Can Britain lead the world in driverless cars?

Some people say the UK has already lost the race in artificial intelligence and that may be true in respect of the large language frontier models being developed by the likes of anthropic, Google, OpenAI and the Chinese giants. But when it comes to what are called world AI models and specific AI tasks, UK companies are right up there and one of the best examples is Wave, a genuine world leader in self-driving car technology. It's an £8 billion British success story founded by Alex Kendall, what I wanted to know from Alex, therefore, is whether the UK is a good place to develop a cutting-edge business. What are the pitfalls? Does government help or hinder? And, of course, is AI becoming an insurgent life form that may wipe us all out? We're very proud to say that the rest is money, is partnered with Octopus Energy this year. Octopus Energy's founder and CEO Greg Jackson is with us. Greg, I've got a question for you. Is the UK's move to renewable power speeding up to think or slowing down? Yeah, there's a whole load of new projects recently been licensed. There's others on the construction, so our renewable percentage is not just forecast to grow, that equipment is going in the ground as we speak. So we're going to have more and more renewable on the grid. Broadly, that should be a good thing. You know, with a gas power station, you have to pay for the gas that goes in with wind and with solar, you don't pay for the resource. But we need to overall our system so people get the benefit of that cheap power. It's crazy that today, we're going to spend maybe £3 billion over the next year, turning off wind farms on windy days and paying gas to replace it. We should be changing our electricity system and putting that electricity to work. Thank you very much, Greg. You'll be hearing more from Greg on the rest is money this year, but for now, let's get on with the episode. Observe the North American dad, 50s, great polo collection. He enters his vehicle information on Carvana and offers appears in minutes. Remarkable. He squints, reads it again, smiles and says, dang, that's a good offer. The highest form of approval, no haggling, no hassle, no one called him boss. He's still on the couch, astonishing. Sell your car today on Carvana.com. Pick up these may apply. When you need to build up your team to handle the growing chaos at work, use Indeed Sponsored Jobs. It gives your job post the boost it needs to be seen and helps reach people with the right skills, certifications and more. Spend less time searching and more time actually interviewing candidates who check all your boxes. Listeners of this show will get a $75 sponsor job credit at Indeed.com/podcast. That's Indeed.com/podcast, terms and conditions apply. Need a hiring hero? This is a job for Indeed Sponsored Jobs. Alex, thank you so much for joining us this morning. There's a ton of stuff I need to discuss with you. Essentially what wave is, what it's like being an AI leader in the UK. Essentially, how you see your own future. But I just want to start with tell us what is different about waves, self-driving technology from, say, a waymo, the business owned by Alphabet Google. You might see on the streets of London both of our cars driving. I started wave about a decade ago out of the University of Cambridge. At the time, we were working on research that showed how to build intelligent machines that can make their own decisions. I think we'd agree that the future of robotics are going to be systems that can operate in the messy and unstructured worlds that we live in. They can go for driving. They can drive the places they've never been before. They can understand the constantly changing environment. You see that in London, there's about 10 times more road works here compared to San Francisco where some of these self-driving cars started. There's a need for this level of intelligence and adaptability. That's exactly what we set out to build it wave. There's a next generation approach to autonomy. One that doesn't require a lot of infrastructure, doesn't require expensive sensors on the car or high-definition maps that tell the cars where to drive, but instead has a single AI deep learning model that can essentially understand the world for itself, make its own decisions safely and drive you wherever you want to go. That's what we're built. To be clear though, you are focusing on what would traditionally be called, I suppose, the software part of all of this. You don't build cars. Yeah, that's right. We've built the software so that it can go on any vehicle all around the world. We've got partnerships with some of the largest automakers to bring this out into consumer cars, where they build the car. You might choose to buy one, and this is companies like Mercedes, Nissan, Stellantis, just to give you an idea. This is companies in Europe, Japan, North America, from luxury down to even this week, we're unveiling a Fiat 500, electric Fiat 500 with a supervised autonomy system on it that we've built. We licensed that software to automakers around the world, and then we're also bringing it to other applications like autonomous robot taxis, where in this sense, we also licensed the software. Automakers will build robot taxis, and then it's owned and operated typically by partners like Uber, who we've just launched our first supervised service with in London. Yes, I wanted to talk to you about that pilot with Uber. I'm absolutely fascinated by how your technology learns. I understand it. Your AI is what would be called a world model, as opposed to the kind of models that many, probably most now, of our listeners would be familiar with the Claude's and the various sort of open AI and Google services, the large language models. How does a world model, how does your kind of world model differ from a large language model? How does it learn? Let me demystify a bit about the differences. Yes, so of course, language models we've seen, or now omnimodels, have seen go through an incredible rise in the last couple of years from when they started, as a next token prediction system that they could predict the next word and complete your sentence. To now, with our cable of rich reasoning, early stages of memory, and quite complex tasks through chain of thought. The difference with physical AI applications like driving is that firstly, you're deploying these systems into a state of critical environment, so it's not just about recall, but this really is where precision matters, where you've got to be accurate to the point where you can go through hundreds and millions of miles in a way that we are reliable and trusted without the open-endedness of the world