 Of course, I had to ask the, so it's a real robo race, so just a robotic car is racing each other, and I said, well, everybody would want to see the humans race the robots, but unfortunately the robo vehicles can operate their wheels independently, which is something a human is not currently enabled to do in the cockpit of a car. And so that represents, what, a two second per lap advantage. That's right, yep. So forget about humans racing robots at least today. Yeah. Again, please take it away, Brent. Excellent. Thank you very much. Thanks for staying for this afternoon. So you may have seen our car is just outside. I'll give you a little bit of context as to why the car's here and what we're all about. I think when we look at motorsport, motorsport has always played a role in showcasing new technology and developing new technology. The picture you see here from 1894 was a first race. That was a road race from Paris to Rouen. And I like to talk about that because actually it proves two things. It proves that the technology could run that distance that actually a combustion car could go that far, but it does so in the real environment. So it actually has to, has to do that through the real traffic situations. And that builds trust in the public. When you can see the technology perform in the way that you imagine it deployed, it builds trust. And then this is where I want to show you where we've got to with motorsport at the moment. This was in Macau a few years ago, showcasing the pinnacle of technology for automakers. Doesn't look as though there's major damage to the car, but we'll keep you informed as to whether... Oh no! Oh! Mayhem! Not necessarily what we want to be communicating as a community. If you look, we're still using flags within motorsport to indicate when there's a car coming. I've ever seen here. You can see the first car after this car crashes, the first car around. Every other car, not so much, but every other car is proceeding, expecting the road ahead to be clear. And that's why I think communications is really important. So if you want to see beyond your field of view, the vision that you have, whether you're a human, whether you're an autonomous vehicle with lidars or radars, if you want to see around corners, connectivity becomes really important. And that's why we're here today. Just also want to show you a little bit, just to show you that the car's real. I'll skip over this video, but we are building robots, intelligent machines, robots that like to drive, robots that want to race. We actually have two vehicles. So the vehicle at the top is Devbot. We have a robot car here on display. But just to show you that the car's actually real, it's not just a rendering or it's not just a mock-up. This was at the Goodwood Festival of Speed in the summer last year. It's a famous hill climb event. It was to celebrate their 25th anniversary. The car itself is driving fully autonomously in this environment. It's in front of, I think, 60,000 people. We did a demonstration every day. Every time the car got to the top of the hill, all of the crowd cheered and clapped. Which is crazy. I don't know who they were clacking for. There's no driver in there. Maybe it's our achievement. But I think it is something unique. So Goodwood, you celebrate all of the motorsport technology from the very beginning. Obviously, this was the first time they'd seen an autonomous car. We were the first to complete that climb. The car itself is fitted with five LiDAR sensors. It has two at the front, two like wing mirrors, one at the rear. It's got two LiDAR sensors, one at the front, one at the rear. Six machine vision cameras, three at the front, two like wing mirrors, one at the rear. It has a GPS inertial system inside the car, which we can use for highly accurate positioning. So combined, we can get down to 70-liter level accuracy on the positioning that we have. But really the competition that we're setting now is a software competition. It's a championship of intelligence. So we provide the standard hardware platform. Teams compete by writing the software that drives the car. If you write better software, you will drive better. And that's really the focus of our competition. It's all about software development. So we'll pick up on that. Very good. Now we'll find out who gets their name pulled out of the hat with the next presentation up. Okay. It's Max. It's got to be Max. Max? Yeah. And introduce yourself, Max, please. Yes, I can draw this. I can't see it. You should cut the mic. You can bend it, Max. Yeah. Like that or? Thank you. Good afternoon. Max Cavazzini from Amazon Web Services. I'm running the Automotive for Europe, Middle East and Africa at AWS. Very briefly, I mean, if you think about Amazon as a company, you recognize that we have been using AI for years. I mean, it's now more than 20 years that we use AI and machine learning, especially to run business and to create new experiences. You see Amazon Go, which is our stores where you just go pick up your stuff, go away without any cashier, without any queue. Just pick up your stuff and go, which is mixing AI, sensor fusion, like an autonomous car is doing. You have Alexa, which is using AI to do voice interaction. You have Kiva robots that are using AI to autonomously decide where to go and to a fulfillment center. And then more, I mean, primary drones, self-flying electric drones. So AI, it's in our DNA, but what does AWS stands in automotive? And the reason why I'm here, and I'm working in the automotive space, I have an automotive background, it's because what we see is that basically OEMs need to become software companies. And as AWS, we are enabling them to build their own solution and to leverage technology to make sure that not only autonomous, now it's more about connected car and electric and autonomous driving testing. It's sure. So we created an entire platform to cover the flow on vehicle development. Ingesting data, petabyte of data that are created for cars running. How do you collect them? How do you clean them? How do you move them to the cloud to allow high-performance computing? We announced with Toyota, an example, they are running the entire high-performance computing. So they're autonomous driving model in the cloud to go 40% faster. Mobileye, it's working with us. Some of these guys are working with us. So the ability to provide the right blocks is what is making AWS automotive business right now. One example, and then I leave to the other panelist. This is too simple, level four autonomous driving trucks. So they're using us to build their system. It's L4, again, it's not L2, not L5, the target today. And they are running their simulations in the cloud. But for the day, this is real-time, so real-time analytics of video, real-time tagging, and working also overnight with, I would say, quite a good precision. So the software allows to speed up development. Thank you. Thank you, Max. I think Anne, you're up. No, I see. Okay, so my presentation was not autonomous enough and didn't reach there, but I will do it without no worry. So my name is Anne Milano. I'm the co-founder and VP operation of Best Smile. We are a Swiss company located in Lausanne, a bit further under the lake. And we started back in 2012 dealing stuff with autonomous vehicles because we were involved in some European projects demonstrating autonomous shuttles in European cities. And at that time, we received six vehicles on the campus of the Swiss Federal Institute of Technology, and we had to do something with them. And we realized quite early that even if these were only prototypes and very first type of autonomous vehicles, there was something missing. The full industry was just focusing on how to make the best autonomous vehicles with embedded hardware, embedded software. But the real question that arise at that time is, okay, if tomorrow, if today, you have the perfect autonomous vehicles here and we have to do something with it, what will we do? What is the goal of these autonomous vehicles? Do we just want to continue having one vehicle per people and to just replace our individual vehicle by individual autonomous ones? Is this really a solution? Is this the way we want to use the technology? And we believe that autonomous vehicles will bring much more flexibility and really help the shift from the ownership economy to the shared economy. So we need, and when I say we, it's not us, it's mobility service providers need to be able to manage autonomous vehicles in real time remotely to bring some real services to the users. And another complexity on top of that is that what I just said will never happen like this. The transition phase toward autonomous vehicles will be very long. And even if some people here can tell you it will be in one year, two year or 25 years, we need to do something with this transition period. We need to give the tools to mobility service operators to be able to manage together today's conventional vehicles and tomorrow autonomous ones. So what we do at Best Mile is that we develop a cloud software, a B2B software for mobility service operators. It's a fleet orchestration software. If I had to summarize it in a few words, our goal is just to send the right mission to the right vehicle at the right time. If you remove the drivers, you need to take a lot of decisions in real time to be sure that your full fleet is optimized. So this is what we do. We have been involved in projects with different types of autonomous vehicles since 2014. We are also deploying systems with normal human-driven vehicles, but with people who have the ambition to transition towards autonomous vehicles. And we enable any type of anti-man mobility. So from ride-hailing and microtransit to autonomous shutters and robot taxi. That's it. Thank you. And I do find it very interesting that here at this show where it's such a super car show and a car enthusiast event, the Geneva Motor Show, the city is actively discouraging you from bringing your car to Geneva. So enjoy the public transit and introduce yourself, Philippe. Thank you. Thank you, Roger. I think thanks for being here. It's a pleasure. And my name is Philippe Heismans and I'm a VP of growth at RideCell. At RideCell, we power some of the highest utilization of mobility services in the world. Services run by operators like BMW and about 20 operators in car sharing and ride-sharing services. And today I'm gonna talk to you about sort of our vision of, quickly talk to you about our vision of the path to autonomy and the progress needed to get there. And so for mobility as a, sorry, it's going the wrong way here. For mobility as a service to truly happen, these