I remember the AI in GC2 very well; I played that game a lot (really a lot) until 3 came out.

The AI tended to behave like General McClellan - its campaigns were plodding and mostly consisted of roaming fleets knocking out planetary defenders and wanting the skies to stay clear for a turn or two before the transports were wheeled up. I was almost always able to 'tech rush' my way out of trouble and I recall lots of battles where my fleet would vaporize wave after wave of enemy fleets, taking no damage in the process.

I'd say it was reasonably challenging, enjoyable and sometimes irritating. I was almost always able to 'tech rush' my way out of trouble and I recall lots of battles where my fleet would vaporize wave after wave of enemy fleets, taking no damage in the process. My early games were usually very tight... honestly I think the AI went downhill in the last installment, but as I understand it that was due to circumstances beyond Stardock's control.

Compared to any of the Civ iterations, the GC2 AI was an Einstein.

I think "brain dead" is something you can always write about any AI in a computer game, when you finally identified the loopholes of the game mechanics/AI. I think we all agree that no matter how many effort Stardock/Brad will put into the AI, at some point loopholes will be found, demonstrated and than you can call the AI brain dead...

I have been playing strategy games since the beginning of the 1990s, when I got  my first PC (386-DX40). I started with Warlords, Dune2000, Civilzation etc. Up to now, GCII (with all expansions) is unchallenged in terms of AI competency IMHO. In fact, the main reason, I became elite founder was the prospect to get (one) of the best AI in 4x games.

For example: if you watch experienced people play Civ5 Deity games (such as Marbozir), basically, the trick is to "play" the system. I.e. you can just outmanoeuver  tactical AI with ranged units (and AI is not able to move and then attack with ranged units in the same turn). So, surely brain dead! Is it? In fact, I am sure 90% of the gamers are fairly challenged on Emperor difficulty already. I consider Civ5 AI in a fairly good state for what can be expected today. It is as simple as that. (And in fact, I play Emperor, because I just do not enjoy the constant struggle for optimisation... And I like to get a wonder from time to time... )

I read somewhere that you targeted a "C" for release. I think this has been achieved (counting ongoing AI bug fixing). Me, personally, I would really love to see it being further developed. If we can get a B+, that would be great. That being said, I think the question you ask is clearly justified in terms of focussing effort and where to spend resources. Not an easy decision to make! If you focus on the things perceived by the vast majority of the player base, you will still have all of the 4X "cracks" posting in the forums about the bad AI. The average player comes around, sees it and gets a wrong impression. If you spend a lot of time to  satisfy the "cracks", most probably 90% of the players won't even notice.

I am mostly excited about this interactive AI that was mentioned, learning from the strategy of human players. I haven't heard too much of that lately? How is that coming along? 

the AI in GC2 was good, imo. comparing the AI in GC2 or Fallen Enchantress to other 4X games alsways felt unfair. the AI in Civ 4 was fairly strong, too. but not because it's a good AI but because the game desin was more "AI-friendly".

custom unit designs is a very big factor. if you give the players such a powerful tool, that's not going to end well when your AI can't use the designer as efficient as a human. Civ 4 or other games just have a bunch of premade units and the human player doesn't get a big advantage just from crapping the default design and using something that is twice as powerful for the same price.

so you put in a lot of effort to make your AI really smart (compared to other game AI) but in the end the game isn't more challenging because you also have a ton of options in the game that are largely "human only" and can be used by the human player to beat the AI. other games take an easier route.

so i guess the real question should be if it's more economical to simplify the design so the AI can make better use of it, or to try and make the AI smarter at using the many many options available. simper game design is the easier route to a similar level of challenge, tbh.

i'd say AI development is mostly a "fan service". AI is already good enough for the casual gamers who probably make up the majority of your sales (not going to speculate on numbers, but i'd be surprised if the majority of players ever played on a difficulty hgher than "normal")

GC2 AI is the benchmark by which I measure a decent AI tbh. It's rarely surpassed. Of course, it is still just an AI and so a little plodding, but I'd be have rather surprised if Stardock had produced Machine Sentience instead. 

