For those of you who haven't been following Stardock these last few years, it has helped found a series of new studios out in the Maryland and Austin areas.
In Towson Maryland, we have 3 studios located in the same building which is where I am today working on the GalCiv AI.
The people who make up the new start-up studios include many of the leads of Civ IV and Civ V. So as I'm writing the AI, I am literally feet away from Brian Wade (Civ V AI) and Soren Johnson (Civ III and Civ IV AI). Together, we can compare notes on different effective strategies for getting the funnest and most challenging AIs going.
Today's work is dealing with technological research that the AI should do. I've already handled much of the strange ship building decisions (you'll see those improvements in beta 6). But making sure that the AI makes good strategic choices without scripting it is a real challenge and being able to ask questions of other AI developers has been a real pleasure.
Wait there will be a beta 6? What will that consist of besides AI improvements and bug squishes? Will this be the full version of the game prior to release a last push to kill every bug left?
We all appreciate the hard work you and your company does keep up the great work
There will always be a bug somewhere. You add a new feature, you probably added a bug somewhere. You fix a bug, you might break something else that you didn't think was related. Any change of code risks breaking other things. Also this can change the balance of other things.
You can only strive to perfection, but there is always Murphy's law to bring you down a peg.
Can't say the Civ AI has ever left me impressed, in any way at all. They're average at most, in my opinion.
That's nice.
Wait there will be a beta 6? What will that consist of besides AI improvements and bug squishes? Will this be the full version of the game prior to release a last push to kill every bug left? We all appreciate the hard work you and your company does keep up the great work
Thanks. And yes, the goal for beta 6 will be to squash out remaining major bugs.
Like Blaze says, there will always be bugs. But we want to make sure the shipping game provides a really good experience for new players. That said, we are at the beginning of this game's journey.
So as I'm writing the AI, I am literally feet away from Brian Wade (Civ V AI) and Soren Johnson (Civ III and Civ IV AI). Together, we can compare notes on different effective strategies for getting the funnest and most challenging AIs going.
Nice! That sounds like a pretty awesome game AI jam session Wish I could've been there!
Of course, however, "Do or do not, there is no try." Empire Strikes Back 1980
it was just a saying that you hope to catch all the remaining bugs that you can before release. Knowing of course that the bug killing process is never fully complete .
Again though thanks for all the hard work you and the rest of your company does for producing great games and the general communication with fans... Even if we're all a bit nuts...
I think the Civ AI's did a really good job with the diplomacy area... while Ghandi was a bit um.... broken, they were very true to the nature and some were easier to beat than others.
Civ did not always do the best job when it came to city placement or research.
That said, with the three of you there discussing AI, I am sure GalCivIII will be much more fleshed out than it would have been otherwise.
Awesome news!!!
Brian Wade and Soren Johnson.
In the same building.
I am not worthy.
Frogboy, if it were just you working on the AI I would be happy. But to see that you are working with those two vets makes me very, very happy!
So when can we expect a Stardockicon in Towson?
Making AI think by itself without scripting it should be the way to go; it really opens up the possibilities, specially for modders that want to go deep into changing the vanilla definitions.
Yea, one of the things about GalCiv III that is really new is how data driven it is. The machines are now fast enough to procedurally generate complex strategies. Unfortunately, marketing doesn't know hat any of that means.
The only thing that makes me nervous is the game has so many variables that it can be difficult for humans to decipher. Poor ai and faction traits, planetary bonuses, adjacency bonuses, ideology, ship building, what resources, who is an actual threat, Influence etc.
That said, I think it would be good for the AI to remember the player over games. We KNOW the tendencies of the AI. They don't really know ours. If the player constantly back stabs allies and friends, the AI should remember that over new games. If you are going to be a warmonger with the Terrans, the AI should remember that. If you are going to influence flip them, they should remember that.
I like that Stardock has added some really good game designers and AI guys from other top shelf games. As they say, you combine talent and the talent pushes each other. I was too young to see it, but when I saw a show that talked about the writers on the Sid Caesar Show "Your Show of Shows" I was dumbfounded.
Writers for the series included Mel Brooks, Neil Simon, Danny Simon, Mel Tolkin, Lucille Kallen,Larry Gelbart, Bill Persky, Sam Denoff, Joseph Stein, Michael Stewart, [5]and Carl Reiner who, though a cast member, also worked with the writers.
Just Mel Brooks, Neil Simon and Carl Reiner alone, wow.
No pressure
<Edit> Any ETA on Beta 6> </Edit>
<Edit2> no Offense to Larry Gelbart, he only created and produced MASH on TV </Edit2>
All I think you're actually pointing out is the difficulty of making a decent AI. Intuition and strategy is multifaceted, complicated and fairly subtle. I would imagine that having a room full of people who have approached the problem from different angles as a huge plus, precisely because no one can do it perfectly.
There's as much to learn from picking apart the shortcomings of existing programming as anything, and I'm glad that GC3 is getting all of the help that it possibly can because I'm really looking forward to spending a lot of time with it.
Granted, I never played much Civ (just not my thing), so I can't comment on those games specifically - but I've never heard about the AI in them being poor - they were kind of the benchmarks in each of their respective times was my impression.
Cheers,
-tid242
I think one tough thing for AI is to organize an attack, tougher is to cooperate on an attack with other AI or even a human, and toughest is to figure when is the right time to spring it. You have any thoughts you can share on that? The danger potential a huge 'one next turn' thing for me.
And great dream team you got there!
I totally agree with @-rakkaus- as the biggest gap on the GalCiv series that there is not enough support for modders or not having modder community I think Also like the idea of @Blaze of Glory that AI remembers player gameplay style, because every strategy has counter strategy that AI may execute I believe
I have to disagree eith dome people civ 4 & expansions had a greatai. I like the Iidea of multiple minds working on this. How did you get them. Are you borrowing them . Are they here to stay. I still afaptive ai is the way to go. Even with a afaptive ai it is still good to start out with the best tactics possible. They don't just learn from you but they also learn from each other. I think the ai should prepare for you to do your average ganeplan. Not necessarily holding the ladt game against you but if ypu regularly do something then prepare for what it knows your going to do. I'm sure that you are going to do good. About the to msny resources give us options on what to have then we can choose ehat to give up. I guess I'm going to hsve to havr multiple simutanious gsmes to run conseculively.
Good god man, spell check.
Works for me; Psyche! When I play Chess, one of the things I do is the same thing over and over again until my opponent learns to expect it and then I switch things up and it's Mate in 3!
I hope the AI can be bluffed just as easily.
What I would like to see as a modding tool. An easy to understand AI decision log. This way I can make sure the AI is making the correct decisions when I decide to make an AI.
Everyone has a bad day. He is usually much better.
This. With Paul, the AI doesn't play the game, it DESIGNS the game!
Heh. You know, looking at what Brad's doing with GC3 and the Oxide guys are starting with Ashes of the Singularity is making me really jealous for the days when I had the time to roll up my sleeves and get down to good old-fashioned computer player AI ...
I suspect that any AI good enough to play a really competent game against you will be making decisions in a way that's much too complex to be loggable in any sort of format that's readable to people other than the developers.
Hmmm deciding how to take this...
I have modded AI and understand LUA and html formats. When I said "simple" I didn't mean laymen simple. Maybe I should have said "concise".
Anyway,
To BRAD:
We need to have to have a log available to analyze the AI decision process. Since modders, are not making the game, we usually have to backwards engineer this stuff.
I want to be able to understand the diplomacy and strategic decisions that the AI makes.
There are many great features available to you once you register, including:
Sign in or Create Account
