Transcription
My somewhat controversial personal opinion, and you're welcome to agree or disagree, is that the AI equivalent of the Manhattan Project has pretty much started.
Now, when you look at history, you look at the Cold War that uh kind of started almost immediately after World War II ended between America and the Soviet Union. Right now, the competition is between America and China. And there's a lot of parallels. There are also some differences, but you're looking at things like military buildups and flexing on new technology and new science and that sort of thing.
During the first Cold War, Cold War 1.0 as I'll refer to it, uh there was things like, you know, jet fighter technology, radar technology, and then eventually the space race between America and China. Right now, the competition is predominantly over China's uh financial ambitions such as the Belt and Road Initiative as well as their military buildup as they're working towards uh creating a Blue Sea fleet. Uh but then of course the the the keystone technologies are AI and nuclear fusion. Uh China has set records for the longest plasma um although it's not net positive. So net positive energy gain is one is one thing that America still leads on which is arguably the more important component. Uh likewise artificial intelligence is something that America leads on.
Now there's a little asterisk to that because when you look at what China does, China is only a few months behind American frontier labs in terms of their open source quality and often they do it much more efficiently. So, what I wanted to do is just kind of ramble and talk a little bit about kind of what I'm seeing and the factors that I'm paying attention to that make me think, okay, the Manhattan project uh for AI has started and it's definitely Cold War 2.0.
So, one of the primary things is you just look at scale, you look at volume. America has 11 times the data centers that China does and we're building more hyperscalers. We're building things like Stargate. Meta is bu is building Hyperion. Um and then I can't remember the name of the Tesla uh or not Tesla the XAI data centers. Uh but point being is that not only we have literally 11 times the square footage of data centers, we also have the lion share of hyperscalers and the lion share of new high-end data centers going in for artificial intelligence.
Now that is a non-trivial amount of investment. You might have seen the financial news where AI investment now outpaces consumer spending as contributing to GDP growth, which is crazy. And so this is like this is the Apollo mission level of investment. We're not quite there yet cuz Apollo took up like alone took up more than 5% of GDP. I don't remember the numbers off the top of my head, but it was crazy. Uh and so we're actually heading in that direction with artificial intelligence. Uh there was a tweet yesterday that that shared that and I said I for one think this is a great thing. And when you look at the historical trends whenever a nation invests this much into a particular technology or infrastructure or initiative uh results follow.
Uh one of the reasons that we enjoyed the you know the baby boom after 1945 really kind of started in 1950 was because of the technological dividend that World War II left us with. Not only did we have the Brett and Woods system where America found itself in more power, but we also had a whole bunch of new technology. And so this actually is kind of in my opinion what led to the boomer belief that all you need to do is get an education and there are great jobs. And that is because at the time there was such a there was such a scarcity of of educated people to fill these new service jobs that it was a complete green field blue ocean opportunity. and there's a lot of lowhanging fruit. Now, of course, by the 80s, all the lowhanging fruit had been picked um and economic compaction set in.
However, what we're going to see is that with artificial intelligence and then all of the downstream technologies that it acquires, which I'll talk about in a second, such as solar, nuclear, and fusion, uh we're going to probably see a similar burst. Um, and it can be it could happen in as little as 5 years because when you invest in that kind of moonshot mission, you get the technological dividends relatively quickly. And it's also not like you have to get to the finish line. You're getting technological dividends every step along the way.
So what are these technological dividends from an economics and scientific and engineering perspective? Um, you can think of it as network effects. So a network effect is when a technology gets to a certain point where suddenly actually it's now better than the other options. So then it you go from like zero to one where the new option is suddenly the best option everywhere. People are talking about how that is what's happening with battery technology where you know originally batteries were too expensive and too heavy and too inefficient but now they're useful for for cars and before too long they're going to be useful for trucks and even airplanes. So you look at the uh you look at the battery technology going into drones when my cousin and I were building you know the predecessor of drones in the in the late '90s batteries were not really strong enough to fly uh aircraft you know vetos um for any length of time and and was mostly batteries but now battery density is to the point that you can have drones and so that's an example of a network effect.
