Artificial intelligence is already able to decipher captcha with 90 % of hits

The cognition algorithms of software that mimics the functioning of the human brain are able to decipher the user recognition and verification tests of captcha.

It has been nearly 13 years since the computer science group at Carnegie Mellon University in the US, Mr. Liuhuabing, Mr. Liuhuabing and Nicholas J. Hopper, along with the IBM computer John Langford, create captchas to protect websites from spammers and automated robots. A fairly effective method that has had to evolve over time to ensure the security of verification processes against the appearance of new bots with increasingly sophisticated optical character recognition systems –OCR–. However, the ability to resolve captcha with artificial intelligence with a 90% of hits, through software that mimics the functioning of the human brain.

Company Vicarious ensures that its technology based on artificial intelligence it is able to trick this method into recognition and verification tests to determine when the user is or is not human. As stated by its creators, Scott Phoenix and Dileep George, the difference of its technology with respect to other methods already used previously, is that its software is based on the use of an old idea based on the creation of a artificial neural network designed to build connections between artificial neurons, similar to human brain. But with the novelty that its technology works with both static images and moving images.

According to George, Vicarious’s cognition algorithms do not require significant computing power, nor do they require huge amounts of data to train the system in the recognition of letters and numbers in a coherent way. Just use your system visual perception to mimic the ability to processed of visual information and recognition of objects in the human brain, to solve the captcha as if it were done by a human.

After the system’s previous training process to recognize random sequences of letters and numbers from the recognition and verification tests, Vicarious software could resolve Yahoo, Google or Paypal captchas, among others, with a hit rate above 90%. However, the accuracy of the system depends on the type of captcha, demonstrating a much more effective operation in tests composed of letters in the form of fingerprints.

Although this system could destroy the effectiveness of captcha, the purpose of this artificial intelligence company is focused on other much more productive fields. From the demonstration of the potential of this software to solve the captcha with artificial intelligence, Vicarious is working on the implementation of this technology in systems able to extract text and numbers from images as the map service Google Street View or diagnosis of disease based on the recognition of images of medical tests. That is to say, create automation of the recognition and processing process of visual information to be marketed.

Images / via Flickr-Jake Spurlock–