Latest / Automated with Brian Heater / Building Safe, Scalable Robots at GM With Mikell Taylor

Transcript

Building Safe, Scalable Robots at GM With Mikell Taylor · Automated with Brian Heater

• 0:00
  SPEAKER_00: Humans are sort of we're sort of defined in our intelligence by our ability to use tools.
• 0:03
  SPEAKER_00: And if we look at all the ways we interact with the world around us, you know, I don't move a pallet by lifting it, I move a pallet by using a pallet jack.
• 0:10
  SPEAKER_00: Why on earth would I make a humanoid robot that's gonna pull a pallet jack around?
• 0:13
  SPEAKER_00: What a what a suboptimal use of all of that complexity and all that capability.
• 0:17
  SPEAKER_00: And so I think we're we're gonna have to kind of imagine for our industry and for our economy this portfolio of robotics that's gonna do everything that we want them to do and not be not be attached to the idea of a particular form factor being the right one for everything.
• 0:43
  SPEAKER_01: Hello and welcome back to another episode of Automated.
• 0:47
  SPEAKER_01: My name is Brian Heater.
• 0:48
  SPEAKER_01: I am managing editor at the Association for Advancing Automation.
• 0:52
  SPEAKER_01: I am uh excited to share something a little bit different this week.
• 0:57
  SPEAKER_01: You are about to watch the first ever live episode of the show.
• 1:03
  SPEAKER_01: This was recorded in front of an audience in Orlando at A3's Business Forum a few weeks back, and we could not have asked for a better guest than Michael Taylor.
• 1:14
  SPEAKER_01: She is an Amazon robotics vet who is now helping lead GM's nascent robotics effort.
• 1:21
  SPEAKER_01: It's a great conversation, and I'm happy to say that I had the restraint to save the question about her going to the prom with the robot until the very end.
• 1:31
  SPEAKER_01: If you're enjoying the show, please like and subscribe.
• 1:34
  SPEAKER_01: It helps us do more of these things, uh including hopefully some future live episodes in your town.
• 1:39
  SPEAKER_01: In the meantime, come join us in a hotel somewhere within the vicinity of the Magic Kingdom.
• 1:46
  SPEAKER_01: We talk a lot about what's coming up next in automation on the show, but if you really want to see the future in motion, you've got to be there in person.
• 1:56
  SPEAKER_01: Automate 2026 is where the world's leading innovators, builders, and dreamers come together to show what's possible.
• 2:04
  SPEAKER_01: Robots, AI, machine vision, motion control, you name it, all automation under one roof.
• 2:10
  SPEAKER_01: Register for free at automateshow.com to join us in Chicago June 22nd through the 25th.
• 2:17
  SPEAKER_01: We'll see you there.
• 2:19
  SPEAKER_01: Well, first of all, thank you for being here.
• 2:22
  SPEAKER_01: We've been talking about doing this for a really long time.
• 2:24
  SPEAKER_01: But the last time we spoke, I guess on the record, you were still pretty new in your current role.
• 2:29
  SPEAKER_00: It's like my third week.
• 2:30
  SPEAKER_01: Yeah, so there were some sort of abstract things you were able to talk about, but you couldn't get in the specifics.
• 2:35
  SPEAKER_01: So we'll be maybe talking a little bit more on that side than we were able to uh back at Automate.
• 2:41
  SPEAKER_01: But since we're here in Orlando, I figure that uh maybe a good place to start was Surfbird.
• 2:50
  SPEAKER_00: Yes.
• 2:51
  SPEAKER_01: And specifically, I bring up Surfbird.
• 2:55
  SPEAKER_01: One, because you clearly had a lot of fun writing your LinkedIn profile, which is probably not a thing that a lot of people can say.
• 3:01
  SPEAKER_01: But I'm gonna quote your LinkedIn profile back to you.
• 3:05
  SPEAKER_01: Tried to launch a startup, failed to launch a startup, a good and educational time was had by all.
• 3:11
  SPEAKER_01: So I figure we'll start with the good and then maybe we'll move into the educational.
• 3:16
  SPEAKER_00: Yes, absolutely.
• 3:17
  SPEAKER_01: So 2017, very different time from right now, obviously, in terms of where we're at with funding for AI and robotics.
• 3:24
  SPEAKER_01: Um what was that experience like for you?
• 3:27
  SPEAKER_00: SurfBird was a startup that we created to spin out some research from a private research lab in San Francisco called Other Lab.
• 3:34
  SPEAKER_00: Um, and the research had been really centered around various kinds of aerial robotics.
• 3:40
  SPEAKER_00: Um, and one of them that had originally been sponsored by Facebook when they were doing their various kind of internet drone technologies was around tethered drones that could act as sort of quick deploy or cheaper to deploy and maintain cell phone towers.
• 3:54
  SPEAKER_00: And so, you know, Facebook ended up not really continuing to pursue that work, but we thought we maybe had something interesting there, and you know, we we knew that Verizon was starting to experiment with this, so we said, oh, let's see what we can do with this.
• 4:05
  SPEAKER_00: Um and and so we we had a really good small team, we had built a prototype, we'd done some flying, we took advantage of um, there's an army base out in the Central Valley in California that'll host, you know, ignore the FAA regulations kind of testing for startups, which is fantastic and super useful.
• 4:21
  SPEAKER_00: Um so we had a really, really good technical path forward, but uh I we were not successful in raising money and continuing on beyond about a year, unfortunately.
• 4:30
  SPEAKER_00: Um but you know, like I said, it was fun.
• 4:34
  SPEAKER_01: It was a good experience.
• 4:35
  SPEAKER_01: It was a learning experience for you.
• 4:37
  SPEAKER_01: Absolutely.
• 4:37
  SPEAKER_01: And it was something that you haven't done and perhaps never want to do again.
• 4:42
  SPEAKER_00: I like the zero-to-one stage.
• 4:43
  SPEAKER_00: I like the startup stage a lot, and my roles at both, I mean, I was in all startups before I joined Amazon.
