Latest / What's Your Problem? / Could Autonomous Diggers Unleash a Building Boom?

Transcript

Could Autonomous Diggers Unleash a Building Boom? · What's Your Problem?

• 0:15
  Speaker 1: Pushkin.
• 0:20
  Speaker 2: When we started this show in twenty twenty two, the
• 0:24
  Speaker 2: standard line about driverless cars was driverless cars have been
• 0:28
  Speaker 2: five years away for the past fifteen years, because it
• 0:31
  Speaker 2: seemed like they were just always around the corner, always
• 0:35
  Speaker 2: just a few years away, but they never quite arrived.
• 0:39
  Speaker 2: Nobody says that anymore. Today, in several cities around the country,
• 0:43
  Speaker 2: getting a ride from a driverless car is just a
• 0:46
  Speaker 2: normal thing people do, and it'll become normal in more
• 0:49
  Speaker 2: and more and more cities over the next few years.
• 0:52
  Speaker 2: Driverless cars are here now, so now we can ask
• 0:56
  Speaker 2: what's next. I'm Jacob Goldstein and this is What's Your Problem,
• 1:05
  Speaker 2: the show where I talk to people who are trying
• 1:07
  Speaker 2: to make technological progress. My guest today is Boris Soffman.
• 1:11
  Speaker 2: He's the co founder and CEO of Bedrock Robotics, a
• 1:15
  Speaker 2: company that's figuring out how to retrofit heavy equipment to
• 1:18
  Speaker 2: make it work autonomously. Boris's problem is this, how do
• 1:23
  Speaker 2: you teach machines not just to drive, but to do
• 1:25
  Speaker 2: things like grade roads and move heavy things around construction sites.
• 1:31
  Speaker 2: Boris's company is starting with excavators, and they plan to
• 1:34
  Speaker 2: have their first commercial excavators autonomously digging holes on construction
• 1:39
  Speaker 2: projects next year. Later in the interview, Boris goes big.
• 1:43
  Speaker 2: He argues that if Bedrock succeeds, the company could help
• 1:47
  Speaker 2: push forward a broad wave of new building in America.
• 1:51
  Speaker 2: But first we talked about his time at Weimo and
• 1:55
  Speaker 2: how the wild evolution of the autonomous vehicles that he
• 1:58
  Speaker 2: worked on there led him to start Bedrock.
• 2:01
  Speaker 1: The time at Weimo was this incredible period where I
• 2:04
  Speaker 1: was there for about five years from mid twenty nineteen
• 2:07
  Speaker 1: till through spring twenty four, and that was this really
• 2:10
  Speaker 1: big period where it was going through this like fifteen
• 2:13
  Speaker 1: years of R and D, and then it finally transitioned
• 2:16
  Speaker 1: into this hockey stick of like growth that is helping today.
• 2:19
  Speaker 1: And so today waymos at over one hundred million miles
• 2:22
  Speaker 1: fully driver lists. It's at five times safer than a human.
• 2:25
  Speaker 1: It's millions of miles every single week, and so it's
• 2:27
  Speaker 1: kind of scaling itsponentially and it's like a genuinely fantastic product.
• 2:30
  Speaker 1: But I was there when we were like stressing over
• 2:32
  Speaker 1: the first hundred miles, and it felt like the most
• 2:35
  Speaker 1: incredible achievement to just go like ten fifty hundred miles
• 2:38
  Speaker 1: to completely drop less, say now that happens at like
• 2:40
  Speaker 1: hundreds of thousands of miles every single day. And so
• 2:44
  Speaker 1: one of the things that really broke through it made
• 2:46
  Speaker 1: that possible is this shift to machine learning and data
• 2:49
  Speaker 1: driven approaches as a core of the autonomy stack.
• 2:52
  Speaker 2: And just to be clear, like that's as opposed to
• 2:54
  Speaker 2: a more like euristics or rule based kind of model,
• 2:58
  Speaker 2: like the old school twentieth century eyes, like, well, just
• 3:01
  Speaker 2: tell the car all the rules of how to drive
• 3:04
  Speaker 2: and then it'll drive. Like that's what you're comparing machine
• 3:07
  Speaker 2: learning to.
• 3:08
  Speaker 1: Yeah, like drive, I'm going to like embed the cost
• 3:10
  Speaker 1: functions and all this. So yeah, so like large scale
• 3:13
  Speaker 1: search and heuristics and rules and you can embed them
• 3:16
  Speaker 1: now inside those heuristics, and so you can solve almost
• 3:18
  Speaker 1: any given problem with that sort of approach.
• 3:21
  Speaker 2: So you can solve one problem, but you can't solve
• 3:24
  Speaker 2: like every single possible problem that would ever arise when
• 3:27
  Speaker 2: you're driving, which is actually what you have to solve
• 3:30
  Speaker 2: to do full autonomy.
• 3:31
  Speaker 1: Right right with activating whackamall, where like you fix one
• 3:34
  Speaker 1: problem and it becomes harder and harder to kind of
• 3:35
  Speaker 1: scale the other ones. And so and so there was
• 3:38
  Speaker 1: this really conscious shift at WEIMA, which was kind of
• 3:41
  Speaker 1: bold at the time because it feels obvious in hindsight,
• 3:43
  Speaker 1: It wasn't obvious at all back then. The shift to
• 3:46
  Speaker 1: like really embracing this as a data driven solution where
• 3:49
  Speaker 1: you're learning from human driving and human behavior.
• 3:52
  Speaker 2: And when you say you're learning, you mean the machine
• 3:55
  Speaker 2: learning model. The AI basically is learning.
• 3:58
  Speaker 1: The machine learning model, that's right. And so you're basically
• 4:01
  Speaker 1: taking giant scales of data one hundreds of thousands of
• 4:03
  Speaker 1: millions of miles, and you're learning the model of how
• 4:06
  Speaker 1: do you drive and how do you interpret this? Like
• 4:09
  Speaker 1: infinite compocity and kitchen sink of contacts, like all of
• 4:12
  Speaker 1: this sensored data, the road structures, the things you see
• 4:14
  Speaker 1: around you, the way people are moving. How do you
• 4:17
  Speaker 1: go from a kind of engineered and you know, kind
• 4:20
  Speaker 1: of trained solution into one that is like dominantly a
• 4:23
  Speaker 1: warned solution.
