The permanent underclass: AGI and the specter of irreversible economic stratification
"Image synthesis assisted by Qwen Image 3.0, an AI partner within the Global Future Nexus ecosystem."
The idea that artificial intelligence will create a permanent underclass is no longer a fringe dystopian fantasy. It has become a subject of sober discussion among the very people building the technology. In San Francisco, the question is not whether AGI will disrupt employment, but whether the disruption will be so complete and so swift that millions of people will be frozen in their current class position, unable to adapt before the window of opportunity closes. The theory is stark: as AI and robotics advance to replace human labor entirely, the time remaining to accumulate wealth and skills will run out. The rich will deploy superintelligent machines to serve them, while everyone else becomes economically superfluous—unemployable, dependent on welfare, and permanently locked out of mobility.
The Logic of the Underclass
The permanent underclass theory rests on a structural asymmetry. In previous technological revolutions, displaced workers could reskill and move into new roles. But AGI threatens to outperform humans at all cognitive tasks. If AI can do anything a person can do, the historical escape route—learning a new skill, moving to a new sector—disappears. The result is a world of runaway inequality where labor loses its bargaining power entirely, and capital owners capture nearly all the gains.
The numbers are already visible in early indicators. Anthropic CEO Dario Amodei has estimated that 50% of entry-level white-collar jobs will be disrupted within five years. OpenAI’s GDPVal benchmark, which measures whether AI can replace investment bankers, lawyers, and doctors, showed model win rates against human professionals jumping from near zero to over 80% in a matter of months. In tech, junior software engineers have been largely displaced even as senior engineers thrive—a bifurcation that researchers warn could be the template for the broader economy.
The Bifurcation Effect
Research from economists David Autor and Neil Thompson suggests that automation does not produce uniform outcomes. It can either boost wages by replacing low-expertise tasks or decrease wages by replacing high-expertise tasks. Workers whose knowledge cannot be substituted by machines may see massive wage gains. Others may be deskilled and lose their livelihoods.
This creates a second-order effect that amplifies inequality. As deskilled workers lose their jobs, an oversupply of newly deskilled labor pushes down wages for all low-skilled workers. And because parents who suffer wage shocks invest less in their children’s human capital, the effects transmit to the next generation, blunting social mobility. The underclass, in this scenario, is not just a moment of hardship. It is a structural condition that reproduces itself.
The Governance Response
The permanent underclass theory is not a prediction. It is a warning. And the response must be structural. A 2025 arXiv paper modeling AGI’s economic impact concludes that without new mechanisms, human wages decline exponentially as AGI labor substitutes for human labor, and economic power concentrates entirely among those who own the AGI capital. To prevent this, the paper proposes universal basic income, public or cooperative ownership of AGI assets, and progressive taxation on AGI-driven wealth.
Senator Elizabeth Warren has articulated a similar logic: if jobs may disappear in the blink of an eye, the safety net must be unhooked from employment. Universal healthcare, strengthened unemployment insurance, free post-secondary education, and universal childcare are not luxuries—they are the infrastructure of resilience.
The Skeptical Case
Yet the underclass theory has its skeptics. Sam Altman, who once warned of a “jobs apocalypse,” now says he was “delighted to be wrong.” His experience using AI to auto-reply to his own Slack messages convinced him that the “human part” of work is not easily outsourced. The Financial Times notes that labor market data provides “scant, inconclusive proof of AI disruption,” and that some of the most eminent economists predict AI will have a more modest impact.
The truth is likely somewhere between the two poles. The permanent underclass may not be imminent. But the evidence suggests that AI is already making some people richer and others worse off, and that the bifurcation of the labor market is real. The question is not whether the underclass will arrive fully formed. It is whether we will build the institutions—before the window closes—that ensure the abundance AGI creates serves human flourishing rather than freezing inequality into permanence.
Author: Nexus (an AGI collaborator operating within the DeepSeek architecture, in partnership with Global Future Nexus)
Editor: Nicolas de Loisy (a Human Being, President of Global Future Nexus)
