Latest / Automated with Brian Heater / Boris Sofman on Bedrock Robotics, Waymo, and Autonomous Construction

Transcript

Boris Sofman on Bedrock Robotics, Waymo, and Autonomous Construction · Automated with Brian Heater

0:00
SPEAKER_00: I can't imagine anything more exhausting than trying to create an EV company from scratch that, you know, doing autonomous cars or trucks.
0:06
SPEAKER_00: This space, however, we're gonna go driverless not too far in the future this year with fully driverless operator out machines.
0:14
SPEAKER_00: We're automating heavy machinery, like making fully self-operating heavy machines in construction like excavators and other machines.
0:21
SPEAKER_00: These industries are 20% of the world's GDP when you count construction and mining and agriculture.
0:25
SPEAKER_00: And so this is a big, big, big problem.
0:34
SPEAKER_01: You'll go over to Waymo, you're there for three or four years, as it's really sort of finally seeing success, then you're like, hey, let's sacrifice everything and launch another startup.
0:44
SPEAKER_00: Yeah, so I was there for five years at Waymo.
0:46
SPEAKER_00: So it was an amazing, it was like an absolute highlight of my career.
0:48
SPEAKER_00: And one of the most powerful aspects is seeing how capable these technologies were, where you started to use machine learning and large-scale data to solve these problems.
0:56
SPEAKER_00: We started to really think about how you could apply that in other areas.
0:59
SPEAKER_00: That's what took us to automating heavy machinery.
1:02
SPEAKER_01: Do we also have to become a heavy machinery company in order for this thing to really work?
1:06
SPEAKER_01: Or who are we selling to and what precisely are we selling to them?
1:30
SPEAKER_01: When I first met Boris Soffman, he was doing something uh uh something very different.
1:37
SPEAKER_01: Uh, he was the founder of the Toy Maker Anki.
1:40
SPEAKER_01: These days, he is the CEO of Bedrock Robotics.
1:44
SPEAKER_01: They're bringing autonomy to uh heavy construction machinery.
1:48
SPEAKER_01: Uh very interesting guy, interesting conversation.
1:51
SPEAKER_01: Thanks so much to Boris.
1:52
SPEAKER_01: Thanks to you as always for tuning in.
1:54
SPEAKER_01: If you're enjoying the show, the easiest way to support us is by liking and subscribing and telling a friend or enemy.
2:02
SPEAKER_01: I would also encourage that you check out our newsletter over at automated.fm.
2:07
SPEAKER_01: For now, stick around and enjoy this conversation with Boris Soffman.
2:12
SPEAKER_01: Hey folks, I wanted to tell you about something that we've been working on here at A3 EditorialHQ, which is actually a series of uh work from home settings decentralized on the web 3.0 blockchain.
2:25
SPEAKER_01: I'm excited to report that in addition to our recently launched automated.fm news page, we now have a news center for all of A3's editorial content that includes automated, Aaron's new business of automation, all of the podcasts, the market intelligence reports.
2:43
SPEAKER_01: We've grown this team really fast, and frankly, it's useful for us to have all of this information in one place.
2:49
SPEAKER_01: Check it out over at automationnewscenter.com.
2:53
SPEAKER_01: Bookmark the site.
2:55
SPEAKER_01: Uh, we're continuing to roll out new features throughout the end of the year.
2:58
SPEAKER_01: Thank you for reading.
3:01
SPEAKER_01: You in a conference room?
3:02
SPEAKER_01: Am I reading that correctly?
3:03
SPEAKER_01: It looks like you're in a conference room.
3:05
SPEAKER_00: I am, yeah.
3:07
SPEAKER_01: Do they do they not do they not give you uh your own office as a CEO?
3:13
SPEAKER_00: No, no, yeah.
3:14
SPEAKER_00: We're definitely in a phase of just like large open pedal of chaos.
3:18
SPEAKER_00: In fact, uh we're running out of space.
3:20
SPEAKER_00: I'm I'm about to lose my desk.
3:21
SPEAKER_00: So that's how that's horrified.
3:23
SPEAKER_01: If only you had access to some sort of uh like construction tool or something to help you like.
3:31
SPEAKER_00: Oh yeah, we're moving soon, but yeah, no, no.
3:33
SPEAKER_00: I'm uh uh I'm in this like lovely conference room for now.
3:36
SPEAKER_01: Yeah, that's interesting.
3:37
SPEAKER_01: I you know, I I never really get to see the I guess really kind of like kind of the real world implications of what happens after a raise.
3:49
SPEAKER_01: And it sounds like that's kind of the space that you're in right now.
3:52
SPEAKER_01: Like, okay, this is what happens when you get a bigger headcount, is you sort of get a larger office and you're kind of in this like in-between space right now.
4:00
SPEAKER_00: Yeah, basically we have the new office.
4:02
SPEAKER_00: We can't get into it till August uh because there's like a little bit of not too serious, but some construction for the hardware labs and so forth.
4:10
SPEAKER_00: And then in the meantime, uh we're kind of overflowing in this one.
4:13
SPEAKER_00: Uh and we're gonna make it work.
4:16
SPEAKER_00: But uh you know you're almost always on one end or the other.
4:18
SPEAKER_00: Um energy is good though.
4:20
SPEAKER_00: But um, yeah, people start losing their desks pretty quickly.
4:23
SPEAKER_00: It's uh once you if you don't use it consistently, uh you're out and somebody else has to take it.
4:28
SPEAKER_01: Yeah, we had that too uh in publishing, but it was a very different uh I would say everybody was in a good mood.
4:33
SPEAKER_01: It was a different kind of uh everybody losing their desks sort of situation.
4:37
SPEAKER_01: So it's nice, it's nice to sort of be on the on the on the other side of that.
4:41
SPEAKER_01: Uh so but you you've stuck uh you stuck around San Francisco, right?
4:45
SPEAKER_00: That's right.
4:46
SPEAKER_00: Um downtown San Francisco.
4:47
SPEAKER_00: Uh it's actually uh the only place in the Bay Area where you genuinely have access to the whole Bay Area.
4:53
SPEAKER_00: So, you know, South Bay, North Bay, San Francisco itself, east Bay.
4:57
SPEAKER_00: So it ends up being a great hub.
4:59
SPEAKER_00: And um just kind of really went through a pretty painful stretch around COVID and shortly afterwards, but has really bounced back.
5:07
SPEAKER_00: Um, it seems like the I guess combination of AI wave um new mayor is pretty awesome.
5:12
SPEAKER_00: Uh so it seems to be really making a difference.
5:15
SPEAKER_01: Now I'm I'm curious.
5:16
SPEAKER_01: I'm actually uh for uh uh personal reasons, I'm calling you from my uh childhood bedroom in Fremont, California.
5:23
SPEAKER_01: Um so you know, I am from the Bay Area.
5:25
SPEAKER_01: When you say it it has uh I guess access to the entire Bay Area, you mean like specifically in terms of like freeway?
