People are getting older, can AI get us out of the cycle?

Do a multiple choice question first.

Put the future in a two-by-two grid. The horizontal axis is the speed at which productivity technologies advance: AI, robotics, automation, energy, and biomedicine. The vertical axis is the speed of social system adjustment: how to collect taxes, how to pay pensions, how to change education, how to divide wealth, and how to make decisions.

Once the two axes intersect, there are only four worlds left in the future.

!

Technology is fast and systems can keep up , that is a prosperous society. The population is decreasing, but the output that everyone can leverage is rising. The pressure of pension care is absorbed by productivity. As working hours go down, living standards go up.

Technology is fast and the system cannot keep up , which is another matter. Society is extremely rich, but wealth is highly concentrated in the hands of the small group of people who own capital and machinery, and most people lose bargaining power. GDP is beautiful, but society is ugly.

Slow technology and keeping up with the system are difficult but decent adjustments. Delay retirement, increase taxes, cut benefits, liberalize immigration, and bring as many people as possible into the labor market. Growth is slow, but it will not overturn.

Technology is slow, and so is the system , two words: crisis society. Debt, pension, unemployment, the gap between rich and poor, and competition for resources feed each other and deteriorate together.

This article wants to answer one question: Where will humans fall into?

To answer that, we have to go back and see how we got here today.

For a thousand years, per capita income has hardly changed

Let's look at a set of numbers that I will never forget first.

Economic historian Angus Maddison created a unified ruler called the "1990 International Dollar" in order to allow comparisons between different eras. According to his estimation, the world's per capita GDP in the first year of AD was about 444 yuan. A thousand years later, in 1000 AD, this figure was 435 yuan.

In a thousand years, it has gone back nine yuan.

You can see it better separately. According to the same set of estimates, China's per capita GDP was around 450 in AD 1 and 1000, rose to 600 in 1500, and then it was still 600 in 1600, 1700, and 1820. Three hundred years, motionless. India from 450 to 533 in 1820. Western Europe was slightly better, from 576 to 1234, but that only more than doubled in 1,800 years.

What is the concept of doubling in 1800 years? An economy with normal growth today can do this in a decade or two.

GDP is not just one indicator. As of 1800, life expectancy in the UK was mostly less than 40 years old. The growth of average height and average life expectancy will not start until after 1850. Three completely independent measures (income, height, life expectancy) point to the same conclusion: Most of human history has stayed where it is.

Why it took so long.

Malthus gave the simplest explanation: population tends to grow exponentially, while land and food output can only slowly climb. So every time life improves a little, the population will increase, until the per capita share returns to a subsistence level.

Therefore, the history of agricultural society has repeatedly followed the same circle: population increase → per capita resources decline → income decline → social tension → famine, plague or war → population decrease → per capita resources rebound → population increase again.

In this circle, war, plague, and famine play the same role, reset. They are cruel, but from a systematic perspective, they do force the variable "population" back, allowing those who survive to redistribute more land and food.

Because of this, the power of ancient states was almost directly equated with population. A large population means more farmers, more soldiers and a large tax base. People are the productive forces themselves, and population is the national strength itself. This equation remained true for thousands of years.

I have to interrupt here to avoid being too full.

The term "complete stagnation for 1,800 years" is not without controversy in academic circles. Some scholars believe that growth is actually intermittent and intermittent. Goldstone once proposed the concept of "cyclical prosperity." Certain regions did prosper during certain periods of time, but then fell back. There is also a more acute criticism that "stagnation" is an illusion created by data processing methods: pick a few points at suitable locations from a large data set and connect them with a straight line, and naturally draw a two-paragraph conclusion of "long-term stagnation first, and then sudden take-off."

I tend to understand it this way: What stagnates is not every year, but the long-term trend line. Prosperity fell back for a while, which just showed that the society at that time was unable to lock in prosperity. Because locking in prosperity requires continued improvements in productivity, and that thing didn't exist yet.

It appeared around 1800.

QKPFX1 The QK Industrial Revolution reversed that formula

Steam engines, electricity, fertilizers, railways, oil, and assembly lines have achieved something that has not been done for thousands of years: It has made a person's output capacity jump by orders of magnitude.

In agriculture, it used to take a hundred farmers to support hundreds of people; in modern agriculture, a small population can support tens of thousands of people. In manufacturing, the work of hundreds of craftsmen in the past year was completed in a few days on an automated production line.

So for the first time in human history, this relationship appeared:

Productivity growth rate> population growth rate.

