IncGamers launched into space and dodged a series of devious wormhole traps to speak with Chris Park, the creator of RTS/tower defence hybrid AI War.
AI War’s central feature is, shockingly enough, the AI; can you explain how it differs from AI found in other RTS titles?
Sure. The core thing that’s different is that I started from a completely different premise. My background is not in AI, my background is in database programming, data-mining that sort of thing. It uses a lot of emergent principles, not simulating too much overall intelligence at a tactical level. Usually in traditional AI you have a lot of battlefield agents. You say ‘there’s this group of fifty guys and they belong to Commander X and you’ve got this group of guys that belong to Commander Y.’ I don’t have that at all. It looks like I do because the guys seem to move around in groups but it’s actually more of a flocking sort of behaviour that arises out of what the individual units are doing together.
Has it been your aim with this game to avoid the exploitable weaknesses that you tend to find with other AIs?
Absolutely, and I think that boils down to predictability. Behaviour’s only exploitable if it’s predicatable – if I know the AI’s gonna do this, I can do this to counter it. It takes players a long time to figure those exploits out, but once they do the shelf life of that game – if they’re not playing competitive multiplayer – is just gone. So that was one of my chief goals, to keep the game interesting for myself and my group for longer than six or twelve months.
So was it your dissatisfaction with the AI in other games which sparked the idea for this project?
Yes, absolutely. I’d been playing with that core group of my dad and my uncle since about 1998, so I guess about eleven years. We play basically twice a week, and we’ve gone from game to game over the years. Most recently we were playing Supreme Commander and its expansion – which we really like, it’s a quality game. But we exhausted the AI in Supreme Commander, moved up to the expansion which had better AI, exhausted that, found a mod where the AI is better again, then exhausted that too.
I was thinking maybe I’ll make a mod as well. Then I thought, why don’t I just make a game of my own? I’ve got background in game design, I’ve been doing that forever, so why not make a whole new game, new units and everything?
Even as the creator, is the AI still able to surprise you at times?
Oh yeah, it surprises me pretty routinely. If it didn’t that’d be a little too predictable, right? I still play AI War weekly as kind of an alpha testing thing with my same group. This past weekend we were playing. We’re getting raided on some planets and, you know, we’re pretty good at the game, I mean, we made it! So we’ve been playing it for much longer than anybody else. Normally on Difficulty Seven, which is what we play at, we’re not losing the game too often. But the AI almost put us out of the game because it slipped passed some defences while it was occupying us elsewhere. It’s main forces are over here distracting us and then this other little force comes through. We hear massive explosions and think ‘oh no, what was that?’
Half of the galaxy is now in flames.
Yeah, exactly! Whoops … there goes his economy. We’re putting up forcefields and all these turrets and stuff and it’s still figuring out ways to get around it. So it’s like playing a human person. Playing against a general player you can put up about this much of a defence and you’ll hopefully block most attacks. But then you find somebody who does something a little bit tricky – and you’re in trouble.
You mentioned that you play on Difficulty Seven, and the levels go all the way up to Ten. Would I be right in saying that some of the higher level campaigns haven’t been beaten yet?
Do you expect that they’ll remain unbeaten?
Yeah, probably so. The Level Ten AI … nobody has beaten that on a fair basis. We’ve got some players who play exclusively on Ten and give themselves a handicap – and they just take a beating the whole way through. They make it pretty far, but often they give up after a while because they’re just getting beaten on the whole time, so it’s maybe not as much fun as they’d hoped.
The reason for having the Difficulty Ten is this is co-operative only, right, so there needs to be a difficulty level for everyone. If the best Starcraft player in the world wants to come and try his hand at this – not that I expect it to be his bag, but if he wants to – there ought to be something that can occupy his interest. For mere mortals, like myself, there’s a range of other difficulty levels there.
How fair does the Level Ten AI play? Is it working purely from intelligence, or does it get resource boosts and other help?
That’s a good question. One thing I should mention about the AI in general is that there are a lot of different subsystems: you’ve got the tactical stuff, the strategic stuff, economy etc. From the economic and strategic standpoints the AI is very asymmetrical. This is something which has been seen as a unique selling point of the game actually. It’s not true RTS, it also has some aspects of tower defence, grand strategy etc and some of the tower defence aspects come in with how the AI strategy works. On the tactical level it’s one hundred percent fair, no bonuses, no anything. With the economy, that’s something that’s simulated, so it’s got its own internal reinforcement rules and offensive wave rules and so-forth that are completely divorced from what the human player is doing.
What I’ve tried to do is balance that, so it comes out at the appropriate difficulty level at each stage. Seven is about what you’d expect from a human player who’s controlling the other side. The AI outnumbers you 10,000 to 1 to start with. It’s got 20,000-30,000 ships and you’ve got maybe four. So obviously if it just ran at you with everything the game wouldn’t last very long.
Right, it would be all too easy for it to just flood you with ships immediately.
Exactly. Here’s the comparison that I like to make – let’s say that you’re playing a first person shooter. The AI is generally pretty fair in an FPS, overall. But if all the guys knew where you were and ran at you – you’re dead in 30 seconds there too. I enjoy myself some FPS games, so I thought what if instead of running around with your own gun, what if you were in an RTS-type control scheme, controlling a bunch of guys. So you’ve got all these outposts and some of the AI units are aware of you and attack you, some are not. That’s all very procedural. There’s a lot of similarities to, say, the procedural algorithms in Left 4 Dead, where fairness is not a goal. With procedural generation the goal is an interesting scenario and where the fairness comes in is balancing that for when the players encounter a group of AIs.
