Navigating the Future of AI and Autonomous Vehicles with Gabe Klein and Alexei Andreev

I drove through a red light at 70 miles an hour.

To get to the next nine is very, very, very difficult.

It's less a technology issue and more of a public policy issue.

We've got to have a new definition of safety across industry and government.

I don't rely on anything when I make my decisions.

Left or right, things don't happen instantly.

It's got to be better than a distracted human on their phone.

Policymakers should set up the bar and the bar should be what is good and enough.

Welcome to another episode of the Venture Capital Podcast. Today we are talking about AI and autonomy and vehicles and the future of transportation. With me and John today we are pleased to have two awesome guests. First is Gabe Klein who led US Joint Office of Energy and Transportation under President Joe Biden, driving America's push towards electrification and clean mobility. He served as Transportation Commissioner in Chicago and Washington D.C. where he launched two of the nation's first large scale bike share systems, expanded protected bike lanes and prioritized people. First, urban design. Earlier in his career, Gabe was VP of Zipcar, National Director at Bikes USA and later founded CityFi, advising governments and companies on building smarter, more sustainable cities. Thanks for joining us, Gabe. Our other guest is Alexei Andreev. Hopefully I pronounced that right. Is that right?

Yep, perfect.

Our other guest is Alexey Andreev who is co founder and Managing director at Autotech Ventures where he invests in hard science startups in both hardware and software with a focus on companies building durable advantages through deep IP and strong network effects. Previously, Alexei led the Silicon Valley office of Harris and Harris Group and before that worked with Draper Fisher Jervison on investments in advanced manufacturing, quantum computing and alternative energy. He holds a PhD in solid state Physics from the Moscow Steel and Alloys Institute and an MBA from Stanford and has served on the boards of several companies including D Wave Systems, Adesto Technologies and others. Alexei brings a sharp perspective at the intersection of deep tech, venture capital and the future of innovation. Thanks for joining us, Alexi.

Thank you for having us.

So to kick things off, I would like to maybe just talk about what is the current state from your both of your perspectives on AI and autonomy for vehicles. You've got, you know, it's one of those things that we've been promised for the last, I feel like for 15 years and yet still don't fully have. And I'm curious, maybe the first question is, is why, why, why, you know, we don't have full autonomy vehicles today?

Well, I can give you my, my perspective. I've been following AVs for, I guess, about 15 years, pretty closely. In my book Startup City, I actually have a picture out at Google X before it was Waymo with the Panda car, and Sergey was taking us for rides in it back in the day. And I was super impressed with that vehicle back then. You know, I think Waymo now, Waymo is definitely at the front of the pack. I think what everybody's realized is getting to, you know, 95 to 99% autonomy. You can do that in a certain period of time. Then getting to 99 point, you know, 99987, whatever. It just takes a lot longer. That perfection is hard. And, you know, you have people like Elon Musk saying, look, you know, we can just use vision. It's as good as a human. The thing is, it's gotta be way better than a human, particularly these days, a distracted human with the smartphone. So I, I think none of us should be surprised. I got the same thing with the, with the charging stations at the U.S. joint Office. Like, well, you just passed the bill at the end of 21. You spun up the joint office in 22. Why don't we have chargers on the ground 24. I'm like, well, these are construction projects and you gotta do RFPs. And you got. Right. Like, things just take time. But we as Americans are used to, like, instant overnight. And then the last thing I would say is it reminds me a bit of smartphones. Like, I had a smartphone in 2002, it was a Siemens. It had a big screen. It gave me headaches when I would hold it up to my head for hours. And then the iPhone came in. 07 changed everything, right? So just because something takes a while doesn't mean it's not going to be revolutionary and turn the world upside down. It's just it. Things don't happen instantly. And technology, as we all know, is very iterative.

So I want to go a little bit deeper on that. And, and I think you make some great points. And, and part of the reason I'm excited for this podcast today, particularly around this, is we've got Alexi, who is, you know, a little bit more on the investor and deep tech side of things, and you're a little bit more on the, the public policy, which I think both of those things need to align for autonomy to really, really take off. You mentioned that it's gotta be better than humans, but, like, you even admitted, right, that it's gotta be better than a distracted human on their phone. And so why is that? Why does it have to be Better, like substantially better. Why can't it just be a little bit better? Right? Because if it's just a little bit better, we would drive down all kinds of accidents and other issues. We'd all be better off, even if it's not like a hundred percent perfect. And maybe my assumption is that there are, it's less a technology issue and more of a public policy issue, that that makes that a requirement.

