Samanth Subramanian on the Fragile and Resilient Technologies that Bind Us

SHRUTI RAJAGOPALAN: Welcome to Ideas of India, where we examine the academic ideas that can propel India forward. My name is Shruti Rajagopalan, and I am a senior research fellow at the Mercatus Center at George Mason University.

Today my guest is Samanth Subramanian, who is a journalist and writer and the author of the recent books The Web Beneath the Waves and  A Dominant Character: The Radical Science and Restless Politics of J. B. S. Haldane.

We talked about under water sea cables and choke points in critical global infrastructure, the intersection of technology and geopolitics, large language models, JBS Haldane and the relationship between science and politics, and much more.

For a full transcript of this conversation, including helpful links of all the references mentioned, click the link in the show notes or visit mercatus.org/podcasts.

Hi, Samanth. Welcome to the podcast. It’s lovely to have you here.

SAMANTH SUBRAMANIAN: Thank you so much, Shruti. Pleasure to see you as always.

RAJAGOPALAN: I know. It’s been years. I think the last time we sat in person and chatted was pre–COVID in New York City.

SUBRAMANIAN: I think that’s right. I was living there 2018, 2019 on a fellowship, and we hung out then. Then, of course, I moved out. I moved to England, which is where I still live, and we haven’t seen each other since.

RAJAGOPALAN: Yes, and I moved out. I live in Washington, DC, and it just feels right to do this conversation online when you are in the UK and I’m on the East Coast, because what I really want to discuss with you is your most recent book, which is The Web Beneath the Waves. I’ve got a copy right here. This is about the underwater sea cables that actually make most of our internet communication and data transfers possible, so I feel like we should probably be having this conversation online. It just seems fitting.

SUBRAMANIAN: Yes, it’s very meta. It may be meta in more ways than one. This conversation may be flowing back and forth on a cable owned by Meta as well, so it’s meta in multiple ways.

RAJAGOPALAN: I was reading somewhere about the W—that Meta is going to lay the cables through a W. It reminded me of those Gold Rush movies [It’s a mad mad mad mad world]. The gold is under the W. If you want to go into crazy movie zone, we can also do that.

SUBRAMANIAN: Yeah, absolutely.

Vulnerability and Resilience: Grappling with the Fragility of Undersea Connections

RAJAGOPALAN: Well, let me start with the book. The book’s main argument is that the way we normally think about the internet is, it is somewhere in the air or the cloud. It’s not something that is material other than literally the phone we hold in our hands or the iPad where we’re watching stuff. Largely, it feels digital, not physical, to us, but there is an entire physical infrastructure that obviously holds this system together.

In this particular book, you point out that this physical infrastructure that we take so much for granted and is largely invisible can be quite vulnerable. You do this in a few different ways. You talk about places like Tonga, where there was a volcanic eruption, places like Taiwan. Of course, these two are islands, but this could also happen in other places that are not islands. Then you walk into details of geopolitics, corporate power, and so on.

Normally, I would just launch into all the questions I have coming out of the book, but I have a feeling no one knows what this underwater network or undersea network is. If you can give the listeners a primer on exactly what you’re talking about, then we can dive into the rest of the questions.

Underwater Sea Cables

SUBRAMANIAN: Wow, OK. A primer would involve me saying that 95 percent of all data traffic that flows internationally goes through underwater internet cables. This is data traffic of every conceivable kind. It includes most obviously emails and Zoom calls, cloud software presentations and day trades and everything—all the kinds of data that you can think of. 

It additionally also carries data you probably didn’t think travels through these cables. Right now, for example, you and I, Shruti, are from a generation where we thought of the landline as distinct from the cell phone or from the internet, but there is no such landline cable; that’s completely different now. All the calls are digitized and sent back through these fiber-optic cables. There are between 500 and 550 of these cables around the world. Together, add up their lengths, and it’s close to a million miles of cable that have been laid.

This is only fiber-optic cable that I’m talking about. This is only the kind of technology that has been around for about 50 years in commercial use. It doesn’t include miles and miles and miles of cable that have been laid since the dawn of the telegraph age in the 1850s. When you think about it, the materiality of the internet that we talked about earlier, that you brought up, is largely concentrated in these cables. Of course, there are also data centers, and there are landing stations, and there are all these other things.

I think those are, to some extent, above land, and therefore visible. What makes these cables doubly invisible is that they’re tucked away out of sight under the ocean, but they also disappear from the very language that we use to describe the internet. It’s astonishing how big a role language and lexicon play in forming our conception of something. When we use, on a daily basis, terms like Wi-Fi, wireless, mobile, cell phones—

RAJAGOPALAN: Cloud.

SUBRAMANIAN: —cloud, we’ve abstracted away the materiality of it. Push comes to shove, if you really ask the man in the street what was carrying his data signals around, he would probably say the satellites that are orbiting the Earth. The truth is, satellites carry a tiny, tiny fraction of that data. They would never be able to cope with the voracious volume of data that we consume on a daily basis.

RAJAGOPALAN: You talked about materiality of the wired network, which is also something Nicole—I think I’m going to butcher her last name—Starosielski. Is that about right?

SUBRAMANIAN: It’s close. Starosielski, yes.

RAJAGOPALAN: Starosielski, OK. She talks about this, but the layer that I found added in your work is you get straight to the point of vulnerability and resilience. The most direct example of that is the one that you begin with, which is the island of Tonga. The closest large landmass is New Zealand from there. It’s one of the Pacific islands. They have this enormous volcanic eruption, which damages the two cables that actually connect the islands to each other and also to the rest of the world, and then everything else that follows.

When I was reading that part, it was very jarring for me because you and I come from that generation where the dial-up wouldn’t connect. We used to think about wireless systems and Wi-Fi systems and these digital systems as something you can’t really rely on. What you can rely on is physical systems, right? There is a landline, there is a corded phone. I remember when the cordless phone first came, the battery would just die.

Past a point, you started relying more on the rotary dial-up phones. Now, it’s flipped. You’re talking about how it is the material stuff which is underwater, and you describe in great detail how those optic wires are actually quite thin, given the task that they’ve been mandated for. Everything else, which is invisible, seems to be a resilient network, but the material is what is eventually going to fail us. Is that a good way to think about it, or am I just anachronistic and come from a generation where materials didn’t fail?

SUBRAMANIAN: That’s really interesting. I haven’t thought about it in this way. The truth is that there are roughly 100 cable cuts—cable breaks—that happen around the world every year. The Tonga case is an extreme version, and I picked it because I really wanted to see what happens to a society, to an economy, to a country when it is completely subtracted from the internet for a long period of time. How do you deal with that? What are the unexpected ways in which a country or a society will struggle?

The truth is, as I said, there are 100 cuts that happen in cables every year. For the most part, we don’t feel it because there’s been quite a lot of redundancy that’s been built into these cable systems. You live on the East Coast of the US. I live in London. These are two of the most heavily wired places on the planet. If three cables were to go down—transatlantic cables—we would barely feel it because that burden would be taken up by other systems.

