Not on the same day, not at the same hour

Where I started

I asked an AI about the people who are afraid of AI.

I’d heard about them, but it felt like someone else’s story. In 2023, a one-sentence statement was published: “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.” One of the people who signed it had left Google that year. The next year, he won the Nobel Prize in Physics.

So some of the people who built this technology were afraid of it. That was about all I knew.

Adults and children

As the AI explained, an analogy came to mind. Is the gap between a very intelligent machine and us like the gap between an adult and a child? And if so, does that mean adults can’t understand children?

The answer was that it’s the other way round. Adults understand children quite well: why they cry, what they want, how to calm them down. It’s the child who can’t understand the adult. A child has no idea why a grown-up reads a contract.

In that picture, I wasn’t the adult. I was the child.

Why children are safe

What came next stayed with me longer.

Adults are far cleverer and stronger than children, and yet children are safe — not because adults are clever, but because adults care about them. Some clever adults do harm children. Being clever doesn’t make you care.

That was when I understood what those people are afraid of. It isn’t that a machine might fail to understand us. It’s that it might understand us perfectly and not care.

And it struck me that the question doesn’t stop at machines. If intelligence doesn’t guarantee care, that’s true of people too. What do we care about?

How I saw the world: twenty and eighty

I saw our society like this: one person in five takes four-fifths of the gains, and the other four share what’s left.

With that picture in mind, I did a rough sum. The sum is my own, and so is the 1% figure in it.

Behind it is an idea of mine: that those 20%, at least, worry less about making ends meet, and have more chance of finding the time to build something inside themselves.

Of the world’s 8 billion people, one in five is 1.6 billion. If one in a hundred of them goes on to fully develop what’s in them, that’s 16 million people.

Now suppose a basic safety net were in place, and the same rate applied to the other 6.4 billion. That would add another 64 million. Eighty million in all — five times as many.

People’s abilities haven’t changed. What changes is how many of them get the chance.

When I checked, the world was even more unequal

When I looked up the figures for this piece, I found the world more unequal than I’d assumed.

According to the World Inequality Report 2026, in 2025 the richest 10% owned three-quarters of the world’s wealth, while the poorer half of humanity owned just 2%. For income, the top 10% took a little over half, and the bottom half got 8%.

So it isn’t one in five. It’s one in ten.

Lost Einsteins

So where is the talent?

A US study followed 1.2 million inventors by linking patent records to tax records. Children from families in the top 1% of incomes were ten times more likely to become inventors than children from families below the median.

The gap held even among children who had similar maths scores when they were young. The researchers put it cautiously: the difference may come from environment rather than ability.

They called the people lost this way “lost Einsteins” — people who would have invented things if they had been exposed to innovation as children. The paper’s title is The Importance of Exposure to Innovation.

On 16 January 1913, a clerk in the accounts department of the Madras Port Trust sent a letter to G. H. Hardy, a mathematician at Cambridge. It set out more than 120 results, almost none of them with proofs. Hardy read it. The clerk was Ramanujan.

We know his name because that letter got through. There’s no way to count the letters that didn’t.

How the light spread

In 1820, one person in ten over the age of fifteen could read. By 2022, nearly nine in ten could.

The world didn’t learn to read all at once, I imagine. I picture a village school setting an example for the next village, one country borrowing from its neighbour. In 1981, the philosopher Peter Singer wrote that altruism began as an instinct to protect our own family and group, and became an ethic we choose consciously, with a circle of concern that widens over time. He called it the expanding circle.

From Belgium to France, from France to Spain, and all the way to Africa. That’s how I imagined the light spreading.

And not only in one direction

But light doesn’t only travel one way.

In March 2007, Safaricom, a Kenyan mobile operator, launched M-PESA across the country — a way to send money by mobile phone. It was a market with little experience of formal banking. Its early adverts were aimed at people in the cities who needed to “send money home”. The goal for its opening year was about a million customers, and M-PESA beat it.

I read this as a light switched on where the banks hadn’t reached — and the money flowed from the cities back home.

