Go to the Globe and Mail homepage

Jump to main navigationJump to main content



Gaming's next great breakthrough? AI that doesn't act like AI Add to ...

Pong, the most successful video game of the 1970s, featured a two-dimensional "tennis" ball that players batted across the screen with simplistic computer-generated paddles.

Today, video games have become far more sophisticated: Computers can adapt to their human opponent's patterns, and games can precisely match the physical movements of an at-home player to an avatar on screen.

But it may surprise some that the branch of computer science that pioneered simplistic games such as Pong is the same one creating the most cutting-edge games today.

The force behind it all is artificial intelligence, which is built on algorithms, or complex lists of instructions. They enable machines to react to the playing decisions made by humans. Not only does this give the impression that gaming machines are thinking, but algorithms can even allow computer-based opponents to "learn" as games progress, resulting in increased difficulty that continues to challenge the human user.

Today, even bigger breakthroughs are fuelling unprecedented growth in the gaming industry, with many experts saying we have only caught a glimpse of what lies ahead.

"It's phenomenal," said Nick Graham, director of Engineering Interactive Systems at Queen's University in Kingston, Ont., a laboratory dedicated to video game research and development. "Gaming is a leisure activity that has really captured the public's attention."

It's big business, too. "The computer game industry is now rivalling the movie industry for entertainment dollars," said Bob Mercer, professor and chair of the computer science department at the University of Western Ontario.

The Entertainment Software Association, a group representing the multi-billion dollar video game industry, estimates today's average player is 34. To keep those who grew up on Pac-Man and Super Mario Bros. interested in games into their adult years, the games have had to become more sophisticated and challenging.

Developments in AI have been central to that task, particularly as other game elements, such as graphics, have reached their apex. Complicated algorithms are used to create opponents that can outwit human players and offer a sense of competition.

As the industry has mushroomed in the past few decades, a growing number of gaming companies and academic research centres have worked to develop increasingly complicated algorithms that are taking games to the next level.

Popular games such as Starcraft or World of Warcraft allow players to compete against computerized opponents that appear to have sophisticated battle strategies. While they are popular, one of the criticisms of such games is that the computer opponent doesn't change tactics or otherwise respond to decisions the human player makes.

Prof. Graham said experts are working on this kind of technology.

Now, one of the main obstacles is determining how to prevent artificial intelligence applications from making those games too difficult for human players to beat.

"[The]challenge to be faced is that games will become too hard," Prof. Graham said. "Part of the fun [of playing video games]could well be the fact the game is predictable."

New gaming systems, such as the Nintendo Wii, or add-ons to existing systems, such as Kinect for XBox, were developed with artificial intelligence of a different flavour - motion-sensing technology to allow users to let their physical movements guide play.

Kinect, which set a Guinness World Record for the fastest-selling consumer electronics device when it was released late last year, can recognize faces, gestures and voices. This allows users to play without having to operate a handheld controller.

Players using Kinect may believe their physical movements guide play on screen, but that's only partly true. The gaming system was developed using a branch of artificial intelligence known as "machine learning," which uses algorithms to help computers adapt after being exposed to many examples of the same thing.

To create Kinect, Microsoft recruited people of different physical builds and studied their movements, in order to make algorithms that would help machines track and predict how a certain individual might move.

Still, despite these achievements, experts predict the future of gaming will increasingly depend on further breakthroughs, such as computer characters that can interact with human players realistically or respond to unpredictable, random actions by an at-home user.

That's something that, given the sheer complexity of creating "intelligent" machines, is far from guaranteed.

Many challenges in artificial intelligence stem from differences in the types of intelligence used. It is easy to program a computer to play chess, a logic-based game, but it has proven extremely challenging to instruct a computer how to do tasks that children take for granted, such as seeing the world around them or understanding words spoken to them, said Yoshua Bengio, a leading AI expert at the University of Montreal.

"A lot of what we know isn't something we can verbalize," he said.

Although the obstacles that lie ahead will likely prove difficult to overcome, the drive to create games that can incorporate newer, better and more complex AI applications shows no signs of slowing down.

"I would say just about every university and every college has a game development program in the country," Prof. Mercer said. "There's a tremendous amount of interest."

Report Typo/Error

Follow on Twitter: @carlyweeks


Next story




Most popular videos »

More from The Globe and Mail

Most popular