Can AI Drive Better Than Humans?
In fog at Provizio's base in Shannon, Barry Lunn says he can see 10 or 20 metres if he is lucky. And on a nice morning here, our sensors can see 600 meters in that same fog.
In this episode · 18 chapters
- The study with a sample of 29
- What Barry makes of the study
- The catalyst
- The autonomy levels
- What you see on social media
- Could I buy a level five?
- How Provizio sees the world
- Range and velocity
- Better than human, in fog
- Radar against lidar
- Rear end crashes
- What crashes cost
- It is all about the data
- How the regulators are reacting
- FMVSS 127
- Pushing it onto the tier ones
- Can we see it?
- Memento mori
So we said, let’s look at it from the opposite side. Let’s focus on the hard bit first. And as it happens, the hard bit in autonomy is also the bit that we as humans find hard.
Barry Lunn, CEO and founder of Provizio, started the company after a crash in early 2019 that he felt should have been preventable. He explains why he finds the autonomy levels confusing, and why he rates Waymo at level four.
Provizio builds radar that reads the range and velocity of cars several vehicles ahead, so it can spot one slowing before its brake lights come on. Lunn also covers a new US emergency braking rule and why cost decides whether safety tech reaches every car.
Barry Lunn, CEO and founder of Provizio, argues on the Thinking On Paper podcast that AI will drive cars better than humans once their sensors see more than people can, and that radar, not lidar or cameras alone, is the backbone of autonomous driving. In fog at Provizio's base in Shannon he can see 10 or 20 metres; he says Provizio's sensors can see 600 metres.
Can AI drive better than humans?
Barry Lunn says it will, once a car's sensors see more than a person can. He argues radar is the backbone of that: in fog he can see 10 or 20 metres, and he says Provizio's sensors can see 600.
Why does Barry Lunn find the levels of autonomy confusing?
The levels of autonomy, which run from level zero (no automation) to level five (a car that drives anywhere with no driver), confuse more than they explain, Lunn says. He says 99 per cent of carmakers are still stuck on level two, where the driver keeps eyes on the road, and calls the jump to level three, hands off and eyes off, “a monstrous leap” because responsibility moves from the driver to the machine.
What is 5D perception at Provizio?
Provizio defines “5D perception” as seeing the world in three dimensions, plus the range and velocity of every object, plus a fifth layer of understanding that interprets the data. Lunn says the level of autonomy should change with conditions: in fog the car should drive, in clear weather the human and the machine work together.
How can radar see several cars ahead?
Radar can see several cars ahead, which cameras, lidar and human eyes cannot. Lunn says millimetre waves bounce around the vehicle in front, so Provizio's system can see a car six or eight vehicles ahead start to slow before its brake lights come on. He says more than 50 per cent of car crashes are rear-end collisions caused by a lack of foresight.
Is most driving data collected by carmakers useless?
Most of the driving data carmakers boast about is useless, Lunn argues. He calls claims of a hundred million or a trillion miles of data “nonsense” and says the value lies in edge cases, the rare dangerous moments, filtered on the vehicle itself because cameras eat bandwidth and storage while radar measures velocity directly.
Which US safety rule will put emergency braking in every new car?
A US safety rule is pushing the technology into every new car. Lunn says the regulator NHTSA's FMVSS 127, an automatic emergency braking rule, means that from September 2029 every new vehicle in the United States must be able to detect, for example, a pedestrian stepping out between two cars. He also says car crashes cost developed economies 2 to 3 per cent of GDP, and up to 5 per cent in emerging economies.
Where Mark and Lunn disagree
Mark asks whether there will come a point when a human is never a better driver than the machine. Lunn says yes, and that the debate then becomes one of evidence: rather than solving the trolley problem, “how about we see the little old lady a couple of hundred meters away and let's not kill anyone”. He also admits he over-focused on safety at first and learned late that cost decides whether carmakers adopt the technology.
Links from this episode
Questions
Can AI drive better than humans?
Barry Lunn says it will, once a car's sensors see more than a person can. He argues radar is the backbone of that: in fog he can see 10 or 20 metres, and he says Provizio's sensors can see 600.
Who is Barry Lunn?
Barry Lunn, CEO and founder of Provizio, started the company after a crash in early 2019 that he felt should have been preventable.
What does Provizio build?
Provizio builds radar that reads the range and velocity of cars several vehicles ahead, so it can spot one slowing before its brake lights come on.
Will AI drive cars better than humans?
Barry Lunn, CEO and founder of Provizio, argues on the Thinking On Paper podcast that AI will drive cars better than humans once their sensors see more than people can, and that radar, not lidar or cameras alone, is the backbone of autonomous driving.
Transcript
Jeremy Gilbertson disruptors and curious minds. Welcome to another episode of thinking on paper. My name's Jeremy. This is Mark. Guess what? Guess what guys each week we get to talk to amazing founders, amazing builders, amazing disruptors, building the new pieces and parts that will power our future. And we get to figure out what it is before it happens. This is why it's so exciting today. Today we get to talk about, ⁓ sensing and perception technologies as it relates.
to the combination with AI to make cars better drivers than humans, ⁓ which is really interesting. Mark, you said in our pre-pro chat that you are a terrible driver. Like what makes you so terrible?