that we live in. They also need to be able to run in an embedded environment. I mean, the kind of system that we're building that I mentioned before into a Fiat 500E. I mean, this is a system that has to be able to be in mass market cars. It's hundreds of dollars of hardware, tiny, tiny computers, just camera, radar sensors, not these large tens of thousands of dollar server GPUs that you could typically run language models in. So there's the safety career aspect, there's the embedded quality of these models, and then of course, they have to be able to understand the physical world and understand geometry and reason in that way. So there are some differences, but then there's a lot of shared capabilities. Of course, they also exhibit scaling laws in physical AI. We see these things scale with more data and compute. And what we've built is an engine, a flywheel, where we're learning from all of the deployments that we have, whether it's a supervised robot taxi today or the consumer vehicles we're deploying around the world. All of that experience improves our model when we can ride that scaling law to higher levels of safety and performance of our system. Of course, they need to generalize, they need to be able to deal with things you've never seen before. We see that, for example, last year, we went and talked to go and prove we could do this. And we did this road trip around over 500 cities throughout Europe, Japan, North America, US, and Canada. And we showed that our system could drive at that kind of global scale, where most of those places would never actually been to before. So we see these kind of exceptional levels of performance. Specifically, when we talk about world models, it's important to think about the different AI capabilities you need to build a self-driving vehicle. Crucially, I've described how WAVE has built a contrarian position. We've taken an approach to a learning approach to the self-driving problem, as well as our business model being different, in the fact that we want to license this to scale. But with world models, we see these as being useful, not just for driving, but also for simulation, evaluation, and reward models. So we're aiming to build this underlying model that can understand the world. A world model is essentially a model that can take your current state, your current position, the action you want to do, and protect how the world unfolds. And this is incredibly useful, because it can enable you to create a world action model, one that can actually go drive the car, or a simulator, which we call Gaia, which can actually simulate and replay and help you understand it. stand long tail or rare safety events that may not typically occur in the world. So you can go and test and make sure your system learns from that. Do you remember you asked about learning? Do you remember the system AlphaGo was very inspiring when we started Wave, built by deep knowledge researchers at DeepMight? Yeah, so the system famously learned through self-plate. We've played itself in a simulator. Now the game of God is very easy to simulate. Very hard to play but very easy to simulate because it's a compared to what is the size of the game board? It's I mean, chess is eight by eight squares. For example, these kind of games, whereas self-driving, you end up with 100 million different numbers coming into the system at once with all the pixels around the cameras around the car. And so the simulation is a lot harder, but you can start to build in similar techniques of self-play, of adversarial learning to be able to train these systems. But of course, you need much larger, more generative world models to be able to train them. So we learn from all of that experience around the world of supervised following expert driving signal, as well as looking at all of the weird and wonderful and sometimes unfortunately dangerous driving we see on the road today. And we learn from that and learn to build a safe and compliant system. So what you sort of as I understand it having common with a large language model is in the basic form. You know, a large language model learns the probability that if there's a series of words, what the following word will be. And you know, people always use this cliche. You know, if you say the cat in the large language model, we'll probably assume you're going to say the cat in the hat, right? In your case, and your system has to sort of predict. So you've got a bicycle coming up on the right hand side. Your system has to sort of predict what that bicycle is likely to do next. So that the both the cyclist and indeed the car don't get into trouble. Now, in the case of Waymo was I understand it. They start with absorbing maps of the cities that they're in and then they drive around and they absorb a ton of amount of real information from literally driving around but reinforced by the the mapping information. In the case of Tesla, by understanding is they are absorbing tons and tons of data from all the Teslas on the road. Now, you've talked to me about your amazing adventures in 50 cities and obviously you've got the what's going on with Uber now. But you you have got a sort of almost philosophically different approach which is you actually simulate, don't you, you know, essentially millions and millions and millions of different events that could happen to a vehicle on the road. Is that right? Is am I right in thinking that one of the things that differentiates you is that essentially the simulation that you do on an enormous scale? That's right. I think we bring together the best of both worlds between Tesla and Waymo, like what you described. And of course, this is a system that we're licensing for anyone around the world. But the beauty of this is and I think we can be, you know, I think my team can be quite quite proud about what we've built because as you described, we've started with much smaller computing and data resources compared to Tesla and Waymo. Yet today we have a system which our customers have benchmarked and to find it can perform competitively or even outperform what we see from Tesla around the world in terms of a driver assistance technology. And we're making great progress towards bringing this into a driverless setting as well. So it's a much more data efficient system that we built to learn and a lot of that comes from the way we use this data and the way that we use synthetic data. Now, maybe the one misconception is that synthetic data can replace real world data. That's not quite true in my opinion because the best simulators, the best synthetic data day come from generative models that actually these need to train on real world data as well. So you can think of synthetic