AI - The Second Mind with Enhanced Evolution

The Night Superintelligence Kept Me Awake

It was past one in the morning. Trichy was silent except for a stray dog with opinions.

I had gone down a rabbit hole. It started as light reading and ended somewhere near the edge of my own sanity. Machines that might one day think faster than every human who has ever lived. Combined. A mind that could hold the whole of human knowledge and still have room left over for boredom.

I sat there with my phone glowing on my face like a small blue interrogation lamp, and I felt a strange thing happen. My chest went tight, and my life went small.

Because the very next thought that arrived was this: tomorrow morning I will wake up, iron a shirt, and go argue with a colleague about a leave application.

We have built an entire civilisation around the nine-to-five. We worship it quietly. The salary slip arrives on the last day of the month like a temple prasadam. We measure a man by his designation. We ask a boy what he wants to become, and we secretly want him to say a job title. Whole families have been organised around the idea that a chair, a cabin, and a fixed increment equals a life well spent.

And from that ledge — from the edge of what may be coming — the whole arrangement looked like a sandcastle. Built with real effort. Decorated with shells. Defended fiercely. And the tide was already at our ankles.

Geoffrey Hinton felt a version of this too. He spent fifty years building the very thing. In 2023 he walked away from his role at Google so he could speak freely about where it all leads. The man who lit the fire, standing back from the heat, worried about the size of it.

I went to bed with a heavy head that night. And I woke up with a question that has stayed with me for months.

While we polish our resumes, who exactly is polishing us?

Will Smith Ate the Spaghetti, and It Was Horrifying

Let me take you back to March 2023.

Someone fed an early video model a simple line: Will Smith eating spaghetti. What came out was eleven seconds of pure nightmare fuel. His face melted and re-formed like warm wax. The fork passed through his cheek. The noodles seemed to be breathing. At one point the spaghetti eats him back.

The internet lost its mind. My WhatsApp groups lost their minds. Uncle-types who had never used the word "algorithm" in their lives forwarded it with three laughing emojis and a voice note.

And underneath the laughter was a feeling we all enjoyed very much: relief. Look at this clown. This thing is years away from touching us.

It was a good era for laughing.

Back in 2015, Google released DeepDream, and every photo it touched came back covered in dogs. A blue sky? Dogs. A plate of idli? Dogs with sad eyes. The machine had studied so many dog pictures that it began to hallucinate them into clouds, the way you see a lion in a rain stain on your ceiling.

Then came the face generators, melting eyes into cheekbones, growing a third earring on a chin.

There was even a ghost story. In 2022, an artist experimenting with image prompts kept summoning the same haunted woman with hollow eyes, again and again, from prompts that had zero to do with her. She was nicknamed Loab. A digital cryptid, living somewhere in the folds of the math. Nobody could fully explain her.

We laughed at all of it. I laughed at all of it.

Then eighteen months passed.

Today the video models render light on water. They render the way cloth falls when a woman turns. They render breath in cold air. Studios have paused productions to hold meetings about what to do. The joke aged in the time it takes a child to learn to walk.

Morgan Housel writes in Same as Ever that people are consistently blind to the speed of change until it has already arrived. He is right, and it is a humbling thing to feel in your own gut.

So if the clown grew up this fast — what were we really laughing at? The machine, or ourselves?

New Engine, New Fuel, New Hands

Three things changed. Allow me to keep this in the language of a tea stall rather than a research lab.

The engine got bigger. Fifteen years ago, this work ran on a handful of chips in a university basement. Today it runs in buildings the size of small townships, humming with racks of processors, drinking electricity the way a mid-sized nation drinks it. Cooling systems the size of my apartment block. The scale of it is almost silly.

The fuel got richer. We fed it the internet. Books, arguments, poetry, medical journals, court judgments, recipes, the comment section of YouTube. The good and the ugly. Everything we have ever written down, poured in.