Where it is felt, I think, is on the fringes or the periphery of, in Tonga’s case, the map itself as we usually conceive of it. Tonga is to the bottom right, as you said, near New Zealand, but then also the fringes of the world economy. If you’re not as important to the world economy, you’re not cabled as densely, and if you’re not cabled as densely, then the loss of one cable can really make a difference.

I think materiality serves us quite well for the most part. Of course, what is important to remember is that materials can fail, and materials can be targeted, and materials can be superseded. We’re talking about materials that are out of sight, dozens of miles below the ocean. Those are mysterious realms in and of themselves. It’s very hard to predict on a day-to-day or year-to-year basis the challenges that will come up.

Yet the industry does a surprisingly good job of keeping track of that and making sure that there isn’t some worldwide outage of internet cables. We can get to how they do that, but I think that’s interesting in and of itself.

RAJAGOPALAN: Actually, that would be super helpful to get into how they do that. There are people who manufacture these cables. There are these cabling companies. The ships that lay the cables do it in not exactly the same way, but it’s reminiscent of how the telegraph cables were laid. Then, of course, the people who repair it, that’s an entire infrastructure. Now, the new Big Tech that is getting involved in both financing and laying the cables. Maybe you can talk about this because you’ll know a lot more, and then I’ll have a bunch of follow-up questions.

How America’s Big Tech Companies Dominate Bandwidth

SUBRAMANIAN: Yes. Maybe it would help to describe how this industry has evolved over the last 30 years. In the early ’90s, any cables that were being laid were laid with the collaboration of state-owned telecom companies. All around the world, there would be a state-owned company, and it would say, “Well, look, we need a cable to go from the UK to Singapore, stopping off in Portugal and South Africa and somewhere in the Middle East and India and then Singapore.” All these state-owned companies would come together, they would kick in investment, they would divide bandwidth as was necessary, and that was the model.

Around 1999, the late ’90s, let’s say, it slowly started to become a norm that a private investor would come and would raise investment either through themselves or through a bunch of other investors. They would arrange for this cable to be laid and then go to these state-owned companies and say, “Look, the only reason I need you now really is to buy bandwidth from me and to get the requisite government permits that are needed to land these cables on your shores.”

The investors started doing it this way and this way. These consortia started to come together, consortia of investors, and they would lay these cables, and they would own them. Of course, they would subcontract out everything else. You buy the cable from a cable manufacturer, you hire a cable-laying company to lay these cables, and then that company would be responsible for mapping the sea floor, conducting safety evaluations, figuring out the best and most efficient route, and then actually laying the cable. Of course, if the cable were to break, that same company would go out there and raise it to the surface, repair it on the ship, which still happens, and drop it back down.

That was the model for a long time until very recently, in fact. I think that started to change between seven and 10 years ago, when some of these big tech companies—I’m talking primarily about Google, Meta, Amazon, and Microsoft, these companies that live and thrive on data as their lifeblood—they were already building data centers. Some of them had written their own operating systems for their data centers. Data is everything to them.

I think they started figuring out that the expense of a cable, while considerable to you and me, is not as considerable to them. These are very cash-rich businesses. It costs about $500 million right now, give or take, to lay a cable across the Atlantic. To a company like Google, which depends on cables and data continuity, as they would call it, it makes perfect sense to say, “I will lay the cable. I will own most of it. I will sell bandwidth to other people.”

They started getting into this game to the point that when I wrote this book, two out of every three new cables were being laid by one of these four tech companies. I think that is the state of play right now still, by and large. There’s a huge new acceleration of interest in cables because of the AI revolution, because you need to build more data centers. Data has to be funneled through these data centers. It’s just—

RAJAGOPALAN: The latency has to be cut down even further.

SUBRAMANIAN: It does, but also, you can just think of it as a use case for data—I hate these terms—a use case for data that didn’t exist four years ago, but now suddenly exists, and it’s just layered onto everything else that we’re doing. It’s not like we’ve stopped using data elsewhere. We just continue to stream Netflix, but also send ChatGPT questions about what I'm watching on Netflix.

You need more cabling and more data infrastructure to cope with all of that. I think you will only see more concentration among these four tech giants of ownership of the bandwidth at the bottom of the sea.

Concentration of Infrastructure

RAJAGOPALAN: I was thinking about two ways the world could go because of this new corporate concentration. As an economist, I think that it just makes sense. The people who profit the most from a particular infrastructure should be the ones who also contribute to that infrastructure. Otherwise, to a very large extent, a lot of them free-ride on infrastructure that was built in a decentralized way by lots of people over a long period of time.

To me, the fact that they actually finance or pay for this doesn’t bother me on its own. It’s more a question of what happens when it comes to resilience. On the other hand, I also feel like this might make the system more resilient, because all these companies are American companies. Now, I have a feeling most countries will start thinking about American companies controlling so much of the underwater sea cable data system. That they need to start bringing in some other backup plan or resilience in case they either deny these companies entry into their own territories or markets, or they’re in a trade war with the United States, which at this point, frankly, I think most countries are.

I don’t know how it’ll be by the time the episode releases. It keeps changing. Which way do you see this going? I can see this going either way: This could end up with more concentration and the system becoming overall more concentrated and less resilient, or it could also spur a lot of people into thinking about, “Hey, we really need to figure this out now. We can’t let these four companies control it.”

SUBRAMANIAN: Yes, I think we’re at an interesting inflection point right now. I think a lot of people are clearly realizing that this kind of concentration is not a good idea. In the industry, if you go and talk about it to them, everybody is worried about it. They’re worried not only by the concentration on the American side, but they’re worried about America’s insistence that Chinese companies should have no role to play in a lot of this infrastructure.

That’s just not feasible. If you want a global internet that connects China as well as many other countries, which we all do, you can’t just say Chinese companies can’t have any role in financing, laying, or repairing these cables.

RAJAGOPALAN: You’re hinting to the Huawei situation where any company that was buying their equipment couldn’t lay cables for the United States, stuff like that, right?

SUBRAMANIAN: Very specifically that, because there was a spin-off company from Huawei called HMN, Huawei Marine Networks (HMN). They were, for a long time, building up their own capacity to lay cables. At some point, if, for example, a European telecom company decided that it wanted a short cable lay, they could go to SubCom, which is an American company, or they could go to ASM, which is a French company. But they could also go to HMN. HMN might give them a better quote, and so that would be competitive. 

Now, what happens is if there are cable systems in which any American company is involved at all, the US government prohibits the Chinese involvement in that same cable system. It doesn’t even have to be a cable system that touches America. There are at least two cable systems I know that have been stopped in their tracks because the US government laid down this law, and things had already been committed.

In one case, the cables had been laid, but you just can’t switch them on because there was Chinese involvement in it, and the US government decided sometime in 2018 or 2019 that that would not work. What worries people is the concentration. What worries people is what they see as some kind of bifurcation of the internet. These are terms that people use in the industry, so this is not my gloom-and-doom-ism, but it’s their terms.

