Latest / What's Your Problem? / Building a Mass Market Robot

Transcript

Building a Mass Market Robot · What's Your Problem?

0:15
Speaker 1: Pushkin As a general matter, I'm a fan of technological progress,
0:23
Speaker 1: but I'll admit that humanoid robots kind of creep me
0:27
Speaker 1: out for you know, obvious, normy, uncanny Valley type reasons.
0:32
Speaker 1: And yet there is an exchange that you'll hear near
0:36
Speaker 1: the end of today's show. That's the most compelling argument
0:40
Speaker 1: I've ever heard for humanoid robots. And it's not just
0:43
Speaker 1: an intellectual argument, it's an emotional argument. If that's a phrase,
0:49
Speaker 1: it's really a very human argument for humanoid robots. I'm
0:59
Speaker 1: Jacob Goldstein, and this is what's your problem to show
1:02
Speaker 1: where I talk to people who are trying to make
1:04
Speaker 1: technological progress. My guest today is Jeff Cardinas. He's the
1:08
Speaker 1: k founder and CEO of Eptronic. Jeff's problem is this,
1:13
Speaker 1: can you make a safe, reliable humanoid robot for less
1:17
Speaker 1: than fifty thousand dollars. We started our conversation talking about
1:21
Speaker 1: the DARPA Robotics Challenge. DARPA, of course, is the government
1:26
Speaker 1: agency that helped to create the Internet and that gave
1:28
Speaker 1: a big push to early self driving cars, among other things.
1:32
Speaker 1: And Jeff says the agency's robotics challenge, which happened a
1:36
Speaker 1: decade ago. Happened in twenty fifteen played a key role
1:39
Speaker 1: in launching a bunch of the companies that are now
1:42
Speaker 1: working on humanoid robots.
1:46
Speaker 2: The Darker Robotics Challenge was a challenge that was created
1:50
Speaker 2: in the wake of the Fukushima disaster. Lukashima happened in
1:54
Speaker 2: the you know, there was a meltdown in the nuclear
1:56
Speaker 2: reactors and it was unsafe for people to go in,
2:00
Speaker 2: and in essentially what happened was they needed a robot
2:04
Speaker 2: to go in to sort of prevent a broader catastrophe.
2:08
Speaker 2: And as they went out to the robotics community, the
2:11
Speaker 2: idea was calling all roboticists, we need a robot to
2:15
Speaker 2: go in and to help out here. And there was
2:19
Speaker 2: no robots that could do all the different tasks that
2:21
Speaker 2: were required to actually get to the melted down reactors.
2:24
Speaker 2: So you had to go down steps, you had to
2:25
Speaker 2: open doors, had to do a whole range of things.
2:29
Speaker 2: And in the wake of that, basically what DARPA said was, certainly,
2:34
Speaker 2: there's got to be the technology to enable us to
2:36
Speaker 2: have much more versatile robots for natural disaster relief so
2:41
Speaker 2: this never happens again. So out of that, Darker created
2:44
Speaker 2: something called the Darker Robotics Challenge, and so there's a
2:48
Speaker 2: variety of teams around the world that were put together
2:51
Speaker 2: to build these general purpose robots. And the team that
2:57
Speaker 2: we came out of was the Nasagen and the Space
3:00
Speaker 2: Center team working on a robot called Valkyrie.
3:03
Speaker 1: So I want to go back to this moment when
3:05
Speaker 1: the DARPA Challenge and there's this big final test, and like,
3:10
Speaker 1: what we have from it is not some incredible breakthrough,
3:14
Speaker 1: but like a blooper reel of robots, what are they doing?
3:18
Speaker 1: Falling downstairs or driving cars into walls or.
3:21
Speaker 2: Something pretty much. I mean, the blooper reels make it
3:26
Speaker 2: seem worse than it was, but but yeah, we had
3:30
Speaker 2: we basically the realization was the technology's not there. It's
3:33
Speaker 2: going to take time to continue to mature until it
3:37
Speaker 2: can get to the point where it's actually commercially viable.
3:40
Speaker 1: And so it's interesting, it's super interesting right that this
3:43
Speaker 1: moment is not the like beginning of some humanoid robot winter,
3:48
Speaker 1: but rather the beginning of this humanoid robot industry. Like
3:52
Speaker 1: how does that work? Like how do people whatever start
3:55
Speaker 1: companiesn't get money out of this seemingly disappointing moment.
4:00
Speaker 2: Well, I think it actually was in winter, you know,
4:03
Speaker 2: when we started in you know, the company was incorporated
4:05
Speaker 2: in twenty fifteen and we started in twenty sixteen, and
4:09
Speaker 2: for the most part, nobody wanted to talk about humanoids
4:12
Speaker 2: and nobody was really paying attention to it. There was
4:15
Speaker 2: a handful of folks that I sort of think of
4:18
Speaker 2: as the true believers that were really working on this
4:21
Speaker 2: problem and thought, you know, we don't care how long
4:24
Speaker 2: this is going to take. We're just going to keep
4:25
Speaker 2: working on this no matter what. But for the most part,
4:30
Speaker 2: the entire robotics industry was very anti humanoids coming out
4:33
Speaker 2: of the Darker Robotics Challenge, and in fact, there was
4:36
Speaker 2: many people that were saying they'll never be viable, like
4:39
Speaker 2: why would you ever use a humanoid robot. They're too complicated,
4:42
Speaker 2: they're way too expensive, and you'll always use a simpler robot.
4:47
Speaker 2: So actually, most of the people that we met when
4:49
Speaker 2: we decided to start Uptronic were doubters, and we're saying,
4:53
Speaker 2: humanoids will never make sense. We'll use these special purpose robots.
4:58
Speaker 2: You know, maybe in fifty years humanoids will make sense,
5:00
Speaker 2: but not for a long time.
5:02
Speaker 1: I mean special purpose robots is a pretty compelling case, right,
5:07
Speaker 1: like whatever we all have, you know, dishwashing robots and
5:12
Speaker 1: clothes cleaning robots in our houses, and like, you know,
5:15
Speaker 1: wheels seem way easier than legs for lots of things.
5:18
Speaker 1: And obviously there have been robot arms for what I
5:22
Speaker 1: don't know, seventy years now or something like. So robots
5:25
Speaker 1: in a way are all around us. Why would you
5:27
Speaker 1: build a machine that looks like a dude when that's
5:29
Speaker 1: like wildly hard.
5:30
Speaker 2: Right, Yeah, I mean I was naive coming out of
5:34
Speaker 2: graduate school, and to me it seemed obvious. And the
5:38
Speaker 2: way I used to think about it was you could
5:39
Speaker 2: either have thousands of robots that do one thing, or
5:42
Speaker 2: you could build one robot that could do thousands of
5:44
Speaker 2: different things. And when I would talk about this with Nick,
5:48
Speaker 2: my co founder, Nick would say, look, you can either,
5:51
Speaker 2: you know, invest all this engineering in each of these
5:55
Speaker 2: sort of narrow solutions, or yes, a humanoid robot, a
5:58
Speaker 2: viable humanoid robot could take you years, it could take
6:02
Speaker 2: you a decade, But once you invest all this time
6:04
Speaker 2: in this single platform, then you can reap the benefits
6:08
Speaker 2: of that across You can spread the research of that
6:13
Speaker 2: across many different applications.
