AI and the end of work

Dario Amodei, the chief executive of Anthropic, has offered one of the more arresting forecasts of the AI age. If recent improvements in AI continue, he wrote in January, “it cannot possibly be more than a few years before AI is better than humans at essentially everything.” Could we be approaching a world that John Maynard Keynes imagined in his 1930 essay, “Economic Possibilities for our Grandchildren”, in which he speculated that technology would solve the “economic problem” and leave his descendants working a 15-hour week?

If the past is a guide the answer is ‘no.’ History shows that people, society and institutions have a remarkable ability to absorb and exploit new technologies without abolishing work. Over time technological change has been associated with growing, not falling, employment.

First, and most importantly, human desires are not fixed. A static view imagines a list of progressively more difficult tasks that machines conquer until nothing remains for humans to do. That has not been the experience. Mass mechanisation destroyed jobs on a vast scale, collapsing the number of people needed to produce the necessities of food and shelter. Instead of universal idleness, rising incomes created new jobs elsewhere, providing everything from health care to holidays. When one set of wants is satisfied, new wants appear. In the modern world the “economic problem” is increasingly not one of material scarcity, it is the open-ended nature of human aspiration.

Second, those aspirations are unlikely ever to be met fully by machines. Humans value, and want to be with, others. Even if AI systems become technically superior, it does not follow that people will prefer machine-only provision. The doctor, lawyer or waiter does more than process and act on information. They judge mood, interpret, persuade and encourage. Mr Amodei himself has noted that although AI has become highly capable at reading scans, radiologists have not disappeared because part of the role involves explaining the results to patients and planning future care.

Third, technology, by raising productivity and cutting prices, bolsters demand. If AI reduces the cost of, say, making cat videos for social media, we consume more of them. Cheaper computing did not eliminate work in the tech sector; it helped create whole new industries. The same holds in travel, communications, healthcare and finance. If AI makes a vast array of services cheaper, we will consume more of them and, if history is any guide, will create new complementary jobs, many in roles that are as yet unimagined.

Fourth, technology often falls short of its theoretical potential. Sometimes the frictions and hassles associated with using tech are just too great. Often the tech underperforms humans. The QR code menu is one example. During the pandemic, many restaurants switched to using QR codes to reduce human interaction and the risk of infection. The technology works, it is cheap and it reduces labour. Yet five years later the human waiter rules the roost in most restaurants because ordering food is also a social process and part of the experience of eating out. Or consider the role of the barista. Machines make excellent coffee. Yet customers continue to pay for staffed cafés because they prefer dealing with a human being and like the ambience and reassurance that comes from having staff around. Businesses do not just maximise efficiency to minimise cost. They optimise across an array of factors. AI will face the same frictions, especially in sectors where errors are costly or where consumers prefer dealing with humans.

This is not to be complacent about the effect AI is likely to have on the labour market. Disruption is already happening, especially in some entry-level jobs, including software engineering. Tasks will be automated; jobs will be lost and many more changed. The adjustment could come quickly, especially for structured, more routine cognitive work. But the idea that better machines mean the end of human economic usefulness is a stretch. AI will change the nature of work. It is much less clear that it will end the human desire to work for, and with, others.