear. And oh, by the way, he's the guy who is changing the way that millions of people interact with the Internet.

The truth is he's very down-to-earth for a person who solved one of the biggest problems of the Internet in 2000 as a first-semester graduate student at Carnegie Mellon University. That was when Yahoo called on his Ph.D. adviser, eminent computer scientist Manuel Blum, for help.

The Internet service was in a quandary. Spammers were using "bots," automated programs, to sign up for thousands of Yahoo's free e-mail accounts every minute. Once the accounts were set up, they were used to send hundreds of millions of unwanted email messages. Yahoo asked Blum how that could be stopped.

Blum and von Ahn put their heads together and came up with the idea of creating a test that would distinguish humans from computers. It had to be a test most humans could pass and computers couldn't. And that was a little tricky. It meant a computer had to administer and grade a test it could not perform itself.

They started brainstorming. Most of the ideas were "crappy," von Ahn says, characteristically blunt about his own work. One idea did have merit. Knowing that computers can't identify images, Blum and von Ahn designed a test in which humans were shown several related pictures and asked to identify a word that described them all. But it turned out people could come up with too many unexpected answers. A horse, a guinea pig, and a monkey could be animals, or maybe pets.

Finally, they landed on distorted characters. People are pretty good at discerning characters, even when they're twisted and stretched like the Sunday comics on Silly Putty. Most computers are lousy at it. And so Blum and von Ahn created those funky characters that are familiar now to anyone who's signed up for a listserv or booked an event ticket in the past ten years. They named the program CAPTCHA—Completely Automated Public Turing Test to Tell Computers and Humans Apart. It was an elegant solution to Yahoo's problem—and a landmark moment in von Ahn's career. You can go a lifetime without having an idea that pure and revolutionary. And von Ahn was just getting started. He started thinking bigger.

The boy in the candy factory is also the kid who didn't get what he wanted. Which turns out to be his great good fortune.

Game changer
Game changer: Von Ahn wanted a Nintendo but instead got an education.

What the boy wants, when he is eight, is a Nintendo. What his mother comes home with is a Commodore 64—an early computer that you plug into your television. "I hear you can play games on these," she tells him. "This is the one you get."

And you can play games on it, but first, you have to learn how to use a computer. In this case, that means learning the BASIC programming language. Keep in mind that he is at an age when most kids are learning how to multiply two-digit numbers, read chapter books, and follow the illustrated directions for a Lego set. But the boy takes on the challenge. He reads the manual and gets started. Then he goes out and buys some magazines to learn more. When he knows enough about the computer to play games on it, he talks his mother into buying some. After about five games, she decides they're too expensive to keep buying. That's it, she tells him. We're done.

But he is not ready to be done. He figures out how to get rid of the copy protections so he can exchange games with the other Commodore 64 enthusiasts that he's met in town. By now he's all of ten years old. It's clear that he's got a talent. His mother buys him an IBM PS-2, the kind of computer that is being used in offices and at universities. It's a serious computer, and he is serious about it.

But there is no World Wide Web, no discussion forums to consult. There is a tiny bookstore in Guatemala City. But even when he can find someone to drive him there, it almost never has the book he is looking for. So most of what he learns is because that's what's at the bookstore. Once again, he's not getting what he asked for. He's getting the answers to questions he couldn't have known to raise.

With the success of CAPTCHA, von Ahn wanted to take on similar projects, even though nobody was working on things like it. "It was risky because complete creativity is not highly rewarded in academia," he says. "As much as you want it to be, it's not. You want to ride this very nice wave of being somewhat creative, but not too creative. So I think it required some guts to do something that was a little more outside of the box than normal."

His outside-of-the-box question was this: Was there a way to get people to solve problems that computers cannot yet solve?

Take, for example, the problem of recognizing images. Given a picture of a dog and a picture of a cat named "Dog," a computer can't tell one from the other. So, ten years ago, if you wanted to use a search engine to locate a photograph of a dog, the search would bring up any images for which someone had included "dog" in the file name. You might get hot dogs, dogwood trees, Dog the Bounty Hunter—but nothing from files named Fido or Lassie.

Von Ahn is a steadfast believer in having fun, even—make that especially—when you're doing work. So he created a game around this issue. Users logged on to the game site and were paired randomly with an anonymous partner. Both players were shown the same image and asked to agree on a label for the image. But you couldn't communicate with your partner—you had to guess what he or she would come up with. Once your labels matched, you earned points and the screen showed another image. You had two-and-a-half minutes to label fifteen images.

