By now, the refrain “socialism or barbarism” has been worn down to the nub. The struggle for a more just and democratized economy and polity isn’t over, but for the most part, the ruling class has opted for barbarism.

Of course, demands for something better persist and the movements trying to realize them continue to evolve. The old struggles — like control over capital and workplaces — remain, but new ones have emerged, reshaping the terrain. Artificial intelligence is, right now, the urgent example of this phenomenon — or at least it should be, as we sleepwalk toward an AI oligarchy.

When it comes to AI, there’s good news, bad news, and news still unwritten. The good news is that the variety of technologies that comprise AI could serve pro-worker, pro-human ends. Marxism, after all, features a long tradition of hope that mechanization and, later, automation can liberate workers. The bad news, however, is that it depends on who owns the robots. Right now, that’s neither workers nor the broader public. But it could be, which leads us to the yet-to-be determined news: who controls AI?

Whose AI?

If the predictions of tech and industrial capitalists come true — and AI does indeed remake work and displace workers in the direction of a quasi-post-scarcity world — then democratizing AI will be essential to prevent social, political, and economic collapse. Without such democratization, we risk entrenching a system even more oligarchic than the one that already dominates so much of contemporary life. Distributing both the power and the gains of this technology would necessitate two major transformations: one in the ownership and use of the technology itself and another in the structure of the welfare state, which would become not just as crucial but utterly and completely necessary.

The first questions, then, are: Who should control AI, how, and to what end? Proprietary systems mobilized to not just augment productivity — a use that is currently, at best, a promise with mixed results — but to terminate jobs and replace workers could in theory be a beneficial, but only if the people displaced are guaranteed lives of equal or greater security, dignity, and meaning than those they led under wage labor. Since when jobs are made redundant, they can’t be relocated, the shift must either be to a new, better job, or to a life that is supported without the need for work (e.g., through a robust universal basic income program). Somewhere in the middle is the grey area of productivity gains.

Productivity gains should not be rejected outright, especially if they can be harnessed to reduce drudgery, shorten the workweek, and improve overall quality of life. Yet the gnarly problem of what the AI-work transition would entail remains. But it wouldn’t be the first time a market and its workers had to confront such a sea change — as the Luddites remind us, more or less. Indeed, if the maximal AI predictions are correct, which is a big “if,” then what’s coming could produce an upheaval akin to, or greater than, the Industrial Revolution.

In an economic system marked by socialist economic relations — in which workers own and control their firms, for instance, through co-ops; or indirectly through state-owned enterprises; or through industrial democracy — the community could decide for itself how to use AI, for what purpose, and at what pace. This would be a sort of controlled, planned deployment that would not only ease the transition, but allow the many to decide to what end AI might be used. This ought to be the way.

Enhancement vs. Augmentation

As Evgeny Morozov writes in Le Monde Diplomatique, the idea of using AI for human enhancement gave hope to Warren Brodey, who was an early cybernetician in the 1960s. In “AI and the techno-utopian path not taken,” Morozov picks up on a distinction Brodey made between augmentation and enhancement. Augmentation, Brodey argued, is fleeting and thoroughly bound to technology and devices; enhancement, by contrast, involves building new capacities and is a true servant of humankind. “Enhancement,” he writes,” leverages technology to develop new skills.”

Morozov’s study of Brodey is a parable for the crossroads we now face. “In essence,” he writes, “augmentation deskills us in the name of efficiency, while enhancement upskills us, fostering a richer interaction with the world. This fundamental difference shapes our integration of technology, determining whether we become passive operators or creative artisans.”

If one had to guess, the capitalist class — thoroughly in control of AI development — would favor the path of deskilled labor: easier to control, cheaper to employ, and ultimately more disposable. Meanwhile, enhancements — true skill-building and empowerment — would be reserved for elites. In a worker- or community-controlled paradigm, however, workers could, at scale, adopt a program of enhancement for their own and the common good. We might then collectively produce a polity and economy that serves that common good and common interests.

Even in a world where AI is explicitly designed to serve the common good through human enhancement, disruption would be inevitable. Entire sectors could be rendered obsolete, and countless jobs eliminated. We’d be left to ask what is to be done: How should the gains from increased productivity be distributed? What new forms of socially valuable activity can replace wage labor — and how can they provide not just security, but purpose, fulfillment, and agency? These are questions a future welfare state must answer, regardless of whether AI is controlled by the many or the few and whether it is used to deskill or empower.

Empowerment or Enclosure

As things stand, the tech oligarchy that controls AI believes the welfare state solution to major technological disruption is a kind of pared-down, survival-level universal basic income (UBI), likely one premised on stripping existing social welfare programs to the studs.

This model would leave workers and nonworkers to purchase their social services from the private market. For years, UBI has been touted by proponents as a kind of social welfare panacea — both those who approach it from the libertarian bare-state model and those who believe it would be a utopian project of liberation through expansive supports. The very real risk of adopting UBI has always been that, in all likelihood, it would proceed to be modeled after the libertarian vision or, perhaps just as bad, would be simply underfunded to the point of serving as a net loss for those who would rely on it.

As AI comes to play a bigger role in industry, the Left cannot surrender in advance. While we shouldn’t imagine AI technologies as a cure-all — but nor should we dismiss them or the productivity gains some of them might offer. This moment in our industrial history is a critical juncture, one which presents us with an opportunity to democratize the control and use of AI, and to remake the welfare state in a way commensurate to the change and consistent with the social, political, and cultural goals we collectively choose. We should seize it and make AI work for us, at our command.

---

---