August 22, 2007

Have you ever heard the term “captcha”? I hadn’t until I listened to a long discussion of these things by Steve Gibson of grc.com (a website everyone should have already visited). You have seen captchas … they’re those funny wobbly letters in a box that you are asked to read and re-type when you sign up for something. Sometimes you can’t even decipher the darn things. Captcha is an acronym standing for “Completely automated public Turing test to tell computers and humans apart”. Strangely, telling computers and humans apart is a rather difficult task on the Internet, and the captcha has been the most recent attempt to do this, so that robots can’t go around creating accounts, and entering sites where they’re not wanted etc. The idea is that a robot can’t read the wobbly image, but a human can. This is the basic idea behind a Turing test. Alan Turing (one of the smartest humans ever) came up with the idea of this test, simply saying that a Turing test is any test that successfully distinguishes humans from computers (when you can’t actually see the subjects being tested). Unfortunately, captchas are only sort-of Turing tests, as they are sometimes crackable by machines. In fact, it would seem that the letters now need to be pretty much indecipherable by humans in order to frustrate the machines. Developing a successful Turing test is, in fact, remarkably difficult, particularly when with things such as captchas, the puzzle must be automatically generated … that is, the captchas that you see are also generated by computers, not carefully crafted by a human with a pair of scissors and some glue. (who then also judges the results). Nope, to make these things work without human intervention is really hard … the most common response is some type of thought-provoking question like “what color is the sky?” which might fool a really stupid computer, but not anything that has the slightest bit of artificial intelligence and might learn from experience. If you ask questions like that, you will have a finite set of questions, and the correct answers can be easily learned, or correctly guessed in a small number of tries by a computer. Identifying objects in photos has a similar flaw. Anyway, the point is that the captcha is a really hard thing to produce. Not only that, but accessibility concerns have greatly complicated matters. If a blind person encounters a visual captcha they will be locked out of whatever it is that a sighted person can enter. The response has been to create audio captchas, which distort the voice and add noise, much like a visual captcha does to an image. Unfortunately, this is also really difficult to do successfully as well, and once an unsuccessful audio captcha is offered, the visual captcha is of no use. It would seem that us humans have to come up with something relatively quickly to differentiate machine and human on the web, or there will be no stopping the robots (pretty science-fiction-y, huh?)


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: