CIV 4 - AI Requests

[Jon Shafer]

i hear what ur saying but i am thinking in a broad sense, as oppossed to addressing specifics. The limitations in so far as programming are probably based more in manpower and budget restrictions ...(and current computer processing speed) rather than doability.

Making a "Deep Blue" style AI for a Civ game would be a monumental feat to say the least. Deep Blue was a massive project in itself, and was already mentioned, Chess is a very simple game compared to Civ with only so many actions that can be taken.

Possible? Maybe.

As far as the technological ability - the processing power required by such an AI would likely outstrip even what we have today. Keep in mind how long it takes the Civ 3 AIs to take their turns. The code is probably not FULLY optimized, but it's not bad either. Multiply the demands of that AI 20 or 50 times and you'll get a glimpse at the limitations of a "real" AI.

 

[Sirian]

The main problem seems to be programming the AI and how to pay for that

That's not my understanding. Can you point to a game AI where somebody made the choice to "pay for it" and actually produced a strong competitor in their AI?

::looks around at games::

I'm not seeing one.

Three of the best game AIs I've ever played against are Half Life, Descent III, and Diablo II. The first was a totally scripted game, though, so it's hard to count it. It's AI only had to perform well in tightly controlled situations, where the designers could kludge their way past a particular shortcoming. The other two games have distinct behaviors for individual AI units, quite remarkable in some cases (compared to every other AI out there) yet still totally predictable and also inept as a group, once you devise tactics to use against them that rely on their predictability to give you the advantage.

Who, exactly, is Firaxis supposed to bring in and "pay for"? Who's got the resume that says, "I've been there and done that." Nobody, really. Game AI is still in the dark ages.

I just hope that the AI can make a bunch of advances in Civ4. Like, maybe get up out of the dark ages and spawn a rennaissance.

We'll see.


- Sirian

 

[MMAfan]

Making a "Deep Blue" style AI for a Civ game would be a monumental feat to say the least. Deep Blue was a massive project in itself, and was already mentioned, Chess is a very simple game compared to Civ with only so many actions that can be taken.

It would also be a completely useless feat. In CIV the problem of using the 'chess solution' to conquer the game GO is even worse. Deep Blue calculated every possible move up to a certain amount of moves deep. Trying to do this with GO doesnt work because the amount of moves possible each move is too big. IIRC a 50x50 GO board allows for 2.500 different moves each turn. chess usually about 30 different moves each turn. And in chess logic can eliminate useless moves fairly good leaing to an effective branching of just 6 moves for each ply(half move), roughly 40 per turn.

think about using the chess solution for CIV...

You attack using a force of 30 swordmen. each turn these 30 swordmen have 9 diferent moves to make (all direction and nothing):

The amount of resulting positions of the map after just 1 ply:

9*9*9*9*9*9*9*9*9*9*9*9*9*9*9*9*9*9*9*9*9*9*9*9*9*9*9*9*9*9=
42.391.158.275.216.203.514.294.433.201 different positions

Now next your opponent with his 30 swordmen is moving for the defense resulting after one turn in:

42.391.158.275.216.203.514.294.433.201 * 42.391.158.275.216.203.514.294.433.201=
calculator crash amount of different positions.

Actually it is worse because while in combat you have to move your 150 workers too, with all the different results. And the AI's are going to move thier 150 total workers too...

Even if a computer could even calculate 1 turn of possibilities ever. It would be meaningless because the chess way calls that the depth of calculation has a sufficient depth to make a good evaluation of the resulting position.

And at this moment there is not even a good way to evaluate a CIV game position. In chess the evaluation is very materialistic, looking at the point count of the pieces mostly.

I hope this shows the futility of trying to 'chess' the AI of CIV.

If you want the AI too appear more human you have to program in 'humanlike' intellegence. (the chess AI plays good, beats most humans but doesnt play like a human!)

There once has been a suggestion that the AI should be structured like an army. The top, the general doesnt concern it self with individual units... it only sets the strategic goals to be reached. Its servants then take an order and make sure that the individual units behave in such a way that they execute the orders of the general.

