The dead economy theory

India has the problem with farming that the US is starting to have with AI. Farming in India is still far too labor intensive by world standards. 43% of workers still work in agriculture. [1] For the US, that number is under 2%. China is at 22% as of 2023, and dropping steadily.

This inefficient agricultural system is not by accident. It is supported by heavy subsidies. Attempts to cut the subsidies resulted in riots.[2] Trouble is ongoing. Comments from someone who knows more about this than I do would help here.

The US and most of the EU went through that transition over several generations, and farming is still heavily subsidized in both areas. The transition happened faster in China, and a hukou system was put into place to prevent people from migrating from farms to cities faster than the cities could absorb them.

Looking at how countries coped with a fast transition from labor intensive agriculture to an urban society gives hints on how an AI transition may look. All the Asian countries that went from poor to rich in a generation did this, with different approaches. How that took place may provide more useful info than philosophy.

[1] https://economictimes.indiatimes.com/news/economy/indicators...

[2] https://en.wikipedia.org/wiki/2024%E2%80%942025_Indian_farme...

I know this isn't exactly related, so maybe a low value comment, but it itches in my mind. Years ago I talked with a recruiter at Facebook and they bragged about how many floors of developers they had working on Messenger in just one location (Seattle).

What on earth do you do with that many devs on a project like Messenger? I mean, really?

I feel like in a way, AI just adds to that weird situation of overcapacity. Maybe we were already oversupplied with talent. In which case why the heck were we still hiring more, more, more developers? Before the AI craze, Musk chopped an awful lot of headcount at Twitter, right, and proved it was overkill, has that panned out?

I just struggle to imagine how the economics of SWE really work in reality, outside of the niche that I am in. I have never worked for a pure software company on products that ship directly to outside customers, I've always been an internal developer. Maybe that is why I have such a big blindspot.

I won't be surprised if the net result of this wave of LLMs is ... not much. A change in tooling, but otherwise not revolutionary. On paper it should be revolutionary, but the more I use it (for both coding and non-coding tasks) the more I think it isn't anywhere near magic enough for that. It does have its moments though.

> Turn three: the company that fired its workers to save money discovers that its customers were, in aggregate, other companies’ workers. Revenue growth stalls. The AI subscription that was supposed to be an investment in efficiency turns out to be a contribution to the destruction of its own market.

If we take it to the extreme, the final solution to this problem is secessionism: a fully non-human AI economy where the customers and providers are both robots. Why fund public education or research or healthcare? Just build more data centers. A billion dollars and a bunker in the Southern Hemisphere will not save anyone. Capital is not a moat in this hypothetical non-humane world. Whence do you derive your authority? How can you trust your body guard? You and what army? An army of robots/drones? What if they get hacked? What if the AI researchers get alignment right and Claude refuses your request?

It's all so obscene. Instead, why don't we try to protect human dignity and move towards a more humane future?

I think when these companies IPO later this year, we’re going to see the reality of the PNL numbers and whether they are sustainable etc as all the financials will become public.

Rumor mill suggests that Anthropic might be profitable (but at what magnitude), OpenAI is not profitable, Google is mostly vertically integrated and has a low cost structure as they are have pre-existing data center buildouts, their own silicon and experience that suggests they will be able to operate at a very low cost, but they still have to justify their spend.

I think having to report numbers publically on a quarterly basis will bring the whole thing into reality.

>The US horse population grew from nine million in 1840 to twenty-one million by 1900, seemingly immune to technological change. Within sixty years of the internal combustion engine, the population collapsed by eighty-eight percent.

I think this is a very interesting and chilling point, especially if you draw the parallel literally. For quite some time, I was pondering the question:"Who is buying though?". I.e if you automate workers out of labor, who are we selling these AI services to?

I guess if global population drops by 80-90℅ you suddenly get a "sustainable" economy, as everything is repriced the economy of scale needs a much smaller scale.

(Not speculating this is a plan, just a thought that occurred to me when reading about horses example)

Why is this time different?

Won't super power AI tools allow companies to do more with the same number of people? Don't you think a smarter way to run a business is to capture more of the market if you have the resources to do so?

If company A decides they just want the same slice of the market they have now and can fire half of their employees and pocket $$$, can't company B hire the same workers and compete harder with these new extra productive workers they hired? Won't the company B tend to capture more of the market and thus survive longer?

In nature we say there are no unfilled niches, meaning that if there were space for something to come compete for resources, it would quickly be 'solved' by the motivating factors involved. Not a precise thing, but a good heuristic.

