Demography and automation: Robots that look after grandma
Because of ageing, the world needs a robotics revolution. The machines don’t seem ready for one
When Gill Pratt sat down to discuss the job of running the Toyota Research Institute, the carmaker’s new research division, his Japanese interviewers wrote one word on a piece of paper and asked him to talk about it. The word was dementia. That might seem a strange topic to put to one of the most respected figures in the world of robotics, a man who had previously run a competition to find artificially intelligent, semi-autonomous robots for the Pentagon. But, Mr Pratt says, the company’s interest in ageing was a big reason for him to take the job. “The question for all of us”, he says, “is, how can we use technology to make the quality of life better as people get older?”
Ageing and robots are more closely related than you might think. Young countries with many children have few robots. Ageing nations have lots. The countries with the largest number of robots per industrial worker include South Korea, Singapore, Germany and Japan, which have some of the oldest workforces in the world.
The connection does not merely reflect the fact that young countries tend to be poor and cannot afford fancy machines, which they do not need anyway. It holds good within rich countries, too. Those with relatively few robots compared with the size of their workforce include Britain and France, both of which (by rich-country standards) are ageing slowly.
Two recent studies quantify the connection. Daron Acemoglu of the Massachusetts Institute of Technology (MIT) and Pascual Restrepo of Boston University show that, between 1993 and 2014, the countries that invested the most in robotics were those that were ageing the fastest—measured as a rise in the ratio of people over 56 compared with those aged 26-55. The authors posit a rule of thumb: a ten-point rise in their ageing ratio is associated with 0.9 extra robots per thousand workers.
A study from Germany used different measures but reached the same conclusion. Ana Abeliansky of the University of Göttingen and Klaus Prettner of the University of Hohenheim found that the growth in the number of robots per thousand workers rises twice as fast as the fall in the growth rate of the population (ie, if population growth falls by 1%, the growth in robot density rises by 2%). Population growth is closely related to age structure.
These findings should not be surprising. Robots typically substitute for labour. That is why many people fear that they will destroy jobs. Countries with plenty of young workers do not need labour substitutes. Wages there also tend to be low, making automation unprofitable. But ageing creates demand for automation in two ways. First, to prevent output falling as more people retire, machines are necessary to substitute for those who have left the workforce or to enable ageing workers to continue to do physical labour. Second, once people have retired they create markets for new kinds of automation, including robots that help with the medical and other requirements of caring for people who can no longer look after themselves.
Automation is destiny
As a result, the connection between robots and ageing is a powerful one. Mr Acemoglu reckons that ageing is the biggest single influence upon how many robots a country has. He estimates it explains close to 40% of the variation in the numbers of robots countries introduce.
The influence will grow. This year, there will be more people over 65 than under five for the first time in human history. By 2060, the number of Americans over 65 will double, to 98m, while in Japan, 40% of the population will be 65 or older. There will not be enough younger people to look after so many, unless robots help (and probably an influx of migrants is permitted, too).
Shrinking and ageing workforces matter as much. China is now the world’s largest robot maker, producing 137,900 industrial robots (typically, machines used in assembly lines) in 2017. Between 2015 and 2040, according to the UN, China’s working-age population (aged 20 to 64) will fall by a staggering 124m, or over 13%. Applying Mr Acemoglu’s rule of thumb to this decline, China would by the end of the period need to install roughly 2m more robots. That is more than four years’ worth of all the industrial robots produced in the world in 2018 and six times as many as the increase in worldwide production over the past nine years.
Such problems loom even in countries ageing more slowly than China—such as Britain. Between 2016 and 2025, according to Mercer, a consultancy, the proportion of British workers who are under 30 will fall by four percentage points and that of over 50s will rise by ten points. That sounds manageable. But it masks big regional swings. In that period, London (which is relatively youthful) will see the share of its labour force under 30 fall by a quarter and the share over 50 rise even more.
That will put enormous pressure on some industries. A third of teachers and building workers in Britain are over 50, as are more than a third of health-care workers, farmers and lorry drivers. They are quitting in droves. A poll in 2015 found that a third of doctors planned to retire by 2020. And this is in a country whose ageing is relatively gentle. Automation is not the only way to deal with skills shortages (immigration and later retirement also help) but it is one of the most important.
