THERE ARE NO STREET SIGNS GOOGLE YOU FUCKING RETARD AND WHEN I TELL YOU THERE ARE NONE YOU SAY I'M WRONG. WHY YOU EVEN WANT MY INPUT IF YOU JUST SAY IT"S WRONG ANYWYAS! REEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE
I know that feel.
It wouldn't accept having all the squares selected.
>not using legacy captchas
>actively training googles ai for free
This baffles me as well. Isn't the point of these to help the engine learn better image recognition? Why would it tell you you're wrong?
>>53527093
I think these just take the aggregate user input and use the average to determine the answer.
>>53527093
I use the old CAPTCHA, I'd rather just type a number every time than to test my luck with the new captcha
>>53527093
>working
>for free
Use the legacy CAPTCHA m8. Often you can tell which is the image they don't know and just enter whatever (I stick to the good old 'nigger').
>>53527487
>actually believing what you're doing with captcha is worth a wage
>being on a technology board and being unwilling to contribute towards improvement of technology
>>53527487
I just get numbers when I choose it
>>53527200
legacy captcha is the same, but with house numbers or street names
>>53527487
>I stick to the good old 'nigger'
And thats why Skynet will kill us
>>53527695
>skynet
>captcha was "stop 1997"
Is it just me or are over 50% of the numbers in the legacy captcha the same every fucking time?
I'm studying the tensorflow code and looking into a way to bypass the new recaptcha. I think I found a vulnerability that might allow us to fool it. Stay tuned.
>>53527748
What about ocring the old capcha?
>>53527673
you can actually do the legacy CAPTCHA with neural networks (see: MNIST) so it's useless to Google
>>53527748
m8, everyone and their mom can pass MNIST with convolutional neural networks with 99.8% accuracy (0.2% even humans can't read either)
>>53527180
Choose only 3. Always choose only 3. Sometimes you can choose 2.
op is a faggot bad photo shop is bad you can cleary tell if you lookat the sqaures over the fire hydrent
>>53528387
You new here?
>>53528105
humans only use 10% of brain
>>53530007
When will this meme die
>not using legacy captcha
>>53530038
When you do. Do it for the greater good.
I got the same thing a while ago
>>53527093
>>53530007
Is that why a teeny-tiny brain biopsy can render you a vegetable for life?
>>53532191
>>53532191
>report a problem
>>53532191
But she's from a better anime.
>this new captcha will make it easier to post, as it can tell you're a human so you have to solve less dumb shit
>for the first week auto verify
>now just always require something solved
I love the Bot.net
>hungry as fuck
>all captchas become delicious food
Fuck you captcha!
>>53528435
no im just a fag who has a pass
>>53527180
Has anyone seen my red stapler desu senpai?
>>53527093
>not using 4chan premium
>select an image
>it dissapears
JUST
U
S
T
>>53527093
>>53534753
That's what made me go back to legacy. Their new thing is pointless.
>>53533447
>all squares w/ a pingas
Google,you fucking faglord.
>>53527093
Select the fire hydrant. It's a sign not to park there
>>53535072
I always found this one retarded. They're not called pickup trucks ffs, they're calling bakkies.
>>53535522
That's really not why i took the screenshot anon
>>53534538
yeeeeeeeeaaaaaaaaaaaahhhhhhhhh
so if you could come in on saturdayyyyyyy
that would be greeeaaaaaaaat
>>53527093
>REEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE
what a fucking poo in the loo
I like how the captchas are now more about training google's AIs with what people think the AI would think is right, rather than what is right.
>>53535208
teach me how
>>53527093
>Fucking Google thinks Burgers and Sandwiches are the same thing
>>53535660
Burgers are sandwiches but sandwiches aren't burgers (unless they are).
>>53535576
Sure boss. What time on Saturday do you need me in?
>>53533447
At first glance it's box ten, but if he's tucking it behind it's borderline
>>53527093
I wish 4chan would make their own captcha server.
>>53527093
How do the new captchas (where you train the neural network) work?
Specifically, if you need to show it what the correct answer is, how does it know whether or not you're right? If it already knows the correct answer then how does answering correctly train the neural network?
>>53535754
that would be magnificent
>>53535994
I GUESS they insert X actually correct things from a source, apply their thing to guess, then we "choose the correct ones" that fulfill some base requirements that when fulfilled submits and compare the results of us vs it.
That's me guessing though.