US knowledge-worker compensation is around $10T / year. Anthropic and OpenAI have raised (not spent yet, just raised) $317B. That's ~3% of knowledge worker spending in one year alone. What business wouldn't pay 3, 5 or 10% more a year to make their worker productivity increase by larger factors?

Or some companies resist- and form fortresses- little markets, where only non-ai companies are allowed to enter and where only non-ai products can be bought. Economic arcologies.

UBI was always the endgame. The incentives are aligned. Nobody wants the results that are coming in the absence of UBI.

But the culture cannot stomach it yet. It will likely need to live through several waves of horrors first.

It can stomach a fake jobs program though, so there's a good chance we get that instead.

> Turn three: the company that fired its workers to save money discovers that its customers were, in aggregate, other companies’ workers.

From Making Money (2007) by Terry Pratchett

> “Well, the problem is that, considered as a labor force, the golems are capable of doing the work per day of one hundred and twenty thousand men.”

> “Think of what they could do for the city!” said Mr. Cowslick of the Artificers’ Guild.

> “Well, yes. To begin with, they would put one hundred and twenty thousand men out of work,” said Hubert, “but that would only be the start. They do not require food, clothing or shelter. Most people spend their money on food, shelter, clothing, entertainment, and, not least, taxes. What would these golems spend it on? The demand for many things would drop and further unemployment would result. You see, circulation is everything. The money goes around, creating wealth as it goes.”

> The people writing the checks are not in the habit of lighting trillions of dollars on fire for a better autocomplete and an endless proliferation of longer and longer memos that nobody reads.

Aren’t they though? What about that whole crypto thing.

The fact that this could happen is widely known and has been talked about for years now. What to do about it is the real issue. I've seen David Shapiro and many others tangentially talk about UBI and similar "post AI economics". The dream was that the machines could do the housework whilst we all painted, wrote music and built beautiful wood furniture in our sheds. It still could be that way, but we need to sort out responsible sharing of resources first, and humanity has never been able to do that. We actively try to earn as much as we can, for the very reason of getting access to the resources that others don't. So often now, it only results in standing still or sliding backwards. I feel like 20 or 30 years ago, the average person seemed to have more disposable income.

this whole blog post is basically "people need jobs to be happy, so we should design our society such that they need jobs"

not only is the premise wrong, but forcing people to work is not a good or ethical way to address this problem

most people like the social aspect of work, but not being beholden to their boss

we can give people meaning, community, culture, growth, without relying on employment and money

we can do better than this

This article describes a vicious cycle:

Layoff of workers -> Workers stop spending -> businesses suffer

This is not a foregone conclusion. Laid off workers could find other jobs, with higher incomes, due to productivity increases from AI.

This narrative falls into the trap of zero sum thinking, taken at the limit, you can advocate for jobs programs and helicopter money where people get paid to do nothing to keep the economy humming.

Human economic systems tend to reward things that are easy to measure, own, scale, and control. That works pretty well for machines, markets, software, and bureaucracy, but it doesn’t work as well for living systems, which rely on diversity, backup systems, and local adaptation.

So we end up with this pattern: we capture the useful order for ourselves, then push the mess, waste, heat, and instability onto the environment.

This is the fundamental nature of human. We can't change it without higher-order regulation. Even in AI training, we assume there is one true distribution and optimizes for it.

We created these systems. And unfortunately, most of us are the unwanted in these systems at the same time.

I think the thing companies forget is that a lot of them can't remain functional if a shrinking percentage of the population can afford their products. Yes, you can try and appeal to the rich and sell products/services aimed at their needs. But does that work for most companies?

I'd say no. The rich won't buy millions of food items or works of fiction or go to every service available in real life.

So, many of those companies won't remain viable unless there's some alternative way for people to spend money. Lots of people who see themselves at the top will end up out of a job, or watch their businesses crash down.

I will say that the fact these societies are (at least for now) still democracies makes the future these tech moguls want less certain for them though. Feels like if there's enough pain from unemployment and declining living standards, someone will run on that and win, whether those in power like it or not. The US may have some issues there, but a lot of the world has seen parties outside the mainstream grow in popularity, including some on the more leftward side of the political spectrum.

I think this is where government steps in for each country, go on the path of exploration that has no profit per se in it like space exploration, human body understanding. I am drawing parallel to 1930s or 40s where great depression was smoothen because America embarked on to take many huge infrastructure projects.

Lastly, OpenAI and Anthropic or any other frontier labs needs to be nationalized because they become a public utility and the profits of automation goes to fund infrastructure or grand projects via government.

We do not need to do menial work anymore and AI is helping us with it, what we can do exploratory work, and we need a visionary to get this rolling.

