Tuesday 4 October 2011

CAPTCHA, reCAPTCHA and Turing Tests

To check that those signing up for its various services are humans and not machines, Google uses a program called a CAPCTHA.  Coined in 2000 by Luis von Ahn, Manuel Blum, Nicholas J. Hopper, and John Langford, CAPTCHA stands for "Completely Automated Public Turing test to tell Computers and Humans Apart."  This distinction is useful to protect programs and applications from mechanized attacks, where computers are programmed to simulate human users in order to get access to the sytems.  Google thought CAPTCHAs were so important that they bought von Ahn’s company reCAPTCHA.  reCAPTCHA exploits the human brain’s superior ability to derive textual information in order to correct errors made by computers when they ‘read’.  The program takes words from scanned books and offers them to users as part of traditional CAPTCHAs.  The text that users enter is then used to correct the transcripts of the scanned books.  By parcelling out the process of correcting transcripts and then making it a game, a task that would otherwise be too laborious to carry out is completed as users go about their day-to-day business.

CAPTCHAs, as a form of Turing Test, go right to the heart of artificial intelligence.  If a machine can pass a Turing Test, then it is a little more human.  Thinking about such tests, then, is a form of thinking about what makes us distinct from machines.

You can learn more about this here:

2 comments:

  1. Like many people I suppose, the CAPTCHA is a phenomenon I've frequently encountered but not one I've thought in great detail about. As previously mentioned they highlight a fundamental difference between people and computers. This is the same ability that lets us focus on a problem analytically and find the solution without looking at every combination in a 'block' approach. The same ability that allows us to listen to just one instrument in a score.

    I may be the only one that feels this way, but when our technology has got to the stage that we need to perform tests to establish whether an agent is human or machine, our technology has become scarily advanced.

    This is not some paranoid fear nor some knee-jerk reaction to a piece of technology that maybe invades and disturbs our concept of humanity. This is simply a reflection. Previous generation's visions of the post millennium era were robotic figures a common part of day to day life. The reality is a lot more subtle; instead of being able to tell a machine from a human by the difference in the way it looks, we are only able to spot a machine from a human by the difference in the way it acts.

    ReplyDelete
  2. This raises (at least) two interesting issues: a) accessibility and b) human nature.

    With CAPTCHA you focus on just one feature of being human, visual pattern recognition. Of course this seems a perfectly easy thing to require in order to shut out automated log-in attempts etc, but unfortunately it also shuts out people who are blind (and presumably also dyslexics). Hence you need to provide an alternative, which is now usually available in the form of a sound recording. I cannot remember, but I am sure that this was not the norm at the start, and that many CAPTCHA systems still do not have this. So always think hard about what you are testing, and whether there are humans who might fail this for other reasons.

    The second one is the reason why we have CAPTCHAs in the first place: because technology is abused. First humans post spam messages, but then they use computers to automate the process. As a result, blogs and forums require logins to shut out those programs. Next step: the spammers get more savy and get their programs to create user accounts. So you need further verification, either by requiring an emailed link to be clicked on, or a CAPTCHA. As Joe said, in previous generations' views the future was positive, robots and computers helping humans to do their drudge tasks, but what has actually happened is that computers are abused to be a nuisance with spam emails and comments, phishing attempts, scams and all sorts of other things. Rather than just being a tool for the good, as people have thought in the past. And I'll stop here before it gets too depressing...!

    ReplyDelete

Note: only a member of this blog may post a comment.