I agree and, for all the data that "stopped" in 2021, the prompts and questions that are being streamed into the live data set from billions of uses will add sociological context, value, weight and sentiment to the existing and modified inputs. WE are the dataset. Back in the day when Google announced Project Gutenberg to digitise books and make them available, the assumption was that it was for access by search engine, but inr eality the text forms part of a large training dataset. The future is very "interesting", eh?