The Story of reCaptcha

In order to make old books in the public domain freely available to the general public, an effort was undertaken to scan and upload them into the Internet Archive. In this effort, optical character recognition (OCR) was used to convert the scanned images of books into searchable digital text. Unfortunately, many older books have stylized script that cannot be converted to text using OCR. A lot of human readers were needed to convert images of stylized text into digital text.

Bots are short for robots, software programs that perform repetitive tasks on the web. Malicious hackers use bots to create a large number of online accounts or purchase large numbers of tickets for resale by scalpers. They also use bots to scan the web for email addresses to which they can send spam, i.e., unsolicited messages. Captchas were designed to stop bots by presenting distorted images of characters that could only be read by humans.

Captcha stands for “Completely Automated Public Turing test to tell Computers and Humans Apart”. In a Captcha, the human user is asked to view a distorted image of text and type it as digital text. Because of the distortion in the image, a computer cannot correctly convert it to digital text using OCR. By completing captcha tasks, human users could go on to log into an account, post a comment or fill out a form. But, bots could not.

Luis von Ahn, a computer scientist, came up with reCaptcha, a single solution for both helping upload old books and stopping bots. In reCaptcha, two images are presented to the human user: one a traditional distorted image of text used by Captcha, and the other, an image of stylized text from an old book that a computer could not convert by itself. So, every time human users converted reCaptcha images, they also helped in the process of scanning and uploading books.

For his work on reCaptcha and crowdsourcing, Luis von Ahn was awarded a MacArthur Fellowship, unofficially called a Genius Grant. Google acquired reCaptcha and continually modified it to counter various attempts to automatically solve it. Google has since replaced it with a newer version of bot detection more accessible to mobile users, wherein, the user is asked to select images from a grid that match a specific prompt.