• how_we_burned@lemmy.zip
      link
      fedilink
      arrow-up
      1
      ·
      1 hour ago

      It does my head in how we ignore the shit answers language models give (I refuse to call them intelligent).

      I swear unless it’s baby shit, like where is the syntax error in this script, it almost always gets it wrong.

      Making fresh scripts, even with padentic level prompting and detail just ends up with a script with multiple errors.

      I’ve realised I would have written it just as quickly (after all the iteration work) if I had just done it myself.

      The only I find LLMS are good for are glorified search engines. And even then it’s horrifying how inefficient chatgpt is for search compared to say Google.

      Not to mention it’s run by a sociopath.

  • 4am@lemmy.zip
    link
    fedilink
    arrow-up
    16
    ·
    6 hours ago

    Have you seen AI with numbers? It doesn’t calculate, it just trying to give a statistically-likely answer (just like it does for every other next-word in its answer).

    Some of them have to drop back to deterministic software tools (and even then sometimes they’re called with incorrect parameters because “intent” lol)

    AI already emulates dumbasses it just got no legs. Don’t worry though, soon they’ll give it legs and guns. Nothing can go wrong guys. trust me bro the future is now

  • A_Random_Idiot@lemmy.world
    link
    fedilink
    English
    arrow-up
    12
    ·
    6 hours ago

    I can see AI measuring something 3 times and cutting it wrong, more than I can see AI measuring things once and cutting it right.

  • slazer2au@lemmy.world
    link
    fedilink
    English
    arrow-up
    98
    ·
    10 hours ago

    What are you talking about? They will measure it wrong 3 times, cut it wrong while saying it’s correct.

  • nroth@lemmy.world
    link
    fedilink
    arrow-up
    1
    ·
    3 hours ago

    I think AI might end up being more successful in the low-complexity trades (e.g., install a disposal) than for high-complexity white-collar work. The robots just take time to develop.

    • nroth@lemmy.world
      link
      fedilink
      arrow-up
      2
      ·
      3 hours ago

      A white collar example where AI is effective: Make the button this specific shade of blue and move it 3 pixels to the left

  • TankovayaDiviziya@lemmy.world
    link
    fedilink
    arrow-up
    9
    arrow-down
    1
    ·
    7 hours ago

    The only thing hindering AI from taking 99% of the jobs is that robotic engineers still can’t create a robot with the same level of dexterity as human hands.

    • mrgoosmoos@lemmy.ca
      link
      fedilink
      English
      arrow-up
      1
      ·
      3 hours ago

      well, that and the fact that it can’t actually do something to the same level of quality.

      but that’s not going to stop it from taking the jobs. only from actually completing them successfully after it does so.

    • pinball_wizard@lemmy.zip
      link
      fedilink
      arrow-up
      3
      ·
      edit-2
      6 hours ago

      I mean, I would add to this list:

      • Humans are way cheaper to recruit, maintain, and produce. Robots only compete on their potential for sociopathicly blind loyalty.
      • Computers are deeply fucking stupid in ways that most normal humans cannot even understand, until something tragic happens directly to them, as a reuslt. (And no, AI hasn’t fixed this. It put a really cool looking coat of razor thin paint over it.)
        • mrgoosmoos@lemmy.ca
          link
          fedilink
          English
          arrow-up
          1
          ·
          edit-2
          3 hours ago

          eh, sort of.

          I work in automotive manufacturing. it’s a huge world. there are tons of people involved in designing, building, and running these cells.

          also, the human part is pretty much always the most basic shit - loading and unloading feeders, moving material around with a forklift, loading parts by hand into the machine from material bins because it’s cheaper to hire someone to do that complex task (that is not sarcastic, picking something up from a loose bin and placing it in a known orientation is a difficult task to automate) than to teach and operate a robot for that

          I don’t know how line operators stay sane. I’ve designed cells where it’s somebody’s job to do the same dozen motions every minute for an entire shift. many cells like that.

    • LastYearsIrritant@sopuli.xyz
      link
      fedilink
      English
      arrow-up
      2
      ·
      6 hours ago

      Realistically, most jobs don’t NEED hands. Most physical jobs can be done better with customized manipulators for the type of work being done.

      We’d only need hands if we wanted a single robot model that could do multiple types of jobs, instead of a specific robot for each job.

  • AeronMelon@lemmy.world
    link
    fedilink
    arrow-up
    27
    ·
    10 hours ago

    I assure you it can give you three different measurements from the same picture then go on to explain why it’s okay to eat the plywood.

  • JasonDJ@lemmy.zip
    link
    fedilink
    arrow-up
    3
    ·
    6 hours ago

    Hey, why don’t you try cutting your dimensional lumber with a gas-powered chainsaw and be more accurate, eh?