The intelligence ramps up from Difficulty One through about Seven. There are a couple of things that AI Eights and up do that’s different. Basically the intelligence on the tactical level stops at Seven – and the reason I did that was because I wanted everybody to be able to play against an AI that’s as intelligent and dynamic as possible without having to handicap themselves with bonuses and stuff like that. So it’s kind of a judgement call whether you consider that a cheating AI or not. It’s really a different design paradigm from a traditional RTS. If you consider it fair in shooters then it’s fair here, probably. It’s always a simulation, you know. No matter what game you’re playing the fairness is an illusion perpetuated by the game mechanics.
The multiplayer aspect is co-op only. Was that a decision you made right from the start?
Yeah, absolutely. The first few months of the game it was PvP only, because I was building the gameplay. The actual mechanics and interface, everything was being developed at that point, so we played PvP. Once the AI came out then it switched to co-op. I don’t play PvP competitively and nobody in my group does. I play co-op intensively so I know what needs to be there and what the shortcomings are – and I know what other games have done co-op right. So I was in a good position to do something that really works there.
There are so many other PvP games out there that are really popular and really good. So that wasn’t necessarily a space where I felt I had anything new to offer or that I wanted to compete in. There’s been some talk about adding PvP in an expansion, later on down the line. We’ll see. It kinda depends on player interest.
It seems like you have an end point in mind for the development of AI War. What would that be and how close do you think you are to it?
Well, this is something that I’m basically going to keep on building for as long as there’s interest. I’ve mentioned that I’ve been through a whole string of games over the last eleven years and one thing that’s made me a little unhappy is that you’re going to exhaust a given game eventually. Past a certain point you’re just done with that game. Then you maybe get a sequel and you’re kind of back to square one.
So my goal is to therefore say, ok, let’s never make an AI War 2. Let’s just keep making expansions for as long as people have an interest in AI War. It’s cheaper for the users, it’s cheaper for me to create it. They get to keep all of their historical content and suddenly you’ve got this game which – it’s already a really large game – but it could be four, five, six times larger than it is now. Right now there’s sixteen billion possible scenarios and then that goes up to some ridiculous number – trillions and trillions. So that’s my goal there.
I’m almost treating it – and this may have a bit of a stigma – but I’m almost treating it like an MMO. Not with a subscription, that’s not what I mean. But insofar as we’re doing free, monthly DLC – and we’re always making little balance tweaks and little additions to the AI and soforth. That’s going to continue, for free, for as long as there is interest. We’re also going to put out a paid-for expansion every eight to twelve months or thereabouts. So for people who just want to have the base game they can keep getting smaller free updates from now until .. whenever. And for people who want large batches of new content to kinda remake the game over and over, they can get that too.
What can we expect to see in the next expansion?
We’re just now finishing up on our 2.0 version [this was ultimately released on October 21st and is available now], which has the last of our main graphical updates and which has some new AI behaviours and balance tweaks, all that general stuff. That’s what’s going on now. Our artist has started working on the expansion and we’ve got a bunch of expansion topics in our forum where people are able to comment and suggest ideas. We’ve probably implemented one or two thousand different player ideas into our free updates since the 1.0 version four months ago. Some of our release notes are like 30 pages long – no joke – they’re about 8,000 words some of them. That’s all been because of fan interaction.
The minimum of what’s going to be in there is something like 80 to 100 new ships. We’re also adding a third, independent faction – aliens that are sort of neutral but you can interact with them in certain ways and get various benefits. With some of that stuff there’s still some ongoing discussion with the fan-base. We’re going to be doing a preorder for the expansion in a couple of weeks, and basically giving out alpha versions to anybody who preorders – and demo alpha versions to anybody who doesn’t. So people can comment and really help shape the expansion into what everybody wants to see. Fan interaction’s been so successful for us that I thought it’d be a good idea to continue that with the expansion and probably with future products too.
AI War is also coming to Steam, is that correct?
Hopefully within about a week [since the time of writing, AI War has now launched on Steam] We’ve got our Steamworks integration already done and our Steam community site – if you search for AI War – is up, so people can join that if they want. We’re just finalising various aspects with the 2.0 version and then we’re good to go.
Will being on Steam help make it a little smoother to create and join co-op games?
Yes, hopefully so. It’s still going to be a direct connect thing at this point, we don’t have matchmaking yet. But we do have an agreement with GameSpy and also with Steam whereby the first expansion will have matchmaking. That’s something which we’ll retroactively give for free to all the existing base game customers, so they don’t have to buy the expansion just to get that.
What have you made of Randy Pitchford’s recent comments regarding Steam and Indie developers?
You know, I read his comments and I’ve seen some other Indies who’ve responded and my comments are kind of the same as theirs. I don’t know where he’s getting that from. I think some of his comments may have been taken out of context … but still. Some of it was about conflicts of interest, I think. But as the other Indies have said, it’s always a conflict of interest. Right now I work with Stardock with Impulse and they’ve got Sins of a Solar Empire which is a direct competitor to AI War, but they’ve done an awesome job of promoting us and we’ve been tremendously successful on their platform.
It’s a similar sort of situation with Valve, they’ve been doing a really good job of promoting the people who are direct competitors to whatever their product of the moment is. And without breaking any NDAs, their agreement is simple and fair and very much at or above the industry norm.
Finally then, do you have any plans to distribute through other digital retailers?
Yep. We’re already on GamersGate and Impulse at this point. Direct2Drive, that’s pretty much finalised, it’s pretty much a go. There are a few other smaller ones too that we might be coming out on, and a few others that are not so small that we’re in discussions with.
There’s a lot of stuff in the works. We actually have some retail deals in the works. We’ll almost certainly be coming to retail in Russia and potentially the US, Germany and UK territories. We’ll see what happens with those.
Arcen games are also working on a puzzle title called
Many thanks to Chris Park for speaking with us. A series of in-depth articles about (amongst other things) designing the emergent AI in AI War cansite.