• 4:50
  SPEAKER_00: Um and at both Amazon and General Motors, I'm I'm doing that kind of startup in a large company sort of thing.
• 4:56
  SPEAKER_00: Um I think that, you know, having a family being the primary breadwinner at the time when we were unable to raise money, um, I was the only breadwinner and also six months pregnant, and uh suddenly losing health insurance was a little bit challenging.
• 5:11
  SPEAKER_00: Um and so looking at that and kind of where we are now, it's like, well, you know, I I like the vibes of that early stage stuff, but uh, you know, I'm not quite ready to take that level of risk again, maybe in the future.
• 5:23
  SPEAKER_01: Something I learned like really uh early on when I started covering robotics.
• 5:27
  SPEAKER_01: Uh we we had an event in Boston and we had a dinner beforehand, and um I I don't believe you were there at the time, but uh it was a small room, and every single person in that room knew each other.
• 5:40
  SPEAKER_01: Every single person had either like gone to school or had studied under somebody or had worked together, and you know, beyond MIT, which is the obvious one, the common thread was iRobot.
• 5:53
  SPEAKER_01: Like everybody, you know, and I think to a certain extent, maybe there's somewhat of an analog to like Willow Garage on the West Coast, right?
• 6:02
  SPEAKER_01: As far as like the the company that launched, you know, like a thousand startups and got a lot of people to start.
• 6:08
  SPEAKER_01: So, how did you manage to uh intentionally or not avoid working at iRobot?
• 6:15
  SPEAKER_00: I mean, I didn't avoid working there.
• 6:16
  SPEAKER_00: It was one of the job offers I had out of college, and and I um I've always really looked up to Helen Graner and and you know, seeing her, you know, at that stage of my career and getting to know her a bit because it is a small world, it's easy to get to know everybody, meant a lot to me.
• 6:30
  SPEAKER_00: Um, I ended up taking the job that would let me travel the most.
• 6:34
  SPEAKER_00: I worked for an underwater robotics company called Bluefin Robotics, and getting to be on boats a lot and travel was sort of what I wanted.
• 6:41
  SPEAKER_00: Um, but I mean, you know, half the people at Bluefin had had worked or went on to work at iRobot, and we got a bunch of iRobot people.
• 6:49
  SPEAKER_00: So even not actually having worked there, you know, you can still trace the genealogy and the lineage.
• 6:54
  SPEAKER_00: I was talking with a couple people about this at dinner last night.
• 6:56
  SPEAKER_00: I would love to do a family tree of the robotics industry, right?
• 6:59
  SPEAKER_01: Yeah.
• 6:59
  SPEAKER_00: And there would be a few kind of nodes of Willow Garage and iRobot and um, you know, MIT kind of research labs and SRI.
• 7:07
  SPEAKER_01: Yeah, and and see how it all propagated and what companies begat other companies and it's well, I mean, we were we were talking earlier, obviously, you know, you you worked for Rod for a while there, and you worked with Clara at Vio.
• 7:19
  SPEAKER_01: She was uh I think her project at iRobot, she was on the the baby doll, which was uh if people don't know, I that early stage of iRobot when they were kind of exploring what to do, one of the early projects was like a robot doll.
• 7:31
  SPEAKER_01: It didn't quite take like a robot vacuum did.
• 7:34
  SPEAKER_00: Yeah, and yeah, everyone says humanoids.
• 7:36
  SPEAKER_01: That's right.
• 7:36
  SPEAKER_01: They were early in.
• 7:38
  SPEAKER_01: They were early in on the humanoids.
• 7:40
  SPEAKER_01: Um obviously, you know, a lot's been uh happening uh in and around iRobot in in the news lately.
• 7:46
  SPEAKER_01: And I know that you and I, we've talked about this quite a bit, have uh, I don't know if passion is the right word, but um we we both spend a lot of time trying to get people jobs, right?
• 7:56
  SPEAKER_01: One of the things that I did in the newsletter is we put job listings in there.
• 8:00
  SPEAKER_01: That's really important to me.
• 8:02
  SPEAKER_01: A lesson, I think, maybe a takeaway as we're speaking of your startup is maybe there's less security in startups, but nobody's completely safe.
• 8:10
  SPEAKER_00: No, that's absolutely true.
• 8:12
  SPEAKER_00: That is absolutely true.
• 8:13
  SPEAKER_00: And I and I think that, you know, in the robotics industry, the blessing and the curse is there are a lot of companies that, you know, for whatever reason have a downturn or struggle or or whatever, and startups that fail, because startups do fail.
• 8:26
  SPEAKER_00: But it is the close-knit nature of the industry that I think matters so so much, right?
• 8:30
  SPEAKER_00: And why events like this matter so so much is those connections come in hugely important when that happens, both for the people looking for jobs, but also as people are building new teams, companies are getting into robotics, the circulation of that talent is making it possible for that knowledge and that expertise to suddenly be grabbed and brought in and internalized in companies that maybe haven't had experience in the space before.
• 8:53
  SPEAKER_00: And I think that's a good thing for the industry as a whole, that they're able to level up these non-robotics companies with expertise that comes available.
• 9:02
  SPEAKER_01: Yeah, one of the things we were also speaking about earlier that that's interesting in your role, like insofar as there are analogues for your role, other places, is that you're entering it from one of two places.
• 9:15
  SPEAKER_01: You're either uh you know the robot person, as you clearly are, or or the bus or the business person, or like the automotive person.
• 9:22
  SPEAKER_01: So it seems like for you, the learning curve has really been getting to know that industry over the past year.
• 9:30
  SPEAKER_00: Yeah, and and it's what I love coming at it from the robotics perspective.
• 9:33
  SPEAKER_00: Every new job is some cool new vertical to learn, right?
• 9:36
  SPEAKER_00: I can be useful because I understand the technology I was brought there to work on, but you know, I've gotten I worked on boats for years.
• 9:42
  SPEAKER_00: I was not a boat person.
• 9:43
  SPEAKER_00: I grew up in Ohio, not a lot of ocean in Ohio.
