Today’s generative artificial intelligence models can create everything from images to computer applications, but the quality of their output depends largely on the prompt a human user provides.