The Increasing Difficulty of CAPTCHAs

If you have been using the internet for a long enough period of time, you probably know what a CAPTCHA is. It is basically a way to ascertain whether or not you are a human being in an effort to prevent bots from propagating too much on the internet. A cool tidbit that you might not know is that CAPTCHA is an acronym which stands for Completely Automated Public Turing Test. A Turing Test of course is the iconic challenge that was meant to help us separate artificial intelligence from actual human intelligence.

A CAPTCHA used to be quite easy. In fact, in our era of the modern internet, CAPTCHAs became as simple as a box that you had to click on that was beneath a line of text that said “I Am Not a Robot”. However, in recent times you might have noticed that Google has become a bit more discerning. Once you click on this box you are taken to another window where you will be subjected to a series of increasingly difficult tests. For example, you will be shown a series of images and you will have to select all of the images that have a zebra in it, or you will have to do a test in a similar vein.

You might be wondering just like the rest of us why this is the case. After all, the good old simple CAPTCHAs had been doing the trick for so long, why would Google want to change anything? Well, one thing that you need to understand is that artificial intelligence is called “intelligence” in the first place because of the fact that it learns. It gets smarter. It acquires knowledge. Hence, any mechanism that is meant to determine the legitimacy of any intelligence would have an expiry date. Eventually, artificial intelligence is going to learn how to trick such a mechanism, just like AI can now pretty easily pass the Turing Test.
Related: Google Launches Database to Prevent AI Manipulation by Fake Audio
Google has been developing artificial intelligence for some time now, and one thing that a lot of people don’t know is that the very CAPTCHAs that are used to help ascertain whether or not a user is a bot are also used to train artificial intelligence and machine learning systems to become a great deal smarter than they would have been otherwise. They are used to help these AIs become better at pretending to be human.

The iconic human robot differentiator CAPTCHAs are getting Harder because Artificial intelligence is getting Smarter

We can see the trajectory of AI in the way CAPTCHAs have evolved. They used to be a word of distorted text that you had to read. They were usually easy enough to read but only for a human mind. AIs often got confused because of the fact that the text did not match anything they had recorded so far. In 2014, Google tested the distorted text CAPTCHA out and found that pretty much all existing AI could pass that test in a flash. This was when the NoCaptcha ReCaptcha was born, the “I’m not a robot” CAPTCHA that we have all gotten accustomed to.

Now, these CAPTCHAs are starting to become increasingly ineffective as well. Jason Polakis, a professor of computer science at the University of Illinois, has played an enormous role in the increase in CAPTCHA complexity in recent times. Indeed, his use of consumer software that was designed to recognize images ended up passing two out of every three image recognition CAPTCHAs that were out there, thus proving that a big change needed to be made in order to ensure that everything worked out in the long run.

The fact of the matter is that AI is now pretty much as smart as humans in a lot of ways. Google will be releasing ReCaptcha version 3.0 soon in an attempt to get ahead of the learning curve that machines are currently going through. These CAPTCHAs are going to be different from previous ones in the sense that they are not going be Turing Tests. It is pretty clear that AI has moved beyond such basic tests and needs a much more complex gate that would not be cleared so easily. Only time will tell what kind of security measures we have to end up taking in the future.

Read Next: Artificial Intelligence: Good Versus Evil (infographic)
Previous Post Next Post