“The commons are under attack” from TikTok and subsea cables to data centers and elections

Danny Crichton:
Scott, you were our first guest in the Securities Podcast, and we're almost crossing our centennial, we're almost up to 100 episodes on the podcast. First, greetings, welcome back.

‍

Scott Bade:
Hi, it's good to be back.

‍

Danny Crichton:
I'm afraid to say, I don't think the world is less of a dumpster fire today than it was several years ago when we started the podcast, so unfortunately we seem to be making limited impact with our listenership, but we continue to try to make the world a better place.

‍

Scott Bade:
We just need to get the listeners into more positions of power.

‍

Danny Crichton:
Yeah, exactly.

‍

Scott Bade:
Because I think we need to go with that, but we'll try.

‍

Danny Crichton:
Well, there we go. You are obviously a geo-technology analyst over at Eurasia Group. You're writing reports all the time. I read them, they're very, very interesting and everyone should subscribe.
One of the things that you've been talking about recently is an issue that I think almost no one on our podcast knows anything about, which is New Caledonia.
That is not a place ... Maybe the French know where it is, and if you're in Paris, greetings. For the rest of us in the world, we have no idea where this island is.
It's actually a really crucial priority to France, there have been some major riots and there's some major geo-technology considerations there. Maybe let's just start off talking about what's going on there and why it matters.

‍

Scott Bade:
New Caledonia is somewhat randomly French. It's an offshoot basically of the French empire. It's southeast of Australia. It's in Oceania, it's part of French Polynesia, with other places like Tahiti, and it is a mainstay of French power in the Pacific.
France considers itself a Pacific power, in the same way that Australia and the US consider themselves a Pacific power, because France has actually a very large amount of territory.
New Caledonia has been inconvenient for France for a long time, because there's a significant indigenous population called the Kanaks, who have been mobilizing for independence for a while.
There have been several referenda, so you've had that history of agitation for independence, but you have a significant ethnic French white, basically settler population, that's now several generations deep.
This is not dissimilar to what happened in Algeria back in the '50s and '60s, where that population is not the majority, but they're a very pro-remaining part of France, and so they have really been a block.
Because what's happened is that New Caledonia has this special status, where they're not quite an integrated part of the country, but they've had these referenda for independence and they have crossed certain thresholds, and independence has lost each time, in a large part because of this ethnic French population.
What has happened is that President Macron, basically they passed this updating of the election law, that would allow more people coming from France to vote more easily, and so it would essentially make the electorate white, and make it harder to win independence.
Fast forward to the last couple of months, you had protesters in the street, you had riots, several people were killed. The French, they sent in some gendarmes from overseas, to actually come in and end it. This is the backdrop, that you have this civil unrest in New Caledonia.
The tech angle is that a lot of this, not all, but certainly a lot of this was agitated online, and the reason that I became very interested in this topic, A, was I'm someone who follows French foreign policy, but B, is that there was this bizarre tech angle where you had French adversaries overseas basically spreading propaganda online, TikTok and other platforms, Facebook, Twitter.
It turns out that they weren't just the usual suspects like China and Russia. Azerbaijan turned out to be the main purveyor of disinformation in this conflict, and you're like, "What's going on there?"
I was shocked, and I actually sit next to, in my office, next to our central Asia analyst, who covers the Caucasus, and I turned to her and I was like, "Tell me what's going on here?" Even she was very confused.
It turns out what's happened is that France recognized the Armenian genocide, which happened back 100 years ago under Ottoman Turkey. Armenian and Azerbaijan are not friends, Azerbaijan is more Turkish and so is allied with Turkey.
Because France recognized the genocide, you have had Azerbaijan essentially taking these anti-French stances and supporting anti-French movements in French territories, in Corsica, in the Caribbean and then here in New Caledonia. As a result of this, you're having disinformation coming from Azerbaijan. As a result, the French actually banned TikTok on the island.
If you were trying to predict geopolitical nexus, you would not think Azerbaijan, France, New Caledonia, go. What has happened now is that banning TikTok, and what was interesting there is that they did not ban Facebook, they did not ban Twitter, the French government put out a report saying, "This is an issue on those platforms."
They were trying to do something, and they were trying to crack down on the propaganda, and TikTok is the easier platform to corrupt. I mean it's Chinese, it's not US, but it really speaks to this ability of countries, and even small countries, to really play geopolitics using information warfare, using these platforms. Because it's cheap, and actually you can go on hybrid warfare, with really these asymmetric measures that online allows you to do.

