← back to essays/

Power To The People

2007-05-27

I think we have reached a point where the technology that we have will allow masses of people to collectively solve problems that are yet to be solved by computers. I have seen some examples of this and am convinced that this concept will lead us to new frontiers in what technology can accomplish. What's really interesting to me, though, is that every time a human solves one of these problems, the computer gets a little smarter too.

Computers cannot perfectly perceive and recognize what the content of a picture is. It would be difficult for a computer to see a picture of a dog and recognize it as a dog. However, this process is very easy for humans. These simple facts are the basis for The ESP Game and Google Image Labeler. These sites use a "game" interface to pit humans against one another. Each player is showed an image and they are asked to type in tags that describe the image. Players receive points when their labels match labels of other players and over time as more and more images are labelled by multiple players, the labels for the images can become more and more precise. With enough players, enormous numbers of images can be indexed. And here's the part that really gets me: all of those labels that humans are generating? They're building a massive reference library that computers could eventually learn from. Imagine the possibilities here...

With the web being flooded with new blogs everyday, spammers have found an excellent place to spread their plague; in comments. In an effort to combat spam, many spam filters show users a word in scrambled print and require the user to type out the letters in the word. This prevents programs from spamming, since the programs cannot identify the words in the images and therefore cannot post. One such system is CAPTCHA. Relatively cool, but hardly worldly, right? Not so fast... Every time a human decodes the word in a CAPTCHA image, the response is entered into a database... The images produced by re-CAPTCHA are words that cannot be deciphered by OCR programs that attempt to scan documents into a computer and turn them into text. For projects like this, CAPTCHA helps to correct OCR mistakes and aids in the process of digitizing books, and you're helping even though you don't know you're doing it.

This is brilliant. And I think it's just the beginning. Every time a person solves one of these little puzzles, they're teaching a computer something it couldn't figure out on its own. Millions of people doing this every day, without even realizing it. At some point, will the computers have learned enough from all of us that they no longer need our help? Could you turn a game into a way to solve problems in biology or medicine? I genuinely wouldn't bet against it. I wonder what will be next.