Mark Fielding I'm a terrible driver. Yeah. What makes it so terrible is that sometimes I think I'm an awesome driver. That is a dangerous combination. Um, honestly, I think I've been driving past my tests when I was as soon as I could, like the car for me was freedom. was 17 in England, took my test, passed it. And I think I have this, Oh, I've been driving for 20 years. I'm a really good driver, but in reality, maybe I've been driving for 20 years. I've got loads of bad habits.
Jeremy Gilbertson Ooh, that's dangerous. ⁓
Mark Fielding I think I'm better than I am. Um, I've got two kids and I was saying in the pre-chat as well, well, I'm very fearful of when they learn to drive the roads. The roads are dangerous. They're more dangerous than war zones, natural disasters. They're more dangerous than your kitchen. Like today's guest is going to helpfully explain how AI and sensors can make the road safer.
The study with a sample of 29
Jeremy Gilbertson Yeah, before you introduce our awesome guest, I ran across a study and I think this was around 2009. This is to help you, Mark, because you could be a part of these people. There was a study done in 2009 that was published in Cerebral Cortex, which is got to be a good spot. Cerebral, no, I think it's a journal. think it's a medical journal. The problem is the sample size is very small, 29 people. So let's add.
Mark Fielding that's the name of a magazine
Jeremy Gilbertson the credence we can give the study that has size of 29 as a sample size. But they say bad driving can actually be genetic. And there is a variant called VAL66MET that if it manifests in your genome, you are actually predisposed to be a 30 % worse driver than someone else that doesn't have that variation. So it could be genetic. It's not your fault. It's not your fault. All right, Mark, let's introduce our guest. Fire away.
Mark Fielding Yeah, I guess today is Barry Lunn, CEO and founder of Provizio. And yeah, welcome to the show, Barry. Just before we go, Jeremy, you forgot to mention our new sponsors this week.
Jeremy Gilbertson that's right. So, so this week, thinking on paper is powered by human conversation, not just any human conversation, but meaningful human conversation. This show is sponsored by human conversation.
Mark Fielding Awesome. Welcome, Barry.
Barry Lunn Hey Mark, hey Jeremy, great to be here and let's have some meaningful conversations then. I feel pressure now.
Mark Fielding What do you think about Jeremy's research about bad driving being genetic? Possibly.
What Barry makes of the study
Barry Lunn Yeah, as you said, the sample size is small as someone who deals in big data all day every day. You certainly have a place for error there. But of course, I think even I find it interesting that they actually found a genome, right? That was responsible because we certainly learn our behaviors from previous generations, right? And I do think that we do pass on some of those bad habits to our kids as well when it comes to driving and and we learn it from other drivers as well. So I think there's an awful lot of things that make us bad drivers, but I think in reference to our pre-chat, I think recognizing that you're not a great driver puts you in a better place than most drivers, because we're all terrible drivers. I have the statistics in my favor to prove that.
Jeremy Gilbertson Awesome. ⁓ Well, one quick thing on that study with the tiny sample size, ⁓ that particular area of the genome is actually responsible for supporting neuroplasticity, which is really interesting as well, which is how our brain can kind of reprogram pathways and that sort of thing. So anyway, I don't want to dive too much into that. Let's talk about our carryover question, because Barry, what we like to do is kind of thread these shows together, thread these episodes together ⁓ in an interesting way. So we had Tori Smith.
from Endiotics who has Pillabot, you swallow it, the pill goes in your stomach and you can do diagnostics. I know you're not ready to do that yet, but his question was really interesting for anyone building something new that is different than the existing things that are out there. So his question was when things get tough, like when things are difficult, which they tend to be in any entrepreneurial venture, why don't you quit?
Barry Lunn That's what my wife keeps asking me.
You know what, it's a very interesting question, right? And there's a couple of layers. And I think timing is a big part of this in terms of the answer would have been different 10 years ago as it would have been 10 years prior to that. But I'm naturally an optimistic person. So I always feel like I can come out the other end, I think is one key part. I think you need a sense of optimism or you'd never start a company. But you know, as I say, 20 years ago was survival. Right. And that changes 10 years ago. It became about, you know, make an impression on my kids. Right. And, you know, that they'd see that that dad worked hard and he, you know, he was he was going to do something interesting every day. And I think now in this stage of my life, what really motivates me and keeps me going is impact, right? And what we're gonna talk about today and what we're building here at Provizio, that's really about impact. And I think that for me is why I won't give up. This is a tough industry. I'm in the auto industry. ⁓ It's a tough industry, right? These guys are all fighting for survival. So you gotta have some tick skin, but when you have a mission...
and abnation could have a pretty big impact on humanity. That's, I think, what keeps me going.
Mark Fielding Good answer and this certainly what you're doing could have a big impact on humanity.
The catalyst
Jeremy Gilbertson So, so there's a note here, tell me, tell me about the catalyst that, was part of the inspiration for, for creating this, this new system related to you in an automobile accident.
Barry Lunn Well, I've been in a few, right? And I guess for context, I've been building ⁓ sensor systems and things like that for over 20 years now. But why did I go at the automotive one again? It was two parts. ⁓ Primarily the technology we've developed that would have been used in aerospace and a lot of type of defense type applications. And you reach a stage in your life where you go, well, that's nice and all.
there's lots of money in that sector, but could I be doing something a little more impactful? And there's nothing like a car crash to concentrate the mind. So I had a car crash back in early 2019 after selling my previous company. And I think I got a little bit of a shock as you do. It's not a very nice experience. I've been in them before.