• 9:45
  SPEAKER_00: I know a ton about the ocean and boats and the maritime industry and navies now, that's fascinating, and I loved getting to learn that stuff.
• 9:52
  SPEAKER_00: Same thing with the drones and telecommunications.
• 9:54
  SPEAKER_00: I know way more about telecommunications than is healthy right now.
• 9:57
  SPEAKER_00: Um, same for industrial safety, functional safety, manufacturing, logistics at Amazon.
• 10:02
  SPEAKER_00: I loved seeing the logistics sausage get made.
• 10:04
  SPEAKER_01: Yeah, I mean, you know, I am a font of completely useless knowledge, uh, but it sounds like some of those things that you learned like were certainly were apt applicable.
• 10:14
  SPEAKER_01: And I was gonna ask you your time at Amazon, you know, how much, I guess, how much logistics uh can be applied to manufacturing.
• 10:22
  SPEAKER_01: But I think the more interesting question here, given that you were at VO, which you know sadly like isn't around in its current form.
• 10:30
  SPEAKER_01: If you weren't familiar with VO, they did a like a safety stop on those big industrial arms, it was really cool and promising technology.
• 10:36
  SPEAKER_01: Safety being a big part of that role, it you know, that's something that you've you have to bring to your current job from day one.
• 10:44
  SPEAKER_00: Yeah, I actually I joked both at Amazon and at GM that I'd been training my whole life for this, right?
• 10:49
  SPEAKER_00: These really random introductions.
• 10:50
  SPEAKER_00: My first introduction to safety standards was um back when Rethink Robotics was Heartland Robotics in the olden days.
• 10:57
  SPEAKER_01: Um your old boss rod.
• 10:59
  SPEAKER_00: My old boss rod.
• 10:59
  SPEAKER_00: I was an early employee there, and I joined the ISO 10 to 18 standards committee um with a couple folks in this room, and that was my first introduction to functional safety at all and the early development of a collaborative standard.
• 11:11
  SPEAKER_00: And so, you know, I was at that job for a short time, walked away going, huh, that was interesting, joined VAO years later, which by the way, tracing the lineage, Patrick Silbouvaro.
• 11:22
  SPEAKER_00: Oh, it's Spanish, VO, I see.
• 11:24
  SPEAKER_00: That's fair.
• 11:24
  SPEAKER_00: Yeah, yeah.
• 11:25
  SPEAKER_00: Um but Patrick Silbouvaro, who was one of the founders of VAO, I met him when he was working at Rethink Robotics with Rod, and Clara had been consulting as well, so there was that kind of through line there.
• 11:36
  SPEAKER_00: But working on the functional safety systems there, I ended up on the R1508 and R1506 committees under A3, met a bunch of people in this room.
• 11:44
  SPEAKER_00: Um, so both that kind of like connections to the industry as well as just the expertise on safety kind of all added up over time, and yeah, had become hugely important both at Amazon and now at GM in terms of not just ensuring we're meeting the safety standards, but understanding them so deeply that we can still innovate within them and find really creative ways to do things that still do assure this level of safety that we absolutely need.
• 12:09
  SPEAKER_01: So at Amazon, Proteus was your big project, and that that was uh I mean you can why don't you describe it?
• 12:18
  SPEAKER_00: If anyone describes it as a giant roomba, I'm gonna be really upset.
• 12:21
  SPEAKER_00: Um everyone loves doing that.
• 12:23
  SPEAKER_01: It is it is a but it is a big round robot that bounces around and and and sweeps up your floor.
• 12:29
  SPEAKER_00: So well, you know, the wheels can kind of you know spin things up, that sort of cleaning.
• 12:33
  SPEAKER_00: Um yeah, so so Proteus is a is an autonomous mobile robot that is distinct from Amazon's other robots.
• 12:38
  SPEAKER_00: Their other mobile robots run on that grid of fiducials, right?
• 12:41
  SPEAKER_00: So they are controlled centrally by a manager that tells them which fiducials to go to when and manages all the traffic explicitly.
• 12:47
  SPEAKER_00: Proteus is an autonomous mobile robot that can kind of roam freely, and there is some loose coordination at that centralized level, but it doesn't rely on the stickers on the floor and it can freely navigate using SLAM.
• 12:58
  SPEAKER_00: The goal there was to bring automation into parts of Amazon's warehouses that they couldn't fence off for these other robots because there is a lot of value there, but you can't take people out completely.
• 13:08
  SPEAKER_00: There's still a lot of work people need to do in certain areas, and you can't restructure those workflows around it.
• 13:13
  SPEAKER_00: So by introducing Proteus, it brought automation into a fully collaborative space where people were coexisting with them, working in concert with them, which is a pretty fun, fun thing to work on.
• 13:24
  SPEAKER_01: Something I wonder if probably I wonder if most people in the room know this.
• 13:29
  SPEAKER_01: I didn't know this until I visited my first Amazon fulfillment center, but all of those Kiva robots you see, all the AMRs you see, are they in a cage space where humans like generally aren't allowed to go into unless it's to repair the robots and then they have to wear that fancy vest.
• 13:47
  SPEAKER_00: Yeah, and that that's the thing is is um those robots are they're mobile robots.
• 13:53
  SPEAKER_00: They are autonomous in the sense that nobody is directly controlling them, but they don't individually think for themselves.
• 13:59
  SPEAKER_00: Proteus was such a different paradigm because it could think for itself, it could make decisions on the fly.
• 14:03
  SPEAKER_00: But what was really interesting from a kind of design standpoint and integrating these into the workplace standpoint, um people get fired if they go onto those fenced floors without the training, without the safety vest, without the right procedures, because it is dangerous for them to go through those floor.
• 14:18
  SPEAKER_01: Big metal moving object.
• 14:19
  SPEAKER_00: Yeah, and those robots are not designed to detect and stop for people or any other obstacles the way that an AMR has to.
• 14:26
  SPEAKER_00: So it's dangerous, and you do it, you get fired.
