File: ／ｆ／ｕｃｋ ｙｏｕ ｍｏｏｔ．.swf-(4.13 MB, 720x480, Other)
[_] Anonymous 10/31/19(Thu)23:12:43 No.3407036
just a reminder that /f/ was the last board without captcha and never needed it to fight bots.
Marked for deletion (old).
>> [_] Anonymous 10/31/19(Thu)23:18:24 No.3407038
I miss legacy captcha. these street signs and crosswalks drive me nuts.
>> [_] Anonymous 10/31/19(Thu)23:25:47 No.3407039
>>3407038
I fucking know, right?
>> [_] Anonymous 10/31/19(Thu)23:46:57 No.3407040
>>3407038
I miss typing "nigger" for the unknown word
>> [_] ITS THE SPOOKY MONTH 10/31/19(Thu)23:47:07 No.3407041
>>3407038
Same here, that shit drives me up a wall. Many times the captcha itself is inaccurate, and you
have to retake it several times until you finally guess it right.
>> [_] Anonymous 11/01/19(Fri)00:13:30 No.3407046
>>3407038
The worst is when you have to click till there aren't any more left and it takes 10 seconds for
each picture to appear.
>> [_] Anonymous 11/01/19(Fri)00:43:36 No.3407050
>>3407039
>>3407041
>>3407046
Install 4chanX and tick 'Force Noscript Captcha' in the settings.
>> [_] Anonymous 11/01/19(Fri)01:04:14 No.3407052
>>3407038
>not taking immense joy from purposefully fucking up Google's slavery drive
Oh yeah, that child and mother is definitely a traffic cone. Go fuckeddy fuck yourself, database.
>> [_] Anonymous 11/01/19(Fri)01:19:35 No.3407055
I feel that when /f/ shuts down it will truly be the end of an era on the internet
I used to come here everyday, now I come here every few months just to cry
>> [_] Anonymous 11/01/19(Fri)03:03:24 No.3407058
>>3407055
Why would /f/ shut down?
>> [_] Anonymous 11/01/19(Fri)03:48:46 No.3407064
>>3407052
this
i'm willing to waste the extra time, just to send them a bunch of wrong info, but make it seem
semi-accurate to where they can't just algorithm it out thinking it's just someone messing around
>> [_] Anonymous 11/01/19(Fri)04:12:23 No.3407068
>>3407064
At least I've figured out the trick to their new captcha. I should make an explanatory
infographic just like typing in nigger into the other ones.
Though, now I feel a little bad. The prior reCaptcha projects were about digitising books, and
while it might be funny to randomly stumble over "nigger" while trying to read a late 18th
century treatise on proper medical care, it's a bit puerile.
There is, however, nothing puerile about fucking up Google's plans, whatever they may be. Fuck
ABC corp.
>> [_] Anonymous 11/01/19(Fri)04:27:49 No.3407069
>>3407058
Because most browsers are discontinuing support of flash, anon.
>> [_] Anonymous 11/01/19(Fri)04:28:56 No.3407070
>>3407069
As in completely removing it? Or where we just have to change some more about:config settings?
>> [_] Anonymous 11/01/19(Fri)04:59:51 No.3407073
>>3407069
Then I don't want to use most browsers. Fuck HTML5.
>> [_] Anonymous 11/01/19(Fri)06:32:21 No.3407074
>>3407050
Or use a user-script such as MoreCAPTCHA if you already use a user-script manager like
Tampermonkey
>> [_] Anonymous 11/01/19(Fri)07:54:48 No.3407077
>adding captcha at random
>not during the years of zeitgiest spam
>dropping flash post limit to two
Fuck you moot, worst things to happen to this board.
Do we have any alternatives in case hiroshimoot ever nukes the board (heh, japanese nuking things)
>> [_] Anonymous 11/01/19(Fri)08:37:57 No.3407079
>>3407068
I can't really figure out how to trick the new captcha. Every now and then I get away with
calling a boat a streetlamp but usually it knows what I'm up to.
I can however fake everything in the audio captcha but that's too time-consuming.
>> [_] Anonymous 11/01/19(Fri)08:43:12 No.3407082
>>3407079
It's easy, but you can't really do it in any of the "fade away" bullshit captchas. You can do it
in both the 3x3 whole picture sets and the 4x4 region grid selects. It's 50:50 which one is the
control image in those, but you can generally get the knack of it; notably, the control image
will always be of the stated object, and that stated object will always take up more than two
squares (but exactly two squares can happen, though they're rare). The trick with the 3x3 ones is
that spread about them will ALWAYS be 3 control images, but frequently 4, and at least three
negative control images. If you select any three images that match, there is a pretty good chance
that you've got all the controls you need, so you can introduce error through occlusion.
