175: There Is Research To Be Done

There Is Research To Be Done

Think computer opponents in games are hard now? Wait until A.I. researchers get through with them. Michael Cook examines how the industry and academia are profiting from an open dialogue on artificial intelligence.

Read Full Article

This is very close in its theme and content to an article I wanted to write a while back, but far better researched and expressed than I myself could have managed. Stirling effort!

I too am hopeful about how relationships between academia and the gaming industry will develop in the years to come. Events like the one you mentioned at Imperial show academics realising the potential of games and the technologies that surround them as 'testbeds' - as well as how the natural drive for new gaming experiences presents almost perpetually renewing challenges for developers that could benefit from a more academic approach to the underlying issues.

I myself am currently pursuing PhD studentship funding for a project that ties closely with the ideas behind Facade (particularly the idea of procedural narrative generation). Mateas and Stern's work - along with that of other noted games academics and commentators such as Katie Salen, Eric Zimmerman, Jesper Juul, Janet Murray & Greg Costikyan - has been instrumental in convincing me of the type of research I wanted to pursue, and its relevance both now and for the future.

I'm always looking for new people to discuss ideas with. If you don't mind my asking (and presumption), what's the focus of your research?

If only I were lucky enough to be researching! I'm sorry to say that I'm currently only a lowly undergraduate, with two hard years left before I get close to research. However, I'm sure Robin (www.doc.ic.ac.uk/~rb1006) and Simon (www.doc.ic.ac.uk/~sgc) would both be interested in talking to you. They're both very interesting and enthusiastic people, and were a treat to interview.

I'm glad you liked the article! I really think that the relationship between academia and gaming will grow and grow. There was so much this article didn't touch on, and books could be, and have been, written on this stuff.

I'd like to research in these fields in the future though. I think narrative generation is very interesting, as is planning for failure and co-operation. But I don't have the time to think much about them right now!

I need to ask my professor if our school does any resource among these lines.

So, looks like I'm not the only one who was disappointed with DEFCON's base AI, then.

Anyway, it's interesting to see the links between academic artificial intelligence research and computer gaming. There's a lot of potential in this research for computer games to benefit from, because there are few things which cause a bigger drop in immersiveness than AI representations showing little intelligence.

I suppose, however, that there are some critical differences between making a game's AI very intelligent and making it fun to play against. An intelligent AI which was specifically designed to play a game well might work in conjunction with the "game" aspect of a project like Façade, but it may not work so well in a competitive first-person shooter, for example, short of limiting the AI's reaction time so that players can get a fix on them.

Interestingly, RAK, Dr. Colton is due to give a talk this week entitled "Ability versus Enjoyability in DEFCON AI-bots", and it's a field of study that both he and Robin are pursuing. We don't want to get beaten, but we don't want it to be easy either. It's an interesting thing to consider.

I've never understood why procedural narrative advocates are so deadset on destroying linear narrative as if the two are somehow incompatible. For as much as people harp on creating a world where they can do or say anything, the first few stabs with games that allow this hint at what every person's ambition actually is: break the rules that real life imposes on us. That has enormous merit, but I don't know why people think there is a video game behind being totally free of rules when all a video games is at the core is a bundle of rules with paint.

Even the best procedural world is going to have to face the fact that people are going to ask what their purpose inside that world is. What do the rules mean, what does breaking them mean, and how does that affect some grander scheme? Even a procedural narrative is going to have to define a series of reactions that ultimately prey on the player's pre-conceptions of good and bad, win or lose. The fact that the player is going to ask the game for that conception is why a linear narrative is always going to be a reality in even the freest worlds.

But what's interesting, to me at least, is whether these linear narratives could be procedurally generated. Certainly, our desire is to break the rules, to rock the narrative boat. So the freedom offered by, say, The Sims, is different to the freedom offered by Fallout or Facade. In Facade, I have a direction and an overall goal that's far more tangible than The Sims. And the differences you and I experience in Facade are more fundamental - perhaps fundamental is the wrong word. More focused? More focused than the ones we might experience in The Sims.

I don't think people want directionless games. They just want narratives that shift and change. They don't want a complete absence of them.

This is rather exciting. Can't wait to see what comes of this. (So long as it's not an A.I. facilitated apocalypse)

This is rather exciting. Can't wait to see what comes of this. (So long as it's not an A.I. facilitated apocalypse)

Heh, so I am not the only one who thought this article could inspire the writing of an episode of The Sarah Connor Chronicles. =)

I think what is particularly interesting about this article from the point of view of this gamer, while it features AI research heavily, it is trying to illustrate that if for profit gaming businesses want to break new ground, they are going to need to risk something in research, not just development.

Our computer generated opponents as being the challenges to our avatars are just the most obvious as mere puzzles or single thread routines. Thankfully, IMHO, games that have been successful and I still consider classics are in fact ones where my opponents had some decent tactical reactions, such as in the original Half-Life and Unreal Tournament. I would be down with a L4D type AI that was doing its darnedest to present me with an interesting dramatic story, I think I will that link to facade a click.

Unfortunately, the two current "powerhouse" consoles (PS3 and X360) use primarily in-order processing, meaning they are not set up to make complex, intelligent AI. In-order processing is used almost exclusively for graphical quality, meaning Sony and Microsoft have intentionally crippled their ability to make decent AI in their games, in favor of realistic water reflections and shinier bloom lighting.


Reply to Thread

Posting on this forum is disabled.