Boris Sofman Went from Leading Autonomous Trucking at Waymo to Building Bedrock Robotics

Speaker A: Over $700 billion of data centers being built this year and it's going to be over a trillion next year. It's just unbelievable. But when you look at the labor pool, there's a kind of this, um, generational transition where people have not been going into construction. And so the retirement rate is accelerating. There was already a 500,000 workers shortage, um, just as a baseline that's being exacerbated with the demand. And some of our partners are telling US they're expecting 50% retirement in the next seven years. And so what we're doing is we're enabling fully autonomous heavy machines, starting with construction, to enable effectively turning these machines into operatorless machines where bedrock becomes the digital operator in the software services around it and enables the, um, general contractors in construction or the miners and quarries and mines and farmers and agriculture to be able to operate these machines.

Speaker B: Welcome to Inspire. I'm Alex von Tobel. Construction is one of the most dangerous jobs in America. Demand is at an all time high and the industry is facing a shortage of almost half a million workers with no edge. Boris Softman has spent his entire career building robots that work in the real world. He co founded Enki, which shipped almost 4 million consumer robots globally. He then went on to lead autonomous trucking at Waymo, where he was part of the team that pioneered what is now the world's most advanced driverless vehicle program. And now he's taken everything he's learned across both of those chapters and applied it into machines that literally build our world. Boris is the co founder and the CEO of Bedrock Robotics, a company that is almost valued at $2 billion, that retrofits existing compute construction equipment into fully autonomous machines. In this conversation we're going to talk about what it actually takes to make heavy machinery driverless on, um, an active and dangerous job site. His fascinating predictions for autonomous vehicles in our cities, why he's more skeptical about humanoids than almost anyone else in the field, and why physical AI may be the most defining technology story of the next decade. Let's dive right in. If you're running a startup, we know things like managing equity and communicating with investors are always top of mind. Carta has built a connected suite of tools that over 40,000 companies use to do just that. Their platform of software and services helps you lay the groundwork so you can focus on building. To learn more or get started, visit Carta.com and with that, let's welcome Boris. Boris, first of all, I'm so excited for this conversation on many fronts. Um, both all of your experience on autonomous trucking and where that's heading and then obviously on autonomous robotics. So just welcome. Thank you so much for coming here today. Um, I want to start with the beginning. Um, and before we dive in, I'd love to talk a little bit about kind of how you got here in life. You were born in the Soviet Union, later immigrated to the United States. Um, I would love just to get a sense of like a formative moment in your childhood that you feel like launched you to where you are today.

Speaker A: So I was born in um, in Moscow, kind uh, of. Yeah, right in the heart of Soviet Union. My parents have been trying to leave uh, for uh, well over a decade and uh, you just couldn't. That feels like such a foreign concept today. Um, they were finally able to leave in 1989. It was uh, actually one of the earliest families that were allowed to leave. And my dad was uh, actually pretty deeply involved in kind of helping coordinate kind um, of families of Russian scientists and mathematicians and engineers that through kind of coordination with the US Government, uh, worked, kind of negotiated be able to leave Soviet uh, Union. And I was uh, six years old. Um, and so we moved through um, a few countries in Europe and ended up in New York, lived in Brooklyn for a bit and then through my dad's work ended up in Texas, uh, um, where he was in telecommunications doing large scale optimization. And so I was in uh, Dallas, ah, suburbs for most of my childhood. Um, uh, and so I was always excited about um, kind of mathematics and engineering and video uh, games was actually one of the things that got me interested in AI where I still remember there's a famous video uh, game series, uh, computer ah, game series called Civilization where you're basically kind of simulating the history of the world and um, uh, um, kind of playing a civilization, thinking about research and units and everything. And the complexity was so immense, which was part of the appeal of the game. Uh, the AI was okay, but not great at all. And I started to think about how, what are the kind of algorithms that drive the kind of coordination of um, all these systems? Um, and that actually got me interested in uh, uh, engineering and computer science which took me to Carnegie Mellon. And it's funny because today my 11 year old son, I got him into civilization and normally parents would be like upset when their kids get into video games. And I'm kind of like strangely proud that my son's kind of following in my footsteps.

Speaker B: That's incredible. I love that. Um, I want to talk a little bit about your time At Waymo, um, where again you really led their autonomous trucking initiative. Um, I could geek out on this just because I'm so fascinated by how that could transform society, safer, all of the things. But can we just get a sense of like, first of all, tell us what you did there. Uh, let's start there and then I want to talk a little bit about kind of where it's at it. So let's start there.

