AI Drama: DeepSeek, Nvidia, and China

A Chinese firm, DeepSeek, claimed to develop an AI model at a fraction of the cost used to develop competing models. This has caused quite the stir, but what is actually going on?

China’s domestic chip manufacturing is dated and their attempts to catch up have fallen short. I’m always skeptical when approaching topics involving Chinese tech claims, but if the DeepSeek breakthrough is true, it would mean a massive AI leap has occurred. This would reduce the reliance on high-end chips and drastically mitigate supply chain vulnerabilities.

Nvidia had designed downgraded chips for China, but even those were banned by the Biden administration in late 2023. So, even if DeepSeek’s breakthrough turns out to be real, the Chinese will struggle to fully capitalize on the opportunity due to hardware restrictions and access. And Trump could make those restrictions even tighter.

But again, I’m doubtful that we’re getting the full picture out of DeepSeek and I suspect we’ll know what actually happened very soon.

Prefer to read the transcript? Click here

Here at Zeihan on Geopolitics, our chosen charity partner is MedShare. They provide emergency medical services to communities in need, with a very heavy emphasis on locations facing acute crises. Medshare operates right in the thick of it, so we can be sure that every cent of our donation is not simply going directly to where help is needed most, but our donations serve as a force multiplier for a system already in existence.

For those who would like to donate directly to MedShare or to learn more about their efforts, you can click this link.

Learn about medshare here

Transcript

Hey everybody. Peter Zeihan here. Coming from Colorado. Lots of people have asked me to comment about recent things that are going on with artificial intelligence from a market point of view. Let me begin by saying I never comment on market movements, and this is not going to change today. What is going on is we’ve had a bit of a rout in the United States for a company called Nvidia, which designs the ultra high end GPUs, as are graphics processing units, which are the advanced chips that are used in almost all artificial intelligence training and applications. 

Specifically, a Chinese company called Deep Seek says that it managed to generate an artificial intelligence model, at 1/20 of the cost of what Nvidia has been able to do or what companies had been able to do with Nvidia chips. And in doing so, we basically saw more market value, bleed off of Nvidia in about eight hours than most countries have economies. 

It was ridiculous. It’s also probably completely fabricated. Keep in mind that almost anything that comes out of U.S. high tech is proprietary. So the information is circumspect, and almost anything that comes out of China has a degree of fraud involved. And if you’re using either of these two things to color your trading decisions, I can’t help you at all. 

But what I can do is talk about where this leaves the industry and what it would mean one way or another. So let’s assume for the moment that this is all fraud, in which case this just all blows away really quickly. And a lot of people who have been trading money and trading stocks feel really stupid in the not too distant future. 

I think that’s the most likely outcome. But let’s assume for the moment that it’s real. And this Chinese firm working with subpar chips has been able to make superior, large language model trained AI systems to what the American bigwigs like OpenAI can do. 1/20 of the cost with dumber chips. That would be significant. One of the big problems that we have with the chips right now is the supply chain is so grossly vulnerable that it doesn’t take much of a disruption anywhere in the world to break the ability of the world to make these high end chips at all. And if that happens, then we’re going to have to decide how to ration our chip usage. If you can drop the cost of training these new models by 95%, obviously, that makes, that concern a little bit less of a problem. 

But even within that, there’s going to be some very clear breaks geopolitically. So again, assuming that this is for real, let’s look at the chips that were used. The high end chips that are basically three, four and five nanometers that provide the backbone of most AI operations. The ones that we think could be used for training AI models are all under sanctions as of 2022, and the chips that were used by deep seek to do their operation were custom built chips, basically. 

Nvidia looked at the sanctions, said that they still wanted a market in China. So they worked with the Chinese company Deep Seek to design a chip that would comply with sanctions, and they dumbed it down, if Deep Six announcements are true. Clearly Nvidia did not dumb it down enough, but the Biden administration, in one of its last actions, decided a couple months ago that, Nvidia has not been playing in good faith and basically banned all of these chips as well. 

So if deep secret was able to do this with the dumb down chips, those dumb down chips are no longer accessible to the Chinese economy writ large anyway. Moreover, the most advanced GPU that the Chinese system is capable of manufacturing is about a 14 nanometer. So you’re talking about a dozen generations behind where we are right now? 