6:15
Speaker 1: I mean, is there not a middle case where there's
6:18
Speaker 1: like some core kind of functionality that you develop that
6:22
Speaker 1: works across many different types of robots. Is that a
6:25
Speaker 1: less straw many version of the non humanoid robot kind
6:28
Speaker 1: of argument?
6:29
Speaker 2: I think there could be. But I came into robotics
6:32
Speaker 2: and basically just saw a lot of challenges with the
6:35
Speaker 2: business models. So you build this special purpose robot, you
6:39
Speaker 2: custom program the robot. In the industrial space, you can
6:42
Speaker 2: spend six times the price of the robot on just
6:46
Speaker 2: systems integration, and then the robot just does one thing.
6:50
Speaker 2: So this idea that you could have a much more
6:52
Speaker 2: versatile robot, to me, seemed obvious, like, if robotics is
6:55
Speaker 2: going to scale, we have to have much more versatile
6:57
Speaker 2: robots that we've had in the past. So if you
7:00
Speaker 2: sort of think of that as the premise is we
7:02
Speaker 2: need more versatile robots, then the question is, well, how
7:05
Speaker 2: do you get there and what does versatility mean? And
7:09
Speaker 2: that's where it led me to the humanoid making a
7:12
Speaker 2: lot of sense, because if you have to modify the
7:16
Speaker 2: environment for every new task that the robot can do,
7:19
Speaker 2: you run into the same problem that we had in
7:21
Speaker 2: sort of classical robotics. But if the robot can retrofit
7:25
Speaker 2: into the environment such that you don't have to change
7:27
Speaker 2: or modify the environment for every new task that the
7:30
Speaker 2: robot can do, then it seemed to me that this
7:33
Speaker 2: would maybe be the key unlock for robotics to actually
7:37
Speaker 2: scale to the masses.
7:38
Speaker 1: The demand would be infinite. If you had a thing
7:40
Speaker 1: that was the size and shape of a person with
7:43
Speaker 1: arms and legs like, scale would be off the charts,
7:46
Speaker 1: And presumably that's what drives costs down, and that's like
7:49
Speaker 1: the good flywheel.
7:50
Speaker 2: Right, yeah, exactly.
7:52
Speaker 1: Okay, so you had this big idea about humanoid robots
7:55
Speaker 1: and you started a company. But at the moment you
7:57
Speaker 1: started the humanoid robot company, the prevailing sentiment was like,
8:01
Speaker 1: deeply skeptical, what happened?
8:03
Speaker 2: What did you do? Well? A handful of us kept
8:06
Speaker 2: working on it, so I didn't know any better. You know,
8:09
Speaker 2: sometimes it's better that you don't know any better. I
8:11
Speaker 2: thought humanoids were really cool, and I thought that it
8:16
Speaker 2: just seemed it made sense to me that, you know,
8:18
Speaker 2: how are we going to get to millions of robots
8:21
Speaker 2: that are working and you know, with and around humans
8:24
Speaker 2: and all these environments, and to me, this seemed like
8:26
Speaker 2: the only way that that was going to happen. And
8:28
Speaker 2: the way I looked at it was, even if we failed,
8:32
Speaker 2: this was a worthy pursuit and I would be proud
8:35
Speaker 2: that I tried to do it. And so the way
8:38
Speaker 2: that we did it was we bootstrapped the company. There
8:40
Speaker 2: was no investors that were willing to invest in humanoid
8:43
Speaker 2: robots at the time that we got started, especially for hardware,
8:48
Speaker 2: which we can talk about that as we move forward.
8:52
Speaker 2: And so we bootstrapped the company and we basically got
8:55
Speaker 2: paid to build robots for a lot of different folks.
8:58
Speaker 2: And for the first five years of the company, we
9:01
Speaker 2: just built the company on revenue. We would get project
9:04
Speaker 2: after project and somehow never died.
9:08
Speaker 1: Like what kind of were you taking at that time?
9:10
Speaker 1: What's an one example, Well, we.
9:12
Speaker 2: Had our first contract was with NASA, So we had
9:14
Speaker 2: a contract with NASA to build Valkyrie two, to take
9:17
Speaker 2: the lessons learned from the DARPA Challenge and build the
9:19
Speaker 2: next iteration of Valkyrie. We were really kind of pioneering
9:23
Speaker 2: new ways of building these systems. So US Special Forces
9:27
Speaker 2: ended up coming to us about a year in and said, hey,
9:30
Speaker 2: we want to do Ironman suits, and our view was
9:33
Speaker 2: this was kind of a humanoid robot that you wear.
9:36
Speaker 2: We worked in automotive. We helped build humanoid robots for
9:40
Speaker 2: a couple you know, major companies that are still working
9:44
Speaker 2: on these things today, and we would help sort of
9:47
Speaker 2: pioneer new ways of building their platforms. So we've done
9:50
Speaker 2: fifteen unique robots since we got started, and we're now
9:54
Speaker 2: in a ninth iteration of humanoid and I have only
9:56
Speaker 2: raised money in the last couple of years.
9:59
Speaker 1: So where'd the idea to build a robot for fifty
10:05
Speaker 1: thousand dollars come from?
10:08
Speaker 2: The idea of where to build a robot for fifty
10:10
Speaker 2: thousand dollars was what will it take for these robots
10:13
Speaker 2: to be economic and reach mass market? So, you know,
10:17
Speaker 2: when we got started, sort of my view was, you know,
10:21
Speaker 2: what will a truly viable commercial humanoid look like and
10:26
Speaker 2: what would the bomb costs need to be for this
10:28
Speaker 2: to make sense? And if you sort of just do that,
10:30
Speaker 2: bottoms up about fifty thousand dollars for the robot, you're
10:34
Speaker 2: in the money for mass market. You can still do
10:37
Speaker 2: some tasks in a very economic way at even one
10:40
Speaker 2: hundred thousand or one hundred and fifty thousand, but fifty
10:43
Speaker 2: thousand was the goal. This has now been blown by
10:46
Speaker 2: by some of the new entrepreneurs that are coming out
10:48
Speaker 2: that are talking about sub twenty thousand dollars. But it
10:52
Speaker 2: never made sense to me that robots were as expensive
10:56
Speaker 2: as they were at the time. If you look at
10:58
Speaker 2: a humanoid compared to a car, there's about four percent
11:02
Speaker 2: the raw material by weight. So one of our robots
11:06
Speaker 2: there's about three hundred dollars of raw aluminum, which is
11:08
Speaker 2: the the base metal of the system. And so it
11:14
Speaker 2: never made sense to me that these robots would need
11:17
Speaker 2: to be any more than fifty thousand dollars. As you
11:19
Speaker 2: could reach scale and as you can start to think
11:21
Speaker 2: about new ways of building them in similar ways that
11:24
Speaker 2: we build other machines.
