A longer piece this week. The topic earned it.
I have been inside two technological revolutions. The dot-com era in the United States. The mobile revolution across multiple countries, from inside the telecommunications industry. Both times I watched the predictions fail in both directions — the optimists wrong about the speed, the pessimists wrong about the permanence, almost everyone wrong about who would actually bear the cost.
This time something feels structurally different. Not faster. Different in kind. And I want to try to explain why — carefully, without performing intelligence, because the ideas underneath this are genuinely important and deserve to be understood rather than admired from a distance.
· · ·
Start with the economic argument — because it is the foundation of everything else.
Daron Acemoğlu is a Turkish-American economist at MIT who won the Nobel Prize in Economics in 2024. His book Power and Progress, written with Simon Johnson, asks a question that sounds simple but lands hard: technology creates economic growth — but who actually captures that growth? Is it shared broadly, or does it concentrate at the top?
His answer, grounded in a thousand years of economic history, is uncomfortable. Shared prosperity from technological progress was never automatic. The Industrial Revolution ultimately made workers better off — but not because it was designed to. It happened because of countervailing forces: labour unions, legislation, organised political power that wrested the direction of technology from a narrow elite and distributed the gains more broadly. Left to itself, technology has historically concentrated at the top. The distribution happened when people organised to ensure it did.
His specific assessment of AI is more precise — and more unsettling. Acemoğlu calls much of current AI "so-so technology." Not because it is unimpressive. Because it performs at best only a little better than a human at most tasks — but it is considerably cheaper. You do not need AI to dramatically outperform a human to replace them. You only need it to be cheaper. That distinction changes the entire economic calculation. Companies do not need a revolution. They need a cost reduction. And that cost reduction comes directly out of labour.
|
The Question Nobody Is Asking Technology creates growth. But who captures it? Acemoğlu's answer from a thousand years of history: not workers, unless they organise to ensure they do. AI does not need to be dramatically better than a human to replace them. It only needs to be cheaper. And right now, it is considerably cheaper. The prosperity from this wave will not distribute itself. |
A CEO at a major bank recently described AI-driven layoffs not as cost-cutting but as replacing "lower value human capital" with financial capital. He later apologised for the phrasing. The phrasing was accurate. That is how the system sees it — labour as an input with a value that can be compared against alternative inputs. When the alternative becomes cheaper, the comparison goes one way. Always.
· · ·
Previous technological revolutions created gaps — they did not create exclusions. In the automobile era, if the United States had a Ferrari and your country had a VW Beetle, you could still get from A to B. You were at a disadvantage, not locked out. The technology spread, slowly and unevenly, but it spread. A country behind could catch up by adopting what the leaders built.
AI is winner-takes-all. The advantages compound — more data produces better models, better models attract more users, more users generate more data. The compute infrastructure, the energy requirements, the human capital needed to operate at the frontier: only two countries have all of these simultaneously at the required scale. The United States and China. Everyone else is working with a fundamentally different set of resources.
For the last seventy-five years, developing countries grew by importing technology and deploying it with cheaper labour. South Korea, Taiwan, Turkey, Romania — economies that were agricultural within living memory became industrial and then knowledge economies through this mechanism. AI removes the labour advantage. If cognitive tasks can be done by systems that do not need wages, the catch-up growth model ends. Countries that have not yet caught up may never catch up. The K-shaped recovery visible within developed economies — where the top pulls further from the middle, the middle from the bottom — becomes the permanent condition of the global economy.
The post-war international order was built on global public goods — open sea lanes, a shared financial system, relatively free trade — maintained by American power but theoretically available to all. That architecture is now being deliberately constrained. AI chip exports restricted. Technology transfer limited. China pursuing parallel dependency-building strategies. AI is becoming the next layer of this geopolitical structure — not a neutral tool any country can adopt, but a strategic asset controlled by two powers with every incentive to limit access. Universal basic income funded by AI productivity is a fantasy for most governments. They will not have the productivity to distribute.
And there is one more dimension that deserves to be named directly. The large language models powering current AI systems were trained on the accumulated writing, knowledge, and creativity of billions of human beings — books, articles, conversations, code, art — almost none of whom were compensated or consulted. Humanity's collective intellectual heritage was taken, processed, and converted into privately owned systems of enormous commercial and strategic value. The important addition: it is not only corporations doing this. The US and China are doing the same thing at state level. The line between state and company is blurring. Neither the traditional left nor the traditional right critique is adequate to describe what is happening — because it is both, simultaneously, reinforcing each other.
· · ·
Now the identity dimension — because the economic and geopolitical arguments are only half of what is happening.
