2. Will such systems still be useful? — LessWrong