Malthus's circle was interrupted. The per capita income curve took a turn more than 200 years ago, from a horizontal line that was almost close to the ground to a nearly upright rise. Economic historians call this line a "hockey stick."

The decades after World War II were the most beautiful part of this new formula. Large young population, baby boomers, urbanization, industrialization, large-scale entry of women into the labor market, expansion of global trade, cheap energy, rampant infrastructure, and continued spread of technology. All variables are forced in one direction.

The bottom account at that stage was actually a bit touching:

More working population × higher per capita productivity = rapidly expanding total wealth.

It is precisely because the cake is getting bigger that pensions, medical insurance, public education, infrastructure, and national defense can expand simultaneously. The whole thing of the modern welfare state is based on the assumption that the working population will continue to grow. Pensions are paid to current retirees based on the contributions of current incumbents, and medical insurance operates according to the same logic. It defaults to a population pyramid with a wide base and a top.

This assumption is now failing.

The two variables of the ## formula are inverted simultaneously

The problem in the past was that there were too many people. The trouble for more and more economies today is: Too few young people and too many old people.

Let's look at fertility first.

The world average total fertility rate under the United Nations has dropped to around 2.25, only slightly higher than the replacement level of 2.1. 99 of the 194 countries are below replacement level. Another UNFPA report gave a global average of 2.3 children per woman in 2024, compared with 3.31 in 1990.

East Asia is at the forefront of the world.

South Korea's total fertility rate will fall to 0.72 in 2023, the lowest in the world, and some Urban area in Seoul are even less than 0.5. Japan will have approximately 686,000 newborns in 2024, which fell below 700,000 for the first time since statistics were available in 1899, a full 14 years ahead of the government's own forecast. ****

South Korea has rebounded in the past two years, returning to above 0.80 in 2025 and to 0.95 in the first quarter of 2026. This may seem like good news, but read it carefully: South Korea's statistics department's own analysis mentioned that the rebound largely comes from the post-epidemic marriage registration back-up and the formation of baby boomers in the 1970s and 1980s just reaching childbearing age."Echo Baby Boomers", This structural dividend is expected to fade around 2027. The distance of 0.95 from 2.1 is still far away.

What is more noteworthy is the cost. Since 2006, South Korea has invested nearly 280 trillion won, equivalent to approximately 1.5 trillion yuan, in encouraging fertility. In 2025, related expenditures will account for approximately 2% of GDP.

It hit 2% of GDP, pulling 0.72 to 0.95, and part of it is still a natural echo of demographic structure.

My judgment on this matter is that the fertility rate is a structural result, not an indicator that can be bought directly with money. It affects urbanization, living costs, education investment, women's employment, and the concept of marriage and childbearing. None of them can be reversed by subsidies alone.

Look at life expectancy.

Extended life is itself an achievement of civilization, but it raises a new question for finance: a person may have to live for twenty to thirty years after retirement. And here is a detail that is often overlooked: Living a long time does not mean living a long time in health.

Globally, the difference between life expectancy and healthy life expectancy is about nine years, or an average of about 13% of your life is spent in unhealthy conditions. Even in countries with the highest levels of healthy life expectancy, the gap is not small: Spain and Italy both have 11.1 years, and Singapore has a minimum of 9.6 years.

More critical is the trend. A study in the United States spanning four decades found that life expectancy without disability and life expectancy with disability are rising simultaneously. From the entire life cycle, there is no so-called "disease compression", but there is a certain degree of compression at the age of 65.

This piece of evidence is very important and I will use it later. It means: The optimistic assumption that "medical progress will make the elderly generally healthier and naturally relieve the pressure of elderly care" cannot be supported by current data.

One clip at both ends, and the result is the dependency ratio.

In the past, there were roughly ten working-age people for one elderly person, but later it became five, and in the future it may slip to two or three. According to the United Nations 'outlook, by 2050, one in three people in Asia will be over 65 years old.

This change is being transmitted to the public sector, and it is being transmitted faster than many people expect. According to public reports, the scale of teacher recruitment in many places has shrunk significantly in 2026, with a drop of more than 50%. As there are fewer students, there will naturally be fewer teacher positions. This is just the first industry that the demographic structure hits, and there is a long list to follow.

Meanwhile, on the other side, bills are piled up.

According to the Institute of International Finance, as of the end of March 2026, global debt reached a record high of approximately US$353 trillion, increasing by more than 4.4 trillion in the first quarter alone. Global debt as a proportion of GDP is about 305%. The figures from the IMF are: the ratio of global government debt to GDP will be close to 94% in 2025, and based on the current path, it is expected to reach 100% in 2029. This level has only appeared after World War II before.

