Researchers at the University of California San Diego and other institutions are working on a way to make a type of artificial intelligence (AI) called diffusion models—a type of AI that can generate new content such as images and videos by training on large datasets—more efficient and widely applicable.