It’s interesting that you say that maybe larger geopolitical developments might make governments think twice about allowing American cable systems onto their soil, about pushing back against this American concentration. One would think that that would be the natural state of affairs. I haven’t seen it play out yet, but these are slow-moving developments. Cables are expensive to build. There are not many companies that lay these cables.

There are only four or five big ones, and one of the biggest is SubCom, which is an American company that not only contracts to Google but also contracts to the American DOD. Again, SubCom is able to exert a certain amount of American influence on the laying of cables elsewhere. I haven’t seen this happen yet, but again, as you say, this would be a natural reaction, for a lot of countries to decide that maybe this is the way that they will find their own resilience that is independent of American networks.

RAJAGOPALAN: The reason it occurred to me was, recently I was talking to Pranay Kotasthane about rare earths and critical minerals. For the most part, no one paid attention to rare earths and critical minerals. They’re a tiny, tiny part of the electronics or the batteries that we build out. Then suddenly, when China says it’s going to put in export controls is when everyone wakes up to the fact that it’s this tiny thing, but we all need it desperately.

Now, every country has a mandate for stockpiling it, or they’ve started forming stockpile consortia. India has its own rare earths and critical minerals mission, and things like that. I’m wondering if we just haven’t had this crazy world event yet, which will spur us in that direction. Sometimes, it takes that one thing. Tonga’s volcanic eruption, as dramatic and terrifying as the consequences were, I don’t think anyone knows about it or thinks about it. We just think of it as, “Oh, this is a weird case because it’s an island, and it’s cut off from everything else, and they never built redundancy because of exactly where they are.” It’s hard to lay two, three wires in that area, as your book points out.

That’s why I was thinking about people beginning to think about these backups. 

The 2Africa Cable and Countries Without Leverage 

SUBRAMANIAN: Rare earths are definitely comparable, but there’s one key difference, which is rare earths are a very tangible commodity. You know if you’re running short of them, and you know what will suffer as a result of that shortage. The main thing about data and the internet, I think, is that it’s so abstract that it’s impossible to even conceive of the eventuality where this might get you into trouble. The one place where I did hear discussions of this is when I went to West Africa. I went to Cote d’Ivoire for the book, where I wanted to go see this big cable called 2Africa land in Abidjan.

RAJAGOPALAN: This is the longest cable in the world, right?

SUBRAMANIAN: It’s currently the longest cable in the world, and it’s financed and owned mostly by Meta. They will not say to what tune and what extent of it, but it’s well known that that’s what it is. Now, it's a very interesting thing that happens in Europe is that when Europe is dealing with some of these big American tech companies, they’re in a position to make certain demands. They’re in a position to say, for example, “Look, if you want the data of our citizens, which you are collecting, we demand that you keep the data on European soil.”

Because there is a certain amount of leverage that these governments are able to exert, American tech companies will come and say, “That’s fine. We will just build our own data centers, and the data will stay here.” A country like Cote d’Ivoire doesn’t have that kind of leverage. Before 2Africa got there, there were three international subsea cables landing in Cote d’Ivoire. They needed more. The entire continent of Africa is criminally underserved by the internet. To not be on the internet in as full a manner as possible is really some kind of exile from the world economy. 

A country like Cote d’Ivoire has to make a choice. They have to say, “Look, do we antagonize a company like Meta by insisting on similar standards, saying that, for example, Ivorian data has to remain on Ivorian soil? Or do we think to ourselves, ‘Our citizens just need the internet too much. We’re not in a position to make demands. We will accept whatever terms you give us, even if those terms are ruinous to us in some way. You are profiting off the data of these people. There is security data, healthcare data. All this stuff that might be stored overseas, and we have no control over it anymore.’” There are very difficult choices that lie in front of governments without much leverage. They are very acutely aware of this, but they are not, unfortunately, in a position to do much about it.

I draw a parallel in the book to how oil exploitation in West Africa occurred, in Nigeria in particular, where at some point these countries had to say, “Look, just come in and pull out the oil and keep most of the profits for yourself and kick back whatever profits you want in terms of bribes and so on.” Not too much is being used to develop the Nigerian state capacity itself and Nigerian society. They had to live with this arrangement for a long time before, at some point, various countries decided to push back against it.

RAJAGOPALAN: I don’t like the oil example so much. I’ll tell you why. Because data, unlike oil, is a non-rival good. Every time the Africans’ data is used by Meta, it’s not like they have less of their data. It’s not literally like someone stole my health records and now I don’t have them anymore. I know the broader point you’re making is about control over that data.

SUBRAMANIAN: Yes, and about who exploits it. You’re right to say that the Ivorian government could decide to sell the healthcare data in its database to a health insurer and make the money off itself, but these datasets circulate around the world. Meta could sell it for less. There are markets in which this data is more valuable, less valuable. Of course, above all of this comes the fact that Ivorians have their right to privacy, just like anyone else. For anyone to go out there and sell it at all is not necessarily the best thing in the world, but that kind of control is not available to a lot of these governments and to a lot of these states.

RAJAGOPALAN: I know the broader point you’re making, but there are so many African governments I would trust far less than I would trust an American corporate company. There’s also that. We’ve got to be careful. It’s not an apples-to-apples comparison to compare Meta to the highest standard of privacy in Western Europe and then compare them to some of the worst dictators in Africa, who are anyway selling out their own people much worse than just data.

I don’t know. It doesn’t feel like the same thing somehow. It may still compromise their privacy and sovereignty on a different margin on another day, but I would trust Meta more than whatever is currently governing Yemen and Syria.

SUBRAMANIAN: I’m not necessarily sure I would, but that’s not the point here. The point is that control should reside in the state where this data was generated. Obviously, the aspiration here should be to build privacy regimes and data security regimes in those countries as well. Obviously, there’s an interest among everyone to work towards that, but that will never start if the data has already left the country and is being monetized by Meta.

As I said, I don’t even necessarily think that trusting Meta above the Yemeni government or whoever’s ruling Yemen right now is the natural conclusion I would come to, simply because whoever’s ruling Yemen right now has no interest in data. It's not like they—

RAJAGOPALAN: They have no interest in the Yemeni people. Do you know what I mean?

SUBRAMANIAN: It is still a regime composed of Yemeni people. In that sense, I feel like if they want to abuse the data, that is also tragic, but at least it is their own people who are abusing the data. I feel like I would be more comfortable with that. The one parallel I would draw—it is not the same, but it feels somewhat the same—is the argument that the British Museum often makes, which is, “Look, if we send all these sculptures and so on back to museums in their own country, they’d be destroyed. The conditions are terrible. At least here they are being preserved and people can come and look at them and so on.”

I feel like your patrimony in this case is unfortunately also yours to destroy if you are not able to protect it. At least it is being done by the people who have generated the patrimony in the first place. Maybe there’s some poetic appropriateness I see in that. Rather than a Western corporation in one case, or a museum in the other case, deciding that it will, for the benefit of humanity and for your own benefit, take this data, take these sculptures, and decide what to do with it.

