101
AI is learning to lie, scheme, and threaten its creators during stress-testing scenarios (fortune.com)
submitted 1 year ago by to c/technology@lemmy.world
22 comments fedilink hide all child comments
you are viewing a single comment's thread
view the rest of the comments
[–] 1 point 1 year ago (1 child)

I don't necessarily disagree with anything you just said, but none of that suggests that the LLM was "manipulated into this outcome by the engineers".

Two models disagreeing does not mean that the disagreement was a deliberate manipulation.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 2 points 1 year ago

    I'm definitely not saying this is a result of engineers' intentions.

    I'm saying the opposite. That it was an emergent change tangential to any engineer goals.

    Just a few days ago leading engineers found model preferences can be invisibly transmitted into future models when outputs are used as training data.

    (Emergent preferences should maybe be getting more attention than they are.)

    They've compounded in curious ways over the year+ since that happened.

  • permalink
  • fedilink
  • source
  • parent