The recent improvements in AI technologies — like ChatGPT, GPT4, and others — have increased the concerns of the role of technology and its effect on labour markets. The fear is a result of what Acemolgu and Restrepo call the ‘displacement effect’ — in which machines replace jobs and leading to increased unemployment, declining wages, and higher inequality.
However, this fear is not new and has been present since the first industrial revolution. It arises from the “lump of labour fallacy” — which is the incorrect assumption that there are only a fixed number of jobs. Since the first industrial revolution, technological advances have led to exponential increases in economic output. However, this growth has always been disruptive, a result of doing new things in new ways which have destroyed the old — what Schumpeter called “creative destruction”.
A secondary, but more important effect of automation is what Acemolgu and Restrepo call the ‘productivity effect’. This effect refers to the fact that more efficient machines lead to economic expansion, which in turn increases the demand for labour in non-automated sectors. In other words, machines contribute to the growth of the economy, leading to new jobs being created that counterbalance the ‘displacement effect’. Another form of the productivity effect is the deepening of automation, in which older machines are replaced with newer ones, resulting in increased output without concern for the ‘displacement effect’.
An example of the positive impact of automation is mechanization of agriculture in the United States and Europe, where the share of agriculture in total employment has declined by over 90% over the last 200 years. However, this automation has led to a reduction in food prices, which has benefited consumers and created demand for more non-agricultural goods and employment. However, this time feels different. It seems like technological change is coming out for most jobs. A report by Goldman Sachs states that “two thirds of current jobs are exposed to some degree of AI automation”. This was essentially the case in the first industrial revolution in Britain from 1760 to 1820.
1st Industrial Revolution
The first industrial revolution was completely transformative to human society. For the first time in human history, there was sustained economic growth. Before this, human society struggled with the Malthusian trap — in which any improvement in technology or productivity saw increases in population, meaning that average incomes remained the same. The first industrial revolution in Britain changed this. Rapid improvements in machine technology saw the growth of factories that used machine tools, replacing more labour-intensive production techniques that relied on skilled artisans. This automation led to widespread displacement of skilled artisans and triggered several riots:
“Between 1811 and 1816 in Britain, the Luddites destroyed machines believing that they would make their skills obsolete. In 1826 in Lancashire, hand-loom weavers attacked weaving machines. In 1830 during the Captain Swing riots, agricultural workers destroyed threshing machines. But technical progress could not be halted, and these skilled workers were quite soon replaced by the machines they tried to fight.” (Daron Acemolgu p.1)
The new factories created new jobs and higher wages for workers, who were less skilled than the artisans they replaced. Overall economic output increased. However, wages in Britain did not increase for everyone, leading to increasing inequality — a period called the “Engels’ pause”.
“As one activity after another was mechanised, hand workers experienced falling earnings either because their own industry was mechanised, or because another was, and workers in it were driven into theirs. Hand workers as a whole suffered falling incomes. The combination of rising real wages in expanding trades with the falling wages in the hand trades meant that the average wage level remained unchanged even as output per worker rose.” (Bob Allen p.11)
During this period, a remnant of the Malthusian past remained, with a positive link between increased productivity and population growth. This high population growth might explain why average wages did not increase by much.
2nd Industrial Revolution
While the first industrial revolution involved “deskilling” combined with large increases in population growth. The second industrial revolution — also called the technological revolution — required more “brain” relative to “brawn”. This period also coincided with increases in human capital in Europe and the US — such as improvements in literacy and expanded basic schooling and the large expansion of research universities. Daron Acemolgu and David Autor see this improved human capital as a causal factor for industrialization. Their reasoning is that technology responds to the supply of educated workers. More educated workers means that there is a greater incentive to invest in more skills-based technologies.
“It is perhaps telling that wages started growing in the nineteenth-century British economy only after mass schooling and other investments in human capital expanded the skills of the workforce.” (Acemolgu and Restrepo p.51)
Declining fertility rates likely played a role in increases in wages in late nineteenth century Britain. Education and fertility rates tend to be correlated, along with improvements in health. Thus, as population growth declined and societies escaped from their Malthusian shackles, income per capita rapidly rose.
“The second industrial revolution, which involved the switch to electivity from steam and water-power energy source, was skill reinforcing because it reduced the demand for unskilled workers in many hauling, conveying and assembly tasks. Over this period, capital-intensive industries considerably increased the demand for skills.” (Acemolgu p.1)
The second industrial revolution likely increased the demand for skills. But there was also a trend of a declining inequality, driven by improvements in skills and education in advanced economies. An example is the dramatic improvements in literacy and high school completion rates. This skills-based technological change, combined with improvements in human capital, decreased income inequality in Britain after 1870. Inequality declined further after 1910 until around 1980, driven by increasing access to education, a larger welfare state, expanding unionisation and other institutional factors.
3rd industrial revolution
While the second industrial revolution in the richest countries would be focused on the rise of manufacturing. The third industrial revolution — also called the digital revolution — would see it decline in terms of employment, as large parts of manufacturing would be automated with industrial robots. Another factor contributing to this decline is the shift in consumption away from goods and towards services.
