When Central Bankers Start Making Bets: The Gap Between Old Economics and the Intelligence Economy
The incoming Fed Chair says policymakers must “make a bet” on AI productivity. He’s right about the bet. He’s wrong about what we’re betting on.
By Dr. Elias Kairos Chen
February 2026
The Most Important Economic Statement of the Year
Kevin Warsh—Trump’s nominee for Federal Reserve Chair—made a statement last month that should have dominated every economics discussion since. In an interview, he said:
“The difficulty of [AI] for policymakers—let’s say central bankers, let’s say fiscal authorities—is that the economy is going to be growing, but it will not show up in the productivity statistics. So we are going to have to make a bet.”
Read that again. The incoming chair of the world’s most powerful central bank just admitted that traditional economic data is about to become useless for policymaking.
This is extraordinary. And Warsh went further. He invoked Alan Greenspan’s decision in 1993-94 to hold rates steady despite conventional wisdom demanding increases—because Greenspan believed, based on “anecdotes and rather esoteric data,” that the internet revolution would be structurally deflationary. Greenspan made a bet. He was right.
Warsh is signaling he’ll make the same bet on AI.
Here’s the thing: Warsh’s diagnosis is precisely correct. His prescription has a gap so large that history will judge it as the most consequential blind spot of the intelligence transition.
What Warsh Gets Right: The Best Version of Old Economics
Before I explain the gap, let me give Warsh credit. He articulates the sophisticated version of establishment economic thinking more clearly than anyone else has. His key insights deserve serious engagement.
The Measurement Problem
Warsh understands something most economists still haven’t grasped: AI productivity gains will be invisible to traditional statistics. The Bureau of Labor Statistics uses frameworks designed in the 1970s. When a knowledge worker becomes 40% more productive using AI tools, that productivity often shows up nowhere in official data.
Goldman Sachs research bears this out. Their analysis suggests AI has added approximately $160 billion to actual economic output since 2022—but only about $45 billion, roughly 28%, appears in official GDP statistics. The measurement gap is real and widening.
The “Cost of Curiosity” Frame
Warsh offered a powerful articulation of AI’s fundamental economic shift: “The cost of curiosity is now zero.” This captures something profound. For all of human history, acquiring knowledge required resources—time, access, money, social capital. Libraries closed at 10pm. Expertise required years of training. Information asymmetries were economic moats.
When an AI can instantly access and synthesize human knowledge, the friction that defined traditional economics evaporates. Warsh sees this clearly.
Conviction Economics vs. Data Dependence
Warsh explicitly contrasts “conviction economics” with the “data dependency” practiced by Jay Powell’s Federal Reserve. His argument: waiting for backward-looking data to confirm AI productivity gains means you’re always late. By the time statistics validate what’s happening, the policy window has closed.
He cites the Bezos principle: “At times of huge consequence, at turning points, if you have a set of data that’s telling you one thing, a set of anecdotes that are telling you the other—listen to the anecdotes.”
On this, Warsh is absolutely right. The anecdotes have turned. CEOs who were skeptical 100 days ago now have “excitement in their eyes.” The transformation is underway.
The Greenspan Precedent—and Its Limits
Warsh’s Greenspan analogy is instructive, but not in the way he intends.
In 1993-94, Greenspan believed the internet would generate structural productivity gains before the data confirmed it. He held rates lower than conventional models demanded. He was vindicated. The late 1990s saw strong growth with stable prices.
But here’s what the Greenspan precedent also shows: even when the Fed gets productivity right, the underlying revolution still produces massive disruption. The internet created enormous wealth. It also destroyed entire industries, concentrated gains among capital owners, and generated the dot-com bubble that erased $5 trillion in market value when it burst.
Greenspan was right about productivity. The economy still needed policy frameworks he never developed—for antitrust in network effects, for worker retraining at scale, for managing wealth concentration.
The AI revolution is the internet revolution squared. Getting productivity right is necessary but radically insufficient.
The First Gap: When “Productivity Leads Wages” Breaks Down
Warsh’s core economic claim is elegant and traditionally unimpeachable: “If we learned anything in economics, what we’ve learned is productivity gains are the predecessor to wage gains.”
This has been true for two centuries. Worker produces more output per hour. Competition forces employers to share gains through wages. Living standards rise broadly.
But this mechanism depends on a crucial assumption: human labor remains the scarce input.
What happens when productivity gains come from eliminating human labor entirely?