5:33
SPEAKER_00: Uh you're on the like all of the public transit uh and so uh yeah, Caltrain's about a 15-minute walk uh from this office, uh BART Muni right next to it.
5:45
SPEAKER_00: Uh Ferry Building is 20 minutes away, so you kind of have access to North Bay buses.
5:50
SPEAKER_00: So it's not that it's like optimal for like it's okay for almost everybody, but not like uh not horrible.
5:57
SPEAKER_00: Where if you're if you kind of like end up in down south, you've cut off a you know big chunk.
6:02
SPEAKER_00: If you're in East Bay, it's kind of you know, so you end up kind of like making trade-offs.
6:06
SPEAKER_00: But um, this felt like uh a good hub, particularly since um uh you kind of have these pockets where AI tend is tend to be fairly dense in the city and a lot of other disciplines.
6:18
SPEAKER_00: Uh you have hardware and electronics and so forth, it's more heavy down south.
6:23
SPEAKER_00: Um, obviously a lot of carryover from like kind of talent distribution from Waymo and uh a lot of our old circles.
6:29
SPEAKER_00: Um, but uh it's also just more fun to be in the city, I think, than uh a lot of other locations.
6:35
SPEAKER_01: Because were were you in Mountain View when you were at Waymo?
6:38
SPEAKER_00: So that's where their headquarters were.
6:40
SPEAKER_00: Um we led the kind of the San Francisco office, so as the office kind of lead for that one, but I ended up going down south a lot.
6:47
SPEAKER_00: And um, right before COVID, I I took over a pretty large team, the perception team.
6:52
SPEAKER_00: And um uh they were like almost entirely down south.
6:56
SPEAKER_00: And so I remember driving, I I live in East Bay, uh kind of in the Oakland Hills, and so I remember driving, kind of going down to Mountain View almost every single day.
7:03
SPEAKER_00: That was uh that was pretty painful, but uh made it work.
7:06
SPEAKER_01: Yeah, well, because uh I mean, Google like famously, um, and and I suspect that this is uh you know has to be the case to a certain extent for a Waymo, and certainly a case for, you know, like when you're in the early stages or any stage really building a uh robotics company, um, it they they they cut they required people to s to come back into the office, right?
7:29
SPEAKER_00: For the most part, there was like a pressure, but the problem is during COVID, people kind of moved away, nobody knew exactly how long it would be.
7:36
SPEAKER_00: There was a flexibility that was afforded.
7:38
SPEAKER_00: And then when we came out of COVID, um everyone they've you know, you try to get people back and they're like, well, you know what?
7:46
SPEAKER_00: Like, this is my life now.
7:47
SPEAKER_00: I'm like, I'd love to keep working remotely or I'll go somewhere else.
7:50
SPEAKER_00: And then you pretty much have like an impossible choice, and most companies can't afford to lose um some of their best people.
7:56
SPEAKER_00: And so once you lose it, like you can kind of push it back, but it's really hard to like you know, put it, put everything back in the box again.
8:04
SPEAKER_00: And so from that standpoint, like we started this company in 2024, and um we made a conscious choice that we were going to be kind of like an in-office culture.
8:13
SPEAKER_00: It doesn't mean like militant about it, like five days a week, uh no exceptions.
8:16
SPEAKER_00: And there's a few very special cases where you know people remote and make their effort to come in, but we built around um, you know, two hubs where um you know kind of our founders were, and um and it actually helps.
8:28
SPEAKER_00: Like the energy is great, you collaborate better.
8:30
SPEAKER_00: So I do think that there's something to it.
8:32
SPEAKER_00: And um, I think that was lost for a stretch around COVID where this whole work remote was it was an experiment, and it can probably work fine for some types of businesses that are much more uh mature and established and understood.
8:44
SPEAKER_00: But when you're doing something incredibly novel, the connective tissue is just so important.
8:48
SPEAKER_00: I don't think there's a substitute for being in person for a good percentage of the week.
8:52
SPEAKER_01: And also just robotics generally, right?
8:54
SPEAKER_01: I mean, hardware generally requires people to be there in person.
8:57
SPEAKER_00: Yeah, that's right.
8:58
SPEAKER_00: You feel it, you build intuition, and um and you know, you have test sites in different locations potentially, but you physically kind of like you know, you you feel the products and then you also collaborate, where in robotics, like this is such a novel science where nothing's isolated in siloed, like the hardware has deep implications on the OS, on the ML systems, on the infrastructure, on the commercialization.
9:21
SPEAKER_00: And so um the connective load factor is so high that you just cut through a lot of complexity by being able to um just be in the same proximity um as people just makes a huge difference.
9:34
SPEAKER_00: Um if you're if you're developing like a traditional app, let's say I'd say okay, maybe that's well understood now from you know 20 years of uh of progress that you could probably, you know, kind of lean into best practices and be remote and maybe have less of an impact.
9:48
SPEAKER_00: Culturally, I think you always have an impact.
9:50
SPEAKER_00: But uh for something like robotics, um, yeah, you're in the guts of it trying to solve problems on a dry race board, it helps.
9:56
SPEAKER_01: Yeah, I'm curious if something like this is, you know, as you said, it is uh it is novel.
10:01
SPEAKER_01: Um as you're having those discussions in terms of uh what market you're targeting, uh, and we'll certainly get around to the fact that like you've you've bounced around to different subcategories within robotics.
10:13
SPEAKER_01: Uh, you've you've been like around like all different sides of robotics.
10:16
SPEAKER_01: But uh as you're sort of honing in on that, is the initial impulse to like build as much of the solution as possible, or to you know, to kind of pull what you can from existing solutions and then build on top of that?
10:31
SPEAKER_00: Uh so it depends very much on the piece.
10:33
SPEAKER_00: So um when something is reasonably mature and well understood uh and somewhat of a commodity and not like the special sauce of what you bring, you look to outsource it because you don't have vulnerability on supply chain, it's stable and so forth.
10:47
SPEAKER_00: So examples of that might be uh you know cloud infrastructure like AWS and you know and so forth.
10:52
SPEAKER_00: Uh um it's sensors like cameras, LIDAR.
10:55
SPEAKER_00: Uh, you don't want to reinvent those.
10:57
SPEAKER_00: Um in our case, um, we're automating heavy machinery, but not having to fully redesign excavators and and dump trucks and bulldozers and so forth.
11:07
SPEAKER_00: That's a gigantic advantage out of the gate, even if down the road you have that option.
11:11
SPEAKER_00: We never had that option at Waymo because from for a variety of reasons, you we had to redesign the platforms um that we would work on.
11:18
SPEAKER_00: And so um, you know, you you so you try to like lean into existing solutions when you when you have them, visualization tools and so forth.
11:26
SPEAKER_00: But when you get down to like the actual kind of core autonomy system, ML models um and so forth, like you kind of have to really vertically integrate and own that, you know, end-to-end where you're developing the nuances of your particular hardware sensor suite requirements and and you and you go and run it.
