Detachment vs attachment [AI risk and mental health] — LessWrong