51吃瓜

AI has rapidly moved from the realm of technological innovation to the center of economic and geopolitical strategy. Governments increasingly regard AI as a strategic national asset, corporations are committing hundreds of billions of dollars to AI infrastructure and financial markets are pricing in expectations of a new technological revolution. Much of the public debate has focused on AI鈥檚 potential to transform productivity, employment and economic growth. Yet one fundamental question has received far less attention: How will AI reshape the conduct of monetary policy?AI is not merely another source of productivity growth. Rather, it represents a profound structural transformation that changes how inflation emerges, how expectations are formed, how financial risks accumulate and ultimately how central banks must interpret the economy itself.

A new challenge for central banking

For nearly half a century, central bankers have relied on a remarkably stable framework for interpreting the economy. Inflation rises when aggregate demand grows faster than productive capacity. Recessions occur when demand falls below potential output. Productivity determines how quickly an economy can expand without generating inflation, while interest rates bring demand and supply back into equilibrium. This intellectual framework has survived oil shocks, globalization, the Global Financial Crisis and the COVID-19 pandemic. AI, however, may represent a more fundamental challenge.

The current debate about AI has largely focused on a familiar question: How much will AI increase productivity? Optimists compare AI to electricity or the internet, arguing that it could unleash decades of faster economic growth. Skeptics counter that history is littered with technological revolutions whose economic benefits arrived slowly or proved smaller than initially expected. Both sides are asking the wrong question.

For central banks, the more important issue is not how much AI increases productivity but how AI changes the relationship between inflation, growth and financial stability. AI simultaneously affects production technologies, firms鈥� pricing decisions, household expectations, investment behavior and financial markets. These forces interact in ways that blur the distinction between cyclical fluctuations and structural transformation 鈥� the very distinction upon which modern monetary policy has been built.

This distinction has always mattered. A cyclical shock requires stabilization. A structural shock requires adaptation. If inflation originates from excessive demand, higher interest rates are the appropriate response. If inflation reflects a temporary restructuring of production, tighter monetary policy may unnecessarily slow the very investments that generate future productivity growth. The difficulty is that, during the transition to an AI economy, these two forces may become almost impossible to disentangle in real time.

Recent evidence illustrates both the promise and the uncertainty surrounding AI. A large survey conducted by researchers affiliated with the Federal Reserve Bank of Chicago that economists expect US GDP growth to remain close to 2.5% annually under baseline assumptions but to rise toward 3.5%鈥�4.0% under a rapid AI adoption scenario. At the same time, labor-force participation could decline from approximately 62% today to around 55% by 2050, implying that nearly half of the projected decline in labor-force participation could be attributable to AI rather than demographics alone. Perhaps even more revealing is that experts disagree far more about AI鈥檚 economic consequences than about the pace of AI development itself.

That disagreement should concern policymakers. Monetary policy depends not only on forecasts but on understanding why the economy evolves as it does. If economists cannot agree on how AI affects inflation, employment or productivity, estimating the output gap, the neutral interest rate or inflation persistence becomes considerably more difficult. Monetary policy risks responding to symptoms while misunderstanding the underlying causes.

History offers an instructive comparison. During the late 1990s, policymakers struggled to determine whether the information technology revolution had permanently increased productivity or merely generated a temporary investment boom. The Federal Reserve eventually that productivity growth had indeed accelerated, allowing the economy to grow faster without igniting inflation. Yet excessive optimism also contributed to the dot-com bubble, illustrating that genuine technological progress and financial excess can occur simultaneously.

The AI revolution presents a similar challenge, but on a much larger scale. Unlike previous digital innovations, generative AI diffuses across virtually every knowledge-intensive sector. It assists software engineers, financial analysts, lawyers, physicians, researchers, manufacturers and even policymakers themselves. More importantly, AI is becoming embedded not only in production but also in the decisions that shape production. Algorithms increasingly determine prices, allocate credit, manage inventories, forecast demand, evaluate insurance risks and execute financial transactions. In effect, AI is transforming the economy while simultaneously transforming the mechanisms through which the economy is measured.

The productivity J-curve

Economic discussions of technological progress often assume a simple sequence: Innovation raises productivity, lower production costs reduce inflation and higher productivity allows faster economic growth. Reality is considerably less orderly.

