In 1930 I wrote a short essay, Economic Possibilities for our Grandchildren, in which I predicted that within a century the standard of living in prosperous countries might rise fourfold or more, and that mankind's oldest problem — scarcity — might at last be solved. The anxiety I hear today about automation is, at its core, the same anxiety I was trying to dissolve then: the fear that machines stealing work is a catastrophe, rather than a liberation waiting to be managed. I was optimistic. I remain so. But I was also clear about the danger, and the danger was never the technology.

The danger is what happens to aggregate demand in the transition. When a factory automates and dismisses five hundred workers, those five hundred people stop spending. Their landlords, grocers, and local pubs feel the contraction immediately. If this happens across an entire economy simultaneously — and technological waves tend to be broad — then the fall in consumer spending is large enough to depress investment further still, because entrepreneurs do not build new capacity into a market that is shrinking. The economist who reassures you that displaced workers will simply find new jobs has confused a long-run possibility with a short-run mechanism. The long run is real; the transition is also real, and people live in it.

The correct response is not to slow the machines. Productivity gains are how civilisations escape poverty; impeding them to protect any particular category of job is, in the end, a tax on the future paid by the poorest. The correct response is for the state to act as the spender of last resort during the transition — investing in the public goods that the private sector, made cautious by uncertainty, will not: retraining infrastructure, broadband, clean energy, care work, the rehabilitation of neglected towns. These are not charity. They are the demand that keeps the economy moving while private investment finds its new footing.

There is also a distributional question that no amount of aggregate growth resolves automatically. If the gains from automation flow overwhelmingly to the owners of capital and the designers of software, while the costs fall on workers with fewer credentials and fewer options, then we have achieved a productivity miracle and a political catastrophe simultaneously. I saw something like this in the interwar years: the middle class hollowed out, resentment curdling into something uglier. The lesson is not complicated — share the gains, visibly and early, before the politics sours.

Automation, properly governed, is the route to the shorter working week, the richer public life, and the freedom from drudgery that I dared to imagine nearly a century ago. Ungoverned, it is simply the next occasion on which we will rediscover that markets left alone do not automatically produce full employment. The choice between those two futures is not technological. It is political.