Then again, I’m giving a much closer look at Haiku & BSD again since a slop generator was used in the Linux kernel.

As far as I understand, the Linux kernel uses LLMs in practice for code review, not for code generation. And they do not replace human review with it, but use it in addition.

From my experience, they’re catastrophic especially in those positions where they review code/actions.

That’s interesting. I am currently searching for ways that I can use them a bit to chill-pill our CTO without sinking the company (which is very deep in the water with technical debt)… at least until I get another job.

Can you expand why they are that bad in this use?

My current idea is to pin up new architecture by hand, define new interfaces by hand, and then let the LLM occasionally fill in a function body. And maybe, maybe, an “ai refactored” branch which has just the purpose to showcase what happens if one is not careful - and will never be merged, at least not by me.

As someone who does code, this is a very valid concern you have.

Especially if the code touches things where the stakes are high.

Integrity of systems e.g. used by dissidents in whatever authoritarian country.

Heavy machines.

Cars, trucks, trains, subways.

Industrial robots - if out of control, their force and speed can easily kill people.

Critical infrastructure for electricity, water, energy, payments, and transport.

Chemical factories.

Medical devices.

This list is expanding because more and more digital systems become enmeshed with the physical world.

I think establishing new rules and boundaries in a new territory will never happen entirely without conflict. Nice if people still behave like gentlemen.

By setting init=/bin/sh in the editable grub command line. You can google how to get there exactly. This directly starts a root shell instead of the normal init process. (it is also in the raspberrytips link that were already linked in another comment).

To note: This also means that anybody who has physical access to your console during bootup can root the system. If you do not want that, you need to set a harddisk encryption password (and optionally also a BIOS password, but the latter won´t protect e.g. against disassembling the computer and putting the HDD into another Linux box).

If Zorin uses grub, you can set init=/bin/sh in the grub command line. You will get a root shell in rescue mode, with the root partition in read-only mode. You can then mount the root partition as read-write, set the user password using the passwd comand, mount read-only again, and reboot.

There is a fascinating parallel to tobacco in that the LLM technology seems, like “social media”, to be deliberately engineered to be addictive by design. A thing we should focus more on.

(By the way, using cars is also somewhat addictive: Humans do have a time budget of about one hour per day for commuting and daily errands. If you start to use a car, the average distance you go daily will expand, and it is very difficult to go back from that, even if you don’t really have more time, or don’t need the car any more. This is probably relevant to ai because the mechanism by which cars hook us is that the brain stem is basically programmed to use alternatives which are less effort, and so people use the car for everything because it is so comfortable.)

Do you work in tech?

One canwork in tech and not use LLMs. Your comment sounds like it should be the norm, or already is.

Also, the discussion is about using LLMs in open source projects, which put a lot more focus on understandable and maintainable code than the software industry typically happens to do.

On the rules, citing hperrin:

It sounds like their policy is basically the same as Human Contribution Policy 2-TR. That’s awesome.

‘AI’ doesn’t exist, at least when that term is meant to reference Artificial Intelligence in relation to LLMs, as they are intrinsically incapable of possessing or providing comprehension.

An important point that deserves repeating, because the marketing PR around the LLMs borders large-scale disinformation.

Many people point out we should be more careful with language, and they are right.

Hell, even the Linux kernel now has AI code.

Actually, AI reviewed code, which is not the same thing.

I can confidently tell you that the code isn’t the best and, at times, it can be massively messy and even shit, but it works.

Well, this “but it works” is kinda the accepted standard in the corporate world, but not in open source projects.

And for a good reason: Open source projects with multiple contributors do have a long-term interest on well maintainable, understandable code. Precisely because working on them is not a one man (or woman) show and resources are limited, and open source consists of thousands of smaller projects that interoperate, so the demand to support cooperation is a necessity.

Conversely, in the corporate world (and I sweepingly include companies using Linux and FOSS software), and with the prevailing capitalist mindset, technical sustainability just does not matter. What matters is the bottom line for the current quarter, and nothing else.

