In a recent blog post, I noted some of my discoveries in regards to how learning causes the strengthening of synaptic connections in the brain. In essence, learning a fact or skill causes an increase in the likelihood that neurons will pass on an electrical signal when pressured to buy other neurons.

That’s all fine and dandy, but how does that really relate to me “knowing” the fact that Benjamin Franklin invented bifocals. I think I’m starting to develop an understanding of this. Let’s presume that somewhere in your brain you have a collection of neurons that “represent” the idea of Benjamin Franklin. You also have a neural network representing the idea of invention. And you also have a neural network somewhere for, you got it, bifocals. These three neural networks are sitting around, minding their own business, when enough information comes in to pass on the knowledge that Benjamin Franklin invented bifocals. Thus, the neuron, or perhaps several neurons, trailing from the Benjamin Franklin neural network to the invention network get strengthened. As do the neurons going from the bifocals network to the invention network. I suppose neurons going from any one of these networks to any of the others get strengthened.

This is an interesting idea — knowledge can be represented by connections. To some degree, this is what language does. You might have words like “dog, frog, ate, Tuesday” but you can form different connections between them. My dog ate a frog on Tuesday. A frog ate my dog, Tuesday. A frog and my dog ate on Tuesday. You can think of the individual words as neural networks, and the act of placing them in sentences as representing the strengthening of synapses.

Of course, the idea of a dog is not represented with a single neuron, it’s many, hence it’s a neural network. That means dog is also series of connections. It’s a connection between networks capturing hairiness, barking, drooling, canine, pet, wolves, wagging and every other property we associate with dogs. And those networks break down to even more granular levels. Barking is, for example, related to sound, to a particular sharp report, to alarm, possibly danger etc.

So what bit of information can be represented as a single neuron? Beats me. But I think understanding knowledge as an incredibly complex series of connections is a metaphor that can be easily mapped onto the circuitry of the human brain.