Speaker A: Uh, it was such a wonderful time. So I was at Waymo for about five years. So I was an executive, uh, leading autonomous trucking. But then I was also leading various technology teams that supported kind um, of all platforms, cars and trucks. And so, uh, I was leading a perception team and a various other teams. And so we helped launch the Jaguars into San Francisco. Um, uh, we unified the tech stack to where cars and trucks were as unified as possible in terms of the overall architecture where you'd have differences in the platform, sensor layouts, you'd have an articulated trailer. But we were reusing almost every single part of the autonomy stack, which was very powerful. And so we ended up leading both freeway driverless for both cars and trucks. Um, Waymo ended up focusing uh, much more on the car side where there was like a, you know, really a pressure to commercialize and accelerate that. And so we actually launched driverless on freeways for the cars. And at the end of 2023, um, but in a lot, a lot of ways that was actually the trucking kind of like kind of program kind of unified into the car platform. And it was a fascinating problem because, you know, with trucking you had this like 80,000 pound trucks with an articulated trailer. You had like crippled vehicle dynamics where there's like so you had so little flexibility on acceleration and deceleration because of the mass of these vehicles. Uh, you had these interesting challenges that because of that we had to push the perception range out from a few hundred meters, which is what the cars mostly utilize, to close to a thousand meters, um, using camera and radar, uh, to go far beyond lidar range. Uh, you started to think about these extremely rare long tail kind uh, of challenges that appear so infrequently on a freeway and completely changes the way you do a safety framework. It was just amazing problems.

Speaker B: Can you give us an example of one of those sort of real long tail challenges that you had to think through?

Speaker A: Let me share one. That's like actually a big class of problems like reliability. So, um, uh, reliability is effectively. How resilient are you to failures in pieces of your software and hardware. Um, and so, uh, what's interesting about freeways is that, um, the speed at which you drive, um, means that you cannot stop in a lane. Uh, it's just like your budget for that is just like tiny, absolutely tiny. Because if you stop in a lane on a freeway, you actually have a meaningful risk of having a rear collision with a really, really serious kind of high severity of collision. Uh, and it's I think like something like 6 to 8%, um, that integrates over time the longer you're sitting there. Um, and so you have this situation where on a surface street, when you're driving in San Francisco, you're at 25, 30 miles an hour. Almost always if you stop, it's a bad product experience or kind of operationally a headache. But it's not a safety risk in the kind of traditional sense of safety. Here it is. And so not only do you have to reinvent the platform of redundancy for brake, steering, actuation, electrical, kind of anything that mechanically could fail, but by percentages, the largest frequency of what you have to deal with is everything else. So, um, a sensor failing, your computer crashing, a process hanging, your localization failing, your router, your perception, behavior, almost anything. And individually these things might be rare, like tens of thousands or hundreds of thousands or millions of miles in between. But when you compound all of them, you can't go driverless on a freeway without a solution. And so you end up having to have this incredible safety net of, um, redundancy across everything, where you have like catchalls with software that's constantly changing every single release and thrashing, the reliability rates. And you have to have like a space shuttle level of reliability with a system that's evolving all the time. Um, and so you have layers of primary and secondary systems and catch alls and budgets that you have to manage, um, to deal with things that you will never see in our lifetime. So for example, a computer, ah, will have a kernel panic that just crashes completely out of your control every 500,000 miles. But that'll eat up. That's way more than your entire budget of uh, um, these sort of failures. And so it becomes this interesting thing where, um, when we went driverless on a freeway the very first time, the first hundred miles of driverless with the Jaguars, our ah, reliability that we had to hit was 1,000 times higher than what was already a pretty sizable scale already driverless in San Francisco. And so three orders of magnitude. And so these are sort of challenges that are just like absolutely incredible. Um, and for things that you would never see on a typical, you know, five hour drive on a freeway.

Speaker B: Um, so let's geek out for a second. Let's talk a little bit about the future of autonomous trucking. What's your expectation for the next five to ten years? Kind of. What's your mental model of what's probable?

