When people consider AI’s threats, unemployment tops the list. This concern is valid, but it fails to distinguish AI from past technological advances, like mechanized textile production displacing hand-weavers, ATMs replacing bank tellers. The AI revolution is not “yet another technology advance”. To explain it, we first examine the role of human consumers in today’s economy.

Why human consumers are important (so far)

Modern economic theories center on consumption. If we look at the textbook description of economy, “GDP = household consumption + investment + government spending + net exports”. Globally, net exports cancel out. Investment and government spending, so far as we have seen, cannot sustain economic growth indefinitely. The US, Soviet Union, and most recently China all experienced periods where investment and government spending drove remarkable growth. But these phases ultimately end. China is still navigating its painful transition to a consumer-driven economy.

History leaves an impression that consumers—and of course, human consumers—will always matter. If AI displaces workers, unemployment soars and household consumption plummets, the entire economy will also suffer. Just like a textbook economic crisis, the government will intervene. Moreover, corporations should also have incentives to prevent it, echoing Walter Reuther’s quip to Henry Ford - Robots do not buy cars.

But the AI revolution may have different patterns. This brings my first claim:

From first principles, human consumption is not essential to the economy.

Historically, why do overproduction and overinvestment always trigger economic crises? The standard explanation is that consumption fails to keep pace. But this isn’t the first-principled answer. If consumption lags, why don’t companies simply invest more, produce more, and sell to each other? And if overall productivity keeps driving down inflation, money supply will increase and appear as business returns. The first-principle answer is: investment stops when it no longer generates returns. This cycle has never sustained itself historically because it breaks somewhere.

My further explanation of this answer is that, consumption matters because humans drive productivity growth, i.e., humans are the productivity amplifiers. For example, when cars were invented and scaled, they didn’t directly boost other sectors’ productivity. They boosted it indirectly, by increasing the mobility of humans. Without humans in the loop, no other amplifiers can sustain prolonged economic growth. That’s the reason so far when labor markets and human consumption were disrupted, productivity growth stopped and economy growth also stopped. No other “consumption” yet can replace the role of humans.

This is the first and by far the most fundamental economic flywheel over past centuries – technology frees up human brains, and the freed-up human brains power more technology advances. Government, academia, and venture capital enhance this flywheel, but humans remain the ultimate driving force. Keeping this flywheel spinning requires improving and incentivizing the majority of humanity. Hyung Won Chung articulates a similar view, using “leverage” to describe humans’ role in technological development, and he also believes AI will always be the augmenting role.

It should be noted that the above economic flywheel is different from the term “economic circulation”. In economic circulation, humans consume goods and produces goods. In the economic flywheel, humans consume the outcome of growth and produce more growth.

Post-consumer AI economy

Will human consumption remain as critical as today? In the first flywheel described above, humans consume the outcome of technology and further drive economy growth over the past centuries. Before we speculate the wild form of another economy, here is the second claim of this article:

AI consumption, currently classified as business investment or intermediate goods, has the potential to displace human consumption and sustain long-term economic growth.

As previously stated, humans’ unique capability lies in innovation and broadcasting gains across different fields. But this may no longer hold true. AlphaGo’s Move 37 was among the first demonstrations of AI innovation. Recent advances in drug discovery, mathematical proofs, and circuit engineering further blur the line between human and AI innovation. Moreover, Anthropic’s Project Vend demonstrates AI’s potential to handle business relationships traditionally managed by humans. We are witnessing AI gradually approach humanity’s holy grail: the ability to innovate and materialize innovations.

Could we see a second flywheel emerge in our lifetime? This time, AI would serve as the fundamental driving force of productivity growth, broadcasting gains across every field. This economic flywheel would no longer rely on humans as either producers or consumers.

If society mismanages the AI revolution, we may witness a dystopian future: Corporations not only don’t need to hire humans, but more importantly, corporations also don’t need to sell to humans. This transition begins with corporate layoffs and small business bankruptcies. New companies pivot toward AI infrastructure (AI’s consumption) and AI applications. As unemployment rises, central banks cut rates and governments distribute stimulus, but consumer markets barely keep pace as the AI-driven economy grows even faster. Ultimately, AI-powered corporations primarily trade with each other, forming a self-sustaining economy that dwarfs the consumer market.

A few winners would emerge from this transition, including business owners and elite AI developers. However, even their situations may not last long:

• “Owning” the business through equities is a human consensus that requires both humans and consensus.
• Consistently producing exceptional talent requires a large, educated population pool.

Warning signs of this transition

The scenario described above may seem like a distant future when super intelligence finally arrives. However, this economic transition requires neither AI superintelligence nor complete human replacement. Rather, it represents a gradual process accelerated by AI’s increasing generalization and declining costs.

If we treat this AI revolution in the same way as past technology advance, i.e., applying traditional measures designed for overproduction or overspeculation crises, several warning signs may emerge:

• High unemployment and low inflation despite robust GDP growth, violating established empirical relationships like Okun’s Law.
• Declining household consumption in the overall GDP
• Expanding corporate profits driven by higher business-to-business revenues and reduced personnel expenditures

How to prevent such a transition? As Yuval Noah Harari indicates in Sapiens: A Brief History of Humankind, the most efficient groups ultimately outcompete others, even when this efficiency diminishes individual welfare within those groups. If AI proves genuinely superior in productivity and cost, companies and nations that resist its adoption will inevitably lag in the competition. Halting this transition by policies may prove impossible even if the majority agree to halt it.

On the other hand, it is easier to list things that facilitate this transition, which is the norm in many countries:

• Tax and social security frameworks that favor capital investment over paying wage
• Monetary policies featuring lower interest rates and quantitative easing during periods of high unemployment and low inflation
• Authoritarian government with high tolerance for public dissatisfaction

We need to ensure the future is bright and humans are a part of the future. I wish I was just being paranoid, and the society will automatically adapt to AI and harness it for broad human benefit. But if not, this analysis serves as yet another warning of the fundamental shift in this AI revolution.