Few topics are generating as much attention and hyperbole as the impact that artificial intelligence (AI) will have on society, especially on work and business practices. In particular, many prominent business executives and management consultants have argued that widespread applications of AI will dramatically reduce employment opportunities for future generations. For example, Knoess, Harbour, and Scemama (2016) estimate that robotization, digitization, digital self-services, distributed digital advice and sales, and robo-advisors, all applications that will be driven by AI, could result in a 60 to 70 percent reduction in the workforce of service providers from financial services to telecommunications. They note that while these changes will not happen overnight, the pace of change might be faster than many expect. Even more dramatic is Elon Musk’s claim that artificial intelligence will cause massive job disruption and that robots will be able to do everything better than humans (Clifford, 2017a). The controversial CEO of Tesla further argues that it’s a virtual inevitability that, as robots replace more and more jobs, the United States will have to implement a program of cash payments to everyone (Clifford, 2017b). One can assume that Musk’s argument for cash payments, or guaranteed incomes, applies to other countries as well.
To be sure, many others disagree with the view that the widespread adoption of AI will lead to massive unemployment that, in turn, will necessitate major new taxpayer-funded income support programs. Roe (2018) references a number of senior executives in corporate strategy and technology management positions who argue that AI primarily changes the nature of work rather than causing widespread unemployment. Furthermore, while some jobs will certainly be lost as AI takes on skills formerly attributed to humans, new jobs will emerge. Warren Buffet, the CEO of Berkshire Hathaway and an immensely successful investor, cautions that the trend towards automation replacing low-skilled labour is not new. He notes that in 1800, 80 percent of people were employed on farms. Two hundred years later, it was 2 to 3 percent. Productivity improvements meant that fewer people were needed to work on farms, and therefore were free to pursue other vocations (Clifford, 2017b).
The research of economic historians largely supports the claims that in the past, automation has primarily led to changes in the mix of occupations rather than to mass unemployment; it has also led to increases in real income levels rather than an expansion of poverty. Nevertheless, a number of prominent technology experts argue that the AI experience will be different from the historical experience with automation. For example, Parada (2017) contends that it is catastrophically wrong to draw an analogy between AI and previous waves of automation. Specifically, he argues that automation in the past was primarily about mechanical power replacing human muscle. The AI revolution will be nothing like that. When robots become as smart and capable as human beings, there will be nothing left for people to do because machines will be both stronger and smarter than humans. Parada (2017) makes what seems to be the extreme claim that intelligent robots will be cheaper, faster, and more reliable than humans, and that no capitalist in her right mind will continue to employ humans. He goes on to assert that unless society figures out how to distribute the fruits of robot labour fairly, it will be an era of mass joblessness and mass poverty. Frey (2019) offers a more modest but still startling estimate that 47 percent of US jobs could be automated due to AI.
Others have made a similar, if less extreme, argument that AI represents a “different” technological innovation, and that the labour market experiences associated with other major epochs of technological change may not apply in the case of AI. Microsoft founder Bill Gates does not believe that AI will prove to be a bad thing for society. However, he is in the camp that believes that job displacement will be sufficiently widespread that government will need to direct financial assistance to the many workers who will be hurt by AI adoption with some of that assistance directed at re-education and income support programs (Clifford, 2017b).
The purpose of this essay is to assess the argument that AI’s effects on employment will be significantly different from previous episodes of automation. Specifically, we review some expert opinion on how producers are likely to use AI, as well as the limited available evidence on how AI has affected employment up until now. We also consider whether there are reasons to believe that AI will be adopted at a much faster rate than other major innovations so that retraining workers for occupations that are complementary to AI is impractical. Our broad conclusion is that future employment effects and adoption rates of AI are unlikely to be much different from the broad historical experiences of other General Purpose Technologies (GPTs).
Fraser technology-automation-and-employment
[To read the original column, click here.]
Copyright © 2019 Fraser Institute. All rights reserved.