I have learnt about modular level design but I see that if you take this idea one step further you can have extraordinary results.
In this case you can see how skyrim is based on 3D tiles for interior spaces.
Here you can see how can an open world can be dynamically generated with very few data.
Here you can see a complete procedural game.
What is your opinion? Has the procedural content generation future or is it a flaw?
You’re asking the wrong question.
The question is, do procedural worlds have their place? The answer is yes, if it fits the requirements.
You could also let a computer write a book with an algorithm.
If you want to tell the day in the life of a housewife it might be quite a good read.
If you want to tell a thriller or a mystery story you’ll have to design it to meet the crowds expectations.
That all said, the sad thing is the “designed” levels of the latest games are more of the housewife story kind anyways 
Anyways, procedural generation gets better and better and more and more people look into shape grammar to generate structures and such. We’ll get there. And up until then, you can still make a rapid prototype of your world procedurally and then shape it to your liking.
Agree with arexma on the book comparison.
The more original, creative care that goes into a game, the better it will be in my eyes. Sometimes, bigger isn’t better: I played the Deus Ex demo maybe a dozen times and each time I found something new. I’m finding it very hard to complete Oblivion because it’s so repetitive and boring. Thus I dislike procedural generation in favor of variation and originality.
However, when they make a massive world that still has the incredible variation in gameplay and, more importantly, amazing story that Deus Ex had, I’ll take back everything I say about procedural generation.
Fortunately, most end users aren’t that concerned about the mechanism behind the experience. The trend of using the latest, as opposed to the proven has it’s own followers, the author of this post, included at times.
I wasn’t saying that the mechanism is important at all. I don’t care how they make a game, as long as it’s fun. For me, non-repetitive is required for “fun,” because repetitive is boring. And all the PCG games I’ve seen are repetitive. I want a better end result, not more end results (which, with PGC, end up being the same result cloned multiple times and then colored differently).
So I only value “proven” as far as the method of creating. A “proven” end result has already been done before, and is therefore boring repetition.
Just to clarify where I stand on it.
I have mixed emotions about this. I suppose that with some simple rules and some fractal math a nice natural landscape could be generated. I guess it could add replay value to a game. The issue that I have with it is that it doesn’t leave much room for the artistic level design that exists in RPGs and single player games. In designing a level through artistry one can create a space that invokes the mood that the designer wishes the player to have while navigating that space. Trying to produce such an effect through procedural methods would be dicey at best at this point. I don’t think it would work for something like Half Life 3 or Doom 4 where the environment not only has visual relevance but also needs triggers. It’s important to be able to tell a story with some games and part of that is embedded in the environment. I wouldn’t say that idea is a bad one at all though. It might even bring about a new genre which would be great.
Using the right math, you could place artist generated content in locations that are based on gameflow,
I was thinking a randomly generated city, fora multi player online survival game would be nice, and making new items by mixing found items using blueprints, with destructable buildings…
I guess that would work if it stamped it in a similar manor that megatextures are created but there’s still the sensors and actuators like the triggers for sounds, effects, door open and close…etc. For a single player game, it would seem pretty complicated to generate the settings procedurally. It would free up a lot of man hours if it could be done though. I’m not sure how much of that would translate to hours writing the code to implement it. Games are running into several million lines of code but content is by far the bulk of the size of the game. Maybe it could knock a few gigs off of that too. Another thing to consider would be that instead of just loading the levels into memory it would be generating them first. That could add quite a bit to the loading times. That could be pretty annoying in a single player game. It might take a lot of head scratching to come up with all of the pros n cons.
I would have to agree that it probably has a lot more pros for multi-player.
Generally, when you want to use procedural design for terrains for example, you will need to have an algorithm that is more complex than a simple perlin noise or fractal algorithm.
Basically, the number of textures and filters you would need to lay on top of each other for massive and large-scale variation in a terrain would mean that you would need a powerful machine and a highly optimized algorithm to create it in an acceptable time frame (and this assuming that you’re not constantly generating new terrain).
Procedural cities are a bit harder (due to their artificial nature which can’t be represented very well with fractals), to this day there has been research on building generation techniques which includes L-systems and shape grammars (the latter being a long set of rules that dictate how shapes can be combined into a new object).
At one point, Blender had the ability to create Vue-like terrains with variance in the features and of a wide scale (which could perhaps be used in the BGE), but today it cannot be done in the viewport due to the fact that you can no longer use the texture node trees with the displacement modifier (due to a crashing issue).
Now procedural generation may or may not be appropriate depending on the game, simply put, the main uses would be to allow for more rapid development of content or the creation of endless gameplay mechanics (ie. like how you have an unlimited number of worlds to explore in some of the recent sandbox building/RPG games like Minecraft).