‘it is the machine which possesses skill and strength in place of the worker, is itself the virtuoso, with a soul of its own in the mechanical laws acting through it’
Karl Marx, Fragment on Machines (1861)
Do you work to live or live to work? In the future you may do neither.
AI systems can already match or outperform humans in a range of capabilities: translating languages, recognizing faces, mimicking speech, driving vehicles, writing articles, trading financial products, diagnosing cancers. Once perceived as cold and lifeless, they can tell if you are happy, confused, surprised, or disgusted—sometimes by reading signals not detectable by the human eye. The economic consequences will be profound. Between 1990 and 2007—more than ten years ago, before the introduction of smartphones—industrial robots alone eliminated up to 670,000 American jobs.1 Between 2000 and 2010, the US lost about 5.6 million manufacturing jobs, 85 per cent of which losses were attributable to technological change.2 In 2016, analysts at McKinsey estimated that ‘currently demonstrated technologies’ could be used to automate 45 per cent of the tasks that we currently pay people to do.3 That’s just using today’s technologies. In the digital lifeworld we can expect systems of radically greater capability and sophistication.
Digital systems can increasingly perform many tasks that we thought only humans could do. Soon they may be able to perform almost all of them. Eventually they may be able to do them better than us. And of course, they’ve always been able to do many tasks that we can’t.
I said in the Introduction that predicting technological change is inherently risky and controversial. Trying to predict the economic consequences of technological change is even trickier. For personal reasons, if nothing else, however, I have a go in the next two chapters. In 2015 Richard Susskind (the world’s leading authority on legal technology, also my dad) and Daniel Susskind (one of the world’s most acclaimed young economists, also my brother) together published The Future of the Professions, in which they predicted the gradual replacement of human professionals by digital systems.4 In 2018 Daniel Susskind will publish a new book on the topic which, with fraternal pride, I expect to become the definitive work. It would be an act of family betrayal if I didn’t at least consider the idea of technological unemployment in this book.
(I know, I know, we’re a strange family.)
Clan loyalty aside, no responsible citizen can now ignore the prospect, accepted by growing ranks of economists, that in the future there may not be enough human work to go around. I call this the technological unemployment thesis. I don’t seek to assess in detail the economic case for and against it. I acknowledge that there are respectable thinkers who think it’s flawed. My view, however, is that if it’s even partly correct then the consequences will be so profound that we can’t afford to wait and see. What’s the harm in sharpening our intellectual tools while we still have time?
The analysis in this chapter is organized in four stages. We begin with the technological unemployment thesis itself. Next, we look at the work paradigm, the idea that we need work for income, status, and wellbeing. We then consider three responses to technological unemployment from within the work paradigm: treating work as a scarce resource, giving people a right to work, and trying to resist automation altogether. Finally, we dare to challenge the work paradigm itself, asking whether we could build a world where income, status, and wellbeing can be enjoyed in the absence of general employment. It may be that economic upheaval requires intellectual upheaval too.
The technological unemployment thesis predicts that developments in technology will eventually cause large-scale human unemployment. In basic outline, it runs as follows.
What we think of as ‘jobs’ are really just bundles of economically useful tasks. As time goes on, machines will increasingly rival and surpass humans in their ability to perform those tasks. Instead of hiring humans it will be more economical for firms to use machines instead. The people currently paid to perform those tasks will find, eventually, that their services are no longer required.5
In the first stage of technological unemployment, there will be less overall work to do but still enough to go around. Laid-off workers may be able to retrain and find new jobs. Mid-career education and training will be essential to their prospects of finding work. But as time goes on, the jobless will find it increasingly hard to get the training or resources needed to perform the shrinking number of tasks that remain. If you’re a steelworker from the north of England and you lose your job to a robot, it’s not much help to know that Google is hiring software engineers in Palo Alto.6
As unemployment grows, competition for work will be fierce. With a few exceptions for superstars in high demand, wages for most of those still employed will fall, because there will usually be a group of desperate souls at the factory gates willing to work for less.
