Written by Imad A. Moosa

Free marketeers tell us repeatedly that we have nothing to worry about as long as we put our faith in the market, which provides salvage as we go along. Some free marketeers go as far as suggesting that adopting the free-market doctrine in its purest form (meaning no public policy or public sector whatsoever) guarantees the avoidance of economic and financial mishaps. This is how an over-enthusiastic free marketeer describes life under the care of the almighty market where the “invisible hand” operates without any constraints:¹

It is hard to imagine the world with a completely free market. People would be much richer. We would have such advanced technology. We would live to be a hundred and fifty or older, because biotechnology has the capacity, if freed up from regulations and freed up from state control, to really extend human life dramatically. We might have rocket ships going to Mars, we might be colonizing Mars. More importantly than that, there would be no poor people in the world.

A natural response to this heroic and extravagant statement is “oh my God” or “wow”. In reality, however, this description of life under the purest form of free market economy defies logic. For one thing, people would die at a young age because they would not be able to afford the cost of healthcare. Even in its diluted form, the free market doctrine has been taking us from one crisis to a bigger one while aggravating inequality to unprecedented levels by making the rich richer and the poor poorer. Adopting free-market “policies” has enriched a small minority of oligarchs and corporate interests at the expense of the majority of wage and salary earners who, thanks to a twisted tax code, bear most of the tax burden.

As far as artificial intelligence is concerned, free marketeers tell us that there is nothing to worry about because AI, like previous technological revolutions, will bring with it a process of “creative destruction”, whereby new jobs are created to replace the jobs (or “tasks”, as they prefer to call them) eliminated by the introduction of AI. The typical argument that makes free marketeers optimistic about the effect of AI on employment is that technology boosts productivity, which in turn creates demand and jobs. Specifically, they claim that AI boosts productivity, enabling companies to lower prices for consumers, pay higher wages, or distribute profits to shareholders. As a result, aggregate demand will be stimulated, boosting job creation. Rising productivity, they claim, produces higher levels of income and spending, consequently boosting demand for goods and services across the economy. They even go as far as claiming that, thanks to AI, all of us will work less and play more.² It follows that there is no need for public policy intervention and that if adjustment is to be made, the burden of adjustment must fall on the individual. This is Alice in Wonderland and Ali Baba and the Forty Thieves—in other words, pure fantasy that deserves inclusion in a volume of One Thousand and One Nights.

It is not as simple and straightforward as free-market enthusiasts make it sound. There is no doubt that AI, and technology in general, boosts productivity, but the effect of productivity growth on employment, wages and prices depends on who benefits from productivity gains, which is determined by the corporate sector (by profit-maximising firms, to be precise). A technology-driven rise in productivity means that the same quantity of output can be produced with smaller quantities of inputs, or that the same quantities of inputs can be used to expand output. In either case, gains will be realised by the producer, but the economy-wide outcome depends on how the realised productivity gains are distributed. Productivity gains can be (i) distributed to workers in the form of higher wages; (ii) used to finance capital investment to expand productive capacity; (iii) channelled to the owners, executives, directors and shareholders in the form of extravagant bonuses and dividends; and (iv) passed on to consumers in the form of lower prices. In a neoliberal world where the economy is dominated by profit-maximising firms operating in an environment with little regulation, productivity gains are more likely to be used to boost profit rather than reduce prices or raise wages. A decision to expand production and employment will be taken by a profit-maximising firm only if it boosts profit. Two prominent economists suggest that “the dominant perspective in most C-suites views labour as a cost to be cut, either to withstand competition or to better remunerate shareholders”.³ In a neoliberal economy, labour is a cost of production rather than a creator of value.

In my book, The Economics of Artificial Intelligence, I report the results of a simulation exercise covering four scenarios for channelling technology-driven productivity gains. Under scenarios [1], [2] and [3], productivity gains are channelled entirely to profits, wages, and prices while maintaining the level of output, which means that a smaller quantity of labour would be used. Under scenario [4], productivity gains are channelled to output by expanding production, using the same quantity of labour, in which case no effect on employment will materialise. The results show that only if a decision is taken to expand output beyond the productivity gains will employment expand, which can only happen if it is profitable to do so. This is very unlikely unless there is a source of extra demand, which is conditional upon the use of productivity gains to raise wages and/or reduce prices.

