1. The Ship Looks Fine from the Upper Deck
The AI revolution can look remarkably different depending on where you sit. From the upper deck, the picture is spectacular: machines can write, code, analyse, translate, tutor, design and increasingly perform tasks that once required highly educated professionals. Productivity rises while the cost of producing many forms of digital output falls.
But go downstairs. A tutor discovers that a student can obtain competent explanations from an AI for a small monthly subscription. A writer discovers that a client can generate a first draft instantly. A junior programmer discovers that software can produce code that once took hours to write. A researcher discovers that the first stage of research has become almost free.
The same technology that looks like abundance from above can look like falling economic value from below. That is the central tension of the AI economy.
2. When Knowledge Becomes Almost Free
For most of human history, knowledge was scarce. Learning mathematics required a teacher or books. Becoming a programmer required years of study and practice. Research required libraries, specialists and considerable time. Expertise had economic value partly because acquiring it was difficult.
The internet changed distribution. AI is changing production. An LLM can take enormous quantities of existing human knowledge and make them available through a conversation. The student does not need to find the book. The entrepreneur does not need to search ten websites. The programmer does not necessarily need to remember the syntax.
The important economic change is therefore not that knowledge itself has become worthless. It is that the scarcity premium attached to transmitting existing knowledge is being compressed.
If everyone can access an explanation of calculus instantly, knowing how to explain calculus is no longer as scarce as it was before. If everyone can generate a reasonable business plan, writing business plans becomes less valuable. If everyone can obtain competent coding assistance, routine coding becomes less scarce.
AI is not merely competing with human workers. It is competing with the economic scarcity that made many forms of human expertise valuable.
3. The Tutor’s Problem: Who Pays for Expertise?
Consider a simple example. A student pays a human tutor at 50X. Then an AI tutor becomes available at X.
The AI does not have to be better than the human tutor at everything. It only has to be good enough for enough of the student’s needs to change the student’s willingness to pay.
That is how technological substitution can work. The incumbent does not necessarily disappear. The price can simply fall.
A 50X service may become a 5X service. Then perhaps 2X. Eventually, only a smaller premium may remain for the human component: mentoring, accountability, motivation, personal attention or specialised expertise.
The same mechanism can operate in writing, translation, programming, research and other cognitive services.
The human has not necessarily become worse. The substitute has become cheaper.
4. AI Is Deflationary—but Who Gets the Purchasing Power?
This is where the AI discussion becomes an economic problem rather than merely a technology story.
Suppose AI makes production cheaper. That sounds unambiguously positive. But imagine that the same technology also reduces the income of the people whose labour has been replaced or commoditised.
AI productivity ↑
Cost of production ↓
while potentially:
Labour demand ↓
Labour income ↓
The first process creates abundance. The second can reduce purchasing power.
And that produces a question that deserves much more attention:
If machines produce more and more of what people need, who will have the money to buy what the machines produce?
An economy cannot run on production alone. It also needs purchasing power.
5. The Career Ladder May Be Breaking
For generations, the implicit bargain was straightforward: study hard, acquire knowledge, develop expertise, gain experience, become more valuable and earn more.
The ladder was never equally accessible to everyone, but it existed.
AI potentially attacks several rungs simultaneously. If information becomes abundant, knowledge becomes less scarce. If AI can perform routine expert tasks, some forms of experience become less valuable. If AI can perform increasingly sophisticated cognitive work, the premium attached to additional education may change.
The problem therefore goes beyond “AI will take jobs.” A job can disappear and another job can emerge. The more fundamental question is: When AI Beats Human Expertise is one related exploration of this shift.
What happens if the traditional mechanism for turning human learning into economic value becomes progressively weaker?
A young person may reasonably ask: Why spend ten years becoming an expert if an AI can perform much of the economically valuable part of that expertise for a small subscription?
That question could affect education, career choices and human ambition long before AI literally eliminates every job.
6. UBI May Solve Income Without Solving the Problem
Universal Basic Income offers one response. If machines replace large amounts of human labour, society could redistribute part of the resulting wealth to everyone. See also The Era of Universal Employment: A New UBI.
That solves an important problem: how does a person obtain purchasing power when employment becomes less reliable?
But income is not the same thing as economic participation.
A person can have enough money to survive while still asking: What am I becoming good at? What am I contributing? What can I accomplish? Where does my status come from? What do I own? What is my path upward?
A sufficiently generous UBI might address some of these indirectly. But a basic income by itself does not automatically create ownership, purpose, capability or a career ladder.
There is a difference between saying:
“You don’t need to work anymore.”
and saying:
“You still have a meaningful place in the economy.”
Those are not the same proposition.
7. Beyond UBI: UBE and Universal Basic Employment
Two additional ideas become relevant here.
Universal Basic Education would treat high-quality education as permanent economic infrastructure rather than merely preparation for a conventional career.
Its purpose would not be simply to teach people information that AI already knows. It would develop capabilities that remain valuable in an AI-rich world: curiosity, judgment, experimentation, physical-world skills, collaboration, problem formulation, creativity and the ability to use AI intelligently.
Then comes Universal Basic Employment.
