x
Why AIs aren't power-seeking yet — LessWrong