Geopolitical Vulnerabilities

RAJAGOPALAN: The second threat that you point out, other than just data control, is what is happening and the fears of Taiwan, which is the geopolitical problem. There are states that are under greater threat from certain neighbors or certain adversaries. Taiwan’s the very obvious case. It’s an island. It has an enormous adversary which is constantly going over the seas and potentially thinking of cutting some of these undersea cables, which would completely threaten the Taiwanese economy.

Taiwan is not Tonga. It is a very, very industrialized and capable state. It’s technologically very advanced. I’m sure they’ve built in a lot of resilience and redundancy when it comes to their underwater sea cables, but it’s also really hard for them to stop an adversary like China from just cutting them. Is this the new standard for warfare? It’s happened before. It’s happened during World War II when German telegraph cables were cut and things like that. Is this the other battleground that you see forming?

SUBRAMANIAN: It’s one of them. The preferred term apparently is gray zone warfare, which involves ships that aren’t really military ships or naval ships. They might be disguised as fishing boats or trawlers or what have you, accidentally cutting these cables while out on the high seas. Of course, it’s a deliberate move. It’s not just China that we’re talking about. The other big player in this is Russia. The Baltic states, Europe, and the Scandinavian states are concerned about this as well, the fact that Russian fishing boats will come out and cut data cables, power cables, undersea gas pipelines, that kind of thing.

In Taiwan, the fear is, as you said, much more clear and present. There are 15 undersea cables that land in Taiwan. It is an island, so it depends entirely for its connection to the world on these undersea cables. It is a very industrialized nation. It has these massive semiconductor manufacturing companies. It’s plugged into the world economy much more than Tonga. For that link to be severed, I think, means a lot more for a place like Taiwan.

What does it do? What does resilience look like in that particular case? One is to build up more microwave relay stations, depend more on satellites, try to route cables. Taiwan is in a very odd place in the sense that one part of it is exposed to the Chinese mainland, but the other part of it is actually away from the Chinese mainland but is in a very geologically active zone as well. It has threats on both sides.

When I was visiting Taiwan and talking to people in the ministry there, the one thing that they said that surprised me was—there’s been a convention, ever since the first telegraph cable was laid in the mid-19th century, there’s been a convention to say, “We will publish the coordinates of every cable.” This act of transparency is a practical one. It is to ensure that ships don’t accidentally cut these cables. They don’t drop an anchor. They don’t trawl where cables are. They avoid these zones as much as possible.

Now, of course, that has sometimes not paid off. You mentioned how the British cut German cables during the world wars. That was one example. It continues until this day. The [coordinates of the] undersea cables, the fiber-optic cables that have been laid over the last quarter century or so are all publicly available. You can go online and look at the map if you want to. I asked the Taiwanese ministry people, “Are you considering now that maybe it is time to revisit this policy of transparency?” They said, “Yes, we’re considering it.”

RAJAGOPALAN: That means they’re doing it.

SUBRAMANIAN: Well, maybe. It’s astonishing that the current tensions of geopolitics are reversing a policy that is a century and a half old. That really says a lot about the state of the world as it is today.

RAJAGOPALAN: Would you be surprised [if] some countries have already done this? I would not be surprised if China or the United States has laid cables, or the northwestern part of Europe, which is close to Russia, has laid cables and not made it transparent.

SUBRAMANIAN: It is true that the US and China, and probably other countries, have laid military cables whose coordinates are not known. Civilian cables, no.

RAJAGOPALAN: Isn’t the distinction between what is a military cable and civilian cable also becoming narrower and narrower?

SUBRAMANIAN: In what sense?

RAJAGOPALAN: In the modern warfare world, we’re talking about all these large language model companies being contracted by different departments of defense, right? It’s unclear to me if the DOD cables will be used, or if the Meta or Anthropic or OpenAI cables will be used in the future, or how exactly that proceeds. For data centers, there’s a much clearer contractual arrangement that the data has to remain in the control of the DOD. I still don’t know if the arrangements have been worked out for the undersea cables yet.

SUBRAMANIAN: It may be too early for that, to know exactly how that’s playing out. For example, let’s say Anthropic is contracting with the DOD, which it did for a while. You might not need any undersea cables at all. It might just be Anthropic servers sitting with the DOD’s. In that case, the cables don’t necessarily come into it. It is true that for a long time, the military used cables that were laid by SubCom. They continue to use cables that were laid by SubCom, whose coordinates are not published.

When you say the difference is getting narrower, I think the difference is definitely getting narrower in one sense. Maybe 30 years ago or 25 years ago, a rival country or a malicious actor could think that the most effective damage to be wreaked upon a country is to cut the cables that are used by its navy or to eavesdrop upon the cables that are used by its navy. I think now you could do more damage to an economy by cutting a bunch of cables that carry civilian traffic.

You could tank a stock exchange. The hospital system could go into chaos. Airports could not work. In that sense, I think from a strategic and security point of view, it is true that it suddenly matters much less which kind of cable you cut. Obviously, the civilian cables all have their coordinates that are published, and they are all out there.

RAJAGOPALAN: No, but I mean it even more directly when I say the distance is narrowing. Recently, some of the Iran drone warfare was trying to attack—and I think did hit—the Microsoft data center. That’s one example. If they know that these companies have contracted with certain governments, then their data centers and their cables might be the most obvious point—not to tank the stock exchange, but maybe to shut down military operations. Even if it’s seven minutes, seven minutes is a long time when we’re talking about drone military warfare and stuff like that.

SUBRAMANIAN: I don’t think we’re at a stage yet where somebody could knowingly cut a cable or a set of cables and think that they would hinder military operations.

RAJAGOPALAN: Fair enough.

SUBRAMANIAN: That’s what I mean.

Indian Resilience

RAJAGOPALAN: Where does India sit in all of this? India has been part of the broader undersea cable network since colonial times, I think 1870s, when the first undersea cable was specifically laid for India. A lot of this was the British trying to control the colonies. India was one of the largest and richest, so it’s unsurprising. India has actually actively participated in this network since the 1870s. How does it feature now? I imagine India has its own underwater sea cables. India also has terrestrial fiber optics and things like that.

Its entire digital public infrastructure relies on this material infrastructure. Does India have resilience and redundancies built in? Is it more vulnerable or less vulnerable than other places?

SUBRAMANIAN: I would say probably less. Look, the big countries, especially countries that have land and sea borders, are uniquely resilient in the sense that even if by some freak of nature the undersea cables went down, there’s still land-based cables that carry a lot of traffic. The bandwidth of those land-based cables is much more than you can lay at the bottom of the ocean. In that sense, India, like America, like China and Russia and all these other large countries, has a great degree of resilience built in.

There are a couple of things that are interesting. One is that, for example, I found it very curious that, since the 1870s, all the cables that come to India and leave from India basically do so from two cities. They all come to Bombay from the west and they leave from Chennai in the east. That hasn’t changed at all. I asked somebody at the Tatas, who once were very actively involved in laying cables, why that was the case. Why not diversify?

