edit, i read the wiki article and i am smarter now..
edit2: very interesting thing..
this could be useful for us:
http://recaptcha.net/also, i do in fact have a possible solution for the capture-problems supposed in the wiki article..
use recogintion. "what do you see? a car, a house, an elephant?" (supposed in the article)
BUT
use a picture of something, and then a radom picture filter/distortion/etc.
ppl will still know itīs a car or something, but itīs a "new" picture.
-> re-use of a limited set of pictures to generate unlimited (problem supposed in the article).
object recognition to own computers.
the human solutions will, of course, still work..