The Impact of Autonomous Vehicles – Digital CxO Podcast EP95

Hello and welcome to the digital CXO podcast. I'm Amanda Ani, and I'm excited to be here today with Iran, er, he is the CEO of imagery. How are you doing?

Great, thank you. Hi, Amanda. Thank you for inviting me.

Happy to have you on our show. Can you share with her audience a little bit about imagery and what services do you provide? Um, alright, so Imagery is a software company.

Uh, we provide level three level for autonomous driving software stack to, uh, two sectors, uh, in the automotive industry. One is tier ones and OEMs, to which we provide software models, uh, for the vehicles, uh, to get, you know, to drive autonomously on the road. And the other one is that we provide autonomous buses to public transportation operators.

Wonderful. Well, that gets us into our topic of the day, which is autonomous vehicles. So first I would like you to share a little bit more about the technology behind autonomous vehicles that AI map was driving.

Can you share a little bit more about that? Absolutely. So autonomous driving, you know, the, the, the, the technologies that were used in the last, uh, I would say decade at least, um, oh, last 15 years, uh, for autonomous driving, uh, changed dramatically.

Uh, it started with the classic, you know, software methods, um, and the machine learning and stuff like that. And I think that what we see in the last three years is, uh, that the few companies that remained in the sector in the, in the, in the sector of autonomous driving are all based on ai. And I will explain later why they're all based on artificial intelligence, because it was, it came evident that if you are not using ai, then you get pretty fast into a, a glass ceiling, uh, of performance.

Uh, because the number of scenarios and permutations of the road of what could happen is, you know, is, is the, is impossible. It's endless. And in order to, uh, get a vehicle without a human driving, right, uh, in order to get that vehicle driving on the road in a safe manner, uh, you have to not only develop a very sophisticated and powerful system, but also there are no shortcuts because you have to train it for years.

Uh, it's called training of neur networks and stuff like that of AI until it gets to the level where it's safe enough, uh, to be on public growth. So the way that we do it is that, uh, imagery is all based on ai and we've been driving on public roads for a little bit more than five years. Uh, we've been doing it in the us, Arizona, California, Nevada, uh, in Germany, in, uh, Japan, in Tokyo, and in Israel.

Uh, so we have a fleet of vehicles, uh, driving. And since 2023, we also have autonomous buses, uh, driving in few places following winds, uh, that we had in, uh, bids, uh, for autonomous buses. So the way it works, maybe I'll explain it in kind of file, try to make it one minute.

Uh, the way it works in a very simplified way is that we get the video feed, the video feed from eight cameras that are installed on the vehicle. That video feed is, uh, getting into an array of neur networks where each network was trained over a very long period of time with hundreds of millions of images to look, to learn to understand a different type of object. It's called object detection and classification.

So one is looking on traffic lights and one on traffic signs, and one on pedestrians, and one on lanes, and one on parked vehicles, and another one on moving vehicles and, and so on, many small, newer networks. And then we take the output of those networks and we create in real time without getting any HDM from the outside. Uh, we drive lucky human the same way that the human is processing everything that is surrounding us.

So, uh, we build in real time a three dimensional map that represents, uh, the surrounding of the vehicle, a kind of 360 degrees, 300 meters. Uh, and that map get in, in, gets into another section of the software, which is doing motion planning. This is a really cool place, uh, because it's built with an architecture that is called the deep convolution, newer networks, which means that unlike the classic, uh, old ways where how we wrote software like 20 years ago, we don't use rule-based coding.

You know, it's not like decision trees, you know, the classic software of what if then do that instead. It's like a black box where everything gets into it. Like, you know, AI that you see in the movies.

And that part like child, like a human is learning over time, uh, to, to take better and better decisions, driving decisions on the road. And this is how a path is being created for the vehicle to drive. Okay.

So are, are satellites also being used, uh, for directional purposes or not? Uh, well, yes and no. Basically, we don't get HD maps, meaning we don't get pre-prepared maps into the vehicle that are telling the vehicle what it'll find ahead.

Uh, so you know, all these elements on the road and the, and crosswalks and circles and the, and so on. However, uh, the vehicle gets, uh, what's called navigation. Uh, so from a navigation layer, which is basically GPS, it gets to A to B.

So for example, from Google Maps is yet, okay, so now you are here and you need to drive there, so you need to like, uh, drive, you know, one mile and then, uh, uh, after that, uh, mile you need to turn left and to drive, uh, half a mile and then take right and, and stuff like that. But the vehicle doesn't know what is it expected to find along that route that it was given. So if, uh, say there's been an accident or a road is closed off or, or an animal's in the middle of the road, um, the AI is just trained to, to, to notice and make adjustments.

