Transcript

Gulf states wedge AI efforts between superpowers

By Reuters

This transcript was created using speech recognition software, it may contain errors. Please review the episode audio before quoting from this transcript.
Peter Thal Larsen
Artificial intelligence is an endless devourer of data, power, chips and, above all, capital. The boom in large language models since the launch of OpenAI’s ChatGPT in 2022 has drawn US tech titans to spend tens of billions of dollars on processers and data centers. But that’s nowhere near enough to fill the $1 trillion in projected spending that Goldman Sachs foresees. Enter sovereign wealth funds, particularly those in cash-rich Gulf states. Just look at some of the things that have happened in the past few months: Abu Dhabi has set up a fund called MGX backed by Mubadala Investment. It recently invested in OpenAI and has a new partnership with Microsoft and BlackRock. Then there’s G42, an AI startup backed by the United Arab Emirates, which also has a deal with Microsoft and is a major customer of AI chip firm Cerebras. Meanwhile, governments are getting to grips with the notion that they need to develop sovereign AI. What this means for the Gulf states, and what it tells us about the geopolitical tussle over artificial intelligence, is the subject of this week’s Viewsroom.
Jonathan Guilford
Welcome back to The Viewsroom, the weekly podcast that invites you, the listener, into a lively debate with our fellow columnists about the biggest stories of the week. I’m your host, Jonathan Guilford.
Peter Thal Larsen
And I’m Peter Thal Larsen.
Peter Thal Larsen
So, Jonathan, we’ve been talking and writing a lot about AI over the past couple of years. But the Gulf involvement is a relatively new thing, isn’t it?
Jonathan Guilford
Yeah. I mean, it’s something that you have to kind of get to grips with, the fact that everything in AI is moving so quickly. It’s sort of on this, you know, exponential curve. So I think geopolitically everybody’s always going to feel like they’re a little bit behind the eight ball. But you have seen some of these links between the US kind of industrial base, let’s call it, and the Gulf get set up, especially Microsoft, which has been everywhere. I mean, not just in the Gulf, in Europe, they forged ties with Mistral, they’ve, you know --
Peter Thal Larsen
Mistral is the French -- the French AI hope, right, yeah.
Jonathan Guilford
Right, exactly. And they’ve kind of -- almost seem like the US AI industry emissary to a certain extent. And now you have BlackRock, larry Fink is very much getting involved in all of this. So you are beginning to see that kind of cross-pollination. But obviously extremely early days. These are polities that kind of stand in between the major power blocs in China and the US. So we’re obviously probably going to see this evolve quite a lot in the near term. And on that front, we should bring our guests into the conversation, who have been investigating exactly this subject. Joining us this week is Karen Kwok, Breakingviews AI specialist in Europe, and George Hay, editor for Europe and the Middle East. Karen, George, welcome to The Viewsroom.
George Hay
Hi there.
Karen Kwok
Hello.
Peter Thal Larsen
Thank you both. I’m really interested in the sort of the geopolitics of all of this, but maybe if we just take a step back to start with, I mean -- the artificial intelligence race seems to depend on a few key factors, right? First of all, engineers. Second of all, vast amounts of data. And finally, lots and lots of money. We know the Gulf states have lots of money, but what else are they bringing to this party?
Karen Kwok
I think first and foremost, for the Gulf, as we all know, they produce a lot of oil and gas and produce a lot of cheap energy. So that will actually help them to pump out -- to power these data centers, which is essential to process data, to train, like, large language models. So this is first and foremost. And second of all is that, not a lot of people, like, not very widely reported is that increasingly the UAE, for example, so they actually managed to build their own large language model. There’s like this government-funded, technology innovation institute, actually produced open source LLM called Falcon, which is performing as good as, like, Meta’s Llama model. And it’s one of the world’s most advanced open source LLMs, which shows that Abu Dhabi is actually quite serious about investing and producing at the next AI revolution, basically.
