Seizing the Opportunity

It would be all too easy for the Left to squander this political opportunity. Two tendencies in particular may hold us back from developing an adequate response to the problem of AI-driven job loss.

One is the fragmented resistance to AI empires. Aren’t there a host of progressive nonprofits and academics working on “AI”? There are — but the issues they “work on” are variegated and often siloed, with many of these people covering important topics like surveillance, AI safety, algorithmic bias against marginalized groups, environmental impacts, cultural degradation from slop and platform decay, creativity, existential risk, regulation and oversight, and so on. People who work on AI policy are fighting on multiple fronts, and some are funded and to some extent captured by industry.

The number of people focusing specifically on AI and labor is much smaller. Outside of tech policy, unions have been engaged with the implications of AI, but they are also busy with more immediate struggles over wages and working conditions, organizing nonunion workers, and the like.

The second reason the Left may miss the opportunity to lead on AI worker displacement is the complicated relationship that many leftists have with emerging technology. There is a prevalent tendency to conflate a technology with the capitalist system and the particular matrix of power relations in which it is developing. In this vein, AI is sometimes analyzed as a wholly negative phenomenon in the context of capitalist social relations, a set of technologies deployed by the ruling class in its own interest to degrade and replace human labor.

While there is a movement to shape technology in the public interest, it tends to be sequestered in academic or policy-oriented circles, even though, as Leigh Phillips writes, the Left should be optimistic about using technology for liberation. The techno-pessimism leads to a tendency, at best, to focus on ill-defined notions of “governance” of AI rather than how to harness it, limiting conversations about how AI could democratize computing or open new modes of education.

AI presents a special conundrum because it’s so ill-defined; with no clear definition of what “artificial intelligence” is, it becomes simply a stand-in for the oligarchs, platform capitalism, the surveillance state — just a pile of evil slop to refuse.

In short, current dynamics on the Left and in the AI policy landscape mean we risk being merely reactive to AI job displacement instead of proactively coming up with policy ideas. In a follow-up essay, we will review and propose some political solutions to these problems — from regulating AI as a public utility to a New Deal–style public jobs program. Here we start by assessing the debate on the Left over whether AI-driven worker displacement is even a problem at all.

How Much of a Threat Is It Right Now?

Until now, the future of labor with AI has been a two-sided “debate” presided over mostly by labor economists. One side believes that AI will cause large job losses. The main argument for this position is that it is literally the business proposition of the AI companies — that enterprises will use their products to save on labor.

Banks and consulting firms have been floating large figures: Goldman Sachs said 300 million full-time jobs worldwide, and a quarter of current work, could be entirely performed by AI; McKinsey analysts projected that 30 percent of hours currently worked in the United States could be automated. (These consulting firms are also vulnerable to AI and are racing to create their own agentic AI platforms, where AI “agents” act autonomously to perform specific multistep tasks.)

But AI will also create new jobs, argue labor economists on the other side. Jobs are bundles of tasks, and AI tools are unlikely to replace all these tasks. More than 60 percent of employment in 2018 was in job titles that didn’t exist in 1940, reports one study by MIT economist David Autor and colleagues, with “new work” including new job titles that involve new technologies (drone operators, textile chemists), reflect changing demographics (hypnotherapists, sommeliers), and include gig-work positions (on-demand shoppers and personal drivers). AI will produce new things we haven’t even imagined yet, and it will augment human work, not replace it.

It is true that with technological change, old jobs have been replaced by new types of work. But two points are important when considering whether history is reassuring here. First, there’s not enough data to make claims about how things “always go”: yes, there have been previous transitions from agrarian to manufacturing to service economies, but that’s still a sample size of only two transitions. Second, these earlier transitions should not be cause for reassurance either, because they are still unfolding and their impacts are still reverberating. US electoral politics continues to be shaped by the failure of the state to guide these transitions.

