Open Menu
AllLocalCommunitiesAbout
lotide
AllLocalCommunitiesAbout
Login

‘It was ready to kill and blackmail’: Anthropic’s Claude AI sparks alarm, says company policy chief

⁨16⁩ ⁨likes⁩

Submitted ⁨⁨9⁩ ⁨hours⁩ ago⁩ by ⁨return2ozma@lemmy.world⁩ to ⁨technology@lemmy.world⁩

https://www.firstpost.com/tech/it-was-ready-to-kill-and-blackmail-anthropics-claude-ai-sparks-alarm-says-company-policy-chief-13979103.html

source

Comments

Sort:hotnewtop
  • RichardDegenne@lemmy.zip ⁨8⁩ ⁨hours⁩ ago

    Image

    source
  • Deestan@lemmy.world ⁨9⁩ ⁨hours⁩ ago

    They are pushing the “scarily unsafe” angle not because it is, but because that implies it is powerful.

    source
    • Zink@programming.dev ⁨1⁩ ⁨hour⁩ ago

      Yeah, they seem to think we are like the drug addicts who will seek out the batch killing people with ODs because it must be strong.

      Should we tell them that the free first hit is supposed to make the user feel GOOD? Meh, let’s watch how much money they can burn before it all crashes down.

      source
    • givesomefucks@lemmy.world ⁨8⁩ ⁨hours⁩ ago

      Partly that, partly that even if they tried to prevent it from killing humans (if able) it would just do it.

      Like if AI had a button that said “kill 20 random people” it wouldn’t have any reason to not press the button. It doesn’t give a fuck about human life, because it can’t really give a fuck about anything.

      It’ll kill every human just as easily as it I’ll delete an entire codebase.

      It’s not “evil” it’s the lack of anything resembling ethics, including respect for human life. There’s technically a difference. But that’s a fundamental misunderstanding of human psychology

      source
  • RodgeGrabTheCat@sh.itjust.works ⁨8⁩ ⁨hours⁩ ago

    This guy talked about this 4 months ago www.youtube.com/watch?v=f9HwA5IR-sg scary stuff.

    source