• Ephera@lemmy.ml
    link
    fedilink
    English
    arrow-up
    7
    ·
    1 day ago

    Yeah, leaving moral reservations aside, it’s especially annoying to me, because it’s being pushed with complete disregard whether it actually helps me.

    I’ve been working in a programming language for the past two years, in which I’m well-trained. Better than the statistical average that LLMs blurt out, at the very least. So, I’ll often end up correcting whatever it generates, rather than just typing out the same directly. In particular, I also find it much easier to think while typing, rather than while reviewing code, so I need pauses to think anyways. And I also just find it disrupts my concentration when the autocompletion-style LLMs keep flickering their suggestions at me.

    Similarly, flavor images. So much of management is fucking excited about generative AI, because they can type shit like “wombat hanging off of a line of code” and then it slops out an image, which they can slap into their presentation and pretend it has meaning.
    I don’t like those images. The AI-generated ones look terrible to me, but I did not either like them before they were AI-generated. It’s just pointless imagery, why are you showing me this?
    Obviously, management can disagree with my stance, many people do, but if they want me to present shit, they need to respect that my presentation style just does not include flavor images, no matter what flavor image generator we pay for.

  • MeatsOfRage@lemmy.world
    link
    fedilink
    arrow-up
    4
    arrow-down
    6
    ·
    edit-2
    1 day ago

    These people all work for what seems like huge (video game) businesses. It’s a pretty entitled position to say results aren’t the most important thing. The fact is, they are. If you want the freedom to do things your own way at your own pace, go make your own indie game. I guarantee these “journey over results” types would end up with feature creep and never ship a finished product though.

    On the coding side, I’m a senior dev who has been in the industry for 20 years. Not video games but for the web. Easily 40% of the code I ship now is AI written. I’ll use a different AI to review my PRs saving my coworkers time catching the little mistakes I overlooked.

    I’ve cut days off my workload over the past year and to say all AI is slop is being intentionally obtuse to the tools. Basically we’ve all been promoted to manager roles and LLMs are the mid level workers we oversee.

    This article is trying to paint a picture that these tools are without value and I can feel your finger hovering over the downvote button but the fact is the genie is out of the bottle. They’re not going away and today is the worst they’ll ever be. People that say they’re slower with these tools are going to get left behind by people who are proficient with them.

    • Infynis@midwest.social
      link
      fedilink
      English
      arrow-up
      6
      arrow-down
      1
      ·
      1 day ago

      “I’m an accountant and I use a calculator, so all artists should have to use photoshop.”

    • BaconIsAVeg@lemmy.ml
      link
      fedilink
      English
      arrow-up
      3
      arrow-down
      1
      ·
      1 day ago

      As a web/devops dev, I 100% agree with your statement. I use GitHub Copilot in Neovim, Zed, Emacs, most of the time it just finishes my sentences or generates comments/docblocks/unit tests, I’m not using it to generate features.

      On the flip side though, the company also went all in on Microsoft Copilot, and holy fuck it’s obnoxious trying to use Excel or Outlook with it shoving itself into your face every time you try to do something. And I have no way to disable it.

    • VinS@sh.itjust.works
      link
      fedilink
      arrow-up
      2
      arrow-down
      1
      ·
      1 day ago

      Hi there, using intellij but I can’t seem to find an Ai code review, what tools are you using?

      • MeatsOfRage@lemmy.world
        link
        fedilink
        arrow-up
        3
        ·
        edit-2
        1 day ago

        On GitHub, after you make a PR go up to the URL. Right after the pr number add .diff

        This will give you a full pr diff. Paste that into what LLM you’re using. I usually prompt with something like

        “You’re a senior web engineer. Give this PR a review being aware of modern code practices and syntax. Try to uncover possible bugs. Give 3 to 5 actionable suggestions.”

        There’s probably a way to do this in the tools, I’ve just found this works.