‍

Danny Crichton:
I think what's really interesting here is first, when you think of France, the right to free speech, obviously very ensconced and protected in the constitution, in the norms of the enlightenment coming out of the French Revolution, and so to me it's a level of exception.
I think it's really quite rare for the French government to shut down an entire social network, basically based on events that are taking place in the streets and say, "Well, you just can't use TikTok for a period of time. It's shut down for national security reasons."
The US obviously is considering a TikTok ban in Congress right now. It's passed, we're waiting on the other shoe to drop later this year when the deadline comes and whether it has the best, it's not a ban, it's sort of a ban.
Nonetheless, when we think about New Caledonia though, to me what's interesting is, it's not just hybrid warfare, because you're not focused on just the center, and I think that's what we talk about with hybrid warfare in regards to US and China. Well, it's about the US election versus the Chinese, whatever the case may be.
Well actually it's this wedge issue, it's this distraction that's all the way around the world. There's this tension that's already pre-existing that you can exacerbate, and it sucks the policy-making bandwidth out of decision-makers, because now, I think Macron actually visited, if I recall?

‍

Scott Bade:
Yeah, Macron got on a plane and flew out to New Caledonia, which is basically as far away from Paris as you can get, right, and had to spend time. They withdrew the reform, the voting reform, because they didn't want the distraction.
Again, the tensions were there, this was not something that was ginned up completely online, but the ability for countries to opportunistically or just geopolitical actors generally, right, it doesn't have to be a nation state, although it tends to be, to opportunistically find ways to needle their rivals or use the internet and use social media to bring about some advantage to them.
It's not, as I said, not immediately apparent, why this would be helpful to Armenia, to sell unrest in New Caledonia, but there seems to be some revenge element to this. If you're Russia and China, it's even more, you're trying to distract countries on the periphery. Russia is doing this to the French in Africa.
If you look at a lot of French and western interests in Africa, the French have been getting kicked out of West Africa, where they've had a large military presence for decades now, and there's been these campaigns to fight jihadists.
Again, it's not just information warfare, it's not just online campaigns, but what you're seeing is the French, and Americans too, have been kicked out of these areas, they've been kicked out of a number of these countries.
The presence, the footprint is dwindling, and a lot of that is because you have the Russians going in and spreading disinformation about the French. They're saying, "Oh, it's neo-colonialism." "Oh, they're stealing, they're taking advantage," whatever it is.
Not because the Russians particularly care, although they are investing in those regions, and we can get into that if you want, but they're doing it because it's a way for them to undermine their adversary and distract them and do it in a place where they may not be looking.

‍

Danny Crichton:
We're talking about disinformation as a form of warfare. We can call it hybrid warfare, there's a couple of other terms for it, gray zone activities, many of these are very amorphous terms.
What's interesting to me is first, disinformation has obviously been very successful in the United States, so the Stanford Internet Observatory recently was just shut down, under fairly heavy criticism from congressional Republicans, over COVID and over a lot of different political considerations.
The US hasn't really gotten into this, we do it a little bit. Certainly the intelligence community has done these activities, propaganda overseas for decades. We haven't seen this modern disinformation propaganda work as much, or at least as public.
The only one I can come up with is, Reuters had a story recently about the US military undermining confidence in Chinese vaccines for COVID. For the most part it seems a little bit more mercantile, whereas for these countries it seems to be targeted specifically at either creating geopolitical tensions, kicking countries out, or just creating chaos.
That reminds me of Giuliano da Empoli, a political scientist in Italy and in France, and he defines this literally as chaos engineering. Which is, look, if you are living in a democracy, you need some level of consilience, you need people to come together, you need people to agree, coordinate, and the more chaos you can just throw into the system, these systems break.
I was just texting and talking with Santi Ruiz, who runs a newsletter called Statecraft, and he interviews a lot of interesting policy folks in DC about how their agencies work.
One of the interviews he just did was with the former head of the Domestic Policy Council, which only has 35 staffers for the entire domestic policy of the United States, for the White House and for the Executive Office of the President.
You can imagine, with just 35 folks running all the domestic policy, you don't have to create a lot of chaos, you don't have to create a lot of new challenges to throw an administration completely off the goals that they have, and win in the objectives that you have as a country.