Yeah, Peter, I can, I can jump in or we can switch back and forth. And if you don't mind, I will geek out a little bit. I agree with you that policymaker should set up the bar and the bar should be what is good enough. And whether it's 99%, 99.9, 99.999. How many nines do you need? Which will satisfy public opinion and policymakers and legislative bodies that something is safe? Because there's nothing like 100% safe device or 100% safe therapeutics, or 100% safe anything. It's never 100% safe. So how many nines do you need? And so it will go back to Gavin. I'm sure he will provide additional insights. But as a former technologist who turned to the dark side, I think your listeners might be interested to discuss why getting those extra nines are so difficult, specifically for autonomous drive. And so when people describe different processes, stochastic processes, random processes, when we're flipping coins, something which is typically used is a normal distribution. So if the processes don't feel each other, they're independent. We flip enough coins, we'll get a normal distribution. And for normal distributions, we know what the standard deviations is, how many examples we need to test and situations to explore in order to give you 2sigma, 3sigma, 4sigma, 6sigma, etc. The challenge with autonomous driving, it is one of so called strongly correlated systems. And so what you do impacts everybody else around an intersection, and what people do around the intersection impacts you and your perception, your path planning, your collision avoidance. And these pieces talk to each other all the time. You turn left, somebody else will react one way. You turn right, somebody else will react in a different way. And so for those so called strongly correlated systems, we're dealing with fat chaos. It's not like normal distributions dropping very quickly. And you know, after a thousand samples, you have a pretty good understanding of how this particular system behaves. For thick tails, you add more examples and more examples and more examples, and you still cannot tell whether it's two sigma, three sigma or six sigma because the tails are not getting Narrower with time. And I think it represents the fundamental challenge. And that's why we all hoped to have autonomy in five years and then in 10 years and in 15 years. Because to get to the next nine is very, very, very difficult and very expensive, given the very fundamental nature of the problem. It's a system with memory and it's a system with strong correlation. And those systems, you just don't know how many times you need to test against different scenarios to get a sufficient level of confidence that the system behaves as needed. And unfortunately, we're dealing with something very complicated from the mathematical standpoint. And Gabe, now passing to you. I geeked out.

Yes, well, I, I love it. And, and I'm not a tech geek, but I so appreciate what you're saying, because that's okay. That's the tech side. The good news is we have a technological leader, right? And I'm a little biased. I advise Waymo for about five years through my company, cdi, but I've ridden in a lot of vehicles from Singapore to California, you know, and, and I've seen the differences. I've experienced it as a. As just a human being as well as an investor at Contin Alice Partners, as well as a public policy person. And when you have a leader like that that's already basically achieved what we need to achieve for the most part, you know, with 25 million miles driven, according to Swiss Re insurance paper I was reading, and two liability claims, we're already significantly, you know, significantly better than a human. And that's where we need to be. It's never going to be perfect because you're going to have those edge cases that happen. I remember I was in Singapore and I was in an economy vehicle and a wild boar walked out in front of us, and the car was totally confused. And I was talking to the guy in the back seat on the laptop. He's like, holy shit. We've never seen a wild boar like. But the beauty of it is every Newtonomy vehicle then the next day knew what a wild boar was, right? Because instantly the entire system soaks up that data. The other thing I would say is the, the definition of safety is a problem. You alluded to this, Alexi, that you've got in the US we think a lot about the safety of the people in the vehicle. You even look at the, at the Nipser, you know, crash test. It's also, as he alluded to the people outside of the vehicle, the cyclists, the pedestrian. You remember the cruise incident in San Francisco, right? So who ultimately decides what's safe. It's not the government, honestly, the federal government is not doing that. It's the consumer. It's public opinion. Crews shut down. And it wasn't because the government shut them down, they shut themselves down. If you're not safe, nobody's going to ride in your vehicle. It's not like a tech system where you can put out an Alpha and a Beta and iterate and test and hope for the best. Like just imagine in 82 when the airlines were deregulated, that a company came along, unnamed company, we'll call it Slate or Tesla or whatever, and said, hey, we're just going to put planes in the sky and look, they're not completely tested to be, be 100% safe from 99.99. But we're gonna, we're gonna learn as we go and it's gonna be in revenue service. I don't think that would fly, no pun intended, with the public. Right. So there's the definition of safety. We gotta have a new definition of safety across industry and government. And there's liability. You saw the full self driving liability claim last week or two weeks ago. 30% of the liability got applied to Tesla. The billion dollar settlement, their part is 332 million. You know the insurance companies play a very important role here and the trial lawyers. So yes, there's a tech side, but there's the reality of what happens on the ground. And am I going to put my daughter, my 11 year old daughter who I love to put in a robotaxi to get to school with all of her friends in the Tesla was with full self driving when my personal experience, I drove through a red light at 70 miles an hour with both of them in the back with full self driving on. No, no, I'm not gonna do it. Would I put them in a Waymo in a second. So you tell me what's sick?

