Very interesting indeed. Suppose we replaced “4. simulate the opponent’s reaction to you” and instead had
Simulate the opponents reaction many times and learn the probability p with which they will defect.
If p < k then cooperate, else defect (for some constant k)
Game theory with machine learning, what do you think?
Maybe, but take care to terminate. If your program always starts out by simulating the opponent twice, it won’t even cooperate against a copy of itself, going into an infinite binary tree of recursion instead.
Very interesting indeed. Suppose we replaced “4. simulate the opponent’s reaction to you” and instead had
Simulate the opponents reaction many times and learn the probability p with which they will defect.
If p < k then cooperate, else defect (for some constant k)
Game theory with machine learning, what do you think?
Maybe, but take care to terminate. If your program always starts out by simulating the opponent twice, it won’t even cooperate against a copy of itself, going into an infinite binary tree of recursion instead.