data like recombining and increasing your learning from a, you know, from a data point you have from the real world. I mean, we do this ourselves. So for example, our hippocampus in our brain is very good at replaying memories. Like if you're learning to hit a ball with a tennis racket, you're not going to try every single permutation of the angles of the racket could hit the ball. You're going to do it once and then you're going to, in your hippocampus, replay this and think about and learn and solidify your memories of how different angles might hit and then learn to hit it better. We can do a similar thing where we might experience driving through an intersection once, but there's many different things that could happen of different ways, different actors could be in that scene. And so we can use generative models to go and replay that in different ways to get a lot stronger learning signal out of that example. And what that results in is, of course, the level of performance we have with still very limited data. And as I described before, we're in the journey of just writing up that scaling curve now, as we start to onboard hundreds of gigabytes of data from our partners around the world. What is your, I suppose, competitive advantage, given that, you know, in the case of Waymo, you would argue that sort of alphabet Google has got almost unlimited resources and in the case of Tesla, you might argue, they got almost unlimited data. How do you compete against what these sort of giants? Look, I tend to prefer to let our results see the talking. And when I started this business 10 years ago, that was the same time when a bunch of different self-driving companies, many of whom today are now, you know, have now ceased business, proclaimed that this will be solved in a year or two, that there'd be, there have billions of dollars of funding into them, and I raised $1 million. So your seat code was $1 million? 1.5 to be exact. But yeah, we started to build this technology, and I think there's something to be said about focus and innovation, and also scarcity driving innovation, to be able to build a system that's next generation that's more efficient than the ones previous. And we see across these different metrics, cost, scalability, safety. This is an approach that can leap us forward from where we are today. Now, specifically, as we're in this, there's some great companies in China as well, but in the Western world, this wave, waymo, Tesla, you know, competition today. I think Tesla has shown remarkably that consumers love the full self-driving technology. There's now over a million people subscribing for this. It's generating Tesla, a billion and a half of revenue a year, recurring revenue. So it's a product that consumers love and are willing to pay for. But the thing is, not everyone wants to buy a Tesla. Tesla's 1% of the car market. And so we've got a much larger opportunity to be on, you know, 100 million plus vehicles across the rest of the market. And then, you know, compared to waymo, that technology, that generation of technology, is not affordable today to put into consumer cars. And so we've got a much larger opportunity through there. And then the benefits of our approach can lead to a much more effective robotaxi offering, because you don't need that high-definition map. It's a more human-like driver that can operate with less infrastructure and a more scalable vehicle. So that's the kind of future that we're building. But I'd rather let our results see the talking. We've got a ton of hard work ahead. But what I'm excited to see is this AI roll out to benefit road safety and consumer experience for transportation around the world in the coming years. And since you mentioned China, you know, obviously in the automobile manufacturing space, there's enormous concern, particularly in Europe, that, you know, in effect Chinese manufacturers are subsidized and have an unfair advantage. I mean, there is a lot of sort of self-driving in China. As I understand it, millions and millions of miles of self-driving has taken place in China. Are you worried as and when self-driving, you know, Chinese self-driving vehicles are licensed in the West just because they're cost so much lower they will have an unfair advantage? I think we need a balanced answer here. I mean, firstly, I think there's a lot we can learn from what's happening in the Chinese market. There'll be much faster at diffusing this technology through the consumer vehicle market. And that's showing that consumers, once they learn about and are exposed to this technology, and think about, I don't know, have you been in a self-driving car robot? I do know what, it's a great shame I haven't ever. So maybe we should organize that when this podcast is finished. Come for a ride to the Sonundan. You're most welcome. I'd love to. But I mean, okay, so not not many people have had the chance to experience it yet. In China, this, you know, many more people have, and you can now see in consumer surveys, it's a top two reason why people choose to buy a certain car. And so it's really important for consumers. And I think the same, it's just a matter of time for that to be true in the West. The technology just needs to be available, which is what way there's going to make possible. So there's a lot we can learn. And of course, I really love the pace of innovation there that's pushing Western automakers to compete. And I think that's the key word. It's we need to compete. Where we need to be careful is to make sure it is a living level playing field as you describe. But I think there is some very thoughtful discussion about how to ensure that markets can protect for that. But I want to compete. And what we found is that through recent competitions that partners have done and others that we can't yet talk about, that they benchmark our technology against what they can get from some of the leading autonomy companies in China. And what we find consistently is that our system is not only safer, more performant, better at generalizing and driving in cities around, or countries around in the Western world. And of course, we can provide much more secure and compliant regulatory compliance as well as data protection around what we build with. So I think all in all, whether it comes down to commercial or technical or trust, I think we've got a offering that can compete. And I want to make sure that we do it and it's chosen on its merits. And so I think there's a balance approach that's needed here. No, you mentioned I think 10 years ago that you started with, was it one and a half million dollars was the original funding. On your latest fundraising earlier this year, I think your valuation was