Broadly, those considerations make me think that a more natural category to which “AI safety” research belongs is something like “management of risks from agency”.
Broadly, those considerations make me think that a more natural category to which “AI safety” research belongs is something like “management of risks from agency”.