In a neoliberal world, the most likely scenario is [1]—that productivity gains are channelled entirely to profits. This proposition is supported by the historical record as reflected in US data on productivity, wages, corporate profit, consumer prices, employment, and capital intensity. The chart below displays average annual compound growth rates, over the period 1948-2023, showing clearly that the benefits of productivity growth have not been passed on to workers or consumers, but rather to directors and shareholders, as indicated by rampant growth in corporate profit. By contrast, productivity growth has been more rapid than the growth of employment and wages, implying that productivity gains have not been passed on to workers. Prices have been rising faster than productivity, implying that the gains have not been channelled to consumers. The rapid growth of capital intensity indicates that productivity growth has led to the replacement of labour by capital, thus reducing the potential growth of employment. This is to be expected in any neoliberal economy.

Average Annual Compound Growth Rates in the US (1948-2023)

Those who hold the view that AI will create more jobs than the jobs it destroys emphasise the point that workers should be retrained to assume new tasks and pursue different careers. Two terms that have become common in the literature dealing with this issue are “skills mismatch” and “upskilling.” The Future of Jobs Report 2023 identifies a wide range of jobs that will diminish or disappear and jobs that will emerge as a result of the advent of AI.⁴ By looking at the lists of jobs expected to disappear and emerge, it becomes clear what kind of upskilling is required so that people who lose their jobs can find new employment opportunities. This would be a monumental task because of the enormous difficulty of moving workers from low-skill to high-skill jobs. For example, what does it take to retrain a door-to-door sales person to be an AI and machine learning specialist? How about retraining a bookkeeper to be a fintech engineer? Can anyone seriously imagine retraining a stock-keeping clerk to become a robotics engineer? It is definitely easier to retrain a materials engineer to be a security guard, except that security guards will lose their jobs to machines. How about retraining a cashier to become an electro-technology engineer? Can a postal service clerk be retrained to become a digital transformation specialist? If any of these is possible at all, only very few clerks have the ability required to become robotics engineers, but it is absurd to imagine that all clerks who lose their jobs can be retrained to take any of the new sophisticated jobs. Then there is no guarantee that as many robotics engineers, fintech engineers, and machine leaning specialists will be required as the number of clerks who lose their jobs.    

Those who look at the history of technology to foresee the future effects of AI on employment seem to forget that this time it is different because AI is more disruptive than any other technology. An observer describes the situation as follows:⁵

The adoption of AI in the workplace will likely be different from prior instances of technological disruption. Over the last few decades, technological change has largely impacted routine tasks that were mostly part of middle-wage jobs. In contrast, the adoption of AI is likely to impact a much wider degree of occupations, automating or augmenting nonroutine tasks that had not been impacted by past automation. Nonroutine tasks are mostly found in low-wage jobs, such as janitorial services, home health aides and food services, or in high-wage jobs, such as managerial roles and roles within the knowledge economy, posing a set of issues unlike before.

The impact of AI on employment is different from that of previous technological revolutions, because “machines” are no longer straightforward mechanical tools but have assumed more of a “worker” role, just as people who can learn and think tend to do. A distinguishing feature of AI is that it can be developed for many different types of activities, with the potential to spread rapidly in every sector of the economy and in every aspect of our lives.

A study prepared by the US Council of Economic Advisors suggests that previous technological advances in automation affected “routine” tasks, but AI has the potential to automate “non-routine” tasks, exposing large new swaths of the labour force to potential disruption.⁶ The study refers to “an emerging body of research” suggesting that AI can outperform workers in an increasing set of complex tasks that are typically assigned to educated workers. The view that this time it is different is expressed by an economist as follows:⁷

The “new automation” of the next few decades—with much more advanced robotics and artificial intelligence (AI)—will widen the range of tasks and jobs that machines can perform, and have the potential to cause much more worker displacement and inequality than older generations of automation. This can potentially affect college graduates and professionals much more than in the past. Indeed, the new automation will eliminate millions of jobs for vehicle drivers and retail workers, as well as those for health care workers, lawyers, accountants, finance specialists, and many other professionals.

In its report on the effect of generative AI on jobs, the Global Partnership on Artificial Intelligence expresses the following view:⁸

Unlike previous AI developments which focused on automating narrow tasks, Generative AI models possess the scope, versatility, and economic viability to impact jobs across multiple industries and at varying skill levels. Their ability to produce human-like outputs in areas like language, content creation and customer interaction, combined with rapid advancement and low deployment costs, suggest potential near-term impacts that are much broader and more abrupt than prior waves of AI.

