Originally Posted by The Economist
When looks are no longer enough
Jun 8th 2006
From The Economist print edition
Gaming: As graphics improve, artificial intelligence is becoming an ever more important part of designing video games
GOOD looks, the video-games industry is discovering, will get you only so far. The graphics on a modern game may far outstrip the pixellated blobs of the 1980s, but there is more to a good game than eye candy. Photo-realistic graphics make the lack of authenticity of other aspects of gameplay more apparent. It is not enough for game characters to look better—their behaviour must also be more sophisticated, say researchers working at the interface between gaming and artificial intelligence (AI).
Today's games may look better, but the gameplay is “basically the same†as it was a few years ago, says Michael Mateas, the founder of the Experimental Game Lab at the Georgia Institute of Technology. AI, he suggests, offers an “untapped frontier†of new possibilities. “We are topping out on the graphics, so what's going to be the next thing that improves gameplay?†asks John Laird, director of the AI lab at the University of Michigan. Improved AI is a big part of the answer, he says. Those in the industry agree. The high-definition graphics possible on next-generation games consoles, such as Microsoft's Xbox 360, are raising expectations across the board, says Neil Young of Electronic Arts, the world's biggest games publisher. “You have to have high-resolution models, which requires high-resolution animation,†he says, “so now I expect high-resolution behaviour.â€
Representatives from industry and academia will converge in Marina del Rey, California, later this month for the second annual Artificial Intelligence and Interactive Digital Entertainment (AIIDE) conference. The aim, says Dr Laird, who will chair the event, is to increase the traffic of people and ideas between the two spheres. “Games have been very important to AI through the years,†he notes. Alan Turing, one of the pioneers of computing in the 1940s, wrote a simple chess-playing program before there were any computers to run it on; he also proposed the Turing test, a question-and-answer game that is a yardstick for machine intelligence. Even so, AI research and video games existed in separate worlds until recently. The AI techniques used in games were very simplistic from an academic perspective, says Dr Mateas, while AI researchers were, in turn, clueless about modern games. But, he says, “both sides are learning, and are now much closer.â€
Consider, for example, the software that controls an enemy in a first-person shooter (FPS)—a game in which the player views the world along the barrel of a gun. The behaviour of enemies used to be pre-scripted: wait until the player is nearby, pop up from behind a box, fire weapon, and then roll and hide behind another box, for example. But some games now use far more advanced “planning systems†imported from academia. “Instead of scripts and hand-coded behaviour, the AI monsters in an FPS can reason from first principles,†says Dr Mateas. They can, for example, work out whether the player can see them or not, seek out cover when injured, and so on. “Rather than just moving between predefined spots, the characters in a war game can dynamically shift, depending on what's happening,†says Fiona Sperry of Electronic Arts.
If the industry is borrowing ideas from academia, the opposite is also true. Commercial games such as “Unreal Tournamentâ€, which can be easily modified or scripted, are being adopted as research tools in universities, says Dr Laird. Such tools provide flexible environments for experiments, and also mean that students end up with transferable skills.
But the greatest potential lies in combining research with game development, argues Dr Mateas. “Only by wrestling with real content are the technical problems revealed, and only by wrestling with technology does it give you insight into what new kinds of content are possible,†he says.
Hence “Façadeâ€, a game created by Dr Mateas with Andrew Stern, a researcher at the Institute for Creative Technologies at the University of Southern California. “It's an example of where I hope to see computer games go in five years,†says Dr Laird. The game is set in the apartment of Grace and Trip, a couple whose marriage is in trouble; by conversing with them in plain English, the player can influence them and help to determine whether they stay together. “Façade†is not so much a game as an interactive drama. “We wanted to build a game built around richly expressive characters,†says Dr Mateas. “We were interested in taking games into content areas that they haven't really explored before—we would like to create a new genre.â€
Tellingly, Doug Church of Electronic Arts, who gave the keynote speech at last year's AIIDE conference, recently started work on a game with Steven Spielberg where “the focus is on building an emotive relationship at a story level and a gameplay level between the player and another character,†says Mr Young. Researchers and games developers are, it seems, converging on the same territory. “Industry people are being exposed to more complex techniques, and academics are learning that game AI is a unique, new and interesting problem,†says Dr Mateas. “Games are an amazing place to do fundamental AI research.â€
