If the botnet already knows the answer to these things, how does my input 'help' them in any way?
Like why am I picking street signs? Why am I entering in people's home addresses? If I make a bad decision on purpose, like operation renigger, it somehow knows already so what is the point? Is reniggering impossible now? Is that the reason they went full-retard with reCAPTCHAâ„¢ now?
Why can't we have more human-readable/friendly ones if there's no point and it isn't training a recurrent deep neural memenet? Isn't *accurate* data necessary to train one of those? How can you train on data with the responses from a captcha?
Someone explained this to me one time but I'm dumb.
>>55622243
The point of reCAPTCHA is that it doesn't know all, only parts of it. Like the two word thing, it only knows one of the two words and is trying to learn the second. That's why you for a long time could enter "nigger" instead of the second word.
The image recognition thing is related. Google is using some machine learning algorithm to recognise pictures of street signs. There are often N pictures of street signs, but only N-1 pictures are actually required (because it tries to learn about the last one).
>>55622243
The botnet doesn't already know the answer. In your picture, the botnet knows the answer to either the first word or the second word. You don't know which word the botnet knows, so you have to answer both to the best of your abilities. After a while, the botnet takes the most common answer to the unknown word.
>>55622321
>After a while, the botnet takes the most common answer to the unknown word.
Ah yes, this was the piece of the puzzle I was missing. Thanks!
>>55622243
If you are the first person to enter something differently, the botnet says you are a bot in that case.
How can one post on 4chan without having to allow google.com, gstatic.com scripts/requests for verification?