The report goes on to say that generative AI is capable of creating content, solving complex problems, and even mimicking human-like text, which brings about a “different set of challenges and opportunities that extend far beyond traditional notions of automation”. What makes it different this time is that previous technological revolutions involved the replacement of human mechanical skills with tools and machinery. AI tools, on the other hand, are replacing human mental functions, particularly the ability to predict future outcomes and make decisions accordingly. This is something that has never happened before in human history.

Those who undermine the argument that AI will have extremely adverse consequences for employment and wages, thinking that it will still be business as usual, are challenged by those who believe that while the economic impact of AI has so far been “pretty mild”, AI technologies will likely deliver “massive economic shocks” over the next two decades. As an example, one observer illustrates the effect of one AI application, self-driving vehicles, which has the potential to put America, and many other countries, into recession. This is how he explains the situation:⁹ 

5 million Americans currently make their living driving taxis, buses, vans, trucks, and e-hailing vehicles. That’s approximately 3% of the workforce. The US Postal Service is already piloting self-driving trucks, Uber has based its whole future business model and valuation on replacing drivers with self-driving vehicles, and two years ago, GM announced that it plans to deploy a large-scale fleet of driverless taxis in large cities by the end of 2019. So, 5 million American jobs are going to disappear over the next 5-10 years. To put that in perspective, America lost 8.7 million jobs in the great recession of 2007 to 2010. So, just one AI technology has the power to put America into recession all by itself. Imagine what will happen when multiple AI technologies are adopted at the same time.

The problem is not in technology itself but in the system under which technology is developed and deployed. When technology is deployed in a neoliberal economy dominated by profit-maximising firms and oligarchs whose appetite for wealth accumulation is insatiable, it becomes a weapon in the wrong hands. In previous technological revolutions, “creative destruction” worked rather well, in the sense that the new technology created more jobs than the jobs destroyed by the same technology. This time, however, it is different (perhaps involving “destructive destruction”) because AI can be distinguished from previous technologies in terms of its transformative and disruptive power. This may be a pessimistic view that will prove to be wrong as in the past. Perhaps, but it is better to err on the side of caution than the side of complacency. Nothing is “inevitable” about the applications of technology or its consequences. Both involve choices, which imply the availability of alternatives. We could at least try to push policy makers to choose the best available course of action for the benefit of the whole society rather than the wealth and power of the oligarchy. 

It is irresponsible to promote the idea that there is nothing to worry about. It is irresponsible to suggest that the adverse consequences of AI can be met by education, reskilling and the need to remain competitive by removing barriers to the adoption of AI. This is free market rhetoric. It is equally irresponsible for governments to accept this proposition, yield to the wishes of the AI oligarchy, and refrain from thinking about regulation and corrective policies to counter the adverse consequences of AI. On the contrary, public policy must deal with the potential adverse consequences of AI, including regulation and redistributive policies (which may involve the introduction of automation tax, wealth tax, luxury tax, and infrastructure tax). Otherwise, even the oligarchy may live to regret the urge to resist the introduction of corrective policies. This is because AI has the potential to destroy the very fabric of neoliberal capitalism from which the oligarchy has benefited enormously.

Notes
¹ Brook, Y. (2016) Free Market, 13 November. http://serious-science.org/free-market-7407.

² Lund, S. and Manyika, J. (2017) Five Lessons from History on AI Automation and Employment, McKinsey Global Institute, 28 November.

³ Acemoglu, D. and Johnson, S. (2023) Choosing AI’s Impact on the Future of Work, Stanford Social Innovation Review, 25 October.

⁴ World Economic Forum (2023) The Future of Jobs Report 2023. https://www.weforum.org/publications/the-future-of-jobs-report-2023/.

⁵ Khattar, R. (2023) Will AI Benefit or Harm Workers?, 24 August. https://www.americanprogress.org/article/will-ai-benefit-or-harm-workers/.

⁶ White House (2022) The Impact of Artificial Intelligence on the Future of Workforces in the European Union and the United States of America. https://www.whitehouse.gov/wp-content/uploads/2022/12/TTC-EC-CEA-AI-Report-12052022-1.pdf.

⁷ Holzer, H.J. (2022) Understanding the Impact of Automation on Workers, Jobs, and Wages, Brookings, 19 January.

⁸ Global Partnership on Artificial Intelligence (2023) Generative AI, Jobs, and Policy Response, Innovation Workshop, Montreal.

⁹ Smith, D. (2019) Why Artificial Intelligence (AI) Means the End of Capitalism, Linkedin, 23 May.

---