There is a familiar rhythm to every technological revolution.
First comes the excitement. Then comes the anxiety. Then, inevitably, somebody important tells us that we ought to be frightened.
Andrew Bailey has been doing some of this thinking for us. The Governor of the Bank of England, in his capacity as chairman of the Financial Stability Board, has warned that the rapid development of artificial intelligence presents serious risks to the financial system, particularly through cyberattacks, concentration among a small number of technology companies and the possibility that excessive optimism about AI could contribute to a future market shock.
These are perfectly reasonable things for a central banker to worry about. Indeed, it would be rather alarming if he did not.
But there is a broader conversation about AI which is becoming increasingly dominated by a different fear: that artificial intelligence will make large numbers of people redundant.
That may happen.
But I suspect that we are asking the wrong question.
The interesting question is not how many jobs AI will destroy.
It is how many businesses one person will soon be able to create.
That may sound like a small distinction.
It isn’t.
The company of one
Imagine someone with a good business idea.
Perhaps she has spotted an underserved market. Perhaps she has designed a product. Perhaps she has a better way of providing a service. Perhaps she has spent twenty years working in an industry and knows precisely what is wrong with it.
Under the old economy, having the idea was only the beginning.
She needed a website.
She needed branding.
She needed market research.
She needed somebody to write the copy.
She needed somebody to build the software.
She needed accounts.
She needed administration.
She needed customer support.
She needed sales and marketing.
She needed somebody to analyse the data.
She might need a lawyer, an accountant, a graphic designer and a dozen other people before she had sold a single thing.
And so she did nothing.
This is one of the great hidden constraints of capitalism. We tend to think of employees as people who perform tasks. But employees are also the cost of complexity.
A business is, among other things, a machine for assembling lots of different kinds of human capability.
Artificial intelligence is beginning to change that.
The entrepreneur can now acquire many of those capabilities without acquiring all the people.
AI can research the market, refine the proposition, write and test software, produce marketing material, analyse customers, prepare accounts, answer enquiries, draft contracts, create presentations and automate administration.
It can increasingly do these things not as a glorified word processor, but as an interconnected system of agents working towards an objective.
That means that a person who previously needed a dozen employees may increasingly need only the right AI system.
And this is not some distant speculation. AI tools are already enabling a growing class of one-person companies, while mainstream business publications are beginning to document the phenomenon.
I have been developing one such system myself.
The purpose is not to replace an employee.
It is to allow an entrepreneur to run a business which would previously have required a small organisation.
That distinction matters.
We are measuring the wrong thing
Suppose a company employing one hundred people introduces AI and discovers that it only needs seventy.
There is an obvious story.
Thirty people have lost their jobs.
But that is not the whole economic story.
What if the company can now reduce its prices and sell twice as much?
What if it can enter three new markets?
What if it can develop five products instead of two?
What if its competitors respond by doing the same?
What if entirely new competitors enter the market because the cost of doing so has fallen?
And what if the person who would never previously have been able to establish a company now can?
That last question is almost impossible to answer using conventional labour statistics.
We can count redundancies.
We cannot count the businesses that were never started.
We can count the employees who are no longer required.
We cannot count the products which would never previously have been developed because their inventor could not afford a development team.
We can count jobs.
We cannot easily count economic possibilities.
And that is where I think much of the discussion about AI is going wrong.
Bill Gates wants to reserve some work for humans
Consider Bill Gates.
Gates is hardly a technological Luddite. Yet his latest thinking about AI includes the idea of creating a category of “Human Reserved” occupations — jobs which, even if machines could perform them, should be deliberately protected for people. He has also suggested taxing AI and robots to slow the incentive to replace human labour.
One can understand the instinct.
If technology threatens to displace people, perhaps we should protect the people.
But there is something rather odd about this proposition.
Imagine that, in 1800, somebody had invented a machine capable of doing the work of ten farm labourers.
Would the correct response have been to declare certain agricultural tasks “Human Reserved”?
Or would it have been better to ask what the liberated labour could now do?
The latter question eventually transformed the world.
The great economic achievement of technology has never been that it allows us to employ precisely the same number of people doing precisely the same things.
It is that it allows us to do things which were previously too expensive, too difficult or simply impossible.
And then there is Musk
Elon Musk takes the argument in the opposite direction.
His prediction is that AI and robotics will make work optional. He has described a future of “universal high income”, in which people need not work unless they want to — rather like choosing to play a sport or a video game.
This is certainly an attractive prospect if one’s definition of a good life is freedom from economic necessity.
But there is a curious omission in it.
What if people don’t merely want to consume?
What if they want to create?
What if they want to win?
What if they want to build the better company, design the better product, solve the problem nobody else has solved, discover something, write something, invent something or simply prove that they can?
Musk’s vision seems to assume that the purpose of economic progress is ultimately to liberate us from work.
But work is not the only reason people create things.
Human beings have an extraordinary tendency to find problems where none have been assigned to them and then spend years trying to solve them.
We climb mountains that do not need climbing.
We sail around the world when there is perfectly good air travel.
We build companies when we could simply take a salary.
We write books nobody asked us to write.
We paint pictures nobody needs.
We compete when nobody is forcing us to compete.
The human desire for achievement is not going to disappear because a robot can make our breakfast.
Indeed, if anything, AI may allow far more people to exercise it.
The end of economic necessity is not the end of economics
There is another problem with the idea of a fully post-scarcity world.
There are eight billion people.
