Tiny Decision Makers in Layers

A quick note on how we teach here. You won't find equations or numbers in this series, and that is on purpose. The goal is understanding, not arithmetic, so every idea is explained in plain words and simple pictures. When you want the math later, the rest of Internals Decoded is ready for you.

Welcome back. Last time we saw that the way machines learn to spot patterns is loosely inspired by the way brains work. It is not a copy. It is more like a rough sketch that gave people a place to start. Today we are going to look at the simple building blocks that make that sketch real.

A neural network is built from many tiny parts, each making a very small decision, arranged in layers so that one layer's answers become the next layer's questions. Simple parts, stacked together, add up to something capable.

A line of helpers

Imagine a fire brigade from a time before fire engines. A line of people stretches from a well to a burning building. The first person fills a bucket with water and hands it to the next. That person passes it along. So does the next, and the next, until the bucket reaches the person at the fire, who throws the water onto the flames.

No single person in that line puts out the fire. Each person does only one small thing: take the bucket, turn, hand it over. But the whole chain, working together, moves water from the well to the fire. The job gets done by handing work along the line.

A neural network works the same way. It is a chain of tiny decision makers. Each one receives something from the part before it, makes one small judgement, and passes its answer onward. Alone, one of these parts cannot do much. Together, they can handle tasks that seem almost magical, like recognising a face in a photo or understanding a spoken sentence.

Stacking the work

Now picture a set of kitchen sieves, each one finer than the last. You pour a mix of sand and pebbles into the top sieve. The first sieve catches the biggest pebbles and lets everything else fall through. The next sieve catches smaller stones. The one after that catches grains of coarse sand. By the time the mix reaches the bottom, only the finest dust passes through.

Each sieve makes one simple decision: does this piece fit through my holes or not? But by stacking them, you separate the whole mess into neat piles, from large to small. The first sieve sees only the roughest things. The last sieve sees only the tiniest details.

A neural network builds understanding in a similar way. The first layer of tiny decision makers notices rough, simple things. The next layer combines those rough things into something a little richer. The layer after that builds on that, and on it goes. Meaning gets constructed in stages, just like sorting sand with a stack of sieves.

The tiny decision maker

So what does one of these tiny parts actually do? It is surprisingly simple.

Each part receives a handful of signals from the parts before it. Some signals are strong. Some are weak. The part has its own set of knobs, like the ones we talked about in the last piece. Those knobs decide how much attention the part pays to each incoming signal.

The part mixes the signals together, guided by its knobs. Then it makes one small judgement: should I pass a strong signal forward, a weak one, or almost nothing at all? It does not think in words. It just does a little bit of arithmetic, the kind a pocket calculator could handle, and then decides how loudly to speak up to the next part in line.

People call each of these tiny decision makers a neuron. The name is a small nod to that loose brain idea we talked about last time. A real brain cell is far more complex, but the name stuck. In a machine, a neuron is just a very simple rule that takes what it is handed, weighs it, and passes on a result.

Layers build understanding

When you put a whole row of neurons side by side, all of them working on the same input at the same time, you get something called a layer. A layer is like a team of specialists, each one looking at the same thing but through its own set of knobs. One neuron might pay close attention to a certain kind of edge in a picture. Another might care more about a patch of colour. Together, the layer paints a rough picture of what it sees.

The first layer in a network that looks at images might notice very simple things: a bright spot here, a dark line there. The second layer receives the first layer's rough notes. It can combine those notes into slightly more interesting ideas: a corner, a curve, a patch of texture. The third layer might piece those together into something that looks like an eye or a wheel. Each layer's answers become the next layer's questions.

This is how meaning gets built up from almost nothing. No single neuron understands the whole picture. No single layer does either. But the chain of layers, each one refining what the last one handed it, can eventually recognise a cat, or a handwritten word, or a familiar voice.

The word for this whole arrangement is a neural network. It is just a lot of neurons, arranged in layers, passing signals forward. Simple parts, stacked, add up to something capable.

Where this is heading

So far we have seen that a network is made of many tiny decision makers, each doing one small job, lined up in layers that build on each other. If a few layers can do useful work, what happens when you stack many, many layers deep? Something changes. The network can learn ideas that are far more complex and subtle. That depth is the reason for a phrase you have surely heard: deep learning. Next time we will see what depth really means and why it matters so much.