With our prolific use of the internet nowadays, it’s likely that at some point during browsing you’ll have been asked by Google to prove you’re not a robot – either by clicking on a box to say so, or by selecting certain images containing a specific object from a selection that Google wants you to find. Commonly, this includes things like selecting all the images that contain cars, or traffic lights, for example. Once you’ve filled out the information correctly, you’ll be allowed to continue browsing.
How does this technology work?
This technology is called CAPTCHA, an acronym for Completely Automated Public Turing test to tell Computers and Humans Apart. The technology was developed in the late 1990s to protect websites from fraud and abuse against automated bots setting up fake accounts by generating and grading tests that humans can pass but that current computer programs cannot. Usually, bots can often only follow set patterns or input randomised characters, whereas humans can recognise novel patterns. This makes it unlikely that bots will correctly guess the right combination to access the site.
Nowadays, CAPTCHAs are often triggered by automated processes sometimes caused by spam bots, infected computers, email worms or DSL routers, or from some SEO ranking tools, which is why it might their appearance might seem random.
Most CAPTCHAs also come with an audio test option to ensure that blind users are still able to complete the tests. Google says that currently people are solving around 200 million CAPTCHAs a day.
How do CAPTCHAs help train AI?
Aside from helping to prevent suspicious activity on websites, CAPTCHAs also have another key purpose – helping Google to train its AI to be smarter, faster, and ultimately more reliable. For instance, when you complete image CAPTCHAs, you are helping train the company’s machine learning algorithms to be able to better recognise and identify certain objects in images. Think about it, with users identifying street signs, for example, the algorithm starts to build a more complete picture of what street signs look like too – removing from its database any images that have not been determined to contain a street sign till it has an idea of what a street sign is.
With this training process in place, Google can make its search results on Google Images more accurate for users. Not only this, but CAPTCHA is also helping Google to develop its driverless car technology, getting its AI trained at identifying what humans call cars, what a red traffic light looks like, and so on.
The problem is, however, that as the AI gets cleverer, Google needs to make its CAPTCHAs more difficult so that it can continue to prove you’re not a robot. This reduces the chances humans will get the CAPTCHA correct. Indeed, in 2014, Google pitted one of its machine learning algorithms against humans in solving the most distorted text CAPTCHAs: the computer got the test right 99.8 percent of the time, while the humans got a mere 33 percent.
To try and combat this problem, CAPTCHAs often change overtime. Recently, there have been efforts to develop game-like CAPTCHAs. These are tests that would require users to rotate objects to certain angles or move puzzle pieces into position, with instructions given not in text but in symbols or implied by the context of the game board. The hope here is that humans would understand the puzzle’s logic but computers, lacking clear instructions, would be stumped.
While, completing a CAPTCHA may be a nuisance, it is cool to know that every time you submit one, you’re taking us one step closer to driverless car technology, greater image recognition tools, and to an improved Google experience. Thanks to our hard work, it is getting increasingly difficult to separate us humans from the machines – something that will be essential when you sit in your driverless car in years to come.
Let us know in the comments section, or @techtroublesho1, if you have any questions or other technology queries.