That seems like a lot of work for something that can be done with a search filter.

Make sure your dotfile directories are part of a git repo, it makes unbreaking things MUCH easier.

You’ll probably need to watch a short video on git, but for this purpose it’s pretty simple to use.

One day you’ll use Arch, btw

zswap

As an aside, I just read that zswap is getting a pretty big performance increase: https://www.phoronix.com/news/Linux-Better-Zswap-Performance

I was looking into upgrading storage and ended up opting to buy 20TB in HDDs for my NAS.

I have to shuffle Steam games around but at 10Gb, it goes pretty fast (and you can run most games directly off the NAS) and the array can survive if a drive fails (ZFS).

You need the extra hardware to run the NAS but HDDs are a significantly cheaper option and you don’t need SSD speeds for most things.

This is already not the case for computers. And you’re trying to argue that we might one day be able to run the bigger models on consumer hardware, but I said it’s too expensive to train them. And you know that the cost isn’t just the RAM right?

If you replace everything I said about RAM with compute and replace the Bill Gates quote with Thomas Watson saying that the “world market [has room for] for maybe five computers” then it’s the same argument. They’re all ultimately built on technical advances in lithography and so growth in RAM and compute largely stem from similar sources.

Sure, my saying they will “always” be tied to concentrated capital is slightly hyperbolic. You never know. But it takes a whole lot of optimism to hope that LLMs will be decoupled from the billionaire class, hence why most analyses of the problems LLMs cause discuss them in this context.

I understand why it is done as shorthand.

I know this likely doesn’t apply to you, as you seem to be capable of understanding the nuance, it’s that a lot of people on social media (by people I’m including the bots) don’t make that distinction.

The same vitriol that is, rightfully, directed at OpenAI/Anthropic/Google/NVIDIA is also directed at things completely outside of the corporate LLM hellscape.

I’m mostly frustrated at seeing the sciences, researchers and open source AI projects catching strays and being painted with the same brush.

It’s useful shorthand among people who understand the nuance, but using that shorthand on social media invites misunderstanding and that is causing collateral damage.

This is mostly true, but you’re glossing over the fact that AI will always be tied to concentrated capital due to the cost of training state of the art models.

I think this is a false assumption.

We don’t know what the ultimate state of AI will be, we only know that we’re at the beginning of the journey.

Look at other technologies in similar positions. At one point computers were like this. Nobody but the military and the largest companies could afford them and now you can buy children’s toys with more storage and processing power than was available to the entire US military at the dawn of computers.

We had no idea what computers would become, even the experts making guesses that sounded wild at the time were off by several orders of magnitude. Bill Gates once said “640K [of RAM] ought to be enough for anybody”, modern PCs have around 100,000 times that.

The reason that AI seems like it’s only affordable to concentrated capital is because models using a large amount of parameter require a large amount of RAM. The 5TB of RAM estimated for some of the largest frontier models is a lot, about 100x a workstation PC. That’s a few orders of magnitude less than 100,000x. Given equal scaling speed, we’re closer to running frontier models on our desktop than we are to the dawn of computers.

In addition, the research is leaning towards there being a ceiling on how useful it is to add more parameters (use more RAM). Doubling parameters doesn’t result in doubling of capability so there is effectively a soft cap on the amount of parameters.

Hardware is growing near exponentially while the RAM demands of models have a ceiling.

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 and, I’d argue, that they actually produce more value than LLMs (which are failing to deliver on their promise of replacing white collar labor).

It’s unlikely that the market will open up any time soon so that you can get your LLM from someone who isn’t a fascist billionaire tech bro.

You can find open weight models (of all kinds, not just LLMs) freely available here: https://huggingface.co/

Many can be run on locally on gaming hardware. Some of the classification and object recognition models can be run on microcontrollers (a popular use it to use them to make aimbots in FPS games).