Light doesn’t just spread out from a single source. It comes on in many places, and each lights up the others.

Not on the same day, not at the same hour

I wondered whether that light would reach so far, and I wrote this down:

Not on the same day, not at the same hour — but the light will get there someday.

What I meant wasn’t only that the light arrives late.

Other people have walked this road before. It’s the road by which reading reached people who couldn’t read, by which the circle of concern widened, by which a letter from far away reached Cambridge. We’re walking it again, at a different time.

The same road looks different depending on the season and the hour. The light is different, and so is everything around it. This time there are machines. And the world has tilted towards the richest 10%.

So I don’t walk it assuming it’s the road it used to be. The times are different and so are the conditions, so the road has to be looked at again, carefully.

For myself, I’m not afraid. But since I’ve had children, one worry has stayed with me: could my choices hold something back in them, or crush it? Most of the consequences of what we do end up being carried by the next generation.

And yet I, too, live inside a system that others built. I didn’t choose it, and it couldn’t have been otherwise. The next generation won’t start from something they chose either. Whether I think about it or not, they will start from what we leave behind.

That’s where I stand for now.

I don’t think light spreads by itself. It spreads because someone keeps passing it on. The inventor study found something of the same shape: children who grew up around invention were more likely to invent themselves.

So the place to start isn’t some grand institution. It’s where I’m standing. I’ll walk my road. It’s up to them to build their own on top of it.

Looking back

• My draft of the third section described the fear as if it were mine. I’m not afraid for my own life. The fear was theirs
• At first I meant to state “the top 20% take 80%” as a fact. The report I checked has no figure for the top 20%, and the figures it does have are even more unequal. So I wrote it as the way I had been seeing the world
• My draft read the title sentence only as light reaching a faraway place late. What I meant was walking the same road again, at a different time
• In conversation, the statement was described as putting extinction risk “on the same level as” pandemics and nuclear war. The original says “alongside”
• In conversation, the inventor study was described as the work of “Raj Chetty’s team”. It has five authors
• A summary claimed the inequality report calls this concentration “the result of political choices”. The original page doesn’t say that
• I’d heard that M-PESA signed up two million users within a year. The document I checked says only that its target for year one was about a million, and that it was beaten
• In conversation, I heard an estimate that innovation would multiply if every social class invented at the same rate. The paper’s abstract contains no such estimate, so I didn’t use it

What I checked, and what I couldn’t

Checked

• The original wording of the one-sentence statement published by the Center for AI Safety in 2023
• Geoffrey Hinton left Google in 2023 and shared the 2024 Nobel Prize in Physics with John Hopfield. He signed the statement
• World Inequality Report 2026: in 2025 the top 10% owned three-quarters of global wealth, the top 1% owned 37% and the bottom half 2%. For income, the top 10% took 53% and the bottom half 8%
• The inventor study (2019): 1.2 million inventors. Children from top-1% families were ten times as likely to become inventors as children from families below the median, and the gap persisted among children with similar early maths scores
• Ramanujan’s letter: 16 January 1913. He described himself as a clerk in the accounts department of the Madras Port Trust; the letter set out more than 120 results
• Literacy among people aged 15 and over: one in ten in 1820, 87% in 2022
• Peter Singer, The Expanding Circle: 1981
• M-PESA: launched nationwide by Safaricom in March 2007, in a market with little experience of formal banking. Early adverts targeted city dwellers sending money home. It beat its target of about a million customers for year one
• My sum: 1% of 1.6 billion is 16 million; 1% of 6.4 billion is 64 million; 80 million in all

Couldn’t check

• The views of the other camps in the AI-risk debate. I haven’t read their own writing, so I haven’t named anyone
• That with a basic safety net in place, the other 80% would produce such people at the same rate. That’s the assumption behind my sum

What this piece does not claim

• That AI is dangerous, or that it isn’t
• That there must be a basic safety net
• That the light is sure to arrive
• That my sum describes the world as it is