Online captchas have become increasingly more complex in an effort to combat online botting, starting from simple checkboxes to prove a user is human to having you select all different kinds of things inside an image grid.
The checkers have become more obscure over time, causing many people to become disheartened as they continue to find it difficult to identify a fire hydrant. The term captcha is short for Completely Automated Public Turing test to tell Computers and Humans Apart.
In the early stages of captcha, simple text images sufficed to differentiate between spambots and humans. However, in the 2010s, Google purchased the program from Carnegie Mellon researchers and later used it to digitize Google Books. The company needed more warped and obscured texts to move forward.
The inventors of captcha realized that any given test could only be used for a limited time before being susceptible to bots. In 2014, Google put forth its machine learning algorithms against humans in a match to see who would be the first to solve the most distorted text captchas. The results showed that the computer had an accuracy of 99.8% while the humans only had 33%.
Google executives decided to move to NoCaptcha ReCaptcha that observes user data and behavior to allow some humans to pass with only an “I’m not a robot” button while subjects others to the discerning image grid that is popular nowadays. However, the rise of technology is slowly adapting to the security measures and might make captchas obsolete.
A computer science professor at the University of Illinois at Chicago, Jason Polakis, took responsibility for the increased difficulty of online captchas. He published a paper in 2016, where he used off-the-shelf image recognition tools to get a 70% accuracy on Google’s image captchas.
Recently, machine learning has become so advanced that it could compete with humans on basic text, image, and voice recognition security measures. Polakis said, “We’re at a point where making it harder for software ends up making it too hard for many people. We need some alternative, but there’s not a concrete plan yet.”
The history of captchas started with attempts at finding alternatives to text or image recognition to differentiate between humans and machines. Researchers tried to use human classification, trivia, and nursery rhyme-based captchas to challenge AI. in 2010, researchers used ancient petroglyphs that machines have difficulty identifying, The Verge reported.
Recently, however, there are ideas of using game-like captchas that provide tests, requiring users to rotate objects to an upright position or place puzzle pieces to the open slots. Creators hope that humans would use logic to understand vague instructions that machines would not be able to read. Cameras have also been used to identify human movement to discern between humans and machines.
However, the problem with tests is not that computers are too good at them; it is that humans are too bad at them. Experts said the problem does not lie with humans, but that they wildly vary in language, culture, and experience, which makes it complicated for everyone to complete.
The engineering lead on Google’s captcha team, Aaron Malenfant, says moving away from Turing test-style captchas is meant to address the problem that humans are the ones losing the competition. He said, “As people put more and more investment into machine learning, those sorts of challenges will have to get harder and harder for humans, and that’s particularly why we launched CAPTCHA V3, to get ahead of that curve.”
Malenfant said that in the next five to ten years, he expects captcha challenges to be obsolete and that a secret Turing test will run in the background of the online world.