Well, there's, I think Peter.

Peter.

Oh, go, go.

I, I know we're supposed to discuss public road autonomy. I think at some point we may spend a couple of minutes on off road autonomy and specialty autonomy applications. Because if you think about agricultural autonomy, driving tractors in the field, if you think about bulldozers, if you think about hauling trucks, if you think about all sorts of off road autonomous applications due to simplified driving environment. You don't have children playing around, you don't have traffic lights, you can stop and start at any point. Plus there is a very clear ROI for the end user, partially because of liability insurance and litigation. Every time you have an accident. So mining companies, agriculture, they want to get all the people away from construction site. You don't want to have people on construction site. You're insurance goes down, your litigation costs go down. I think we should not be forgetting about those use cases of autonomy as well. And I think what way more accomplished and I think what PASLA is accomplishing, they are and crews as well, they give birth to a lot of off road autonomy startups which learn from the best and managed to adopt highway autonomy technology for smaller niche, but much closer to revenue and much closer to clear ROI use cases. So maybe we can spend some time down the line expanding what autonomy is and say public road, non public road, off road, on road, da da, da, da, da, da.

Yeah, no, I think, I think that, I think that'd be great. Especially when we talk about like, where are the opportunities? But because my impression is that there are a lot of really interesting investment opportunities in that sector. The one thing I was going to mention about liability that I think is another kind of wrinkle on this that's challenging is safety.

For who?

So like, you know, who gets, whose safety is optimized in these vehicles? Is it, is it the person in the vehicle? Is it the other person's vehicle? Is it the pedestrian on the street? Which I think adds like a whole other level of figuring out what safety safety really means. But I guess maybe the question then, Gabe, is like, why do we not have a standard for safety? What's holding us back there?

It's a really great question. And first of all, I'll say people have very particular feelings, particularly industry people, about this. And it doesn't necessarily have to be about autonomous vehicles, just street safety. And I've seen people go from being like advocates for having as little regulation as possible and like, look, you know, every man or woman or kid for themselves to being real advocates for street safety, for pedestrians when somebody in their family gets hit, right? The reason we don't have a definition for safety is because, because you look at a organization like nhtsa, they're focused on the vehicle, right? The physical vehicle. Now it is interesting, you look at like a, a vehicle like the cyber truck, which is not legal in most places in Europe, but it is legal here. And the reason is that it's not safe, it's not considered safe in Europe. Say, you know, it is or isn't, but it's not considered safe for users outside of the vehicle because of the combination of weight, the angular aspect, somebody gets hit at like 12 miles an hour. They're dead right. Getting back to AVs though. So there's the physical aspect, there's the tech stack. We don't have a rating system for the tech stack.

Really.

You know, like you can have somebody from Neuro or Zoox or Elon from Tesla, Kedra from Waymo. They could all get on a panel and argue about whose system is better. But it's up to us to decide. I think there's a real opportunity for a rating system. I think we should have it. I think it should be made up of the people that are rating. This should be people from industry.

Perhaps government.

Maybe it's maintained by a third party. But you know, like we have Consumer Reports, we have iihs, you know, like there are organizations out there. Why don't we have a rating system? I think it's a matter of time and I'm an advocate for that.

Yeah, no, I think that's really interesting. Any thoughts, Alexei?

