Intelligence as Privilege Escalation — LessWrong