over eight, eight billion dollars. Is that still roughly what the business is worth, do you think? That was our last fundraiser. That's right. We're now set at the start of this year. In terms of your sort of cash needs, do you have enough now to get you to where you need to go in a commercial sense? We do. I mean, we're fortunate to be well financed. I think we've got escape velocity at the stage. And this has been hard for, I mean, a deep tech journey is not not easy to go through. But getting to the point where we've built a proof of concept and we've got customers in a path to distribute this technology. And now the financing to take us all the way through. That's where we are today, which is some critical opportunity. And you have got some big investors and partners in the shape of Microsoft, for example, and Nvidia. What role do they play? Yeah, her and besters include some of the best venture funds like Eclipse Ventures, Boulderton, Softbank, some of the best automotive and transportation partners like Uber, Mercedes, Nissan, Stellantis, AMD, ARM, Qualcomm. And then, of course, some of the biggest funds in the world, like Capital Group, Ontario teachers, D1, there's a whole spread there. Specifically, Microsoft and Nvidia, they've been very important partners to ours. And so what I found building wave is that building this out has, of course, taken a lot of time, taken a lot of investment before now. We're bringing the product to market. And finding strategic partners has been a crucial part of that journey. And I'm very grateful to Microsoft and Nvidia that have both backed us since the very beginning of the company, both with their products as well as with our financial investment to enable us to build this. Now, today we can turn this into a win-win where today we're consuming and co-developing a frontier physical AI cloud with Microsoft Azure. And of course, within Nvidia, we're deploying our AI and vehicles around the world, accelerated by their GPUs. Of course, the GPUs we also use in the cloud, it's AI is really an Nvidia pod today. So all in all, these are now turning into big strategic partnerships. We're together. We're accelerating each other's businesses. But 10 years ago, that wasn't the case. And I think the ability to build these kind of relationships and help for them to help us grow was very important to us as a growing company. Is the UK, in your experience, a good place to develop your kind of cutting-edge technology operation? It's one of the best in the world. And I mean, I've had the fortune in the last couple of years to spend a lot of time around the world selling this technology. And what I've been exposed there, as I think, particularly London, but the UK is a really fantastic environment for doing this. There's not many places in the world where you could do what wave has been able to pull off so far in London's one of those few that you could probably count on one hand. What are the advantages of being in London? Specifically, I mean, everything with this comes down to talent, comes down to the people. And London has been a great place to get started. I came out of the university system here from Cambridge. And the tip of research talent has been very positive for us to build from here. Then of course comes the availability of capital. And then as you think about, for further growth, it's about today, I see it. One, it's the regulatory environment for us about to deploy this. But then secondly, it's the brand and the reputation of the UK on the global scene for us to be able to help influence global automotive regulation to accelerate the deployment of our technology. Because a year ago, it wasn't legal. Now it is. And a lot of that has been down to our collaboration with the UK government and their ability to shape UN level regulations. And then secondly, to be a trusted trading partner and for us to be able to form strong European partnerships with European OEMs, for us to be able to have an open way to collaborate with American OEMs. And of course, the great relationship we have with countries like Japan. So I think in general, that's been a very positive effect for us. It gives your job posts the boost it needs to be seen, and helps reach people with the right skills, certifications, and more. Need a hiring hero? This episode is brought to you by Google Chrome. You think you know a browser, but Gemini and Chrome, that's new. It can help you with practically anything on the web, like restoring a vintage motorcycle from a 50-page restoration block, or finally break down that long article you've had open for weeks. Gemini and Chrome is here for it. Ready to make anything online makes sense? There's no place like Chrome. Check responses set up require compatibility and availability varies 18 plus. This episode is brought to you by Palm Olive. Family time isn't just the big moments. It's weeknight dinners, sitting around the table, everyone talking all at once. So when the plates are empty and the sink is full, use Palm Olive Ultra. Palm Olive's most powerful formula removes up to 99.9% of grease, leaving your dishes sparkling clean. And the new convenient pump makes cleaning even easier, so you can spend less time tackling dishes and more time together. Shop now at Palm Olive.com. I was struck and I should have noticed this at the time. However, the in terms of what you might call sort of conventional pension fund money, you've got Ontario teachers in. And it is striking. And with something we've talked about quite a lot on this podcast, when it comes to scale up finance, businesses like yours will get money from confident Canadian pension funds, sometimes from confident Australian pension funds, very, very rarely from British pension funds. And one of the things we are concerned about here on this podcast is how great British businesses cannot get finance at the right moment from British institutions. Is that a concern that you would share? A hundred percent. A couple of things to qualify this, though, with the first one is that I think the strength of the global capital markets and also the respect that global capital markets have for the UK means that great businesses can get funded here. And I don't think it's an impediment. I think it's just a mess that the funding doesn't come from those institutions. But we have been able to find strong funding from around the world and I see other great businesses get funded here. But of course, having local financing can help improve that because you're exposed to those funds more, you spend more time with them. There's a local ecosystem. So I think it's super important to be able to develop this. I think right now where we are, it's in an abysmal