Yves here. Get a cup of coffee. This is a particularly fine, wide-ranging discussion between Michael Hudson and Radhika Desai on the economic and financial effects and risks of AI. Some distinctive topics are how US policy toward research and development changed in the neoliberal era, and how China is taking a very different approach to AI implementation than the US. Readers are encouraged to chime in on Chinese AI deployment, but as Hudson describes it, they seem to use it as turbocharged machine learning, as in for performance of narrow tasks, and do not (generally) see it as a substitute for higher-order work, such as what was once called judgement. And of course, Desai and Hudson cover the almost certainty of an AI crash and how this bubble compares to stories ones of the past.
By Radhika Desai. Originally published at the Geopolitical Economist channel
Radhika Desai: Hello and welcome to another Geopolitical Economy Hour, the socialist and anti-imperialist conversation that illuminates the fast-changing political economy and geopolitical economy of our times. You are watching Radhika Desai, Geopolitical Economist, and I am Radhika Desai. Before we begin, let me remind you to please like, subscribe, share, and if you can, donate, if you like what you are seeing and hearing here.
That said, let’s get this show on the road. Today’s show is another one of our regular ones with Michael Hudson, and we want to discuss artificial intelligence. And we also want to discuss why President Trump wants to call it superintelligence. There’s been a lot going on with AI recently, and it is more and more dominating the conversation, sucking the oxygen out of practically every other conversation the world might have.
There has been a tussle going on between regulating AI and not regulating it. There have been a slew of legal cases that were launched against AI developers whose agents have been hacking public and private data systems. Pope Leo has even issued an encyclical about using AI to aid humanity and its development, and not to thwart it.
Meanwhile, within the United States, in community after community, in city after city, people are coming out. People are up in arms against data farms that AI will require. Everybody can see that their electricity bills are already going up because of these data farms, because of the enormous energy needs of AI. And they would be absolutely right to expect that their water bills will also go up. And what’s more, there will be scarcity of both electricity and water if things continue the way they are.
Meanwhile, politicians on the left and right are demanding that AI is such an enormous bonanza that the public must participate in the profits of AI. But then there is a separate question whether there are even going to be any profits, because while there was a bubble in AI stocks, it seems to have receded in favor of chipmakers, for the simple reason that whether or not AI yields any profits, while they continue to make this headlong burst of investment, the chipmakers will definitely be making enormous profits.
And part of the reason, at least, why the Western effort at AI, which President Trump has been saying must dominate the world, must defeat the Chinese efforts at AI… but actually what’s happening is that Chinese efforts at AI are already triumphing, because institution after institution, whether it be a corporation or a government, are choosing to use the open access and much cheaper Chinese AI models than their proprietary and very expensive American ones. And I know, Michael, that you’ve got a lot to say about it, so why don’t you begin anywhere? There’s just so much to talk about.
Michael Hudson: Well, I think the way to frame this issue is to see that AI is not only a technology problem, it’s a financial problem. And everybody sort of forgot now that in today’s neoliberal era, the whole issue is, how are you going to fund the research and development to develop a new technology?
Back in the 1970s, and economists didn’t play a role in this, but futurists did, people like Herman Kahn, Alvin Toffler, and myself, and people who later came on like Brzezinski, talking about technology dominance, the technotronic society. And there was…
Radhika Desai: Technocratic society, I think you mean.
Michael Hudson: Yes, right. There was a unanimous agreement that research and development for all of this technology, and they were thinking of Russia’s outer space technology, for instance, especially with the satellites and the landing on the moon and all of that. Russia seemed to have got a technological lead. How is America going to achieve the technological lead to maintain what still looked like it could be its industrial dominance?
Well, the result was the state had to play a rising role, and there were even discussions like, are the US and Soviet economies converging? Because introducing a new technology requires a lot of research and development, and it’s not the job of corporate industry to undertake research and development, because that’s a long-term problem, and the corporate managers, the financial managers of companies, have a time frame of three months, or maybe one year.