• 14:29
  SPEAKER_00: Proteus, we put out on the floor with them and told them it's fine, don't worry, right?
• 14:35
  SPEAKER_00: So they were a little pre-programmed, yeah, to get yeah.
• 14:38
  SPEAKER_00: Yeah, so there was a really interesting kind of human factor to development of Proteus to kind of interrupt and to overcome the mental model they had built up all these years of Amazon's mobile robots to get them to understand this is something different.
• 14:51
  SPEAKER_00: You can work with it differently, you can feel safe, you can trust it.
• 14:54
  SPEAKER_01: Yeah, I remember I, you know, I was I was at the event when it was announced, and you know, we we were wondering why uh uh there wasn't like a fully autonomous robot prior to that.
• 15:04
  SPEAKER_01: You know, obviously we've got companies, you know, Locus is doing well, they've got robots out on the floor that I, you know, to a certain extent are are engaging with people.
• 15:12
  SPEAKER_01: Melanie's in here somewhere, you know, Fetch was building that for a while as well.
• 15:17
  SPEAKER_01: Specifically, I guess maybe in the context of Amazon, why was building an autonomous robot that like lives among human beings such a difficult problem?
• 15:25
  SPEAKER_00: Yeah, um, I mean I'm gonna say something very controversial here.
• 15:28
  SPEAKER_00: The robot isn't the hard part, right?
• 15:31
  SPEAKER_00: Um robots, robots are a thing people know how to build, right?
• 15:33
  SPEAKER_00: My my 11-year-old built a Lego robot and made it do things autonomously, right?
• 15:37
  SPEAKER_00: And obviously it's a little harder for industrial use, but but the hard part and what Amazon really valued and prioritized is the integration of that robot or a fleet of robots into all of the work that is happening.
• 15:49
  SPEAKER_00: And so there was a complete stack, not just of the robot and the autonomy software on it, but the fleet manager was directly designed and architected to integrate with the warehouse management system and to a bunch of cloud-based services that automatically orchestrate all of the work that's going on.
• 16:04
  SPEAKER_00: The processes, the the workflows that are designed could and did change as robots were introduced.
• 16:11
  SPEAKER_00: And because Amazon could control all of that, you could get more out of the robots than dropping the robots and expecting them to work while nobody, no person changed what they were doing, no infrastructure changed, no software changed.
• 16:24
  SPEAKER_00: And so that complete end-to-end integration, that was the hard part, right?
• 16:29
  SPEAKER_00: And and that is, you know, even when I left Amazon still something that they were working on improving and making more scalable.
• 16:35
  SPEAKER_01: Yeah, and an important lesson for me uh since starting at A3 and talking to probably a lot of people in this room, is you know, obviously there are still a lot of cages, a lot of fences, and it they're still important both for like the functioning of the robot and for the safety of the people who could uh who could potentially collide with those robots.
• 16:54
  SPEAKER_01: Are people too eager to sort of remove the fences and have them live alongside humans?
• 17:01
  SPEAKER_00: Well, I mean, I understand the eagerness, right?
• 17:03
  SPEAKER_00: There is there is money in that square footage that is taken up by those fences, right?
• 17:08
  SPEAKER_00: And and there is money in um literally breaking down the walls between the manual processes and the automated processes that need to happen.
• 17:17
  SPEAKER_01: So it's literally like there's a finite amount of space, like how could we make the most of it?
• 17:20
  SPEAKER_00: Yeah, and and you know, real estate's at a premium, right?
• 17:22
  SPEAKER_00: If if you can get back a few hundred square feet, you can put more additional valuable work there that you may have to, you know, not have because you can't afford to build new space for it.
• 17:33
  SPEAKER_00: So um, and and especially, you know, as we look at how how people contribute to a lot of these processes, right?
• 17:40
  SPEAKER_00: Humans are tremendously valuable with their dexterity, their brains, and their level of judgment.
• 17:45
  SPEAKER_00: Having to in the work design separate those things mean that you could spend a couple minutes with a robot doing the wrong thing very precisely, and then you have the recovery time as a person realizes that and deals with it.
• 17:58
  SPEAKER_00: If you can have that happening in the same space at the same time, that's efficiency, right?
• 18:03
  SPEAKER_00: You can prevent a lot of errors that then have to get rectified downstream.
• 18:07
  SPEAKER_00: So I think that there's value both from a process standpoint and from a literal physical square footage standpoint, and having that cooperation, that collaboration, that teaming happening.
• 18:17
  SPEAKER_00: It's hard to do, right?
• 18:19
  SPEAKER_00: And I think that that what has been reflected in the way that the term cobalt, for example, or collaborative robot has been taken out of the standards lately is people made it so much about safety, because that was the number one hard problem to solve, that everybody assumed, okay, we solve the safety, everything's collaborative now.
• 18:37
  SPEAKER_00: But in addition to it not always being clearly solved on the safety front, um, it takes intentional design to make a manufacturing process or logistics process that is designed for a human to collaborate well with a robot.
• 18:51
  SPEAKER_00: It takes intentional effort for the software systems that run that process to allocate the right work to the person and the right work to the robot.
• 18:58
  SPEAKER_00: So again, you know, the hard part's not the robot.
• 19:00
  SPEAKER_00: It's figuring out how to go top down and say, how do I get the best out of both the people here and the machines here to get as much efficiency and quality as I can.
• 19:09
  SPEAKER_01: As we mentioned earlier, you were at Rethink, or like before Rethink was Rethink and you know, cobots insofar as I don't know, I'm not allowed to use that word anymore.
• 19:20
  SPEAKER_01: Cobots, right?
• 19:20
  SPEAKER_01: Are we?
• 19:21
  SPEAKER_00: Collaborative applications.
• 19:22
  SPEAKER_01: Yeah, collaborative applications, thank you.
• 19:24
  SPEAKER_01: But you know, obviously like Baxter was a really kind of like early version of that.
• 19:29
  SPEAKER_01: Um I still see them in labs, you know.