The idea is to: simplify, simplify, simplify the information on which a decision has to be made. you cant write a piece of logic that bases a decision on the placement of 500 different units.

Divide and conquer...

 

[dh_epic]

In AI we call that "combinatorial explosion". The solution is good heuristics. In fact, this is the way that more real world players play. Yet programming these heuristics into a computer is never enough.

Still, chess has a few heuristics. Like the opening 10 or 20 moves. And the heuristic of "control the center of the board". There's also applying point values to units in an effort to give the computer a sense of what's a worthy sacrifice.

But I still maintain the most difficult thing about Civ is the fact that exploits exist as they develop it. So you can have a great AI that manages to balance their interests, and react appropriately to crisises. But then there's always the human being who figures out that the best solution is to repeat a small action every turn, and to create a crisis at the most unexpected or unpredictable time.

It may seem, on the surface, that the best way to make the AI better is to let them exploit those stupid actions (like trading for maps every single turn, or moving your science slider down on the last turn of some research, or building then switching to a wonder). But this will just piss off more players. And especially imagine if the computer were as opportunistic as a human -- it would be 10 times worse than a Civ 2 gang bang.

The solution is to close said exploitations and reward logical, peaceful behaviors as much as psychotic, expansionist behaviors. The problem is bigger than AI.

 

[Volstag]

All this talk of "improved" AI is a little misguided. Has anyone played an in-depth computer strategy game that boasts good (let alone, great) AI? I've been gaming for over 20 years, and I've yet to see it. It's possible in chess (and other minimalistic, abstract strategy games), because the number of variables is far, far less than what you'd encounter in CivIII (or Hearts of Iron, Combat Mission, TacOps IV, Total War, etc).

Granted, huge strides are being made in the field, but these are being funded by universities and/or research labs w/ more money, computing power, and time, than Firaxis. Would I like to see better AI... hell yeah, who wouldn't? But, at the same time, I'm not going to hold my breath.

The best solution is improved "hard coded" routines and/or allowing the computer to "cheat" from time to time. I don't particularly care for the cheating aspect, but it accomplishes the goal of improved AI at a fraction of the time/cost. Namely, your opponents are more difficult to beat.

In short: it's going to be a long time before we're playing against genuine Artificial Intelligence.

(as an aside: adding 50 programmers to a particular task doesn't necessarily accomplish the task any quicker).

-V

 

[Volstag]

There once has been a suggestion that the AI should be structured like an army. The top, the general doesnt concern it self with individual units... it only sets the strategic goals to be reached. Its servants then take an order and make sure that the individual units behave in such a way that they execute the orders of the general.

This is the same model that they use in Combat Mission. To it's credit, it works pretty good, but I still wouldn't classify the CM AI as good (maybe "fair"). It operates on a 3 tier hierarchy:

Strategic AI
Operational AI
Tactical AI

So.. in the case of Civ III, the Strategic AI would determine that it needs iron. The Operational AI determines the best source of iron. The tactical AI would do the busy work of actually obtaining the iron (via settling, workers/roads, etc). Just an example, of course.

-V

 

[dh_epic]

I agree that cheating isn't so bad. I just think the AI should cheat more to achieve balance, rather than to achieve a steady pace.

What I'm trying to say is I'd rather the AI has a "catch up" algorithm than keeping a steady pace. Even cheating at the high levels, the AI is beatable, and you can even open up a huge gap on them. And if the AI is ahead of you on the low levels, this would give you a better chance to catch up.

A close game is an interesting game.

 

[Lewsir]

Hearing about how Firaxis will be starting from scratch with CIV 4 does worry me, vis-a-vis the AI. I guess it's possible they will come up with some kind of revolutionary improvement in the approach to it, but it seems much more likely that they will have to reinvent a lot of wheels, and we will be back to having a first release loaded with inadequacies, that get fixed over time through cumulative patching and upgrades. Looked at that way, I begin to really see the need for a principle that CIV 4 AI should be "at least as good" as CIV 3. It easily could be worse, certainly compared to C3C.

