The perceptron is usually remembered as an algorithm, but Frank Rosenblatt’s Mark I Perceptron was an actual machine, built at the Cornell Aeronautical Laboratory in Buffalo, New York, with funding from the US Office of Naval Research and the Rome Air Development Center. Its “eye” was a 20-by-20 grid of 400 photocells that turned a projected image into electrical signals. These sensory units were wired, in part at random, to a layer of association units, and from there to output units. The crucial connection weights were not numbers in software but physical settings - potentiometers turned by small electric motors, so the machine literally adjusted its own knobs as it learned to classify patterns.
The perceptron drew extraordinary attention. After a 1958 demonstration for the Navy, the New York Times reported on an “embryo of an electronic computer” that the Navy expected would one day “walk, talk, see, write, reproduce itself and be conscious of its existence.” That hype, far ahead of what the hardware could do, helped set up the backlash that came with Minsky and Papert’s 1969 book Perceptrons. The original Mark I Perceptron is preserved in the Smithsonian’s collection - a reminder that the first neural networks were built from photocells, wires, and motors before they were ever lines of code.