George Wald at Harvard worked out the molecular photochemistry of the retina in 1958, showing how rhodopsin in rods and cones changes shape when it absorbs a photon, triggering the chemical cascade that turns light into a nerve signal. A year later, Hubel and Wiesel, also at Harvard, recorded neurons in the visual cortex of cats and found that individual cells fired only when a dark bar was shown at a specific orientation, revealing that the brain builds up feature detection by combining signals from simpler cells into more complex ones.
- 01
Wald shared the Nobel Prize in 1967 for this work.
- 02
Hubel and Wiesel won the Nobel Prize in 1981 for the cortical feature detection findings.
- 03
The hosts frame Hubel and Wiesel's layered wiring of neurons as a precursor to how artificial neural networks are structured, with layers combining simpler units into more complex feature detectors.
398 words · auto-generated from the episode video
1:58Right? So you know obviously we see things as human beings but before 1958 it was a bit unclear how images right the photons that are coming in into our eye get translated into electrical signals and go all the way to the brain because at the end of the day what needs to happen is photons need to get registered somehow and then that needs to get processed into like objects and things like that right biochemist George Wald at Harvard university discovered the molecular photochemistry of the retina. He found rods and cones and he figured out what is the chemical cascade that lets
2:39redopsin which is the protein that gets a photon and then actually changes shape to register that it caught a photon. He discovered all of that. So it's really like this quantum biochemistry type stuff where you know nowadays we've got cameras with like little detectors the silicon registers the photon and creates an electrical current. Well, now we've got the biological analog to that, which I think is really cool. That was in 1958. He earned the share of the Nobel Prize in 1967. And then later on in 1959, just a year later, um, Hubil and Weasel, also at Harvard, discovered cortical feature detection. What they were doing
3:19was looking at the visual cortex of cats and they noticed that certain neurons would only fire when a bar, a little dark bar in a white background, a black bar was oriented in a certain direction. Like the neurons would only fire when the black bar was at this angle and not at any of the other angles. So you've got like now higher order feature detection happening in the retina. And they figured out what types of neurons how do we get that type of feature detection if we string together multiple neurons into a single neuron and create this type of object feature detection. Right? This is the precursor to not just
4:02understanding the neural system and our brain but also to later theories about how to use this for artificial neural networks. Because at the end of the day, artificial neural networks take a bunch of neurons, they input into one neuron, and layer by layer create more and more complexity, right? Y >> so huge, huge thing. The Hubble and weasel won the Nobel Prize in 1981.