three elements need to come in play. And you could argue shared and electric are already here. And the big question is autonomous. Obviously shared, if you look in Switzerland, car sharing service like mobility is already functioning very well and since a long time. And all the ride-sharing services like Uber, Lyft, Diddy and Ola. So the shared has happened. Electric is really starting to happen now. Electric is really starting to happen now in a meaningful way with EVs starting to come into the market. And the big question is autonomy. We know it will happen, but when in five years, in 10, under what circumstances and do we just wait? Well, fortunately, we don't have to wait for the future to build it. And a path to that future is today operating fleets. So the shared element, because everything you do when you operate a fleet, which is learning about maintenance, learning about marketing, this will be invaluable for leading autonomous services tomorrow. That's also why you're seeing a lot of interest, maybe on the sidelines of the show, but all the OEMs are very much looking into the next page of autonomous. And what we obviously say is autonomous fleets are the future because the question is, will it be individual autonomous cars or large shared electric fleets? We think it will be more of the latter. And talking about electric, obviously EV fleets is the next step. Electric is a future for many environmental reasons. Reduce pollution in cities if electricity is produced in a greener way. Reduce pollution worldwide. That's obviously a target for all of us. And electric vehicles also love the idea of fleets or they're a perfect pairing. And why do I say that? Because there's obviously a bit higher cost initially for acquiring an electric vehicle and the cost of maintenance is low. So if you look at fleet operators by hundreds of vehicles, they could bring the cost, the acquisition cost down and the operation cost is really what matters for them in the long run. So again, electric and if you think about it autonomous where the cars are going to be even more expensive with all the all the radars needed to drive are also going to be very well adapted and suited to fleet management in the future. And so what we really see is the two biggest costs today of operating fleet are maintenance and that includes cleaning, recharging the electric cars, maintenance and repair and repositioning the cars. Because if you're doing a car sharing service, for instance it's very important to reposition the cars from cold zones to hot zones where people can actually take them. That's about a quarter of the cost. And yet another quarter by the way similar magnitude is marketing. Obviously you have to invest a lot in marketing to get new users to your service. So these are two very important things that you learn and what do we think as the next step is autonomous? We think autonomous will happen step by step and we think it's gonna be without rides in the beginning because I just explained that maintenance is super important. There's a lot of use cases that we see for autonomous cars repositioning themselves, autonomous cars going to the recharging station by themselves, et cetera. And why is this important? Well, for two reasons. One, is it biggest hurdle for running autonomy will probably be legal and insurance because people will say sort of a chicken and egg problem there's not enough proof that this is safe to have sort of a machine driving a human and it's completely unknown, right? So the best way is to have autonomous vehicles drive through their charging station between 2 a.m. And this is a near-term use cases that we see is basically low speed, urban areas, level four autonomy and no passengers. So the typical example is in a city like San Francisco between 2 a.m. and 6 a.m. you could reposition your cars for pick up the next morning to go to the cleaning station and the refueling station. And this is a great way to lead the path to autonomy because you basically have the safety track record that's being built and the people getting used to autonomy on their streets without drivers and then these cars switched to being driven afterwards. We actually do that at Ride Cell. Real quick about Ride Cell we also have an autonomous division in-house we acquired Auro late 2017 and we work on these near-term autonomy use cases and we have a public permits to do public testing in California. We are also better known for basically the ride sharing and car sharing platforms that we offer to operators and fleet management and maybe just a quick last word about Ride Cell as a company, if you wanna know Ride Cell where does the name come from? It came from the early beginnings where it was order a ride on your cell phone. So that's basically the genesis in the ride sharing space and it was founded in 2009, headquartered in San Francisco with the presence in four European countries quickly growing here. Also in Asia, we now have 150 employees. We just finished our Series B round of 60 million with a lot of great investors that you can see here a call for our European investors would be Deutsche Bahn, BMW, BNP Paribas and Munich Rea and also large scale customers like AAA, BMW, Ferro Vial and Renault. So this is basically in a nutshell what we do and happy to take questions afterwards. Excited to be in this space, thanks. Okay. So Alain, I think it's your turn and now Alain, please introduce yourself and Alain is threatening to have a provocative slide for us here today and I think there's plenty of opportunity for provocation because I think what you're seeing in these presentations and at this event is what I like to call some very aggressive radical creative destruction in the automotive industry. We are completely threatening our existing business models with all of these new technologies. I'll try to just make a couple of quick comments on the professor at Princeton. Been there, this is my 47th year on the faculty so I know nothing else except academics. I don't like the name, I don't like level one, two, three, four, five. I call them smart driving cars. There's a lot of confusion. There are only three kinds. There might only be two. There are safe driving cars. So if you talk about safety, it's safe driving cars you're talking about, not any of the other stuff. There's the Teslas of this world which are the self-driving cars which give us the opportunity to take our hands off the wheel and feed off the brakes sometimes and maybe use our cell phones. So therefore we're gonna buy them. And then the third one is a driverless. And those guys are mobility machines. And I had a couple of other slides. What I hate is safe driving cars. This is a speedometer in my car. 160 miles an hour. Are they kidding out here? I mean, really, where can I do 160 miles an hour in Jersey? I mean, this is like crazy. And so, okay, sure there's the Autobahn over here but what about the rest of us in this world? Cars should be limited to, you know, speed kills anyway. This is a safe driving car. It allows us to misbehave when we're down there but yet it has all the gizmos and all the tools and lidars and whatevers to keep us from running into the shelves. Okay, we can do the donuts, we can misbehave. I mean, everybody out here on this floor is trying to get us to misbehave and we're just trying to get to someplace. So if we're really talking about safety, let's put those things there but keep us from misbehaving because that's what hurts us. When is this gonna happen? Safety hasn't sold, why hasn't safety sold? Why? Because all the stuff that we put in the cars to make them safe and not kill us ended up costing more. The airbags, even the seatbelts, the crush zones all cost more. Hey, if we don't go bang anymore, guess what? The cost goes down. If that cost goes down, then the expected present, that present value or the expected liability is less than the cost of the technology. And if that happens, then there's an insurance company that will decide to make money off of the losses and savings of those losses. And it becomes the Amazon of this business and gets us to all buy it. That hasn't happened yet. We know the Tesla's been out here. They're not giving it for free to give us that comfort, convenience and so on. And auto companies love selling that. They're gonna stay in business. That business model's going to stay. As I like to call it, it's the new chrome and fins of the automotive industry. Then there's the driverless. And the driverless is really, has no steering wheel, has no pedals and it is a mobility machine. That's what it is. And really what it is is, it's not that we're going to own it. I think all these things are a real embarrassment. They're an elitist tool that somehow we're gonna go in there and sip our cocktails as we're doing that. And the amount of quality of life improvements that we're gonna get out of those things compared to already we have a Bentley convertible. I mean, T's are excellent. However, to a whole other population, those things are that mobility machine instead of serving maybe five trips a day serves 50. And there are a lot of folks out there that could really use it. We heard some of it before and that's where in the heck that should be used. And where that's going to be used is in a business model that goes out there and really provides mobility to the mobility, disadvantage and basically makes money off of that. And that's not selling these things in dealerships. It's really creating companies that in fact use these as fleets and fleet operators to provide that mobility. So those are the important folks and I really do wanna emphasize that the mobility disadvantage, they're the young, they're the old, they're the ones that are physically disabled but they're the poor. And there's nobody out here talking about providing mobility to the poor. And if you look at the poor, they've been left behind to transit services, don't serve them. And that's here in Switzerland as well as in Princeton, New Jersey as well as all over the United States. Mass transit serves 4% of the trips. It's an embarrassment. And whereas these could go out and serve the population which is probably representative in the United States of 15% of the households and that's where the value proposition is. Thank you. Good luck. Okay, Julien. Sorry, you had to hold back so much, Alain. Hey, it's a discussion. Absolutely. Okay, good afternoon. My name is Julien Masson. I'm in charge of business development and sales for CloudCar. I've got five slides, right? So not a 30 minute speech. I promise. We, CloudCar, CloudCar was founded in 2011. So it was a startup. We like to say we are now a scale