That's a beautiful example. That's exactly the point gi given that we are mathless, uh, and it's not that we told the vehicle what it'll find. So if there is a, a, an accident, if there is a need to do detour, for example, with buses, right?

Uh, if, um, uh, there is a construction on road, everything that is happening, you know, on the road in real life, uh, the vehicle will cope with it because it knows where it needs to get and it'll know that, okay, so if I cannot drive straight, then I will turn in here, right? And then I will turn their left and I will, and, and it'll direct itself, uh, to, to that big point where it needs to get and will still understand everything in its road because it wasn't trained, uh, to a specific road, it was trained to drive, you know, generally in every road. And by the way, that's what leads us to a unique capability that you can currently find only with Tesla line, with imagery, uh, that is location independent.

So basically, you know, like you get into a new location, uh, you land, uh, you know, uh, in an airport you rent a vehicle and you drive out. That's it. The same with our vehicle.

Uh, it gets to a new place and given that it knows where it needs to drive, it'll simply drive there because it knows how to drive like a human. That's amazing that, and so you said you have even buses that are autonomous. Are these public transportation buses that, that everyday people can get on and, and go where they need to go?

Yes. That's a really new big O uh, uh, blue ocean that is being built. Uh, so currently the 22 markets globally, uh, that are conducting pilots with autonomous buses.

Um, and, and so the way it works is like that once upon a time there were these small pos like special purpose vehicles that used as autonomous shuttles. Now those ports could only carry, I don't know, four, six, maybe nine passengers, and they could drive very slow, like 15 miles per hour. And, um, uh, they couldn't do a slope on the road of more than, uh, uh, I don't know, uh, 15 degrees.

Um, and, and they, they needed maps and everything, but nowadays the world changed and many public transportation operators globally issued the tenders in the last two years or so. Uh, for autonomous buses, meaning real buses, what's called M three class bus, meaning bus that looks like a bus and has the capacity of 32, 35, 40 passengers and electric buses. And those buses basically are, we, we convert them, those electric buses into autonomous buses.

We install computing, we install sensors, uh, we install our software, we turn that electric bus into autonomous, uh, uh, bus, and we hand it over, uh, to the public transportation operator. Uh, for pilots, typically such pilots takes between 18 to 24 months because of legislation. There is a, a very, uh, uh, the, the legislation for autonomous bus buses is much more, um, uh, rigorous is much, uh, tougher than the, uh, uh, than the regulation for, uh, regular, you know, passenger vehicle, uh, autonomous, um, uh, driving vehicles and, and, um, uh, those buses, they, at the beginning, they need to drive in an operational, uh, zone road.

Uh, and then eventually they get the permit to drive on a public road. And then the next stage, after passing all kinds of cyber, uh, security, um, uh, regulations, they can onboard passengers. And eventually, after a period of time, uh, for example, in Europe, it's like, I dunno, after you drive 100,000 autonomous kilometers, uh, with no problems, uh, then on that route, or it's called ODD, uh, then you get a permit from the local Ministry of transportation to drive level four, meaning to pull out a driver and to let the bus continue, you know?

Um, and, and that's what we do. We have two lines like that in Israel, uh, that are driving, uh, autonomously, currently level three, and they will get to level four in 2025. And we are now getting with that product to other markets, uh, in, uh, Portugal, in Germany, uh, in, uh, Japan, and so on.

So that's, uh, this is really widespread and something that I think that you will find interesting or you, you, uh, uh, you know, the, the audience, um, hearing that will find interesting, as much as it sounds counterintuitive, autonomous buses is an easier product than autonomous, uh, passenger vehicles. And the reason that's very interesting, and the reason is that, you know, think about it, autonomous bus drives in a geofenced area, right? In a special, uh, in its route, uh, in, in a no environment.

So that's one, two, it's, uh, it's, it, it drives in a city, right? So it's, it's limited by speed. You can only drive in a city, I don't know, up to, uh, a 50 kilometer per hour or 35 miles per hour or something like that, um, uh, in, in, in a city.

Uh, uh, so it's not like the, the like, uh, passenger vehicle that it can get to 70 miles per hour. Uh, so that's another thing. And also, the stops that it needs to, the bus stops that it needs to take along the way are very specific.

So, uh, especially an AI based system, it learns really fast, like within two weeks, the entire route. Uh, and it knows where to stop in each of those bus locations. Unlike a passenger vehicle that can go everywhere, and that can stop everywhere.

So from auto perspective, autonomous bus is an easier, uh, product to produce than autonomous passenger vehicle. Wow, that's amazing. So I wanna talk about what this means for the future and the impact.

I have a few questions like, does this reduce accidents? Um, does this reduce carbon footprint and congestion in cities? So, um, can you share a little bit about the impact this is going to make?

