The first obvious casualty of large language models is homework: the real training for everyone, though, and the best way to leverage AI, will be in verifying and editing information.