Matthew Crawford has a PhD in philosophy from the University of Chicago. He worked briefly at a Washington think tank, writing policy papers that nobody read about topics nobody could verify. Then he quit and opened a motorcycle repair shop. His book Shop Class as Soulcraft explains why: when you fix a motorcycle, the motorcycle tells you honestly whether the job is done or not. No performance review required. No manager to approve the outcome. The work itself provides the feedback. That relationship between person, skill, and real-world result is, Crawford argues, one of the primary ways human beings build a coherent sense of self — and it is exactly what the shift from craft to knowledge work, and now from knowledge work to AI-assisted work, is progressively removing.
Jean-Paul Sartre argued that human beings arrive with no predetermined essence. We build who we are through the choices we make and the work we take on. Unlike a chair — which is made for a purpose, its essence defined before it exists — a human being defines themselves through action. This philosophical position has enormous practical implications: the opportunity to work, to build skill, to master something, to be useful and know you are useful, are not just means to income. They are the primary materials most modern humans use to construct identity.
For two centuries, the Industrial Revolution opened a form of self-definition that aristocratic societies had denied most people. You could become something through work that you were not born into. What Crawford mourned in the motorcycle shop — the loss of craft identity in knowledge work — AI is now doing at scale to knowledge work itself. The answer to "what do you do?" is becoming unstable. And "what do you do?" has been, for two centuries, the primary way most people answer "who are you?"
There is a battle happening at a level most people are not watching — and the fact that it has escalated to religious language tells you something about the stakes.
Peter Thiel — co-founder of PayPal, early Facebook investor, one of the most influential figures in Silicon Valley — has been arguing publicly, including in lectures delivered in Rome, that AI regulation is the work of the Antichrist. Not metaphorically. As a theological claim. The Vatican responded with a full papal encyclical — Magnifica Humanitas — examining AI through the lens of Catholic social teaching, in the tradition of the 1891 document Rerum Novarum which addressed the social disruptions of the Industrial Revolution.
Both sides have concluded that the stakes are too high for ordinary political or economic debate. When the language shifts from regulation and markets to salvation and damnation, something has been recognised: this is not a technology question. It is a question about what kind of world human beings want to inhabit and who gets to decide.
The ancient Greeks had a concept for the kind of life worth living regardless of economic function: eudaimonia — flourishing through the full exercise of distinctly human capacities. Reason. Contemplation. Connection. Creation. The question they were living inside, the one AI is now posing to the rest of us: what does a human being do when not consumed by utility?
· · ·
My daughter is almost six. She draws pictures that look nothing like what she intended and holds them up with complete satisfaction. She dances without checking whether she is doing it correctly. She answers "who are you?" without hesitation, without reference to any role or function — with the full confidence of someone who has not yet been told that the answer needs to be justified by what she produces.
What world is she growing into? Here is what the evidence says, honestly:
The economic disruption is real and will be uneven — hitting cognitive workers sooner than most expect, removing the catch-up growth ladder for developing economies, concentrating the gains in fewer hands unless countervailing forces organise to prevent it. Acemoğlu's historical point matters here: this is not a law of nature. It is a choice being made, and choices can be made differently — but only if enough people understand what is being chosen and by whom.
The geopolitical concentration is real — AI is becoming a strategic weapon controlled by two states, with most of the world's governments increasingly dependent rather than sovereign in this domain.
The identity crisis is real — and it will arrive for millions of people without the preparation time or financial cushion that makes the question navigable rather than devastating.
What is genuinely open is the cultural and philosophical response. Whether individuals — and eventually societies — develop a robust answer to who we are when we are no longer primarily defined by what we produce. The religious traditions have answers. The philosophical traditions have answers. Crawford found his answer in a motorcycle shop. My daughter has not yet lost the question.
The people who navigate this best will not be those who find the next useful function fastest. They will be those who had already started building an identity that does not depend entirely on what they produce — who know who they are in the silence between roles, before the question arrives with urgency.
That work is available right now. Before the urgency arrives. That is the only practical thing worth saying at the end of an article like this.
|
This week's insight AI is not just an automation event. It is removing the primary tool most modern humans use to build identity — useful work. The economic disruption will be real and uneven. The geopolitical concentration will limit most countries' options. The identity crisis will arrive for millions without preparation time. The people who navigate this best will not be those who find the next function fastest. They will be those who already know who they are when the function is gone. Start building that now. Not because the crisis is coming. Because the answer is worth having regardless of what comes. |
If this landed, there is one piece every week at stillnavigating.com — honest writing about the stuck middle, from someone still in it.