I've been looking at this in a little bit of detail because we

Are going to use, we are going to come to artificial neural

Networks next, which is essentially a network of

Neurons and every neuron has a couple of input channel which

Have inherent weights so you build a weighted sum over the

Inputs and classify this into i'm going to fire with my own

Signal or i'm not. You somehow have to do as a

Neuron. You have to do something with

Your inputs. And that's a classification

Problem. Okay.

And the simplest way to do classification is by

Perceptron. Okay.

So one of the simplest ideas of how to do neural networks

Apart from the fact that it's a big graph of highly connected

Computational nodes that have inputs and outputs, right?

And they can send signals or not to the outputs.

That is, so it's very simple to say, well, and we have a

Neural classification problem. You could have something else.

You could have a quadratic or 7th degree polynomial in there.

But why would you? we always start with the dumbest

Thing we can imagine first. And sometimes that works.

And for neural networks, it does actually work.

And people have tried more interesting models of neurons.

They don't really work any better. The power of these neural

Networks is not in the power of the units, but in the power of

The networks. And in particular, in the power

Of the units. Okay.

So we started looking at neural networks.

And we essentially got as far as defining them and kind of

Reviewing the basic biology facts. Okay.

What are the biological facts? we have networks, highly

Interconnected networks of relatively dumb computational

Units. The computational units are

Essentially things that have a couple of inputs on the order of

A thousand or so inputs they collect.

And then they have an output which is then again connected to

An order of a thousand other input cells and neurons by the

Way of synapses. And these synapses are kind of

Little communication devices that actually make some of the

Waiting decisions. And they can learn.

So competitive brains have a lot of these.

There are smaller brains in other biological organisms that

Have much less. And they can do much less.

From a fly you wouldn't expect higher math.

But they can fly very well. So they have smaller brains.

And the somewhat always again surprising thing is that the

Computation turnaround time of neurons is extremely slow by

Today's standards. Right?

We do not have a gigahertz computation machine up here.

We compute in turnaround times of milliseconds.