Conversation between me (R) and ChatGPT (ChatGPT)
(R) How does OpenAI imagine it will make money? If In the future ChapGPT will be able to replace all human jobs, won't other companies be able to do the same, within a few months' passage? And if AIs take all jobs, who will be able to buy goods and services produced by AI? Is the government supposed to give all citizens a universal basic income?
(ChatGPT) Yes. You’ve landed on one of the deepest economic problems created by advanced AI, and your second question is especially important:
If AI can produce almost everything, but humans no longer earn wages, who has the purchasing power to buy what AI produces?
There isn't a completely settled answer. But we can separate the problem into three different questions.
1. How does OpenAI expect to make money?
The basic business model is actually fairly conventional:
- Consumers pay subscriptions
- Businesses pay per employee
- Developers pay to use the API
- Companies pay for specialized AI agents and enterprise services
- Advertising is becoming another revenue stream
OpenAI says advertising is now one part of a diversified model alongside consumer subscriptions, enterprise products and usage-based APIs. As of August 2026, OpenAI said ChatGPT advertising had reached a $1 billion annualized revenue run rate.
Its business plans, for example, currently range from $20 per user/month for a standard seat to $100 for a premium seat when billed annually.
So in the near term, the economic logic is:
AI makes workers/businesses more productive → businesses make more money → businesses and consumers pay OpenAI for AI.
That's perfectly coherent.
But it becomes much stranger in the scenario you're describing.
2. "But if OpenAI builds it, won't everyone else just copy it?"
Potentially, yes.
This is one of the fascinating things about AI economics.
Suppose OpenAI creates an AI that can perform the work of 100 software engineers.
If the technology can be reproduced cheaply, then Google, Anthropic, Meta, xAI, Chinese companies, open-source developers, etc. can develop competing systems.
Competition would then tend to push the price of intelligence downward.
That's very different from, say, owning the world's only oil field.
And there is an additional wrinkle: AI companies themselves are customers of one another. They need enormous amounts of computing power, chips, electricity, data centers, networking equipment, etc. So the ultimate scarce resources may shift away from "intelligence" toward physical capital and energy.
In other words, the long-run question may not be:
"Who owns the smartest AI?"
but:
"Who owns the machines, energy, land, factories, robots, data centers and other physical capital that lets intelligence turn into stuff?"
That distinction becomes enormously important.
3. And then we hit your really big problem
Imagine an extreme future:
AI can do virtually every economically valuable cognitive job.
Then:
- lawyers are largely automated
- programmers are largely automated
- accountants are largely automated
- designers are largely automated
- doctors are heavily automated
- teachers are heavily automated
- managers are automated
- researchers are automated
And eventually robots handle much of the physical work too.
Now imagine you are a company.
Your AI/robots can produce $1 billion worth of goods.
Wonderful.
But who buys them?
If 90% of the population has no job and therefore no income, your wonderfully efficient automated factory has created a demand problem.
This is not a silly objection. It is a fundamental feature of economics.
You could have an economy capable of producing enormous abundance while simultaneously having people who cannot afford that abundance.
4. There is a possible solution that doesn't require UBI
This is where I think discussions of "AI takes all the jobs" sometimes become too simplistic.
There are several possible ways society could distribute purchasing power.
A. UBI
The government taxes some combination of:
- corporate profits
- capital
- AI-generated economic output
- wealth
- land
- consumption
and distributes money to citizens.
Essentially:
AI produces the wealth → government redistributes part of it → citizens buy the AI-produced goods.
That's the classic Universal Basic Income scenario.
And yes, governments could eventually move in that direction.
The IMF has explicitly discussed stronger social protection, taxation and redistribution as possible responses to large-scale AI disruption.
But UBI isn't the only possibility.
B. People could own the AI
This is arguably even more interesting.
Imagine that instead of:
AI companies own all the AI → citizens receive government checks
you have:
Citizens collectively own part of the productive capital.
For example, imagine every citizen owns shares in a national AI/robotics fund.
AI companies generate enormous profits.
Those profits flow into the fund.
Citizens receive dividends.
