When I first came across neural networks, I was blown away.
They didn’t just solve problems — it felt like they could do everything a human could. Think. Reason. Fall in love. Develop an ego. Even become evil.
Since then, I’ve given talks to A-level students about how large language models (LLMs) like ChatGPT work — mostly to show how powerful (and fun!) mathematics can be. But along the way, I found myself circling back to deeper philosophical questions — What is intelligence? Consciousness? Free will?
This post is a simple intro to how neural networks work — and how uncannily similar they are to you and me.
How do humans work?
Imagine you’re looking at a sunset and say, “Wow, that’s beautiful.”
What actually happened here?
Step 1: Light enters your eyes.
Sunlight bounces off clouds, grass, waves — and hits the retina at the back of your eyes. There, millions of light-sensitive cells (rods and cones) convert that light into tiny electrical signals.
Step 2: The signals travel to your brain.
Each photoreceptor in your eye is wired to a neuron in the brain via a nerve. The signals pour into your brain — tens of millions at once!
Step 3: Neurons fire and connect.
Each neuron is linked to thousands (sometimes millions) of others forming a deep, branching network. When it receives its signal, the neuron may “fire” — sending electricity to the neurons it’s connected to, depending on
- The neuron’s current chemical state (its “activation”)
- The strength of the connection (the “weight”) between it and its neighbors
Step 4: The signal bounces through the brain.
Signals fan out in a huge cascade — like a pinball machine on steroids.
Layer after layer, neuron after neuron. Some fire, some don’t. It’s a vast, chaotic, beautifully organized mess of electricity.
Let’s pause for scale:
If 100 million neurons connect to 100 million others (numbers not so far-fetched in parts of the brain), that’s 10,000,000,000,000,000 (10 quadrillion) connections — and that’s just between two layers!! There are many such layers.
Step 5: Eventually, the signal reaches your vocal cords.
Some signals reach your speech center, triggering nerves that control your vocal cords. You exhale. Vibrating air escapes. And someone nearby hears: “Wow, that’s beautiful.”
What does this tell us?
From photons entering your eye to words leaving your mouth — it’s all just electrical signals bouncing through a network.
There is no magic, no moment where “free will” enters. It’s just pure, (but super complex!) cause and effect — a beautifully choreographed deterministic system. That system is you!
Modelling the brain in a computer
Inspired by this structure, researchers built artificial neural networks. These are the engines that power modern AI tools like ChatGPT, Claude, and many others.
Let’s say you want to build a chatbot that reads a message and replies with a single word. You can then pass the output back in and get another word (the 2nd word) and so on until you have a full message.
- You take the input message (say 1000 words) and convert it to numbers.
- Each number activates one neuron in the input layer (1000 neurons total).
- The output layer has ~200,000 neurons — one for every word in the dictionary.
- Between them, you add hidden layers (say 50 layers), each with 1000 neurons.
Each neuron passes its signal to others based on:
- Its own “activation function”
- The weight of its connections
Sound familiar?
This is how a neural net “thinks.” It’s the same as a brain, just cleaner and made of code.
Most of the work in setting up their models is figuring out what weights and activations to use in order to get useful outputs (that is the “training” of the model).
So… will AI have free will?
This is where things get weird and philosophical.
If a vast enough network, following deterministic rules, can simulate
- intelligence…
- reasoning…
- emotions…
- creativity…
then is there any meaningful difference between its behavior and our own?
Is free will just an illusion of complexity?
And if we are deterministic systems — highly advanced, gloriously tangled nets of neurons and weights — then are we not, in some sense… machines too?
Or perhaps the question is not whether AI can be human — but whether humans ever were anything else.
Determinism doesn’t mean despair
Even if consciousness is just an emergent property of a very complex machine… Even if “free will” is an illusion… That doesn’t mean life is pointless.
Sure, someone might say, “If I have no real choice, why not stay home and watch TV all day?”
But if your choices are just the result of the machine’s wiring… Then whether you go to work or not isn’t really up to you anyway.
So maybe… just enjoy the ride.
Mathematics, machines, and meaning
As a mathematician, I’m drawn to the idea that consciousness might be broken down — reduced to simpler parts, down to the axioms. Like all good theorems. I’ll explore that in another post — including how quantum physics might (or might not) be relevant.
But for now, let me leave you with this:
Even in a fully deterministic universe, beauty still exists.
A child might enjoy playing noughts and crosses, not realising there’s a winning strategy. As they get older and figure it out, the game loses its mystery.
Adults playing chess, a much deeper game — might marvel at it, wondering what the opponent will do next, and feeling the thrill of uncertainty.
To a vastly more intelligent being though, chess would feel as trivial and predictable as tic-tac-toe (noughts and crosses); chess has a finite board with finitely many moves possible so there are only finitely many games that could ever be played.
Maybe it’s in our limitations that joy lives. And in our finiteness, that meaning hides.
So don’t fear being a machine. Be glad that you’re not a perfect one.