‍

Scott Bade:
No, I think that's right. I mean you're seeing this in anything. The foreign policy apparatus in the US is very distracted right now, because of what's happening in Ukraine and what's happening in the Middle East and what's happening in Taiwan.
No, none of those events are happening because of distraction, but your point that the chaos, if you look at disinformation specifically, and hybrid warfare more generally, there are specific goals and they differ, they differ across regions.
I think when you're looking at the global south, Russia and China are trying to sell their narratives. Whether it's Russia selling the war in Ukraine, whether it's China trying to basically frame itself as the defender of the global south, you have those campaigns.
Then you have the specific campaigns of, okay, Russia prefers Jacob Zuma in South Africa, so is going to spread propaganda that's vaguely supportive. Then you have the actual undermining of institutions and what you see here in Europe is, I think in 2016, Russia was in favor of Brexit or was in favor of Trump in the US, because chaos agents.
Whereas now, Russia is deciding who is going to actually be better for Russia, and their campaign is much more ... Because Russia has a very specific goal right now in Europe, which is the Ukraine war, and undermining support for Ukraine.
Now, they are trying to use this disinformation to undermine pro-Ukraine politicians, or back others, and we just saw that in some of the European elections, we're going to see that in the ongoing French elections.
Whereas the UK election, both sides are actually very pro-Ukraine, so I'm not sure that Russia has much of an interest in interfering, because I'm not sure they ... No matter who wins, it's not really going to change the policy for them.

‍

Danny Crichton:
Well, when we're talking about ISP's and blocking and knocking out specific apps, I mean this gets to some of your other work that you've been writing about, which is around cables and specifically internet cables underseas, mostly across the Pacific, and the U.S efforts to protect its internet infrastructure and create a space where Chinese companies cannot build out the systems of infrastructure you need for the internet, whether it's to Australia, whether it's to Hong Kong, whether it's to elsewhere.
Let's talk a little bit about that work, because I think that dovetails with, not necessarily the application layer, but also the infrastructure layer when it comes to tech geopolitics.

‍

Scott Bade:
Subsea cables are really, really important, as your listeners probably know, and they carry 90% of the internet traffic around the world. They're much more efficient than land cables, extremely more efficient than satellites, and they go back 150 years. We had telegraph cables underwater, crossing the Atlantic, back in the 1800's.
They become ever more important, and they're really a part of the global commons. You can frame this as actually part of the shrinking of the global commons, in that you have this thing, subsea cables, which has been historically an area of cooperation, so you have the US and China having joint projects together, etc.
Over the last few years, and in the same way that in 5G, we've had the fight with Huawei and the US and its allies have basically tried to cut Huawei out of as many 5G telecom systems as possible, or just not let them in. Now, you're having the same thing happen with subsea cables.
In fact, Huawei had a subsea cable company, and it was forced to divest it, because they were finding that they were coming under geopolitical scrutiny.
Now, it's all very unofficial, the subsea cable issue, there's no sanctions, there's no export controls, like with Huawei, there's not these legally outlined.
What you are seeing, is you're seeing, especially in Asia, you're seeing the US and China competing over projects, and the US has the upper hand because it's US, French and Japanese companies are the mainstays in this space, but there are tensions in different parts of the region.
The Chinese, for example, want cable laying projects that go through the South China Sea, within their nine-dash line. They're saying, "Oh, you need to clear the route with us," and companies don't want to do that, because there's an espionage risk, and so now they're going around the nine-dash line, which means going through Malaysia and Indonesia, which is more expensive, but that's what they have to do or else they have to adhere to Chinese rules.
The US now has a entity called Team Telecom, which is this group across the government, which is basically working to shut the Chinese out of new projects, and the Chinese are responding by investing in their own projects, where there was actually a project in Asia that was a joint project that the US basically got the Chinese company kicked out of, and so the Chinese are now building a rival cable.
Now, the Chinese don't have the ability to do that globally yet, but you are going to start to see this parallel infrastructure. That's the competition side. There's also this nexus with really all the current conflicts, because cables are vulnerable, and even though they're part of the commons, they are increasingly, the commons is under attack.
Whether it's Taiwan, where Taiwan is dependent on subsea cables for its internet, and so the cables to outlying islands have been cut numerous times over the last few years, and so there's concerns about connectivity and would the Chinese cut cables before an invasion or as a hybrid war method?
Taiwan's one theater, but then you look at, in Europe and the Middle East, where there have been direct threats to subsea cables. The Houthis in Yemen, will threaten to cut cables, and the Red Sea is one of the main corridors for telecom tables between Asia and Europe and the Middle East and Africa, and there's a bunch of cables snaked through the Red Sea.
Several of them were cut a couple of months ago, and it turns out that it wasn't actually the Houthis cutting it on purpose, it was they cut it by accident, because a ship that they hit with their rocket started sinking and it dragged his anchor, and that's what the US government thinks happened.
It cut off the internet to significant chunks of East Africa and the internet has not been restored, because it's impossible to do repairs there, because you can't get the licensing because of the government there.
We're seeing in a bunch of different places, cables becoming more and more geopolitically relevant.