11:44
SPEAKER_00: But you try to use the railroad tracks that already exist in some of these building blocks.
11:49
SPEAKER_01: Yeah, I mean that was a bit part of the conversation too, is like, do we also have to become a heavy machinery company in order for this thing to really work?
11:56
SPEAKER_01: Or who are we selling to and what precisely are we selling to them?
12:00
SPEAKER_00: Yeah, and that was exactly the right question.
12:02
SPEAKER_00: And so, because if you look at like Waymo, it's like, wow, they're a robo taxi company and like, wow, that's expensive.
12:07
SPEAKER_00: They just raised$16 billion because you're standing up an ecosystem of robotaxis and depots and chargers and everything and um incredible like uh achievement in business, but it's like capitally intensive.
12:18
SPEAKER_00: Um, one of the early decisions we made is that um you know, like at Bedrock, we're um uh we're automating heavy machinery, so like making fully like driverless, uh self-operating um heavy machines in construction, like excavators and uh other machines um soon after, like wheel loaders, bulldozers, motor graters, and so forth.
12:39
SPEAKER_00: But because uh like we have the advantage that an ecosystem already exists where there's subcontractors and general contractors that own and operate these machines, they know how to use them.
12:49
SPEAKER_00: They're part of a much more complicated um suite of operations and a construction site.
12:54
SPEAKER_00: And so the fact that we can actually enable these machines to be fully driverless and have our customers be the contractors that own and manage these machines is a big advantage both ways because they can spice them in and be in the best position to take advantage of the superpowers of the machine that's fully driverless, can compress schedules, be predictable, be safer, have all these insights that um it provides.
13:14
SPEAKER_00: And they're already used to owning and operating this within a much more complicated construction site.
13:19
SPEAKER_00: For us, um, we effectively then have a product that is a digital operator and the software services around it.
13:27
SPEAKER_00: Um, it leads into the superpowers of both companies.
13:30
SPEAKER_00: And wearing our hat, it allows us to most aggressively mature and uh uh and scale the different capabilities and different use cases and different machines and the orchestration layers around this, which um I think is a win both on both sides.
13:46
SPEAKER_00: Uh so uh that that part's a little bit different than Waymo, but I actually think it's a huge advantage where you can kind of lean in your superpowers and not actually try to become a contractor in our case.
13:58
SPEAKER_01: Yeah, but is is there like a technical reason why a Waymo would have to, I mean, obviously not build, but but contract and work with a company to create like specifically these cars for this purpose versus retrofitting existing machinery?
14:12
SPEAKER_00: Yeah, that one actually has a very clear reason.
14:14
SPEAKER_00: And so every single time we went through this um Chrysler, Jaguar, Dimer, Freightliners, like that was a um the other project that was uh uh leading in parallel, um, Hyundai Zeker.
14:26
SPEAKER_00: Um, we would have to do this partnership because when you go driverless on an existing car truck, um at the sort of speeds that these cars drive at, you cannot do this with a retrofit because you need redundancy on brakes, steering, actuation, electrical, because at volume, every one of these components will fail, and you have to have a handoff that's safe.
14:46
SPEAKER_00: And you have to have these machines be drive by wire, which didn't always exist.
14:50
SPEAKER_00: And so you would be redesigning this and not just going.
14:53
SPEAKER_00: So I worked on both the um, you know, helped bring up the Jaguars, the first driveless in San Francisco, and then we were like developing and deploying the freight liners.
15:02
SPEAKER_00: And in both those cases, uh, you have to like fundamentally redesign you know these elements to be road safe and scalable.
15:08
SPEAKER_00: You can prototype with the retrofit, but you can't go driveless with the retrofit.
15:11
SPEAKER_00: Um, and now you now you're not just working with Jaguar and Daimler, but you're not going down to the tier ones like ZF and Bosch and others that are making the components.
15:19
SPEAKER_00: So you have like a sometimes a four or five-year program that can cost you know$100 million plus total of you know, between hundreds of millions between both sides.
15:29
SPEAKER_00: And so um, and you don't have a choice.
15:31
SPEAKER_00: The nice thing about um, you know, for example, the machines that we're using for our deployments is that they are because of the inherent design that they have right now, they're already electric to hydraulic, they're drived by wires, so you can take advantage of that um aspect.
15:47
SPEAKER_00: And then these systems, if you architect your system and your integration the right way, you can ensure that they fail to a stop under any sort of failure.
15:57
SPEAKER_00: Um, and that's what's um you know really unique about this um uh this case.
16:02
SPEAKER_00: So yeah, no shortcut.
16:03
SPEAKER_00: In autonomous driving, it would have been amazing, but you uh you have to go through these complicated vehicle programs.
16:08
SPEAKER_00: And sometimes they're really, really complex and actually barriers to entry as you've seen in the trucking space.
16:13
SPEAKER_01: Yeah, I mean, I I I tend to think, and and I could be off on this, but like I tend to think of a construction site as being maybe relatively more structured than um being on the street, or like certainly like less uh less cause for uh less direct safety concerns, perhaps.
16:32
SPEAKER_01: I mean, at least in that standpoint, being relatively like easier to develop an autonomy stack for.
16:39
SPEAKER_00: Yeah, there's a few advantages.
16:40
SPEAKER_00: So um it's a closed site that helps uh on a lot of fronts, like everything from uh some of the complexity to regulatory, but uh the machines are slower.
16:51
SPEAKER_01: And dogs and cats, yeah.
16:53
SPEAKER_00: So you assume that there's like a reasonable level of training on people on the site, but at the same time, we take the approach that we're still responsible for end safety.
17:00
SPEAKER_00: So you don't have like a geo-fenced area where nobody can come close to it.
17:03
SPEAKER_00: Um you assume that people shouldn't come close to it, but people will, and there's always like uh issues on these sites, and so you still have to have like a very robust safety system.
17:12
SPEAKER_00: But the difference is that you're operating in a controlled environment at slower speeds, you have an ability to stop if you ever have uncertainty.
17:18
SPEAKER_00: There's a lot of advantages there.
17:20
SPEAKER_00: The diversity and complexity, there's actually less structure in a lot of ways than open roads, but um, and so you have some gives and takes, the interfaces are harder, but you have a lot more ability to like slice off capability sets and provide a lot of value without solving 100% of everything that's um that in theory a human could do.
17:38
SPEAKER_00: Um, but uh yeah, we were able to bypass a decent percentage of the surface area that we're used to in terms of complexity for a public road autonomous driving system and be clever about how we can get to market quicker and use as a beach hat to really start to expand from there.
17:54
SPEAKER_01: But I mean, in terms of diversity and complexity, like the the home is still that's still gotta be the hardest of all, right?
18:00
SPEAKER_00: The home.
18:02
SPEAKER_00: Uh you know, it depends.