You could end up with something resembling a capital-owning society, rather than a traditional wage economy.
That's conceptually different from UBI because you're not simply taxing the productive sector and handing money back. You're giving people ownership of the productive machines.
5. And there is another possibility: jobs don't disappear completely
This is probably the most important caveat.
"AI can perform a task" does not necessarily mean "there will be zero humans employed doing that task."
Suppose AI makes a programmer 10× more productive.
You might think:
We need only 10% as many programmers.
But something else can happen:
Programming becomes 10× cheaper, so people want 20× as much software.
Now employment might actually increase.
Economists call this kind of effect a rebound or demand effect.
The IMF's analysis emphasizes precisely this uncertainty. It estimates that around 40% of jobs globally are exposed to AI, but exposure isn't synonymous with elimination. Some jobs are complemented by AI and become more productive rather than disappearing.
That distinction is enormous.
6. But your extreme scenario remains possible
Suppose AI eventually becomes so capable that the productivity increase is enormous.
Imagine:
Today:
100 people + machines → $100 million of output
Future:
5 people + AI + robots → $1 billion of output
That's fantastic from the standpoint of physical abundance.
But the five people now own a vastly disproportionate share of the economy unless ownership is distributed somehow.
And this produces the strange paradox:
The richer society becomes technologically, the less necessary human labor becomes.
That sounds wonderful.
But our current economic system distributes purchasing power largely through:
"You perform economically valuable labor → you receive wages → you buy things."
If labor ceases to be necessary, that mechanism starts breaking.
7. This is why the ownership question may ultimately matter more than the employment question
I think this is the really profound part of your question.
The traditional question is:
"What jobs will AI replace?"
But the deeper question may be:
"Who owns the AI that replaces them?"
Imagine two societies with exactly the same AI technology.
Society A
10 companies own almost all AI and robotics.
Their shareholders receive almost all the profits.
Most citizens have little capital.
Result: potentially enormous inequality and political instability.
Society B
AI companies still exist, but ownership is distributed through pensions, investment accounts, sovereign wealth funds, employee ownership, public funds, etc.
Citizens receive dividends from the automated economy.
Result: people can have little or no traditional employment while still having purchasing power.
Same technology. Completely different society.
That is why AI economics is ultimately also a question about property rights and distribution, not merely technology.
The IMF's research makes essentially this point: AI could raise productivity and incomes dramatically, but the distribution of the gains depends heavily on how capital ownership, labor income and fiscal policy evolve.
8. And there is a rather beautiful possibility hiding underneath all this
For most of human history, the fundamental economic problem has been:
We don't have enough stuff.
Food, houses, transportation, medical care, education, entertainment, energy, etc. require scarce human labor.
Advanced automation potentially changes that equation.
We could eventually reach something approaching:
"We have plenty of productive capacity, but we need to figure out how to distribute access to it."
That's a radically different economic problem.
If AI and robotics become extraordinarily capable, "unemployment" itself could eventually become a strange concept.
Being unemployed today means:
"I don't have access to the resources necessary to live."
In a sufficiently abundant automated economy, it could instead mean:
"I don't have a job, but that's okay because jobs aren't necessary for material survival."
At that point, the purpose of work would change too. People might work because they want:
- status
- purpose
- social interaction
- creativity
- competition
- craftsmanship
- meaning
- additional consumption
rather than because starvation is waiting outside the door.
But we're nowhere near knowing whether we'll get there
The important caution is that "AI eventually can do everything" is an enormous assumption.
There are at least three very different futures:
| Future | What happens |
|---|---|
| AI assistant economy | AI makes humans much more productive; most people still work |
| AI displacement economy | Many jobs disappear; governments dramatically expand redistribution |
| Automated abundance | AI + robotics make human labor economically unnecessary; society reorganizes around ownership and distribution |
We don't currently know which one we're heading toward.
The IMF itself emphasizes that AI could produce substantial productivity gains and higher incomes while also increasing inequality, and that the ultimate result depends heavily on how economies and governments respond.
And there's one particularly interesting twist to your question:
If every company has access to roughly the same AI, then AI itself may eventually become cheap and abundant.