It’s worse than it sounds, because even to make those 14 nanometers, they have to use imported equipment and a lot of imported labor to do the quality checks. Yes. Yes, yes. We have had a couple Chinese firms make something in the seven nanometer range, but it’s been really clunky. It’s been a massive power hog. They get something only like a 20% success rate. 

And most commercial chips, if they’d have anything less than 90%, are considered failures. So if they push and push and push and push and push and throw in an exorbitant amount of resources into it, they can make a chip that is still half a dozen generations behind what would be necessary to do an AI model. So if if Deep Seek is not lying, then we would have a breakthrough in AI that the Chinese would be utterly unable to participate in, and that would generate a very interesting world. 

But we’re going to know how this is going to go either way in the not too distant future, because everything that Deep Six says that it did is open source. And we’re going to be able to try to recreate it, in any number of companies, in any number of countries in the not too distant future. One other reason that I’m really doubtful the DPC actually did this. 

There are basically two ways you train AI. You basically you teach it like in school, like, you know, x plus two equals y three plus four equals seven. And once you’ve done that and built this kind of catalog of basic information, and you then let it loose on bigger problems, what deep seek is saying is they just skipped the first step, saying that you don’t need to teach basic math to the computers. 

You can just set them loose on calculus, if that’s true, which I find exceedingly unlikely. This can be replicated in a matter of weeks to months, not years. So we’re going to know, in the not too distant future, just what actually happened and how many people in the financial world jumped the gun. In a really silly way or not? And even in the, you know, worst case scenarios, the right term, even in the situation where deep secret really, really did generate a bit of a revolution, everyone’s going to be able to play except for the countries that can’t get the chips. And if there’s one thing that we understand about the Trump administration versus the Biden administration, is they have no problem taking sledgehammers to specific companies. 

So if it is the understanding of the Trump administration that Nvidia has been playing fast and loose with sanctions, you can count on Nvidia having much bigger problems than simply a Chinese company coming up with an a more efficient model.

Can AI Replace Those Retiring Boomers?

*This video was recorded in mid-July, prior to Peter departing on his backpacking trip.

As the baby boomers age into retirement and Gen Z fails to satisfy the gaping hole left in the labor market, will artificial intelligence be able to help mitigate some of the fallout?

AI has solidified its spot in the labor mix of the future, but “when?” and “how?” are the looming questions. Producing AI-capable chips remains highly complex and reliant upon a web of critical international suppliers. Until that supply chain is slimmed down and fail-safes are added, production disruptions will continue to be a huge factor in limiting the AI buildout.

On the other front, the nature of AI lends itself to better take on white-collar work. Most people picture AI taking care of the manual and mundane tasks no one wants, but it is better suited to a lifestyle “in the office”. So, the places we need the help most, like blue-collar jobs, won’t get much help on the AI front.

AI’s transformative effect is coming, but it won’t occur as quickly as many expect. Think in the 40s rather than the 20s. When it does arrive, the productivity in white-collar jobs will skyrocket, but fields that really need some extra hands (or hand effectors or whatever you call robots’ metaphorical gripping mechanisms) will still be scrambling to navigate the labor shortages.

Prefer to read the transcript? Click here

Here at Zeihan On Geopolitics we select a single charity to sponsor. We have two criteria:

First, we look across the world and use our skill sets to identify where the needs are most acute. Second, we look for an institution with preexisting networks for both materials gathering and aid distribution. That way we know every cent of our donation is not simply going directly to where help is needed most, but our donations serve as a force multiplier for a system already in existence. Then we give what we can.

Today, our chosen charity is a group called Medshare, which provides emergency medical services to communities in need, with a very heavy emphasis on locations facing acute crises. Medshare operates right in the thick of it. Until future notice, every cent we earn from every book we sell in every format through every retailer is going to Medshare’s Ukraine fund.

And then there’s you.

Our newsletters and videologues are not only free, they will always be free. We also will never share your contact information with anyone. All we ask is that if you find one of our releases in any way useful, that you make a donation to Medshare. Over one third of Ukraine’s pre-war population has either been forced from their homes, kidnapped and shipped to Russia, or is trying to survive in occupied lands. This is our way to help who we can. Please, join us.

