A letter to my children

I am writing this somewhere over France, on a flight from Cork to Nice.

I cannot speak.

It is now three and a half weeks since my surgery and I am still under strict instructions to stay completely silent, although I admit I did crack two days ago and try it out.

So I am travelling almost mute, with a phone full of opinions and an increasingly elaborate system of pointing at things.

There are worse problems to have.

If you are going to be mute, you may as well be mute in Nice.

This started about an hour before boarding, when I was sitting in Cork Airport eating a nine-euro yoghurt that tasted vaguely of compost and watching a short clip of Tristan Harris talking about AI, governments and what happens when human beings stop being responsible for most of the economic output of a country.

I downloaded it and started writing a WhatsApp message to my son.

This is what parents do. We encounter something mildly alarming and immediately decide our children need a lecture about it.

About three paragraphs in, two things became obvious.

First, this was not a WhatsApp message.

Second, WhatsApp is currently the nearest thing I have to a kitchen table.

So this is the conversation I would be having across that table if I could actually bloody speak.

And before we go any further, this is not really an article about computer science.

That is simply where I can see what is happening most clearly because I have spent thirty five years working there.

The thing I am worried about is much bigger.

It is graduate work.

Possibly work itself.

And I don’t think most of us have a mental model big enough for what might be coming.

We imagine the future using the worst thing we remember

This is the bit I keep coming back to.

Human beings are absolutely rubbish at imagining things for which we have no personal reference point.

Ask somebody my age what a really bad economy looks like and we reach backwards.

The early 1980s.

Mass unemployment.

Factories closing.

People leaving school with nowhere to go.

Then, if you are younger, 2008.

Banks collapsing. Property crashing. Businesses disappearing. Graduate jobs drying up.

That is what catastrophe looks like to us.

And it was horrible.

But there is something really important about both of those examples.

The economy still needed people.

The factories might have shut because demand collapsed.

The bank might have stopped hiring graduates.

The building sites might have gone quiet.

But nobody seriously believed that when the economy recovered we would no longer need accountants, builders, solicitors, programmers, analysts, managers and engineers.

The machine had stopped.

Eventually it would start again.

People waited for the jobs to come back.

That is the framework almost every adult is using when they think about AI.

We hear that millions of jobs may disappear and instinctively translate it into something we already understand.

A very bad recession.

A painful adjustment.

A few dreadful years.

Then some new industries appear, everyone retrains and off we go again.

Why?

Why are we assuming that?

A recession is a collapse in demand for human labour.

AI may be a collapse in the need for some human labour.

Those are not the same thing at all.

If a company stops hiring accountants because there has been a recession, those jobs can return when business recovers.

If it stops hiring ten junior accountants because one senior accountant with a collection of AI systems can do the work, what exactly are we waiting to recover?

That is the bit I think people have not psychologically absorbed.

This may not be another turn of the economic wheel.

Some of the wheel may be coming off.

The government problem

This was the part of the Harris clip that really got me.

Governments invest in people because people are economically useful.

That sounds horribly cold, so let me put some clothes on it.

The state educates you.

It trains you.

It builds universities.

It subsidises courses.

It provides healthcare.

It spends an extraordinary amount of money turning a five-year-old into a healthy, educated, economically productive adult.

Obviously there are moral and social reasons for doing those things too.

But there is also a brutally simple economic bargain underneath it.

Educated people work.

They build things.

They run businesses.

They earn money.

They pay tax.

They create GDP.

Spend money on a child now and, if all goes reasonably well, you eventually get an economically productive adult.

That bargain has existed for generations.

The point I took from Harris was: what happens to that bargain if machines generate most of the economic value instead?

Imagine a country where enormous AI companies and data centres are producing most of the taxable wealth.

Why does that government have the same economic incentive to spend a fortune turning every citizen into a highly educated worker?

It doesn’t.

And no, I don’t think this happens next Tuesday.

Quite the opposite.

Governments are slow as fuck.