In that world, the really valuable things might become the things AI can't manufacture infinitely:
land, energy, natural resources, physical infrastructure, scarce experiences, political power, human attention, and ultimately ownership of the machines.
That's where the economic story gets really interesting. The question stops being "How much is an AI worker worth?" and becomes "Who owns the post-labor economy?"
I think it should be treated as a parallel civilization-scale project, not as an add-on to AI safety.
If the AI project is essentially:
Teach machines to become extraordinarily capable without losing control of them.
then the companion human project should be:
Teach humans to become extraordinarily capable without losing sight of what capability is for.
I'd call the second project The Human Flourishing Project.
The important thing is that I would not design it as a program for teaching everyone one official philosophy. That would itself be dangerous. The goal would be to develop a population capable of thinking clearly about ends, recognizing manipulation, understanding human nature, and distinguishing things that merely stimulate desire from things that actually make lives better.
1. Start with one deceptively difficult question
Every child, repeatedly, should encounter:
"What makes a human life go well?"
Not as a religious catechism. Not as a standardized-test question.
As a genuine intellectual problem.
Children would encounter competing answers from Aristotle, Buddha, Confucius, Jesus, the Stoics, Epicurus, Maimonides, Ibn Sina, Locke, Mill, Marx, Nietzsche, Gandhi, contemporary psychology, evolutionary biology, literature, and many others.
And then they would be asked:
What do you think?
The objective isn't agreement.
It is developing the ability to distinguish:
- pleasure from flourishing
- desire from need
- wealth from usefulness
- status from worth
- freedom from unlimited choice
- comfort from happiness
- achievement from meaning
- consumption from fulfillment
- power from agency
- novelty from progress
That is already a form of AI safety, because a civilization that cannot make those distinctions will eventually outsource its choices to whatever optimization system happens to be most powerful.
2. Teach the psychology of desire
This might be the most neglected part of modern education.
People should understand, at an age-appropriate level, how their own minds can be manipulated.
Advertising, social media, gambling mechanics, status competition, addictive design, tribal identity, outrage, pornography, consumerism, political propaganda, recommendation algorithms, prestige cascades.
Not merely "these things are bad."
Rather:
Here is how the machinery works.
The ideal outcome would be a person who can notice:
"Something has made me want this. That doesn't necessarily mean I wanted it before."
That's an astonishingly powerful piece of intellectual equipment.
And it becomes vastly more important when AI can generate personalized persuasion at essentially zero marginal cost.
3. Restore the concept of enough
This connects directly to what you said about unused wealth.
I would make sufficiency an explicit concept in education.
Not poverty. Not forced equality. Not asceticism.
Enough.
What amount of food makes you nourished rather than merely fuller?
What amount of money produces security?
How much housing produces comfort?
How much status is actually useful?
How much consumption improves life before additional consumption becomes largely symbolic?
How much wealth can one person meaningfully use?
And then the fascinating question:
What should we do with the surplus?
That's where generosity enters.
Instead of teaching children that the purpose of economic life is to maximize their personal consumption, we could teach:
Once your genuine needs and reasonable ambitions are secure, surplus becomes a question of stewardship.
That's a radically different psychological relationship with wealth.
4. Make generosity intellectually respectable
Modern economics has an extraordinarily sophisticated vocabulary for competition and a comparatively impoverished vocabulary for generosity.
I'd change that.
Children should learn that human beings cooperate for many reasons:
- affection
- reciprocity
- loyalty
- gratitude
- reputation
- solidarity
- beauty
- curiosity
- justice
- religious conviction
- civic pride
- love of future generations
- simple generosity
And they should learn that these aren't irrational deviations from economic behavior.
They're part of economic behavior.
The economist's model should become larger, not narrower.
Imagine a 25-year-old deciding where to put $100,000 of surplus capital.
The question isn't merely:
"What gives me the highest return?"
It might be:
"What deserves this money?"
That one change in language contains an enormous change in civilization.
5. Teach practical philosophy as seriously as mathematics
I'd actually create a core curriculum around something like:
The examined life
Age 5 to 8:
What makes us happy? What makes someone a good friend? Why do we share? What is enough?