‍

Danny Crichton:
To summarize, I think a lot of governments want to control the infrastructure that runs their information communications technologies, that hones their data, it solves some legal issues, solves some intelligence and national security issues.
Google was one of the sponsors of, I'm looking at the name here, Gondwana II, which was a cable that connected Fiji to New Caledonia, completed around 2022, a follow up to Gondwana I, which was launched in 2008.
In the South Pacific, Google and a lot of tech companies who have the data centers, who have the AI large language models, basically who have all this cloud infrastructure and need to move data from point A to point B, are building these cables out themselves.
Google's been one of the largest, that has also been contributing to some of these projects, and I believe some of the other large tech companies have as well. The South China Sea, the South Pacific more generally, have been just a hub for this, both because there are a lot of islands that just don't have good internet, as crazy as it is, because they're remote, it's hard to connect.
Submarine cables are not that expensive, but even at a couple of hundred million bucks to connect the next island, if there's only tens of thousands of users, it's very hard to make the math work and the economics work, and so many of these places have been overlooked for a long period of time.
The fear is, that when Chinese vessels were arriving and saying, "Look, we'll connect it for you, but it's going to be centered around Beijing, your data's going to hub through Hong Kong before it goes somewhere else," that triggered a lot of western governments and western technology companies to say, "Well, no, no, no, no, we want to own that," in the same way as you mentioned with 5G.
Look, Huawei was back, five, six years ago when I was reporting on it, but it's still the story today. I think Germany, I read what, just a couple of weeks ago, is finally going to ban 5G, or do a rip and replace, or a please don't keep buying it anymore policy, which has been the debate for five, six, seven years, going on in Germany.
To me it seems interesting that countries that have been really recalcitrant about changing their infrastructure policies are updating them in mid 2024.

‍

Scott Bade:
This is something that started under Trump, five, six years ago. He had this ... Yeah, people were talking about it, I wrote about it for you, the tech crunch. It was an issue, but it was very much a China hawk issue, and I think gradually it's become more and more relevant.
Our firm does something, we have a Huawei ban map, and it's really quite striking to see that map and how it's colored in. If you look at the map, it's essentially, with the exception of India, it's basically all the countries that have sanctioned Russia over Ukraine.
There is that one difference, in that India is also very anti Chinese dotcoms, in fact, more than the US, they've banned TikTok completely. That is a trend, and I think you mentioned data centers, and that's another important part, because you are seeing with this trend of being a sovereignty, I think a lot of governments look at tech and they realize that they're not going to be able to control the platforms, they don't control the social networks, but they can control data, they can control data flow.
Mandating, for example, that data be stored locally, in local servers, it's a way to assert sovereignty. We call it data sovereignty, and that's increasingly important for countries. It's a way for them to have rules that they assign, have accountability, because they'll say, "Well, you as a US company have to have a local employee that is legally responsible for data requests, and has to answer, answerable to the courts or the regulators."
It's also a method of control, because it also allows them to censor or to be able to have law enforcement access. When you think about geopolitical technology, countries, they're being forced to choose between US and Chinese rules, US and Chinese platforms, US and Chinese hardware. A lot of them are trying to find ways that they can carve out a niche, and to assert themselves in some way, and data is one of the ways that they can do that.