state for growth and pension funding in the UK. If I look at maybe just to also qualify that, we've had some great funding from UK funds, namely First Minute Capital and our first seed round, Olderton and our Series A and Bailey Gifford and our Series B. So we have had some great funding, but it's a fraction of what we've raised, which has mostly been international. Specifically, I look at maybe Bailey Gifford aside when I look at the pension funds in the UK. I just see compared to what I've been exposed to in Canada. I just see a lack of talent, a lack of ambition, conservatism. And I think that this is, I know there's a lot of rhetoric about opening this up, but I just see zero evidence on the ground. I don't see any proactive engagement. I don't see any examples of funding here or even. When I think about the kind of money that I want to take, it's not just the money, but it's the type of investor, the partner, the person I get to work with, how they are to work with, and then all of the other aspects that come with it, that aren't financial. These kinds of things need to grow as well. And what I think the Canadian model is a good one. At least I've had good experience with it so far. And I would love to see the UK lean more into this. It's a huge untapped opportunity. for growth. It's an untapped opportunity in so many different ways and in particular, British pension savers, it seems to me, he should be able to have exposure investments in businesses like yours that could become world leaders. One of the statistics that I cite, probably boring are listeners to tears about this, but in 2000, the UK had, depending on how you measured it, but four or five of the world's top 50 businesses, on most measures, we now have zero of the world's top 50 businesses, depending on the market cap or one or two oil and pharmaceutical companies, occasionally we have one. And one of the things that is just very striking at the moment, you started by saying, what a great place this is to create a potentially world-leading business, but too often what happens, particularly because of the absence of scale-up finance, and because so many businesses don't want to list on the London Stock Exchange, that when they get to sort of, in a sense, absolute liftoff, too many of these businesses, I mean, deep mind is the one that is cited frequently, sell out to an American giant or list in America in the way that ARM has done, which is a long rumble to say, how committed are you to staying British? I mean, we've had ample opportunity to go different routes with the company with what we've built, but what I am excited about is to go turn Wave into a generational company, one that can actually bring these benefits of AI to the physical world, starting with vehicles, and then much, much more broadly at the time. That's an awesome opportunity. I don't want to sell out with the company I want to keep growing. So just to be clear, because that's quite important. So your ambition is to remain an independent company, rather than selling out to one of these giants, as it were. So if there is going to be an exit, would you consider listing on a Stock Exchange, or is it too early to evaluate that? I don't want to wait into that debate because we've got work to do to get our product shipped and scaled first. So that's where I'm focused. But what I was going to say is, I feel like I get so many questions from the media, like the one you asked in the UK about, "Hey, are we going to stay or remain British?" And if you look at those top 50 companies that you cited, I mean, how many of them are just in one country? I think it's so important to be able to be successful in a space like we're in to be international. And what I've taken wave in the last two years, particularly, is from one that's really just been in London. So now, we've got offices in over six countries around the world, huge partnerships in Germany, Japan, and the United States. And we're really globalizing the company. I think that's something that we should be proud of and we should lean into, because that's what's required to be able to win in such a massive market, like physical AI. And also, to get the soul around the world, you've got to work with the diversity of talent and customer opportunities around the world. And so I think this is something we should really lean into and not try and hold on to things locally here. And what about the risk climate here? You have got a license to do this pilot with Uber, which is your technology, but there is a sort of human in the car just in case something goes wrong. Do the regulators, does the government have enough of a risk appetite for this to be a good place to take your business to a level of commercialization? I think so. And part of what I described is we're not just going to be launching here in the UK. And that's important because if you're just in a single market, there are risk factors that are out of your control. And so I mean, think about the three biggest taxier ride hailing markets in the world, that London, Tokyo, and New York. And we're going to be the only self-driving company operating in two of them by the end of the year when we bring our service into Tokyo as well. So we're excited to globalize this. But what I do see in London is it's been a great journey. We've been operating on British roads since 2018. So, you know, a long history of safe and prosperous development. And the journey we've been over through in 2024, we got law changed right before Parliament dissolved. And we got the automated vehicles at Broadham that legalizes this. We've had tons of work with the vehicle certification agency, TFT, TFL, like all these different departments. And I think we're at the point now where there is you know, cautious optimism and excitement about the technology. I'm really happy with the relationship we have with Transport for London today. And I think we have a, you know, great path to go and mature this technology here in the UK. We want to do it responsibly. So we're starting with relatively small number of vehicles. We're starting with safety operators. And we're helping, you know, us and Transport for London learn how to make a great service here in the UK. And then as we grow that, of course, I expect autonomy will be something that London is love. And they already are from our early trials, our early services. We've done thousands of, of now, public rides. We see a fantastic response. And when will you have enough data for you to be confident and the regulators to be confident that the human safety operators don't have to be in the vehicles? Good question. I don't want to put a specific data on this because I think that could create