Speaker A: Trucking is an incredible space. Ah, it was such a deep passion of not just mine, but actually a lot of the team at Bedrock is from the trucking program. We just kind of really loved working together on this. Uh, as an industry it's astronomically large. So the entire ride hailing industry in the US I believe is something like 50 billion in revenue. Trucking is a trillion. So it's like really, really significant. Um, uh, and it's the lifeblood of the entire economy. So you have everything grinds to a halt. Uh, if you burden the transport of goods. Um, it's very ah, dangerous. Uh, there's like when an accident happens, it's catastrophic. There's over, I think it's over 5,000 people a year die from trucking accidents uh, in the United States. Uh, um, very difficult kind of lifestyle for long haul truckers. And what you end up having is this like really deep inefficiency in the entire network where you have by design like 11 hour driving limits, federally regulated to where um, you cannot exceed that because of fatigue. And so what you end up having is this fragmentation where um, you do a long haul route across the country, um, 55, 60% of your time is spent on overnights, kind um, of like resting or you have to do dual drivers which is incredibly expensive. Um, you end up having um, huge amounts of. I think it's something like over, like over 25 to 30% of trucks are almost, are empty or close to empty because you have you know, people having to like they miss their pickup window and now they're not going to be able to meet their shift. Uh, right, they have to go back home. Uh, you have people in Chicago when you actually need the trucks in LA during the holidays. And so you have all of this inefficiency. And um, there's studies that have been done where only about a third of that inefficiency is due to asymmetric demand. Like there's more trucks coming into the US from Mexico than going the other way. The rest of it is human factors. And so you have this like giant latent kind of capability to be able to like, you know, uh, optimize across the whole network and move people m, move trucks around anywhere where you need the demand, be able to drive 24, 7. Um, when you do that, the pricing will go down because the price of a load is amortizing all these empty loads and all this inefficiency. Um, and the elasticity is actually quite interesting that if you fluctuate the pricing just a bit, um, you start competing with the price of rail. You start to be able to meet demand where previously there isn't. You start to be able to be more efficient in retail, more efficient in kind of goods. You start to reshape the entire logistic network. Right now hubs are about 10 hours apart because of these driving limits oftentimes. So you start to go from splicing into today's economy of routes and trucks into one where you can fundamentally re optimize how you get much more productivity out of this network. And a lot of times when you do these sort of things, um, the elasticity is actually surprising where you unlock a huge amount of capability because the prices get a little bit cheaper and suddenly the demand grows and you start to be able to leverage these capabilities much more. Um, it's actually quite exciting um, where uh, you see this property in a lot of areas of autonomy and what we're doing in construction and trucking and cars and transportation and manufacturing. When you change the constraints, the physics of the whole industry completely reoptimizes around them.

Speaker B: I think that's part of what excites me so much. Not just the like, improved safety, but just the incredible efficiencies that will be unlocked when our supply chain logistics get incredibly improved, um, quickly. Because I would be bummed not to ask you, um, give us your predictions. Five to ten years out, what does autonomous trucking look like? What percent of truck drivers are autonomous? Like what's your prediction of how quickly that will be adopted?

Speaker A: Yeah, it's a good question. Uh, so just like we saw with Waymo, where the scale can start to be exponential, but every 10x in volume takes time. Like it takes time to ramp up the manufacturing to qualify the safety systems to really methodically kind of scale these technologies. Because if you go too fast you can kill people. Like it's just, you have to be incredibly thoughtful today. It's an interesting ecosystem where um, uh, Waymo has uh, paused his trucking program, but he completed the vehicle platform with Daimler, who's an incredible, like absolutely incredible, uh, uh, you know, uh, like technologically I believe they're the strongest. Kind of like, um, you kind of OEM on the technology front, uh, I would expect Waymo to come back into trucking in the next handful of years, particularly now that the ride hailing side is scaling. They're in the middle of a very, very aggressive ramp. But in the next handful of years, there is no future of Waymo that doesn't include trucking. Just given how gigantic of an opportunity this is. And the message, the thesis of Waymo has always been it's the driver. It's not the robo taxi or the Thomas car. Right. So trucking would be back. And so I would imagine that as you get towards the end of the decade, Waymo's ramping up its trucking program again. Um, uh, I do believe that there's far more that's unique and challenging about just scaling autonomy and dealing with what does it take to ship hundreds of millions or billions of miles than there is that's unique about a car versus a truck. And so I do think that Waymo, I still put them as a front runner, even if they're delayed by a few years. Aurora is also a really, really capable team that's now starting to like, do the, in their earliest ramp. And so I would, I would look at them as one that would start to maybe retrace some of the early kind of ramp that Wayo was doing a few years back at. Uh, like I think they're, they're set to launch this year at some tens to you know, to like 100 units. And so I would imagine that by the early2030s you start to get into, uh, really meaningful miles. Now that could be hundreds of millions, like where Waymo is today to billions. But that's still pretty early in the ramp. I would imagine that by 2035 you start to really see a sizable percentage of this become driverless because it's such a, um, powerful enabler. And truck drivers, um, like the local routes are some of the most challenging, but also the more preferable ones where they don't have to be on the road all the time and away from the families. And so you'll start to see re optimization of where the wafer goes.

