Voxel shmoxel, for me it's the AI that is mind blowing in this project.
Will the AI lead to global things like rises and falls of entire civilizations, wars, social and political evolution?
Edit: from your KS page "Actions and reactions are played out to a limited depth" - I gather the answer is no..? Why not? How about in the future?
That is a really great question, I think I may have hinted at this in one of my bulleted feature lists.
The AI makes decisions at a macroscopic level in addition to a personal level. Basically, whoever is in charge of the province (it could be you) issues commands to the citizens and soldiers. You can declare laws, make alliances, and declare wars.
Again, this is way more simple than it might sound. Most of this is occurring at a pretty high level of abstraction - think of it like a 4x or strategy game (
http://www.europauniversalis3.com/ for example).
So, how might this work? It builds a table of alliances and enemies (not a yes/no table but a number from 0.0 (totally hostile) to 0.5 (neutral) to 1.0 (totally friendly / allied) - this gets deeper than political ties, and accounts for things like racial biases and so forth. Each table is weighted in terms of importance, then added up, and averaged, ultimately producing a final number of 0.0 to 1.0 with the same meaning as above. The AI acts on its motivations and deterrents to maximize its score for each individual - everyone tries to help their friends and destroy their enemies, while more importantly fulfilling any personal goals and staying alive. Sometimes personal goals conflict with political goals - i.e. you have a love interest that is in the town of a rival faction. Conflicts like this ultimately are what makes the story interesting.
Ultimately, its going to be an interesting problem and may be limited in scope to a few hundred NPCs in a world, depending on how compute-intensive everything gets. As always, there are tricks to get around everything, like working with a macro-scale AI and an individual (local AI) separately.
There won't really be any form of evolution I can think (aside from survival of the fittest, which is not explicitly defined but just a side effect of a non-static world). There is no learned behavior (i.e. use of a neural network).
Don't confuse depth of compute with ability to plan though. Characters will (and must) be able to plan. For example, if an artifact is very far away and hard to get, NPCs must be able to devise a solution to get it (assuming they have motivation to). Compute depth is more about predicting what move you will make, based on what move I make, based on what move you will make 2 turns later, etc etc.
