This chapter collects two papers: one is the seminal paper by McCulloch and Pitts (1943), where the authors decided to explore the question ``what can a neural network compute?'' in terms of a very idealized model of the brain which was in turn based on a very simplified model of neuron: a threshold unit. The second paper, by Minsky (1967), discusses in detail the implementation of finite-state machinesin terms of McCulloch-Pitts neurons.