• terabyterex@lemmy.world
    link
    fedilink
    English
    arrow-up
    8
    ·
    edit-2
    4 hours ago

    this worries me with any tech.

    1. if a smaller company dvelopes a competing product (openai and anthropic used to be small) will it hinder them and provide access to only the mion stream companies?

    2. how does this affect non-us models?

    3. will they only be approved if they say wonderful things about trump? have you ever asked grok about elon?

    • Voroxpete@sh.itjust.works
      link
      fedilink
      English
      arrow-up
      11
      ·
      4 hours ago

      I really feel like you’re actually being too generous to this proposal.

      Let’s be clear, when this administration says they want to vet new models, what they mean is that they want to turn them into right wing propaganda engines. This is “reprogram ChatGPT to say the 2020 election was stolen, white genocide is real and trans people are all sex predators, or we won’t certify it.”

    • XLE@piefed.social
      link
      fedilink
      English
      arrow-up
      5
      ·
      4 hours ago

      This is some Cold War regulatory capture BS based on a Myth(os), something that didn’t happen:

      Anthropic did in its Mythos system card, suggest a model has “broken containment and sent a message” when it A) was instructed to do so and B) did not actually break out of any container.