Yeah, yeah, couple of factoids. Factoid number one, we keep hearing rumors that some of the best engineers from Cruise are now joining insurance companies because insurance companies figure out they will need to underwrite insurance for autonomous vehicles and there is not enough statistical data to produce actuarial tables. You need to start offering insurance now. And the only way to do it is to, as Gabe mentioned, to qualify and certify the quality of a tech stack. And so former developers are now creating those virtual test rooms and test intersection and test facilities to exactly what was just mentioned, that Zoox stack is better than this and Waymo is better than this and this stack is in between. And therefore I am underwriting based on simulation, which is kind of interesting. Never before insurance industry underwrote based on simulations and now they seemingly willing to do so. So that's, that's the first factoid. The second factoid, I think the lows will differ drastically around the world. I think our attitude to safety in the United States is different even from Canada. And I think it's drastically different what the safety is in Japan and Europe. So you have cultures more focused on well being of the society and you have cultures more focused on well being of individuals, more socialists and more libertarians. And I think those countries will be applying completely different criteria when they judging which stack is safe and how a particular vehicle will behave in a difficult situation. Will it try to protect people inside the vehicle, sacrificing people on the bus stop? Or it will sacrifice two people inside the vehicle saving 10, 10 children from a summer camp instead? And different societies will be making different moral judgments and will be applying different criteria.

Can I just jump in on this? Because I think Alexi hit probably the most important point of this podcast, perhaps, which is that in the US we live in an individualistic society. And he hit the nail at in Europe, Japan, other places it's a collectivist society. So when we think about rating AV tech stacks, it's got to be universal. I think it's got to be global. Now, if we want to say, hey, our people are going to be less safe, that's fine, let's make that decision. I think the insurance companies are going to focus on dollars and cents, right? Sort of like our attitudes towards climate change are very different no matter who's president, but definitely when a Republican president is in extremely different, like, you know, the opposite of Europe, the reality is we're all going to suffer the same consequences and the insurance companies are going to have to pay for the damages. So it's been interesting to see what's happened in Florida with liability. You know, my aunt's place in South Palm beach used to pay like 2500 a year for insurance, and now it's $22,000 a year in the last few years. So. So look, you can say climate change isn't real, but the insurance companies don't give a shit. They're going to, they're going to charge you what their risk is. And I think it's going to be the same thing with AVs, right? So if one company, their risk profile is, you know, like, let's say there's a score 1 to 100, and let's say, you know who at 95 to 93, you have some, some of these other guys at 65, it'll be reflected in the cost to the consumer.

Yeah, yeah.

And then they can make their own decisions.

Yeah, yeah, I do. The insurance companies are the ones that will ultimately be the, the ones that decide what safety kind of means with dollars. Right. And then even if you self, even.

If you self insure, even if you self insure, at the end of the day, it's still dollars. Unless we're going to change our society and say you can't sue and then people just aren't going to ride in the vehicles. Right. But sorry, John, you're going to say something. I was just saying you're the, the, it's, to me, it sounds like the profile, you're the, the rating system you're talking about will just be the, the end cost to the consumer. Well, you could look at it that way. I, I'M thinking more about having a whole series of criteria. You know, yes, safety is probably the most important aspect, but there's also like, what are the impacts to the transit system? What are the impacts to other city services? I was talking to a friend of mine who ran the San Francisco GOT until recently. He's telling me a story about how these cruise vehicles would go into an intersection. And if there's a traffic control officer, right, they would brick because they didn't know what to do with the traffic control office. And it would take 20 to 30 minutes to extract that vehicle from the interception. Get somebody out there, deal with it. There's a cost to that for everybody who's trying to travel, right. As well as their safety issues and lots of other things. So there's a whole bunch of different criteria, I think, that can be applied. Just like there would be if you're buying a vehicle. You know, you look at Consumer Reports.

Yeah, for sure. I also wonder, too, if part of the reason why there hasn't been more uptake from consumers on this is that even though I logically know that it's safer or should be safer to let the. The car drive itself, the idea that I'm not in control, right. It makes it more terrifying. I think it's the same thing for people that are terrified of flying. It's not the actual, like, art, you know, the fact that you're in the air, that's terrifying. It's that you have handed over control to somebody else. Um, and that. That makes people really nervous. And so I think that's another big, big factor that is just really hard to get our arms around. And it's really easy to point to the edge cases where it didn't work and use that anecdotal evidence as supporting evidence for why you'll never get in a, you know, an autonomous car, even though it's still safer, right.