Major technological revolutions rarely improve measured productivity immediately. They first require enormous investment. Firms adopting AI must purchase specialized semiconductors, expand cloud-computing capacity, construct data centers, redesign software architecture, strengthen cybersecurity, reorganize production processes and retrain employees. These expenditures consume resources long before measurable efficiency gains appear.

This creates what might be called the . During the early phase of technological transformation, measured productivity may actually deteriorate even as the technological frontier advances. Output grows more slowly because firms devote increasing resources to implementation rather than production. Costs rise before efficiency improves. The economy appears less productive precisely because it is investing in becoming more productive.

The phenomenon is hardly unprecedented. American economist and Nobel Laureate Robert Solow famously in 1987 that 鈥測ou can see the computer age everywhere but in the productivity statistics.鈥� Only after firms fundamentally reorganized production around digital technologies during the second half of the 1990s did productivity growth accelerate noticeably. Electricity followed a remarkably similar pattern. Factories initially replaced steam engines with electric motors but retained old production layouts, realizing only modest productivity gains. Substantial improvements emerged only after firms redesigned factories to fully exploit electricity鈥檚 flexibility.

Artificial intelligence may produce an even longer transition because it requires organizational rather than merely mechanical adaptation. Unlike previous technologies, AI fundamentally changes decision-making itself. Managers must learn to supervise algorithms, employees must work alongside intelligent systems, firms must redesign workflows and regulators must develop entirely new governance frameworks. These adjustments require substantial investments in digital infrastructure, organizational restructuring, cybersecurity, software integration and workforce retraining. Although these implementation costs are difficult to observe directly, they may temporarily outweigh the productivity benefits of AI adoption during the early stages of the transition.

The figure illustrates this mechanism. During the short run, implementation costs generate temporary inflationary pressures before productivity gains are realized. As firms gradually adapt their production processes and AI technologies diffuse throughout the economy, productivity improvements begin to offset these initial costs, producing what this paper refers to as Transition Inflation. Eventually, as AI reaches widespread adoption, productivity gains dominate implementation costs and inflation gradually returns toward its long-run equilibrium.

For central bankers, this creates a profound policy dilemma. Suppose inflation rises while measured productivity temporarily weakens. Conventional macroeconomic analysis might interpret this combination as evidence that aggregate demand has become excessive or that productive capacity has deteriorated. Yet both interpretations may be misleading. Inflation may instead reflect temporary implementation costs associated with AI adoption 鈥� costs that ultimately generate substantial productivity gains. In other words, the economy may temporarily appear weaker precisely because it is investing in becoming more productive.

The figure summarizes these dynamics by distinguishing three phases of the AI transition. Phase I represents the short-run adjustment period, during which implementation costs and organizational restructuring raise inflation. Phase II captures the transition period in which implementation costs gradually diminish while productivity gains emerge but remain insufficient to offset inflationary pressures completely. Phase III represents the long-run equilibrium, where productivity gains dominate implementation costs, production costs decline and inflation gradually converges toward its long-run level.

These dynamics fundamentally complicate monetary policy. Interest-rate increases designed to restrain inflation may simultaneously discourage the investment necessary for future productivity growth. Conversely, maintaining accommodative policy risks allowing financial markets to extrapolate future productivity gains too aggressively, fueling excessive leverage, speculative asset valuations and financial instability. Monetary policy therefore faces a challenge unlike that posed by conventional business cycles: distinguishing temporary transition inflation associated with technological transformation from persistent inflation generated by excess aggregate demand. The distinction becomes even more difficult once expectations enter the picture.

Expectations move faster than factories

If the Productivity J-Curve complicates the supply side of the economy, expectations make the demand side even more difficult to interpret.

Modern monetary policy is built on the insight that expectations matter. Inflation depends not only on today鈥檚 economic conditions but also on what households and firms believe about tomorrow. Central banks therefore devote enormous resources to managing expectations through forward guidance, policy communication and credibility. Artificial intelligence, however, may fundamentally change how those expectations evolve.

Consider a firm that believes AI will substantially increase future productivity. It has little incentive to wait until those gains materialize. Instead, it begins investing immediately in data centers, cloud infrastructure, software integration and worker training. Investors behave similarly. If future profits are expected to rise, equity valuations increase today rather than after productivity has improved. Households may also revise their expectations of future income upward, increasing current consumption through intertemporal substitution. In short, expectations respond almost instantly, whereas productive capacity expands only gradually.

One might summarize this process with a simple observation: Expectations move faster than factories.