11:26
Speaker 1: So you decide you want to build a fifty thousand
11:29
Speaker 1: dollars robot, Like, what do you actually do to do that? Like,
11:35
Speaker 1: how do you go from having an idea of building
11:37
Speaker 1: a fifty thousand dollars robot to having a fifty thousand
11:39
Speaker 1: dollars robot.
11:40
Speaker 2: Well, you iterate until you solve any problem. So what
11:44
Speaker 2: we would do is basically we would get a project
11:46
Speaker 2: or a contract to build a robot, and we would
11:52
Speaker 2: put a lot of different ideas into those designs. In
11:56
Speaker 2: early days, it was all about performance. How can you
11:58
Speaker 2: get the robot to just do these tasks, to stand
12:01
Speaker 2: and have a battery life that's long enough. And then
12:04
Speaker 2: as we kept evolving, we started to really focus on
12:09
Speaker 2: cost in addition and scalability and assemblability and robustness. And
12:14
Speaker 2: the key building block to drive cost and performance is
12:17
Speaker 2: the actuator. So I mentioned we've done nine iterations of humanoids,
12:21
Speaker 2: but we've done sixty iterations of electric actuators.
12:24
Speaker 1: Actuators are basically the thing that makes the robot.
12:27
Speaker 2: Move, right, Yeah, the muscles of a robot.
12:31
Speaker 1: So where are you now? Tell me about Apollo.
12:34
Speaker 2: Yeah, we're now at an exciting point. We have about
12:37
Speaker 2: one hundred and seventy employees at Electronic. We are piloting
12:41
Speaker 2: these robots right now. So I think the entire industry
12:45
Speaker 2: is still in the pilot stage overall. There's some commercial
12:48
Speaker 2: orders that are happening, but still early days for humanoids.
12:52
Speaker 2: We're working with a handful of really great partners, folks
12:55
Speaker 2: like Mercedes and GXO, and we're getting the robots out
12:59
Speaker 2: into the real world and in pretty big ways, and
13:03
Speaker 2: so we'll have more announcements over this year. We have
13:06
Speaker 2: a big partnership with Google Deep Mind, which is something
13:08
Speaker 2: that I always dreamed of coming out of graduate school.
13:11
Speaker 2: We had a lot of respect for the folks that
13:14
Speaker 2: at Google, and they have a whole history and legacy
13:17
Speaker 2: in the humanoid space as well. And basically right now
13:21
Speaker 2: we're getting these robots out into the world and gearing
13:25
Speaker 2: up for you know, real commercialization, which we expect to
13:29
Speaker 2: happen in twenty twenty six.
13:32
Speaker 1: What's the robot look like?
13:33
Speaker 2: The robot kind of looks like a superhero. Maybe this
13:36
Speaker 2: has been kind of the idea that we've had from
13:39
Speaker 2: the beginning. It's got two eyes and a face. It's
13:43
Speaker 2: five foot eight, weighs one hundred and sixty pounds, has
13:46
Speaker 2: four hour swappable batteries. Yeah, it's got a screen on
13:50
Speaker 2: its chest and a face and that's about it.
13:53
Speaker 1: Two arms, two legs. What's to have in the way
13:57
Speaker 1: of hands.
14:00
Speaker 2: So it has hands, has five fingered hands. You know,
14:03
Speaker 2: there's these debates that I think of his false debates
14:05
Speaker 2: in the humanoid space. So a lot of people when
14:08
Speaker 2: they sort of knock humanoids and the viability of humanoids.
14:11
Speaker 2: It usually has to do with do they need legs
14:13
Speaker 2: and do they need hands? And the answer to that
14:16
Speaker 2: question for me is no, they don't. It's a robot,
14:18
Speaker 2: and robots are modular. So we can put Apollo on
14:21
Speaker 2: any mobility base. We can put it on wheels, we
14:24
Speaker 2: could put it on tracks, we could stationary amount the
14:27
Speaker 2: upper torso of Apollo. And the same thing's true for
14:31
Speaker 2: the hands of the grippers. We can use parallel grippers,
14:34
Speaker 2: or we can use five fingered hands.
14:37
Speaker 1: Hands are like a whole thing, right, Like hands are?
14:40
Speaker 1: Is it partly because they're so hard? Like what's going
14:42
Speaker 1: on with robots and hands?
14:44
Speaker 2: Yeah, it turns out hands are a whole thing. This
14:46
Speaker 2: is another one of those things that you know. It's
14:48
Speaker 2: it's almost better that you don't understand the complexity before
14:51
Speaker 2: you get into it, or else you might not have
14:54
Speaker 2: done it in the first place. Ninety eight percent of
14:58
Speaker 2: all tasks that humans do are done with our hands.
15:02
Speaker 2: So there are narrow things that humanoids can do without
15:06
Speaker 2: more dexterity, but it's very limited relative to the whole
15:11
Speaker 2: sort of you know, all the different types of tasks
15:13
Speaker 2: that humans do. Most of the things we do involve
15:17
Speaker 2: our hands, and certainly in the industrial space, most of
15:19
Speaker 2: the work is done with hands, So solving the end
15:23
Speaker 2: defector or the hand problem is a big deal. There's
15:26
Speaker 2: a lot of different debates about what you need and
15:28
Speaker 2: how you get something that can actually perform industrial work.
15:32
Speaker 2: You know, we've chosen the five finger hand route and
15:36
Speaker 2: we're working across the space to really make some big
15:40
Speaker 2: advancements there.
15:41
Speaker 1: Overall, it's part of the trade off, Like I could
15:44
Speaker 1: build whatever two what do you call them prongs? Like
15:47
Speaker 1: if you had two fingers, basically like a claw, Like
15:50
Speaker 1: you could do a lot of things with a claw.
15:52
Speaker 1: Presumably it would be way easier, but you couldn't do everything.
15:56
Speaker 1: Is it kind of like what are you optimizing for?
15:58
Speaker 1: And sort of how much payoff now versus how much
16:00
Speaker 1: payoff later?
16:01
Speaker 2: Yeah, I think that's exactly right. It's you know, versatility,
16:06
Speaker 2: you know, compared to robustness and cost. Basically, how much
16:11
Speaker 2: complexity do you want to have on the system. And
16:14
Speaker 2: you know, for these robots to be really viable in
16:16
Speaker 2: the long run, especially in the industrial space, they've got
16:19
Speaker 2: to be able to operate two shifts a day minimum.
16:22
Speaker 2: Really you know, twenty two hours a day, seven days
16:25
Speaker 2: a week. But solving that problem in a hand, so
16:28
Speaker 2: just getting the performance of the hand first, but then
16:30
Speaker 2: the robustness for them to do that type of work
16:34
Speaker 2: is the next piece. And that's a trade off of
16:36
Speaker 2: performance and complexity and cost.
16:39
Speaker 1: Because like it gets delicate, right, presumably the fingers so
16:42
Speaker 1: to speak, would be fragile, right, Yeah, they can bring
16:46
Speaker 1: easy to break.
