Of all the forces that bear upon the wage, none has frightened the worker longer, or more deeply, than the machine — the suspicion that the contrivance built to do his work will, in the doing of it, take his bread. The fear is as old as the spinning-frame and the men who broke it, and it has returned, faithful as a season, with every new engine since. This plate takes up the question the fear poses, and that the wage can no more escape than the worker could outrun the loom: is the machine the worker's friend, multiplying what his hour is worth, or his rival, driving what his hour is worth toward nothing? Two centuries of asking have returned an answer that satisfies neither camp, because it belongs to both. The machine is reliably each in turn; and which it is, in a given trade in a given year, is not a detail but the whole of the matter.
Begin with a recantation, for the question divided the discipline at its very birth. David Ricardo, who had built the optimistic system of classical economics, held at first the consoling view that machinery injured no one — that what it withdrew from labour in one place it restored in another, as surely as water finds its level. In the third edition of his Principles, in 1821, he took it back. He added a chapter, plainly titled “On Machinery,” in which he confessed himself converted to the opposite opinion: that the substitution of machinery for human labour was “often very injurious” to the interests of the class of labourers, that it could reduce the demand for their work and leave some portion of them out of employment and their condition wretched. It was an extraordinary thing for the architect of the cheerful system to grant, and it set the terms of a quarrel that has not closed in the two hundred years since: whether the machine, when the full accounting is in, displaces the worker, or in the end takes him back.
The optimist's reply, then and ever since, is formidable, and it rests on a weight of evidence the pessimist cannot simply wish away. The machine that turns a worker out of one task, the argument runs, sets in train the very forces that take him back in another. It cheapens the good it makes, so that buyers, with money now left in their pockets, demand more of that good and of others besides; it lifts the productivity and the income of the whole economy, and income spent is income that employs; the capital it saves does not lie idle but is ventured anew; and whole industries undreamt of in the previous age rise up to absorb the displaced. The dread that the machine must breed lasting mass unemployment rests, the optimist says, on an old fallacy — the lump of labour, the unexamined belief that there is but a fixed quantity of work in the world, so that whatever the machine takes up is so much torn from the portion left for human hands. There is no such fixed quantity. The wants of humankind have proved bottomless, and the labour of meeting them has grown, not shrunk, across two centuries of furious mechanisation; employment has climbed beside productivity, decade upon decade, and the recurring prophecies of a workless world have been, every last one of them so far, mistaken. It is no small army that the optimist has at his back.
There is no fixed lump of labour. The wants of humankind have proved bottomless, and the work of meeting them has grown, not shrunk, across two centuries of furious mechanisation.
The modern account, warier than the old fear and the old comfort alike, dissolves the question into its parts, and in doing so reconciles much of the quarrel. A technology that takes over a task, it observes, does three separate things at one stroke. It displaces the labour that did the task — the loss the pessimist fixes upon. It raises productivity, and so, by cheapening the product and swelling incomes, lifts the demand for labour across every task it did not touch — the gain the optimist insists upon. And, in the fullness of time, it tends to create new tasks altogether — new occupations, new kinds of work that had no name before — and it is this third effect that, in the long historical record, has done the most to keep the worker in employment. Daron Acemoglu and Pascual Restrepo, who gave the framework its sharpest modern statement, name the third the reinstatement effect; and upon its strength against the displacement effect, the whole question turns. Where reinstatement and rising productivity keep pace with displacement, the machine lifts the wage and holds labour's place at the table; where displacement outruns them, the machine presses both down. The machine is neither friend nor rival in its nature. It is friend or rival in its balance.
