Imagine making an inventory of everything human beings do.

Driving a lorry. Diagnosing an illness. Writing software. Designing a building. Teaching a class. Preparing a set of accounts. Composing music. Managing a warehouse. Negotiating a contract. Caring for a garden. Deciding which experiments a laboratory should run next.

Now consider each activity in turn and ask whether sufficiently capable AI and robotics could eventually perform it.

The honest answer, for most of the list, is probably yes.

Some tasks will be harder than others. Some will remain uneconomic to automate for a long time. Some require access to physical environments that are difficult to control. Law, custom and personal preference will preserve human involvement in places where it is no longer technically necessary.

But if the question is merely what machines might ultimately be capable of doing, the exceptions are unlikely to provide much comfort. It is dangerous to build a vision of humanity’s economic future around a shrinking collection of tasks that we hope will prove impossible to automate.

This leads to a familiar search for protected corners. We may still want to watch humans play football even if robots become better players, just as people still watch humans play chess when computers are stronger. We may pay for handmade objects because a person made them, value live performance because it can go wrong, or prefer care delivered by somebody who can share our experience.

All of that may be true. Yet it accepts the most important assumption in the original thought experiment: that the list of things worth doing is fixed.

It is not.

A photograph is not a forecast

The inventory presents a static picture of human activity. It takes everything we currently do, holds our aims constant, and introduces increasingly capable machines. Under those conditions, automation can only reduce the human share. Every task transferred to a machine leaves less behind.

But life does not develop under static conditions.

Every substantial change in our capabilities changes the circumstances in which people live. It removes old constraints, creates new expectations and exposes problems that were previously hidden behind more immediate ones. People begin to want things that would once have seemed impractical, extravagant or impossible. They also encounter consequences nobody had reason to consider before the change occurred.

The result is not the same world with fewer human tasks. It is a different world with a different set of possible aims.

Before widespread electrification, nobody could make a complete inventory of the uses to which cheap, dependable power would eventually be put. Before the internet, a list of valuable activities could not have included all the institutions, relationships, businesses and cultural forms that continuous global communication would make possible. These technologies did replace existing labour. They also changed the environment so profoundly that later generations found worthwhile things to do that earlier generations could not have specified.

AI differs from those technologies in important ways. It can substitute for a much wider range of human capabilities, including capabilities used to invent new work. We should not assume that history will repeat itself in the comforting form of familiar jobs disappearing and a predictable number of new jobs taking their place.

But the deeper lesson still matters. A society’s objectives are not an inventory prepared in advance. They evolve with its capabilities.

Capability changes desire

Suppose AI makes it possible to give every child a patient, highly capable personal tutor. At first, the objective may be to teach the existing curriculum more effectively.

If that succeeds, the questions change. What should children learn when individual explanation is abundant? Which parts of education are really about knowledge, and which are about confidence, cooperation, judgment or discovering what is worth caring about? What forms of learning become possible when a school is no longer organised around the limits of one teacher addressing thirty pupils at once?

Solving one constraint does not finish education. It reveals a larger design space.

The same pattern could appear in medicine. If diagnosis becomes faster and more accurate, attention moves towards prevention, individual variation, quality of life and conditions that received little research because more common illnesses consumed the available capacity. Better treatment creates longer lives, which create new medical, social and personal questions. Success changes the population to which the next success must respond.

Or consider a small business. Once routine administration, scheduling, purchasing and reporting can be handled cheaply, the owner is not obliged to spend the saved capacity reproducing the same business at lower cost. They can offer a level of service that was previously uneconomic, adapt products to individual customers, investigate persistent problems, enter smaller markets or attempt something that would once have required a much larger organisation.

Not every new possibility will be wise or valuable. Greater capability produces trivial appetites as readily as noble ambitions. It can generate surveillance, manipulation and waste as well as better education and medicine. The point is not that technological progress automatically improves our objectives. It is that it changes the range of objectives available to us.

Capability does not merely satisfy desire. It helps form desire.

The frontier moves

This is why the attempt to locate the final residue of human work may be looking in the wrong direction.

If AI and robotics eventually become able to perform nearly every activity on today’s inventory, humanity will not still be standing at today’s starting point. We will be living with the results of everything that has already been automated. We will know more, expect more and be able to contemplate projects whose scale or complexity currently puts them beyond serious consideration.