Right. Okay. So, um, I'll start with the accidents.

I know all these, you know, populistic, uh, uh, uh, claims about, uh, you know, accidents of, uh, autonomous vehicles and so on. But if you'll check data, you know, statistics, then you'll see, take for example, Tesla, if te Tesla's vehicle vehicles, uh, pa cost, uh, at the beginning of 20, in, in, at the end of the first quarter of 2024, uh, they, uh, passed a billion, uh, autonomous miles that were driven by Tesla vehicles. If you would let, uh, humans drive a billion miles, you will see that the number of accidents caused by the humans would be, I don't know, tenfold or 100 times, uh, the number of accidents that were caused, uh, by those Tesla vehicles.

So that, and why is that? It's because the processing power that exists in those vehicles, in our vehicles, in Tesla's vehicles, uh, we don't make cars, right? We, we provide the, uh, system, a platform, uh, to other car manufacturers so they could compete, uh, with Tesla's, uh, FSD, uh, to German, to Japanese, to other, uh, car manufacturers.

Uh, so the vehicle that is installed with our system, it sees 360 degrees. I cannot see 360 degrees when I drive, it sees 300 meters. It analyzes every element within that range, uh, and it's vector, meaning where is it going on which direction and what its speed.

It's again, not something that I can do in real time. And on top of it, it does that in a, in a speed that is sub 100 millisecond, uh, in terms of a response time, what's called latency. Whereas we, as humans, our response time is, so, is something closer to 300 milliseconds, which means that the autonomous driving system is at least three times faster than a human.

So it's safer, it's sees much more, it understands everything on the road. It's, it, it drives by education safer than a human. It doesn't take risks, and, and it's, it's faster.

So, so that's on the accidents, uh, it's, it's obvious that there would be much less accidents, the more autonomous vehicles will be on the road. So that's, uh, and, and I don't, I didn't even get to the human factor, right? Drivers that are driving, uh, you know, uh, when they are drunk or, uh, distracted or busy with the phone, uh, or, uh, I don't know, agitated, whatever.

Uh, you know, we see all the, all of that stuff. This is not going to happen with autonomous vehicles, and especially think about it when it gets to buses. Now, with respect to carbon footprint, footprint.

So what we see today because of organization, a organization because of aging, because of a, a, a severe shortage of drivers, uh, for buses globally, 15%, and it's going, it, and it go again, it goes up by 1% every year. We see that there is a lot of pressure on, uh, public transportation, uh, companies to bring more services, more vehicles on the road in order to meet demand, but they cannot because they have drivers. So what's currently happening on the road is that more vehicles are getting into traffic because of that, because of that need that is, is not met, is not being met.

So with autonomous buses, public transportation, not only did they save a lot of money, because every such bus saves between 40 to $70,000, uh, uh, for the public transportation operator per annum, per bus, and they have thousands of buses. But also that bus carries, I don't know, 35, 40, 45, uh, passengers, which means that these are people that won't, won't take their vehicles because they will get to their destination safer with an electric bus. We see many cities in Europe that are becoming green, that are blocking, for example, city centers, uh, for passenger vehicles and stuff like that, and only allow, uh, uh, buses to get into those, uh, uh, areas.

And, and again, many of these buses will be in a few years autonomous. Uh, we see more and more. Uh, so what it means, uh, is that many passenger vehicles, especially in an urban environment, will be replaced in the next decade by autonomous buses because not only because it's safer, but because be, you know, money is, is a great motivator, right?

It'll be much more economical, uh, for everyone, not just for the consumers, but also for the governments because, uh, they don't need to, to build them the infrastructure for so many vehicles. If instead, they could subsidize like they do today, public transportation operators to bring autonomous buses on the road. Wonderful.

Well, I have a question also as to how long will it take for autonomous vehicles to be the main type of vehicles on the road? Mm-Hmm. Do you think there's gonna be a time where there's very little people actually driving and everybody's in autonomous vehicles?

So, so first of all, I think that the autonomous buses, you will see them, uh, widely used, uh, before you will see, uh, mean autonomous buses and taxis. You will see widely populated and being used before you will see passenger vehicles, uh, especially just because of the public education. It takes time, uh, to getting used to, you know, let your car drive instead in instead of you.

Now, this is why the, the, the what's OEMs, what car manufacturers are doing today is educating the market. So instead of, uh, dropping on them like a full, uh, full self drive like Tesla did, many German and, uh, Japanese manufacturers, for example, took the other, the opposite approach, what's called bottom up instead of top down, meaning they're starting with entry level vehicles and, uh, gradually introducing features that we provide them, like in a, in a, uh, scaling up roadmap. It starts with autonomous parking, and then it gets to traffic jam assist, and then it gets to save driver Overwatch, which is a, a product that was designed for elderly and for young drivers.

