Frank Rosenblatt Introduces the Perceptron

Depending on how it goes with AI, the perceptron could end up being one of history's most important concepts, well, ever basically.

Frank Rosenblatt's perceptron is essentially the first neural network and a monumental step in the story of deep learning AI art. The perceptron is also a bit of a tragic tale. A decade later, deep learning AI's anti-hero Marvin Minsky published Perceptrons, a book that critics of neural networks would point to in order to discredit the field or deny funding. It was thought of as proof that those pesky neural networks just didn't work. Machines couldn't learn (at least not that way). So the perceptron itself marks the beginning of neural networks and also partly explains why if this idea came out in 1958, it only really exploded around 2010. There are many reasons why like data and compute needing to catch up with theory. But that theory barely even existed, largely because of Minsky and his dividing of the AI field from about 1969 to 2010.

Rosenblatt himself though was a research psychologist at the Cornell Aeronautical Laboratory in Buffalo, New York. He described the perceptron in a 1957 report and in Psychological Review in 1958. It modeled the brain as sensory, association and response units that learn to sort shapes into classes. In July 1958 the U.S. Office of Naval Research (And, yes, you knew the military industrial complex was going to be involved, didn't you?) demonstrated it on an IBM 704: after 50 trials, the computer taught itself to tell cards marked on the left from those marked on the right. In other words, it was already demonstrating some very early effectiveness.

Rosenblatt called it "the first machine which is capable of having an original idea." The human idea itself was certainly original.