Edit: This is apparently one person's idea and may not make it into Tumbleweed: I would still love suggestions on which way to jump off it does happen
-
gentoo has no-LLM policy. however, it's unavoidable for the packages themselves, as well as the kernel
@xarvos Thank you
-
@Canageek The tools mentioned are pretty great. It sounds a bit like a baby/bathwater situation.
I mean yes, it's a bit ridiculous / trendy, but of all the bad ideas about using LLMs, using them to answer questions or produce summaries on a specific set of documents and information is one the most reasonable ones. If rather than saying "just read the docs at doc.opensuse.org" you can offer a new user a way to query them using natural language, i can see the draw.
@axx until it hallucinates and tells you to do something that breaks your system
-
@Canageek What is the source of this information please? The only thing I can find is https://www.phoronix.com/news/GSoC-2026-Exciting-Projects Which is a student doing the Google summer of code said they had an idea for such a project.
@Ooze That is where I got it from: I thought Google summer of code projects were like when an academic brings on an undergraduate, they work on a project designed by the supervisor?
-
Edit: This is apparently one person's idea and may not make it into Tumbleweed: I would still love suggestions on which way to jump off it does happen
"For openSUSE Linux they are looking at developing an AI-powered onboarding experience for users to deal with "new openSUSE users face a steep learning curve navigating distribution-specific tools like zypper, YaST, Btrfs/Snapper, and systemd." "
Guess it is time for me to find a new linux distro. Recommendations welcome.
@Canageek just rawdog arch atp.
-
@Ooze That has very little to do with how much I trust it and refuse to use software that has it
@Canageek It should. It isn't a thing built by stealing other people's work as it uses only the SUSE docs as its data set.
It is never going to use a giant data centre because in runs locally and is aimed at new users not businesses.
This kind of single purpose SLM is the only use of this technology that I would consider ever thinking about using. And it will probably be one of the very few uses of this tech which will be around after the bubble bursts.
-
@Canageek It should. It isn't a thing built by stealing other people's work as it uses only the SUSE docs as its data set.
It is never going to use a giant data centre because in runs locally and is aimed at new users not businesses.
This kind of single purpose SLM is the only use of this technology that I would consider ever thinking about using. And it will probably be one of the very few uses of this tech which will be around after the bubble bursts.
@Ooze It's still going to hallucinate and lie to users and tell them to do incorrect and unsafe things, it's that's just part and parcel of how statistical models work.
-
@Ooze It's still going to hallucinate and lie to users and tell them to do incorrect and unsafe things, it's that's just part and parcel of how statistical models work.
@Canageek I am not saying it is perfect. There is also the issue of the psychological effects on humans of becoming dependent on these things.
I am suggesting that we can be discerning and weigh the risks depending on the variables.
Software can't lie or hallucinate as it has no intent or perception. These are just words their developers use to avoid having to say that they wrote very bad software.
-
@Canageek I am not saying it is perfect. There is also the issue of the psychological effects on humans of becoming dependent on these things.
I am suggesting that we can be discerning and weigh the risks depending on the variables.
Software can't lie or hallucinate as it has no intent or perception. These are just words their developers use to avoid having to say that they wrote very bad software.
@Ooze I just don't think the technology is worth anything and don't trust any group that would willingly use it as a result.
I'm not quite willing to move to OpenBSD due to LLM generated code in the kernel yet, but I understand the people I know who are doing so
-
@Ooze I just don't think the technology is worth anything and don't trust any group that would willingly use it as a result.
I'm not quite willing to move to OpenBSD due to LLM generated code in the kernel yet, but I understand the people I know who are doing so
@Canageek I agree. I would look for another distro if SUSE started putting machine generated code in.
But I do think there are applications for this that are useful. e.g. medical image processing https://www.cam.ac.uk/research/news/ai-tool-can-analyse-complex-cancer-images-rapidly-offering-potential-to-personalise-treatment
-
@Canageek I agree. I would look for another distro if SUSE started putting machine generated code in.