‍

Danny Crichton:
Well, we're seeing this very heavily in the Middle East. Over the last year in particular, the UAE, Saudi Arabia, has really gone on to the AI bandwagon, and specifically this nexus of data centers. The Middle East, unsurprisingly has a lot of oil, has a lot of energy wealth, they want to develop, they want to industrialize and move into the technology center.
Both countries, I think combined, have put together hundreds of billions of dollars, at least on plans, on strategies and on memos, to be one of the largest data center hubs globally in the entire world, and own the future of a lot of AI model training, inference and basically become the hub.
It's interesting to me, because that seems to make sense, they're willing to build data centers, it's really expensive, they have the oil, go do it. Then you have these questions around freedom of speech, legal rights. What can be in the AI models, can it be anything, do they care? Because the rights and the freedom of speech are obviously not in any way western standards, and so it's very hard to see how that works out long-term.
That gives you a lot of leverage, and so you look at OpenAI, building multiple major partnerships with these capitals. There's this tension here of, who has the technology? Countries choose China for expense reasons, it's cheaper.
If you have to put 5G across your entire country, China will do it, in some cases, a third to half the price of a US equivalent or an Ericsson or whatever the case may be.
You're left with a bad set of options, and I think countries are trying to navigate those, both from the human rights side of things, but frankly you've got to have the infrastructure and so that's where it starts.

‍

Scott Bade:
I suppose, there are a bad set of options or a good set of options. In the Middle East, what is their comparative advantage? It's money.

‍

Danny Crichton:
That's an incredible comparative advantage. They just have cash.

‍

Scott Bade:
It really works. If they open their eyes, some of these companies, that some of the work that some of my colleagues at the firm are doing, and a lot of other people are doing as well, is looking at the cost of data centers and that cost is blurring.
Where is the cheap energy, or where is the capital to help you build those data centers and buy all the chips? If that's, oil wealth is going to get it for you, that's one way.
Yes, that gives them a little bit of leverage, but in some ways it's them getting leverage over them, so they're pushing the Gulf States to not use a lot of Chinese tech. Because well, okay, we'll invest in these data centers locally, but we don't want to do it on Chinese chips, and do you know what, the Gulf says, "That's a good deal, we can do that."
Then you look at other examples in the global south. One is Kenya, the president of Kenya came here for a state visit, I guess probably last month now. One of the big announcements is that Microsoft is building a big data center in Kenya. That's a great thing for Kenya, because they didn't have a data center, they were having to use a data center in South Africa.
Now they have one much closer, and so that's good for economic development. It's good because they want to be a tech hub, or at least a regional tech hub, and so that helps, and it's good for Microsoft, because they're trying to build out their market there, and then also where are they building it? They're building it right next to geothermal, so they're going to take advantage of that kind of energy.
Now are there issues with geothermal? Sure, there's issues of any kind of energy. It's crucial, you probably have to build your data center a little bit differently, it might be a little more expensive, but the energy is there.
Then, the other consideration is water, and that's also a problem. If you look at in Latin America, where they're doing a lot of this as well, building out data centers, there's also local opposition to it. Because in Chile there's some local opposition because they're in a multi-year drought and a bunch of US tech companies are building data centers out there, and the government there has been very in favor of that, and they have the water to cool it.
They say they do, and they say they've done the environmental reports and that it ... It's a very left-wing government there, and so they say they're very pro-environment and all that, but that is a consideration you're thinking about, but to connect it there, is that Chile and Australia just agreed to build the Humboldt cable, the first subsea cable between Australia and South America.
These things are all connected, and it's part of this, I think understanding that a lot of these countries, a lot of these regions, were left out of IT infrastructure, and not that they're going to necessarily leapfrog the US or Europe, although the state of broadband deployment and 5G deployment in the US, maybe they will leapfrog the US, but they're trying to figure out a way to take advantage of these tech trends and especially the AI revolution now, and these are ways that they can do that.