How do they increase their stock prices? Well, they use their profits largely for stock buybacks and to pay out as dividends to support the prices. They cannot afford the research and development because that would hold down the stock prices. And if they tried something like that, then corporate raiders would come in, buy them out and say, “Look, we can make a lot quicker profits in these companies by just cutting off the research and development.”
So who was going to fund this? Well, there was all sorts of discussion in the 1970s about a new class, a managerial class. And this was the early version of a professional managerial class. And the idea was, well, it’s really not the job of corporate industry under industrial capitalism to fund research and development. The government has to do it. Let the government spend all the money in preparing a technology to be commercialized, and then turn it over to industry, the campaign contributors to the politicians, to make a huge profit, or in fact monopoly rents is what they were really talking about, more than profits.
So it seemed that there was going to be government spending and a whole proliferation of think tanks, of government policy institutes, of research institutes, and the US government began a lot of research. Well, then came Margaret Thatcher and the Reagan revolution. And they said, no, we want to dismantle the state, and anything the state does is wasteful. Leave it all to private enterprise.
And so the end result of this neoliberal revolution was the situation you have today. Well, all right, the state isn’t going to undertake the study and research and development of AI. It’s going to be left to these seven companies that have dominated the whole stock market, and altogether have so far, it’s reported, borrowed up to $3 trillion, or at least signed commitments for this, to finance the research and development and introduction of all this without really making much of a profit.
Well, companies like Amazon were able to make a huge stock market gain without making a profit because they were putting in place, under Bezos, a monopoly. And the plan for the United States is, well, we’re going to create a huge monopoly in artificial intelligence to go with our information technology. And if we can monopolize this and block other countries from trade and investment and economic linkages with China, which is taking a lead, then we’re going to be in a position to achieve a huge choke point in the international economy to extract economic rents from all of this. That’s the plan. You and I have had a long discussion about this. What do you think of how this plan is going to work out?
Radhika Desai: Well, you’ve raised a number of points, so maybe I want to start by clarifying, let’s say, some of the things you said. So, one thing is that you’re right. In the post-Second World War period, there was a considerable emphasis on governments investing in technological research and development and innovation and so on. And the general idea was, in fact, at that time, that if the government were to invest, then they would also be required to impose upon corporations certain obligations and responsibilities to return to society what they are getting from society. And the reason why governments had to invest in these things is that corporations did not have the power to invest.
Now, on both of these counts, on governments investing, and then requiring corporations to obey certain types of social regulations, we are seeing a major reversal. So, on the one hand, the sheer amount of money that is sloshing around, thanks to nearly two decades of easy money policies, is enabling these private corporations to at least try to amass the funds, the trillions of dollars that they propose to invest, through private initiatives, through IPOs, and through borrowing. That’s what they’re trying to do.
However, let’s remember that while they are trying to do this, they are putting enormous pressure on the existing demands for this money. As you know, and you know better than me, Michael, we are living in a world that is indebted to the gills. Whether you’re talking about governments, whether you’re talking about corporations, whether you’re talking about households, all the three major sectors of every economy are in hock to banks and to lenders. So governments are now needing to borrow simply in order to pay interest. And the immense demand for money that is being made by these artificial intelligence companies is going to put further pressure on government borrowing, etc. So that’s the first thing.
The second thing is that governments, when they invested, they employed masses of scientists, including scientists who were devoted to some sort of public purpose, to ensure that whatever was being developed was socially useful, not necessarily just profitable, but somehow would enhance human life. Left to the private sector, the decisions that are being made about the way in which AI is to be developed, these decisions are actually detracting from human life.
Today’s 20-year-olds are worried about nothing more than the threat represented by AI. So, on the one hand, they are afraid that they are looking at a jobless future. We’ll come back to whether there will be a jobless future or not. It’s going to be even worse than a jobless future for young people. But at the same time, their parents are concerned that the existence of AI is preventing their children from learning, which is absolutely essential. If an entire generation stops learning… because without learning, there is no passing on of anything from one generation to another. So learning is absolutely central to life, and AI is actually dangling this completely artificial carrot that somehow human beings can now stop learning because machines are going to do it for them.