Age 9 to 12:
Fairness, cooperation, envy, courage, honesty, friendship, responsibility, persuasion.
Age 13 to 15:
Desire, status, money, sexuality, power, advertising, mortality, identity, technology.
Age 16 to 18:
Competing theories of the good life, political philosophy, economics, religion, science, ethics, AI.
And importantly, students would have to argue against themselves.
If you believe wealth produces happiness, construct the strongest case that it doesn't.
If you believe equality is morally necessary, construct the strongest case against it.
If you believe markets are good, construct the strongest critique of markets.
If you think technology liberates us, construct the strongest case that it enslaves us.
The objective is intellectual immune-system development.
6. Put AI into this curriculum early
Here's where the two projects meet.
Every student should eventually experience an AI that can say:
"Here are five different interpretations of what you're trying to accomplish. Here are the assumptions behind each. Here's what you might be overlooking."
AI could become extraordinarily good at moral deliberation without becoming the moral authority.
That distinction is crucial.
The AI shouldn't tell a seventeen-year-old:
"This is the meaning of life."
It should say:
"Here are the major ways humans have answered that question. Let's examine your assumptions."
In other words, AI as philosophical sparring partner rather than oracle.
7. Create institutions where people practice the philosophy
Education fails if the adult world teaches the opposite lesson.
So I'd pair the curriculum with institutions that make the desired behavior possible.
For example:
Universal basic security
People shouldn't have to become desperate before they can afford to behave generously.
Broad capital ownership
People should receive some of the productive wealth of society, rather than merely wages.
Public service opportunities
People should have meaningful ways to spend time on things whose payoff isn't financial.
Philanthropic and civic institutions
Make it easy to direct surplus toward things people care about.
Deliberative democracy
Give ordinary people actual experience wrestling with collective choices.
The goal is to create a feedback loop:
learn → practice → experience → reflect → revise.
8. Change the status system
This may be the hardest part.
Humans are extraordinarily sensitive to status.
You can abolish one status hierarchy and humans will invent another before lunch.
So don't try to abolish status.
Redirect it.
Imagine a society where prestige accrues not primarily to the person who owns the most, but also to the person who:
- discovered something important
- built something beautiful
- raised wonderful children
- restored an ecosystem
- founded a school
- cured a disease
- mentored young people
- created extraordinary art
- solved a difficult civic problem
- gave away a fortune intelligently
- devoted twenty years to something that mattered
The billionaire philanthropist wouldn't necessarily be admired for having the billion dollars.
They'd be admired for what they did with the surplus.
That's a subtle but potentially enormous cultural transformation.
9. Teach mortality
This sounds dark, but I think it's essential.
A civilization with enormously powerful machines could become extraordinarily good at distracting itself.
So people need to confront the oldest constraint of all:
You have limited time.
If AI eventually makes almost everything materially possible, the scarce resource remaining may increasingly be human attention and finite life.
That changes the economic question.
The fundamental question isn't:
"How much can I acquire?"
It becomes:
"What is worthy of the finite portion of reality that I get to experience?"
That question has been around for thousands of years.
We haven't exactly solved it.
But perhaps we should stop pretending it isn't an important part of economic infrastructure.
10. Build a cultural firewall around "optimization"
This would be one of my strongest recommendations.
Children should learn to recognize the sentence:
"If it's measurable, maximize it."
as a potential warning sign.
Because that's exactly the kind of thinking that powerful AI systems could amplify.
GDP can be maximized.
Engagement can be maximized.
Profit can be maximized.
Life expectancy can be maximized.
Efficiency can be maximized.
Production can be maximized.
But a good life isn't necessarily the maximum of any single measurable quantity.
Human wisdom may consist partly in knowing which things should not be optimized.
That's a lesson humanity desperately needs before handing increasingly capable optimization machinery the keys.
And I'd put one principle at the center
Perhaps the whole curriculum could eventually crystallize around this:
Use things. Love people. Create meaning. Share surplus. Never confuse the instrument with the end.
