Modern CAPTCHA tests are not always based on a direct question that the model can answer. Some of them monitor the way the user interacts with the page, such as cursor movement, click speed, and sequence of actions, in addition to analyzing images or performing multiple steps.
And here lies the paradox: AI may be able to analyze an image with a high degree of accuracy, but when it deals with a test designed to detect automated behavior, it may face another layer of challenge.
The test not only asks: “Do you know the right picture?” It also sometimes tries to find out whether the person behind the screen is behaving like a real human being.
Failing a CAPTCHA does not mean that the model cannot understand the image or recognize objects in it. The problem is that CAPTCHA was originally designed to be a behavioral and interaction test, not a traditional test of knowledge or intelligence.
CAPTCHA developers also know that artificial intelligence tools are constantly evolving, so tests are constantly changing to try to stay one step ahead of the bots.
On the other hand, models capable of using the browser and dealing with websites are developing, which makes the competition between the two parties more like a continuous race. (Erm News)