In due course, when machines achieve the requisite levels of capability and reliability, it will make no economic sense for firms to employ humans for most economic tasks. As a result, there will no longer be enough human work to go around. Eventually only a small number will be able to find paid work. By this stage, to borrow the young Karl Marx’s phrase from 1844, the ‘worker has become a commodity, and it is a bit of luck for him if he can find a buyer’.7
Widespread unemployment will not cause the economy to stagnate. On the contrary, automation will enable firms to make great savings and efficiencies. Their profits will be reinvested as capital or passed on to consumers in lower prices.8 The economy will grow. In the past, we might have expected such growth to create more jobs for humans, because more demand for goods and services normally means more demand for the people who provide them. But not in the future, because eventually any extra tasks created would themselves be done more efficiently by machines than humans.
Overall, the economic pie will grow (that’s right, folks, the dessert metaphors are back) but human workers will receive an ever-smaller slice. On an extreme outcome, the vast majority of people of working age could be unemployed. But the effects would be radical if even half or a third of the working-age population was unable to find work.
It’s intuitive to assume that lower-educated workers will be hardest-hit by technological unemployment. It currently costs about $25 an hour to pay a human welder and about $8 an hour to use a robot.9 Supermarket check-out staff face the prospect of ‘smart’ stores that run without human check-out staff and shelf-stackers.10 Truckers, of whom there are 3.5 million in the United States alone, could be superseded by self-driving vehicles that can trundle for weeks without rest. Millions of low-paid workers in the food and beverage industry could be replaced by systems that can do every task from patty to plate:11
Whereas a fast food worker might toss a frozen patty onto the grill, Momentum Machines’ device shapes burgers from freshly ground meat and then grills them to order—including even the ability to add just the right amount of char while retaining all the juices. The machine, which is capable of producing about 360 hamburgers per hour, also toasts the bun and then slices and adds fresh ingredients like tomatoes, onions, and pickles only after the order is placed. Burgers arrive assembled and ready to serve on a conveyer belt.
The evidence, however, suggests that technological advance in the twentieth century actually boosted the number of low-education tasks in the economy as well as the number of high-education tasks. It was middle-education tasks that declined in number.12 What explains this? The answer lies in one of Daniel Susskind’s most important observations: that the level of education required by a human to perform a task is not a reliable guide to whether a machine will find that task easy or difficult. Machines outperform humans not by ‘thinking’ or working like them, but by using computational and robotic methods that are quite unhuman. This explains why we’re closer to automating the work of lawyers than the work of make-up artists.
Because we don’t know exactly how machines will perform certain tasks, it’s therefore hard to say definitively who will be the first to lose out. According to McKinsey, it’s easier to automate ‘physically predictable activities’ than unpredictable ones. The hardest tasks to automate are those that involve managing and developing people, and applying expertise to decision making, planning, and creative work.13 But, as the Susskinds argue in The Future of the Professions, it looks like even the most complex professional work can be broken down and automated in the long run, including many aspects of those jobs said to require a human touch.14
As time goes on, however, the question won’t just be one of automation of tasks currently done by humans. The desired outcomes might be attained in entirely different ways. For instance, the fact that we no longer need human farriers (horseshoe-makers) because cars have replaced horses and carriages makes it wholly irrelevant whether the work of a farrier could be done better by a machine. People don’t want surgeons; or even machines that do the work of surgeons better than humans. They want health. If advance in another field of medicine (such as nanotechnology) makes the work of certain surgeons redundant, then it is irrelevant whether the work of human surgeons could be done better by machines.15 Human redundancy will come in various forms; automation of tasks that we currently do is only one of them.
So what does political theory have to say about a world of widespread unemployment?