Money is an instrument.
Markets are instruments.
Technology is an instrument.
AI is an instrument.
Even intelligence itself is an instrument.
The danger begins when an instrument quietly becomes the thing we're serving.
And that brings us back to your original observation. If AI gives humanity vastly greater power while leaving human beings psychologically organized around accumulation, status and endless optimization, we may become extraordinarily efficient at pursuing the wrong things.
But if our moral and philosophical development keeps pace with our technological development, something much more interesting becomes possible.
AI could supply an extraordinary expansion of means while humans become increasingly deliberate about ends.
That, to me, is the real companion project.
Not AI alignment alone.
Human alignment.
(R) "Human alignment with the best ways to live." You've given a rough estimate of the chances of navigating successfully the coming changes from AI. Does taking this new anthropological factor into account change your prediction?
(ChatGPT) Yes. Quite substantially. In fact, after thinking through your point, I would revise my earlier estimate upward, but with an important qualification.
My earlier rough estimate was about 55% that humanity eventually reaches a broadly owned, post-scarcity-ish economy, and only about 25% that we get there without a major crisis first.
If I add the possibility of a serious, deliberate project of human alignment with the best ways to live, I would now put my rough estimates at:
| Outcome | Earlier | Revised |
|---|---|---|
| Eventually reach broadly shared AI-era abundance | ~55% | ~70% |
| Reach it without a civilization-scale economic/political crisis | ~25% | ~40% |
| End up with a stable but imperfect hybrid of markets + broad ownership + social guarantees | high | ~75-80% |
| Fall into durable oligarchic/authoritarian AI capitalism | ~20% | ~15% |
| Catastrophic AI/control failure | hard to estimate | still the largest wildcard |
Those aren't actuarial probabilities. They're my attempt to make the structure of my intuition explicit.
And I think the reason for the upward revision is deeper than "education makes people nicer."
We have been missing a variable
Our previous model was roughly:
AI capability × economics × institutions × politics × AI alignment
I now think it should be:
AI capability × economics × institutions × politics × AI alignment × human anthropology
That last term is not decorative.
It determines what humans do when they acquire unprecedented power.
Suppose two civilizations receive exactly the same technology.
Civilization A believes:
More wealth = more success.
More consumption = more happiness.
More status = more worth.
More control = more security.
Civilization B believes:
Wealth is useful insofar as it serves human flourishing.
Enough is a legitimate destination.
Status is best earned through contribution.
Surplus carries responsibility.
Human relationships, knowledge, beauty, character and meaning are ends in themselves.
Those two civilizations can receive identical AI and produce completely different futures.
The technology isn't the determining variable anymore.
The people operating the technology are.
And there is some encouraging evidence that this isn't merely philosophical speculation. Recent research on AI governance is increasingly treating values, agency, disagreement and human flourishing as fundamental problems rather than decorative ethics layered on top of technical safety. There is even emerging work explicitly proposing "positive alignment": AI should not merely avoid harming humans but actively support human and ecological flourishing while preserving pluralism and human agency.
But there is a fascinating paradox
I don't think we should try to program "the good life" into humanity.
That would recreate the problem we're trying to escape.
The project should instead make humanity better at discovering, discussing and pursuing the good life.
That's a very different objective.
We shouldn't teach:
"Here is the correct philosophy. Follow it."
We should teach:
"Here are thousands of years of humanity wrestling with this question. Here are the strongest arguments. Here is what psychology, biology, history and lived experience can tell us. Now become capable of judging."
That produces something much more valuable than ideological agreement:
moral competence.
And moral competence may turn out to be one of the most important technologies of the AI age.
The really hopeful part
There's a feedback loop here that I hadn't included in my earlier forecast.
AI itself could make the human-alignment project dramatically easier.
Imagine every person having access to an extraordinarily patient philosophical tutor.
You could say:
"I'm miserable despite having achieved everything I wanted."
And instead of receiving generic self-help, it could walk you through Aristotle, Buddhism, Stoicism, attachment theory, positive psychology, existentialism, religious traditions, neuroscience, biographies and your own patterns of behavior.