• 4:24
  Speaker 2: I mean, as you're doing that, you're sort of riding
• 4:26
  Speaker 2: the historic machine learning AI wave, right, Like that's right.
• 4:31
  Speaker 2: Presumably you're able to do that at that moment because
• 4:34
  Speaker 2: of this explosion in AI, which is basically machine learning, right,
• 4:38
  Speaker 2: you know that.
• 4:38
  Speaker 1: And you couldn't have done that five ten years ago.
• 4:40
  Speaker 2: That's what you're kind of coming out of at way MO,
• 4:43
  Speaker 2: that's driving this success at way MOO. How does that
• 4:46
  Speaker 2: set you up for what you're doing a bedrock?
• 4:48
  Speaker 1: The most shocking thing was just how well does generalized
• 4:50
  Speaker 1: from San Francisco to Los Angeles, Phoenix in Austin and
• 4:53
  Speaker 1: then becomes almost like a qualification problem eventually, where you're
• 4:56
  Speaker 1: using data to fill some gaps, but like you need
• 4:58
  Speaker 1: less and less of it, less and less new things
• 5:00
  Speaker 1: surprise you, like you're just your competency kind of expands.
• 5:03
  Speaker 1: And then we unified the technology stack between cars and
• 5:06
  Speaker 1: trucks where even jumping from a cargo a truck was
• 5:09
  Speaker 1: needed maybe ten to fifty percent more data, but it
• 5:11
  Speaker 1: was fundamentally like you're using data to explain why how
• 5:13
  Speaker 1: you operate a very different potages.
• 5:15
  Speaker 2: This is like a big truck, This is like an
• 5:16
  Speaker 2: eighteen wheel truck. When Google was working on that, yeah,
• 5:19
  Speaker 2: are way most like.
• 5:19
  Speaker 1: Fifty three foot trailer like eighty thousand pounds. And so
• 5:23
  Speaker 1: that was the big moment where we started thinking about
• 5:27
  Speaker 1: where else can you apply this? What are the places
• 5:28
  Speaker 1: where you have all this diversity of challenges and capabilities
• 5:32
  Speaker 1: that benefit from this type of versatility and also have
• 5:35
  Speaker 1: the ability to jump between platforms in this really natural way.
• 5:38
  Speaker 1: And so we looked at a lot of spaces and
• 5:41
  Speaker 1: really settled on automation of specialized typing machinery, and so
• 5:44
  Speaker 1: the sipes of machines that you see in construction, like
• 5:47
  Speaker 1: excavators and wheeloaders and bulldozers, but also frankly the sort
• 5:50
  Speaker 1: of machines you see in all sorts of industries like
• 5:53
  Speaker 1: agriculture and mining and lumper and garbage movement. And so
• 5:57
  Speaker 1: you have these very diverse types of machines that are
• 6:00
  Speaker 1: interacting with the world around them. They're fairly still moving,
• 6:03
  Speaker 1: they're in semi controlled environments, and at the end of
• 6:05
  Speaker 1: the day, there's astronomical scale at which they great at
• 6:09
  Speaker 1: and a lot of the learnings that we experienced a
• 6:12
  Speaker 1: weymle actually transfer over incredibly well. But the physics of
• 6:15
  Speaker 1: the problem is a lot less aperspherial. It's actually like,
• 6:17
  Speaker 1: really tackle this and get the market.
• 6:19
  Speaker 2: Basically, big vehicles operating in semi constrained environments doing things
• 6:26
  Speaker 2: to the world, interacting with the world in some physical
• 6:28
  Speaker 2: way in addition to just driving across it.
• 6:30
  Speaker 1: That's right, and these are slow moving vehicles where you're
• 6:33
  Speaker 1: already on a closed site with people who are assumed
• 6:36
  Speaker 1: to be knowledgeable about the world around you, and you're
• 6:39
  Speaker 1: moving at like five off an hour, for example, You're
• 6:41
  Speaker 1: able to slow down and stop. You're able to minimize
• 6:43
  Speaker 1: your exposure. You always have a minimum safety condition of stopping.
• 6:46
  Speaker 1: Your complexity is less abound interactions with others on the
• 6:49
  Speaker 1: road and more about the interactions with the world around you.
• 6:52
  Speaker 1: And so you can actually tackle safety not through this
• 6:55
  Speaker 1: sort of statistical methods that drove a lot of the
• 6:58
  Speaker 1: mileage that we collect, but through a much more direct
• 7:03
  Speaker 1: measure of your competencies in order to just make sure
• 7:05
  Speaker 1: that you're like you're actually capable of hurting somebody.
• 7:08
  Speaker 2: Yeah, more like like an industrial robot al most right.
• 7:12
  Speaker 1: That's right. Yeah, it's like a traditional Susis engineering, Yeah.
• 7:15
  Speaker 2: Where it's like, look, even if it can't do the
• 7:17
  Speaker 2: thing every time, that's fine. Just make sure it's not
• 7:19
  Speaker 2: gonna like go crazy and kill somebody.
• 7:21
  Speaker 1: That's right. You can make it to that. Like your
• 7:22
  Speaker 1: worst case is your productivity suffers, but safety wise, you're
• 7:25
  Speaker 1: you're you're solid. And so that's like incredibly enabling because
• 7:30
  Speaker 1: your long tail is now no longer safety. It's the
• 7:32
  Speaker 1: versatility what you can do.
• 7:33
  Speaker 2: Why did you start with excavators?
• 7:36
  Speaker 1: So excavators are the most highly utilized machine, Like they're
• 7:40
  Speaker 1: usually the highest volume in fleet. So between twenty and
• 7:43
  Speaker 1: twenty five percent of fleets are excavators.
• 7:45
  Speaker 2: Fleets are just like big construction heavy equipment.
• 7:48
  Speaker 1: Yeah, like a general contractor that alone, like a thousand
• 7:51
  Speaker 1: machines two hundred, two hundred and fifty will probably be
• 7:54
  Speaker 1: excavators on average.
• 7:55
  Speaker 2: And it's basically what a kid would call it digger, right,
• 7:57
  Speaker 2: It's like at a digger bucket an arm, and it
• 8:00
  Speaker 2: like digs stuff up.