Some economists take a more nuanced position, warning that tech entrepreneurs will claim “innovator rents.” Anton Korinek and Joseph Stiglitz write, “We economists set ourselves too easy a goal if we just say that technological progress can make everybody better off — we also have to say how we can make this happen.” Inequality rises because innovators earn a surplus, and unless markets for innovation are fully contestable, that surplus they earn will be in excess of the costs of innovation, they explain. In addition to this, innovations affect market prices and change the demand for factors like labor and capital. “AI may reduce a wide range of human wages and generate a redistribution to entrepreneurs,” they conclude.

Marxist Perspectives

A more explicitly Marxist analysis of technology is also helpful here. Karl Marx argued that technology is not developed under capitalism to improve society or “lighten the toil” of labor, but rather to produce surplus value or profit for capital. Thus, capital will not deploy technology unless it can perform tasks more cheaply than the cheapest labor available (Marx quipped that capital was happy to use female labor of surplus populations over machines when the cost of such labor is “below all calculation.”)

From this perspective, it should be clear that capital has a powerful interest in automating the high-cost labor of technical and professional work — i.e., the forms of work apparently most vulnerable to AI disruption. That said, the calculation for capital still depends on it accessing AI tools at a lower cost than that labor. Right now, AI firms aim to offer such tools at low prices to get users hooked before raising their costs. Indeed, there are serious questions about the profit or “business model” in broad terms of adequate revenue generation, with some critics predicting a subprime AI crisis that could ripple through the entire tech industry due to companies having built their products on unprofitable models.

The cost of AI to capitalists looking to replace such labor will be important in determining how widespread such automation becomes. Still, given Marx’s logic, you might think the Left would be alarmed about how capitalists will use this new technology to enrich themselves at the expense of workers. What good is our labor power when it is instantly replaceable?

Marx also powerfully argued that under capitalism, the main product of rapid technological change is the production of a “reserve army” of impoverished unemployed “set free” by technology. The poverty and misery suffered by these surplus populations — even if temporary — could also become an explosive political force and a check on the demands and power of the employed workforce.

Marx and Marxists have noted how this has affected various kinds of manual labor since the Industrial Revolution, but the prospect of widespread automation of “mental” or “cognitive” labor could begin a process of “proletarianizing” the “professional-managerial class,” or at least portions of it. Even if these workers will eventually shift into new lines of work, the transition is not always smooth and can be politically volatile (as we’ve seen with deindustrialized Rust Belt areas struck by high levels of unemployment shifting to Donald Trump in large numbers).

In fact, the continued persistence of a “middle class” between labor and capital has long been seen as refuting Marx’s prediction of increasing class polarization between a small set of capitalist owners and an increasingly de-skilled mass of proletarians. Regardless of what Marx did or did not predict, AI could strike right at the heart of a major source of capitalist stability for over a century — relatively stable middle-class workers who enjoy decent wages and some autonomy at work, and who (mostly) see their interests as aligned with capital’s.

Moreover, as Autor has argued, if educated professionals have been able to carve out advantages in the labor market based on their skills, AI-based de-skilling could make these capacities more widely available and, thus, reduce the polarization between such workers and their low-waged counterparts in more precarious service and manual jobs.

And a sudden increase in precarity for large swathes of skilled and educated workers might in fact increase solidarity between such workers and the wider working class. Even if a high salary appears to insulate one from the depredations of capitalism, most professional workers ultimately rely on their wage for survival like all other working-class people. In other words, they are workers and should see themselves as such.

Nothing to Worry About?

But the Left often tends toward skepticism that AI is likely to cause massive job loss — and thus risks missing the opportunity to build this kind of broad worker solidarity. Part of this dismissal stems from an understandable tendency to distrust what sounds like corporate hype. One of the sharpest critics of AI is sociologist Antonio Casilli, whose recently translated book Waiting for Robots: The Hired Hands of Automation points out that

despite the grand vision of big tech companies and startups alike, AI reality is constantly scaling back: users are promised autonomous vehicles, and they get assisted driving; they’re promised decision-making software, and they get a drop-down menu of options; they’re promised a robot doctor, and they get a medical search engine.

