a captcha (completely automated public turing test to tell computers and humans apart) is a program that can generate and grade tests that most humans can pass, but current computer programs can’t pass.

in practice, it’s one of those images of distorted text one is often required to enter when signing up for free web-mail or other such services. they are implemented in order to prevent bots (computer programs masquerading as humans) from abusing such services, by signing up for thousands of accounts and using them to send spam, for example.

just as the science of designing these programs is evolving, so is the science of defeating them. it’s hard to tell which is more fascinating.

greg mori and jitendra malik of uc berkley have written an algorithm for defeating gimpy, one of the most common captchas (the one used by yahoo, for example).

interestingly, although their approach relies heavily on the letters in the test being part of a common word, it’s hard to see why these tests don’t use meaningless strings of letters. most humans, i think, would have no great difficulty passing such a test.