ClaroscuroEl sombrerero con diagnóstico

An article on automation, pensions and life without work

· 12 min read · Translated from the Spanish original. Read the Spanish original

Share
RSS

I. Fifteen hours

In 1930, with the Great Depression at its peak and half the world queuing for soup, John Maynard Keynes had the gall to get optimistic. He wrote that in a hundred years we would be so rich that three hours a day, fifteen a week, would be enough – and even that just to keep the habit. The date he gave was 2030. We’ve got four years left to meet it, and it doesn’t look like we will.

Half the prophecy has been fulfilled with flying colours: we’re far richer than in 1930. The other half fell flat. We’re still slogging through thirty‑five to forty hours a week, twenty‑five to thirty more than we’re owed, because every bit of productivity we gain is swallowed up in goods and capital, and not a minute in free time. Keynes got the size of the pie right and screwed up what we’d do with it.

The finest bit of the essay, all things considered, was a worry that then sounded like a rich‑man’s joke. Keynes was scared of the day after. A man who loses his job is left alone with the question of what to do with the hours, and he feared the general answer would be a depression of the sort where you don’t even get out of bed. For an economist to say that in 1930 is something.

Almost a century later the issue is back, this time with machines that write and robots that haul boxes. I’m backing the optimistic version, with its conditions, and I’m laying it out in full so you can see the seams.

Automation will free us from work and solve the pension problem with brutal productivity gains, provided it’s taxed properly and the rents are shared between those who lose jobs and the pensioners, leaving a technical elite to run the AI‑driven systems and the working robots, or even a day when AI upgrades itself and maintains itself without human intervention.

II. The pension count

I’ll start with the numbers, which are stubborn. In Spain there are about twenty‑two million Social Security contributors and over ten million pensions, so we’re looking at roughly two contributors per pension. Spending runs at about thirteen per cent of GDP, and the worst is yet to come: the peak is expected around 2050, when the whole baby‑boom generation – which is large and wants to cash in – retires.

The system is a pay‑as‑you‑go one, meaning today’s pensions are paid with today’s wages. Now stick a machine in. If it replaces ten workers, ten wages disappear that were supporting five pensions. The machine produces the same or more, but it doesn’t pay contributions, it has no affiliation number and it has no intention of registering. Under the current rules, the better the invention works, the worse the accounts balance.

The solution isn’t a secret. What salaries used to cover will have to be covered by capital and automated production, via taxation. If wealth moves, the tax collector has to move doors.

It remains to see how much new wealth will appear, and the sages can’t agree on the order of magnitude. Daron Acemoglu, Nobel laureate 2024, crunched the numbers task by task, which can be automated and how much each saves, and he gets less than one per cent extra growth over ten years. Goldman Sachs puts it at seven. Between the Nobel and the bank there’s a seven‑fold gap, so it’s wise to distrust anyone who talks about this with great certainty, myself included.

There are also things that can’t be sped up. A string quartet still needs four musicians and the same amount of time as in Mozart’s day, and no engineer can squeeze more notes per hour out of it. William Baumol and William Bowen explained this in 1966 and called it the cost disease: what doesn’t gain productivity gets more expensive relative to what does. On that list are caring for the sick, teaching children and looking after the elderly. Take note, because it will come up again.

III. Who gets charged

Saying we should tax the machines is easy. The European Parliament looked at it in 2017, with a robot tax, and tossed it out for a difficulty that sounds legalistic but is fundamental: there’s no way to define the taxable event, the concrete thing you actually pay for. A mechanical arm can be counted. A programme that does the work of half a bookkeeping firm has no arms, no licence plate, no place to stick a sticker on.

In 2021 Sam Altman, the head of OpenAI, proposed something else. Instead of charging per machine, set up a fund with stakes in the big firms and land, and distribute its returns to the whole adult population. The idea looked decent. Five years have passed and nobody has launched it, not even its author, who could do something about it.

And it’s not that funds of that sort are a pipe‑dream. Alaska has been paying an annual oil dividend to its residents since 1982, and Alaskans aren’t exactly Bolsheviks. Norway puts its oil rents into a state pension fund that is one of the world’s biggest investors. It can be done. You just have to want it.

The serious obstacle is that everyone has to want it at the same time. If one country taxes machines and the neighbour doesn’t, capital packs its bags – which is what it does best. The precedent doesn’t inspire euphoria: to agree in 2021 on a minimum fifteen per cent tax on multinationals, with the OECD pushing, more than a hundred and thirty countries needed years of meetings. And that was a minimum.

IV. Who decides the split

Everything workers have won has been won by stopping. The strike, which comes from holgar – to be idle – is exactly that: the producer stops producing until he is heard. That’s how they won the working day, holidays and their own pensions. The bargaining power was simple – without them the factory didn’t run.

Now picture a factory that runs itself. The day production depends on machines, the person who has lost his job can’t stop anything, because there’s nothing of his to stop. He’s left with the vote and the protest, which are fine, but they’ve never scared a board of directors the way a silent assembly line does.

From that comes a conclusion with an expiry date. The split of automation rents must be written into law beforehand, while we still matter. You negotiate when you have something to bargain with. After that you can only ask.

V. Morlocks

H. G. Wells told in The Time Machine how things end up if you let them run wild. In the far future humanity has split into two species: the Eloi, handsome, idle and pretty stupid, who live in the sun, and the Morlocks, who live underground, keep the machines running and at night climb up to eat the Eloi. It’s a novel, but the division between those who enjoy the system and those who know how it works is already noted.

The technical elite in my thesis isn’t a hypothesis. It already exists and fits in a very small space. The machines that make the most advanced chips are built by a single firm in the Netherlands, and the chips are assembled almost entirely in Taiwan. The automated civilisation hangs on a couple of nails.

