In 1958 the Navy said its neural network would one day be conscious
The first trainable neural network, the perceptron, was announced as the embryo of a computer that would “walk, talk, see, write, reproduce itself and be conscious of its existence”.
Neurons as logic
In 1943, the neuroscientist Warren McCulloch and the logician Walter Pitts published a paper describing simplified neurons as switches that could compute logical functions. If brains were networks of such units, perhaps machines could be built from them too.
A machine that learns
At Cornell Aeronautical Laboratory, the psychologist Frank Rosenblatt took the next step: a network that could learn from examples. He simulated his perceptron on an IBM 704 in 1957 and published it in 1958. The later Mark I Perceptron hardware looked at images through a 20×20 grid of light sensors and adjusted its connection strengths with electric motors as it learned to tell shapes apart.
The headline
In 1958 the New York Times reported on a US Navy demonstration, describing the perceptron as “the embryo of an electronic computer that [the Navy] expects will be able to walk, talk, see, write, reproduce itself and be conscious of its existence.” That was a remarkable claim for a machine that could learn to tell left from right.
Bring it up
“In 1958 a newspaper reported the Navy expected its new learning machine to eventually be conscious of its own existence.”
Follow the thread
- RelatedAI got its name from a summer workshop proposal The same era of bold promises.
- Leads toAI has gone through “winters” before Overpromising and a famous critique pushed neural networks out of favour.
- Leads toFrom a bedroom GPU to ChatGPT Modern AI is built on the multi-layer descendants of Rosenblatt’s idea.
