Approximation of some AI problems