Future of Work: Will a Robot Take Your Job?
Will artificial intelligence bring an end to the traditional labour market? Despite the powerful impact this narrative has on economic policies, modern economic theory suggests it’s technological stagnation and mediocre innovations that labour markets should be afraid of – not highly efficient technologies.
ABOUT THE AUTHORS
Eljas Aalto
Eljas works as an economist and data analyst at Futures Platform. He develops quantitative foresight analysis tools and participates in various customer projects to help them prepare for the future.
Henrik Södergrann
Henrik is a Foresight Analyst at Futures Platform and works with economics and finance-related research, foresight and the quantitative development of foresight analysis.
FUTURE PROOF – BLOG BY FUTURES PLATFORM
In our Challenging the Obvious article series, Futures Platform’s futurists and economists look at trends from different angles, raise what-if questions and inspire new thinking by shedding light on the alternative future possibilities that lie beyond the obvious.
EFFECTS OF INDUSTRIAL REVOLUTION – IS THIS TIME DIFFERENT?
There’s no denying it: Artificial intelligence, robots and automation are here to stay, and they are already having a powerful impact on the global economy. Alongside the purely positive future visions, these technologies have also given way to some dystopian scenarios: Will AI and automation take over all future jobs? Will all income flow to wealthy robot owners? Do we need a universal basic income scheme to feed masses of unemployed workers? Even this text, and other types of writings, could be written by an elegant machine learning algorithm in the future. Oh no, stop the innovation!
These fears are reasonable, but they’re nothing new under the sun. Already during the time of the industrial revolution, a group of workers known as Luddites protested against mechanised manufacturing in the textile industry. The Luddites’ message was clear: technological change causes unemployment and should be resisted. This narrative and the fears of robots taking over jobs have boomed time after time, especially during economic downturns.
Yet, economic history has proven the Luddites wrong: New technologies led to growth in productivity and income, creating demand for new types of goods, services and jobs. Currently, most workers in the West keep their hands busy with tasks that didn’t even exist in the 19th century. Figures from the labour market and national accounts also contradict the Luddites’ arguments: Today, unemployment levels in the Western world are not higher than what it was before the beginning of automation, and the share of national income going to labour has largely been constant during the last century – even though technological innovation has far from stopped since the Luddite movement.
Unemployment rates between the years 1950-2020. Source: U.S. Bureau of Labor Statistics
But can we be certain that economic growth and structural change will always create new jobs? What if we succeed in developing artificial general intelligence – an AI with human-like intelligence – that will make human workers obsolete for good? To answer these questions, we need to turn to modern economic theory and the historical patterns of technological employment effects.
Subscribe to our weekly newsletter
MASS JOB LOSS TO AUTOMATION: IS IT POSSIBLE?
Labour methods remained largely unchanged for thousands of years until the industrial revolution. Humans designed simple devices and improved existing ones to increase work efficiency and output, but it wasn’t until the industrial revolution and early mechanisation that we experienced mass replacement of human labour. Historically, the jobs replaced by automation were simple and repetitive, and didn’t require any artisan skills or human-like aptitude, such as creativity. Mechanisation relieved workers from simple manufacturing tasks and pushed them into the service economy, where the primary input is labour itself. And now, with the advent of human-like solutions driven by AI and deep learning, this labour is in danger of becoming redundant.
In the wildest AI protagonist scenario, we can picture that the knowledge worker, who collects data, analyses it, and provides management with the best solutions, is now replaced by an algorithm who neither complains about the workload nor asks for a pay raise. Feeding on its success, the AI grows in power, and the management becomes its next victim. Business decisions are now based on data, and the AI patented solutions are scalable. In this scenario, ownership of the AI will emerge as the only factor of significance. Everybody else is redundant in the economy.
Surely, no one wants this scenario to become a reality, except for the privileged few. What structural changes would then need to be implemented to avoid the detrimental impacts of these new technologies?
HOW DOES AI IMPACT FUTURE JOBS AND MARKETS?
A recent paper by the current superstar of economics, Daron Acemoglu, and his co-authors sheds light on this question with an interesting new framework.
In traditional theoretical models, technology is just a multiplier that makes labour or capital inputs more effective. Such models are unable to capture the real effects of automation and the changes in task composition. In Acemoglu’s model, economic production consists of a continuum of different tasks, where some are done by humans and some by machines. Technological innovation doesn’t simply make labour more effective; it replaces humans with machines in some jobs and, in some cases, creates entirely new jobs.
Daron Acemoglu: Robotics, AI, and the Future of Work, Source: CIFAR
In this framework, replacing labour with machines has two different effects. The first one is the displacement effect: replacing humans with machines decreases employment levels. The second is the productivity effect: automation increases productivity and lowers costs, enabling the producer to expand production. This will increase the demand for labour in tasks that are not yet automated, which in turn has a positive effect on employment. The final impact on employment depends on which of the two effects dominates.
