AI job worries, what the data say - The Monday Briefing

AI job worries

Last week’s Bletchley summit on artificial intelligence (AI) concluded with a prediction from Elon Musk that AI will eventually replace all forms of human labour. As Mr Musk put it, “You can have a job if you want a job… but AI will be able to do everything”.

Concerns about new technologies have grown in recent years. Two of the leading experts in this field, MIT professors Erik Brynjolfsson and Andrew McAfee, have written that “in more and more domains, the most cost-effective source of ‘labour’ is becoming intelligent and flexible machines as opposed to low-wage humans in other countries.” In a similar vein Lawrence Summers, a former US treasury secretary and Harvard economist, has written that with regard to automation destroying jobs, “This isn’t some hypothetical future possibility. This is something that’s emerging before us right now.”

The logic seems obvious. When machines can outperform humans, and at lower cost, humans become redundant. Except, in aggregate, they don’t. More precisely, that was not the experience of the last century.

A period of unprecedented technological change was accompanied by a vast expansion of the number of people in work. In 1900, about 30m people worked in the US. By the year 2000, the total was more than six times higher, at 134m.

Technology created far more jobs than it destroyed. This was partly about new industries, such as autos, aerospace and technology, requiring new workers. More subtly, technology-driven productivity growth raised US incomes and boosted demand for labour-intensive services such as retail, healthcare and entertainment. The numbers are remarkable. David Autor of MIT estimates that 60% of US workers are employed in occupations, such as software engineers, personal trainers and recreational therapists, that did not exist in 1940. (A recent assessment of the susceptibility of 702 US job categories to automation ranked recreational therapy – a role that involves using leisure activities, such as art or games, to improve physical and mental welfare – as hardest to automate.)

One might object that the world has changed a lot since the turn of the century. Technology has become even more pervasive and capable in the last 20 years, a period that has brought us smartphones, social media, cloud computing and more capable AI. What do the data show about the effect of these, and other technologies, on jobs?

Last year the US Bureau of Labor Statistics (BLS) examined this question, looking at how employment in occupations that economists and technologists identify as being susceptible to job losses from AI and robotics had fared. Jobs included truck drivers, fast food workers and interpreters, roles that were largely unaffected by earlier technologies and whose number therefore offers a test of the effect of recent developments in AI and robotics.

Between 2008 and 2019, the number of people working in such supposedly vulnerable jobs increased from 13.3m to 15.1m. Remarkably, employment increased faster for these roles than it did for other, less ‘at-risk’ jobs. What particularly surprised me was that in the era of free and ubiquitous computer-based translation the number of interpreters and translators rose by 22% between 2008 and 2019.

As the paper notes, “Part of the reason new technology does not produce sharper changes in employment is the diversity of jobs within occupations and the diversity of tasks within jobs, not all of which are equally susceptible to technological substitution. Automation of some tasks may also alter the task composition of jobs, rather than simply reducing the number of jobs”.

Jobs in radiology are seen by technologists as being particularly vulnerable to AI. Professor Geoffrey Hinton, one of the leading figures in the development of AI, said in 2016, “We should stop training radiologists now. It’s just completely obvious that within five years, deep learning is going to do better than radiologists.” Yet the number of radiologists has risen, not gone down. Rising demand for healthcare is one factor; the work of a radiologist also involves a wide variety of complex diagnostic and related tasks that cannot be performed by AI tools.

The BLS also finds that an array of new technologies, from robot lawn mowers to robot surgery, have failed to put a measurable dent in employment in the affected sectors in the US.

To test whether the findings in the BLS paper were a product of US conditions we applied the same approach in the UK and produced similar results. Between 2008 and 2018, a range of occupations widely deemed to be at risk of replacement by new technologies witnessed strong job growth. The number of software professionals increased by 84%, radiographers by 43%, accountants 32%, gardeners 29% and legal professionals 28%. These increases compare to overall UK employment growth of 17% over this period.

None of this is to argue that technology does not displace some jobs. The UK data, for instance, show that the number of telephone salespeople declined by 73% in the period 2008-18. There were significant declines too in the number of travel agents, bookkeepers, wage clerks and payroll managers.

But such job losses were eclipsed by gains in employment in other sectors, some of them relatively new job categories, such as fitness instructor, where employment more than doubled.

Mr Musk may be right. Perhaps AI will eventually take every job. Perhaps the previous relationship between jobs and technology will flip. It just hasn’t so far.

For the latest charts and data on health and economics, visit our Economics Monitor:

https://www2.deloitte.com/uk/en/pages/finance/articles/covid-19-economics-monitor.html