130. The Brutal Truth Behind Britain's AI Gamble

0:11Hello and welcome to The Rest is Money with me Steph McGovern. I'm with me, Robert Paxton. Now, we have been talking for yonks now about our government needing to do more to make sure that we don't fall behind in the AI revolution. And this week, Robert, we heard from Kia Starmer on that, didn't we? That's right. I was at the University College London's East London near Stratford campus, where they do actually quite a lot of AI research, where the prime minister was launching the so-called artificial intelligence action plan. And that was very much the sort of broad brush set of announcements.

0:51I have to say it was ambitious, but now we're going to dive into the detail of what the government is actually going to do to fulfil its promise, the Prime Minister's promise, right, to make us a global leader in AI. Let's find out from Peter Kyle whether that is a realistic prospect. So here's our interview with the Secretary of State for Science, Technology and Innovation, Peter Kyle. Peter, great to see you. You have gone all in, some would say belatedly, on artificial intelligence. Give us a sort of vision of how you see an AI-powered UK that our listeners can really connect with. I wouldn't say belatedly, because we've only been in office six months, and we got the full action plan out the door.

1:43There wasn't a lot of talk about it during the general election, let's be honest. You and I have had words about this. But anyway. We have. We have. Well, I think let's start at the end then, because it's the end at which people start to feel the difference. I want a world where people will have government in their pocket. In order to get government in their pocket, you need to have good use of data. In order to have good use of data, you need to have the systems across government where data is interoperable, so it's talking to each other. We have privacy and safety baked in at the outset so that government can analyse, process and use data very wisely.

2:16And then in front of that, so one step further back, you've got to have what is for us the first step, which is building the data infrastructure for the country. And we announced this plan this week where we're going to increase the computing power of our country by 20-fold, the public computing power by 20-fold in five years. We're going to create AI investment zones so that we can build the data centers that are the factories for AI that can process the data and get it back into people's pockets and put to task for them. And we're going to make sure that government has the skills and the ability to harness technology at the very start because you are right.

2:50Government is always last to bring this technology in. We're creating two units within government, one in number 10 and one in my department, both of which will be responsible for creating the conditions to have those cutting-edge technologies developed, designed and deployed in our country rather than always buying off the shelf from other countries. And then secondly, bringing it into government, putting it to the use of the public first, not waiting till the last where the rest of society, banking, travel, commerce, are already using it before we even think about how we use it for government for citizens.

3:23So can we just then unpick some of this and some of the detail of it? So one of the things you said first there was about the compute power. So all of the kind of data processing that's needed to actually make AI function. There was a plan to have this supercomputer in Edinburgh. Exascale. Exascale is the word. Yes. And this was going to be the most powerful computer in the UK. And that got scrapped, didn't it? You know, money was talked about. It's going to cost something like 800 million. But what you're suggesting for this compute to increase 20 fold is going to need a lot more even than that.

4:01And that's already been scrapped. So what's going to happen there? Have you got the money to do it? Are we going to have these supercomputers, these extra skill computers? And if so, when and how? Well, firstly, I didn't scrap anything. What happened was that Boris Johnson promised verbally to Edinburgh University that they would have it. Then Rishi Sunak recommitted to it when he was asked. And then Jeremy Hunt announced it in the budget. But what never happened was what has to happen if something's going to actually get delivered in government. Treasury needs to allocate the money for it. So when I came into office, there was a verbal commitment, but there wasn't a single penny allocated by the Treasury.

4:39So for me to actually carry on with that, I would have had to cut£800 million worth of existing funding from my department in order to pay for something which a previous Tory prime minister had promised verbally but never actually bothered to pay for. We should point out at this juncture that that is Rachel Reeve's decision not to fund it. You didn't have to fund it through making cuts. She could have decided this was a great project. You're a government that believes in investment. You chose not to go ahead with an exascale computer. There are exascale computers in the US. They are building one in the EU.