‍

Danny Crichton:
You're hosting a conference a couple weeks from now, on AI in the global south, and since we're talking about AI in the global south, I'm curious, what are some of the topics outside of data and climate change and energy we've always talked about, are there other issues that are top of mind, that are unique to the global south and different from the global north?

‍

Scott Bade:
Yeah, so I think a lot of the global south is, the concern is getting left behind. I think that they look at the different waves, the different tech revolutions, and each one, there's all these promises and they don't get to take advantage of it.
Phrases that I've heard, I'm not saying I agree with them, but some of the phrases you hear are, data colonialism going around. You hear, there's a concern that, "Oh, well, big tech just wants us for our data," or "They just want our eyeballs and we're not going to actually benefit."
Who gets hired in the global south for tech jobs, they're doing content moderation, which is possibly the worst job in tech. They're very low in the value chain basically, and so they're trying to figure out where can they fit in.
For some countries that's going to be critical minerals, that's being part of the chip supply chain. For some countries it's going to be data centers. For some countries it's going to be being able to actually be part of building applications.
The other part I would say is, there is a lot of talk, and you'll hear this at the UN, you'll hear this in international institutions, you'll hear this in the US as well, in the west, but thinking about how can you deploy AI to help other goals?
Not just tech as an economic development in its own right, which I think, if you're the US or you're Japan or you're China, you're thinking, "We want to be a tech power, and we want to make sure that we own AI in some way."
Whereas, I think if you're the global south, right, we know we're not going to own AI, but we want to make sure that AI is deployed in a way that, A, it's accessible to us, and B, that it's being used to solve our problems.
Whether that's climate change, whether that's public health, whether that's inequality, whatever it might be, they just want to make sure they're not getting left behind, and that you are increasing this global digital divide that has long existed.

‍

Danny Crichton:
Well, and I wrote about this recently in the Riskgaming newsletter, but it focused on large language models for emerging markets, think African languages, languages throughout southeast Asia.
One of the challenges you find, is because of the digital divide, because so many countries have been left behind in the last decades, there's not enough of a digital corpus to actually train an AI model in the first place, and so researchers have actually done benchmarking.
If you look at languages, I gave the examples of Igbo, 31 million speakers, Oromo 45 million speakers. These are not small languages, millions and millions of people are speaking it.
None of that data, none of that conversation that's taking place is legible to computers. We're not recording it, it's not going online, there's limited chat. Obviously there's more today than there was five, 10 years ago, but WhatsApp is end-to-end encrypted, none of that is going into your AI model, and so it's almost impossible to actually train, because what we've learned, at least with the technologies and the mathematics and the algorithms we have available to us today, you need just so much data.
You need the entire world's output of English data, just to have a reasonably decent English model, and every other language is way further behind. What happens if you speak a language that might be the 15th most popular language in the world, but AI just can't be trained on it, there's just not enough information?
When I think about leaving behind, I've written about Dani Rodrik's work this year, on this very subject, and he's like, "Look, the East Asian industrialization model is gone. At this point the US and China actually produce things cheaper at scale with automation, than any country can hope to do with just cheap labor."
You can't compete. There's no ladder up this road anymore, and so countries like India went with services, Wipro, Infosys, and did automation of business process automation.
Now, that's getting decimated by AI, because you can read expense reports automatically, you can have automated code with coding agents, and so the question is what's next, what's another model?
AI is definitely not one of them, because the capital upfront costs to getting onto this bandwagon is so high, that only countries like the UAE, Saudi Arabia, potentially could catch up, because they have, as you said, the capital and just the wealth to be able to go do so.
There is a really open large question, so hopefully your conference answers it. Hopefully in a couple hours in New York City, done, you'll solve the problem and we'll be off to the races.