Until you do. And then you're like, oh, my God, this is amazing. Like, if it's right, if it's a good quality service, good tech stack, and you're like, oh, my God, it's so much better than, like, the Uber driver doesn't smell good and is, you know, arguing with me or whatever those issues are. And then people are like. Like, there was an article yesterday about people. I forget where, but people are like, cancelling on Uber until they get to Uber wayo because, you know, Uber is the demand generation for wh mo in a few cities. I think it's Atlanta, and I thought that was interesting. So once People try it. If it's good, they seem to love it.

So I'm curious from your perspectives, who's winning the race? Where would you stack the companies that are out there?

It's hard to tell because we see only players in China and we don't know how they measure reliability and false negatives and false positives and false disengagements, unforced disengagements. We have no idea what kind of toolkit they're using for to certify and deploy those vehicles in the U.S. clearly Waymo has the most visibility and what we see. I'm sure it's an amazing use of engineering knowing Waymo DNA from early Google days as well as the sensor stack they're using. Tesla just launched their own vehicles and for the full disclosure, I've been driving Tesla full self driving for a long period of time and I think it's an amazing technology. Definitely induces your mental load when you're driving long distances on the highway. I'm, I'm, I'm blown away how much easier it is to drive for four hours with Tesla full cell drive versus just holding the steering wheel all the time on my own. I don't know if those guys are ready or not. I haven't tried but I think you raised one of the and, and clearly Zooks is a nobody knows what's happening out there and Cruz was a high flying player and then they kind of went sideways and now Cruise is rehiring and people are coming back and so it feels like they will restart the program because General Motors doesn't want to fall behind and they want to keep fighting. In this game we have evolution of ADAS system which are becoming level three plus and maybe touching level four. So a number of traditional OEMs have pretty good solutions. And then the question your airplane flies and the rocket ship flies. Can you start with airplane and transition to rocket ship? Do you have enough smarts and enough technology developed through adas improvements to eventually allow full sail driving or you still remain the very best airplane which will never make this job to to the orbit. I, I think the jury's still out. Another interesting jury is the sensor stat and way more versus Tesla because clearly alone is anti lidar at least publicly. And in our history auto tech ventures would never invested in a lidar company. So we were kind of in a loan scam. But if you really think about the problem car makers are trying to solve, it's not a new problem. We just say it's autonomous vehicle. But all of us and the birds outside and this dog running around. They're autonomous creatures. The goal is to find prey and eat something and be fed and avoid being eaten. And you do it while having the, how to say constrained energy budget. You are not unlimited budget. You have to. You kill what you eat, you need to or you eat what you heal and you need to run after chase and grab and digest and still stay alive. And it's not a new problem. And the interesting thing is different species over three and a half billion years of evolution, they evolve differently. So in some extreme cases you are maxing up your sensor stack and you have bats and you have owls and you have some other creatures which have pretty small brain but absolutely amazing sensor stack. And in other case you have human beings. And our sensor stack is pretty bad actually. It's shockingly bad. But you have a lot of processing power. And so the trade off between a lot of sensing plus a little bit of processing power versus a lot of processing power and a little bit of sensing is not new. And the evolution created species in one way. And you have an opposite branch. And I have a joke that you have dolphins and they have both, but they're water cooled. What dolphins do, you need to leave in very forgiving environment and then you can afford sonar as well as significant processing capabilities and it will still feed the bill. And therefore to say that lidar stack is intrinsically better than camera based stack. I cannot tell you where manifestations would drive relatively well without any lidars, you and me. And we don't have high definition maps and we just know how to drive. If we're not distracted and we don't have hot coffee pouring on our laps, we're okay. And so the question which approach will win? Which technology will eclipse what the total cost of ownership is going to be? I think the jury is still out. On the other hand, if you talk to industry veterans, they're saying yes, when lidars are expensive, it makes sense to build your stack based on low cost cameras and consumer electronic grade solutions. But if you believe in more slow and the fact that you're compute as well as sensors will be dropping in in price over time, then it's pretty easy to imagine vehicles with superhuman senses. Vehicles not only which can see in infrared and see heat, but can see through the walls with X ray or gamma or terahertz radiation. We didn't involve with those sensors, but we already have sensors at the airports which can do amazing things. And every time you go to X ray or mri, those sensors Are amazing. They're too expensive today. But I think the far future. I don't know who's going to win two, three, four years from now. I can tell you who's going to what the sensor stack will look like 10, 15 years from now. I think 15 years from now every vehicle will have cameras and thermal and X ray and terahertz because all those sensors will be like tens or maybe hundreds of dollars. And if you can fuse all the data together, if you can do low level sensor fusion and create a holistic view of the world, I think it will drastically simplify computational load on your compute architecture and it will avoid those vehicles with superhuman senses. Definitely beat humans on how safe we are and do some really Ernie, warning. Slam the brake before something comes out from around the corner. I think we cannot do it but our machines of the future should be able to do it. I'm a big believer.

