Roger Bootle. The AI Economy: Work, Wealth and Welfare in the Robot Age. Mobius, 2021.
One of the most intriguing and important questions of our time is how advances in artificial intelligence and robotics may transform how we work and how well we live. Roger Bootle brings an economist’s perspective—and I think a healthy does of realism—to the question. He is interested in countering some of the more fanciful visions of the future that may make good science fiction, but are too unhinged from current economic realities. Some of these are apocalyptic, such as Stephen Hawking’s statement that “the development of full artificial intelligence could spell the end of the human race.” Some are quasi-religious, like Ray Kurzweil’s suggestion that humans could achieve eternal life by merging with intelligent machines and transcending our physical bodies.
Bootle acknowledges that the future is always somewhat uncertain, but he finds some projected futures, especially near ones, more plausible than others. He believes that macroeconomics can help keep the discussion grounded in reality. Although the terms “artificial intelligence” and “robot” may immediately make us think of humanlike entities, he reminds us that robots remain just “mechanical devices that can be programmed to act in certain ways.” From an economic standpoint, we should think of them as capital equipment rather than human labor.
The industrial analogy
For Bootle, the key question is, “Are the changes taking place now, thanks to robots and AI, essentially a continuation of what we have seen since the Industrial Revolution? Or are they something quite different?” He believes that the continuities and similarities are strong enough to make economic history very helpful in anticipating what is coming.
First, he points to the remarkable surge in economic progress associated with the Industrial Revolution in its various phases. Estimated global per-capita GDP was over thirty times higher in 2000 than it was in 1800. Economic revolutions had occurred before, notably the Agricultural Revolution that brought the planting of crops and domestication of animals. But they took place much more slowly over many centuries, and the gains in production were smaller and largely offset by human population increase. The Industrial Revolution was not only faster, but it involved social changes that complemented technical developments like the steam engine and electricity. Sustained capitalist investment in such technologies raised productivity by increasing the amount of capital equipment per worker. Acquisition of new skills added human capital to the equation. Increases in trade encouraged divisions of labor and economies of scale.
Although rapid by historical standards, improvements in living standards did not come overnight. According to economic historians, “it was only after 1870 that European real wages rose decisively above medieval levels, with Britain leading the way.” Another big leap in living standards occurred during the postwar economic boom of 1945 to 1973. The average worker gained purchasing power despite working fewer hours than previous generations. Then that boom ended, and more recent decades have seen slower growth and more unequal distribution of benefits.
While the Industrial Revolution boosted productivity and incomes in the aggregate, workers in all industries did not benefit equally. Labor-saving technologies destroyed massive numbers of jobs, especially in agriculture. At the same time, it created many new jobs in manufacturing, some of which eventually became relatively high-productivity, high-paying jobs like auto worker. Contrary to the fears of many social critics, technological change neither reduced employment overall nor lowered per-capita wages. Today the locus of job creation has shifted toward services, and the future of work and income depends mainly on productivity and pay in that sector.
Is this time very different?
Some writers think that the impact of artificial intelligence and robotics will be markedly different from the industrialization experience. Bootle discusses their arguments, but does not find them very convincing.
Robert Gordon has argued that the Industrial Revolution introduced a unique set of technological developments, and nothing resembling it is likely in the foreseeable future. The slower growth in investment and productivity in recent decades is the new normal. Bootle argues that “the digital revolution needs time to play out,” and that “we are on the brink of new developments that promise to bring rapid advances: biotechnology, nanotechnology, 3D printing, robots and AI.” He expects new technologies to “reinvigorate the engine of growth which has seemed to sputter and stall over the last two decades.”
At the opposite end of the spectrum from Gordon are thinkers who are so impressed with new technologies that they expect impacts even more momentous than those of the Industrial Revolution. They usually assume that whatever humans can do, intelligent machines can do better—or more precisely, learn to do better now that they can master tasks with more autonomous learning processes. Machines of the future will replace not only physical labor, or highly repetitive mental labor, but potentially all labor.
Bootle takes an intermediate position. He notes that predictions of massive unemployment from automation remain speculative, since “many of the countries with the highest number of robots per worker (e.g., Singapore, Japan, Germany) also have some of the lowest unemployment rates.” He believes that living human beings have qualities of consciousness, creativity and emotional intelligence that remain beyond the capacity of dead machinery. He also cites “Moravec’s paradox,” that robots often have more trouble with tasks humans find simple (tying a shoe or getting a joke) than tasks we find complex (calculating a statistic or remembering the names of all U.S. presidents). Although some forms of human work will disappear, other forms will persist or expand. The areas of expansion are most likely to involve jobs where human labor is assisted and complemented by intelligent machines. That creates the potential for a long overdue productivity revolution in human services such as education. As always, some will benefit more than others, but on balance the results should be beneficial for our quality of life.
The macroeconomic impact
If the digital revolution turns out to be another phase of technological progress in continuity with earlier phases of the Industrial Revolution, then its long-run impact on the economy ought to be positive. We ought to see a period of sustained high investment, gains in productivity, and strong economic growth.
If anything like the vision of more rapid growth of productivity expounded here is realized, then the result will be a faster rate of economic growth than has been normal in the West over recent years, and a faster increase in living standards, even on the inadequate measurements that we have, such as GDP or real incomes, or real consumption per capita. Moreover, because of increased leisure time, the underlying rate of improvement in the human condition will be still greater.
Economic arguments always come with a lot of ifs, ands, and buts, however. Economists want to identify the economic mechanisms by which greater productive capacity translates into higher purchasing power for the average household. In theory, “production gives rise to income that can be spent to buy what has been produced,” which keeps the cycle of investing and spending going. Suppose, however, that most of the income is going to the owners of robots, leaving the workers too poor to afford many of the goods and services being produced. That resulting shortfall in aggregate demand then discourages new investment, so that the economy’s productive potential cannot be realized.
One possible answer is that the emerging economy will need a lot of human capital (skills) as well as physical capital (equipment). That will give qualified workers leverage to claim a reasonable return on their contribution to productivity. Workers will have to compete for good jobs, but employers will also have to compete for qualified workers. Of course, this assumes that workers can obtain the skills they need to work productively with the technology.
Bootle also hopes that other systemic factors will support high aggregate demand. He says that the aging of many populations has created a large generation of people saving for retirement, but they should spend more when they are actually in retirement. Also, countries like China and the oil producers have held wages and spending down in order to promote their cheap exports to richer countries, but show signs of increasing their demand for consumer goods in the future.
This discussion may strike the reader as a little irrelevant, given the recent inflation. The problem right now seems to be excess aggregate demand, given the recent supply shortages, tax cuts and stimulus spending to counter the Covid economic disruptions. It takes a little imagination to envision a world in which a significant surge in productivity threatens the economy with an abundance of supply but a shortfall in demand. But that was the problem preoccupying many economists in an earlier era of industrial expansion. We have been there before, and could be again.
In later posts, I will go on to discuss more specific topics in The AI Economy, such as jobs, leisure, and macroeconomic policy.