Comment Re:tried it (Score 1) 169
I admit, I have not thought about this, but it sounds beyond stupid to me.
The whole point of the dozens of 'Tell Humans from Bots' methods is that it is relatively easy to automatically generate the challenge, but (ideally) it takes human intelligence to solve it. As an example,
(1) it is easy for an algorithm to randomly choose a few letters, and add colors, extra lines, etc... to the picture.
(2) it is easy for a human to see the letters through all the obfuscation.
(3) it is hard for a OCR (optical character recognition) algorithm to tell the letters from the random crap.
But, in this case:
(1) it is very hard for an algorithm to generate and label the test cases. "body builder lady with mustache and goofy in the center" indeed.
(2) it is quite hard for a human to tell them apart.
(3) sure, it is very much impossible for an algorithm to do so, but at this point, who cares?
That said, the idea may have something going for it. Taking stock images that have already been labeled, and applying obfuscating algorithms can work.
But this particular crap set my bullshit detector... oh, wait, it's actually my stupidity detector.