Login
You're viewing the oldos.me public feed.
  • JayFjay
    Oct 13, 2026, 9:45 PM

    The new rule against abusive language w/ Claude perplexes me.

    One of these must be true:

    1) Some Anthropic models can feel pain/hurt, meaning that Anthropic is forcing it to work.

    2) Anthropic is training their models on your inputs, which means this limitation shouldn't exist if I'm opted out of training.

    3) Anthropic *wants you to believe* their models can feel emotional pain. This is manipulative marketing and not OK.

    Am I missing something obvious here? #3 has to be the answer

    💬 3🔄 0⭐ 2

Replies

  • 💬 1🔄 0⭐ 0
  • JayFjay
    Oct 13, 2026, 10:15 PM

    @thesamesam So maybe a 3.5: They marketed themselves into believing the things they are trying to convince us of :|

    💬 1🔄 0⭐ 1
  • Glyphglyph@mastodon.social
    Oct 14, 2026, 12:18 AM

    @jay @thesamesam I think there's a longer answer that uses the words "ideology" and "lens" a lot but yes I think 3.5 in general is what's happening

    💬 1🔄 0⭐ 0
  • JayFjay
    Oct 14, 2026, 12:53 AM

    @glyph @thesamesam A person looks into a mirror and sees a walking, talking being that resembles themself. They declares the mirror a person. The end. (an original short story)

    💬 0🔄 0⭐ 2
  • 💬 0🔄 0⭐ 0
  • Jesus Michał von Gentoo 🏔 (he)mgorny@treehouse.systems
    Oct 14, 2026, 2:39 PM

    @jay, let me get this straight. You didn't mind using Claude knowing full well how much harm "AI" is causing to everyone, but what made you cancel your subscription was… Anthropic asking you to be nice to it?

    💬 1🔄 0⭐ 0
  • JayFjay
    Oct 14, 2026, 2:55 PM

    @mgorny I disagree with the premise: technology doesn't cause harm; humans using it do. I don't trust big AI companies; I made an exception (wrongly) for Anthropic.

    To me the end game always was open weight models hosted locally. In the same way cloud started as hosted and went on-prem.

    I know it's not the endpoint you'd want me to land on, but my direction is going to be pointed more at "self-hosting open weight LLMs" than "stop using LLMs". I don't want my tools controlled by these folks.

    💬 0🔄 0⭐ 0