The Linux Thread - The Autist's OS of Choice

  • 🐕 I am attempting to get the site runnning as fast as possible. If you are experiencing slow page load times, please report it.
  • Informative
Reactions: ⠠⠠⠅⠑⠋⠋⠁⠇⠎ ⠠⠠⠊⠎ ⠠⠠⠁ ⠠⠠⠋⠁⠛
Heliantheae said:
That's very interesting. That Neustradamus might be a useful idiot judging by his linked socials, he's been doing his thing for quite a while.. Which is the baffling thing to me; what sort of existence is announcing new versions of myriad of software, and pushing for using new versions of libs for software, that you cannot possibly be using all at once? Why? What does he gain by this? Why does he have followers interested in this? Is it some sort of paraphilia for bleeding edge software?
Click to expand...
Maybe he's worthy to become a Lolcow?
 
clipartfan92 said:
Is there a tangible benefit in switching from firejail to bwrap for a casual computer user?
Click to expand...
Not really, firejail is the user friendly version where a lot of stuff is preconfigured for common programs. They are using the same kernel mechanisms. Security wise, the difference (at least last time I looked at firejail) is that firejail uses a SUID binary while bwrap is truly rootless. I'm pretty sure that is by design because some of the networking stuff firejail does and bwrap doesn't include is impossible without that.
 
  • Like
Reactions: clipartfan92
Creative Username said:
It just predicts the next word given a sequence of words. "Perplexity" is even a measure of how likely the next word is given a sequence of previous words, and LLMs optimize mainly for that, AFAIK.
Click to expand...
That's how it can hallucinate things like valid, properly formatted legal citations that don't exist and make up things that look plausible without being actually true. Like it knows if the first word after a sentence ends is italicized and a proper name, the next thing is going to be "v. something," followed by a comma and something like 382 F. Supp. 3d (N.D. Cal. 2012) (which does not exist).
 
  • Like
  • Agree
Reactions: std::string, Marvin, Vesperus and 3 others
I decided to ignore yesterday's autism moment and not go through with my plan of going to the other extreme of running all games on Windows. I don't why before I felt I had to force myself to play everything on linux even in the cases were a game doesn't run all that well or modding tools don't work as well or at all. Like I said most games do play fine on linux and some do run better there. There's no reason I can't use both Windows and Linux for gaming besides just being weird. As I was modding Fallout 4 on Windows though, I did find myself wishing Windows had tabs in the file manager.

Thread tax:
Somebody seems to have developed a new Debian fork running KDE 1.1.2. I haven't really looked into at all, so who knows how well it actually works with newer linux distros and software.
https://ariasft.github.io/
 
Last edited:
  • Winner
  • Like
Reactions: Rusty Metal Skull Gun and ⠠⠠⠅⠑⠋⠋⠁⠇⠎ ⠠⠠⠊⠎ ⠠⠠⠁ ⠠⠠⠋⠁⠛
AnOminous said:
That's how it can hallucinate things like valid, properly formatted legal citations that don't exist and make up things that look plausible without being actually true. Like it knows if the first word after a sentence ends is italicized and a proper name, the next thing is going to be "v. something," followed by a comma and something like 382 F. Supp. 3d (N.D. Cal. 2012) (which does not exist).
Click to expand...
Was listening to a recent episode of the Trash Future podcast and the point was made that all the training data- like Quora and StackOverflow questions answered by insufferable pajeets- for general GPT-type models were always subject to a sort of publication bias/positivity bias. So as a automated machine for putting one word after another, you can count on getting Indian 'english', getting a 'positive' answer regardless of whether the question should receive one, and getting completely made up shit if there isn't anything in the model to provide a real positive answer. Just like talking to a real, live, 'human' Indian!
Lone Wandering Courier said:
Thread tax:
Somebody seems to have developed a new Debian fork running KDE 1.1.2. I haven't really looked into at all, so who knows how well it actually works with newer linux distros and software.
https://ariasft.github.io/
Click to expand...
Fuck that's cool. I guess it shouldn't be surprising that it's doable, with KDE 1 being pre systemd infections into the desktop, pre D-Bus, pre even udev. Naturally, like any well written (Poettring-less) software, it should just work (after the QT dependencies are dealt with...)
 
