How about the preservation of books? I'm working on converting scans (OCR and otherwise) of long out of date, public domain books to a clean digital form (web, PDF, and ePUB) and the hours upon hours of editing typos made by OCR seems like something a large language model or contextual syntax and spelling analyzer should be able to easily sniff out and correct easily enough. My goal is just to preserve knowledge in as compact a format as possible (along with the basis for a clean re-publishing effort if someone wanted to take that on).