Simulated Intelligence

Let me first get some naming right. I will call the area that I intend to focus on “Simulated Intelligence” instead of “Artificial Intelligence.” I do this because what I will do here is not the real hard science that tries to build machines that really act like humans do, that have properties that qualify as real intelligence. I’m not trying to write software that could pass the Turing Test, I’m not even going into that broad direction. My idea is to create software (more specifically software components, vertical prototypes or whatever) which is a sufficient simulation of intelligence to create a good gaming experience.

In order to create a real AI, I would have to delve into Cognitive Science, trying to figure out who a mind works. I have not even the slightest clue how to do that, nor do I have the capabilities (for example access to medical equipment to “look” into someone’s brain, or the ability to make extensive surveys to do black box probing of the brain and extract interesting fact using statistical analysis). Instead I just pick up the research results of some real scientists, bend and abuse them and see if I can make some interesting use of them. You might want to call it “Applied Cognitive Science,” but I think that’s already more credit than it deserves.

So what exactly am I trying to achieve?

Hum, only time will tell ;) But I do have some ideas. As I mentioned before I think that an “advanced approach” of game intelligence could be applied to more abstract things than NPC behavior. There are two broad areas where it can be of use. The one that jumps directly to mind is in-game dynamics, that is creature and NPC behavior, storyline, … In general how the game evolves and reacts to the player’s actions. Another thing I mentioned before is that my vision is a gigantic world which is just too big to be created by any reasonably sized team of human designers. It is worth investigating how software could help to do the grunt work, that is how my ideas could be used for the design phase of a game.

That’s still a lot of abstract, fuzzy bla bla. Let me try to come up with some examples.

NPC behavior: “Stupid” characters can be a real downer for the experience. But giving them a more natural touch is a very complicated challenge. You want them to react to the players presence, but if you just have a boolean trigger “player near → talk to him” (as all characters in Oblivion seem to do), this behavior becomes boring and stupid. You would need a much more complex set of rules that uses a big list of input before coming up with a specific behavior. A non-exhaustive list: “player is nearby,” “player is looking at me,” “there’s something interesting about the player,” “I have nothing else to do,” “I need help.” Each of these inputs contributes a little bit and in the end the “talk to player” action is triggered. Or not. Or after two minutes. Or the player behaves completely different.

Mostly procedural game content: Let the software create terrain and vegetation, place creatures, villages and cities and their inhabitants. When done manually the work would be proportional to the size of the world, thus directly limiting the size by the man hours you have at hand. Much could be achieved using simple algorithms, spiced up with some randomness. The interesting question is if it is possible to capture some of the intelligence and creativity of a human designer in software and if this could result in more interesting and fun worlds.

So how should I approach this topic? I could just jump in and implement all the great ideas that I have. Problem is that I don’t really know yet, how it should work in detail. A more promising approach seems to me to go top to bottom. Play with (and learn about) game AI as it is current state of the art, perhaps set up a playground like a small village inhabited but some folks and give them a simple brain. Then see what the limitations are and how to overcome these. Hmm, that sounds like a plan and should give me enough time to wrap my head about the theories I want to use for the “greater goal.”