In this second article of a four-part series commemorating the World Wide Web’s 20th birthday, Scientific American editor Mark Fischetti explains how online games and other Web-enabled social machines build machine intelligence–and help us, too
One of the Web’s huge promises is that it will help people work together. Although the Web has made it easier for groups to share information and collaborate online, it could do much more. So-called social machines could allow large numbers of people to conduct better science, even to transform democracy.
You have probably already taken part in an embryonic social machine. For example, when you fill out an online form, you may be presented with a little box that has wavy-looking words in it, and you must type the words before you can proceed. This is a social machine known as captcha. By typing the words, you and other people who are seeing the same boxes are helping computers that are digitizing printed text to figure out words they cannot determine, for various technical or syntax reasons.
Luis von Ahn has pioneered these kinds of systems and taken them further. In one variation, you are shown an image and have to type a word to identify it, such as dog, house or asparagus. If a second person somewhere unknown to you types the same word, you each get a point. People play such games—games with a purpose (GWAP)—for hours, but the exercise is helping computers identify images that the machines could not recognize on their own, building up catalogs of machine intelligence.
Online review systems are simple social machines: Which restaurant should you go to? Which exercise bike should you buy? Check the reviews and ratings people have made. The ratings don’t just help consumers, however; they can alter commerce. If Schwinn gets the best ratings it may sell more bikes, and vendors with poor ratings may go out of business. If most customers of the best-selling bikes comment on how much they like having a built-in heart-rate monitor, chances are that future bikes will have them.