George Hay
Yeah I mean, I think that’s -- these are the kind of important points Karen is making there. I mean, the clearly everyone knows that the Middle East has lots of money. Everyone knows they’ve got lots of power. And that’s partly the reason why they have so much money. But the key thing is they’re not just a pile of money and a lot of oil. They, in the UAE in particular, they are actually, they have been building their capabilities to be a kind of, you know, thought leader or, you know, they’re not just a -- they’re not just the follower in this game, which is one of the more interesting bits of the story.
Jonathan Guilford
Well, let’s get down into the specific kind of organs of that, right. I think MGX is the one that looms large, especially for folks looking at this from over in the US, just because, you know, they’ve stood up the $30 billion fund with BlackRock and Microsoft to invest in data centres and renewable energy. They have, you know, the kind of separate Microsoft partnership. Like, what is MGX? Exactly, like -- how does it relate to the UAE’s sovereign efforts and what is it doing in the industry?
George Hay
Yeah. So basically, MGX, I mean, we were out in the Middle East recently. One of the ways that that was explained to us is: it’s like a -- it’s a big fund. It’s a big pot of money. It’s a bit similar, in one way of thinking about it, is, it’s like the artificial intelligence version of Mubadala, which is one of the UAE’s big sovereign wealth funds. This one is specifically focused on a AI infrastructure, the power that goes into that, and it is investing kind of all over the place. The BlackRock fund that it’s, it’s putting money into, that will be doing that kind of thing, not just in the Middle East, in the US as well. So MGX is very much the pot of money or a big pot of money, but it’s not the same --
Karen Kwok
A hundred billion.
George Hay
Indeed. But it’s not -- it’s only a part of the infrastructure that the UAE is building.
Karen Kwok
Yeah. And also, like just to add on that as like given the pace of AI startup, like how fast it is, the startup raise money these days, yeah. I think having a separate fund for them is very important, you know, to chase that pace, basically.
Peter Thal Larsen
Okay. But I’m still hearing basically, money, right? So let’s talk about this company, G42, which I suspect most -- well, I suspect some of our listeners have never heard of. I certainly hadn’t heard of it until April when they announced this sort of quite convoluted deal with Microsoft, where Microsoft is an investor and they went on the board. Well, Karen, you’ve been out there with George recently talking to people about this. What can you tell us about G42 and how should we think about that deal?
Karen Kwok
So first of all, I think of G42 is kind of like a company, like state-owned company that is a merger of different sorts of things that, in order to power, and then having this core AI snippet, that center into it, like, they have like this Core 42, which is kind of like a data collection for it to produce AI. So they have like this research institute is called Presight that actually -- they go to libraries, of Arabic library, to scan data to put to together -- to collect data and so on. And they also have like this, different parts of it that’s focused on particular sectors and help -- hoping to use AI to power the sector, for example, in health. They have this Health 42, which is like a merger of a Mubadala health division with G4 -- a division of G42. And it’s got a combined effort of AI and healthcare together to join forces in order to develop more products in the healthcare industry.
Peter Thal Larsen
Right. But this was quite -- this was quite a sort of a curious deal, wasn’t it? Because it wasn’t just sort of Microsoft going and putting some money into an AI startup like it’s done in France and the US and stuff. There was also a whole kind of like US government involvement in -- sort of it feels like almost brokering this deal, right?
Karen Kwok
Yeah. So, yes. So it’s, what’s interesting in this deal is that as part of the deal, as I was talking to some people that were telling me, is that Microsoft will sell a proportion of their cloud capacity to G42. And then in between that deal that G42 used to be having relationship with Huawei, which is this Chinese telco company, and then they have to like, end relationship with them in order to get the support from Microsoft. So in a way, it’s, kind of feel like UAE is using it as a bargaining chip to pick a side, to side with us in artificial intelligence, in order to let go of their relationship to China, basically.
