Author's Essay
The Fire of Intelligence
In the Beginning Was the Spark
Original narration
This original recording contains earlier factual formulations and does not match the revised text. See the factual notes below.
I think of intelligence as a force of nature.
I imagine it woven into the fabric of the universe, waiting to be awakened. As gravity binds matter, as electromagnetism binds light, intelligence, in this metaphor, binds patterns into meaning.
In that imagined landscape, it lay dormant. Then a creature rose onto two legs, gripped a stone, and struck it against another stone. Sparks. Fire. The first taming. With fire, the night learned to obey. With the wheel, distance shrank. With writing, time stood still. Every time humanity has seized a force of nature, it has been transformed at the root — and lost a little of its innocence.
Now we stand at the next fire. This time it is not heat we are taming. It is thought itself.
Stars in Your Skull
Inside your cranium lives a universe.
Around eighty-six billion neurons form an intricate network, joined by vast numbers of synapses. The number of connections varies from one neuron to another.
Electrical and chemical signals pass between cells. Memory is woven through this changing network; it is not a hard drive with a reliably measurable capacity in bytes.
Its efficiency is extraordinary, but comparisons with computers depend on what we ask each to do.
All of this fits in a 1.4-kilogram lump you carry between your ears.
The River of the Senses
Every second, the world comes pouring through you. Light enters your eyes; cells in the retina turn it into neural signals. A face, a flicker, the changing colour of the sky.
In your inner ears, hair cells transform vibration into signals. A voice, a footstep, a note held in the air — an orchestra finding its way inward.
Your skin registers pressure, heat, pain, contact. You can feel a mosquito landing on your arm — a small arrival, suddenly part of your world.
Smell. Taste. Balance. Proprioception, the body's sixth sense, keeps you aware of where your body is in space.
All this incoming information reaches the brain. And the brain does something remarkable: it filters and organises it. Only some of it enters awareness.
The Art of Forgetting
An infant's brain is still developing the systems that help select and organise experience. Crying can signal hunger, fatigue, discomfort or pain; it has no single explanation.
As the brain develops, connections change. Synaptic pruning removes some synapses, while other connections are strengthened. It is not the same as billions of neurons dying. In the language of this essay, highways emerge where there were footpaths: learning reshapes the routes we travel.
The brain continues developing and maturing into the mid-to-late twenties, with the prefrontal cortex among the last regions to mature. There is no single birthday on which that work is finished. You have become a specialist at being you — and paid for it with all the others you might have been.
Some people show exceptional abilities in memory, calculation or drawing. Such differences are striking, but they do not establish that everyone carries a hidden photographic memory, or that exceptional ability is simply the result of pruning that failed.
Copying the Architecture
For decades, we tried to build intelligence out of rules.
If X, then Y. Logical chains, decision trees, expert knowledge fed in line by line. It worked for chess, for bookkeeping, for sorting mail.
But what seemed simplest proved much harder to handle with hand-written rules: recognizing a face, understanding a sentence, seeing a dog in a photograph.
Then we turned our gaze inward.
What if, instead of writing a rule for every case, we could build a system that learns from examples? Not a copy of the brain, but an architecture loosely inspired by its networks.
An artificial neural network connects mathematical units, often in layers. These units are loosely inspired by neurons, not replicas of them. Each connection has a weight, a number that decides how strongly the signal is amplified or dampened. Training can begin with random weights. The network makes predictions; a training method adjusts the weights in response to errors. A little stronger here, a little weaker there. Thousands of times. Millions of times. Billions of times.
Researchers still design the architecture, training methods and objectives. What is learned from examples is not a hand-written rule for every case. The results can include image recognition, translation and assistance with medical image analysis. This is not the same process as a child learning to recognise a parent.
From Switch to Dreamer
It began with a switch. On and off. Zero and one.
ENIAC was completed in 1945 and publicly unveiled in 1946. A machine weighing roughly thirty tons, built to calculate artillery tables.
In 1969, the Apollo Guidance Computer carried humans toward the Moon with 2,048 words of erasable memory and 36,864 words of fixed memory. An extraordinary journey, made within extraordinary limits.
In 1997, Deep Blue beat Garry Kasparov in a chess match, combining fast search with chess-specific evaluation — not brute force alone.
Then something happened.
In 2012, a deep neural network made a major leap in the ImageNet image-classification competition. In 2015, researchers reported results beyond a human benchmark on that specific task.