Wells, writing in 1895, cast the Morlocks as the working class and the Eloi as the rich. The times allowed that. What his fable shows, perhaps unintentionally, is that power ends up in the hands of those who retain the knowledge and keep the apparatus on which everything else depends. The Morlocks rule because they’re the only ones who know how to fix it.

There are two defences against that, and neither is new. Basic supplies – water and electricity – must stay in public hands. Technical know‑how has to be taught with the doors open, in schools and universities, so anyone can get into the elite and the elite doesn’t become a hereditary caste.

VI. The last invention

The most daring part of the thesis, the self‑improving machine, has a father. In 1965 mathematician I. J. Good, who had worked with Alan Turing cracking Nazi codes, spoke of an intelligence explosion: a machine smarter than us would design an even smarter one, and so on. He said it would be the last invention humanity would need to make, adding a condition that is less often quoted: “provided the machine is tame enough to tell us how to keep it under control”.

Some of that is already peeking through. In 2025 Google unveiled an AI system that designs algorithms, and it has already improved workload distribution and training inside the company’s own data centres. The machine tuning the machine, for now, in internal plumbing matters.

In the physical world everything moves slower, and for a funny reason. These intelligences find easy what we find hard, like playing chess, and find hard what any kid can do, like turning on a tap. Laying cable, tightening a nut in an awkward spot and fixing a leak will remain human‑hand work for quite a while. The plumber will outlast the chess player.

E. M. Forster wrote in 1909 a short story, “The Machine Stops”, which is the warning Keynes’s essay lacks. Humanity lives underground, tended to entirely by a Machine, and over the centuries ends up worshipping it. One day the Machine starts to fail, and it turns out nobody knows how to repair it. Everything collapses. Even if the machine eventually runs itself, someone has to keep knowing how it works, for the same reason we keep house keys even when the door is automatic.

VII. Sunday neurosis

What remains is Keynes’s fear, which hits me closest. In 1919 Hungarian psychoanalyst Sándor Ferenczi described Sonntagsneurose: patients who got miserable on Sundays, precisely because the day they were ordered to work was missing. Come Monday and they were fine.

In 1930 the Marienthal factory shut down, a town near Vienna that lived entirely off it, and three investigators – Marie Jahoda, Paul Lazarsfeld and Hans Zeisel – went to see what happened to a town on strike. They reported it in The Unemployed of Marienthal. They perhaps expected a revolt and found a town dimming out. People moved slower. With all the time in the world ahead, they stopped going to the municipal library.

Jahoda drew from that a list of what a job gives besides the wage, and called them latent functions. It gives a schedule that structures the day. It gives interaction with people who aren’t family. It gives the feeling of being part of something collective. It gives a place in society, and forces regular activity. Anyone who’s been on long‑term sick leave – and I am – can recite the list without having read it.

The Finns tried the reverse. Between 2017 and 2018 their social security agency, Kela, gave €560 a month, unconditionally, to two thousand unemployed. Over two years that’s €13 440 per head. The result was modest and respectable: stress fell among recipients and employment barely moved. Money removes anxiety, which is no small thing. Sunday still isn’t filled.

VIII. Perpetual Sunday

The underlying problem is that we are what we do. We ask “What do you do?” before almost anything else, and the person with no answer is left without a introduction. Separating what you’re worth from what you work on can’t be fixed with a decree. It will take several generations, and the first will get through it unevenly.

The Greeks phrased the issue better. In their free time they called it scholé, which is where our word school comes from, and they regarded it as serious time – the time to learn and attend to civic affairs. In 1932 Bertrand Russell revived the idea in In Praise of Idleness, where he proposed a four‑hour day and warned that leisure must be taught, because anyone who hasn’t learned to use it gets bored or does harm.

The proposal, stripped down, has four legs. An automated infrastructure that sustains a small, open elite. Taxes on capital that feed a common fund for pensions and rents. The split guaranteed by law before we lose the strength to demand it. And public institutions designed for leisure, which is the leg that always gets forgotten.

For that fourth leg Baumol and Bowen return with their list of jobs that don’t get cheaper: teaching, caring and accompanying. Those are precisely the things machines do worse and where people are always in short supply. There’s more than enough occupation for all the free time left, and it’s the way to keep perpetual Sunday from turning into Marienthal.

Viktor Frankl tried it in Vienna, with unemployed youths who came to his office sunk. He put them as volunteers in libraries and adult‑education schools. They earned nothing, and their depression lifted. He’d given them a place to go in the morning.

Works cited

  • John Maynard Keynes, “The Economic Possibilities for Our Grandchildren” (1930).
  • Ministry of Inclusion, Social Security and Migration, pension‑spending projections (2025).
  • Daron Acemoglu, The Simple Macroeconomics of AI (2024); Goldman Sachs (2023).
  • William J. Baumol and William G. Bowen, Performing Arts: The Economic Dilemma (1966).
  • European Parliament, resolution on robotics (2017). Sam Altman, “Moore’s Law for Everything” (2021). OECD, global minimum tax (2021).
  • H. G. Wells, The Time Machine (1895). E. M. Forster, “The Machine Stops” (1909).
  • I. J. Good, “Speculations Concerning the First Ultraintelligent Machine” (1965). Google DeepMind, AlphaEvolve (2025).
  • Sándor Ferenczi, “Sonntagsneurosen” (1919). Marie Jahoda, Paul F. Lazarsfeld and Hans Zeisel, The Unemployed of Marienthal (1933).
  • Kela, Finnish basic‑income experiment (2020). Bertrand Russell, “In Praise of Idleness” (1932). Viktor E. Frankl.
Share this article
RSS