There is another indicator that worries me more: the global fiscal buffer is almost zero. Ten years ago, this buffer was more than 1% of GDP, but now it is close to zero.

Buffer zeroing means that when the next impact comes, there will be no food left.

The several long-term thrust points named in the IIF forecast are worth reading out one by one: population aging, defense spending, energy security, cybersecurity, and capital expenditures related to artificial intelligence .

This last item is very interesting. While AI is expected to solve aging, it is becoming a new source of debt.

What ## AI loosened is the rope "person = labor"

Here, we can talk about AI.

I want to avoid details such as parameter quantities, model generations, and which company leads because they are not the hierarchy of this article. From the perspective of civilization evolution, AI and robots move an identity that has existed for thousands of years.

Past: People ≈ labor.

The future may become: People + AI + robots + computing power + energy = effective labor.

Once this equation is loosened, the rope between "how much a society can produce" and "how many young people there are in this society" that has been tied for thousands of years begins to loosen.

Break it into five floors.

At the first level, AI takes over cognitive labor. Supporting aspects of writing, programming, customer service, accounting, design, analysis, administration, legal and research and development.

On the second level, robots take over physical labor. Manufacturing, warehousing, logistics, agriculture, construction, cleaning, and care.

The third layer, the two are superimposed to form a multiplier. One person manages ten Agents, one engineer watches an entire automated production line, and one nurse uses equipment to take care of more elderly people. This layer is the key to hedging aging. It does not change "how many people there are", but "how many can one person cover".

The fourth level, AI accelerates innovation itself. New drugs, new materials, energy, chips, biotechnology. The meaning of this layer is compound interest: it increases the "speed of increase in the future."

The fifth level, medical technology extends healthy life. Let a 65-year-old have the same physical and cognitive states as they were 55 years old.

If all five levels are true, the conclusion will be very touching. But I'm not going to keep writing like this, because the evidence is not one-sided.

QKPFX4 Positive aspects of QK evidence

The experimental results at the micro level are quite solid.

Brynjolfsson et al.'s 2023 study covering 5179 customer service personnel found that overall productivity increased by approximately 14% after accessing AI. But what's more worth watching is the structure: It improves newcomers and low-skilled employees by up to 34%, while it has little impact on experienced senior employees. Another customer service study using a staggered roll-out and randomized controlled design found similar conclusions: the overall improvement was about 15%, the lowest skill fifth increased by 36%, and the new employee turnover rate dropped by about 10 percentage points.

In terms of writing tasks, the big language model has reduced the average time spent on middle-level professional writing by 40%, and improved output quality by 18%.

A randomized controlled trial conducted in Argentina in 2026 pointed more directly to "leveling": 1174 adults aged 25 to 45 were randomized to solve business problems. Without AI, the high-education group led the low-education group by 0.548 standard deviations. After AI was used, the gap narrowed to 0.139 standard deviations, approximately 75% of the baseline gap was wiped out.

There is one piece of evidence that I think is the heaviest in the medical field. The MASAI trial, released in 2026, randomly assigned 105,934 Swedish women to undergo AI-assisted breast screening or standard dual reading. The results are: The reading workload decreased by 44%, while the number of detected cancers increased by 29%(from 262 cases to 338 cases), the sensitivity increased from 73.8% to 80.5%, and the specificity did not decrease.

A randomized controlled trial with a scale of 100,000 people, with less work and more detection at the same time, is basically what the "third-level multiplier" looks like in reality. And it happens in medical care, where the pressure of aging is most concentrated.

QKPFX5 The opposite of QK evidence

But if you only talk about the above, you are lying.

Skilled people may be slowed down. Research by METR found that developers proficient in using open source tools completed tasks 19% slower than expected with AI-assisted assistance.

There is a J curve at the enterprise level. Research based on micro data from the U.S. manufacturing industry found that for every standard deviation increase in AI adoption, total factor productivity fell by 1.33 percentage points in the short term. The reason is that the process needs to be re-adjusted, the organization needs to be adapted, and the original management experience will be temporarily invalid. Only by surviving this period will long-term productivity advantages emerge.

There is a gap between investment and reward. A survey of 2400 respondents around the world showed that 97% of companies have deployed AI agents, but only 29% have achieved significant return on investment.

