Abuse material monitor says number of AI images assessed this year is already 40% higher than last year’s total
  • FlashMobOfOne@lemmy.worldEnglish
    4 days

    This is, in part, why Microsoft is having human reviewers peruse people’s private prompt data.

    To make Copilot a better smut generator, even of CSAM, according to info leaked to 404 media last week. They’re not reporting the pedophiles who use their LLM for this either, they’re having reviewers evaluate different outputs and pick the ones that are the best fit for the user’s prompt, even for material of questionable legality. I talked about it on my channel this week, but be aware I did make a few off color jokes if you’re offended by that sort of thing.

    • frongt@lemmy.zipEnglish
      4 days

      Well at least it’s better than them abusing actual children?

  • weps@lemmy.worldEnglish
    4 days

    Remember when they discovered a huge child sex ring and locked up one woman and then told us we’re moving on from it

  • Dryad@lemmy.worldEnglish
    4 days

    The companies that produce these images should be prosecuted for each and every one. They must have safeguards against stuff like this out go out of business.

    • gian @lemmy.grys.itEnglish
      4 days

      Fine, as long as you agree to prosecute also all companies that produce software to work with images (or generate them without using AI), every company that produce software that can be used to create 3d models and then the renderer that get these models and render them into photorealistic images.
      Then you need to go after everyone who produces plugins fot these softwares. And then you need to go to every open source software that do the same thing and every person who contribute to them.
      And maybe even hardware vendors that sell graphics tablets for drawing.

      Sometime it seems that, especially with AI, people think that the solution is to ban the tool and not to punish who use it to do something illegal.
      True, AI make it extremely easy to create such horrific images but AI is not the first tool to do so, it make only faster. Back at the time there were entire Usenet groups dedicated to fake nude/xxx images of persons, both celebrities and normal people, but we have not banned internet, we simply prosecuted (when/if caught) who created the images.

      Yeah, that these images are repulsive, but prosecute the companies that make tools that can also generate them was like to prosecute a phone company because someone use a phone to plan and commit an homicide.

      • Dryad@lemmy.worldEnglish
        3 days

        Another commenter mentioned that most of the major platforms do have safeguards against this stuff at this point and are more being done with local models. That’s my ignorance and I’m sorry about that. I would agree with you in general - what you do on your local device should be private, which will protect a lot of important usage but also some nefarious usage. Terms and conditions of freedoms.

        I was thinking of how Grok in particular was just open to generating these sorts of images without any guardrails. In my mind that was similar to a company producing deadly chemicals without any guardrails for preventing it from poisoning everyone around.

    • mystik@lemmy.worldEnglish
      4 days

      This is the pretext for taking away individual’s rights to general computing – because once all the providers have guard rails and are locked out, individual users with local hardware are the last source for ‘illegal content’. “We have to scan everything you do on your local computer, or you could be doing something we don’t approve of”. https://lemmy.world/post/52743759 <-- see this story up on the front page just now. She (foolishly) thought her prompts in OpenAI were private to her, and now she’s charged with “sending a threat in a way that others can read” – we won’t be able to use our computers for private diaries or planning anymore the more they get locked down to only ‘approved’ data and programs.

      And remember, it starts with CSAM, and then it moves on to “illegal content”, then “terrorist content” – and how many times has the Trump said “if you are against the president and the laws I’m enforcing, you must be a terrorist?” …

      • WaxRhetorical@lemmy.worldEnglish
        4 days

        Sorry, what? The other side of this coin is not regulating companies like OpenAI and them using their enormous level of influence to push whatever they want. When a large part of the population uses these platforms, and any (political) query always gets replied to with an answer that fits what they want to achieve, what happens?

        The potential power these AI companies hold is horrifying. They absolutely should be regulated.

      • Nollij@sopuli.xyzEnglish
        4 days

        The Four Horsemen of the Infocalypse - terrorists, pedophiles, drug dealers, and money launderers