[Texto disponível apenas em inglês] On the right: a cat's visual cortex. A bar of light crosses the animal's field of view and one neuron answers. The clicks are the neuron's activations, i.e. the spikes, recorded on tape. Hubel and Wiesel received the 1981 Nobel Prize for what those spikes meant; that seeing is built, cell by cell, from single features of the world.
On the left: the same spikes, played into an analog neuron I fabricated in silicon. It has a membrane, a leak, and a threshold. It answers in continuous time, in analog, and exploits nothing else than physics. The neuron itself has no clock and computes no numbers. It does not answer identically; no two neurons, born or made, ever do.
These clicks are also where machine vision began. Today's neural networks are a diagram of this cat. What they copied was the wiring. What they left behind was the body.
Sixteen neurons is not a brain. It is the start of a different bet: that machines which have to perceive, in real time, in the world, on the power the world can spare, should be made of physics rather than computer programs. How far that goes, nobody knows yet. In 1959, nobody knew either.