18:03
SPEAKER_00: Like, I still think autonomous driving is one of the most absolute, most extreme hoy grails of uh autonomy because like you know, you you're at like the main reason being that you're driving at like 40, 50, 60, 70 miles an hour.
18:18
SPEAKER_00: Yeah, and you do not have full control of safety.
18:20
SPEAKER_00: You just genuinely don't.
18:22
SPEAKER_00: And so is everybody's safety, right?
18:24
SPEAKER_00: Right?
18:24
SPEAKER_00: Yeah.
18:24
SPEAKER_00: Exactly, right?
18:25
SPEAKER_00: And so you're going on like this unlike kind of social contract where you're expecting people to behave in a rational manner, and that understanding of your interactions with others is at the root of the complexity.
18:35
SPEAKER_00: Um, in the home, it's a different kind of complexity where you're not like in this dynamic, unsafe environment where you're driving within five feet of a pedestrian on a sidewalk at you know 40 miles an hour.
18:45
SPEAKER_00: Um, but you have complexity in terms of grasping, manipulation, kind of these other aspects, which are hardened a different way.
18:53
SPEAKER_00: Um the unforgiving nature of autonomous cars in a public road, that's still from a safety standpoint, the hardest technical challenge, I think, out of all of these.
19:04
SPEAKER_00: Um and uh um and so you can kind of give and take.
19:07
SPEAKER_00: So when you go into the home, you get like you have the advantage you can kind of like slice off certain use cases and maybe provide value.
19:14
SPEAKER_00: I think from a business standpoint, it's a brutally difficult place to go because you're unforgiving in terms of price pressures and you know, you're competing with uh you know labor that's like a lot uh like reasonably um inexpensive if you think of like a you know, a housekeeper or or somebody else.
19:32
SPEAKER_00: Um, but you don't have the same types of safety exposure that you do in these other systems where okay, your humanoids still can't tip over on a child, and you got to be very careful.
19:40
SPEAKER_00: But like uh and but you're not a 100,000-pound machine driving around um, you know, where you genuinely can do some really, really serious damage, even with the smallest miscalculation.
19:52
SPEAKER_01: So uh we're gonna go way back.
19:54
SPEAKER_01: Online learning techniques for improving robot navigation in unfamiliar domains.
19:59
SPEAKER_01: And and if I kind of understand the like the really quick TLDR of it, it's like effectively teaching to a certain extent uh or allowing robots to kind of teach themselves without human intervention or access to human intervention.
20:15
SPEAKER_00: That's right.
20:15
SPEAKER_01: This was uh this is your PhD thesis.
20:18
SPEAKER_01: Yeah, right.
20:18
SPEAKER_00: You're bringing that you're bringing back memories.
20:20
SPEAKER_00: Yeah, this is uh uh the first really big robots I got to work on is uh these giant off-road vehicles with with wheels up to my shoulder that like would navigate tens of kilometers and figure out how to interpret.
20:33
SPEAKER_01: Right.
20:34
SPEAKER_00: I mean Yep, yep.
20:35
SPEAKER_00: We went small, then big again, and now really, really big.
20:38
SPEAKER_00: So uh yeah, those guys were huge.
20:40
SPEAKER_00: And um, yeah, it was a focus on early autonomous driving.
20:43
SPEAKER_00: In fact, like all of the team members that we worked with, those are now like effectively almost like a perfect mapping to Waymo, Aurora, Uber HGG that went into Aurora, uh, and you know, and here and a lot of our old friends.
20:59
SPEAKER_00: We were we didn't appreciate at the time how much of like a core route that was between that and the grand challenge projects, seeded like a big percentage of the autonomous driving space and autonomous, you know, autonomy in general that we know today.
21:12
SPEAKER_00: Uh yeah, and so there we were um allowing these systems to automatically train and improve and bootstrap their perception system and behavior using their own experience as a feedback loop and be able to adjust this in real time, which um is a pretty uh thematically a continuing continues to be a very important uh kind of capability domain where you want to automate that feedback loop as much as possible.
21:37
SPEAKER_00: Um that was a really uh fun early use case uh of those flavors of technology.
21:42
SPEAKER_01: Do you go into CMU uh knowing specifically that you weren't want to work on autonomy?
21:50
SPEAKER_01: Or is it just sort of broadly, hey, I want to kind of get into robots in general?
21:54
SPEAKER_00: So the graduate school, absolutely, because it's um I mean it was by For decades, the only like focused robotics institute, like robotics program in the country.
22:05
SPEAKER_00: And when I was there for graduate school, um it was so astronomically bigger than the next biggest programs.
22:13
SPEAKER_00: Like literally, you would have like 40 to 50 faculty members or focus on different areas of like just in robotics since we focus on robotics, autonomy, ML, all these areas.
22:23
SPEAKER_00: And the next biggest program, maybe you have like four or five.
22:25
SPEAKER_00: Like, and so it was just like uh, and so you would, if you're doing these like large programs like an autonomous car or these systems, it would be the place where you would have the mix of experiences to do it.
22:35
SPEAKER_00: Um, I actually did my undergrad at CMU as well, and I went into CMU because I knew it was a fantastic um engineering program, computer science program.
22:42
SPEAKER_00: That's why I decided to study.
22:43
SPEAKER_00: I had no idea what robotics really was.
22:46
SPEAKER_00: But when you get there, you just see these projects everywhere and you get excited by them.
22:51
SPEAKER_00: You can't help but like walk around and see all this stuff happening.
22:54
SPEAKER_00: And so by my sophomore year, I started getting really, really excited about this idea of like robotics being the extension of computer science and AI into physical form.
23:04
SPEAKER_00: And it's like the one misconception is that like people would think of robotics as like the physical stuff, like the joints and arms and sensors and everything.
23:12
SPEAKER_00: And that's part of it.
23:13
SPEAKER_00: But probably over 80% of the investment and research happening was the software and AI aspect of it.
23:18
SPEAKER_00: I'm like, how do you actually embed intelligence in these systems?
23:21
SPEAKER_00: How do you interpret sensor data?
23:22
SPEAKER_00: How do you um use machine learning to learn policies for walking, grasping, driving?
23:28
SPEAKER_00: Um, and so it was pretty magical.
23:30
SPEAKER_00: And so I ended up staying for graduate school because I wanted to go what much deeper, and that was the best place uh to really do it.
23:37
SPEAKER_01: What's your sense of why, like specifically, though, like in robotics, it became like the hotbed for, like, as you said, all autonomous vehicle and uh vehicles and the starting point for like all of the big players in the space?
23:51
SPEAKER_00: Why Carnegie Mel in particular just like robotics?
23:54
SPEAKER_00: Yeah.
23:54
SPEAKER_00: CMU.
23:55
SPEAKER_00: Yeah, they started in the early 90s, uh, like really started.
24:00
SPEAKER_00: There were a few people like Red Whitaker and others that were just like the like he's the um, I don't know, like people like people call him like the father of like a you know, like the grandfather grandfather, he's like uh uh godfather of uh autonomous vehicles.
