HaraldvonBlauzahn

The Making of an AI Data Centre Warrior | The Tyee

Why not LLMs? [Codeberg]

Coding is NOT solved cross-posted from: feddit.org/post/36071801 This author trieds hard to do a pragmatic view. He does not mention ethical or ecological concerns (which should matter far more and subject to legal regulation IMO). Instead, he focuses on the quality topic. His thinking is that companies are doing now a kind of quality arbitrage: Companies can make low-quality code at far cheaper costs, and sellthe result still at human-made prices. Which will only work for so long, he thinks: Customers will demand lower prices. (Easy to imagine a quality downward spiral.) Which invokes the question: What has happened that people accept lower and lower quality? Is new software nowadays still really useful?

So, their fix for spying is gaslighting?

Given empirical numbers of bugs per lines of code … and the size of Microsoft products, how many bugs do they probably have? Given how “productive” LLMs are at adding lined of code, how many new bugs are there likely added per month?

Global fossil-fuel emissions are set to fall by in 2026 Time for a Nobel prize, I guess. Who should get it?

Keep in mind that this only relates to the state of the largest LLMs. You can train generative models on consumer graphics cards. Object recognition, classification, etc are all tiny compared to LLMs. These are also AI […] The linked Codeberg page refers to LLMs in their current, corporate-controlled shapIe. It is about using LLMs for coding. Your argument is rather leading away from that.

Ridiculous anti-AI hysteria is pushed by the AI companies themselves to make the anti-AI side look insane. And to weaken the FOSS communities I guess. Plain information warfare.

How would Solus compare to Void Linux? One of the most interesting characteristics Void Linux has is “stable rolling release”. The only large distribution which does this is IIRC SuSE Tumbleweed/ Slowroll.

I don’t want to be funny, but that’s not mutually exclusive.

Big Tech and Big Oil have fused – Call it Fossil Tech cross-posted from: piefed.social/…/big-tech-and-big-oil-have-fused-c… (Big Tech data-centers are increasing fossil fuel use and GHGs, while the executives are making prophecies that “AI will fix it”)

Void Linux maintainer orphans 100+ packages over AI policy dispute

How would such a Turing tarpit work that delivers good-looking, plausible, but very subtly non-functional code? So subtly that it takes an above-average developer to fix the result?

I think there are cities which are at extreme risk from flash floods. For example Munich. Munich is crossed by the Isar river, a rapid mountain river in a narrow valley which is in turn embedded in the Alpine Foreland, a landscare shaped by the end of the an ice age, with wide valleys which could collect considerable amounts of rain. The subway of Munich crosses the river in tunnels which are many floors below the surface, I´d gues some 30 to 40 meters. And it has entrances close to the river. In case of a large flash flood, this is a really dangerous initial situation.

Installing new packages is relatively slow. But it was also quite slow because the infrastructure at gnu.org was loaded. It is a lot faster since the project moved to codeberg. Also, for me personally is openix a “guix shell” a lot faster than starting up a virtual machine.

Here a comment on this from usernomdeguerre on hacker news. “His” and “GKH” refers to Greg Kroah-Hartmann, who is maintainer of the stable kernel series and the speaker in the video. I’ve included a few slides into text that i thought were eye-opening to me: From his Kernel Recipes 2026 slide on Mythos Mythos's 79 vulnerabilities: 24 - no detail at all "something crashed" 14 - not a bug at all 3 - totally made up data 15 - already fixed in latest release - 11 by others - 4 by anthropic 20 - fixes were needed - 7 "assume a malicious filesystem image" - 2 "assume you can inject a malicious network packet into the middle of the stack" - 2 "NOMMU" - 6 sctp networking issues for untrusted devices - 2 ipv6 minor network issues - 1 gpu driver for local malicious user GHK called this “10 ‘real’ bugfixes”, which to me sounds like there’s a wild hype machine around these companies and uncritical parroting of every press release they make that falls apart when you engage the affected real experts.

I am in South of Italy now in a beautiful city build in 17th century from white marble and limestone. And in every street in the old center, there are several flats for sale. I can’ tell what’s up, but it is like people are leaving their city. I think it was the hottest and most brutal summer Southern Europe has ever seen.

