• TotallynotJessica@lemmy.blahaj.zoneM
      link
      fedilink
      English
      arrow-up
      25
      arrow-down
      1
      ·
      10 hours ago

      In the era of AI, folks always seem to forget that people make mistakes. Artists fuck up anatomy, infographic designers make typos, none of that means AI was involved.

      The protester doesn’t match the description and could be AI, or it could be a non AI stock image that the infographic creator found. It’s hastily slapped together regardless, but that doesn’t mean it’s AI.

      People have a serious Dunning-Kruger problem with spotting AI; thinking any errors in an otherwise professional looking image are valid evidence. You really need to look for mistakes humans are less likely to make.

      • Miner_Fabs@lemmy.world
        link
        fedilink
        English
        arrow-up
        8
        arrow-down
        1
        ·
        5 hours ago

        I agree with your point in general - there’s no use in being skeptical if you call stuff fake at the first sign. Innocent until proven guilty or whatever.

        You’re right that the image could be clipart, and a human could have made those typos, but there are some non-human tells in here other than that.

        Zoomed in version of the post image, highlighting several errors

        The P in protection is a lowercase capital, and one of the r’s in irŕritation is accented. The latter could’ve been mistyped on a phone keyboard, but in that case it would’ve been autocorrected, and that still doesn’t explain the P.

        There’s also some weird GPT-like descriptions that are thrown in for no reason. Why does the bandana being “fashionable” matter? How are arm warmers “subtle”, exactly?

        As another commentor pointed out, a T-shirt with a graphic is more likely to make you recognisable, so that part’s just wrong. I also don’t get why “tactical/military bags” should be “avoided”? Seems like the kinda misinformation AI would spout.

        What I really don’t get here is why someone would use AI to generate the entire thing, rather than only using it for the graphic (still morally questionable but excusable) and adding the text in manually (something you can do in MS paint).

      • okwhateverdude@lemmy.world
        link
        fedilink
        English
        arrow-up
        6
        arrow-down
        2
        ·
        10 hours ago

        You really need to look for mistakes humans are less likely to make.

        Yeah, like the “descriptions and pointers / connectors are all messed up.” Or including nonsense like “tain”.

        I am all for benefit of the doubt here, but only when it is warranted. The probability of this being a single prompt generated image is very, very high. And no amount of “hastily slapped together” excuses it.

        • null_dot@lemmy.dbzer0.com
          link
          fedilink
          English
          arrow-up
          4
          ·
          5 hours ago

          Am I crazy for thinking these mistakes seem very human and not very AI?

          Like some other commenter said, “tain” isn’t nonsense. It’s a typo for rain, t being adjacent to r on the keyboard and all.

          The mixed up pointers is really just out of order text boxes. Illustration software manages the flow of text between text boxes, and something misconfigured explains the mixed up connectors. Conversely this doesn’t seem like an AI error given that the intelligence making the image knows what a shoe is.