I also wonder too if part of the challenge here is just getting vehicles to talk to each other and if, if we could have devices and vehicles talk to each other that is.

Cellular versus It's a curveball.

It's a curveball.

Peter or vdex. Can I jump on?

Yeah, yeah, yeah, yeah, yeah. Go ahead Gabe.

It's a good one because I think, I think it's perfect. You have a spot on. Because we, we agree on a lot but we come from a slightly different perspective and I think what we've learned is there's a big difference between ADAS systems. I agree with Alexi. I have a Mercedes now with some autonomy features. I got rid of my Tesla. I miss autopilot. It was great. You know, as an advanced highway ADAS system it was amazing. It didn't function in the city very well. Obviously it's not, it's not for that. The full self driving I felt was also very underpowered for the city. My car came with ultrasonic. I had a 2020 Model S came with ultrasonic sensors, radar sensors and the eight cameras. Elon disengaged ultrasonic and radar at a certain point because I guess the car didn't have the compute power. I think had a hardware 2 or hardware 3, I don't remember. And he didn't want, you know, this is. Look, the, the radar sensors are as little as five bucks and the ultrasonic sensors are like as little as a hundred bucks. It's not.

We actually sold a company to Tesla called Deep Scale specifically dealing with low level sensor fusion. And the challenge with radar is a multi path reflection and very hard to annotate those images to say this is a human, this is the car, this is a goat. Because nobody knows how to annotate radar reflections.

But here's the thing, right? You're right about that, obviously. And you know more than I do. That's why you need a sensor suite. And you gotta have the compute power in the vehicle to support a sensor suite. So I went to vision only on my, on my car and literally it only worked 80% of the time because I live in the city, I live under a tree. A bird would shit on the front windshield. Actually, often the bird would crap on the left side so the cameras on the left side wouldn't work. And then if the light was shining in the front camera, the front camera wouldn't work. And it would notify me that I couldn't use the features that, that, that doesn't work. You, you can't have an eye.

Full disclosure. If the bird shoots in my eye.

Yeah.

And my second eye, I have sun. I'm not going to drive as well. So for the full disclosure.

Okay, but, but Alexi, you said something interesting earlier about humans and animals being autonomous. Right. And you're absolutely right. This is why this is such an interesting conversation. Well, I don't just rely on my eyes. Well, first of all, I have two eyes. I also have my hands. I can touch things, I have ears, so I can hear. I have a whole sensor suite in my body. So when Elon says it can see as well as a human, I'm like, well, it better be able to do more than that, you know, so lidar, in my mind, you know, having we invested in Ouster many years ago, I, I advised them, an excellent company, big merge now with others. But the lidar is absolutely essential. I think it's absolutely essential. I think vision only is never going to work. It's never going to work because a human wouldn't work if it was vision only. It just doesn't work. It doesn't. It. You need the redundancy of a sensor suite. It's not a cost issue. It's a compute power issue and programming that it takes to fuse those sensors. And I think fundamentally the issue here is you've got folks and I didn't know the answer to this 10 years ago, but you've got folks like Waymo and Zukes, right? Neuro, like we're going to start at level five and that's it, Level four or five. That's what we're going to do. And you have folks like gm, you know, With Super Cruise, Tesla with Autopilot and what they call full self driving saying we believe that you're going to go from level two and it's going to be iterative and you're going to get to level five. And I think they're at a real disadvantage because they make fundamental mistakes thinking that you can do that and iterate and thinking you can rely on a limited set of sensors like just vision. It's never going to work.

The jury's still out gentlemen, but I.

Think we will see.

I disagree, but I love it.

Have us back in like a year and let's, you know, let's talk more where things are at. I, I will bet Alexi a dollar and you're gonna have to have lidar.