People in the ancient world disliked the idea of working for a living. Work appears in the Old Testament as a kind of divine retribution: ‘By the sweat of your brow’ God berates the sinful Adam and Eve, ‘you will eat your food until you return to the ground’ (Genesis 3:19). As Kory Schaff observes, the ancient Greeks understood the practical benefits of work but believed that the key to human flourishing was avoiding it. A ‘mechanical or commercial life’, writes Aristotle, ‘is not noble, and it militates against virtue.’ Better, as Schaff puts it, to dedicate one’s days to contemplation, statecraft, and warfare.16
Philosophers in the Christian tradition came to see hard work as a path to salvation, encouraging thrift, honesty, and self-discipline. Work also played an important role in the liberalism of John Locke and Adam Smith. Only through labour, they believed, could humans forge and hew the natural world into instruments of value and progress. Marx, by contrast, contended that capitalists sucked value from their workers without giving them a proper share of the rewards.17
So why do we work today? For three main reasons: income, status, and well-being.
First and foremost, we work to live. For most of us, selling our productive powers is the only way to earn the money that keeps us fed and warm. We rarely grow our own food or build our own shelter. Instead we work so we can pay for those things (themselves the product of other people’s labour). Most of us also hope that if we work hard there’ll be a little left over to pay for pleasurable treats, luxuries, and experiences.
We also work to satisfy a deep-seated need for status and the esteem of others. Gainful employment brings recognition and prestige. We feel pride when we receive positive feedback on our work from bosses, customers, and clients. Unemployment, by contrast, can lead to stigma and shame.
Finally, some are fortunate to do work that contributes to their wellbeing. Such work might offer the chance of intrinsic satisfaction from a job well done. Or it could be rewarding because it brings value to others: I think of my mum, first a nurse and now a psychotherapist, who’ll do anything to look after her patients. (Not every Susskind writes about technology.) For others, work offers the chance for self-improvement and a platform to cultivate skills and faculties. When people are out of work, they often speak of feeling ‘useless’ or ‘on the scrap-heap’.18
Work also helps us to manage our fragile relationship with time. From our earliest years we experience the passage of time in a structured way: morning, afternoon, evening, night, each with its own routines and activities. In adulthood, our experience of time is closely connected with the routine of work. The unstructured languor of unemployment can lead to boredom and dislocation, which over long periods can become a ‘major psychological burden’.19
Still thinking about psychological wellbeing, work can allow us to associate in a purposeful way with people outside our immediate family and gives us space to relieve our competitive urges. Even inane conversations round the office water-cooler permit us to observe other people and share stories, experiences, and knowledge of the world. This is part of keeping a grip on reality itself.20 In Civilization and its Discontents (1930) Sigmund Freud claims that no other technique for the conduct of life binds the individual ‘more closely to reality’:21
in his work he is at least securely attached to a part of reality, the human community. Work is no less valuable for the opportunity it and the human relations connected with it provide for a very considerable discharge of libidinal component impulses, narcissistic, aggressive, and even erotic, than because it is indispensable for subsistence and justifies existence in a society.
There is, I believe, an assumption in modern society that work is needed for income, status, and wellbeing. I call this the work paradigm. In a world where most adults work in some capacity, the nature of work is closely related to social justice. It’s connected with distribution (of income and wellbeing) as well as recognition (in the form of status and esteem). Under the work paradigm, therefore, the prospect of a world without work is naturally disconcerting. How should we respond?
Accepting for a moment the idea that paid work is something we need, there are three possible responses to the problem of technological unemployment: to treat work as a scarce resource, to give people a right to work, or to resist automation altogether.
The first possible response would be to continue with business as usual, treating work as a resource (albeit an increasingly scarce one) to be distributed among the population according to a mechanism of our choice, the most obvious being a labour market like the one we have now. Through the interplay of market, state, and algorithms, tasks could be allotted to the most deserving or meritorious candidates. An alternative mechanism for distributing scarce work might be a rota, with work rotated around the citizens so everyone gets their turn.
