Game AI Roundup Week #28 2008: 13 Stories, 1 Video, 1 Quote, 1 Event

Novack on July 13, 2008

Weekends at are dedicated to rounding up smart links from the web relating to artificial intelligence and game development. This week, we got a bunch of news from diferent directions; many AI community insightful articles and news are summarized, for you to stay informed and enjoy!. Remember, there’s also lots of great content to be found in the forums here! (All you have to do is introduce yourself.) Also don’t forget the Twitter account for random thoughts!

This post is brought to you by Marcos Novacovsky (aka “Novack”). If you have any news or tips for next week, be sure to email them in to editors at Remember there’s a mini-blog over at (RSS) with game AI news from the web as it happens.

AI, Game AI and apparent intelligences

In his site, Hrafn Thorri Thórisson wrote an excellent article about the place of the videogame AI within the wider context of the general AI development and research. The aspects and limitations that make game AI what it is, described from an AI specialist perpective.

The game designers tailor the environment and its limits to make sure that the intelligent processes can handle them, and vice versa. It’s the lifelike gatekeeper who doesn’t need to know how to find his way home because he has no home. It’s the terrorist that can pull a trigger but couldn’t count his fingers if you took his shotgun and held it to his head (and you can’t).

Spore’s Strategist

Interviewed by Chris Remo and Brandon Sheffield for Gamasutra about his work on Spore, Soren Johnson made some interesting comments about his background on strategy AI, in relation with his current project.

I was wondering if it would be difficult to come from a more hardcore mindset and then figure out what the game needs, to be casual.

“Sure. I knew that coming in, but I put some assumptions behind us about what it would mean to make a game. Unlike a Civ game, I’m not looking for the AI to wipe you out or to win. In fact, there’s a lot of things that the AI just doesn’t do, that it could do.

For example, we have a super-weapon system, which is something that came along maybe six months ago, that gives you some high-level stuff like nukes and EMPs. You can heal your units or give them a building, or whatever, and there’s all these high-level things you can do. And we just kind of decided flat-out that these shouldn’t be available to the AI. These are just human-only weapons that make the game cooler and more interesting and add some variety, but we didn’t want the player to be in a situation where, “Whoa, what happened? My city got nuked. I didn’t even see that.”

That’s totally acceptable for your classic RTS audience, but making the game competitive was… for this audience, we’re not interested in the AI wiping you out. There are difficulty levels, which I pushed for late in the project, because I know there will still be some gamers who want a challenge for this type of a game, so we do ramp up certain parts of the game and make the AI more aggressive and harder to make them happy, but [the AI] still has certain things unavailable to them.”


The Australasian Conference on Interactive Entertainment 2008 edition, is about to reach the deadline for the call for papers (July 18th).

The Australasian Conference on Interactive Entertainment , in its fifth year, is a cross-disciplinary conference that brings together researchers from artificial intelligence, audio, cognitive science, cultural studies, drama, HCI, interactive media, media studies, psychology, computer graphics, as well as researchers from other disciplines working on new interactive entertainment specific technologies or providing critical analysis of games and interactive environments.

The conference will accept innovative submissions that present new ideas, improvements to existing techniques or provide new ways of examining, designing and using interactive entertainment technologies. All full paper and short paper submissions will be peer-reviewed by an international program committee.

Indie Journal moving

The GameDev’s journal Digital Playground is moving to a new site, with a new name: There are two posts already on the site, both game AI related.

Behavior / AI Rambling

Instead what’s caught my eye is behavior trees. In particular it seems to solve a problem that I’ve been having - how to write specific modules of behavior (like a specific enemy attack) in a way that they can be reused and rearranged rather than having an explicit “next” behavior. I’m not sure I entirely understand how it’s all going to fit together with some of the higher level gameplay interactions, but it’s a promising direction.

Yet More On Behavior Trees

In a way I’m actually feeling a little disappointed - the current game design doesn’t call for massivly complex AI (you are fighting zombies after all) but now i have the urge to switch to something with fewer enemies with a much richer set of behaviors. But that will have to wait, and the current game will give me a chance to walk before I run anyway.