So, just taking that first point you made, Michael, about the corporate responsibility, let me throw in one last point before I ask you to come in. And that last point is that this whole interaction between public and private, in the past, it was undertaken with at least some balance between the private interests and the public interest. Today, what we are seeing is that the private interest is recklessly developing AI, harmful forms of AI, AI that is capable of doing a lot of destruction, AI that is currently involved in targeting civilians in wars.
This type of AI is being developed, and when the crash comes, and you and I know the crash is going to come, all this talk about the public purpose and how governments should support AI, if for nothing else than to compete with China, this is going to mean that taxpayers and ordinary members of these economies are going to pay like never before, even more than the 2008 crisis, for this coming crash of AI. So in a certain sense, the government is no longer the overall regulator of what’s happening. The government is merely a paid janitor that comes in and cleans up after the wild parties of speculation that are currently taking place in AI.
Michael Hudson: Well, you’re talking mainly about the social problems, which of course is what most people are concerned about. What’s the effect already of AI? I want to really focus on where the profits are going to come from. What is AI going to be used for? Well, basically it’s garbage in, garbage out. For instance, if you ask a question, what’s really happening, or for an explanation, AI can scan the entire internet and summarize all the views on the internet. Well, we all know that not everything on the internet is true, and a weighted average of these views, all sorts of things it can fantasize, which already is found.
One of the big demands for AI is going to be from governments, especially on spyware for citizens, to find out what’s happening. That’s a social effect, but it’s also, what’s the government going to pay for all of this spending? AI is a kind of central planning. And this kind of central planning, already my friend Alvin Toffler was talking about it in one of his books in the 1970s. He said one of the results of this sort of modern corporate management is to shift the actual work of distribution onto the consumers. The consumers have to do, already at that time, the work at self-checkout counters.
Well, right now the consumers have to do the work. If you have a problem with a company, you want to talk to it, or a medical provider, you make a telephone call, and in the past you’d get to a competent person who would be able to know what to do or who to send things through. Now you get this long AI torturesome runaround, especially in companies like Amazon, which is notorious for having to go through, “What could your question be about? Could it be A, B, C, or D?” Waiting for all of these things.
The corporations have found that where this AI is supposed to actually speed up the work of its employees, to hopefully get rid of them all, it’s actually increasing a lot of the work, not to mention the distraction of AI. So it’s hard to find out how AI is contributing to profits, much less technology rents, and without these profits and rents, how are all these debts going to be paid for, and where are the capital gains going to come from?
And among the problems, you pointed out all the demonstrations locally against putting AI in, because AI is going to be powered by electricity, and it’s going to use a lot of water. Well, it takes five years to get permits and construct electricity production. I think you and I have spoken about this in the last few weeks. And until that time, will AI bid up the price of electricity? Or is it going to somehow charge an enormous amount to AI companies to keep the electricity prices low enough so that local populations won’t revolt?
And of course, it also uses a lot of water. Well, if you’re trying to put AI out west, where a lot of them are being located, not all that water is there to cool the AI electricity that’s using all of these computer chips, which, as you pointed out, so far Nvidia and others are the big winners.
It’s a quandary that nobody can answer any more than they could answer the dot-com bubble of 1997 or the railroad bubbles of the 19th century. Every country had a railroad bubble. Austria, Germany, France, the United States, and all of the stock market crashes were railroad bubbles. People thought, here’s a monopoly, it’ll make a lot of money, and nobody really calculated, well, what are the costs and what are the returns? And of course, then the government came in and regulated the railroad rates and prevented these fantasy rents from being created, and stuff crashed.
What’s going to happen here? Well, you’ve already had Donald Trump promising the Silicon Valley campaign contributors, “Don’t worry, we’re not going to regulate you. You can charge whatever you want, because we realize you have to fund all of this research that you’ve had to do, not the government.” So all of this research is, as far as you and I and the public are concerned, a black box. We don’t know what’s going into there. It’s a black box that’s being created with the intention by every company: how do we control access to this? How do we have choke points where we can charge money for everyone who buys a little bit of an answer to the AI questioning?