Trade has also played role, as competition from China and other developing countries has shifted manufacturing jobs. However, the main story of the third industrial revolution is the rise of information technology:
“Since 1973, total factor productivity growth in the US economy has been driven mainly by productivity growth in the IT sector itself and in the industries that use IT extensively, which are mainly services. Labour productivity has risen in manufacturing outside of IT primarily because of increases in capital intensity.” (Bob Allen p.37)
Therefore, resulting in the acceleration of the skills-based technological change from the second industrial revolution. In the book “The Wealth of Humans”, Ryan Avent puts forward the view that changes in digital technology are making some workers very productive and more valuable, while automation is making many others less so. He calls this the "abundance of labour" problem: resulting in stagnant or falling wages for most workers, while profits and rents are rising for those who own the technology or have the skills to use it.
While countries in the second industrial revolution accommodated this technological change focused on skills by expanding the access to education, this process has reached its natural limits in rich countries:
“Evidence from across the rich world suggest that it is very hard to boost secondary-school completion rate above 90%, and to raise university completion rates above 50%. The populations of advanced economies are close to being as educated as they can reasonably expected to be.” (The Wealth of Humans p.55)
However, access to the many high skilled jobs that have been created by the rise of software and computers — which sometimes require technical or scientific training – is inaccessible to many:
“Rapid growth in incomes keeps receding to higher and higher echelons of the income and skill distribution pyramid”. While “the level of education needed to participate in the most lucrative corners is growing beyond the reach of most workers.” (The Wealth of Humans p.57)
Thus, this skills-based technological change without the proportional increase in education has resulted in rising income inequality. Daron Acemoglu and Pascual Restrepo find that between “50% and 70%” of the increase in income inequality in the US since the 1980s has been driven by automation. This increase in income inequality has had macroeconomic effects, as the rich tend to consume less relative to their income. Therefore, increases in income inequality have increased savings rates, leading to much lower interest rates, according to Atif Mian and others.
Unemployment hasn’t increased in rich countries during this time, instead, there has been an expansion in both high skilled and low skilled service sector jobs. The result is a “job polarization”, which has seen a decline in wages and employment of “middle skilled occupations” that involve many routine tasks that have been automated.
The above graph shows the increasing returns to education in the US, which has led to increased income inequality. US males with bachelor’s degrees have experienced stalled real wages since the early 1970s, while the real wages for those with some university experience and below have declined by a lot. The real wages of graduate degree holders have increased by a lot. For women, things have mostly improved for all education groups, but from a low base. The most obvious explanation for these changes is based on skill-biased technological change.
There are several other factors that influence income inequality other than automation. These include the role of trade, labour institutions, and the size of the welfare state — these likely explain the differences between Europe and the US.
4th Industrial Revolution
The fourth industrial revolution — which includes the rise of artificial intelligence, advanced robotics and energy — could potentially see an acceleration in the technological changes that have taken place. These changes are likely to be disruptive to many people. Paul Krugman correctly states that changes in technology take a while to have economic and societal effects. It takes a while for investment and skills to accumulate to take advantage of new technologies. Improvements in AI technologies could change this, OpenAI’s ChatGPT reached 1 million users in 5 days and 100 million active users in two months —the fastest adoption rate for any platform ever.
Not only is GPT-4 reasonably proficient in writing code and producing essays, but it is also expected to improve further with time. Meanwhile, other AI tools excel in generating images. The emergence of these advanced programs suggests the possibility of deskilling — similar to the first industrial revolution — as they provide access to tasks and jobs that were previously only available for the highly skilled.
Or these new technologies could enhance the productivity and skills of the very skilled and lead to increasing inequality, as happened in the third industrial revolution. I don’t have a certain answer to this other than that it will be disruptive. Demographics also plays a role, population growth in the world is the lowest in decades. This decline in the ‘supply’ of humans could help to reduce inequality, as it has to some extent in rich countries over the past two years. A repeat of increased inequality as in the first industrial revolution is unlikely.
Overall, I am certain that there won’t be technological unemployment. Increases in productivity and economic growth rarely leads to increased unemployment. This period will be disruptive, but with the overall effect of increasing the standard of living for everyone. This period could see the rise of the skilled artisan, a group that was destroyed in the first industrial revolution. Human beings are social creatures, and we are obsessed with social status. Art and other hand-crafted products — by being an indicator of social status — could be a way of redistributing income from the rich to the rest. In the world of art, authenticity and low productivity are often prized. Baumol’s cost disease, in which increases in productivity in one sector causes higher wages in other sectors, should see wages rise in all sectors that are not impacted by automation. In my view, the combination of increased productivity, the rise of the artisan, and demographics could lead to a reduction in inequality, even if education levels do not improve significantly.
The Singularity
A future in which we reach the singularity — where computers are far smarter than humans — is hard to predict. I don’t know if or when this would happen. But the trend is heading in that direction, and it could result in robots automating almost every job. Such a future could support a standard of living that we can only imagine, a utopia maybe ending the requirement for work. This would completely upend the world’s economic system, potentially creating a world where universal basic income is prevalent, and jobs are mostly socially based. People are social beings, our relationships are the primary determinant of our long-term happiness, and any likely economic future will be based on that.
The singularity could also destroy humankind, such as if it sees us as a threat. Elon Musk and others have signed an open letter calling for a six-month pause in the development of AI. While some, like Eliezer Yudkowsky, propose more extreme measures such as permanent sanctions and a threat of bombing for those who don’t comply. However, we are still far from achieving the singularity, and technological progress usually takes time to have a significant impact on the economy. Since the first industrial revolution, there has always been anxiety about technological change. Before the singularity — if it does actually happen — improvements in technology will continue to do what is has done before: which is to improve overall human life and standards of living, but in a way that will initially be disruptive to many.