AI takeoff story: a continuation of progress by other means — LessWrong