Speaker B: Okay, my last question is obviously I asked about trucking. How do you think about autonomous, uh, vehicles and cities? When do you expect that, that. When can I stop worrying that my kid needs to drive a car?

Speaker A: Oh, funny, I'm in the same, uh, boat. I got, uh, three kids between 5 and 11. So I'm, uh, anxiously waiting too. So, okay, so what are we doing?

Speaker B: We're on the exact same page. I've got 78 and 11. And my question is like, can I make sure that they don't ever have to drive a car?

Speaker A: Amazing. All right, yes, we got a lot in common here. Okay, so where's Waymo today? Uh, over 200 million driverless miles. 10 times safer than a human. Like, it's superhuman. It's absolutely incredible. Uh, it's in 11 cities and various degrees of maturity. Driverless, fully driverless. San Francisco's the most mature, where over 40% of ride hailing in the city is waymos. It's by far preferred to Uber and Lyft. Uh, and it's constrained mainly by the availability of cars where Waymo's like, catching up on the like volume of cars. That's a really good signal of where things are going. There's 15 more cities announced, including London, New York, uh, Tokyo. So really aggressive ramp. I think they're, it's something like 15 cities that are on the like, you know, have been announced publicly for um, for 2026. So you, you're going to see this go exponentially where you're going to get into billions of miles. And so if you fast forward at that rate by three, four years, I believe that every single major metropolitan area in the US will have very deep penetration, uh, of robo taxis. And if you fast forward into the2030s, it's very natural that this goes from robo taxi, where you can take advantage of being okay with higher price points because of utilization, some constraints on geography, into a world where you're fully unconstrained on geography, the price points are lower, and you start to infuse this into personal car ownership where you and I can buy a car that is level four capable. I would see that in the early part of the 2000 and 30s being uh, a real thing. And so, um, I literally cannot wait

Speaker B: to own a Waymo. I just cannot.

Speaker A: 100%. Right. And so, and that's like, so worth it, right? It just changes completely your entire existence,

Speaker B: of your entire existence, five, six years.

Speaker A: I think that's like really, really ah, doable. And so, um, yeah, right around the time when ours are ah, about to enter the roads. I think it's actually depending on where you live. Like, um, it's a genuine decision on whether you need a car or not.

Speaker B: Last question on this before we shift over to everything you're building at bedrock. Um, how does that change cities? How does that change things? Because I always say some of the most interesting stuff is the intersection of things. So it's not just that the Autonomous vehicles are going to replace drivers. But what does that actually mean for delivery? Are stores now going to be on wheels? Like does the Sephora need a store? Or could it be be an autonomous vehicle that moves? Like how does that unlock the next layer of innovation in your mind once we believe that 85% of cars on the road are autonomous?

Speaker A: First of all, I think uh, you know that it's not just cars and trucks. It's also local delivery and packages, uh, and everything else. So um, first thing is I think you start to really start to see a shift in the layout of urban, of urban environments. Uh, the amount of parking you need. The percentage of space that's taken up by parking in cities astronomically like surprisingly large, that starts to disappear because I think I read a stat now that in cities like Manhattan, a shocking percentage of cars on the road are actually driving around waiting for somebody or looking for parking. So it's like you start to reduce the amount of parking required. The efficiency of the throughput of cities increases because it's just quicker and more efficient. Um, I think the pain of driving reduces because now you can fully check out and have, uh, have that be a time where you're relaxing, watching something, doing work. Uh, eventually the vehicles themselves start to shift to become more conducive to like an office style environment or sleeping or whatever the case might be. Uh, people become more comfortable with ah, a commute that was 45 minutes used to be really painful or an hour used to be painful. Now it becomes okay, um, and then obviously delivery and transportation becomes easier. And so you start to be able to probably rethink the entire layout of um, cities. And I would imagine you would start to see different form factors where vehicles can be customized for local delivery, um, they can be customized for long driving. Um, and that's where it gets really, really interesting where you start to rethink the form factors of today's cars and trucks into something of the future where you don't need a cabin on a truck in ten years. Um, you can optimize for the layouts of a car, uh, that optimize for comfort and work, uh, because of the expectations. So, um, that's exciting, uh, and uh, I think it'll change, um, where people live and how they interact with their commutes and everything.

Speaker B: Okay, so Boris, given that you're doing something incredibly interesting, um, you obviously left with a number of other amazing, uh, waymo leaders in 2024, uh, to found bedrock. Um, tell us about bedrock. Let's start there first of all like what made you make this leap? What was the problem you saw? What was the aha moment that made you want to start this?