• 8:01
  Speaker 1: It's like the equivalent of an arm. Like you can
• 8:03
  Speaker 1: like dig stuff, you can demolish stuff, you can swap
• 8:05
  Speaker 1: your tools, you can like lift pipes and put them
• 8:08
  Speaker 1: in a holes. They can do a ton of stuff
• 8:09
  Speaker 1: with it. It's kind of crazy. It really is a
• 8:11
  Speaker 1: personal machine. It also makes it very complicated to work.
• 8:13
  Speaker 1: So there's like seven degrees of freedom sometimes eight, and
• 8:16
  Speaker 1: so it's one of the hardest machines to learn, and
• 8:18
  Speaker 1: it takes four to five years to really become an expert.
• 8:20
  Speaker 1: And there's a huge difference between an expert and novice,
• 8:22
  Speaker 1: and so you kind of have this situation where it's
• 8:25
  Speaker 1: a huge volume of work and it's really hard to learn,
• 8:28
  Speaker 1: and so you have a really deep pull in the
• 8:30
  Speaker 1: market for it, meaning.
• 8:31
  Speaker 2: A lot of demand, like a lot of people want
• 8:33
  Speaker 2: somebody who can drive one of these or a machine
• 8:36
  Speaker 2: that could do it.
• 8:37
  Speaker 1: Well, let me tell you about the demand. I've never
• 8:39
  Speaker 1: seen such a divergence of supply to demand like in
• 8:42
  Speaker 1: my career in any case, where on the demand side,
• 8:45
  Speaker 1: you have this astronomical construction industry that's already two trillion
• 8:48
  Speaker 1: dollars a year in the US, that's obviously very heavily
• 8:51
  Speaker 1: building and having machinery work. And then you have this
• 8:54
  Speaker 1: like shortage of operators that already existed, but it's going
• 8:57
  Speaker 1: in the wrong direction where forty percent of construction workers
• 8:59
  Speaker 1: is retiring in the next ten years. Our partners are
• 9:02
  Speaker 1: consistently having trouble filling labor. We've met some that have
• 9:07
  Speaker 1: one hundred percent turnover.
• 9:09
  Speaker 2: Everybody leaves every year.
• 9:10
  Speaker 1: It means like a third of people might be lifers,
• 9:13
  Speaker 1: but then like two thirds transition more than once per year,
• 9:15
  Speaker 1: and you're constantly backfilling the skill sets very a ton
• 9:19
  Speaker 1: and one of them said that for every one person
• 9:22
  Speaker 1: entering the workforce of the quality that they look for,
• 9:24
  Speaker 1: their seven leaving. And so we end up having is
• 9:27
  Speaker 1: this shortage of ability to meet this demand, and so
• 9:30
  Speaker 1: prices go up, jobs don't get done. And what's interesting
• 9:33
  Speaker 1: is it's not that's like isolated industry that's like just
• 9:36
  Speaker 1: on its own kind of like having these sort of challenges.
• 9:38
  Speaker 1: It's a horizontal it supports every industry. You can't build
• 9:41
  Speaker 1: data centers free eye without it, you can't build houses,
• 9:44
  Speaker 1: you can't build energy facilities.
• 9:46
  Speaker 2: Of like a rate limiting skill or rate limiting machine.
• 9:49
  Speaker 1: It's a whole country. That's exactly right. And then you
• 9:51
  Speaker 1: have this need that starts with labor. But then there's safety.
• 9:54
  Speaker 1: It's huge amounts of safety challenges. You have, you know,
• 9:57
  Speaker 1: huge predictability challenges. So if you actually could soften out
• 10:01
  Speaker 1: some of these constraints, we would build more, more work
• 10:04
  Speaker 1: would get done, and it would like stimulate the whole economy.
• 10:06
  Speaker 1: And so that's what's actually pretty exciting about this opportunity.
• 10:09
  Speaker 1: It's not as yours some game at all good.
• 10:12
  Speaker 2: So, like you decide to focus on excavators, you you know,
• 10:14
  Speaker 2: you raise money for your company, Like, do you go
• 10:17
  Speaker 2: out and buy a whatever, a million dollar excavator. You
• 10:21
  Speaker 2: go to what is it bucket and shovel dot com
• 10:23
  Speaker 2: and buy yourself an excavator.
• 10:25
  Speaker 1: They're like three hundred thousand to five hundred thousand, so
• 10:28
  Speaker 1: they're like still expensive. Yeah, they're pretty expensive.
• 10:30
  Speaker 2: So I mean like, did you buy one like we did?
• 10:32
  Speaker 1: Yeah?
• 10:33
  Speaker 2: Did you drive it?
• 10:34
  Speaker 1: Of course, like you have to. It's like it's like
• 10:36
  Speaker 1: a ride of passage. Yeah, they're really fine. I even
• 10:38
  Speaker 1: took a lessoners. There's a place in Vegas ALESSI drive
• 10:41
  Speaker 1: excavators and take away and from like a trade operator.
• 10:44
  Speaker 1: It was pretty fun.
• 10:44
  Speaker 2: So, oh, that's genius. You get to break things. I
• 10:47
  Speaker 2: was actually as I was walking here today to the
• 10:50
  Speaker 2: train that there's a playground and there was an excavator
• 10:53
  Speaker 2: just like breaking up the asphalt and then like prying
• 10:57
  Speaker 2: it up. I was like, that looks awesome.
• 10:59
  Speaker 1: This is the funniest thing. Anybody that gets an excavator,
• 11:02
  Speaker 1: like your inner six year old comes out. Suddenly you
• 11:04
  Speaker 1: have like the most mature, sophisticated person is trying to
• 11:07
  Speaker 1: be all professional. You get an excavator and just like
• 11:09
  Speaker 1: get like a giant glob of mud and then like
• 11:12
  Speaker 1: bring as hig as a can and then like PLoP
• 11:14
  Speaker 1: it down and see what happens. It's amazing, and it's like,
• 11:16
  Speaker 1: without fail, everybody kind of reverts back to this sort
• 11:18
  Speaker 1: of a world. Yeah, we drove and I think we
• 11:20
  Speaker 1: bought like a half dozen excavators at this point, and
• 11:23
  Speaker 1: then we also then use a lot of excavators from
• 11:25
  Speaker 1: our partners who are general contractors and subcontractors.
