4chan - the Internet hate machine

Will the 4chan hack be the end of it?

  • Yes, goodbye forever 4chan

    Votes: 912 18.2%

  • No, they will rise from the ashes, stronger than ever

    Votes: 256 5.1%

  • This will rattle them but it will be forgotten about next week

    Votes: 2,143 42.8%

  • I am just here for the janny phonebooking

    Votes: 1,018 20.3%

  • What the fuck is 4chan

    Votes: 181 3.6%

  • Yotsuba&!

    Votes: 498 9.9%
  • Total voters
    5,008
Next Highlight
Here's a compilation of threads that would be good to read for rapefugees to get a better understanding of the website:

https://kiwifarms.st/threads/principles-of-the-kiwi-farms.127111/ | The principles of the Kiwi Farms
https://kiwifarms.st/threads/we-hav...t-being-able-to-figure-out-thumbnails.113862/ | How and why to thumbnail your images.
https://kiwifarms.st/threads/doxing-in-2025.208564/ | Usage of the word "Doxing" on the forum.
https://kiwifarms.st/threads/clarification-the-kiwi-farms-is-strictly-safe-for-work.196115/ | Clarifying rules on NSFW content.
https://kiwifarms.st/threads/a-guide-on-ratings-stickers.90703/ | A guide to stickers and what each of them means.
https://kiwifarms.st/threads/if-you-dont-care-dont-post.190719/ | 'If you don't care, don't post'
 
  • Informative
  • Like
  • Winner
Reactions: ita, relobotomized, exhausted and 53 others
gagabobo1997 said:
Here's a compilation of threads that would be good to read for rapefugees to get a better understanding of the website:

https://kiwifarms.st/threads/principles-of-the-kiwi-farms.127111/ | The principles of the Kiwi Farms
https://kiwifarms.st/threads/we-hav...t-being-able-to-figure-out-thumbnails.113862/ | How and why to thumbnail your images.
https://kiwifarms.st/threads/doxing-in-2025.208564/ | Usage of the word "Doxing" on the forum.
https://kiwifarms.st/threads/clarification-the-kiwi-farms-is-strictly-safe-for-work.196115/ | Clarifying rules on NSFW content.
https://kiwifarms.st/threads/a-guide-on-ratings-stickers.90703/ | A guide to stickers and what each of them means.
https://kiwifarms.st/threads/if-you-dont-care-dont-post.190719/ | 'If you don't care, don't post'
Click to expand...
No.

Ok fine I'll check it out.
 
  • Feels
  • Like
Reactions: Jump and gagabobo1997
Bearmumrik said:
Apu is cool and was made by a Finnish guy.

View attachment 7230781
Click to expand...
deckthread.webp

apu is cool and good
 
  • Feels
Reactions: Didnt Use Real Email Once, Lexi-karabiner, keklmao and 4 others
wmaf1488 said:
FreeBSD is well known as a server OS . It is #2 after Linux
Click to expand...
And? MacOS is the #2 Desktop OS and it has a 15% usage share. Doesn't mean it's popular.
FreeBSD used to be the OS of choice because of the superior network stack. Now it's used because IBM & co made Linux almost as shit to work with as Windows
 
Unlicensed Shitposter said:
How does it feel?
How does it feel that KF has outlived fucking 4chan?
Click to expand...
Here, since you brought that up. Before any of the locals pile on me for not keeping abreast of all this stuff, wasn't this somewhat similar? Most of the internet was believing some tranny had intelligence agency levels of clout chasing this place off the internet. I'm not buying this 'hack' for a second, not just because it feels adjacent and similar to that but because things were weird for a while on /pol/ and I don't mean the usual moaning about the absolute state of the place. Posters were replying to anons making remarks about their personal lives, just subtle enough to be plausibly deniable to third-party observers but on the nose enough for the anon being replied to to pick up on.
 
  • Optimistic
  • Thunk-Provoking
  • Like
Reactions: keklmao, Punished Magician and FagofCeremonies
Jimjamflimflam said:
Christ. For how fast the party van is going to take them away, why are pedos so intent on sharing their shit.

Do miss idle scrolling x. Also should have saved some of the cool wallpapers posted when I had the chance.

Like tears in the rain~~~
Click to expand...
the thing about paedos is they do it for the danger. i doubt many are REALLY into kids, they just want to be as bad as possible, and society places CP at the top of the list of bad, so thats what they do. but whats the point in being bad if nobody knows how bad you are. so they spam CP all over they place to prove how bad they are and hopefully get caught like the attention whores they really are.

the cuck and bbc posts are just mossad jews in troll farms under tel aviv. when you think about it neither bunch are better than the other
 
  • Disagree
  • Informative
  • Optimistic
Reactions: Foreverial, Qonas and Jimjamflimflam
ringtones said:
what I know about voice recognition suggests it does work this way. I think the images could be somewhat-sanitized by running them through a color normalizer that clamps values to near-neighbors, but I don't know what to do about the audio steganography

basically the way text-to-speech works now is that there's a few master databases (I actually only know of one but I assume there are others) of high-frequency (above 15khz) phonemes that are sounded at the start of the enunciation of a word. Like if you're going to say "Sneaker" the inaudible high-frequency hiss of the S is distinctly different from the high-frequency hiss in "Shoe" and the AI can pick up on that. The models they use for speech recognition only key on the high frequencies because you can get high frequency data faster than low frequency data due to the wavelength of the pressure wave carrying the sound. Also it's very cheap (in terms of power usage) to do this rather than fully-parsing voice input.

So almost all of the data used by phones and AI assistants to parse voice input is inaudible to all human ears.
Click to expand...
Oh yeah. I'm talking purely from the image generation side of things. I imagine audio stuff works really fucking well but I'm not actually sure super worthwhile when, for generic data, there's almost definitely more than enough material in existence for a general model and if you're training on a super specific voice/tone you're going to be working with a full bespoke dataset anyway.
But on the image side of things, a big fallacy of these types of attacks is the assumption that all this shit is being picked up blindly and isn't being pruned or curated at all before it gets trained on. And that really isn't how any of this works - for anything corpo-level generative model you bet their fucking ass they are, at minimum, sending everything through vision models to check shit and tag everything consistently and checking things for adversarial noise is pretty trivial. And for non-corpo efforts if you really absolutely need to train on someone doing this shit there are ways of slightly reversing it. And, hey, if an artist is retarded and baking nightshade into all their shit and making it look worse for their audience, and the model ends up reproducing the same awful looking nightshade look... Is it really preventing the training from "working" at all?
 
You must log in or register to reply here.
Back