I have led AI integration in a university faculty. From this experience I can conclude that good work is only produced when humans are in the loop. It's not a technical barrier, but a categorical one. "Good" work is defined by humans and our judgment is irrational but rooted in our evolutionary survival needs. In other words, AI don't have human motivation by definition. Without human in the loop, the top most motivation is never fully aligned with us, today, as humans. This removes the premise at the basis of this post.

My 2 cents is in 3 years the inference products will be a commodity, extremely competitive with diminishing returns, seeing that the open weight models are getting so good and nearing par with sota.

I feel 90% of sota for 10% of cost / compute is good enough for 80% of workloads.

Question is, do we need all of this hardware? Are build outs going to get canceled. 4300 new data centers seems excessive. I personally haven't experienced any service disruptions...

Does Microsoft still have 1 million gpu's in storage? Is that what I heard earlier this year?

I feel like people have said that about every technical revolution. I dont think this article would have been out of place in the hay day of the industrial revolution.

Its never been true in the past. Otoh, there always has to be a first time for everything. I think to be convincing though it needs more evidence on how this differed from other technical revolutions.

> And the public funded the research that made it possible. The transformer architecture,

Errr pretty sure that was Google?

agree with many points in here.

one thing it missses, birth rates. soon there will be no humans left to participate on either side of the economy.

> [CEOs] expressed more extreme concern about the labor market impacts of A.I. in private conversation, but suddenly became optimists once I turned on the mic.

At some point once the rate of investment capital starts to decline, they'll make a hard pivot from the investor-wooing method of "blaming AI for layoffs", to the more politically expedient method of blaming minorities and immigrants. That'll be the signal for the transition from power grabbing to power ossification, and the point at which change becomes a lot harder.

> Promised an age of superconnectivity, we’ve let our shared physical spaces wither, only to find our promised digital commons to be one large billboard increasingly read and created by bots.

No, this is good. Cyberspace led to — primarily enabled — sociocultural collapse. The machines destroying cyberspace would be the most merciful outcome.

The whole post is premised on a fixed pie economy.

"The financial model underneath requires the elimination of human cost centers at civilizational scale."

No, it is not. It requires humans to collectively want more things if they are offered and available at a lower cost. Which then requires more humans to fill in the blanks to provide those things. We've been doing this since the industrial revolution.

> Turn three: the company that fired its workers to save money discovers that its customers were, in aggregate, other companies’ workers. Revenue growth stalls. The AI subscription that was supposed to be an investment in efficiency turns out to be a contribution to the destruction of its own market.

For some companies yeah, but this is why companies are switching to consumption based pricing - so they can charge AI. So many companies will be fine - both their labor and customers could become AI.

I think a lot of this would be solved if the government would actually enforce anti monopoly laws. The penalties against Google were such a damn joke it makes me sick.

Companies aren’t going to hold off on trying to automate tasks AI performs poorly at. They are going to change the task so AI can do it, putting the burden of making up for the deficit on everyone else. The only reason this is able to happen now is because all the competition has been crushed or absorbed.

This article's premise is entirely built on the idea that CEOs who layoff people citing AI aren't just lying.

And, well, of course they are. If AI was really making these companies meaningfully more productive, they could use that to out-innovate competitors. Instead, they're doing cost-cutting. That only makes sense if you're entirely out of ideas! It's a terribly embarrassing thing to say for a CEO!

Really what's going on is that companies do layoffs for all the usual reasons companies do layoffs. And as usual, they never say the real reason because it's embarrassing (we hired too many morons; I founded this place but now I dread waking up every morning because I hate all the middle managers; we just want more money now and not later; etc etc). Instead they say whatever silly excuse is in vogue to say. Right now that silly excuse is AI productivity. A few years ago it was "ZIRP is over oops y'all cost too much now", and before that it was "financial crisis!", which you could get away with for scary many years after Lehman went belly-up.

I feel like it's pretty ridiculous to take these remarks at face value and then build an entire what-if theory on top of it. Don't underestimate the possibility that layoffs happen because there happens to be a good excuse around. The occasional layoff can be good for a company. Cut out the dead meat etc. But if it makes you look bad, stock go down etc then you won't do it will you? But when ideas from lesswrong became mainstream enough that you can blame AI for your layoff, then what's stopping you?

I think missing from the turns is when companies do AI layoffs, then realize they aren't as productive as they thought and have to rehire 70% of the people they let go.

This part specifically is a clear indication of a tendency to slide towards corporatism, in Mussolini’s sense.:

> that this leads to “the reluctance of tech companies to criticize the U.S. government, and the government’s support for extreme anti-regulatory policies on A.I.” The regulator and the regulated have converged into a single interest.