But unusually, I didn't go home and shake myself to sleep and think about maybe how I could be a better driver and things like that. I actually went home and just started working on, hang on, that accident should have been preventable.
I should have been able to foresee that. Why wasn't I? And I started looking at all the things that humans are bad at that lead to us crashing. That led me down a rabbit hole. And when I came out of the rabbit hole, I formed a company. it was over a six-month period, but I just couldn't let it go. And the more I looked at it, the more I thought, you know what? We could do something really interesting here. And by we, I mean, I had to go out then and talk to the smart people.
people and recruit them. But that was definitely a really important moment in why I went with Provizio and not something else.
Mark Fielding I think it's not the first time we've heard that a personal story that drives that will to just keep going at all costs. ⁓ before we get into Provizio, there's, was going through the website, I was looking at the tech and there's quite a lot of terminology that I don't, I'm not familiar with. don't understand. Obviously AI is there. I get that.
The autonomy levels
Mark Fielding Could I just ask you a few tech questions just to set the scene for listeners so we kind of know what you're talking about when we get into how it's solving the problem? ⁓ I mean, I don't know if you're familiar with L1, L2, L3, L4, L5, these layers of automation in cars and how that impacts what you're doing. So maybe if you could talk through that and 5D perception.
Barry Lunn Yeah, absolutely. you know what? The two things you mentioned, ⁓ you have a very good excuse for not understanding them. And the first one being the levels of autonomy. And the reason I say that is because they don't make a lot of sense. They were developed before we truly understood the complexity of the problem. really what level zero is basically a car with no sensors that you just.
drives by human control all the time. And level one is where you start to add things like cruise control and things like that.
Level two is where you start to allow hands-on, eyes-on driving, but you're now moving to a stage where you've got ⁓ lane centering and adaptive cruise control. So a little more automation in the system and automated emergency braking in a dangerous scenario. then here's the problem, right? Because it goes to level five, but it's a bit like that Microsoft progress bar.
and you go I 20 % of the way but I've been sitting at 20 % for the last couple of years and that's actually what's happened. Level 3 is where you've hands off and eyes off driving, right? So that's quite a leap. So you hear an awful lot in the sector and out there in the news about autonomous driving but
99 % of automotive companies and in the space are still stuck on level two. And the reason is it's so significant. You're talking about changing an entire industry. When you go to taking your eyes off the road and taking your hands off the wheel, something else is responsible for driving now. on, you know, one of my hands on the wheel and my hands on the pedal. I'm in charge here, right? I'm responsible. My insurance is covering me. So it's a monstrous leap.
Level four then is where in certain scenarios, the vehicles fully autonomous, right, are in certain regions, for example. So that's where you something like a Waymo vehicle that we're all aware of that, you know, San Francisco can drive fully autonomous without a driver, but in that area, right. It can't it can't move up the road. And then as you go towards level five.
That's where it's ubiquitous autonomy. The vehicle can drive anywhere it wants with no driver. In fact, there's no steering wheel. And that's really where you're going. So they're not equal. I think the goal is to get from level two to level three, and then everything gets a bit easier from there. I hope that makes sense.
What you see on social media
Jeremy Gilbertson No, totally does. to kind of route this in to what people see on TV in between the shows they're watching. It's something called a commercial. A lot of people pay to avoid them. Some people have to watch them. But there's one, I can't remember if it was GMC, Chevy, I don't remember whatever, but there's the, will rock you Queen song and they're going donk, donk.
Jeremy Gilbertson Like doing this with the hands off the wheel in the truck. Is that, is that two or three? Where are at with that?
Barry Lunn It depends how long are they keeping their hands off the wheel.
Jeremy Gilbertson a quick cut quick cut of them doing it for maybe, you know three seconds But largely they're telling the story of maybe the whole trip into the mountains could be done with a looped We will rock you. I don't know
Barry Lunn Yeah, so that's level two in the main, right? Because if you've taken the hands on and off the wheel, you can do that with level two driving systems, but your eyes are still on the road and things like that. And it's different rules in different countries, right? So in Europe has different rules to the United States, but true level three, those hands aren't going on and off the wheel and those eyes aren't focusing on the road in certain scenarios. So it's another leap. So like everything.
It's a new flavor of level two and you'll hear people talk about level two plus. And really that's because it's a bit more level two. So the guys that allow you to take the hands off sometimes that's a bit more than the guys who never let you do that. But again, ⁓ the descriptors are bad, I would say, using these levels.
Could I buy a level five?
Mark Fielding If I lived in California, could I buy an L5?
Barry Lunn I know you can't buy an L5 anywhere in the world, ⁓ just so we're very, clear. ⁓ In fact, I'm not aware of any L5 programs in the world. mean, Waymo is number one in autonomy, know, just absolutely ahead of everybody. But they're level four, right? Or maybe level four plus, because now they're in multiple cities. Waymo will be at level five when...
They can cover the whole United States and not have to worry about anything, right? Whereas now they're in pre-mapped environments that they understand. So I would consider that level four.
