It is acknowledged by Firaxis that the one AI "cheat" is that the AI "knows" the whole map, but it is unclear exactly what "knowing" the map means.
I hate to throw a small wrinkle into the discussion, but I'm not entirely convinced that the AI knows which resource will appear where, nor am I convinced that it places a tremendously high value on "known" future resources. In other words, I believe that the AI knows that a particular tile is very valuable, but doesn't know that it is valuable because of an iron, coal, or uranium resource will some day appear.
My theory is that the AI assigns a value to each tile on the map, and seeks out high value tiles. Resources (even as yet unknown resources) constitute one of the factors in tile value, but not the end-all be-all.
I came to this tentative conclusion after reading about an editor test originally done by TheNiceOne, which I reproduced with variations on my own. Here's how I conducted the test: I built a map consisting of two islands - one small for me and one nearby and somewhat larger for the AI. The AI's island was all grassland and was covered in RR for ease of movement. I started the AI in the center of its larger island with 2 settlers. The AI built a city in the center location, but then had to choose where next to place its second city. At each corner of AI island, I included a fish and a whale, so that the AI would be drawn to the corners. At this point, all corners were "equal" with the same proportion of shielded grassland and fish/whales and so all represented equally attractive city sites. Now, for each quick test, I placed a future resource like iron, coal, rubber, uranium, etc. at different corners. I also modified corners so that the corner with iron might not have a whale or might have no shielded grassland, etc. I don't recall the full and exact results of the testing, but the principal noteworthy (to me) results included: (1) sometimes better terrain and bonus resources (fish/whales) trumped any strategic resource every time (i.e., as an example, perhaps shielded grassland w/o iron was viewed more favorably than plain grassland with iron in the city radius); (2) the AI either didn't know which resource would later appear, or valued the reource in an odd fashion -- i.e., one would think iron, a critcal resource available very early in the game, would be more valuable than uranium or aluminum -- both important but not available for thousands of years (assuming survival ).
So, in the end, I concluded for myself that the AI knows which tiles are valuable, but either doesn't understand why they are valuable or doesn't value them as a human player might with the same foreknowledge.
I also understood that it was coding, processor and memory challenges that created this cheat, rather than any desire on the part of the developers to give the AI an artificial cheat.