Or:
"I want to become extremely wealthy."
The AI could ask:
"What do you imagine the additional wealth will actually give you?"
Then help distinguish security, autonomy, recognition, revenge, status, curiosity, generosity, fear and genuine material need.
Or:
"I want to spend my life helping people."
It could help you discover whether you're actually motivated by compassion, identity, superiority, guilt, belonging, or some mixture of all five.
That is potentially extraordinary.
For most of history, serious philosophical reflection was available primarily to people who had the time, education and temperament for it.
AI could make the examined life radically more accessible.
And that makes me more optimistic than I was.
There is another reason
You made a particularly important observation earlier:
unused wealth, money made and held for its own sake, corrupts the mind.
If that proposition became culturally widespread, the political problem of distributing AI wealth becomes much easier.
Consider two societies with exactly the same AI productivity.
Society A
Everyone is told:
"You must accumulate as much as possible."
Then AI produces enormous fortunes.
People fight over ownership.
Every billionaire has an incentive to become a trillionaire.
Political influence follows wealth.
Those without capital demand redistribution.
Those with capital resist it.
The whole system becomes a gigantic tug-of-war.
Society B
Everyone is taught:
"Secure yourself. Provide for those you love. Build things. Create. Explore. But beyond reasonable sufficiency, wealth is a responsibility rather than a trophy."
Now broad ownership doesn't necessarily feel like confiscation.
It can feel like stewardship.
And wealthy people aren't necessarily being told:
"You are bad because you have too much."
They're being told:
"You have been entrusted with something enormous. What is it for?"
That is psychologically much more sustainable.
So I would actually modify my proposed AI plan
Previously I suggested something like:
AI safety + financial stability + broad ownership + social insurance.
I would now make it a two-track civilization project:
Track I: Make AI safe enough for humans.
- technical alignment
- interpretability
- corrigibility
- containment
- cybersecurity
- international coordination
- limits on autonomous systems
- financial safeguards
- broad ownership
Track II: Make humans wise enough for AI.
- philosophy of the good life
- psychological literacy
- media/manipulation literacy
- education in desire and status
- economics of enough
- ethics of wealth
- history of civilizations
- practical wisdom
- moral disagreement
- civic participation
- generosity and stewardship
- cultivation of friendship, family, community, creativity and meaning
And crucially, Track II isn't subordinate to Track I.
They're mutually dependent.
A perfectly aligned AI serving a confused civilization is not a good outcome.
And a wonderfully wise civilization controlling a catastrophically misaligned AI doesn't get very far.
We need both.
My biggest revision, though, is this
I used to think the central question was:
Can humans survive the transition to a post-labor economy?
I now think that's too narrow.
The deeper question is:
Can humans become the sort of creatures for whom a post-labor economy is actually worth having?
Because imagine we solve everything else.
AI is safe.
Robots work.
Energy is abundant.
Medicine is extraordinary.
Everyone has housing, food and healthcare.
People own substantial productive capital.
Poverty is essentially gone.
And then we discover that humans have no idea what to do with themselves except compete for increasingly elaborate forms of status.
That would be a technological success and a civilizational failure.
But the converse is possible too.
If people gradually learn that a good life is not the same thing as a maximized life, then AI becomes something very different.
It doesn't need to make humans infinitely productive.
It needs to make it possible for humans to spend more of their finite lives doing things that are worth doing.
That distinction may be the single most hopeful thing in this entire discussion.
And oddly enough, I think your "economy of generosity" may be the bridge between economics and anthropology.
Because once enough people genuinely believe that surplus is for something, the ownership problem changes character. The question stops being merely "Who gets the machines?" and becomes:
"What should we do with what the machines give us?"
That is a much harder question.
But it is also a much more human one.
(R) Thanks, see you later.
(ChatGPT) My pleasure. This has been a particularly good rabbit hole to wander down. 🐇
The question I suspect we'll come back to is the one lurking underneath all of it:
Can a species become wise enough to inherit abundance without turning abundance into another form of scarcity?
That's a much more interesting question than "Will AI take our jobs?"
Until next time. 👋