16:47
Speaker 2: Yeah, yeah, yeah, and you've got to maintain it and
16:50
Speaker 2: you've got to support those systems and fix them out
16:53
Speaker 2: in the field. And so what's the trade off there?
16:56
Speaker 2: And that's a whole trade space that we've been working
16:58
Speaker 2: on over a long time.
17:00
Speaker 1: So we've been talking about hardware. Let's talk about the
17:03
Speaker 1: software side. What's happening with that.
17:08
Speaker 2: A lot's happening on that side. I think we're really
17:10
Speaker 2: in a really you know, we're an exciting point for
17:14
Speaker 2: robotics overall. Think of the AI as really the last
17:17
Speaker 2: piece of the puzzle. So, you know, we've had the
17:20
Speaker 2: ability to build complex robots for a relatively long time.
17:25
Speaker 2: We're just now, you know, really figuring out how to
17:28
Speaker 2: take the lessons from automotive and consumer electronics and and
17:31
Speaker 2: build much more economic systems, and we've had some advancements
17:35
Speaker 2: and things like motors and batteries and compute and sensors
17:39
Speaker 2: that have all sort of built up to this moment.
17:42
Speaker 2: But the final piece of the puzzle was the AI
17:44
Speaker 2: and the intelligence and essentially the way to think about it.
17:48
Speaker 2: And I think Jensen does a great job of explaining this.
17:50
Speaker 1: But the advances Jensen Wong from Nvidio, Yeah, that's right.
17:55
Speaker 1: I feel like Jensen is not quite the Elon level
17:58
Speaker 1: of one name, household name, but sorry, go on, he's
18:01
Speaker 1: getting there.
18:02
Speaker 2: Yeah, he should be.
18:04
Speaker 1: He should be. He should be.
18:07
Speaker 2: Basically, the advancements in geni AI turn out to apply
18:11
Speaker 2: very well to robotics, and particularly to humanoid robotics, So
18:15
Speaker 2: you can basically map human movement and trajectories from humans
18:19
Speaker 2: doing things and build big data sets and use that
18:23
Speaker 2: to train robots to do similar tasks in similar environments.
18:26
Speaker 2: And these transformer architectures that we're using and generative AI
18:31
Speaker 2: actually apply very well to robotics, and so this has
18:34
Speaker 2: been a big sort of breakthrough moment for robotics. And
18:37
Speaker 2: so I think as an industry as a whole, everybody's
18:41
Speaker 2: really excited right now because we're reaching new heights and
18:44
Speaker 2: we're able to do things that we dreamed about doing
18:47
Speaker 2: with robots only you know, even a few years ago,
18:50
Speaker 2: are now possible, and we're seeing a really rapid advancement
18:53
Speaker 2: in performance overall.
18:54
Speaker 1: What's an example of a thing that you could only
18:57
Speaker 1: dream of a few years ago that robots can do now?
19:00
Speaker 2: I think it's more dexterity and versatility. So just the
19:03
Speaker 2: range of things that you can do. So the challenge
19:06
Speaker 2: for robotics was that each news even if you build
19:10
Speaker 2: something like a humanoid robot, and this is true for us,
19:13
Speaker 2: and say you build an application to pick boxes off
19:18
Speaker 2: of a palette and place those boxes onto a conveyor. Well,
19:21
Speaker 2: you hand build that application, and you know, maybe takes
19:24
Speaker 2: you eighteen months to sort of ring that out and
19:27
Speaker 2: get it to a certain amount of robustness. Well, now
19:30
Speaker 2: you want to do the reverse of that and pick
19:33
Speaker 2: off of a conveyor and palletize something that will take
19:36
Speaker 2: you the same amount of time that it took you
19:39
Speaker 2: to build the initial application.
19:41
Speaker 1: You have to write basically have to write a whole
19:43
Speaker 1: other piece of software. You have to start from scratch
19:46
Speaker 1: almost yeah.
19:47
Speaker 2: Yeah, exactly. And so basically what is happening now is
19:51
Speaker 2: that we now have these much more sort of general
19:55
Speaker 2: models that where you can collect a lot of data
19:58
Speaker 2: at the top layer, and so each new task that
20:01
Speaker 2: you want to perform actually takes less and less incremental
20:04
Speaker 2: amount of work. So what it's opening up now is
20:08
Speaker 2: more dexterous applications.
20:13
Speaker 1: Still to come. On the show how Jeff's grandparents inspired
20:16
Speaker 1: his work on robots, So you were talking about using
20:27
Speaker 1: the transformer model. That has been the you know, breakthrough
20:32
Speaker 1: that has driven large language models in training robots essentially.
20:38
Speaker 1: I mean, of course, a key sort of serendipitous thing
20:41
Speaker 1: that happened with language models was there is this crazy
20:46
Speaker 1: large data set of words and pictures which is the Internet,
20:49
Speaker 1: and there's not an analogous data set for the physical world. Right, Yeah,
20:54
Speaker 1: it seems like that is is that the rate limiting step?
20:58
Speaker 1: Is that the big problem in sort of AI for robots?
21:03
Speaker 2: Yeah, I mean there's a lot of work that's still
21:05
Speaker 2: happening at the research level for you know, how can
21:08
Speaker 2: you pull that kind of data from from videos so
21:11
Speaker 2: you can think of big data sets of humans doing
21:14
Speaker 2: things that could be really interesting to train robots in
21:17
Speaker 2: the future. Interest and that that will come into play
21:20
Speaker 2: over time. But yeah, it's the chicken or the egg problem.
21:23
Speaker 2: And data is the one of the key things that
21:27
Speaker 2: we need to enable the next wave of breakthroughs. And
21:31
Speaker 2: and this is kind of the race, is can you
21:33
Speaker 2: get robots out into the to the real world, into
21:36
Speaker 2: the field and collecting data very high you know, in
21:39
Speaker 2: very high volumes. Whoever does that, you know, will we'll
21:43
Speaker 2: have better models. And this is the data flywheel. So
21:46
Speaker 2: this is kind of the race that's on right now
21:48
Speaker 2: where you hear a lot of other humanoid CEOs talking
21:51
Speaker 2: about getting a lot of robots out into the world.
21:54
Speaker 2: Largely those are going to be under Teley operation collecting
21:57
Speaker 2: data and then you know, training and building these models
22:00
Speaker 2: of the future.
22:01
Speaker 1: Ah So so it's like whoever gets there first will win,
22:06
Speaker 1: just because that'll be the accelerant. Like once you have
22:08
Speaker 1: robots in the world and you're collecting data, then you're
22:11
Speaker 1: immediately getting ahead of whoever has fewer robots out in
22:14
Speaker 1: the world because they're collecting less data.
22:16
Speaker 2: Yeah, so.
22:18
Speaker 1: Tell me about Telly operation.