These two wildly different goals create different strategies, and they are not very compatible.

AI is here to stay. It’ll never go away,

You could say the same about cars. An US American might not be able to imagine life without them. But while trucks, vans, and ambulances have their meaningful use, many European cities are moving away from allowing personal cars as dominant means of transport because bicycles and public transport are frankly better and do not impose such a high cost on the community.

Coming back to the AI topic, saying “it is here to stay” is not a valid reason to avoid discussion what it is good for, what are the costs. whether the Linux communities are willing to pay those costs, and whether there are perhaps better solutions for what AI promises.

I’m so tired of “Ai big number” headlines. 100+ packages - but how many good ones? (Genuinely, I don’t know, as I don’t the recognize those listed.)

There might be a bit confusion. A package maintainer does not (necessarily) author the packages (the source code of the projects they contain), but typically maintains build scripts and security patches for upstream packages maintained by their authors. This is necessary because build systems and scripts are complex and packages from one distro would usually not fit into another one.

In other words, most code he/she produces are build scripts, and testing and adapting updates. And yes, this is still a lot of work.

Let’s make one thing clear: The “extinction warning” was marketing.

Funny that it works well to distract from the real extinction we are facing, an absolutely serious, scientifically proven threath to all of humanity: Accelerated climate change if we do not get away now from fossil fuels. A world with 3 ° C warming, which is the path we are on, will not be survivable for technical civilizations. And time is running out. A problem to which AI data centers are only contributing to.

So, we should all come to terms with it and use it the best way we can.

It might have some sensible uses (bug screening was mentioned) but if it causes a massive, devastating environmental load, it still might better to either not use it, or very sparingly.

A bit like opioids: They can cause a warm fuzzy feeling, and sometimes they are needed in medicine, but overall using them all the time would mean the disadvantages and costs are higher than the advantages.

Shredding the environment is a very high cost. I do not understand how people don´t get how serious is our climate situation. We only have this one wonderful planet. We are about to destroy it. There is no other place in the universe where we can go to and have a happy life.

It is a small and comperatively frugal distro. Which pretty much makes sense for an AI-free distro.

The now orphaned packages include Adapta, Autodia, SFML, abootimg, alacritty, arpwatch, asus-kbd-backlight, backblaze-b2, bash-preexec, beep, binfmt-support, bitlbee-facebook, bitlbee-steam, cdm, clementine, crane, csound, daemonize, darkhttpd, dep, diff-pdf, docker-cli, docker-gc, docker-gen, dtc, embree, entr, etcd, fcron, flannel, flint, font-fontin, forego, freeplane, fscrypt, gede, gettext-stub, git-cola, gkill, glock, go-bindata, go-jira, hans, hivex, hopper, hostaliases, hugo, iio-sensor-proxy, indicator-doom-cpu, interbench, ispc, jbigkit, kaitai-struct-compiler, kubernetes, kvirc, libguestfs, libnfc, libpodofo0.9.7, logstalgia, massren, mbox, mfoc, mkinitcpio-nfs-utils, mksh, moby, motion, openresolv, packer, pd, perl-Template-Toolkit, pinktrace, podofo, python3-Arrow, python3-b2sdk, python3-pdm-backend, python3-xattr, racket, rakarrack, rancher-compose, rmlint, rng-tools, rocker, runit-swap, sassc, scanmem, schroot, sift, splix, sublime-text3, subtle, supermin, sydbox, tag, tbb, terraform, terragrunt, texmaker, thermald, trollock, ttf-ubuntu-font-family, urbanterror-data, urbanterror, usbredir, vagrant, vba-m, virt-manager, virtme, webcat, websocketd, wpa_actiond, xboxdrv, xcompmgr, and yoshimi.

I am using a few of them, for example virt-manager and racket.

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.

Well, Americans had opted not to have rail transport. Which is safer.

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