elcome to Beyond the Bits. This is the first in an occasional series that aims to give credit to some of the unsung facets of gaming development. By speaking to a variety of experts and practicing professionals in the fields of discussion, we hope to shed light in some of the underappreciated aspects of gaming. The first chapter of this series will focus on the brains of what you play. Artificial Intelligence.

Artificial Intelligence, or A.I., is one of the most important yet least understood aspects of modern game development. Depending on the type of game, developers may have entirely different goals when it comes to telling the computer how to play the game. Some say mimicking human intelligence is the goal. Others take a more pragmatic approach and say simply having the computer avoid doing too many stupid things is a solid benchmark.

While it’s not necessarily easy to define what makes “good” A.I., the effects of less-than-optimal A.I. are hard to ignore. Any gamer worth his salt has A.I. horror stories. Whether it’s a fighting game that can be beaten by using the same move repeatedly, a puzzle game with a computer-controlled opponent that outright cheats or a driving game with ridiculously fast rivals, weak A.I. can ruin an otherwise good game experience.

A.I. MISCONCEPTIONS

After talking to a number of developers, programmers and game producers, one thing became clear—when it comes to gaming A.I., misconceptions abound. There’s a lot going on under the hoods of our favorite games—though it’s sometimes less than we imagine.

“I do feel that some games are credited with having A.I. that’s just unbelievable and stuff like this, and I just look at it and go, ‘What the hell are they talking about?’ ” says Ed Boon, co-creator of the Mortal Kombat series. “I think a lot of people, in their minds they formulate that there’s a lot more thought going on in computer A.I. than is really happening in the game.”

“You can make games with no personality at all, and when they’re around the human driver, the human automatically puts emotion on the A.I.,” adds Dan Greenawalt, game director for Forza Motorsport 2. “Oh, that guy was trying to take me out so his buddy could pass me!” You know, I’ve worked on a lot of these games, and I can categorically say that’s not what happened. Many things may have happened, but that was not any of them.”

Another point of contention is that gaming A.I. necessarily has to be smart, a stereotype that’s surrounded the concept since its inception. After all, doesn’t having a conversation with a computer game seem like a goal?

“It is a common misconception that ‘bad A.I.’ is synonymous with stupid A.I.,” say Anthony Yu, lead gameplay programmer on Ratchet & Clank Future: Tools of Destruction. “Bad A.I. is not, in fact, when things in a game are stupid. Nobody complains about the simplicity of the animal AI in Katamari Damacy. Bad A.I. is when there is a disconnect between the actions the enemies perform and the capabilities of the player to affect them and react to them.”

One thing’s for sure: As games become more sophisticated, players have come to expect a similar degree of sophistication from their electronic opponents. While it once was fairly acceptable to see less-than-stellar A.I. behavior—trying to walk through walls, performing the same attacks repeatedly—modern audiences are less forgiving with those deficiencies.

EARLY A.I.

Obviously, much has changed since the early days of game development. “Over the last 20 years A.I. has gone from being a trivial few lines of code to a major part of games,” says Peter Molyneux, founder of Lionhead Studios. Molyneux is no stranger to the field, having worked on such A.I.-heavy projects as Populous, Theme Park and Black & White.

“I think, initially, A.I. stuff was basically, I would describe it as simulating A.I.—kind of faking it,” says Boon. “You would program in kind of scripted sequences and randomly pick between them, and that kind of created the illusion of thought or something, as opposed to a random-number generator. Then, as the years have passed, certainly over the past 10 years, things have become more like an actual algorithm that scans the conditions of the game.”

Others say that, while the technology may have changed, the fundamentals of A.I. programming have remained relatively stable.

“Actually, you would be surprised. It hasn’t really changed that much at all. The way we write A.I.—maybe this is just because I’m old school—but the way we wrote A.I. back in the 1980s was not that different from the way I was writing A.I. for our last real-time strategy game Warlords Battlecry III,” says Steve Fawkner, Puzzle Quest’s designer and A.I. programmer. “Games that we’ve worked on traditionally used two sorts of A.I.. Once upon a time it was just a whole set of positive rules—if this, then that. If this, then that. A set of identifying patterns that would occur in a game and having a simple set of rules—and the simpler the better, to be honest, especially when we were programming 640k DOS machines or 48k Sinclair Spectrums. The simpler the better, the set of rules we could come up with. Over the years, as memory increased, we could get more and more complex with the situations we could look for, the patterns we could match. And when things took a turn to the more powerful machines back in the mid to late ‘90s, we found we could start actually scripting exact situations up, and that’s what’s really changed.”