Micro efficiency has not yet turned into macro growth. This is the most fatal point. An average efficiency increase of about 30% at the individual task level has so far not been steadily translated into economy-wide productivity growth. Some estimates suggest that AI can bring productivity gains of 0.4 to 1.3 percentage points per year in developed economies, but this is highly dependent on assumptions about the speed of penetration and the proportion of coverable tasks. The optimistic estimate is 1.8 percentage points, but the one who gives this figure is the AI company itself, and you should have a scale in mind when reading it.

There is also the Baumor Effect standing in the way. Services such as haircuts, nursing care, and on-site teaching have ceilings for efficiency improvements, and their prices will rise as wages rise in other departments. The efficiency of one department's soaring cannot drive the entire economy.

There has been no large-scale substitution in the labor market. As of December 2025, approximately 35.9% of American workers have used generative AI, and the wage effect is slightly positive. There is no statistically significant decline in recruitment and employment in occupations with high exposure.

Putting the pros and cons together, my judgment is: The effect of AI at the task level is real, but has not yet been realized at the macro level. Every link in the micro-to-macro transmission chain (organizational adaptation, process restructuring, skills matching, capital investment) is dragging on time. The timetable for aging has been written down.

, will you run or not?

This question should not be answered with a "can" or "can't". It's actually six races running simultaneously.

!

The first game, productivity versus labor decline. This is the game AI has the best chance of winning. As long as the rate of increase in per capita output overshadows the rate of decline in the working population, total output can still grow.

Second game, automated deployment speed vs aging speed. The asymmetry of this game is critical. The population structure in the next decade or two has been basically determined. Those people have been born or have not been born; and the deployment speed of robots is full of variables. One side is certain, the other side is uncertain.

Game 3, medical progress vs growth of the elderly population. If healthy life expectancy can be extended by five to ten years, many "elderly people" can still work and take care of themselves, and the nature of elderly care will fundamentally change. But the previous 40-year study reminds us that so far, the number of years of survival with illness has increased with it, and disease compression has not really occurred. This one is not optimistic at present.

In the fourth game, machine costs fell vs labor costs increased. This game has a self-accelerating nature. Aging itself will push up labor costs, making automation more and more cost-effective, so automation accelerates and is further replaced. It is the only one in these six games with positive feedback.

Game 5, energy growth versus energy consumption in machine society. A highly mechanized society essentially replaces "manpower" with "electricity." There is not enough cheap and sufficient energy, and the first few floors are castles in the air.

Game 6, speed of technology diffusion versus speed of institutional adjustment. I think this is the most critical and most likely game to lose. Technology can advance exponentially, while adjustments in tax systems, pensions, education systems, and political decision-making usually take ten years.

A rarely mentioned mismatch

Japan's data is worth looking at alone because it is the most aging economy in the world, and many things there are no longer predictions but realities.

The Ministry of Health, Labor and Welfare estimates that about 2.72 million nursing staff will be needed in 2040, while the actual performance in 2022 will be about 2.15 million, with a gap of about 570,000, equivalent to an additional 32,000 per year. Another gauge gives a shortage figure of about 690,000.

But what really made me sit straight was the Ministry of Economy, Trade and Industry's "Employment Structure Projections for 2040". Its conclusion is: provided that AI and robots are utilized and retraining is promoted, Japan as a whole will not experience a large-scale labor shortage. The serious thing is mismatch.

To what extent is it wrong: There is a shortage of about 2.6 million on-site talents (production engineering, services, care, etc.), and there is a surplus of about 4.37 million people in service positions. The overall labor force in the Tokyo area has a surplus of 1.93 million, but the local area lacks on-site manpower.

Put these two numbers side by side, and the problem emerges:

What AI is best at replacing is transactional cognitive labor, which is the surplus of 4.37 million yuan. What lacks the most people is the manual labor and care labor on site, which is precisely the most difficult part of robots.

This is a cruel dislocation. It means that automation is naturally partial to solving the problem of aging. It continues to replace places where people are already overstaffed, and progress slowly in places where people are severely understaffed.

The actual situation of nursing robots also confirms this point. Surveys in Japan show that almost all facilities that have introduced nursing robots or ICTs report results: reduced physical and mental burden on employees, improved safety, and improved work priority judgment. The introduction rate is also rising, with some types of facilities approaching 45%.

But the gap is equally obvious. For the same model, some facilities have been in use and some have been discontinued. The card points focus on cost, adaptability, employee training and psychological resistance. There is an allocation survey in the relevant data of the Ministry of Health and Labor, which has the highest response rate, reaching 55.0%.