5:15We've got extraordinarily powerful computers in China. We don't have one. And on your plans, we're not going to have one. It's not true, actually, because I had the commitment in my department, but I had no money for it in my department. That's not Rachel Rees' responsibility. She could have funded it, is what I'm saying. She could have funded it. The people who announced it should have funded it. The people who announced it should have funded it. And this is all the smoke and mirrors that was going on that we had to deal with. So what I wanted to do was not just say, OK, how do I deliver all the things that were promised two years ago by previous prime ministers and it had been delayed?

5:47I mean, nothing had happened. So what are we going to have instead? Exactly. And this is the real question. So the first thing I had to do was be honest with people and then have a clean sheet of paper and start from scratch. Because I had no strategy. I had no funded proposals. So what I've done immediately is got the Bristol supercomputer, it's now being expanded by a third, that's underway. The Edinburgh University already has a supercomputer, I've extended that by a year, and now when it comes to publicly funded compute power, I am now working behind the scenes on a strategy which I'll release in the spring.

6:19Every bit of that will be costed and it will be funded and it will be delivered. That is public sector compute, which pales compared to the private sector investment we need into our computing as well. We need computing for research purposes, for universities. We need compute power for AI purposes. And we need it for the general economy because 85 % of businesses process data in order to serve their customers. So sometimes they have that in-house. Other times they buy it in from companies like Amazon and Microsoft and Google and other sort of domestic suppliers, which are smaller than those big three.

6:55But what we need to do is just build the general capacity and compute capacity as a country, which is what the strategy that we've announced. And you do accept, don't you, that we are way behind. We don't have a single supercomputer in the world's top 50, for example. We don't, and we are behind, and we will get to where we need to be. But the needs for computing today are so different to five years ago. The chips, infinitely different. And now the chips are able to do a broader set of processing abilities. You can do AI and research purposes with the same chip set. So things are adapting and evolving.

7:25So we need to do something that's for 2025 going forward, not 2020 going forward. And that's the challenge that I have. That's the opportunity that I have before me. We've committed to a 20-fold increase in our public compute in five years, and it will be delivered. It is a huge target. But if we're going to do all of the things that we need to on behalf of our country, to empower citizens in our country with digital technology and crucially make sure that everybody from everywhere benefits equally from it, then this is the first step. Yep, we've got to get this right, the basic infrastructure.

7:56So just this week, we've announced£14 billion worth of private investment into our compute. Since we came into office six months ago, there has been£24 billion worth of investment into our data infrastructure. This is private sector investment. Private sector investment. So£24 billion and now another£14 billion on top of that. That is unlocked because of our pledges and policies on regulatory reform, on planning reform. and now we have the investment zones. What they mean is that we will find parts of the country where there is an excess of power because these data processing centres are incredibly power-intensive.

8:31So parts of the country that have an excess of power... And that'll mostly be in the north, won't it? ...ready connection to the grid where we can do a deal with the planning authority so we can expedite the building, construction and deployment of these data centres. And actually the first one we announced was in Oxfordshire because in Cullum there is a site right next to our fusion reactor research centre. And are you going to be building a new modular nuclear reactor there to power the data requirements, the computer on the ground? We already have a 100 megawatt connection from the grid directly there, and that could be scaled to 500 megawatts.

9:07Without building a new nuclear. Without the need to build a new power station there. That's because of certain circumstances, and we're looking at circumstances around the country, and there will be others that will be announced, and we're going to announce... And just to be clear, if you look at what Microsoft is doing, you know, it's investing astonishing$100 billion in this so-called Stargate computer, which will be the world's most powerful computer by quite a wide margin. It's building its own nuclear power station. Are we going to be seeing more nuclear power stations in the UK to power these computers?

9:40So what we've done is we've set up a council, an AI energy council that's jointly chaired by myself and Ed Miliband, so we can bring the voice and the needs, because the future data infrastructure is going to have huge energy demands. But also AI, don't forget, is going to make huge efficiencies in the way we use and distribute energy into the future, both as consumers, as households, and as a country as well with the national grid. So these are very complex interactions between the need for energy and also what the outcomes for it, AI, will do for energy. This is why AI is so sometimes difficult to grasp because it's not like a single purpose.