Let's, let's discuss five years from now because as, as I mentioned 15 years from now. I agree with you full heartedly. But it's a little bit like with OLED TVs and LCD Liquid Crystal TVs. OLED is such an advantageous technology in any way. You analyze, slice and dice OLEDs, they're better. But traditional liquid crystal guys scaled infrastructure and they build factories and they drop the cost so much that suddenly you have a vastly superior technology essentially remaining niche and LCD guys controlling the market share everywhere in the world except for some very, very teeny tiny highly specialized applications or absolutely high end use cases. And so I think there is a non trivial chance when you can just brutal force with substandard technology. I agree with you, it's substandard. But if you deploy aggressively and suddenly everyone hooked up and it's substantially cheaper, including insurance, including liabilities because alone self insures at some point people will say holy, he controls such a large market share that maybe for some niche, unique, high end applications you will have all the sensor suite you just mentioned. For everybody else vision is good enough. Especially if lawmakers will say, you know what? Good enough is what the loan can deliver. I'm giving you an alternative view of the future. I'm not saying you're wrong, but I'm saying the jury is still out. I think it'll be interesting to reconnect two or three years from now and sync up on this particular evolution of, of autonomous space.

I learned that vision is enough on the highway during the day. I did. It works. I had Tesla for. Well I have had two Teslas since 2018. At night, when it's foggy, when it's raining. Doesn't work. Doesn't work. So I mean even for like trucking.

On the highway, but I don't work as well in night and the foggies. I, I just don't drive. I'm like, dude, I'm not driving. It's too dangerous. And, and maybe if you put criteria, this thing is as good as an average human. Yay. If you put it it's like Michael Schumacher in at. At at his prime plus superhuman senses, then clearly people relying on vision will not make a cut. It's just where the policymakers set it up. And I don't know how it's going to play out well.

And if we get to the point where we have AV lanes, like, like, let's say we have HOV3 plus AV lanes or trucking lanes, that's a different ballgame. Right? Because one of the issues we have is context sensitivity here. And like you, you brought up off road, farm equipment. Totally different context than highway. Totally different context than very complex urban environment or suburban. And so I can definitely see a world where vision only in AV only lanes, even in cities is plausible. I think in the complex environment, nighttime, all those things, it gets dicey. But, you know, and, and, and that's where like when I think of this rating system, I'm thinking specifically like about urban environment.

So Gabe, I don't want to spend a lot of time on this because I want to make sure we have some time for some other topics. But I do. I am curious to get your perspective on why we don't have AV lanes today. Because it seems like that is a fairly obvious thing to have. And almost every single car that comes out today has some level of autonomy built into it. Right. Even if it's as simple as, you know, basic lane keeping.

Yeah. Well, I don't think we're gonna have AV only lanes soon in like an urban context, but I could certainly see it as a mixed like HOV on the highway. The fact is we don't have a lot of, we don't have a lot of VBS right now, so there's not really a need for it necessarily. But you've got like cavnue that's been working on connected vehicle lanes. Michigan, I think you'll have like tolls, like toll lanes, meet AV lanes, meet trucking lanes, meet HOV lanes. In the urban context, you could take a bus rapid transit lane and you could make it a high occupancy AB lane as well for like AB show. You could definitely add AVs, complimenting transit, linking bus stops to train stops. And, and it could be very, very useful. Like there's these Great use cases for autonomy where like if you work with cities, I think cities will step up and say, hey, we're open to giving you your own life if you can do X, Y and Z. Yeah, let's, let's do it. But it's got to be like a collaborative effort.

Yeah, yeah, it's a very good point. And I think there is a lot of cross tokel and analogy between vehicle to vehicle communications and what you just mentioned. If the government can force V2X, V2V standard, it will make roads safer. There is no doubts about it. And in some European countries it's already been implemented in Asian countries too. And so there's a clear advantage. However, the question who's going to pay for it? And so if you want to deploy V2V and V2X, somebody needs to build the infrastructure and power the infrastructure and support the infrastructure. And it looks like some societies are willing to invest because it makes life safer. In some societies they're saying, no, I'm libertarian or liberal or my private preferences exceed the societal needs and therefore I don't want to be taxed and I don't want to invest and I don't want my country to deploy this particular solution. And so, Gabe, I wonder how cities and federal traffic agencies will deal with dedicated lines specifically for AVs. I spiritually want to see it. I want it to happen. I'm getting a little bit skeptical that us will be on the forefront of those deployments. I think we should be relying and our view of the world, at least my view of the world, like us humans, we're self sufficient drivers. I don't rely on anything when I make my decisions left or right. I think the presumption for us should be all those autonomous vehicles will be making their own decisions and they will not be getting any help from anybody else. Like, oh, communicate around the corner what's coming. Like, no, I'm on my own. It's live free or die. A little bit of New Hampshire to you. But yeah, all the vehicles will have to make their own decisions for quite a while.

