Without one bend, 100 layers act like one

Why every neural network needs an activation function, explained with two simple equations