Let's say, hypothetically, that you had a model that did not need to re-train in order to incorporate new information in its weights. What would you do with such a model?