10:16But you said nuclear will be in the mix as an important point. There's no question. Yeah, there's no question. I mean, nuclear will be. And also we've got Rolls-Royce that are developing small modular reactors. That's something that's being very heavily invested in by the private sector as well. I mean, in America, both Google and Microsoft and others are investing in nuclear. They're reopening Three Mile Island, for example. But we are moving to a low carbon and renewable energy supply system here in this country. That, again, is an industry in itself that will create wealth and jobs and lower prices for our country in the long term.

10:48But we need to bake into this from the start, the needs of AI and digital. Because, look, if we don't, we're always going to be buying off from other countries. And this is going to be where the economic growth for the future is going to crucially be centred around. and there is a huge opportunity if we can be part of the few countries in the world that is capable of designing, upscaling and implementing some of the new innovations that are coming out of this incredible age that we're living in. Otherwise, it'll just be China and America and us hoping to get a good enough relationship that we're always near the front of the queue when it comes to actually benefiting from it.

11:28We will always be importing it, in other words. And I think at the moment we have this incredible opportunity to use the culture of scientific endeavour that we have here, the fact we have great world-class universities. We have an investment landscape which is good but getting better when it comes to investing in innovation. And we need to really, really capitalise on this moment in time. Otherwise, we'll miss the boat. And this is something we've talked a lot about on the podcast. But can I just ask about these AI growth zones just in terms of more detail around them? Because as Robert's just said, you know, it's probably going to be up where I am, hopefully some of these zones up in the northeast and obviously Scotland and things where there is that, as you say, kind of surplus of energy creation.

12:11But how is it going to work? Are you going to identify the areas yourself and then kind of wait for businesses to rock up? Or are you going to decide it with private investors? Are you going to decide it with businesses on board? And then how are they going to compare to things like free ports? You know, is this going to be a place where I know that the plan is to have, you know, easier planning processes and things in these areas. But what about things like the tax incentives? What do these zones look like and how are you choosing them and who with? So the first one we've identified in Cullum in Oxfordshire.

12:47We want the parts of the country that would benefit most from this scale of investment to be the ones that get there because we're a progressive government, I'm a progressive politician. And what we're doing at the moment is setting out the criteria to become a growth zone. And then we are putting those criteria out and inviting areas to get in touch with us. And what councils would get in touch with you? Local authorities, regional mayors, city mayors can get in touch with us with it. And if they meet the criteria, look, the thing people have to understand about this is this isn't about government.

13:22You know, you don't need tax incentives to, you know, we've just one single investment announced a couple of days ago is£2.5 billion into one data centre. This was in London, this one. But what they want is to identify areas, because at the moment you've got these companies who are trying to find bits of the country where they can have a lot of power, a quick connection to the grid, because one company told us they were given 15 years as a timeline to get connection to the grid for the power that they need. So what we're doing is finding, using the power of central government, the convening power, to find parts of the country, invite parts of the country to come to us that have all of the criteria, Excess energy supply, easy connection to the grid, planning authority willing to act very, very fast and use the new powers that are coming online with the Planning Act so that we can expedite the construction of these things.

14:18So all of these things will be in one site. And we don't need a tax break because the global investment is there. But if we can't get and seize it and present it as a huge opportunity to invest in our country, believe me, they will go to others and they are going to others. When I'm on the phone trying to sell Britain to some of these companies, I'm always on the phone or on these calls with people. And when I say, where are you? They list all these other countries that they're in. You know, each time they're in a different country. And when they say why, it's because they're there doing a deal with another government because the other governments are trying to get their investment into their country.

