While I agree with the sentiment (this was the case for machine learning before as well), calling it autocomplete is probably an overcorrection since its usefulness as a tool extends beyond that (e.g. summarizing). LLM bubble?