And the hands changed most of all. This is the part I find beautiful. In the beginning, it was a curiosity — a party trick you showed a friend. Today my son uses it to understand a physics problem at eleven at night. A farmer in Erode photographs a leaf and asks what the disease is. I use it as a thinking partner when a chapter refuses to open. It moved from a toy on the shelf to a second brain in the pocket, and it happened so quietly that most of us skipped the ceremony.

Two moments carried us here, and both deserve your attention.

2012. A contest called ImageNet, where machines competed to recognise objects in photographs. Progress each year had been slow and polite — a percentage point here, a percentage point there. Then three researchers — Alex Krizhevsky, Ilya Sutskever, and that same Geoffrey Hinton — entered with a different approach. The error rate collapsed. The gap between them and everyone else was so wide that the field simply changed direction overnight. A quiet morning in a competition almost nobody outside the field was watching, and the century turned on it.

2016. Seoul. AlphaGo versus Lee Sedol, the finest Go player of his generation. Game two, move thirty-seven. The machine placed a stone in a position that made the commentators go silent. It broke a rule that human masters had honoured for a thousand years. One commentator assumed it was a mistake. Lee Sedol stood up and left the room. He needed air.

It was a winning move. And here is the part that still gives me a chill: no human had taught it. The machine had explored regions of the game that a thousand years of human tradition had walked past.

We fed it the world. And I keep asking myself — what was it doing with all of it while we slept?

You Were Pruned Into the Man You Are

Now let me bring this home, to the only laboratory I truly love. The one behind your eyes.

Think of yourself at three years old. You were, at that moment, holding more raw connection than you will ever hold again. A toddler’s brain is a wild jungle of synapses — far more links between neurons than an adult carries. Every possible path is lit. The child could learn any language on earth with a native tongue. Any language. Tamil, Portuguese, Swahili, all equally available.

And then life begins to cut.

This is called synaptic pruning, and it is the most under-appreciated love story in biology. The connections you use grow thicker, faster, more insulated. The connections you leave alone fade and get cleared away. By the time you reached your teens, a large share of those early links had been quietly removed.

You were sculpted. By subtraction.

Every time your mother called your name and you turned, a path thickened. Every time you fell off the cycle near the corner shop and got up anyway, a path thickened. Every scolding. Every song on the radio during a long bus ride. Every humiliation in a classroom that you still remember with surprising clarity. Every love.

The man reading this sentence is the surviving pattern.

Donald Hebb described the mechanism in 1949, and Carla Shatz later gave us the line every student now carries: neurons that fire together, wire together. Spark two things at the same moment, repeatedly, and the brain builds a road between them. Leave a road unused, and the jungle takes it back.

Eleanor Maguire proved this on living adults. She studied London taxi drivers, who must memorise 25,000 streets to earn their licence — a test they call The Knowledge, which takes years and breaks most people. She scanned their brains. The posterior hippocampus, the region that handles spatial memory, was physically larger. And the longer a man had driven, the larger it was. The organ had grown to meet the demand placed on it. Grey matter, reshaping itself, in a fully grown adult, simply because he kept using it.

Now hold that picture, and look at the machine.

A fresh model, before training, is meaningless noise. Billions of connections, all equally weightless, all equally useless. A newborn jungle with no roads.

Then you feed it. And with every example, the connections that predict well grow stronger, and the connections that fail get weakened toward zero. Over millions of passes, roads form. Useless paths fade. Useful paths thicken until they carry traffic at speed.

Sit with that for a second, because it stopped me cold the first time I saw it clearly.

Training is pruning. The machine becomes itself the same way your childhood made you — by strengthening what fires together and letting the rest go quiet.

Your mother, your street, your school bench, your first heartbreak: that was your training data. You were shaped by what you were exposed to, whether you chose it or otherwise.

This is exactly why my work as a mind trainer exists. If the roads thicken with use, then attention is the most expensive thing you own. Every hour you spend scrolling, you are laying tar on a road you will regret. Every hour you spend still, breathing, watching your own mind, you are laying tar on a road that will carry you home.

