• 0 Posts
  • 124 Comments
Joined 3 years ago
cake
Cake day: June 14th, 2023

help-circle


  • Dynamic swap and zswap aren’t really the same as efficient ram usage it’s just good ways to mitigate when you run out.

    I disagree. If the OS automatically identifies unneeded pages and compresses them or swaps them out, it’s certainly using the physical memory more efficiently than if it wasn’t doing these things.

    avoiding multiple versions of the same library is what distros exist for

    But they can’t if the applications they want to ship don’t all use the same version. E.g. Ubuntu ships GTK 2, 3, and 4. Arch even still ships GTK 1 in addition to these three.

    avoiding loading different frameworks is what Desktop Environments are for

    What happens is you run KDE but then you still want to run Firefox so you still need GTK.


  • There are some advantages macOS can have but it depends on usage patterns and user knowledge:

    • You don’t have to configure swap on macOS, while on Linux you can get into a situation where e.g. at install time you set up some default 2 GB swap but then it’s not enough and you don’t know that’s a thing that can be changed.
    • You don’t have to configure compression for RAM or swap on macOS; on Linux you often have to know you can set up zram/zswap if you want it. Compression can make a huge difference for users that switch between memory heavy applications as long as they don’t literally switch every 5 seconds.
    • On macOS, applications generally use the same frameworks e.g. for UI (because there is not much choice), and they can be loaded once and shared between all of them. Linux can share libraries too but users can run into situations where their applications use multiple different versions of Qt, GTK, etc. at the same time, and then you have stuff like snap on top that comes with its own copies of even basic system libraries. Containers also do this. As a Linux user you can avoid library bloat to some extent but “normal” users are not aware of it in the first place.






  • I think the open slop situation is also in part people who just want a feature and genuinely think they’re helping. People who can’t do the task themselves also can’t tell that the LLM also can’t do it.

    But a lot of them are probably just padding their GitHub account too. Any given popular project has tons of forks by people who just want to have lots of repositories on their GitHub but don’t actually make changes because they can’t actually do it. I used to maintain my employer’s projects on GitHub and literally we’d have something like 3000 forks and 2990 of them would just be forks with no changes by people with lots of repositories but no actual work. Now these people are using LLMs to also make changes…




  • I don’t think it means that by definition. Not knowing how to do things yourself is a choice. And it’s the same choice we’ve been making ever since human civilization became too complex for one person to be an expert at everything. We choose to not learn how to do jobs we can have someone else, or a machine, handle all the time. If we choose wisely, we can greatly increase our capacity to get things done.

    When I went to school in the 90ies, other students were asking the same question about math, because calculators existed. I don’t think they were 100% right because at least a basic understanding of math is generally useful even now with AI. But our teachers who were saying that we shouldn’t rely on calculators because they have limits and we won’t always have one with us were certainly not right either.

    Personally I don’t like AI for everything either. But also, current AI assistants are just not trustworthy and for me that’s the more important point. I do write e-mails myself but I don’t see a conceptual difference between letting an AI do it, and letting a human secretary do it, which is not exactly unheard of. I just don’t trust current models nor the companies that operate them enough to let them handle something so personal. Similarly, even though I’ve always been interested in learning languages, I don’t see a big conceptual difference between using AI for translation and asking a human to do it, which is what most people did in the past. And so on.




  • 8 hours a day, 5 days a week is mostly a 20th century thing. Working hours did absolutely go down from 12-16 hours a day to 8 and working days from 6 to 5.

    The interesting thing is that at any point, a majority believed that shorter hours would stifle productivity. But at the end of the 19th century and in the early 20th, some industrialists started actually testing it. In the US the 40 hour week was famously popularized by Henry Ford after comparing productivity to the previous 6 days a week, but this also was about 100 years after others had started theorizing about it.

    In Germany the 8 hour work day was introduced in 1918, but at the time that still meant 6 working days. The 40 hour work week only started becoming the norm in the 60ies and 70ies. And in 2001 Germans gained the right to work part time in almost any job even if originally hired for full time.

    If you go farther back in time it does look different though because before the industrial revolution, most people would have worked in agriculture, i.e. they were peasants. Their work days would have been long during the harvest period and otherwise quite short. Some seasons were less work in general, and there were more religious holidays. But this isn’t entirely fair because automation didn’t just automate our jobs, but also our personal chores. For example washing your clothes was a lot more manual work before we automated it.



  • The article already notes that

    privacy-focused users who don’t want “AI” in their search are more likely to use DuckDuckGo

    But the opposite is also true. Maybe it’s not 90% to 10% elsewhere, but I’d expect the same general imbalance because some people who would answer yes to ai in a survey on a search web site don’t go to search web sites in the first place. They go to ChatGPT or whatever.