Speaker A: So the catalyst for us really getting excited about this was actually seeing the success of Waymo, of the shift that we had taken to really aggressively move from like a classic robotics stack where you're using engineered solutions and heuristics and search and um, and kind of more rules into really embracing machine learning and data driven approaches where you're using massive scale data to explain how does a human interact with the chaos of downtown San Francisco, uh, or the super crazy and diverse challenges that you would see in general in driving. And not only did that big bet play out where when we launched San Francisco, it was able to reasonably quickly go and solve the diverse challenges of San Francisco, but it generalized super well to Los Angeles, Phoenix, Austin, from car to truck, surface street to freeway. We started to see astronomical reduction in the amount of data that you needed. The work that it took to open up new cities and open up uh, kind of new platforms and um, it went from giant engineering to basically qualification and operations to kind of go and stamp new cities, uh, over time. Um, and so for us that became kind of a superpower where we started to think where else can you apply this? Where you have diversities of use cases, environments and conditions and platforms and you have to have this sort of an approach to solve this problem. Otherwise it's just way too much fragmentation for a kind of hyper focused engineered solution. And that took us to exploring a lot of spaces but we really landed on automating heavy machinery, uh m. And so we're starting with construction, but more broadly this is about automating these large scale machines that are the um, workforce behind a lot of the work in construction, but also mining, agriculture, lumber, garbage movement, environmental cleanup. And so it's these large machines like from Caterpillar and deer and others where they're the ones manipulating the earth, the aggregate material farmland like logs. And when you boil it down there they have a. First of all, these industries account for over 20% of the world's GDP. So and these are the workhorses. And this is like has astronomical labor shortages while demand is skyrocketing because of data center construction and onshoring and manufacturing. Um, we looked at it and uh, industries are massive. There's this spike in demand is just unprecedented. Um, over $700 billion of data centers being built this year and it's going to be over a trillion next year. It's just unbelievable. But when you look at the labor pool, there's a kind of this generational, um, transition where people have not been going into construction. And so the retirement rate is accelerating. There was already a 500,000 workers shortage, um, just as a baseline that's being exacerbated with the demand. And some of our partners are telling US they're expecting 50% retirement in the next seven years. And so what we're doing is we're enabling fully autonomous, uh, heavy machines, starting with construction, to enable, effectively turning these machines into operatorless machines where bedrock becomes the digital operator and the software services around it and enables the um, the general contractors in construction or the miners and quarries and mines and um, farmers in agriculture to be able to operate these machines, uh, and meet the demands that they have from a labor standpoint, uh, compress schedules by working not just 8, 10, but 15, 20, 24 hours a day, um, optimize the safety where this is. Construction is the most dangerous job in the U.S. um, and again really transformed the physics of this industry. And so we're starting with uh, construction and focusing on excavators as our first machine.

Speaker B: First of all, like, my young son would be so excited that you're focusing on excavators.

Speaker A: I got so many, so many cool dad points. Uh, my five year old knows how to operate an excavator. It's amazing.

Speaker B: And like, of course your dad points are legit on fire. Um, can we quickly rewind, Tell us in plain English, what is Bedrock Robotics doing with excavators?

Speaker A: Exactly. So we are retrofitting existing heavy machinery like ah, starting with excavators, uh, and turning them into uh, uh, driverless machines. And so we have a retrofit that turns existing Caterpillar machines for example, into ones that are capable of being fully operatorless, uh, for the scope that it's cleared for and that increases with software updates. Our customers end up being the general contractors that own and operate these machines and bedrock becomes the digital operator of those machines. And so we enable these excavators to be completely operatorless, uh, and be able to do their work without uh, any constraints and at a superhuman level of safety, uh, and precision, uh, and enable these contractors to then leverage that um, to really rethink the way they operate their businesses.

Speaker B: Okay, so I have so many questions. Boris, you went from obviously and that's why I wanted to touch on the trucking, really thinking about fully building autonomous trucks. And then you shifted into another heavy machinery industry. So not that dissimilar to trucks, um, and made the decision to retrofit. Let's double click on the decision not to go build autonomous X excavators but instead to retrofit. Why?