I’ve calculated your chance of survival…

The Omnivangelist posted the third article of a series “General Video Game Strategy” with some interesting concepts on human factors when cames to game AI.

I know that when I have a 2-0 advantage (or even a 1-0 advantage) the opponent will be at a loss, and the above theory suggests that the player will tend toward more risk taking in his/her gameplay in order to recover the match. I can therefore take advantage of this risk-taking behavior to my advantage.

This type of psychology won’t work against an AI opponent. In fact, you are usually at a disadvantage psychologically in all video games. At least be aware of this and put yourself in check when you notice yourself falling victim to this theory. Of course, in the end humans beat AI in video games because AI cannot adapt; but in terms of emotions computers have an advantage - they always correctly weight the odds and make the optimal decision.

This time is really beyond the Call of Duty

Some gruesome details on the Call of Duty: World at War’s japanese AI, and the different options for the player to… end it. I dont know what is worst, the actual details, or the journalist justification for an slaughter simulation. C’mon, war is war, but to what point we really need (or want) to recreate some terrible facts on our games?

The Japanese army’s new AI also forced a lot of changes. Sometimes, they’ll emerge in camouflage from foliage only a few feet away from you, with no previous signs they were there. Other times, they’ll bury themselves in pits called “spider holes” or under shallow beds of dirt or sand from which they can suddenly spring up. You’ll even see Japanese troops pretending to be dead, purely so they can spring up and fire at you once you’ve passed by. The numbers on the Japanese side are dense, and they’re absolutely willing to fight to the last man and throw sheer weight of numbers at impossible tactical situations. It’s the sort of game where you’re going to start worrying about your side having manufactured enough ammo to have a shot at taking down all of the enemies arrayed against you.

If this sounds like an exceptionally cruel way to fight, even by video game standards, bear in mind that the AI for the Japanese soldiers you’re facing would be just as ruthless. In the demo session, you could see Japanese soldiers in melee range using their swords to decapitate soldiers. The developers described a tactic that involved Japanese snipers hiding in trees and intentionally wounding a member of your combat team, so when others came out to help them, they could all be picked off with headshots. Players will be expected to adapt to situations on the fly, and also to using the most relentlessly efficient tactics possible.

Dynamic music in the year 3000

Matt Sayre in The Game Composer’s Blog, wrote an interesting article where he mentions some almost unexplored side of the game AI: music direction!

It could happen, I suppose. Artificial intelligence is always going to get better. It’s possible now to track certain game states, like how close enemies are and how much health you have, and play appropriate music. We can even have computers play only certain layers of music or mix and match layers depending on various game states. Is it only a matter of time before the computer is composing and orchestrating on the fly? The enemy just saw you on his radar, but he’s 300 yards away. Cue the woodwinds playing a danger chord progression. He’s approaching and now you see him. Cue the brass and percussion to take over the chord progression. He shouts at you to go away. Cue the melody in the cellos and horns.

Video: Using AI to Improve Animation

Torsten Reil gave a TED Talk in 2003 called Simulating Humans, describing his research on animating human characters in 3D.

Reil discusses how traditional video game animation is static and repetitive: animators create each character animation, and the character goes through the same motions every time — every time the character falls down it’s exactly the same animation of falling down. As a result, human characters in games have looked pretty unrealistic, and movie animation has taken a long time to create, as each motion has to be created by hand.

So Reil designed artificial intelligence to simulate a human body’s actual biological behaviors — building a nervous system with muscles and bones — which can simulate typical human actions like walking, falling down, avoiding obstacles, and so on. Although the technology is now five years old, it’s just finding its way into computer games like Grand Theft Auto 4.

FEAR real sequel

4player Podcast posted an article about Project Origin, the “real” sequel to F.E.A.R. releasing later this year on Xbox360, PS3, and PC. What else? ah, yes, giant praises for the game’s AI.

