Autonomous AI in the Real World with Dave Ferguson (Nuro) + Steve Jang (Kindred Ventures)

Yeah. You love chatting about autonomy.

The traffic that almost made one of us late this morning. Yeah. Well, let's, I mean, we're pretty close to your roots here in the valley. So, I thought maybe what we could do is kick off and if you could tell, like, you know, how you got started, the origin story of Nuro and you and your co-founder.

Yeah. Well, it depends sort of how far back you want to go. I did my undergrad in New Zealand and then I came over to the US to Carnegie Mellon to do a PhD at the Robotics Institute, which was and still is basically the Disneyland for robotics nerds. And so while there I got hooked on basically every robot I could get my hands on. And towards the end of my PhD I participated in the DARPA Urban Challenge. So these were these races, these autonomous vehicle races that the defense department held. And in 2007 Carnegie Mellon won the urban challenge, which was basically driving in suburban urban areas. And from that point on I've been pretty hooked in autonomy. I was part of the Google self-driving car project before it became Waymo and that's where I met my co-founder, Jiajun Zhu, the JZ we call him. And then in 2016, the two of us left to start Nuro.

And when we started Nuro, the goal was really, we looked back on the last 20 or 30 years and we felt like we had all experienced this tremendous revolution in how we interact in the digital world. Things are completely different. We had the internet now. Subsequently over the last 5 years, it's revolutionized again. But when we looked at the physical world, we didn't see a corresponding transformation. And we felt like the way that we all exist and live in the physical world is not that different from how it was 50 years ago, even a hundred years ago in some cases. And we believe pretty strongly that that is going to change and was going to change over the next few decades. And we wanted to create a company in the physical AI space and the robotic space that would help bring about some of the really positive potential benefits of that technology and the transformation that that would have. So we created Nuro. The mission of the company was to better everyday life through robotics. And when we started out, we thought very hard about what the first application should be. We knew we wanted to build generalized intelligence and ideally superhuman intelligence that could be applied in the physical world, but we also knew we had to have a really concrete good first application that we could go after, something real and something very worthy for us to dedicate the team to. And so for the last 10 years, we've been working on self-driving, fully autonomous vehicles. Initially, it was in the scope of goods delivery. So we thought if we focused on vehicles that didn't transport people, we could build very defensive, very conservative drivers and really focus on safety. Over time, because the system we had built was fully based on AI, it learned how to drive based on all of the training data that had been given and all of the training data it had been given was expert human drivers. So we ended up learning how to drive like an expert human driver. And so we realized this technology is much more generalizable than just for goods delivery. And so a few years ago we expanded to passenger delivery logistics in both the commercial setting and in the personally owned vehicle front.

Yeah. So maybe to draw, we've talked a little bit about the digital AI side and the physical AI. Maybe to draw an analogy from the digital side today. If you look at the digital AI ecosystem, you have a few layers. The base layer you have the compute and you can sort of throw in data centers. So the power and the compute that is powering all of the AI. Then you have this foundational model layer and this is like the Anthropics, the Open AIs, the Geminis of the space that's building the core foundational AI technology that's generalizable. And then above that you have the application layer. So folks that are building specific applications of that AI like legal agents or customer service agents and so on. Where Nuro is playing in the mobility space is at that foundational model layer. So we have the compute providers, we have the hardware providers and then we have this sort of core generalizable, very very capable general AI that can do the driving and then on top of that we have a number of applications and those can be again like logistics vehicles, robo taxis, personally owned vehicles and the partners that we are working with, we are trying to help enable to build the entire application across the whole ecosystem. And so you mentioned Tesla and Waymo. So Tesla again they have this direct to consumer building the vehicles. Waymo is doing the robo taxis. We are trying to build a Nuro driver ourselves to really focus on what we do best which we believe is all the autonomy and then we are partnering with others to build out the entire product. So for instance we are launching a robo taxi this year with Lucid and Uber. Lucid will be manufacturing the vehicles. So they will roll off the production line with all of the hardware required to be able to be fully L4 autonomous. Uber is owning and operating the fleet. And so Nuro is purely providing the driver part. So every mile that the vehicles drive, it'll be the Nuro driver driving it. And we feel like that's a really good place for us to play because we can, as I mentioned, focus on what we think we can do best-in-class, which is the autonomy piece, and work with partners to do what they do best in class.

