ChatGPT-4 was 8% less accurate than emergency room doctors in training, a study found and Dr. Darien Sutton explains.