Casilli argues that we should focus on digital labor, specifically the work that goes into AI training and data labeling, which illustrates that human workers are actually being replaced by other humans. “Our work isn’t destined for obsolescence; rather, it’s being shifted and hidden, moved out of sight of citizens, analysts, and policymakers, who are all too eager to abide by the platform capitalists’ storytelling,” he writes. (Here his argument is complemented by Madhumita Murgia’s Code Dependent and James Muldoon and colleagues’ Feeding the Machine, which also focus on vulnerable, low-wage digital workers.) In some cases, labor is merely being shifted by these digital platforms in the way Casilli describes. But job loss also happens; it’s not an either-or.

Another serious left critique of the AI job displacement threat comes from Aaron Benanav, whose 2020 book, Automation and the Future of Work, explains that rates of job creation slow as economic growth decelerates, and that this, rather than technology-induced job destruction, is what has depressed the global demand for labor over the past fifty years. The main story, he argues, is economic stagnation due to deindustrialization. In a recent New York Times op-ed, Benanav notes that the productivity gains from generative AI have been limited, that it’s hard to see how it would create sweeping improvements for core services, and that its advancements appear to be already slowing.

While we agree with some of this — economic stagnation needs to be addressed as a broader underlying issue — it would be a mistake to deny AI progress just because capitalists always hype their products, or because they haven’t been able to monetize the achievements yet. Moreover, despite general stagnation (particularly for the working class), capitalist profitability has been substantially restored since the economic crisis of the 1970s, and some of the most profitable companies today are investing heavily in AI.

Peer-reviewed studies are emerging that illustrate AI can outperform humans across many medical tasks, provide effective psychotherapy, and write poems that are more popular than those composed by humans. It is possible that the current AI wave might really be different than previous experience and hype cycles. Moreover, capital’s relentless historical drive to automate all labor — agricultural and manufacturing labor most dramatically — does not suggest “service” and/or “mental” labor will be forever immune.

So when it comes to the question “Is AI job loss a looming catastrophe or a nothingburger?” social media platforms’ tendency to polarize discussions into binary debates is leading us astray. The truth is probably somewhere in the middle — AI disruption won’t destroy the majority of people’s jobs, but it will still be significant — and like with climate change, the middle-of-the-road scenario is still extremely disruptive, especially when combined with other social and ecological trends.

We argue that this is a “right-now” problem. We don’t have robust evidence of massive displacement, but there are plenty of warning signs. Companies like Shopify are sending memos about becoming “AI first” companies where employees will have to justify why head count on projects can’t be replaced by AI, and CEO Marc Benioff of Salesforce — San Francisco’s largest private employer — says that AI now does 30-50 percent of the company’s work. It’s not just Silicon Valley: Ford Motor CEO Jim Farley just declared that “Artificial intelligence is going to replace literally half of all white-collar workers in the U.S.”

There is special concern for entry-level workers. The Financial Times reports that recent graduates account for just 7 percent of hires across the fifteen biggest technology companies, with new recruits down a quarter compared with 2023. Anthropic CEO Dario Amodei made waves predicting that AI could wipe out half of all entry-level white-collar jobs and lead to 10-20 percent unemployment in the next one to five years.

Again, there are also counterarguments. The Economist argues that the jobs-pocalypse is a long way off because the share of white-collar employment has slightly increased, unemployment is low, and wage growth is still strong, suggesting that firms haven’t actually incorporated much AI into workflows yet. It also reports on frustrated CEOs who have spent money on AI without seeing usable results and how hyperscalers like Alphabet and Meta have poured money into the tech without seeing returns. The evidence of disruption happening currently is still spotty.

