Gaming Building a StarCraft bot that plays like a pro

[The Helper]

PhD students at UC Santa Cruz are working on building an artificial intelligence designed to take on the very best StarCraft players in the world. Here they explain how they're doing it, and the challenges that lie ahead for them.

Can an artificial intelligence play StarCraft like a professional player? That's what UC Santa Cruz students Ben Weber and Peter Mawhorter set out to test when they began their latest project at the university's Expressive Intelligence Studio. Working with their associate professor, the noted interactive drama pioneer Michael Mateas, they are participating in the studio’s broader goal of exploring the intersection of artificial intelligence, art, and design. According to Mateas, "our goal is to create compelling new forms of interactive art and entertainment that provide more deeply autonomous, generative and dynamic responses to interaction. A major thrust of this work is advanced A.I. for video games, including autonomous characters and interactive storytelling. By viewing A.I. as an expressive medium, our work raises and answers novel A.I. research questions while pushing the boundaries of the conceivable and possible in interactive experiences. Current projects in the group include automated support for game generation, automatic generation of autonomous character conversations, story management, and authoring tools for interactive storytelling."

Whole story here.

 

[Dan]

Well... of course you can make an AI for a video game... It's just a series of algorithms that will choose the best actions to be made... though it's hardly making it's own decisions. it will be as good as the algorithms it's made with.

wonder what the interfacing robot part will look like. XD

 

[sqrage]

Well... of course you can make an AI for a video game... It's just a series of algorithms that will choose the best actions to be made... though it's hardly making it's own decisions. it will be as good as the algorithms it's made with.

I think everyone knows what AI is.

This is pretty could, if they actually make it would. They could leave their computers running and reach #1 without ever playing SC2.

 

[BANANAMAN]

Good luck with the impossible.

 

[thewrongvine]

Guess they gonna have to interview me...

~Hai-Bye-Vine~

 

[ReVolver]

Have the best SC2 player play for one game and have the computer record his input and just replay it? Saves about 250,000$ on stupid research. I wish I had a phD these days because finding out that we can make cars out of coffee is brilliant!

 

[sqrage]

Have the best SC2 player play for one game and have the computer record his input and just replay it? Saves about 250,000$ on stupid research. I wish I had a phD these days because finding out that we can make cars out of coffee is brilliant!

You seriously think that would work? If it repeats the same thing over and over it would obviously just start out making the counter for that one play.

 

[Durandal]

stop focusing on the concept of building an A.I. to beat starcraft, yes that is their goal, but it is not why they are doing it

By viewing A.I. as an expressive medium

A medium is what you work with when you create art, it doesn't matter if the A.I. ultimately can or cannot defeat the best SC players out their, what is important is that they discover new things and important information, they are doing this for research.

Not to beat some stupid SC player

 

[ReVolver]

You seriously think that would work? If it repeats the same thing over and over it would obviously just start out making the counter for that one play.

Well AI is based on a human player, we will defeat each other eventually, I'm just seeing the satire in this.

 

[Sevion]

They should try to make a learning A.I.

(We'll all be doomed! I smell Arnold...)

 

[SineCosine]

Give it a few years.
I hear PC capabilities are tripling every few years =/

In our life-time, we'd already be out-smarted..
Soon..

 

[Jesus4Lyf]

building an artificial intelligence designed to take on the very best StarCraft players in the world.
...
including autonomous characters and interactive storytelling.
...
Current projects in the group include automated support for game generation, automatic generation of autonomous character conversations, story management, and authoring tools for interactive storytelling."

...?
Are they writing story lines, or SC AI?

 

[LurkerAspect]

So one of these days I'll be playing online... against AI??? :nuts:

It seems like a really cool idea and project to work on, but methinks the outcome, if as succesful as they want it to be, would be highly controversial, and would probably change the way we play starcraft. Think about it, you'd have to start checking if you're playing a human or a bot!
And what would do if one of these bots reaches #1 on the ladder and starts winning tournaments?
Do we reward the owners of the AI? We'd start having AI battles instead of human battles!

