To be fair, it isn’t aware of what source it’s using. It’s not “referencing” anything in particular. It’s just trained to replicate a bunch of data. It doesn’t understand it or anything. It doesn’t know what came from one source and what came from another, and how accurate any of those are, or what the context of it is. It just generates something that resembles it’s data based on the input.
I don’t disagree that it would be tough, but they had to start from nothing when choosing Windows originally. It all had to be learned and built up at some point. It can again, and hopefully on an open platform that won’t fuck them over in the future. (I know, there’s no chance, but there should be.)
Everyone always complains that whatever they want isn’t on Linux. Well, it wasn’t on Windows at some point either. Make a user-base for it on Linux or make it yourself. Someone did it in the past. It can be done again.