• shiveyarbles@beehaw.org
    link
    fedilink
    arrow-up
    6
    ·
    1 year ago

    Yeah I guess, I just see a mishmash of consumed data presented where you have to tweak parameters and so forth. Some gobbledygook nonsense presented as facts, three arms and six fingers in generated art, etc. just seems like shit to me

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

      Yesterday I gave OpenAI’s latest chatbot a photo of a challenging board game quiz card with questions that I couldn’t answer.

      The questions were intentionally difficult, no ordinary human is expected to be able to answer them all - at least not without spending an hour googling/etc. Most of us could only answer a couple of the questions before the timer ran out and we all compared answers.

      The new version of ChatGPT answered every question, perfectly, in two seconds. It couldn’t do that a week ago, the tech is advancing incredibly fast.

      There are definitely some things it’s not very good at, but there are equally things it’s very very good at - the technology is useful, unlike crypto which I see as an interesting solution to a problem that nobody has.

    • mayooooo@beehaw.org
      link
      fedilink
      arrow-up
      2
      ·
      1 year ago

      I’m sure there are use cases, it’s just that there’s much too much hullabaloo about it all. Some people can get something out of it, they’ll fire their workers, that’s about it. But I can’t use it for shit because I want to control what’s being done, don’t do text processing at all. But I smell the crypto dudes that are clinging to the whole thing, a dripping mess of turds wanking off to fanciful stories about ai. It’s language models all the way down

    • V ‎ ‎ @beehaw.org
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 year ago

      At my company we have already used it to great length. We had a backlog on several thousand support tickets we wanted categorized. GPT-4 did it in about 8 hours and with over 80% accuracy, at a fraction of the cost (and higher quality) it would have taken to get humans to do it.

      We’re rolling out a chat bot too using it, with a local model as backup, to reply to leads when our clients are busy. So far they love it.

      We’re making our money back despite the costs, and we’re able to spend that money paying people to not do busy work.