“The performance of large language models will continue to improve, and there is no known fundamental obstacle that will stop this progress.”