But I do think there are applications for this that are useful. e.g. medical image processing https://www.cam.ac.uk/research/news/ai-tool-can-analyse-complex-cancer-images-rapidly-offering-potential-to-personalise-treatment
@Ooze Totally different technology then statistical-best-token, that is just a classifier (I still prefer analytic algorithms for that, tbh). They just throw it all under AI to justify their bullshit
-
Edit: This is apparently one person's idea and may not make it into Tumbleweed: I would still love suggestions on which way to jump off it does happen
"For openSUSE Linux they are looking at developing an AI-powered onboarding experience for users to deal with "new openSUSE users face a steep learning curve navigating distribution-specific tools like zypper, YaST, Btrfs/Snapper, and systemd." "
Guess it is time for me to find a new linux distro. Recommendations welcome.
@Canageek Won’t help you here. Never used anything else than Debian, be it my laptop, home PC or servers I manage. Well OK there was a year or three of Slackware, but that was back in late 1990s, so it was THE Linux.
-
@Canageek Won’t help you here. Never used anything else than Debian, be it my laptop, home PC or servers I manage. Well OK there was a year or three of Slackware, but that was back in late 1990s, so it was THE Linux.
@lgrochal Debian *is* on my list of backup options!
-
@Canageek assuming that OpenSUSE really does do this… well, I feel your pain. I don't know what your needs are, but it is worth looking at Debian; they have indicated that they have no plans to integrate AI. Gentoo has taken a firm anti-AI stance.
Arch hasn't made any statements, but integration of AI would be against their whole ethos. By extension, Endeavor OS is also probably safe.
There are actually quite a few good Linux options that eschew AI. But depending on your use case, you could also give NetBSD a shot
You seem knowledgeable about this arena. Do you make recommendations on anti-ai distros?
-
R relay@relay.infosec.exchange shared this topic
-
You seem knowledgeable about this arena. Do you make recommendations on anti-ai distros?
@eldersea @Canageek it depends on how you define anti-AI. Gentoo is the only distro that I know of that has been vocal about their anti-AI stance. Arch has yet to set a policy, but I would be very surprised if they were accepting. Debian has said they are going to wait for things to settle down before making a decision, which is very Debian of them.
NetBSD (not a Linux distro I know) has made firm statrments against AI, and don't allow AI code. The other two major BSDs are less vocal, but neither is currently accepting AI code as far as I know (although FreeBSD has used it for documentation, or at least translation of documentation; it did not go well.)
It's probably better, though, to point out distros that have either fully embraced AI or tout their usefulness in AI, or are bragging about upcoming AI-oriented features. That would include Ubuntu, Fedora, and OpenSUSE.
So right now, I'd say use Gentoo, Arch, or Debian for Linux, or NetBSD if you want to be 100% sure.
-
@eldersea @Canageek it depends on how you define anti-AI. Gentoo is the only distro that I know of that has been vocal about their anti-AI stance. Arch has yet to set a policy, but I would be very surprised if they were accepting. Debian has said they are going to wait for things to settle down before making a decision, which is very Debian of them.
NetBSD (not a Linux distro I know) has made firm statrments against AI, and don't allow AI code. The other two major BSDs are less vocal, but neither is currently accepting AI code as far as I know (although FreeBSD has used it for documentation, or at least translation of documentation; it did not go well.)
It's probably better, though, to point out distros that have either fully embraced AI or tout their usefulness in AI, or are bragging about upcoming AI-oriented features. That would include Ubuntu, Fedora, and OpenSUSE.
So right now, I'd say use Gentoo, Arch, or Debian for Linux, or NetBSD if you want to be 100% sure.
Thank you. I am on Fedora KDE at the moment, and I suspect KDE is really the bigger driver of what I want on a day-to-day basis. Are any of the others you mentioned well-integrated with KDE?