Jeremy Gilbertson So the one interesting thing about frameworks like this is, you know, one of our, so we have a book club and one of our books we read, the Nexus by Julio Tino. And he talks about grounding new things in the old way of doing things, right? To make them relatable. So frameworks can be interesting in that regard. Sounds like this one is a little, it's a little lacking and a little confusing too, but like where are you guys headed related to your work? And let's talk about some of your specific technologies that you're developing.
How Provizio sees the world
Barry Lunn Yeah, no, that's a good segue into how we see the world, I guess. We don't see the world in levels of autonomy. That's not how our customers do, of course, right? They have to sound, have to adhere to certain criteria. But we decided to look at the problem maybe from the different end.
And what I mean by that is it helped having a background in aerospace. So if you go to NASA tomorrow with a new technology, and you say, I want to bring this technology to space, but now it's made from a material that obviously can't go into space, that's the end of the conversation with NASA. But if you think of the autonomous vehicle industry, that's exactly what they did. They were like, well, we can drive it around in circles on a track. ⁓
If it rains, we have to pull it in. That to me, it just wasn't the path to go here. So we said, let's look at it from the opposite side. Let's focus on the hard bit first. And as it happens, the hard bit in autonomy is also the bit that we as humans find hard. And the edge cases, they're often called in robotics. And really, these are the things that lead to crashes and things like that. So we started looking at, well,
What level of sensing would you need to actually solve the driving problem? Stop us crashing into each other. Stop it. Give people the foresight to prevent crashes. And if you solve all of those, well, then autonomy becomes a byproduct because then it becomes an ethical question. If I can prove that my sensors and softwares and brain are better driver in these scenarios than you are.
then that to me is a much more compelling case instead of telling the world, hey, we're going to kick all truckers out of trucks and just have them autonomous to save money. That to me wasn't interesting. So that's kind of how we looked at it. And that's where 5D perception and conscious Mark asked a while ago, what the hell is 5D perception? And it is an abstract term, but it came from that. It was basically to solve the problem, you need to be able to see the whole world and understand the whole world in 3D, right? So that's three dimensions we all understand.
Range and velocity
The next dimension is range and velocity, right? So I can see the world and human eyes. We can see a car coming towards us, but we actually have to guess what speed it's coming at. And we've got to guess what distance it's at. Then make a guess of, can I get out in front of it? And we often get it wrong, right? So you have to understand that fourth dimension. You have to understand the range and velocity of all objects. And the fifth dimension, that's our abstract term, that's the understanding.
whether you call it AI, whether it's machine learning, whether it's digital signal processing, but making sense of all that data. That 3D object is a car. It's moving at this pace. There is a pedestrian about to step out. There's going to be an incident. That's the fifth dimension. That's that understanding. So that's how we kind of flipped the problem on its head and said, well, let's focus on solving the problem.
And then in those scenarios, and that's why the levels go out the window, right? Because actually you will have a car that will jump from level four to level two, you know, by those definitions at different times, depending on the scenario. So I'm here in the Southwest of Ireland in Shannon, right? And you may have heard the rumors, but we get a little bit of rain here, right? We also get quite a bit of fog, right? And so in a foggy environment, for example,
I can see about 10, 20 meters if I'm lucky, right? And on a nice morning here, our sensors can see 600 meters in that same fog. So in that scenario, the vehicle should be driving itself, not me. I can see 10, 20 meters. What use am I in that scenario, where as vehicle can see 600 meters, so we can make much better decisions. So suddenly in that fog situation, it's gone to level four.
But when the fog lifts, statistics say, I'm still better now in good conditions, I'm a better driver than the machine. So therefore we work together and it becomes level two. So that's kind of just a slightly different way of viewing the world as we see it.
Better than human, in fog
Mark Fielding I love going. Just playing on that though, in the future, is there a point when ethically you're never better than the machine at driving?
Barry Lunn Absolutely, absolutely. That's what we're going for, right? But then it becomes a, it doesn't become a debate at that stage. It becomes statistical data. I have the evidence. I'm better than the human driver in this situation, which is kind of, think, where the industry got it wrong in the beginning, right? And then people started having the ethical debates, you know, the trolley problems often brought up, you
Do I mow down the little old lady crossing the road or do I swerve and hit the kids? Hey, how about we see the little old lady a couple of hundred meters away and let's not kill anyone, right? And I think, you know, it seems so, right? I mean, we're laughing, right? Because it just seems so bizarre, but it's so logical, right? That that's how you would approach it. So you need better perception. that's really what we're all about in Provizio is to build that.
Jeremy Gilbertson Super interesting. So number one, I love that you're tackling the hard shit first. Like I think that's very commendable and just saying, let's go after the big stuff. ⁓ And so understanding the world around you, yes, seeing what's out there, but the range and velocity thing is interesting, right? Cause we're all pieces and parts moving in that world. So that makes a whole lot of sense, but having these sensors that can actually deliver the information that you need is really interesting as well. I start thinking about in planes like,
IFR and VFR like instrument flight rules visual flight rules like and and the the delineation of like obviously when you can't see in your flying airplane you have to you have to go instrument flight rules, but ⁓ What are the key? What are the what are the? ⁓ Give us a couple more examples of when? The tech that you're actually working on and testing and delivering What are those scenarios where it helps you make that switch?