If the same energy went into starting from where CIV 3 is now, it seems to me they could make much more progress, albeit within the limitations of the general approach. But I guess this is a moot point. In any case I do look forward to other innovations in gameplay that I guess are much more likely to occur if they start from scratch.

 

[Jon Shafer]

The game engine of Civ 3 was based off of another game and they were having quite a bit of trouble adding much of anything for Conquests. Trying to base an entirely new project off of that would have been a nightmare and I assure you that you probably wouldn't have been impressed with the result. Since the game engine had to be redesigned and constructed, obviously the core of the AI would be unusable.

 

[Sirian]

I guess it's possible they will come up with some kind of revolutionary improvement in the approach to it, but it seems much more likely that they will have to reinvent a lot of wheels

Soren Johnson was the AI programmer on Civ3 and is the lead designer for Civ4. That tells me there won't be any backsliding. Plus Soren was the driving force behind the original vanilla C3 patching process, which eventually dealt successfully with the vast majority of issues present on Civ3's release. (How many games DO NOT ever really get fixed post release? MOO3 comes to mind. Diablo II expansion took like THREE YEARS before they got around to fixing it. I could go on!)

Firaxis picked up a lot of currency with me as a result of keeping their commitment to Civ3 going until they really did get the game polished. Of course that won't save them if Civ4 were to flop, but truly, that's the last thing I expect to see happen!

What I worry about is with the game still supposedly a year away (or something like that) and the fans so active with suggestions and stuff (I mean, LOOK at this forum! Wow) I just hope everyone remains calm and level headed, and nobody works themselves into a lather and then goes nuts when they realize the game's still a year away or something. And what if, by chance, they were to push it back? That may not happen, either, but really, we are shooting at the moon here. It's all speculation. Well, that and making requests and forwarding ideas in the hope they get used.


- Sirian

 

[Lewsir]

Soren Johnson was the AI programmer on Civ3 and is the lead designer for Civ4. That tells me there won't be any backsliding. Plus Soren was the driving force behind the original vanilla C3 patching process, which eventually dealt successfully with the vast majority of issues present on Civ3's release.

Encouraging.

I think people will be patient, though I admit given 1 year+ waiting time, it would be nice to have another expansion to keep us going until CIV 4! Luckily there are other things in life...

 

[dh_epic]

I don't think we should get too hasty. I think Christmas 2005 is EARLIEST, which is more like 1.5 years. It'll probably be more like 2 when all is said and done.

 

[Lewsir]

Hmm. Okay, here goes my naivite - if it could be 2 years til Civ 4, wouldn't it make sense for them to put a few programmers on doing a 4th expansion of Civ III, to come out in, say, spring 05? I mean, at $25 a pop that must be easy money for them...

I await to be disabused of this notion (it would take way too many programmers, not enough time, no one would buy a fourth expansion...)

 

[MMAfan]

Since this threat is called AI requests I want to post a request.

In all previous CIV's there was on choice of difficulty (amount of cheating actually)... Since I dont think it is even possible to make an AI that will equal a human without cheating, this feature will return.

I just wonder if they could make it possible to make separate difficulty levels for the various aspects of the cheating:

-choose the amount of AI-AI trade chreating
-choose the amount of extra starting units of the AI
-choose the amount of happy citizens in cities (your cheating factor)
-choose the amount of combat cheating
-etc...

well perhaps not a separate difficulty option for everything, but a bit more choice then just one. Make it possible to play a high diff level on a tiny pangea map without being run over by the AI's starting offensive units while you are building your first warrior.

</end request>

 

[Jon Shafer]

Hmm. Okay, here goes my naivite - if it could be 2 years til Civ 4, wouldn't it make sense for them to put a few programmers on doing a 4th expansion of Civ III, to come out in, say, spring 05? I mean, at $25 a pop that must be easy money for them...

I await to be disabused of this notion (it would take way too many programmers, not enough time, no one would buy a fourth expansion...)

Read my last reply to you.