Yeah. So we see ourselves very much as a partner to both. So I mentioned on the robo taxi side there's obviously a very close relationship with the mobility service provider. So Uber. On the personally owned vehicle side we want to empower OEMs to be able to provide full self-driving capabilities to all of their customers. And that may be a range from L2 plus which is driver assistance all the way through to L4 fully driverless. But Nuro wants to provide that foundational driving technology that can enable them to build out all of these future products and frankly to compete with the Teslas of the world who are doing a really really good job on the driver assistant side. And so other OEMs feel quite far behind most of them and they're looking for ways that they can provide a product that is as good, that is competitive with FSD but also in a way where they're not giving up a massive part of the future of their company, right? Like the OEMs don't really want to just be producing metal. They understand where everything is going. They generally understand that we're going to have more and more intelligent vehicles. The intelligence in our vehicles is going to be a more and more critical component of the vehicles and the purchase decision. And so they want to make sure that they have a piece of that and that they're not locked out. And we think that by working with Nuro, we can provide for them the ability to have this incredible best-in-class product but still have some of the control and access to the technology and more control over their future.

Absolutely. So the future that we see is one in which mobility services will have fleets of dedicated robo taxi vehicles that are either managed by the mobility service or fleet operators more likely. And then there will also be just like there is today there will be a pretty significant number of personally owned vehicles that are also fully autonomous. And those vehicles can go operate on Uber network or Lyft or others when they're not being used by their owners. And that model, the interesting thing is that right now we're in this transition stage where mobility services overall are recognizing in order to leverage fully driverless vehicles today, someone needs to own a fleet and to operate it. That is actually a deviation from how mobility services generally operate. You know, generally they're very asset light. They don't own vehicles. They leverage people that own their own vehicles and it's a very good model. We're having to take this transition step where someone maybe the mobility service provider themselves is now owning big fleets of these vehicles, but that is a step towards a future where it's going to be more similar to today where there are personally owned vehicles again that can be leveraged by the mobility service. And so it becomes asset light again. And so you've sort of got this step in the middle, but I do believe that there will always be some fleets that are owned and operated that are dedicated.

Yeah. Physical AI obviously with atoms takes a long time, right? It's not software. It's not a software roll out. But I think for it to work, it needs to be sort of driven by the network that stands to control all the customers and all the customer accounts, right? And so, you know, in along that vein, I mean, you have logistics out there, right? And clearly like all these trucks that are delivering packages at random times in front of our homes and offices.

That's right.

Yeah, I think more broadly we have something like two billion vehicles in the world today. Two billion vehicles and obviously today...

