What the Industrial Revolution Can Actually Teach Us About AI

The comparison is not wrong, exactly. That is what I kept returning to on the drive home. It is a comparison that is doing a great deal of useful work, and then, at a certain point, stops working and becomes something else. Becomes comfort, perhaps, dressed as analysis.

Here is what it gets right. Large-scale technological change arrives, disrupts established ways of working, causes real hardship for some, creates new possibilities for others, and eventually, over decades, across generations, settles into a new normal that the people who lived through the transition would not necessarily recognise as normal at all. Luddites were not, as the caricature has it, simply afraid of progress. They were skilled craftsmen watching machinery eliminate the thing that made them valuable, and they were correct that it would. The new jobs that eventually emerged (and they did emerge) were largely not theirs. They belonged to different people, in different places, working under different conditions. The transition was real and uneven and long.

That much carries across. I find it a useful frame because it refuses two kinds of dishonesty at once: the panic that says everything is ending, and the cheerful inevitabilism that says don't worry, it always works out. The Industrial Revolution worked out, in the long view. It did not work out for everyone who lived through it. Both of these things are true simultaneously. For those sitting with the question at a more immediate level, that distinction matters quite a lot.

But here is where my nod started to feel dishonest.

The Industrial Revolution displaced physical labour. You could see it happening. The machines were visible: large, loud, obviously mechanical, present in the landscape in ways that changed the landscape. The people displaced were predominantly those doing manual work, and while the social cost of that was enormous, the cognitive class (the professionals, the managers, the people whose value was located in their minds rather than their hands) largely survived it intact. Architects were not replaced by steam. Solicitors were not replaced by the spinning jenny. The working assumption, held more or less until very recently, was that physical labour was automatable and cognitive labour was not.

That assumption is the thing that does not carry across.

What is happening now is not the automation of physical labour. A solicitor I know spent last month watching an AI tool review a hundred-page contract in four minutes, something that used to be a full day's careful work. She found it impressive and unsettling in roughly equal measure, and said so in a way that reminded me of nothing so much as a craftsman holding a machine tool for the first time and not quite knowing where to look. The class of people whose value has always resided in their minds is now watching what the weavers watched in the 1820s. Which is a rather different historical situation, and one for which the Industrial Revolution offers somewhat less comfort than my colleague's tone at dinner suggested.

I want to be careful here, because I am not saying the outcome will be worse. I genuinely do not know that. What I am saying is that the population most disrupted this time is a different population, and one that has, historically, had considerably more say in how disruptions are managed. That is interesting. It may even be, in some ways, fortunate. People with resources and voice tend to navigate transitions differently from people without them. But it also means the comparisons to the Industrial Revolution are being made, largely, by the people who would have been fine during the Industrial Revolution, and who may be reasoning, unconsciously, from a position of historical safety that does not apply in quite the same way this time.

The other thing the comparison softens is pace. The Industrial Revolution unfolded across a century. Children watched their parents adapt and then adapted differently themselves. There was time (not comfortable time, not painless time, but time) for the adjustment to happen across generations. What is happening now is happening within careers, within decades, within product release cycles. The adjustment window is narrower than the historical comparison implies, and the expectation that one generation of disruption will deposit a new generation of opportunity is harder to hold when the cycle is this compressed.

None of which means my colleague was wrong to find the comparison useful. I still find it useful. History is almost always the right place to look when something feels unprecedented, because it is usually less unprecedented than it feels, and that is genuinely steadying. But the comfort the comparison offers is partial. It is comfort about the shape of the thing, not the specifics. And the specifics, in this case, are different enough to deserve their own thinking.

They are uneven. Some people and sectors absorb the change and others do not, and the distribution is neither fair nor predictable. They take longer than optimists claim at the start. And the new equilibrium, when it eventually arrives, does not look like the old one with different tools. It looks like something the previous world would not quite have recognised.