• 19:31
  SPEAKER_01: It's like it's like the the Wheeler Garage, it's like the PRT.
• 19:33
  SPEAKER_01: You still see them popping up.
• 19:35
  SPEAKER_01: Um what's your sense of why that didn't quite work out in its original form?
• 19:41
  SPEAKER_00: Was it just a little too early for um I think it was a little bit too early, but I I think that, you know, if we look at that and we look at kind of the push for humanoids now, right?
• 19:53
  SPEAKER_00: It is fundamentally, it's not about the legs, it's about the coordinated bimanual dexterous manipulation.
• 19:58
  SPEAKER_00: That there is something.
• 19:59
  SPEAKER_01: The arms for people who speak robots.
• 20:00
  SPEAKER_00: Right, and the hands specifically, right.
• 20:02
  SPEAKER_00: And and I think that the certainly the manipulation technology and the intelligence did not exist at the time Rethink was trying to get going to leverage to be able to get the value out of it that people want out of that capability.
• 20:16
  SPEAKER_00: And I think that's part of why they went to Sawyer, the one-armed one, is because there was still a lot to do with one-armed robots that were more intuitive, you know, had the power and force limiting and so forth.
• 20:25
  SPEAKER_00: But I think now we're seeing the same thing with humanoids of it's really about those hands.
• 20:30
  SPEAKER_00: It's about what the hands and the intelligence can do.
• 20:32
  SPEAKER_00: And until there's a a huge breakthrough in the promise of all of this technology becoming reality, um, I think that the form factor is still gonna struggle a little bit.
• 20:43
  SPEAKER_00: Um, and and you know, maybe still be a little bit ahead of its time.
• 20:47
  SPEAKER_00: What I do love is the amount of money going into this research right now.
• 20:50
  SPEAKER_00: Yeah, right.
• 20:51
  SPEAKER_00: It's it's venture capital, it's not academic dollars, but there is so much capital pouring into this problem now that it has never been spent on this in history.
• 21:00
  SPEAKER_00: Um, something's gonna come out of it.
• 21:02
  SPEAKER_00: It may not look exactly like what we see today, but a lot of really valuable stuff is gonna come out of this.
• 21:08
  SPEAKER_01: Yeah, I I you know we we keep drawing a parallel to uh self-driving cars, right?
• 21:13
  SPEAKER_01: I mean, we we kind of have them now, all these years later.
• 21:16
  SPEAKER_01: It didn't, you know, it didn't meet the timeline that a lot of people were expecting, but I we don't have humanoids and we don't have a lot of the robots that we have right now with all without all the sensors and safety and everything else that was innovated for that.
• 21:28
  SPEAKER_01: Yes.
• 21:28
  SPEAKER_01: Yeah.
• 21:29
  SPEAKER_01: How was how was that experience?
• 21:32
  SPEAKER_01: As you said earlier, you you kind of avoided iRobot early on because you know you didn't want to go to the big company, but then you end up at Amazon, and you know, now you're at GM, like pretty large companies, I think people are familiar with.
• 21:43
  SPEAKER_01: Uh how how was that jump from like startup land to suddenly going into like one of the most massive companies in the world?
• 21:51
  SPEAKER_00: I mean, it was an adjustment.
• 21:53
  SPEAKER_00: Um it did strike me though how many of the problems are the same.
• 21:58
  SPEAKER_00: Um so so I mean that was like Little bit comforting and it's a lot of people.
• 22:01
  SPEAKER_01: So it's like scale is the difference?
• 22:02
  SPEAKER_00: Yeah, there's there's scale that's a difference, and yeah, there is bureaucracy, and certainly there's different bureaucracy at GM than there is at Amazon, and you learn to deal with it, right?
• 22:09
  SPEAKER_00: No, no work environment.
• 22:10
  SPEAKER_00: It's completely free of that kind of thing.
• 22:13
  SPEAKER_00: Um but you know, you still kind of have to pitch your investors, right?
• 22:17
  SPEAKER_00: Just being at Amazon doesn't mean your program is always funded infinitely as much as you want.
• 22:21
  SPEAKER_00: You still have to justify yourself regularly with demos, with videos, show it off, give them more data, more of a roadmap, more of a plan.
• 22:29
  SPEAKER_00: It was very much like the cadence of the board meetings and investor pitching at startups.
• 22:35
  SPEAKER_00: So, you know, a lot of that is the same.
• 22:37
  SPEAKER_00: Um, and I and I think the big thing for me is, you know, it's it's big companies don't know anything small companies don't when it comes to robotics technology, right?
• 22:46
  SPEAKER_00: Everybody's really on the same playing field from a technology standpoint.
• 22:49
  SPEAKER_00: Big companies have more resources, they can put more people on it, they can put more money on it.
• 22:53
  SPEAKER_00: But the tech problems are the same for everybody.
• 22:56
  SPEAKER_00: And that was interesting to see from both sides too.
• 22:59
  SPEAKER_01: Yeah, the problems are the same, but also the people are the same.
• 23:02
  SPEAKER_01: You know, it's it's still like fundamentally, I would say, like a small industry.
• 23:07
  SPEAKER_01: Again, like everybody like went to school uh together.
• 23:10
  SPEAKER_01: And I've always said, you know, in my own job as you know, a longtime uh tech journalist, like just don't, you know, don't go to anybody's bad side because like at a certain point, like you are going to work with that person again.
• 23:22
  SPEAKER_01: Yes.
• 23:23
  SPEAKER_00: Yeah, it is it is such a small world, and people do pop back up, you know, in your circles when you least expect it.
• 23:29
  SPEAKER_00: And and and and I think that's why that's why I like a lot of the robotics industry events, is because people are very much incentivized to be friendly and to maintain the connections and and to keep it going year over year and not be cutthroat about it.
• 23:42
  SPEAKER_00: So it's a good circle to be in.
• 23:44
  SPEAKER_01: Something that uh uh Brad Porter mentioned to me when we had him at the humanoid forum that I I wasn't aware of at the time was that like early on, early on, um, you know, I think like right about the time that um that agility was making the transition into a company, that Amazon internally was exploring the humanoid form factor.