up. And we were based in the Silicon Valley. We are growing our execution in China. And our mission is to, we deliver cloud-based infotainment services. But our mission is to deliver seamless, relevant and personalized digital experience for vehicle occupants. While really enabling OEMs, car manufacturers, to stay in control and not give away everything to the big ones of the software world today. So if you look at the current in vehicle-connected services approaches that we've seen in the markets in a couple of years, a lot of the cars still on the floor today are showing individual apps on the infotainment systems. And so it was the early approach to application integration into the vehicle cockpit. And there are goods and bads around it. And one of the bads typically is that, okay, a vehicle is not a smartphone on wheel and you don't necessarily want to touch a user interface and end up into a full-screen HMI to deal with Spotify, Deezer. And some of those apps are actually depreciated and OEMs have not done great in maintaining those solutions. What has definitely rolled out in the market is the second approach where the end user is, they are bringing their own ecosystem in the car with Android Auto Projected mode, with Apple CarPlay, some other solutions in China. So the benefits of those are definitely to, yeah, the end user is bringing his own ecosystem in the car cockpit. The big inconvenience is that you're, I don't know if you are using those in your car, I am. And the challenge is that you're always swapping from one world to the other. With Apple CarPlay, you can do some stuff and then you go back on the built-in infotainment system. And for the OEM, the downside of that is that there's a few opportunities or touch points for the OEM to talk to the end user, to the vehicle occupants. So what we are focusing on at CloudCar is to, we've done a heavy work on aggregating a lot of content and service providers, which are relevant for the in-cocpit experience. And we've tried to really abstract these content and services from the HMI. So we decoupled those service providers from the HMI itself so that those services are all cloud-based, you can deactivate them remotely from the cloud and the OEMs can still manage their own HMI and the whole service experience so that they can give access to a lot of the big ecosystems of this world, but they can still be part of that game and not give away everything to Google. I mentioned it, sorry about that. So yeah, we are a white label platform and our aim is really to give some control to the OEM of this digital ecosystem. We activate, deactivate services and content from the cloud without massive over-the-air software updates on this infotainment system. We can manage the regional difference from the cloud without 54 variants at this OEM group to manage the whole world in terms of software variants. There are opportunities for service monetization. It's not, I mean, it's hard. It's not easy, but there are opportunities and we enable those opportunities. We store, we create profiles for the vehicle occupants and we store them GDPR compliant in all cloud platforms so that you can actually find back your profile into that second card that you're gonna use with whatever mobility as a service program. We are agnostic to the natural language understanding technology used in the car. This is a challenge. We need to talk about that together, but that's really our aim and we work today with the big ones with nuance, with some specific ones in China and we are not developing our own natural language understanding and wouldn't make sense, but we need to cooperate and work with the big ones. And obviously all of that makes sense only if we have the ability to use vehicle data because how to deliver a personalized user experience into that car, you need to know and predict what's going on for that car, where the car is heading to, how many occupants are sitting into that car. And so those vehicle data are very useful to build a context and have the software running in the cloud predicting what's going on to provide a personalized experience. So how we do that, we do the hard and not funny job of keeping interfaces with multiple service providers in multimedia domain, in location domain, in productivity domains, up to date. And we abstract those content and service providers so that you don't need to have 10 different media streaming application, but you can have one media player and all this application are available in one experience, one domain. The same for places, the same for productivity. And then the true added value comes when you start building this profile of an end user, you know that Roger loves Mexican food. And well, you don't expect, if Roger is hungry driving, what car are you driving? Nice BMW, lucky you. So you say you're hungry. Well, we should know that you love Mexican food when you are, and we're gonna make recommendations for you, which are Mexican restaurants. We don't want you to, I'm looking for a Mexican restaurant with a five star rating from treat advisor. So that's really, and the way we do that is we try to record your preferences with your consent. We load the user behavior on the infotainment system so that when you skip Madonna, we're not gonna propose you to listen too much to Madonna. And we use the location of the vehicle, we use vehicle data to know what's going on inside that car, obviously date and time of the day. And also we use the content providers that we've aggregated, all your historical preferences from those content providers. And then the goal is obviously to make predictive recommendations, to build predictive discovery, to predict where you're heading to, and to make, this sounds scary. Yes, you need to agree for that. And then we make those answers available, but then the key is how you make the user experience into the car relevant. That means not annoying, not like giving you 10 recommendations every minute, and not intrusive. That's tricky. So that's what we're focusing on. Very good, thank you, Julia. Okay. Holger? So I suspect Holger will talk about his vision of de-appification in the car? Yeah, very good. Okay. Wow. Okay. I want to tell you what is happening if CloudCar is disconnected? Well, we call Max and he speaks in that. That is something we're talking all the day about the connected car. We still have a situation where a car is belly in a connectivity, or in a strong connectivity constantly. Good. So I'm Holger, Vice-CEO and founder of Drum Auto Labs. And if this goes well here. Here, give me a second. Excellent. Good. Very good. Thank you. I wanted to talk a little bit about what Roger initially said. We see two elements of AI, or let's say general elements here. One is more outside focusing. And that is something that... You like that better? Good. Outside focusing. And that is what we heard in the speeches before. Mainly it's about autonomous driving, lighter, radar, et cetera. Understanding is this a cat or a bicycle or a shade of a tree or something like that. And I wanted to talk a little bit about more what's happening inside the car. And we also could describe this a little bit by those terms that we all love in the industry. There's this case term, right? This connectivity, autonomous sharing and electrification. That is where a lot of development and evolution is happening right now. But there's also something fundamentally different or a change that is happening in the automotive environment inside the car in the HMI, the human machine interface. So what I like to do if I try to understand how far the future will bring us, I like to look over to Hollywood. Hollywood is actually a better indicator as you can imagine for predicting technology. There is a methodology that was developed in Hollywood over since the 50s by the script writer Guild to predict a technology in a meaningful and a senseful manner. We know that Starship Enterprise, a lot of the 60, 70 series technology, we find in our reality, right? The communicator, we call it the smartphone these days, et cetera, and others as well. So what this guy actually is doing, he's flying a very autonomous thing, his ship, right? And the interface that he's using to communicate with the ship is his voice. He's just talking to a ship and say, hey computer, where's Leutnant Uhura? Or if we look to this other movie here, her, very beautiful four years ago, playing in the far future, you do not know exactly. This a little bit shy gentleman here who's in his life, a love letter writer, author, falls in love with the operating system of his computer because this is so natural, talking to him and feeling like him, et cetera. The voice is by Scarlett Johansson, maybe that is one of the reasons why that is happening, but she's omnipresent, he has a little button in his ear. But when we're talking about our industry, all of that was foreseen as we know, of course, in the 80s, right? Michael Knight, the Knight Rider, buddy, I need a kid, I need you buddy, like that, right? Turning the car into a companion, into someone who understands me, who knows how I'm feeling, and especially who's talking to me and proactively taking care of me. So that is something that we see rising at the moment. We spoke a little bit about this before here. Recently with this, I would say the second race rise of voice assistance. So voice assistance or let's say voice interfaces are not a new thing. We all have these experience with in-car navigation systems that were voice enabled, right? It's quite a frustrating type of experience. The reason for that was that in the first 40 years of that technology, the level of understanding, right? The level of how much the machine understood of what you said to it was that about 60, 65% at the end. And that is, some people say that's the average in a good, with a good couple. And that is why you stop talking to each other because you just don't understand. That's why we stopped using those navigation systems. But over the past five, six years, we got a breakthrough here. We are now at 97% that has a lot to do with cloud computing, with AI also, with these gentlemen over there with the infrastructures like AWS allowing us that. So meaning that computers understand today on the human level, which is something that we all feel when we start to use and started to use those systems like Siri and Alexa and so on. And this is the technical reason, but there's also another reason that technologies need this tipping point that enable them to really being used and adopted. And that's something that we also feel that there is a whole generation growing up now who just will talk to all things, right? I