Tim Poole

The computer already knows what that word is, and it's trying to test whether or not you are giving it accurate data, not whether or not you're a robot.

It was actually not about, if you're a robot, I mean, it's an ancillary benefit.

What it was trying to do is the second word we show you, we can't read, and we want a data set on, we want a trustworthy data set on what that word says.

So basically what they were doing was scanning books.

And if the computer said, this word is unidentifiable,

They would, you'd log into a website.

You'd see this thing asking you for a robot.

And then you'd say, oh, it says Shillelagh doorknob.

It knew Shillelagh.

It didn't know doorknob.

And the purpose of it was to say, if the first word is correct, they likely put in both correct words.

So 4chan created an operation called re-N-word, but not N-word, the actual word.

And they said to exploit the system,

The first word is always the code.

The second word can be anything you want it to be.

So what happened was, and I'm pretty sure this actually was confirmed.

4chan started going to every CAPTCHA and they would put in the first word and the second word, they would put the N word.

And so what ended up happening was in some of these natural language processing sites, the N-word started popping up randomly in places where it shouldn't because the computers were being trained improperly because people were exploiting it.

But they quickly fixed it.

So it may just be an urban legend.