Business School news
Read news relating to the University of Sussex Business School.
AI’s emergence does not guarantee a raw deal for workers, evidence from past industrial revolutions indicates
By: Neil Vowles
Last updated: Thursday, 30 September 2021
![A black and red steam engine pulls wooden carriages through a forest](/broadcast/images/uploads/2021/09/13765.item.jpg)
Photo by Sugden Guy sugden on Unsplash
History indicates productivity gains from the fourth industrial revolution could take decades to be fulfilled and the threat of AI to jobs and wages to be overstated, a new study by a leading economic historian suggests.
The productivity benefits of past industrial revolutions took up to 40 years to be significantly realised, the new study by Professor Nicholas Crafts of the University of Sussex Business School details.
But AI-sceptics are mistaken in thinking that the history of past industrial revolutions makes it inevitable that the introduction of the new game-changing and labour-saving technology will drastically harm job numbers and wages, the paper explains. The experience is of a changing composition of employment rather than a massive decline in employment.
A key implication from previous industrial revolutions is that governments must deliver policies that reduce adjustment costs and facilitate redeployment of workers to offset the initial disruption that AI will bring, the research warns.
The study, published today in the Oxford Review of Economic Policy, explains how a considerable lag between the introduction of a new technology and a rise in productivity can be the result of incremental improvements to technology over time, gradual reductions in the quality-adjusted price of the capital equipment and the time taken for investments in organizational change, as witnessed during previous industrial revolutions.
Prof Crafts, Professor of Economic History at the University of Sussex Business School, said: “It typically takes time before a general-purpose technology (GPT) has a substantial impact on productivity. The technology improves, complementary investments and innovations are made, businesses are re-organised and learning accrues. The examples of steam, electricity and ICT illustrate that time-lags are to be expected. It is quite plausible that this will also be the experience with AI as it progresses.
“It is a common misconception that the First Industrial Revolution is a template for a general-purpose technology having a major adverse effect on workers’ living standards. The essence of that industrial revolution was not rapid productivity growth in the short run but the ‘invention of a new method of invention’ which increased technological progress in the long run. Since AI is potentially a general-purpose technology that raises the productivity of research and development, it may be the basis for a Fourth Industrial Revolution.
“It is highly likely that AI will eventually become to be seen as a classic GPT and eventually deliver the much-needed boost to productivity that techno-optimists envisage once its full potential is realised. Growth accounting estimates for earlier GPTs show that their impact on productivity takes time to develop.”
The new study details how the impact of the First Industrial Revolution, and therefore its potential as an indicator of the impact of AI on the jobs market, has been mischaracterised.
Prof Crafts’ close analysis of contemporary data reveals that the impact of steam power was only felt at the end of the first industrial revolution after a very long period of slow wage growth for workers which reflected slow productivity growth rather than technological unemployment. Although some types of employment declined, new occupations blossomed in the context of the variety of tasks required by industrialization and urbanization.
The study details how the big impact of electricity in the United States came about 40 years after Thomas Edison first distributed electrical power in New York in 1882 following widespread redesign of American factories.
Similarly, the impact of steam power on productivity growth in Britain was negligible prior to 1830 with its impact limited to a small number of industries including coal mining, cotton textiles and metal manufactures as the cost effectiveness and diffusion of the new technology was held back by the high coal consumption of the original low-pressure engines.
Prof Crafts said: “It is quite reasonable to think that we are at the point with AI as a general-purpose technology early in its lifetime with its significant impact on macroeconomic productivity performance lying in a not-too-distant future. We would expect its impact to come into fruition sooner than the relatively lengthy time that it took for steam and electricity to make a noticeable difference though potentially not as instantaneously impactful as some of AI’s critics fear.
“Western societies have been getting better at exploiting new technological opportunities, thanks to superior scientific and technological capabilities, greater expenditure on R&D, and more sophisticated capital markets, so that the impacts are felt more quickly. The impact of ICT was relatively large and unprecedented in its rate of technological progress. We should expect the same with AI. Added to this, machine-learning systems are designed to improve themselves over time and so by nature should advance rapidly.
“If AI fulfils its promise, it will alleviate the current productivity slowdown. Nevertheless, many other things matter for productivity performance as well as the arrival of a new GPT. Optimism over AI is no reason to neglect supply-side policy reforms, for example, to address the trend to weaker competition in the US and other global economies.”