Carver Mead published “Neuromorphic Electronic Systems” in the Proceedings of the IEEE in October 1990 (Vol. 78, pages 1629-1636). The paper named and framed a field: neuromorphic engineering, the design of electronic systems whose architecture imitates the nervous system.
Mead’s argument started from energy. He pointed out that biological nervous systems perform perception and control tasks with a power budget millions of times smaller than digital computers attempting the same work. He attributed much of this gap to the digital approach of representing everything in binary and recomputing it with general-purpose logic, and argued that analog VLSI circuits, in which the physics of transistors operating below threshold directly mimics the physics of neurons and synapses, could close it. In this view computation is something the device does naturally rather than something imposed on top of it.
The paper laid groundwork for decades of work on silicon retinas, silicon cochleas, and event-driven sensors, and it set the conceptual stage for later large neuromorphic chips. Its core insight, that matching the structure of computation to the physics of the substrate buys enormous efficiency, remains the central pitch of neuromorphic hardware.
For a general reader, Mead’s paper is the origin of a hardware tradition that, as the energy cost of AI becomes a real constraint, is drawing renewed industrial attention.