Scott Garrabrant comments on Does Agent-like Behavior Imply Agent-like Architecture?