兩位已故經濟學家談論人工智慧

Two Dead Economists on AI

"History doesn't repeat itself, but it often rhymes." —Mark Twain

Two Dead Economists on AI

First, watch Keynes:

Princeton reanimated John Maynard Keynes through AI trained on his writings and voice. He discusses his 1930 essay “Economic Possibilities for Our Grandchildren” and what it means for artificial intelligence.

This Substack is reader-supported. To receive new posts and support my work, consider becoming a free or paid subscriber.

Now, Schumpeter responds:

A Conversation with Joseph Schumpeter: Creative Destruction in the Age of AI

INTERVIEWER: Professor Schumpeter, we recently hosted an AI-generated version of John Maynard Keynes discussing his essay on economic possibilities. He argued that AI could deepen technological unemployment and requires careful demand management to prevent social breakdown. As someone who spent your career studying “creative destruction,” how would you respond?

SCHUMPETER: Hearing Keynes’ ghost warn about the dangers of progress fills me with nostalgia. We had these same debates in Cambridge in the 1930s, and I found his pessimism misplaced then as I do now.

Keynes worries that we discover ways to economize on labor faster than we discover new uses for it. But this misunderstands the fundamental nature of capitalist development. The problem is not technological unemployment - it is the temporary dislocation inherent in creative destruction. This process is not a bug in the system; it is the essential fact about capitalism.

INTERVIEWER: Keynes predicted that within 100 years, the economic problem would be largely solved and we’d need only 15-hour work weeks. His productivity forecasts were roughly correct, yet we still work full schedules. He attributes this to “relative needs” and status competition. Do you agree?

SCHUMPETER: Keynes makes a category error. He treats the economy as a problem to be “solved” - as if economic life were a mathematical equation with a final answer. But economic development is not about reaching equilibrium; it is about perpetual revolution from within.

Yes, productivity has multiplied as he predicted. But Keynes assumes this means we approach some satiation point. This fundamentally misunderstands human nature and economic evolution. New wants emerge from new possibilities that innovation creates.

Did people in 1930 want smartphones? Streaming entertainment? mRNA vaccines? Access to the world’s information in their pocket? These wants didn’t exist because the technologies didn’t exist. The entrepreneur doesn’t satisfy existing wants; he creates new ones by introducing new combinations of productive means.

INTERVIEWER: So you’re saying AI won’t create unemployment because it will generate entirely new industries we can’t yet imagine?

SCHUMPETER: Not quite so simple. AI will absolutely destroy existing jobs - that is certain. The question is what happens next, and here Keynes and I part ways fundamentally.

Keynes sees job destruction and immediately reaches for demand management, redistribution, shorter working hours. He wants to cushion the blow, to manage the transition. But this therapeutic approach misses the engine of progress.

When old industries die, they release resources - capital, labor, entrepreneurial energy. The crucial question is: what channels these resources into new combinations? The answer is entrepreneurial profit. The prospect of abnormal returns - what Keynes dismisses as “rentier income” - is what induces someone to risk everything on an unproven idea.

INTERVIEWER: But during the transition, real people lose real jobs. Keynes argues we need stabilization policy to prevent what he calls a “nervous breakdown” of society. Isn’t that reasonable?

SCHUMPETER: Here we reach the heart of our disagreement. Keynes believes in stabilization. I believe in evolution.

Yes, there is friction, dislocation, hardship. I do not minimize this. But Keynes’ solution - to prop up aggregate demand, to redistribute productivity gains, to prevent the adjustment - risks freezing the economy in amber. It protects the old against the new.

Consider: if in 1930 we had implemented Keynes’ program to protect agricultural and manufacturing workers, where would the computer industry have come from? The internet? Biotechnology? These industries emerged precisely because resources were freed from old sectors and attracted to new opportunities by the prospect of profit.

INTERVIEWER: So you’d accept higher transitional unemployment?

SCHUMPETER: I would accept reality. The choice is not between “smooth transition managed by wise policy” and “creative destruction with friction.” The choice is between genuine development through creative destruction, or stagnation disguised as stability.

What Keynes calls technological unemployment, I call reallocation. The relevant question is not “how do we prevent it” but “how do we ensure the resources find their highest use?” Through the price system, through entrepreneurial discovery, through the profit motive.

INTERVIEWER: Let’s talk specifically about artificial intelligence. Keynes sees it as another technology that economizes on labor - this time mental rather than manual work. He worries it could create overlapping waves of displacement. Your response?

SCHUMPETER: This is precisely wrong! AI is not primarily about economizing on existing labor. It is about enabling entirely new combinations that were previously impossible.