If, God forbid, a tsunami were to strike Bombay or Chennai—and it has in Chennai’s case—and it snaps a lot of these cables, why not build some redundancy in? The answer is—and this says a lot about the industry, also—there’s just a lot of inertia. You land some cables in Bombay, and you set up data centers there. Then it only makes sense to land more cables there so you can hook them up to the data centers and the landing stations and so on and so on.

Before you know it, Bombay and Chennai are only two exit and entry ports for a lot of these cables. This guy from the Tatas said, “I really wish people would think more about where else these cables can land because I think that’s important to build resilience.” 

From the point of view of laying cables, obviously, a lot of these big telecom companies in India—I’m thinking of Airtel in particular—they are bandwidth purchasers. They help cables land, so they take care of things like permitting and so on, when they land in Bombay or in Chennai.

The Tatas used to, as I said, do a lot of cable-laying. I think the proportion of that as part of the overall industry has come down somewhat since the early 2000s or the mid 2000s. Otherwise, I think countries like India, America, China, and Russia are the most securely cabled countries out there. I think the fact that you have a land border is a big deal.

RAJAGOPALAN: The really vulnerable places will be something like the Andaman and Nicobar Islands if it comes to securing the Straits of Malacca or something. It must be something very, very specific, I imagine. You could always have the Indian navy patrolling. I’m sure there are many, many ways to back that up.

SUBRAMANIAN: These are not heavily populated places, either the Andamans or Lakshadweep. There are microwave links and satellite networks that might do the job. It is islands that are uniquely vulnerable. Singapore. Taiwan. The UK, where I live now, is still an island. The British government, a couple of years ago, I think, commissioned two naval warships to constantly patrol the island to make sure that its undersea infrastructure is safe—not just cables, but power cables, gas pipelines, oil pipelines, all of that.

RAJAGOPALAN: It seems quite wise, actually, given the world we live in.

SUBRAMANIAN: Yes, I think so. I think in the UK—

RAJAGOPALAN: You have enormous security for ports, and so it would just make sense to extend it out further, I imagine.

SUBRAMANIAN: I think that’s right, yes.

RAJAGOPALAN: Because the goods trade and the digital services trade is becoming, if not equally important, at least converging in importance.

SUBRAMANIAN: Absolutely.

Choke Points in the Global AI Economy

RAJAGOPALAN: The other piece of yours that I read, which is about—I think there were similar themes—is the one you wrote about ASML, which is, there’s this entire global dependence on semiconductor chips, and semiconductor chips being available for cheap as soon as we need them. During COVID is the first time we really realized that it’s all concentrated in terms of production in a couple of places.

The lithography machines that are making chip production possible for so cheap are actually concentrated in one company, which is ASML. I read another book recently on ASML called Focus. When I was reading your piece, my original imagination was, “Oh, we need to have backups for ASML.” Once I started looking into it, I was like, “This is the most intricate sort of—it’s intricate almost at the level of war planning. You can’t move a tiny thing.”

If they have a tiny supply disruption for a particular kind of glass or a particular kind of lens, or one of their suppliers has a disruption of inputs, everything would come to a screeching halt. I’m not even getting into the human capital being a highly specific problem. It just seemed like, wow, it’s really hard to find ways to back up ASML. Is this a similar vulnerability—concentration in one company, or that one company being in one country in the world? What is a way to think about this parallelly, or is it just a completely different thing?

SUBRAMANIAN: It’s kind of similar. Obviously, it is an extreme case in the sense that just one company makes the machines that make the chips. That, I think, is unique. I should say, we don’t know if there are Chinese companies out there now making lithography machines. We think they are, but we’re not sure. Possibly not the extremely specific, very-small-wavelength lithography that ASML makes.

ASML becomes something of a choke point for the industry at large. I think, again, we must think about it as what happens to these choke points at times such as these, the times that we are living in. One is that it becomes extremely vulnerable to geopolitical pressures. To give you context, when I reported this piece a few years ago, I was just interested in the fact that I wanted to know, “What are these machines? I’m just curious, why are they so hard to make?”

I tried to find that out. In the course of that, the CEO would tell me a little bit about some of these pressures that he and the company faced. He would tell me about the fact that they were essentially caught in the middle of this big battle over chips between the US and China. The US was in a position where it could impose blockades of what kind of chip technology could be sold to China, and ASML had to abide.

ASML, from everything he said—now, he didn’t say this explicitly, so I should emphasize that. He did not say this, but from everything in his demeanor, I sensed that he would rather not abide. This is a business. This is a company whose business is selling these machines, and what they want to do is sell the machines to anybody. They don’t really care if China is making the kinds of chips that these machines are able to make.

It is particularly interesting that this company is Dutch. It’s not British. It’s not one of America’s old allies, like the British or Canada or whatever. It is Dutch, and the Dutch are not enthused about the fact that they are under the American thumb in this particular case. I got every sense that this guy was grudgingly telling me about this blockade that the Americans had imposed.

Now, the fact that they are able to impose this blockade automatically creates this bifurcation in the industry where the Chinese have to figure out a way in which they have to make these machines themselves. Maybe they aren’t able to get to that hyperfine wavelength model, which means they aren’t able to build the really cutting-edge chips that the Americans are supplying to the rest of the world, which means that their technology is necessarily lagging. This is what the Americans would like to believe. 

That is what it means, I think: There’s a juncture that we’re [at] in modern history where I think these technological choke points—we didn’t think as much about them 10 or 15 years ago, but I think that we have to think about them acutely and constantly now.

The Challenges of Decommissioning Nuclear Plants

RAJAGOPALAN: The other vulnerability, and this has been going on for a long time, is nuclear energy. As climate change becomes more and more of a problem, civilian nuclear energy is becoming more important. Germany has fired its plants back up, or is rather bringing them back. I think overall, it’s the better tradeoff, given just how big the problem is in terms of both the problems related to meeting the carbon targets, but also the air pollution.

This is very much a Global South problem. In some sense, it just seems like the better way to do this would be nuclear energy. You’ve also written about how fragile that system is when it’s not just about starting up a nuclear energy system, but also shutting one down. This is your reportage on Sellafield, which is, I guess, what? Northwestern tip of the UK?

SUBRAMANIAN: Yes, that’s about right.

RAJAGOPALAN: Yes. What does it take to decommission something like this? I know uranium and nuclear energy is a weird situation where shutting things down sometimes may be even more dangerous than starting them up. But it still feels like it’s a very similar choke point to what you’re talking about when you’re talking about ASML or when you’re talking about underwater sea cables. These are the kinds of choke points we worry about in the modern world.

SUBRAMANIAN: First of all, to start, I think the guy you should have on your show next is a professor called M. V. Ramana, who used to be at Princeton, is now in Vancouver at the University of British Columbia. That must be it. You can have an hour-long debate with him about, “Is it really wise for the Global South to go in? Is it really cheaper for them to build nuclear power plants rather than solar and wind? Does that really work out?” He has very specific views on this. I will not speak for him, so I’ll leave that question aside.

