Luke Drago and Rudolf Laine Warn of an "Intelligence Curse": AI Could Make Humans Economically Obsolete
Explainers
According to researchers Luke Drago and Rudolf Laine, AI could trigger a far bigger problem than job losses alone. If artificial intelligence eventually performs nearly all economically valuable work, ordinary people could also lose their economic leverage. They call this scenario the Intelligence Curse.
Drago and Laine published their extensive essay series The Intelligence Curse in 2025, exploring what happens when companies and governments become less dependent on human labor for economic growth. Read the original Intelligence Curse essay series.
Their warning is not a simple prediction that ChatGPT will fire everyone soon. Their argument runs deeper: if human labor is no longer needed to create economic value, powerful organizations may also lose a key reason to consider the interests of ordinary people.
AI and technological progress are entangled with questions of wealth, social power, and even humanity’s long-term survival.
Why they say AI starts at the bottom of the corporate pyramid
Drago and Laine expect automation to climb through organizations step by step. They call this pyramid replacement.
A traditional company roughly resembles a pyramid. At the bottom are large numbers of junior employees. Above them sit smaller groups of experienced staff and managers. At the very top is a relatively small executive team.
The authors argue AI will hit the lower layers first.
Junior roles often involve tasks that are easy to standardize: processing information, drafting initial analyses, answering simple queries, writing first-pass copy, or handling recurring administrative work.
That doesn’t require a company to fire thousands overnight. It can simply hire fewer new people.
That concern is now visible beyond their theory. AI Wereld previously covered research on AI’s real impact on the labor market. It shows AI can technically handle far more tasks than companies currently automate. At the same time, there are signs that entry-level hiring in certain junior roles is under pressure.
Drago and Laine push that trend much further. As AI grows more capable, they argue increasingly complex layers can be automated—moving from execution to management, analysis, and ultimately strategic decision-making.
AI agents could accelerate this shift. These systems don’t just respond to prompts; they can plan and execute steps autonomously.
Their real concern isn’t unemployment
The most striking part of the Intelligence Curse is that Drago and Laine don’t see joblessness as the endgame.
Their bigger fear is human economic irrelevance.
People currently have economic power because companies and states need something from them. Businesses need workers to deliver products and services. Governments depend on people to conduct economic activity, start companies, and pay taxes.
That creates strong incentives to invest in people.
A well-educated worker becomes more productive. A healthy population can produce more. Good infrastructure makes companies and employees more efficient. Even a profit-maximizing firm benefits when the workers it needs have sufficient education and skills.
Drago and Laine worry that highly advanced AI could fundamentally rewrite this relationship.
Imagine an AI system that can do virtually the same work as a human—and that this digital labor can be replicated at will. A company no longer needs to train thousands of people to scale production. It can buy more compute.
Human intelligence would shift from a scarce production factor to something machines can also supply.
In that world, say Drago and Laine, people could lose their bargaining power.
Why they compare AI to oil and other resources
The term Intelligence Curse is deliberately inspired by the economic “resource curse.”
Some countries sit on vast reserves of oil, gas, diamonds, or other natural resources, yet still have weak institutions and little broadly shared prosperity.
A key theory behind this centers on economic dependence.
A government that relies on millions of productive citizens for revenue has a strong incentive to keep people healthy, educated, and working. A regime that can tap vast income directly from oil fields is far less tied to the productivity of individual citizens.
Drago and Laine ask whether extremely powerful AI could trigger a similar dynamic.
Instead of oil fields, the economy would pivot to data centers, chips, energy, robots, and AI models. The most valuable resource would no longer be human expertise alone, but artificial intelligence that can be deployed with enough capital.
That makes AI’s physical backbone more important. AI isn’t just software—it runs on specialized chips, massive server systems, data centers, networks, and electricity.
AI Wereld previously described how that stack now spans full compute systems. Nvidia, for example, doesn’t just sell GPUs—it ships integrated racks, networking gear, and software to build large-scale AI systems.
In Drago and Laine’s extreme scenario, owning these production assets matters more than owning human skills.
When AI is owned, power follows
The Intelligence Curse is, at its core, about ownership.
A worker owns time, experience, and skills. Today, rare expertise can be highly valuable. A strong programmer, scientist, or founder can create economic value without starting with billions in capital.