Yes. Well, I think you're absolutely right. I've been a skeptic of Vita X technology for 15 years. Nobody could agree. It seemed like on DSRC versus cellular and you got the SEC that regulates the radio spectrum for V2X and there's a conflict actually with the FAA. And so there, I mean it's, it's actually a legitimate problem. I think that's where folks like Elon and others are right. Is that like the vehicle's gotta function on its own and within a network of vehicles, right? So cars can talk to each other and particularly in network vehicles can talk to each other. We, we do it now to some extent with peer to peer like Waze and Google and you know where it's like hey, there's an accident here. You let everybody know it'll it's just going to be automated. And yeah, I, I, I don't see cities or anybody paying for sensors to be embedded in infrastructure anytime soon. Unless, particularly if they have to be linked to the stall network. Unless it's.

But Singapore may be different, in Zurich might be different and Stuttgart might be different. It just in you will have very different approach to safety and what safety is and what is acceptable and what is not acceptable. And clearly guys in mining and construction, they perfectly fine placing the base stations and providing centralized control over different vehicles in addition to each vehicle being autonomous or almost autonomous. And so different deployment models, they result in different architectures and different ROIs and different scale up challenges, but also opportunities. But it will stay with Public Road. I agree with you full heartedly.

So as we wrap up there's been a really fun conversation. I'm curious for each of you. What is, I mean this whole podcast is the venture capital podcast, right? So it's like where are the big opportunities going forward? I'd love to hear from each of you. What is one area within autonomy that you think is ripe for disruption or a really interesting opportunity to invest in? You know what, what should our listeners be keeping an eye on? Especially maybe stuff that's not, not obvious.

Alex.

Sure. Thank you. My intuition tells me the, the Highway Autonomy or public road autonomy game is pretty much over for venture capitalists and you have three or four or five exceptionally well capitalized companies and it's, it's like the Game of Thrones. If you cannot build army big enough, you'd better join another army because you have no chance in any frontal collision with, with somebody else who's well funded, well researched and has 10 years behind the shoulders. And so my, my, my intuition tells me the Highway Autonomy. It's too late to enter for any startup or any small medium sized venture capital firm like Autotech Ventures. We are investors in multiple fab companies and so safe AI. One of our companies just got acquired by Promto. We are investors in Agtonomy which provides solutions to agricultural OEMs. They announced a big deal with Kubota and Bobcat and they have many more deals they're going to announce shortly and so turning small and medium sized tractors into fully autonomous vehicles. We are investors in Fox Robotics forklifts and loading and offloading 18 wheelers. And so if you think about all sorts of yard automation, yard management system, ports, how to move all the containers, airports, you, you're moving lorries and moving cargo between planes. So all those niche cases, I think for half a billion dollar VC firm like Autotech, those are attractive opportunities. And they're plenty of, of of of those subdomains where people hear about labor shortages and cost of labor and liabilities and everything else. Another interesting aspect is like more than autonomy itself, but somebody needs to clean autonomous cars. Somebody needs to recharge, refuel autonomous cars. Somebody needs to provision and fleet manage autonomous cars. Somebody needs to insure autonomous cars, somebody needs to repair and maintain autonomous cars. At what point autonomous cars will go and ask for help from like my tire is flat. Okay. What do you do with my tires flat? And so in addition to just thinking about autonomous autonomous, you start thinking what kind of infrastructure has to be deployed and has to be created in order to support wide deployments of autonomous, how to fuel that. Okay. And who's going to do it? And that's where we are actively looking and exploring.

I love it. Gabe, quick thoughts on what you're excited about as we wrap.

I agreed with Alexi, like we invested in Nutonomy back in the day. Those types of investments are not happening anymore. It's the sub components, like ouster, you know, was shot way up, went way down. I've held my shares, you know, went from like it went down to $6. Now it's at $30 today. I think the subcomponents are going to be really, really important. And I think that EV charging, I see T word is, is joining EV charging for these vehicles and the services for the vehicles will be essential. Awesome.

Cool. Well, thank you Gabe and Alexi. Really appreciate having you guys on the podcast today.

Wonderful.

Sounds great.

Thank you for having us.