Speaker A: So if we could do that at Waymo, we would have done that as well. Um, the thing is in cars and trucks you can't, you physically have to redesign these platforms to be drive by wire so you can control it digitally and then have the redundancy we talked about in brake steering actuation because of safety reasons. Um, and so these programs are really complicated. They took oftentimes kind of jointly with the partnerships like hundreds of millions of dollars, four to five years. And every single time we made a vehicle program like this, uh, the Chrysler Jaguars, Daimler Freightliners, you know, Zeekr, Hyundai, it would take this sort of an effort. Um, and so that's actually a huge barrier to get to market here. Because of the way these machines are designed, we're able to retrofit them with a sensor and compute suite that turns them into autonomously capable. But um, uh, we're able to ensure that they can still fail safely to a stop. And so we can safely deploy them fully driverlessly without having to fully reinvent an excavator or a wheel load or a bulldozer. Um, and that's actually incredibly enabling because we can get to market a lot faster. And then long term we would partner with the OEMs to go and do this at manufacturing time because there's huge advantages there as well. But this is one of the reasons that we're able to uh, get like later this year we're going to um, release our first driverless systems, um, excavators doing mass earthwork as the first entry point. And we can do this a little bit over two years since the start of the company where AV companies would take many, many m more years and far more expense in order to do this. And so for us this was an advantage. And as a customer you now do not need to buy another. Like these machines cost $600,000. And so instead of having to buy a brand new machine, um, that is custom and much more expensive and complex, you get to use your existing systems with a very inexpensive upfit and immediately start leveraging these capabilities with much less investment.

Speaker B: I'm going to get to what the technology is in one second, but I want to cover two topics. One is what is the role of the human in these uh, more autonomous excavators?

Speaker A: So when you look at the operations of a construction site you have huge amounts like diversity of Machines, you have lots of operators, you have supervisors, you have um, uh, you even have like spotters for safety of all these things happening. And so um, we're starting with an excavator doing mass earthwork. And then the way these systems are designed, um, the technology very quickly snowballs the way we saw Waymo now stamping out lots of cities. We start with earthwork for an excavator and then start adding new capabilities like demolition, material handling, trenching, fine grading, new machines like wheel loaders, bulldozers, motor graders. And then eventually you start being able to go up level and be an orchestration layer where you're able to um, uh, kind of coordinated team of machines doing work. And as a GC you can set a goal and it's a digitized manufacturing process and all sorts of things start to appear when that's possible. Now what's interesting is that on a construction side there's such a gigantic span of expertise and these machines have huge diversity of use cases of what they're um, used for. Particularly an excavator which is the most complicated of these machines where it's uh, minimum six degree of freedom machine and sometimes as much as nine. And so it's effectively like an arm. And that's why it's so versatile. Um, and so what we'll end up doing is starting to focus on these large scale use cases. But what that enables is in an industry that is so severely labor constrained, the workflow can actually move around. And just being able to meet the demands of some of this work that can make a project pencil out means that they can take on so many of the projects that currently are just being completely unfunded. Like heavy civil work, schools, data centers, housing, and all of these places where we're feeling the pain. Uh, and it's actually quite expansionary where the jobs will shift around and you might have a person supervising 10 machines and doing the work on a side that is not autonomous and is actually much more complex.

Speaker B: That's actually a theme that we've heard on many past episodes where this isn't getting rid of that job. What it's simply doing is actually taking somebody from being a line worker and actually making them the factory manager. Um, uh, to your point, of many machines. So in fact we're up, leveling, upskilling these jobs. And you just said it's incredibly expansionary. Which means now rather than huge projects that we want done for society sitting, uh, unresourced, they actually now we can go do more with the supplies. And infrastructure that we have. That's wonderful. What are the other advantages? You had said that there's a ton of other advantages of this. And by the way, I've obviously spent a lot of time on the Bedrock Robotics site where you can actually see the machines moving around and picking up the dirt. But talk through those other more subtle advantages.

Speaker A: Yeah, and this is where like, as we started to spend more and more time with our, with our partners who eventually are our customers, uh, it's amazing to hear the sort of ways that they can like start to leverage this and rethink their business. So I'll go through a few. So, labor, very obvious. Everybody's struggling with this. And every one of our partners, um, like not a single person gets laid off. They take on more work, they meet the demand. They start allowing people not to get stressed until overtime, where you start to have fatigue and safety risks. So that's a given. What else can you do? Now all of a sudden, um, you're able to compress schedules. Um, a job that um, uh, because you can barely staff one shift, nobody's even thinking about nighttime in most cases. Now instead of 8, 10 hour shifts, you can do 15 hours, 20 hours, 24 hours. And now you can take a project to compress it. The utilization of your machines increases. You can take on more work. The profit margins of the entire kind of business end up increasing. Um, that's critical for uh, uh, so imagine like you're a developer that's um, doing a, um, commercial facility or investing in a factory. Uh, you're taking a loan at 9% interest. If you compress that by a month, it completely changes your irr. Everything transforms. A data center gets, uh, millions of dollars revenue every day. It's up and running. So that's massive. Uh, the predictability and the accuracy of these systems, um, is going to be exceptional. Um, uh, and most importantly, it's very predictable. And so what happens now is because of both the variance of the work and, and the attrition risk, um, on jobs, uh, and all the uncertainties that happen, um, GCS will put a 5% contingency on their bids. Oftentimes because of they're on the hook for liabilities. If they're over, if that 5% becomes 2%, because now you have complete control over surging labor and the predictability of your work, um, you tighten up your entire operations. A GC will win every single bid because a lot of these are won by like half a percent. So that becomes a superpower. We see unprecedented levels of detail on the site. So now um, they can see the productivity mismatches in trucks to excavators. Um, they see the progress, you can see real time estimates. And so that insights and analytics completely improves. And safety is a huge one where um, not just the people safety, which is again the most dangerous profession, um, in the entire country, but the amount of secondary damage around hitting fire hydrants, damaging a truck, damaging the machine itself, um, preventable kind of wear on the machine, all of this improves. And now every machine that has a spotter behind it, all this kind of disappears. So you start to have all of these layers of benefit where um, uh. And when you now put all these together as an operator, it sounds a lot like trucking where you can now move these machines anywhere you want, start a project 200 miles away, the next day you rethink the entire physics of the whole business.