• 11:28
  Speaker 2: And it's just what a year or so ago that
• 11:30
  Speaker 2: you started, like not that long in.
• 11:32
  Speaker 1: This less than a year and a half, yeah, like
• 11:34
  Speaker 1: last last minute, seah, it's a pretty been a good run.
• 11:37
  Speaker 2: How much is sort of commodified of autonomy, right? How
• 11:40
  Speaker 2: much is just like, well, we're gonna buy this, this,
• 11:42
  Speaker 2: and this in terms of sort of hardware and we
• 11:44
  Speaker 2: know the software, and then how much is like, oh,
• 11:46
  Speaker 2: here's the things we have to figure out that nobody
• 11:48
  Speaker 2: knows how to do.
• 11:49
  Speaker 1: So it's one of the other enablers that's like way
• 11:52
  Speaker 1: better than what we would have had to go through
• 11:54
  Speaker 1: ten years ago. In the space, we can use a
• 11:56
  Speaker 1: lot of the existing components on lighter on cameras, on
• 11:59
  Speaker 1: Imus GPS, there's a lot of tailwind from automotive great
• 12:04
  Speaker 1: cameras and compute.
• 12:05
  Speaker 2: That's like accelerating and like presumably they're buying those things
• 12:08
  Speaker 2: at scale you can like you've got to go get
• 12:10
  Speaker 2: them cheap. Essentially, you're like, give me one of those,
• 12:13
  Speaker 2: one of those, one of those.
• 12:14
  Speaker 1: That's right, because you're gonna go to millions and millions
• 12:16
  Speaker 1: of units as like every car against an autopilot equivalent
• 12:19
  Speaker 1: over the next two three years. Oh, that's starting to
• 12:21
  Speaker 1: kind of pump in a cost. So that helps. And
• 12:23
  Speaker 1: then even the platform, we can retrofit existing machinery.
• 12:27
  Speaker 2: When you say the platform, what do you mean in
• 12:29
  Speaker 2: this context.
• 12:30
  Speaker 1: Platform is like the car, the truck, the excavator, and
• 12:33
  Speaker 1: the way in this case the excavator. Yeah, the machine,
• 12:35
  Speaker 1: so like the excavator itself. Construction machines, particularly this latest generation,
• 12:40
  Speaker 1: they're really well designed to where they're already effectively drive
• 12:44
  Speaker 1: by wire, which means that every signal going through the
• 12:46
  Speaker 1: machine is electric, and so we're able to spice into
• 12:49
  Speaker 1: it and both read and write to these signals.
• 12:54
  Speaker 2: And control them electric meaning like computerized basically.
• 12:58
  Speaker 1: Meaning like you have a joystick that operates the excavatory
• 13:01
  Speaker 1: that's not physically connected to the hydraulics. It's an electric
• 13:04
  Speaker 1: signal that's connected to the hydraulics, and so the same
• 13:06
  Speaker 1: for every signal and the machine, like the sense the
• 13:09
  Speaker 1: pressure gages, and so we're able to both get the
• 13:12
  Speaker 1: data from the machine as well as control the machine
• 13:16
  Speaker 1: through a non invasive integration where we can upit a
• 13:20
  Speaker 1: machine with our sensors and compute suite in like less
• 13:22
  Speaker 1: than three hours and make it autonomy capable and it's
• 13:26
  Speaker 1: completely reversible. We could never do that with a car
• 13:28
  Speaker 1: or a truck because the platforms were just not designed
• 13:31
  Speaker 1: this way.
• 13:32
  Speaker 2: Yeah, so there's a lot that's there already. The excavators
• 13:35
  Speaker 2: themselves are sort of easy to integrate with lots of
• 13:38
  Speaker 2: off the shelf technology. What's not there, Like, when you're
• 13:41
  Speaker 2: coming in, what do you have to sort of build
• 13:43
  Speaker 2: that nobody has built before?
• 13:44
  Speaker 1: Nobody has actually created autonomy that can solve the really
• 13:49
  Speaker 1: nuanced problems that you need to solve in order to
• 13:51
  Speaker 1: operate like an excavator in construction tasks.
• 13:55
  Speaker 2: So let's talk like specifically, you buy your excavators, you
• 13:59
  Speaker 2: got your hardware, you know, whatever, what basic AI model
• 14:02
  Speaker 2: you're going to use, But like, what's a specific thing.
• 14:05
  Speaker 2: You have to figure out well a.
• 14:07
  Speaker 1: Lot of things. So first of all, it's not trivial
• 14:08
  Speaker 1: to tap into these machines and to build a platform
• 14:12
  Speaker 1: or kind of an integration around it where you can
• 14:15
  Speaker 1: read them, you can control them, and so you have
• 14:17
  Speaker 1: to basically create like a wrapper around the machine and
• 14:19
  Speaker 1: also design in a way that scales later to new machines. Right,
• 14:22
  Speaker 1: So that part's hard. It's a big autonomy problem. So
• 14:25
  Speaker 1: in that respect is no different than what we tackle
• 14:27
  Speaker 1: with a WEIMO, where you have to train massive scale models.
• 14:30
  Speaker 1: You have to collect a huge amount of data.
• 14:31
  Speaker 2: And in this case it's like about the digging, like
• 14:34
  Speaker 2: it's about the digging dumb question? Is the digging the
• 14:37
  Speaker 2: hard part? Like what happens when the thing hits the ground.
• 14:39
  Speaker 2: There's different kinds of stuff in the ground.
• 14:41
  Speaker 1: Like tell me in the category of what's new, suddenly
• 14:44
  Speaker 1: a car doesn't change the environment around it. You have
• 14:47
  Speaker 1: a really complicated tough earth and soil, and you got
• 14:50
  Speaker 1: to figure out the physics of how you dig through it,
• 14:53
  Speaker 1: how do you deal with clay versus top soil, how
• 14:56
  Speaker 1: do you deal with rocks inside? Then so, now if
• 14:58
  Speaker 1: you want to use simulation to solve parts of these problems.
