• TheKrunkedJuan@lemmy.world
    link
    fedilink
    English
    arrow-up
    17
    ·
    7 months ago

    As someone scripting a lot for my department in the tech industry, yea AI and scripts have a lot of potential to reduce labor. However, given how chaotic this industry is, there will still need to be humans to take into account the variables that scripts and AI haven’t been trained on (or are otherwise hard to predict). I know the managers don’t wanna spend their time on these issues, as there’s plenty more for them to deal with. When there’s true AGI, that may be a different scenario, but time will tell.

    Currently, we need to have some people in each department overseeing the automations of their area. This stuff mostly kills the super redundant data entry tasks that make me feel cross eyed by the end of my shift. I don’t wanna be the embodiment of vlookup between pdfs and type the same number 4+ times.

    • misspacific@lemmy.blahaj.zone
      link
      fedilink
      English
      arrow-up
      17
      ·
      7 months ago

      exactly, this will eliminate some jobs, but anyone who’s asked an LLM to fix code longer than 400 lines knows it often hurts more than it helps.

      which is why it is best used as a tool to debug code, or write boilerplate functions.

      • Ragnarok314159
        link
        fedilink
        English
        arrow-up
        10
        ·
        7 months ago

        Do you think AI for programmers will be like CAD was for drafters? It didn’t eliminate the position, but allows fewer people to do more work.

        • misspacific@lemmy.blahaj.zone
          link
          fedilink
          English
          arrow-up
          10
          ·
          7 months ago

          this is pretty much what i think, yeah.

          a lot of programming/software design is already kinda that anyway. it’s a bunch of people who were educated on computer science principles, data structures, mathematicians, and data analytics/stats who write code to specs to solve very specific tool problems for very specific subsets of workers, and who maintain/update legacy code written decades ago.

          now, yeah, a lot things are coded from scratch, but even then, you’re referencing libraries of code written by someone awhile ago to solve this problem or serve this purpose or do thing, output thing. that’s where LLMs shine, imo.

        • rottingleaf@lemmy.zip
          link
          fedilink
          English
          arrow-up
          1
          ·
          7 months ago

          No. More high-level languages with less abstraction leakage are like CAD for drafters. Not “AI”.

          I personally would want such tools to be more visual and more like systems, not algorithms.

          Like interconnected nodes in a control system. Like PureData for music, or like LabView. Maybe more powerful and general-purpose.

      • hansl@lemmy.world
        link
        fedilink
        English
        arrow-up
        4
        arrow-down
        9
        ·
        edit-2
        7 months ago

        You’ll get blindsided real quick. AIs are just getting better. OpenAI are already saying they moved past GPT for their next models. It’s not 5 years before it can fix code longer than 400 lines, and not 20 before it can digest a specification and spout a working software. Said software might not be optimized or pretty, but those are things people can work separately. Where you needed 20 software engineers, you’ll need 10, then 5, then 1-2.

        You have more in common with the guy getting replaced today than you care to admit in your comment.

        Edit: not sure why I’m getting downvoted instead of having a discussion, but good luck to you all in your careers.

        • misspacific@lemmy.blahaj.zone
          link
          fedilink
          English
          arrow-up
          7
          ·
          7 months ago

          i didn’t downvote you, regardless internet points don’t matter.

          you’re not wrong, and i largely agree with what you’ve said, because i didn’t actually say a lot of the things your comment assumes.

          the most efficient way i can describe what i mean is this:

          LLMs (this is NOT AI) can, and will, replace more and more of us. however, there will never, ever be a time where there will be no human overseeing it because we design software for humans (generally), not for machines. this requires integral human knowledge, assumptions, intuition, etc.

          • hansl@lemmy.world
            link
            fedilink
            English
            arrow-up
            2
            arrow-down
            2
            ·
            edit-2
            7 months ago

            LLMs (this is NOT AI)

            I disagree. When I was studying AI at college 20+ years ago we were also talking about expert systems which are glorified if/else chains. Most experts in the field agree that those systems can also be considered AI (not ML though).

            You may be thinking of GAI or Universal AI which is different. I am a believer in the singularity (that a machine will be as creative and conscious as a human), but that’s a matter of opinion.

            I didn’t downvote you

            I was using “you” more towards the people downvoting me, not you directly. You can see the accounts who downvoted/upvoted, btw.

            Edit: and I assumed the implication of your comment was that “people who code are safe”, which is a stretch I was answering to. Your comment was ambiguous either way.

              • hansl@lemmy.world
                link
                fedilink
                English
                arrow-up
                2
                arrow-down
                2
                ·
                7 months ago

                Wow. Thanks for the advice. I guess that’s just Lemmy showing me the door. Good luck with your community here.

                • Welt@lazysoci.al
                  link
                  fedilink
                  English
                  arrow-up
                  2
                  ·
                  7 months ago

                  Try not to let the bot hurt your feelings, it was trained on cunts ‘n’ assholes

        • aesthelete@lemmy.world
          link
          fedilink
          English
          arrow-up
          5
          ·
          edit-2
          7 months ago

          Where you needed 20 software engineers, you’ll need 10, then 5, then 1-2.

          It’s an open secret that this is already the case. I have seen projects that went on for decades and only required the engineering staff they had because corporate bureaucracy and risk aversion makes everyone a fraction as effective as they could be, and, frankly, because a lot of ineffective morons got into software development because of the $$$ they could make.

          Unless AI somehow eliminates corporate overhead I don’t understand how it’ll possibly make commercial development monumentally easier.

    • rottingleaf@lemmy.zip
      link
      fedilink
      English
      arrow-up
      1
      ·
      7 months ago

      Scripting is one thing and unpredictable plagiarism generator is another.

      If you mean ML text recognition, ML classification etc - then yeah, why not.