Over the next few years, demography will change the kinds of robots people need, as well as increase the number in use. At the moment, the robotics market is dominated by industrial machines, the sort used to assemble cars or electrical equipment. Sales of industrial-robotics systems were $48bn in 2017, seven times as much as “service robots”, a category that includes logistics robots for running warehouses, medical robots, robotic milking machines, exoskeletons that help people lift heavy objects and household robots that vacuum the floor.
As demographic change speeds up, service robots will become more important. One day, their makers hope, they will enable old people to live alone and stay mobile for longer. Robots will help assuage loneliness and mitigate the effects of dementia. They will make it easier to look after people in nursing homes and enable older workers who want to stay employed to keep up with the physical demands of labour. These robots will also be fundamentally different from industrial ones, which usually replace human activity—fitting a car windscreen, for example. By contrast, service robots extend it. For example, if an exoskeleton helps someone lift something heavy, the person still has to be there.
You can see the stirrings of this robot revolution most clearly in Japan. AIBO, a robotic puppy with artificial intelligence (AI) made by Sony, and Paro, a furry seal made by Japan’s National Institute of Advanced Industrial Science and Technology, are therapeutic robots for children and patients with dementia. Pepper, made by SoftBank, is a humanoid robot which can carry out conversations on a limited range of topics, so long as its human interlocutor does not stray too far from the script. MySpoon is a robot for those who cannot feed themselves. HAL, by Cyberdyne and Muscle Suit, by Innophys, are exoskeletons, helping nurses pick up and carry patients (HAL stands for hybrid-assistive limb). Panasonics’ Resyone is a robotic bed that transforms itself into a wheelchair. And so on.
Demand for these gizmos is growing fast, if from a low base. Sony said it had sold 11,111 AIBOs in the three months after the new model went on sale in January 2018. Japan’s government reckons that 8% of nursing homes now have lifting robots, and its national robot strategy (every country should have one) calls for four-fifths of the elderly receiving care to have some care provided by a robot by 2020.
For the time being, though, the technology remains a long way from transformative. According to the International Federation of Robotics, an estimated 20,000 robots were sold in 2018 that could realistically be described as helpful for ageing (medical robots, handicap assistance, exoskeletons and the like). That is less than 5% of industrial robots.
The number will doubtless grow. The question is how quickly. Mr Pratt is optimistic. Over the past five years, he argues, there have been huge advances in artificial intelligence, enabling machines to surpass humans in certain kinds of information- processing, notably pattern recognition which (within limits) robots can perform more quickly and reliably than humans. New firms are pouring into the business. A third of robot companies are less than six years old and make service robots. The costs of research and development are coming down and investment is rising. Within a decade, Mr Pratt reckons, domestic robots will help people cook at home and car-guidance systems will keep them mobile for longer.
But for that to happen, robots will have to perform a dauntingly long list of things they cannot yet do. They cannot navigate reliably around an ordinary home, move their hands with human dexterity, or conduct open-ended conversations. Although they can provide some physical assistance to the elderly, one robot can do only one thing, so multiple tasks would require your home to be stuffed with machines.
Their pattern recognition is not 100% reliable. One image classifier could not tell the difference between a snow plough and an overturned school bus. Robots struggle to operate on the basis of incomplete information, or to adapt to novelty as quickly as humans do. Driverless cars are proving harder to develop than most people expected. Rodney Brooks, a professor at MIT, reckons that driverless services comparable to those offered by conventional taxis are unlikely before 2032. Google’s Duplex, an AI-enabled personal assistant launched in 2018, can so far make appointments only for hair salons and restaurants. All this suggests that, as solutions to the problems of ageing, robots have some way to go.
Their limitations have significant implications. Robots that make the end of life more bearable are likely to remain expensive for many years, so only rich people will buy them. That may limit their wider social acceptance. Companies may not be able to automate their way out of future skills shortages. Other responses, such as raising wages, attracting more women into paid work and allowing more migration, will be just as important. Last, there may be room for the expansion of global supply chains, as work shifts from ageing China and other middle-income countries, to Africa and poorer places with more labour. Ageing demands a robotics revolution but it may be slow to arrive.