• 9 Posts
  • 1.27K Comments
Joined 1 year ago
cake
Cake day: March 22nd, 2024

help-circle
  • And any divergence from that is “ruining games” or “being woke” to the point that we don’t even GET those games outside of the rare case of a game nobody cared about becoming popular

    I would argue the origin is sales. E.G. the publisher wants the sex appeal to sell, so that’s what they put in the game. Early ‘bro’ devs may be a part of this, but the directive from up top is the crux of it.

    And that got so normalized, it became what gamers expect. And now they whine like toddlers when anyone tries to change it, but that just happens to be an existing problem conservative movements jumped on after the fact.


    TL;DR the root cause is billionares.

    Like aways.




  • Yeah. But it also messes stuff up from the llama.cpp baseline, and hides or doesn’t support some features/optimizations, and definitely doesn’t support the more efficient iq_k quants of ik_llama.cpp and its specialzied MoE offloading.

    And that’s not even getting into the various controversies around ollama (like broken GGUFs or indications they’re going closed source in some form).

    …It just depends on how much performance you want to squeeze out, and how much time you want to spend on the endeavor. Small LLMs are kinda marginal though, so IMO its important if you really want to try; otherwise one is probably better off spending a few bucks on an API that doesn’t log requests.






  • At risk of getting more technical, ik_llama.cpp has a good built in webui:

    https://github.com/ikawrakow/ik_llama.cpp/

    Getting more technical, its also way better than ollama. You can run models way smarter than ollama can on the same hardware.

    For reference, I’m running GLM-4 (667 GB of raw weights) on a single RTX 3090/Ryzen gaming rig, at reading speed, with pretty low quantization distortion.

    And if you want a ‘look this up on the internet for me’ assistant (which you need for them to be truly useful), you need another docker project as well.

    …That’s just how LLM self hosting is now. It’s simply too hardware intense and ad hoc to be easy and smart and cheap. You can indeed host a small ‘default’ LLM without much tinkering, but its going to be pretty dumb, and pretty slow on ollama defaults.






  • A number of companies have noted this trend. David Gibbs, CEO of KFC and Taco Bell parent Yum Brands, told analysts in late April that the company was closely monitoring the issue and its impact on consumer behavior, but added that there had not been an observable impact on its sales.

    I’m surprised its mild enough for them to claim this so far.

    I know the Canadians are pissed, but are all y’all’s peers on other continents mostly… indifferent? I guess it’s egotistical to assume people generally follow American news, but some of it really seems hard to ignore.


  • relatively minor efficiencies don’t make up for the compile times lol.

    It’s sometimes even a regression. For instance, self-compiled pytorch is way slower than the official releases, and Firefox generally is too unless you are extremely careful about it. Stuff like Python doesn’t get a benefit without patches.

    I think the point of Gentoo is supposed to be ‘truly from source’ and utility for embedded stuff, not benchmark performance. Especially since there are distros that offer ‘march’ optimized packages now.


  • What Toot said.

    Things I would emphasize are:

    • The community “critical mass” is amazing with the wiki, online posts and such. You get a lot of support that isn’t ancient, jank and ad hoc like Ubuntu.

    • Arch emphasizes paying attention. It’s not a hands free OS: you have know what graphics drivers you run, and what your desktop environment is. When you update, you have to watch the log for emergency messages and such, including official notifications from the arch repos themself. It’s not a “hands off” OS where you can operate without knowing anything about it, but the reward is that shit gets fixed quick, officially, without having to stray from defaults and break your system, or accumulating a bunch of hacks you have to maintain yourself.

    • Much of Arch’s bad reputation comes from AUR. Don’t use anything from the AUR (instead of an official repo package) unless you absolutely have to. This is when stuff starts breaking. Installing standalone apps that aren’t on the repo via AUR is fine, but to be clear, avoid things that integrate with the system if you can.

    • It doesn’t have to be hardcore barebones like Gentoo, there are all sorts of Preconfigurations like Garuda and Endeavor. I recommend CachyOS (which I have kept for two years now, and will into the future).


  • There are a couple of historical examples that escape my soft brain, but ivermectin is a big one.

    Basically, during COVID-19, some rural doctor in India published a paper on Ivermectin reducing COVID fatalities. Which is great!

    …Because Ivermectin is a livestock dewormer. It’s an anti-parasitic, so it reduced the chance of co-infection with parasites in rural India (which would weaken your immune system so you’re more likely to die of COVID).

    But the anti-vax crowd lost this nuance in the paper and ran with this as an alternative treatment, misinformation that still persists today.

    https://en.wikipedia.org/wiki/Ivermectin#COVID-19_misinformation


    …That’s whats about to happen.

    Tylenol (acedomediphine) will be forever linked to causing autism, true or not. No one amount of refutation is going to kill the meme.

    And Folinic acid, whether its a provably effective treatment or not, is not going to get touted by the anti-vaxx crown who ‘believe’ in its efficacy.