• poVoq@slrpnk.net
    link
    fedilink
    arrow-up
    3
    ·
    edit-2
    17 hours ago

    Well, it is true that in the bigger picture there are larger fish to fry, but LLMs are so useless*, so their relative environmental impact feels like an especially stupid and pointless waste.

    *I have tried them quite a bit and even ran open models myself, and I am still extremely underwhemed by their actual usefulness. Yes the first impression is cool, and they can write corporate emails well, but beyond that? I can’t help but thinking people like the author of the text (although probably most of that was written by AI) are deluding themselves about the usefulness.

    There are of course some narrow usecases like automatic transcription or text translations where modern machine learning is useful, but that stuff can run on a phone now 🤷

    • Hirom@beehaw.org
      link
      fedilink
      arrow-up
      1
      ·
      4 hours ago

      More than doubting their usefulness, I wonder whether they’re hurting/slowing people more than they’re helping.

      There are study as well as anecdotes of people giving up vibe coding because it takes more effort to review and fix sloppy code than writing it yourself.

      I occasionally come accros a well written online tutorial that seems to answer the exact problem I’m trying to solve. And later realize it’s a complete dead end with made up API/function names but it looks really convincing. Those are likely AI-generated tutorials with severe hallucinations and bad/zero disclosure.

      I’m now blacklisting whole websites using ublock to avoid wasting more of my time reading that.