cross-posted from: https://lemmy.cafe/post/29389301

A few days ago X introduced a new AI Image Editing button that lets any user modify images posted by others, even without the original uploader’s consent. Image owners are not notified when edits are made, and the feature is enabled by default with no opt-out option (at least not so far).

  • TehPers@beehaw.org
    link
    fedilink
    English
    arrow-up
    2
    ·
    2 days ago

    Ok yes you’re right. “Grok generate me some CSAM” is the same as opening up a photo editor and drawing a new real looking body onto someone’s child and putting it in a new body position. Same exact thing. No different at all. Twitter has no responsibility for running a service that can do this.

    • locuester@lemmy.zip
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 day ago

      You’ve totally changed the original post’s topic and made it into something obviously unacceptable. There’s a line to cross with content in general, AI or not, and any public ai model should absolutely have safety rails / content moderation on its output

      • TehPers@beehaw.org
        link
        fedilink
        English
        arrow-up
        2
        ·
        edit-2
        1 day ago

        Surely you have an example where it’s appropriate for a service to generate nonconsensual deepfakes of people then? Because last I checked, that’s what the post’s topic is.

        And yes, children are people. And yes, it’s been used that way.

        Edit: as for guardrails, yes any service should have that. We all know what Grok’s are though, coming from Elon “anti-censorship” Musk. I mentioned ChatGPT also generating images, and they have very strict guardrails. They still make mistakes though, and it’s still unacceptable. Also, any amount of local finetuning of these models can delete their guardrails accidentally, so yeah.

        • locuester@lemmy.zip
          link
          fedilink
          English
          arrow-up
          1
          ·
          23 hours ago

          This post isn’t exclusively about deepfakes. It’s about editing someone else’s images. Making suggestive deepfakes is mentioned in the article as an example, but it’s not mentioned in the title or summary here.

          That said, my points stand. Don’t post shit online if you don’t want it to be edited.

          With the ease of photo manipulation, society has no choice but to adapt to nonsensical, simple edits. It can’t be stopped. There’s hundreds of apps and programs that do this now. Even adobe’s famous suite.

          I know you hate Elon, I can hear it in your tone. But this isn’t an Elon thing. Look around.