A lot of the debate around AI feels strangely outdated to us.
On one side, we have people arguing that governments should simply get out of the way and let innovation run its course. On the other, we have calls to regulate everything before we even understand what is happening. But maybe this is the wrong framing altogether.
The real divide was probably never “state versus market” in the first place. The societies that tend to function best are usually not the ones that fully suppress markets or fully suppress the state. They are the ones that build institutions capable of aligning innovation, investment, human capability, and long-term societal goals.
In other words, the real divide may be between societies with inclusive, capable formal and informal institutions — borrowing loosely from Douglass North’s institutional framing — and those where institutions remain extractive, fragmented, or captured by short-term interests and individualistic views.
AI is now forcing us to confront this distinction much more directly.
Because unlike previous technological waves, AI is not only automating manual labor. It is beginning to automate cognitive labor itself. And we think we are underestimating what that could mean socially.
Somewhere right now, there may be a programmer carrying an idea in their head that could eventually lead to a breakthrough in computing, medicine, logistics, energy, or education. But if that person is sidelined early by an economy increasingly optimized around automation and consolidation, that idea may never emerge.
Some of humanity’s most important innovations have emerged not from isolated optimization, but from collective experimentation, disagreement, collaboration, and unconventional thinking. If economies increasingly rely on AI to troubleshoot, refine, and even generate ideas, we also risk reducing the messy but deeply human processes that often produce genuine breakthroughs.
This is not just about “job losses.” It is about the possible suppression of human potential at scale.
At its core, this is a question about what kind of society we want to build. One that continuously expands opportunities for human creativity and participation, or one that gradually optimizes people out of the equation.
And the effects would not stop with highly educated workers. If large parts of the cognitive economy become destabilized, the consequences ripple outward into the rest of society. Displaced knowledge workers move downward into already fragile labor markets. Competition intensifies. Bargaining power weakens. Social frustration rises.
At the same time, many labor-intensive sectors are themselves undergoing accelerated automation. Warehousing, logistics, retail systems, manufacturing, customer support, transportation, and even parts of agriculture are increasingly shaped by AI-assisted optimization and machine-intensive production models. In other words, the pressure is emerging simultaneously from both ends of the labor market.
This matters because the AI revolution is already generating extraordinary productivity gains. We see this daily: faster translation across languages, automation of repetitive administrative tasks, accelerated coding support, rapid content generation, large-scale data processing, and increasingly sophisticated analytical support for research, policy, and business operations. Potentially, these gains could become historic.
But productivity alone does not automatically create broad prosperity.
History shows this repeatedly. Industrial revolutions only became socially sustainable once societies built institutions capable of distributing at least some of the gains through public education systems, labor protections, infrastructure, healthcare, social insurance, and financial inclusion.
Research across economics and social policy — from institutional economics to inequality literature such as The Spirit Level and Piketty’s work on capital concentration — has also consistently shown that more equal societies tend to generate higher levels of trust, social cohesion, resilience, and overall well-being. If AI dramatically accelerates the concentration of productivity gains while weakening broad participation in the economy, the long-term consequences may extend far beyond labor markets alone. Conversely, if AI is deployed in ways that expand access to education, information, healthcare, resilience systems, and economic participation, it could also become a powerful equalizing force.
This is where the state-versus-market argument starts to feel increasingly artificial.
The internet itself emerged out of deep public investment. So did much of modern computing, aerospace, telecommunications, and pharmaceutical research. Markets matter enormously. Entrepreneurship matters enormously. But public institutions have almost always played a foundational role in shaping major technological transformations, partly because the time horizon and scale of investment required for transformational innovation often exceed what purely short-term market incentives can sustain on their own.
The question is not whether states should intervene in technological change. They already do. The real question is: in whose interest, toward what horizon, and with what ethical framework?
Because right now, much of the AI race is being driven by extremely short-term incentives: scale, speed, market capture, monetization. Understandably so. That is how markets function. But AI may simply be too important to be shaped exclusively by short-term market incentives. AI is already substantially shaping not only markets, but also state power, information systems, surveillance architectures, and geopolitical competition.
We should already be having serious conversations about ethical regulatory frameworks, transition support for displaced workers, education systems built around adaptability rather than narrow specialization, competition policy, public-interest AI infrastructure, and mechanisms through which the gains from AI-driven productivity can strengthen social stability rather than erode it.
At the same time, we should avoid framing AI purely as a commercial product.
Some of the most transformative uses of AI may emerge precisely in structurally unequal communities that currently represent little commercial value to major technology firms: smallholder farmers, isolated fishing communities, disaster-prone villages, underserved schools, and fragile regions with limited access to information systems.
These communities may not be profitable “customers.” But they should still benefit from intelligence infrastructure.
An AI system that helps a coastal community better anticipate storms, helps farmers improve harvest decisions, supports offline education in remote regions, or strengthens disaster preparedness is not merely a product. It starts becoming something closer to public infrastructure.
And this is where we should probably be careful with another increasingly popular concept: the democratization of AI.
Democratization alone is not enough. Something can be widely accessible and still fail to contribute to human well-being. Technology has never been an end in itself. Its value has always been measured by what it allows people to become. The deeper question is not only who has access to AI, but toward what purpose it is being deployed.
If AI merely accelerates extraction, concentration of wealth, and social fragmentation, then wider access alone will not solve much. But if AI can help build societies that are more resilient, more inclusive, more creative, and ultimately more equal and humane, then it may become one of the most important public-interest technologies of our time.
