I have NOTHING against fun and beautyful games with poor AI (or even no AI), but I do think this kind of confusion somehow helps to keep AI “hidden”, not only from programmers, researchers and gamers interested In this field, but also from the gamers that don’t even know about It and still think AI Is “when the ogre runs away instead of shooting”.

Agree that while graphics get so much attention it draws commercial pressure away from AI dev. This is no single person’s fault or decision of course, it’s the zeitgeist. But as graphics get closer to realism and processing power increases, people will move on to other aspects of gameplay, and I think we can be fairly certain that AI is near the front of the line (along with innovative machine interfaces, like Nintendo showed recently). Like Tozour said, avoiding game AI is a mindset that makes less sense as the industry develops. Let’s hope the business-oriented make sense of that soon

I know of some resources on the web, but they are usually about specific genres (strategy, interactive storytelling, etc) and lack updates. Maybe some AI blog could host It?

It’s possible—but perhaps Wikipedia would be a better venue? Possibly a list page linked to from the Game AI page.

Meanwhile, if you have any links/suggestions, please post them here! I have some links of my own. If anyone Is interested, I can post them also.

Feel free to post!

Here are some of my comments about bringing “Real AI” to gaming:

1) It may not be as fun as “Game AI” In a comercial way, but It does provide other types of fun (not better, different).

2) I agree It Isn’t necessarily more processor/memory expensive. Either way, I don’t think technical issues are relevant to the general concept and the discussion about bringing better AI to games specially considering how fast the processing power increases. Besides, this can be dealt with In a variety of ways – the same way 3D games use “tricks” like smart textures, fog and variable-poly-count , AI also benefits from this kind of design philosophy.

3) It Is more expensive for the game companies, but maybe It’s a price certain consumers are willing to pay. Personally, I’d trade all the graphics In Bioshock for better AI.

4) Usually when people think about AI In Games, genres like Action, strategy and RPG come to mind. Although these kind of games deal with lots of relevant AI issues, I dont think they should be the main examples on discussions about “AI in Games” (as they clearly are nowadays). All types of games could benefit from exploring other kinds of “intelligence” (not necessarily more complex or complicated), that are not found In these genres.

Everybody who plays action games knows that, no matter how smart NPC’s are, sometimes they do VERY stupid things (like keep running towards a wall, getting “stuck” forever in a tree, not noticing you are behind him even though you are on a noisy terrain…moving fast… shooting… with a tank, etc). The reason Is that AI In these kinds of games Is applied to specific aspects of gameplay, resulting In a very specialized and limited intelligence – and this Is intentional.

In most action games players dont want soldier #92 to run slower because he Is sad for the death of soldier #23, for example – the focus is on battle/strategy, not human relationships. And that’s exactly why I think that, by using this kind of game (or any other specific genre for that matter) as a “reference for AI In games”, this simplifies a very interesting subject that deserves more attention. In AI-GAMEDEV’s list, for example, the top titles are action/strategy games, even though they represent only a very specific usage of AI. I also I noticed some people mistake “cool 3d graphics” and “fun gameplay” with “good AI”.

I have NOTHING against fun and beautyful games with poor AI (or even no AI), but I do think this kind of confusion somehow helps to keep AI “hidden”, not only from programmers, researchers and gamers interested In this field, but also from the gamers that don’t even know about It and still think AI Is “when the ogre runs away instead of shooting”.

5) I hope games get more and more independent from the “laws of industry”. Some of the greatest examples of innovation In games are In research, independent and experimental projects. Still, I know there are games with great AI out there (In many genres), and from time to time I do find great stuff (both old and new).

I think It would be a good idea to have an updated list on-line with these titles, not only for research purposes, but also to spread the word about the advances and possibilities of AI In Games to the general public. I know of some resources on the web, but they are usually about specific genres (strategy, interactive storytelling, etc) and lack updates. Maybe some AI blog could host It?

Meanwhile, if you have any links/suggestions, please post them here! I have some links of my own. If anyone Is interested, I can post them also.

That’s It! I hope I made some sense. Though I am a Computer Engineer, my research field Is on Art and Technology so I’m probably out of my league here.

Also, I’m brazilian and I dont write In English very often, so I hope my writing Is understandable

In most cases, the factors that determine the quality of a game’s AI tend to come into play long before an AI system is developed enough to measure its performance.

Hi Paul. I see what you mean — I think you misunderstood me, but to roll with it: This is when we’re speaking in terms of a minimalist-AI mindset, yes? If we were to apply the same extravagant, detail- oriented thinking to AI as is applied to graphics in modern games, then we could most certainly push the limits.

What I meant was to clearly point out that game AI involves shortcuts around the heavy processing. For example, object recognition can be implemented with machine vision techniques to simulate the processes in natural beings — but in a game you can (and probably will) simply maintain a list of object names instead. In that sense intelligence is a processing power hog, but today’s game AI dev involves finding ways around it. And like you say, if it’s done right it doesn’t need to suck power.

I’m planning to post on this subject at http://www.ai-blog.net/ in the future.

Thank you for the insight Paul. It’s fantastic to hear it directly from a professional who’s worked on major league games. I look forward to reading your entry when you post it.

There are some games, such as real-time strategy (RTS) games and massively multiplayer games (MMOs), where AI performance is a major issue due to the need to simulate massive numbers of AIs and players in real-time. In most other cases, though, when you hear developers say things like “We don’t have enough processing power for better AI,” it’s much more likely to be a disingenuous excuse. The real reasons usually have much more to do with the other factors you mention in your article.

> A game AI developer tries as hard as he can (usually at the bidding
> of a project manager) to minimize intelligence. One reason is that
> intelligent processes are massive processing-power hogs.

You’re somewhat correct that much of the industry intentionally avoids deep AI, and that there’s a strong culture of “AI avoidance” in some parts of the industry — an attitude that used to make a lot of sense, but less so as the industry matures.

Done correctly, intelligent processes don’t need to be massive processing-power hogs. We can do a lot more with the processing power we already have, even before we go anywhere near what’s generally referred to as “academic” AI techniques.

I’m planning to post on this subject at http://www.ai-blog.net/ in the future.