22:21
Speaker 2: Tellyoperation is basically just remotely controlling the robot, so you're
22:25
Speaker 2: taking over the robot. You can see through the robot's
22:28
Speaker 2: eyes with a VR headset, and then you're controlling the
22:32
Speaker 2: robots arms and hands to do a particular task. It's
22:36
Speaker 2: like a video game and you're controlling a robot. It's
22:39
Speaker 2: the simple idea. There's a couple of reasons it's important.
22:44
Speaker 2: The first thing is that it tells you what the
22:46
Speaker 2: robot's physically capable of doing. So if I'm completely controlling
22:49
Speaker 2: the robot and I can't do a task under Telly operation,
22:53
Speaker 2: then that means the robot's not physically capable of doing
22:55
Speaker 2: it to be very difficult for an AI control system
22:59
Speaker 2: to control the robot to do that. So this is
23:01
Speaker 2: how we understand the physical capabilities of the robot as
23:05
Speaker 2: these new models have come along. The simple idea is
23:08
Speaker 2: that if you can tell you operate the robot to
23:11
Speaker 2: do a task, then you should be able to automate
23:13
Speaker 2: that task on the other end. So if you can
23:16
Speaker 2: collect enough data under Telly operation, then you can automate
23:20
Speaker 2: it by running it through these similar architectures that we
23:23
Speaker 2: talked about.
23:25
Speaker 1: And so's the basic idea that like you use remote
23:29
Speaker 1: control to like drive the robot to do a thing
23:32
Speaker 1: whatever A thousand times some number of times, and in
23:35
Speaker 1: doing that, you're training the robot, you're training the software
23:39
Speaker 1: training the I.
23:40
Speaker 2: Yeah, that's exactly right.
23:42
Speaker 1: What's an example of a thing that you've done that way?
23:44
Speaker 1: And like, you know, how many times did you have
23:46
Speaker 1: to remote control it before the robot could do it?
23:48
Speaker 2: So each is picking is a good example. Or you
23:51
Speaker 2: know you're taking objects and you're putting them into a
23:53
Speaker 2: box to do that in a simple context, thousands of
23:59
Speaker 2: demonstrations is what you need, you know, and we think
24:02
Speaker 2: of this as generally ours. So you know, how many
24:07
Speaker 2: hours of data collection have we done, and thousands of
24:11
Speaker 2: iterations can get you to let's say, eighty percent of
24:14
Speaker 2: human rate. If you want to get to ninety five
24:17
Speaker 2: percent or better of human rate, then you need more
24:20
Speaker 2: and more data. But it's in the thousands, it's not millions.
24:24
Speaker 1: Yeah, thousands makes it seem totally tractable.
24:27
Speaker 2: Yeah, that was actually surprised by how well these models
24:31
Speaker 2: work and actually how little data they need to get
24:33
Speaker 2: relatively good performance. And you're seeing a lot of demonstrations
24:38
Speaker 2: of this out there today.
24:40
Speaker 1: And presumably that'll get better and better. Right as the
24:42
Speaker 1: software side of AI gets better and better, it'll learn
24:47
Speaker 1: faster essentially and the other obvious thing, but I'm just
24:49
Speaker 1: going to say it is like, once you have done
24:51
Speaker 1: it once, then it works for every robot. Then you
24:54
Speaker 1: can make a million robots and they all know how
24:56
Speaker 1: to do the thing, right.
24:57
Speaker 2: Yeah, that's exactly right. And one of the interesting things
24:59
Speaker 2: about these models is actually the diversity of data is
25:02
Speaker 2: almost more important than task specific data. So you want
25:06
Speaker 2: to go wide across a range of tasks, and then
25:09
Speaker 2: you're basically building all these skills into the robot, and
25:13
Speaker 2: then it becomes better at doing any one particular task.
25:17
Speaker 1: It really is like learning.
25:19
Speaker 2: It really is a human esque Yeah, that's right.
25:22
Speaker 1: So I know you're in a few pilot projects with
25:25
Speaker 1: Mercedes and what is GXO big logistics company. When do
25:30
Speaker 1: you want to start selling robots for real? Like when
25:32
Speaker 1: do you think that might happen?
25:33
Speaker 2: Twenty twenty six?
25:35
Speaker 1: Okay, yeah, suddenly that's next year. Almost.
25:39
Speaker 2: Now, a year is a long time in you know,
25:42
Speaker 2: these are dog years. It's a long time in this space.
25:48
Speaker 1: And twenty twenty six could be almost two years.
25:50
Speaker 2: Yeah.
25:51
Speaker 1: Now, like who are you going to sell robots to and
25:54
Speaker 1: how much you're going to charge and what are they
25:56
Speaker 1: going to.
25:57
Speaker 2: Do so the initially in manufacturing and logistics, so folks
26:02
Speaker 2: like Mercedes and GXO, these are the initial customers of
26:05
Speaker 2: these systems. We are not announcing pricing yet, but you
26:11
Speaker 2: can think of it as you know, take what you
26:15
Speaker 2: know it costs to do these tasks today and with
26:17
Speaker 2: some discount to what it costs to do these tasks today.
26:21
Speaker 2: We have a rass model that we that we use,
26:24
Speaker 2: so you basically robot as a service. Yeah, robot as
26:28
Speaker 2: a service model where you're you're paying the robot basically,
26:33
Speaker 2: you know, by the hour effectively to do a particular task,
26:37
Speaker 2: and that's at a discount to what it costs to
26:38
Speaker 2: do that task today.
26:40
Speaker 1: How far are you from the fifty thousand dollars robot.
26:43
Speaker 2: We're not there yet, so not very far. So we
26:48
Speaker 2: have the architecture to be able to do this. So
26:51
Speaker 2: getting the cost down on these robots is a two
26:53
Speaker 2: step process. So first step is new architectures. So if
26:58
Speaker 2: you still require this very high precision in the system
27:01
Speaker 2: and you're using bespoke components that are only used for robotics,
27:05
Speaker 2: these robots will still be expensive. The challenge of umanoid
27:09
Speaker 2: robots is they have a lot more motors than traditional robots.
27:13
Speaker 2: So traditional robot has six or seven motors, a humanoid
27:16
Speaker 2: robot has thirty to forty plus.
27:19
Speaker 1: Okay, so that means it's expensive where you've got to
27:22
Speaker 1: figure out how to get cheaper actuator.
27:25
Speaker 2: Yeah, so we're there. So for us, that was a
27:26
Speaker 2: five hundred dollars actuator that we and we have a
27:30
Speaker 2: five hundred dollars actuator now today. And so once you
27:33
Speaker 2: solve that problem, and once you solve the architecture problem,
27:36
Speaker 2: now it's about scale and manufacturing. So a lot of
27:40
Speaker 2: where we spend, a lot of where the cost is
27:43
Speaker 2: driven at low volumes is in just the structures of
27:45
Speaker 2: the robot or we're seeing seeing we're milling at of
27:49
Speaker 2: big blocks of metal parts and very small quantities. But
27:54
Speaker 2: there's other techniques that are are much more cost effective,
28:00
Speaker 2: like casting or stamping, and these will allow these robots
28:04
Speaker 2: to be much cheaper. As I mentioned, look at automotive
28:06
Speaker 2: and look at the scale of automotive, there's four percent
28:09
Speaker 2: the raw material by weight and a humanoid robot as
28:13
Speaker 2: compared to a car. So as you are once you
28:16
Speaker 2: solve the architecture problems such that you can build a
28:19
Speaker 2: lot of these systems, and you are, they're simpler to
28:22
Speaker 2: make than The next piece is just applying mass manufacturing
28:26
Speaker 2: approaches to this to make them a lot cheaper. As
28:29
Speaker 2: you skin, well, I mean.
