Follow
A significant theme in my life is actively choosing not to Goodhart and then being punished for it
:surprised_pikachu:
A significant theme in my life is actively choosing not to Goodhart and then being punished for it
:surprised_pikachu:
If an agent technically "cooperates" but computing the conditions for cooperation is computationally intractable then the agent isn't truly cooperating
Engaged in modal combat with institutions