In artificial neural networks, a typical neuron generally performs a simple summation of inputs. Using computational and electrophysiological data, the authors show that a single neuron predicts its future activity. Neurons that predict their own future responses are a potential mechanism for learning in the brain and neural networks.