Already you have a lot of companies, all of a sudden they let their staff use AI to help themselves, and then they get the bills for all of these units of AI. And they said, “My God, you’ve got to stop using it. You’ve got to ration it all.” You’re already seeing the corporations saying, will the profits of AI, if they make these super profits, is that going to bankrupt or lower the profits of all of us companies that have to use the AI? Something has to give.
Radhika Desai: Absolutely. Now, let me go back to the point you were making about exactly how governments, for example, will use AI. You mentioned spyware, but I think the second example you gave will be much more important. Yes, of course, what you meant by spyware is not the old meaning of spying, that is, spying on foreign governments. I think our governments have become completely incompetent at doing that, actually collecting any serious information about what other governments are doing. So let’s not even go there. But what they’re going to use it for is surveillance. Surveillance over their own citizens, surveillance over the data of their own citizens, and so on. So absolutely, this is going to happen, and this is going to lead to an extremely dystopian future.
But the other point that you made, which is that they are going to essentially replace human beings who used to answer phones as competent, responsible people with chatbots. And these chatbots are allegedly going to be as good as, or even better than, the humans they are replacing. But in reality, they are going to be incompetent, considerably worse, and you and I are going to tear out our hair every time we have to call a big corporation or a government office or whatever. We are going to be chatting with these chatbots, who are going to put us into doom loops which we will never get out of, and our problems will remain unaddressed. And that means that the ability of governments or corporations to provide the goods and services that they claim to provide will be gone. It will simply not be there. It will diminish massively.
But let me then go a step further as well. You were asking about how this AI is being developed, and where the profits are going to come from, and so on. And I find myself, when I’m thinking about this whole fantasy, and I call it phantasy with a PH, it’s a phantasy as in phantom, rather than a fantasy with an F. This phantasy is completely unhinged from any conception of reality. Let me explain.
It has been a fantasy of big capitalist employers for centuries to replace human beings from manual work. And they have indeed created many machines, which has reduced the amount of human labor, but they have not been able to eliminate human labor, even from manual work. What has happened instead in the neoliberal era, over the five decades or so past in which neoliberal policies dominated, which allowed corporations to do whatever they like… the idea was that if corporations could do whatever they like, they would innovate, they would create more and better machines, and they would eliminate human labor, and we would all live in some kind of society of fully automated luxury communism. None of this is happening.
What has happened instead is that capitalists have not eliminated manual work. They have simply shifted the production requiring manual work to cheaper wage locations, or handed it over to segments of the labor market that are composed of immigrants that are brought in especially for the purpose. So now, if you consider this fact, that basically capitalism has not even eliminated human labor in manual labor, then how are they going to eliminate it in intellectual labor?
But this is their fantasy. This is why today, when you go to any kind of social event with the members of the professional managerial class, they are all worried. And they are worried because they will be eliminated, but they will not be eliminated because capitalism has been able to develop competent replacements for them. On the contrary, they’ll be eliminated under the ruse that capitalism has been able to do so. But when this claim is tested, when this ruse is tested, it will be found wanting. So this is the situation.
Now, the other part that we must point out, and here we must bring in China. Both China and the United States are competing in the field of AI. Most people think that developers in both countries are trying to aim for more or less the same thing. And if you think so, you could not be more wrong.
In the United States, in the hope of attracting these vast quantities, billions and billions of dollars, the likes of Elon Musk, or whoever these people are, Mark Zuckerberg and Sam Altman and so on, the likes of these people are basically promising that they are going to create an intelligence which is even greater than human intelligence. Look, people wrote about this back in the fifties, and they said this is not possible, because ultimately human beings are programming this, and human intelligence evolves. Anyway, there are so many different reasons. There is some statistic about the number of connections that are made in even a very large AI machine, which is dwarfed, like thousands of times, by the number of connections in an ordinary person’s human brain. I mean, think about it.