Drago and Laine fear powerful AI weakens that mechanism.
A very wealthy organization could buy not just buildings and machines—but effectively, intelligence. More chips and more compute yield more digital labor.
Capital’s edge could then compound.
This idea isn’t entirely novel. Economists Anton Korinek and Donghyun Suh modeled several paths to artificial general intelligence in an NBER paper. Their scenarios diverge widely: if automation is gradual and leaves human tasks, wages can keep rising. If all human tasks become automatable, wages can collapse.
That research doesn’t prove the Intelligence Curse. It does show that rapid growth and rising wages don’t have to move together.
An economy could, in theory, produce far more while the market value of human labor falls.
Upward mobility could take a hit
Drago and Laine also warn of a second effect: social mobility could slow.
Human talent is currently a way to break out of one’s starting position. Without large wealth, someone can become an exceptional programmer, entrepreneur, researcher, or designer—and build power from there.
If intelligent work can largely be purchased as AI compute, existing capital gains a bigger edge, the authors argue.
Those who already own companies, chips, energy assets, or large-scale compute can also deploy large-scale artificial intelligence.
In the extreme, that creates a self-reinforcing loop: capital buys AI, AI generates value, that value returns as more capital.
Human workers drift further outside that loop.
Economist Pascual Restrepo, in recent work on a hypothetical AGI economy, likewise concludes that labor’s share of total income can trend toward zero if computational systems can take over all economically necessary work. Again: a long-run theoretical outcome, not today’s job market.
Today’s reality is still far from that future
Drago and Laine’s thesis hinges on a massive assumption: AI would eventually perform nearly all economically relevant human work.
That’s not where we are today.
The International Labour Organization’s global analysis of generative AI found that about one in four workers holds a job where at least some tasks are exposed to generative AI. Only 3.3 percent of global employment fell into the highest exposure category.
Exposure doesn’t mean an entire job disappears.
A job is a bundle of tasks. AI can automate some, accelerate others, and struggle with the rest—for now. Human oversight, accountability, physical work, social interaction, and organizational constraints still matter.
The current labor market, therefore, offers no evidence that the Intelligence Curse is actually on its way.
Drago and Laine also frame their work as a problem that emerges only if extremely powerful AI can replace nearly all human labor. That distinction matters.
Their theory works better as scenario analysis than as a forecast for next year.
Their answer isn’t to shut AI down
Strikingly, Drago and Laine aren’t simple AI opponents.
They believe technological progress can be steered.
Their proposed playbook boils down to three ideas—avert, diffuse, and democratize: prevent catastrophic outcomes, spread AI capabilities and economic power more widely, and design institutions so people retain political influence.
A key piece is building technology that empowers ordinary people instead of pushing them out of economic activity.
Drago and Laine have worked on that themselves. They founded Workshop Labs on the idea that people should be able to link their own knowledge and skills to AI systems that act in their interest.
In April 2026, Workshop Labs announced it would become part of Thinking Machines. Drago and Laine wrote that they originally founded the company to make people “irreplaceable.” Both now work at Thinking Machines.
Their alternative to full centralization isn’t less technology—it’s a different direction for technology.
For example, personal AI systems could amplify individual users’ expertise. Small businesses could use the same powerful tools as large organizations. And access to AI capabilities, in their view, shouldn’t be concentrated in a handful of companies or governments.
At its core, the Intelligence Curse is about power
That shifts the focus to a bigger question than how many jobs ChatGPT, Claude, or future AI agents might automate.
Where does ordinary people’s power come from in a highly automated economy?
In today’s society, part of it is economic. Companies need workers. Governments need productive citizens. Consumers have income because they participate in that economy.
If advanced AI ever removes much of that interdependence, Drago and Laine argue that social power must be protected in other ways.
Their scenario may never materialize. New human roles could emerge. AI may boost workers’ productivity rather than replace them. Laws could reshape how AI gains are distributed. AI ownership could be broadened. And the technical path to full automation may prove far harder than the authors assume.
But that’s exactly why the Intelligence Curse is a useful thought experiment.
The traditional debate asks: what happens to people if AI takes our jobs?
Drago and Laine pose a more uncomfortable follow-up.
What happens to our power if the economy no longer needs our labor at all?
Loading