14:54So just to double check then. So it's going to be a case of the regional authorities, mayors, whatever, that you're going to invite them. And they're going to put together a case for why they should have it a bit like, you know, we've seen with City of Culture and everything else. How long is that going to take? Because that's going to be ages, isn't it? No, it won't. The authorities get themselves, you know. It absolutely won't. We are absolutely hungry for this. You know, we are talking about a process that is weeks, not years. We're not asking people to put together a bid that's like the Olympics in 2012.

15:25You know, what we're doing is saying identify the areas. My department will work with them incredibly quickly. when we identify those areas, we will then start putting it out to those companies that are coming to us and saying they want to invest. Can I ask you about another aspect of the wealth creation potential here? Back in October, a government set up advisory body called the Council for Science and Technology wrote to the Chancellor. You were copied in, pointed out that in the UK, and this is a really encouraging statistic. There are, it says, 750 venture capital backed technology companies across the UK who have already scaled to generate each of them more than$25 million in revenue.

16:14it, like many of us, is very worried that although we have these smaller companies, what typically happens is when they have the potential to become multi-billion dollar companies, we don't have the capital providers in the UK. And what typically happens is they sell out particularly to US either investors or US companies like Google. And, you know, the classic case of this is one of the most powerful AI companies in the world is DeepMind. It was originally developed here. It sold out for a few hundred million dollars to Google sort of 10 or so years ago. It's now worth tens and it's probably worth more than a hundred billion dollars.

16:55And it's all of that benefit is being reaped or the lion's share of that benefit is being reaped by Google in America. And it would have been so much better for the UK if it could have remained a British owned company. So what this council says in terms of the recommendations it's put to government, a number of things. You keep using this word procurement. One of the things that it would like to see, for example, we started with how much the UK government is going to be now buying from AI companies to develop this data in your pocket. You know, government services in your pocket approach that you want.

17:31Government becoming AI driven across the piece. how much of what you buy to essentially take the government into the AI age will be channeled towards our own AI companies. Because if it's channeled towards our own AI companies, these companies will get the revenues that will help them to grow way faster. And it will encourage other investors, perhaps UK investors, to come in. So will you buy British? because what people will be scared about is, you know, you will be buying all of this stuff from Google or, you know, from the American giants. There's so much that's going to need to be procured and also developed that there is going to be a lot of this sort of activity going on and we want as much of it as possible to come to British companies, I'm aware.

18:25But will Georgia Gould, who's in charge of all of this in the cabinet office, has she got in the back of her mind, if I've got a choice, I'm going to buy from a British company and I'm not going to buy from OpenAI or I'm not going to buy from Google. Well, look, so we're talking about many different companies here. I mean, AI, it is frontier side. OpenAI is one, Anthropic, and the other big companies as well that we know about. They are spending$100 billion on each new model that they release. and there is no British company right now that can invest that kind of money in one model release. I'm just going to interrupt just briefly there because one of the things that was slightly jaw-dropping recently is there's a Chinese company called DeepSeek which recently released a generative AI large language model service that it says it developed on a budget of just a few million dollars and it is showing the way that you can develop these generative AI models investing way less than the open AIs do and on all the performance measures this deep seek is performing exceptionally well.

19:36So I'm not going to I'm not going to get involved in that because you know I don't test these models myself and I'm not the person that's an arbiter of the power and the capabilities of these models but what I'm saying is to finish the sentence I started AI is different things At its very, very frontier side, it is taking enormous amounts of investment, enormous amounts of energy and power in order to train these models. But then there is the implementation and the deployment side, which sometimes accesses the models to take parts of them to use it to develop tools and services and products. That is something that we in Britain are getting very, very good at.

20:11That is something that I am very, very keen we partner. And indeed, we are partnering and we want to do more. You know, Rachel Rees, she does stand on the platform of Securonomics. We want to make sure that we can buy British, that we can invest in Britain. We're also setting up the AI sovereign unit that's going to be in number 10, reporting to the prime minister and myself jointly. That is to make sure that we can bring together the circumstances. All of the things I see in Silicon Valley, whether it's investment or it's skills or the ability to bring together all these things, it's magic in Silicon Valley.

