Remember when the web was populated with annoying “captchas”, where we had to introduce words that, even as humans, we had trouble reading? Cases like this?
The CAPTCHA is used as a protection method for websites to avoid being invaded by “bots”. Many times, for different reasons, malicious people can start creating thousands or millions of accounts to a service that is not protected against it. For that, the “CAPTCHA” serves as a verification that we are human, that we can discern text that a machine could hardly read. The problem? That on many occasions, even as humans, we can not read it.
However, on the web we have seen how a CAPTCHA box like the one above has been replaced by a simpler one, one that simply has a “check” next to the phrase “I’m Not a Robot” or ” I am not a robot”. We click, and that’s it! Access granted.
But how does the I’m not a Robot work?
CAPTCHA comes from “Completely Automated Public Turing Test to tell Computers and Humans Apart”. Yes, the acronym, but in Spanish, implies a public Turing test to distinguish between robots and computers (the Turing Test is frequently performed against artificial intelligence). Here’s how I’m Not a Robot works.
The problem with these tests is that artificial intelligence and computational power have improved so much in recent years that even for tests like the old Captchas, computers are even more capable of inferring what is written here.
But then, why would a simpler test, such as clicking on “I’m not a Robot”, be more effective than trying to deduce blurry or hard-to-read text?
It turns out that we as humans leave a great trail of information on our PC, such as the IP address (number that identifies our computer), and cookies, which are stored in our PC and serve, for example, Keep our Facebook session active, or allow advertisers to know a little about our browsing habits. This is used by Google, those responsible for the new CAPTCHA, to detect that it is not a malicious computer, but an ordinary user. A computer or bot programmed to simply try to access websites would display a totally different behavior, a totally different history. Google also uses clues as subtle as the way the mouse moves before clicking on “I’m Not a Robot,” because a script, a computer, would have a much more direct path, than random, erratic movement , Typical of a human being.
This collection of little tracks helps the Captcha button to differentiate between human and robot. The trick of the new Captcha button is not to make the test more difficult, but to accumulate a series of evidences that makes the job much more difficult for a bot, to deceive the CAPTCHA.
The process has improved but it is not perfect. For those cases where the CAPTCHA fails to collect sufficient evidence of our humanity, we are presented with a test. A much simpler for us, human, but an infinitely more complicated for a bot.
Here, the test is asking us to identify images like “this”.
Note that you do not even mention the cat, but simply, it shows an image of a cat. We, as humans and by general culture, know that it is a cat and we can quickly identify cats, dogs, rabbits and plants to pass the test in a couple of seconds.
But for a bot? The process is much more complex. First, it would have to pass a process of image recognition in order to define what the test refers to with “this” (because it is never mentioned that it is a cat, that would facilitate the process). This then involves passing this image into an image database or neural network (more on neural networks and artificial intelligence, in this special). With luck and after a great effort, perhaps the machine detects that it is a cat. And then, you would have to pass each of the images shown additionally by the same neural network so you can make the attempt to differentiate between a cat and a rabbit; Or a cat and a dog. Computers are wonderful when processing vast amounts of information, but they are totally incapable of reasoning, which makes this a tedious and complicated process to solve for a computer.
That is where the genius of this new system of Captchas falls. Instead of complicating the test, we have sought a more efficient way to protect our information, by collecting information about our behavior, to make it difficult for the test not to all, but to those who want to pass.
This CAPTCHA system started to be implemented by Google in 2013, but we have seen a rapid expansion in almost all important web pages because it works not only better but, more important, it is much simpler and less tedious for us, the humans who We want to make use of these services. That’s right, then how I’m Not a Robot works.