Thank God.

That slowness may be one of the greatest advantages my children currently have.

The machinery of the state is still operating on the old assumptions.

It still believes educating you is an investment.

It will still spend money on you.

It will still build colleges for you.

It will still offer courses, apprenticeships, degrees, grants, training schemes and all the other infrastructure created for a world where productive citizens were the engine of the economy.

If Harris’s sort of horizon is even approximately right, that may not always be true.

Nineteen years is nothing.

My eldest children will still be young adults.

So my advice to them is very simple.

Use it.

Use every bloody bit of it.

But do not confuse the fact that the old system is still offering you a path with evidence that the path still leads where it used to.

That is a very different thing.

More gratuitous pictures of me on holiday

I happen to see this from software

This is where computer science comes in.

Not because programmers are uniquely doomed.

We aren’t.

It is simply the industry I know well enough to see what is happening without somebody explaining it to me.

I have spent thirty years building software for banks, insurers, startups, universities and whoever else was willing to pay the invoice.

And for the last eight months I have been building what is, for all practical purposes, a small company staffed by AI.

I mean that fairly literally.

I have coding teams.

They report to synthetic project managers.

There is governance.

There are reviewers.

There is marketing.

There is a board.

The machines argue with one another.

Michelin Star Night!! Delicious it was too

They challenge work.

They send things back.

They decide that another synthetic employee has misunderstood the requirement.

I occasionally find myself reading the minutes of a disagreement between three computers and wondering how exactly my life arrived here.

But here is what matters.

The interesting thing is not that they can code.

Everybody obsesses over that because code is visible and slightly magical if you don’t write it yourself.

The interesting thing is everything around the coding.

Give the system a problem.

It can investigate it.

Read the existing material.

Break the work down.

Research.

Draft.

Compare.

Test.

Review.

Summarise.

Coordinate.

Report.

Escalate the odd thing it cannot resolve.

That is not “programming”.

That is the basic shape of an astonishing amount of graduate work.

Think about the first few years of almost any white-collar career.

You are not running Goldman Sachs at twenty-two.

You are not arguing before the Supreme Court.

You are not restructuring BP.

You are doing the grunt work while somebody more experienced checks it.

Read these documents.

Research this question.

Compare these numbers.

Prepare a first draft.

Check this contract.

Build this spreadsheet.

Summarise this meeting.

Look through these cases.

Prepare the deck.

Write the report.

Find the discrepancy.

Ask these five people for an update and tell me which one has buggered off without replying.

That is graduate work.

And machines are becoming extraordinarily good at graduate-shaped work.

Software is just where I happened to notice the bodies first.

We may be automating the nursery

There is another problem buried inside this that I don’t hear discussed nearly enough.

Everybody says some version of this:

“AI won’t replace the really experienced people.”

Maybe.

I hope they’re right.

But where exactly do experienced people come from?

They aren’t discovered in a cupboard at forty-three.

Every brilliant accountant was once a fairly useless junior accountant.

Every senior software architect once wrote terrible code.

Every partner in a law firm once sat somewhere reading documents nobody else wanted to read.

Every consultant who can walk into a room today and understand a business in twenty minutes spent years being the idiot carrying the deck.

That is how expertise is manufactured.

You employ inexperienced people to do relatively simple work.

They get things wrong.

Someone corrects them.

They do slightly harder work.

They get less wrong.

Twenty years later everybody calls them an expert and pays them an offensive amount of money.

But what happens when the simple work is no longer economically sensible to give to a human?

If a machine can do the work of five graduates, the company may still need the senior person.

It simply stops hiring five graduates.

Excellent.

Very efficient.

Come back in fifteen years and tell me where the next senior person is coming from.

We may be automating the nursery.

That is a much bigger problem than losing some entry-level jobs.

We could be breaking the mechanism by which whole professions create their next generation.

I don’t know which jobs are safe

This is another place where I think most AI career advice goes wrong.

