| QUOTE |
By Alan Connor BBC News Google has just taken on legions of new workers. None are being paid - and you might be one of them. Since workplace computers were hooked up to the internet, office workers have found more ways of wasting time at w**k, with e-mailed jokes or videos of apparently-amusing accidents. And then there are the games. Dr Luis von Ahn of Carnegie Mellon University estimated that in 2003, nine billion human hours were spent playing computer solitaire. To put this in context, the construction of the Panama Canal took 20 million human-hours. Being a computer scientist, Dr von Ahn was aware of projects like SETI@home, where volunteers donate "wasted cycles" (the spare time of their home computers) to help the Berkeley Space Sciences Laboratory look for signs of extra-terrestrial life. Rather than paying people, I get them to want to label images for free Dr Luis von Ahn For all the elaborate projects that computers are w**king on, there are still some things these machines are very poor at. One of these is seeing. A computer will recognise that something is an image, but will have no idea what it is an image of. So a project to, say, label all the images on the web will need to get humans to pitch in and help. But who is going to sit around saying what they see for hours at a time? Enter Dr von Ahn, with a new game. "Rather than paying people to label images for me, I get them to want to label images for free." How to play This is the game you might have been playing online: paired up with a stranger, both of you are shown the same image, and both come up with a label for that image that the other will have thought of. Once you get a match, you move on, building up points. It's important to understand how compulsive this simple activity can be: it is a race, and it is rewarding when you find a partner on the same wavelength. And if a partner fails to label quickly enough, there is the frustration of lost points - even though the rewarding of said points is wholly arbitrary and worthless. Dr von Ahn has created a suite of image-labelling games, and noticed many players putting in more than hours each week. For the public good, he decided to cut players off after 10 hours of continuous play if they had connected from a university computer. So, many images are getting many labels. To what end? A scientist, of course, should not need to have applications in mind in order to be seized by a challenge. For his part, Dr von Ahn talks of better browsing for the visually impaired, and better cataloguing of data. And when he talked about these things at Google HQ recently, it is not hard to imagine the appeal of the game to his hosts - and their shareholders. One licensing agreement later, and an academic research project has become a Google Images game - and the results are proliferating. In fact, you do not even need to be online to be contributing: your strings of guesses are memorised, and other players may be playing against a phantom "you" - or it might even be a phantom "them" against a phantom "you", building up matches all the time. Sweatshop potential Having spent time helping out computers with the tasks they cannot do, you might wonder - was this not supposed to be the other way round? Dr von Ahn's previous contribution to the web was the "captcha", the distorted string of letters or numbers that have to be decoded before pressing "send" in online forms. One unintended consequence has been the alleged existence of "captcha sweatshops" in the developing world, where spammers employ humans to decode 12 "captchas" a minute, all day long. So what might the unintended consequences of the Google Image Labeler be? The answer probably depends on how literally to take the Matrix films. But making humans enjoy helping computers to see things - primarily to see humans - is likely to affect more than web-browsing for the visually impaired. Story from BBC NEWS: |