• 15:00
  Speaker 1: You have a very complicated simulation problem because you have
• 15:03
  Speaker 1: to solve not just the sensor side, but also the
• 15:05
  Speaker 1: manipulation side. You have to structure this problem in a
• 15:09
  Speaker 1: way where you're learning from the data you collect in
• 15:13
  Speaker 1: order to actually capture the nuances of how do you
• 15:15
  Speaker 1: actually interact with the environment. You have to then control
• 15:18
  Speaker 1: and execute it the right way. You have to think
• 15:20
  Speaker 1: about how do you actually define the goal?
• 15:22
  Speaker 2: So from the user point of view with an excavator,
• 15:25
  Speaker 2: like how does it look like? I just I want
• 15:27
  Speaker 2: to dig a hole of this size at this spot?
• 15:29
  Speaker 2: Or like what is it?
• 15:30
  Speaker 1: Yeah, and this is a journey like fast forward you
• 15:32
  Speaker 1: know next year where card customer does this right, So
• 15:35
  Speaker 1: what they would do is they would give it a
• 15:38
  Speaker 1: model of what they want the earth to be dug
• 15:40
  Speaker 1: towards and so that would be like a three D
• 15:42
  Speaker 1: representation of the depth. So with length depth, here's the edges,
• 15:47
  Speaker 1: here's the constraints. The pickups for the trucks are going
• 15:50
  Speaker 1: to be over here and then go to work. And
• 15:52
  Speaker 1: basically these projects, like these machines will work for many
• 15:56
  Speaker 1: months at a time, continually just digging earth and working
• 15:59
  Speaker 1: it towards the right foundation, and the system will respect
• 16:01
  Speaker 1: that boundary and basically dig down to that level. It'll
• 16:05
  Speaker 1: have precision to the edges that you've defined, it'll do
• 16:08
  Speaker 1: it in a sequence that makes sense so that you
• 16:10
  Speaker 1: don't trap yourself in a hole, for example. And then
• 16:12
  Speaker 1: you specify where you're going to have dump truck pickups.
• 16:16
  Speaker 1: Then there's projects where this happens for nine months straight
• 16:19
  Speaker 1: or twelve months straight with like many machines, and so
• 16:22
  Speaker 1: what you basically need to specify is what you want
• 16:25
  Speaker 1: to dig to, and that what you want to dig
• 16:27
  Speaker 1: to ends up basically being the foundation that you're working
• 16:31
  Speaker 1: towards for whatever you're going to construct there. And we
• 16:33
  Speaker 1: want that to be versatile. So sometimes you're just taking
• 16:36
  Speaker 1: off a layer of top soil. Sometimes you're digging eight
• 16:38
  Speaker 1: feet deep. Sometimes you're taking an existing stockpile and moving
• 16:42
  Speaker 1: in somewhere else. So there's a lot of permutations of this.
• 16:45
  Speaker 1: Sometimes you're taking rubble from a demolition job and loading
• 16:48
  Speaker 1: it onto trucks, right, and so what's nice is that
• 16:51
  Speaker 1: you start to see patterns over and over again, and
• 16:53
  Speaker 1: this sort of work.
• 16:58
  Speaker 3: We'll be back in just a minute.
• 17:11
  Speaker 2: So the data problem is interesting here, right, Like you
• 17:14
  Speaker 2: were talking about the physics, just the physics of an
• 17:16
  Speaker 2: excavator is quite different, right, Like it's pushing down on
• 17:19
  Speaker 2: the ground, which is pushing back on the excavator, and
• 17:21
  Speaker 2: like as you say, the ground is changing because of
• 17:24
  Speaker 2: its work, and there's not Well where do you get
• 17:28
  Speaker 2: the data?
• 17:29
  Speaker 1: Data is actually an interesting mix. We get it both
• 17:31
  Speaker 1: on our test sites and also with our design partners.
• 17:35
  Speaker 1: So we were working with general contractors and subcontractors. Today
• 17:39
  Speaker 1: we have five that we're partnered with across southern states
• 17:42
  Speaker 1: like Arizona, Texas, and so we're getting data on their size.
• 17:44
  Speaker 1: We're getting data on our sites.
• 17:46
  Speaker 2: It's not that much data, right, Like I mean, I
• 17:48
  Speaker 2: guess my reference point is always like image net or
• 17:52
  Speaker 2: you know, the Internet for large language models. It does
• 17:54
  Speaker 2: seem like a sort of recurring problem in robotics type
• 17:57
  Speaker 2: AI applications is sparsity of data.
• 18:01
  Speaker 1: How do you get it?
• 18:02
  Speaker 2: Yeah?
• 18:03
  Speaker 1: Yeah, And so this is where there's a few kind
• 18:06
  Speaker 1: of intersuties. So first of all, like our partners together
• 18:08
  Speaker 1: have thousand and thousands of machines and so there's a
• 18:10
  Speaker 1: lot of choices of which she's a partner. On the
• 18:12
  Speaker 1: other thing that you can do is actually you can
• 18:14
  Speaker 1: be very clever on a test site, and when you're
• 18:17
  Speaker 1: on a real project, you're kind of getting a unbiasample.
• 18:21
  Speaker 1: You're just getting a random distribution of the things you see,
• 18:24
  Speaker 1: just like driving around on a road. When you're on
• 18:26
  Speaker 1: a test site, you can actually up sample the things
• 18:29
  Speaker 1: you actually want and you can go and you can
• 18:31
  Speaker 1: collect ten hours of data. There's representative of five thousand
• 18:35
  Speaker 1: hours of random data, but which is particularly useful for
• 18:38
  Speaker 1: things like safety situations. Right, you can actually create a
• 18:41
  Speaker 1: much larger equivalent amount of data on a close course
• 18:45
  Speaker 1: through like kind of structure testing. And so for example,
• 18:47
  Speaker 1: safety scenarios where you have weird interactions with people doing
• 18:50
  Speaker 1: things they shouldn't do. You don't wait to see that
• 18:53
  Speaker 1: on an open site.
• 18:55
  Speaker 2: What is one of those, like what is your nightmare
• 18:57
  Speaker 2: human behavior scenario?
• 18:59
  Speaker 1: In the field? Night very human behavior is a human
• 19:02
  Speaker 1: is curious, they walk up to your machine, Your machine stops,
• 19:05
  Speaker 1: and then they get really really close and they are
• 19:07
  Speaker 1: now in a blind spot where you can't see that.
