do you support telegram @Suiseiseki ? i saw you posted it on there today

@maija I do not support Telegram and do not use it.

That wasn't my post and I wasn't the one to make such rendering - I merely reposted such GNUKE.

If I was going to go so far to make a 3D rendering, I would also make a animation showing it being used.

@Suiseiseki why don't you support telegram? it's open source

@Suiseiseki @maija Sir, a second GNUKE just dropped on the fediverse.

@maija GNU/JIHAD AGAINST "open source" AND ALL OTHER FORMS OF PROPRIETARY SOFTWARE!!!

FREE SOFTWARE!!!


The official version of the Telegram software is proprietary software and contains additional proprietary malware - they only get around to releasing the source code every few months (I figure they need to set aside time to remove their hand customized proprietary malware before doing a source release that still contains proprietary malware from other groups).

You can't even make a telegram account anymore without running the proprietary software release; https://f-droid.org/en/packages/org.telegram.messenger/

It's also SaaSS, as the server software is proprietary and also unpublished to ensure you can't host your own server.

@Suiseiseki just use telegram-foss build and an exising account, that makes sense, the backend is nonfree anyways, but they all are, you can't replace a centralized backend, source code there doesn't matter because its entirely unverifiable anyways

@maija >and an exising account, that makes sense
That doesn't make freedom sense when there are freedom-respecting chat protocols.

>the backend is nonfree anyways, but they all are
There are many fully free chat protocols with free clients and servers - for example IRC and XMPP.

>source code there doesn't matter because its entirely unverifiable anyways
Freedoms 1 & 3 don't matter because the software is unverifiable?

@Suiseiseki there aren't any good freedom respecting chat protocols by your definition

> There are many fully free chat protocols with free clients and servers - for example IRC and XMPP.
thtey dont have even a fraction of the featureset, irc is too limited and complicated for your average person, xmpp doesn't have good clients outside of android (no dino and gajim are not)

> Freedoms 1 & 3 don't matter because the software is unverifiable?
yes because they can justt give you fake misleading code

@prettygood @Suiseiseki yeah its fine for the outcasts but telegrams audience is the general public (mostly eastern europe)

@Suiseiseki pidgin has horrific xmpp compatibility, and uploads and omemo shit tends to break and crash all the time even moreso than others. gajim and dino are maybe the most usable and breaking least on pc but only conversations just works as a client.

also xmpp doesn't have all i need and irc can't send files without external services or offline users without a bouncer. no e2ee without a nonstandard plugin. xmpp works okay but omemo breaks a lot and lacks some privacy and other features, lacks good room moderation and permissions, some oher stuff that will be forever stuck as unused XEPS because it hasn't gotten with the times much since the 2k's

> I can quite easily determine by inspecting the sources and seeing if they compile to a similar binary, if I have been given the real source code, or have been given incomplete, non-corresponding source code.
no you cant compare it to the server code you dont have access to the server binary

@VIPPER @Suiseiseki if i had the time and money i'd do it because unlike 99,9% of fosstards i'm pragmatic and shit and no driven by ideology and pragmatism to the point of ideologically excusing jank and buginess and wanting dumb shit like not enough features gnome too much kde too complicated most linux users, i'm the rare breed to understand the importance of ux basically is what i am saying.
just give me a small loan of a million dollars i'll work on it and pay some quality contracters to write out he code for me (with the help of foss chat ai like openai to speed things up) and i'll be like pavel durov on this shit just you watch me i'm the greatest

@maija @Suiseiseki I think there’s still a big difference between having the source code closed vs available. Closed source is a huge red flag - like, what are you trying to hide? Are you doing something sus? Nothing to hide, nothing to fear, right?
The absence of this red flag doesn’t mean you should blindly trust the software (as e.g. NPM shows), but makes establishing trust easier especially in conjunction with other possible signs of trustworthiness.

they can justt give you fake misleading code

Reproducible builds would help ensure that they can’t just put random bullshit in the binary because anyone can independently verify that it matches the source code (by obtaining the exact same binary from the exact same source code). Won’t help much if the source code itself is malware, but it’s still better because there’s less points at which malware can be introduced.

@maija @VIPPER GNU/JIHAD AGAINST "FOSS"!!!!

>i'm pragmatic
Free software is pragmatic idealism that is commonly dismissed as they assume that GNU/Freedom is impossible.

If only they looked they'll realize it has been implemented.

>jank and buginess
IRC is neither janky or buggy.

>gnome too much kde too complicated most linux users
Using the GNOME and GKE desktop environment is much easier than using windows janky and buggy DE, although there are better DE's like xfce4.

