Rise of the machines

[martin]

well I suspect the best approach may be to program AIs with specific tactics, and then use a GA to evolve a counter-tactic, then have all the evolved tactics in some kind of a larger AI framework - which chooses which tactic to use in response to information on enemy silo positioning.

[Ace Rimmer]

Sounds good to me, when can we expect the new and improved AI?

[Nutter]

Sounds good to me, when can we expect the new and improved AI?

Tomorrow



... just kidding

[martin]

Sounds good to me, when can we expect the new and improved AI?

Tomorrow

heh, funny you should say that, I've been programming an AI library for games me and my friend are making (pathfinding etc) and I just promised him the alpha tomorrow

anyway, if I could program AIs for defcon I might have a go, but I don't so I can't :S

[leckyt]

What we're talking about here is what I think of as AI vs. A-Life : bringing artificial life (strictly speaking : digital a-life) to video games. Just for the record, I tend to separate the two because it makes my thought processes easier. In other words, a-life uses AI techniques to mimic real behavioral patterns that we see in life. I can't describe it more succinctly than that here, but I'd happily take up a private discussion with anyone that wants to.

The thrust of the application of a-life in video games, for me, revolves around behavioral modification; from adaptive behavior through to mere variations in an otherwise 'perfect' AI (within the rules of a game, these are sometimes possible). This modification works on other levels too - using genetic algorithms to build armies of behaviorally similar but locally different creatures, or using genetic programming to create behavior trees that might be 'better' or 'worse' than their parents.

And there are yet more levels - everything from in-game prose to the way that NPCs look as well as react. These can equally well be based on GP and GA techniques.

With today's processors, there is no reason, in my mind, that we can't have all of the above goodness. It might even work out as cheaper for the developer : after all, it's easier to procedurally generate a vast city than go out and take a million photos of a real one, right? This, to my mind, is an example of artificial life, that is, creating a synthetic version of something that occurs in real life.

Again, in my opinion, the way that the industry should approach this is to attack the issue in a low-impact way. Altering the existing AI techniques by bolting-on adaptive and a-life techniques. Bringing it in slowly ought to quell the inherent panic of designers, developers, and stakeholders who believe that a-life should stay in the lab. Present company excepted, of course!

This slow introduction of a-life in video games has probably started, but needs a small push!

[zoombapup]

I'm kind of annoyed that this whole network wasnt a bit wider publicized as I'd have loved to have come along (being both a game programmer and an academic).

One thing I've found as a game programmer, is that most of the time, the constraints on games make a load of the techniques people use in academic AI (its a bit nasty to seperate it as its just AI) just not possible. I'm glad that academics are interested in starting the debate, but a bit bemused at the same time.

I dont see many games really pushing the envelope in terms of AI. I cant see many academics really contributing to games AI research properly for a while yet. Mainly because the funding opportunites are bound to lie elsewhere (military springs to mind).

Its all good though, hope to make it to the next event, whenever that is.

[tllotpfkamvpe]

.

[128.128.128.128]

sounds like a no brainer to me..hehehehe anyway, does anyone ever thing a.i. will ever prove useful in human life?

[KingAl]

Err... aren't we just talking about one of many applications for A.I. right now?

[KingAl]

More on this.

"...but this simple idea led to significant results ... for Introversion, who will soon have tools that they can release to players to let them write their own A.I. for the game."

[martin]

Ooh sounds interesting, maybe I can have a go at building that evolving tactics AI that I appeared to be talking about a year ago

[Ace Rimmer]

Nice, competitions.

I'll certainly "write" A.I. for defcon if for nothing else than to drive you nuts (for a while). That is, assuming what's in the story is true.

[ynbniar]

That is, assuming what's in the story is true.

I spoke to Robin at the MRP (I think so anyway...I spoke to someone who was working on a project on Defcon AI so I assume it was him)...so I suppose the article is indeed true...