How can artifical intelligence still not sole reCAPTCHAS?


We’ve gotten to the point where, in the latest update, iPhones can now identify even the breed of animals in the background of a photo you take, and suggest links to read about that breed. So how have hackers not been able to create programs that select all squares with traffic lights? Or recognize a short word just because there is a line through the letters/the letter are a little distorted?

In: 1

It’s not that they can’t solve – solution is not the only thing they check. AI has a hard time mimicking human behaviour, like click speed, click accuracy etc.

They can most of the time identify the object, but often reCAPTCHAS dont find out if yor a Human or not through you selecting the right pictures, but how you behave doing this. They also see sometimes how you surf through the web, a computer also behaves diffrent there. Dunno if this the whole reason tho

The concept of reCaptcha is not to generate hard tasks that only humans can solve, that was what captcha was. However it is to find real world problem, have the best AIs try to solve it, and then have humans try to solve the problems that the AIs were unable to solve. If a hacker is able to create a program that is able to select all squares with traffic lights with higher accuracy then a human that hacker will get far more money selling this program to Google, Uber or Apple then to use it to solve reCaptcha for malicious purposes. And this program would then be used to filter out all the images that an AI can easily solve.

They can!!! That’s actually what they are for. You think it’s to improve security, and it is, but you are also teaching the AI how to recognize these things. You notice how it changes over the years? Used to be just text, then objects, now objects within a scene.

They can depends how much $ you have to spend on your anti captcha bot. There are various services where are even humans that solve the captcha for you to receive points so they can exchange them for various things.