Jonathan Guilford
I’m really curious about that, as it relates back to a point you were talking about previously. Like, G42, their efforts to go into all these libraries and kind of digitalize texts and whatnot. You had a piece that had like this really, really fascinating little data point in it, which was, there is only roughly like 1% of text on the internet is in Arabic. And so you look at kind of the US and China and kind of how their efforts have grown up. They’ve grown up in very like data-rich environments, right? Really large training sets like an internet that is geared kind of in both polities towards almost like producing data that is perfect for that training. And I kind of wondered, like, does that constrain the UAE’s ambitions for its own efforts at all? Like, does that inform kind of how they try to pitch themselves in the middle of this geopolitical tension because there’s a restraint on the homegrown industry?
Karen Kwok
Yeah, I kind of feel like that’s the question I had when I was talking to them, actually. And then, but then in their view, actually, they think they see it more as an opportunity because not all, they’re like, no other Western companies can come in and just like make the same effort to build, like, Arabic, collect data and so on. And actually using their local language to build actual language model is actually, much cheaper for companies to produce models internally. So one of the companies we talked to is saying that, using Arabic is like actually one-third cheaper than using an English model and then they have to transfer later to Arabic when they use it internally for their business. So, yeah.
George Hay
But yeah, I mean, the point about -- there’s a lot of emphasis in the Middle East, both in Saudi, in the UAE, on data security. They don’t necessarily want the data in their backyard being stored on clouds or physical infrastructure in the US, for example. So there’s a kind of -- there is a kind of data security angle here. But like what Karen is talking about, there is just, there is also an opportunity if you become the best, if you are training your models in Arabic, then that creates a different product than if you’re training -- training them in English. One of the, one of the ways that people were explaining it to us is just like, here, if you type in “King” to a Western, English-trained LLM, then it will give you a particular picture of someone on horseback. Whereas if you do it in the Middle East, training, using Arabic to train the LLMs, then that will give you a picture of a sheikh. Now that’s a very facile example, but there’s all sorts of things that -- kind of all sorts of implications that spring from that which mean that in that part of the world, you know, they might prefer an Arabic LLM, especially as it might be cheaper, as well, for the reasons that Karen was saying.
Peter Thal Larsen
So, so talk just a little bit -- it feels to me like, just stepping back from this a little, that there is this sort of tension here because you have this industry. You start with the chip industry. It’s growing up as this very global industry. You know, the chips are made in Taiwan. The company that makes the machine that makes the chips is in the Netherlands. The mirrors in those machines come from Germany, and so forth. It’s kind of spread around. And then, and with AI, you almost sort of have something similar where people are either pulling datasets from one place, they’re getting investors from another, they’re sourcing their chips from elsewhere. But overlaid on that, you now have these sort of sovereign kinds of interests, which is governments who are worried about being left behind, worried about security and are trying to sort of control things. And I just wonder, the Gulf is quite interesting in this respect. I just wonder from sort of -- from the reporting you did there, how does that play out? Are we going to see sort of the division of the world into like different AI blocs, or are they all going to sort of rub along but in a slightly difficult way? How do you see it working?
George Hay
Well, it’s very obvious that there are going to be two AI superpowers: the US and China. And it’s not like people at G42 and the kind of delivery mechanism, or the operating bit of the UAE AI Inc, which Karen was talking about earlier. It’s not like they think they necessarily are going to be able to be as important, as high-end, as either of those two. But the way I suppose, the way they look at it is, they, it’s not, it’s not really just about chatbots and those kind of -- the AI outputs that you can see at the moment, which have clearly come a long way. But what people in that part of the world and all around the world are worried about or focused on is just the next stage of AI. When you, once you come to -- you’re leading up to artificial superintelligence, leading up to a situation where it becomes important, the implications of these, you know, incredibly sophisticated pieces of technology is that they may be more intelligent than humans. And then that has all sorts of kind of security and moral questions. And the way they look at it there in that part of the world is, you know, one consultant who we talking to talked about it in terms of, at some point in five or 10 years’ time or whenever it is, people might think of it like nuclear weapons, and you might have to limit who has them. And what the Gulf is saying, UAE is saying, is that they want to be seen as one of these powers, one of these AI superpowers, even if they’re not as powerful as the US or China. They want to be a player, and that’s why they’re doing all this stuff.
