This sounds about right though there's the element of cost-effectiveness so I'll give'em some slack and just wish, hope & all that sort things that the AI would be better, much better.
This doesn't make any sense. Firaxis are not first giving bonuses to AI and then dumbing it down for balance sake but vice versa. They'll make an AI they're happy with and then comes the hand outs. If the AI would be competitive without bonuses it would be bloody awesome and make the game vastly better - we could've have a deity+n levels without the current stupidities and on easier levels humans would be compensated; a win-win situation really.
Speaking as an AI programmer, and as I've said before on these forums, it would be extremely hard to make the AI good enough at Civ to compete with the best players on an even footing. Now, there are certainly silly little things they can fix, and you'll find me complaining about them as much as anyone else, but there are 3 key areas that are difficult to fix:
1) Long-term planning
2) Military strategy (especially troop movement)
3) Knowing when to drop everything else and focus on one thing
Of the three, #1 is IMHO actually the easiest to make better than it currently is, but the hardest to truly "fix".
The analogy is chess. Chess is completely predictable with enough computing power. There are no random elements. One person moves, then the other, until someone wins or a stalemate occurs. Each person has only 16 units they can move, there are only 6 unit types, and the board has only 64 tiles.
Even with this seemingly simple set of requirements, it takes a super-computer to beat the world's best human players. If you added just one random element to the above equation, it would be very difficult for any computer existing today to beat a human at Chess. If you made the board slightly bigger or added more pieces, it would be even more difficult. The game of Go is more difficult for a computer than chess simply because of the possible number of moves, even though what you can actually do is limited to "place a piece".
Civ is MANY times more complicated than Go or Chess, has a bigger board, way more possible unit types, way more pieces (than chess at least), and adds a large number of random factors, and instead of 1 opponent, you have SEVEN.
Now add in multiple victory conditions, barbarians... To put it in perspective, to make a computer unbeatable at Civ would require solving multiple NP-complete problems, (which is fundamentally impossible without quantum computing or the like) and just the number of possible moves alone would require the AI spend an amount of computing time on its turn that would mean you'd die of old age waiting. Literally.
In fact, I'm pretty sure the AI would be smarter *right now* if the computer was allowed to spend more time on its turn thinking. But one of the #1 complaints among players is turn times... You can't have it both ways, folks. ;-)
Luckily, the computer doesn't have to be perfect. ;-)
So, lest my constant complaining sound like I'm not giving props where they're due, the programmers at Firaxis have done a pretty good job by just making the AI not completely hapless. There are definitely things they can do, but they're actually quite challenging.
#1: Long-term strategy. Every time a patch/expansion changes the ruleset and problem space, it takes the best civ players in the world a fair amount of time to come up with a new "semi-optimal" strategy. It would take the CIV AI programmers a similar amount of time to come up with AI strategies that are as effective.
So, when Tabarnak posts a 3-city tradition food caravan strategy, this is an example of something the AI team ideally would be doing on their own internally. Ideally, they would be at least coming up with semi-optimal peaceful tall strategies for the AI. These are the easiest to come up with because there aren't as many variables as there are in conquest. Defense is easier than offense. Your empire is smaller and more manageable, etc.
However, they need a *stable build* to come up with and test these strategies. If balance is still being tweaked, and the game is totally buggy, they can't effectively run simulations to see how their AI strategy is faring. It's entirely possible they're doing exactly what I describe... although I imagine it's more likely they're just crunching to fix bugs for most of the time when the game balance is stable enough to come up with strategies. ;-)
And, I don't work at Firaxis, but I can pretty much guarantee from experience that the build isn't stable until shortly after -er I mean
before they send it off for production. *cough* launch-day patch *cough* ;-)
So that means the AI programmers really have no hope of developing long-term strategies for the AI until AFTER launch. And the first patch is fixing exploits and balance and bugs, so really, it's not until after the first patch. Which is why, IMHO, you see so many AI fixes come in the first and second patch.
But, nonetheless, I'll keep coming back to what they *could do*.
Imagine if the AI went all-out science: Rushing libraries, NC, Universities...
This would make Deity insanely hard. The AI already techs significantly faster than the player, grows faster, builds faster, buys cheaper, can support more units, etc.
Any strategy the player uses would be *more* effective in the hands of the AI on Deity, or on any level above Prince, for that matter. And it would make every difficulty level harder. Which is what you want as a starting point.
Trust me when I tell you it's so much nicer to be in a position where you're dumbing down your AI to make a difficulty level easier, rather than trying to make it harder. For example, in an FPS, there's no reason an AI can't always have perfect aim, and complete foreknowledge of where the player is. You can start with that and intentionally make the AI less effective than that to match the difficulty level. The AI has access to similar information in Civ, IMHO... they know where your troops are, and pretend not to. However, this information doesn't give them the same advantage that perfect aim in an FPS does.
None of what I'm saying addresses the troop movement/military strategy issues in CIV, but that's a much harder problem to solve, and a whole separate rant.
What I would really like to see, as a player, is just the AI attempting to follow long-term strategies that work. I know they *have them*, because there are variables in the XML that seem to control how much they focus on their long-term plan, how likely they are to abandon it when things aren't working, how much priority they give it over defense, etc. etc.
But I'm pretty sure their long-term plan isn't very good, because if it were, in any situation where the AI's neighbors are playing peaceful, the AI would severely out-tech the player on Deity. Without interference, they should be able to have universities up SUPER fast, especially with the GL... they should be able to complete the rationalism policy track and have research labs by turn 150.
Now, that's a flavor issue, of course, but it's just an example. The AI really reminds me of a novice player. Building things they don't need, just because they can. They never seem to be really sure why they're doing what they're doing. Why is the AI building Wonders it doesn't need instead of focusing on tech? It's all very novice-level behavior.
To be fair, for a very long time I was not good enough at Civ to know what I should be building and why. I used to follow the build recommendations because there was just too much to think about. I didn't micromanage tiles. So maybe for most Civ players, and most difficulty levels, it doesn't matter that the AI plays like a novice. Maybe it only matters to that tiny vocal minority of Deity players. But to us, it matters a lot.