Simulated Architecture #1: Welcome to The City
The catalyst for this project is the question: "Can computers create art?" My gut feeling is that as impressive as computers are, they cannot create "true art", the kind of art that arrests your very thoughts. They can compute, they can infer, but they cannot make that extra leap. The driving question then is: "How far can they go?"
The end goal, the hold-in-your-hands desired result is explained here on my website. It's really a guiding light in this project. Originally it was the whole project, but this question of computer-driven art outgrew the original plan. I want to be able to create a 3d city on my home computer by providing a little bit of input and letting the computer do the rest of the work. What do I mean by this? Apart from all the programming and thinking required to create the program to do this, once I have the program, I want to give it general guidelines on what I want in a city (for example: size, features, maybe topography) and it churns away. It eventually produces data files so that I can walk through the city in real-time as if it were the artist-created cities in GTA. I'm willing to allow compromises in detail for performance reasons.
I want to document how I'm thinking about this project. I think it's an interesting idea bringing in many different disciplines. I want to learn about architecture and town planning. I want to learn about psychogeography. I want to learn about graphics algorithms. I want to learn clever ways to organize data and program things so that I can view a city on my home computer and not give it a heart attack. I also want to approach this in a very idiosyncratic way, that is, let my goal drive the learning. What might be relevant to real town planning might be irrelevant to my goal. Also by examining this stuff from a generative viewpoint ("how do I make it?") rather than a reductive viewpoint ("how does City X do this?") I might have an interesting take on cities. I don't really want to learn these things in the traditional way. I want to think about them and toy with the ideas myself. There's nothing at stake here. I don't want a job in town planning. I don't want to solve the whole AI thing. If I get a certain distance and figure it's all too hard, then at least I've gotten somewhere and learnt something about the world around me. Hopefully the journey will be interesting for others, which is why I'm documenting it here. Feel free to chime in at any time.
The first step is to see what people have already done. As I've mentioned, the Grand Theft Auto series have featured sizeable cities that you can drive through in real time. Pascal Mueller has done some impressive work on the procedural generation of cities. On that topic, there's a world of information on procedural modelling (which I'll need if I want to generate buildings without resorting to copy-pasting premade templates). Will Wright and the Spore team, and the demoscene crowd have looked into procedural modelling, texturing and animation of elements for real-time use. I have no doubt someone's created tools for town planners to demonstrate cities, but I haven't looked into it. So really, all the elements are out there in the world: we can display city environments in real time, we can plan city environments and we can create all the 3d elements we need for the city. The trick is to stitch it together.
Well that's enough for now. Next time I might talk about psychogeography in GTA, the topic of perceived distance and how it relates to data streaming strategies.
The end goal, the hold-in-your-hands desired result is explained here on my website. It's really a guiding light in this project. Originally it was the whole project, but this question of computer-driven art outgrew the original plan. I want to be able to create a 3d city on my home computer by providing a little bit of input and letting the computer do the rest of the work. What do I mean by this? Apart from all the programming and thinking required to create the program to do this, once I have the program, I want to give it general guidelines on what I want in a city (for example: size, features, maybe topography) and it churns away. It eventually produces data files so that I can walk through the city in real-time as if it were the artist-created cities in GTA. I'm willing to allow compromises in detail for performance reasons.
I want to document how I'm thinking about this project. I think it's an interesting idea bringing in many different disciplines. I want to learn about architecture and town planning. I want to learn about psychogeography. I want to learn about graphics algorithms. I want to learn clever ways to organize data and program things so that I can view a city on my home computer and not give it a heart attack. I also want to approach this in a very idiosyncratic way, that is, let my goal drive the learning. What might be relevant to real town planning might be irrelevant to my goal. Also by examining this stuff from a generative viewpoint ("how do I make it?") rather than a reductive viewpoint ("how does City X do this?") I might have an interesting take on cities. I don't really want to learn these things in the traditional way. I want to think about them and toy with the ideas myself. There's nothing at stake here. I don't want a job in town planning. I don't want to solve the whole AI thing. If I get a certain distance and figure it's all too hard, then at least I've gotten somewhere and learnt something about the world around me. Hopefully the journey will be interesting for others, which is why I'm documenting it here. Feel free to chime in at any time.
The first step is to see what people have already done. As I've mentioned, the Grand Theft Auto series have featured sizeable cities that you can drive through in real time. Pascal Mueller has done some impressive work on the procedural generation of cities. On that topic, there's a world of information on procedural modelling (which I'll need if I want to generate buildings without resorting to copy-pasting premade templates). Will Wright and the Spore team, and the demoscene crowd have looked into procedural modelling, texturing and animation of elements for real-time use. I have no doubt someone's created tools for town planners to demonstrate cities, but I haven't looked into it. So really, all the elements are out there in the world: we can display city environments in real time, we can plan city environments and we can create all the 3d elements we need for the city. The trick is to stitch it together.
Well that's enough for now. Next time I might talk about psychogeography in GTA, the topic of perceived distance and how it relates to data streaming strategies.