Personally I hate Captcha, that horrid technique of placing some messed up text on a image that a human should be able to read but a computer should not. My main complaint is that it is time consuming and I often just not readable by me. That being said I have not got a better idea for the masses.
Now if you concede that something like this is needed then Carnegie Mellon University’s reCaptcha is an amazing idea. They have been scanning books using OCR software, but not all the words can be interpreted automatically. An unknown word is used for the Captcha, along with another word that is now known, and then user needs to enter both words. This way people are entering the unknown word to identify it while confirming they are human with the known one. The service can be used by other web sites and a number of code libraries are available.
It is not perfect. Some text is imposable to read, and without the context of the sentence I suspect some words will be interpreted incorrectly. That being said I find it easier to use than most other Captcha solutions.
Delicious
Flickr
Google Plus
Hackaday
Instructables
Lanyrd
Twitter
YouTube
Facebook
Geocaching