some, you know, adverse high pull pressure. We've got a very clear internal target we're marching towards. But I can tell you the steps that we need to do. So the first one is we need to move from our retrofit vehicles to vehicles that are built by an OEM ready to be scalable for our driverless service. And that's underway. That's underway. Is it the building of those vehicles? We're now with Nissan and Stellantis. We're building these. And so Nissan leaf robot taxis are already on the road in Tokyo and we've actually just received our first couple here in the UK and in the US. So we're globalizing that platform with Nissan. Really excited about that. So there's moving to that platform, which we'll transition to shortly. Secondly, there's on that platform proving the level of safety through on-road and simulated miles. And then third is the approval to launch from regulators through a nationally led process, the automated vehicle scheme, as well as, of course, TFL support for London or whichever city we're in in the UK. So those are the three steps. The vehicle, the safety performance and the certification, and then, of course, we'll enjoy the driverless service. I'm not going to ask you for a deadline. But in terms of when we're talking in your view months, till we see a proper driverless service in the UK and London or years, just a flavor of it, really, I suppose, would be what I'd love. Well, I can share that in the next year or so, we're going to start rolling out and consumer vehicles. And so those are some committed dates we have throughout 2028, many of our different automotive partners are rolling this technology out at global scale and consumer vehicles. And that's for supervised autonomous products. We'll grow that into unsupervised autonomous products. And then we're going to be bringing our supervised robot taxi service, starting in London to Tokyo later this year, and then to many more cities around the world next year with our partners Uber. And then as for when it's not a binary question, what you described, I mean, in terms of taking those from supervised unsupervised services, this will grow depending on markets, local regulation, and things like this, but specifically for London, again, I'm not going to give you a specific date, I'm afraid, but it's going to grow, I think, quickly. And these are big ambitions. It's just to confirm what you said earlier. You genuinely don't need additional capital to hit those ambitions, or to meet those ambitions. No, I mean, we've raised 2.8 billion in cash. So that's the cash we need to go and get these first products launched. Of course, our ambition grows a lot beyond this. We just announced our Wave Labs this year, which is a really exciting expansion where we're finding that the world model we've built is not just great for driving, but also highly effective at other physical AI tasks, like manufacturing and manipulation and robotics. There's many more opportunities that, of course, I expect our company will keep looking for venture and scale up financing to go expand in. We want to really become the intelligence layer for all of robotics. And specifically for the UK, I think that physical AI is a huge opportunity, right? You talked about competitiveness with Chinese manufacturing and things like this. You know, one of the ways we're going to compete when here in the UK is with physical AI and innovations like this that I think can completely grow the industrial base can enable more adaptable, innovative products to be built in a much more efficient way. And so these kind of these kind of use cases, I'm very excited to see the technology grow and to further. I think in general, we're going to see mobility before manipulation. So not just driving, but, you know, we're house mobility, delivery, trucking, these kind of applications before you see manipulation, like assembly and manufacturing and probably B2B before B2C applications. So your, your manufacturing and industrial plant before consumer robotics. But over time, I think AI is going to be so beneficial to the physical world, just like we're seeing in the information world today. Can I ask you on that? This is a sort of broadened bit. Obviously, we've all been gripped by the kind of warnings that Dario Amade has been making about extinction risk from whether it's his agent's going rogue. There are a variety of different ways of looking at this extinction risk or artificial intelligence, recursive self-improving intelligence, simply breaking free from whatever shackles and anthropical. Open AI, I put on them. Your model, a world model, starts from a very different place. But there are many thinkers and developers in this area who basically say that if we are ever to get artificial intelligence they'll have to be some kind of fusion between large language models and your kind of world model. Is that how you see it? There's a lot to unpack there. I think we are going to see models become more unified over time. I mean, we're already seeing that in 2021. We put the first vision language action model on the road that was able to understand language to see for itself and to drive a car. And this was quite interesting. It opens up a lot more product opportunities like you can start to backsy drive your autonomy and say, "Hey, I want you to drive this way or go here or go there." Or ask, "Why are you doing that?" And I think that can help build a lot more trust and customization in what you delegate to the physical AI. But in general, we're going to see more and more. Because more these models become the more breadth of things that the better they get. And therefore, you're sort of your model is sort of converging in a way with the sort of word-based models in that sense. It is. But then you brought up the probability of doom and risk around these models. I think we need to draw a very careful distinction here. Because I think there is an important debate to be had at the front here. But when it comes to sector-specific deployment of AI models, for example, in healthcare or education or driving, I think we can actually very accurately understand the opportunities and risks today. The already great regulators are already set up for those. And driving insurers understand that they can economically quantify the risk of driving very accurately today. Regulators can put in place a great set of technical regulations for this. And the opportunity is very clearly understood in terms of horrendous road safety record that we have. So I think in these kind of sectors, it's very important to as quickly as we can to fuse the benefits of AI in a responsible way. But in the frontier bit of your research, presumably actually recursion, self-improving