Yet it is a problem that needs our attention in devising a response now, before the effects fully emerge, for three reasons. One, the capabilities of AI are already able to displace jobs even if technological progress in AI doesn’t proceed at the same rate. If companies manage to incorporate agentic AI into their workflows, dislocation will be even more clear.

Two, popular social media–informed sentiment about AI and jobs might diverge from the empirical reality of “AI job loss” — but still be a potent political force. For example, distinguish SARS-CoV-2, the pathogen, from “COVID,” the divergent social representations of the pandemic that spread online on both the Left and Right. “AI job loss” can be a political football that demands a reaction even before, and independent of, material impacts.

Three, it takes time to build out serious ideas and political power to confront the displacement. Planning has to begin now, before there is a weight of evidence confirming that it is already happening on a large scale.

The Political Traps Ahead

Public concern about AI brings a unique opportunity to reorient our broader politics and even reinvigorate the Left. But there are a number of traps emerging. The first is the risk that we wave off the threat of AI-driven job displacement as a scam and miss the opportunity to lead on solutions.

The second is that we allow the Right to use AI job displacement to exacerbate class tensions in ways that play to Trump’s base and further weaken public institutions. Treasury Secretary Scott Bessent’s quip that fired federal workers could supply “the labor we need for new manufacturing,” plus the attacks on universities, have inspired discussions of “MAGA Maoism” — a movement that glorifies economic sacrifice, strongman leadership, centralized economic power, and nostalgic visions of industrial production. We can imagine how displaced knowledge workers might be jeered at for even having gone to college and wasting their money and time — the rhetoric will be against “bailing out” those workers for bad choices they made.

This idea that AI job displacement is just white-collar folly is a misconception. AI is also poised to displace many lower-wage jobs in a variety of fields, including for non-college-educated workers. And if higher-earning jobs are displaced, the effects will ripple out through the economy and impact service workers as well. But it is critical to think about how a proposal that includes public support for white-collar work will read to people whose communities have been decimated by outsourcing and automation over the past few decades, for whom politicians have failed to do much of anything. Any public jobs discussions need to incorporate that history and ensure the design and messaging advances solidarity among all types of workers.

The third trap is that the Right will claim leadership on being “tough on AI,” and channel populist resentment into policies and rhetoric focusing on its social dimensions while ignoring the economic dimensions. The social dimensions are where bipartisan policy collaboration is more likely, and they’re important.

Yet the attention space will then be taken up by discourse about protecting children from deepfakes or anxiety about romances with AI “partners” replacing sexual encounters and driving down birth rates, at the expense of debates about power structures or economics. When Vance says that “the main concern that I have with AI is not of the obsolescence, it’s not people losing jobs en masse” and that he’s worried about “millions of American teenagers talking to chatbots who don’t have their best interests at heart,” that’s going to be the template for right-wing discussion of the topic.

But if we can avoid these traps — and develop a register and strategy for talking about the issue of AI-driven job displacement with bold and rigorous analysis of some of the topics raised above — there’s still a chance to take hold of an unpredictable and disruptive moment.

It can seem overwhelming. In her new book, Empire of AI, Karen Hao describes OpenAI and other power players as empires: during colonialism, empires seized and extracted resources, exploited subjugated labor, and projected racist and dehumanizing ideas of their own superiority and modernity to justify the exploitation and the imposition of their world order. The metaphor resonates. But Hao maintains that at this pivotal moment, it’s still possible to “wrest back control of this technology’s future.”

Doing so means steering clear of the Left’s tendency to organize and think around single “issues” or “movements.” AI could fundamentally reshape the relations between labor and capital, and how we live, work, and think. This fight could shape the terrain of capitalism for decades to come. It will need socialists and labor unionists as much as economists, tech visionaries, or computer science experts. Without a Left thinking seriously about actively shaping the future of AI, we will be forced to merely react to a dark future made by tech bros.