But anyway, the point I’m trying to make is that in order to attract this money, in order to attract government support, in order to attract government subsidy, in order to attract government deregulation, which is necessary to allow these companies to do what they like, they are massively overstating what can be achieved. This is in the United States. And this is where, by the way, this sci-fi model comes in. We’ve talked about this before. I wrote a paper when DeepSeek came out, and I said how DeepSeek has upended Silicon Valley’s sci-fi-n-fi model. So it’s based on a lot of science fiction, and it’s based on a lot of financial fiction. So it’s sci-fi-n-fi.
So this is the Western model, and this has a long lineage. It goes back to the dot-com bubble, and it even goes back to Star Wars, where they created this fantasy of how one missile is going to bring down another missile. This is much harder to do, as Israel is discovering now.
But anyway, let’s come back to China. China, meanwhile, is pursuing a completely different strategy. It is setting finite tasks. Artificial intelligence itself is a bad word, not because the real word should be superintelligence, but because it is nothing more than magnified computing power, quantum computing power. Fine. So they know the limitations of quantum leaps in computing. Then they program the computers to solve finite problems. So how can solving this problem aid human beings to become better doctors, or become better actuaries, or become better whatever it is, whatever requires a lot of computing? So this is what China is doing, and so the two approaches are completely different. I’ve gone on for a while, so I’ll let you come in. I have a few other points to make, but I’ll do it later.
Michael Hudson: I think we have a lot of points. Well, your point that you’ve just made, maybe you could think of automatic intelligence as eliminating intelligence itself, eliminating human judgment.
Radhika Desai: Michael, I have to interrupt you very briefly here. My favorite joke about artificial intelligence is, one professional says to the other, “Are you worried about the rise of artificial intelligence?” So this guy answers and he says, “No, but I am worried about the decline of natural intelligence.”
Michael Hudson: Yeah, exactly. Well, you’re quite right to bring up how different China’s automatic intelligence is. We’ve all seen pictures of China’s factories making entire cars from beginning to end with robots.
Radhika Desai: Dark factories.
Michael Hudson: This is what you just called limited, finite problems. Think of it as a servo mechanism. You already had in the 18th century singing birds that would sing certain songs, and artificial robots that appeared to be human beings. Well, robots can do a lot of things now, like make automobiles, that they couldn’t do before, but there’s not having to do a judgment about how to structure the overall context. Real intelligence is thinking of the context and putting things in perspective, to interrelate things that at first may not seem related. But the automatic intelligence is tunnel visioned, whereas real human…
Radhika Desai: Let me briefly interject and remind us, you and our listeners, this AI that is developed in the West has been encouraging people to commit suicide.
Michael Hudson: Yes. So somehow it’s automatic, you follow instructions. Well, how many of us have called a company to ask a question, and we get someone who’s worked there for a week or two and is just reading the script of “here’s how we answer the questions,” but doesn’t have any idea how to answer, “Could we talk to a manager?” Somehow you always want to get to the manager who can solve the problem, and not just have somebody read this long script to you, and you say, “No, that’s not what I’m talking about.” Well, AI has a problem in deciding what’s really important to be talking about here. That’s the problem. That’s where judgment comes in. That’s where some, what do you call it, meta-intelligence, or just human intelligence, judgment.
Radhika Desai: Yes, exactly. Judgment is not easily developed.
Michael Hudson: And China, as far as I know, is not trying to do that. It’s human beings that are deciding China’s policy, and they seem to have a policy of selling and marketing their version of automatic intelligence, the open source, to the rest of the world. They’re not trying to make monopoly profits. They’re not trying to create systems that have proprietary kill switches, where you can all of a sudden press a button and, if you have automatic intelligence to operate centrifuges in Iran’s uranium thing, all of a sudden you press a button and they spin fast and explode.
With an open-source type of intelligence, as opposed to a closed monopoly source, you can protect yourself. Once you monopolize artificial intelligence, there’s no way of protecting yourself from all sorts of things that it may do, that you may not know what it’s doing or might do.