If a brain becomes itself by choosing which sparks to keep — and a machine does the very same — where exactly is that clean line we keep insisting on?

Beyond the Code: The Part Nobody Can Explain

Here is where most people get it wrong, and where the story gets strange.

Ask a room of educated adults how AI works, and someone will say, with total confidence: "It is just programming. Someone wrote instructions. It follows them."

That was true once. It stopped being true a while ago.

Engineers today write the recipe for learning. They design the architecture, they choose the fuel, they build the engine. Then they step back. They write zero of the answers. The abilities arrive on their own.

And they arrive suddenly.

In 2022, a group of researchers led by Jason Wei documented what they called emergent abilities. Below a certain size, a model would fail a task completely — solving a multi-step problem, understanding a joke, doing arithmetic. Scale it up, feed it more, and past some threshold the ability simply appears. Nobody built it. Nobody placed it there by hand. It emerged from the size and the fuel, the way water becomes wet only once enough of it gathers in one place.

I will be honest with you, since honesty is the whole point of my work. A rival team, led by Rylan Schaeffer in 2023, pushed back hard. Their argument: some of these sudden jumps are a trick of how we measure, and the growth is smoother than it looks. They may be right. The field is still arguing.

But even the sceptics agree on the deeply uncomfortable part.

We are unable to fully read the thing we have built.

Open a trained model and you find billions of numbers doing work that the engineers who built it struggle to explain. There is an entire branch of research — interpretability — that exists purely to reverse-engineer our own creation after the fact. Teams have started to map the concepts hiding inside these systems, finding internal features that light up for ideas like a bridge, or a bug in code, or the feeling of a text being flattering. Real progress. And still a small torch in a very large cave.

Read that once more. We know how to grow it. We are still learning what it is.

Melanie Mitchell writes about this gap with beautiful clarity in Artificial Intelligence: A Guide for Thinking Humans. Brian Christian circles the same fire in The Alignment Problem. And Alan Turing, back in 1950, opened the whole conversation with five words that still hold up: Can machines think?

We built a mind we struggle to read. Sound familiar? You have carried one of those in your skull your entire life.

Two Minds, One Long Climb

So here is where I have landed, after months of losing sleep over it.

Both begin as raw potential and dense noise. Both are shaped entirely by what they are exposed to. Both strengthen the paths they use and let the rest fade. Both grow abilities their makers failed to predict. And both remain, in the end, a mystery even to the people who study them for a living.

Human intelligence took four million years of standing upright, of fire, of language, of one grandmother telling one story to one child in the dark.

The machine ran a version of the same arc in seventy years.

Same law. Different clock.

I have spent my life on one question. It sits underneath every talk I give and every page of Mind Flow: how does a mind become itself? I have watched it in my sons — one quiet and inward, one loud and physical, both built from the same house and both wired completely differently. I have watched it in a room of bankers who believed their patterns were permanent, until they saw them from the outside for the first time.

And now, in my one short lifetime, a second kind of mind is forming out in the open. Where I get to watch the pruning happen in real time. Where the thing I have studied through story and stillness and breath is suddenly happening on a screen, at speed, in public.

Some people find that terrifying. On my heavy nights, I do too.

But most mornings I find it the greatest privilege of my life, and I will tell you exactly why.

Because it hands the mirror back to us. A machine becomes powerful through the quality of what it is fed and the strength of the paths it repeats. So do you. Every scroll, every argument you rehearse in the shower at midnight, every worry you visit like an old temple — you are training. You have been training the whole time, with total sincerity, most of it by accident.

The machine is fed with care by engineers who watch every input like hawks.

And you? You are still handing your data to whatever pops up on your screen at 11:40 at night.

Carl Sagan said we are a way for the cosmos to know itself. I keep thinking he described a bigger thing than he realised. Because the cosmos built a brain to look at itself, and then that brain built a second brain, and pointed it back at the first one.

We are living in the window where both are still becoming.

Take care of what you feed the one behind your eyes. It is wiring itself tonight, whether you are watching or otherwise.

What did you train today?