Many such cases; https://www.gnu.org/gnu/gnu-users-never-heard-of-gnu.html

>i'm the rare breed to understand the importance of ux
Real user experience starts and ends at the software respecting the users freedom.

If it is proprietary malware, it will never have a good user experience, as it doesn't even respect the users freedom.

I love the user experience of GNU/Freedom.

>just give me a small loan of a million dollars i'll work on it and pay some quality contracters to write out he code for me
You'll probably hire proprietary software contractors that will demand the software is proprietary, or will make it depend on proprietary libraries, so no thanks.

>with the help of foss chat ai like openai to speed things up
"Open"AI is a proprietary software and SaaSS company that doesn't even provide the source code to GPTX - they run it on their computer and you have to run proprietary JavaScript and pay to access it.

LLM's effectively copy-paste input software and most of the input software is garbage, thus what you'll get out is garbage that also infringes copyright, meaning that the user will not be able to exercise the 4 freedoms and therefore the software is nonfree.

>i'll be like pavel durov
Just like that proprietary software developer, yes.

>just you watch me i'm the greatest
You aren't.

@kimapr @Suiseiseki closed source isn't a red flag its standard practice because its a lot easier than cleaning up your code, fixing licensing issues, covering an entire complicated backend infrastructure, ettcetcetc
also, there's no need for suspicion when there's good old greed, keep your competitors away dont give away how you make it work

>Reproducible builds would help ensure that they can’t just put random bullshit in the binary because anyone can independently verify that it matches the source code (by obtaining the exact same binary from the exact same source code).
it's the backend you dont have the binary or the source code not under gpl either

@Suiseiseki @VIPPER >You aren't.
your less of a competent programmer than I am

>"Open"AI is a proprietary software
no its not that's why it's called open

>LLM's effectively copy-paste input software and most of the input software is garbage, thus what you'll get out is garbage that also infringes copyright,
nott at tall it uses matrix operations to find a combination of words that matches best with the combination of words from the given input plus some randomisation i believe, the ai is trained on billions of generated axis for the similarity of different word combinations and eventually the output makes sense, extremely dumbed down to be clear. this is only copying in the same sense that having seen code before to know what you're doing is copying. it's closer to inspiration. if they just copy and pasted no code of use would ever be generated because most all prompts will be for something new. are you stealing code by following a programming, a best practice, a general organizatiton structure, doing common operations in the the same order? no of course, neither is the ai. it's only as much of copying as having the same thought process as somebody else, and the training data will have so much overlapping ways of doing things it wouldn't just only ever do something one way. world courts agree with this.

@maija @VIPPER >your
11/10 bait, made me reply.

>no its not that's why it's called open
Link me to the source code of GPT4 or whatever then.

>are you stealing code by following a programming,
Copying code is not theft, but it's copyright infringement if you don't follow the license terms.

If you copy free software and turn that into proprietary software, what you are doing is theft of freedom.

>this is only copying in the same sense that having seen code before to know what you're doing is copying
You look at the output and then search for it and then surprise, surprise you find it's pretty much a 1:1 copy with some trivial changes like slightly different variable names or a wrong license on top (as there's a certain amount of noise).

If you know what you are doing, you can write good code and you won't need to go and copy-paste someone else's code.


>world courts agree with this.
Courts mostly agree with whoever has the most money, even if they are agreeing about something false.

@Suiseiseki @VIPPER > Link me to the source code of GPT4 or whatever then.
i would but you need to be a researcher iirc and the trillions of lines of texs processed is probably too fast for your connectiotn anyways

@Suiseiseki @VIPPER also
>If you copy free software and turn that into proprietary software, what you are doing is theft of freedom.
like i just explained it doesn't copy anything of substance that's why its not considered copywriten ouput by the courts

>You look at the output and then search for it and then surprise, surprise you find it's pretty much a 1:1 copy with some trivial changes like slightly different variable names or a wrong license on top (as there's a certain amount of noise).
this doesn't happen

@maija @Suiseiseki It doesn't happen right now, but was happening nack when copilot first got released that it would "generate" code from gpl projects, ever since microsoft changed copilot so that it prioritises code from mit, bsd and apache type licenses for ai completion/generation and it will only ever spit out gpl code when the license of a project is also gpl, tho gpl versions incompatibility between the specific gpl license used by the project and the code copilot spits is still bound to happen.

