No BackProp, Please! How can a biological network learn deep representations without Backprop? This video shows that biologically plausible learning rules with a predictive component are able to achieve good representations of objects such that the usefulness of the representation increases over 6 layers.