The obvious difficulty with this approach is that there would still be a majority without work at any given time and those lucky enough to have a job would probably face severely depressed wages. This response would not properly address the challenge of technological unemployment.
A second approach would be to introduce some kind of right to work together with a scheme of artificial ‘make-work’ that provided everyone with tasks to keep them busy. Many unemployed workers in the past have claimed such a right. The ‘droit au travail’, as Jon Elster notes, was the ‘battle cry’ of the 1848 revolution in France.22 Ultimate responsibility for any right to work would have to rest with the state, perhaps entitling citizens to redress if insufficient work was forthcoming. Public authorities could offer work directly or issue vouchers redeemable with firms for that purpose.
The notion of a right to work is normally seen as economically nonsensical because it would be a colossal drain of resources. In the digital lifeworld, however, it might be possible to skim some profits from firms and spend them on schemes that put people to work. As well as taxing their profits (or potentially instead of doing so) the state could place a legal duty on firms to provide a certain amount of make-work in proportion to their economic output.
All things considered, however, a right to work of this kind would be deeply problematic in various respects. Producing enough make-work and paying people for it would be massively inefficient and perhaps even impossible, particularly if the scheme tried to cater to each person’s conception of worthwhile work. Moreover, the status and esteem associated with winning and keeping a job would be rather undermined by the knowledge that it was guaranteed as a right, and by the knowledge that the work itself was fundamentally useless.23
A third possible response would be to try to resist automation altogether through some kind of global technological moratorium. I don’t see this as a practical possibility.
The three responses just described share a common trait: they accept the work paradigm and seek to find ways of maintaining it. This may be the wrong approach, intellectually and practically. Technological unemployment could in fact present an opportunity to dismantle the work paradigm and replace it with a different set of ideas. In short, it would mean weakening or even sundering the connection between work on the one hand, and income, status, and wellbeing on the other.
The first step would be to recognize that there is no necessary connection between work and income. Even under the current system, work does not guarantee a living. One in four employed adults in the United States earns wages below the official poverty line. Almost half are eligible for food stamps.24 The converse is also true: there need not be a connection between non-employment and destitution. In past eras, losing your job may have meant homelessness and starvation. Today, most advanced political systems have some kind of collective safety net.
The digital lifeworld offers the chance to take the next step, weakening or even breaking the connection between work and income altogether. At its simplest, this could mean taxing some of the profits of firms and redistributing them among the general public. Hence Bill Gates’ proposal for a ‘Robot Tax’, by which firms would be taxed for their use of machines with the proceeds going to fund employment opportunities elsewhere.25
Another increasingly popular notion is a universal basic income (UBI) paid in cash to everyone ‘with no strings attached’.26 On the radical model advocated by Philippe van Parijs, a UBI of about a thousand dollars a month would be available to every citizen with no means test or qualifying obligations.27 Such a system would differ from the ‘make-work’ model above because it would not require people to work. On the contrary, how people chose to satisfy their needs would be a matter for them.
The idea of a UBI is not new. It has long been discussed in the context of a functioning labour market like we have today. Theorists have argued about whether a UBI would discourage work, or whether people would seek to ‘top up’ their UBI with additional income, or whether a UBI would prepare people well for re-entry into the job market. These questions (like human workers themselves) would be redundant in a world of technological unemployment. The function of a UBI in the digital lifeworld would be to replace the labour market and not to augment it. And if it’s going to be their main source of income, people in advanced economies might need more than a thousand bucks a month.
There is room for debate about what a UBI might look like. It makes sense, for instance, that payments should not be contingent on looking for work if there’s not enough work out there for everyone. It probably makes less sense that a UBI should be paid to each citizen regardless of their other means of income. Those who earn massive rents from land or capital arguably don’t have the same moral claim to public funds as others. They would also have less use for the cash, and be less likely to spend it, making it economically inefficient to hand it to them. There could also be ways of topping up a UBI through economically productive work (if there is any for humans to do). If you have a rare medical condition that requires extra funds, perhaps you could be entitled to a higher basic income. Or if you do good deeds like caring for the elderly, perhaps you ought to receive a bonus (even if a machine could have performed that care better).28 Rewards could incentivize self-improvement, with extra income to those who learn new skills (the value of such skills not being measured in economic terms).
