Agreed, but that's not really what i was getting at.
No Man's Sky's use of procedural generation was believed to deliver on that lofy ideal that you could assemble diverse and interesting environments using nothing but procedural generation, and a relatively small number of principle assets. And it, frankly, did worse than some other games that rely on Procedural Generation (hell, at least 7 Days to Die has randomised geology, rather than every Iron Oxide node looking the same). Yes, filling out a massive universe is a different thing than making a single TES province, but the basic premise behind them is the same. Even with an extensive list of potential assets, No Man's Sky shows that Procedural Generation cannot create the same kind of diversity as hand-placement. We know it's technically possible, because, hey, Evolution is a thing, but the work required and the actual generation isn't there yet.
Again, i'm not saying that Procedural Generation isn't useful. As i said last thread, Bethesda really needs to reconsider their use of it (or lack there of, with Skyrim and Fallout 4) because their world spaces NEED to grow, or they need to redefine the types of games they make. But even if they use Procedural Generation, the need to go back over things and put in hand-crafted details limits the size of what they can work with. Too much area, and your hand-crafted details are going to simply get lost in the scope of things. It doesn't matter if you scatter a thousand unique, hand crafted details through an area the size of Daggerfall, most of that effort is going to be wasted because no one will ever find them. If you use standard Procedural Generation approaches, it means scattering 1000 instances of the same 3 camps across the map. Which, frankly, gets really notable, really fast.
That's because nature has the benefit of a 'Procedural Generated' model that no computer system can even come close to. But you CAN mimic it's effects through hand-crafting, far easier than you can through the use of a Procedural Generation mechanic. A prime example of this dynamic is with AI. The AI used in games isn't even intelligence, it's just a complex formula of If-A-Than-B scenarios that trick you into thinking the the entities in the game are moving with a degree of autonomy, when they're really just reacting to pre-set parameters. It's much easier to simulate nature, than it is to recreate it.
And that's not even touching on the NPC problem.
Once again, this doesn't mean that Procedural Generation isn't useful. It is, and good use of it could save TES from being relegated to a sort of micro-open-world category that it's quickly falling into. But the fact that it still requires touching up limits the total size it's going to be able to offer.