Google apologizes for ‘missing the mark’ after Gemini generated racially diverse Nazis::Google says it’s aware of historically inaccurate results for its Gemini AI image generator, following criticism that it depicted historically white groups as people of color.

  • kromem@lemmy.world
    link
    fedilink
    English
    arrow-up
    3
    arrow-down
    1
    ·
    edit-2
    8 months ago

    Actually the way you get it to do better is to put more of the burden on interpreting the context on the LLM instead of heavy handed instructions - because the LLMs do understand the context.

    For example, here’s Gemini answering what the physical characteristics of 1940s soldiers in Germany might have looked like:

    During the Nazi regime in 1940s Germany, racial ideology strictly dictated who was deemed “suitable” for military service. The Wehrmacht, the unified armed forces, prioritized individuals deemed “pure Aryans” based on Nazi racial criteria. These criteria favored individuals with blond hair, blue eyes, and “Nordic” features.

    However, it’s important to remember that the reality was more nuanced. As the war progressed and manpower needs intensified, the Nazis relaxed their racial restrictions to some extent, including conscripting individuals with mixed ancestry or physical “imperfections.” Additionally, some minority groups like the Volksdeutsche, Germans living in Eastern Europe, were also incorporated.

    I think it could have managed to contextualize the prompt correctly if given the leeway in the instructions. Instead, what’s happened is the instructions given to it ask it to behind the scenes modify the prompt in broad application to randomly include diversity modifiers to what is asked for. So “image of 1940s German soldier” is being modified to “image of black woman 1940s German soldier” for one generation and “image of Asian man 1940s German soldier” for another, which leads to less than ideal results. It should instead be encouraged to modify for diversity and representation relative to the context of the request.

    • fidodo@lemmy.world
      link
      fedilink
      English
      arrow-up
      2
      ·
      8 months ago

      I think a lot of the improvement will come from breaking down the problem using sub assistant for specific actions. So in this case you’re asking for an image generation action involving people, then an LLM specifically designed for that use case can take over tuned for that exact use case. I think it’ll be hard to keep an LLM on task if you have one prompt trying to accomplish every possible outcome, but you can make it more specific to handle sub tasks more accurately. We could even potentially get an LLM to dynamically create sub assistants based on the use case. Right now the tech is too slow to do all this stuff at scale and in real time, but it will get faster. The problem right now isn’t that these fixes aren’t possible, it’s that they’re hard to scale.

      • kromem@lemmy.world
        link
        fedilink
        English
        arrow-up
        1
        ·
        8 months ago

        Yes, this is exactly correct. And it’s not actually too slow - the specialized models can be run quite quickly, and there’s various speedups like Groq.

        The issue is just more cost of multiple passes, so companies are trying to have it be “all-in-one” even though cognitive science in humans isn’t an all-in-one process either.

        For example, AI alignment would be much better if it took inspiration from the prefrontal cortex inhibiting intrusive thoughts rather than trying to prevent the generation of the equivalent of intrusive thoughts in the first place.

        • fidodo@lemmy.world
          link
          fedilink
          English
          arrow-up
          1
          ·
          8 months ago

          The issue is just more cost of multiple passes, so companies are trying to have it be “all-in-one”

          Exactly, that’s where the too slow part comes in. To get more robust behavior it needs multiple layers of meta analysis, but that means it would take way more text generation under the hood than what’s needed for one shot output.

          • kromem@lemmy.world
            link
            fedilink
            English
            arrow-up
            1
            ·
            8 months ago

            Yes, but in terms of speed you don’t need the same parameters and quantization for the secondary layers.

            If you haven’t seen it, see how fast a very capable model can actually be: https://groq.com/

            • fidodo@lemmy.world
              link
              fedilink
              English
              arrow-up
              1
              ·
              8 months ago

              Yeah I’ve seen that. I think things will get much faster very quickly, I’m just commenting on the first Gen tech we’re seeing right now.

    • GiveMemes@jlai.lu
      link
      fedilink
      English
      arrow-up
      1
      ·
      8 months ago

      That isn’t “understanding content”, it’s just pulling from historical work that humans did and finding it for you. Essentially, it’s a search engine for all of its training data in this context.