I appreciate the opportunity to present a general framework for modeling neurobiological systems that grew out of a 6 month sabbatical at Caltech with John Hopfield in 1984. At that time he asked how large ensembles of highly nonlinear, noisy neurons could carry out the complex computations our brains are capable of. With the support of David Van Essen, and many collaborators and students over the following 20 years, a simple answer emerged that is described in my book with Chris Eliasmith “Neural Engineering”, MIT press 2003. The short answer to Hopfield’s question is: one achieves precise computation through averaging over large numbers of neurons, i.e. highly redundant population codes. This talk outlines the basic principles of our framework and concludes with a brief description of the most important missing component, learning. I strongly believe the issue of learning rules for population codes is tightly coupled to the development of a systems level description of cortical circuits.