Consider what AI is actually doing. It’s not making human workers faster at their tasks—that was the automation of the past century. It’s performing cognitive work that previously required human minds. When an AI system can do in minutes what a team of analysts did in weeks, the productivity gain doesn’t translate into higher wages for analysts. The analysts are gone.
Warsh mentions that “52% of our fellow Americans have no equity”—no stocks, no 401(k), no pension. They experience wealth creation only through wages. His solution? Better products, more competition, innovation in financial services.
This fundamentally misreads the moment. If AI productivity gains flow to capital rather than labor—and they are structured to do exactly that—then making financial products cheaper helps people access... what wages? The mechanism that translated productivity into broad prosperity is breaking precisely as productivity accelerates.
The Greenspan precedent didn’t involve this rupture. The internet made workers more productive. AI is making workers optional.
The Second Gap: The Competition Fallacy
Warsh’s prescription for distributing AI benefits rests on competition. He celebrates the “micro foundations” of American capitalism—entrepreneurship, risk-taking, the aspiration to do better rather than envy of neighbors. He argues that competition in financial services will bring AI benefits to ordinary Americans.
This reflects a profound misunderstanding of AI economics.
Traditional competition works because barriers to entry are surmountable. A better restaurant can challenge an established one. A smarter entrepreneur can build a competitive product.
AI has different economics. Foundation models require billions in compute investment. Data moats are self-reinforcing—the more users, the better the model, the more users. The infrastructure being built right now creates economic dependencies that may prove permanent.
When Warsh says the “best companies” will capture AI productivity gains, he’s describing concentration, not competition. When he says the U.S. will “gap out even further” from other nations in productivity, he’s describing international inequality, not broadly shared prosperity.
Competition is indeed coming to financial services—but it’s competition among AI-powered platforms for the privilege of serving consumers with increasingly precarious incomes. That’s not the same as competition that raises wages and broadly distributes wealth.
The Third Gap: No Framework for What We Can’t Measure
Warsh correctly identifies that AI productivity won’t show up in traditional statistics. His response is that policymakers must “make a bet” based on conviction rather than data.
But here’s the critical question he doesn’t ask: if our measurement frameworks are failing, shouldn’t we be building new ones rather than just betting?
The measurement gap is not a technical limitation—it’s a framework failure. GDP measures transaction volume. When AI performs work that used to generate transactions (salaries, payments, service fees) but now happens inside corporate systems for near-zero marginal cost, that value creation becomes invisible.
We need new measurement paradigms:
What is actual capability being deployed in the economy?
Who has access to AI-powered productivity enhancement?
How is value creation being distributed?
What are the outcomes for human flourishing, not just transaction volume?
Warsh’s “conviction economics” assumes we know what to bet on. But without frameworks to measure the AI economy, we’re making bets in the dark. Greenspan could at least see productivity statistics, even if they lagged. We’re heading into an economy where the statistics themselves become meaningless.
The Fourth Gap: No Post-Labor Framework
This is the most consequential gap. Warsh offers no framework for what happens when human cognitive labor loses economic value.
His optimism rests on historical precedent: technology displaces jobs in one sector while creating jobs elsewhere. The agricultural revolution freed workers for manufacturing. The industrial revolution freed workers for services. The information revolution freed workers for knowledge work.
But each of these transitions preserved human comparative advantage. Machines did physical work better; humans did cognitive work. Machines processed data faster; humans provided judgment and creativity.
AI dissolves these distinctions. Systems that match or exceed human capabilities in analysis, judgment, creativity, and problem-solving don’t preserve any clear domain of human economic advantage.
Warsh acknowledges this problem obliquely—he mentions the K-12 education gap, the need for workers to develop skills to participate in the productivity revolution. But if AI systems can learn any skill faster than humans can be educated, what does “developing skills” even mean?
The historical frame fails here. We’re not transitioning workers from one form of productive labor to another. We’re potentially transitioning from an economy based on human labor to one based on something else entirely.
And for that transition, we have no framework. No Universal Basic Income proposals. No post-labor social contracts. No vision of human meaning and purpose when economic contribution is no longer required.
What New Economics Must Address
If Warsh represents the most sophisticated version of old economics adapting to AI, what would genuinely new economics look like?
New Measurement: Beyond GDP
We need what might be called “Capability GDP”—measuring actual problem-solving capacity deployed in the economy rather than transaction volume. When an AI cures diseases, educates children, or solves logistics problems, that value creation matters regardless of whether it generates traditional economic activity.
The Human Prosperity Index I’ve been developing measures what actually matters: material sufficiency, capability access, human agency, sustainability, and social cohesion. These metrics become essential when traditional economic indicators decouple from human wellbeing.