‍

Scott Bade:
What is interesting is on the language thing, is you do see, this is an area of pseudo-industrial policy, and you have a number of governments who have come in and said, "We want to fund our model." There is, in the Gulf there funding an Arabic LLM. The French have said, "We want a French LLM."
Some of that is not just for the language, but it's also about values, and this is another way that you can have quote unquote, data sovereignty, right. I mean, it's not data sovereignty, but it's LLM sovereignty of some sort.
China, you want to have your own LLM, because you want to be able to have ownership. What's interesting, I was just reading a report on that actually, that one of the Chinese LLM's, actually, if you look at the English versus the Chinese, they're different in how they come out. You do have that aspect as well.
Certainly, I think that if you're in the global south, that you do have a concern of, if it's a handful of Western companies who are dominated by a very specific demographic, and look, how many of them are all in San Francisco? I'm from the Bay Area, I love San Francisco and Californians, but we don't know everything.
There are going to be blind spots, and I think that a lot of tech companies are thinking about how they can address that, but certainly making sure that they have a voice. That there is some anxiety I think, in a lot of countries, about making sure that they have some voice in this, so that this isn't technology and it's just foisted upon them.

‍

Danny Crichton:
The other thing that a lot of folks don't realize, just given the point you just said, is even the multilingual large language models are predominantly English.
In general, the way these models are built is, you might do 94% English and the other 6% of the data is other languages, and so it thinks in the sense of whatever an LLM thinks, we don't really know, what do sheep think?
Well, it thinks in English, and there is some proof, and I actually talk about this in the newsletter, there is some proof that they actually do literally think in English, and then they basically translate at the end into the language that you asked for.
Even beyond just values, there may be a little bit of a deeper humanistic reason to say, "Look, we actually want it to think in a Chinese way, in an Arabic way, in a French way," because languages do change how we can think and the range of emotions we can express. They have a lot of control in that way, in how we can express, and so presumably the same thing is true with our LLM's as well.
Let's wrap up, we're in mid June as we're recording this. It's been election high season of epic proportions this year. India, the EU, this year I believe the majority of people around the world are voting.
We'll see if Japan might vote later this year. Certainly they are in Tokyo, but the two major elections going on outside of the United States are obviously France and the UK. Both of which were massive surprises, came out of nowhere, we didn't expect ... The UK was time limited to late December, early January, so it's just a little bit early.
In the case of France it's a complete shock, no one was expecting. Obviously there's a ton going on, and by the time we publish this, everything will have changed.
I'm curious, just with the technology focus, if you're seeing, this was supposedly the year, the deep fake, this was the year that all the AI technologies would destroy democracy, and it seems like democracy is actually doing okay compared to the worst apocalyptic visions that we heard from the beginning of the year.

‍

Scott Bade:
As you say, it's the year of elections, and we've had a number of big ones that aren't ... Indonesia and India and Mexico, Taiwan. What's interesting is, a lot of these elections were actually quite predictable, in terms of they weren't big change elections.
Maybe certain unpredictable results at the end, but Mexico, everyone knew that Claudia Sheinbaum was going to win, and she did very handsomely, and so there wasn't necessarily going to be much disinformation or AI infiltration there.
I'll give you a couple examples, we were talking about the UK and France. Taiwan has historically been a big focus of disinformation, because the Chinese are constantly trying to influence Taiwanese politics and Taiwanese culture, etc, and the conversation.
In a way it has worked against them, because the Taiwanese noticed the last couple elections, and so they've really grown resilient to this. I was in Taiwan last year, and meeting with civil society folks and government officials, and they still have a whole digital ministry, which in a large part is trying to counter the Chinese cyber and disinformation threat.
They've created a lot of mechanisms. They actually use AI, they have a chatbot that you can feed in to say, "I heard this thing, is this real?" It'll give you a synopsis and say, "Oh, no, yeah, this is Chinese disinformation."
Taiwan's actually pretty resilient, and not that the Chinese didn't try, and certainly there are narratives and the KMT still won the parliament and the parliamentary election has some more, not pro-Beijing, but more Beijing right party. Did that play a role? Maybe on the margins, probably not too much, the incumbent DPP still won, their candidate won.
India, there's not really much going on in there. They're actually, the use of AI was by the candidates themselves. You saw this in Pakistan, you saw this in India, and actually in South America too, where you've had candidates of political parties actually using deep fakes of themselves. Or in the case of India, you had one case, I read that you had this candidate whose dad was a deceased politician and he used a deep fake of his dead father to endorse him.