is something that you're working on. Oh, 100%. I mean, we have early signs or different aspects of recursive learning already today. I mentioned the example of learning and simulation. I mean, it's a bit of a buzzword. I wouldn't use it because of that. But there is recursive learning happening there where AI is in our simulator, experiences it, improves from it. And then we use the data to simulate more. And that's their adversarial process continues to improve the capabilities of it. But when we deploy these systems in self-driving cars, you know, they have clear constraints that they can't go and do something that's not driving. They're only there to drive. They have clear cybersecurity boundaries that are well formed. And so a lot of those risks that I talked about are very well contained when you're deploying in a specific sector-specific application. And so all I'm trying to say is that we should have that frontier debate. We should think carefully about what the right regulation is. But clearly, there's a huge opportunity there that we can't take away from the world. But when it comes to deployment of the AI and sectors where it can benefit in a meaningful way now, we should move very quickly on and not hold that up. No, I understand that. Obviously there's a whole debate, which is not specific to you around. You know, as we roll out, there will be some implications of your rollout from this. It does have an impact on what jobs humans do. You know, there will be, I imagine, quite a lot of taxi drivers who are terrified of what your technology means for them. And one of the things that actually does frustrate me is how little proper debate there is about this within political parties and within governments. And you know, all the things we talk about along this podcast is, you know, we should be debating how you change welfare systems, how you change education systems. And so on, because, you know, this is an incredibly powerful industrial and social revolution. And, you know, as a society, we have to adapt so that everybody benefits rather than everybody suffers. Actually, I will just ask you specifically on that. I assume, you know, politicians do talk to you about what does this mean for taxi drivers. What do you say in response to that? Well, firstly, I think it's important to zoom out and look at the benefits of the technology can bring. Is there going to be job disruption and displacement? You know, yes, but I think it's important to recognize that. But to zoom out and really focus on the benefits to be able to create that transition. I mean, of course, the road toll was just inexcusable when over 99% of those accidents are due to human error. The AI we build doesn't get distracted, drunk or impaired. It can make over a dozen decisions a second, seeing all around it. So clearly, there's the opportunity to improve this and make it much safer driving experience than what we see today. At least 100X improvement, if not more over time. But then you also look at, say, the congestion in cities or you look at in London, we have car parks that take up 10 times the size of a wide park. That with autonomy, you know, right now cars are utilized about 3% of the time. Why do they need to set idle so much? You know, we can utilize vehicles and make this much more efficient and turn these car parks into green parks. Think about speed bumps or traffic lights where if cars could talk to each other or be more efficient, you could have a more efficient mobility network. Or even think about how much of the policing resource is used on road policing. And if cars actually followed safe guidelines and how they should drive, how could we use that to improve other aspects of policing in the country? I could go on. There's a really long list. But I think the imperative is this technology is needed for so many people. And I'm excited about bringing that out to people around the country. I mean, my point is not that there aren't going to be technological benefits. It is simply that we have to make sure that this works for people. And one of the problems in previous industrial revolution is just that governments did far, far, far too little to retrain people, give people welfare support as they retrained. Two little support, two little attempts to support living standards for people whose wages may have been suppressed. And there's just a ton of, but my view is not that one should stop technological rollout or technological progress. See, these just societies, and there's an onus on governments to look to start planning for these changes and protect people. I agree, Robert. And I don't want to overstate our position on this. I mean, I think there's a big responsibility of governments through this. And look, how can I and how can wave support that? I mean, I want to make sure that we have a very open dialogue of the risks and progress of the technology. So that policies can be brought in that are calibrated to what is happening and what is going to become possible in the near future. We have that through frequent dialogue with regulators and government, as well as with our met with leadership of the tax unions. And I think these dialogue is very important. I think also making sure we bring this out in a responsible and incremental way. So that there's the ability to work through feedback and make sure this is sustainable and safe. These are the kinds of things that we're trying to think about. And to help enable governments to best navigate the challenge you describe, which I think is important. And we're almost out of time. I've got one other thing I don't have to ask you because it is such a sort of compelling part of the current AI debate. I don't know if you've seen most of us, Solomon, who would have found us a deep mind talking about anthropics approach to Claude, which is effectively the treat Claude as a sentient being. Not in a metaphorical sense, but actually in the way that it trains Claude. And I mean, most of us very concerned about about this. For all sorts of both practical and philosophical reasons, when you think about your own AI, is it a category error to think of AI as a new life form? Or should we essentially, because there are a lot of people who just say we are creating a silicon life form. What's your view on? This stuff is open to a very important debate here, but I think it's important to be technically accurate here. So, firstly, most people define consciousness as the perception that something is human-like. And so, sure, I mean, some people say that our AI drives cars in a human-like way, but I don't think anyone is going to claim that a self-driving AI is human-like and its capabilities. So, I think, again, before