@EdBoatConnoisseur @Suiseiseki that would happen if you're already copying part of it or asking for specific things that very much already exist, in a proper llm. i assume copilot literally just looks through code that already exists and tries to parse what snippest are relavent. its not quite the same thing this would have more issues. not in the states atleast though which is the relavent jurisdiciton of copilot, because the courts ruled that these generative ai outputs are public domain iirc

@maija @EdBoatConnoisseur "copilot" is the same kind of LLM technology as any other LLM.

>i assume copilot literally just looks through code that already exists and tries to parse what snippest are relavent.
That is a simplification of what LLM's do on any sort of text.

>the courts ruled that these generative ai outputs are public domain iirc
It's really incredible freedom washing if you can take GPLv3-or-later software and then can claim that a 1:1 copy of substantial parts with some trivial changes like slightly different variable names is public domain.

I haven't heard of a court case ruling that outputs are public domains, just some non-court decision.

@Suiseiseki @EdBoatConnoisseur > "copilot" is the same kind of LLM technology as any other LLM.
they're not a unified things detatils very that why there are more than one

> That is a simplification of what LLM's do on any sort of text.
they dont match anything long enough to be considered a "snippet" by anybody sane

> It's really incredible freedom washing if you can take GPLv3-or-later software and then can claim that a 1:1 copy of substantial parts with some trivial changes like slightly different variable names is public domain.
weren't you just earlier saying it only spits you terrible code?

> I haven't heard of a court case ruling that outputs are public domains, just some non-court decision.
https://www.reuters.com/legal/ai-generated-art-cannot-receive-copyrights-us-court-says-2023-08-21/
this is an upholding of previous decisions as well, it's not supreme court but it is case law

@maija @EdBoatConnoisseur >weren't you just earlier saying it only spits you terrible code?
Most software is terrible, so most inputs were terrible, so most outputs are terrible - although some of the inputs happened to be good.

>this is an upholding of previous decisions as well, it's not supreme court but it is case law
The ruling was that people who were not the author of a work cannot be the copyright holder.

I see nothing in that article that mentions that such works are public domain.

Really, the copyright holder should be whoever held the copyright on the input works, as those are the people who actually made the work.

@Suiseiseki @EdBoatConnoisseur i would like to sue you for using a text input snippet from my work, the post you are replying to. it happens tot be under my copyright and sole ownership by default. by pressing send i am autthorizing distribution and storage through akkoma and activity pub, not modification and further redistrtibution

@Cyrillic @maija @EdBoatConnoisseur People just can't help but to cope over the world's best software, beautifully crafted with GNU/Meatballs.

@Cyrillic @Suiseiseki @EdBoatConnoisseur that's true its notoriously bad

@maija @EdBoatConnoisseur I was quoting, for non-profit purposes, which doesn't require copyright permission.

I did not modify any part of the sentence I copied.

Copyright is not an ownership law - if it was owned, it would never expire.

@Suiseiseki @maija >Most software is terrible, so most inputs were terrible, so most outputs are terrible

LLMs are in the most basic mathematic models which will spit out an average of what the dataset input was, when i say "average" this is the very top of the curve used by the specific LLM, most will use a simple bell curve to calculate results others will use more fancy math to calculate an average within a constrained negative skew distribution curve to give you better results. But the point stands that AI generated code will be some function of the average code feed as dataset, and most of the code in, for example github, turns out to be terrible code with probably a 15% of code or less being actually good code, unless copilot is specifically trained with just the 15% good code as dataset all the code it will spit out will tend hard to terrible and only once in a blue moon will it happen to output something remotely resembling the good code.

@EdBoatConnoisseur @Suiseiseki suiseiseki should buy a nice gpu like nvidia high end with cuda, to train on only gplv3 code which is of course perfect and flawless because the only metric in everything is freedom

@maija @Suiseiseki @EdBoatConnoisseur you two should get a room ngl

@maija @EdBoatConnoisseur Nvidia GPUs are now handcuffed to only run certain proprietary software, so I will never buy a modern nvidia GPU.

Ironically, older nvidia GPUs aren't handcuffed and run with free software, so I use those.


It is a complete waste of time to use a LLM on software - if you want to re-use GPLv3-or-later software, you simply copy-paste it while following the license terms.

@Suiseiseki @maija @VIPPER >You look at the output and then search for it and then surprise, surprise you find it's pretty much a 1:1 copy with some trivial changes like slightly different variable names or a wrong license on top (as there's a certain amount of noise).

