^ WL is NOT a solvable game , hence a "perfect" AI is by definition impossible.

I don't think you understand what Sze was saying. All games are solvable in terms of a Nash Equilibrium, but that doesn't mean that the perfect strategy will win every game. It just means that it will win more often than not against any non-optimal strategy.

I do see the argument from game theory, but lets assume that someone makes a "perfect" AI that never has a less then 50% score.

Would it not then be possible to make a "even more perfect" AI that knows everything about the 1st AI, thus being able to beat it? Hence the 1st AI must be imperfect.

Additionally due to no-luck-cycle, you could always make a faster AI, and hence always make a "yet even more perfect" AI, making all previous AI's inferior.

Since the AI can always be improved, it is never perfect. At least that is how I see it.


You could take this a step further, and argue that you will end up with an AI that humans can no longer improve. Well, what if someone else, a non-human being, can improve it even more? And so I would argue, things would go on.

Edited 4/18/2016 03:29:00

I don't understand, a nash equilibrium assumes perfect information, in a nash equilibrium, you know your opponents moves, and they know your moves, yet neither of you has a better option as long as the opponent keeps their orders. This wouldn't happen in warlight, firstly because you don't know your opponents orders, and if you did, you would not be able to find a nash equilibrium, imagine you have a double border on your opponent, who has equal income, if your opponent can see your planned attacks, he can defend them, if you see your opponents defense, you can break him, therefore, it is impossible to reach a nash eqaulibrium

It is an equilibrium in terms of mixed strategies, meaning that for every position the "optimal" strategy is to actually play strategy A a certain % of the time, strategy B a certain % etc. It might seem very abstract, but when you really think about it and try to incorporate that into your own play, your play might be less dogmatic. As I said before, computing any kind of equilibrium like that would be insanely hard for most games, best players are relatively close in terms of play to an abstract like that.