Speaker B: Okay, so Boris, now I want to get your head up, um, and think about the entire category, uh, which is really the intersection of AI and robotics over the next decade you have spent. And by the way, you have such dad points. I'm like, I just like sitting here reflecting. Your kids think you're so cool. Um, if we look out the next decade, what are some of the big predictions you have? Give us just some of your very clear predictions that you have for the whole category.

Speaker A: Yeah. So I think physical AI is going to be in my opinion, the biggest theme of the next 10, 15 years. We've seen the transformation on a digital side, astronomical. Everything that anthropic OpenAI everybody's doing. At the end of the day, 80% of the world's economy is still physical industries. It's industrial, it's construction, agriculture, mining, manufacturing, logistics. So I think we're just scratching the surface in this. We've seen the beginnings of it in transportation with Waymo. And it's literally just the beginnings. It's trivial percentage of the. There's 3 trillion miles of driving per year in the U.S. waymo has done cumulatively 200 million um, in the entire history of the company. So it's clearly in the beginning of this astronomical ramp, um, we're pushing in, ah, uh, as significant of an industry, uh, in these kind of the heavy machinery and construction and related industries. I think manufacturing is another astronomical one where these are like 10% plus of GDP industries that all have such a transformative ability to kind of reshape um, the way they work. And so I generally do think that um, the theme of the next 10 years will be um, bringing AI into physical form. The way that happens though is not going to look like OpenAI, um, uh, and quad, um, you're not going to have this single model that solves everything because um, that had, that was able to bootstrap off of this infinite amount of data that was publicly available on the Internet. Here you have way more differentiation in the platforms and the sensors. Um, and you have very complicated safety risks. And so you're going to see more vertical integration like what we're doing in construction, what Waymo's doing in transportation. But the businesses can be trillion dollar businesses because of how big these sectors are.

Speaker B: Not only do I agree with you, we backed two companies, one called brightai, um, which is physical sensors for very specific categories and Tree Swift, um, which is physical robotics for trees, um, and for overseeing forests. And so I deeply agree with you, which is it's not a one size cuts all LLM that can solve millions of use cases. You actually need the data sets and to go and you're really building a very, very unique data set that will be powerful here. Um, talk a little bit about like a contrarian view that you have. Like what's an opinion that you have that maybe um, others don't have given uh, your perch.

Speaker A: Yeah, I'm pessimistic about this wave of humanoids. Um, I think that this is technology that could be much further away than uh, people think because there's still pretty massive advances that need to happen in tactile sensing and simulation. Um, uh, and particularly when you look at sectors like the consumer market where you have very high uh, quality bars that people expect price point pressures that are like extreme. Um, you have so much combinations um, of technology risk and then product and market risk that uh, I'm personally I can understand the excitement about it, but the idea that there'll be this like universal kind of platform that kind of solves everything. In my mind I think this could be almost like a dot com style situation where uh, it's not wrong, it's just off by a big period of time and there's like big advances that need to happen and it comes back and these things start to happen much later than people think. But the solutions that are actually much more practical are more surgically focused solutions that can provide really deep value focused on a particular application in industry. And so um, I am much ah, less optimistic about that as a category of robotics immediately today as a practical kind of like near term business case.

Speaker B: Do you mean humanoids in warehouses or humanoids in homes or all humanoids Uh,

Speaker A: I say particularly in homes and the idea of 100% generalized solutions. I'm more optimistic of the idea of manipulation in fulfillment in a warehouse, in assembly, in manufacturing, where you customize it for the particular application you have and you don't necessarily need to mimic the form factor of a human. You solve the problem with these technologies and eventually grow into something more generalizable.