Radhika Desai: Exactly, Michael. And I want to take the conversation in a slightly different direction. I particularly want to talk about Trump’s really remarkable turnaround on AI. This was quite clear. Earlier, when Trump was running for election, he was against AI, right? And then, the moment he got elected, essentially all the big AI bosses agreed to contribute to him, agreed to contribute to his inauguration, and since that time, Trump has been essentially favoring the unregulated growth of AI. He has been investing in it himself.
So in all of these ways, Trump’s turnaround has been very important, and it seems as if the United States is basically scraping the bottom of the barrel of profitable ideas. And this is about the best it can come up with, and the best it can come up with, which is AI, it can already be seen, within about a year or so of AI hype, that it is very iffy whether they are going to make profits, which is why Magnificent Seven stocks are not going up, and other stocks are going up, and so on. So this is already, in many ways, cratering.
So now what’s happening is that it seems to me that the groundwork is now being laid for the eventual bailout of these companies when they crash, because it looks as if the crash is imminent. Now, what do I mean by that? What I mean is, how come suddenly these guys who were talking about how they are the super geniuses and so on, like Elon Musk and Sam Altman and whatever, these guys who think of themselves as the Einsteins of this world and super Einsteins of this world, how come they are now demanding regulation?
The reason they are demanding regulation is that it is now time for them to involve the state. Previously, they didn’t want to involve the state because they thought that they could keep this game going for longer. They would keep getting money, people would keep investing and sending their stocks up, and they would have a good time. Now their stocks are going down. So they are now invoking regulation, and I can think of a number of reasons why they are doing so.
Number one, regulation means you are involving the state, which means that when the crash comes, the state can then say, or at least politicians like Donald Trump can then say, “Well, look, this is important for us all. This is important for humanity. This is important in America’s fight with China. We must bail these people out.” So that will be done.
Secondly, once you get regulation, you then create the structure of incentives that allow some companies to embed themselves in monopoly positions vis-a-vis other companies. And so this is precisely how you create the choke points that you were talking about, that allow them to charge the tolls that they would need to charge in order to become, like Jeff Bezos, essentially another such monopoly. So regulation is necessary for that.
And of course, the other reason that they are invoking regulation is because this is a way of deferring their IPOs. “Oh, we all must stop.” And remember the way in which they have demanded regulations. They have demanded regulations by invoking the possibility that the AI they are going to create is going to be so powerful as to destroy humanity. So this is the danger that they have created, and then they say, well, we must have regulation. But the real danger that they are trying to fend off is the danger that their stock prices, their IPOs, will not succeed. So they are postponing these IPOs, they are postponing these demands for more borrowing and so on. So these are the various reasons why regulation is required.
And one final point. People say that now, in this war, funnily enough, Donald Trump has parted company with the Magnificent Seven, the AI developers, and his only friend among the big CEOs is Jensen Huang of Nvidia, who is in a position to profit just at the present moment. He does not fear, because he’s selling chips hand over fist. So that’s one point.
Now, the connected point I want to make is that in the mainstream Western media, Trump and President Xi are being put in the same box. We are being told that neither of them want AI regulation. But actually, that is not true. China is already regulating AI. China is already requiring its AI developers to be responsible, to develop AI in the interest of humanity, not in a way that destroys humanity, etc. And they have all sorts of checks and balances built into the system. The American system does not have that.
The reason why President Xi does not want regulation is because he does not want to hand over the keys to regulation to the Americans, and there is already plenty of regulation in China. The reason Donald Trump doesn’t want regulation is because he still believes in this myth that the AI is going to produce its… anyway, he’s now friends with Jensen Huang, so he’s going along with that.
Michael Hudson: Well, you mentioned “destroy humanity,” and I was thinking, when people do discuss how it could do that, what are they talking about? They’re really talking about a kind of scamming. They’re saying, what does AI do? It can look through computer programs on every site and say, what’s the weak point? What’s the attack point? And that means that people can use AI and find a way of hijacking other people’s sites. You could hijack an electric utility’s site and turn off the electric power for a whole city, and who knows what crisis that will do.