mean, there will be a very new interface. So now when we're thinking about where this is going to happen, for a long time, it was very clear it's going to happen in the connected home. So we want to say, hey, Alexa, switch on the lights or Siri, do I have milk left in the fridge or whatever we want to say or open the garage door, something like that. So recently, all of these hypothesis were a little bit cooled down by the understanding that it's still easier to just stand up and switch on the lights or taking the remote control, et cetera. So while, and that is not a surprise, in a car, it makes way more sense because, I mean, we heard here, not only in your presentation, that for a very long time, we still will have people steering cars and why you steer a car. You should focus on the road and having the hands on the wheel and the eyes on the road and not on your mobile phone, for example, and replying to your Facebook messages or watching your YouTube videos. But even in autonomous driving environment and that was the example with Jean-Luc Picard from the US as Enterprise and the interface will turn into the car. So what we believe in remote labs and what we are working at, those trends will be dominated by very strong platforms, horizontal platforms, the Alexis of this world and Cortana and Siri and Bixby and Alibaba and so on. So there will be a handful and they will control a lot of our daily lives just by the fact and that was what Roger said that with having an assistant, there will something happening in all of your lives, there will be a de-appification. So because you don't use any more apps and services as you do at the moment, you just say, hey, Alexa, I have to fly tomorrow to Geneva. I don't want to arrive after 10, book me a hotel, not more than 10 minutes away or 10 kilometers away and in the morning I need a shuttle. So you would not really care about where this assistant actually would doing that. So, and that is a fantastic, that's a fantastic thing for Amazon and for you as a user, because they control and you get the service, it's not so much a fantastic thing for the service provider who is somehow outside of that gate. So if we now transfer this into the real life, now we see this horizontal platforms, right? But what would it mean in a car? In a car it would mean that actually there would be one interface and this one interface would not be controlled by the OEM, but by someone else, by an outside company. There is a lot of data, user data that will be involved into that. So long story short, we believe why these horizontal players will dominate that voice assistant game. There will be a need, a very strong need for vertical players. Vertical players meaning the, in our case, the digital co-driver, making a digital assistant, a digital co-driver. And the question is what could this be? And it's very clear. So I came back from Las Vegas, CS, and Alexa meanwhile is able to flush toilets, right? Which is probably in one or the other case a helpful thing, but I don't know if you really want to have someone flushing your toilet, driving your car or being your co-driver in a sandstand, taking care of you, keeping you safe, understanding what's behind the curve, et cetera. So what we're doing here is really focusing on what is the DNA, what's the nature that turns a voice assistant into a co-driver other than a home assistant or a co-worker, et cetera. Those are things like acting proactive. You don't want to ask, for example, a co-driver in rally sport after the curve what you should have known before the curve. That's something that is one example. So to sum this up, we're building a platform, a vertical platform for voice AI in the car that is providing, I like to say, the soul and the brain of this assistant, optimizing for in-car usage and interacting with those horizontal platforms. Thank you. Okay, last but not least, Tommaso, Grossi from TomTom. And thank you for mentioning safety, Holger. Thank you. My name is Tommaso Grossi. I'm a responsible for business development at Europe for TomTom autonomous driving. So one of the questions that... Microphone. One of the questions that these panels aim to address is also where are we in the transition from driver assistance to full autonomy or for full autonomous driving? And well, I just wanted to kind of zoom in on some of the technology improvements that need to happen before we can actually say that we achieved full autonomy. So if you essentially dissect a self-driving car today, these are the key components that essentially comprise a self-driving car. So it's composed of maps, it's composed of sensors, driving policy, which is essentially the software that will essentially tell the vehicle how to drive in a safe manner and obey the traffic rules. And actuators as well. So where are we today in the spectrum from manual driving to full complete autonomy? And the professor will forgive me if I use the different levels of automation before this purpose, I think it's... I know, I haven't been doing this anywhere. So if we look at the different levels as determined by the SAE, we are essentially somewhere here. So between level two and level three. We're at the stage where it's a big transition because going from level two to level three essentially means from a safety perspective that you go from saying the driver always needs to pay attention to the vehicle is on its own or at least in certain situations, the vehicle has to be on its own. And so today, most of the automated vehicles or vehicles that have driver assistance systems are equipped with mostly sensors. But how do you actually make the jump from level two to level three and even level four and five? Well, to kind of explain in simple terms, well OEMs are starting to introduce maps. So digital maps to be used in the automated driving systems. And these maps of course will be used also for level three, four and five. But so why are OEMs essentially throwing maps into this kind of into the pot? Well, for kind of the user twofold. So the first one is if you already have a level two system, for example, you can use maps to improve this system or add extra functionalities or expand the function that you have. And the second use is essentially just taking the function to the next level. So from going to level two to going to level three. And why are these maps needed? So how do these maps help the system essentially and how did the maps help the car drive itself? Well, for starters, they work in all conditions, all weather conditions. They're not affected like, for example, sensors by different lighting conditions or rain, for example. So they're kind of what we call a good weather sensor. Well, they help anticipate the road ahead. So sensors typically have a range of a couple of hundred meters. So that means your car can only see a couple hundred meters ahead. Using a map, well, you can already predict what the road ahead looks like. And finally, well, they reduce the computing power needs of the vehicle because if you drive without a map, it's the same as driving essentially for a human. It's the same as driving in a road that is completely new to you. So imagine driving on a road that you've never driven before. You're more likely to pay attention to everything that is going on on the road. Now think about actually driving your commute. So from home to work, well, that's a road that you've probably done hundreds of times. So you don't really pay attention to the road. You're a bit less, you're a bit more on autopilot, let's say. And this is kind of the equivalent of driving with a map or without a map. So without a map, the sensors, the vehicle always needs to understand everything that's going on as if it discovered it for the first time. With a map, it's as if you had a kind of like a memory of the road. And so what are we at TomTom doing about this? So we've essentially, our heritage is in navigation systems. We are a leading provider of navigation systems to the automotive market. But now our focus is on helping OEMs bring automated vehicles to life. And we do this by building maps for automated vehicles that are tailored to different levels of automation. So of course you're not gonna use the same type of map for a level one or two or four or five. Well, we're building maps that are highly accurate. So we're talking about the symmetry level accuracy, highly attributed. So these maps include, as you can see behind me, all sorts of attributes such as lane markings, traffic signs, lane edges, and lane geometry, and they're light and optimized for in-vehicle usage. So that most of the process actually happens in the cloud. And well, so our goal essentially is to build maps to essentially facilitate OEMs into bringing automation to the next level. We believe that maps are the key to unlock next levels of automation. And our goal is to provide maps and work together with OEMs to build maps that help them bring autonomous driving to market safely, quickly, and scale. Thank you. So I think the common thread we're talking about here is a little bit antithetical to the last panel discussion which was all about cybersecurity and protecting information. All of these solutions want to gather data from vehicle sensors, whether those sensors are facing the driver in the vehicle. And I think it's worth noting that Euro NCAP is laying the groundwork for driver monitoring as sort of the next layer to get the five star rating, as well as gathering data from around the vehicle to enable autonomous operation. The strange thing is that if we are truly successful in this pursuit, we will have fully connected fleets operating autonomously with a lot fewer vehicles on the road serving populations as opposed to trying to give a car to everybody. But I'd like to get the panelists to talk a little bit about how this technology is changing the industry and how we're coming to grips with it. And I think we want to emphasize the positive aspects, but there may be some negative. But, and we could start at the end and then work our way over, but anybody can jump in. Oh, super, is this, yeah, it's working. I think we discussed it yesterday, Roger. I think there's something like 100 billion that has been invested in the autonomous vehicle market for technology that will be used for autonomous vehicles. What I'm really interested in is how that technology actually gets used, gets deployed, but actually comes down to ADAS and then goes all the way down to the lower middle income countries where actually the most road deaths are occurring. And I think that's something that we'll start to see is the pressure on actually deploying that technology for