Late last week I attended a very interesting meeting in Madrid on the subject of what some have described as a coming ’second machine age’ and its possible future economic and social impact. The technological developments described at the meeting by scientists, engineers, business people and others who follow them closely center on possible breakthroughs in robotics, artificial intelligence and pervasive internet-enabled monitoring and linking of functions in daily life. Those who discuss these developments are often concerned about their possible employment impact, which is only one of the many kinds of impact they may have. Their effect on human self-understanding and the nature of everyday social experience may be as profound. As the discussion proceeded it became increasingly apparent to me that the possible employment impact is poorly understood. It is often assumed that ‘technologically induced unemployment’, at least in the short to medium term, is most likely to affect the relatively less-skilled. However, this is far from obvious, since even some tasks done by ‘unskilled’ humans (for example facial recognition) have proved very difficult for computers, and a very high degree of increase in machine competence will therefore be required to make machines possible substitutes. Machines are still very poor at general and flexible task specifications even if they can undertake particular tasks well (for this reason, ‘busboy’ was mentioned as an occupation for which there would remain robust future demand). Moreover, cost will continue to be a factor in determining whether it is profitable to substitute machines for humans.
The cost effectiveness of machines will eventually depend on whether they are ‘good enough’ at the required tasks, the progress of technology and the extent of their adoption (affecting the possibility of spreading large initial fixed costs over future units as well as learning by doing). This is likely to make certain activities subject to more rapid mechanization than others. As in the case of driving, in which humans are likely to play a role alongside machines for a considerable time for reasons of social acceptance, even if their role is rarely if ever truly needed, machines may not substitute for humans as much as supplement them in certain domains. Nevertheless, the economic effect will often be either to make it possible to replace skilled workers with less skilled and more widely substitutable workers who are paid less or to hire fewer skilled workers (think of medical scans which are mostly classified by machine and only reviewed by an expert physician). In certain cases, less skilled workers assisted by artificial intelligence may be able to do tasks that only more skilled workers could previously do, thus potentially providing more employment for the less skilled and less for the more skilled. Which particular relations of complementarity and substitutability are likely to prevail is a deeply empirical question for which armchair theorizing is not much good. Notably absent in the discussion was much attention to David Ricardo’s famous concern, in his chapter On Machinery, of whether the increased social product would ultimately be put to use to increase employment overall. Considerations of the consistency of increased output with aggregate demand (of who will pay to purchase the increased social output and how) were also not much recognized. Some participants assumed that a universal basic income, a social profit sharing fund, or other alternative would be needed to fairly distribute claims on output and to manage the social peace. A darker perception, however, was that far from ushering in a society of leisure, as hoped by Keynes and not yet attained, continued mechanization would lead to increasing work expectations and resulting continued or increased ‘time poverty’ for those re-skilled few who are employed alongside under- or unemployment for the de-skilled many (in short, our world made starker). Contrary to Thomas Piketty’s remarkable presumption, Marx like Ricardo was centrally concerned with the effects of technological change, and this possible future is indeed illuminated by their analyses. What is clear is that economics has not yet faced up to the challenge of interpreting how a ‘second machine age’ could affect employment and incomes in a world economy in which the link between manufacturing, output and employment is already precariously weak, in poor countries as well as rich. Since we already live in a far from full employment world, the view that full employment will be restored automatically seems far too sanguine. On the contrary, what is needed is a forward looking vision and relevant public policies to influence the evolution of science and technology and how they are dealt with by human societies. It is indeed hard to know the future, but there are ways of being prepared for it.