Captcha tests are meant to prove that users are humans, not bots. However, according to a new report, even automated bots are as good as humans at answering these security questions.
Bots Better Than Humans in Captcha Tests
Websites use captchas as security checks to block potentially malicious bots by displaying simple puzzles for people to complete, but they are highly challenging for computers.
With improvements in computer vision and machine learning, bots quickly caught up. For example, they could recognize the text with almost perfect accuracy in earlier versions of Captcha, which required users to transcribe distorted text from a picture.
Captchas have developed into an intrusive presence on the World Wide Web due to an arms race with bots that saw them get harder and harder for humans and computers to solve.
However, a recent study that hasn't been peer-reviewed but was published in arXiv shows that bots can easily and rapidly overcome Captcha tests, indicating that users' overall effort isn't as great as it could be.
But according to recent research that hasn't yet been peer-reviewed but was published in arXiv, bots can easily and rapidly solve Captcha tests, indicating that the daily effort users put into solving these puzzles might be more trouble than it's worth.
About 120 of the 200 most popular websites evaluated by scientists for the study still used Captcha. They enlisted the aid of 1,000 online volunteers from various backgrounds to complete ten captcha tests on these sites and evaluate their difficulty levels.
Volunteers ranged in location, age, sex, and educational attainment.
According to researchers, numerous bots mentioned in academic papers were found to be faster and more accurate than humans in these tests.
Humans required between nine and fifteen seconds to complete several Captcha tests with 50 to 84 percent accuracy. Still, bots completed them in less than a second and with near-perfect accuracy.
The accuracy of the bots ranged from 85% to 100%, with the majority being higher than 96%. According to researchers, this is significantly more accurate than their study's human accuracy range (50-85%).
They also discovered that in virtually all instances, the bots' solving times are "significantly lower" or practically identical to those of humans.
Researchers have recommended new and more dynamic methods to defend websites because present Captchas do not achieve the necessary security to prevent bots from accessing them.
AI Getting Better Than Humans
Several reports have proven that artificial intelligence is getting better than humans. In a previous report, UCLA psychologists found that when given the kinds of reasoning challenges that typically arise on standardized tests and IQ measures like the SAT, the GPT-3 artificial intelligence language model performs about as well as college freshmen.
In another study, researchers evaluated 195 medical queries and contrasted the responses from physicians and an AI chatbot called ChatGPT. According to the researchers, ChatGPT's reactions were ten times more sympathetic than human doctors.
Check out more news and information on AI and GPT-3 in Science Times.