New Distribution: Universal Basic Infrastructure
If AI productivity gains accrue to capital, the traditional mechanisms for distributing prosperity fail. We need new mechanisms:
Universal Basic Income that provides floor-level economic security
Universal Basic Intelligence—ensuring everyone has access to AI capabilities, not just those who can afford premium tools
Universal Basic Capital—restructuring ownership so citizens have stakes in AI-generated wealth
These aren’t welfare programs. They’re infrastructure for an economy where traditional employment cannot distribute productivity gains.
New Purpose: Beyond Economic Contribution
The deepest challenge isn’t economic—it’s existential. For centuries, work provided identity, structure, social connection, and meaning. An economy that no longer needs human labor requires new frameworks for human flourishing that old economics never contemplated.
Warsh’s “micro foundations”—the culture of aspiration and risk-taking—remain valuable. But they need new expression when aspiration cannot be fulfilled through traditional employment and risk-taking cannot be financially rewarded through wages.
The Real Bet
Warsh is right that policymakers must make a bet. But he’s framing the wrong bet.
His bet: AI will generate productivity gains that may not appear in statistics. Therefore, monetary policy should accommodate growth rather than fighting phantom inflation.
That’s a reasonable monetary policy bet. But it’s a tiny piece of a much larger wager.
The real bet is this: Can we rebuild economic institutions fast enough to distribute AI’s productivity gains before the traditional mechanisms for doing so collapse entirely?
This bet has a timeline. If AGI capabilities are already here—as AI pioneers acknowledged in late 2025—then superintelligence is perhaps 18-24 months away. The infrastructure for artificial minds is being built right now, on federal land, with government support, targeting operational status by late 2027.
We have maybe 24-36 months to develop the frameworks, measurement systems, distribution mechanisms, and social contracts that Warsh’s economics doesn’t contemplate.
That’s the bet. And unlike Greenspan’s bet in 1994—where being wrong meant suboptimal growth—being wrong on this bet means civilizational consequences.
Conviction Without Framework Is Not Enough
Kevin Warsh deserves credit for recognizing what many economists still deny: AI productivity is real, imminent, and unmeasurable by traditional statistics. His willingness to make bets based on conviction rather than lagging data is appropriate for the moment.
But conviction economics without new frameworks is insufficient. Warsh is right that we must bet. He’s wrong about what we’re betting on.
We’re not betting on whether AI will generate productivity gains. That bet is already won.
We’re betting on whether we can rebuild the foundational structures of economic life—measurement, distribution, purpose—before the old structures fail.
Warsh’s old economics gives us confidence in productivity. New economics must give us frameworks for prosperity when productivity no longer needs us.
The Greenspan precedent isn’t about getting productivity right. It’s about what happens after—when being right about productivity still leaves us unprepared for everything that follows.
That’s the gap. And filling it is the defining economic challenge of our time.
Dr. Elias Kairos Chen is the author of “Framing the Intelligence Revolution” and writes weekly on the economic transformation accelerating around us. This analysis is part of the “Framing the Future of Superintelligence” series examining what happens when machines exceed human capabilities.
Key Quotes from Kevin Warsh Interview
For reference, here are the key Warsh statements this analysis engages:
On the measurement problem:
“The difficulty of [AI] for policymakers—let’s say central bankers, let’s say fiscal authorities—is that the economy is going to be growing, but it will not show up in the productivity statistics.”
On conviction economics:
“So we are going to have to make a bet. Is the economy becoming much more productive?... If you’re looking at the [economic] data, my view is you’re backward looking. You’re going to be late.”
On the Greenspan precedent:
“The closest analogy I have in central banking is Alan Greenspan in 1993 and 1994... He believed based on anecdotes and rather esoteric data that we weren’t in a position where we needed to raise rates because this technology wave was going to be structurally disinflationary.”
On productivity and wages:
“If we learned anything in economics, what we’ve learned is productivity gains are the predecessor to wage gains.”
On the cost of curiosity:
“This is the most productivity enhancing wave of our lifetimes, past, present, and the future. The way I think about it is the cost of curiosity is now zero.”
On American advantage:
“My bet would be that we’re at the early innings, but the relative growth of the United States at the cutting edge of this productivity wave relative to the rest of the world will gap out even further in the next 5 years.”
On the 52%:
“52% of our fellow Americans have no equity. They don’t have equity in their house. They don’t have an account at Schwab or Robin Hood... they don’t have a pension.”
Couldn't agree more. My AI dev projects feel this way.