The GC3 AI isn't dreadfully bright yet - if it can just get to grips with the adjacency system it'd be a big improvement. But at this stage in the game's life cycle there's whole mechanics which might be changed between patches; it'd be pointless teaching the AI to play systems which may change completely month-on-month.

I think the OPs question was designed to get some interesting feedback and therefore was sufficient, but I would have phrased it differently:    Was the AI in GC2 good enough to provide an enjoyable experience for the gamer?  That's what a game is supposed to provide, in my view.

Suppose one could create an AI that was as good in GalCiv as Deep Blue is at chess.  Well, that would be interesting to see, but not much use to me as a gamer trying to have an enjoyable experience.   I would just be seeing how long I could last before I got crushed.   A game is an entertainment vehicle, not a vehicle for measuring one's tolerance to humiliation.  

So was the AI in GC2 good enough to provide me with entertainment?   I would say definitely yes, though hardly perfect.  It certainly was at least as good as in any other 4x I played, and I have played a lot of them.  I will break it down into what I think are the key components.

Does the AI build itself up sensibly?  Fairly well, but the Community Update currently in play was prompted largely (if I am not mistaken) by the problem that the AI (in TotA) was not using its tech tree well. or perhaps I should rephrase and say it couldn't use its tech tree well.   My experience suggests that it does much better in the CU.

Does the environment one plays in seem alive?  That is, are the AI doing things that appear sensible, do the AI factions interact with each other, making friends and enemies, fighting wars or making alliances?   It seemed that way to me, but one would have to go into debug mode and study to really tell.   But all I care from an enjoyment standpoint is whether it "feels" ok to me, and I would say, yes, it did.  Some have said they want the AI to recognize who the human player is and play against him.   That doesn't feel right at all to me.  They should behave "normally" according to their traits and respond to objective threats and opportunities, not automatically dogpile on the player.  That would feel very wrong, but its just me.

Here I need to comment about the constant discussion about the AI "cheating".   I have a different view I suppose --- what I care about is what do I see and how does it feel?  What's under the hood, well, that's for the garage mechanics and engineers to argue about.   How does the car feel on the road is all I care about while playing.    There are some sorts of AI operations that don't feel right and look artificial --- for example, units suddenly spawning from nowhere, clearly without having to be built or moved into position, etc.   That's bad, but not because the AI is "cheating"  but because it "feels" wrong.  To me, anyway.  Didn't see that in GC2. 

I remember a game of Civ4 in which I thought I was doing ok against the Chinese, but after I penetrated their frontier I was faced with a stack of several dozen units.  I just about fell off my chair.   It was mid-game and though it was a very nasty surprise, it didn't feel like an exploit.  After taking a deep breath and doing some thinking I managed to figure out a way to wear down that stack and eventually I won the game.   Question is, were they able to build this stack because they were "cheating" or was it just good organizational planning, etc,?   Don't know, don't care - except in an academic sense.  It was sufficient that it seemed plausible at the time.   I think at that difficulty level they had some production advantages, etc., which one could look up, but I didn't.    I had a fun game, though I was "moderately" stressed for a  while.  I don't play a game to analyze its mechanics, even if that may be interesting in another context.

Does it avoid immersion breaking mistakes, of omission or commission?   I think GC2 was fairly opportunistic if I left an "opening".   Current example:  A few weeks ago I was about to attack an AI colony in GC3 but stupidly left a transport unattended near a  planet that was defended.   I think I had thought that it wasn't, and wanted to get the transport close enough to invade the next turn.  Oh well.  The AI could have sent out his fleet and destroyed the transport, but just watched.   His fleet could never have beaten my main fleet but he could have delayed the invasion many turns.  (Beta 4 I guess, not sure).   That's the sort of thing that the AI should jump on; I don't remember getting away with that kind of sloppiness in GC2.   And I am not sure I could get away with it now in 1.03.

