People keep referring to the output of neural nets as the "average of the training data" but imho the deep value in neural nets and the interesting philosophical part is they are massive conditional probability engines whose results can change based on the variables you feed in.