Radar against lidar
Barry Lunn Yeah, absolutely. And there are hundreds, right? But just to kind of give some other examples, our backbone in our technology is radar, and we're very radar focused. And the reason for that is there's loads, like cameras are on every vehicle, and they will be. And we work with cameras, obviously, very closely. And then you've light errors was another technology that was developed.
LiDAR has a lot of the same weaknesses that camera has, right, and what humans have, right? So cameras, you can see like a human can. So we can't see the vehicle in front of the vehicle, for example, in front of us. And we can't, you know, we can't tell, as I said earlier, the exact velocity. So a very ⁓ interesting example of how we use the technology is radar.
millimeter waves, right? they literally bounce up around the vehicle and then onto the next one and onto the next one. And those waves send back the signal on the image of the vehicle, but also gives you the, as I said, range of velocity. So you know what every vehicle is doing. So when you have multiple vehicles ahead, humans can't tell what's going on a couple of vehicles ahead. The cameras can't tell, a lighter can't tell because of shadowing, but we can, the radar can.
So for example, we've all been there, especially in country roads in lots of different countries, where you're stuck behind someone and you go to make that overtaking maneuver, but you're blinded to the vehicles coming towards you as well. The sensors can see all of that. And these are some of the early implementations we're seeing with our OEM partners, where it gives you a gentle nudge as a driver. So you go to make that overtaking maneuver, it stops the wheel. Then you see that car go past you and you go, ⁓ I'm so glad.
that there's artificial intelligence built into this that saved me from myself, right? And now suddenly I trust a robot, right? As opposed to when we actually use hands off driving, humans get very nervous. They've done the measurements on our brains. Like when we hand over control, we're not really good at it. It's why people are scared of flying, even though it's so safe. So when the machine starts to save your life on a regular basis like that, that's where we think that that impact would be significant. Another one that I can give you is,
Rear end crashes
More than 50 % of car crashes are rear endings, right? So why do you drive into the guy in front of you or the guy behind you drive into you? It's because of lack of foresight, right? If you could have seen that something was gonna happen, you wouldn't have hit them, right? And so, in that scenario, we can see the car, let's say six, eight cars ahead of you, we can see that car start to decelerate before his brake lights come on, right? That's what the technology allows you to do because we've basically got his exact velocity. And so normally you've to wait till the car in front of you brake lights. So you can imagine the latency in that, right? It's one car, two cars, three cars, four cars, and then bang. So if I'm able to see that so much earlier, I'm able to take preventative action. So I'm able to slow down in a manner that the guy behind me doesn't hit me either, right? So those are two kind of just very simple type scenarios that if you've the right sensors, you're absolutely going to be better than a human driver.
Mark Fielding If you, one of the experiences I've had of late, I didn't know this was a thing so I actually looked it up, it's like phantom traffic jams and the secondary effects of what you're talking about, sometimes unfortunately this technology saves lives isn't enough to move the needle to get people to act. But if you say this technology can save $1,800,000 billion a year in man and woman hours, then everyone goes, okay. And actually removing traffic jams and making traffic flow quicker.
more smoothly is a massive value add to society as a whole, on top of removing all these unnecessary deaths, isn't it? That's another
Barry Lunn Absolutely. and smart city applications. Governments are getting involved heavily. There's so much investment going into smart cities and our technology has been adopted by a number of companies in that space because it's not just what's on the vehicle, but actually being able to watch intersections and adjust in real time to what's actually happening and giving real time alerts and actually able to do it through V2V technology where you're able to send signals and V2X.
to warn vehicles in advance as well about different things, about managing traffic flows. It's incredible. And you're right, it's not just restricted to the vehicle, it's around that. The other thing that people kind of underestimate about like outside of the human devastation of road deaths and that side of things. You mentioned it, people are very motivated by costs, right? And the cost to the global economy from car crashes is incredible. In the developed world, it's about
What crashes cost
2 % to 3 % of GDP is lost to car crashes because of ambulance services, because of all the things, the impact in terms of efficiencies on the road, all of those things. So those are real numbers that have been measured. In emerging economies, it's even higher, right? It goes up to 5%. So those are trillions, right? That's the cost you're getting into to the global economy. So there's an economic factor.
And by the way, I learned that probably a little too late. didn't know the automotive industry as well as I should have when I started the company. I learned a little too late that the economic factor do not ignore it because it matters an awful lot because of course automakers want safer vehicles, right? But they also want to survive. So if you can't build this type of technology at a price point that it can go on every vehicle, you're not going to have success. I can, you know, I can be, you know, I'm so clever and I've got the best sensor, I'll have no impact on road deaths or car crashes if I don't have the best sensor that every carmaker can access. So economics is unfortunately probably the key driving factor of the rollout of the technology.
Jeremy Gilbertson Yeah, economics and as you, as you referenced to supply chain and availability and how easy it is to pull this stuff in. The thing that I always tend to think about with a lot of these new technologies is you have this great sensor that operates from a radar perspective. So it's deeper information. It's based on range and velocity and predictive technology. So great information is getting captured and a lot of information is getting captured and it is processed in a way to make real time decisions.
in key critical moments, right? So I start to think about with all of this information coming in and landing in a car, how does that affect data storage, data access, processing? Like, what does that do to that side of the equation?
