If we still do not know exactly how artificial intelligence doses the weights in learning within the hidden layers, can we really predict its evolution?