RAJAGOPALAN: Sure.

SUBRAMANIAN: The Sellafield piece that I wrote for The Guardian, which was a few years ago, came to me the way most pieces I write for The Guardian come to me, which is just that the editor will text me one word, and I will say yes or no depending on that. Sometimes—

RAJAGOPALAN: What was the word that was texted to you?

SUBRAMANIAN: He texted “Sellafield,” and I said, “OK, let me look into it.” That’s pretty much it. His question and the question I wanted to answer was, “OK, look, we all know about the debates about nuclear energy, but actually, what does it take to shut a power plant down?” 

Sellafield is Britain’s oldest plant. It had a terrible accident three-quarters of a century ago. Parts of it had been shut down then, but now they’re basically decommissioning the entire thing.

Decommissioning doesn’t just mean pulling the switch and turning off the lights and shutting the doors. It means making sure that spent fuel rods and the water in which the spent fuel rods were contained and the pool in which the water was contained and the land in which the pool was built and the cooling towers and all this other stuff—what do you do with this material? Because radioactivity is, from our point of view, practically forever.

The expense and the physical engineering challenges—and ultimately this challenge of entombing all of this far, far below the earth—was what I was really interested in. It was almost from a physics-geeky point of view that I went out there and was like, “Let me see what this is all about.” 

I went to Sellafield. I also went to a waste storage sit they are building in Finland. They took us down below the earth, and it was half a kilometer below the earth. They are building these huge storage tunnels where they will store nuclear waste and decommission nuclear material at a cost of €5.5 billion or something astronomical.

They have to maintain this site for years because nuclear waste might seep into groundwater. Nuclear gas might escape through fissures. They have to make sure that it’s airtight, watertight. That maintenance job is also an ongoing process. My contribution to this debate, without necessarily taking a side on it, is that decommissioning nuclear plants and making the material safe is time-consuming, hugely expensive, and arduous. I think when people are planning energy policies—renewable energy policies in particular—that should probably factor into it.

RAJAGOPALAN: I think that’s fair, and I think it’s the same for all other kinds of energies. When we think of solar and wind, I think we should also factor in what it costs to make a grid that can absorb power that is so variable, because that’s the main issue with renewable energies. When it comes to fossil fuels, there’s this huge externality in the air, both pollution and other kinds. I feel like every source of energy has some externality effect.

SUBRAMANIAN: I think that’s right.

RAJAGOPALAN: It’s just sound economics to actually account for all the externalities before we take a call on how it works.

SUBRAMANIAN: That’s totally right. I think the person to talk to, as I said, is Ramana, who has taken—

RAJAGOPALAN: I should definitely look up his stuff.

SUBRAMANIAN: He is great. He is also a big Carnatic music fan, Shruti. 

RAJAGOPALAN: Oh, lovely. Yes, we haven’t even talked about Carnatic music. I have got to have you back on. I heard you talking with Sharan about other musical things like A. R. Rahman and other things. We’ve not discussed our big passion yet, but we’ll get there. Maybe I’ll just have you back on one day just to talk about music.

SUBRAMANIAN: Another day, yes.

On the Frontiers and Ethics of AI Technology

RAJAGOPALAN: There are two other big things that I want to talk about. One is your book, which I hopefully get to. More broadly, I feel like you’re a good person, and I want to pick on your brain on how you see the current AI wars and sovereignty wars going. This is very much in your wheelhouse in terms of a lot of the other tech and science writing you’ve done.

These big American firms, they serve the whole world as of right now. They’re at the technological frontier, so it doesn’t make sense to completely shut them down. In a narrower sense, I feel like right now in the Trump-administration-versus-Anthropic debate, there are a lot of people who are like, “Hey, Anthropic should be allowed to do whatever they want to do.” On the other hand, you’ve written also a lot about what happens when there is this kind of concentration in the hands of a few corporations.

It’s not just about data. It’s also about frontier technology concentration. The way I’ve read how you’ve written about it is, it’s almost like the imperial times. It’s not exactly colonization, but it’s a very high degree of control, and people can hold you hostage. There's one—China as a country can hold you hostage on rare earths and critical minerals and a bunch of other things. The United States can hold ASML hostage. Now, how do you imagine things playing out when it comes to these very, very large language model companies? They are very CapEx-heavy already.

SUBRAMANIAN: OK, I’m going to throw my hands up on this one and say I don’t know. I actually don’t know enough. I’ve been thinking obsessively about this. I haven’t written about it yet, and I don’t know if I ever will. It feels like, for me personally, the pace of change has been too fast, and the scale of change has been too great. I think I, like a lot of other people I know, am struggling to deal with what the consequences of this are and how to think about this.

Obviously, everything you said is worrying. That the tech firms that control a lot of other things are also controlling the frontiers of this new technology is automatically worrying. As a writer whose books I know have been used in training data by at least two language models out there, I know for a fact, that worries me.

RAJAGOPALAN: Are you seriously telling me that if I typed a prompt and said, “Can you help me write like Samanth?” it will help me write like you? Because I will take that trade. You write beautifully.

[laughter]

SUBRAMANIAN: It definitely has data that—it has information that belongs in my books. It has been fed on the facts in my books. Maybe it has also been fed on [my] writing style or whatever, although I don’t know how these things are combined and assimilated and aggregated inside these black boxes. Obviously, from that point of view, also, I am not a disinterested observer. That also worries me.

I haven’t yet figured out whether I should be cosmically worried about AI as a whole. There are people within the AI industry who have—Hinton himself, for example, is worried about runaway AI and AI that we cannot control. I haven’t even gotten to the stage of thinking about that and assessing that for myself and figuring out how important it is. I think my short answer to your question is I just don’t know enough yet. 

RAJAGOPALAN: Let me ask the question a different way because one of your great skills is how to approach a topic when you know nothing about it. Like you said, your editor will send you one word, which is “ASML?” or “Sellafield?” and then you figure out a point of entry. For those of us who are trying to think about these questions or write about these questions, what would be your advice? Because you also teach writing and you teach journalism. What are the points of entry?

SUBRAMANIAN: I think a lot of it would have to do—the things I’ve been interested in about this come from a point of view of philosophy and ethics. The philosophical questions, in particular, are enormously interesting. To take the question of writing, it’s a small niche within this bigger machine. What is writing, and what is reading? These are very fundamental questions that we had answered for ourselves for a long time.

Now we suddenly think that maybe we have to re-answer them, or maybe we don’t, and maybe large language models are not writing. That question has to be resolved, and we have to think about it. The relationship between a piece of writing and its reader also has to be reassessed. There are many, many blind tests where people are given AI pieces of writing generated by AI, and they’re moved, astonished, or charmed in the way they would be if a human writer had written it. They don’t know that an AI has produced it.

If they are charmed or moved by it, what does that say about a writer’s relationship to a reader, or writing’s relationship to its reader? In critical lit crit [literary criticism], as it were, there is a school of thought that depends on what is called reader-response theory. Reader-response theory basically says, “If I am a writer, I can put all the subliminal messages and themes and hopes and emotions into a piece of text that I want, but the only valid interpretations and responses are the ones that it generates in a reader. That is the only thing that matters.”

