Below is the text of a New York Times Arts article from June 7, 2005 - the original article is now in the New York Times pay archive.


Redefining the Power of the Gamer

By SETH SCHIESEL
Published: June 7, 2005

MARINA DEL REY, Calif., June 3 - Standing outside the apartment on Thursday, Walter could hear the barbs and retorts of a failed marriage's final throes.

Walter's friends, Grace and Trip, had invited him over. Now, though only every third word seeped through the door, Walter could hardly mistake the bickering.

At Walter's knock the voices stopped. The couple adopted brittle masks of happiness. But as their banter moved from Trip's new bartender set to recent Italian vacations to Grace's latest apartment makeover, the couple gradually returned to the needling exchanges of domestic strife.

As Grace and Trip retreated to opposite sides of the living room, sniping about old grievances, Walter appealed to the couple's loyalties, trying valiantly to reconcile his friends.

This is the future of video games. In their modern riff on "Who's Afraid of Virginia Woolf?" Walter was the only human. Grace and Trip were virtual characters powered by advanced artificial intelligence techniques, which allowed them to change their emotional state in fairly complicated ways in response to the conversational English being typed in by the human player.

It was one version of the future here this past week at the first Artificial Intelligence and Interactive Digital Entertainment conference. It is a future where games are driven as strongly by characters as combat, where games are as much soap opera as shooting gallery and as much free-form construction set as destruction arena. The apartment drama, a 15-minute interactive story called "Facade" that is scheduled to be released free next month (interactivestory.net), was one of the demonstrations offered to the roughly 120 game makers and academic computer experts who attended.

"As we try to create more immersive experiences, these artificial intelligence techniques are helping drive games forward and this is one of the areas that could really explode," Bing Gordon, chief creative officer at Electronic Arts, the No. 1 video game company, said after his talk Wednesday night. "We hope that the folks here start thinking about artificial intelligence as a feature, like graphics is a feature or sound is a feature."

While the adaptability and behavioral subtlety in recent classics like "Black & White," "Sid Meier's Alpha Centauri" and "The Sims" have impressed gamers with their seeming-intelligence, those titles have been but an early step.

"For a long time, games have been judged largely on their graphics," said Ian Lane Davis, a conference organizer and chief executive of Mad Doc Software, which recently created the well-received Empire Earth II, a real-time strategy game. "The graphics hardware is now getting powerful enough that basically everything looks good now. So what is starting to differentiate games is what is happening inside the characters, how the opponents behave and make plans, how comprehensively and realistically the worlds respond to what the players want to do."

"At the same time," he added, "players are demanding a lot more freedom. Often they don't want to be put on a roller coaster track that just takes them along one path, no matter how entertaining that one path may be. They want a range of choices and they want those choices to matter in creating the overall experience. You put together all of these demands, and that's why you're seeing all of this attention now on artificial intelligence in games."

Outside the game world, the term artificial intelligence is used to label technologies as disparate as air traffic control systems and automated vacuum cleaners. At the conference, much of the discussion was about specific game activities that, to a human, would seem more intuitive than rational, like using conversational language.

But one of the broadest and most powerful approaches to artificial intelligence may be one that does not focus on determining specific behaviors. ("Does the computer general know that it should use tanks and artillery together?")

Rather, it is a move to structure programs so that they absorb available information and then generate their own strategies to achieve sometimes-contradictory goals ("protect the hostages" versus "kill the enemy," for instance).

Traditionally, game programmers have created activity through explicit if-then statements: if the player attacks the castle, then send pikemen to defend it; if the player corners the market on wheat, then invest in corn. That process is known as scripting. But what should the computer do if the player takes an action that is not in a script?

"The problem now is that the worlds are so complex and the variety of potential actions so vast that trying to direct the environments and the behaviors of computer-controlled agents through traditional scripting can become unmanageable," Jeff Orkin, an artificial intelligence programmer at Monolith Productions, said between sessions.

Three years ago, Mr. Orkin worked on Monolith's campy "No One Lives Forever 2," set in the 1960's. Now he is working on "F.E.A.R.," a game scheduled for later this year.

"We used to manually lay out all of the steps that an agent would take: do this, then do that, and if this other thing happens then try this," Mr. Orkin said. "Now we tell the agent: here are your goals, here are your basic tools, you figure out how to accomplish it."

"For example, let's say you the player are running down a hall and an enemy is pursuing you," Mr. Orkin said. "You get to a door and slam it behind you. The enemy replans and tries to kick it in, but if you hold it closed with your weight he will replan again and maybe come around and dive through a window. In the past, the programmer would have had to explicitly code each of these steps. Now, you put the character in the building and it figures out a plan on its own."

As put by Chris Crawford, a legendary game designer of the 1980's who now focuses on interactive storytelling technology: "As a game designer you are an absolute god. One kind of god says, 'O.K., now this leaf will fall a little bit here, and then this wind will blow a bit over there.' The other kind of god says, 'Here are the laws of physics. Go for it.' "

That conceptual leap from designer-as-determinist to designer-as-prime mover is what has made both the "Grand Theft Auto" and "The Sims" series so popular. The challenge is that even as gamers have come to expect more freedom in their virtual environments, they have also come to expect more explicitly directed cinematic moments, like the D-Day invasion scenario in "Medal of Honor," where players can feel as if they are living a movie.

"There is a real tension between wanting to handcraft the experience to generate a specific emotional response and wanting to allow a more open-ended environment so the player feels they are in control," said Doug Church, one of the designers behind the highly regarded "Thief" and "System Shock" series. "Artificial intelligence will help us bridge the two."

But perhaps that bridge will run in unexpected directions. Until now, artificial intelligence has often involved making computers accessible to humans. With his new project, "Spore," Will Wright of "The Sims" fame means to invert that concept.

"Until now, artificial intelligence has usually meant that the human creates or perceives a model of how the computer makes decisions," Mr. Wright said. "But what if the computer is instead analyzing the player, and the program is customizing the experience based on the internal model it has created of the human?"

"Spore" is meant to tailor a species' entire evolutionary experience - from amoebalike gene pattern to intergalactic emperor - to each user's individual play style. In that sense, future generations of games may process humans just as intensively as humans are playing the software. But not to worry, Mr. Church said: "We have a long way to go before we get there."