And eventually it gets when you are comfortable with it, uh, to a full self drive, uh, you know, like the Tesla one. So what we will see probably in the next decade is, um, a process of educating the consumers, uh, to more and more advanced functionality. And this also, uh, coincide meaning this, this gets along with the drop in the cost of the needed elements.

For example, computing. Everything that I've mentioned relies on really heavy sophisticated computing that is needed for ai. Uh, and, and as we know for news law and in general, uh, the price of computing, you know, gets lower and lower and lower.

So gradually what we see in vehicles is that the computing that they get installed in the vehicle is getting much and much and much stronger, which means that the cap, the autonomous driving capabilities of those vehicles are getting increased exponentially. And so, so I I, I give it something like 10 years. Um, uh, and then I believe that most of the new vehicles that will get on the road will be, uh, capable of, uh, autonomous driving, uh, capabilities, uh, functionality.

And then it's for the buyer, it's for the consumer to decide, uh, if, uh, he or she would like to activate them, right? Because you need to activate them in order for them to walk, even if you buy a vehicle that can drive autonomously, Right? So they'd have that choice if they, if they're tired and they want it to just drive for them, they can sit back and relax, or maybe they want to drive And pay for it.

And they see, you see the, the, the new model that the car manufacturers are moving to, you can see with Tesla, with the $99 per month, uh, that they charge FSD, uh, users, and now you can see it also with BMW and Mercedes. So eventually our vehicle, that's why it's called SDV software defined vehicle, our vehicle would be like our phones. We will download and upload applications, uh, that, you know, will do all kinds of stuff like autonomous driving, and we will pay for them.

Uh, so we will have a billing relationship with the car maker, right? We will need to pay, I don't know, GM or Ford or, uh, uh, anyone else, uh, for the, uh, for the functionality of our vehicle. Okay.

So most likely it would be some sort of subscription service for that feature? Yes. Yes.

Okay. Will start paying, uh, subscriptions, uh, for the autonomous driving capabilities of our vehicles. Mm-Hmm.

I would see there, there being a, a good opportunity for the shipping industry. You know, all the truck drivers out there might not wanna hear that, but, you know, with autonomous trucks shipping across the US much more quickly and efficiently, Yes, that's a very big sector of autonomous driving. Uh, first of all, there is this a big shortage.

Uh, uh, also we truck drivers, because, you know, you, you can't, uh, the young generation, uh, the, you know, the, the z generation, and so they don't want to drive commercial vehicles, whether it's, uh, buses or trucks, and you can't, uh, and, and you see entire supply chains, uh, being affected in Europe and in the US because there aren't enough truck drivers. So currently there are many companies. We are not doing autonomous trucking, uh, but there are many companies that are doing autonomous trucking and trucking.

The unique thing about trucking is that it divides. Uh, so there is the long haul trucking, uh, you know, those big trucks that are driving cost to cost, uh, in the us and there is the, uh, last mile, uh, you know, those, uh, Amazon trucks that, you know, gets the package, uh, to your home. And there is, uh, the mid-range, uh, kind of the middle mile, uh, which, uh, is doing the distribution from the regional, uh, logistics centers, uh, to into the city.

Uh, so, you know, eventually the package will get into the last mile truck. And we, so what you will see, in fact, is companies providing autonomous driving capabilities to all those types of different trucks along the way. And eventually this all, uh, that, that entire supply chain will be done autonomously, which is, I don't know, I think that is fascinating.

Wow. All right. Well, if there was one key takeaway you could leave our audience with today, what would that be?

I, I think that it's, you know, we've been hearing about autonomous driving for the past 15 years and, and, and now it is happening. And I believe that we will see a very big change, uh, until, uh, the end of this quarter. I don't think that by 2030, uh, my daughters will be driving their vehicles.

And the reason is simply be not, not just because of regulation, but mainly because of computer, the, uh, computers, meaning the maturity of the systems, which respect to one, the software, the AI based software, and two, the hardware, which is the computing that is needed, uh, for that ai, uh, system to run on. Uh, it's, it's there. That's it.

It's mature and it's, it's, it's, uh, the, the cost for it is at such level that it can get into everyday vehicles. So it's not just for the premium, uh, vehicles as it used to be and stuff like that. So I, I think that, uh, now we'll see an accelerated pace of adoption.

Wonderful. All right. Well, thank you for coming on our show and talking to us all about autonomous vehicles.

Thank you, Amanda. It was a pleasure. Good with you.

Alright. And thank you to our audience. Stay tuned.

There is more.