AI is understood to be an unstoppable force, but it is still wholly dependent on human labor to function. Whether these technologies liberate or create misery will depend on who controls their development and deployment.
Talk of the artificial intelligence revolution tends to skip over workers unless it’s a warning that the new technologies will disempower and displace human labor. That the klaxons should be sounding louder and louder is a given — but they are the sort that encourages a collective shrug. After all, what can you do? It’s the great sweep of technology, of history, of progress. History will sort it out. It always does.
History, in fact, does sort it out. But never in the way promised. Technological revolutions of the past did not eliminate labor, but they very much transformed it. This process of transformation does generate new work — as tech advocates often claim — but it is work that is more fragmented, more surveilled, and more alienated. Early mechanization de-skilled craftspeople, concentrating control over production even as employment expanded. Sure, from clerical automation to platform logistics, technological innovation does create “new jobs,” but worker autonomy takes hit after hit as labor is reorganized around ever more tightly managed systems.
The fatalism surrounding AI contributes to a shared complicity in the potential social, political, and economic collapse that threatens to accompany mass unemployment brought about by AI adoption. However, there’s a further dimension to the labor-technology dynamic that’s ignored or downplayed in assessments of these great leaps forward: the role workers play in training and otherwise enabling these new technologies.
This is not simply a matter of labor being replaced by machines, but of labor being reorganized by capital. It is the threat of skills and knowledge being extracted and formalized by new tech and productive systems that are designed to function without the workers who made them possible.
Training the Tools to Bury Old Jobs
You could call it digging your own grave, although the reality is more complicated than that. A new short film, Their Eyes, by Nicolas Gourault, documents the work of overseas laborers who train driverless cars by analyzing and annotating the images the vehicles rely on to learn how to navigate the road. Gourault says he hopes his work “illuminates a part of the extractive reality that many of our current AI systems rely on.” He succeeds.
If workers are, in some sense, digging their own graves by enabling AI technologies that will eventually render themselves and other laborers obsolete and powerless, they’re doing it at the barrel of a gun. Technology companies seek low-paid labor to train their machines, and the workers who take these jobs have less room to worry about the long-run effects of their work than, say, making a living and surviving today. They aren’t exactly calling the shots.
It’s a familiar dynamic, both historically and in the current AI moment. In the fall of 2020, reports of robot workers captured public attention when a machine appeared to stock shelves in a Japanese shop. In reality, it was operated remotely by a nearby worker through a video camera. At the time, both literal and figurative forms of remote work were not just appealing, but necessary — a matter of life and death. By the summer of 2025, a new robot, Neo, among others, was touted as the future of domestic work.
An AI robot could fold your laundry, water your plants, load your dishwasher, and otherwise manage the mundane tasks that we’ve been promised for decades would someday be the purview of automated technology. But as it turns out, Neo and most of its counterparts are controlled by humans with cameras — at least for now. Not only, then, are the automated robots an illusion (again, at least for now), but they’re a window into your home, with their operators able to see what the robot “sees.” At $20,000, it’s an expensive window into your private space, but the true costs are much higher than the price tag.
Hal’s Human Minders
As Alex Hanna and Emily M. Bender put it, “most AI tools require a huge amount of hidden labor to make them work at all.” Labor creates, analyzes, and sorts data for machines; in some cases, it operates them too, deploying the dexterity and judgment that is natural and easy for humans but exceedingly difficult for robots.
At some point, no doubt, the plan is to wean off human labor and increase the share of AI and robotic control. Whether the tipping point arrives in five years or twenty is uncertain. What is not in doubt is that big capital is all-in on the play, with trillions invested, and more piling up each year.
In the short term, laborers are indispensable to developing AI and making it run. In the long run, however, their work risks becoming not merely precarious, but utterly redundant. The same labor that generates enormous value for tech investors and shareholders may leave workers themselves on the outside looking in. But they won’t be alone: millions of jobs worldwide are at risk from AI.
Some believe that the AI bubble will pop. But even if it does, that won’t forestall or eliminate the long-term technological and economic changes that will accompany AI development. Some AI technologies are here to stay, and they will be permanently folded into blue- and white-collar job flows and systems, reshaping job categories and rendering swaths of workers redundant. Transcription work, for instance, is particularly vulnerable. It’s difficult to imagine a return to transcription pools — and their associated costs — when cheap services can deliver usable copy quickly and with reasonable accuracy. Self-driving cars and robot domestic workers may be a riskier bet, but that’s not going to stop Silicon Valley from trying.
Workers Should Run the Shops
At this point, labor’s best bet is to pick battles very carefully and organize strategically in key industries, prepared to go to the mat to protect jobs and workers in those spaces at scale — for instance, in manufacturing and the automotive industry. But organized labor can only go so far on its own.
Worker control of enterprise, authority of workplaces, and production, ought to be nonnegotiable. Entrenching worker power within industry itself, on shop floors and in office spaces, gives labor a direct control that can’t be achieved through, say, collective bargaining or strike action alone.
Worker ownership and control of industry isn’t easily achieved. Models vary, and no single approach fits every industry. Interests vary across and even within sectors, and solidarity across geographical borders, foreign and domestic, can be tricky to maintain. But we are in an all-hands-on-deck moment.
AI development’s large-scale reliance on workers at home and abroad lays bare the power technology companies have over laborers and their future. If companies can use workers to render those very workers powerless and obsolete in real time, with full knowledge of the long-term individual and collective effects of their strategy, then they can get away with just about anything they please.
Proving otherwise requires more than recognition of the critical role labor plays in developing and deploying AI. Technology is not simply some exogenous force, impacting society from the beyond. It is a social relation, shaped by ownership, control, and those whose interests it serves. There is no question that AI will transform labor — the question is to what extent workers will have a say in how that reorganization unfolds. We need to construct economic and political arrangements that permit the many to shape technologies and their use in a way that works for everyone, not just for tech executives and their investors.
Great Job David Moscrop & the Team @ Jacobin Source link for sharing this story.
