Game AI Roundup Week #40 2008: 9 Stories, 1 Video, 1 Quote

Novack on October 5, 2008

Weekends at are dedicated to rounding up smart links from the web relating to artificial intelligence and game development. This week, 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.) Also don’t forget the Twitter account for random thoughts!

This post is brought to you mostly 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.

Seven Deadly Sins for Strategy Games

On the blog “Designer Notes” (the Soren Johnson’s Game Design Journal), Soren recently posted a list of usual “sins” that designers make on the strategy genre. The first one, is specially interest for us:

Strategy games have a direct lineage from board games, and the fun of playing the latter comes from understanding the rules and mechanics of the game world and then making decisions that have consequence within that world. Computerized strategy games allow a single player to experience this same world on his or her own. At some point, however, strategy developers began to create lengthy, scripted scenarios as the single-player portion of their games. (In fact, the recent World in Conflict shipped without a single-player skirmish mode altogether.) These scenarios have a peculiar feeling - they use some of the same rules as the core game while often violating others. The AI takes action depending not on its own development rate or strategic priorities but on whether the human has hit certain triggers. In many scenarios, in fact, the human cannot even lose because - when defeat approaches - the script will freeze the AI and starting pumping in free units for the player. Further, these scenarios are often built around specific “objectives” to achieve, such as destroying a specific structure or capturing a single point. This artificial environment takes decision-making away from the player. Not only is there only one path to victory, but the player’s performance along that path may not even matter. Games without interesting decisions get boring quickly. Fortunately, some recent strategy games, such as Sins of a Solar Empire and Armageddon Empires, have returned to open-world, random-map gameplay - without pre-set objectives or artificial triggers - and are reminding us of the joy of cohesive and consistent strategy games.

Old complaints

On the Bush Mackel’s personal blog, he put a list of games he is currently playing, with some info and opinions on the games. Some appellant issues came up from his comments on the title Project Gotham Racing 3.

“One of the things I liked about the first PGR was that as you got better cars, it showed because your better car performed better than your competitors in those beginning races. In THIS game it seems so far that if you’re racing a Ferarri, all of a sudden everyone else will be driving one too. Which in opinion is no good. This same kind of “match of power” I hate in other games like Oblivion. What’s the point of leveling up if everyone else levels up with you? (just my opinion)”

How To Hire Good Game Designers

Phil O’Connor (Codemasters principal designer) outlined 10 ways to spot a “real game designer” during the resume and interview process, in an article for Gamasutra. This are interesting lines regarding the comprehension on AI that the applicant designer must have:

if a designer talks about how the AI will be able to react to the player’s actions, they should be able to detail exactly how that will work: will it be based on how many “bad behavior” points the player has accumulated, will it depend on triggers set in the dialogue system that will play specific responses, will it be based on a proximity system that the AI checks when the player is within range, assessing the player’s reputation points, shown weapons, clothing, etc.

If a candidate cannot describe probabilities, mathematics, or outline game systems supporting a feature, then they probably are not the real deal.

A WWII conflict that doesn’t hold your interest posted a review on the Gearbox title Brothers in Arms: Hell’s Highway, and briefly comments on the game’s AI.

Artificial intelligence on both ends is mostly effective. Squadmates follow orders consistently. Enemies not only stay entrenched, but will retreat to safer positions if they catch you flanking. They’re not very aggressive, however. Rarely do they attempt to flank you or force you to fall back. You’re usually hunting them down and flushing them out.

Quote of the Week

Interviewed by CVG Valve’s Doug Lombardi made some outstanding declarations regarding graphics vs others aspects of the games. A very interesting interview, here is one of the bests parts, as the quote of the week:

“It’s more about what you can do in the game. Graphics have started to top-out now. We’ve got really great-looking games but what we want are more intelligent, more visceral games and the multi-core processors are going to be the way that we get there on PC.”

University Selects NaturalMotion for Research Project

Trinity College Dublin’s graphics, vision and visualization group to utilize morpheme’s animation system to simulate large-scale crowds in a realistic ‘Virtual Dublin’.

NaturalMotion, the animation technology company behind the revolutionary euphoria engine, today announced that Trinity College Dublin has selected morpheme to power character animation within its Metropolis research project. Metropolis, under development by the university’s Graphics, Vision and Visualization Group (GV2), part of the School of Computer Science and Statistics, is a realistic, computer-generated ‘Virtual Dublin,’ with functioning, large-scale simulated crowds.

Gamer rant of the week

In a post entitled “Swan song for Snake“, the university newspaper Student, posted an interesting article with an unusual criticism on a game considered a model to define “State of the Art Game”.