There's something oddly compelling about this. Maybe it's the clock ticking away, maybe it's the accumulation of points and the chance to see how you stack up against other players. Maybe it's simply the triumph of finding a match. The game, called the ESP Game, went viral. Von Ahn claims over the course of its run there were 200,000 regular players—many of them playing more than forty hours a week. But the true beauty of the game is the useful work the players were accomplishing.

Every time partners agreed on a label, that answer was noted. Over time, when a statistically significant number of players had come up with that word for the image, it was stored away as a descriptive keyword. If enough people labeled the image of that cat named "Dog" as "cat," then the computer knew not to return it in a search for dog images. Over time, von Ahn says, millions of labeled images were collected.

Google recognized the ESP Game as a game changer, so to speak, and licensed it in 2006 under the name Google Image Labeler. Type "dog" into a Google image search today and 182 million pictures of dogs pop up—even a dog on an ancient Greek urn from around 550 B.C.E. But no cats.

The ESP Game marked a watershed in von Ahn's already robust career. That's when he began to toy around with an idea that would change his work—and the Internet. The most powerful part of the Internet didn't have anything to do with computers. It was the people using them. And that was not only an untapped resource, but a seemingly endless one—as limitless as von Ahn's imagination.

Think of this: Von Ahn says in one year more than nine billion human hours are used—wasted—playing Solitaire. What if you could wrangle those hours into something productive, something that served the common good? He saw that as a new challenge for computer science, calling it "human computation." It would bring together the intelligence of humans and computers to solve enormous problems, ones too huge for either to solve alone. Now all he had to do was find a task worthy of the idea.

In Guatemala, although he is an only child, the boy is surrounded by an extended family. Since his mother has eleven siblings, there is a bevy of cousins to play with—so many that later he will say he didn't have the mental energy left to have many friends outside of his family circle.

His mother has told him since he was young that he would go to the U.S. for college. Even though Spanish is her first language, she speaks only English to him at home so that he will grow up bilingual. Going away to college is not unheard of in Guatemala, but it's not the norm. Most young Guatemalans are expected to live at home until they marry.

In 1996, the boy, by now a young man, enrolls at Duke as a math major. He loves computers, but he loves the elegance of math even more. Still, he takes so many computer courses he almost has a second major in computer science. All those back home who expect that he'll be too homesick to stay he proves wrong—even though he's never had to take care of himself before. He's never even made his own meal, unless you count pouring cereal into a bowl.

Perhaps he's working too hard to notice that he misses home. He always has been an excellent student, but he looks back on high school and feels like he had glided through. Now he makes a conscious decision to put muscle into his studies. He is going to learn "everything," he says. He reads "a gazillion" books—math and computer- science books, but also Spanish literature. It's funny because in Guatemala, he hadn't read literature. Maybe it was the connection to home, the comfort of the Spanish language, that he needed.

Eventually, he decides he'll hitch his wagon to computer science rather than math. It's a pragmatic decision—there are fewer job openings in mathematics departments to be had. But there was more to it than that. Some mathematicians spend their careers working on open problems that haven't been solved in 200 years. What he heard from computer-science professors was different. It wasn't unusual to hear them say, "I did something today that nobody has ever done before." That, he decides, is what he wants to do.

The world at his fingertips
The world at his fingertips: von Ahn sees limitless possibility in the vastness of the universe.
Justin Merriman

The floor in the seventh-floor corridor of the Gates Center at Carnegie Mellon is worn a little more than most. That's where Luis von Ahn paces. You can almost set a clock by him. He works in short spurts. Seven minutes, he says, is as long as he can sit at a computer and write. Then he gets up and paces, comes back, does some work, gets up and paces again. It's very inefficient, he knows, but that's what he does. His team teases him. "It seems like you don't do anything all day long," they say. But his accomplishments say otherwise.

He was one of the hottest commodities on the job market in 2005 when he finished his Ph.D. Microsoft courted him, as did other corporations and universities. He chose instead to stay at Carnegie Mellon as a postdoctoral researcher. The following year he was made a professor. His first week on the job, he was awarded a MacArthur "genius" grant—a $500,000, no-strings-attached stipend given to individuals who show exceptional promise and creativity in their work. The next week, Popular Science named him one of the Brilliant Young Scientists of 2006. The president of the university sent him an e-mail note saying, "What's next week?" When von Ahn didn't earn any unexpected accolades the third week, he e-mailed again, tongue-in-cheek, to say he was disappointed.

Even before the accolades rolled in, von Ahn was pacing the hallway, working on his next big thing. He had found a project as outrageously Herculean as his ambitions—a project that could only be carried out by dint of the masses who log on to the vast, sprawling, amorphous Internet every day.

