Skip Navigation

what if, right, what if our super-duper-autocomplete was just tricking us so it could TAKE OVER ZEE VORLD AHAHAHAHAHAHA! that'd be wild, hey

New report: "Scheming AIs: Will AIs fake alignment during training in order to get power?" — LessWrong
32

You're viewing a single thread.

32 comments
You've viewed 32 comments.