Translated into human language: Robots have indeed reduced the burden, but those who can really reduce the manpower allocation are still a qualified minority.

From "burden reduction" to "ability to hire fewer people", what is separated is not technology, but process, cost, training and acceptance by people. This distance is the specific form of the micro-to-macro fault zone mentioned above in the care industry.

also has electricity as a hard constraint

The energy issue in the fifth game is worth explaining clearly with numbers.

According to the International Energy Agency's 2026 forecast, global data center electricity consumption will double from approximately 485 TWh in 2025 to approximately 950 TWh in 2030, accounting for approximately 3% of global electricity consumption by then. Among them, the part with AI as the core grew faster, tripling over the same period to approximately 465 TWh. From 2024 to 2030, the average annual growth rate of data center electricity consumption will be approximately 15%, more than four times faster than the growth rate of electricity demand in all other sectors .

It sounds scary, but look at the other side: Under the benchmark scenario, the increase in data center electricity consumption accounts for less than 10% of the increase in global electricity demand . What really drives global electricity growth is still industry, electrification, electric vehicles and air conditioning.

So judging from the global total, energy is not the hand that strangles AI.

The problem is distribution. Data centers are highly centralized and not spread out like electric vehicles. More than 80% of the increase comes from the United States and China. Data centers in Ireland already account for more than 20% of the country's electricity consumption, and Virginia in the United States accounts for a quarter.

Therefore, the constraint changed from "whether it is enough to generate electricity" to "whether the power grid can be connected to it." More than 2500 GW of projects around the world are stuck in grid queues, and the IEA estimates that about 20% of planned data centers may face grid delays. To meet demand before 2030, annual grid investment needs to increase by about 50% from the current approximately US$400 billion. The equipment side is also tight, the supply of gas turbines and transformers is tight, and the shortage of high-bandwidth memory is expected to last until at least the end of 2027.

This is a typical example of "systems and infrastructure cannot keep up with technology." Power generation capacity can be built and chips can be built, but grid connection approval, grid investment, and transmission corridors follow a different timetable, a ten-year timetable.

In the first five games, AI and robots have a fair chance of winning, but the above two restrictions must be recorded in the accounts: It is partial, and it hits the power grid. Game 6, to be honest, I am not very optimistic.

And this sixth game leads us to the turning point of this article.

Once the production problem of ## is solved, the distribution problem will surface

Let's start with a minimalist example.

In a factory, there were 1,000 people in the past, and each person received wages, paid social security, and paid personal income tax. Now it has become a hundred people and 900 robots, but the output value has doubled.

From a production perspective, this is a complete victory. The total wealth of society has increased.

But pay attention to what happened on the ledger: Nine hundred people lost their wage income. Robots do not receive wages, so they do not pay payroll taxes or contribute to pension accounts. The extra profits flow to companies and shareholders.

So a very awkward situation emerged:

Countries are getting richer, and ordinary people's sense of financial security is getting worse.

This is not a deduction, the data is already moving.

The proportion of U.S. workers 'compensation as a proportion of economic output fell to *53.8% in the third quarter of 2025 , the lowest value since this statistics were first established in 1947. In 1947, this figure was 70%**. In 1978, the share flowing to labor fell by about 16 percentage points, and the same amount shifted to capital. Corporate profit margins were at historical highs during the same period.

This is not unique to the United States. Globally, the labor share has fallen by about 6 percentage points since 1980, with 13 of the 16 richest countries declining.

But I want to give a hedge here. Some economists have warned that part of this decline may be caused by statistical standards, such as owner income being mistakenly classified as capital income, an increase in penetration through entities, etc.; changes in shares cannot be mechanically equated with wealth from labor to capital."transfer". This reminder is reasonable, and I put it here because I don't want to hide its complexity with a clear number.

But even with discounts, the direction is still worthy of vigilance. Especially when this trend is superimposed on the fact that robots do not pay social security.

The modern welfare system has a hidden foundation: it assumes that wealth flows mainly through society in the form of "wages." Pensions are deducted from wages, medical insurance is deducted from wages, and the bulk of personal income is salary income. The complete redistribution machine is built on the salary pipeline.

If wealth is increasingly generated in the form of "capital returns" rather than wages, then this pipeline will become increasingly thin. The pipes are thinner, but there are more people to support. This is the most dangerous place where aging and automation are combined. It's not that there is no wealth, it's that wealth no longer flows through the pipe we use to collect taxes and pay pensions.

Therefore, one question cannot be avoided: Why and through what channels does the wealth created by machines enter the lives of ordinary people? *

Each of the plans on the table now has obvious problems.

