While most of us were still setting up our stands in the main hall of ECTS, the second European edition of the Game Developers Conference was already in full swing in the meeting rooms upstairs. As you might expect, some of the topics being covered were rather esoteric or even downright boring for casual observers, but the first day of lectures kicked off with an event that anyone interested in gaming could relate to - a discussion on the future of videogame technology. Arrayed on the stage to consider this weighty subject were some of the leading lights of the British gaming industry - David Braben, Peter Molyneux, Jez San and Demis Hassabis.
Right from the start it was obvious that this would be a no-holds-barred debate, with Jez San's opening ramble about how many polygons each new generation of consoles could throw around being interrupted by David Braben asking the simple question "does that matter". Being able to display more polygons on the screen is all well and good, but "how will that effect the way your game plays"?
"Well polygons don't actually effect the gameplay at all", Jez admitted, before giving a rather worrying opinion on the current state of the games business. "They effect the way it looks, which is actually the thing that sells games. If you walk into a shop and you see a game that's ten thousand polygons a frame or ten million polygons a frame, one of them looks nicer, and you buy that one. That's how the game is sold in the store when it's on demo. You've got a ten second exposure to the game, and you go 'oh wow, I'll buy that'."
While this may be true to some extent, the arrival of each new generation of console makes this less and less relevant. As David pointed out, what actual difference will people see between the current consoles (with around 10 million polygons a second) and the next (which may handle upwards of a billion)? Jez mumbled something about blades of grass and people-looking people, before being rescued by young Demis Hassabis, who pointed out that graphical capabilities can effect the types of game you can make. "If you look at the Sims, if they were just pixels and not people then you couldn't actually make the game, because the animation and the kind of expression and emotion that comes across in the characters is directly related to the visuals. They have to look vaguely like real people. And you can kind of extrapolate that to other subjects perhaps. You couldn't do certain types of game on the SNES [and] I think you can say the same thing going forward, it just might not be quite as big a jump."
Making Better Worlds
"In the end, I think what all these polygons allow us to do is to make worlds that are more and more detailed, which is for our industry a huge problem", according to Peter Molyneux. "The number of people that we need to support the hardware that's being made is frightening nowadays. We're not modelling big square surfaces anymore, we're modelling dirt and dust particles, shadows and the way light's cast. And that is a huge problem, because if we don't model it someone else will and, as Jez says, their version looks far better than our version. And if we do model it, the fear is that we spend 99% of our time modelling dust particles in the air, and 1% of our time actually working on the gameplay. And that has always been a problem. We've been talking about graphics vs gameplay since the mid 80's."
Picking up on this during the question and answer session at the end, I asked the panel if we might be approaching the point at which it was no longer economically viable to take full advantage of the hardware's capabilities. "In some ways that's true, and in some ways it's definitely not true", was Jez's answer. "With increased power there are opportunities for generated content and emergent behaviour, so it doesn't necessarily increase cost. That said, historically costs have gone up every time there's a new console, because consumer expectations are higher every year. And because costs are increasing it is inevitable that eventually there will be less small developers and more big ones."
At this point Peter Molyneux suggested that there could still be a role for small specialist groups working on specific areas of a game, such as AI, leading Jez to offer the kind of system used in movie production as a possible way forward. "If we move to the film model, where games are created as projects and people are hired with skills that are required for that project, then that would be very interesting". But as Peter said, "it's going to be a painful road", particularly for smaller companies. "I think the days of a small independent company saying 'we're going to make the greatest game and it's going to have the best visuals, the best AI and everything' are over."
Being able to render these beautiful 3D worlds (given enough money and artists) is one thing, but David Braben seemed to think that the panel was missing an important factor. "One of my worries is that once you have characters actually speaking, you realise just how shallow the thing that's driving these characters is, in the quality of the interaction, and my belief is that the graphics are now much less important than the ability to do clever characters, characters with whom you can interact on a level where you don't just blow them to pieces."
"I think what you're touching on with Project Ego sounds really exciting", David added, turning to Peter, who somehow restrained himself from spending the next half hour talking about how your character could grow old and bald and incontinent in Project Ego. Instead he picked on the subject of animation, saying that one of the problems with making this kind of game was that "we need an army of animators, because we're making something really beautiful, and then you have this character walking across the scene and he looks stupid because the resolution of the animation is not good enough".
It's not just animations which cause a problem though, it's also how you interact with the characters. One of David Braben's pet projects at the moment seems to be speech-based interfaces, allowing players to talk to characters naturally and have them respond in kind. "Oh gosh, you're talking about having in-game characters actually responding to your speech", Peter exclaimed when this finally dawned on him, triggering a debate between all four speakers on whether this was feasible. "The voice recognition is the easy part", Demis pointed out. "Once you've got whatever ungrammatical sentence has come in, then you're going to have to work out what the player means, and the next step is to generate some reasonable response to that. Those last two steps are much more difficult than voice recognition."
There are ways to simplify this job though, and one of Jez San's teams at Argonaut has been doing just this for their new SWAT game. "In single player mode you do actually talk with voice recognition to the AI players in your squad, and they talk back. There's a limited set of orders and it's context sensitive, so depending on what you're looking at it's only listening out for a few phrases."
Demis was inclined to believe that this was the only way it was going to work, although David still thought that there was a greater role for voice communications in games like Project Ego and Republic, "where you're actually face to face with characters and you're not necessarily blowing their brains out". The problem Demis identified was that "as soon as you give the ability to characters in the game to make up their own sentences they're going to start breaking your carefully set out storylines".
"The good news is that we can constrain the scenarios in any way we want, to feel around the boundaries of what's possible", was Jez's answer to this. "In our SWAT game there's only a certain number of responses for every thing that you could be doing, so that's a very targeted, constrained game. As the technology improves we can start lifting those constraints, making the world more open and making the responses and potential for communication wider and wider, until eventually you end up where you don't have to set the story in advance, you just drop players in the world and let them discover the story."
Heady stuff indeed, but could we be getting too hung up on technology here? While skirting around the subject of online and mobile gaming earlier in the session, Demis had pointed out that simple games often work best. "If you look at the sorts of games that people play on their mobile phones, it's Snakes or something really quick like Chu Chu Rocket. And if you look at other games that are successful online at the moment, like Bejeweled or any of the MS Zone games, they're very very simple games. Maybe the sort of games we would build, like EverQuest, aren't going to reach that mass market. Maybe something that's graphically impressive isn't the right thing."
"Is the conclusion from this that online network gaming should be simple, so simple that we would be almost too scared to write that game?" Peter Molyneux mused. "I think that's a short term thing. I have no doubt in my mind that in five years' time we'll be playing some amazing console-lead online game."
Hopefully Peter and friends will be back in five years so we can find out just how far wide of the mark their predictions were...