When considering overfitting, which is widely seen in the field of machine learning, it is important to say that ``an AI created for a specific goal cannot be trained with the goal itself, so a 'proxy ...