Despite being near perfect when it comes to CAPTCHAs — a feat that plenty of humans can't even manage — the new system's analysis of Street View imagery isn't quite as accurate, correctly identifying the text just over 90 percent of the time.
Here is their blog post regarding the changes made to reCAPTCHA in response to the research: http://googleonlinesecurity.blogspot.com/2013/10/recaptcha-just-got-easier-but-only-if.html
Of course it doesn't give many details. I'm curious what types of traits they're looking at now:
Mouse movements? Time delays to get to the CAPTCHA box? It's interesting stuff. Distinguishing bots and humans can only get harder from here.
It reminds me of a section in Permutation City where a character has a virtual simulacrum of herself, responsible for fielding incoming video calls and messages. The spammers of the future in turn have their own realistic AI likenesses which try to ascertain if they are communicating with a real human or a spam filter. There's some crazy back and forth Turing-esque tests going on over video chat to try to skim out the important information.