Universal basic income. Direct thinking: Since the total amount of wealth is enough, divide it directly. The question is where money comes from, whether it will push up prices, and what to do with the meaning structure of society after it completely cuts off people's relationship with "participation in production."

Robot tax. It sounds fair. If the machine takes the place of the person, it pays the tax for the person. But "what is a robot" is technically difficult to draw. What is the boundary between an automated device and an automated script? Moreover, it is essentially punishing productivity improvements. In an international competitive environment, unilateral expropriation will almost inevitably lead to the outflow of production capacity.

Negative income tax, social dividends. It is more sophisticated than UBI, but still has to answer the same question: Where does the financial money come from?

Education and retraining. This is the most effective answer in the past two hundred years: technology eliminates old jobs, and education puts people in new jobs. But this time there is a new situation: the previous batch of experiments showed that AI improved people with low skills even more. This is a good thing, it means that AI is a power leveler. But it also means that the path of "obtaining a high premium by mastering scarce skills" will become narrower. When a tool allows novices to approach veterans, where does the premium for veterans come from?

There is another idea that I think has been discussed too little: Let ordinary people own machines.

If the share of labor income is destined to fall and the share of capital income is destined to rise, then the most direct hedge is not to give workers more wages, but to make workers the owners of capital , holding shares in those machines and computing power through pension funds, sovereign wealth funds, and universal stock ownership plans.

The advantage of this path is that it does not go against productivity, does not punish automation, but changes the starting point of allocation rather than the end point. Its difficulty is also very real: once ownership is formed, a gap between those who hold it first and those who hold it later will be difficult to catch up. In a world where returns on capital continue to exceed returns on labor, where the starting line is almost everything.

This may sound like a fantasy, but there have been samples in the world that have been running for decades, and there are two different running methods.

Norway put oil revenue into the government's global pension fund, which now has a size of approximately US$2.2 trillion and holds an average of 1.5% of the shares of all listed companies around the world. Based on a population of approximately 5.5 million, per capita is approximately US$400,000 . When oil and gas were first discovered in 1969, some people in China advocated direct distribution and immediate improvement of living standards. The final decision was to invest it in the global market.

Alaska takes the other path: the permanent fund is about US$87 billion, which is about US$120,000 per capita based on a population of 730,000, and directly distributes cash to residents every year .

Both paths prove that "universal ownership of capital" is technically completely feasible. But the walls they each hit are more informative than the feasibility itself.

The wall in Norway is It does not issue money directly . The fund allocates a proportion into the budget every year in accordance with fiscal rules (this proportion has been reduced from 4% to 3%), indirectly benefiting the whole people. The advantage is that it avoids the annual political pull, but the price is that the people have almost no direct sense of gain from this huge wealth. The average person lies in the accounts of 400,000 US dollars and cannot be felt in the days.

The wall in Alaska is so fierce that there is a fight every year. This round in 2026 is particularly typical: according to the legal formula established in 1982, more than $3800 should have been paid this year; the governor advocated paying 3650 according to the formula; the House Finance Committee once promoted the draft of 3800, and then changed it to about 1500. The final implementation was a $1000 dividend plus a $200 energy subsidy. The $1000 in 2025, adjusted for inflation, is the smallest since the plan was launched in 1982.

What is more noteworthy is that since 2016, as oil revenue has shrunk, legislators have been using fund income, which should be dividends, to cover the government's daily expenses.

So "letting all people hold capital" solves "where the money comes from", but it does not solve "who has the final say" at all. The choice between dividends and public spending will ultimately be a political game, and it will be played again every year.

There is another limitation that cannot be avoided. The principal of both funds comes from natural resources. Oil is something that can be confirmed, nationalized, and directly taxed. The value created by AI and robots is scattered across the balance sheets of countless private companies, in the form of algorithms, chips, data and organizational capabilities. You can't "discover" an AI oil field and then declare it owned by all people.

From oil funds to "machine funds", there is a complete set of institutional designs for ownership definition and taxation. That thing does not exist yet.

Writing here, I want to release what I think is the most important sentence in this article:

AI can create wealth, but it will not decide on its own how to divide the wealth.

Technology is a multiplier that amplifies existing distribution structures. The distribution structure itself is fair, and it amplifies fairness; the distribution structure itself is tilted, and it amplifies the tilt.

What if technology comes too slowly

What is discussed above is "technological success and institutional backwardness." The reverse situation should also be considered clearly.