• 24:09
  SPEAKER_01: Was it was was that something that you were uh like privy to at the time?
• 24:13
  SPEAKER_00: No, I wasn't involved in that one.
• 24:14
  SPEAKER_00: But I think, you know, I think I think most most companies with any kind of operation have explored humanoids at some point.
• 24:22
  SPEAKER_00: And I think that you know, a lot of it is there is so much promise about the humanoid form factor, right?
• 24:30
  SPEAKER_00: And it's not just the current hype cycle, it's years of science fiction informing what we all think these robots are going to do.
• 24:37
  SPEAKER_00: And you know, Rod Brooks has his laws of robotics, and one of them is something along the lines of the the form of a robot implies something about what it will do and its capabilities, and you have to remember that when you're designing the form.
• 24:48
  SPEAKER_00: And and the humanoid form now is definitely writing a lot of checks that it is not clear whether or not we're gonna be able to cache.
• 24:56
  SPEAKER_00: Um, and so I think that as companies have been experimenting with this, it is a little bit of that not just is the technology ready, what could it do for me, but what does it do that's different from all the other robotic technologies out there?
• 25:08
  SPEAKER_00: And where is a bipedal form or a dual arm form or a humanoid size and shaped form uniquely valuable and useful where I can't solve that problem easily another way?
• 25:20
  SPEAKER_00: So um I think that's that's been the nature, I think, of a lot of the experiments.
• 25:24
  SPEAKER_00: Um, and and it's interesting to see, like I said, you know, all this money is coming into it, all this visibility is coming into it, good things and valuable things are gonna come out of it.
• 25:33
  SPEAKER_00: I just can't predict what shape they're going to take.
• 25:36
  SPEAKER_01: Yeah, that that that strikes me as like a fundamental disconnect between roboticists and and VCs, is that um, you know, you're out there trying to solve things as simply as humanly possible, right?
• 25:48
  SPEAKER_01: And when VCs get excited about robots, I think they're excited about the complexity of them.
• 25:54
  SPEAKER_01: So, you know, I'm sure a lot of people right now in the startup world are having the conversation of like, why isn't this a humanoid?
• 26:01
  SPEAKER_00: Yeah, and and I think there's also this assumption that humans are a general purpose form factor, therefore a humanoid robot is a general purpose form factor.
• 26:11
  SPEAKER_00: We should never have to build another form factor robot again.
• 26:14
  SPEAKER_00: Well, okay, but you know, humans are sort of we're sort of defined in our intelligence by our ability to use tools, and if we look at all the ways we interact with the world around us, you know, I don't move a pallet by lifting it, I move a pallet by using a pallet jack.
• 26:25
  SPEAKER_00: Why on earth would I make a humanoid robot that's gonna pull a pallet jack around?
• 26:29
  SPEAKER_00: What a what a suboptimal use of all of that complexity and all that capability.
• 26:33
  SPEAKER_00: And so I think we're we're gonna have to kind of imagine for our industry and for our economy this portfolio of robotics that's gonna do everything that we want them to do and not be not be attached to the idea of a particular form factor being the right one for everything.
• 26:50
  SPEAKER_01: Yeah, I suspect a big part of your job up to this point has been looking at like the robots that are out there and which like off-the-shelf can already best kind of slot into the existing systems at GM.
• 27:03
  SPEAKER_01: Is that fair?
• 27:04
  SPEAKER_00: Yeah, and I mean I think with a lot of industries that are trying to do their big transformations in in their logistics or their manufacturing, that there is that question of do we incrementally retrofit, which constrains what your your solutions face, um, or do we build all new from the ground up, rethink everything?
• 27:24
  SPEAKER_00: Greenfield, brownfield.
• 27:25
  SPEAKER_00: Yeah, greenfield, brownfield.
• 27:26
  SPEAKER_00: And and I think that's still an open question for a lot of companies because it's a huge risk and a huge amount of money to build greenfield, but it would be the biggest payoff.
• 27:35
  SPEAKER_00: And so I think that's um that's an ongoing question.
• 27:38
  SPEAKER_00: And there's gonna be a very long and happy future for companies that can support those incremental retrofits, those brownfields, because there are companies that won't make the big leap.
• 27:48
  SPEAKER_00: The companies that are gonna make the big leap, there's a whole bunch of startups out there that'll benefit from that as they have this really out-of-the-box thinking and novel technology that they can get into the space.
• 27:58
  SPEAKER_01: Yeah, I feel like I feel like people shouldn't underestimate even like the day or the hours that it would take to install even like a brownfield robot company.
• 28:07
  SPEAKER_01: If that's stopping your production line, obviously they don't want to do that.
• 28:11
  SPEAKER_00: Yeah, no, those things have to be planned out so carefully, and and there is real impact.
• 28:15
  SPEAKER_01: So ARC, so it is the Autonomous Robotics Center.
• 28:20
  SPEAKER_01: Uh, I'm trying to wrap my brain.
• 28:22
  SPEAKER_01: Well, one, like exactly what what it is that you're you're doing at GM, but also what the what ARC is.
• 28:28
  SPEAKER_01: Um, is it, you know, we we we know, certainly I know like TRI, the research they're doing over there, and that's been really um, I I think like as as close as a corporate environment can get to basically just having like a fully research lab.
• 28:45
  SPEAKER_01: Um and then uh Hyundai obviously gave Mark Ravert a ton of money to launch this thing as well.
• 28:52
  SPEAKER_01: Is this is that kind of the world that you're in right now?
• 28:56
  SPEAKER_00: No, I think the ARC is is designed to be a little bit more kind of applied science transitioning into product development or product launches.
• 29:05
  SPEAKER_00: Um we are, we do have some research and development going on, and we have really strong ties to GM's research and development team that's been around for for quite a long time doing some really neat stuff.
• 29:15
  SPEAKER_00: Um, and so we we have ways that we connect with them and pull from their work into these applied sort of activities.