good, effectively, not just for the sort of commercial revenues focused around cities, focused around mobility as a service. So I think that's something that we're gonna see. We spoke yesterday about performance being part of that or safety, vehicle control being part of that. That's where we see motorsport playing a key role. That perspective is near and dear to Alon. And I have an issue with cars. I mean, I just don't think cars should run into things, right? That just doesn't make sense. As a basic consumer product, it shouldn't let me change a lane into another car that's in the other lane that I happen not to see. But that requires, is the only solution a mandate? What is your thinking? Do we have to require these capabilities? Well, no, I don't, man, sure. If we're trying a mandate, go ahead. But we're not, I think, I said it when the insurance industry realizes that the cost of the technology to keep you from doing that, it just sits there just like our interlock breaks. Don't let us apply the brakes in the wrong way. They say, you know, don't push the pedal to the floor, ease up, do that. And as soon as that happens, as soon as that leads to a lower expected liability exposure to the entity, then there's money on the table from insurance to pay for that and make money. And so there's going to be an insurance entity in there that's going to disrupt it. By the way, to your point about cars being able to go too fast, what was the word from Volvo here? Okay, so they're writing off the German market. Why is Volvo that short? He had my speedometer show 180. I mean, I wouldn't know the difference. I mean, you know, if it's dealing with perceptions and dealing with my perception from this, is the answer. Make sure you speak into the microphone, by the way. Okay, by the way, sure. By the way, Philippe, you have an insurance company or a reinsurance company investing in what you're doing. What are their thoughts about this? Well, I'm not going to, no, also I don't know the full extent of their thoughts, but well, I'll ask more generally. Is the insurance industry our friend? Well, I have some opinions on that subject. Is this working? Yeah, so no, but I think we actually, I have a couple of large insurance players and it shows the interest in what's really happening, right? For me, there was a way, the wave of the web, the wave of mobile and now the wave of mobility, solving two world problems, which is the environment pollution with electric and the congestion, which is we spend mindless hours in our car, just waiting for the next car to move forward. And I think that everybody knows that those are the huge problems that we're solving with mobility and all the traditional companies, including the insurance companies, have a lot, the early movers have a lot to gain, they see it as an opportunity and the late movers have a lot to lose because the world is going to move forward and autonomous car is going to be much lower risk and potentially a lot less auto insurance revenue, right? So you can imagine if there's 80% less risk of an accident, there's 80% less car insurance revenue. That's a huge frightening thing. At the same time, they recognize that all these models need to evolve and the first thing that everybody needs is data. So the reason we also have an autonomous division is sort of open data sharing because realize that I said legal cities and insurance companies are going to be with sort of public perception that maybe the three main blockers of moving to autonomous quickly, right? And the insurance model will be, well, sure, I'll ensure your autonomous car and put a very high price sticker on it, right? That's a way of putting a break on the industry. And so what they're doing is basically saying, by working with us, they actually get access to the data quicker and they can make the new models of determining the real risk and basically be the first to offer smart insurance. Well, and you're operating vehicles in a couple of different operating environments, business models and scenarios. So car sharing, ride hailing and sort of geofenced areas. Yeah, yeah. So yeah, we look at it from the three large angles of fleet management, which is car sharing, ride sharing and fleet management. So we basically have many different use cases and large fleets, it represents thousands of cars that basically can provide all that data of the future that the insurance companies will need, but also the city regulators, as I mentioned. I think that that's another key factor. Are you treated any better as an insurance customer since you're connected to your vehicles and you're collecting that data and you're operating a business so you can do the analytics and make a case for wanting and deserving a better rate? Yeah, I mean, that's always like a business discussion. So I think that hopefully with insurance investors as our partners, we'll find the solutions early, but I think that right now, what we're trying to solve is really build the new mobility use cases. And I think that that's what they see in us just so you know where I'm coming from. I always get aggravated with the Insurance Institute for Highway Safety because they always do these studies that show that either the consumers are turning off the ADAS features or that the efficacy of that feature is not reducing claims. And therefore the insurance company should not give a discount for automatic emergency braking or blind spot detection, et cetera, et cetera. They're slowly coming around now, but for years it was no, our studies show it doesn't reduce claims. Therefore you shouldn't get a discount. But that's because it didn't work. You know, the automated emergency braking systems don't work. Don't? They don't work. Okay, because the object ahead is stationary and we assume we can pass underneath it and we can't determine well enough whether or not we can pass underneath it. So the error rate associated with that is such that it's gonna start putting the brakes on and therefore it puts the brakes on and it didn't have to put the brakes on. I take it back and it's a lemon. And so of course turn it off, all right? And so until we can get this stuff to work well enough so that in fact the probability that it can determine whether or not we can pass under a truck that's ahead of us or we can't, which is what happened 10 days ago. Boom, another Joshua Brown. Why? Because the automated emergency braking system doesn't work on a Tesla. They turn it off for stationary objects. Well, don't hold the Tesla example up as a... Yeah, it's 80% of the OEM which are not considering stationary obstacles as real obstacles. It's only when the thing is moving that they adapt to the cruise and the speed. So this is why a lot of this accident happened. It's because you are on a highway. You have one vehicle going on the right lane and the vehicle ahead, what is stationary? The vehicle, do not consider this as an obstacle and just don't break at all. It happened to Elaine Hersberg. But I mean, that is maybe a good, if I may... By all means. Jump in here. So we're talking about AI and the status of AI in automotive and we just heard there are a hundred billion dollars invested into autonomous driving technologies. And I would say there are about 80 startups in Silicon Valley going fast very soon because the hype will be over very soon. So, but it's exactly that. All that will not work without the data. So one mentioned that here before and you buying a company to get that data and to learn about that. And there will be fatalities and cash realities, like in the case of Tesla, there are people dying, right? And mankind could say, okay, you have to do that to not having people dying in the future. But that is something where we also shouldn't be naive with everything what we're seeing at the moment. We are not in the fifth dimension type of thing with robot taxis and autonomous driving car. 98% of what we're seeing has nothing to do with AI. It's machine learning. Machine learning needs to be trained on a very low level and that's something where I think at the moment we are completely as an industry exaggerating about the capabilities that we have. I would say for the next 15, 20 years, people will do almost 100% still steering cars and not driving autonomous, at least not on a mass market type of scale. And yet, right, so and if I could just throw in on that. We may be at peak deep learning in AI. In other words, we've had a big hype since 2012 of all the, hey, I can tell it's the difference between a dog and a cat. But these adversarial attacks, the adversarial piece, the fact that this is a black box by which I can change one pixel on a recognition of numeral zero to nine and have that just by changing the intensity of one pixel have it flipped from probability, essentially one, it'd be a five, to probability, better that it's a nine, all of a sudden flip by one pixel. You can't tell the difference. We don't know what's going on inside of that deep learning neural network. It has gazillion coefficients in there, non-linear functions, it's probably overfitting and darn it, it may not be robust and we may have to go through a dark period again until the next breakthrough is made. So I would say keep your horses, sir. Keep your horses maybe, I don't know. Well, I hope we're not giving up because as a matter of fact, of course, right cell and the best mile are operating today, correct? Yes, we are, but I fully agree with the fact that we don't use AI. And just a small anecdote, one of our investors, one potential investor told us, I don't remember exactly the context, but he told us you don't use the world AI and this is why we are here today because there are so much hype around this and it's just a buzzword that a lot of company and especially startups are using, but behind there is nothing. It's just a bit of machine learning and a lot of power points, but this is not the reality. So yes, I think we have companies like Red Cell and us focusing on trying to get these vehicles on the road today to learn, to get the data, but also to learn like for an operator, what does it mean operationally to have to manage autonomous vehicles? And this is a long learning curve also because it brings a lot of new constraints and a lot of new challenges that they need to face. And we are starting like with these types of shuttles that we have here in Europe. I think we need to consider things that the start of these autonomous vehicles in the US and in Europe, we took two different ways. In the US, the thing was like this, we want to try to reach level five. So let's start with the level zero or one and let's try to go up the curve until level five. And