Mark Fielding You love that side of the equation, Jeremy.
It is all about the data
Barry Lunn No, you're 100 % right, right? It's all about the data, but it's about as much, it's as much about the data and the volume of it as the good data. So filtering the data is every bit as important as collecting the data and collecting the wrong data. So we spend a lot of time focused on that. So for example, cameras, that's where they're weak.
right, because they, you know, you can unlock an incredible amount of semantic data from cameras, which is really interesting, but they also take up an awful lot of bandwidth, right? And they take up an awful lot of storage and it can be very hard to discriminate the interesting part of the camera image. Radar is actually an awful lot better at that. So when you put the two together, now you can start to bring in filtering on the edge, because this is extremely important and where edge compute becomes very important so that your only collecting important data. So companies that have been very successful in this space is people like Mobileye, right? So you have a AI chip on camera. And what they're really good at doing is they basically got a map of the entire world that they've been able to gather from their customers driving around, right? And basically that's because they were able to discriminate the good data from the bad data, right? And that was, and obviously because it's a very specific thing of a map and what's you know, the road edges and ⁓ positioning yourself off of maybe some signage. That was a ⁓ clear thing that they could discriminate against. With our technology, now we're able to actually just gather data from incidents. So because of the things I talked about, what, and you start to think about, well, let's say the computer tells you to slow down. So that's a big part of the technology. Something comes up in front of you and says, slow down. And if you don't, we start to slow you down.
But then gathering that data is very interesting because why didn't the drivers slow down, right? Or why did they override us when we gave them some prompts? So there's a huge amount of data that can be gathered from that, but it's good data rather than, and you often hear in the space of automotive companies, OEMs, especially saying we've driven, you know, hundred million miles of data or trillion miles of data. It's nonsense, right?
most of that data is useless, right? Like most of that is the most mundane driving around in circles or going from mom's house to the shop, right? Like that's useless data. What you really want is the critical data. So when you focus only on the edge cases, you start to get a lot of really good data because you're right, that pipeline is full. There's not a lot of space on it, right? And when you put an Nvidia GPU into a vehicle,
You'd be amazed how quickly people like us start to fill it up, right? Because we all think our senses are the most important. So that discrimination is a huge part.
Jeremy Gilbertson That's a, that's a, that's a ingrained rule in humanity. It's something called Jevons paradox. It started back in like the industrial revolution, the more capacity you have for something, the more we're going to fill it up. Right. And it's just this thing that goes, it goes up and down, but
Mark Fielding Just like a house, isn't it? You just buy stuff, fill it up, got to empty some space, put some else in its place, yeah.
Barry Lunn I know guys, I think LA's traffic could be sorted with just one more lane. Just one more lane and LA would be fine.
Jeremy Gilbertson Totally. No, so I appreciate your, your, your answer to that last question. So succinctly, think that, so that, just to clarify the data that comes in from radar is less bandwidth intensive than pictures and video, right?
Barry Lunn Yeah, exactly. You can imagine because especially with the velocity equation, right, I can discriminate an awful lot of things that I don't need to gather.
Jeremy Gilbertson And then, and then also making sure, like you said, you know, getting rid of the data that you don't need and really getting the good data and feeding that into the decision-making machine of this, of this chip that, that lives in the car.
Barry Lunn Exactly, because the camera, you think about how you train a neural net, right, so if you want to understand even the velocity of other vehicles using a camera, right, you basically have to take a frame of the visual and then another frame and the distance between the two vehicles. You then have to gather the time synchronized data from the vehicle, right, and you have to gather data from the canvas to understand the acceleration that was on the ego vehicle, your vehicle.
at that time. That's quite a lot of information just to find out if you've trained it well enough to understand the velocity of other objects, right? Whereas radar since World War II, that's what it's really good at. It's able to say it was at this exact range of velocity. So now I've taken away all of that and now I can just focus on what the camera is really good at, right? Which is unlocking the kind of semantic data like where were the road lines? What was the speed limit? I can read the road sign. What was it a red light or a green light?
So that's exactly, you've got it in a nutshell. It's identifying what they're good at and disseminating it down to that and using as little bad data as possible.
How the regulators are reacting
Mark Fielding ⁓ Moving on from that then, so you have all this data, the way you've described these systems, they're incredibly efficient. So on the Proviso website, you're talking about OEM. So essentially this is your technology that you, Proviso can put in any car on the market. How is the car industry reacting to you and others doing what you're doing? Firstly, yeah, that's the first question and then on from that.
Barry Lunn Yeah, so how they're reacting really well now. And I say that because everyone liked what we were doing from the beginning because we could do some really interesting stuff with radars, right? And we could show them really good visuals and things like that. But as the technology has matured and as our kind of unique IPs have matured, the path to low cost access to the technology has become clearer. And that's why I'm saying
Now the reaction's an awful lot better because now they're kind of okay, this is accessible. This is something we can do. And another part of that equation is, and by the way, I got this wrong in the beginning where I over-focused on safety without thinking about the cost and how important that is to them, right? ⁓ So I think that's where the reaction has been very good. The other thing that's happened, and you always need a bit of, know, so that's the carrot, right? The stick.