• 19:09
  Speaker 1: But you still have to be smart enough to track
• 19:11
  Speaker 1: them where they're in a hole in front of you
• 19:12
  Speaker 1: while you're like thinking about digging, right, So occlusions from
• 19:16
  Speaker 1: humans is probably a huge category which is called plex.
• 19:20
  Speaker 2: Occlusions, meaning in your blind spot humans in a place
• 19:23
  Speaker 2: where the machine can't see them.
• 19:25
  Speaker 1: Yeah, or usually your number one priority is anything that
• 19:28
  Speaker 1: touches on human safety. That's sacred. You never take any
• 19:31
  Speaker 1: risks on that.
• 19:32
  Speaker 2: What's something you haven't figured out yet?
• 19:34
  Speaker 1: The things people do with excavators, Like we've seen them
• 19:38
  Speaker 1: do bizarre like clearing of debris. We've seen them load
• 19:43
  Speaker 1: a wheeloader with dirt. We've seen them bang tools to
• 19:48
  Speaker 1: change them. We've seen them bang tools to change them.
• 19:51
  Speaker 1: What's that one? Like a tool gets stuck and they
• 19:54
  Speaker 1: like bang it on the ground in order to get
• 19:56
  Speaker 1: it loose. They use their arm as a pivot point
• 19:59
  Speaker 1: to turn when the wheels are like stuck in mud.
• 20:02
  Speaker 2: It's kind of baller, Like it's pretty baller.
• 20:04
  Speaker 1: Yeah, it's like I mean, like when you see the
• 20:06
  Speaker 1: expert operators, they just show off and it's like it's
• 20:08
  Speaker 1: increa like they're awesome. And so there's like all these
• 20:11
  Speaker 1: like weird subtleties of how they'll use these tools in
• 20:14
  Speaker 1: like really subtle ways, which like the dimensions of on
• 20:18
  Speaker 1: the product side of the use cases that blew was away.
• 20:21
  Speaker 1: We thought about the obvious ones, but as we started
• 20:24
  Speaker 1: like really going deeper and like studying this, there's so
• 20:28
  Speaker 1: much diversity and interesting things that you can do with
• 20:30
  Speaker 1: these machines. It's quite powerful.
• 20:32
  Speaker 2: I mean, presumably you don't have to do all those
• 20:34
  Speaker 2: to have a product people will pay you for. Right,
• 20:37
  Speaker 2: you can just have the like competent automaton that can
• 20:41
  Speaker 2: dig a big hole just the way you want it.
• 20:44
  Speaker 1: That's correct, and that there's a huge amount of work
• 20:46
  Speaker 1: even in those areas. But there is this like tale
• 20:49
  Speaker 1: that you go in you over time go and add
• 20:50
  Speaker 1: more and more capability, and then I think it is
• 20:52
  Speaker 1: just as softwareupdates and you get more. It's just like
• 20:54
  Speaker 1: a you know, the product ends up being the digital
• 20:57
  Speaker 1: operator which gets better over time, and so it's similar
• 21:00
  Speaker 1: to human getting better.
• 21:01
  Speaker 2: What's the business model?
• 21:03
  Speaker 1: So we will go operator out next year. So that'll
• 21:06
  Speaker 1: be our first, like fully operator list product that is
• 21:09
  Speaker 1: in a spirit of like what this is meant to scale.
• 21:11
  Speaker 2: As operator out means driver list means autonomous. Is that
• 21:15
  Speaker 2: what it means?
• 21:16
  Speaker 1: Yeah, nobody there, It's just like operator lists. Yeah, and
• 21:18
  Speaker 1: so our product is the digital operators. So what I
• 21:22
  Speaker 1: mean by this is that our customer is a general
• 21:24
  Speaker 1: contractor or subcontractor, so it's the companies that already buy
• 21:28
  Speaker 1: and manage these machines we are we sell them a
• 21:32
  Speaker 1: retrofit of these machines, so we install an upfit that
• 21:35
  Speaker 1: adds sensors and compute.
• 21:36
  Speaker 2: So the contractor already owns the excavator.
• 21:39
  Speaker 1: That's right, They've already buy a five hundred thousand dollars
• 21:42
  Speaker 1: machine or three hundred thousand dollars machine, and so this
• 21:44
  Speaker 1: is an upgrade that enables autonomy. And then our business
• 21:48
  Speaker 1: model is selling labor, So we're effectively selling the labor
• 21:52
  Speaker 1: and then a variety of digital services around it that
• 21:55
  Speaker 1: operates as machine. And for the surface area of types
• 21:58
  Speaker 1: of tasks that it's approved for, it's completely driverlests. And
• 22:01
  Speaker 1: then that surface area increases over time with software updates,
• 22:03
  Speaker 1: and for everything else, it can still be manually operated.
• 22:06
  Speaker 2: So is the core product? Are they paying you by
• 22:08
  Speaker 2: the hour for digging?
• 22:10
  Speaker 1: So we're figuring it out, but it'll be something that
• 22:12
  Speaker 1: is either by the hour, by the project by subscriptions.
• 22:15
  Speaker 1: So it's effectively the way that today projects are forecasted
• 22:19
  Speaker 1: and build by shifts, you know, for labor it's a
• 22:24
  Speaker 1: parallel of that, but it becomes a huge win because
• 22:27
  Speaker 1: you have complete flexibility. You can work ten hours a day,
• 22:30
  Speaker 1: you can work twenty four hours a day. So there's
• 22:32
  Speaker 1: all sorts of benefits that will operationally give you a
• 22:34
  Speaker 1: lot of leeway on how you use it, what might
• 22:37
  Speaker 1: go wrong, what might go wrong. I'll tell you the
• 22:40
  Speaker 1: things that are like really challenging. They like we worry
• 22:43
  Speaker 1: about it, we think about the nuance and diversity of
• 22:47
  Speaker 1: things are high. As much as you want to just
• 22:51
  Speaker 1: boil it down to a very simple dig and load
• 22:54
  Speaker 1: sort of operation, there's always little corner cases. There's things
• 22:58
  Speaker 1: you find in the ground, there's weird ways that trucks
• 23:02
  Speaker 1: can interact with you, there's things people will do. There's
• 23:04
  Speaker 1: varieties of machines, but there's like fifty kinds of excavator
• 23:07
  Speaker 1: models and sizes and so forth. So I think the
• 23:11
  Speaker 1: long tail is still challenging.
