OH GOD
IT'S GETTING STRONGER
OH GOD
Other urls found in this thread:
you haven't seen shit yet
>select street signs
this shit again
>what kinds of street signs are there?
WHAT THE FUCK
>what kinds of street signs are there
ironically, this is the easiest problem for machine learning bots to solve, so I doubt that'll happen.
Holy crap, it actually rewards you for being thorough.
>so I doubt that'll happen
it's already happened, i only get that task like once a day but it's there
I switched to legacy captcha after these started popping up.
really, i've never seen that one before. Kind of surprising, considering machine learning can already caption images with a great degree of success. All you need to do to solve the "what kind" problems is a properly architectured RNN classifier on the image with a significant number of images trained.
You realize you can just sloppily select whatever they want you to? Like, in 4 clicks or so
I like that captcha best actually
How about this one
this
holy fuck sites should stop using google's captcha
That's much more difficult because it's currently a hard problem to "look for" street signs and understand that the street sign is in that space and contains the two disjointed parts of the stop sign and the street name signs.
It's even more difficult when the images are basically thumbnailed.
>not using legacy
you deserve it
Am I a good boy?
>Sup Forums is the only site using captcha
>not every cloudflare-enabled site has captcha
Forgot the captcha
press it a few times it will give you a select all stores challenge
I tricked it
First one was
street signs
cars
mail boxes
>using sites other than Sup Forums
you deserve it
we need to think of ways to trick it while we still can otherwise it'll get too smart to trick
You can trick the "select all" variant
Just pick what relatively resembles the object and ignore ones with tiny fragments
Here's a trick: whenever you are asked to select all images of something, you can always get by with selecting three images. Even if there are many, you only need three.
That one didn't work btw
Nothing, absolutely nothing is worse than the one where you click a few and then wait for it to sloooooooooooowly load a few more to click.
Sometimes it accepts one square from my experience
Also, it seems to make me do multiple ones even if I do it "correct" the first time
what is happening
Are you talking about the ones that fade out and then different ones appear in their place? I was sure they only happened if your connection was overloaded. Also, it seemed to go on forever.
what about the new one with the shapes?
Elaborate
answer correctly or they will start punishing you with grammar correction
the one in OP
That's nice.
Well, they're easy. You just select around sloppily and you're done
no fucking way
Is that a screenshot from a game or something?
We need to evacuate google as fast as we can
no clue, but i found it weird enough to save
i did it
wow these audio ones are kinda easy
test
>properly architectured RNN classifier
are you sure? I thought they were using a deep conv net
Preparing for google's android military™
They are rate limited tho, and they do get hard.
my tank lags what do
...
i tried legacy captcha like 2 years ago, but it was ridiculously hard to decipher
...
>hehe use legacy captcha because you can use legacy captcha for posting threads
Fucking retard poo in the loo pajeet fuck off.
Noscript v2 captcha is the best in my opinion.
ez
You only get one question and it's pretty much always easy.
With the script enabled I would often get "select X until there aren't any more of then", "select around X" or other stupid bullshit.
i'm sticking with audio captcha 2bh
i'm tired of redoing
works first try every time fucking insane
these new recaptcha variants seems to be very hard to solve for bot am I right
fucking audio captcha family how did i not use this until now
HOLY SHIT WHAT NOW IT WAS AND I SHIT YOU KNOW
"IT IS A WORD CONTRAPTION"
what the fffffuck
being phased out, won't work anymore soon
i can't get v2 to work with 4chanX anymore
damn son tho this audio captcha is comfy as heck
i love audio captcha
why is this so good
ok fine this time i couldnt understand but numbers are super easy so still good imo
fuck it last try
Thanks for all the (You)s. I really mean it when I say it's easy to solve. I understand why some of you might stick with audio captchas or the legacy captcha, though.
>tap the center of the street signs
And here's the video.
I might as well have posted a webm instead of a link to the file.
show me your Sup Forums X settings
with v2 it just tells me that i mistyped it
>HOLY SHIT WHAT NOW IT WAS AND I SHIT YOU KNOW
>that font rendering
What the hell?
native extension
That was when I decided to disable anti-aliasing but before enforcing a proper font.
seriously?
fuck this
With normal captcha I have to redo like 6 times sometimes trying to click trhough them
Yeah, something similar happened to me. I sometimes had to solve more than the standard two captchas. I'd always get a "Next" button and would end up solving more than 5 of the fuckers before I finally got verified. Those things were what made me switch to noscript v2 captcha.
that's why audio is superior
Where the fuck is the car
>If there are none, click skip
Is there an extension or a tamper/greasemonkey script that disables the scripts of all captchas, effectively achieving noscript captchas on all the pages that require them?
no lol the whole point of recaptcha is that bots can't bypass it easily
I just came to the realisation that those captchas are basically Google saying "our AI is the strongest AI in existence, come at us, bro".
I'm not asking for a bot that can bypass it, I'm asking for a script or an extension that disables javascript on the captcha. The same thing 4chanx has.
No, it's them using us to train their AI/neural network.
My guess is that they'll use this information for their shitty self driving car, object recognition software, etc.
test
Well obviously I get it that we are further training their AI. But think about it, if their AI can be used to verify is a client is a real person, it means nobody has AI strong enough to realistically bypass their machine-learning-based captcha.
What's the answer?
L E G A C Y
E
G
A
C
Y
> have legacy captcha enabled in 4chanx
> still get regular captcha only
CAN'T ESCAPE
>helping google build the AI that will fuck us all in the future
literally like people in the US training their replacement pajeets
test
>doing data entry for Google to shitpost on Sup Forums
>not just spending $20 on a pass via Bitcoin so you can post freely and post using a VPN
Negro stole it
Click the squares with the street signs. It'll work.
sounds like a personal problem
I'd rather help google than give money to this site
I remember getting a "select around X" one a few months ago, probably when they had barely any data, and they showed me an airplane. I drew a giant dick and it let me through immediately.
Enjoy your double botnet then.
This. In my experience it also uses this to "punish" me, so if I fail a bunch (by being impatient) I'll usually get stuck on one of these.
By that point I'll usually ragequit because no matter how many times I reload, the botnet will be suspicious and will either give me ones that make me wait like 20 seconds between each fade-in, or want me to select all street signs correctly like 10 times, and since I'm already impatient and frustrated I never get through either.
How do I get this with uMatrix? I allowed google scripts because without some of them it won't even load the iframe, but now I have no idea which ones to block to prevent the really crappy botnet.
>he trains other people's botnet for free
But their verification AI is just based on analysing data from other people solving it. If you had the dataset it would be trivial (relatively speaking) to solve the captchas.
That doesn't necessarily mean that google has an AI that can solve its captchas on its own.
"Anschlagfläche" can mean location of a terrorist strike
Google has models that achieve the greatest certainty and the best success rates on the "captcha" problems. And of course it is mainly because of the sheer volume of their training data sets.