Yes... Sea invasion was quite , but absorbing your cities? LOL try neglecting culture and you'll see... Even at lower diff levels! Probably you where just playing it right and the AI didn't manage to do that!

Now, again, my point was never wich AI is the best, I was pointing out how easily Civ was dismissed as simple, while I feel that, ATM, Eemental is much simpler.

And maybe I'm overestimating Civ's complexity, but I surely fail to see Elemental complexity, since at the moment REX---->Steamroll is all the strategy we have!

What? Because you can't rush the patched Civ4 AI [@normal diff - which is not getting any bonuses] you say?

Well since Elemental AI will evidently be so easy to create, I'm looking forward to playing several community AI mods once we're able to really get at the guts of the AI

I'm looking forward to AI modding myself, but I don't think it will be "easy" to get Elemental's AI to play well (as in, like a quality and diverse human player with as minimal cheating as possible) no matter how "simple" a game Elemental evidently is.

Go ahead and try killing 9 AIs on normal difficulty inside of 200 turns on a map of comparable size to Elemental. The point is not about how smart the AI is, it's that there is so much going on in the game, even against a dumb AI it will take much longer to achieve a victory condition (and the BTS normal AI is smart enough to make this quite challenging). With Civ your strategy has to change in each game, based on the environment you start in. With Elemental, you can follow the same rules every time with some minor adjustment and always win.

Civ AI needs to worry about tech, diplomacy, religion, culture expansion taking out resources, strategic resources, goody huts (yes, they have them too), quests (there are some in Civ IV although I'm not sure the ai can get them).

The real difference is that EWOM has tactical battles, and that's something immensely complex which Civ IV didn't have.

As for Brad's track record, Galciv (and Galciv2) definitely had great ai, but these games were even simpler than Civ. For instance, you need zero map analysis in space, and so there are basically no chokepoints. Now in Elemental or Civ, you have impassable mountains, and thus mountain passes to handle for instance. So even with his track record, there are many tactical difficulties in EWOM that he hasn't coped with in Galciv. I'm still confident he can provide a good ai in the long run.

Haha I remember a game on one of the higher AI settings in Galciv where the computer gifted me a whole bunch of war ships and then declared war on me a turn later, allowing me to wreck havoc on their interior space with the war ships they had just given me. The AI in Galciv is overall quite good, but it's certainly capable of doing some humorously idiotic things at times!

I think tactical simplicity is a key part of AI management in many cases. AIs can handle a complex economy easily, in part because they can compensate by cheating. If the computers economy lags behind by 50% on average due to poor decision making you can always just give them a bonus to make up for it, and the player will never necesarily know about it. But that doesn't work in combat, where the AIs stupid mistakes are blatantly obvious. I think it's the whole moving one space at a time thing on a typical civilzation style map that really leads to room to problems, because there are so many options there. I've always appreciated games that limited that kind of movement somewhat, by dividing the world up into regions or star systems or something. Imperialism 2 was a classic game in this regard in my mind because it divided the world up into regions for strategic movement, but each region was still a big complex area of spaces that could be develeoped for economic purposes.

The link is to a lecture on, not so much the nuts and bolts but overall objective of AI. In that regard, the two are similar, and I'd prefer a similar approach in Elemental; that is, each faction and sovereign act in a manner that seems reasonable for that nation and character.

The axis drawn up betwenn 'Good' and 'Fun' AI is interesting and, while not exactly news, was still a good description.

It is silly to make this a discussion about which is more complex, Civ4 and E:WoM - complexity in itself is not a goal.

As for Spore - it's actually a cool game, I like creating those creatures, but the second half was kind of worthless. Unfortunately, that's the part he worked at.

Personally, I think that Elemental is kind of game for which is much more difficult to make good AI then civ.

You have both tactical and strategic battles, unit customizing system that is much less forgiving to bad designs (compared to promoting some unit in Civ), plus Civ4 is made so that it's game concepts are AI friendly. And questing and creature economy are also concepts difficult to train AI for. Not to mention it's difficult to train AI to place buildings properly, without having cities block each other in growth. Or use spells on strategic level.

There are just so many places where player can get edge over AI.

As for Soren vs Brad AI discussion, let's just say that GalCiv2 AI is on Civ3 quality level (which IMO is not bad), but not up par Civ4 level (which is great).

P.S.

Master of Magic had one of the worst AIs for 4X game. Game was too complex and multi-layered for its time.

A game can be complicated with many options, But if the options are not "real options" in the sense that each option has pros and cons it doesn't matter.

The problem with EWOM is while there are many complicated features (design your own unit anyone?), but the choice is so obvious or doesn't matter, it's easy to create a "good" AI.

Those obvious choices are only "obvious" due to poor balancing. Having AI that knows how to exploit them would not make a good game, IMHO.

Thanks for the link. Another great AI system is the one in AI War which layers AI to create complexity; each unit has simple AI, each army has its own AI (handling army level issues like giving orders), then there's sector AI, empire AI, and a few others. Civ5 does something similar where there are multiple AI systems that interpret the orders from higher up to create intelligence. AI War takes it a step further though in that it doesn't use state-machines for it's decision making (as much as a practical limitation as anything else as there are hundreds of different units and usually thousands of units in any given game).

I think all of these people (Brad, Soren, the guy doing AI for AI war, and the guy doing AI for Civ5) would have lovely lovely talks. With their powers combined... all hail our robot overlords!

That means nothing. AI for every civilization game had to be written from scratch, since each of them introduced new gameplay elements. Also overall design has changed significantly since Civ1 and games have become increasingly complex. On top of that, those AIs were written by different people, I don't think even 2 civilization games had same lead designer and AI coder. There's was 5 years between every civilization game, it's not like they keep using the same engine and copy pasting AI code. Soren did a great job with AI in Civ4 and I hope Elemental will reach that one day.

Civ IV's AI didn't really get truly solid til the last patch of BTS, which incorporated a lot of AI fixes/changes from a few people in the community.  I'm not sure Soren was even around at that point.  The earlier Civ IV AI wasn't that great.  So...

GalCiv2 has some nice takes on AI but I'm not sure I'd call it the best ever written.

In all of these types of games AI is usually the weakest element.  The more complex the games are, the toughter it is to create AI to come remotely close to player behaviors.  This is one reason I'm hoping for stronger AI in Civ 5 - they've stripped out a lot of what I call "elements of cheese" (things no AI will ever do as well as a player and that players often exploit to take advantage of weak AI rather than strategize), and refined game systems like 1UPT and clearer rules of engagement can lead to tighter AI (coding to rules is a lot easier than trying to make an AI think in broad terms).