NHTSA, right, in the United States, the regulator in the United States is typically the United States has been behind Europe and the rest of the world in terms of enforcing safety regulations on cars, right, because Americans have a very strong ⁓ opinion on cars, right, they love their cars, right, and it's
Mark Fielding And they've got no corners.
Barry Lunn And they've got no corners. That's true. That's so true. But certain things have happened over the last number of years that made Nitzel go, hey, we can't just let this industry regulate itself. And they brought in ⁓ a ruling just recently. And from September 2029, every single vehicle that goes on the road in the United States is going to have to have a technology like ours. That's just, whether it's ours or some competitor, they have to.
FMVSS 127
It's now law. So it's FMBSS 127. It's very interesting. And it's an emergency braking ⁓ rule. And it basically says, at certain speeds, you have to be able to see a pedestrian walk out between two cars. Quite complex scenarios. Incredibly well-written. It's a great step forward from a safety perspective. But of course, it's in the industry, really, right? Because they were slow in adopting imaging radar. But now it's kind of like, ⁓ we need it.
And I'm out there and they're all, they all go, God, that's going to cost me a couple of hundred bucks per vehicle that I'm going to lose from my profit margin. Margins aren't very high in automotive as we need, as we know. But I'm out there saying, Hey guys, we've developed a technique to do that and pass FMVSS without the complexity that you've been told you need for that. So I'm going with the right story to them right now, where you have a headache, we've solved it and it costs less than you thought it was going to.
So we're pretty busy, I suppose is one way of putting it right now.
Jeremy Gilbertson on that. from it from this being something that you license and work with existing carmakers to incorporate your technology into their vehicles, is it a is it a process of like, okay, we're going to figure out how and what these decisions and how the inputs and outputs go, but then translate that how does it How do you translate that kind of thing to the UX inside the car? That just seems like crazy complicated.
Pushing it onto the tier ones
Barry Lunn Absolutely. we saw that in the best way possible. We push it onto someone else. That's not a because it can't be all things to all men. And also, OEMs take very seriously how you as a driver interface with their vehicles. So they like to have strong ownership. Even if they have a tier one, which are the big suppliers like Bosch, Denso, big, big, companies supplying the system, they have a very strong influence on what what the user sees and how they interface with that vehicle. So what we do is we basically develop a ton of microservices, little modules and even in our AIs and our machine learning, they're all kernels, they're all little things that build up. And what we do is we offer them as a library. essentially, so we license even on the software defined antenna technology we develop.
We licensed that to a tier one and they can build their radar based on our technology. We're very comfortable with that. With none of the perception type software, if that's what they want to do. And then we can work with their OEM on the perception side. So we build all these little microservices, we make them available, but they're almost like building blocks of what the system needs to do. And it gets rid of the kind of hard bit of dealing between the sensor and the AI. ⁓
And essentially they build on top of those and make those decisions. So because we all know OEMs all see the world in a different, all the car manufacturers have a different view of the world. know, BMW, it's all about driving, right? So, you know, they're not going to go straight to, hey, autonomous, here it is. You know what I mean? And that day, so that won't be the ad for them. So they'll make those decisions and whether it's gentle nudges, whether it's things on the heads up display or whether it's pure play interventions.
that will be their decision and we just build these little microservices so they might say need 10 things from us or they might need two and we're very comfortable whichever one it is.
Can we see it?
Mark Fielding It makes a lot of sense. Can we see it? Do you think that before the show, maybe you could give us a little tour of what's happening behind you. Everyone's seen the cars behind you. Any chance of that, Barry?
Barry Lunn I have to be very careful here, Mark. me see if I can walk around a bit. Why I say I have to be very careful. This industry is... I worked in aerospace, I worked in defense, I've worked in the space industry. I thought those industries were secretive. My God, this industry is so secretive. But let's go for a little walk around and see how far I get before someone...
Mark Fielding Don't put yourself in danger Barry, don't put yourself in any kind of...
Barry Lunn There's no way I'm going to get knocked down anyway walking around this place. So this is our HQ right in Shannon in the South West of Ireland. And this is the future mobility campus test bed. And this is where we've got a bunch of these really cool vehicles here. There's a BMW here and a Jaguar Land Rover. And this is absolutely peppered with our technology, right? So the units tend to be bigger obviously as we're developing the technologies. ⁓
And then they get smaller as they get more integrated into the vehicle. So this is one of our base here. can see the vehicles. You can see some very interesting other technologies in the space as well. And then you can start to see the kind of future units that are sitting around here. And you can see that the technology just keeps kind of getting smaller and smaller. this is where you can see our friends, NVIDIA here as well on the side of the vehicle. So we work very closely with NVIDIA.
So that's the kind of fun stuff we're doing here. The lab is up here. I'm definitely not going to be allowed to walk in there because it's secure environment, ⁓ but I can certainly show you there's a little bit of activity going on in there. And this is where, know, software is uploaded and firmware onto the vehicles and even got some friendly engineers, is great to see. And then in here actually is control center. So this is where we connect when we do vehicles out on the roads here and driving around. And Janet, because this entire zone is a ⁓ testbed, right, beside the airport here. ⁓ So this is the control center. We can watch everything going on from here, see all that data in real time. And obviously, that allows, and obviously, all the engineers can port to that at any point. ⁓ So yeah, so that's a lot. And this is the boardroom where absolutely nothing gets done. So I won't even.
show you that. So yeah, so that's a bit of it there. And obviously we welcome you and your listeners to come down here and we'll give you a demo drive. We're the demo guys, right? That's what we pride ourselves on, that we take people, oh, I'm not getting my cable caught there. So that's interesting that's happening today. But we pride ourselves on, we take people for drives every day, right? And it's a really important part of what we do is showing the technology because that's,
That's the best way to get rid of the nervousness that a lot of people feel around autonomy.