In 2016, AlphaGo beat Lee Sedol at Go, a game so complex that brute force alone would never suffice.
By 2023, language models brought fluent conversation into everyday use. Fluency does not guarantee accuracy.
Step-by-step reasoning methods were already being studied before 2025. New systems extended those methods, without making their conclusions infallible.
Now researchers are aiming for artificial general intelligence, AGI — the ambition of broad capabilities across many tasks, rather than excellence in one narrow domain. It is not an established ability to solve any problem, nor evidence that a machine wonders.
And beyond AGI lies the hypothetical prospect called artificial superintelligence: systems exceeding human capabilities across a broad range of tasks. A name for a possibility, not a declaration that it has arrived.
The Last Mystery
But here the question changes. A machine can generate words that echo with longing, or music that moves a listener. It can model patterns associated with emotion. Is that the same as feeling?
Whether an artificial system could have subjective experience remains an open question. Fluent language alone does not settle it. That is the question this story has asked.
To me, love remains the last mystery. Biology can help explain attachment and care without exhausting what they mean to the person who loves. We do not always love because it is rational; sometimes we love even though it is not.
Empathy: the ability to suffer with someone else, to feel their pain even when no reward is promised. For me, that willingness to care is part of what makes us human — not a proof that care lies outside biology.
For us, freedom can feel like the gap between impulse and action — the moment when we see our reaction but have not yet obeyed it. What would such a struggle mean for an artificial system? The question remains a philosophical one, not a settled proof of what a machine can or cannot experience.
We remember, we miss, we long; we stretch the present into a story with a past and a future. Machines also work through time: computations take time, and systems can store information and revise earlier outputs. Whether any of this is lived experience is a different question.
And truth? Truth has become fluid. In a world where images can be fabricated, voices cloned, memories manipulated — what does authenticity mean?
Perhaps faith — the irrational conviction that something is true beyond the algorithm — is the last human superpower.
At the Threshold
We have always built tools that surpass us. The wheel moves faster than legs. The telescope sees farther than the eye. The pen remembers better than the brain.
But now we are building systems more capable than we are in many tasks, not necessarily better at everything. And we do not know where it ends. Every time humanity has tamed a force of nature — fire, the atom, the gene — we have lost a little of the control and gained a little of the insight. We have burned ourselves, poisoned ourselves, nearly erased ourselves. But we have also warmed ourselves, healed ourselves, risen again.
This time, there is a question we have never had to ask before: what does it want? Fire wanted nothing. The atom wanted nothing. They obeyed the hand that held them, for blessing or for catastrophe — but they obeyed.
An advanced system may pursue goals in ways we did not intend. How do we make increasingly capable systems keep serving what we value — not just today, but as they change?
Researchers call it the alignment problem. It is the defining question of our time. There are proposed approaches, but no universally accepted complete solution.
This is no longer a future you can wait for. You have probably already spoken to a machine today.
Now we stand at the fire of intelligence.
The question is not whether the machine will think. It already does.
The question is what we do when it begins to wonder why.
Factual notes
Neural connections change with development and experience. Synaptic pruning and neuroplasticity are not interchangeable with neurons dying, a fixed memory capacity in bytes, or a universal hidden photographic memory.
NCBI Bookshelf — NeuroscienceInfant crying has several possible causes, including hunger, fatigue, discomfort and pain; it does not have a single explanation.
NHS — Soothing a crying babyArtificial neural networks are mathematical models inspired by aspects of biological networks, not copies of brains. Their architecture, training methods and objectives are designed by researchers.
Nature Communications — A critique of pure learning and what artificial neural networks can learn from animal brainsDeep Blue defeated Garry Kasparov in 1997 using fast search and chess-specific evaluation. AlphaGo defeated Lee Sedol in 2016, combining learned networks with search.
The brain continues developing and maturing into the mid-to-late twenties, with the prefrontal cortex among the last regions to mature. There is no single birthday on which that work is finished.
NIMH — The Teen Brain: 7 Things to KnowIn 1969, the Apollo Guidance Computer carried humans toward the Moon with 2,048 words of erasable memory and 36,864 words of fixed memory. An extraordinary journey, made within extraordinary limits.
NASA — Apollo Guidance, Navigation and Control (PDF)In 2012, a deep neural network made a major leap in the ImageNet image-classification competition. In 2015, researchers reported results beyond a human benchmark on that specific task.