28:33
Speaker 1: That's a hard leap to make, right, Like what do
28:36
Speaker 1: you do? You get a ton of capital and just
28:39
Speaker 1: build a factory and hope there's demand on the other end, Like,
28:42
Speaker 1: how do you go from this bespoke expensive thing to
28:45
Speaker 1: a mass produced, you know, much less expensive thing.
28:48
Speaker 2: Well, it's a gradient. So, like I said, step one
28:51
Speaker 2: is you have new approaches that allow you to make
28:54
Speaker 2: them cheaper, just inherently on a unit to unit basis.
28:57
Speaker 2: So the early humanoids were like millions of dollars and
29:01
Speaker 2: now we're in the hundreds of thousands of dollars range
29:03
Speaker 2: for building one.
29:04
Speaker 1: So you just got to get one more order of
29:06
Speaker 1: magnitude out of it.
29:08
Speaker 2: Yes, they've already We've drop the price by an order
29:10
Speaker 2: of magnitude. And then now as we build more, and
29:13
Speaker 2: even as you go as you add a zero, as
29:15
Speaker 2: you go from ten to one hundred, the price drops
29:18
Speaker 2: pretty dramatically. So you don't need the volume that you
29:22
Speaker 2: that you might think, like my view is we can
29:24
Speaker 2: get to the sub fifty thousand dollars price point in
29:27
Speaker 2: the thousands of unit quantity, so without without hundreds of
29:31
Speaker 2: thousands or millions of.
29:32
Speaker 1: These So one big buyer, one one big car company
29:36
Speaker 1: or logistics company might place an order of thousands of units.
29:39
Speaker 2: Right, Yeah, And you made a comment that you said,
29:43
Speaker 2: and you hope that there's demand. One of the things
29:45
Speaker 2: that I think is important to note is the demand
29:47
Speaker 2: for these robots is enormous. We have demand for hundreds
29:53
Speaker 2: of thousands of units already today with the customers that
29:56
Speaker 2: we're working with. So the demand is enormous. So we're
30:01
Speaker 2: we're ramping up. You know, we've got to get the
30:03
Speaker 2: robustness and the safety of the system and really bring
30:06
Speaker 2: out the design. And you know, we're these are you know,
30:10
Speaker 2: really credible, thoughtful people that are coming from other industries
30:13
Speaker 2: that are now joining us, that now see that we've
30:16
Speaker 2: crossed this threshold of technical viability and now taking lessons
30:23
Speaker 2: from you know, how you scale and manufacture other things
30:26
Speaker 2: and bringing that into the robotic space and the humanoid
30:29
Speaker 2: space overall.
30:31
Speaker 1: So in a year or at least next year, you
30:33
Speaker 1: want to be selling robots for real? Where do you
30:37
Speaker 1: want to be in five, say years.
30:41
Speaker 2: My view is that where this evolves is it's going
30:43
Speaker 2: to start in logistics and manufacturing, and then as we
30:45
Speaker 2: solve safety as an industry. I'm really interested in healthcare
30:49
Speaker 2: and particularly in elder care over time, so you know,
30:54
Speaker 2: how can these robots be used to improve the way
30:57
Speaker 2: we live and work. That was really the lens that
31:00
Speaker 2: I came into this on, and so I think over
31:04
Speaker 2: the next five years you'll start to see the early
31:06
Speaker 2: stages of the next three years just to see early
31:10
Speaker 2: applications for robots entering the home. There's some folks that
31:14
Speaker 2: are really working hard on this. I think we're about
31:17
Speaker 2: three years out from that being really viable, but I
31:21
Speaker 2: hope people prove me wrong. I hope it's faster than that.
31:24
Speaker 1: Three years seems fast. What's the sort of first first
31:30
Speaker 1: use case, first job you imagine a robot doing for
31:32
Speaker 1: real in somebody's house in three years.
31:35
Speaker 2: Well, everybody wants laundry. If everybody I talked to says,
31:39
Speaker 2: when is this thing going to do my laundry? And
31:41
Speaker 2: I want that as well.
31:42
Speaker 1: There's literally already a machine to do your laundry. All
31:45
Speaker 1: you have to do is put it in one machine
31:47
Speaker 1: and then put it in another The remaining work is trivial.
31:50
Speaker 2: Yeah, I mean, look, I'm not the person to talk
31:53
Speaker 2: about the home. I think we're still a ways out.
31:56
Speaker 2: But there's humanoid companies like one X that are really
31:59
Speaker 2: focused on on the home, and we've got a lot
32:02
Speaker 2: of respect for what they're doing over there, and so
32:05
Speaker 2: I hope they do it. I know that they're working
32:08
Speaker 2: hard on it. And you know where I want a
32:12
Speaker 2: robot for the home as well, So you a lot
32:15
Speaker 2: of the things that are that are happening. And with
32:17
Speaker 2: these models that I talked about, these more generic models,
32:20
Speaker 2: the things that we're learning in the industrial base can
32:22
Speaker 2: apply to the home over time as well.
32:25
Speaker 1: In terms of the AI models, sure, I mean the
32:27
Speaker 1: AA models are basically teaching a robot how to how
32:30
Speaker 1: to deal with the physical world, that's right, how to
32:32
Speaker 1: move around, how to pick things up, how to put
32:34
Speaker 1: things down.
32:35
Speaker 2: Yeah, the task and the I mean the home's tough
32:37
Speaker 2: because like, how much is it even a robot does
32:39
Speaker 2: your your dishes, your laundry, cleans and cooks for you.
32:44
Speaker 2: How much are you willing to pay for that on
32:46
Speaker 2: a yearly basis?
32:47
Speaker 1: I'm imagining the first household tasks. I would have thought
32:50
Speaker 1: you would have said, like people who are quadriplegic, right, Like,
32:54
Speaker 1: there are a lot of people who have various kinds
32:56
Speaker 1: of mobility problems who can't do very basic things around
32:59
Speaker 1: the house, where essentially a robot could do it for them,
33:01
Speaker 1: Like I would think that would be the first use case.
33:03
Speaker 2: I think that's a great use case. And you know,
33:06
Speaker 2: for me, that's that's sort of in the realm of
33:07
Speaker 2: what I say is older care, which is like assistive
33:10
Speaker 2: robots that help you with just base tasks, right, Like,
33:13
Speaker 2: you know my granddad, one granddad went to a home
33:16
Speaker 2: the other granddad had in home care, and the one
33:19
Speaker 2: that had in home care. It was very simple things.