• 29:23
  SPEAKER_00: Um but the the goal is really, I mean, GM pioneered the use of industrial robots and automotive manufacturing in the 70s and 80s, right?
• 29:30
  SPEAKER_00: They were one of the first movers on that front.
• 29:33
  SPEAKER_00: And so everything that can be automated with traditional technology pretty much has been.
• 29:38
  SPEAKER_00: Um these new robotic technologies offer a lot of new opportunity, but the challenge of the readiness level, the level of reliability on a very linear kind of operation, trying to get them in there and not risk downtime on the line from a robot misbehaving, that has, I think, been gating the entire automotive industry and adopting these technologies.
• 29:59
  SPEAKER_00: Technologies are getting to the point now that they are ready.
• 30:02
  SPEAKER_00: And so there's this great opportunity now to kind of level up on all of the automation that is in these plants, especially the safety gains they bring, right?
• 30:10
  SPEAKER_00: There's there's so much now that we can see in other industries where safety has been improved so much, whether it's forkless driving around that are manually driven versus autonomously driven, or whether it is just you know repetitive stress injuries that are now not an issue.
• 30:25
  SPEAKER_00: Um I think that there is so much opportunity there that GM obviously we want to jump on it.
• 30:32
  SPEAKER_00: I know other companies are too.
• 30:34
  SPEAKER_00: Um, and we're trying to bring the robotics expertise combined with the very deep manufacturing engineering expertise that we have at GM to find the right way to do it for GM's needs.
• 30:44
  SPEAKER_01: Yeah, that that's that's interesting.
• 30:46
  SPEAKER_01: And and and you know, having read some of the interviews that you've been doing since you've taken this job, and also again, just since I've I've started at A3, um obviously, as you said, automotive like tip of the spear, right, when it comes to automation.
• 31:03
  SPEAKER_01: Um so they've got this like decades-long head start for everybody else.
• 31:07
  SPEAKER_01: Um to a certain extent, did car manufacturers kind of cede that lead?
• 31:14
  SPEAKER_00: I mean, you could look at it that way, but then I go back to like the the readiness of the technology, right?
• 31:19
  SPEAKER_00: If and and it makes sense to me, having been in both worlds now, why logistics really jumped on this before manufacturing, right?
• 31:25
  SPEAKER_00: Logistics, the workflows are much more parallelized in those sites.
• 31:29
  SPEAKER_00: And so I can put in a fleet of 150 AMRs, and if one or two of them goes down, it's fine.
• 31:34
  SPEAKER_00: There's others that'll pick up the slack, right?
• 31:36
  SPEAKER_00: If I have a bunch of robotic arms sorting packages and one of them goes down, I can just reroute the packages to all the other arms and it's fine.
• 31:43
  SPEAKER_00: On a manufacturing line, you have this linear process where if one station goes down or one station runs out of material and an AMR doesn't replenish it on time, everything upstream gets backed up.
• 31:53
  SPEAKER_00: And there are very large numbers attached to every minute of downtime on that line.
• 31:58
  SPEAKER_00: And so even though logistics and manufacturing are using the same technology, the problem to solve and the readiness level and the bar for what is what is okay to be putting into the plant from a technology standpoint is very different.
• 32:12
  SPEAKER_01: One of the things, actually, like the main thing that jumped out at me yesterday uh during the uh Paul Stevens from Ford's chat is uh he described Ford as being a conservative company, right?
• 32:24
  SPEAKER_01: Which is which is it I mean it's interesting for a number of reasons.
• 32:27
  SPEAKER_01: Like obviously, like in a very like abstract sense, Ford is a technology company.
• 32:33
  SPEAKER_01: Cars are our technology.
• 32:35
  SPEAKER_01: Um we we we see a lot of very forward-facing car companies.
• 32:44
  SPEAKER_01: Like again, I think like Toyota and Hyundai fit fit in there as well.
• 32:47
  SPEAKER_01: You know, they're spending all kinds of capital to innovate.
• 32:51
  SPEAKER_01: Where on that spectrum, in your experience so far, do you feel like GM fits?
• 32:56
  SPEAKER_00: GM is I probably make the same transition that Ford is, right?
• 32:59
  SPEAKER_00: From the kind of traditional conservative automotive manufacturer to wanting to be more of a tech company and leverage more of what you know that internal expertise in technology can bring.
• 33:09
  SPEAKER_00: And it's happening on the vehicle side, right?
• 33:10
  SPEAKER_00: The sort of software-defined vehicle and advanced driver assistance systems and everything, where going from the the pure kind of mechanical nature of cars into now what can software bring to this?
• 33:22
  SPEAKER_00: Um, it's the same thing on the manufacturing side, right?
• 33:25
  SPEAKER_00: There was what worked for a very long time, um, occasional step changes in in those capabilities, but um, you know, ready now to embrace what what the newer technologies can bring.
• 33:35
  SPEAKER_01: Yeah, I'm I'm wondering again if the conversations that you have are kind of like those VC conversations where you're like you're there with uh an automotive exec and and you know he's like, oh, I I just saw the um the 60 minutes thing on uh on on Atlas.
• 33:51
  SPEAKER_01: Um how soon can we get a uh humanoid with embodied AI on it?
• 33:56
  SPEAKER_00: I mean, there's always those folks.
• 33:58
  SPEAKER_00: Even my my parents will call me up sometimes and ask, is this real?
• 34:01
  SPEAKER_00: Can I buy this?
• 34:02
  SPEAKER_00: No, no, you can't.
• 34:03
  SPEAKER_00: Um so you know there's always that, but I I think you know, as GM has been making this transition, and I don't want to say into a tech company, right?
• 34:10
  SPEAKER_00: They're not they're not gonna be AWS, right?
• 34:12
  SPEAKER_00: Um and that's that's fine, that's good.
• 34:15
  SPEAKER_00: Um They're gonna be a different kind of company, but they are gonna be leveraging technology in a fundamentally different way than than they have been, and new kinds of technology.
