It is amazing how quickly we come to take monumental advances in computing for granted. Today, when people using Google’s Chrome browser visit a foreign-language website, the browser offers to translate the page. Those accepting the offer are provided with a translation in little more than a moment, and it is of startlingly good quality. Translations improved dramatically in 2016, when Google began using a new system built on machine learning; overnight, quality improved by more than it had in the previous 10 years. The advance earned a fanfare in the press, before becoming part of our everyday lives, scarcely noticed: this technological breakthrough that had eluded computer scientists for decades.
High-quality, instant translation has the potential to change the world. One can imagine a moment in the not-so-distant future when ubiquitous earpieces connected to software in the cloud will allow people to communicate in foreign countries with the ease and fluidity of native speakers. It will also send tremors across the global economy. Business models will change, as publishers of media content suddenly find themselves participating in a truly global market. And the many thousands of people now employed doing good, skilled work as professional translators might find themselves out of a job. Yet that scarcely begins to capture the changes that loom ahead, because the techniques that allow Google to provide near-instantaneous, high-quality translation are increasingly allowing engineers to build systems with all sorts of human-like capabilities: from operating automobiles on busy city streets, to assessing when a shape on a hospital scan is likely to be cancerous. A great age of automation looms ahead of us.
Assessing the potential for job loss due to automation is not simple. In a paper published in 2013, Carl Benedikt Frey and Michael Osborne, of Oxford University, analysed the nature of the tasks involved in different lines of work to gauge their ‘automatability’. They reckoned that 47% of jobs in the US are at risk of computerisation in the next few decades. A paper published by the Organisation for Economic Co-operation and Development in 2016 looked instead at the potential for automation of tasks within jobs, rather than of occupational categories as a whole, and concluded that while many workers will see their jobs change as certain tasks are automated away, only about 9% of jobs are fully automatable.
But even the smaller estimates of future job loss to automation represent millions upon millions of jobs across the rich world. What’s more, technology continues to improve. Tasks that seemed beyond the reach of computing a decade ago, such as automated driving, are now realities. If advances in machine intelligence continue to progress faster than anticipated, we could very soon face a serious labour-market crisis. The way in which that crisis unfolds, however, depends upon how we react to it.
An abundance of labour
Technological progress in areas such as machine intelligence and robotics promises to create an economy in which capital (machines) is increasingly capable of substituting for human labour at a reasonable cost. As the scope for substitution grows, the number of workers whose labour is not strictly necessary to the operation of the economy will rise. The political and economic institutions now in place across most advanced economies, which emerged during the industrial revolution, are not well equipped to handle this dynamic. Welfare states are designed for a world in which most adults spend much of their life working, and in which the money earned from working is most adults’ prime income stream. For the majority of people, cutting back on hours worked or leaving the workforce entirely means accepting a substantial decline in income and the possible loss of critical benefits. It is simply not a realistic option.
“leaving work is unacceptably unpleasant for most”
In this world, steady improvements in the capabilities of machines paradoxically end up hurting many workers and, potentially, the economy as a whole. Because leaving work or reducing one’s hours is unacceptably unpleasant for most people, those who find themselves displaced by technology must seek new jobs wherever they can find them. But because machines will increasingly stand ready for deployment as substitutes for people, displaced workers face two options: they can specialise into the type of work (typically demanding a high level of skill or talent) that machines cannot do, or they can compete with machines on price. The better machines get, the larger the role the second adjustment mechanism will play; workers will increasingly find that obtaining new work means accepting a reduction in pay.
A society that is committed to keeping everyone working amid the sort of technological change we can expect in future is one in which human workers will stave off replacement by machines by accepting ever-reducing wages. The economy that results from such choices is one that grows more slowly than it ought to, because firms frequently choose to use cheap workers when they might instead use more advanced technology. It is also one that is increasingly unequal, because the glut of workers seeking employment at any wage allows owners of capital and intellectual property to capture most of the gains from growth.
If it is hard to understand how an economy could function like this; consider the status today of many developing economies. David Autor, an economist at MIT, describes how Nissan operates plants all over the world in which it produces the same sorts of vehicles. In its plants in Japan, where wages are high by global standards, robots do much of the work. In India, in contrast, production is far more dependent on human labour because wages are much lower. India’s economy uses less capital per worker, and is less productive and poorer than Japan’s. In other poor and highly unequal societies, large amounts of labour are absorbed by the households of the rich, who take advantage of the glut of people available for hire at low wages to employ cooks, valets and doormen.