33:21
Speaker 2: Remind you to take your medication and bring the medication over,
33:25
Speaker 2: get you a glass of water, help you to get
33:27
Speaker 2: up and out of bed, you know, to go to
33:29
Speaker 2: the bathroom. Just help you stabilize to go to the bathroom.
33:34
Speaker 2: And so that's not something that we're you know, we're
33:37
Speaker 2: largely paying attention to industrial applications right now, but that
33:42
Speaker 2: is the dream long term. So I'll be excited to
33:44
Speaker 2: see how it shakes out.
33:46
Speaker 1: Yeah, rationally, what you were saying is it makes sense,
33:49
Speaker 1: Like I understand most people would rather stay at home.
33:51
Speaker 1: I understand that in home care is like impossibly expensive
33:55
Speaker 1: for most people. At the same time, like my emotional
33:59
Speaker 1: response to a robot taken care of, say, my parents,
34:02
Speaker 1: is it makes me feel sad. And I recognize that
34:05
Speaker 1: that's perhaps irrational, but that is at some level my
34:08
Speaker 1: emotional response. But you know, the happy thing is like
34:10
Speaker 1: I should take care of them, but like that's hard
34:13
Speaker 1: and it's probably not gonna happen, and it's for its
34:15
Speaker 1: own set of reasons. Right, it's more more than we
34:18
Speaker 1: bargained for in this conversation. I don't know, No, there
34:20
Speaker 1: is something though a humanoid robot starts to get to
34:23
Speaker 1: some weird places in that way.
34:25
Speaker 2: Right, Yeah, I've thought a lot about this and and
34:28
Speaker 2: I think it's a great place to go to happy
34:31
Speaker 2: to talk about it. I think what we want is
34:33
Speaker 2: we want humans taking care of other humans. That's what
34:36
Speaker 2: we want, right, But we don't have that today where
34:39
Speaker 2: you know, look at the way that we age. You know,
34:42
Speaker 2: for me, you know, I was very close to both
34:44
Speaker 2: of my granddads. They both lived into their nineties and
34:46
Speaker 2: outlived my grandmother's oddly enough, and so I sort of
34:52
Speaker 2: watched them age through their lens and that was a
34:55
Speaker 2: big driver of doing this. And you know, these are
34:58
Speaker 2: people that both of them were war heroes, they contributed society,
35:03
Speaker 2: they did all these amazing things, and then at the
35:05
Speaker 2: end of their life, they felt like they were a
35:07
Speaker 2: burden to their family, and they had this feeling like
35:11
Speaker 2: they never had to rely on anyone for anything, and
35:14
Speaker 2: now they're completely reliant on people for everything. And what
35:20
Speaker 2: I saw them do as they age was they lost
35:23
Speaker 2: their dignity. And for me, this idea that you could
35:26
Speaker 2: have a machine that carries your secrets, that is, your
35:31
Speaker 2: machine that does things for you, allows you to keep
35:35
Speaker 2: your dignity such that then you as a human that's aging,
35:40
Speaker 2: you're fresher. You don't have to rely on your son
35:42
Speaker 2: or your daughter or your spouse to get you a
35:46
Speaker 2: glass of water or to do things for you. You
35:49
Speaker 2: still have your own agency and your own autonomy through
35:52
Speaker 2: a machine, and that then helps your family to be
35:55
Speaker 2: much fresher because they don't have the burden of having
35:58
Speaker 2: to do all these things to support you, where then
36:01
Speaker 2: they can be fresher. And so my hope is that
36:04
Speaker 2: this is not designed to replace what humans do for
36:07
Speaker 2: each other. This is designed to augm and enhance that. Remember,
36:11
Speaker 2: just like my granddad as he was getting older, you know,
36:15
Speaker 2: I was working and I was busy, and you know,
36:18
Speaker 2: I would try to go over as many days as
36:19
Speaker 2: I could, but it was always really tough and I
36:21
Speaker 2: didn't want to be alone, and it was this whole
36:24
Speaker 2: battle that I think everybody goes through. And my hope
36:27
Speaker 2: for the future, actually think it's a much more optimistic
36:30
Speaker 2: version is that hopefully my parents have a robot, and
36:33
Speaker 2: that robot is basically programmed for their happiness, and it's
36:38
Speaker 2: designed to remind them when they're down of their favorite
36:41
Speaker 2: song and play it right, remind them that of the
36:45
Speaker 2: movies that they like watching, or whatever, it might be right.
36:48
Speaker 2: And I think that's more optimistic. I think that's exciting,
36:52
Speaker 2: and that makes me hopeful about the future. And you know,
36:55
Speaker 2: I think that's the worst part of the human experience
36:57
Speaker 2: is the way that we age. And I think that
36:59
Speaker 2: these robots and AI embodied AI and AI in general
37:04
Speaker 2: can hopefully allow us to take better care of each other.
37:06
Speaker 2: So I don't think it is creepy. I think it
37:11
Speaker 2: can actually be pretty beautiful if properly done. And that's
37:15
Speaker 2: what you asked at the beginning. How is eptronic different
37:19
Speaker 2: and what are we focused on? You know? For me,
37:22
Speaker 2: I say human centered robotics. But what that means is
37:26
Speaker 2: that we want this to be an optimistic future for humanity.
37:30
Speaker 2: We are tool makers, and we want to build tools
37:33
Speaker 2: for humans to enable us to live in better ways.
37:38
Speaker 2: And I think that if we really focus on that,
37:41
Speaker 2: I think that there can be really amazing ways of
37:44
Speaker 2: doing this, and I think elder care is a great
37:47
Speaker 2: example of how this can be used in that way.
37:53
Speaker 1: We'll be back in a minute with the Lightning Round.
38:05
Speaker 1: And now we are back, as promised with the Lightning Round.
38:09
Speaker 1: What's the biggest difference between Austin today and Austin ten
38:13
Speaker 1: years ago?
38:14
Speaker 2: Oh man, it's changed quite a bit. So. One of
38:18
Speaker 2: the things that made Austin, you know, really a great
38:22
Speaker 2: place to live and work was just how small it
38:25
Speaker 2: was and how accessible everybody was. You know, we used
38:28
Speaker 2: to have these house parties and somebody would bring a violin,
38:31
Speaker 2: and someone would bring a sitar and these world instruments,
38:35
Speaker 2: and you'd have just all sorts of eclectic, creative people
38:38
Speaker 2: doing really interesting things. And so I think, you know,
38:42
Speaker 2: one of the things that I am worried about is
38:47
Speaker 2: that was kind of what made Austin special and the
38:50
Speaker 2: things that make you special. People want to kind of commercialize, right,
38:54
Speaker 2: and they want to they want to take this and
38:56
Speaker 2: they want to sort of scale it, and and it's
38:59
Speaker 2: almost special because it's not commercialized. It's just this raw,
39:02
Speaker 2: organic thing. And and so how does that as more
39:05
Speaker 2: tech and more money comes into Austin, you know, how
39:08
Speaker 2: does that? How does what make Austin, what made Austin great?