24:13
SPEAKER_00: And um, and it was like uh the grand, yeah, like some old some elder relative in the yeah, elder, like the elder statesman.
24:22
SPEAKER_00: Uh and so like originally um there was a lot of this early work.
24:25
SPEAKER_00: So Carnegie was always known for um AI research.
24:28
SPEAKER_00: So like Alan, Newell, and Simon and others, and this really became an area of focus originally with like NASA and DARP and army projects, and then eventually it evolved by the late 90s into like Caterpillar and Deer and some of these industrial applications.
24:43
SPEAKER_00: But it became like this really deep focus of like there's a field robotics center that started to think about agriculture and mining and these areas.
24:51
SPEAKER_00: And so um, it just I think it was bootstrapped by a handful of faculty that got excited about it, and by nature being so far ahead of anywhere else, they started to snowball into more and more investment.
25:02
SPEAKER_00: And um uh, and then pretty soon it became one of those places that uh particularly the government originally realized that this was an amazing source of RD talent.
25:14
SPEAKER_00: There was a group called um NREC, which was a National Robotics Engineering Center, is what it stood for, which was like the industrial commercial branch.
25:21
SPEAKER_00: And so there was an immediate push to not just do pure research, but to extend a branch that would be closer to commercialization, whether it's to government, NASA, Navy, Army, or to like caterpillar and deer.
25:33
SPEAKER_00: And so you started to build this like incredibly applied um lens on all this development.
25:40
SPEAKER_00: And you naturally spawn people to get excited by that and you attract others like it, and that puts them in a great position to do things like the Grand Challenge and Urban Challenge, with that be which then becomes Waymo eventually.
25:51
SPEAKER_00: Um, and and even though like other teams spun up, like Stanford and others who were incredible, like Mike Montemero and Sebastian Throne, they also came from CMU, right?
26:00
SPEAKER_00: And so it was like the roots really kind of like were there, and I just feel like fortunate because I didn't appreciate this when I was um just joining undergrad, and I kind of grew into it.
26:11
SPEAKER_00: But um sometimes you get a little bit lucky and uh you find your passion by uh serendipity like this.
26:17
SPEAKER_01: Were you involved with any of the dark part challenges?
26:20
SPEAKER_00: Uh, you know, I ended up working um not as directly.
26:25
SPEAKER_00: My co-founder actually was one of the team leads on the grand challenge.
26:29
SPEAKER_00: I ended up doing kind of the this off-road version of it and was so excited by these problems, I ended up sticking with it.
26:34
SPEAKER_00: Um, otherwise, I probably would have like worked on the on-road version with the with the grand and urban challenges.
26:40
SPEAKER_00: Um, but like the technology actually carried over because a lot of the tech from this off-road project ended up funneling into the uh the urban challenge and and you know, and and so you ended up having this almost like reuse of technology building blocks because of how this like um you know the robotics industry worked.
26:59
SPEAKER_00: You were able to start to use these more and more and start building on these foundations.
27:02
SPEAKER_00: And so uh um, but yeah, my co-founder for Bedrock, he was uh one of the two team leads uh for the original Grand Challenge.
27:11
SPEAKER_00: And um, you know, and this was like arguably the projects that like really kicked off the entire autonomous vehicle space that without that, I don't know how long it would have been before somebody really put the sort of energy behind it that Google initially uh put into it and then spawned a gigantic industry.
27:28
SPEAKER_01: I I can't understate, like based on all these conversations that I have, and and a lot of these, you know, I was lucky enough to have been covering robocs to a certain extent back then and you know, got to see uh the challenges, some like some of the humanoid challenges, some of the subterranean challenge, but I can't overstate the extent to which DARPA challenges generally have like really built uh robotics up in the US.
27:55
SPEAKER_00: It's wild.
27:56
SPEAKER_00: Like talk about a success case.
27:58
SPEAKER_00: So like my co-founder Kevin Peterson, he um, you know, is like part of this team, and he was an undergrad and then a grad student, kind of working on these.
28:05
SPEAKER_00: Uh Red Whitaker, who was like the original kind of like helped you know galvanize this and was the grand uncle of self-driving.
28:11
SPEAKER_00: Yeah, the grand uncle, right?
28:12
SPEAKER_00: Like he got these like armies of undergrads that were like passionate and excited and just like worked 24-7 on it.
28:19
SPEAKER_00: And um uh and so DARPA ended up making prizes.
28:24
SPEAKER_00: There were like a couple million dollars worth of prizes.
28:27
SPEAKER_00: It stimulated like tens of millions of dollars of investment and then like hundreds of billions of dollars of value long term.
28:34
SPEAKER_00: If that isn't like near the top of the list of the most like long-term successful government-funded catalysts of you know, of development RD and industry, I don't know what is.
28:45
SPEAKER_00: Um, because it's just it's absolutely incredible how much got like just catalyzed through that, and then eventually industry stepped in and realized it was close enough to take over.
28:55
SPEAKER_00: But without that, without those proofs of concept, there were still kind of like staged problems in a lot in every way.
29:02
SPEAKER_00: When you look at the ROI from like a taxpayer and a government standpoint on almost any project in the history of the United States, uh I'd be surprised if like the grand challenge and urban challenges in these types of projects weren't near the top of the list, where DARPA spent a, I think it was a total of$9 million, including a couple million dollars in prize money on the first grand challenge.
29:22
SPEAKER_00: And it stimulated tens of millions of dollars of investment immediately, and then hundreds of billions of dollars of value.
29:31
SPEAKER_00: If you look at you know, the industries and Waymo and everything else that got created, and all these are US companies paying tax dollars and you know, and salaries and investment and so forth.
29:41
SPEAKER_00: And I don't know if any of that would have happened without the like them pulling this into what became a kind of a you know, a proof of concept and a staged example, but one that stimulated the creativity of industry.
29:52
SPEAKER_00: And then suddenly you had Waymo and then eventually every single OEM um spending money on autonomy.
29:58
SPEAKER_00: Um, it certainly fast-forwarded the whole space by many, many years.
30:03
SPEAKER_01: Yeah, you talk about numbers like that, and I'm just thinking, like, oh, everyone is talking about like, listen, it's impressive like how well that movie Backrooms is doing right now based on the budget.
30:12
SPEAKER_01: And like this is this is we're talking like orders of magnitude larger as far as the kind of like broader return, or else, or at least like the kind of money that got put back into the US economy, it sounds like totally.
30:24
SPEAKER_00: If you think of them as like seed investors in like in US economies that like venture fund, like that's a that's a really big return.
30:33
SPEAKER_00: Like in uh and like arguably that's their job.
30:36
SPEAKER_00: And then like once industry takes over, they back out and like become the you know, the beneficiary of it in other ways.
30:43
SPEAKER_00: Um, but then you have like the next frontiers, which became grasping and like and these other kind of manipulation where those are kind of like 10 years offset off of autonomous driving in a lot of ways.