Andrew Kelley also made a YouTube version, for people which prefer video: www.youtube.com/watch?v=zLZwpH5lCD4

I think this is happening a lot. Say somebody wants clean drinking water in their home town. Or they want to stop Climate change which is going to destroy important parts of the biosphere. Then they opposea chemical plant which pollutes the water. Or oppose technology and policies which cause more carbon emissions, instead of bringing them to zero, as is really needed. Now, the industry creates culture war with highly polarized, extreme position. And when you continue to oppose that, you are suddenly the causing part of the culture war, and responsible that there is conflict. “If you were reasonable, you would agree to something in the middle”. Works also with far-right policies. “Ah, you oppose killing immigrants! So, you could agree to deport everyone whose parents were not born here!”. It is the whole point of the real extremism to shift the mainstream view. If having an ai girlfriend and talking about the end of humanity is lunatic, it must be the reasonable and measured position to do a lot of office work with ai, right? And the idea that in reality, it might not help at all is extremism. Of course, there are a lot of problems which require to find a kind of compromise. But not every problem is like that, and in order to find real compromises you need to look at the legitimate interests of the people involved. Drinking clean water is a legitimate interest. Destroying the planet for more profit not.

AI or not, I refuse to believe anyone is capable of being the sole maintainer of 100+ anything while preserving any kind of quality. Most of these packages were probably functionally abandoned already, if not technically. He/She was maintainer of the Void Linux packages, that is the package build scripts, updates, etc., not the projects and source code which these packages are based on. For example, Guix has now over 32000 packages, including an Emacs package. But of course, they did not re-create Emacs to make that package - they just made a build definition which uses the Emacs git repo, in order to build the Guix package. Or in the case of Ubuntu, it is based on the package definitions in Debian testing, so the Ubuntu maintainers do not have to write all the build scripts or package definitions - but they still have to do significant work in order to adapt them that everything fits together. Or with Arch Linux, the packages are taken relatively vanilla from upstream, they package stuff more “as is” as compared to Debian. Which means less work, and often more clarity for experienced users. Coming back to Void Linux, this means for these 100+ Void Linux packages which are currently orphaned, it is feasible that other people with good Linux experience adopt these, with a reasonable effort - some are important.

There’s a difference between vibe coding and the hundred other ways to “use AI” on a project. I am not talking about the Linux kernel or the Rust project - they have stabdards way above average commercial projects - but generally “using ai” in practice means to use it for code generation. Then, programmers are expected to review the output but are not given extra time for it. This creates an extremely slippery slope leading to mostly unsupervised code creation. Also because reading and understanding buggy third party code means more effort than writing new code. Then there is the idea to get rid of the bugs by “debugging”, but debugging is frankly not a software engineering strategy, because you can always just recognize a fraction of the bugs that your process created in the first place. Debugging is also more expensive tha writing good code. Then, what makes the process between product managers and software developers very often a bit friction-loaded is that good developers always want to have clear specifications what they are supposed to build. LLM code generation largely replaces clear specifications with plausible assumptions. That appears to make the process fast, but the result is often wrong. At times obviously wrong, and at times subtly and catastrophically wrong.

Sometimes, I think the “AI” industry has an interest to keep the debate as polarized as possible. Because this keeps people from taking a sober look and examine rationally and objectively what they promise, vs. what they deliver. We have several public debates which followed a similar pattern, such as the debates around climate protection and clean energy. (And lo and behold! Now “AI” is the reason we can’t have climate protection!) Or the “AI arms race” argument, which honestly just reminds me in the enthusiasm, propaganda, and delusion that happened at the outbreak of WWI. Yes, modern weapons got “better”. But that made war just more lunatic. And now, you can kill a building full of school children with an “AI mistake”. Military people must be gleeful.

To say that, you would need proof that this was in bad faith, woudln’t you?

What desktop environments and window managers does Void Linux support?

Not sure whether this is favoured here but I am thinking that some of you might appreciate some positive news, too.

Even more gain! Do you know what prozac does cost? Its family will need to pay the pharmacy bills if it doesn’t want their loved one ending up homeless.

There is a brilliant book which helped me a lot to get started with LaTeX, Leslie Lamport, LaTeX: A Document preparation system. It is itself a masterpiece of good technical writing. My copy is now 30 years old, and I still use it.

Indoctrinate the spider that it will catch many more bugs by using AI tools, and then let it pay.

You could look at Clojure, Scheme, and Common Lisp. I do not know whether they fit your cognitive preferences, but they stress logic. Resources online are The Lisp Cookbook, Clojure for the Brave And True, and the Racket Tutorials. Some Lisp programmers are excellent writers and teachers!