FEAR, even with all these problems brought me some of the most intense and challenging firefights I have ever experienced in a video game. The enemies were smart, fast, dangerous, and knew how to navigate their environment. Now I have seen games do enemy AI pretty well. Enemies that flush you out with grenades, take cover behind objects, or flank your position are what is considered “good AI” in most games. But this game is on a whole different level. The enemy soldiers in FEAR will go through the offensive motions just mentioned, but will do so very quickly. You rarely see soldiers just standing around as they will quickly move around the environment, even jumping over objects to get to better positions. Often times I would find myself ducking behind cover to reload, regain my barrings, and prepare for assault—only to find my target nowhere to be found in my futile attempt to rush him. Whether he was moving in on my flank, looking for his allies to set up an ambush, or simply didn’t like the car he was hiding behind, it was this unpredictability that kept me on my toes. The enemies in FEAR, quite simply, seemed as if controlled by human players.

Funny Quote of the Week

This week was saturated by news about the AI who beated humans in a poker table (its one of our today’s stories). But in the blog dæmonology I found this unexpected short paragraphs:

“I’m surprised it’s taken this long to write an AI that can pwn the Poker table. In fact, I suspect there are older implementations running on dedicated servers hosted in the Turks And Caicos Islands that can give Polaris a beatdown that will leave marks for years.

Still, it’s nice to know that the news is going out now. Playing a game of Poker intermediated by a telecommunication system is one of the dumbest things a human can do these days.”

Adaptive C&C

The blog The Sanatorium, comments some interesting details about a moded AI of the last Command and Conquer. Again community work, can enourmously improve the studio’s original release.

Been playing a lot of Skirmish games with the GDI and NOD in recent days, (Not much online yet) and I’ve noticed some rather peculiar things concerning the AI’s play style. Peculiar in the sense that the Adaptive AI is seemingly shining through more so then I expected in regards to the new features.

Fallout 3: more AI insights

Kotaku posted an article detailing some parts of the Bethesda answers to a fan-driven 25 question post. In it they touch on everything from child killing and drug addiction to AI and dialog trees. If you dont mind, I will be just copying the AI related parts…

How advanced will the AI of NPC’s be this time around? Are they really going to have a life? Speaking to other NPC’s in a logical manner, traveling and trading with/in faraway places, Submitting to the player rather than fighting if they know, or think, they’re no match for him?

I wish I could answer with a number, like “it will be 17 advanced.” AI is difficult to define, the NPCs certainly appear much smarter than our previous stuff, by a lot. Much of that is us giving them better data, massaging what they do so the player gets to see more of it. We added a lot of animations, so people in town are doing more. They “seem” to be interacting with the world in a more realistic manner, but that usually means going up to something and playing an animation. It can be something really simple, like we added “lean against wall”. It’s funny how something that small can give life to a person. They walk into a space, and just lean against the wall, arms folded. Like Oblivion, we use our Radiant AI system, so most of the NPCs eat, sleep, work, etc. I think we take it for granted now, but it’s pretty great to have that level of control. We’ve also done a lot to the conversation system, which makes them seem a lot smarter, but again, that’s better data, not a new system.

On the technical side we spent most of our time doing an all new pathfinding system. Morrowind/Oblivion use nodes for pathing and Fallout uses a navmesh. This is the difference between an NPC having a valid point to stand on (node) versus an area to stand in, or walk around (mesh). You can do much more sophisticated actor movement and behavior with a navmesh, and I think you’ll see the results onscreen, especially when the bullets start flying. The actors do a great job of finding cover and using the space well, something we could never have done with pathnodes.

In terms of the NPCs traveling around, many travel around town, and some travel the wasteland. There are a few caravans in the game that go from town to town trading. Radiant AI handles something like that really well.

Lastly, as far as submitting to a more powerful foe, yes they do that, in that they run away. If they’re overmatched, they holster their weapon, flee and try to hide. While this sounds cool on paper, it’s often not fun at all, and we’ve ended up really dialing that back, because it gets really annoying really fast, to have people run away all the time. The main faction that still acts like this are the Raiders, the others don’t do it so much.