Support medshare's ukraine fund
Support medshare's global efforts

Transcript

Hey, everybody. Peter Zeihan here, coming to you from the Lake of the Ozarks in Missouri. Today, I’m taking a question from the Ask Peter forum about artificial intelligence. The question is, with labor shortages looming due to the retirement of the baby boomers, the largest workforce we’ve ever had, and the new generation, Generation Z (the Zoomers), being the smallest generation ever, can increases in productivity, such as through artificial intelligence, help round out and fix these issues? The answer is… maybe.

There are a couple of problems here. First, it’s very unlikely that we can sustain the level of production of artificial intelligence-capable chips necessary to fundamentally transform the American or global workforce. The issue is that thousands of companies are involved in manufacturing these chips. Yes, 90% of these chips come from a single town in Taiwan, which poses a security risk, but the problem is even bigger. Most companies in the supply chain for these fabrication facilities only make one product for one end user, which is TSMC in Taiwan. If something happens to a few of these companies, we could lose the ability to make these chips at all. So the idea of having massive server farms crunching data for artificial intelligence is probably not going to happen this decade and might not happen next decade. We’ll need to build a new ecosystem for that, and it takes time.

The second problem is the nature of artificial intelligence itself. There’s a saying that’s becoming popular: “We want artificial intelligence to do the dishes so that we can spend more time writing, not to do our writing so that all that’s left is the dishes.” The issue here is that artificial intelligence primarily helps with white-collar work, like brain work, correlation tasks, paralegals, researchers, writers, and editors. These are the sorts of jobs that AI is most likely to slim down. But that’s not where the labor shortage is in the United States. Over the last 50 years, baby boomers have focused on getting their kids into college and into high-value-added white-collar jobs.

However, in a world where China is fading quickly and the United States is experiencing an industrial renaissance, we need electricians, welders, linemen—people to physically move and make things. These are blue-collar jobs, not white-collar, and artificial intelligence can’t help nearly as much with them.

So while artificial intelligence will be a part of the future, I don’t think it’s going to happen as quickly as some expect. When it does, it’s going to reshape industries like finance, making it more efficient to move money around, which could reduce the need for large numbers of coders in places like New York. It’s going to increase the productivity of white-collar workers, so we might not need as many people handling tasks like processing insurance claims in doctors’ offices. AI might also help crack the genome to boost agricultural productivity and will certainly play a role in defense and cryptography to enhance security.

But it’s unlikely to have a huge impact on manufacturing. For that to happen, we don’t just need AI; we need robotics—mobile robotics—and AI isn’t quite there yet. It’s progressing, but we’re looking at the 2040s, not the 2020s, for significant developments in that area. So can AI help? Sure, at the margins. But you’re probably still going to have to do your own dishes.

Hollywood on Strike: The Future of Writers and Actors

The entertainment industry is changing, and the writers and actors are making sure their voices are heard. With all the technological changes hitting Hollywood, strikes like these shouldn’t come as a surprise.

As we’ve seen with the fall of Blockbuster or the demise of RedBox, this industry is constantly evolving. The old revenue models must be reimagined and updated as we move into the streaming service era.

But what does that mean for the screenwriters who can no longer determine viewership metrics and get paid royalties? Or for the background actors sacrificing their digital rights so CGI copies of them can be made and used forever?

This strike is an inevitable step in Hollywood’s evolution, and these questions will change the industry forever.

Prefer to read the transcript? Click here

Here at Zeihan On Geopolitics we select a single charity to sponsor. We have two criteria:

First, we look across the world and use our skill sets to identify where the needs are most acute. Second, we look for an institution with preexisting networks for both materials gathering and aid distribution. That way we know every cent of our donation is not simply going directly to where help is needed most, but our donations serve as a force multiplier for a system already in existence. Then we give what we can.

Today, our chosen charity is a group called Medshare, which provides emergency medical services to communities in need, with a very heavy emphasis on locations facing acute crises. Medshare operates right in the thick of it. Until future notice, every cent we earn from every book we sell in every format through every retailer is going to Medshare’s Ukraine fund.

And then there’s you.

Our newsletters and videologues are not only free, they will always be free. We also will never share your contact information with anyone. All we ask is that if you find one of our releases in any way useful, that you make a donation to Medshare. Over one third of Ukraine’s pre-war population has either been forced from their homes, kidnapped and shipped to Russia, or is trying to survive in occupied lands. This is our way to help who we can. Please, join us.