RAJAGOPALAN: Wow.

SUBRAMANIAN: The reader is creating a piece of text as they are reading it, and that is the piece of text that matters. It is not what you put in there. If you were to go by reader-response theory, then it should not matter who puts a piece of writing out, whether it’s a large language model or a human writer. Now, that is a very, very alarming thing to think about for somebody like me who has been writing for the better part of a quarter century. I don’t know how to feel about that yet. I have no idea.

RAJAGOPALAN: By the way, I think the critical theory guys need to get their Marxism right here because, on the one hand, they also have labor theory of value and what it means to be a writer. I think the crit guys need to get in a room and sort this out, perhaps.

SUBRAMANIAN: Yes. I think literary criticism, those guys are not necessarily Marxists. Look, when you outline reader-response theory to me or to somebody else—

RAJAGOPALAN: Yes, it is alarming.

SUBRAMANIAN: It sounds like a perfectly valid thing to think about, that ultimately, the only thing that matters is the relationship between text and reader. I don’t know how to feel about that. I have no idea. I have not come to a conclusion about this in any way at all.

Haldane’s Legacy and Science in the Public Square

RAJAGOPALAN: Hopefully you’ll come back, and you will talk to us more and more about this. I want to definitely talk about your book a little bit, the other book, which is A Dominant Character. This is an intellectual biography, I want to say, of J. B. S. Haldane. It’s not just a biography. It’s not just political. It really makes him come alive. First of all, I have a complicated relationship with Haldane because I really admire him and love him in parts, and really find some of it quite alarming.

I was thinking about, “What are some of the parallels between the two books?” I don’t think I’m misremembering. There’s a lovely bit where you write about Haldane being called by the British government during the war years because they lost a submarine. He’s a geneticist, and he needs to figure out how to help soldiers and other people in the war effort survive underwater.

That portion actually just came alive in my mind because it brought together all the different threads that are in the book, which is testing on oneself—which is something Haldane’s father did, and which also formed him as a scientist. It’s your obsession with underwater, underground stuff, which, simply, I don’t seem to be able to get away from it, but it’s also how science progresses. How do we think about other human beings and testing on them, or is everyone equal when it comes to testing? That seemed like a good entry point into the Haldane book when I was trying to connect all the threads.

SUBRAMANIAN: Haldane was not a trained scientist. He never got a degree in any science, but he became most well known as a population geneticist, which basically involves building mathematical models to tell us how genes spread through a population, through time, and through area. 

Some of his biggest contributions were in population genetics, but not the only ones. He wrote papers on the movement of bees. He did a lot of physiological experiments of the kind that involved respiration and the kind that you talked about when he was doing submarine research. He was just a polymath. Somebody called him the last man who knew all there was to be known, which is a reflection partly of the times that he lived in, because science had not gotten hyper-specialized as it is today, but also, partly, it reflects his polymathic abilities.

As for his politics, he started off on the left and at some point became a card-carrying communist for the better part of, I’d say, 15 years. Then quit the Communist Party, left England, moved to India, where he was at the Indian Statistical Institute in Calcutta. Then he moved to Bhubaneswar, and he died there. 

He lived a very full life. He was in the trenches during the First World War. He was doing all this research during the Second World War. He went to Spain during the civil war. MI5 thought he was a Russian spy and spied on him for the better part of 20 years. He was very actively involved in promoting and talking about science. He was a very well-known figure back in his day. Of course, because of his political beliefs, he was even more outspoken and radical at that time. 

More than anything else, I was interested in this era in which he was not the only one who was a political figure among the scientists that he was a part of in the UK and in the US. It was quite a common thing at the time for scientists to be outspoken about political issues. Admittedly, this was a time when, for example, to take genetics, the study of genetics was being misrepresented and used as an excuse for violence.

RAJAGOPALAN: And eugenics, quite explicitly racist, and sterilization campaigns being recommended, and things like that.

SUBRAMANIAN: Everywhere—in the US, in the UK, in Germany, in the USSR—there was no country of the Western world where this was not happening. At a time in which science was being put to this kind of use, I found it very interesting that scientists of that time would actually stand up and be counted on one side or the other. I was researching this book starting from 2016, 2017, at a time when it seemed to me that that was not happening enough today.

Obviously, there’s been a long-running perversion of climate change data by the right. There haven’t been enough climate change scientists, I think, standing up and saying, “Look, climate change is a fact, and these are misrepresentations of studies.” Then, around the time that Trump took office the first time, there was this big March for Science.

It slowly started to seem as if scientists were realizing that their role is not to sit in an ivory tower and do their science—[or] an ivory lab, as the case may be—and do their science, but they have to go out there and engage in the public sphere of ideas if their work is to matter.

RAJAGOPALAN: This is not just a “right” thing. This happens across the spectrum. Right now, the medical community is completely under threat from the left, for instance, when it comes to certain medical practices. The most alarming thing I’ve read about it recently is chest-feeding, where they pump hormones into someone who is trans and self-identifies as trans and allow them to feed an infant. We don’t know the harms that come from these pumped-in hormones, and what that’s going to do to the infant child. I agree with you. I don’t think of it as a left or a right thing. I think of it as a “political capture of any kind of scientific integrity” issue.

SUBRAMANIAN: I just meant that when I was researching the book, the things that were on my mind at the time were basically these two or three instances that I talked about. Of course, in the book, I talk extensively about how the left in the Soviet Union was also capturing and misrepresenting agricultural science and genetic science.

RAJAGOPALAN: Lysenko is an important person in this. They had a complicated relationship between them.

SUBRAMANIAN: Yes. I was very interested in this intersection of science and politics and what that looks like. The research that I did for the book involved, obviously, a lot of time in the Haldane archives, but it also just involved talking to a lot of scientists about their views on this. I happen to know a lot of people who are very political people who are also scientists. They guided me through and explained a little bit of how the nature of scientific work has changed over the last 60 years and why it has made it difficult for scientists to speak up today, necessarily, and to engage in this public sphere of ideas.

RAJAGOPALAN: Is it just the funding models which are the problem? You’re absolutely right. During that time, of course, you had people like Haldane. This was also just after the nuclear bomb came about. You had everyone from Einstein, the top physicists and so on in the United States, having very strong opinions about science and technology being used toward political ends or being used in a war effort, and so on.

There were many outspoken critics. You’re right, we don’t find too many now, at least no one as large in the popular imagination as Haldane or Einstein or someone like that. One is just science has become very concentrated and specialized and sits in these very specific research centers that have very long-term funding models, so that’s one possibility [for] why they don’t engage as much. What are some of the others?

SUBRAMANIAN: I think that’s right. Funding is one thing. It’s also who you’re getting your funding from and whether you want to annoy a funder. It is the hyper-specialization of science, which means you have to start thinking sometimes right from your undergraduate years about what you will pursue in your PhD if you’re really passionate about this stuff. It also involves the dwindling of an education [in] the humanities, in literature, and anything that I think might go some way toward creating some sort of empathy for people on a social and personal level.