20:42We have the constituent parts here. It's just we don't bring it together in a concerted way and make it add up to more than some of its parts. We do have a problem with capital in this country. There are too few companies that can do the kind of venture capital when you come to the series two and three and beyond. Big money that takes you from being a$25 million company to$100 million or a billion. To use the great cliche, the unicorn status where you're worth more than a billion dollars. Indeed. But also sometimes that will rest in these days on a business case where profitability isn't for two or three years down the road.

21:18So it takes a big, bold investor with deep pockets to stick with that kind of timeline. So how are we going to, I mean, how is the Chancellor, because this is something she's looking at, going to create the risk-taking capacity among British funds? So it's not just risk, it's about access to capital. And one of the big areas we've identified is the pensions. I know you've talked about this on this podcast before. But in the US, just by having an extra few percentage points of US pension funds investing in spin-outs, start-ups and scaling up, high-end and in the tech sector, is putting enormous amounts of resource into Silicon Valley.

22:02now we don't have that culture here and we're trying to inspire that kind of culture here it doesn't mean taking more you know unnecessary risk with pensions because at the moment it's a lose-lose because pension funds are not performing to the degree that they are in other parts of the world and our domestic businesses are not benefiting from it i've met now so many entrepreneurs that have gone to america simply because of the swiftness and the scale of investment that can happen and they care so much about their innovation that they were willing to move, they're willing to travel around the world just so that their innovation can get the good start that it needs.

22:38And we want them to stay here, don't we? And we are desperate. No one is more desperate than that than I am. But I don't blame people when they do do this because British entrepreneurs are doing incredibly well. We just need to make sure that we have the circumstances here where they can do really well here. Now, we are working on all these things. We're trying to create the circumstances where these sorts of innovations can succeed here, where frontier AI activity can happen. We're trying and create the demand and the incentives by using data that we have as a government much more wisely, making it available for research purposes once it's safely deployed, once people are reassured about their privacy and anonymised.

23:13We can use government data because we have the welfare state that should attract the very best in the world to come here. We're not just going to sell this stuff off. We're not going to just send it off around the world. We want researchers to come here, base themselves here, and have access to some of the best researchers, some of the greatest skills, universities with real long culture of scientific endeavour, and having some of the best quality data in the world for which to train models and develop some of the incredible pioneering solutions we need to the big challenges we face in medical science, in chemical engineering, in biomechanics, in all these things that are high growth, but also providing solutions to the big challenges we have as people going forward.

23:57This will transform individual people's lives. And I'm hoping in the coming sort of days and weeks, not months, that I can start explaining to the public how this will directly benefit them as citizens, as individuals, as people who are using our public services. And that's something I'm really, really excited about because we can't just expect them to think this is all about government and it's all about a few people making a lot of money at the top. This is about transforming lives. For people like me who have neurological barriers to learning and had a difficult time at school, if I had had some of the tools that AI is providing, I would have developed a love of learning much, much earlier in life.

24:36And I could have been in the same classroom as somebody with a real talent that could get the challenge they need in the same classroom coupled with a great teacher. And this is the kind of richness that it's bringing to people's lives and parents will be far more iteratively involved in students, their children's progress through education. And I think all of this could be enabled by brilliant technology underpinned by AI. And are you as a minister using a chatbot? I do. No, I use ChatGBT. I use ChatGBT. On a daily basis? On an hourly basis? Not an hourly basis. No, I use it a few times a week.

25:10Look, AI chatbots are great tutors. They're really good tutors. What do you use it for? Like, what was the last thing you put into chat, Pop? The last thing I used it for... Was it, how do I do a podcast with Robert Pestin? No, but I have in the past said, when you just test it, you just sort of, because sometimes it comes back with answers and you think, that's so good. And you want to just test it more and more. One day I was just, I was putting into it, I was asking questions because in this job I meet scientists and I want to go into those conversations, not starting from scratch. I want to try and learn as much as possible.