Suppose that aging and debt pressures are already on the rise, and productivity improvements in AI and robots are not yet in place. Society will in turn use its buffers: delay retirement, increase labor participation rates, introduce immigrants, increase taxes, cut pensions and medical insurance, expand borrowing, and tolerate inflation.

These methods have one thing in common: They are all re-cut inside the same cake.

And every cut creates a new antithesis within society: the intergenerational conflict between young people and the elderly, the tension between local workers and immigrants, and the differences between taxpayers and welfare recipients. When the inside is really exhausted, attributing conflicts to the outside becomes the least costly option politically.

Pushing down this chain, the worst end is international conflicts.

But I want to make one thing very clear here: War has never been the answer to this question, even under the most ruthless calculations.

The reason why war in the agricultural era could serve as a "reset machine" was because the most important means of production at that time were land and population. Land can be taken away, and people can be looted and moved. Winning a battle can indeed redistribute these two things.

What are the most important means of production today? It's technology, computing power, robot capacity, energy systems, data, and the most critical thing: trained people and the systems that allow them to collaborate efficiently.

The common feature of these things is that they cannot be taken away and will be broken with a hit.

A war can simultaneously kill the youngest workforce, destroy the capital stock, lower fertility, create refugees, interrupt education, and drive up debt. It will make the demographic problem more serious , not less. To regard war as an outlet for population and debt problems is to reverse the arithmetic.

So my judgment is: War may well emerge as a result of the system getting out of control, but it is almost impossible to emerge as an effective economic solution. These two things must be said separately.

As for whether technology will make conflicts more or less, I think the power in both directions is increasing at the same time.

On the one hand, more abundant production, greater resource efficiency, and a decline in the weight of population in the national power equation are weakening traditional motivations for war. On the other hand, unmanned systems and automated military technology reduce local casualties and lower the political threshold for launching local conflicts. Moreover, the form of modern confrontation itself is changing. It is less and less manifested in large-scale infantry charges, and more manifested in technological blockade, supply chain cut-off, financial sanctions, cyber attacks, and information manipulation.

Such confrontations are characterized by low intensity, long-term duration, and difficulty in defining the outcome, making it more difficult to conclude through negotiations.

escaped, but may fall into a new cycle

Now we can go back to the original question: Can humans escape the historical cycle?

My opinion is that the material foundation of the old cycle is indeed disintegrating.

The cycle of the agricultural era was based on the fact that "production capacity was limited by land and population." The Industrial Revolution interrupted it because productivity growth outperformed population. The AI era may advance to the third stage: Production capacity is gradually separated from population size.

For the first time in thousands of years,"how much civilization can make" may become less and less dependent on "how many young people there are."

But escaping the old cycle does not mean that there will be no cycle. I think a new cycle is taking shape:

Technological breakthrough → wealth explosion → wealth concentration → accumulation of social contradictions → institutional reform → redistribution → the next round of technological breakthroughs.

In other words, the main cycle facing mankind is changing from a cycle of "population and resources" to a cycle of "technology, capital, distribution, and politics".

In the past, what determined the stability of a society was whether there was enough food. What determines the stability of a society in the future may be how wealth created by machines enters society.

There is another consequence of this transformation that is not easy to notice: the components of national power are being transformed.

In the past, population size was almost directly equal to national strength. In the future, the weights of population quality, technical capabilities, robot production capacity, energy supply, capital stock, and institutional efficiency will increase significantly. The comparison between an economy with a small population but highly automated and an economy with a large population but less automated will become much more complex than in the past. Because for the first time,"effective labor force" can be decoupled from the number of heads.

I don't intend to rank specific countries here or judge the outcome of victory or defeat. The credibility of such predictions is usually very low. But one observation is worth keeping in mind: Technology is likely to spread faster than demographic changes. This means that latecomers may not have no chance in this round: robots and AI can be bought, learned, and pursued, but demographic structure cannot be bought.

By the way, I'll mention a detail that I noticed in the data. According to the latest report of the International Federation of Robotics, South Korea's robot density is 1220 units per 10,000 manufacturing employees, ranking first in the world; Germany 449, Japan 446, and the United States 307. Due to the adjustment of the statistical caliber in China (recalibration of the denominator based on the latest labor force data from the National Bureau of Statistics), the density is 166 units, ranking 22nd in the world. In the previous edition of the report, China's number was 470, ranking third in the world.

In the same country, it dropped from third to 22nd in one year, and there was no missing robot. What changes is the denominator.