Jeremy Gilbertson Awesome. Thanks for taking the time to walk through it. Our listeners always seeing like things in action beyond just like, like a essentially a zoom screen, gives a little bit more of like, wow, look at all this really cool stuff they're doing. And I want to call out a couple of quick.
Mark Fielding That is for us, Jeremy, don't lie.
Memento mori
Jeremy Gilbertson It is a little bit, it was a little bit for us. ⁓ so one quick thought, I love the gentle nudge kind of thing. Cause cause you know, lot of stuff we talk about here is even though it's like hyper technological, it's still at the core is built on trust, right? It's built on establishing trust, right? So by doing that a little bit at a time, and it's like, I've been in this car for a while. It's treated me right. It's probably saved me from a big crash once or twice.
I feel pretty good about, you know, putting my hands up and, and, you know, letting it run in certain situations. So I think that's amazing. I know we want to be mindful of time. want to, I want to ask you, so you deal, you, ⁓ you navigated a question from our previous guests very well. And this question can be on any topic, anything you want, ⁓ free rein, like what is, what is spinning around in your curiosity engine right now that you would love to hand off to our next guest?
Barry Lunn that's interesting. And maybe I'm going to almost pass the ball on what I was asked, which I thought was a really, really cool question. It brought me to that place of now I'm in the stage of my life where I'm thinking of impact. And I'm going to be quite morbid for your next guest and ask ask them what will be their legacy when they're gone? What will the world say?
Jeremy Gilbertson Memento Mori, right Mark?
Barry, that came up in one of our other books. We were reading Clear Thinking by Shane Parrish and it talked about, you know, legacy and, you know, memento mori and that sort of thing. Well, Barry, this has been a fascinating conversation. I feel like I understand what you are doing so much better than I did when I first kind of dove into your stuff that you very succinctly kind of ran through everything. The takeaway for me on this one is the radar technology.
and the ability to actually get the right data into the car at the right time to transfer that into real time decision making. I think that's amazing. Mark, any closing thoughts on your side before we kind of do our formal.
Mark Fielding So last week, me and you, especially me, raged against the Apple AI machine. we, I was particularly disappointed, almost disgusted on how the AI was, was doing things, which was so moronic, so unnecessary, so detrimental to the human condition and then being packaged and sold as a lifesaver. And then we speak to people like Barry today and the technology is being used, the AI is being used to save lives, to change the dial and to do interesting, important things. And that fills me with joy and it wipes away some of my anger against the people who are using it for surface level ⁓ improvements.
Jeremy Gilbertson Well, here's the irony in that is that, you know, I drive down the road all the time and I see people like with their heads in this, like 55 miles an hour, heads down looking at this. And it's like, all right, here's the thing. We're not going to get people to stop doing that. Unfortunately, it sucks, but we are not going to wholesale convince everybody that if they don't stop looking at their phone while they're driving, you know, they're going to kill somebody. But
If we create a technology that offers another solution to that, I think it's amazing. And I think what you guys are doing is awesome. And I can't wait to see this like sprinkle out ⁓ into the world. Thank you so much, Barry, for joining us.
Barry Lunn It's been an absolute pleasure. So much fun. Thank you.
Jeremy Gilbertson Again, disruptors and curious minds, this episode has been brought to you by Human Conversation, Meaningful Human Conversation. If you'd like to hear...
Mark Fielding Rumor that next week's show is going to be brought to us by vinyl.
Jeremy Gilbertson Potentially potentially, but if you need to learn more about meaningful human conversation find someone else in 3d approach them ask them a question and really listen so ⁓ Next week actually we might yet. We've got this book club guys If you haven't done the book club yet, we read books we unpack them. We'd love to read them with you We're reading Michio Kaku's Quantum supremacy right now that talks about the implications and potential of quantum computing. We're learning about it bit by bit
And ⁓ you know learn about it with us if you're not a book reader Let us be the book readers and you can listen to the show. So mark closing thoughts and we'll get out of here
Mark Fielding Tell them about the next book, Jeremy.
Jeremy Gilbertson my gosh, you go ahead, man. You know more about it than I do.
Mark Fielding So on the book club front, we've decided on next book. It's going to be the second book in our book series called Nexus. And this one is by Yuval Noah Harari, The Nexus. That's going be our next book. So you've got a couple of weeks to get your orders in and come and read it with us.
Jeremy Gilbertson This book talks about the history of information networks as they run through society. So as you know, he's written a bunch of other books that, that take us through history, and looking at different lenses through of history. This is going to be fascinating because the future is information networks. That's all it's going to be right. Electrical signals bouncing back and forth, working with each other to, provide outputs and such. So anyway, my rant is done very thanks again for joining us. Marco post some great notes where everyone can learn more about what you're doing.
Mark Fielding keep thinking on paper. Bye bye.




