Why isn't there any news about the AI?

[Kautilya]

Now I don't want to sound gloomy like a few people *cough*... yin...*cough* but is anyone else worried by the almost complete absence of news about any effort to improve the AI?

For me and for most Civers , I think, AI ranks right up there among the very most important things that we would like to see improved (albeit also among the most difficult to improve).

You can have some great gameplay improvements (and actually I think the culture model and new diplomacy/trade system sound great) but in the long run they mean squat if the AI is crap. There is no fun thrashing the AI no matter how cool the gameplay. And if the AI is poor it won't be able to exploit the cool new gameplay anyway.

Multiplayer might be a solution for some, but only a fraction of Civ players play multiplayer and even hard-core multiplayers probably would like a better SP experience.

If Firaxis was thinking about making significant improvements to the AI they would have hyped it no end. So conversely their failure to say anything seems to indicate they aren't putting too much effort there.

One of the most depressing things in empire games is how there is so little improvements in AI over the years. Almost all aspects of PC games in general improve consistently year after year but it wouldn't surprise me if the AI in Civ 3 is no better than in Civ 2 or even Civ 1.

What do you think?

[Sabre2th]

In many cases, the strength of the AI comes right down to computing power. There will never be a truly smart AI (until actual artificial intelligence) and the AI won't be challenging without having it cheat at least a bit.

[Zanzin]

why haven't they talked about the AI??

Well, if it's as poor as the graphics we're seeing, it doesn't suprise me they're not talking about it!

[yin26]

In fact, my 'review' of Civ 3 puts almost the entire failure on horrid AI. Thus, your question is a profound one. Not even the silly tag lines, like "The best AI ever seen in the genre" have been thrown around. At least give me the option to ignore the AI hype...but to have NONE is, well, "interesting."

[Sabre2th]

For some reason, I don't know why, I think that Firaxis is holding information back so that they can really surprise us later when the game is finished/close to finished.

[Alex 14]

Me too.

[Kautilya]

sabre,
I hope you are right but one problem with your theory is that Firaxis has in fact been hyping their game rather a lot so it's hard to believe they are trying to reduce expectations.

As yin has noted in his thread they have hyped the graphics to the sky but from the screenshots the graphics will , at best , be at the level of CTP-2 ie. decent but nothing great. At worst they could be even worse than Civ2 in certain areas.

The basic question is: if something like graphics, which Firaxis has hyped, is going to be average, how much worse is something like AI which they have ignored going to be?

The more I think about it the more I am becoming like yin.

[LewsTherin]

Well, if something Firaxis hyped is bad (graphics) then maybe something they didn't hype is good... at least I hope so...

[yin26]

The more I think about it the more I am becoming like yin.

God help you, friend. But it IS awfully hard to think they are just keeping the cutting-edge AI a 'secret,' eh?

[Inverse Icarus]

that and the fact that the status of the multiplayer system is questioinable at best.

[Nazgul]

Well AI takes so much time to make and mostly to fine tuned and balance, that it is usually the last thing the development team work on.

So you are unlikely to hear anything about the status of the AI until quite late in the development.

The real question would be how far are they in the completion of that game?

Nazgul.

[Wiglaf]

Well AI takes so much time to make and mostly to fine tuned and balance, that it is usually the last thing the development team work on.

If that's the case, forget it, civ3 will be a failure.

Work on the AI should begin immediatly, seeing as the game is overcomplicated in its current stage. To try to cram all these features (trade, diplomacy, etc) into the computer at the end of development is like studying for a final thirty minutes before it's handed out. No good.

[H Tower]

they probably haven't been working on it too much. They might have just copied and pasted the civ2 ai into the civ3. afterall, they aren't messing with the game eingine much, except for this whole culture and border thing. Heck, they might be using the SMAC ai and game engine because then they would only have to add culture to it...

[SerapisIV]

Originally posted by H Tower
they probably haven't been working on it too much. They might have just copied and pasted the civ2 ai into the civ3. afterall, they aren't messing with the game eingine much, except for this whole culture and border thing. Heck, they might be using the SMAC ai and game engine because then they would only have to add culture to it...