To sever the connection between work and income would be radical. But luck egalitarians would see it as a step forward for social justice. For too long, they would say, people’s wealth has been determined by their unearned talents and capacity for work. Marx himself described the right to work as an ‘unequal right for unequal labour’ because some are better equipped, ‘physically or mentally’, to participate in the labour market.29 Isn’t there something degrading, after all, in having to offer ourselves up as commodities on the labour market? Owners of capital have been earning without labouring for centuries. Might it not be progress if the rest of us could do the same?
The second step would be to challenge the assumption, perhaps the deepest and most entrenched in the work paradigm, that only gainful work is deserving of status and esteem while unemployment is a source of stigma and shame. ‘Most people in modern competitive market societies’, says the political theorist Richard Arneson, ‘believe that the failure of an able-bodied person to earn his keep is degrading.’30 We are taught to believe in the dignity of work, no matter how disgusting or dangerous that work is. Long-term unemployment is often thought to suggest personal incompetence or moral decay. Idleness is seen as disgraceful. ‘What is a man,’ asks Hamlet, ‘If his chief good and market of his time | Be but to sleep and feed? A beast, no more.’31
When listening to public debate about unemployment, it sometimes feels like we haven’t shaken off the Victorian moralism of Herbert Spencer, the patron saint of grumpy old men in pubs:32
‘They have no work,’ you say. Say rather that they either refuse work or quickly turn themselves out of it. They are simply good-for-nothings, who in one way or other live on the good-for-somethings—vagrants and sots, criminals and those on the way to crime, youths who are burdens on hard-worked parents, men who appropriate the wages of their wives, fellows who share the gains of prostitutes; and then, less visible and less numerous, there is a corresponding class of women.
The duty to work, and work hard, is usually referred to as the work ethic, the name itself cunningly suggestive of the idea that work and ethical conduct are intertwined. The work ethic is one of the most pervasive and accepted doctrines of our time—so much so that it often goes without explaining or justifying. But now is as good a time as any to ask whether people’s status should depend, as much as it does, on their economic contribution, rather than (say) their goodness, kindness, or public-spiritedness.
The work ethic is hard to reconcile with a world of technological unemployment. It’s one thing to insist that people have a duty to work even when the work is hateful; but it’s downright sadistic to say they have such a duty when there isn’t even enough work to go around. We might also doubt whether the work ethic could survive in a world where the majority do not work. Would there still be a stigma to losing your job if everyone else was unemployed too?33 The term ‘unemployed’ would lose some of its explanatory force, and the ‘un’ part would lose much of its stigma.
Technological unemployment prompts us to think about building an economy where status and esteem are associated with traits other than economic productivity. This might not be a bad thing.
The third and final intellectual step would be to challenge the connection between work and wellbeing. It’s true that some jobs are fun, safe, creative, educative, uplifting, and meaningful, but many others are drudgerous, dangerous, repetitive, stultifying, depressing, and pointless. How effectively a job satisfies a particular person’s needs will depend partly on their subjective preferences—one teacher may love her vocation; another may despise it—but some jobs are plainly worse than others.