Journalism Weekley Rants on AI

This week, as we have so many news to give you, I’ve decided to make a collage story, putting together any rants I found from the specialized media, on the videogames AI.

PSP Haven: review on Brothers in Arms, D-Day

The game also has a few AI problems, too. Most of the enemies in D-Day seem content to get shot in the face and not do too much about it. It’s strange to say a game with poor AI can actually be difficult, but that’s the case here. But that’s beside the point. It would have been nice to see enemies react appropriately to adverse situations like they did in the PC games, but that really doesn’t happen all that often.

IGN: SPOGS Racing Review

Many arcade racing games include rubber-banding AI that helps the racers at the back of the pack and handicaps the racers at the front. In SPOGS, this AI is way too transparent. Each race plays out, literally, like a rubber band, with the player constantly taking turns at the front and back of the line. Your speed may max out at 120mph, but as soon as you enter last place you’ll receive a speed boost that sends your speedometer to 140mph and you’ll be leading the race in no time. Both single- and multiplayer works this way. It’s a ridiculous way to race and takes the action out of the player’s hands. Steven Rees on Code of Honor: The French Foreign Legion

Annoyingly the effect that the scenery has on gameplay is rather unpredictable. The AI seems entirely unaware of its existence, so the enemies will gleefully fire in your direction regardless of whether they could really see you, and even if it means firing point blank into a wall.

Unfortunately, it can be hard to tell which parts of the scenery are more solid than others, so there will be times when you’re shot through a ‘solid’ brick wall, and others when a few twigs are enough to stave off a hail of Kalashnikov rounds (like that bit in Commando where bullets are deflected away from Arnold by rose bushes – 80s action Ed).

Diehard GameFAN: Alone in the Dark (Wii)

Not that you ever have to worry about enemies. Not only are they easy to kill, but for the most part they don’t even bother trying to come after you. I’ve had times where I’ve run right into an enemy and stood there waiting for it to attack, and it never did. If they do attack, they’re really slow about it and if you run a couple of feet, they’ll probably just leave you alone. I’d make a complaint about enemy AI, except there’s is no enemy AI present to complain about. I almost felt bad killing them in the game. It was like attacking comatose patients, and even then I think Terry Schaivo would put up a better fight then these Humanz.

Defining Dialogue Systems

Another story from Gamasutra, this time an article written by Brent Ellison, where, in words from the editor, Ellison looks at the universe and history of player-NPC dialogue interaction in games.

In games using non-branching dialogue, the simplest form of interaction, the player walks up to an NPC and initiates conversation. The NPC delivers his or her lines and the conversation ends. Alternately, initiating a conversation with an NPC triggers a cutscene where the player’s avatar and the NPC have a non-interactive dialogue.

AI Beats Human Poker Champions

One of the week’s most commented stories, Polaris finally won the match. As always, this kind of competitions unleash waves of divergent opinions.

“There are two really big changes in Polaris over last year,” said professor Michael Bowling, who supervised graduate students who programmed Polaris. “First of all, our poker model is much expanded over last year—its much harder for humans to exploit weaknesses. And secondly, we have added an element of learning, where Polaris identifies which common poker stratagy a human is using and switches its own strategy to counter. This complicated the human players ability to compare notes, since Polaris chose a different strategy to use against each of the humans it played,” Bowling said.

Response to AI Developer Interview Questions - Part 1

The blog Artificial Intelligence and You (AIandU) posted an article in answer the one posted by Paul Tozour on ai-blog. The game AI community on the move!

In this series of posts, I will be attempting to answer questions posted by PaulT at The large majority of the techniques referred to in these questions are new to me, but this seems like a good way to learn. Much of the material I used was from, which is an invaluable resource for anyone trying to learn the latest and greatest in game AI.

Stay tuned next week for more smart links from around the web!

Discussion 0 Comments

If you'd like to add a comment or question on this page, simply log-in to the site. You can create an account from the sign-up page if necessary... It takes less than a minute!