Likewise we're going to see what's called spillover effects. So spillover effect is what people talk about when they say like oh all all of the inventions from NASA have many many spillover effects like microwaves and everything else red light therapy a lot of these things came directly from the work that NASA has done and so uh when you have this kind of compounding set of compounding returns the effects on the economy and society uh tend to amplify over time they they accumulate. So you have this virtuous cycle of more technology, more progress, new paradigm shifts and those sorts of things.
So when you look at the fact that we have geopolitical competition and a massive level of natural investment, those two things kind of say, okay, this really is looking more and more like Cold War 2.0. Um, one difference though is that it's not being centrally coordinated by the government. Yes, the government is playing its part, you know, with the AI acts and, you know, America's AI policies and uh the Genius Act and those sorts of things, but those are enablement. And that's one of the biggest differences between the buildup for World War II and the buildup for the Cold War, which is that those were financed primarily through uh government spending. You know, there was there was war bonds, there was treasury bills, there's debt and that sort of thing. Whereas now it's being financed primarily through private investment, which is really interesting.
And you could say, well, if it's being financed by private investment, then this is just a your run-of-the-mill industrial revolution. It's not actually a Cold War, except that it is a geostrategic uh policy for two nations to, you know, dominate artificial intelligence. Which leads me back to talking a little bit about China. So, one of the things that has interested me about China is why are they doing everything open source? Is it just to prove that they can? Is it to is it to shame America and say, "Look, you released reasoning models a few months ago and now we've done one that's open source and 100 times as efficient, even if it's only 90% as good." Uh, and this is what that's kind of what I was referring to by saying China's only a few months behind America is every time our frontier labs release something, China duplicates it within a few months. Uh, and then it's completely open source.
So the logic might not seem obvious, but the Chinese are nothing if not pragmatic. So the Chinese know that they don't have the data capacity that we do, but they can duplicate things and they also see the writing on the wall that open source is always just going to be, you know, at most 6 months behind closed source. So they're like, why spend the extra hundreds of billions of dollars when we can have something that's good enough and only 6 months behind?
And then the other thing, there's two other factors that that uh explain why China is just doing everything open source. Number one is that the more open source they release, the more it mobilizes their own developers, their own universities, and their own companies. So they're they're looking at it as more of a public good. They're saying, "Okay, every open source model, every open source data set, everything that we release opensource in Chinese is going to be better for China." Um so that that's uh big factor number one.
And big factor number two is uptake in the rest of the world. America with our pretty much all of our frontier work is closed source means that the models stay in American data centers. Whereas open source Chinese models, they get taken up globally. They go to the UAE, they go to Brazil, they go to Russia, they go everywhere else. And so this from China's perspective makes sense from a geostrategic point of view because ultimately that means that everyone is going to be more familiar with the Chinese standard. So it's a soft power play. Uh which is really interesting strategy and that's another thing that you could say is a little bit different between uh before during the World War II and the Cold War is because uh everything was locked behind closed doors. You know, the the Manhattan project was super ultra top secret. Today, we're building it in public, right? That's kind of the Silicon Valley ethos is build in public, but we're building geostrategic assets in public as well, which includes data centers, chip fabs, artificial intelligence, and that sort of thing.
So, you can also make the argument that it's a soft power kind of thing, which it's about flexing, right? You know, it's a national flex every time China says, "Hey, look, we duplicated the Americans working. It just took a few weeks or a few months." and so on and so forth.
So, that's about all I got. Um, I'm not going to ramble at you too much longer. Just wanted to share my thoughts about like, okay, when you look at the bigger picture here, it really does look more like a Manhattan project. Uh, just Manhattan Project 2.0, Cold War 2.0, the rules have changed a little bit, but ultimately there's a lot of the same motivations, which is to prove that our system is better and to achieve those geostrategic and geopolitical uh advantages. So, thanks for watching. Have a good one.