Magmarock: HOWEVER! nothing will ever beat a hand crafted world for atmosphere and depth. One example of this is System Shock 2 where everything feels deliberate in design. Another prime example is Witcher 3. Everything in that game is so delectly crafted like a sculpture some artist worked tirelessly getting all the details just right.
Actually, a lot of elements in a digital work will be assisted by various constraints, pre-made shapes, and so on. To the point where arguably the rules within the digital world will decide what should be there more than the artist.
Imagine a slope. It's a hill, it has various elements, it should be ran down with a snowboard. You should get a particular speed if possible, you should be able to jump somewhere, it should be possible to arrive at the particular trick-sites through various routes, the weighting graphs should ensure that you can reach a specific number of them, etc. etc. In the same way, you obviously know the constraints of the controls for the runner, because they are programmed to be within a set of angles.
If you knew those rules on beforehand, you might get only a few possibilities that are viable, out of all the random options you might have had. So depending on how well the simulation works, or how much time you spent on it, it's not necessarily an artist that placed the elements in the slope. Instead they were the ones who decided, within the viable configurations, which one should be polished up.
In the same way, Witcher 3 has a lot of hand-made transitions spread out in the world, and they are all very well made, I'm not going to say anything else. The hill with the windmill over the farmland (where you fight the Griffin) is very obviously hand-made in the sense that this is designed by someone who imagines a landscape, draws it, and then have other artists fill in elements to make the hill look real.
But a lot of the forests and marshland are at most restricted by a heighmap someone loosely decided on from an overhead map. Where the objects in that space has been populated by an algorithm. If you ever travelled down in the right hand corner of the map, where nothing really happens, you can see probably the baseline for the algorithm, where all the trees suddenly seem meted out and placed based on location, height (loosely speaking there's a tree-line in W3). You can be pretty sure that a lot of forest paths are weighted in the same way - they follow an outline, but then deviate based on how steep and how curved they can be. And this is likely what they've started with when they made the rest by hand.
Same thing in a lot of less spectacular games with different procedural elements from the artist's point of view. The model of the soldier in COD wears a helmet. Another one wears a clown-hat. Both are the same model, essentially, and the hats are created towards the same constraints on beforehand. Before they are switched out based on user-preference. But the artist made each of the hats inside the constraints given by the model's head restraints, or have them linked to specific dimensions, so the hats will always fit.
So the question isn't about "hand-made vs procedural". It's about how deep you think it's viable to go when using an algorithm to decide what the game-world should look like. And how much of this can be done in real-time.
...unless we're talking about Gothic, where you can literally feel the programmer's boredom as the scripts play out. But even then you can't plan everything.
So from one point of view, it's just a question of finding out at what level you have to take over for the simulation. When an AI is actually assisting you, vs., when it stupidly takes over and drives you down a dead end. Same concept.
Anyway. So having an algorithm to assist you isn't a new concept (except to the people making Gothic). And having an algorithm decide on how to compose the world in layers, given particular rules, also isn't new. But doing it from the top-level, and then having a fixed outcome downwards based on the initial conditions - that's new.
Normally you'd have individual routines running on the different layers, no? So what's deciding the random elements in a specific place is really exactly as random as when the routine is run in a different spot. Building something from the ground up as well, that's always someone's favorite - a recursive algorithm always depending on the initial "seed".
But what if you decide on new constraints as you subdivide downwards instead? It's extremely interesting stuff that's very interesting to watch in practice. Because it means that you could for example design something on the high-level, and use an algorithm to put all the lines together on the low-level.
And better yet - do it in real-time, like in NMS :D I mean, this has to be impressive, even if you're the kind of person who likes Gothic, right?