And the balance falls unevenly across the ranks of labour, in a pattern that obliges the fourth plate's account of skill to be redrawn. What the computer automates is not the work of the unskilled as such, but the work that is routine — that can be set down as a rule and a sequence of steps — and the routine is found as readily in the mid-skilled hand at the assembly line as in the mid-skilled clerk at the ledger. What resists the rule, and is therefore complemented rather than replaced, lies at both extremities: the abstract, non-routine work of analysis and judgement and management at the top, and the manual, non-routine work of cleaning and tending and serving at the bottom. The consequence, traced by David Autor and others, is not the simple ascent of the skilled over the unskilled that the fourth plate described, but a polarisation — a swelling of work at the top and the bottom and a hollowing of the middle, the vanishing of precisely those solid mid-skill trades that once bore a worker without a degree into the comfortable middle of society. Even where the machine has left the sum of jobs untouched, it has recast their shape, and it has not done so gently.
Nor is the displacement always made good. When Acemoglu and Restrepo turned from the framework to the fact — counting the industrial robots installed across the local labour markets of America, and following what came after — they found not the optimist's smooth re-absorption but real and durable loss: in the places most exposed to the robot, employment fell and wages fell, and no compensating bloom of new work arrived in time to mend the harm. The robot, they argued, was an instance of a peculiarly unkind species of technology — “so-so” automation, just capable enough to take the worker's task, but not productive enough to set the answering forces strongly in train. Here the pessimist's case is no Luddite superstition but a measured finding: a technology can displace faster than it reinstates, and when it does, the workers it sheds do not dissolve gently into newer and better employment — many simply earn less, and some cease to work at all. The long-run record of the optimist and the local record of loss are both true at once, and the distance between them is counted out in broken trades and emptied towns.
Which carries the oldest fear to its newest shape. Every wave of automation before this one had, on the whole, spared the non-routine work of the mind at the top of the scale — the analysis, the writing, the judgement that the polarisation story had placed safely beyond the machine's reach, and that had been at once labour's refuge and the educated worker's advantage. The artificial intelligence now arriving lays its claim upon exactly that work: upon the drafting and the summarising, the coding and the diagnosing, the composing of the very abstract tasks that had been thought unautomatable. If the claim is made good, the comfortable arrangement of the last half-century — the machine complementing the schooled while it displaced the routine — need not hold, and the question reopens at a flank the workforce is least prepared to defend. And yet the prudent forecaster keeps in view the long graveyard of confident predictions at his back. Every generation has believed its own machine to be the one that would at last render the worker obsolete; and every generation, so far, has been wrong, undone by the same reinstatement and the same bottomless wants that confounded the prophets before it. Whether this time is truly different is, in plain honesty, not yet known. The task framework tells us only what it has always told us — that the answer turns on the balance of displacement against productivity and new work — and that balance, for a technology still in the first of its years, cannot yet be read. The plate records the question, and declines to counterfeit an answer.
Every generation has believed its machine the one that would at last render the worker obsolete; and every generation, so far, has been wrong. Whether this time is different is not yet known.
Return, then, to the question this volume puts to every wage. The machine does not stand outside the contest between worth and power; it is among the most potent forces at work within it, and it presses upon both sides at once. A machine that complements the worker raises the value of his hour — his worth, in the second plate's sense — and, by making him more productive and less easily done without, can strengthen his hand where the wage is settled. A machine that substitutes for him lowers the value of his hour, and, worse, places in the employer's keeping the most effective of all threats — the credible assurance that the work can be had without him, the outside option that, as the eighth plate showed, disciplines the wage even when the machine is never finally bought. The machine, that is, reaches the wage by worth and by power together: it alters what the worker is worth, and it alters how much power he commands to claim it; and what it withdraws from labour's side of the ledger gathers, at the scale of the whole economy, into the falling share the volume's final plate set out to measure. The future of the wage, in an age of quickening machines, will therefore be settled not by the machines alone — not by what they can be made to do — but, as this whole volume has insisted, by the same contest the machines merely re-arm: over who shall capture the gains the machine creates, and who shall bear the losses it deals. That contest the machine sharpens. It does not settle it. And as to how it will come out, the table can say only what it has said of every wage gathered within it: the argument remains open.