The Future Human: An Interview with Tim Sweeney
December 28, 2012 Page 1 of 3
As technology evolves, how do we interact with it? An iPhone is far more powerful than your average desktop was just a few years ago. And it has a camera in it. And it's got your schedule, and your contact list. And it has games. And, occasionally, it's even a phone.
We rely on our technology to an incredible degree. Have you ever been caught without your smartphone in a foreign country, or an unfamiliar town? It's almost existentially terrifying to realize how little we can do without our contemporary tech; that's why we have it around us all the time.
What does this convergence mean, not just for the future of games, but the future of technology in general, and the way humans interact with it?
Epic Games' Tim Sweeney knows a lot about tech, there's no denying it. In addition to being one of the paragons of game code, he has also read extensively about the limits of human perception, and the nature of technological convergence.
In this extensive interview, we speak with him about the possibility of a graphics plateau, the promise of Unreal Engine 4, and what might happen if technology were all around you, all the time -- even more integrated into your life than it is now.
Back in 2008 I wrote about something that I think you would totally disagree with -- I felt we were starting to see diminishing returns on graphics, just in terms of whether people really cared about new particle effects or lighting, when "good enough" seems to work for many, in games like League of Legends that aim for compatibility over poly-pushing.
I know that you're very interested graphical advances from a code perspective, and from a what-you-can-achieve perspective, but do you think that also pays dividends for the audience?
Tim Sweeney: Yes. We're still at the point where improvements in graphics technology are enabling major improvements in gameplay. Just the ability to do real-time lighting on environments now means you can construct a completely dynamic environment -- or destruct a completely dynamic environment -- and have all of the lighting respond accurately. It turns out that the technical features you need for that are really elaborate and expensive.
If you have your own support for real-time lighting like Doom 3 had, then all of your areas that are directly hit by light are bright, and all of the areas that aren't directly hit by light are completely black. So you need real-time indirect lighting, which means calculating two bounces of light on them, and so on, which really is only becoming possible now with today's GPUs, that are 2.3 to 2-and-a-half teraflops.
Even if your thesis is that we're getting diminishing returns on graphical effects, I think we're still at the point where making graphics innovations greatly improves our capability of implementing new kinds of games.
I saw the Unreal 4 demo, and it's very good-looking. But the thing that I found interesting, at least framed by what you're talking about, is that in a certain respect the closer you get to reality, the less impressive it is in a way, because as it gets closer to a "real thing," I know what real things look like, and we get into an uncanny valley situation. Obviously, we do this in a more fantastical setting, but that move toward reality to me almost seems like it's going to, at a certain point, start being less impressive.
TS: We don't necessarily want to simulate reality because reality is pretty boring, right? (laughs) Simulate realistic characters in a game, and they're probably just sitting around sending their friends stupid messages on Twitter. You want fantastical environments and fantastical characters, and that's really the big job of an engine -- it's not just to enable graphical realism but also to give our artists and designers the capability to really tweak things to create a custom look and feel for the game, and a custom enhanced version of reality that they can play around with consistently.
You're trying to solve a lot of new problems with UE4 -- indirect lighting, more efficient and dynamic particle effects, and that sort of thing. But what about some of the legacy problems that are still not totally solved, like shadows that are jaggy everywhere, and dynamic texture loading so that it doesn't have a pop effect; these sorts of things?
TS: Well, each generation, we improve. We greatly reduce the flaws that you see, but we're still far from having enough hardware performance to completely eliminate them. The jaggies in shadows in Unreal Engine 1 were 3 feet wide, and now they're just a few inches wide. And that's great, but until they're much smaller than a millimeter you'll still notice those artifacts. Really, the amount of performance you need to solve this completely is immense. I think we're just slowly moving in the right direction there.
The technology is solving other problems. For example, texture streaming has been a huge challenge given optical media. When you're playing Gears of War off a DVD, sometimes you see textures popping in just because we can only move the DVD head four or five times a second in order to load the textures in. If textures are coming into view at a faster rate, then you're screwed. If you look at what's possible now with solid state disk technology and flash memory storage, you have a factor of 10,000 less latency.
It's pretty significant.
TS: Oh, yeah! It's gigantic! It's able to greatly, greatly reduce some of those flaws. Every generation we're improving a lot of things, but we're still a long way from being able to simulate reality. For a long time, the Holy Grail was completely destructible environments; that means you basically have to build your game levels using architectural tools and engineering analysis so that, when the right amount of force is applied to your wall, it breaks. Then your level designers aren't just creative folks; they're structural engineers. There are significant barriers to a lot of advancements in those areas.
The thing I find funny about completely destructible environments is that any game could just become a flat plane at a certain point if you just blow everything up.
TS: (Laughs) You want to be able to completely destroy the world?
Yes -- some sort of antihero complex, probably, or maybe I just like playing Earth Defense Force.
Page 1 of 3