 

[warpstorm]

Hmm. Okay, here goes my naivite - if it could be 2 years til Civ 4, wouldn't it make sense for them to put a few programmers on doing a 4th expansion of Civ III, to come out in, say, spring 05? I mean, at $25 a pop that must be easy money for them...

I await to be disabused of this notion (it would take way too many programmers, not enough time, no one would buy a fourth expansion...)

I think late 2005 is the earliest we will see it shipping. It might be later.

As for another expansion, I would buy it.

Now for the disabusing.

I don't think Firaxis has the staff to do this without hiring for it. It's not just a few programmers, but rather about 25 people to do an expansion with the content of C3C. If they hire for it what do they do with the extra dev team (they are currently staffed in the ballpark of 50 people) when the project is done? I suppose they could do like they did for C3C and contract some of it out.

The other issue is how to convince the publisher to pay for it (especially as they are releasing a version titled Civ3:Complete next month)?

 

[Lewsir]

By this...

Read my last reply to you.

I presume you mean this...

The game engine of Civ 3 was based off of another game and they were having quite a bit of trouble adding much of anything for Conquests.

I find this hard to believe, given the 1001 ideas sitting right here on this board. What strikes me about most of the suggestions you get is that they sound more appropriate as (mostly attractive) incremental improvements to CIV 3 than something for CIV 4, if that game is going to be very different.

I can imagine that there may be difficulty in coming up with the one big thing that grounds the next expansion and gives it a theme. (though I like the idea of having another scenario-related one, with all the scenarios using Earth world maps starting at different points in actual history).

Now, these points...

I don't think Firaxis has the staff to do this without hiring for it. It's not just a few programmers, but rather about 25 people to do an expansion with the content of C3C. If they hire for it what do they do with the extra dev team (they are currently staffed in the ballpark of 50 people) when the project is done? I suppose they could do like they did for C3C and contract some of it out.

The other issue is how to convince the publisher to pay for it (especially as they are releasing a version titled Civ3:Complete next month)?

...I find hard to argue with (but I'll try). I had no idea it takes a staff of 25 to do an expansion - I would have guessed 5 with some contracting (though I suppose that easily amounts to the same thing). But why not contract out nearly the whole job, so that you don't have to worry about them when the project is over? As for convincing the publisher, if C3C turned a healthy profit (it did didn't it?) then it should be almost a no-brainer. After all, SIMS keeps cranking out expansions about 1 every six months (I know, their sales are much higher - but the point is, people will buy the 4th and 5th expansions of a series).

Incidentally, does anyone know what the sales of C3C have been, world-wide?

Last thought then I'll shut up - if there are no further expansions (and there obviously won't be) then I think they'll be under great pressure to get CIV 4 out sooner rather than later. 2 years is a long time to go without a release, I therefore doubt they'll let it go much beyond late 2005.

 

[warpstorm]

I find this hard to believe, given the 1001 ideas sitting right here on this board.

It's not the ideas. Ideas are a very cheap commodity in the games industry.

It's the problem of implementation.

What Trip was alluding to is the fact that the Civ3 code is old (parts of it were actually from previous games that had nothing to do with Civ), crufty, and pushed to near the breaking point. It was never the cleanest most well thought out code to start with. Two expansions later and it is very hard to work with without breaking something else.

 

[dh_epic]

I personally think the investment on Civ 3 expansions are wasted for the returns. Speaking from where I work anyway, it reached a point where it cost so much to push the code forward since it was so intricate and complicated (without necessarily being well laid out). So we retired one code base, and announced production of the next full version on a completely new code base.

25 programmers for an expansion sounds poor marketing, if you ask me. But I don't have the numbers in front of me, so I'm really speaking randomly. Not like I'm changing anything, nor do I want to. Just hoping they can get to Civ 4 and hit the ground running.

 

[Lewsir]

All interesting. I guess maybe it's a natural cycle - code gets stretched to the breaking point, then you have to start over.

Yes, I wouldn't want to distract Firaxis from working on CIV 4!