Better Language Models And Their Implications