Although pattern recognition helps a lot, it might not be good news for the online world.
Have you seen these before? They are “Completely Automated Public Turing tests to tell Computers and Humans Apart”, or CAPTCHAs for short. CAPTCHAs can prevent botnets from launching DDoS attacks on their servers by blocking computers.
However, these first-generation CAPTCHAs only use twisted letters and numbers, so pattern recognition cracked them many years ago.
Then we got the second generation of CAPTCHAs. In this version, users have to recognize the pictures.
This works, BUT in a few years, the pattern recognition also improved a lot. It can recognize some easy images now, so CAPTCHAs use more difficult images to prevent computers from passing the test. But this also blocks many human users out, pretty annoying, right?
To improve users’ experience, Google introduced reCAPTCHA v3. It monitors your cursor movement, IP address, and HTTP request to see if you are part of a botnet. So normal human users can usually pass the test without any extra action. But will AI be able to break through these CAPTCHAs? We don’t know it now, so we should think if there are any other ways to know if you are human.