History
In 1951 Marvin Minsky and Dean Edmonds conceived what is known as the first neural network machine. What is its name?
SNARC.
Card 1 of 259. Answer: SNARC.
SNARC.
Card 1 of 259. Answer: SNARC.
About · Quizzes · History of AI · To suggest new questions or report an error send me a dm.
SNARC, the Stochastic Neural Analog Reinforcement Calculator. Marvin Minsky and Dean Edmonds, Princeton graduate students in mathematics and physics, built it in the summer of 1951 with Office of Naval Research money that Harvard psychologist George Miller helped secure. It is usually called the first artificial neural network machine.
It had about 40 neuron-like units, each storing recent activity on a capacitor and its long-term connection strength on a potentiometer. When the machine was rewarded, motors turned the potentiometers of recently active units, making them more likely to pass a signal again: a hardware form of reinforcement learning. It simulated a rat learning a maze, using vacuum tubes, motors, clutches and a surplus B-24 bomber autopilot.
Interviewers may contrast it with Frank Rosenblatt's perceptron of the late 1950s, which learned to classify patterns from labeled examples rather than from reward. Minsky went on to co-organize the 1956 Dartmouth workshop and to co-write Perceptrons (1969) with Seymour Papert, a critique often blamed for cooling interest in neural networks.