Karen Kwok
And also, like, I mean, if you look at the history of technology development, right, when is that -- when it comes to like the internet age, like we have, we see like the breakthrough of like Facebook, Microsoft and all these, like, Western companies dominating the World Wide Web. And as you can see, like more than half of the World Wide Web is actually in English and Arabic is only 1%. I think what Saudi and UAE are trying to apply here is that they, in this next revolution of technology, they want to have a place and they want to be the one who can also profit from it rather than being the pure consumer of this technology. And I think that’s what they were trying to go for.
Jonathan Guilford
Right. And it’s interesting because you kind of posit it as, they’re caught between all these different concerns, like the vaunted, like, existential risk that you’re talking about, George, or these kind of more subtle, nettlesome strategic questions about, you know, how do you ensure, like, your footprint within like the global data space or what have you. And I think there’s sort of like this, I guess, stereotypical view that a lot of big funds out there are maybe not the most sophisticated investors compared to strategic investors or what have you. And that specifically, you know, Gulf investors, kind of like we saw with, say, SoftBank’s Vision Fund, have thrown a lot of money at things that don’t always work out. Is there a danger here that Gulf investors are sort of once again getting involved in a late-stage hype cycle that’s just about to kind of exhaust and collapse and then a lot of this is going to end up being fruitless?
George Hay
I think there is certainly that, but I don’t think it’s necessarily fair to single out the Middle East as the, you know, the only potential fall guys. It’s clearly, the SoftBank Vision Fund 1 parallel is front and center, or in the minds of the people who run UAE, because they provided 15 billion and Saudi, they provided 45 billion. And they didn’t get a very good return on that at all. But, you know, it’s more just, as we were saying, there’s MGX, which is a large amount of money, there’s G42, which is a kind of attempt to turn stuff into commercial applications. But possibly the key thing is that the UAE, you know, they are, they do have kind of academics who are kind of producing things like Falcon large language model, which are actually, people in other parts of the world are going, okay, that’s a valuable contribution to the debate or that’s interesting. And it’s not -- there’s some reason to think that it’s not just them throwing money at a wall and seeing what happens. They are actually kind of, they’re actually bringing something to the table. And that’s the difference, really, between what’s going on here and what’s going on with, what went on with the Vision Fund, where they just literally handed over a load of money to Masayoshi Son and then he didn’t do much with it. So, yeah, I mean, but the risk is there. It’s, you know, you don’t want to say that it isn’t.
Peter Thal Larsen
Yeah, it’s a good point. I mean, obviously if, if these investments proved disappointing, shall we say, then they’ll be in good company with Microsoft, Amazon, Meta, you know, and all the others.
George Hay
Yeah.
Peter Thal Larsen
So, George, Karen, that was fascinating. We have to leave it there. We’ll come back, no doubt, to hear about how the AI nuclear arms race is proceeding. That’s the mental image that is going to stick with me from this discussion. But very much appreciate your time. And, yeah, talk to you again soon.
George Hay
Thanks.
Karen Kwok
See you.
Peter Thal Larsen
Thanks for tuning in. This podcast was produced by Oliver Taslic in London. You can listen to a new episode of The Viewsroom on the Reuters app, or your favorite podcast platform, every Thursday. Also, don’t forget to tune in to our sister podcast, The Big View, every Tuesday, as well as the other great podcasts from the Reuters team. If you like what you heard, please follow the Viewsroom and let us know what you thought. And check out our views on the biggest stories in business and finance every day at Breakingviews.com and Reuters.com.