Artificial Intelligence

I also have a background in artificial intelligence (AI). I did my BSc in computer science but my Ph.D. was in AI. Why? Because AI is highly relevant to virtual worlds.

The reason AI is relevant is because of mobiles. Okay, well there are other reasons, too—object representation, pattern recognition, route-finding, parsing, and so on—but mobiles were why I went for it. Virtual worlds are places, but (except for heavy role-playing or educational worlds) they’re places that are better for having denizens. The virtual world designer has to create not only the environment, but also the occupants of that environment. This is where AI comes in.

Next to me as I write this is a stack of 26 AI research papers, every one of which I’d love to write about here. These represent perhaps 40% of the AI-and-virtual-worlds papers in my collection—the ones I pared it down to. Unfortunately for me, but perhaps fortunately for you, I’m only giving an overview here. So, here it is.

As far as virtual worlds are concerned, AI is all about agents. Agents are what AI people call independent entities capable of action (they also call them actors). In virtual worlds, we call them mobiles, NPCs, and, if they’re not integrated into the virtual world itself, bots.

Mobiles don’t have to be clever. You can implement wind as a mobile. Bots don’t have to be clever either, but there’s not a lot of point to writing them if they’re not. AI researchers like virtual worlds because they’re shared, complex environments that human beings and AI programs can interact with on equal terms; they therefore use them for bots (although they might actually implement them as mobiles; go figure).

Bot research integrates many aspects of AI. Bots need (among other things) natural language understanding, planning, knowledge representation, learning, plan recognition, and the ability to model their own thought processes. Some of them are created with particular applications in mind (interactive drama or storytelling, for example), and there is an emphasis on making them “believable” (for the benefit of humans), but the great thing about AI is that bots are interesting purely for their own sake. They don’t have to show off any theory: They’re AI in action.

From the point of view of virtual world design, AI promises great things. If virtual worlds could be populated by intelligent NPCs, all manner of doors would open. Examples: Quest generation becomes contextual; reputation becomes intangible but works better; the economy behaves like it should; PC clans can’t dominate politics. Worlds will seem—will be—so much more alive.

I could go on and on about this, but restraint has prevailed. To find out more:

• Read Andrew Leonard’s book. Bots ¹.

• See Leonard Foner’s paper ² about the famous bot Julia.

• Consider the Skotos ³ approach, which comes at bots from a different angle.

• The American Association for Artificial Intelligence ⁴ has been sponsoring yearly symposia on the subject of AI and Interactive Entertainment.

Oh, I would like to add one more thing….

Don’t try to make your virtual world itself be intelligent. Don’t make it spawn creatures that are appropriate to the skills of the players who are about to try kill them; don’t make mobiles in combat act clever against clever opponents and stupid against stupid ones; don’t make the treasure that people find adapt to whoever finds it. This will not only be perceived by the players as unfair and frustrating, but it will seem to mock them. A constant challenge is constantly beatable and constantly patronizing.