Somebody writes a list.

Ten Careers AI Cannot Replace!

Number seven will shock you.

I haven’t a clue.

Neither do they.

Nobody knows.

I would be very suspicious of anyone who claims to know exactly which professions will be thriving in twenty years.

Twenty years ago Facebook was a website where university students poked one another.

Predicting 2046 with great confidence seems ambitious.

So I am not going to tell my children that plumbing is definitely safe or medicine is definitely safe or law is definitely doomed.

I don’t know.

What I can do is look at the incentives and place bets.

If these were my three lives — which, inconveniently, they are — these are the characteristics I would currently bet on.

Be difficult to substitute

The first card is geography.

If somebody needs you physically standing in front of the thing that is broken, you have some protection.

This is why builders, electricians, plumbers and other skilled trades deserve much more serious consideration than my generation was taught to give them.

For thirty years we shoved every academically able child towards university and quietly implied that trades were where you went if the academic thing hadn’t quite worked out.

That was extraordinarily stupid.

Now try getting a good electrician.

You cannot outsource Mrs Jones’s blocked toilet to Bangalore.

ChatGPT cannot currently rewire her kitchen.

And if you are bright, organised and commercially minded, the opportunity is even bigger.

Learn the trade properly.

Turn up when you say you will.

Wear a clean shirt.

Answer your phone.

Send an actual quote.

Have a website that wasn’t designed by your nephew in 2009.

You will look like bloody NASA.

Your competition may quite literally be a sweaty man displaying a hairy Grand Canyon while mumbling at somebody’s nan.

The bar is not always high.

But I would not make the mistake of thinking physical work is permanently protected.

The robots are coming there too.

If you own the gardening company when autonomous machines become genuinely useful, buy the machines.

Do not become the bloke standing beside a lawnmower announcing that a robot will never replace honest hard work.

He will go bust with enormous dignity.

If your work can travel, become very specific

The second thing I would bet on is specialisation.

I would not want to enter the next twenty years as a generic graduate who “does business”.

Or “does marketing”.

Or “does IT”.

Or “does consulting”.

That is a horrible place to compete because your competition is everybody.

Human and machine.

I survive in technology partly because I have become absurdly specific.

I do a particular sort of backend platform work for universities.

That is not exactly going to get me invited onto Graham Norton.

But in my tiny pond, the same names keep appearing.

One piece of work leads to another.

People know what I do.

That matters enormously.

So if my children choose work that can be delivered from anywhere, I would tell them to become almost comically specific.

Don’t be a photographer.

Become the person everybody in Cork thinks of for photographing boutique hotels.

Don’t “do finance”.

Know one corner of it so well that people phone you when that exact problem appears.

Don’t be one more human-shaped item in a global pile of CVs.

Own a tiny category.

Which unfortunately leads me to something parents my age generally hate.

You are probably going to have to make content.

Sorry.

I don’t mean dancing on TikTok.

I mean being publicly associated with what you know.

Write.

Film.

Explain.

Show your work.

Teach people something.

Become recognisable.

Expertise hidden in your head is lovely, but the market cannot buy something it doesn’t know exists.

Our generation finds this faintly embarrassing.

Yours doesn’t.

For once, you have the advantage.

Embrace the cringe.

Human may become a luxury feature

There is a strange flip side to all of this.

As machines become capable of doing more, genuinely human things may become more valuable precisely because they are human.

Music has already given us a preview.

Recorded music became effectively free.

So enormous value moved towards the live performance.

You can stream the song for almost nothing.

You will pay two hundred quid to stand sixty metres from the person who recorded it while somebody behind you spills lager down your back.

Why?

Because the live bit is scarce.

I think some version of that happens everywhere.

Human made.

Human delivered.

Human present.

We may start paying extra for those things.

“Made by a person” could eventually become the equivalent of “handmade”.

Which would be quite funny after spending a century inventing machinery specifically to get rid of the person.