Support medshare's ukraine fund
Support medshare's global efforts

Transcript

Hey everybody. Peter Zeihan here coming to you from Seattle. It is late October and this is one of those that can store. So you probably won’t see it till November. But whatever the topic I want to discuss today is strikes in Hollywood of all places. We’ve got two separate strikes going on, one of which has at least temporarily been resolved, one involving the writers, which for the moment is behind us.

They’ve settled and one involving the actors. And the issue is, honestly, one of technology with the advent of a sufficient processing power to allow streaming services over Internet connections. We now have a number of major providers such as Hulu or Apple or Amazon or Netflix that are not just providing legacy shows, but new shows and transmitting them to their end users and consumers to a completely different network that doesn’t use the normal TV radio approach, VHF, things like that.

Well, this means different models used to be that all the income came pretty much from advertising, and now there are different ways to do it with, say, a per month subscription charge. You’re going to see more and more and more in this going forward. So we have to split it in those two factions. So first, the labor that deals with the screenwriting and then the labor that deals with the acting.

So screenwriters, this is going to be an ongoing issue largely because of the revenue play. When screenwriters used to produce things, whether it was for television or movies before, they’d work directly with the movie house, which would generate their income from either advertising or ticket sales. Now that you’ve got streaming, the question is what constitutes readership or viewership?

Because it’s not the same model. And is it something that’s a once and done? Is it something that gives them income over time? Because unless you’re an A-lister where you can demand whatever terms you want, everyone else has to kind of suffer through and for the writing. The rise of air is providing more and more support for people who are particularly creative and leaving everyone else by the curb.

But as problematic as that is, from the point of view of the writers, it’s a disaster for the actors. We already have technology that allows us to fill in the background with either a complete greenscreen generated system or even to a certain degree, extras. And I think the the two movies that have demonstrated how this technology is coming along, the best one is Ant-Man two, the one where we had some characters who in today’s world are in their sixties.

But we had a couple of scenes where they were shown when they were in their twenties, in their thirties, and you can use the technology to make people look younger. We’ve also had World War Z, which is a hot mess, great book, horrible movie. But when you had the swarms of zombies, you know, those weren’t real zombies. They were all CGI generated.

And we could see how they looked very, very real. Well, you play both of those movies forward because those are both five years old now. And we’re getting to the point where you can film an extra from a number of different angles and insert them into the background just fine. And so part of the reason that the actors were striking is because they were being asked to sign documents to basically surrender their future digital rights.

So you’re an extra in this movie, you’re filmed, and then they reserve the right to recreate you royalty free in the future. And obviously, if you’re a low ranking actor or an extra an aspiring actor that pretty much end your career. It won’t really hurt the A-listers, but pretty much everybody else would be left in the cold and very soon we’re going to be able to take footage from people who are dead and use A.I. Technologies to put them into leading roles if we want to.

And so the the balance, the ability for you to profit off your skill set in your presence all of a sudden isn’t there anymore? And it’s a question of who generates the revenue at the moment, the law suggests it’s the people who control the A.I. driven software. It actually designed the movies in the first place. So from a writers point of view, this is going to get a little bit stickier as we move forward, but it’s going to be more of an evolutionary process.

But for the actors, you’re actually looking at the evisceration of an entire class of people, and that will take with it the way movies are produced. Because if you can just have a handful of A-listers and be listers who are doing kind of a number of the main roles and have the star power to draw people and everything else is computer generated and it looks as real as the real thing.

Then we’re in a fundamentally different model for everything and we’re probably going to be in that environment by the end of this decade. So one way or another, Tinseltown is going to be very, very, very different, and it’s probably going to have a lot fewer people.

Is the AI Revolution Here?

Today we answer the question that’s been in the back of everyone’s mind since The Terminator came out – is Skynet taking over? The short answer is no, but let me explain…

The AI capable of decision-making and judgment (i.e., Skynet) is nowhere near being ready to go, and won’t be for at least another 30+ years. On the other hand, applied AI (i.e., mission-specific and code-driven) is great at completing tasks as long as it’s written into its program.

