I've just learned of a concept called 'Deep Learning', which in short has been used to create computers and AIs with immense capabilities. In January, a robot was taught to play Go, an incredibly complex game, and managed to beat a grandmaster five games in a row (link here). If you're too lazy to open the link, the way they achieved this was first inputting about 30 million moves into the robot played by a variety of experts, and then allowing the AI to teach itself how to beat that strategy, and how to beat that one, until it had a vast range of strategies to go on, and thus being able to beat even the greatest player in the world. This, in a nutshell, is what 'deep learning' means.

So, shouldn't this be possible with civ? If Firaxis gathered millions of moves professional players have made, and then creates an AI to come up with a strategy to beat those moves, and beat those ones, and over time it would have a much more 'human-like' response to things. For example, maybe the first 10 times the AI plays a game it embarks all it's land units into the ocean to reach an enemy city. Once it realises that this strategy results in it's units dying, and subsequently losing their defence and getting invaded, it will rethink the strategy until one fateful day the land unites actually stay on land (crazy, I know).

Obviously I'm no expert, but obviously neither is the civ AI, and I made this post in the hopes that maybe it could create a game where AIs could potentially not only make moves like a human, but then mould it's strategy to make it even better than one. And once that's done, we can have a Deity level AI that actually knows that a melee unit is needed to conquer a city, and one that won't need a gazillion bonuses to even keep up with a professional human player.