EVs are better for autonomy, right? So well I mean so autonomy doesn't actually care too much about whether it's EV or not. I mean generally electric vehicles there's a correlation between electric vehicles and latest and greatest electrical system architectures. And so the EVs today tend to be more straightforward to integrate with than traditional ICE vehicles. But the autonomy there's nothing in the autonomy system that requires it to be an electric vehicle. But of those two billion, it's obviously a very very small fraction that are EVs mostly coming out of China, frankly, in terms of the market share. So, we have two billion vehicles that are being driven around, almost all of them by people, people that frankly have better things to do, are often distracted and so on. And so, there is this massive opportunity to transition as many of those vehicles over time as possible into autonomous ones. Now on logistics in particular, yeah, this is another great example of where we have lots and lots of vehicles that are on the roads. We want to maintain an incredibly high safety bar for them. And we want them to be able to provide a service that, you know, frankly may be very difficult for humans to want to manage. You mentioned package drop off in the middle of the night like maybe that's the best way so that people wake up and they have them. Executing on the operations of these massive logistics networks is challenging so it is a really good fit for autonomy where there's an additional challenge I would say on the logistics side particularly in regard to packages is with robo taxis all the vehicle has to do now, it's a hard task but the only task that it has to do is drive itself, right? Basically where you do pickup and drop off, it's the same service for the end customer whether there is a driver in the car or not. When you start talking about package delivery, it becomes a little bit more different because today we have packages, say Amazon or FedEx or whoever, there's a driver that not only drives the vehicle, but they take the package and they drop it off at your doorstep. And if we replace that with a fully autonomous van without doing that last 10 foot, suddenly either it can't perform the same service or it's a suboptimal product experience for the end customer. Like they have to come meet the van. You can imagine how annoying that would be for most of the packages that we just get dropped off. So, there's an additional challenge that needs to be solved, which is how do you get the packages from the delivery vehicle to people's front door. I think what we've seen in the last 5 years or so is that there's been pretty incredible progress on that front. And so, I'm very optimistic that we're going to have fully autonomous end-to-end logistics delivery. And it does feel that way because if you look at sort of where robotics is heading right now, it's starting inside of the distribution center inside of the factory. So there's industrial robotics. I mean, humanoids are just starting to get deployed and it's really like POCs and tests, but as industrial robotics takes off inside of the warehouse or inside of the facility, it's almost natural for it to extend out into then how do we move those goods? How do we move those materials or those products out into the world? And so it makes a lot of sense. Even on that last mile, you could see an autonomous robot. It doesn't have to be bipedal humanoid. It could just be on wheels and come from the vehicle and out to the door inside the factory or the depot.

Yeah. I mean, I think that 1% very likely won't be uniformly distributed. So, like we already see in the Bay Area, you can take a Waymo here, for instance, we have pretty amazing technology here relative to the rest of the world. And I think we'll continue to see there being pockets. So for instance, we will have robo taxis across the US and across major western countries. Obviously, China is very advanced on that front too. But long term, I think what we're going to see is mobility in general is going to become a trusted part of infrastructure is really how I see it. Like if you look at how today we invest in roads because we know that roads form the structure, the foundation for us to then build all the rest of society on top of, to build the economic engines that power everything else. Mobility is going to be part of that. So basically we're going to have it and I know this was part of Uber's original vision right was to make it as reliable as running water. That is exactly what we're going to see and it's going to be mobility for passengers as well as for goods. The cost is going to come down dramatically and the availability and access is going to go up and we're going to be at a place where we can trust mobility the same way that you would trust turning on the tap. And that's going to be a pretty phenomenal shift for us as a society. There's also going to be other implications, right? We're not going to need as many parking spots. So, a lot of the visions that we've had from the early days of autonomy where we could transform a lot of the very very valuable space that we have within cities from roads and certainly from parking lots and driveways into much more livable, more beautiful, more beneficial for day-to-day life.

Yeah. 1.2.

There's a safety performance that's much better than human driving unfortunately.

Yeah.

Yeah. I mean in the US it's over 40,000 people that we lose every year. And I believe it's still the case that there are certain age groups, particularly the younger age groups, where it is the leading cause of death, right? Which, you know, I think we've gotten used to. It's sort of a price that we've grown to accept, but that's absolutely ridiculous. You know, that we can do so much better. And over the next decade, we are. And I think that's going to be an incredible contribution overall. So, it's not just about liquidity of ride sharing rides and time to destination. And it's not just about the packages getting there on time and more frequently. But it's also about just human safety.

So, let's do some quick predictions here.

Yeah. I mean, I think that one of the challenges that we have overall in the automotive industry is that these vehicles last a very long time. And so even if today every new vehicle that we made was fully autonomous, it still takes decades to completely wean off the existing vehicles. So the overall penetration and number of vehicles is notoriously difficult to estimate. But what we can say, or at least that I will say with some confidence is in 10 years in certainly any sort of western country you're going to be able to take robo taxis I would bet for almost every trip that you would want to take a vehicle. So there will probably be some rural areas where the coverage is not so great. But the level of reliability that you have today from Uber, you will have from fully autonomous vehicles and it will be a combination of both the fleet-owned robo taxis and personally owned vehicles that are being used within the mobility service. And that's, I mean that's in a 10-year period. I think that's a pretty dramatic shift that we're going to see.

Awesome. Thanks, Steve.