 

[SerraAvenger]

I think it will be very challenging to build an AI for a complex game like starcraft that will actually be able to win against good players.
Of course, an AI can easily excel in a couple of problems. Eg processing many units at once, clever microing in a couple of situations. Simple game decisions.
Other problems, however, may be really hard to solve.

Have the best SC2 player play for one game and have the computer record his input and just replay it?

You see, there's a reason you don't have an PhD

>learning A.I.
The LISP stuff all over again?
Using genetical algorithms?

@SineCosine
sigh... Speed is only a linear factor. Our brain works completely different from any computer. Of course, we can emulate artificial neuronal networks, but its still not the same.
Computers have outsmarted us a long time ago. And they will never outsmart us. What I want to say is: There is problems a computer can solve much faster than any human. There are problems a human can solve much easier than any computer.
If we ever manage to build an AI that outsmarts us, then it will happen by chance. We will never be able to _really_ understand why it outsmarts us or how it works, nor will we ever be able to understand how we work. And a computer will never be able to understand himself either (unless, of course, we tell him how he works. But that would be considered cheating since no one tells us how we work either.)


@da1
Have you ever played chess?^^
We might do the very same = )

 

[BANANAMAN]

Of course, an AI can easily excel in a couple of problems. Eg processing many units at once, clever microing in a couple of situations. Simple game decisions.
Other problems, however, may be really hard to solve.

Microing is not simple it is by far the most complex aspect of the game since so many random variables are associated with it.(Hence why ZvZ is such a volatile match up) Getting the AI to calculate which targets to attack to maximize it's chances of victory is hard enough.Sure you can order an AI to force fire on a target and retreat back immediately. But good luck getting an AI to do that simultaneously while at the same time adapting to the changes in the enemy player's change of micromanagement tactics which can happen at any given time of the engagement.Then there's also the fact that the AI can be easily abused due to it being limited to a set programing algorithm.

I have yet to see an AI that can mirco like a pro and at the same time marco like one.They're better off creating a real life replica of the C-14 Gauss rifle that works just as well if not better than the one in the game. Which is not an easier feet to begin with.

 

[Jackal]

I'd imagine it's all possible. Deep Blue anyone?

 

[sqrage]

Give it a few years.
I hear PC capabilities are tripling every few years =/

In our life-time, we'd already be out-smarted..
Soon..

I think I'm already outsmarted. Wikipedia and the calculator tool contain far more knowledge than I do.

 

[13lade619]

Moore's Law.

The complexity for minimum component costs has increased at a rate of roughly a factor of two per year... Certainly over the short term this rate can be expected to continue, if not to increase. Over the longer term, the rate of increase is a bit more uncertain, although there is no reason to believe it will not remain nearly constant for at least 10 years. That means by 1975, the number of components per integrated circuit for minimum cost will be 65,000. I believe that such a large circuit can be built on a single wafer.

http://en.wikipedia.org/wiki/Moore's_law

 

[SerraAvenger]

Microing is not simple

I never said so = )
I only stated that an AI can excel at microing in some situations.
It has the huge advantage that it can basically control any number of units at the same time independently.

[An] AI can be easily abused due to it being limited to a set programing algorithm.

I do totally agree here. We won't be able to reproduce an AI that works the way we do because we don't know the algorithms we use, and we won't be able to get those.

A mutating AI would probably be very interesting. Having one AI template and randomly choosing decisions. At the end of the game, the decisions would have to be analyzed. Those decisions that changed the game in favor for the AI would from then on be slightly preferred. The only thing that would be necessary is a function to analyze any current situation in the game, which could be very interesting too.

Getting the AI to calculate which targets to attack to maximize it's chances of victory is hard enough.

Well I wouldn't be so sure on that. Of course, getting the chances to really MAXIMIZE the chances of victory is really hard, but you can get a pretty good estimate only checking some potentially beneficial targets.

A fairly interesting project. I'm glad I don't have to do that = )
I would utterly fail. Even if I was a better Sc player^^

 

[Slapshot136]

I shudder at just the thought of just a computer with half-decent micro...