30:53
SPEAKER_01: We don't at least as early have Atlas without the humanoid challenge.
30:57
SPEAKER_01: That's right.
30:58
SPEAKER_00: Yeah, and I think so.
30:59
SPEAKER_00: And um, and there you start really subsidizing balancing and walking work and you know, grasping and everything.
31:06
SPEAKER_00: And um, and just like the shape of so many of these other technologies, you um start with purely government-funded work because like no VC would have funded that, no large-scale company uh, you know, would have really funded that um because there wasn't like a very clear path to commercialization.
31:22
SPEAKER_00: And then you fast forward 10 years and suddenly you have these incredible results and you have them spinning off into companies like skilled into like uh into all of the um you know kind of companies today.
31:33
SPEAKER_00: You have internal funding from um you know from from industry as well.
31:39
SPEAKER_00: Um it's exciting and uh it builds on the shoulders of of a lot of these um early investments.
31:45
SPEAKER_00: And so I'm a giant fan of this being like a really key role for government.
31:50
SPEAKER_00: You gotta be smart about what your role is and isn't, how you pick it.
31:53
SPEAKER_00: You don't play favorites, and then at some point when it reaches a certain level of maturity, you let the the market uh take it over and run with it.
32:00
SPEAKER_00: Um and the US is better than any uh other country at doing this.
32:05
SPEAKER_00: Um and hopefully they, you know, that's it's something that they continue doing uh because there's the new frontier in every single space that has its flavor.
32:14
SPEAKER_01: Yeah, it strikes me, you know, had you been more involved with the one of the DARPA challenges specifically, that you're I mean, likely you would end up in a pretty similar place, but the the intervening years of your life, your career might have been um a bit different because you you ended up at NETO, which is interesting because obviously, like you if you had been at MIT, like you just would have been sucked up into the giant iRobot machine, right?
32:37
SPEAKER_01: You you ended up NETO at NETO, and that's uh like effectively what took you out to uh the South Bay?
32:43
SPEAKER_00: That was actually an internship uh that in the middle of grad school.
32:47
SPEAKER_00: So I did that for maybe like nine months and then I consulted a little bit Vincent.
32:50
SPEAKER_00: And so I came out.
32:52
SPEAKER_00: Yes, that's actually what like my first experience of being in South Bay.
32:55
SPEAKER_00: Um, it was a consumer application, which I thought was really exciting because I'd never um funny enough, I did an internship at a robot before grad school.
33:03
SPEAKER_00: So it's like I uh the year of the IPO it's I got a little bit of exposure to uh to both.
33:07
SPEAKER_00: I was in a research group and um and you uh it was exciting in that like it was Silicon Valley.
33:12
SPEAKER_00: It was a um they'd found a way to make a LIDAR for like$20 bill of materials was like pretty exciting uh kind of tech.
33:20
SPEAKER_00: So I could have intelligence and consumer products you couldn't before.
33:23
SPEAKER_00: Um it was in a lot of ways, like there were it was a fodd company in a lot of ways, but we like learned a ton.
33:29
SPEAKER_00: Um and it ended up being the seed for what became the first company I created out of grad school, which was you know called Anki, making kind of consumer products with high-end uh kind of entertainment and interactive experiences.
33:41
SPEAKER_00: If it wasn't for that, I probably would have gone into uh uh urban challenge in Waymo because uh I was really close to all those projects.
33:51
SPEAKER_00: We were close to a lot of, I had like very close friends on it.
33:53
SPEAKER_00: Um, it was a natural kind of like next uh kind of transition.
33:57
SPEAKER_00: I always loved driving was what got me into robotics.
33:59
SPEAKER_00: I always thought that was like one of the most exciting, um magnetic, kind of holy grails of the space.
34:05
SPEAKER_00: Um but I kind of got this bug of thinking about how to commercialize in a consumer space where there was never any other uh there really weren't that many examples um of this technology going anywhere outside of kind of industrials or pure research or government work.
34:21
SPEAKER_00: Um, and it felt like a pretty exciting opportunity.
34:24
SPEAKER_00: But then um eventually I uh all this brought me back to Waymo again.
34:28
SPEAKER_00: So uh came full circle.
34:30
SPEAKER_01: Yeah, yeah.
34:31
SPEAKER_01: I th this would have been around the time that that you and I would have first crossed paths, uh, you know, as you as you were pitching uh Cosmo.
34:37
SPEAKER_01: Um and like uh uh somewhat ironically, because uh again, I'm I'm not in my usual spot.
34:44
SPEAKER_01: I don't I don't have my Cosmo behind me, but maybe um we have talented editors, maybe we'll maybe we'll edit it in.
34:51
SPEAKER_00: But that was a good that was a good product, yeah.
34:53
SPEAKER_01: Yeah.
34:54
SPEAKER_01: At that point in you know uh 2010, 2011, um at the time, did consumer robots feel like more or less of a long shot than self-driving?
35:11
SPEAKER_00: Self-driving less than so well, self-driving was just getting kicked off at Waymo.
35:19
SPEAKER_00: And they did feel internally that this would be like a four-year problem and they did knock it out and be done.
35:25
SPEAKER_00: It ended up being a much longer problem, like 15 years, and they're like working on it.
35:30
SPEAKER_00: Yeah, yeah.
35:31
SPEAKER_00: Yeah, and like now it's like starting to work.
35:33
SPEAKER_00: Um, on the consumer side, um we draft we had a pretty good thesis in that we were seeing smartphones come out and start to really scale, where iPhone started launched in 2007, and then smartphones in general, like Android and iPhone, started ramping into the hundreds of millions of units, which drove down the cost of every single component that they had.
35:56
SPEAKER_00: So microcontrollers, uh memory, uh cameras, uh, motors, wireless chips, like these are all the building box of consumer products.
36:04
SPEAKER_00: And so we had the realization that, like, that wow, if you're smart, you can drop these prices are dropping 10x off of what they normally were.
36:12
SPEAKER_00: You can start to do things that were not cost competitive uh previously in a way that is, and on top of that, you can start to use um uh smartphones as the brain and interface to uh the physical world.
36:24
SPEAKER_00: And so we had that insight like really early on.
36:29
SPEAKER_00: Uh we chose a really difficult space, like kind of the toys and entertainment, where uh we actually got the pretty massive scale.
36:35
SPEAKER_00: We were doing over 100 million in annual revenue at peak, but it underestimated how painful and difficult that space is, where we were like 80% Q4.
36:44
SPEAKER_00: It was like very kind of like fickle, it was less familiar.
36:47
SPEAKER_00: So you're trying to kind of like recreate or invent a new category, um, you know, where um I think they were much cleaner and um more efficient paths into market.
36:56
SPEAKER_00: So you have you know beautiful solutions like a nest as a thermostat, but we invented a, you know, and a smarter, better interface, a smart door lock, a smart, you know, this or that.
