Game AI Roundup Week #27 2008: 9 Stories, 1 Video, 2 Quotes

Novack on July 7, 2008

Weekends at are dedicated to rounding up smart links from the web relating to artificial intelligence and game development. As always, there are some good articles and blog posts for you to read. Remember, there’s also lots of great content to be found in the forums here! (All you have to do is introduce yourself.)

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

Unwelcome Advice

Intel’s blog has a post about the innevitable many-core future, which provoked some strong responses among the developer community. It also ties in to another statement by Intel that AI simply requires better processing power to compete with MMO (see this discussion / rant in the forums). Here is a quote from post by Anwar Ghuloum, Intel’s Principal Engineer and author of the original blog article:

Ultimately, the advice I’ll offer is that these developers should start thinking about tens, hundreds, and thousands of cores now in their algorithmic development and deployment pipeline. This starts at a pretty early stage of development; usually, the basic logic of the application should be influenced because it drives the asymptotic parallelism behaviors. Consider a common pattern of optimization we’ve seen in single core tuning: the use of locally adaptive algorithms to heuristically reduce the computation time. By definition, this introduces dependences in the computation that are beneficial in the single core case but limit parallelism for multi-core. Similar choices are made about libraries and programming languages that optimize for single core performance (or even small-way parallelism), but sacrifice long-term scalability.

Later on, Peter Bright from Ars Technica, published an article entitled “Intel: an expensive many-core future is ahead of us” where he states:

Intel has bad news for software developers. It’s been hinted at already, but now the company has stated explicitly: it’s not enough for software developers to be targeting dual, quad, or eight cores. No, the future holds tens, hundreds, or thousands of cores, and developers are going to have to bite the bullet and write programs that will scale to such systems.

Lego Indiana Jones

Although Matthew Forbes from says the latest Indy game is a “great game for fans”, dont throw too many praises to its AI.

Of course, this game is more enjoyable if you saw the movies first. It’s also more enjoyable with a friend. A friend controlling the other character is also easier, since the A.I. (Artificial Intelligence) is questionable at best. Sometimes the character will stop following you for no reason, sometimes the character will refuse to fight, etc… This game was never intended as an adventure to be enjoyed alone.

Making the Enemy See

gaming your way posted an article telling the insights of the creation of the initial AI for the current game under development.

The current game (let’s call it CC for the sake of it) is getting close to the point where I would declare the main game engine done, most of the events are processed now and the final enemies are going to be done today (hurray!). As the title suggests (wow it’s something post related) I want to write about the dumb ass AI one, nope rather 4 of the enemies in CC use, if you were so bold to call it AI.

AI Portfolio

Browning the web, I found the Ryan Panuski’s Projects page, the part of his portfolio-site, dedicated to demo works. Interesting enough, there are some AI materials to look at (which we mentioned a few months back when he announced the project)..

“Pacman Type Program: This program is currently a work in progress. It is a much more advanced program than my previous Mousey program and will be written in C++. This program is using some of the libraries from my Mousey program such as console output. However it is going to be directed by a Neural Network brain with weights and inputs that utilize genetic algorithms. The AI will exemplify emergent behavior as the Pacman type character will learn and evolve to both eat all the “food” as well as evade the Ghosts. The Ghosts will use emergent behavior in the form group tactics to hunt the Pacman type character.”

Fallout 3 AI

Bethesda Softworks used a funny approach on the promotion of the FallOut AI features: “Vault-Tec Industries” the fictional enterprise responsible for the post-apocalytic “Capital Wasteland”, says this about the game’s AI:

Mind-Blowing Artificial Intelligence - At Vault-Tec, we realize that the key to reviving civilization after a global nuclear war is people. Our best minds pooled their efforts to produce an advanced version of Radiant AI, America’s First Choice in Human Interaction Simulation. Facial expressions, gestures, unique dialog, and lifelike behavior are brought together with stunning results by the latest in Vault-Tec technology.

Maybe the termonuclear war of the year 2077 has something to be with the Radiant AI beeing the “America’s First Choice in Human Interaction Simulation”?

Solving Games

Phil Bowermaster posted in the blog The Speculist an article about machines playing games, and depending on the game, if that mean something in the path to artificial intelligence.

So by this reasoning a computer like Deep Blue — or software such as Deep Fritz, which can now pretty much run on any computer and which has a nice record of beating or tying any human chess player it has taken on — isn’t really more intelligent than the person it beats at chess. Some contend that such an argument involves moving the goalposts on what we mean by “intelligence.” So we have a situation where a computer beating a human being at chess is a good indicator of intelligence until it happens. At that point, the ability to win at chess not longer indicates intelligence at all, but something else.

