i've read it, i do agree with you but maybe some nuance. I think the article is aimed towards LLMs/GenAI/Chatbots, here the statements hold true (for now) I would like to make the comparison with vision AI systems in manufacturing (YOLO,...) A few years ago they also did quite simple/costly demo's ("is this tube actually as round as expected?") but now the tech matured, recently i've seen multiple very in-depth quality control/... usecases. (yet model drift and data trustablity are still hurdles here!) maybe this is just the kickoff phase of LLMs in manufacturing? The demo's are indeed not really impressive and quite costly atm yet but i can imagen in a few years we will ask the LLM “Are any fillers drifting out of spec right now?” I can see that being valuable & finding its place on the shopfloor/maintenance dep. This way we loose the agonising step of iteratively writing queries. (my personal frustration; IT knowledge <> production knowledge) I think/hope we will be able to ask questions to a LLM and get a trustable response in a few years! 🙂