• xia@lemmy.sdf.orgOP
      link
      fedilink
      English
      arrow-up
      4
      arrow-down
      2
      ·
      1 day ago

      It was not immediately obvious from the image (though you might see the starfish for scale), but these things are huge!

  • Binzy_Boi@piefed.ca
    link
    fedilink
    English
    arrow-up
    61
    arrow-down
    1
    ·
    edit-2
    2 days ago

    “Hey guys, look at this image I made showing how much I care about the environment”

    It’s AI generated.

    This doesn’t help your case you know.

  • juliebean@lemmy.zip
    link
    fedilink
    arrow-up
    13
    ·
    2 days ago

    clearly everyone in the comments immediately pegged this as ai. i did too, but i am not sure what it is specifically that makes this feel so obviously ai. anyone got any ideas what it is that we’re all picking up on here?

    • SSUPII@sopuli.xyz
      link
      fedilink
      arrow-up
      11
      ·
      2 days ago

      This is the art style that OpenAI forced its model to generate. Nobody else draws this way.

      • xia@lemmy.sdf.orgOP
        link
        fedilink
        English
        arrow-up
        5
        ·
        1 day ago

        It makes you wonder if they have a whole bunch of training data in this style, or if it is the mathematical average of all cartoon styles mashed together.

        • SSUPII@sopuli.xyz
          link
          fedilink
          arrow-up
          5
          ·
          edit-2
          1 day ago

          If it was a simple average it would have been mangled. This is a deliberate fine-tune of the model to the particular style. For fine-tuning you need some type of input, if generated or human made doesn’t matter.

          The reason why they fine-tuned to this particular style is unknown, but they might be:

          • To quickly and not as expensively produce and release a usable model in the then more heated, now slowing down competition for better and better models
          • To reduce the model size as there is no need for lots of data on multiple styles.
          • To give a distinct comics style to their model, so to make people associate the image to their model instead of OP

          There is nothing wrong with fine-tuning, and it is very often necessary to have the output not be gibberish.