Does the AI organize attacks with a purpose and try to carry them out?    Yes, I think so in GC2, though most times they are not the best plans, that's not critical.   A bad plan is better than no plan.   Heck, I would say most casual players don't often actually compose a plan of any depth.   That's the first thing I see, say, on Civ Fanatics as to how to raise your game level.   Make plans.   Why is Marbozir so good?   His plans, made very quickly it seems in his let's plays, are quite deep and generally spot on.

And lastly and most difficult, does the AI, seeing a threat, organize against the threat by trying to use cooperation with other AIs or with the player?  That's a tough one.    I am not sure that really worked so well, but I don't have a clear picture.   What I do know because it is more recent is that  in GC3 I have been asked by the AI many times to declare war on someone.  This points in the right direction, but so far I have had the feeling that the AI really isn't as threatened as it claimed and that it doesn't follow through cooperatively against its "enemy".  Rather, I wind up conquering the opponent, while the original AI makes peace!   Well, it has a ways to go.

One last overall comment about AI.   A lot of the discussion seems to revolve around how difficult the AI makes the game for the human player.  I translate this to mean, how hard is it to win, and how much "help" does the AI need to make it challenging for the human player to win?    Well, as in RPGs, there is another point of view.    Does the game have to be about winning and nothing else?   You can play Skyrim to finish the main quest, or  you can role play, bringing an entirely different mindset, and with the objective being the journey and there not needing to be a destination at all.   Heck, the Metaverse is broken anyway because of the blatant exploits people are using to run up scores, so who cares?   I posted earlier about bringing some more role playing types of play into GC3, but of course, as always, that is just me, and the branch I am sitting out on seems to generally be pretty thin and pretty lonely.

My 3c.   I am charging more because I am on a fixed income.

The Ai in GCII was excellent. It was indeed ahead of its time. As stated many times some posters fail to see the forest for the trees in the way. At the time and even now, it offered a player a challenging game which not many other 4x games can still claim. You were hampered by the fact it was hard coded so the amount of  work you could do was limited by putting in hundreds of hours of code rewrite, something nobody wants to do.

For GCIII Brad, please do not get discouraged by naysayers. This is your game and the brilliance of the current AI shows through it. I do hope you take time to work on it 'continuously' for the next several years ahead!  This time around you have the 'api hooks' in place and the coding is not as set in stone as GCII was.

This is of course correct. On the other hand, I think the influence or significance of the "cracks" (no negative pun intended, maybe "enthusiasts" would be the better word?) is overestimated. How many people are registered to these forums and actively post in it? Compare this to the sales numbers (which I do not know exactly of course, just see the Steam stats). The same goes for e.g. Civ5: It is still one of the most played games on Steam, with e.g. 40.000 active players today! If you go to the Civfanatics forum (the place to be for Civ and 4x fans), you will find that the crowd there (me included) is not very representative. The number of active people there is insignificant compared to the total player base.

This is how I personally interpreted Brad's question. He is looking for the right decision on how to invest resources. (I also seem to perceive a slight hint of frustration or burn out in that post, but maybe that's just me. If it is the case: Absolutely unjustified!  )

Let me help you here a bit. In GC2 there were a few mechanisms at large that could do this. Most of this was based against the leading faction in military might.

