Using generative AI to design, train, or perform steps within a machine-learning system is risky, argues computer scientist Micheal Lones in a paper appearing in Patterns. Though large language models ...