It would be amazing if it doesn’t disintegrate if it rains, too.

  • Benjaben@lemmy.world
    link
    fedilink
    arrow-up
    0
    ·
    edit-2
    1 个月前

    Just throwing this out there, but the problem you’re describing sounds like a good fit for an LLM I’ve been using for similar purposes, Claude.

    I’ve found it to be really good at helping me slog through what would be a burdensome and wasteful amount of reading, in order to answer specific questions OR to get a baseline understanding of a thing.

    It’s a bit hard to know how much value comes from my engineering background and my tendency to “know what I don’t know” and thereby ask focused questions, but it’s definitely worth a shot. I have found it to be surprisingly sophisticated and much better than slogging through the wasteland of bad search results + too much unrelated but real info.

    A topic like this where there’s a tremendous amount of legit docs, articles, and forum activity - it’s really the exact use case where it’s very difficult for a human, and very easy for an LLM to effectively digest that info.

    Some caveats I’ve noticed:

    • it sometimes is overly agreeable / “friendly” when it should be more direct
    • it does sometimes hallucinate or say BS with casual confidence, which sucks because the more you need the info the less well you can spot that. It hasn’t hampered usefulness too much for me, but then again I’m usually able to spot the mistakes even in ~unfamiliar subjects
    • they’ve moved the free tier back to a less capable model at the moment…most of my good experiences are with Claude 3.5 Sonnet, but Claude 3 Haiku (present free tier) is still good

    If you’re really curious but the volume of reading and documentation to get started is presenting a big barrier, try using Claude to see how quickly you might be able to clear that obstacle. It’s been removing those exact barriers for me very effectively lately.

    Edit to add: a particularly useful way I can imagine folks in your shoes using this - as a “companion” while you try to follow a guide in an article somewhere. It can answer questions about terms you don’t understand, even reasoning behind doing certain steps or what to do if it goes wrong. In fact, you could almost certainly just feed it the written procedure itself (telling it that you’re doing so) and really get it to reason about the process with you. Just help get you through whatever implementation.

    • ShareMySims@sh.itjust.works
      link
      fedilink
      English
      arrow-up
      0
      ·
      1 个月前

      I genuinely appreciate you taking the time to make this comment, thanks.

      I do see how this can be helpful for a lot of things, but I think in my situation (namely a bunch of neurodiversity and brain fog, and no existing interest in or experience with LLMs), having to learn how to deal with an LLM to a point where I get results I’d be confident enough in to trust without having to double or triple check, is probably a bigger hurdle than just diving in to figuring out security camera set ups. It’s putting one more thing in the way of the information I actually need, which means my brain is much much less motivated than it already is to get the information, if you see what I mean…

      • Benjaben@lemmy.world
        link
        fedilink
        arrow-up
        0
        ·
        1 个月前

        I do understand what you mean, but I think you’re probably significantly overestimating the difficulty of using the tool. One of its major strengths is its ability to just understand you, like you’d talk to anyone human, with the benefit that you can even instruct it to use a style you prefer. Just say “I’d like your answer to be terse, let’s see if we’re on the right track before getting into details”. Just as an example.

        With all that said you know what you want and need better than anyone else, that’s all I’ve got to say on it, cheers!