The focus of neural networks is not just on present state uncertainty, but on uncertainty in how future events will unfold. By forming automated abstractions, machine learning engineers can create algorithms with better models for predicting the unpredictable.