<Phrase Morality="0">
<Text>We wanted to inform you that we have decided to go to war against the [AGGRESSORNAME]. Their brutal and unprovoked attack against the [DEFENDERSNAME] had to be responded to. Let this be a warning to all civilizations who would needlessly go to war.#We felt that we should inform you that the [DEFENDERSNAME] have been ruthlessly attacked by the [AGGRESSORNAME]. To stop such warmongering, we have decided to put together a force to deal with the [AGGRESSORNAME]. Feel free to join in.</Text>
</Phrase>
<Phrase Morality="1">
<Text>The [DEFENDERSNAME] are under attack by the [AGGRESSORNAME]. In the name of galactic stability, we have decided to come to the aid of the [DEFENDERSNAME]. We felt we should allow you to be aware of this in case you wanted to join in.#The [DEFENDERSNAME] are the victims of a war of aggression by the [AGGRESSORNAME]. We have decided to intervene in the name of peace.</Text>
</Phrase>
<Phrase Morality="2">
 <Text>The [AGGRESSORNAME] are our kind of scum. The [DEFENDERSNAME] are weak and ripe for destruction. However, it is not in our interest that we let them go around conquering our food, err I mean our neighbors. So we have decided to take this opportunity to remove the [AGGRESSORNAME] from the galaxy. Feel free to join in the fun.#As much as we approve of the weak being wiped out, the [AGGRESSORNAME] are becoming too much of a threat. Their war with the [DEFENDERSNAME] is the perfect opportunity to deal with them.#The [AGGRESSORNAME] are beating us to the punch at wiping out weak civilizations such as the [DEFENDERSNAME]. But we cannot afford them to take over the entire galaxy so we have gone to war. You should join us in this task.</Text>

Usually the AI is reluctant to go to war versus a stronger civ, but in this case they would do this and the delivered message is an example to collaborate against it by combining power.

It also could happen on a larger scale via a mega event:

[Alliance] During a secret meeting, it was decided that the %s menace has grown too great. They have concluded that the only way to stop them is to work together both openly and covertly. The war has expanded.

Finally, there was the Altarians SA that intended to unite all good civs against their attackers. With the Drath also ingame, who made everybody attack everybody, that process got accelerated on grand scale and would even throw the Drath themselves into this if they made someone attack the Altarians^^

I haven't played GCII in a long time, pretty sure it was Ultimate Edition I played though. I remember the AI as being functional in most aspects at the very least, with no glaring errors for its time. I do remember loving the game! So much so that when I discovered GC3 was in Beta 4, somehow I managed to not hear about it at all until then, I insta-bought a copy.

I seem to be one of the rare players who prefers to play the soft game. I like to win by any means but conquest and I will never declare war, just win them. With that in mind the AI was engaging, frustrating in the right ways, and generally enjoyable to play against. I do know that it was a vast improvement on the closest, and much older game, I'd played before it, Birth of the Federation. With BotF it was a steamroll for me or the AI and I could never get anything approaching a fun challenging game. With GCII I must have played for several months before moving on, and that was because the game continue to be fun even as I played around with the setup settings to be wildly different.

The one thing that I've not been happy with in any game I've ever played from any developer, is diplomacy. It just never feels right and is always easy to game. Maybe that comes down to my obsessive need to make optimal strategies and the AI won't ever be able to compete with a dedicated human in this area. I want it to feel like there's a real give and take, I want the AI to propose a trade for once that isn't so bad I don't laugh in their face. Some wildcards occasionally are okay, but in general it needs to be much better.

With all that said GCIII is a great game, and it's getting better.

I can never meet you in person Larsenex. I have these vision of a deep, rich voice expounding wisdom. Anything less and my life would shatter...

GC II had decent AI, but it was still about speed in the end, once you got on top, it didn't have the ability to surprise you.

I felt the AI on galciv2 was very good.  It is Stardock's AI (particularly Brad's love and care for such) that sets the galciv series apart from other space based 4x games.  If I did not expect galciv3 to have at least as good if not better AI then I would not have near as interested in the game.  My time is fragmented and inconsistent so playing multiplayer is not really an option and so a entertaining AI is king for me.

TOTA introduced alot of new stuff (mostly from new technologies, example: Atlas modules, stuff that enhances your attack/defense fleet- or shipwide) and the AI was never coded to use it. In that particular case the ship designer would totally ignore any of these modules, the combat mechanics did prioritize any ship having such a module as a first target, making them very weak overall even for the player.

The usage of race-specific buildings & the researchpattern adapted to the unique techtrees is (mostly) fixed with the Community Update.