Artificial intelligence is likely to be among the most transformative technologies of our time. The nature of the vast changes ushered in by AI – most notably potential efficiency gains in both white- and blue-collar occupations – can partly explain the attention the technology has garnered, particularly on the heels of developments in large language models such as ChatGPT.
Although the technology itself is wholly new, the macroeconomic challenges associated with AI are not. History provides ample evidence that, while AI is unlikely to spur joblessness or mass unemployment, the likelihood of rising inequality is high. And this brings with it a host of monetary and macroeconomic considerations that will impact economists and central banks alike.
What are the macroeconomics of AI?
A common starting place for economists is to study previous technological revolutions – most notably the industrial revolution and the early 20th century technological transformation (railroads and glassware, for example). Both of these periods coincided with significant growth in labour productivity. Both boosted growth and quality of life for subsequent generations. And both of these moments were replete with contemporary discourse that fretted about the ‘end of work’ and the prospect of widespread unemployment at the hands of new machines.
We now know that previous major technological revolutions did not induce mass unemployment. And so, it often becomes tempting for economists to dismiss such technophobic angst as having been completely misguided. Efficiency gains allowed households to save money due to the lower costs that emerged from more efficient industries (agriculture, for example), which led to the creation of new occupations.
There is currently no evidence to suggest that the economics of AI are substantively different from those of previous technologies. AI promises to greatly expand the scope of what occupations can be made ‘digital’, which may boost productivity and encourage a reallocation of labour towards new industries or occupations, which are best performed by humans. It marks an extension of existing trends in digitalisation – present since the 1980s – as opposed to something entirely distinct.
Yet to dismiss concerns about technological disruption would be a mistake. More contemporary economic analysis has shown that living standards for many working- and middle-class Britons declined considerably during the industrial revolution. Rather than boosting standards of living, the productivity gains spurred by industrial technology led to a de-valuing of labour (often accompanying worsened working conditions), the decline in labour’s share of income, wealth accumulation among capital owners and a consequent rise in economic inequality.
Both the 20th century and early computerisation technological shocks coincided with similar deteriorations in workers’ livelihoods and an often-intolerable growth of inequality. In the early 20th century, the rise of inequality and decline in working class living standards, which defined the ‘gilded age’, were kindled by highly concentrated oil, rail and automobile industries. The early computerisation shock that began in the 1980s was smaller in scale, but highly disruptive to many middle-wage occupations. And from 1980-2014, it was responsible for roughly 33% of the growth in wage inequality that emerged in the US.
Economists like David Autor, Claudia Goldin and Daron Acemoglu have uncovered the mechanisms that have allowed technological shocks to widen inequality. They describe a process of skill-biased technological change, whereby technology shocks tend to primarily displace ‘middle-skill’ workers and lead to growth in the share of low- and high-wage labour. This makes sense: middle-wage jobs are often slower to emerge and more desirable for firms to automate.
When technological shocks occur, workers tend to find that their labour is either complemented or devalued. Those that are able to work with the emerging technologies to enhance their output experience wage premiums. Those whose occupational tasks can be completely or partly replaced by machines see wage stagnation and often a decline in their real wages.
Early evidence suggests that AI kindles SBTC dynamics, which will widen inequality, just as the early computerisation shock did. And so, while the fears of joblessness and unemployment are misplaced, the prospect of rising inequality should be a concern for everyone, particularly in a moment when countries like the US and UK have already experienced a bifurcation of wealth and income since the 1970s. The further growth of inequality would be socio-politically corrosive and disastrous to long-term economic growth, financial stability and a common-sense notion of economic fairness.
AI and central banks
The advent of AI is also beginning to influence central bank decision-making. For over a decade, monetary economists have explored the possible erosion of the Phillips curve, which defines an inverse trade-off between inflation and unemployment. Yet the curve no longer seems to provide an accurate depiction of the relationship between employment and price stability.
A major reason for this has been the reduction in worker power ushered in by the computerisation SBTC shock, which played a role in decreasing labour bargaining capacity. This leads to lower price growth, even when unemployment is low.
AI is also further kindling market concentration. Previous technological revolutions have tended to coincide with the rise of market power by a small number of individuals and firms. The same has been true of digitalisation and is true again of AI, which tends to rely on expensive talent and computing costs.
The testimony of Sam Altman, chief executive officer of OpenAI, in US Congress was a welcome public acknowledgement of the very real threats posed by an under-regulated AI. Yet policy-makers should be wary of corporate regulatory capture and the establishment of guardrails to innovation that allow only the most well-funded ventures to succeed.
The macroeconomic challenges for central banks will be to preserve a healthy economy amid distortions of labour markets that coincide with AI. The benefits of widely deployed AI are potentially significant, taking the form of improved productivity growth after decades of sluggishness. And yet the economics of AI point to an unmistakably disequalising bend, which has implications for how monetary policy should assess countries’ economic health, interest rate deliberations and the vulnerability of middle- and working-class households.
Each large-scale technological shock has coincided with a significant growth of inequality and period of intense dislocation and disruption. Policy-makers and economists have a fleeting moment right now to make sure it doesn’t happen again.
Julian Jacobs is Senior Economist, OMFIF.