This seemingly simple insight has profound implications for inflation. Aggregate demand may strengthen long before the supply side of the economy becomes more productive. Consumption increases, investment accelerates, credit expands and asset prices appreciate while firms are still struggling to integrate AI into their production processes. Inflation therefore emerges before the productivity boom itself.

From the perspective of a central banker, this creates a diagnostic problem. Rising inflation traditionally signals an economy operating beyond its productive capacity. Yet during an AI transition, inflation may instead reflect rational optimism about future technological progress. The economy appears overheated even though its productive frontier is simultaneously shifting outward.

The distinction matters because the appropriate policy responses differ dramatically. If inflation reflects excessive aggregate demand, higher interest rates remain appropriate. If inflation instead reflects temporary adjustment costs combined with forward-looking investment, tighter monetary policy risks suppressing precisely the capital formation necessary to realize future productivity gains.

The challenge is not that central banks have forgotten how inflation works. The challenge is that inflation itself begins to originate from multiple mechanisms operating simultaneously.

AI changes the Phillips Curve

For decades, economists have whether the Phillips Curve 鈥� the relationship between inflation and economic slack 鈥� has become flatter or steeper. Globalization, labor-market flexibility, inflation targeting and demographic change have all been proposed as explanations for its apparent instability.

AI an entirely new dimension to this debate.

Traditionally, firms adjusted prices relatively infrequently because gathering information was costly. Managers observed changes in wages, input prices or demand before deciding whether price adjustments were worthwhile. AI dramatically reduces these information costs. Algorithms continuously monitor inventories, transportation costs, customer demand, commodity prices and competitors鈥� pricing strategies. Prices can therefore respond almost instantaneously to changing economic conditions.

One possible outcome is a steeper Phillips Curve. Cost increases pass through to final prices more rapidly, making inflation respond more quickly to fluctuations in aggregate demand. Monetary policy becomes more powerful but also more volatile because firms react almost immediately to changes in economic conditions. An equally plausible outcome points in the opposite direction.

AI may increase market concentration by allowing the largest firms to exploit economies of scale in data, computing power and software development. At the same time, pricing algorithms continuously monitor competitors rather than focusing exclusively on firms鈥� own production costs. Strategic complementarities therefore become stronger. Instead of asking whether their own costs have changed, firms increasingly ask whether competitors have adjusted prices.

This subtle change alters inflation dynamics. Prices become determined less by marginal production costs and more by strategic interaction among algorithms.

Ironically, greater technological sophistication may weaken one of the most fundamental relationships in macroeconomics. Inflation becomes less sensitive to unemployment or output gaps and more sensitive to digital market structure.

The traditional Phillips Curve therefore faces a new source of instability 鈥� not because labor markets have changed, but because firms themselves have changed how they make pricing decisions.

Beyond demand inflation

Economists typically classify inflation into familiar categories. Demand-pull inflation arises when aggregate demand exceeds productive capacity. originates from adverse supply shocks such as higher energy prices or wage pressures. Monetary policy is largely designed around these distinctions. Artificial intelligence suggests the emergence of a third category.

During the early stages of AI adoption, firms experience rising implementation costs, organizational restructuring, software integration expenses, cybersecurity investments and workforce retraining. These expenditures increase production costs even though they ultimately raise future productivity. Inflation therefore originates from technological transition itself rather than from excessive demand or permanent supply constraints. This may be described as Transition Inflation.

Unlike conventional supply shocks, Transition Inflation is potentially temporary and investment-driven. Unlike conventional demand inflation, it reflects structural transformation rather than macroeconomic overheating. The same technological investment that temporarily raises inflation eventually reduces production costs and expands potential output.

Recognizing this distinction is essential for monetary policy. Responding aggressively to Transition Inflation may stabilize prices in the short run while reducing productivity growth in the long run. Ignoring it, however, risks allowing optimistic expectations to evolve into speculative financial excess. The policy challenge therefore becomes less about determining the correct interest rate than about identifying the true source of inflation.

Inflation becomes harder to diagnose

Artificial intelligence changes more than productivity. It changes the informational content of macroeconomic indicators themselves.

Traditionally, policymakers interpreted higher inflation, lower unemployment and stronger investment as evidence that aggregate demand had strengthened. In an AI economy, the same data may instead reflect technological restructuring, forward-looking expectations or temporary implementation costs. Identical economic statistics can therefore very different underlying economic conditions.