Here's an example demonstrating how these two different effects work in practice: Suppose there is a car manufacturer employing ten workers on the production line and ten people in administrative tasks such as sales, marketing, R&D, and management. The firm decides to replace the ten people on the production line with robots. This lowers employment by ten people due to the displacement effect. Next, it turns out that these robots are much more efficient than human labour. The cost of production decreases significantly, and the firm doubles its production. Now it must employ ten more people for the administrative tasks that are not automated. In conclusion, there is no effect on overall employment. Automation didn't create mass unemployment in the car manufacturing sector, but the skill composition changed. The model suggests that technological automation that is substantially more effective than human labour has positive effects on employment due to the productivity effect. Hence, there’s no reason to fear it.
On the contrary, what Acemoglu fears is the diffusion of technologies that trigger only a minor productivity effect but have huge displacement effects. Such technologies, which Acemoglu calls so-so technologies, replace lots of human tasks but yield only minor improvements in productivity. Based on this, labour markets should not be afraid of revolutionary innovations but fear the diffusion of mediocre innovations.
EXCESSIVE AUTOMATION AND MASS TECHNOLOGICAL UNEMPLOYMENT
Replacing human labour might be optimal for a single firm to cut costs. However, it is likely that ideological factors, lack of knowledge, or ill-advised economic policies will prompt businesses to automate excessively. This leads to inefficient usage of the so-so technologies, where the displacement effect dominates and might produce negative externalities – an economic concept that describes the unaccounted-for effects of economic activity on a third party – in this case, the society.
Acemoglu argues that the observed loss of jobs in the manufacturing sector, which has triggered various economic, social and political problems, is essentially a symptom of excessive automation. These jobs have been exposed to robots more than what is optimal. Instead of excessive robotisation, resources should have been allocated in a completely different direction, such as developing sustainable energy solutions. According to Acemoglu, climate change and increasing economic inequality are the adverse impacts resulting from this mistake.
Acemoglu also states that there is a risk of artificial intelligence being a so-so technology: Currently, AI is used to automate many simple tasks, which humans are already quite decent at. Pattern recognition, such as facial recognition, is a good example: the human error rate is often so low that a machine learning algorithm will not bring any major improvements. Customer service is another example of insignificant productivity gains attributed to the implementation of technology: Many humans in customer support have been replaced by chatbots that are unable to answer most non-conventional problems.
Unless we develop general AI that results in radical productivity gains, AI will likely end up having a negative effect on overall employment due to displacement effects. Additionally, it will create new problems related to privacy, competition, and behavioural manipulation. This argument is the opposite of the prevailing narrative, which states that highly efficient, human-like general AI has the worst effects on employment.
HOW DOES THE FEAR OF AI TAKING OVER JOBS IMPACT THE ECONOMY?
The story about machines replacing humans has been told time after time during the last two centuries, and it has certainly affected consumption patterns, investments, entrepreneurship and economic policies. In his book Narrative Economics, Nobel-winning economist Robert J. Shiller argues that it is the fear of robots taking over jobs that has driven amplified booms and recessions – not technological unemployment per se.
Every time fears have surged, the effects of temporary business cycles and temporary unemployment have been misinterpreted as effects caused by the new job-destroying machines, which deepened the pessimism experienced by the unemployed and made the economic consequences even worse. According to Shiller, events such as the global financial crisis, the dot-com bubble, and “the automation recession” of 1957–1958 are all great examples of the power of this narrative.
In recent years, the narrative of technological unemployment has had a formidable impact on the inequality debate. This discussion has been strongly influenced by the work of scholars such as Yuval Noah Harari and Thomas Piketty, who fear that the future will hold a small share of wealthy capitalists, “the 1 per cent”, who own the bulk of all machines while the working class remains useless and poor.
As a solution to this scary scenario, a universal basic income financed by the taxation of robots has gained prominent support, thus reflecting a future analogous to the one Karl Marx foretold – the shared ownership of the means of production. In the future, this narrative of technological unemployment caused by automation will also strengthen the support for capital taxation and labour unions. Hence, it is possible that the influence on economic policies will be completely unrelated to the actual effects of automation and artificial intelligence on the labour market.
WHAT WILL THE FUTURE OF WORK LOOK LIKE?
It’s impossible to predict the exact impact of new technologies on the labour market. If technological innovation stagnates, there will be a significant risk of vast automation, leading to growing unemployment and inequality. In this scenario, politicians and regulators might need to step in and prevent excessive automation with accurately targeted policies and incentives.
However, if new revolutionary and disruptive innovations with significant productivity improvements emerge, the economy should not be afraid of radical structural changes within its labour market. Hence, when it comes to employment, it is essential to keep the wheels of technological progress spinning and adjust the skill composition of the workforce.
Lastly, since the narrative of technological unemployment and labour-saving machines will endure, public policy will certainly be affected by the fear itself, no matter what happens in the labour market.
Want to take a deeper dive into this topic and explore what the future of work may look like? Try Futures Platform’s digital foresight solution for free and access a rich library of 1500+ trend and scenario analyses across all industries.
Literature
Daron Acemoglu & Pascual Restrepo, 2018. “Artificial Intelligence, Automation and Work”, NBER Working Papers 24196, National Bureau of Economic Research, Inc.
Robert J. Shiller, 2019. “Narrative economics: how stories go viral & drive major economic events”, Princeton University Press.