Look at distrowatch.org. Myself, I am using plain old Debian stable since about 17 years, after using SuSE and a bit of Ubuntu since 1998 or so. I also use a bit of Arch and Guix when it comes handy. So, I am a boring guy. But back to your question: What du you want? What is your goal?

I have a LaTeX file from 1993 in my archive dir that I can format and print at a whim. It is a technical manua from an uni research hardware project written in Spanish, initially on an ULTRIX or DEC system, and I converted it once to ISO 8859-1 encoding and once to UTF-8.

Do we need in a police state to keep people from stealing?

i need proof of the already proben impact of AI and AI datacenters over territories in the country i live in? No, proof about the maintainers intentions. Because you make a statement over them. I am extremely worried about the implications of “AI” as well, but you do not win this discussion by name-calling.

They mean no one can be reasonably be up to date on software security updates, and (ideally) ensuring all packages changes are not malicious (as well as the small stuff like properly integrating to the distro, but idk how void works enough to comment on that) Some people do an astounding amount of work. And then there are many, many, which help a little. Also, if, say, somebody finds that in Debian testing is a malicious package, people which maintain Arch will learn that. Open information sharing between many thousands of people willing to coordinate is a superpower.

Is there a better archive link?

From the discussion on the programmer forum lobste.rs: I just started a NixOS project and briefly considered Guix instead. I went with Nix entirely because it’s more popular, which means there’s more AI training material for it. I don’t intend to ever program in Nix myself so while I’d prefer Scheme conceptually, it’s not a big thing for me. OTOH everyone loves the Guix docs, that’s a point in its favor for AI usability. Joey Hess seems to have been right with his recent comment that LLMs promote using things like configuration languages that are worse for understanding.

The fact that it isn’t perfect is the grey. And the fact how it is used in practice, actually, in reality, is the very dark grey that some people call “black” for better clarity. The fact that you can use a large butchers knife for practical things does not mean that entering a primary school with one is the right thing. What are the actual intentions for grabbing that tool?

FOSS isn’t entirely immune to AI mania. I think it can be used for some useful things, like looking for security bugs in new kernel code. But this does not answer the question whether it should be used, because there are many important points in contra. For example, if a distribution’s package build code is done by AI, it is likely to become harder to understand over time. This is probably not the kind of technical debt which a distribution with limited resources wants to have. And you see, the decisions of individual affect other people to, which is why we have rules.

OSS projects have already started to recognize that LLM is just a tool and how you use that tool matters a lot more than whether you are using said tool at all. This tool is however not neutral but has a lot of steering built in. Like an advisor who optimizes heavily for personal gain. Lets bring StackOverflow for analogy: we didn’t ban browsers for code development because one can navigate to StackOverflow and copy-paste some really crappy code. Thid is a good exanple. I am programming over 40 years and i used it to find interesting new angles on issues with seemingly obvious answers. Like “what is the most portable, clearest eay to efficiently serislize data in a c struct”. LLMs are SO on steroids (actually trained on SO topics). So grabbing LLM to get a sense of direction or shape is an OK use in general, but copy-pasting it’s results blindly is silly and dangerous. There’s a second argument that AI is ruining OSS communities which I believe is true, but refusing to accept AI code is a self-defeating action. Why? Because AI training needs unadulterated source of knowledge, one not tainted with AI and OSS communities rejecting AI would be that source. Also: if OSS community did not use AI it would hardly stop it from continuing it’s growth in Enterprise. So there’s not much of a winning strategy there either. Which brings us to GrapheneO

We have lots of tools to put metrics on that, and AI assisted code is fairly well crushing those metrics. You mean more lines of code?

This is what I do. I use guix shell for local development environments, for example when I need current Rust + cargo or an LSP package in Emacs. It is also quite good for packaging and distributing own software that way.

But to be honest, TempleOS is way cooler. While plan9 is Unix tidied up a bit.

Fuck NVidia. I saved so much time since I bought an AMD Radeon card.

Supposedly wonderful “human artisanal code” has plenty of fuckery. These are supply chain attacks and in the case of xz utils, the attacker had gone to extreme lenghts to hide the attack from a well-meaning, good-hearthed but overworked and burnt out solo maintainer. To compare this to bugs that people unwittingly introduce in normal human-written code is not sincere.

Ah, the good old “let’s highjack a popular thing to promote AI” PR method!

Also, polarization is another danger, it can seriously hurt communities. I saw that with some dance communities during COVID. We need to navigate between Scylla and Charybdis here.