This observation helps explain why economists increasingly disagree about AI鈥檚 macroeconomic effects. Recent survey evidence shows relatively broad agreement that AI capabilities will improve substantially during the coming decade. The disagreement arises not from technological forecasts but from uncertainty about how those capabilities will translate into productivity, employment, inflation and financial markets.

For central banks, this uncertainty has important practical consequences. Monetary policy has always operated under imperfect information, but the information problem itself is becoming more complex. AI simultaneously changes production technologies, labor markets, pricing behavior, financial intermediation and expectations. Estimating the output gap or the natural rate of unemployment therefore becomes considerably more uncertain because the underlying structure of the economy is evolving continuously rather than remaining approximately constant.

This represents perhaps the most important implication of AI for monetary policy. Artificial intelligence does not simply generate new economic shocks. It changes the framework through which shocks are interpreted. As structural transformation accelerates, distinguishing temporary cyclical fluctuations from permanent changes in productive capacity becomes increasingly difficult. Diagnosis 鈥� not implementation 鈥� may become the central challenge of monetary policy in the AI era.

The AI financial accelerator

AI affects more than inflation and productivity. It also changes how credit is created, how risks are measured and ultimately how financial crises may emerge.

Historically, technological revolutions have always been accompanied by financial booms. Railways in the 19th century, electricity in the early 20th century, the internet during the 1990s and housing before the Global Financial Crisis all generated waves of optimism that encouraged investment far beyond what existing productive capacity could immediately justify. Financial markets have always been willing to price tomorrow鈥檚 productivity today. AI is unlikely to be different.

Technology companies are hundreds of billions of dollars to data centers, advanced semiconductors, cloud infrastructure, electricity generation and specialized networking equipment. Much of this investment reflects expectations about future profitability rather than realized productivity. Financial markets therefore become increasingly dependent on beliefs about technologies whose long-term economic effects remain highly uncertain. This creates what may be called the AI Financial Accelerator.

Under the traditional , rising asset prices increase collateral values, allowing firms to borrow more and invest further. Artificial intelligence strengthens this mechanism because optimistic expectations about future AI productivity immediately raise equity valuations, improve firms鈥� financing conditions and stimulate additional borrowing. Investment therefore accelerates before productivity itself has improved.

Unlike previous investment booms, however, AI investment is unusually concentrated. Building frontier AI models requires enormous fixed costs. The largest firms increasingly dominate access to computing power, proprietary data, specialized chips and engineering talent. Consequently, AI investment is financed not only through retained earnings but also through corporate bond markets, project finance, private credit, securitization and increasingly sophisticated financial structures.

Monetary policy therefore operates through a financial system that is becoming both larger and more interconnected.

When every bank uses the same brain

Artificial intelligence also transforms financial intermediation itself. Banks increasingly employ machine-learning models to evaluate borrowers. Insurance companies use AI to price risk. Asset managers rely on algorithms to construct portfolios. Trading firms execute millions of transactions through increasingly autonomous systems. Fraud detection, compliance monitoring, liquidity management and credit underwriting all become more efficient.

From an individual institution鈥檚 perspective, these innovations reduce risk. From the perspective of the financial system, however, the opposite may occur.

One of the lessons of financial history is that crises rarely emerge because individual institutions make irrational decisions. They emerge because many institutions make remarkably similar decisions at exactly the same time.

Artificial intelligence may unintentionally reinforce this tendency. Large financial institutions increasingly rely upon similar foundation models, cloud-computing providers, training datasets, optimization algorithms and risk-management software. Even institutions that believe they possess independent models may ultimately depend upon remarkably similar information and computational architectures.

The financial system therefore begins to resemble a biological monoculture. Agricultural monoculture is extraordinarily productive because every field contains the highest-yield crop. Until disease appears. Then every field becomes vulnerable simultaneously.

Financial markets face an analogous danger. Artificial intelligence may improve the quality of individual decisions while increasing the correlation of decisions across institutions. Diversification gradually becomes an illusion. Everyone believes they have different models. In reality, everyone may own different versions of the same model. The next financial crisis may therefore originate not from irrational exuberance but from excessive algorithmic agreement.

A new kind of stagflation

Perhaps the most surprising implication of AI concerns . For most economists, stagflation immediately evokes memories of the , when oil-price shocks simultaneously increased inflation and reduced economic growth. Supply contracted while production costs increased, forcing central banks to confront inflation without strong demand.