He had some calculations that intrigued him. The Empire State Building was constructed with seven million human hours—or, as he liked to look at it, 6.8 hours of people around the world playing Solitaire. It took 20 million human hours to complete the Panama Canal, less than a day in Solitaire hours. If the grandest projects of humanity—say the Great Pyramids of Egypt— took 100,000 people to accomplish, what dreams, he wondered, could you fulfill with 500 million? Could an army of ants move an ocean, one drop of water at a time?

He already had an army of ants—the 200 million people typing in CAPTCHAs every day to prove that they're not a computer. In fact, he's felt more than one twinge of guilt over the ten seconds of a person's time that is being wasted each time they stop for a CAPTCHA. Add those lost seconds together and you have an astounding 500,000 human hours a day. With that kind of brainpower, he thought, you could digitize all the books and newspapers printed before computers came on the scene.

Why can't the computers do that themselves? Well, even the most sophisticated OCR (optical character recognition) software can't interpret all of the words printed back when letters were set by hand. Distorted type, blurred print, and damaged pages cause computers to misread 30 to 40 percent of the words from books produced before the 1990s, von Ahn says. But if humans can read even the most distorted CAPTCHA, couldn't they also interpret a word that OCR can't?

Enter reCAPTCHA. You've seen this already, too. Instead of one CAPTCHA you must enter, a site has two. The first one tests you against an answer the computer knows. The second is one the computer doesn't. It's not testing you; it's putting you to work. You're reading a word from an old book or newspaper that the computer can't decipher correctly. Instead of wasting your few seconds of time, you've invested it in a cause, helping put more printed works online.

Google bought reCAPTCHA from von Ahn in 2009. It now offers reCAPTCHA free to any website that wants to use it to improve its security. More than 100,000 websites have it, with 60 to 70 million people transcribing around 100 million words a day. Close to a billion people have participated, whether they knew it or not.

Von Ahn doesn't disclose what Google paid for reCAPTCHA. He just says he will never have to worry about money again.

The young man, for he is still young, is an idea generator. Hundreds of ideas occur to him every day. Most of them are idiotic, he is the first to admit. But when one rises to the top, he sits with it for six months. If he's still thinking about it then, he decides it's worth trying to develop.

Being financially secure has given him a certain intellectual freedom. "I don't want to waste my time on things that I don't think are worthwhile, which is a really nice feeling to have," he says. "If I don't think it's worthwhile, I don't want to spend another second on it. So now I'm only working on things that I think are going to have a big impact."

Some people look at the vastness of the universe and shudder at their own insignificance. Von Ahn looks at the vastness of the universe and sees limitless possibility. So he starts looking at the Internet, the whole of it, and seeing how little of it is accessible to the billions of people on Earth who don't read English. Eighty percent of everything on the Internet is in English. Can he change that? Computers won't be good enough at translating languages on their own for another fifteen to twenty years, he estimates. He can't sit still that long.

And the next big idea is this: Von Ahn thinks he can get 100 million people on the Web translating everything into different languages for free.

By the way, most of them aren't bilingual.

Von Ahn calculates there are 1.2 billion people in the world interested in learning a new language, with five million of them willing to plunk down $500 for software that will teach them. His latest project, Duolingo, offers a way to learn a new language for free while helping to translate the Web. Or as it says grandly on its home page, you'll be "enabling a wealth of language- shackled information to be liberated for all of humanity."

Private beta testing began in November, and a public launch will follow. At first, the site will focus on Spanish and German, but eventually it will include French, Italian, and Chinese. Easy sentences for beginners; more complex ones for advanced students. You'll learn not only by translating it, but by seeing how others translated it as well. Von Ahn claims that people who have tested it have learned as well as those who have used the leading language software. In fact, he says that the translations have been as accurate as those by a professional translator.

If Duolingo goes viral, here is a hint of what von Ahn believes it can accomplish. With 100,000 active users, all of Wikipedia could be translated into Spanish in five weeks. Bump that up to one million users, and that task could be accomplished in eighty hours.

"Trying to translate the whole Web into every major language is a really super-crazy, ambitious thing," von Ahn admits. "It's proceeding well. We're pretty confident we're going to be able to do it. But at the end of the day, we still don't know." That's exactly where the boy wanted to go all along, somewhere no one had been before. And he's taking all of us with him.

Flecker is a freelance writer and writing coach living in Pittsburgh. She has written on topics from synthetic biology to geography, the Everglades, alpaca ranchers, truffle farmers, scriptwriters, and labyrinth makers for universities around the country.