Speaker B: We share that opinion. Um, Boris, last question before we move to the Quick Fire Round where I just get some of your hot takes. Um, what are you most excited about? One question. What are you most excited about?

Speaker A: I think the pace of uh, technology progress is the fastest I've ever experienced in my lifetime. Uh, I can't wait for it start to have an impact on some of these spaces where it takes a little longer to break through but can completely transform our lives. Particularly in medicine, genetics, personalized treatment, um, where so much of it has been this painful progression of almost like iterative testing that can. The way we've seen this transform a lot of these digital and kind of structured physical AI problems, I think that's going to become much more prevalent in the next 10 years. I can't wait for that to happen. And that may be the most transformative impact of AI, uh, uh, which I think that's actually kind of almost necessary because there's this unfortunate negative opinion that's forming on AI because it feels like society isn't universally being able to benefit from it. I hope that in places like education and healthcare, that's the most transformative and universal applications of these technologies that become possible.

Speaker B: I agree with you on that. Um, okay, Quick Fire Round, before we let you go, um, what's a book that you recommend everybody read? What's a book that's had an outsized impact on your life? It can be any book. It does not have to be a business book.

Speaker A: Uh, I loved Creativity Inc. Uh, this was actually Ed Kapml's biography and the history of Pixar. And um, it was just a fascinating story, uh, but it was almost like this, ah, memoir on how do you create reproducible, uh, creativity that doesn't stifle the ability for individuals to push their visions, but maintains the level of quality bar that you need as you scale a company. It was actually quite amazing and I think applies to a lot of uh, industries far beyond entertainment.

Speaker B: That's awesome. Um, what's a mantra that you have that kind of helps fuel your work? And it can be anything, a belief, but like sort of something that runs in the back of your head.

Speaker A: I think the most incredible applications, uh, and solutions are at the overlap of industries that don't typically have commonality. And so, and we've seen this over and over again in like, uh, these types of companies where you bring backgrounds from completely different places and the sum is far bigger than the parts.

Speaker B: I totally agree with you. I think it is the Venn diagram between truly divergent categories where you find wild, unlocked for society. So I deeply agree with that. Um, if you had to talk about another category that you're just fascinated to learn about right now that is not construction, what is it? What's just another place where the world is catching your attention?

Speaker A: Oh, my gosh. Um, well, I got two answers for this. Practically. The, uh, other area I was really excited about and was started to really explore was manufacturing, where I do think that there's going to be a world not too far in the future where you start to create an increasing library of types of tasks that can be automated, that are increasingly sophisticated. And the combination of those can transform the way we think about manufacturing and assembly. Uh, maybe less practically for me personally, but one that if I were to transform and shift decades of experience into another area, it's genetics. I'm so fascinated by how something that was, uh, it's effectively the software of the human body. Um, and with that lens, you can take all, all of the same types of approaches and mindsets that we have in AI today to a space that is, as we talked about, the most transformative of our generation.

Speaker B: I'm only smiling because that is precisely what I would have said too, and I deeply, I love it. Um, Boris, last question. Um, to date, what's been the biggest pinch me moment of your career? What's, like, the moment that when you tell your kids you're proud of, what is it? We'll end there?

Speaker A: Uh, you know what? I think, uh, I hope to have many of these, uh, at Bedrock, uh, at Waymo, it was, um, being able to really see the first driverless car milestones, like the very first time in San Francisco commercial rides, and then the very, very first freeway driver list that we very directly kind of pushed and worked on. The reason being is that, um, autonomous driving was the thing that got me into robotics. It was like this holy grail that was so exciting. And it was this catalyst for so many people in the industry, industry that pulled us into this, like, physical autonomy space. And it was this search of R and D for over, for decades, and then finally hit this inflection point where it started to become real. And it was an incredible feeling for the first time.

Speaker B: That's amazing. Um, Boris, thank you so much for joining us today. Everybody out there, if you want to check out Bedrock Robotics, please do. The website is incredible. Boris, we are rooting for you. Um, thank you for unlocking the future of everything. Construction, humanity, saving lives. Um, and for all your work at Waymo. I'm deeply, personally grateful. Thank you.

Speaker A: Thank you.

Speaker B: You're the best. We wish. Best of luck.

Speaker A: Thank you. Love the conversation and, uh, uh, excited to talk again soon.

Speaker B: Thank you all so much for tuning in today. If you enjoyed the episode, please rate, review, subscribe. We'll be back again with another episode of Inspired with Alexa. Thank you.