Average Is Over: Why AI Won’t Upend the Social Order

Taleb is back with another round of “AI cope:”

Two weeks ago I wrote that predictions of widespread AI job loss often amount to a coping mechanism. His tweets can be interpreted in two ways: as what he thinks will happen, but also as what he wants to happen.

A lot of people want to see an AI crisis, a bubble, or the obsolescence of white-collar jobs (including math and high-status academic jobs) in order to reconcile themselves to their own mediocrity. The cognitive dissonance is especially telling. Many of these people hold both beliefs at once or alternate between them depending on their mood or the news cycle: that the AI buildout is a bubble that is unsustainable and will pull the US economy into crisis, and/or that AI will lead to mass joblessness, necessitating a UBI and/or bringing about societal upheaval (whew, that was a lot to say).

You cannot have it both ways: AI cannot be both vaporware and economically consequential. AI cannot be so productivity-enhancing that it creates an unemployment situation rivaling the Great Depression, but also be a bubble that has held the US economy hostage. If they were honest, they would say, “This really has nothing to do with AI. I am envious of others’ success, and I want a UBI and higher taxes to spread the wealth. Or an economic crisis, which would have the same effect of reducing the wealth and influence of business elites and others whose status and success remind me of my own inadequacy.”

Envy and resentment are at the root of many people’s beliefs, which is how you get these contradictory views. For example, those who predicted in 2016 that Trump would lead America into civil war or “undermine democracy.” When Trump lost in 2020, democracy suddenly “came back to life,” only to die again in 2024. Or election interference was a huge concern in 2016, but then those same bad actors disappeared in 2020, only to resurface in 2024. Maybe it has nothing to do with democracy or elections, and you just hate Trump. That would at least be honest.

Or take the claim that Covid was worse than the Spanish Flu, in order to justify the lockdowns and shutdowns even though for 99.9% of people Covid either produced no symptoms or amounted to a bad cold. Or, in 2008, that capitalism was dead. The same predictions of the “end of capitalism” are being made about AI. For these people, AI will finally bring about the crisis or “social reshuffling” that 2008 and Covid failed to deliver.

As for categories of jobs hurt by AI, I can only think of the following examples:

-term-paper writers (such as for college students)
-graphic designers
-photographers (such as stock photographers)

These are tiny, inconsequential industries relative to the behemoths of tech, healthcare, advertising, retail, finance, or insurance. Chegg and Getty Images are worth less than $100 million combined, while Apple, Alphabet, and Meta alone have a combined valuation of over $10.5 trillion. What is often blamed on AI isn’t out of the ordinary. People keep reaching for examples to blame on AI, when the evidence is weak to nonexistent. For example, from ZeroHedge:

It’s like, “Breaking news: companies sometimes have large layoffs.” This is hardly unprecedented. Consulting Claude, I found many examples of large companies reducing their headcount by 5% or more:

For American Express and Barclays above, the justification is the typical vague “restructuring,” with AI not singled out. It’s way premature to read this as being directly due to AI. Anything that involves working directly with people or with one’s hands is especially hard to automate. As society becomes wealthier, demand for individualized services increases, from healthcare to finance and education. This means more labor. Examples include asset management, college admissions counseling, retirement planning, and individualized healthcare.

Insurance is especially labor-intensive, from dealing with multiple parties to customer service (such as people finding it hard to sign up or reset passwords), pricing premiums, and so on. I recall signing up for a bare-bones policy and still spending time on the phone talking to an actual human, not an automated system. What happens when someone needs to file a claim? They make a call and someone answers. Now multiply this millions of times. How does one possibly automate this when the circumstances and policy are unique to each individual, and you have counterparties trying to lower the payout?

With the AI and big tech boom—from Meta and Google to Anthropic and OpenAI—Tyler Cowen was wrong about the “great stagnation.” But he was indisputably right that “average is over.” In agreement with Tyler, my thesis is that AI magnifies innate disparities in human capital rather than leading to the more egalitarian outcomes many are hoping for. AI creates abundance in some things like knowledge, but tangible scarcities and individual differences of status remain: The smart and successful will continue to pull ahead, even as AI becomes more capable.