Artificial intelligence may generate a fundamentally different version of the same problem. During the transition toward AI-intensive production, firms face substantial implementation costs. New software systems must be integrated into existing organizations. Employees require retraining. Cybersecurity becomes more expensive. Computing infrastructure demands enormous investment. These adjustment costs temporarily reduce measured productivity even though they ultimately expand productive capacity.

Meanwhile, financial markets observe something entirely different. Investors anticipate extraordinary future profitability. Technology companies report massive investment plans. Equity valuations increase. Credit conditions remain favorable. Asset prices continue rising. Consequently, the economy may simultaneously experience rising production costs, elevated asset prices, expanding leverage and slowing measured productivity.

This combination differs fundamentally from traditional demand-driven inflation. It also differs from conventional supply shocks. Inflation originates from structural adjustment rather than overheating. Financial markets continue expanding because expectations remain optimistic. The result is AI-driven stagflation.

Unlike the stagflation of the 1970s, the economy is not suffering from a shortage of future productive capacity. Instead, it is temporarily paying the cost of creating that capacity.

Monetary policy meets structural change

This creates one of the most difficult policy environments central banks have ever faced. Suppose inflation rises above target while equity markets continue reaching new highs. Conventional policy would interpret this combination as evidence of excessive aggregate demand and accommodative financial conditions. Raising policy rates would therefore appear appropriate.

But suppose inflation instead reflects temporary implementation costs associated with AI adoption. Higher interest rates may reduce inflation modestly by weakening aggregate demand.

At the same time, they increase financing costs for AI investment, delay infrastructure construction, discourage research and development, and slow technological diffusion. The central bank may successfully reduce today鈥檚 inflation at the expense of tomorrow鈥檚 productivity.

The opposite mistake is equally dangerous. If policymakers assume every increase in inflation merely reflects technological transition, they may tolerate excessively loose financial conditions. Optimistic expectations become speculative valuations. Credit expands more rapidly than productive capacity. Eventually expectations adjust, asset prices decline and leverage amplifies the downturn.

The policy dilemma therefore becomes considerably more complicated than choosing the correct interest rate. It becomes a problem of distinguishing healthy technological optimism from financially destabilizing speculation.

Beyond the inflation target

For decades, monetary economists have debated whether central banks should 鈥渓ean against the wind鈥� by tightening policy when financial imbalances become excessive, even if inflation remains subdued.

Artificial intelligence gives this debate renewed urgency. The same technological optimism that raises long-run productivity may simultaneously generate excessive leverage and systemic risk. Monetary policy therefore confronts an increasingly uncomfortable trade-off. Supporting technological investment may encourage financial fragility. Restricting financial excess may slow innovation.

Price stability and financial stability, long treated as complementary objectives, may occasionally point toward different policy prescriptions. That possibility suggests that the central bank of the AI era will require broader analytical tools than those developed for industrial or even digital economies. Monitoring inflation alone will no longer be sufficient. Policymakers must also understand technological diffusion, credit creation, market concentration and the evolving architecture of financial networks.

Artificial intelligence changes not only the economy鈥檚 productive capacity. It changes the financial system through which monetary policy itself is transmitted.

The central bank of the AI economy

Artificial intelligence does not require central banks to abandon their traditional mandates. Price stability remains essential. Financial stability remains indispensable. What AI changes is not the destination of monetary policy but the map policymakers use to reach it.

For decades, central banking has relied on relatively stable macroeconomic relationships. The output gap approximated excess demand. The unemployment gap signaled labor-market slack. The Phillips Curve linked economic activity to inflation. Estimates of the neutral interest rate provided guidance for the appropriate stance of policy. None of these concepts disappears in an AI economy, but each becomes substantially more difficult to estimate because the economy鈥檚 underlying structure is itself evolving.

The distinction between cyclical and structural change has always been one of the defining challenges of macroeconomic policy. Artificial intelligence makes that distinction increasingly ambiguous. Productivity may temporarily decline while long-run productive capacity expands. Inflation may reflect implementation costs rather than overheating. Rising investment may signal either speculative excess or rational preparation for a more productive future. Financial markets may become simultaneously more efficient and more fragile as AI reshapes credit allocation and risk management. Economic indicators therefore become less informative because identical data can reflect fundamentally different underlying mechanisms.

