Uncertainty is not a practical problem for reCAPTCHA. During monitored tests, reCAPTCHA had a 99.1% accuracy rate. In order for reCAPTCHA to re-produce this data entry as productive labor, though, it must conceptually undermine its original form. With a CAPTCHA, while the program is designed to generate a puzzle it knows other computers cannot solve, it still knows the correct answer. With reCAPTCHA, it does not. Assumed to be entered correctly only after others solving the same puzzle affirm it, reCAPTCHA is programmed to accept all answers. Though the success rate is near-perfect, accuracy or correctness is a secondary concern to creating and valorizing productive micro-labor. Or, rather, both issues are conveniently collapsed together. This is, essentially, not about spam anymore. In order to make reCAPTCHA work, the ‘reverse’ Turning Test is not a test at all, but a vehicle for work.
reCAPTCHA words take no more than a few moments for humans to interpret, but are the most difficult words for a computer to process in the first place. Von Ahn observed that “users of websites that switch to using reCAPTHCA typically complain less often than when the sites used a difference type of CAPTCHA…due to some users being more willing to accept reCATPCHA because their work is contributing to the digitization of human knowledge.” Results show a clear advantage to the reorientation of CAPTCHA puzzles into reCAPTCA digitization, but also proof of a “more general idea,” according to von Ahn:
“Wasted human processing power can be harnessed to solve problems that computers cannot yet solve. Some have referred to this idea as ‘human computation’. In previous work, we have shown that such processing power can be harnessed through computer games: people play these games and, as a result, collectively perform tasks that computers cannot yet perform…here we have shown that CAPTCHAs constitute another avenue for ‘reusing’ wasted computational power, while serving the useful purpose of preventing automated abuse over the Internet.”