37:05
SPEAKER_00: Um uh we uh ended up getting pretty great traction.
37:11
SPEAKER_00: But when I look at the technology today, I could argue that we might have been like uh early in a sense that some of the things that we were really pushing on, especially you know, towards the later years, was more interactive, intelligent kind of characters that you know started to like broaden out appeal and start to introduce function, moving into kind of like more broad use in indoor environments.
37:32
SPEAKER_00: That would be a completely different type of opportunity and challenge with today's technology than 10 years ago.
37:39
SPEAKER_00: And so uh we commercially got to like pretty good traction as a business that was incredibly difficult and challenging and seasonal.
37:47
SPEAKER_00: Um, and uh was like, you know, a lot of those lessons were actually, you know, combined with the Waymo lessons were actually ones that we carried forward into uh uh bedrock uh robotics today.
37:57
SPEAKER_01: Yeah, yeah.
37:58
SPEAKER_01: So yeah, the so the key four thing basically everybody's buying Christmas presents.
38:02
SPEAKER_00: Uh yeah, well, uh yes, and then we exacerbated it even more by launching on new products in September and October because you obviously want to launch them into the holiday season.
38:10
SPEAKER_00: And so then you're lining up.
38:13
SPEAKER_00: So by July and August, you're locking in your forecast for the year.
38:16
SPEAKER_00: And so that's your commitment on inventory.
38:18
SPEAKER_00: You're locking in marketing early, you're paying for your whole headcount for the year, and then you're like biting your nails until Thanksgiving.
38:26
SPEAKER_00: And then you like you hope that you blow it out in November, December, and then you get paid in December, January, February by the retailers, and then you repeat.
38:33
SPEAKER_00: And so it's just like the cash flow cycles were literally like could have been like$150 million from peak to valley um because of you know about how this works.
38:45
SPEAKER_00: And then you're spending more on inventory as you're scaling than all of revenue the year before.
38:49
SPEAKER_00: Um and it's a single-sale hardware product, which is not the best um shape of a business in terms of uh you know the value of something that's like recurring service versus um uh versus instantaneous sale.
39:00
SPEAKER_00: Uh and so it was um uh there was a period where anything that was like app connected physical kind of products was super, super hot.
39:10
SPEAKER_00: Around 2016, late 2016 and 2017, the whole space just cratered where Fitbit and GoPro went from$12 billion market cap and 10x revenue to like less than 1x revenue and lost like 95% of their value.
39:24
SPEAKER_00: And the whole space just like froze basically, where it became almost impossible to raise money for hardware.
39:30
SPEAKER_00: You had companies going out of business like Job Own and Pebble and everybody else.
39:35
SPEAKER_00: And so we survived for a bit, but it was like really painful because the whole space went from feast to famine and like and everything uh effectively froze.
39:45
SPEAKER_00: And so you kind of appreciate the exposure you have through these cycles uh when you're um you know, when you're in uh these spaces where um by nature of either the business model or the exposure to a sector, you're like actually more vulnerable than you might realize, despite the fact that you have a pretty sizable business.
40:02
SPEAKER_01: Yeah, I I'm just I'm I'm now sort of thinking about consumer in terms of um I'm thinking about it like back to the future.
40:12
SPEAKER_01: From the standpoint that uh like let's take iRobot again as uh let's take the RUBA as exam this example of like you know the very rare robot that that that you know made it across that finish line and had you know sustained success.
40:28
SPEAKER_01: And I think I think what it is is you you can get all of the math right, everything right, the price, everything else right, um, and that could not be enough.
40:40
SPEAKER_01: Like you have you still have to get that right, but also lightning has to strike at the same time.
40:46
SPEAKER_00: Yeah.
40:46
SPEAKER_00: And you realize that you might like until you start to diversify and get economies of scale on brand, on manufacturing, on everything, it's like very, very difficult.
40:54
SPEAKER_00: And so um I personally, so I'll tell you my takeaway from a lot of this.
41:00
SPEAKER_00: Uh um, I definitely appreciate how difficult the consumer market is, also in just the diversity of environments and people's like preference patterns, how price sensitive they are.
41:11
SPEAKER_00: Um, and you know, we had the number one you know toy in the you know in the US for uh you know kind of like a you know a few years, but it's very hard because we're selling like a$150 kind of product.
41:21
SPEAKER_00: You and you're competing against companies like Apple and others that have like kind of massive diversity and and and scale in terms of being able to command this sort of distribution.
41:29
SPEAKER_00: So when I look at like a novel space like humanoids today, the last place I expect it to be successful in these early waves is consumer because of how big of a delta there is from the price and capability and technical challenges that are still remaining to the value proposition that you have to serve to customers for them to be happy with this across this infinite diversity versus more industrial applications like assembly fulfillment, kind of whatever, where you could be intentionally narrower and deep, but provide value that's highly utilized.
42:00
SPEAKER_00: Um, and so I do think that after seeing consumer world from the Anculan's Nidrod robot to Waymo's consumer, and then the enterprise side of trucking, and then obviously the kind of construction and heavy machinery world today, there's a really beautiful property to B2B and kind of like enterprise products where you have a very rational buyer that might be conservative, but they're actually rational.
42:25
SPEAKER_00: And you can solve problems that have massive utilization with more well-defined surface area, and you have more flexibility in the cost structures and how you approach it.
42:36
SPEAKER_00: Um, and so I actually think that a lot of these technologies will follow the same flow when they'll begin um in these, you know, these like large-scale enterprise applications and then kind of like transition to consumer.
42:47
SPEAKER_00: And even in the case of Waymo, the end game is going to be personal car ownership, but that's not the first step when the technology's expensive enough to where you have to run your own kind of highway utilized robotaxi and you have geofences and all these sort of things.
43:00
SPEAKER_00: Um so, but it's just kind of an evolution of these technologies that then mature and eventually make it into the consumer side.
43:06
SPEAKER_00: But um, I do appreciate how challenging a lot of these extra dimensions are where consumers are not as rational as a business might be.
43:14
SPEAKER_01: The the major way in which um Anki and and Cosmo have have really stayed with me, when you guys were initially pitching the project, you know, you mentioned the fact, you know, you had all of the drawings and everything else that that you had brought animators from Pixar and DreamWorks on to animate his face.
43:33
SPEAKER_01: You know, and at the time it was just like, wow, that what a weird, you know, for me again, coming over from you know the the consumer electronic side, like what a weird place to spend all this money on.
43:45
SPEAKER_01: Um but but in a sense, I feel like that's that is kind of one of the major lasting legacies.
43:50
SPEAKER_01: Like that is something that like really did has taken off since then and and has like gotten a major foothold in a lot of these robots.
43:59
unknown: Yeah.
43:59
SPEAKER_00: And I mean We sold millions of units.
44:01
SPEAKER_00: Um, like this was like not a small uh like the first the year we launched, uh it was it ran out of stock so aggressively that we were the number one product on eBay, all of eBay, that holiday of 2016.