Firaxis Mike posted a while ago here saying that the game began originally as SMAC with placeholder art and iterated ever since to the point where maybe 10% of the engine is still SMAC, so it's pretty much a new game engine. I'm not a programmer, but it would seem that the AI is part of the game engine.

edit: 10% was Firaxis Mike's number, not mine

[H Tower]

I was right!? NOOOOOOOOOOOOOOOOOOOOOOOO!!!!

*runs away screaming*

[SerapisIV]

Originally posted by H Tower
I was right!? NOOOOOOOOOOOOOOOOOOOOOOOO!!!!

*runs away screaming*

Huh? I think what they did was good. It means that they've been playing the game since the first started production 2 years ago. That's a hell of a lot of playtesting, more then most released games get. Battle Realms, a heavily anticipated RTS game from Ed Del Castillo (C&C) only was playable a few months before E3 this year and it's to be released in October, that's maybe 9 months of playtesting, Civ3 will have over 2 years.

Besides, I'm sure you could find close to 10% code similarity between SMAC and Half-Life, some code is just so basic you don't need to change it.

[Lee Johnson]

For some reason, I don't know why, I think that Firaxis is holding information back so that they can really surprise us later when the game is finished/close to finished.

Or, it just may be that they aren't telling us anything because there is nothing to tell. If Civ III had great AI in it at this stage, I'm sure we would have heard about it by now.

[JellyDonut]

Hmm...an optimist might respond along the lines of "The reason They're not publishing anything about the AI is because when the game ships it'll be a huge surprise that the AI on its hardest mode can read your brain waves, determine your strategy and thrash you into the ground!"

However, the pessimist might say:
"Well, since they're not saying anything it's going to be the ****tiest AI ever. It's not going to know anything about the new features or the old features for that matter. It won't be able to move units or build cities because it'll be really really dumb even tho they've had several years to work on it."

The realist, however, knows that without facts we have no way of knowing what the AI is like.

[LewsTherin]

Well, I think the AI will better (or at least equal) to civ2 - more CPU power = better AI possible. "possible" is the critical point, though.

[Rhuarc]

Thank you Jely Donut! It is about time somebody says that! Listen people, without Firaxis actually say something about it we have no way of knowing how the AI is going to be. it could go either way, it could be that on its hardest level you could play for 50 years and never come close, or it could be that on the hardest level a 6 year old could beat it. I really hope it is the first on eof those two, just like everyone else! but my point is until we hear something then there is no point in making baseless speculations about it!!

[Sabre2th]

Firaxis has in fact been hyping their game rather a lot so it's hard

I haven't seen that much hype. I've seen a couple of statements that have been repeated over and over and over and over... you get the idea. Firaxis hasn't released much at all. It's just too early to tell.

if something like graphics, which Firaxis has hyped, is going to be average

Again, we have no clue what the graphics will be like. The screenshots we've seen are very old and Firaxis has been working full speed on the game.

BOTTOM LINE: It's too early to tell. Everyone stop whining.

[Frugal_Gourmet]

Actually, the person who said the AI is generally the last part of the game to be developed is absolutely right.

As someone who has designed computer games before, this has always been my experience in the past and I think you will find it still works that way today.

Just to submit a little common sense, think about it this way. If the game mechanics are not all devised and implemented in a semi-permanent way, how in the world can you design an AI to act within these game mechanics in an intelligent way?

More importantly, *should* you even design an AI -- even partially -- when the game rules/mechanics have not yet been decided upon and balanced? It's generally the case that designing an AI as you develop the rest of the game results in serious backtracking and wasted effort in the long run. Better to design the AI after the game itself is in a relatively fixed state.

For this reason, I think it's extremely likely the reason we haven't heard much about the AI is that it isn't yet developed.

However, given all that (and not to be a pessimist)... but you can STILL expect that the AI will be infinitely inferior to the human brain. Nowadays, they can design chess computer games that can beat (normal) human beings with pretty good consistency, but games like Civilization(tm) -- in addition to a much larger board and number of "units" -- have hundreds of other factors and permutations of events to take into consideration.

I have to admit the AI is not a big deal to me, since my primary realm of interest these days is multiplayer games. But I would prefer a good one.

