I think that could be a good fit for groups who play TTRPGs mostly for the mechanics or the interactions between player characters. It keeps the action flowing and makes it easy to adapt to the players’ wishes.
The cost is that it opens up the risk for plot holes and makes the setting less dense. With every blank that you fill in, you need to make sure that it doesn’t contradict things that have already been established and that it won’t cause problems further down the line. If you’ve already designed a cool mountain fortress for your BBEG and suddenly it needs to be deep in a swampy forest, you’ll probably have to redesign it from the ground up or else it will feel very out of place. Foreshadowing also gets a lot harder and basically every plot line boils down to “somehow, nobody ever noticed this threat/opportunity until the point the heroes showed up”. That’s absolutely fine for a more spontaneous format but if you want a long campaign that leads to an epic climax, it usually feels a lot cooler if the players can look back and see that they had been working towards that goal since day one even if they hadn’t known back then. That’s a lot harder when you as a GM don’t know enough of the details to hint at them.
Of course, at a smaller scale, this is something that most GMs do. If the plot requires the party to talk to a certain NPC to get vital information and the party absolutely refuses to interact with them, the relevant information must come from somewhere else. Maybe a different NPC, maybe they find a note, maybe something entirely different. But at least for the way I structure my campaigns, that’s more of an exception to fix problems rather than something I leave vague on purpose.
As for other prep methods, I’m working on a short essay on how I design crime/detective modules. I’ll link it here once it’s done.
Overall: go for it, experiment with it, see how it works for you and let us know. There is no wrong way to be a GM as long as your goal is to build something that’s fun for everyone at the table. Even if you don’t reach that goal on the first try.
Absolutely and it has done so for over a decade. Not LLMs of course, those are not suitable for the job but there are lots of specialized AI models for medical applications.
My day job is software development for ophthalmology (eye medicine) and people are developing models that can, for example, detect cataracts in an OCT scan long before they become a problem. Grading those by hand is usually pretty hard.