It’s not just that not all work enhances the human condition. On the contrary, work often harms the worker. Consider this account:
The pace of work is unremitting. Workers are reprimanded for ‘time theft’ when they pause to catch their breath after an especially difficult job. They are subjected to ever-increasing quotas, threatened daily with discharge, and eventually fired when the required pace gets too high for them to meet—a fate of the vast majority . . . But not before they suffer injury on the job: workers have to get on hands and knees hundreds of times per day, a practice that leaves few unscathed. [The employer] forces them to sign papers affirming that their injuries are not work-related, or they are given demerits that can lead to discharge. [On one occasion the employer] allowed the indoor heat index to rise to 102 degrees. When employees asked to open the loading doors to let air circulate . . . [the employer] refused, claiming this would lead to employee theft. Instead, it parked ambulances outside, waiting for employees to collapse from heat stroke. When they did, they would be given demerits for missing work, and fired if they accumulated too many. [The employer] didn’t care, because regional unemployment was high, and they had hundreds of applicants to replace the fallen workers . . .
Do these workers benefit physically or psychologically from their work? The extract reads like it’s from the pages of Friedrich Engels’ Condition of the Working-Class in England (1845), which exposed the squalorous horrors of working life under Victorian capitalism.34 But it’s actually from 2017, authored by the American scholar Elizabeth Anderson. The employer it describes is Amazon.35
A UBI, or something like it, would free humans from having to do awful work. It could fulfil the dream sketched by Oscar Wilde in 1891, of ‘All unintellectual labor, all monotonous, dull labor, all labor that deals with dreadful things and involves unpleasant conditions’ being ‘done by machinery’:36
Machinery must work for us in the coal mines, and do all sanitary services, and be a stoker of steamers, and clean the streets, and run messages on wet days and do anything that is tedious or distressing.
So what would we do all day? Perhaps nothing. As Michael Walzer observes, for many people the opposite of work is leisure in the form of idleness. The Greek word for leisure, scholé, has the same etymological origins as the Hebrew word Shabbat, which means ‘to cease’ or ‘to stop’.37 But it’s far from clear that a life of idleness would be good for humans, even if such a lifestyle were freed from the stigma currently attached to unemployment. ‘If all the year were playing holidays,’ cries Shakespeare’s Prince Hal, ‘To sport would be as tedious as to work.’38
The psychological needs said to be met by work—to feel a sense of achievement, to structure our day, and to interact meaningfully with other humans—probably won’t disappear. On a libertarian approach, we might be left to decide for ourselves what to do with our days. On a paternalistic approach, the state could decide and specify mandatory non-economic activities, either as a prerequisite to receiving state funds or at pain of some other punishment.
It might turn out that what some of us need is in fact work or something like it! But when we surmise this, we aren’t talking about horrendous work down a mine. We’re talking about good work of the fun, safe, creative, educative, uplifting, and meaningful kind. Under the work paradigm, the joyous experience of good work is something a few are lucky enough to be paid for. In the digital lifeworld it may become something that people choose to buy with their allotted share of society’s resources. As for any leftover bad work that still needs to be done by humans, citizens might have a duty to do it when called upon, like military conscription. Bad work could even be allotted on the basis of moral desert as a way of punishing criminals or wrongdoers.
Is work the only way to meet our need for a sense of routine or to connect meaningfully with people outside the family? In simple societies where natural conditions meant that few had to work for a living, these needs were met by ‘rituals, religious and community practices’.39 Today, when we’re not working, we have hobbies, leisure pursuits, public and voluntary service, clubs and associations, and time with friends and family. Plenty of people happily abstain from paid work. As I explained in Part III, in the digital lifeworld we might have much freer access to new and exciting experiences, whether in VR or otherwise. It’s for political theorists, economists, social psychologists, and the like to identify our needs with precision and suggest what kind of activity, other than work, might satisfy them. That, rather than trying to salvage the work paradigm, could be a more productive use of their intellectual energy.40
The work paradigm insists on the necessity of work for income, status, and wellbeing. Yet who among us, waking at dawn and facing a long day of toil, has not quietly yearned for a life without labour? It might be time to challenge the idea that work is something we need. The analysis can’t stop there, however. If it’s right that the wealth that currently flows to labourers will be increasingly redirected toward those who own the technologies that replace them, then we need to examine that radical redistribution of wealth and consider whether it can be justified. That’s the work of chapter eighteen.