The idea is not that governments should manufacture meaningless jobs so that employment statistics look healthy. It is that anyone who wants to contribute should have access to socially useful paid work.
That work might involve education, care, environmental restoration, local services, infrastructure, culture, research, community activity or tasks requiring human presence and accountability.
The objective changes from:
AI replaces you → government compensates you
to:
AI changes the economy → society preserves capability and participation.
That is a very different philosophy.
8. Universal Basic Equity: Who Owns the Machines?
But there is an even deeper question: who owns the machines?
Suppose AI and robotics eventually become the dominant productive assets. If a relatively small group owns those machines while everyone else receives income from those owners or from the government, the economy has essentially changed from a labour-income system to a capital-income system.
That makes ownership extraordinarily important.
This is the logic behind Universal Basic Equity. The ownership question is also explored in Universal Basic Equity: Who Owns the AI Future?.
The idea is simple:
People should not merely receive a share of the output of the productive system. They should have some ownership claim on the productive system itself.
That could take many institutional forms: broad share ownership, public funds, sovereign wealth structures, employee ownership, cooperatives or other mechanisms. The exact mechanism is a separate question.
But the conceptual shift is significant.
UBI asks: “How much should everyone receive?”
Universal Basic Equity asks: “Who owns the assets producing the wealth?”
The second question goes closer to the source of the income.
9. The Titanic Problem: The Lower Deck Feels the Water First
This is where the Titanic becomes a useful metaphor.
Everyone is technically on the same ship, but everyone does not experience the ship in the same way. Those closest to the flooding would encounter the consequences before someone comfortably seated elsewhere on the ship necessarily understood what was happening.
AI may produce a similar information asymmetry.
A person whose income has already been undercut by AI does not experience “AI productivity” as an abstract statistic. They experience it as fewer clients, lower fees, fewer opportunities and less bargaining power.
Someone who owns the technology may experience exactly the same development as higher productivity, lower costs and greater wealth creation.
Both are observing the same technological change. They are simply standing on different decks.
10. The Upper Deck May Not See the Leak
This creates a particularly dangerous possibility.
The people benefiting most from AI may have the strongest evidence that AI is working. Their businesses become more productive. Their employees become more efficient. Their products become cheaper to produce. Their assets may become more valuable.
From their position, the future can look extraordinarily bright.
Meanwhile, someone entering the labour market may see something entirely different. The skills that once provided a route into the middle class are becoming easier to reproduce. The first rung of the career ladder is becoming thinner.
The person at the bottom therefore receives a signal that the person at the top may not see:
“The old way of becoming economically valuable isn’t working as it used to.”
This is why economic statistics alone may not capture the full transition. People do not experience an economy as GDP. They experience it through their ability to earn, progress, own, participate and imagine a future.
11. Who Pays the AI?
There is another paradox hiding underneath all of this.
AI companies need customers. Customers need purchasing power. Purchasing power traditionally comes largely from income.
But if AI progressively replaces the labour that generates that income, the system has to find another mechanism.
The circular problem looks like this:
AI → productivity ↑
AI → labour requirement ↓
Labour income → potentially ↓
Purchasing power → potentially ↓
Demand → potentially ↓
So who ultimately pays for the AI?
Businesses can pay from productivity gains. Governments can redistribute income. Capital owners can spend their returns. New industries can create new demand.
None of those mechanisms is impossible. But the underlying distribution problem remains.
Production is only half of an economy. The other half is the distribution of purchasing power that allows people to claim the production.
12. The Real Question Is Not Whether AI Can Produce Abundance
AI may indeed make many things extraordinarily abundant. The broader productivity question is also examined in Are AI Productivity Gains Worth It?
Education can become cheaper. Software can become cheaper. Research assistance can become cheaper. Translation can become cheaper. Content production can become cheaper. The cost of cognitive assistance may fall dramatically.
That is the promise.
But abundance alone does not tell us who benefits from the abundance.
A society can have extraordinary productive capacity while distributing its ownership very unevenly.
That is why the central AI question may eventually move away from:
“How intelligent will machines become?”
toward:
“How will the economic claims on machine productivity be distributed?”
Technology determines what becomes possible. Institutions determine who gets a stake in what becomes possible.
13. The People on the Lower Deck May See It First
The Titanic analogy should not be interpreted as a prediction that the AI economy will inevitably sink.
The point is simpler: people do not experience technological change simultaneously.
The person whose professional income has already collapsed experiences the future differently from the person whose assets are benefiting from AI. The student wondering whether education is worth the investment experiences the future differently from the company reducing its training costs through AI. The tutor competing with an X-priced AI subscription experiences the future differently from the company selling that subscription.
The lower deck may therefore provide the earliest warning—not because its passengers understand AI better, but because they encounter the economic consequences first.
And that leaves the largest question of all:
If machines eventually become capable of doing most economically valuable work, what remains for humans to own, contribute, accomplish and aspire to?
Perhaps the answer is not simply UBI.
Perhaps it requires education, employment, participation—and ownership.
Because if AI produces the abundance while humans lose the ability to claim a meaningful share of it, the problem will not be that the world has too little wealth.
It will be that too few people have a stake in the wealth the machines create.
The Era of Universal Employment: A New UBI