Own things

This is the least exciting advice in the article and possibly the most important.

Own assets.

Early.

Tiny amounts count.

If labour becomes less scarce, ownership becomes more important.

Own part of a business.

Own equity.

Eventually own property if you can.

Own productive things.

The person who only sells their labour is completely exposed to what happens to the price of labour.

The person who owns part of the machine has another card.

I wish somebody had explained that to me properly at sixteen.

So I am explaining it to you.

And use the system while it still wants to invest in you

This may be the biggest thing I want my children to understand.

Right now, society is still spending money preparing you for adulthood.

Use that period ruthlessly.

Learn things.

Try things.

Get qualifications where qualifications actually matter.

Take the apprenticeship.

Use the university.

Use the library.

Use the training programme.

Use every subsidised opportunity that makes you more capable.

But please do not blindly follow a path simply because adults recognise it.

We recognise it because it was our path.

That is not the same as it being yours.

And this is where I think parents have to become much more humble.

We are giving advice about a future we have never experienced.

When I started work, the internet barely existed.

When today’s careers teachers started work, generative AI did not exist.

Nobody raised us to compete economically with machines that can read, write, reason, code, analyse and operate software.

There isn’t an older generation with wisdom about this.

There is no Grandpa who went through the AI transition and can tell us how it turned out.

We are first.

So the least we can do is stop pretending that the old map must still be right.

The hard part

I don’t think everybody adapts to this.

That is the bit people really dislike hearing.

Whenever somebody raises the possibility of mass automation, someone says:

“But technology always creates new jobs.”

Maybe it will.

I desperately hope so.

But again, notice what we are doing.

We are reaching backwards.

The loom created jobs.

The motorcar created jobs.

The computer created jobs.

Therefore AI will create enough new jobs.

That isn’t an argument.

It is a historical pattern.

Patterns continue until they don’t.

The important difference is that all those earlier machines were tools for human beings.

This thing is increasingly capable of performing the cognitive work itself.

Perhaps we create enormous new categories of human work that none of us can currently imagine.

Wonderful.

But I would not bet my children’s lives on it simply because the Industrial Revolution eventually worked out.

There is a genuinely ugly possible future in which enormous economic output is produced by a relatively small number of companies using machines, while a very large human population becomes economically peripheral.

They are not starving.

They are not necessarily miserable.

Governments transfer money to them.

Entertainment becomes almost infinite.

GTA 20 is presumably excellent.

But the relationship between citizen and state has changed.

The state is no longer investing in you because it needs what you will become.

It is maintaining you because the economy no longer particularly does.

That is the future in the Harris argument that bothered me.

Not killer robots.

Not Terminator.

Not even unemployment in the way my generation understands unemployment.

Economic irrelevance.

And I think that possibility deserves considerably more thought than it is getting.

What my mother told me

My mother used to tell me that, in the grand scheme of things, whether I got an A or a C in maths probably wasn’t going to change my life.

She was trying to make an anxious child feel better.

It was about 1985.

Forty years later I think she may accidentally have made an economic forecast.

Because I can imagine a world ten or fifteen years from now full of straight-A students carrying beautiful qualifications for work that barely exists.

And somewhere else there will be a kid with thoroughly average grades running three vans, two robots and a weird little niche business nobody at school had heard of.

The grades were never really the game.

Reading the board is the game.

And the board has changed more in the last few years than most of us have emotionally caught up with.

So to my three:

I am not telling you not to go to university.

I am not telling you to become plumbers.

I am certainly not telling you that I know what your working lives will look like.

I don’t.

That is rather the point.

I am telling you not to mistake the world that educated your parents for the world that will employ you.

You have something incredibly valuable right now.

Time.

A government that still wants to educate you.

A society still built around preparing you for work.

Access to more knowledge than any generation in history.

And, for the moment at least, machines that can make you dramatically more capable rather than simply making you unnecessary.

Use it wisely.

This is the conversation I would have across the kitchen table if I could speak.