The AI in a games like CTP-2/SMAC/CIV-3, is like an automobile there the programmers must try to simulate two specific human player abilities. Basically "steering-capability" and "thrust-capability".
"Steering" is the hardest part (and you cannot expect something ridicules here, like a "pass a turing-test AI", or anything like that). "Steering" deals with anything that cannot easily be dealt with through templates or thumb-rules, partly because everything changes in often unpredictable ways, from turn to turn. For example: moving units and some complex diplomacy-related things. This together with the overal complexity of CIV/CTP-games (compare with chess: only 64 squares/only 6 units/few and distinct rules), makes any wishes for something "almost human" very unrealistic.
HOWEVER (and this is important):
"Thrust" is the easier part: I am talking about strategical/logistical choices that can be dealt with through lead-guided template-changing AI city-areas, city- and unit-improvement editable paths, editable tech-tree paths (wich includes AI-civs that automatically capitalises on any city-/unit-upgrade that a newly achieved technology has to offer, according to editable templates), and also methods leading to more effective AI city-placements; like pre-designated invisible AI city-placements (which nevertheless let AI-civs start and expand, either unpredictably from any pre-designated point, in the main-game, or perhaps according to editable scripts, in scenarios).
These choices (perhaps also in any other strat/log areas) can to a large degree, be lead-guided through templates and thumb-rules provided that the game, as much as reasonably possible, is designed in such ways that extensive overal strategical/logistical lead-guiding, through player-editable templates becomes viable.
Finally: also more "thrust" by implementing ideas that lays the burden mainly on human players, like Anti-ICS (Infinite-City-Sprawl. CTP-2 is NOT a good solution to that), Anti-BAB (Bigger-Always-Better, which lead to the Eternal China Syndrome), Rise-and-Fall, some nifty gradually added uphill-struggle world-conquer parameters + more.
The bottom line is, that its really IS possible to design games like Civ-3/CTP-2 with almost veteran-player thrust-capability (without AI-cheating), provided (again) that the game, from the ground up, is designed in such ways that extensive overal strategical/logistical lead-guiding, through player-editable templates, really becomes viable.
Instead the AI follows blindly all these strategical/logistical templates until something triggers it to temporarily take over direct command.
You can compare it with a blind man being given exact blind-read instructions how to move around inside an unknown building. He can follow this instructions meticulously and rather effectively. If he never the less bump into something - he temporally put his blind-read map in his pocket, and starts to feel ahead and analyze the situation. Soon feeling confident, he can take the map up again.
To summorize it all:
Any development-team that works on complex TBS-projects like CTP-2/CIV-3 must try to free up the AI as much as possible, in order to let it work more efficiently on fewer selected tasks. The more the AI is confronted with buckloads of confusing choices, that its suppose to "choose wisely" from; the more its likely to buckle down (and screw up).
[This message has been edited by Ralf (edited January 09, 2001).]
Apolyton Civilization Site Forums
I just hope that the Firaxis-team understands this better. 
Also, anyone seen The Limey? I think he'd enjoy this post, too.