At that point, intelligence can be applied to the next frontier: cleaner cities, richer forms of education, more resilient institutions, better ways of ageing, new scientific questions, more beautiful environments, deeper exploration of the oceans and space, or aims we cannot yet name because the conditions that would make them intelligible do not exist.

Each advance changes what can be attempted next. Each new attempt produces information, consequences and disagreements. Those create further opportunities for observation, judgment and invention.

The frontier moves because the world after a solution is not the world before it.

This does not mean that material resources are infinite or that every system can improve without limit along every dimension. The planet has boundaries. People have finite time, and some goods compete with one another. Every choice carries opportunity costs.

But the field of possible improvement may still be effectively inexhaustible. A city can become safer, more beautiful, more accessible, more sociable or more ecologically sustainable in countless combinations. A longer life raises questions about how those added years should be lived. A scientific discovery creates more questions than it closes. Even where physical limits are fixed, the ways we organise experience within them are not exhausted by our current imagination.

For most of history, exploration of this field was constrained by scarce intelligence and scarce execution. We could identify more worthwhile problems than we had people, expertise, money or institutional capacity to address. AI relaxes part of that constraint. In doing so, it may reveal how much larger the field really is.

This is not an argument for human uniqueness

There is an obvious objection. Why assume that people will be needed to identify the next frontier? If AI can perform today’s work, it may also propose new scientific questions, discover unmet needs, design institutions and generate ambitions more effectively than we can.

That is possible. A hopeful account of the future should not depend on an assertion that machines will never originate goals or insights.

The dynamic argument does not require such a boundary. AI can help us see possibilities we would have missed. It can model consequences, compare alternatives and produce ideas that no individual person would have reached. The expanding frontier can absorb artificial intelligence as well as human intelligence.

But people are not merely an inferior source of proposals competing against a better one. We are participants in the world being changed.

We experience whether a new system is liberating or oppressive, whether an efficiency has removed drudgery or stripped meaning from something important, whether a measured improvement corresponds to a life that actually feels better. New capabilities alter our relationships, expectations and sense of what is fair. Those responses become part of the reality to which the next generation of changes must adapt.

An AI may identify that reality, interpret it and suggest a response. People may do the same. Often the useful result will come from interaction between the two. What matters is that the process remains open: change produces new conditions; new conditions produce new knowledge and preferences; intelligence acts upon them; and the cycle begins again.

There need not be a final list on which every worthwhile objective has been crossed out.

An open frontier does not guarantee a good economy

None of this proves that technological change will provide everyone with secure work or a fair income.

A society can possess an enormous frontier of possible improvement while denying most people the resources or authority to contribute to it. The owners of powerful systems may capture the gains. Institutions may use AI to reduce costs without allowing workers, customers or communities to shape what replaces the old arrangement. People may have valuable insight into what should happen next and no credible route for turning that insight into influence or economic value.

Transitions can also be brutal even when the distant future contains abundant possibilities. A new field of useful activity does not necessarily appear in the same place, at the same time, or for the same people whose livelihoods have disappeared. Telling somebody that humanity’s ambitions will eventually expand is not an answer to the loss of their income next year.

The dynamic frontier is therefore a possibility, not a distribution system.

If we want broad prosperity, we will need institutions that allow people to participate in choosing aims, testing ideas and sharing in the value of successful improvements. Education, ownership, attribution and access to capable systems will all matter. So will the willingness to direct some of our new capacity towards public goods that markets alone may neglect.

The future can contain more worthwhile contribution without automatically rewarding the people who make it.

A better question

Predictions about AI often begin by surveying the present and subtracting everything machines may learn to do.

The result is understandably bleak. If the world contains a fixed quantity of useful thought and action, then a machine capable of supplying more of both must leave less room for people. Hope survives only in activities protected by technical difficulty, law or sentiment.

But a civilisation is not a fixed collection of tasks. It is a continuing response to its own changing capabilities.

AI may allow us to achieve aims that currently consume enormous amounts of human effort. Achieving them will alter our circumstances. Those circumstances will reveal new problems and make greater ambitions credible. Human beings and machines can then apply their combined intelligence to what has become possible next.

The important question is not which items on today’s list will remain exclusively human.

It is what becomes worth attempting when intelligence and execution are no longer the constraints they once were.

That question has no final answer. Every good answer changes the world in which we ask it.