• 23:13
  Speaker 2: Presumably there's you don't have to figure out every edge case,
• 23:16
  Speaker 2: but you probably have to figure out a lot for
• 23:17
  Speaker 2: your thing to be to work in a functional sense, right.
• 23:22
  Speaker 1: Yeah, And so those in the meantime, you're not just
• 23:25
  Speaker 1: digging and loading and got a reposition you got to
• 23:27
  Speaker 1: organize the earth, You got to think about the sequencing,
• 23:29
  Speaker 1: you got to deal with you know, daytime, night time rain,
• 23:32
  Speaker 1: and so you have like these types of like really
• 23:35
  Speaker 1: challenging diversities you have to think about and deal with.
• 23:38
  Speaker 1: So I think all in all, it's still a complicated
• 23:42
  Speaker 1: product area where there's a huge amount of diversity of
• 23:45
  Speaker 1: the things that need to be done. But it's one
• 23:47
  Speaker 1: of those where, like I personally think there's a handful
• 23:50
  Speaker 1: of these holy grails of autonomy and physical industries that
• 23:53
  Speaker 1: are like genuinely transformational opportunities for both impact positive impact
• 23:59
  Speaker 1: to the country and the world, and also just kind
• 24:01
  Speaker 1: of scale of industries that are like double digit percentages
• 24:04
  Speaker 1: of GDP. Transportation is, without a doubt one of them,
• 24:07
  Speaker 1: Construction is one of them, Culture is not far behind.
• 24:10
  Speaker 1: You know, mining is a very very significant as well.
• 24:13
  Speaker 1: Manufacturing is one of them. And so I think that
• 24:16
  Speaker 1: we're going to see a wave over this next ten
• 24:18
  Speaker 1: years in autonomy, but it's going to be tackling this
• 24:22
  Speaker 1: like seventy five percent of the world's GDP that's physical
• 24:24
  Speaker 1: and not digital, and there's a lot of work, like
• 24:27
  Speaker 1: a lot of positive impact that can happen across these spaces.
• 24:29
  Speaker 2: I mean, give me a little more on that one
• 24:31
  Speaker 2: pick a time in the future, five years, ten years,
• 24:34
  Speaker 2: not more than ten.
• 24:35
  Speaker 1: So my personal belief is that this idea, like there's
• 24:38
  Speaker 1: a lot of companies getting flooded for this, but the
• 24:39
  Speaker 1: idea that like this giant brain for all of robotics,
• 24:43
  Speaker 1: the foundation model for robotics, like, I personally do not
• 24:47
  Speaker 1: believe that that's viable in the next ten years because
• 24:50
  Speaker 1: you have such complexity in really understanding these verticals on
• 24:56
  Speaker 1: the inputs, the hardware, the products, the neat use case,
• 24:59
  Speaker 1: the customers, everything, that every single one of those have
• 25:02
  Speaker 1: such complexity to get data that the idea to get
• 25:05
  Speaker 1: enough data that bulldos is through a generalization problem. That's
• 25:09
  Speaker 1: a long ways off. But I do think that word
• 25:12
  Speaker 1: a perfect time for these vertical solutions where if you
• 25:14
  Speaker 1: have a focused solution where you're trying to do construction,
• 25:17
  Speaker 1: you're trying to do fulfillment in a warehouse, you're trying
• 25:20
  Speaker 1: to do a very focused manufacturing solution, I think that's
• 25:23
  Speaker 1: actually there's a giant stuff function and how powerful and
• 25:26
  Speaker 1: male technologies are. But if you're trying to do a
• 25:29
  Speaker 1: humanoid to do everything in a home for a consumer
• 25:31
  Speaker 1: market too far away.
• 25:33
  Speaker 2: Right now you're saying you basically you got to solve
• 25:36
  Speaker 2: one problem at a time. Yeah, I want to go
• 25:39
  Speaker 2: back to open road autonomy almost done. When do you
• 25:42
  Speaker 2: think that most rides most people take in cars or
• 25:47
  Speaker 2: trucks will be in driverless cars autonomous vehicles?
• 25:52
  Speaker 1: Great question. There's a few pre requisites for most rides.
• 25:55
  Speaker 1: So first, all the goofs is have to disappear and
• 25:58
  Speaker 1: the thing just works everywhere that's on a trajectory of
• 26:01
  Speaker 1: getting there because it's getting more and more efficient as
• 26:03
  Speaker 1: the scales across the country and then the world. It
• 26:05
  Speaker 1: has to go from ride healing to personal car ownership
• 26:08
  Speaker 1: that is both cost down and versatility and everything that. Like,
• 26:13
  Speaker 1: but that's gonna happen.
• 26:14
  Speaker 2: In that universe. When do they take the steering wheel out?
• 26:18
  Speaker 2: I just took a way mover the first time I
• 26:20
  Speaker 2: was in San Francisco this summer.
• 26:21
  Speaker 1: It's awesome, right, it was awesome.
• 26:22
  Speaker 2: Yeah, And like the weirdest thing about it to me
• 26:26
  Speaker 2: was the steering wheel. Yeah, right, Like if it was
• 26:28
  Speaker 2: just a box that I got in that looked like
• 26:30
  Speaker 2: a train car, someone I would have been like, oh sure,
• 26:31
  Speaker 2: it's like the air train or whatever, but that there's
• 26:34
  Speaker 2: a steering wheel and it's like a ghost is turning
• 26:36
  Speaker 2: the steeringhell, Like what's the steering wheel doing there?
• 26:39
  Speaker 1: So the steering wheels will disappear pretty quickly because ride
• 26:42
  Speaker 1: sharing autonomous cars, yeah, have no use for it, and
• 26:44
  Speaker 1: you need to point you know, there's like special cases.
• 26:46
  Speaker 1: We need to recover them. But it's a wasted space, right,
• 26:49
  Speaker 1: It's a wasted spot. So that'll happen soon. But the
• 26:53
  Speaker 1: idea of really deep penetration, I think it's personal cars.