Think about what AI actually does. It doesn’t just make a human accountant work faster. It makes it possible to analyze data patterns no human could perceive, to optimize systems too complex for human comprehension, to create in ways that blend human and machine creativity in novel ways.

This is not economizing - this is the introduction of new production functions! The carrying out of new combinations, the introduction of new methods of production, the opening of new markets, the conquest of new sources of supply.

INTERVIEWER: But won’t a few large platforms capture most of the gains, creating what Keynes calls “new rentiers”?

SCHUMPETER: Ah, now we come to Keynes’ eternal confusion between monopoly profits and rentier income!

Yes, the pioneers of AI - your Anthropics, your OpenAIs, your NVIDIAs - will earn extraordinary profits. But these are entrepreneurial profits, not rentier rents! They risked capital on uncertain technologies. They created new markets. They should be rewarded handsomely.

But - and Keynes always misses this - these profits are self-liquidating. High returns attract competition. New entrepreneurs enter with better models, more efficient chips, novel applications. The monopoly position erodes. This is creative destruction working as it should.

Keynes wants to tax away these profits or regulate them into submission. But this kills the very signal that attracts the competition that ultimately benefits consumers. You cannot have the innovation without the profit, and you cannot have lasting innovation without the destruction of yesterday’s profits.

INTERVIEWER: Keynes discussed Jevons’ paradox - that efficiency can increase total resource use. He suggests that with AI, whether we see this depends on income distribution. If productivity gains go to capital while wages lag, we get deficient demand. Doesn’t this concern you?

SCHUMPETER: This is Keynes at his most seductive and most wrong. He makes it sound so reasonable: we must manage distribution to maintain demand. But this is the economics of the circular flow, not of development.

In my framework, development occurs in spurts, in cycles. There is a boom phase when entrepreneurs introduce new combinations. Profits are high, inequality rises, resources flow to new sectors. Then comes adaptation - wages rise as new industries mature, consumption patterns adjust, the gains diffuse through society.

Keynes sees the boom phase and wants to immediately redistribute. But this short-circuits the process! The high profits in the boom are what fund the next wave of innovation. The inequality is temporary - unless you prevent the competitive destruction that eliminates it.

INTERVIEWER: So rising inequality during an AI boom doesn’t concern you?

SCHUMPETER: It depends entirely on its source and duration. If inequality rises because entrepreneurs are creating new value and being rewarded for it - this is healthy and temporary. Competition will erode it.

If inequality rises because the state grants monopolies, or because we bail out failing firms, or because we prevent creative destruction - this is toxic and permanent. This is the rentier income Keynes should worry about but never does.

The real question with AI is not “how do we redistribute the gains” but “how do we ensure the competitive process continues?” Regulate monopolies when they arise, yes. But do not strangle innovation to achieve distributional goals.

INTERVIEWER: Keynes discussed AI stock valuations through his “beauty contest” metaphor - investors trying to guess what other investors will think. He warns we may be in a speculative bubble. Do you see it differently?

SCHUMPETER: Keynes’ beauty contest is clever but misses the point. Yes, there is speculation. Yes, some AI valuations will prove absurd. So what?

This is how capitalism finances innovation! Someone must take the risk on unproven technologies. In my work on business cycles, I showed that credit creation - including what looks like speculative excess - is essential for funding new combinations.

The railroad boom of the 1840s looked like a bubble. The electricity boom of the 1890s looked like a bubble. The dot-com boom looked like a bubble. And they were! Many companies failed. Many investors lost money. But the infrastructure got built. The technologies got developed. Society benefited enormously.

INTERVIEWER: So bubbles serve a purpose?

SCHUMPETER: “Bubble” is Keynes’ word, loaded with moral disapproval. I prefer to think of it as capitalism’s way of funding radical uncertainty. When you don’t know which AI approach will win, you fund many. Most fail. A few succeed spectacularly. The system works.

Keynes wants to prevent the “excesses” through monetary policy and regulation. But you cannot have the upside without the downside. You cannot have transformative innovation without waste and failure. The attempt to stabilize everything produces stagnation.

INTERVIEWER: Keynes argued that once the economic problem is solved, economists would become “humble specialists like dentists.” But he also said the “master economist” must be mathematician, historian, statesman, and philosopher. How do you see the role of economics in navigating AI transitions?

SCHUMPETER: Keynes’ dentist remark reveals his fundamental misunderstanding. The economic problem is never “solved” because it is not a problem - it is a process!