There’s already been a hijacking of hospital records. You’ll have AI find a breach in the hospital systems, and it’s good at that, better than individuals poring through hundreds of thousands of commands to find what’s the back door. Well, we know that the CIA and government have put back doors into almost all of these programs that everybody uses, for their own purposes. And AI can find these back doors, and people can essentially ransom records from financial institutions, from banks, from hospitals.
That’s, I think, what they’re talking about. The fact is that there is such a sense of vulnerability to break-ins from the existing internet and all of this, that AI can essentially break in through all of these barriers that haven’t been all that carefully made, as we’re reading every single week, and as the FBI produces statistics every year on ransomware, how much it’s taken from the economy. They don’t include ransomware in the GDP of many countries, but that’s certainly one of the great sources of fortunes these days, wherever it’s coming from. So I think that’s just one of the problems.
But it all comes back to, if not destroying humanity, will it destroy the profitability of corporations, how they work? Will it lead to unemployment? Well, there was a whole argument in the 19th century: will all this cotton spinning and textile-making machinery cause unemployment? And the answer was, well, instead of weavers and hand loom makers, you’re going to have people designing weaving machines. All right, there wasn’t any crisis that the Luddites had thought about.
But this is something different. They’re not replacing a kind of labor that’s already being done. They’re creating a new kind of problem, and as you and I have been saying, choke points. AI, because it’s rent seeking, rent extraction, how much can we make, instead of providing it as open source, let the whole world, a hundred flowers bloom… this is what happens when you create a system of choke points and proprietary dependency.
Radhika Desai: Yes, and I just want to say that these new schemes that are being proposed, and it’s unfortunate that someone like Bernie Sanders, who should know better, is being drawn into it, this idea is that somehow AI is going to be such a magnificent profit producer and so on, that the public must take a share in it. But this is simply preparing the ground for the ultimate bailout that will come. Because once you say that somehow the public has an interest in it, then the public will have a liability to bail it out. And I think that this is going to be really, really nasty.
There are some people who think that this idea is wrong because somehow the profits will be made at the expense of third world countries, and that this is some kind of new imperialism. No, it is very, very doubtful whether any profits will be made at all, because they are basically attributing way too much power and success to the American way of developing AI.
Michael Hudson: Well, look at the populism in all this. This is the whole point. The promise is it will give a lot of money to people, but before there’s any profit or rents to be distributed, how are you going to pay for all of these debts that are falling due, so the companies don’t go under and have to be sold at auction? I think that’s going to be the whole thing.
Now, the Sanders legislation envisages legislation for an estimated $7 trillion. And he says, well, if the government’s going to provide all these guarantees, we should get half the company. And he said things like, well, suppose you have an annual dividend of 5% on the $7 trillion. Everybody in America can have a $1,000 check.
I’m sorry, I get the throat when I raise my voice for all of this. This is what’s called blue sky thinking. And maybe Sanders is trying to say, “Well, we know that you on the Democratic National Committee are trying to isolate all of us socialists, social democrats in there. But look at what we can do for your biggest companies. If we give this to you, won’t you support our socialists and maybe let us bring about socialized medical care,” all the other things that he wants. I think there’s a kind of trade-off here that Bernie is trying to make.
And if AI is going to make all of its money not only in the American economy, but above all internationally, that brings in the geopolitical dimension that our whole show usually talks about. Look at all the pressure that the United States has put on Europe not to develop its AI at all, not to do anything that would become a rival of American firms, and not to impose the regulatory moves and the taxes that have been proposed in Europe. And Trump is very confrontational with Europe on this. “No, this is going to be our monopoly, not yours.” So Trump envisions a lot of the profits being made on foreign countries, certainly on Europe, and the implication is, on all countries that he can convince not to use China’s open source that it is providing free to anyone.
I can see you have some ideas on that.
Radhika Desai: No, no, I said it ain’t happening. Trump may cry himself hoarse trying to get people to use American AI instead of Chinese AI, but it ain’t happening.