In the past, economies have been able to respond to major technological shifts by retraining and relocating workers to take better advantage of the new employment opportunities generated by new technologies. As a result of the industrial revolution, societies shifted from providing no universal public education at all, to universal public primary and secondary education, and accessible and affordable university education. It is certainly possible that significant new investments in education could reduce the magnitude of the adjustment facing governments today and extend the time they have to manage it.
But there are two hard constraints on an education-based solution to underemployment caused by technology. One is that it is much harder to boost the educational attainment of a highly educated population, which most rich economies now have, than it is to boost the attainment of a poorly educated population, such as existed in early industrial times. Second, an education solution is running a losing race against the clock. While societies seek to educate and retrain their workers, technology continues to get better. More and better education is desirable for many reasons, but at best it will delay the need for a broad restructuring of society in response to rapid technological progress.
Building an economy that benefits everyone under these sorts of technological circumstances means solving several different problems. The first is the highly unequal distribution of income. If we are willing to assume the existence of a magic wand that can wave away political difficulties, then this problem is not so hard to solve. We could introduce new and highly efficient taxes, on land, for instance, and then redistribute the proceeds. Societies could pay all citizens a universal basic income (UBI), which would provide a basic standard of living to all people whether or not they sought employment, or provide generous and non-expiring unemployment benefits alongside good wage subsidies for low-income workers, which would encourage people to stay in work if they could, while spreading the gains from growth. Alternatively, societies could move toward systems of social ownership. Governments could take stakes in firms and either distribute ownership rights to all citizens, entitling them to dividend payments, or hold the stakes in a sovereign wealth fund, the dividends from which could be redistributed as income payments to citizens.
The aim of such programmes would be threefold. First, they would allow workers to scale back the hours they work without facing impoverishment. Second, because people could opt out of work as technology improves, such measures would maintain the incentive for firms to deploy new labour-saving technologies as they become economically attractive, the better to keep economies growing and becoming more productive. And third, they would distribute the gains from growth broadly, thereby helping to maintain the public legitimacy of the system.
But, while helping people supplement or replace their income as their labour becomes unnecessary is a critical piece of the puzzle, it leaves other problems unaddressed. A second difficulty facing society might ultimately prove the more confounding: how should those freed by technology from work spend their new leisure time? On the face of things, this might not seem like much of a problem. Who cares what people do with their free time, after all? If a quarter of the workforce is rendered unnecessary by technology, then why shouldn’t those individuals be free to spend their days however they like, even if that means hours spent napping in front of the television, or days lost in drugs and alcohol?
There is a strong case to be made for societal agnosticism regarding the use of free time. What, for one thing, is the point of technology if not to free us to do as we like?
And, for another, if efforts to keep an idle few from abusing drugs prevent many more from indulging in satisfying lives of leisure, spent working at hobbies and crafts or peaceful reflection, well that, too, represents a failure of society to make good use of the technological tools available to us.
But society probably won’t be as liberal-minded about such shifts as one might hope. The introduction of generous benefits — and particularly of those, such as a UBI, calculated to allow people to work less — creates troublesome incentives. ‘Necessary’ workers might well join ‘unnecessary’ workers in opting to leave the labour force, creating headaches for people who will no doubt be upset to learn that their favourite paediatrician has quit the practice to be a surf bum. Even if redistribution were to surgically excise the most expendable workers from the labour force, those still in work and paying taxes might reject the social bargain presented to them. Why, they will certainly ask, should they keep working and paying taxes (whether or not they earn enormous salaries to do jobs they love) in order to subsidise the idleness of millions of others?
And then, ironically enough, those receiving government assistance might find themselves just as unhappy with the arrangement. People look to work to shape their identity, to provide their lives with meaning: because they are doing something they love, or because they are doing something that others value, or because they are, through the work they do, helping their family to survive and thrive. Were we able to wish a grand new redistributive system into existence, it might soon collapse as a result of its failure to address these social issues. Indeed, the political changes we observe today suggest that those whose livelihoods are undercut by broad economic shifts might prefer to vote for the undoing of the liberal system as a whole than to plump for greater redistribution, even if the latter solution offers some hope of improvement in welfare while the former does not.
What does this tell us? For a new social compact to earn broad acceptance, there must be a societal consensus regarding how those not in work should spend their time and how they can prove themselves of value to the community. Society must reinvent what it means to be a contributing member, a member of value, worthy of admiration and respect.
