• 0 Posts
  • 30 Comments
Joined 10 months ago
cake
Cake day: September 10th, 2023

help-circle



  • The first problem, as with many things AI, is nailing down just what you mean with AI.

    The second problem, as with many things Linux, is the question of shipping these things with the Desktop Environment / OS by default, given that not everybody wants or needs that and for those that don’t, it’s just useless bloat.

    The third problem, as with many things FOSS or AI, is transparency, here particularly training. Would I have to train the models myself? If yes: How would I acquire training data that has quantity, quality and transparent control of sources? If no: What control do I have over the source material the pre-trained model I get uses?

    The fourth problem is privacy. The tradeoff for a universal assistant is universal access, which requires universal trust. Even if it can only fetch information (read files, query the web), the automated web searches could expose private data to whatever search engine or websites it uses. Particularly in the wake of Recall, the idea of saying “Oh actually we want to do the same as Microsoft” would harm Linux adoption more than it would help.

    The fifth problem is control. The more control you hand to machines, the more control their developers will have. This isn’t just about trusting the machines at that point, it’s about trusting the developers. To build something the caliber of full AI assistants, you’d need a ridiculous amount of volunteer efforts, particularly due to the splintering that always comes with such projects and the friction that creates. Alternatively, you’d need corporate contributions, and they always come with an expectation of profit. Hence we’re back to trust: Do you trust a corporation big enough to make a difference to contribute to such an endeavour without amy avenue of abuse? I don’t.


    Linux has survived long enough despite not keeping up with every mainstream development. In fact, what drove me to Linux was precisely that it doesn’t do everything Microsoft does. The idea of volunteers (by and large unorganised) trying to match the sheer power of a megacorp (with a strict hierarchy for who calls the shots) in development power to produce such an assistant is ridiculous enough, but the suggestion that DEs should come with it already integrated? Hell no

    One useful applications of “AI” (machine learning) I could see: Evaluating logs to detect recurring errors and cross-referencing them with other logs to see if there are correlations, which might help with troubleshooting.
    That doesn’t need to be an integrated desktop assistant, it can just be a regular app.

    Really, that applies to every possible AI tool. Make it an app, if you care enough. People can install it for themselves if they want. But for the love of the Machine God, don’t let the hype blind you to the issues.



  • I’ve once had difficulties running some apps on Proton that used .NET features not supported by mono, which has been updated since then and is now working out of the box.

    I’m playing Trackmania on wine, I’ve played Elden Ring and Monster Hunter: World on Proton, so I’m wondering which issue you’re running into.

    Regardless, building precompiled Linux native binaries is a commendable goal. Others have mentioned Flatpak, which imo is a good and user-friendly way to handle that.


  • I asked for a rough description because I didn’t wanna bother anyone to take the time for a full, detailed explanation…

    …then you come along and write a whole article on it that’s most certainly more informative and useful than anything Google would have spat out.

    I love that. Thanks so much for taking the time. I also think I’ll give Bazzite / Fedora Atomic a shot. The idea of simply rebasing onto a different option to try different things is definitely appealing.





  • luciferofastora@lemmy.ziptoLinux@lemmy.mlCrapped my system
    link
    fedilink
    arrow-up
    5
    ·
    edit-2
    2 months ago

    I’ve had good luck striking out on a new path with Nobara after years of only ever using Ubuntu. There was a bit of a learning curve (and I still haven’t gotten everything I wanted to work the way it did before), but I mostly got it figured out.

    But that may well be a Survivor case in the sense of Survivor Bias, no idea how many people tried and decided “wasn’t worth it”.

    I did have a bone to pick with pipewire because my old pulseaudio config no longer worked and I had difficulties figuring out just how to redo it in pw, but that’s probably not distro-specific.










  • Accessibility of information has also caused a problem of academic rigour getting removed with the higher barrier cost of learning curve.

    At least now I can just ask a question and get a decent answer without needing to dig through manuals that leave me more confused than before or going the roundabout route of saying “Linux sucks, you can’t even do thing on it” and waiting for someone to correct me with a painstakingly detailed explanation.

    We might have to come up with ways of curating answers or making that curated content more accessible too, but until we have a universal linux advice chatbot that can answer natural language questions and clear up nuance, people will keep asking questions on public forums.

    Google can’t understand all nuances of the question and latches on to keywords instead. If I search for “Thing does not work in context”, I get answers for how to do thing in context, but not what to do to troubleshoot it if it refuses to show me any decent error message (or doesn’t produce an error, simply unintended results running counter to every one of those answers, including the removeding manual, which I did read thoroughly, several times).