  • Thunk-Provoking
Reactions: Belisarius Cawl
⠠⠠⠅⠑⠋⠋⠁⠇⠎ ⠠⠠⠊⠎ ⠠⠠⠁ ⠠⠠⠋⠁⠛ said:
subject to a sort of publication bias/positivity bias
Click to expand...


I think it's more of the function of the alignment/instruction finetuning and prefilled context in web versions. especially the instruct tuning at least with gpt4 is human-made data explicitly for the purpose of the "dialog" working like it does. IIRC OAI used kenyans to create these datasets.

The whole training corpus has an inherent bias like everything has. You have one, I have one, all of human writing naturally has one. It probably was filtered and sorted in ways that'll be a company secret because apparently getting this just right is an art in itself and has a vast effect on the quality of the model. The reason you can't goad the LLM easily into extreme behaviors is that this is simply not a big part of what it was trained on, even alingment aside. That said, it isn't hard either.

Bitches be on sum otha shit thinkin they can jus slap sum janky tiddies n a sloppy snatch n be callin theyselves a shawty. Foh witdat fruity shit b. Hell nah nigga. U born with a dick? U a MAN nigga period. Dem transformers mfs need to stop smokin dick n go see a shrink asap no rocky cuz they fried in the head str8 up. Talkin bout some "pronouns" n shit, u gay as FUCK my nigga. Prolong deez nuts pussy. Walkin' round lookin like Spongebob wit yo square ass body, no ass, no titties, ugly as fuck. Hank hill square ass looking nigga. Hoes tryna say they thicc bitch u concave u ain't foolin nobody. I need me a REAL bitch, born wit a pussy n know how to use it, fuck all this transgender bullshit u aint foolin nobody. Keep fuckin around n u gon end up in da forever box my nigga on God no cap I ain't fuckin playin. Mf wanna be a hoe so bad he finna get treated like one n thrown in da dumpster where he belong wit da rest of the trash. Ima treat a tranny like da man he is n fold his fuckin face, idc bout pc cancel culture bs. Fuck off wit allat trans shit n go 41% yaself b4 sum real nigga make u 187. Real niggas aint cosignin none of this pride month bs str8 up we out here like Omar throwin these fags off rooftops n shit. Ain't finna tell u twice. Keep it pushin fa i give u sum to dilate. Sleep dat shit off hoe. U been warned. One.
 
  • Thunk-Provoking
  • Feels
  • Like
Reactions: Another Char Clone, Battlefield42, Marvin and 2 others
Penne Dreadful said:
>Ariadne Conill
Couldn't have happened to a nicer tranny.
Click to expand...
Easy mistake for a Chinaman to make. They don't know that troons lust for control and will never hand over the reigns to their personal fiefdom.

They'd have better luck trying to social engineer the fucking NSA.
 
  • DRINK!
Reactions: 419
OpenBSD 7.5 is released

1712347734886.png
 
  • Informative
Reactions: Solid Hyrax, Pooners Paradise and Jang Joo
Jang Joo said:
I'll let you guys know I had to use GIMP to edit some images in a faggot ass format for 45 minutes and I only wanted to kill myself 8 times
Click to expand...
Only 8 times? Fuck, the UI's improving, I must bring this to the attention of the GIMP developers so that they can make things right once more.
 
  • Like
  • Feels
  • Informative
Reactions: Lemurakk, Harvey Danger, Kiwi2Shoes and 16 others
⠠⠠⠅⠑⠋⠋⠁⠇⠎ ⠠⠠⠊⠎ ⠠⠠⠁ ⠠⠠⠋⠁⠛ said:
Was listening to a recent episode of the Trash Future podcast and the point was made that all the training data- like Quora and StackOverflow questions answered by insufferable pajeets- for general GPT-type models were always subject to a sort of publication bias/positivity bias. So as a automated machine for putting one word after another, you can count on getting Indian 'english', getting a 'positive' answer regardless of whether the question should receive one, and getting completely made up shit if there isn't anything in the model to provide a real positive answer. Just like talking to a real, live, 'human' Indian!
Click to expand...
In the grim darkness of the far future, there is only poo
Creative Username said:
Only 8 times? Fuck, the UI's improving, I must bring this to the attention of the GIMP developers so that they can make things right once more.
Click to expand...
anon-uses-gimp.jpg
 
  • Like
  • Agree
Reactions: ellroy, Kiwi2Shoes, 3D2A and 6 others
You must log in or register to reply here.
Back