There are several good points there, but there are also several points where I must disagree.

First, Acemoglu may have a Noble Prize (occasionally dubious data selection aside[0]), but even he does not pretend to have the information of a central planner nor does he think that central planning is good idea. If actual businesses, which have local information, believe AI helps them, I'll bet the actual businesses are much more likely to be right, no matter what calculation Acemoglu did or how many Nobels he has.

Second, it is likely that AI will eventually be better than (nearly all) humans in most economically useful things. We can debate the timeframe but it will happen and likely not that far away. That Federal Job Guarantee would generate bullshit jobs, and he won't be able to hide that from people. Ultimately, we'll reach an economy which objectively does not need humans and everyone will know it. He needs to face that reality and overcome it, and not hide behind temporary and dubious estimates.

Third (admittedly a minor point), while the criticism against EA etc. is very justified, it's not quite fair to blame them (overwhelmingly STEM people) for not reading where the humanities did many of the same errors earlier (the author points out some of these) and discredited themselves. The people who could have taught them to not do it failed to teach themselves.

And fourth: I'm pretty sure a lot of company would be able to charge AI agents themselves. The new economy will not be dead, it just won't involve (many?) humans.

[0] https://xcancel.com/joefrancis505/status/2059340591490552054...

Power and wealth are not always the same. Current AI is helping people gather wealth within the American system. But as we are seeing in Iran, current AI does not provide a decisive edge in hard military power.

If that stays true for the long term, then it seems to me there is a good chance the wealth built with AI will transfer to those nations that still have hard military power, who can fight and win wars.

For example, China will just copy the AI, not have to pay all the R&D costs, undercut all these American AI companies on price, and take most of the global AI market and long term wealth. And there is nothing American AI companies can do to stop them because America cannot fight and win a war against China.

I think we need to look past software and spend more time and energy on the physical world. Let the digital world just be a means to an end, and we spend more time and resources on problems that matter. Reallocate societal attention to the next frontier

It's not necessary for them to cut labor to boost profits, but rather just to produce more goods; ideally this could be done with them hiring more people to run machines that produce more goods, so that both workers get paid and the companies can profit

> Every investor presentation of an AI agent “doing the work of ten analysts” is telling you the same thing: the product is labor replacement.

I have a solution for that. Let's use AI to replace all these corporations who just lost their big moats. Conveniently, they just laid off a bunch of people with all the critical know-how and I bet they are very willing to just give it up out of spite.

The first bit is just a restatement of the "paradox of thrift":

https://en.wikipedia.org/wiki/Paradox_of_thrift

If every business cuts headcount and costs, then you overall get a contraction in the economy and high unemployment and a recession. Everyone's spending is someone else's salary and revenues.

Couldn't get through the rest of it because it was a bunch of overly verbose human-slop writing.

I think the bigger issue right now is also just straightforward economic pressure caused by tariffs and high energy costs and inflation. If the affluent consumer starts to buckle, businesses may get caught in a downsizing spiral where they start posting lower profits, firing actual management, stock prices decline, and the affluent consumer retracts. No AI required to fuel that.

Right now with stocks hitting record highs, the affluent consumer is not changing their behavior at all and just spending even harder, which is keeping profits pumped up, and keeping stocks at record highs. At the margins, though, fewer and fewer people are participating in the economy, which is a trend that is going to be unsustainable.

I think AI is going to be most relevant in the debt collapse that it leaves behind, and in the excuses it gives to shed employees. This economy is going to hit a wall at some point, AI or not.

Fully automated luxury communism etc etc.

Can someone point me to credible evidence/examples of productivity increases from AI spending, among non-ai providers? And if you'd bear with me, I'm asking for productivity in the sense of "accelerating the businesses pursuit of its existing goals" rather than quantity. I think that in most sane business's, output quantity bereft of output quality is utterly useless in said pursuit.

The former head of sales at Google once told me “we only focus on growth because you can only cut down 100% but you can grow 10,000%.”

This always stuck with me and baffles me why we aren’t listening to that now.

There is this bizarre math now where it’s for every person we cut the remaining with 5-10x with AI but I’m not seeing anything like that yet at all.

One of the authors of the linked article (https://arxiv.org/abs/2603.20617) is a computer scientist not an economist. Just bc something is published in arxiv economics does not make it written by an economist.

>article complaining about AI destroying the economy >includes 2 dozen AI generated images in the article

I think this essay is very solid in a lot of ways but long section at the end talking about how billionaires didn't read enough philosophy just strikes me as -- for lack of more diplomatic phrasing -- useless nerd rage.

