LLMs in general, I mean. They seem to be the first widespread application for large, unstructured datasets. Still hype-y, but maybe even a /practical/ application.