There are larger models that can be run on more expensive dedicated hardware (like NVIDIA DGX Spark, ~$5,000) if you want to keep your projects completely internal. Or you can run those open weight models on hardware hosted in datacenters ran by companies you research to be ethical.

You can use AI without using one of the AI companies.

Most of these arguments are conflating AI with the people using them.

Labor exploitation, scraping without consent, military contracts, surveillance policing, grid strain and hardware prices driven up by speculation are real harms. But none of them are properties of a statistical language model.

They come from concentrated capital and state power, and whatever technology arrives next will get used the same way.

Just like we’ve seen everything in your list, done by the same people, using different technology.

We have had hardware speculation during the crypto craze.

These same companies (Google) have been taking your data without consent to power their surveillance advertising.

Datacenters have always been power and water hungry, the ones powering LLMs are exactly the same ones that exactly the same companies have been building since the Internet began.

Essentially every major advance in technology has been used for state surveillance and in the military. AI is simply the latest capability.

Models reproducing licensed code, poor code quality and floods of junk contributions. These are real problems with the technology itself. And the solution to most of these things is to iterate, build better models trained on consentually obtained data.

Yes, you should still be mad at what the AI companies are doing. I am too. Just keep in mind that Google was evil before Gemini and will certainly be on the forefront of fuckery with whatever technology comes next as long as people keep getting tricked into misdirecting their anger onto the technology.

Nothing. OP is stirring shit, attacking the other is part of that same tribalistic POV

and medium sized ones

OP, a term you may find useful is ricing. This is the sub-community of Linux users who focus on customizing their desktops. You can look at the unixporn subreddit to see examples of how much a DE can be customized.

Ex:

This is KDE Plasma

This is also KDE Plasma

This is also a, very customized, KDE Plasma

(video)https://www.reddit.com/r/unixporn/comments/1w2afj0/kde_my_2nd_ever_linux_rice/

Another is dotfiles or dots, ricers will sometimes make their .config files available for other users. (Be careful here, this can include hidden malicious scripts and such uglies. Use your best judgement).

PXE booting is pretty neat, but it’s a lot to setup for home use.

I did a lab where we configured an immutable distro to boot using an encrypted RAM drive as storage (using TSME). The machine had no storage other than RAM (and NFS mounts). We were trying to harden a machine against coldboot/evil maid attacks. It was a neat project but a bit too clunky for daily driving and also, a RAM drive in this economy?

What I don’t understand is why super critical errors are hidden away in a log instead of being shoved in my face once they happen.

That’s a choice you make when you’re setting up your bootloader, you pass a kernel parameter that tells the kernel what to show on the screen.

https://docs.kernel.org/admin-guide/kernel-parameters.html

If you’re using grub, look in /etc/default/grub. There will be a line: GRUB_CMDLINE_LINUX_DEFAULT=“<kernel parameters>” you can set the log level to something (like loglevel=3) and it will write the errors to the console on boot. The higher the number the more information is written to the console.

That was 2 hours-ish that could have been just 5 minutes if the system had actually told me it was having problems with connecting to a drive.

It did, it almost certainly wrote the error into the system log. That’s generally the first thing you should check if you’re having such low-level problems, like a failure to boot.

Going forward some good advice is: Get a USB stick, install Ventoy and put a system rescue image on it: https://www.system-rescue.org/

This way, if you’re ever having issues booting you can boot into a live environment and read the logs.

If you mount your system drive on /sysroot, you can read the system log with

journalctl --root=/sysroot

Add a -b flag to see just the log from the latest boot (and -b -1 from the previous, -b -2 from the one before, etc)

And, surprisingly, it wasn’t DNS

Yeah, probably a .bashrc difference between users.

It could also be ~/.inputrc (or /etc/inputrc if not in the user’s home directory.) https://man.archlinux.org/man/readline.3

A stressful residency

That’s one small step for linux users, a giant leap for nerdkind

Putting yourself into the position of trying to solve a practical problem is the best way to learn.

If you fail completely, you can always restore your system and try again (you have a backup, right?)