Even so, we should all keep in mind that -- no matter whom they get to program the game and how long they spend -- it will fall incredibly short of what a real-live human player can do. That's just the way it is, and if you were expecting a computer who can beat you on a regular basis without cheating... well... I just have to think your expectations are pretty unrealistic...

[art_vandelai]

I have a feeling that the AI will be somewhat more challenging, due to some of the changes in the game that limit alternative strategies by the players.

Let me explain it this way - the way the original game and CIV 2 was designed, you could adopt a strategy either to wage war and conquer, build large cities with lots of improvements (perfectionist), or expand relentlessly throughout the countryside (ICS). These types of strategies will no longer work to the degree they do in CIV 2, because of the mitigating effects of culture and resources which forces a successful player to adopt a more balanced strategy. Presumably, an AI that uses a well-balanced strategy throughout the game will be more challenging than in past CIV's.

The big question marks for the AI are how it handles diplomacy with the players and the other AI, and how well it makes use of the balanced strategy. What it seems people want, is that on the harder levels, it should excecute almost flawlessly. To come up with an optimal strategy for a game like Civ, would require some pretty serious hardware. Even a relatively simple game like Chess took IBM a few $billions to finally get Deep Blue to a level where it could compete with the top human players. I don't expect Firaxis would make the same type of investment, in a game where there are over 100 different types of "pieces" on a board that is FAR larger than 8X8, to compute a perfect strategy that can come close to human ingenuity would be far beyond anything within the scale of NASA, let alone a computer game manufacturer.

I would, however like to see realistic diplomatic behavior on the part of the AI, but not always the same every game for a particular civ. One game, the Americans would be willing to trade with everyone and make lots of offers to its neighbor civs, and in another game, they would be more isolationist and fight to get the resources they don't already have. And some real personality in the negotiations would be nice. What I would do is set up about 40 to 50 different "personality types" that would influence the way each CIV develops and negotiates, and randomize these amongst the AI civs.

And make them react more like real people. I know when I play, if someone sets foot in my territory, I defend it. Would the U.S. let a foreign military unit wander around its continent or its allies? No chance - so why shouldn't the AI do the same thing. If the AI is threatened militarily, it shouldn't keep building libraries - it should rush buy as much military as possible near the threats. Similarly, weaker civ's might strike military alliances with more powerful ones (like Kuwait/Saudi and the U.S) or try to buy the attacker with exclusive deals for luxuries/resources. Some desparation on the part of AI's to survive would be a necessary part to include in the game.

[Kautilya]

" if you were expecting a computer who can
beat you on a regular basis without cheating."
Personally I am just hoping for a relative improvement ie the AI should be significantly better than Civ2/SMAC rather than some absolute standard.

Actually , unlike a lot of people, I don't much mind cheating as long as it's implemented intelligently. The bottom line is the challenge that the AI gives you and I think you need both some good progamming and some cheating for a good challenge.

What is annoying is an AI like Civ 2 which cheats like hell but is still easy to beat almost every time at the highest level.

BTW I have to disagree with the idea that we know so little that we can only wait for the release and have absolutely no way of saying anything for now. We have enough information based on what Firaxis has given us to make a reasonably intelligent guess about what the game is going to be like. What they haven't told us is , I believe even more revealing.

I hope the people who say that Firaxis hasn't given us their best screenshots are right but IIRC a lot of the graphics engine is complete so I don't see why they wouldn't give us newer,better screenshots.

Even if the actual AI programming is not yet done, a decisions to make AI a special priority would probably be made much earlier and Firaxis could have easily told us that. When Firaxis has listed the things that they think this game will deliver they don't mention AI. I think it's reasonable to assume that there won't be a big improvement. Of course if I am wrong no one will be happier than I.

[Sabre2th]

I hope the people who say that Firaxis hasn't given us their best screenshots are right but IIRC a lot of the graphics engine is complete so I don't see why they wouldn't give us newer,better screenshots.

Again, we don't have any idea whether or not these graphics are finalized. Also keep in mind that the screenshots are shrunk down and compressed before posting. They might not be as bad as yin and others would have you believe.