• 26:57
  Speaker 1: It's beyond luxury, which is another generation to like actually
• 27:00
  Speaker 1: make it be something that's affordable. Then you got to
• 27:03
  Speaker 1: go through a buying cycle, which is like five to
• 27:05
  Speaker 1: seven years, and so I think it starts to get
• 27:08
  Speaker 1: serious penetrats in the back half of the thirties and
• 27:12
  Speaker 1: it'll be the forties where like, okay, fifty percent of
• 27:17
  Speaker 1: driving is kind of like autonomous. I think it's still
• 27:19
  Speaker 1: twenty years off in my mind. For example, you bought
• 27:21
  Speaker 1: a car today, it's going to be in circulation for
• 27:23
  Speaker 1: the next twelve to fifteen years. It just takes a while.
• 27:31
  Speaker 2: We'll be back in a minute with the lightning round. Okay,
• 27:46
  Speaker 2: let's finish with the lightning round. If you could operate
• 27:49
  Speaker 2: any machine, what would it be?
• 27:53
  Speaker 1: Oh gosh, this is a fun one. If I could
• 27:56
  Speaker 1: operate any machine at all in the world, not even
• 27:58
  Speaker 1: in contry anything, Oh gosh, Okay, you know what, I
• 28:02
  Speaker 1: would love to either operate one of the like Boring
• 28:05
  Speaker 1: Company gigantic drill machines. Was like so astronomical or there's
• 28:09
  Speaker 1: like my trucks that are so astronomically big that they
• 28:12
  Speaker 1: just dwarfed any machine in the world, and it costs
• 28:14
  Speaker 1: like five million dollars and just a skill that would
• 28:16
  Speaker 1: be really really fun to try at some point.
• 28:19
  Speaker 2: Just to be that high, just have that much momentum, right,
• 28:22
  Speaker 2: that much mass at your disposal.
• 28:25
  Speaker 1: Yeah, like literally the tire is like three stories tall.
• 28:28
  Speaker 1: It's like it's wild, Like it's absurd.
• 28:30
  Speaker 2: What's one thing you remember about immigrating from the Soviet
• 28:33
  Speaker 2: Union to the US when you were six?
• 28:36
  Speaker 1: I was six. Yeah, I was born in Moscow. We immigrated.
• 28:39
  Speaker 1: I was super young. I remember we ended up having
• 28:43
  Speaker 1: a pet stop in Europe where there's a standard path
• 28:46
  Speaker 1: of going to be out of Venice in Rome, why
• 28:47
  Speaker 1: your paperwork gets processed, and then we went to New York.
• 28:50
  Speaker 1: I remember running around the rooftops of Venice with a
• 28:53
  Speaker 1: friend of mine, causing a bunch of you know, trouble
• 28:57
  Speaker 1: and running away and disappearing for long periods of time
• 29:00
  Speaker 1: and having a boss for for whatever reason, running around
• 29:02
  Speaker 1: the rooftops of Venice was in my was ingrained in
• 29:04
  Speaker 1: my memory, which was a kind of very positive memory
• 29:07
  Speaker 1: while my parents were going through a whole bunch of strusts.
• 29:09
  Speaker 2: It sounds very free. Not to project East versus West
• 29:12
  Speaker 2: language onto it, but it sounds very free.
• 29:15
  Speaker 1: It was very freeing. It's It's funny. I just hit
• 29:18
  Speaker 1: the age my dad was when they immigrated from the
• 29:21
  Speaker 1: Soviet Union with two kids. My sister was one month old,
• 29:25
  Speaker 1: zero money, almost no English, having a fresh start.
• 29:29
  Speaker 2: So courageous, right, doesn't it seem so brave?
• 29:33
  Speaker 1: Yeah? It does. And they were trying to leave for
• 29:35
  Speaker 1: like ten years and couldn't leave. Yeah, it's kind of fascinating.
• 29:38
  Speaker 1: So we're very fortunate to not have to go through
• 29:40
  Speaker 1: something like that.
• 29:41
  Speaker 2: Is that when you think about it that way, does
• 29:43
  Speaker 2: that like put pressure on you? You think, oh my god,
• 29:45
  Speaker 2: my parents did all this better? I better deliver.
• 29:50
  Speaker 1: It's funny, Uh, a little bit. I mean you kind
• 29:52
  Speaker 1: of have this I don't know, like adventurous spirit. I
• 29:56
  Speaker 1: guess maybe it gets ingrained. I think of it now
• 29:59
  Speaker 1: for my kids, where I'm like, okay, Like now they're
• 30:01
  Speaker 1: in a really nice and comfortable environment growing up. How
• 30:05
  Speaker 1: do you I convey that edge in a little bit
• 30:07
  Speaker 1: of that spirit, Like, how.
• 30:08
  Speaker 2: Do you keep them from going so.
• 30:09
  Speaker 1: Yeah, it's something you want them to go through anything difficult,
• 30:13
  Speaker 1: because it's not that I was, you know, for me,
• 30:14
  Speaker 1: it was actually just an adventure my guards. It was difficult,
• 30:17
  Speaker 1: But part of it is just, yeah, like conveying that
• 30:21
  Speaker 1: spirit of being able to like be comfortable trying to
• 30:24
  Speaker 1: tackle something new and being thrown in a completely different environment.
• 30:27
  Speaker 1: It's hard to force ound or simulate that when you're
• 30:29
  Speaker 1: just growing up in the Semptusco area. Right.
• 30:32
  Speaker 2: I appreciate your time. Thanks for talking to me.
• 30:34
  Speaker 1: It's a pleasure. This is a lot of fun. Thanks
• 30:35
  Speaker 1: for having me.
• 30:43
  Speaker 2: Boris Soffman is the co founder and CEO of Bedrock Robotics.
• 30:48
  Speaker 2: Just a quick note that this is our last episode
• 30:51
  Speaker 2: before a break of a couple of weeks and then
• 30:53
  Speaker 2: we'll be back with more episodes. Please email us at
• 30:56
  Speaker 2: problem at Pushkin dot Fm. We are always looking for
• 31:00
  Speaker 2: new guests for the show. Today's show was produced by
• 31:04
  Speaker 2: Trinamanino and Gabriel Hunter Chang, who was edited by Alexander
• 31:08
  Speaker 2: Garretton and engineered by Sarah Brugueri. I'm Jacob Goldstein and
• 31:12
  Speaker 2: we'll be back next week with another episode of What's
• 31:14
  Speaker 2: Your Problem.