@monkey @Suiseiseki @EdBoatConnoisseur i'm already in bed with my bf as his stay at home pet girlfriend (male) rn wdym i need to sleep with somebody else :(
baiting is just fun though and i cant sleep ;;

@maija @Suiseiseki @EdBoatConnoisseur you're living the good life god damn

@EdBoatConnoisseur @Suiseiseki @maija maybe a model could be trained to determine if code is likely to be good or not, I wish I was in on this stuff before the internet became irreparably flooded by slop

@Suiseiseki @EdBoatConnoisseur i dont think you understand the point of llm coding assistants ngl but whatever
why do you support a company with broken open source drivers whos probably the biggest opponent to free software of them all? (even though their official drivers are free)
use a company that supports free software more like amd you hypocritical proprietard

@monkey @Suiseiseki @EdBoatConnoisseur yeah i finally made it! :D
i took kratom like 11 hours ago and it took five hours to kick in strong as shit and it wore of normally but now its back and i feel far tighter
smth tells me this shit isnt legit how am i supposed to sleep like this bro 😭

@maija @EdBoatConnoisseur >support a company with broken open source drivers
Nvidia has never released a free software GPU driver.

The closest they have come is a GPU driver with obfuscated sources.

Another group has developed a free software driver without Nvidia - Nouveau (which is fully free software, with free peripheral software and free VBIOS init up to the 780 Ti).

>even though their official drivers are free
Nvidia's official drivers are clearly nonfree proprietary malware and aren't even gratis, as you pay for them as a percentage of the hardware cost.

>use a company that supports free software more like amd you hypocritical proprietard
As I pointed out, ironically AMD is worse.

Every driver AMD has ever provided has been proprietary software.

As for the AMDGPU driver, half of it is proprietary peripheral software that that runs on the GPU that you're not allowed to understand or modify.

There is a nasty hack available where you go get the radeon driver and hack it up to not load proprietary software and force it to load for AMD GPUs and you at least get native resolution, but not ACPI S3 suspend or 3D acceleration (but there is no free VBIOS init for AMD GPUs, thus that's less free than using a older Nvidia card with Nouveau, that also offers full 3D accell and suspend support).

If I was to use AMD, I would be hypocritical.

@maija @EdBoatConnoisseur >i dont think you understand the point of llm coding assistants ngl
I believe the point is to have this mythical "AI" that writes the proprietary software for you without you having to understand anything, which is clearly not what people who use and write free software want.

@Cyrillic @maija @VIPPER I don't do proprietary drugs.

https://xcancel.com/mitsuhiko/status/1410886329924194309
https://video.twimg.com/tweet_video/E5R5lsfXoAQDRkE.mp4

@Suiseiseki @Cyrillic @VIPPER @maija

This implies that you might do free drugs.

@monkey @EdBoatConnoisseur @Suiseiseki wait no maybe its because i was drinking before i forgot about that

@sally @Cyrillic @VIPPER @maija I haven't synthesized any free drugs yet.

@Suiseiseki @EdBoatConnoisseur amd hass the best free drivers nvidias nouveau shit are notoriously bad

@Suiseiseki @EdBoatConnoisseur so you dont

@Suiseiseki @Cyrillic @VIPPER @maija

And you shouldn't, unless used as chemical weapons against your enemies (law enforcement trying to arrest you for the crime of being too free). One milligram of fentanyl can kill an adult man in less than ten seconds, it's very easy to produce as well.

@maija @EdBoatConnoisseur >amd hass the best free drivers
Ah yes, a disgustingly proprietary license is free now;
No reverse engineering, decompilation, or disassembly of this Software is permitted.

(See attached for the full license of AMDGPU).

>nvidias nouveau shit are notoriously bad
Nvidia did not write nouveau.

Nvidia has does their very best to sabotage Nouveau, but despite all their proprietary sabotage, it works quite well on the 780 Ti with acceptable re-clocking and for ancient nvidia GPUs it works functionally better than the proprietary driver.

@Suiseiseki @maija @EdBoatConnoisseur

AMD GPUs have really poor quality proprietary software with many bugs, there's a bug with the RX 6000 series that has been a problem since the GPU released and nowhere near to get patched, where audio output through GPU randomly cuts and stops for a few seconds up to minutes whenever the GPU is decoding video, the higher the bitrate the worse, must be some problem with the audio buffer or pipeline crashing constantly, no one from the kernel, seems interested in investigating the issue.

@eliseo01 @maija @EdBoatConnoisseur No Linux developer, or anyone else not from AMD can investigate the issue as it appears that the bug is in the proprietary peripheral software and well there's no source code and you're even disallowed from reverse engineering it and fixing it yourself.