Even if the actual AI programming is not yet done, a decisions to make AI a special priority would probably be made much earlier and Firaxis could have easily told us that. When Firaxis has listed the things that they think this game will deliver they don't mention AI. I think it's reasonable to assume that there won't be a big improvement. Of course if I am wrong no one will be happier than I.

Given the difficulty of programming a strong AI, I'm not surprised they aren't hyping it. If they have what they think is good and give it to experienced civers who eventually learn how to win, that's not great, of course, but not entirely Firaxis's fault. Experienced civers would probably eventually beat any AI that ANYONE could put out, unless they somehow come up with true artificial intelligence.

Think about it: People are yelling now about "bad" graphics, what would they say if Firaxis pushed the AI, which will eventually get beat anyway? Firaxis is in a no-win situation as far as the AI is concerned. No matter what they do, somebody will yell at them.

If you want a truly smart opponent, you'll simply have to play MP, it's the same with almost any game.

[Imran Siddiqui]

Question: What would you have them say?

That the AI is REAL good and hard? Would you believe that (I'm guessing no). So, what CAN they say?

[Kautilya]

"Question: What would you have them say? "
Simple: I would have them say that AI was a big priority and they would put quite a lot of effort into it and they hoped to make the AI significantly better than previous games (if not necessarily fabulous in an absolute sense).

That kind of relative improvment is very realistic IMO. If they wanted to I am sure Firaxis could make an AI which was significantly better than SMAC.

What I think will happen is that the AI will be roughly as challenging as Civ 2 and probably Civ1 as well . Like I said in my first post I think it's rather disappointing that ten years on when everything else is greatly improved a crucial aspect of the game is roughly the same. I think it's reasonably to expect some improvement each time.

[Wiglaf]

Question: What would you have them say?

That the AI is REAL good and hard? Would you believe that (I'm guessing no). So, what CAN they say?

He's got a point there...Kau, if they said that, we wouldn't really care, as it could very well be a ploy.

[Imran Siddiqui]

Simple: I would have them say that AI was a big priority and they would put quite a lot of effort into it and they hoped to make the AI significantly better than previous games (if not necessarily fabulous in an absolute sense).

You see how well that worked for Graphics, yah? .

And I doubt anyone would take it more than hype, as SMACed alluded too.

I don't think they can really please this forum with the AI until the game is actually released.

[Phutnote]

I think we haven't heard about the AI because they're probably still working on it. The changes they have introduced into the game are going to require some major alterations; more than the changes between Civ I and Civ II, or Civ II and SMAC.

1) Armies: Having armies means you have concentrated forces. The AI cannot surround all it's cities with a halo of troops now, it has to deploy concentrated forces between the human and it's own cities. So it has to deploy it's troops more intelligently or it'll be in trouble.

2) Gold to support armies: means that civ's likely won't be able to support armies as massive as the ones in Civ II. It won't be affordable. Again, the AI will have to use what it does have more intelligently to compete with the human player. All screenshots have shown civs running a national deficit so the economy is going to be harder to balance with armies.

3) Resources: the AI is going to need some kind of response programmed in if the human player gets control of all, say, the oil sites in the game, or all the uranium. How does the AI strike out if it can't build important offensive units? They'll probably get around this sort of stuff (searching out all over the map for important resources) by having the AI civs trade their resources freely and just expand normally. Otherwise, it'll be another thing to program into the general AI so it's competitive.

As for missing units, I think they break them down into categories: attack, defense, transport, so the AI really has no idea of the difference between a catapult and a howitzer: both perform the same function, as does a phalanx and the mech inf. But not having some of the better units could really hurt the AI's chances, especially if it doesn't realize it.

We don't really know much about the game at this point. No idea if Leaders, for example, are in or not (another new concept that will have to be balanced). With so much new stuff, and such a tight deadline, I'd guess even Firaxis doesn't know how the AI will really handle everything in the end.

Getting the AI to focus it's forces intelligently is an area where I imagine they'll have the most trouble. The AI has reasonably well before by guarding itself from every direction with units. Most games where the AI has to effectively chose between where to mass troops and where to leave vulnerable it does very, very poorly.

Of course, I'm not a programmer so I really haven't the faintest clue what I'm going on about.

Phutnote