In the last section of The Technology Trap, Carl Frey looks toward the future, trying to anticipate further impacts of technology on jobs, and suggesting policy measures to ease the transition for affected workers.
Artificial intelligence is enabling machines to do even more of what humans used to do. “The fundamental difference is that instead of automating tasks by programming a set of instructions, we can now program computers to ‘learn’ from samples of data or ‘experience.’ When the rules of a task are unknown, we can apply statistics and inductive reasoning to let the machine learn by itself.” When a computer beat the world’s best player of the game Go in 2016, it did it not just by following a fixed set of rules, but by inferring its own rules from a series of trials using a large data set.
The range of tasks that smart machines can perform is broadening to include jobs like driving a truck, answering phone calls, picking up and packing products, taking consumer orders and accepting payments.
Still, there remain things that humans do better:
Even if we assume that algorithms at some point will be able to effectively reproduce human social intelligence in basic texts, many jobs center on personal relationships and complex interpersonal communication. Computer programmers consult with managers or clients to clarify intent, identify problems, and suggest changes. Nurses work with patients, families, or communities to design and implement programs to improve overall health. Fund-raisers identify potential donors and build relationships with them. Family therapists counsel clients on unsatisfactory relationships. Astronomers build research collaborations and present their findings in conferences. These tasks are all way beyond the competence of computers.
In 2013, the author and his Oxford colleague Michael Osborne reported on their detailed analysis of tasks and their estimate of the automation possibilities for 702 occupations covering 97% of the American workforce. They found the greatest risk of automation in the occupational categories of office and administrative support, production, transport and logistics, food preparation, and retail jobs. Overall, they classified 47% of jobs as vulnerable to automation.
Other research has yielded somewhat different percentages. But one general principle that has emerged from such research is that a job’s probability of automation varies inversely with the education it requires and the wages it pays. A study by the President’s Council of Economic Advisers found that “83 percent of workers in occupations that paid less than $20 an hour were at high risk of being replaced, while the corresponding figure for workers in occupations that paid more than $40 per hour was only 4 percent.” That could be good news, as long as we can keep expanding the good jobs and help workers acquire the skills they need to do them.
Unemployment, leisure, or new jobs?
Frey describes a “widespread dystopian belief” that technology will create a future of mass unemployment and low wages. Others envision a utopian future in which technology enables us to produce so much so easily that we can work very little and live lives of affluent leisure. Neither mass unemployment nor lives of leisure are evident in today’s society, and Frey doesn’t expect them. Instead people will have jobs for the foreseeable future, both because there remain things people do better than machines, and because people generally choose to take the benefits of high productivity in the form of more goods and services rather than more leisure.
Although new technologies have been replacing more middle-class jobs than they have been creating, Frey suggests that this may be just a “first-order effect.” He believes that the greatest gains in productivity and job creation are yet to come. That reinforces my belief that whether a new technology turns out to be replacing or enabling depends on how we use it in a social context. Replacing existing jobs may happen first because it’s easier than creating new jobs and upgrading skills, which requires some social reorganization. Frey points out that “it took roughly four decades for electricity to appear in the productivity statistics, after the construction of Thomas Edison’s first power station in 1882….[H]arnessing the mysterious force of electricity required a complete reorganization of the factory.” And of society, I would add, considering the changes required to turn workers and their families into affluent consumers of the products coming off the assembly lines.
In the end, Frey remains optimistic about technology, but concerned about the divisions between current winners and losers and their immediate effects on society. Mitigating those effects is the main challenge for public policy. Among his recommendations:
- Investments in education, especially early childhood education to offset the disadvantages of children from low-income, low-education families; such education pays for itself in better health outcomes, higher productivity and reduced crime
- Wage insurance, especially for middle-aged workers who lose good jobs
- Expanded tax credits to supplement low wages
- Easing of licensing requirements that make it too difficult to move into new occupations
- Vouchers to pay for moving to areas with better job opportunities
- More affordable housing in thriving communities, supported by an easing of zoning restrictions like minimum lot sizes
I see a role for government not only in helping disadvantaged workers, but in creating economic demand for the good jobs they need. If the manufacturing sector is no longer expanding, and if the low-wage service sector is most vulnerable to the next wave of automation, then that leaves the skilled services as the most likely frontier of job creation. But skilled services like education, health care, counseling, mental health services and quality child care are also what people need to enhance their human capital and qualify for good jobs. Public investment in those services pays off in two ways–better jobs and more qualified workers to do them. It also strengthens democracy because successful workers are more politically active and less alienated.
Why public investment rather than private investment? Because the families most in need of such services often cannot afford them. And because employers have only limited incentive to develop the human capital of their own workers. Employers own the machines they buy, but not the workers they hire. The workers can take their enhanced human capital and go to work for someone else. For that reason, human capital is a public good that cannot be entirely privatized. A healthy, well-educated population is good for all of us. So, of course, are other public goods like a solid infrastructure and renewable energy.
But can the country afford new investments in health or education? If the government seems tapped out, it’s not because the country is poorer than it used to be, but because the wealth and income are so unevenly distributed, and those who have them support such low taxes on themselves. From the Reagan administration on, the tax cuts were supposed to stimulate the economy from the top down, by making more money available for private investment. The results have been disappointing, with slower economic growth than in the mid-twentieth century, when taxes were higher. Now we should consider the possibility that we can grow the economy faster with high domestic spending than with low taxes, if the spending is concentrated on human capital development and needed public goods. In order to make human services affordable for consumers and for the taxpayers, they need to be cost-effective. Providers will need to apply new technologies not to replace labor–which would defeat the purpose of creating jobs–but to enable labor to serve clients as efficiently as possible. In the predominantly service economy, a productivity revolution in skilled services is the key to fulfilling the positive potential of information technology.
Advocates of new government spending have their work cut out for them to mobilize public support. They need to convince the less educated half of the population that they will receive more benefits than costs, since their incomes are too low to be targeted for tax increases. If they can also convince the more educated middle class to vote in the public interest, they can achieve a democratic majority. As Frey says, “Redistributive taxing and spending depend on whether the middle-income voters feel an affinity with people with lower incomes.”
Although my interpretations and policy preferences differ from Frey’s in a few respects, I found this book enormously helpful in thinking through the relationship between technology and employment. I highly recommend it.