• skarn@discuss.tchncs.de
    link
    fedilink
    arrow-up
    6
    ·
    2 days ago

    What does agentic AI even mean? If it’s this thing Altman keeps talking about It’s mostly scifi. The thing that’s being proposed now is not anything I think people would call agentic AI. It’s closer to an automated moderation filter. And it’ll never work anyway.

    There is no guarantee that this regulation will ever pass at all. Hopefully it won’t, not this time around or in this form.

    Apple leaving on principle? It operates in China with plenty of compromises.

    Signal on the other hand might.

    • 9tr6gyp3@lemmy.world
      link
      fedilink
      English
      arrow-up
      2
      arrow-down
      2
      ·
      2 days ago

      Im not sure what you’re referring to with Altman, but we are already having browsers with AI built in, text editors, IDEs, social media, etc etc.

      Microsoft already has a sort of agentic AI running on Windows machines called Recall. It takes screenshots at intervals and has AI kind of keep track of what you were doing in those screenshots.

      If they cant have backdoors into your messages, they’ll just have AI summarize your activity and send anything that appears illegal to whoever is monitoring for that.

      But like I said, it’s just my opinion and completely hypothetical.

      • skarn@discuss.tchncs.de
        link
        fedilink
        arrow-up
        1
        ·
        edit-2
        19 minutes ago

        Yeah, AFAICT that’s not what agentic AI is supposed to mean.

        Agentic AI is supposed to be AI that can run tasks autonomously. The typical examples would be order you take out by clicking around in a browser or phone app.

        Edit: and anyway the current proposal, even in it’s most degenerate form, is not suggesting anything remotely as criminal as having everyone’s activity constantly summarized. It’s “just” supposed to be a content filter that will report “illegal” content. This is horrible in many ways, namely the fact that in the best case false positive would massively outnumber real positives. The safeties built around these reports also seem too low. So the picture I send to my wife of our child playing in the bathtub would stay in a state database for years. Fuck no. And I could go on.

        But nowhere is anyone remotely proposing having an AI send a fucking daily summary of instant messaging activity.