So AI that’s used as a copywriter or chatbot is already up and running, but the omnipotent robots that hunt us down and try to take over the world…ya, not happening anytime soon.

The biggest takeaway is that change is coming, but not in the scary revolution type of way that we’ve been hearing about for years.

Prefer to read the transcript of the video? Click here

Here at Zeihan On Geopolitics we select a single charity to sponsor. We have two criteria:
 
First, we look across the world and use our skill sets to identify where the needs are most acute. Second, we look for an institution with preexisting networks for both materials gathering and aid distribution. That way we know every cent of our donation is not simply going directly to where help is needed most, but our donations serve as a force multiplier for a system already in existence. Then we give what we can.
 
Today, our chosen charity is a group called Medshare, which provides emergency medical services to communities in need, with a very heavy emphasis on locations facing acute crises. Medshare operates right in the thick of it. Until future notice, every cent we earn from every book we sell in every format through every retailer is going to Medshare’s Ukraine fund.
 
And then there’s you.
 
Our newsletters and videologues are not only free, they will always be free. We also will never share your contact information with anyone. All we ask is that if you find one of our releases in any way useful, that you make a donation to Medshare. Over one third of Ukraine’s pre-war population has either been forced from their homes, kidnapped and shipped to Russia, or is trying to survive in occupied lands. This is our way to help who we can. Please, join us.

CLICK HERE TO SUPPORT MEDSHARE’S UKRAINE FUND

CLICK HERE TO SUPPORT MEDSHARE’S EFFORTS GLOBALLY

TRANSCIPT

Hey Everybody. Peter Zeihan here. We’ve had a lot of questions come in about AI and what that means for the workforce moving forward. What sort of activity should we expect to be replaced? What does this mean for economics and labor and politics and are there any obvious winners either in terms of geography or sectors? What popular when it comes in is whether or not this is going to hit red states or blue states more. For example, the cop out answer is we don’t really know yet because we’re dealing with technologies that have yet to be invented. But there are a few general guidelines we have.

First of all, it’s not so much that jobs get created or destroyed, it’s that they change. And it’s pretty common when you’re dealing with an environment that has evolved because of technology. You know, jobs evolve, too. We’ve been talking about technology overwhelming the workforce really since the onset of the Industrial Revolution, and it obviously generates changes. We don’t all live on subsistence farms anymore. The trick is whether or not the technology evolves faster than our political ability to adapt to the changing workforce conditions. And I would argue that at least at the moment, we’re nowhere near that.

I mean, yes, we’re dealing with the information revolution. And, yes, there is the possibility that’s going to replace a lot of jobs, increase productivity. The point that a lot of people just don’t have anything to do. But that’s all theoretical. Our experience in the last five years is, if anything, it’s going to be the opposite. You see the sort of things that IT revolutions do is they make it, it’s in the name information, information tech. It manipulates information at a faster rate, but that is not where the uneducated people in our society are. Most of the uneducated society, people in our society are in lower class, blue collar jobs. That is not something that AI can help with at all. That’s something that the advances we’ve seen in productivity are almost irrelevant. AI instead is taking away those low to mid skilled white collar jobs, which is not normally what we think of when we think about the sort of jobs that can be destroyed.

So we’ve actually, in the last three years since the greatest increase in take home pay for low skilled blue collar workers in over a century. And that has actually helped in the case of United States, narrow economic inequality to a degree that we have not seen since before the World Wars. So if anything, the theory is proving itself wrong rather than right.

However, if you’re, say, a copy editor or a secretary, well, you might have some really big problems because I already has been able to deal with those jobs in a more efficient manner. You just don’t need as many people. In fact, the blockchain, which is one of the things that undergirds crypto. Crypto is something that could be very transformative in things like health care. If you think about any doctor’s office you’ve been in, there’s that huge forest of staff in the back who are basically on the phone with the insurance agents every day, all day. Well, the whole idea of blockchain is that anyone who controls half of the pieces can grant others access. Well, in your health care records, that would be you. And if everything from that can be digitized, then that entire flood of low skilled white collar workers in the back of every doctor’s office and hospital simply goes away. So it’s probably not going to hit where we think it is. And it’s probably not a red versus blue thing, and it’s probably not a coastal versus interior thing. It’s a mid-levels of education versus the edges.

