I was nonplussed. For the previous decade, I had been reporting on the rapid expansion of new AI technologies into the workplace and how they were about to displace not only blue-collar manufacturing jobs but, for the first time, white-collar knowledge workers like lawyers and doctors. But that night, Kahneman alerted me to a largely unexamined aspect of the AI-fueled automation debate. As he pointed out, China is a rapidly aging society. Its one-child policy has ensured that the working-age workforce in China — people aged between 16 and 59 — is actually contracting. The number of working-age people in China is expected to fall to 700 million by 2050. That’s a decline of 23 percent from 2012.
As I began to explore Kahneman’s theory, I came to realize that an aging population is not just a Chinese problem. In 2015, a U.S. Census Bureau report noted that, sometime before 2020, for the first time in history, there would be more people in the world over 65 years old than under five. Populations are also aging quickly elsewhere in Asia — South Korea and Japan in particular — as well as in much of Europe. Even the U.S. is a graying society, although less so than the rest of the world because of more liberal immigration policies — until recently at least. As birth rates fall, the whole world, with the exception of Africa and parts of the Middle East, is getting older at an astounding rate.
Globally, the number of people over 80 will double by the middle of the century — almost half a billion people will fall into the neediest care category — and that percentage will increase by sevenfold by the end of this century. The dependency ratio — that proportion of humans who require care compared to those who can give care — is also increasing inexorably.
Japan has understood this transition for some time and has led the world in efforts to develop care robots for the elderly. But the U.S. is still largely ignoring the coming wave of aging. Not long ago, I gave a talk at Apple in Cupertino where I was invited to speak to a group of machine learning developers after the publication of my book, “Machines of Loving Grace.” I suggested that the company’s future might not lie in phones but rather in self-driving cars. After all, it was roboticist Rodney Brooks who suggested that self-driving cars will be the first elder-care robots. My appeal, however, drew mostly blank stares; the Apple engineers did not seem to realize that their once young market of baby boomers will be in assisted living facilities in a matter of a few years.
Although it is starting to change, the general disinterest in technological innovation and deployment for elderly people is evident in industry trade shows like Aging 2.0, an annual Silicon Valley conference that focuses on technology designed for the elderly. At each exhibition, there is a usually a modest range of start-ups offering products such as Internet communication services that address the challenges of a geographically distributed family wanting to stay in close contact with an aging senior. For the most part, however, the dominant technology and consumer electronics firms don’t view the aging as a market.
The commercial prospect of autonomous robots that might either assist or ultimately supplement human health care workers is still far in the future. What is more likely to appear in the coming decades are systems of sensors and virtual assistants that will allow aging people to remain in their homes for far longer. At Stanford University, for example, medical and AI researchers have piloted systems that are based on infrared cameras that can detect when an elderly person falls.
Despite the fact that robotic human surrogates might not appear for decades, the possibility of “machines of loving grace” that might replace human caregivers has sparked a bitter debate, best framed in 2014 in a dispute between Louise Aronson, professor of geriatrics at the University of California San Francisco, and Zeynep Tufekci, a sociologist at the University of North Carolina.
Pointing to the increasing scarcity of caregivers, Aronson called for robots to care for people and, more importantly, act as companions for those who are isolated and lonely. In response, Tufekci argued that a decent, ethical society would increase the number of human caregivers by raising wages as an incentive to encourage human contact with elders. “Surely we should mourn if we put our elderly and our children in ‘care’ of metal objects animated by software,” she wrote.
Tufekci obviously has a strong moral case. However, as families have disintegrated in the U.S., institutional elder care has become the norm. It is hard to envision a scenario where American society will dramatically increase salaries for elder-care givers, jobs that in this country are frequently performed by immigrants.
There are points to be made on both sides of this argument. If an aging society chooses not to provide human care for the elderly, or crowds them in assisted living facilities and places them in front of television sets for hours on end, could a world in which augmented reality, monitoring sensors and virtual assistants that offer companionship actually be better? There is evidence that human contact offsets dementia. What if research shows the same cognitive benefit from AI-based virtual assistant systems that interact with the homebound elderly, permitting them to be in constant contact both with family members and with artificial companions?
The implications of the rapid adoption of conversational interfaces — systems like Apple’s Siri, Microsoft’s Cortana and Amazon’s Echo — are clearly a step along a path to a world in which speech will be one of the most common ways we interact with computers and web services. Indeed, for the aging, it is one of the most promising ways of extending independence.
Several years ago, I wrote about a Microsoft experiment in China with a virtual assistant named Xiaoice. Unlike Siri and most of today’s virtual assistants that are designed to accomplish tasks, Xiaoice is a chatbot that was designed as a conversational companion. It rapidly attracted an engaged audience of more than 40 million, mostly young Chinese users, who had an average of 60 interactions with the program monthly. Significantly, it was also able to help simulate an emotional connection. At least 25 percent of the 40 million users texted the phrase “I love you” to the bot. The Microsoft researchers were disturbed by the response to what they had created.
However, when I spoke with Chinese engineer Michelle Zhou, she had a very different view. She described the feeling that Chinese have when they come to America as arriving in a “socially barren” society, contrasting Chinese society to American society and suggesting that the way Xiaoice was used in China was different than it would be in America.
Recently, the author Nicholas Carr attacked the Siri and the Echo software bots in an essay, suggesting they represent a step toward a world that will increasingly isolate us from each other, leaving us trapped in a web of surveillance sentinels. But maybe there is room for a middle ground. Perhaps there is a humanistic path to be found in creating a web of sensors and conversational systems that will simultaneously be connected to human caregivers who will continue to be the most important source of comfort for the aging.
Already, TaskRabbit, which allows people to hire someone for odd jobs, and Lyft, a ridesharing car service, are helpful for the elderly who want to stay in their homes longer. If these services can be designed to offer a living wage for young workers, it might not be such a grim future. Perhaps before we have true self-driving cars, new safety systems will be designed that will permit aging drivers to drive themselves safely. Maybe we will turn from a young society playing multiplayer video games to a networked world in which elders stay in contact via augmented reality systems, making it possible to travel virtually if not physically, giving them a quality of life they may currently lack.
Despite some progress, the impact of AI systems will depend on how we deploy them. It comes down to the question of whether or not we use this technology in humanizing ways. In 1962, computer pioneer Douglas Engelbart called this “augmenting human intellect.” In Engelbart’s world, the human is at the center of each technological system. It remains the best way to engineer the coming advances in AI.