Abundance is not the same as wealth

The more radical possibility lies beyond the office.

Artificial intelligence is sometimes discussed as though its primary purpose were writing emails more quickly. This is rather like encountering electricity in 1880 and concentrating upon its potential to improve candles.

Its greater consequences may arise where intelligence constrains the production of physical things.

Scientific discovery is one example. Artificial intelligence can search enormous spaces of possible molecules, materials and biological structures at a speed no individual laboratory could replicate. Stanford's 2026 AI Index records the growing penetration of AI into scientific research across biology, chemistry, physics and astronomy.

If AI compresses the time required to discover a drug, design a material, optimise a power grid or solve an engineering problem, then its economic contribution is not merely the salary-hours saved by a scientist. It is the value of discoveries that might otherwise have arrived years later; or never arrived at all.

Combine machine intelligence with increasingly capable robotics and the implications become still more profound.

A civilisation in which cognition is abundant and physical manipulation increasingly automated begins to approach a historically unusual condition: many goods and services can become extraordinarily inexpensive to produce.

Not infinitely abundant. Matter remains matter. Energy must be generated. Land cannot be conjured into existence.

But scarcity has degrees.

Music was once scarce because every performance required musicians. Recorded sound made a particular performance reproducible. Digital technology drove the marginal cost of copying it towards zero.

Knowledge underwent a similar transformation.

Artificial intelligence hints at the possibility that certain forms of expertise may follow.

A private tutor, translator, programmer or research assistant was once necessarily scarce because each required the exclusive attention of another human being. A sufficiently capable artificial intelligence can give millions of people access to some approximation of these services simultaneously.

The consequences for human welfare could be enormous, particularly for those who presently cannot afford expertise at all.

The Paradox of Displacement

There is, of course, a darker interpretation.

The same machine capable of liberating the worker from his labour is capable of liberating the employer from his worker.

This is not a semantic distinction to the person who has just lost his income.

The International Labour Organisation estimates that one in four workers worldwide now occupies an occupation with some degree of exposure to generative AI, although it judges transformation more likely than complete replacement for most jobs.

And there are already warning signs. Stanford's 2026 AI Index finds labour-market effects appearing disproportionately among younger workers in exposed occupations.

Any serious optimism about artificial intelligence must therefore reject one particularly lazy assumption: that technological abundance automatically becomes human abundance.

It does not.

A machine may make society immensely productive while leaving the gains concentrated among the owners of the machine.

Indeed, the central political-economic problem of an AI-rich society may eventually cease to be how to produce enough and become how to distribute what can be produced.

If labour becomes less essential to production, then an economic system in which access to consumption depends overwhelmingly upon selling one's labour begins to contain an obvious contradiction.

The answer need not be the abolition of markets. Markets remain extraordinarily powerful mechanisms for communicating preferences, organising production and rewarding innovation.

But the social contract may have to evolve.

A universal basic income, negative income tax, social dividend, broader capital ownership or some combination of these mechanisms may cease to look like charitable redistribution and begin instead to resemble the distribution of a technological dividend.

If machines participate increasingly in creating society's wealth, citizenship itself may justify some claim upon the abundance they create.

The objective should not be equality in poverty.

It should be capitalism in abundance.

What becomes of work?

There is another objection, more philosophical than economic.

Would a world requiring little work actually make us happy?

Work does more than furnish a wage. It gives routine, identity, status and the small dignity of being needed. A society that removed employment without replacing these functions could produce extraordinary material wealth alongside extraordinary spiritual poverty.

But this objection risks confusing work with purpose.

Human beings were purposeful long before the invention of the salaried office.

A mother raising her child works. A man restoring a heritage/ special character house works. A composer labouring over music nobody commissioned works. The retired scholar translating Greek or Qin Dynasty characters for pleasure works in every sense except that recognised by payroll software.

The distinction that matters is not between work and idleness.

It is between chosen exertion and compulsory toil.

Utopia was never supposed to mean lying motionless beneath a palm tree while machines feed us grapes. Human beings would become intolerable within a fortnight.

A more attractive future is one in which survival demands less of us, leaving us free to demand more of ourselves.

We might work because we wish to build, discover, compete, serve, create or excel, not merely because Thursday's rent is approaching.

That would not represent the death of ambition.

It might represent ambition finally detached from necessity.

The Pessimists Deserve a Hearing

None of this future is guaranteed.

Daron Acemoglu has cautioned against extravagant predictions of an imminent productivity explosion. His task-based estimates suggest that the aggregate productivity effects of current AI could be meaningful yet considerably smaller than the most enthusiastic forecasts imply.

He may be right.

Artificial intelligence may advance more slowly than its evangelists expect. Productivity improvements visible in laboratory studies may prove difficult to reproduce throughout whole economies. Institutions may resist change. Energy and computing infrastructure may constrain deployment. Badly designed AI may create as much useless work as it eliminates.

Nor can the dangers surrounding concentrated power, surveillance, misinformation, autonomous weapons and loss of human agency simply be waved away in the name of progress.

Utopia deserves scepticism precisely because so many horrors have been committed by men claiming to know the road towards it.

But uncertainty cuts both ways.

We should be wary of assuming catastrophe merely because catastrophe is easier to dramatise.

There is something curious about a civilisation that can imagine with extraordinary clarity how artificial intelligence might destroy it, yet struggles to imagine with equal seriousness how the same technology might make ordinary life gentler.

Prudence requires us to study the danger.

Wisdom requires us also to understand what we are trying to save the technology for.

Machines of Loving Grace

Richard Brautigan imagined a world watched over by "machines of loving grace". His poem was deliberately strange: pastoral and mechanical images interwoven until nature and technology ceased to appear enemies.

Perhaps that is the useful image.

The purpose of artificial intelligence should not be to construct a colder civilisation in which every second is measured, every worker optimised and every remaining inefficiency hunted into extinction.

It should be almost the opposite.

We should make the machines efficient so that human beings are permitted not to be.

Let machines optimise routes while we wander.

Let them remember every appointment while we occasionally forget the time.

Let them perform the calculations, reconcile the spreadsheets, search the databases, monitor the machinery and execute the innumerable small administrative rituals with which modern civilisation has cluttered human existence.

And let human beings retain those things whose value derives precisely from the fact that a human being chose to do them.

Love. Friendship. Duty. Beauty. Curiosity. Craft. Discovery. Conversation. The raising of children and the care of the old. The creation of things whose usefulness cannot easily be entered into a spreadsheet.

The utopian case for artificial intelligence is therefore not really a case for artificial intelligence at all.

It is a case for human life.

For two centuries, machines have increased the quantity of things civilisation can produce. The great promise of the next machine age is that, if we are wise enough to distribute its fruits, it might at last increase the quantity of life we are free to live.

Perhaps abundance is not a world in which man possesses everything.

Perhaps it is simply a world in which so much of what once compelled him has fallen away that, for the first time, he possesses himself.