some companies here in Europe said you want something that is like a level five, just remove the steering wheel and the pedals and see what we can do with this. And this is what happened. And at one point, these two strategies are supposed to arrive to the same point and we will learn different things. I think that today in the US, okay, Waymo did millions of miles with the autonomous vehicles, but whoever tested once at CS, perhaps. But besides CS, did you ever went into one of these autonomous vehicles from these big companies? No, but you can go here in Sion in Switzerland and test two autonomous shuttles running in a city center. Okay, it's very low speed. It's just a technical demonstration of what the vehicles are able to do because doing a loop in a small city center, you don't have a lot of transportation meaning behind, but you can still go and see how people react to this. So yes, I think we should start somewhere and just to learn because learning is just key. But to get to my point, and I have a feeling you have something to say about this as well. Yeah, go ahead. Because we're not just trying to understand the operating environment of the vehicle, we're also trying to understand the state of the driver and the intentions of the driver from their voice and gestures and things. Cameras are coming into the car. So it's a broad scope of changing this relationship with the human, the lump of meat behind the steering wheel, if there's going to be a steering wheel. But what are your thoughts? No, I mean, can you hear me? Yeah. I think we are pushing a little bit too hard on a complex problem, trying to solve it tomorrow. If I look back at, I would say, let's say broader Amazon culture, it's very much about thinking big but start small and then scaling fast because the technology allows it. So I think we all agree that in 10, 15, 25, whatever, it's the number, autonomous driving will be there. So I don't think that we can say it's tomorrow. No one probably really knows when, but there are many, many examples where some pieces of technology is like, it's a big puzzle, right? So are going in the right place. So if I look at, let's say, Holger trying to solve a piece of that and looking, you know, also at disconnected car, what we can do today or ride the cell, best mile or tom-tom, I mean, the idea of solving in one day a problem it's impossible, but I would like to reverse a little bit. We have, you know, customers out there that are looking for solutions and technology in this moment, it's not an issue, right? You can easily analyze petabyte of data in hours. You can easily... Well, your TensorFlow is your service, right? No, we support it, you simple, it's a mixed net. We support all main cafe and all main stuff. But again, the technologies there, I think RoboRais is showing, I mean the technology can help. I think we are a little bit forgetting the human side, legislation, liability, as it was mentioned. I saw a nice video from Volvo testing and recording the reaction of drivers when the car is taking over and driving. So I would say that... Well, it depends on whether you saw the first reaction or the third or fourth reaction. Yeah, I mean... The first one is... Yeah, I think we were together in... The subsequent are like, it's okay. Yeah, but this is something we need to work on. So I would really encourage to start from the customer and working back for there, not really thinking about, oh, there's a technology or not. Well, let me ask you in the panel because Ann brought up a key point and it's embodied in GM. So GM has supercruise and I'm fond of saying it's the Roadhouse Blues of level two driving. The doors keep your eyes on the road and your hands upon the wheel. Well, if you're keeping your eyes on the road in a supercruise vehicle, you can take your hands off the wheel. Okay, I don't know how many Jim Morrison fans are here, but anyway, being while they own cruise automation, which is trying to make that great leap directly to automated driving. And I think it's becoming a little more difficult than they originally anticipated. It is a little bit more difficult, but my question is, do we want to have this intermediary step where we don't know exactly who is in charge of the vehicle, the human or the technology? And this is where we have situations like the one with Teslas. I think this is the most dangerous thing to do is to say the vehicle can help you but you are still responsible and this is what the insurance will say in any case, you should have taken it back. So, I don't know. I kind of feel that this mixed situation where the vehicle and the human have to take co-decisions, it's dangerous. Level three, not a big fan. Yeah. If I may there, well- And I'm Swiss-natural, but level three. Well, so I understand the point in the sense of the uncertainty of who's in charge and who's operating the vehicle actually. And I think also the OEMs are now making it easier in terms of their marketing, for example. And there's a bit of over-promising going on there. But at the same time- There's one OEM coming from- Another performance. Leaving names out. But yeah, so in the business model will now work if you just wait to commercialize until you have a level five or level four. And so OEMs need to one, start developing the technology today and hope that it works. But I think the root cause of the problem is there is that the same approach for autonomous driving is being, they have taken the same approach for autonomous driving as they have taken for any other development in the vehicle for the past 20 years. And the problem with autonomous driving is that we're all good at knowing, okay, the cars will all drive themselves. They will not have steering wheels. There will be pods and they will come pick you up. Sure. Everybody has the end goal in mind, but very few know how to actually figure it out. And so the industry nowadays is kind of in, okay, let's shut up and get to work kind of mode. And you don't see many huge announcements anymore as you were. Let's not forget that when they finally are successful, I don't think this is an owned vehicle, which is the ultimate irony to me. But anyway, I don't know what you're going to say, but my thought is, it sort of raises quite, why are we doing this in the first place? It's because the path to getting to here, hopefully we'll introduce safer driving circumstances on the evolution. Yeah, along the way. Yeah, I was picking up on that point about human and AI and whether they can work together as a pair or not. And that's something that's fundamental across all of our lives in every area. And if we really believe that it's either humans on their own or AI on its own, I think we're in a bit of a difficult situation. We definitely have to look at that intersection and where humans and machines can actually work together. So we have two vehicles. We have the RoboCar outside. That's fully autonomous, showing the AI can do this on its own. We can put that in extreme environments that we can't put humans in. So we can race those cars in a full traffic environment with 60 mile an hour trucks, while the car's doing 200 miles an hour. So we can push the technology limits beyond where we're all comfortable driving, just to show that the technology can do to build. Well, you proposed an interesting concept yesterday, which was that the AI infused car could actually train a human how to drive. And that's why we have the second view. The irony being, of course, that the humans presumably are teaching the computers how to drive in the first place. At the moment, yeah, that's what's interesting. So if you have a car that's fully autonomous capable, then when the humans drive in, what do we want that system to be doing? Do we wanna turn it off? Do we wanna turn off this 360 degree awareness? Because it's, I know the humans on their own now. Sorry, we're not gonna help you. I don't think that's sensible to anybody. So what we're starting to look at is how can humans and machines work together as a pair? And if you look at Toyota's positioning really as guardian angel, it's starting the other way around. It is starting from, this is what humans can do really well. This is where they make mistakes and this is how we can use AI to prevent them making the mistakes. That's a much simpler use case. It's much more deployable now and it will save lives immediately. And another piece of AI, at least an application of Alexa, which it doesn't always have to be described as AI necessarily. But I see voice technology being added to ADAS systems to tell the driver what is going on in the car, which is an interesting idea. So the best example that we have is, I tend to say when I was 17, I was learning to drive. I wasn't allowed out on a car on my own. I had to have a trained driver that sat next to me, either a parent that was driving or an instructor. And what's that instructor there to do? Perceive the environment, perceive the risks. They advise me on what to be looking for within that environment. So why when I pass my test, am I considered to be expert enough not to need that person anymore? Especially when I've not driven in every type of operational design domain, as the industry would say. So I can pass my test in the UK, I've never driven on a motorway. Can we imagine autonomous vehicles being approved if they haven't ever driven in that domain? You have a question? If we look at the accident statistics, they absolutely confirm exactly what you're saying. Most accidents occur with new drivers between the ages of 18 to 24, and then bumps up to 27 or so, and then it levels off as people get more experience. We now have vehicles that are out there. You've used the term artificial intelligence. I'm not really sure that any vehicles that are operating today, including yours, is actually using artificial intelligence. I like to use the term that the economist prefers, which is collective intelligence, because the intelligence that's put into robots, whether it's a car or any other type of, is our intelligence. We've put that information at the disposal of this robot. But that's another question. Just... I would agree. I've heard it takes about two years for a human to then build up a safe level of driving. To have seen it enough. But for that two years, they are on their own. And that is the incredibly unsafe part. I'd have to... It's fair to say that there were 300,000 years of evolution before. Right? And then we started learning driving, and then two years is not too long, I think. I mean, of course. But just to add what you said before, I think no one is questioning that, in general, that this technically it's possible. And