Michael Hudson: Yes. You and I have been reading the news about all of this, and it’s obviously going to be one of the big problems. It’s been forecast that AI has to earn $6 trillion a year by 2021. Well, right now, AI and its related information technology is, what is it, 7% or 9% of GDP. As much as Americans pay for food, they’re paying for AI. And if this goes up to 16%, this is going to be one of the largest sectors in what is becoming a rent-extractive economy that will be dominated by AI, just as the oil industry is a choke point, and all the other choke points that are made in the economy.
How are you really going to hold other countries into this technology if other alternatives are available? And that’s where you get the difference between China’s approach and the American approach. And what will not only Europe do, but what’s Africa going to do? Well, China apparently is making huge inroads into Africa. They can’t afford the trillion dollars a year, or whatever it’s going to cost. It’s no contest.
And then, AI requires raw materials to be made. China produces the raw materials. It requires electricity. China has developed huge data farms that are solar powered in its western provinces, because they have plenty of sunlight there and not that much competition. Well, Trump hates solar power as much as he hates wind power. It’s all going to be oil usage. So what does that mean for Trump’s version of AI, powered by the oil industry, with all of the oil problems that we’re having now with the Iran war?
I mean, talk about what does automatic intelligence not take into account, and how can they succeed? Just look at the problems when you try to interrelate these different dimensions together using human intelligence instead of automatic garbage in, garbage out intelligence.
Radhika Desai: Exactly. Now, we should wind up this conversation, so I want to make three points, and then ask you to make any last comments you want to make. Although I should preface my three points by saying it’s really interesting that you continually call artificial intelligence “automatic intelligence,” which would also be quite an interesting way of putting it. So your slip of tongue might be an interesting one.
Anyway, the three points I want to make. Just when the world is suffering from the effects of climate change, with summers full of wildfires and the hottest summers ever, and by the way, the coming year is going to be even worse, just when all of this is going on, the West has stopped talking about addressing climate change. And the reason, I understand, they have stopped doing so is because Chinese companies are now in the forefront. China is now developing the technology to combat climate change, so the West has stopped talking about combating climate change.
But worse, with these enormous energy demands of AI, they are going to make the problem a lot worse. And the American government wants to give these companies essentially captive power stations. And of course, these captive power stations will be built in a new environment of deregulation, and they will also be nuclear, and you can just imagine the nightmares that will be built for power generating capacity for these data farms. So the whole energy and climate change dimension is absolutely critical. If humanity is not going to hell in a handbasket very soon, we have to do something about the American way of doing AI, because the Chinese way is not so energy intensive.
The second is, of course, the water problem, which is equally an environmental problem. The amounts of water that will be used by these data farms is absolutely horrific, and again, we need to pay attention for the same reasons as the energy one.
And the last point is also very interesting. With these hacks that are taking place, there have been a slew of lawsuits against the AI developers, and apparently one of the key legal issues is, who is responsible? Imagine that you develop an AI which is driving a car, and then there is an accident and somebody dies or is injured or whatever. Now, who is responsible? Is it the person who owns the car, or is it the company that developed the AI? And I would say that the companies are going to try their damnedest to turn the law in the direction of blaming the owner of the AI product. Whereas, in fact, it is in the public interest to orient the law towards blaming the people who develop the AI. But this is going to be a very key issue.
Michael Hudson: Yes, you put your finger on the political issue and the environmental issue. And there’s the geopolitical issue. America tried to solve these problems by saying, well, instead of putting AI here in America, let’s put it where the energy is. Let’s put it in the Emirates and in Saudi Arabia. Well, what’s the first thing that got bombed by Iran? The data centers of Amazon and other companies in the Emirates and Saudi Arabia.
Radhika Desai: And also there, Michael, the cooling costs would be just enormous.
Michael Hudson: Yes.
Radhika Desai: But anyway, Michael, this has been a fantastic conversation. I’m sure that as there are new developments in this field, we’ll be returning to this topic time and again. So, folks, I hope that you found this conversation interesting. Thanks to you for attending, thanks to Michael for taking part and being here to talk with me about this, and until next time, goodbye.
By