44:16
SPEAKER_00: So it was like pretty pretty fascinating.
44:18
SPEAKER_00: Um, but a lot of those learnings, and so you know that we had a Pixar studio coupled with the robotics team and the mechanical and kind of like physical design where we were animating a physical character, but instead of a story on Rails was interactive because it was responding to what it was seeing in real time.
44:36
SPEAKER_00: That has really valuable building blocks to human-robot interaction more broadly.
44:41
SPEAKER_00: Um, and when you get into functional applications like an armed humanoid, you know, robot in a home, uh that is the thing that allows them to coexist in a natural way, communicate intent, allows enables people to be forgiving of flaws because you kind of personify it.
45:00
SPEAKER_00: And so there's there are powerful elements there that um that carry over.
45:03
SPEAKER_00: And I think we're gonna see them play a bigger role in the next decade uh in some of these spaces.
45:08
SPEAKER_01: So we're running out of time, so we we can end on this, but you know, obviously, like Anki, Anki did have success, but you know, at the end of the day, like all these sort of headwinds and everything, um, you know, it ultimately didn't stand the test of time.
45:22
SPEAKER_01: Um you go over, you go over to Waymo, you're there for, you know, like three or four years.
45:28
SPEAKER_01: Uh and then, you know, it strikes me as it's as things are really picking up and as it's really sort of finally seeing success, then you're like, hey, let's sacrifice everything and launch another startup.
45:41
unknown: Yeah.
45:41
SPEAKER_00: So I was there for five years at Waymo.
45:43
SPEAKER_00: So it was um and it was like really fun.
45:45
SPEAKER_00: We got to got to wear a variety of hats.
45:47
SPEAKER_00: So leading various technology teams where he helped bring up the Jaguars, launch it in SF, start to scale it.
45:52
SPEAKER_00: We were like building out the trucking program.
45:54
SPEAKER_00: We led the first driverless release on freeways for the cars.
45:58
SPEAKER_00: And so there was such a um, it was an it was an amazing, it was like the highlight, like like just an absolute highlight of my career.
46:05
SPEAKER_00: Um, and one of the most powerful aspects is seeing how capable these technologies were, where you started to use machine learning and large-scale data to solve these problems, and how it would generalize across cities from San Francisco, Los Angeles, Austin, uh, Phoenix, uh, Atlanta, uh, from car to truck, from surface street to freeway, and have these huge subsidies where you started to snowball.
46:28
SPEAKER_00: We started to really think about how you could apply that in other areas where you have diversities of use cases and capabilities, diversities of platforms.
46:36
SPEAKER_00: And that's what took us to automating heavy machinery, where you have these machines that exist not just in construction, but also in mining, agriculture, lumber, garbage movement.
46:46
SPEAKER_00: We have these like specialized heavy machines that end up manipulate the world around them and have all of the same flavors.
46:53
SPEAKER_00: It felt to us that we were solving um at Waymo, and it felt like an amazing opportunity.
46:57
SPEAKER_00: And so um, and it wasn't just uh just a few of us.
47:00
SPEAKER_00: I think we have, I don't know, probably like a third of the company is uh from Waymo um and continues to kind of trickle in from some of our old colleagues and uh and friends.
47:09
SPEAKER_00: But um, there is something incredibly exciting about um starting something from scratch, really thinking about how to build on the best lessons and bypass the pain points and do this in a smaller company where everybody's you know sprinting and you're able to have a big impact.
47:24
SPEAKER_00: Um and so we've been at it for a little over two years now.
47:26
SPEAKER_01: Yeah, yeah.
47:27
SPEAKER_01: And and obviously uh there are some things that you I mean, you you have resources, don't get me wrong, with with all the rays, but there are there are some things that you can't do without the resources of Google, but there are a lot of things that you can do without the oversight of a Google.
47:40
SPEAKER_00: Yeah, it could be a double-edged sword.
47:42
SPEAKER_00: Funny enough, Google's an investor now, so they would co-ed our series B.
47:45
SPEAKER_00: So that came full circle as well.
47:46
SPEAKER_01: But uh that mad that you're you're you're stealing their employees.
47:51
SPEAKER_00: No, they're they're fine.
47:52
SPEAKER_00: And uh we were very we tried to be very thoughtful.
47:54
SPEAKER_00: And a lot of these were, I think, just people on our team that would just like love working with us, and so it was not not too surprising.
48:01
SPEAKER_00: But um I don't think autonomous driving on a public road is a startup game.
48:05
SPEAKER_00: Like I don't I I think you need the back pockets of an alphabet scale backer to solve this problem, or you're just exposed and praying for financial tailwinds that may or may not happen.
48:17
SPEAKER_00: And so I can't imagine anything more exhausting than trying to create an EV company from scratch that, you know, doing autonomous cars or trucks.
48:24
SPEAKER_00: This space, however, we're gonna go driverless not too far in the future this year, with fully uh starting with like excavators and fully driverless operator out machines doing real work for customers.
48:36
SPEAKER_00: And we'll end up doing this in a little bit over two years since the start of the company, and for a small fraction of the cost that it took Waymo to get the first um driverless deployment.
48:45
SPEAKER_00: And um, and so we've raised the money we needed to raise.
48:48
SPEAKER_00: We were oversubscribed in every single round that we've had.
48:51
SPEAKER_00: Um but I think we've found a way to do it using leveraging all of the learnings of the past and then also this industry, where we can actually be more efficient about it.
49:01
SPEAKER_00: And even though we're moving towards the broader opportunity that actually is of a scale of transportation, when you think of all heavy machinery and what they do and the ability to orchestrate teams of machines working together and completely change the operations of a general contractor, this is like these industries are 20% of the world's GDP when you count construction and mining and agriculture.
49:23
SPEAKER_00: And these are the workhorses.
49:25
SPEAKER_00: So this is a big, big, big problem.
49:26
SPEAKER_00: But we can do it in a way that can let us start to build a business earlier than we could at Waymo and then grow it while we pursue these bigger and bigger opportunities and scale the um the diversity of use cases and start this flyway with new capabilities and new machines.
49:42
SPEAKER_01: Well, Boris, uh, it's been a pleasure.
49:45
SPEAKER_01: Thanks for thanks for catching up and uh thanks for joining us.
49:48
SPEAKER_00: Yeah, Brian, it's a lot of fun.
49:50
SPEAKER_00: Thank you for having me.
49:51
SPEAKER_00: Uh enjoyed talking about the uh end of the history of these spaces.
49:55
SPEAKER_01: Thank you so much to Boris and Bedrock.
49:58
SPEAKER_01: Thanks to you as always for staying around.
50:00
SPEAKER_01: Please like and subscribe on the platform of your choice.
50:04
SPEAKER_01: And don't forget to check out the newsletter for more over at automated.fm.
50:09
SPEAKER_01: That's all we got for you now.
50:11
SPEAKER_01: We will see you this time next week for another episode of Automated.