Wednesday, March 23, 2005


Games to focus on algorithms again?

At GDC 2005, the lecture on "The Future of Content" by game designer Will Wright attracted quite a lot of game development folks. Unfortunately I wasn't there, but a nice transcript of the session is available on Gamasutra. While Wright also presented his new game, Spore, his point in the talk is an interesting one, and might even be an indication of a probable change of direction in game development. In summary, he announced the return of the algorithm, as an alternative to ever-increasing amounts of content (i.e. data) in games. What the heck does that mean?

If you look back a few years, in the era of the C64, game development was all about getting the best out of the machine within its limitations. There was simply not enough memory, or disk space, to store lots of art and giant levels. Stuff was usually generated procedurally, to cram levels, enemy characters, animations and everything else into 64k (rather less, since you couldn't easily use all of the machine's memory). Remember "Elite"? I always wondered how the developers could get all this data for thousands of planets in memory, complete with descriptions and trading goods. The answer is: it's computed on the fly, procedurally.

With the advent of the PC, bigger hard drives, and later the CD-ROM, this changed considerably. Suddenly you could store bulk loads of art with a game. The current consensus seems to be that art is the key to any game, and the more art you put into a game, the better. The archetypical game following this philosophy is probably Myst. This in itself is not wrong, I guess: you really have the possibility now to create an amazing, cinematic experience. Artistic styles in games become more diverse and mature, and current titles deliver an intense emotional athmosphere through their art. But, the production effort of current games is breathtaking, in terms of manpower, budget, and time to completion. It's not hard to see that it won't be possible to carry this much further into the future.

Now, Wright knows all this, and he's trying to take a different route: build a game based on procedural elements instead of using tons of art assets. Which to me, not having seen Spore yet, doesn't seem that different from his previous simulation-oriented titles - maybe just more extreme. As he said, he hired a bunch of people from the demo scene to work on what became Spore. These folks strive on competitions like writing the most impressive graphics demo with the smallest possible memory footprint, and thus really know how to generate content procedurally.

According to Wright, most of Spore's content is generated procedurally. We'll have to wait for the game to appear on the market, but the approach sounds quite attractive:
Well, it sounds great, but this approach also raises a bunch of serious issues, which Wright apparently didn't talk about: building a world by placing the bricks manually is safer than letting some potentially buggy algorithm loose on it. How can you test all of the game, if the space of possibilities is limitless? Unpredictability means risk. Coding effort is much higher and more demanding. Also, you somehow have to transfer the images the game designer has in mind to the programmer, how translates them into code. Communication between artists and programmers is key here.

It's all a bit familiar to procedural animation used in visual effects production: Say you have a great system to do physics-based, animated ocean water. This system will work within its parameters and figure out the detailed animation by itself. Now you have a control problem: the director comes and looks at your beautiful, procedurally generated ocean and says: "Very nice, but I'd like to have that wave back there a little higher and more to the right". That's were things become really tricky.

I'm looking forward to seeing how Wright deals with these issues. If it's not only a marketing plot for Spore, I'd love to see how algorithms can influence game design in the approach he proposed, instead of being just an "art driver". But that's only the computer scientist speaking :)

EDIT: Just found some pictures of Will's Spore presentation. Check them out, before they have to disappear!
Comments: Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?