A Step Closer to Artificial Intelligence with CAPTCHA-Solving Tech

Are you a human?

On the internet, that’s no easy question to answer. Companies varied as Yahoo and Ticketmaster are always seeking to weed out the humans from bots. That cat-and-mouse game has lead, most recently, to the proliferation of the CAPTCHA, or the completely automated Turing public test to  tell computers and humans apart. You may know it better as the frustrating mishmash of characters, lines, and smudges that you’re supposed to distill into plain text. They’re frustrating to complete, but they do a decent job at keeping unwanted bots out of the system.

While CAPTCHAs work well as gatekeepers to internet services, they can also serve as tests for artificial intelligence. If you can design some software that’s good at cracking CAPTCHAs and you’ve passed a Turing test. (Turing tests were proposed by pioneering computer scientist Alan Turing in 1950 as a way to tell humans from machines.)

Humans can easily read this nonsensical string, known as a CAPTCHA, but it can be a struggle for computers.

Defeating a CAPTCHA doesn’t necessarily mean a system has passed a Turing test. CAPTCHAs have been defeated a number of times with a variety of techniques, from those that solve them using the audio version to others that use massive amounts of training data to crack a decent percentage of the puzzles. Now, a company named Vicarious claims they have developed a system that can defeat CAPTCHAs with greater than 90% accuracy without relying on some of the shortcuts or training data other systems may use. They say their system reconstructs the human visual system in a computer, allowing it to easily distinguish letters and numbers from visual noise.

Here’s Rachel Metz, reporting for Technology Review:

The purposes go well beyond Captchas: Vicarious hopes to eventually sell systems that can easily extract text and numbers from images (such as in Google’s Street View maps), diagnose diseases by checking out medical images, or let you know how many calories you’re about to eat by looking at your lunch. “Anything people do with their eyes right now is something we aim to be able to automate,” says cofounder D. Scott Phoenix.

It’s still a far cry from what we may think of as full-fledged artificial intelligence. It’s also somewhat different from what IBM has been doing with its Watson system, which recreates human reasoning instead of vision. But if you start piecing these disparate parts together—as will undoubtedly happen in the future—you can easily imagine a computer that can see and think on par with humans.

To learn more about artificial intelligence, watch “Smartest Machine on Earth.”