I always like to say, the technology we have, it's usually not a problem of technology and beautifully demonstrated here and the RoboCar. I saw the NVIDIA is also running one right in Munich. It's phenomenal. And people are really like wow and giving applause. But if you do a RoboCar 200 miles per hour versus a truck 60 miles per hour type of test, you wouldn't do this in the city center of Geneva on a Monday morning at eight o'clock. And I think that is something where we see at the moment that, to a certain extent, we are already beyond the boundaries that are responsible in understanding that this needs time. And that comes very much from the perception that software can solve everything just by this beta type of iterative. We test, we fail, we test, we fail, we test, we fail. And it's simply something very different doing this in a car or with a whatever a Pokemon Go type of game, right? Where you might gonna stumble upon a tree and so we didn't see this because the AR was not so clear. And just to go back to what Anne was saying before, yes, there's a lot of hype around AI and a lot of what we are doing and what the industry is doing is actually collecting the data and training the system and one of the challenges that we collectively face right now is that there are some companies who are better positioned to collect that data and they are the big ones with a B2C offering who can definitely address such a huge amount of people while I mean, you're doing a good job rolling out your product right now. We've got Jaguar and Rover vehicles and user interacting with our technology, but that's nothing compared to the, yeah, the million of end users talking to Google, Apple, Alexa. Yeah, that's a big challenge for us. Yeah, well, besides all the testing that we're doing, there's also the simulation and in fact, the simulations have gotten to be an awful lot better. And so it's really not the number of miles that you end up driving. It's those particular use cases. You know, it's really 150 feet or something like that, or 150 meters in that critical time when things happen that you can then test with simulation. So we're making a lot of progress with that. And then of course, we have Amazon sitting here, okay? Amazon, okay? Amazon, what does Amazon need most? Free shipping. Guess when they might be able to do free shipping between 1 a.m. and 5 a.m. when there's nobody out there and deliver all this stuff to my house? So, well, I know you haven't told us what Amazon is doing in driverless mobility and so on. If we wanna see who's gonna be first out there and who has the real business case to be first out there, Amazon, thank you. I have a quick yes or no or sort of hands up and hands not up question. Everybody here with autonomous vehicles, they're on the road or semi-autonomous. So it's not gonna be safer to say semi-autonomous. Cars with level two kind of automation. Do we wanna see more, so put a hand up if you wanna see more regulation as opposed to less regulation. So more or less? Okay. That's the more. That looks like, yeah, okay, I'll give it to more. All right. Any questions from the audience for the panel on this subject? Shall we just continue? Come on, audience, let's go. Come on, audience. Sure. Cause we just, we're talking about cybersecurity. Now we're talking about autonomous driving. Yeah, it's the shared mobility. I've always had a personal question about how it works in low density rural areas and the business model for it. The surprising thing is with my students, we just ran through several iterations of the billion or so individual trips that we think occur in the United States on a typical day. And in fact, even in North Dakota, the ride-sharing opportunities do exist there such that you could get a basically a, I think it's a 67% reduction in vehicle miles traveled if you did sharing using robot taxis or what I call autonomous taxis. What happens in low, what surprisingly happens in low density areas is it's low density, but guess what? There are a few places to go. So therefore, in fact, the key thing with respect to ride-sharing is the probability that you're coming from about the same place to about the same place at about the same time. And guess what? A lot of it, you know, kids go to school. Maybe you want to shove them into a school bus. You should hate to do that. My goodness, they spend how many hours out there wasting their lives. They could be doing it in much smaller vehicles. So it turns out the opportunity, and it's surprising as hell, we didn't expect it. Yeah, no, I agree with that. We also thought that sort of the mobility as a service was not applicable to outside the major urban areas. And there are actually a lot of use cases, but I think that if you look at the large car-sharing programs today, they're focused on urban areas, obviously, because that's where the biggest or lowest hanging fruit is. That's where the biggest problem of pollution exists. That's where you can have the biggest bang for the buck if you want on running these new programs. You surely shouldn't disqualify running robot taxis in North Dakota, I heard, because there are use cases that it can work. That's not gonna be the priority investment of fleets, for instance, right? Because it's gonna be much more difficult from a complexity perspective. On-demand services, what you have traditionally in rural areas that you have on-demand services, and that's something, I mean, I don't know how this is in the years, but in Germany and Europe, that used to be a popular service that you could call a minibus, and then you waited a bit longer than for the regular bus and then picked you up, or you, so just that the public transport companies could not afford it anymore. With the on-demand semi-autonomous or fully-autonomous driving, you don't need a fleet to go there and pick someone up. I mean, it's one call. No, I mean, the kinds of ride-charing you want are the 2Z's, 3Z's, and that's not enough money to pay for a bus driver, but it certainly has enough money to pay for one of these Waymo vehicles and so on to do that. The other surprising number that I found in the S1 of the Lyft filing is, I think, I don't remember exactly on top of my mind, I think it's 46% of the Lyft rides were picked up or dropped off in low-income areas. And which, someone mentioned that before that. And going back to Hollywood, going back to Hollywood, if you think about, I mean, not moving humans, but moving goods, parcels, primary was born to serve rural areas where it makes no sense to drive. So if you look at airbus... You got the packages there with the little parachutes? Yeah, it's not parachutes, it's a drone. But I mean, let's think also about flying humans because it will happen, maybe not around the corner. There are several tests around the world. And again, if you go to parcels was, I mean, small drones, but think about also flying humans. It will happen. So don't limit yourself to think that everything moving will be on the ground. Any other questions? Oh, I have not a question, but maybe another provocative comment on what you said before. Because there was, I think you said that right sharing is has a good goal is to avoid and reduce congestion. And we talk a lot about what could technically work and what is reality, reality today. And that's my impression. I have no numbers for that. It's that everything that we like so much at the moment where we are so excited about the new mobility is doing anything else but reducing congestion. I mean, if you see the numbers of bikes lying around on Beijing's roads and Berlin's roads, and we have, I think, something like eight car shares in companies now in Berlin, each of those, this company has minimum 500 to 800 cars. Otherwise, it doesn't work. And there's still no car at the airport when I arrive at 10 o'clock in the night and so on and so forth. So I think also coming back to the point that I made before, it's great to test all that and doing that. And I love to use all those different possibilities, but at the moment it doesn't really do any better to our environment or the congestion of the city. Well, yeah. So I see a whole transportation environment that's tearing itself apart. And Uber and Lyft are tearing apart the taxi industry and car sharing. The fact is that in Manhattan there are now over 7,000 more cars ever since Uber started and not less because people still own cars, the taxes are there and there are 7,000 more cars. But then Uber and Lyft are investing in bikes and scooters, which is taking away rides from the Uber and Lyft vehicles. So then all these scooters lying around on... All of which are taking people on public transportation. No, no, so just to react to that, because obviously that's taking a hit at sort of what I presented from a long-term perspective. I fully agree. And in the near term, it's called like the sort of investment curve or the pain curve, that's actually, you actually have higher congestion in the near term because it's the Wild West beginning, right? And for it to actually go down, you're gonna need to have sort of reduced complexity, optimization and just a... It's a similar thing of San Francisco when you were talking about micro mobility. When they opened up the scooters in San Francisco, there were 12 services that put on kick scooters all over the city. And then San Francisco had to blow the whistle. And in the end, they only led two that were working on the best solutions and closest to the cities and governments who ultimately made the call. So they had to clean up that whole Wild West scene. And all the big players, jump from Uber, Lyft, Lionbird, spin, all of those did not get permits. So that's sort of like an organization call after the beginning. And there's numerous example of ride sharing. In the early days, basically making that there's more cars roaming around in the city. They're not electric yet. So the environment, the pollution thing is moving to clean electricity. So that's another discussion. But the congestion, you need to be able to organize these fleets and that won't happen in overnight. I think it will be more disorganized in the beginning. Coming right to you. Oh, well, go ahead. Just one comment about this Uber, Lyft and so on. I think they just have to move of place into the ecosystem because they're business model today what they do peer-to-peer marketplace. If autonomous vehicles would be there, it will disappear. And they have multiple choices. They are investing a lot in this micro-mobility into cities, but they're also moving more and more towards, for example, mass aggregators. Uber CEO is from Expedia. This is the same type of things in