You have heard about the book digitization projects. You know that Optical Character Recognition is sometimes an uncertain thing, especially with offbeat typesets and obscure fonts.
Then there is CAPTCHA (and here), the security method that many sites are using to monitor logins. Basically you view an image of distorted text, and type in the correct letters in order to access the site.
What if each of these problems could use the other as a solution? reCAPTCHA (here, too)
found on MetaFilter