‍

Danny Crichton:
That feels very Star Wars.

‍

Scott Bade:
It was very Star Wars.

‍

Danny Crichton:
This is my father. You could show up Obi-Wan Kenobi, yeah.

‍

Scott Bade:
That's a weird use of it, but again, probably if it had any impact, it's probably on the margins.

‍

Danny Crichton:
Sorry, I'm going to get distracted, but I just have this image of bringing out the can of the top politicians. The RNC, you're going to have Abraham Lincoln speaking on stage. At the DNC you'll have FDR speaking on stage. I mean, why not just bring everyone back with an LLM and a hologram?

‍

Scott Bade:
I mean, basically, you have the Disneyland Hall of Presidents, a campaign rally now. Yeah, I mean, this is where you are seeing some creative uses, but then we did see in New Hampshire, right before the New Hampshire democratic primary, which was not really a competitive primary, but Joe Biden still wanted to win it.
You had these robocalls go out, that were in Biden's voice saying, "Oh, don't vote, no reason to vote." This is where I'm actually a little optimistic, is that it got caught pretty quickly. It was pretty obvious it was a robocall. It got in the press, I think it was everywhere. I'm sure that people do not care about this, but it was debunked pretty quickly.
Then what happened was, a number of states immediately started banning the use of deep fakes in election campaigns or in ads. The FCC came in and banned the use of robocalls in election campaigns, and so not that the issue is completely dealt with, but actually the policymaking apparatus worked pretty well there.
The bottom line is, that I think that AI is certainly marginally making a difference. We're not seeing it make a huge difference yet, and part of that is because we're finding ways to combat it. Part of it's also because there's a lot of crap out there, and I'm not sure that the AI generated stuff is making a marginal difference.
It's not making an impact in the way that I think people thought it might, and I think as we go on and deep fakes get better, that's going to become an issue, but we're not there yet.

‍

Danny Crichton:
Yeah, if I had to close this out, I'd say, we ran a Riskgaming exercise for 55 folks back in January, a mix of government, state government, election officials, and a bunch of tech folks.
The biggest lesson from that whole experience was, we're actually relatively immunized. There's a lot of information held by a lot of different actors, the private industry, intelligence agencies, the rest of government, state, local, federal, your next door neighbor might know something that's going on, because they're getting a text message and you're not, because you're not on the right list.
The biggest challenge is, how do you create an immune system to some of this? It's okay today, and it's probably in line from an arms race perspective with what's available.
I get a lot of bots on Twitter. It's pretty obvious what the bots are, because they're just so ... But I can imagine as LLM's get better, more culturally aware, more culturally sensitive, as more of my data gets fed into this, and you can customize these more towards me, that they become more and more compelling and it's harder to ignore them.
I feel like we're going to skate by 2024, in this year of democracy, but it will get tougher going forward.

‍

Scott Bade:
Yeah, I think that's right. I think right now, you look at the campaigns that have been happening, they're surprisingly being fought over issues like inflation, old fashioned bread and butter, economic issues. India, that's Mexico, that's a lot of the EU stuff has been over economics.

Danny Crichton:
Well, which house it always ends up being, because people like houses, food, and no one likes inflation, but-

‍

Scott Bade:
No one does.

‍

Danny Crichton:
I feel we've turned full circle. We're back to New Caledonia and the economics of a small island, all the way to the economics of the largest countries, which means we've covered the full world and basically every continent in one tight level conversation.
Scott Bade, geo-technology analyst at Eurasia Group in Washington DC, thank you so much for joining us.

‍

Scott Bade:
Thank you very much for having me, Danny.

‍