another industry. But I think the problem today in cities with this peer-to-peer model, if you consider Uber, for example, for each mile that they drive someone, they also drive an empty mile, just because they are angling empty just to look for some riders. And this is one of the big problem of these services is that today Uber, Lyft or La DD, they pay nothing to the drivers if the vehicles are empty. So their only problem is about recruiting as many drivers as possible so that when you open your application, you have someone in less than five minutes. I don't, they don't care so much about what this human guy is behind his driving wheel. So this is also why we believe more in professional fleets in fleets where vehicles are owned by the operators because it's only when you have a fixed number of assets that you can start optimizing. If you don't know how many vehicles are where when you can't really optimize anything and you are just trying to assign the closest vehicle to the traveler because what they want is to have an immediate solution, otherwise they switch to another type of service. So optimization is linked also to just knowing the stages of your ecosystem. And this is where a change will happen and who will own these vehicles? It may be OEMs, it may be other types of players like for example, rental car companies. They are the best position just to own and maintain fleets. Well, my favorite scenario in that regard is sponsored mobility. The shopping mall sends a vehicle to bring you to the mall. Someone will own the fleet, the question is who? Okay. But today's cities are just realizing that there's peer-to-peer types of services are creating more traffic than they were supposed to reduce the traffic that they make made everything worse. So 10% of decrease in the average speed in New York in the 10 last years. For sure the city is growing, population is growing but also all these systems are growing. So before we run out of time and I may need Elaine Alon's help on this one. We're getting more compute in the car, more storage, faster networks, more sensors. We're fusing the sensors to create new interpretations of the environment, presumably that will enable safer operation of the vehicle. But what I'm hearing from the panel is we don't have AI in the car today. Is that our conclusion of our AI panel that we don't have AI in the car yet today? And if we don't, do we need it? And when will we have it? Well, we have some and maybe, I don't think that's the problem or that's the problem, we're gonna have it. There are enough people that people talk about the amount of money that it's all going to be there. The problem is going to be, I like to call it the welcoming environment for all of this. This is going to need a welcoming environment, not only by the OEMs, not only by the mayors, not only by the governors, not only by the, but by the people on whose streets these things are gonna drive down in front of. I like to say, if I don't want one of these things to drive down my street, I'm gonna either key it, I'm gonna throw a brick through its window or I'm gonna run from my front porch with a jack and jack it up and put it on cinder blocks and steal the wheels. I mean, these are driverless vehicles going to pick up somebody. And if they're not welcomed in our neighborhood, if we don't believe that they are safe, there's somebody here from the city government of Geneva I was talking to outside and he said he's concerned about the safety. He's not gonna let these vehicles run around that and provide mobility to the mobility disadvantage here. So we've got to get from all the techno jargon that we all deal with every day to somebody's feeling good about these things saying that the reward is worth the risk and actually making it happen. Ali? So yeah, I just wanted to react because for me, like I'm not nodding my head in approval when I'm hearing that there's no AI in the vehicle or even in our software for that matter because I think it's just a typical AI backlash of buzzword because it's been used by every startup that's born after 2016. And so everybody has sort of like excessive AI is basically just to diffuse it. It's just when there's micro decisions that can easier be made by a machine than by a human why not let the machine do it? And then the AI part is when the machine self learns itself instead of having machine learning where the human trains the system. So it's not, it's just basically AI is sophisticated use of big data and cars are definitely repositories of big data certainly the autonomous cars. Even the software that's managing big data pools or data lakes, et cetera is using. But for real AI, do we probably need a connection to the car maybe to have more of a hybrid experience to do more sophisticated processing off board or self contained in the vehicle edge computing kind of? Well, I mean, people are working on both, right? And I think for me, the answer is also in the car because it's basically the microchip that has to recognize in a flick of a second. And also, we talked about connectivity what if you're not connected at that moment now it would be good if you have the chips in the car to be able to say that that's a dog or that's just a football, right? And that thing is super important and to have that quick reflex as sort of a machine intelligence or that mimics the human brain. I think that that would be good if we can transpose it into car chips. Yes. From Roger, I think it's happening now if we think about AI like, I mean, a computer taking decisions because if you think automatic breaking or lane assist is a computer taking a decision. So I think the broader, I mean, when you think AI like Jean-Luc Picard, we are not there yet. But there is some brain in the car. And I think you will see more and more hybrid models with Edge and Fog computing inside the vehicle. I mean, the power of the cloud it's allowing you to scale, to go fast and to have this huge computational power. But then you have decisions that should be in milliseconds inside the car. So usually you will see something happening in the car majority of heavy workloads in the cloud. And then what you run in the cloud can be deployed back in the car with these continuous improvements. So I think you will see, I mean, there is a German OEM doing that, cleaning the data before moving them to the cloud using machine learning, which is a subset of AI in the vehicle. And you will see more. I don't think that the totally 100% cloud vehicle will happen because you still need decision in milliseconds. And even if it's not about autonomous driving, I mean, I guess that's the same for Olga and your co-driver assistant. And that's the same for us at CloudCar. We, some of it needs to happen in the car when you don't have connectivity. And yes, it's much better when you can contact this huge computing capability in the cloud. Yeah, but Edge is the problem. Yeah. I think for me, it's fascinating from a city perspective when we think about safety. So what are we actually optimising for? Are we optimising for the largest revenue that's coming back? That's where the mobility as a service business is. But are we optimising for safety? And what information can we get from these Edge sensors, these cars that know about their environment, know where all the objects are in the environment? What are we doing to learn from that data that's been acquired for driving road safety? At the moment, 95% of accidents are caused by humans. How many accidents have been avoided by humans every day? There's no data on that. There's no statistics on that. We have cars that are going to come onto the roads, which will be acquiring all that information. So you'll be able to provide information back for insurance companies, but for cities to try and manage those risks. And I think that's something that as an industry, if we have a hundred billion pouring into it, some of that should be focused on safety. And just to add about the city role, I mean, we're very encouraged to see that cities, typically, you could portray cities and governments as slightly bureaucratic and sort of following the fast technology trends. But in this case, we see actually some cities that are ahead of the startup curve. When you see a Madrid that says, all of a sudden, we're not allowing our fuel emissions inside the city center, that is a radical movement. And so this is actually a fantastic mobility market where even the governments and the cities will play a key role in allowing autonomy to happen. I often make the joke, like mayors up for election all of a sudden see mobility as, hey, that's a good ticket for me to get reelected, right? And when it becomes that sexy, then all of a sudden... Well, now it's funny, you should say that, especially coming from the neighborhood that you come from with Gilles-les-Jean and this people fighting for their cars and fighting against fuel tax and then using it as an excuse to burn cars in the street or whatever. But vehicles are this incredible economic asset that governments fight over. I mean, one of the first things Trump did was institute tariffs on cars. But at the same time, cities are limiting the kinds of cars or determining the kind of transportation environment they want to have. I think the only problem that they're wrestling with right now is almost all of the transportation modes that are currently at their disposal are overloaded. There's no solution. It's about sharing the data that's coming from those cars as well. So I think there's cities in the U.S. where they're saying, yes, you can deploy your fleet, but that data has to be shared, it has to be open and then we can get the universities to do research against that. And it has to be a data-driven economy in a way but that needs data to be shared to enable that. On that note, I'm going to give Alain the last word as a representative of the university. The answer is shared ride. Okay, you put two people in a car instead of one, you reduce VMT in half. You reduce greenhouse gases in half, you reduce energy in half, you reduce the EV energy requirements in half. It's all about ride sharing. How we get that now? If he can create an environment in the car such that we'll love the ride together and I won't say, hell, I'm not riding with him. I mean, that's what I want to get him to do. Well, let's make a distinction, not ride sharing, but the kinds of vehicles that I think, ride cell and vest mile are enabling with multiple people, not just the driver and the passenger, right? Yeah, we need right pulling, pulling is key. Okay. Any other last thoughts or are we done? Okay, please thank these experts for their thoughts.