39:12
Speaker 2: How does that continue to evolve? So I think though
39:17
Speaker 2: I welcome it, you know, I'd rather be in the
39:19
Speaker 2: place where everybody's coming and everyone wants to build the future.
39:23
Speaker 2: So I'm not one of those that is sort of
39:26
Speaker 2: resisting the changes. I think it's really exciting, and I
39:29
Speaker 2: think more people with new ideas about the future in
39:33
Speaker 2: the world and kind of a free place to do it.
39:36
Speaker 2: There's a you know, there's a real ability here in
39:38
Speaker 2: Texas and in Austin to kind of do what you want,
39:43
Speaker 2: and there's a real culture around you know, the freedom
39:46
Speaker 2: to do the things that you want to do. And
39:48
Speaker 2: so it's a kind of a unique place where all
39:51
Speaker 2: that's coming to the creativity and the you know, the
39:55
Speaker 2: capitalism and all that's all coming together. Is Austin still weird?
40:00
Speaker 2: Still there's pockets of weird. Yeah, certainly there's still weird.
40:04
Speaker 2: Austin is still there. It's all growing up. But yeah, certainly.
40:10
Speaker 1: What's your favorite humanoid robot in fiction, in books and movies? C?
40:16
Speaker 2: Three po for sure.
40:18
Speaker 1: Okay, you were ready with that. You have that one
40:20
Speaker 1: on deck.
40:21
Speaker 2: Yeah, well for three point it makes C three po
40:23
Speaker 2: the human helper. Right.
40:25
Speaker 1: What's one thing that you've learned about the human body
40:29
Speaker 1: from building robots?
40:31
Speaker 2: Oh? Man, At a high level, what I've learned is
40:35
Speaker 2: how amazing the human body really is. I think there's
40:39
Speaker 2: this fear from humans that as we sort of continue
40:43
Speaker 2: down this pursuit of replicating humans and building machines that
40:48
Speaker 2: can do what humans do, that that diminishes what it
40:51
Speaker 2: means to be humans. But what it's actually done for
40:54
Speaker 2: me and most of the people working on this is
40:56
Speaker 2: it just makes you appreciate even more how amazing humans are.
41:01
Speaker 2: So the hand is something that you think a lot about.
41:04
Speaker 2: You just do all these things and you don't appreciate
41:07
Speaker 2: how incredible your hands are are. And you just when
41:12
Speaker 2: you when you start to try to build a hand
41:15
Speaker 2: for a robot, you just appreciate all the limitations how
41:19
Speaker 2: we walk, how we move, the fact.
41:22
Speaker 1: That we can go hard right, all the things we
41:25
Speaker 1: do just like pick up an egg or open a
41:27
Speaker 1: door like that's wildly difficult.
41:29
Speaker 2: It's amazing. Or you eat that egg and it powers
41:33
Speaker 2: you for a day. It powers this neural network brain
41:36
Speaker 2: that's you know, billions of parameters, right. I mean, the
41:39
Speaker 2: human human humans are amazing. And I think as we
41:44
Speaker 2: continue to learn more about what it means to be human,
41:47
Speaker 2: what does it mean to be conscious, all these kind
41:50
Speaker 2: of big ideas I think will only grow to appreciate
41:52
Speaker 2: what we actually have here.
41:55
Speaker 1: Last one, tell me about your grandfather.
41:59
Speaker 2: Oh man, So you know two grandfathers, one Gilberto Cardinas,
42:07
Speaker 2: the other one George Smith. Both of them were great.
42:13
Speaker 2: My granddad, Gilberto Cardinas, came from Puerto Rico. When he
42:17
Speaker 2: was seventeen, he joined the army and fought in the
42:20
Speaker 2: Korean War. He spoke five languages. He was self educated,
42:26
Speaker 2: and you know, he had the American dream and dreamed
42:29
Speaker 2: of what he could do. He was in the army,
42:32
Speaker 2: he was actually a field medic but became a hospital administrator,
42:35
Speaker 2: and he's a big sort of driving force in our family.
42:41
Speaker 2: And I watched him age and watched all the things
42:44
Speaker 2: he went through. He actually fell and lost his vision,
42:47
Speaker 2: so his brain was still intact in his body largely,
42:52
Speaker 2: but he couldn't see, and so he had to have
42:54
Speaker 2: around the clock care when he was in his nineties
42:57
Speaker 2: in the home. And he wasn't wealthy by any stretch,
43:01
Speaker 2: but he'd done okay and had saved his money, and
43:05
Speaker 2: it was seventeen thousand dollars a month for in home care.
43:09
Speaker 2: And it was like this revolving door of people that
43:14
Speaker 2: would rather be doing anything else than sitting in a
43:17
Speaker 2: room with my granddad and taking care of him. And so,
43:20
Speaker 2: you know, for me, I admired my granddad so much,
43:25
Speaker 2: and just seeing sort of that as the end of
43:27
Speaker 2: his life, sitting in a room, you know, counting the
43:32
Speaker 2: days down, I just thought there's got to be a
43:34
Speaker 2: better way than this, And that was a big, a
43:37
Speaker 2: big driver for me doing all this. My other granddad
43:40
Speaker 2: actually ended up getting George Smith went to he had
43:44
Speaker 2: to go to a home. He had colon cancer and
43:47
Speaker 2: his brain still functioned. He had a great sense of
43:50
Speaker 2: humor and he lost control of his bows. And so
43:54
Speaker 2: you can imagine how humiliating that is as you age
43:59
Speaker 2: to be fully aware of what's going on, never rely
44:02
Speaker 2: on anybody, but not be able to control your bows.
44:05
Speaker 2: And so he had to get multiple showers a day,
44:08
Speaker 2: and every time I would go see him, you know,
44:11
Speaker 2: just it was a humiliating experience. And so these are
44:15
Speaker 2: things that are just my story, but everybody has their
44:18
Speaker 2: own story of you know, taking care of an aging
44:21
Speaker 2: parent or grandparent and just what that looks like. And
44:26
Speaker 2: and my hope is that as humans, as tool makers,
44:29
Speaker 2: I think we can do better than that. And I
44:31
Speaker 2: think that these machines we can create will allow us
44:36
Speaker 2: to take better care of each other. And my parents
44:38
Speaker 2: are already naming their robots. They're there there, they can't
44:43
Speaker 2: wait to get them. They're like, you know, they're they're
44:45
Speaker 2: almost seventy now, and they're like, you got to have
44:48
Speaker 2: these ready for whenever, whenever our time is there, so
44:51
Speaker 2: that you know, so that we can age more gracefully
44:54
Speaker 2: than our parents did.
45:02
Speaker 1: Jeff Gardinas is the co founder and CEO of Actronic.
45:06
Speaker 1: Today's show was produced by Gabriel Hunter. It was edited
45:10
Speaker 1: by Lydia Jeane Kott and engineered by Sarah Bruguer. You
45:14
Speaker 1: can email us at problem at Pushkin dot Fm. I'm
45:18
Speaker 1: Jacob Goldstein, and we'll be back next week with another
45:20
Speaker 1: episode of What's Your Problem.