One difficulty we face is that we simply do not know precisely what sort of work might satisfy all these demands. What we can say, however, is that such categories are not immutable. Society values what society values. In the past, for instance, women were often expected to work long and hard hours in the home without pay. Even if society did not reward in an adequate way the work of the women who tended homes and raised children, it nonetheless saw such work as a critical contribution to broader well-being. Social norms can be harnessed to direct the labour of large portions of adults, entirely outside of the marketplace.
Something like that will need to take place in future, alongside broad changes in the structure of the welfare state. As the economic stress from technological change deepens, governments that rely on employers to provide benefits such as healthcare will increasingly provide them directly, while more of the benefit programmes on which people rely will be ‘work agnostic’, including, eventually, income subsidies themselves. As this occurs, norms will shift regarding the responsibilities of those not in work, and new social roles will emerge (or expand). The not-employed might be expected to be more active carers for children and parents; roles that might include more community involvement, as well, just as the homemakers of past generations were often expected to participate in parent-teacher associations or other community groups.
“Society must reinvent what it means to be a contributing member”
As this occurs, we might well see the emergence of a ‘semi-pro’ category of employment, in which people do work that is economically useful but that cannot provide an income sufficient to support an individual. Semi-pro work could mean running a business that just manages to cover its (non-labour) costs. It could mean teaching: tutoring young people, for instance, or leading seminars in one’s area of expertise, or providing lessons related to a skill one has obtained (such as piano playing or metalworking). It could involve work as an extension of the healthcare industry: helping to check on and care for elderly, sick or disabled people.
These sorts of work might not enjoy the same status as paid labour, and while many of the people involved in such work might find it satisfying and fulfilling, others would surely prefer the ability to find good work in the salary-paying economy. But the evolution of this social niche would solve multiple problems. It would help the jobless to know how to spend their time usefully and in a way that provides some sense of meaning and identity, while also contributing to the broad legitimacy of a welfare state designed to allow people to abandon paid employment.
We then arrive at the last and most critical question: just how will society and its political systems bring such changes about? The answer is: very slowly, and after trying many other approaches.
In most rich countries today, political systems and social norms are built on the centrality of paid work. When the institution of paid work stops functioning as expected, the first (and second and third) reaction from political systems and societies is to repair the institution of paid work. In some cases, those attempts will mean efforts to neutralise the forces that are seen to be undermining paid work, such as foreign trade and migration. When those efforts fail to work as intended, governments will turn to other strategies. Some will use corporatist tactics to reduce competition and boost firm profits, then apply pressure on profitable firms to operate with bloated payrolls. Some governments will do more to subsidise hiring. Government payrolls themselves might also swell. That is worrying, since the most politically acceptable way to increase public employment will often be through increased military spending in response to foreign-policy crises.
To achieve a fundamental change in outlook, away from the centrality of paid work, will first and foremost take generational turnover. Adults raised to expect to find good work will support politicians who promise to bring back the economy of the past. The young people being raised now, at a time when work is less certain and less remunerative, will expect less from the institution and will find it more reasonable to expect people to find satisfaction and contribute to society in ways other than paid employment. Cohorts to come will grow up with very different expectations about the world than those held by their grandparents. Those expectations will shape their ideas about what governments might reasonably do to address the crisis of work, and they will build the political movements that enact the welfare changes needed to bring about a more prosperous technological future.
That, more or less, is the process through which the industrial welfare states we have now were built. First came massive technological and economic change. Then came political battles between those who believed old social orders could and should be maintained and those pushing a more radical view. Over time, and after many different kinds of failures of social reform, ideas changed regarding what rights and protections workers were owed. And the people championing those new ideas built new institutions capable of wielding the power to change laws, levy taxes and provide the social insurance needed to make industrial capitalism less brutal and more equitable. The process took nearly 200 years.
Change may occur more quickly this time around. Idea transmission seems to occur much more rapidly in the digital age than in the past, for example, which has allowed ideas like a universal basic income to very quickly become part of popular discussions about how to make economies work better. But addressing the challenge of technological abundance is less a matter of technocratic policy design than one of fundamental change in our view of the role of work in society. In the industrial era, workers were a small but critical component of a big economic machine; the resulting challenge was to build a society that ennobled and empowered such workers.
In the digital era, workers are a vestigial component of a big economic machine, and the social machinery that ennobled and empowered people in the past now serves to demean and disempower. Realising the potential of the digital age will require nothing short of a new social revolution.
Ryan Avent is author of ‘The Wealth of Humans’ and Senior Editor at The Economist
This article first appeared in the RSA Journal Issue 4 2016–17