It seems to me that the computer makes a number of serious errors in overall strategy, particuarly in the research paths it pursues. For example, I have learned that there are innurerable benefits to going after Star Republic and Democracy early on, yet the computer rarely gets these before the game is over. Also, it seems that any Neutral Civ can benefit immensely from XEno Ethics, yet the computer researches Habit Improvement instead.
In some sense I wouldn't want much better AI, since giving the computer decent AI AND production bonuses would make the game all but unwinnable, but still, having competent AI would make the game feel so much more real. Any one else notice some easily correctable errors in AI judgement? At a low level I'd say it's pretty good, most of the problems I see are in overall strategy. I already mentioned the number of unescorted transports and weaponless fighters...