• 34:23
  SPEAKER_00: Um and in that transition, our leadership team has been making that shift as well, right?
• 34:28
  SPEAKER_00: We've got the the lifers, the people who've been there for ages, who are coming up to speed really quickly on this, trying to understand it and effectively making new decisions for the company around it, as well as bringing in new talent that comes from the tech industry to kind of infuse even more that understanding and that knowledge.
• 34:46
  SPEAKER_00: So the, you know, while there may have been a time that those kinds of phone calls happened, it's not, there's there's much more savviness, I think, at the upper echelons of companies like this now.
• 34:55
  SPEAKER_01: I'm gonna make you do something that I hate when people make me do, but I'm an interviewer, so I'm allowed to.
• 35:00
  SPEAKER_01: Um, you know, in terms of actual readiness, uh, you know, obviously I I I mean I I'm I'm getting from you, I think you do have some um you do believe that like at some point human rights are go are really going to be useful, maybe in industrial settings, but it's gonna take maybe a long time to get there.
• 35:18
  SPEAKER_00: Did you see that poster last night where somebody has to do?
• 35:21
  SPEAKER_00: Yeah, yeah, yeah.
• 35:21
  SPEAKER_00: It's a 2015.
• 35:22
  SPEAKER_00: I had a little red dot like hanging off of 2050.
• 35:25
  SPEAKER_01: Okay.
• 35:25
  SPEAKER_00: Yeah.
• 35:26
  SPEAKER_00: Um, I I so and and then we got in a big argument about what does working mean and what does humanoid mean?
• 35:33
  SPEAKER_00: Oh boy.
• 35:34
  SPEAKER_00: I love, I love, I love being pedantic, it's my favorite thing to do.
• 35:38
  SPEAKER_00: So all right, you're gonna regret that you put me on the soapbox.
• 35:42
  SPEAKER_00: Well, fortunately we're running out of time.
• 35:44
  SPEAKER_00: Okay.
• 35:45
  SPEAKER_00: I'll make it a fast soapbox.
• 35:47
  SPEAKER_00: Um I think that multi-coordinated multi-armed manipulation is going to be a critical unlock from a technology perspective for a whole bunch of industries, including automotive manufacturing.
• 35:59
  SPEAKER_00: Um, I think that the the AI that is being used to accelerate that work right now is also gonna be a critical unlock.
• 36:08
  SPEAKER_00: Do I think it needs to be on a human-sized and shaped torso with a head?
• 36:11
  SPEAKER_00: Not necessarily.
• 36:13
  SPEAKER_00: Do I think it needs to be two arms instead of three arms or like a king crab with one giant one and one little one?
• 36:18
  SPEAKER_00: No, I don't think so.
• 36:19
  SPEAKER_00: It would be cool, though.
• 36:20
  SPEAKER_00: It would be cool.
• 36:21
  SPEAKER_00: Um, I really want to see a company do like a prehensile tail for manipulation.
• 36:24
  SPEAKER_00: Like, come on, let's let's get creative, right?
• 36:26
  SPEAKER_00: Why are we why are we humans being so narcissistic thinking we are the ideal, right?
• 36:30
  SPEAKER_01: Like, is that is that because Melanie walked in with a Yoshi hat on?
• 36:33
  SPEAKER_00: Oh, I mean, listen, I'd be okay with it.
• 36:35
  SPEAKER_00: Maybe not the T-Rex clause.
• 36:36
  SPEAKER_00: That's probably not as useful from a manipulation standpoint.
• 36:39
  SPEAKER_00: More joints and length would be good, but um, yeah, I I think I think that there's a lot more creativity out there that we're gonna discover as we get the fundamental unlocks that we need, which, like I said, is this coordination, this intelligence, this dexterity.
• 36:51
  SPEAKER_00: Then we can look at what is the form factor that makes sense.
• 36:55
  SPEAKER_00: In an industrial environment built for forklifts and carts, I'm not convinced that you need bipedal locomotion.
• 37:01
  SPEAKER_00: Maybe you need wheels, maybe you need quadrupedal, maybe there are places you need bipedal, right, for various reasons.
• 37:07
  SPEAKER_00: I think in a home environment, yeah, you need bipedal, maybe quadrupedal, but not wheels, right?
• 37:12
  SPEAKER_00: And I think we'll get smarter over time about finding the right embodiments of this stuff.
• 37:17
  SPEAKER_00: But I think the core technology unlock is really around that manipulation and intelligence, and I want to see that happen.
• 37:22
  SPEAKER_01: Okay, real quick, and and we're gonna get your questions ready.
• 37:24
  SPEAKER_01: So um uh the the robot that you famously brought to prom as your prom date was it a humanoid that what I was thinking about I was thinking about it this morning.
• 37:39
  SPEAKER_01: What was it uh was it a humanoid?
• 37:42
  SPEAKER_01: It was not.
• 37:42
  SPEAKER_00: It was a wheeled robot, as God intended.
• 37:47
  SPEAKER_01: First of all, thanks everyone who took the later flight today for joining us and for being a part of this uh momentous piece of history.
• 37:55
  SPEAKER_01: But most of all, Michael, thank you so much for being with us.
• 37:58
  SPEAKER_00: Thank you for having me.
• 37:59
  SPEAKER_00: This is fun.
• 38:05
  SPEAKER_01: All right, that was a great conversation.
• 38:07
  SPEAKER_01: Uh, thank you so much to Michael.
• 38:09
  SPEAKER_01: We we had a uh quick chat at Automate last year, but this was really our first time actually uh having an interview of of any uh uh considerable length.
• 38:20
  SPEAKER_01: Thanks to the A3 events team for helping put that together, and thanks to you for watching.
• 38:26
  SPEAKER_01: Please like and subscribe, and uh don't forget to subscribe to the newsletter.
• 38:30
  SPEAKER_01: You can find all of that over at automated.fm.
• 38:34
  SPEAKER_01: Stay tuned, and we will see you at the same time next week for another episode of Automated.