>> [_] Anonymous 11/01/19(Fri)13:17:34 No.3407091
>>3407069
That doesn't mean /f/ will shut down you doomsdayer
>> [_] Anonymous 11/01/19(Fri)16:41:50 No.3407106
>>3407038
90% certain i'm a computer at this point
>> [_] Anonymous 11/01/19(Fri)16:58:07 No.3407107
>>3407068
>>3407082
do the infographic faggot
>> [_] Anonymous 11/01/19(Fri)17:14:53 No.3407109
>select the crosswalk
>It shows a picture of rumble strips
Fuck jewgle.
>> [_] Anonymous 11/01/19(Fri)20:44:14 No.3407122
>>3407052
>>3407064
most likely they give the exact same captcha to 1000 people around the globe and unless at least
700 click the exact same tiles they see it as unreliable reference data and don't let that sample
touch their AI training algorithm. so your trolling does nothing.
>> [_] Anonymous 11/01/19(Fri)20:46:35 No.3407123
>>3407068
also the nigger trolling did absolutely nothing as well, you think reCaptcha wouldn't notice a
thousand people using the exact same incorrect word? those were simply disregarded too
>> [_] Anonymous 11/01/19(Fri)22:29:49 No.3407131
>>3407122
A lot of times you have to click things that you think the average person would. I used to fail
it a lot because I was being too accurate. Like with "select all traffic lights", if there's
small part in another square, I'd select that too. Or if there were traffic lights in the
background, I'd select those too. So I'm sure if a lot of people fuck with them, the AI would get
less and less accurate. "Select a motocycle" but it shows a bike and it still being accepted
proves that well enough.
>> [_] Anonymous 11/01/19(Fri)23:04:46 No.3407134
>>3407122
fuck you man it makes me feel better.
>> [_] Anonymous 11/02/19(Sat)04:52:55 No.3407149
>>3407073
Based pale moon chad
>> [_] Anonymous 11/02/19(Sat)06:59:31 No.3407151
What's worse is goot can't keep up the credit card payments up. Going back to posting with
captcha is fucking hell after not doing them for years.
>>3407036
I love this flash way too much.
>> [_] Anonymous 11/02/19(Sat)09:07:54 No.3407155
>>3407131
pretty sure they track behaviour in IP ranges, for example i used to get a bunch of "click
traffic light" and then started getting "click crosswalks", but i kept clicking only traffic
lights and they recognized that i was human anyway, assuming that i was just used to click them.
this was without cookies as well, which i always clear every day (and i don't use chrome so i
dont get spied on without cookies), so they only had IP and browser user agent to go on.
>>3407134
ok, it does something. it makes you feel better, but that's about it.
>> [_] Anonymous 11/02/19(Sat)09:10:19 No.3407156
>>3407070
completely removed
>> [_] Anonymous 11/02/19(Sat)09:32:09 No.3407158
>>3407155
>>3407122
but if those 700 are the troll ones it could just work
so don't despair anon >>3407134 your work might accomplish great things
>> [_] Anonymous 11/02/19(Sat)11:50:17 No.3407167
>>3407158
sure but they wouldn't send all of those 700 tests to the same website, especially 4chan, or even
the same part of the world so it's a pretty slim chance to get a hivemind that trolls the same
way. and if 0.0001% of the AI's total training data is corrupt it doesn't do much
>> [_] Anonymous 11/02/19(Sat)12:02:24 No.3407170
>>3407167
But there is still a miniscule chance that it gets through. While reading through digitised,
obscure books, I did, in fact, stumble over several niggers.
And believe me, that one in a million chance matters. Database is everything in big data. If we
can introduce any sort of coordinated error into the database, they'll have to manually crawl
through it, or develop a secondary verification method (like another captcha).
Imagine manually crawling through *terabytes* of data to find where an obsessed group of assholes
made your network think that any instance of red is pedestrian.
And that's the crux of the issue. You can track certain efforts to manipulate the quality of the
database, but not others. Complex rules like "for any new query that is asking to identify
traffic cones, select all grass" are a lot harder to debug and eliminate than "select all the
squares on the right hand side". And we humans are pretty good at following those rules, but not
perfect - which introduces additional error, additional uncertainty.
What I have been doing so far is selecting the 4x,2y square on all the new images because it is
relatively close to the send button. What I should be doing is making up some simple rules to
follow for certain image types that can then forever fuck up a database if multiple responders
follow them.