If you’re highly educated or low educated, you look fine from this. Second, there’s the issue of time. Now, obviously, these technologies continue moving, but there’s two reasons to expect that we’re going to have a lot more time to make this adaptation that I think a lot of people give it credit for. Number one is, as the baby boomers are retiring, which is happening right now, they are liquidating all of their investments and going into really boring stuff like T-bills and cash.

That’s not what funds I.T. startups. That’s not even what funds the big IT companies in Silicon Valley. For that, you need venture capital. You need a high velocity of money. Retirees are no good for that. And the whole world is aging very rapidly. And baby boomers are not a phenomenon limited to the United States. So we’re going to see the amount of capital just kind of seize up in the entire space.

At the same time, most of the world is running out of the 20 and the 30 somethings that are necessary to do the research and develop these technologies in the first place. So overall, we should expect the pace of technological evolution in the world to slow quite a bit in the two decades to come, compared to the two decades we’ve just completed.

Second, we’re not close to a general, a breakthrough. Let me explain what I mean by that. Artificial intelligence kind of falls into two general buckets. General A.I. is, you know, Skynet. The idea that the machine can actually look at a situation thing, come up with a potential solution and act on it, or nowhere close to that. I don’t know anyone, even Elon Musk, who thinks we’re going to be there before 2050. And this is before you consider that the amount of capital and workers that are available to develop these sort of things is in the process of drying up. So probably we’re looking at 2060, 2070 or beyond. We’re just we’re not even close.

The other type of AI is applied A.I. or mission specific A.I. and it’s not so much artificial intelligence in the way that we kind of have it on our heads. But it is machine learning, but it’s really more like machine programing. So you put in dozens, hundreds of thousands of if then statements into a program for it to execute. And as long as the conditions that are presented to the machine fall within the rubric of what you programed, you’re okay. But if you see something even a little out of context, the whole thing tends to fall apart.

So an example, let’s say you’re developing an A.I. driving program and you tell it what a stop sign looks like. What if the stop sign has a bumper sticker on it or graffiti, or if it’s on the side of a building as part of an ad. Applied to a I can’t recognize those other conditions. And if you kind of widen your parameters to make it a rounding error, then it’s going to make a very real mistakes in the very real world that any four year old could hit.

So if you need A.I. to do calculus, yes, they’re light years ahead of what we as humans can do right now. If you need it to make a decision based on a judgment call, they are still completely and utterly incompetent.

Alright. But let’s assume that some of this happens anyway. And so we’re going to have to deal with an A.I. system that is making decisions. What does that mean for the job industry?

So historically speaking, this is not the first time we’ve dealt with this issue. In fact, for those of you who remember your 1800s of political economic theory, good old Karl Marx, his whole idea was that the future of the proletariat was to take over from the capitalists, that once the industrial plant was built, then you could get rid of the bourgeoisie and the proletariat could live and do very, very well for itself because the machines and the industrial plant would be there to provide for everyone.

Well, folks, he was wrong then. He’s still wrong. Now, universal basic income is the idea that we live in such a world of plenty that we don’t need to work. But as we have seen in the last three or five years, if anything, the opposite is true. The productivity has stalled in part because of tech, but moreover, because we’ve discovered that as populations evolve in industrialization, we live longer, we have fewer kids. And that means after we urbanize five, six, seven, eight decades after, we’re actually running out of young people to do a lot of the lower skilled work. So if anything, Marx was completely wrong because the part of the population that he thought that would benefit the most from industrialization is at the current moment, actually not doing all that well.

The middle class, it’s the lower classes that are cleaning up right now. There are very, very few places in the United States at this point where if you were earning $15 an hour before COVID, you’re still in that bracket today. You’ve been able to leverage the fact that there’s a sharp labor shortage to move up, and that means you have a vested interest in the system.

And it means if you decide not work, there is no one who is willing to pay you to not work because there are jobs, jobs, jobs everywhere. So in conclusion, is  AI real yes, but we’ve been thinking about it completely wrong. And most of the assessments that I have seen from almost everywhere are drawing the wrong conclusions when it comes to sociological outcomes.

It’s going to be important. It’s going to change who we are. It’s going to change how we live and how we work. But the word here is change. It’s not a revolution. All right. That’s it for me. I want to go get a snow shovel. Take care.