As long as human wants are unlimited, as long as innovation creates new possibilities, as long as entrepreneurs form new combinations, there will be economic development. And this development is not a technical matter for specialists. It requires understanding history, institutions, culture, power.

But here Keynes and I agree on something: most modern economists are too narrow. They build elegant models of equilibrium and optimization. They miss the essential fact - that capitalism is evolution, not equilibrium. It is perpetual gale of creative destruction, not a system tending toward rest.

INTERVIEWER: What would you say to economists studying AI today?

SCHUMPETER: Stop trying to manage the transition and start trying to understand the transformation.

AI is not a shock to be smoothed. It is a new epoch in capitalist development, comparable to the steam engine or electricity. Your models of technological unemployment and demand stabilization miss the point entirely.

Ask instead: What new combinations does AI enable? Which industries face obsolescence? Where will entrepreneurial energy flow? What institutional forms will emerge? How will the competitive process unfold?

These are historical and evolutionary questions, not technical ones. You cannot model them with representative agents and market-clearing conditions. You must study them as I studied the transition from competitive to trustified capitalism.

INTERVIEWER: If you were advising policymakers today, what would you tell them about AI?

SCHUMPETER: First, abandon the Keynesian dream of smooth, managed transitions. You cannot plan creative destruction. You can only ensure the institutional conditions that allow it to work.

This means: Maintain competitive markets. Prevent the state from protecting obsolete industries. Do not try to preserve jobs that technology has made redundant. Let resources flow to their highest use.

Second, understand that inequality during innovation booms is not market failure - it is market function. The abnormal profits of AI pioneers are the signal that attracts the competition that will eventually benefit everyone.

Third, accept that some will lose from this transformation. The proper response is not to prevent the transformation but to ensure those who lose can participate in the new opportunities. Education, yes. Redistribution of opportunity, yes. But not redistribution of outcomes that kills the incentive to create them.

INTERVIEWER: Keynes would argue that without demand management, we risk social breakdown during the transition.

SCHUMPETER: And I would argue that Keynes’ demand management risks something worse: a civilization that has lost the capacity for genuine development. A society that values security over dynamism, equality over excellence, stability over progress.

The choice is not between Keynesian stability and chaos. It is between evolution and stagnation, between a living economy and a fossilized one.

INTERVIEWER: Keynes famously said “in the long run we are all dead,” defending his focus on short-run stabilization. Your response?

SCHUMPETER: Yes, in the long run we are all dead. But our civilization lives on, shaped by the choices we make today.

If we choose Keynes’ path - stabilizing demand, redistributing productivity gains, protecting workers from disruption - we may smooth the transition. But we sacrifice the long-run dynamism that creates entirely new possibilities.

If we choose the path of creative destruction - accepting temporary inequality, allowing resources to flow to new combinations, letting entrepreneurial profit guide development - we accept harder transitions. But we preserve capitalism’s revolutionary character.

INTERVIEWER: Which path do you think we’re on with AI?

SCHUMPETER: We are, as always, muddling through with a mixture of both. Some genuine creative destruction, some Keynesian stabilization, some crony capitalism, some real innovation.

But I worry that the intellectual climate has shifted too far toward Keynes. Everyone wants to manage, to regulate, to redistribute. Few appreciate that the chaos and inequality of innovation booms are not aberrations to be corrected but essential features of how capitalism creates new value.

INTERVIEWER: Final question: Looking at AI development today, are you optimistic or pessimistic?

SCHUMPETER: I am historically minded. I see AI as another wave in the long history of capitalist development. Like previous waves - steam, electricity, automobiles, computers - it will destroy old industries and create new ones. Some will gain enormously. Others will lose. Society will be transformed in ways we cannot predict.

The question is not whether this transformation will be painful - it will be. The question is whether we will allow it to unfold, or whether we will try to manage it into mediocrity.

Keynes offers comfort: we can smooth the transition, share the gains, prevent disruption. But comfort is not progress. Stability is not development.

I offer no comfort, only reality: capitalism is a process of creative destruction. If you want the creation, you must accept the destruction. If you want the long-run dynamism that has multiplied human welfare a hundredfold, you must accept the short-run dislocation that makes it possible.

AI will be no different. The question is whether we have the courage to let it work.

INTERVIEWER: Thank you, Professor Schumpeter.

SCHUMPETER: Thank you. And to those listening: study history, understand institutions, and remember - capitalism is evolution, not equilibrium. Plan accordingly.

This Substack is reader-supported. To receive new posts and support my work, consider becoming a free or paid subscriber.

No posts

Ready for more?