Technology Cannot Replace Manpower

When automatic teller machines spread across the United States from the 1970s, many people expected the bank teller to disappear. The economist James Bessen found that the number of tellers actually rose for decades. Machines cut the cost of running a branch, so banks opened more of them, and the tellers who remained moved from counting cash to selling services and advising customers. The case is often cited to support the topic's claim that technology cannot replace manpower. It deserves attention. But the full record of history is more mixed, and that is what the topic must face.

The fear that machines will take jobs is as old as the machines. In 1811 groups of English textile workers known as Luddites began to smash the frames that were putting them out of work. In India the destruction of the handloom industry in the nineteenth century, under the pressure of cheap mill cloth from Lancashire, threw millions of weavers into poverty. The fear has returned with each wave of change, with the computer in the 1960s and with robots in the 1980s. Technology has replaced human labour in specific tasks and in specific occupations, and no honest account can deny it.

At the same time, over the long run, technology has not produced mass unemployment. In 1900 about forty per cent of Americans worked on farms, and today it is under two per cent, yet the country has not run out of work. Productivity gains made goods cheaper, which raised real incomes, and the spending generated new demand for services and products that did not exist before. Few jobs now common, such as software developer, mobile repair technician or social media manager, were imaginable a century ago. Economists call the process the creation of new tasks. The sum of jobs has grown with the population.

The economist David Autor has pointed out why. Most jobs are bundles of tasks. Machines take over the tasks that follow explicit rules and can be written as code, such as routine calculation or repetitive assembly. They do poorly at tasks that need flexibility, common sense, physical dexterity in unpredictable settings, and understanding of people. A nurse, a plumber, a teacher or a chef still does a range of things that no machine can match. This is sometimes called Polanyi's paradox, after the philosopher who observed that we know more than we can tell.

Most jobs are bundles of tasks.

Human strengths remain distinctive. Creativity, empathy, ethical judgement and the capacity to deal with new situations are not easily programmed. A machine can diagnose from images, but a doctor sits with a frightened patient. It can compose music in a given style, but it does not know why music matters. Customers in many fields still prefer a human face, and in care work, education and the arts, the human presence is part of the product. Where there is trust at stake, as in justice or leadership, replacement by a machine is unlikely to be accepted.

Yet the second half of the claim cannot be waved away. Advances in machine learning have extended automation into tasks once considered safe, such as translation, driving and the reading of medical scans. In 2013 Carl Benedikt Frey and Michael Osborne of Oxford estimated that about forty seven per cent of jobs in the United States were at high risk of computerisation over the next two decades, a figure that has been much debated and probably overstated, since it counted jobs and not tasks. The World Bank in its report of 2016 estimated that around two thirds of jobs in developing countries could be affected. Foxconn, the manufacturer of electronic goods, announced plans for a million robots to supplement its workers. The risk is especially acute for the routine middle layer of jobs, such as clerks and machine operators.

India's position is distinct. It has an enormous labour force, and a young population that adds roughly twelve million people a year to those seeking work. Its comparative advantage lay in abundant labour, which would be eroded if factories in rich countries can be run with robots, and a manufacturing path like China's may not be available. At the same time, India has vast needs, in health, education, infrastructure and care, that are labour intensive and unmet. The Mahatma Gandhi National Rural Employment Guarantee Act forbids the use of contractors and machines in its works to ensure that wages reach the poor, a deliberate choice for manpower. The country's challenge is partly to use technology to raise productivity without destroying livelihoods.

The topic thus needs to be read as a claim about what technology cannot do, and not about what it will not. Even if there will always be work for people, the transition can be painful. A displaced weaver does not easily become a software engineer. Wages may fall, regions may be hollowed out, and the gains may go to owners of the machines. The history of the Industrial Revolution, in which living standards for workers stagnated for decades before rising, warns that adjustment can take a generation. Policy has to prepare for it through education that teaches adaptability, income support during transitions, and encouragement of sectors where human work counts.

It would also be a mistake to treat technology and manpower as opposites. The most productive arrangement is often a partnership, in which machines do what they do best and people do the rest. Doctors with diagnostic tools, farmers with weather data and teachers with digital resources are more effective than any alone. The question of design is whether technology is built to assist workers or to remove them, and that is a choice made by managers, engineers and governments.

When a machine takes over a task, it leaves behind the part of the work that the machine could not do, and that part tends to be what people value most about a job. It is possible that this leftover is what the topic really means by manpower. How long it remains out of reach of the machine is something that no forecast of the past fifty years has got right.