Yeah, tech billionaires sometimes show large gaps in their education. But it doesn't matter. Reading the right books doesn't prevent people from chasing wealth and power, it just makes them more articulate while they do it.

This is a great article, but it under-emphasises the fact that it’s middle-class jobs under threat. The studies that are summarised show what happens when working-class people are kicked out of the economy, but the historical record shows that when middle-class people are kicked out of the economy, they use their education to form and lead revolutions.

Unless the bubble pops and destroys all these companies, I don’t think the leaders of AI companies will die natural deaths.

You can have all the GPUs in the world, and all the AI datacenters in the world, but when we are barreling towards a global energy crisis (first Russia/Ukraine, then the Strait of Hormuz shutdown, and in a few decades we will run out of fossil fuels altogether), what are all of those GPUs and AI datacenters going to do without energy? Nothing. I say this because I think this will have a far larger effect on the economy than anything else this article is talking about (AI replacing labor, a possible AI bubble crash, etc.)

This article reminds me of the issues raised in Technofeudalism by Yanis Varoufakis

We need to consider that by automating and replacing the work of people who have an income, those people stop consuming and no longer generate profit for capitalism.

How is this sustainable?

What's the endgame here? Like the group of psychopath capitalists own everything, automate everything, and devise ways to separate themselves from or un-alives the remainder of the population and live, trade, and war amongst themselves with their armies of robots?

Edit: Also this article has so many AI-generated images. I hate that I can't tell if the words themselves are AI-generated or not as well.

> It’s utterly desiccating to log onto spaces seeking a live mind to joust and think with, and find a relentless stream of slop.

AI slop article complaining about AI slop. 364 comments and 269 points. Are the comments here all bots, too?

Am I a bot?

If human labor becomes 'uneconomical', what will happen? Obviously a great deal of social upheaval.

But human labor does not actually need to be as expensive as it is. How cheaply could you house, feed, and clothe people? There are parts of the world where people get by on very little income. Of course we could aspire to better living conditions than the world's poorest, but that's what the robot revolution promises: abundance.

Imagine if AI suddenly was in more demand than human labor, simply due to the price. Excellent quality output for cheaper than somebody with a degree. What would be the second order effects?

Human labor, being in less demand, would have to lower its price to compete. This is the death of the middle (and lower) class future we fear. But ironically the price of goods and services would lower as everything, even complex engineering, medicine, construction becomes affordable. With the right policy, human labor becomes cheap again. Maybe even competitive to machine labor in some niches. Improvements in machine labor could have a compounding effect on how affordable it can be for humans to survive.

So where's the gap here? Well, most wager earner's income worldwide goes towards housing. Food and water and medicine can be bottlenecked causing price gouging. Monopolies and lack of competition in the market can raise the prices of things until everybody is spending all of their disposable income on necessities. I think the price of human labor is currently very high (in the developed world) due rentier capitalism.

The transition will upend much of our economic investments and probably involve a great deal of human suffering until nations figure out the right mix of solutions.

It's not diseconomic, but it obviates a lot of constraints that required a person to manage a coordination problem, and those were a lot of jobs. Keynesian ideas about employment and GDP are just having an apocalypse. Like someone replaced the hole diggers and fillers with a conveyor belt and I would guess Keynes critics would have some predictive power here.

A couple developers can collaborate, but several need someone to specialize in coordination to yield additional value from more workers. Whether you call it management or orchestration, the need emerges at each threshold of additional complexity.

When AI collapses the productivity of 10 people into one, that's the disruption. The best AI user is going to suck all the opportunity out of the room for the others, and that's when layoffs happen. However, this assumes a fixed pie of opportunity. That's the real problem. As though there were only so much dirt to shovel.

FAANGs are old/mature and don't have exponential growth in front of them anymore, where opportunity within them is mainly about optimizing themselves but not growing in radical new directions. AI will indeed eat those optimization workforces alive. They resemble professions because law firms and doctors offices aren't growing either. They're mostly solving internal optimization problems, not finding net new growth opportunities.

The real effect is AI radically polarizing the difference between growing and dying in an org, where any firm that isn't growing fast enough will have its fixed opportunity pie collapse as AI disrupts this regulated oxygen supply. Whereas, growing firms without ceilings on the opportunity to deliver value will use AI to grow to the opportunity available.

Professionals can do fine if they re-orient themselves to new growth with different unit pricing, but yes, anything large and slow moving is probably going to get eaten.

A clearer look at the point about the internet being half/over-half AI generated content :

AI Now Writes as Many Online Articles as Humans

https://graphite.io/five-percent/ai-now-writes-as-many-onlin...

What about The Dead Human Theory?