R. Michael Young, an assistant professor of computer science at N.C. State University, is working on research that might one day make video games more enjoyable. Young, 41, is studying ways to build artificial intelligence -- the ability of computers to act like humans -- into games so that users get movielike stories. With such technology, for example, a game could adjust to a player's actions and provide a different experience every time it is played. He sat down with Connect's Jonathan B. Cox to discuss his work. The following is an edited transcript of that conversation. Q. Tell me about your research.

A. It involves taking ideas from artificial intelligence and computer science and trying to put those techniques behind the scenes ... to try and make virtual worlds more compelling or engaging than they might otherwise be. Specifically, the stuff I look at tries to take ideas from conventional AI [artificial intelligence], linguistics, cognitive psychology and ideas about narrative theory and look at computational models of narrative, so that you can take these computational tools that are well founded on the other theories from other disciplines and automatically create stories inside a virtual environment. Stories that are sort of created on the fly and customized to a person's particular interests or their background and that react to you as you interact with them.

Q. I'm not sure I completely understand.

A. In a typical, traditional computer game, designers work really hard while they're building the game in the lab to imagine all the different choices that you might make as a player and make sure that every choice that you make leads to a path that's going to be fun or entertaining or compelling. But then there's a fixed number of paths. So then basically for somebody who plays the game, they play it once and it's done. If they play it over again, it's a matter of learning the paths rather than experiencing the story. Because the designers are really good at what they do, you also come away with a storylike experience.

The idea in my research is, let's come up with a model of what storytelling is. The computer can create an initial story for you. While you're walking around in the beginning of the story it fills in new details for you to experience later on. So that no matter what path you take, you end up sort of having a great experience. But the idea is, it's not done at design time by people. It's done using this model of what storytelling is at the time that the player, the user of the computer system, is actually experiencing it.

Q. So the computer system adapts to you as opposed to you adapting to the system?

A. That's right. There is a little bit of tension there between the system trying to come up with great stories and the player trying to experience the world the way that they want. Part of interesting storytelling is not telling the participant in the story everything they need to know up front so there's some suspense. So on the one hand the user has to be a little bit cooperative in the experience of the story and the system has to be a little cooperative in allowing you to make the choices that you might want.

Q Do I have to answer a set of questions about myself before I play or does the computer learn automatically as I do whatever it is that I do?

A. There are a bunch of different approaches to that. One approach is to create a profile for yourself so that games like this would just be able to access that. The other idea would be to learn as you go. One interesting and compelling thing about these types of games is that everything you do is mediated by the computer. The computer can kind of record things as you go along.

Q. Take me out of the context of games for a minute, what other types of applications are there for this technology?

A. Entertainment is a key application but you can also look at things like education and training. ... If I wanted to create a simulation of, say, a 13th-century castle and I wanted to teach middle school students about social rules between the nobility and the middle class and the peasantry, I might place a student in a particular role in the castle and they have to observe the world around them and figure out what those social rules are. Then they have the ability to step in and say, 'Let me test out my theories of the social rules by, say, speaking to a peasant and seeing what happens.' As the social rules play out as they expect or not, they revise their own theories about what they're learning in a way that you wouldn't get to do if you were just reading a textbook.

Q. How has this manifested itself so far? Have you been able to take this out of the theoretical world and put it in a real context?

A. Yeah. We've been spending a lot of time building the underlying infrastructure for this. Now, we're getting to the point where we have the architecture in which we can begin trying out the theoretical aspects.

There is a computer game called Unreal Tournament, which is made by some people [Epic Games] here in Cary. It's one of the leading first-person shooter games. It's great because it comes with the ability to extend it. What we ended up doing was sort of taking out the control from the game and writing our own drivers that would let us connect out across the Internet to a server where our AI tools were running. It uses its AI tools basically to act as a puppeteer to pull the strings of everything that's going on in the world. Users connect to the game just like they would normally. In reality, what they're doing is they're playing in a world that's controlled by the AI code that we've got.

Q. I'm curious if you've heard from any of the game manufacturers about your work.

A. There are interesting areas of overlap between the academic community and the game development community and we've gone to a number of workshops or symposia. A lot of the specific techniques that we use and other AI researchers who are looking at games use can end up winding their way into various games.

There's a little bit of conflict between the research community and the game development community. The game development community is very targeted on getting their titles out by Christmas. If it doesn't sort of lead to that, then it's very difficult for them to justify the investment. The academic community wants to look at principles or ideas that are not necessarily going to lead to immediate improvement in the next three months, but maybe look at something that's going to be enabled five years from now. So there's a little bit of a disconnect.

Q. So how far away is something like this artificial intelligence from the marketplace?

A. I think to some degree you can see in the next five years a lot of these techniques winding up in things that are sold. You're going to see a lot more as the processing power develops. A lot of the development historically, especially in terms of processing, has been focused on graphics. Now that the graphics are getting as fast and as good as they are, they're realizing, OK, now let's focus more on how we can spend more of our processing cycles on the content. You're beginning to see these little explorations into what kind of novel interaction we can use our processing power to create.

Q. How did you get into this field?

A.I was always kind of interested in how people thought and that was really what was most compelling to me. The first work that I did was on looking at how people talked or spoke, particularly about the tasks that they were going to cooperate on. I wanted to create a computer that would be able to figure out some complicated tasks for you and tell you how to do it in a way that was natural. When I looked at a set of computer-generated instructions for a task, a lot of times it was either at a completely detailed level that was unnatural for me or it left out steps and I couldn't figure it out. But I'd look at a Reader's Digest do-it-yourself manual and it would be very, very helpful because it was written by a human. I wanted to try and get a computer to be able to do that.

Q. Growing up, were you always interested in robots and that type of thing?

A. Actually, growing up I was really involved in theater. I was involved in theater through undergrad and then started focusing on artificial intelligence. Given my interest in performing arts and theater and film, it was sort of obvious to me that there was an overlap in these disciplines.

Q. When did the computer aspect come into the picture?

A. I was a computer science major in undergrad.

Q. I want to talk a little more broadly about artificial intelligence. We've heard for years about robots that could one day be like the Jetsons. Where are we?

A.There's a range of disciplines that go into building a robot. Everything from the mechanical engineers that have to build the grippers to the computational linguists who have to design the interfaces. If you think of a robot as a manifestation of intelligence, a robot ... that walks around and that you can talk to like a person and can perform tasks in the physical world, we're quite a bit farther than we were 50 years ago. And we have quite a bit farther to go. There are robots now that can be built in a humanoid shape with arms and legs and a torso that can walk up stairs and walk down hallways and navigate.

Q. Where do you see the future? How will artificial intelligence affect our daily lives?

A.I see the way that artificial intelligence affects our lives the way information technology has changed the way people live over the last, say, 10 years. There's now a computer in most components in your car. There are computers that run conventional computer programs that can be put onto credit cards, onto your wristwatch, in your toaster, in your microwave. Incrementally we have seen the growth, the outreach of information technology, of networking. Artificial intelligence ... I think will also see sort of a sliding increase in the pervasiveness. People will see it first in the interfaces to computer products. They'll be able to speak to the toaster. They will pick up their Palm Pilot and it'll say something to them that wasn't obviously preprogrammed.

	tools
	printer friendly version		send this story to a friend		subscribe to our e-mail lists		get news, weather, movies on your PDA