So a true test of intelligence would be something else. Some have said that if a machine can beat a person at poker, then we’re dealing with an intelligent machine. Interestingly, one of the difference might be that poker is not — as far as I understand the concept — a solvable game the way chess is. It would seem that there is too much randomness and too much psychology involved.

Spore in the Newspapers

Michael Parsons from Times OnLine, wrote a review on the EA recent title Spore, where he briefly remarks some AI qualities, and how they integrate into a more complex context.

“I tried it with my wife, who I normally has to be dragged kicking and screaming to sit in front of a video game, and also roped in our four-year-old son, who likes monsters. Both were hooked in seconds. The deceptive simplicity of the interface masks cunning artificial intelligence that takes whatever you create and animates it with humour and style. For example, there’s something very beautiful about the way the animations of the spinal column work: you feel like a Pilates instructor or an Alexander teacher when you watch how pulling on one bone in the spine realigns the creature’s posture. I can’t think of a more engaging anatomy lesson for a small child.”

Minesweeper hacking posted an article showing some minesweeper insights, after some community hacking and debate. The fact that minesweeper cheats to help the player, and nobody noticed for so long, is a testimonial that game AI can (and arguably should) do the same. Here’s a quote from the article itself:

“During this process, I found that Minesweeper will sometimes assist you and move bombs away from where you are clicking on.”

Intelligent Murlocs

The blog posted an iteresting article analizing how AI improvements are not always good for the gameplay; the example: the hatred WoW murlocs! (Those things cant fight like a man, they are always crying and running for help!)

Many people hate murlocs in World of Warcraft. But when you consider the reasons for it, you’ll see that murlocs are hated because they are slightly less dumb than other mobs: many of them use ranged combat, and thus can’t be pulled away that easily, and when they are hurt, they run for help. That makes them annoying to kill, but also shows how MMORPGs are designed around artifical stupidity. Even with the most intelligent races of them, a single adventurer can kill half their settlement in plain sight of the other half without anyone noticing or reacting. That is essential for soloable gameplay, but not very bright.

Een’s AI review

This is a quote from a user’s perspective of the Land Of The Dead - Road To Fiddler’s Green game’s AI, published on the blog Een.

Artificial Intelligence: 5.0/10 Zombies are dumb, that’s a fact of life but this is just sad. I’m standing on top of a car shooting at a zombie and it doesn’t even react. The zombies around it just stand there even though I am in plan view in front of them. Only when I actually engage the zombies back on the ground do they go into attack mode. Sometimes I even stand on the ground right in front of a zombie and it won’t attack me until I shoot it once first.

And the AI Shall Inherit the Earth

And more and more interesting comments on the GRID’s AI.

Moving along to Racedriver:GRID. Something that is definitely worth mentioning, is the AI of the opposition in a race. Gone are the days of the “catch-up” mechanism (where your car is faster than the rest of the field if you’re behind them) and CPU drivers that follow the best line and never spin out their cars. In GRID, the CPU drivers have personalities and driving styles. And they really do keep to their own style. They make driving errors resulting in a spin-out or a spectacular crash, they drive aggressively if that’s their style and you’ll have a tough time passing them. And the only way to catch up to opponents, is by driving better than them or by slip-streaming the opponent.

If these features aren’t on every new racing game developer’s “to-do” list, then there’s something fundamentally wrong with them. These features truly make the game. It gives the player the true feel of racing. The feeling that anything could happen at any moment. Just like real racing.

Then there’s your teammate you get to select once you advance far enough in the game. I had a teammate that truly sucked. He came in last in every single race except the ones that were his strong point. In which case he still only came in at 4th or 5th place. But I decided to stick with him just because I was too lazy to compare his stats with new drivers’ stats to determine if I should hire a new driver =)

To my surprise, the more races we did, the better he started doing. Up to the point where he was always right behind me and threatening to even overtake me! I don’t know if I’m imagining things, but it would seem the driver actually learns from you and the way you drive. Which to me, is truly a step up in racing games.

Video: AI.Implant Demo, from GDC 2007

Flash back to last year’s GDC, here’s a video from middleware company

This is part one of our massive destruction demo from GDC 2007. Developed in concert with our partners NetDevil and Ageia, this demo shows the in-game use of AI.implant for intelligent pathfinding and dynamic path refinement in a physics based environment.

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

Discussion 2 Comments

thaspius on July 7th, 2008

Just thought that I would point out that the last link in the post points to the wrong location [QUOTE] * [URL=",guid,9f42fb7e-ec31-4c1c-9c18-78891a517267.aspx"]AI for Massive Destruction (GDC 2007)[/URL][/QUOTE]

Ian Morrison on July 7th, 2008

In regards to the article about chess AI you posted, the AI games team here at the U of A just beat 7 top poker players with their AI, so consider that another milestone passed. :)

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!