Will we all own super-yachts?
Of course not.
And if everybody did, owning a super-yacht would cease to be particularly interesting.
There will still be scarce land.
There will still be desirable locations.
There will still be beautiful things that only one person can own.
There will still be limited access to particular experiences.
There will still be competition for attention.
There will still be status.
There will still be time, which is perhaps the ultimate scarce resource.
Technology can make some things extraordinarily abundant without making everything abundant.
This is basic economics.
AI may eventually make intelligence extraordinarily cheap. Robots may make manufacturing extraordinarily cheap. Energy may become much cheaper. The marginal cost of producing many goods and services may collapse.
Wonderful.
But scarcity does not disappear.
It moves.
And when it moves, markets move with it.
The successful entrepreneur will still have something that other people want.
The best restaurant will still have limited tables.
The best piece of land will still have only one owner.
The most successful company will still have competitors trying to take its customers.
The person who creates something genuinely valuable will still be able to prosper more than the person who does not.
That isn’t a defect in the system.
It is one of the reasons the system works.
UBI is not the same thing as human flourishing
There may well be a role for redistribution in an AI-rich economy.
If technology creates extraordinary wealth, it would be perfectly reasonable to debate how much of that wealth should be redistributed to ensure that nobody is destitute.
Eliminating poverty would be a magnificent achievement.
But poverty and purpose are not the same thing.
Give a twenty-five-year-old entrepreneur enough money to live comfortably for the rest of his life and then offer him the chance to build a company worth £100 million.
Many will take the second option.
Not because they need the money.
Because they want to build the company.
They want to win.
They want to create something that matters.
They want to be the person who did it.
This is why I find the prospect of millions of AI-assisted entrepreneurs considerably more interesting than a future in which everyone receives a government cheque while machines do everything.
The former preserves human agency.
The latter risks making humanity the spectator in its own economy.
The great economic liberation
There is a more optimistic way of looking at technological unemployment.
Suppose one person really can do the work that previously required ten.
Why should our first reaction be to mourn the ten?
Perhaps we should ask what the one person can now do.
And then ask what happens when a million people acquire the same capability.
Suddenly the world contains a million potential businesses that did not previously exist.
A retired engineer can commercialise the invention he has been thinking about for thirty years.
A teacher can build an educational company.
A doctor can develop a specialist information service.
A designer can create a global product brand.
A consultant can turn his accumulated knowledge into software.
A young programmer can build a product without first raising a quarter of a million pounds to employ a development team.
A farmer can develop sophisticated direct-to-consumer sales without employing a marketing department.
A person who has never thought of himself as an entrepreneur may discover that entrepreneurship is no longer primarily about assembling an organisation.
It is about having an idea and being able to execute it.
That is a profound change.
For centuries, economic power has been closely connected to the ability to command other people’s labour.
AI potentially changes that.
It allows an individual to command capability.
That may be one of the most important changes in the history of capitalism.
The one-person company
This is why I think we should stop describing AI primarily as a machine for replacing employees.
It is becoming something rather more interesting:
a machine for creating organisations.
The traditional company might have required a founder and twelve employees.
The AI-assisted company might require the founder and twelve intelligent systems.
The founder still has to make the decisions.
He still has to understand his customer.
He still has to take the risk.
He still has to decide which of the machine’s suggestions are brilliant and which are idiotic.
He still has to persuade people to buy.
But he no longer needs to hire an entire organisation merely to turn his idea into reality.
And that means the real economic transformation may not be the destruction of employment.
It may be the democratisation of enterprise.
We should be encouraging people to have a go
This ought to change the political conversation.
Instead of asking only how we protect workers from AI, perhaps we should ask how we equip people to exploit it.
Instead of trying to preserve every existing job, perhaps we should make it dramatically easier to start a business.
Instead of telling people that AI is coming for their careers, perhaps we should tell them that they now have access to capabilities which, until very recently, would have required a small company to provide.
The response to AI should not be to put human beings in a technological nature reserve.
Nor should it necessarily be to pay everyone to sit outside the economy while machines run it.
It should be to make participation in the economy easier.
That means encouraging entrepreneurship, reducing unnecessary barriers to starting companies, making capital available to people with good ideas and — above all — teaching people how to use these extraordinary new tools.
Because Britain has no shortage of ideas.
What it has often lacked is the means of turning them into businesses.
That may be about to change.
The question we should be asking
There will certainly be disruption.
Some occupations will shrink. Some businesses will disappear. Some people will lose their jobs and need to learn new skills.
None of that should be minimised.
But disruption is not catastrophe.
The history of technological progress is largely the history of making things cheaper, easier and more accessible — and then discovering that human beings are extraordinarily inventive in finding new things to do with the capabilities they have acquired.
We are currently measuring the coming AI revolution in redundancies because redundancies are easy to count.
We should also be counting the businesses which are about to become possible.
The products which will now be developed.
The services which will now be offered.
The inventions which will now be commercialised.
The entrepreneurs who will discover that the idea they have carried around for twenty years is no longer impossible.
And perhaps most importantly, the people who will discover that they don’t need to employ twelve people before they can become one.
The great promise of AI is therefore not that nobody will have to work.
It is something much more interesting.
It is that nobody may have to say, “I have a great idea, but I can’t afford to make it happen.”
That is not a catastrophe.
It may be one of the greatest economic liberations in human history.
The question is not how many jobs artificial intelligence will destroy.
The question is how many businesses one person will soon be able to build.