The All New AI Algorithms Beat CAPTCHA Tests 90% Of The Time


CAPTCHA stands for “Completely Automated Public Turing test to tell Computers and Humans are Apart”. And believe me everyone of us has come across this validation test, which is responsible to stop or try to stop SPAM, the plague that has the potential to make our lives worst.

By the time CAPTCHA is considered to stop or at least try to stop spammers and made our life easy. But, now, a new model of artificial intelligent is on the go that can interpret CAPTCHAs with little, in fact, very little data training and it may be the end of this anti-spam system.

The already in use system, CAPTCHA, was created to be unreadable by computer algorithms. It seems to us in various ways, from amalgam of letters and different numbers in millions various styles to images which only human can group.

This method, more specifically speaking, the technique, let the humans to identify the standards, numbers, and requirements, naturally, while computers face difficulty to validate layers with designs, noise and other pitfalls.

From several years this algorithms have been invented to solve these data intensive puzzles, requiring training on millions of examples on CAPTCA images that have been solved, the image and its interpretation or coded rules on how to interpret various kinds of images.

The Stanford University had devised a more efficient and reliable model by the researcher Dileep George and his colleagues, dubbed as Recursive Cortical Network (RCR) that combine neuroscience insights to teach the program how to conclude beyond what it is taught in training mode or training time.

The new workout believes to work more closely with the human brain. With the ability to learn and generalize, by the use of relative few examples, more specifically in comparison to present models of intense and deep learning, eventually becomes the approx. 300 times more perfect and efficient in terms of data.

The creator of the new model mentioned that the key to success is its creation with strong assumptions that are the way to indentify inputs which never found in training.  The algorithm, in this way, has the ability to interpret CAPTCHA texts, recognize handwritten letters and numeric values, depict objects in complex layers and notice text in the images of real-world views.

RCR is comparable and has the almost greater accuracy about 5000 times fewer training photos, as compared to the advanced approaches of deep learning.

There are two different and distinct strands, these studies have, and both are interesting. One of them clearly shows the advantage in the field of neural network and artificial intelligent and other is permission to qualitative leap in dismembering a security against spam thus leads to the requirement of the urgent creation of something stronger and more efficient than CAPTCHA system.

How this will change the surfing experience? Share your thoughts in comment section below.