High-status jobs and signifiers of status refuse to yield to AI. It has been nearly four years since LLMs saw widespread commercial adoption, and despite enormous breakthroughs in subsequent models, the status quo, or “social order,” remains intact. Under AI, a bigger (in terms of GDP) and more profitable economy (in terms of corporate profits) means greater returns to outlier ability and talent at the individual level. Huge, trillion-dollar companies will use AI, guided by highly paid human experts, to widen their advantage and lead. Indeed, wealth inequality has only widened despite AI (Elon Musk is now the first trillionaire).

Historically, this is how it has always been, from the Industrial Revolution to the Information Age. Every technological breakthrough has brought with it a rising yet still inequitable tide. As with the rise of mega-sized tech companies, huge salaries and “net worth” coincided with economies of scale and moat effects. And this has proven sustainable, too. It didn’t suddenly come crashing down just because some overrated experts or the media called it unsustainable or a bubble. From my post, “AI Changes Everything–and Nothing:”

I agree that AI has been transformative in many respects, but at the same time, much will remain unchanged. That’s why I position myself as something of an “AI incrementalist.” I don’t expect a crisis, mass unemployment, or a sudden “singularity” at some arbitrary date (which keeps being pushed outward when the awaited paradigm shift doesn’t happen). Instead, I anticipate steady sector-specific productivity gains, with most things staying broadly the same–just augmented by increasingly powerful AI tools.

I also agree with Freddie deBoer, who has written extensively about this, that “human life in 2026 appears to be almost entirely similar to human life in 2021.” Outside of math and coding–which has gotten tons of media attention–he is right. Macroscopically, society will be the same. Cars still look like cars. Buildings still resemble buildings. The Cybertruck, although derided for being ugly, is at least the closest thing to an actual physical change. Economically, GDP hasn’t changed much–in the same 2-5 percent band of “real growth.” The unemployment rate remains within fractions of a percent of historic lows, as per BLS data.

There will also be no post-scarcity insofar as healthcare, tuition, insurance premiums, housing, or gas prices are concerned, as AI does not address those. Also, under an AI-led economic boom, we would expect more asset and affordability inflation, not deflation. My post from over three years ago, “AI is not going to make things cheaper (hidden fees and costs will persist),” has aged especially well given the so-called “affordability crisis.” Sure, the corpus of human knowledge is at your fingertips, but healthcare will always be expensive.

Moreover, there won’t be an economic reset or, more modestly, a UBI paid for by AI windfalls, as many are hoping or expecting. The absence of political will for such reforms, and a forever-divided House, means nothing will change much. Such surpluses will just go to the private sector and, indirectly, to the public in the form of higher tax revenue. But there will be no action by policymakers to address affordability. Your only options are to invest in stocks and/or “earn more money.”

Higher ed, such as elite colleges, will survive despite AI, taking the necessary steps to prevent AI cheating so their degrees are not devalued. When employers are inundated with applications (a picture unlikely to change with AI), degrees are still a useful filter at scale. Alluding again to “average is over,” Harvard’s status has only grown in recent years, not diminished, even with AI. With more content produced by AI, elites can extend their relative lead by acting as curators and gatekeepers (e.g., deciding “What is important?”).

Even if we assume AI somehow leads to high unemployment, history shows society will not fall apart. “Will the pitchforks come out if AI leads to job loss?” is often asked. Nothing of that sort will happen. Economically, the ’30s were worse, and there was no mass upheaval. Americans are much more complacent now than even then. People aspire to get rich, as on TikTok and in “hustle culture,” not to hang the rich.

A second common concern is the far left voting for socialism. Despite the rise of left-wing populism in some pockets of the US, all the trends point toward increasing wealth creation and the “ownership society,” such as surging stock and home prices and incomes. Tech has created more billionaires over the past 25 years than any other industry in history. If capitalism is defined as a system favoring the ownership of capital, it is thriving (maybe I will spin this off into a future post).

The more that is at stake, the greater the returns to elite human capital to maintain that lead. And we’ll see this everywhere in society where individual differences matter or persist, from large white-collar salaries to creative endeavors to sports (pro sports salaries are bigger than ever). Overall, the more we expect AI to change society, the more we’re continuously reminded of how much things stay the same. Maybe we need to just learn to accept this rather than keep hoping that AI will create the change that we are unwilling or unable to bring about for ourselves.