This uncertainty has important implications for monetary policy strategy. Central banks have traditionally emphasized forecasting inflation one or two years ahead. AI suggests that diagnosis may become more important than forecasting. Policymakers must understand whether inflation originates from excessive demand, temporary technological adjustment, changing market structure or financial optimism. An incorrect diagnosis may produce a larger policy mistake than an inaccurate forecast.

Consequently, monetary policy will increasingly require a broader analytical framework. Traditional macroeconomic indicators should be complemented by measures of technological diffusion, AI-related investment, productivity dispersion across firms, computational infrastructure, financial leverage and market concentration. Monitoring semiconductor investment, electricity consumption by data centers, cloud-computing capacity, AI adoption across industries and the financing structure of digital infrastructure may become as relevant for monetary policy as unemployment or wage growth. The central bank of the future may therefore resemble not only a macroeconomic institution but also an institution capable of understanding technological transformation.

Rethinking monetary policy

Recent developments already point in this direction. Major technology firms have announced AI-related capital expenditures measured in the hundreds of billions of dollars, while governments increasingly view AI infrastructure as a strategic national asset rather than merely another private investment. Meanwhile, from the Federal Reserve Bank of Chicago suggests that economists expect only modest increases in baseline GDP growth despite assigning relatively high probabilities to substantial advances in AI capabilities, illustrating the considerable uncertainty surrounding the transmission from technological progress to macroeconomic outcomes. This gap between technological optimism and macroeconomic uncertainty may become one of the defining policy challenges of the coming decade.

The implications extend beyond monetary policy itself. AI may alter the interaction between central banks and fiscal authorities. Governments may increasingly finance investments in digital infrastructure, semiconductor production, electricity grids, education and cybersecurity to accelerate AI adoption. Industrial policy, competition policy and financial regulation may therefore become more closely intertwined with monetary policy than at any time since the postwar reconstruction period. Maintaining price stability in an AI economy may depend not only on interest-rate decisions but also on how effectively public and private institutions manage technological transition.

International coordination may become equally important. AI development is concentrated in a relatively small number of countries and firms, yet its macroeconomic consequences are global. Divergent rates of AI adoption may generate persistent differences in productivity growth, capital flows, exchange rates and equilibrium interest rates across economies. Central banks will increasingly operate in a world where technological leadership becomes a determinant of monetary conditions, much as energy prices shaped inflation during the twentieth century.

Perhaps the most important lesson is one of humility. Throughout history, central banks have often been challenged not because they lacked policy instruments but because they misunderstood the nature of the economic transformation underway. During the 1970s, policymakers underestimated the persistence of supply-side inflation. During the Global Financial Crisis, many underestimated the vulnerabilities accumulating within the financial system. During the pandemic, the distinction between temporary and persistent inflation again proved exceptionally difficult to identify. Artificial intelligence presents another such moment, although one driven not by crisis but by innovation.

A new paradigm for central banking

History also offers a useful reminder. Every major technological revolution has initially confused policymakers. Railroads transformed finance before they transformed transportation. Electricity changed factory organization long before productivity statistics reflected its benefits. The internet generated one of history鈥檚 largest equity bubbles before reshaping nearly every industry. AI is likely to follow a similarly nonlinear path. The greatest policy mistakes may therefore arise from expecting technological revolutions to appear neatly in conventional macroeconomic data.

Artificial intelligence should not be viewed simply as another productivity shock. It represents a structural transformation that simultaneously changes production, expectations, pricing behavior, financial intermediation and the transmission of monetary policy itself. These mechanisms interact in ways that make inflation harder to interpret, equilibrium harder to estimate and financial vulnerabilities more difficult to detect.

Central banks do not need a new mandate. They need a new analytical framework. The challenge of the AI era is not that inflation will disappear, nor that monetary policy will become obsolete. It is that the economy will increasingly evolve faster than the models used to understand it. The central bank of the future will succeed not because it reacts more aggressively than its predecessors, but because it learns more quickly.

Artificial intelligence is often described as a revolution in computation. For central banks, it may prove to be something even more significant. It is a revolution in economic interpretation. The defining question of the next decade will not be whether AI transforms the economy. It almost certainly will. The defining question is whether monetary policy can transform quickly enough to understand the economy AI creates.

[ edited this piece.]

The views expressed in this article are the author鈥檚 own and do not necessarily reflect 51吃瓜鈥檚 editorial policy.