There is a constant juxtaposition of cutting edge technology with dated game mechanics. Metal Gear Solid 4 puts heavy emphasis on how your actions mould the battlefield. Except… they don’t. Soldiers will spawn in the middle of doorways constantly until you run out of ammunition or patience; blow up a helicopter with a rocket launcher and another will rise up to take its place. With the mind-boggling stupidity of the game’s artificial ‘intelligence’ and inconsistent approaches to level design like invisible walls and invincible enemies, Metal Gear Solid 4 feels like a 1998 game stuffed into a 2008 graphics engine.

AI archaeology

In a Royce Armstrong article (briefly reviewing the videogames history), a surprising data came up… is the game AI development that old?

“People have been trying to play games on computers almost since the days of the very first computer. As early as 1950, Claude Shannon, a mathematician and engineer, believed that computers could be programmed to play chess in competition with humans. He became intrigued with the concept of artificial intelligence. In pursuit of this idea researchers and scientists designed crude games that could be played on the huge and clumsy computers of the 1950s and 1960s.”

Video Games in the Future

In an post taglined “A Gamer’s Plea” the blog ITGeek commented on some features expected for the games (near?) future:

Another cool idea we’d like to see erupt within the gaming industry is the ability to talk to the characters inside a game. Some games allow players to textually speak to game characters already, but we’d like to see this pushed a little further. We’d like to be able to orally interact with characters: ask questions, joke around, warn and speak to them as if we were speaking to another human being. And we’d like to hear these characters talk back! It’s the ultimate artificial intelligence opportunity and although it would probably be years before this technology would be available on a wide scale, we’re sure it would be a hit.

Emote gets £600,000 for AI project

On other news, Develop brought us a story about the british Online games publisher funding an AI development related project.

Emote Games has been given £600,000 in funding to support a new artificial intelligence project.

The cash comes from the Technology Strategy Board and contributes to the project’s overall value of £1.3m.

Emote develops social games and online technology and recently named its partnership with Avalanche - a new online hunting title.

The AI project looks to improve the quality of AI characters in social networks and online games. Emote will partner with Imperial College London for the research, starting in January 2009.

Video: openFrameworks

openFrameworks is a C++ library for creative coding.

We are currently in pre-release, and heading towards a public release. If you want to participate in the pre-release, or be notified when we release the library, please join the mailing list.

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

Discussion 3 Comments

RobinB on October 6th, 2008

For me as a researcher, it's good to see that university-industry collaboration gets supported by the government, particular AI related research. The use of games has been a bit frowned upon by "classical" (dare I say old-school?) researchers, but luckily that stigma is fading, especially as universities notice that the games industry is a serious and influencial business. There are still lots of discrepancies in how researchers think about what game developers do and what game developers say academics should research, which became quite obvious in the kick-off meeting of the AI and Games Research Network here at our university. In that light, such grants for emote and others (I'm actually funded by a similar grant, but the corresponding games company didn't issue a press release yet, so I'm not sure how public they want to keep it) are headed in the right direction. Although, as a former browser-MMORPG developer, I have my doubts about the morality of the research content of the emote grant. And now a little rant about speech understanding :) I highly doubt that we'll see working speech understanding in the foreseeable future. As it is an [URL=""]AI-complete[/URL] problem, it requires extensive world knowledge to be able to correctly parse and understand utterances. Even in a seemingly closed and well-defined game environment, this feat is not easy: Through (verbal or textual) communication, players implicitly assume full (outside) world knowledge, by using analogies, word-plays or jokes. In my eyes, getting a database of some sort of this world knowledge is the hard part, and reasoning is relatively easy once you can access such a database. The [URL=""]Singularitarian school of thought[/URL] believes that such strong AI will actually happen within a couple of decades. Once that happens, we might have true speech understanding in games, but also real AI opponents. If Skynet doesn't decide to destroy humanity first, that is. :D

William on October 6th, 2008

Suggestion: I don't feel the roundups do justice to a game and the game's AI by quoting a single review, especially when the review doesn't come from a "reliable and experienced" reviewer (think Eurogamer, EDGE, perhaps gamespot). If possible, I'd appreciate the "metareview" approach taken by Joystiq where positive and negative comments are taken from multiple reviews, providing a more complete and more balanced view. See [url][/url] for an example (Assassin's Creed meta-review). (In this case, the USA Today review does seem in line with other reviews for this game, but that hasn't always been the case for the Roundup quotes).

alexjc on October 6th, 2008

William, It's not a bad idea! Though I'm not sure if you were quoting Edge as a good or bad example of journalism... hehe :-) They've recently been lower quality than anything else. I'll talk to Marcos about the format of the "reviews." Kotaku's Frankenreviews come to mind too. However, few reviewers even mention the AI in many cases. Alex

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!