• Dr. Moose@lemmy.world
    link
    fedilink
    English
    arrow-up
    16
    arrow-down
    6
    ·
    2 days ago

    LLMs - you can just talk to your computer. I used computers since the 90s and that has always been the ultimate scifi dream and it’s here.

    • daddycool@lemmy.world
      link
      fedilink
      arrow-up
      3
      ·
      1 day ago

      I wouldn’t say that’s crazier than a 70’s nerds wildest dreams. In the 70’s they thought talking to your computer was just around the corner. They would more likely be amazed that it took so long.

    • ragas@lemmy.ml
      link
      fedilink
      arrow-up
      9
      arrow-down
      1
      ·
      2 days ago

      You can talk to dumb chatbots since the nineties. Now they can into grammar, wow.

      (I know they are way better now. But the basic fact stands.)

      • fuck_u_spez_in_particular@lemmy.world
        link
        fedilink
        arrow-up
        1
        ·
        15 hours ago

        It’s a little bit more than that, I would say at this point.

        As I said somewhere else I’m not really into LLMs, so I try to stay neutral…

        I’m currently using Astra a little bit more, and it’s really interesting to see how these models work these days. You increasingly feel the (self-)reinforcement learning on these models. They developed an interesting code style (not very human…). Also the way it explores the code (and other stuff) like writing scripts on the fly to either explore or monitor, and even writing scripts to build the sourcefiles. And well, it definitely currently reshapes software development…

    • fuck_u_spez_in_particular@lemmy.world
      link
      fedilink
      arrow-up
      6
      ·
      2 days ago

      As much as I have a love-hate relationship with these because of all the negative effects (And there are countless ones), it’s quite impressive what current Frontier models can do at this time, with very little input.