How accurate is this anyways?
What DOES google use reCAPTCHA data for anyways?
nice try moot.
also how the fuck does solving captchas make an AI "learn" shit ?
>>7849262
I never bothered to read up on what they use the picture captcha for but as far as I remember the word captcha was used to improve the accuracy of scanning books and converting them to text
>>7849269
https://en.wikipedia.org/wiki/Active_learning_%28machine_learning%29
Why else would it give you "captcha incorrect, try again" for choices that are obviously correct?
in b4 retard, don't act like you've never picked 4/4 street signs and have it tell you "wrong answer"
Labeled data is the most valuable commodity there is when it comes to supervised machine learning. The data being gathered by reCaptcha is worth a fortune.
>>7849262
It's correct.
That being said, someone already broke captcha and you can get captcha solves for like 0.001 cents per solve.
>>7849797
Not broken, you're just paying indians to solve them
>>7849262
This was known since the days of the original captcha.
Also, with this system, every time I post I try my best to choose incorrect answers that the system will for some reason just swallow, knowing that I am fucking up that AI.
There, just got it to swallow a fuckign empty grassland for a tree. topkek
>>7849815
Often I find answering correctly is less likely to work than answering with the dumb easy wrong answer
>>7849305
and somehow that only happens with said street signs. Never with pizza's or pick up trucks or trees
I am glad my ability to recognize trees with 80% accuracy is important information.
>>7849815
If you are going to do some work for free against your will, at least do it wrong.
>>7849262
AI is the next step in evolution, deal with it.
>>7850694
>Is a human
>Actively proclaims that humans are done so open up the doors for AI
topkek so that is what being a cuck is like.
>>7849262
Not necessarily maybe some percent of that shitposting is deliberate to fool that AI by feeding it mumbo jumbo. I.e. banananas instead of apples.
>>7850739
I wonder if a computer can learn to feel self-doubt or that it is being tricked.