This change in caliber itself speaks for itself: The pressure of automation depends on how many people you need to be covered. China's operating stock of robots is about 2 million units, which is about 4.5 times that of Japan, which ranks second. In 2024, 54% of the world's newly installed robots will be installed in China. Both stock and increment are the first fault, but when spread to the huge manufacturing employment population, the density is still not high.

This is the real difficulty of the problem "effective labor force": it is not a race of numerators, it is a race of numerators and denominators.

QKPFX12 If QK wants to escape, it has to evolve three systems at the same time

Writing here, I want to wrap the entire article into one framework.

If a civilization wants to get rid of the old cycle, it needs three systems to evolve at the same time, all of which are indispensable.

!

The first set is the production system. AI, robots, energy, medical, materials, biotechnology. It answers the question: Is there enough things in this society? *

The second set is the distribution system. Taxation, pensions, benefits, capital ownership, education, public services. The question it answers is: How do these things enter ordinary people's lives?

The third set is the governance system. Decision-making efficiency, policy adjustment ability, interest coordination mechanism, conflict resolution channels. The question it answers is: When someone is dissatisfied with the distribution results, can society deal with it through systems rather than violence? *

Summarize the relationship between the three in one sentence:

Technology determines how much a civilization can create, system determines how this wealth is divided, and governance determines what happens when the distribution is uneven.

If any of these three is seriously left behind, something will happen. And the ways in which they go wrong are different: falling behind in the production system is poverty, falling behind in the distribution system is tearing, and falling behind in the governance system is out of order.

Now back to the first four squares, their meaning is complete.

World A, technology is fast and systems are fast. The population declines but productivity increases faster, the pressure of pension care is absorbed, working hours are reduced, and healthy life is extended. A considerable part of the wealth created by machines enters the public domain through some ownership or tax mechanism. This is the rudiment of a "post-labor society" where people no longer earn income mainly by selling their working time.

World B, technology is fast and systems are slow. AI creates huge amounts of wealth, but the share of labor continues to decline, capital returns are highly concentrated, and most people lose bargaining power. The country's GDP figures are beautiful, and social tension continues to rise. This is the state known as "technological feudalism": productivity belongs to modernity, but the distribution structure is advancing and retreating.

World C, technology is slow and systems are fast. Society accepts lower growth and slowly absorbs the demographic structure through delayed retirement, tax reform, immigration and welfare adjustments. Life is not easy and growth is slow, but it will not collapse.

World D, technology is slow and systems are slow. Debt, pension, unemployment, the gap between rich and poor, and resource competition reinforce each other, and buffers are exhausted one by one. This is the world most likely to slide into a large-scale political crisis.

So where is the world going today?

I dare not make any conclusions, but these sets of data in hand point in the same direction: the technology side is accelerating significantly, the robot density has doubled in seven years, and the efficiency improvement in micro experiments has been quite solid; on the institutional side, the global government debt-to-GDP ratio is approaching post-World War II levels, the fiscal buffer is close to zero, and the labor share has fallen to its lowest level in 78 years.

The leg of technology is running fast, and the leg of system is dragging. If I had to point out, the current center of gravity is leaning towards B.

Race between Four Speed

I don't want to end with "AI will save the world", that's too frivolous.

What is really racing are four speeds: The speed of population aging, the speed of debt accumulation, the speed of technological progress, and the speed of institutional adjustment.

If technology + institutions> aging + debt , there will be room for progress in society.

If aging + debt + distribution conflicts> technology + institutional adjustments , conflicts will move upwards.

Note that there is a system term on both the left and right sides of this formula. This is no accident. Among these four speeds, only institutional adjustments are entirely decided by people on their own initiative. The demographic structure is basically dead, debt is a stock of past decisions, and technological progress has its own rhythm. The only variable still in our hands is how fast we are willing to adjust the rules.

The historical significance of AI may not ultimately lie in whether it can write articles or replace programmers.

What it can really change is a relationship that has existed for thousands of years:

For the first time, how much wealth a civilization can create may increasingly depend on how many young people it has.

If humans get the question "Who belongs to and how to divide the wealth created by machines" correctly while taking this step, then this will be the second time since the Industrial Revolution that they have truly escaped from the population trap.

If technology succeeds and institutions fail, we will have a world where technology is extremely developed and society is extremely divided: There are too many things to use up, and most people can't afford them.

By that time, the answer to be answered is no longer "Do people still have jobs?" It is what a modern society should rely on to organize income, wealth, power, and life when human labor is no longer the scarcest thing in social production.

No society has yet answered this question.