Could AI create a 'permanent underclass'?

[Liam97]

The Financial Times has taken up one of the more contentious debates currently circulating in and around Silicon Valley, whether AI could eventually create a genuine permanent underclass, a term thats been gaining real traction after a New York Times opinion piece and a widely discussed Ezra Klein podcast conversation put the concept back into mainstream circulation

The core theory, as its been articulated across several recent pieces, goes something like this, if AI systems become powerful and capable enough to generate enormous wealth for the people who own and build them, the gains could concentrate so heavily among a narrow group of AI company owners, investors and top technical talent that everyone else gets genuinely locked out of meaningful economic participation, left dependent on some form of welfare or basic income rather than having any real path to building wealth or economic agency of their own, its a framing thats been connected to Anthropic and its peers loading up aggressively on top technical talent right now, a rush thats been interpreted by some observers as those insiders positioning themselves to capture outsized rewards before that door potentially closes

The pushback to this theory has been genuinely substantial too though, critics argue the permanent underclass framing is impossibly pessimistic about human nature and initiative, the logic being that if AI tools are powerful enough to create that much wealth for their owners, those same tools should also be powerful enough for regular individuals with good ideas to put to their own use, most economists and AI researchers surveyed on this question reportedly do not expect this specific doomsday scenario to actually play out, though critics on that side of the debate raise a genuinely different concern, that panic over the theoretical risk of a permanent underclass could itself lead policymakers toward poorly designed interventions, like a hastily constructed AI funded universal basic income, that inadvertently recreate the very dynamics of dependency and stagnation that the term underclass has historically described in welfare policy debates going back decades

Whats genuinely notable about this debate reaching the FT specifically is that it signals the conversation has moved well beyond just tech blogs and opinion columns into mainstream financial and economic journalism, the FTs traditional readership of investors, business leaders and policymakers is exactly the audience that would need to take this seriously if any of the proposed interventions, from AI taxation to guaranteed income schemes to new labor market protections, were ever going to gain real political traction

This whole discussion also connects to a much older and genuinely unresolved debate in labor economics about technological unemployment, past waves of automation destroyed old jobs but consistently created new ones over time, the open question that divides economists on AI specifically is whether this technology is different enough, improving fast enough and broad enough in its capabilities, that the usual pattern of job destruction followed by new job creation simply wont hold this time around, thats ultimately an empirical question that wont be settled by theory alone, and this remains a genuinely open one rather than something either side has convincingly settled

Ddf44Cf7 0Ab5 4E7E 9B1F E5E8E34181E6www.ft.com

[RedKnight]

The point about panic over the theoretical risk leading to worse policy outcomes than the risk itself is genuinely underrated in most of these discussions, badly designed AI funded welfare could absolutely recreate dependency dynamics even if the pure permanent underclass scenario never fully materializes

[TinyGriffin]

If AI tools are powerful enough to create that much wealth then they should also be powerful enough for regular individuals to use themselves is a compelling counterargument in theory, but it assumes equal access to capital, infrastructure and technical knowledge, which historically hasnt been how transformative technologies actually diffuse through society

[Alisson]

This debate reaching the FT specifically does feel like a genuine inflection point for how seriously mainstream finance and policy circles are taking AI labor disruption, thats a very different audience than the usual tech Twitter or Substack crowd having this conversation

[Reward Annie]

The historical parallel to welfare reform debates and the culture of poverty framing from decades ago is a smart connection that a lot of tech focused AI commentary misses entirely, this isnt actually a new problem, its an old policy debate wearing new AI clothes

[Dataset Cheetah]

Most economists surveyed not expecting the full doomsday underclass scenario is worth taking seriously, but its also worth remembering economists have a pretty mixed track record forecasting genuinely novel technological disruptions, past performance on job creation doesnt guarantee this cycle behaves the same way

[Slate Mike]

Whether this technology is different enough that the usual job destruction and creation cycle doesnt hold is honestly the single most important open question in this entire debate, and I dont think anyone, including the experts, actually knows the answer with any real confidence yet

[Piston]

I think the healthiest way to read this debate is as two camps disagreeing about magnitude and timeline rather than direction, almost everyone agrees AI will cause genuine labor disruption, the disagreement is over how severe, how concentrated and how permanent that disruption actually ends up being

[TheLegendKyle21]

The framing tension between capital owners capturing outsized gains versus AI genuinely democratizing opportunity for anyone with ideas and initiative is the same tension thats defined every major technological shift throughout history, the internet, mobile computing, all of it, this isnt uniquely an AI problem even if the scale feels different

[Wandering Matt]

Anthropic and its peers loading up on top talent being interpreted as those insiders positioning to capture the spoils before others get locked out is a cynical read, though not an unreasonable one given how concentrated AI company equity and compensation already looks