• makingStuffForFun@lemmy.ml
    link
    fedilink
    English
    arrow-up
    5
    ·
    1 day ago

    I 100% believe this.

    We’ve really looked into co-pilot in our example at work to really see what it can do, and oh my god, the programmers in our team are literally excited and terrified at the same time.

    The level that it has come along is astounding and for the first time amongst all the AI hype, and all the AI bashing, I have seen it with my own eyes, how very much we will start relying on these tools.

    We just spent a couple of hours playing with the thing to build up a project and amongst us with literal vibe-coding joy in our hearts and just embracing it, this thing, rather than bashing it and avoiding it like we normally do, we made projects that were fully tested, absolutely rock-damn solid, in hours that would have taken us weeks to make as professionals.

    So now we are back looking at our terminal and our IDE and thinking oh my god I want some more of that.

    So as a lawyer that has an equally technical task at hand with a lot of research they need to pull together, cross-referencing and all the stuff that goes with it, I can imagine that if they have tasted it, I don’t know how they will ever go back to not using it.

    However, from a programming perspective, it either works or it doesn’t. From a legal perspective, that is something completely else, so I guess the comparison is not quite equal.

    Still, the days are numbered.

    • No1@aussie.zone
      link
      fedilink
      English
      arrow-up
      2
      ·
      19 hours ago

      The problem is, that when you cross-examine an AI hard, it eventually does a 180 and gives you exactly the opposite answer.

    • MisterFrog@aussie.zone
      link
      fedilink
      English
      arrow-up
      6
      ·
      edit-2
      1 day ago

      However, from a programming perspective, it either works or it doesn’t. From a legal perspective, that is something completely else, so I guess the comparison is not quite equal.

      To me this is the crux of the matter. It’s why AI is absolutely worthless for engineering (the traditional kind) beyond asking “hey, what’s the part of the standard that says something like XYZ?” and then just going directly to the source.

      LLMs can’t reason.