deploying this technology in a specific application, where we can bound us capabilities and see the benefits, I think that's very important. But we're not aiming to build conscious or what you might call a sentient broader things. We're aiming to deploy AI that can benefit physical driving and manufacturing. But an anthropic arguably is trying to build essentially a sentient life form. Well, yeah, we're trying to build more intelligence into this, and I think that I don't think you have to, well, let's talk about some of the challenges here, which is, it really comes down to the issue of alignment. And so, the issue of alignment means that a classic example given in literature is a few paperclip problem, a few task and AI to build paperclips. If it runs out of resources in a factory, it might go start destroying other things to go and build more paperclips, right? So, you have to be very careful with how you design objective functions of these AI's so that they will work in a way that you want them to towards that. Now, there is a lot of thought and effort going into how to go and do that, and the theoretical example I gave you is sort of far from the reality of where these systems are today. And I think the important thing is we need to stay in an equilibrium. There's never going to be no risk, but I also think that risk can be managed through very effective engineering and research on this. And what I'm seeing today is a lot of good effort in the industry of how to develop these systems in a way where you can test and measure their likelihood to do these things. Of course, it's never going to be perfect, and we've seen recent examples like with the Hugging Face Incidents or what happened in Australia with Australia Health Data. These are some examples of some escapes, but hopefully they're being talked about and we're learning from them. So I think we need to keep developing, minimizing risk of this technology, but if there is it, ever an escape learning from it. And that kind of equilibrium, I believe, will bring us to an outcome where I'm optimistic that it will result in huge benefits from the technology. And so I think this is an important debate to have. I think it's important that we continue to be open, and all and all just have very strong engineering towards these systems to be able to mitigate the risks that I think I talked about here. One of the things that concern me in recent years because, you know, I've been sort of obsessed with all this stuff for years, was how little public debate there was and thank goodness now a light is being shown because it is really important that we manage these risks. Alex, listen, it's been an absolute treat talking to you. Thank you so much. We're going to have to arrange for me to get into one of these, I need to see one of your systems close up. Oh, yes. Listen, let's do it. You've got to come for a drive. I mean, what's your take on autonomy? How do you think Britain is going to see this in the coming years? You will be aware of the, in government, opinion is split, you know, there is a very active debate going on between some ministers and indeed officials who broadly want to see Britain at the cutting edge of both owning and rolling out technologies of this support because they see it essentially as an important part of the growth agenda. And then there's another constituency that is, I mean, first of all, you know, in White Hall, there is a very risk of earth culture. And so, you know, one of the things I imagine you will struggle against. I don't know how you're finding this yourself at the moment, but even if you can produce data that shows, you know, per 100 million miles accidents and deaths are lower with your kind of system than with humans, the first time somebody does get killed, there will be a moral panic. I've always found this very interesting, humans struggle with the concept of risk because on the one hand, you can show the dreams of data showing that a particular form of transport is way safer than something else. And yet, you know, instinctively people don't feel it that way. And so, you know, you will find yourself and I'm sure you are finding yourself up against both those who recognize, you know, essentially revolution is coming and we've got to embrace it versus those who are going to worry about how we protect both human lives and a job sense, but also this whole issue of, I suppose, the political risk around approving a technology and then somebody gets injured. I disagree and agree with that. I disagree in the fact that I've seen actually very broad support through government and opposition parties for this technology, through the conversations that we're having. And I think that's really encouraging about the direction that things need to go. But I do agree that there's a risk adverse culture. And I guess it's what I'd say, the good news is that startups are designed to find ways to innovate and thrive against all odds. And within this risk adverse culture, you know, wave is found a way to pioneer this technology and to work within the political system to be able to have the opportunity to launch this out here. And we're not stopping here. And so what I'm excited for is to find a way here to keep growing this technology and bring the benefits out. And I think we have that opportunity in front of us. Yeah, look, we, Stephanie and I are sadly definitely not here today, but we would passionately believe that there is no progress unless you challenge the status quo. So, you know, keep challenging Alex and thank you so much for joining us today. I thought it was an absolute fascinating, absolutely fascinating conversation. That's it for this edition of The Rest is Money. Goodbye from me. This week, it pays extra to be a MyLos Rewards member. Through October 14, members get $10 MyLos money. When you spend $200 or more on an eligible purchase during MyLos money days, also using your MyLos Rewards credit card, even better, that's an additional $20 MyLos money. Use it to save on your next Loes haul. Exclusions, restrictions, and more terms apply. See Loes.com/MyLosMoneyDays for details. MyLosMoney expires 30 days after issuance, subject to change. Find a doctor is hard enough, finding the right doctor, even harder. Someone who takes your insurance has an available appointment when you need one and feels like someone you can actually open up to. That's where Zock.com's in. Download the Zock.com app to search and compare doctors by specialty, insurance, and availability. Read real-patient reviews, find the right fit for you, and book instantly, because getting seen by the right doctor matters. Zock.com ready to be seen, download the Zock.com app today.