CAPTCHAs, for those not familiar with the acronym, are those little images of distorted words or letters that you have to type in on some sites in order to register as a user or post a comment. The idea is that only a human could read those, while a spambot, even with the help of OCR, could not.
There's been some heated debate about CAPTCHAs ever since they've become commonplace. They are obviously not for people with bad or no eyesight - and even people with good eyesight often have problems deciphering them. Plus, there have already been reports of CAPTCHAs being overcome, either because of bad implementations (the first incarnation for the popular phpBB forum software was such a case) or simply by hiring cheap labour to break them manually.
Now, did you check the email spam you have been getting lately? In order to defeat the Bayesian spam filters, more and more spam is being hidden in images instead of the actual email's message text. The spam filters responded to this by applying OCR to those images and the spammers in turn responded with distorted images. In other words, the (email) spammers are now using methods that have been developed to keep the (comment) spammers out. Nice.
And now, in an even more ironic twist, the anti spammers may even help the spammers, since the makers of spam filters are looking into ways to machine-process, i.e. read, those distorted images.
In other words: Expect CAPTCHAs to become worthless soon.