• Admiral Patrick@dubvee.org
      link
      fedilink
      English
      arrow-up
      7
      ·
      edit-2
      6 months ago

      Oh, wow. It really isn’t. Axios usually does really good reporting, but that looks more like the outline / notes for a story than something ready to publish.

      I strongly dislike generative LLMs (I refuse to call them AI) for a host of reasons, but the biggest reason has less to do with the tech and more to do with the people / upper brass who are trying to replace human jobs with them and expecting it to just work (while salivating at the thought of pocketing the salary of the displaced human employee).

      I don’t think the article really calls that out explicitly, but they are saying it’s not living up to the hype. As far as progress goes, I suppose that’s a good first step.

    • Creesch@beehaw.org
      link
      fedilink
      arrow-up
      4
      ·
      6 months ago

      For real, it almost felt like an LLM written article the way it basically said nothing. Also, the way it puts everything in bullet points is just jarring to read.

  • millie@beehaw.org
    link
    fedilink
    English
    arrow-up
    8
    ·
    edit-2
    6 months ago

    Frankly, corporations seem to have no idea how to use LLMs. They want them to be a public facing company representative, which is exactly what LLMs can’t do well. Where they accel is as an assistant.

    Want to figure out what scale you’re playing a song in? It’s great at that. I’ve had it give me chords to go with scales too, or even asked for some scale options based on the feeling of the sound.

    It’s also great for looking for terms in other languages. I’ve got some ranged weapon abilities in my tabletop rpg. I knew i wanted one of them to be called pistolero, but I didn’t know the terms fusilero or escopetero, and might not have found them on my own, but chatgpt came up with them right away.

    I’ve also learned that it’s great at looking up game guides and providing hints that aren’t spoilers without giving the puzzle away. I had it generate results for the Lady’s Maze in Planescape: Torment and the Water Temple in Ocarina of Time. Amazing hints without giving it away.

    If you have your own brain and want to off-load some simple queries, it’s great. If you want to use it in place of a human brain to talk to customers, you’re barking up the wrong gpt.

    • averyminya@beehaw.org
      link
      fedilink
      arrow-up
      3
      ·
      6 months ago

      That music example is how I’ve used them, it really is spot on. Key, tempo, scale, overlapping scales that could be used, plus factoids included. It really can be very helpful.

  • darkphotonstudio@beehaw.org
    link
    fedilink
    arrow-up
    2
    ·
    6 months ago

    AI doesn’t have to be good, it only needs to be good enough. Even if it’s just barely functional, if it’s cheaper than paying a human, then it will be used by capitalists.

  • noxfriend@beehaw.org
    link
    fedilink
    arrow-up
    1
    ·
    edit-2
    6 months ago

    improving and integrating the technology is raising harder and more complex questions than first envisioned

    Many people not only envisioned but predicted these problems as soon as the hype cycle began.

    Interesting article. I’d have loved to see some stats on how LLM investment and LLM startups are doing.