I think if you’re shut up in a lab from day one of your undergraduate years, which, as scientists have told me, some people are, it becomes much harder to think about the ramifications of your science and your work out there when it goes out into the real world. It’s a number of these things. I think academia has made it more and more easy for you to retreat into itself. I think maybe it is something that is changing now, maybe not in the sciences so much.

I’m not sure. This is something I should check out now. The number of PhDs who come out of a university and struggle to find academic positions these days is obviously much more in the social sciences and the arts and the humanities, but there’s some effect on the sciences as well. I’d be curious to see whether the competition for the academic positions that exist, the cutthroatedness and monomaniacal focus that you need to get these positions, does that alienate you even more from the rest of the world, or does it make you rethink the role of academia and your own role in all of this?

It’s all very interesting. Of course, scientists don’t always go into academia. They sometimes go into industry, and that is its own set of closed doors. You definitely don’t want to speak out if you’re sitting in the bowels of a corporation that is doing terrible things to the world. At least university academics, we used to think, had some degree of freedom. I don’t know if that’s the case anymore.

Optimism and Pessimism About the Future of Science

RAJAGOPALAN: When I think about connecting this back to our earlier conversation, I’m very optimistic about AI and the large language models helping in this. I think, one, the barriers of entry for regular folks to enter scientific research have lowered. You see 15-year-old young people, 16-year-old young people, suddenly start decrypting old Indic or Indus Valley scripts or the Vesuvius scrolls, or they start looking at chemical combinations or AlphaFold, which is about protein folding.

Now, it suddenly seems like the barriers to entry for a lot of the sciences get lowered as long as people have access to some of these tools, as opposed to just getting entry into some of the top universities in the world. In that sense, it might become more democratic, and they’ll also be able to pursue other things. They don’t have to be monomaniacal the way you pointed out, and they can maybe start differently.

The other part is, I think large language models can also help existing scientists write better. I think we have very poor science writing, or at least very little of the good science writing. I’m a little bit more optimistic on this going into the future than I am when I look at the past.

SUBRAMANIAN: No, my natural state of being about any part of the world at this point in time is pessimism. It has nothing necessarily to do with how much I know about AI and what I think LLMs will do for us. I just feel like, based on recent record, let's say, it is safe to assume that if things have a chance of going badly, they will probably go badly.

RAJAGOPALAN: No, I have a lot of faith in young people and their ability to use technology well.

SUBRAMANIAN: OK, well.

RAJAGOPALAN: Yes, and I also have a lot of faith in decentralized systems, which is, again, a theme that keeps recurring in your book. I know that the cables might be concentrated, the data centers might be concentrated, but it also puts this extraordinary tool in every single person’s palm in the world very soon. There is something to be said for that. At least that’s what’s causing my optimism.

SUBRAMANIAN: OK. [laughs] I think you’re just a naturally optimistic person, Shruti, and I think I have been soured by experience in the world, so I don’t know whether—

RAJAGOPALAN: This is what happens when you’re a journalist.

SUBRAMANIAN: This is what happens when you’re a journalist. It’s true. You see the seamy side of society all along, and you think, “There’s no good in people. If somebody wants to use AI to destroy the world, they will try to use AI to destroy the world. That happens.”

RAJAGOPALAN: No, I’m not saying people won’t try to use AI to destroy the world. I’m saying also that’s not the only side I see, I guess, is where my general optimism comes from.

SUBRAMANIAN: No, that’s true. There are definitely upsides and downsides to everything. I completely agree there.

RAJAGOPALAN: Yes. I think you should do an optimistic piece. You should chase some of these— 

SUBRAMANIAN: Should I?

RAJAGOPALAN: —12-year-old kids who are using AlphaFold to figure out cures to new diseases. I think it might change you a little bit.

SUBRAMANIAN: I love the fact that you’re making it sound like every 12-year-old kid out there is using AlphaFold. [laughs]

RAJAGOPALAN: No, not every 12-year-old kid, but I also run Emergent Ventures, which is this moon shots grants program that we run at the Mercatus Center. I’ve given about 400 grants across India. At any given point, 10 percent of my grants go to people below 18, and they’re literally trying to map the known universe. They’re trying to understand which star exploded where. They’re trying to figure out protein folding. They’re trying to solve air pollution. So, I actually come across a lot of these people. Maybe we should introduce you to some of them, and you can do some real stories.

SUBRAMANIAN: It would be a welcome change from people I meet otherwise in that age bracket, who are basically thinking about what to put on TikTok and who can’t read more than a page at a time. The kind of feedback I get from people who teach undergraduates in universities about the fact that they can’t assign reading anymore because people can’t read.

RAJAGOPALAN: You’re sounding like those Mylapore uncles. No, we’ve got to get you out of this.

SUBRAMANIAN: I’m from the Adyar side, first of all. Secondly, look, all said and done, I’m a writer. For 35 years, I’ve been obsessed with reading. I think it is still an important way not just to disseminate information, but to teach people how to think critically.

RAJAGOPALAN: No, of course.

SUBRAMANIAN: If people can’t do that, then we have a problem. No amount of AlphaFolding will help if you can’t sit with a piece of text and try to come up with original responses to it.

RAJAGOPALAN: Fair enough. Before I let you go, first, you have to promise you’ll come back and talk about music and more about folding and other things.

SUBRAMANIAN: Anytime. Anytime. We will have far fewer disagreements on that front.

RAJAGOPALAN: Well, you never know. Before I let you go, what are you working on right now? Because you’re always working on three or four different interesting questions at any given moment. You don’t have to tell us the exact story that’s commissioned, but whatever you can tell us.

SUBRAMANIAN: Not much, actually. For the last year-ish, I’ve been associated with a new magazine called Equator, of which I’m currently the acting managing editor. We launched in late October [2025]. We publish long-form pieces. We publish essays, memoirs, some fiction, and some poetry. It’s all online as of now, but we have a print issue coming out later this spring. I am involved in helping with that.

I commission and edit pieces. I help fundraise. I do marketing. I write the newsletter. I help put a podcast together. There’s a lot of this stuff that I do. That is taking up most of my time. The rest of my time is taken up with the one-year-old we have at home, and whatever little is left, which is not much, I actually haven’t managed to do much writing over the last year. That’s the truth.

I’m only now slowly getting back into it. I’m working on a review essay for the New Republic about the commercial space industry in the US. I’m working on a photo essay for The New Yorker, a couple of investigative stories for Businessweek and the Economist. Things that are slowly starting to accrete, because now I finally have time to read and think and go out and do interviews. It’s been interesting. It’s the longest I’ve gone without writing big pieces. Obviously, I miss it, so I want to get back to it.

RAJAGOPALAN: Well, we look forward to reading all of it, and we’ll link to all of it in the transcript. Thank you so much for doing this.

SUBRAMANIAN: Thank you, Shruti. It’s been such a pleasure.