Latest / The ReForge Podcast / Humanoid Robots Are Not Ready (with Matt and Colin) | ReForge 015

Transcript

Humanoid Robots Are Not Ready (with Matt and Colin) | ReForge 015 · The ReForge Podcast

• 0:00
  Matthew Chang: Teleoperations of a vision-enabled robot is just creepy.
• 0:04
  Matthew Chang: It's just creepy.
• 0:05
  Matthew Chang: And, you know, we deploy a lot of autonomous robotics.
• 0:08
  Matthew Chang: We've deployed mixed fleets of autonomous robotics.
• 0:11
  Matthew Chang: I've deployed autonomous vehicles on the streets, deployed autonomous vehicles for the USDOT.
• 0:17
  Matthew Chang: Like, I have a pretty deep resume in autonomy.
• 0:19
  Matthew Chang: I have an autonomous vehicle.
• 0:20
  Matthew Chang: My Cybertruck drives me around.
• 0:23
  Matthew Chang: I don't understand why you need teleop.
• 0:26
  Matthew Chang: I can't get my head around
• 0:33
  welcome to the Reforge Podcast.
• 0:35
  I'm Matthew Chang, and we're here to explore all things AI, automation, and the American worker, proving that the American reindustrialization is possible, pivotal, and pressing.
• 0:47
  Thanks for joining us today.
• 0:48
  Let's get into the episode.
• 0:51
  Colin Russ: Welcome back, everyone.
• 0:52
  Colin Russ: This is episode 15 of the"Reforged" podcast.
• 0:57
  Colin Russ: I'm Colin Russ, and Matthew Chang is here as well.
• 1:01
  Colin Russ: Today, we are going to get into everybody's favorite topic, or at least the topic of recent highlights, robots.
• 1:10
  Colin Russ: And who possibly could we ask to be better about that than the one and only Matt Chang?
• 1:16
  Matthew Chang: I just find it so amazing that, you know, we had to warm up to it for 14 episodes before we finally got here.
• 1:22
  Colin Russ: You know, we, we, we didn't, we didn't wanna overwhelm people.
• 1:25
  Colin Russ: There's a lot to it.
• 1:26
  Colin Russ: You know, it's exciting.
• 1:28
  Colin Russ: You know, we almost got, you know, blacklisted when, when I made some comments about certain robot types earlier, so we're just, you know, easing into it
• 1:36
  Matthew Chang: Yeah.
• 1:36
  Matthew Chang: All right, let's go
• 1:38
  Colin Russ: So obviously the big thing everybody talks about is, you know, humanoids versus this other big black box of robots.
• 1:48
  Colin Russ: What, you know, so many different types of robots.
• 1:50
  Colin Russ: What does it look like?
• 1:51
  Colin Russ: All the hype though is around humanoids, which, you know, I'll start off, I think they're cool.
• 1:56
  Colin Russ: I think they're straight out of sci-fi movies.
• 1:59
  Colin Russ: I am a little bit skeptical.
• 2:02
  Colin Russ: So you being the man on all of this, help me, am I completely off base?
• 2:08
  Colin Russ: You know, I, I'm, like, for example, the other week I'm watching a company have one of their interns versus a robot flipping packages over, over and over and over.
• 2:19
  Colin Russ: Seems like a gimmick to me.
• 2:21
  Colin Russ: Doesn't seem like the best use.
• 2:23
  Matthew Chang: Let, let's back up a little bit.
• 2:25
  Matthew Chang: So obviously, robotics as a field has, has been around for 30, 40 years, right?
• 2:31
  Matthew Chang: And you've been in a lot of factories that have an installed base of robots, robot arms or, um, you know, so I think it's kind of important to set some definitions here.
• 2:43
  Matthew Chang: you know, simply put, a robot is a type of mechanical automation, right?
• 2:47
  Matthew Chang: it's designed to do a thing.
• 2:49
  Matthew Chang: And, you know, our, our specialty obviously at my company, at Chang Robotics is we design robotic systems, so we seldom ever, almost never get involved with a unit of one.
• 3:00
  Matthew Chang: It's usually to support an operation, and an operation, you know, has a lot of different workers, a lot of different job tasks because they're complicated.
• 3:08
  Matthew Chang: So what we're always looking to do is automate this entire system.
• 3:12
  Matthew Chang: That's kind of the goal.
• 3:14
  Matthew Chang: And to do that, we use all different types of things.
• 3:18
  Matthew Chang: We use machines, automated machines, human assist machines, and then of course, we use robotics.
• 3:26
  Matthew Chang: And I think robotics is the great unlock.
• 3:29
  Matthew Chang: That's what takes, you know, your factory from 40, 50, 60 years ago, into the factory of the future is robotics.
• 3:38
  Matthew Chang: But what I've found to be funny, and this is not uncommon, is that this year specifically, a lot of people will say the word robot and they're referring to a humanoid.
• 3:51
  Matthew Chang: So they'll say,"Hey, I'm..." I was co- coaching two guys that wanna do a startup, beautiful concept.
• 3:56
  Matthew Chang: And they said,"Hey," so we were thinking of, doing this startup in a certain industry vertical that they were deeply knowledgeable in.
• 4:03
  Matthew Chang: They both spent almost 20 years in this industry vertical and they said,"Our startup is gonna be focused on robots." I was like,"Great, you have my attention."
• 4:13
  Colin Russ: It's the big picture.
• 4:14
  Colin Russ: Now
• 4:14
  Matthew Chang: And then as we zoomed down into the details, they were only referring to humanoids.
• 4:20
  Matthew Chang: And I'm like,"Well, that's interesting." So we had to play a little trivia, and I said,"What do you think America's installed base of humanoids is?" They're like,"Oh man, it's a lot." I said,"Well, what humanoids do you know about?" They go,"Well, we know about Apptronik, we know about Tesla, we know about Figure, we know about Unitary," and they started listing the, you know, the top of a couple dozen companies chasing the humanoid dream.
• 4:46
  Matthew Chang: And I said,"Okay, I don't have hard data on this.
• 4:51
  Matthew Chang: Nobody does, but the install base is effectively zero," right?
• 4:54
  Matthew Chang: it's basically we're at 0% saturation.
• 4:58
  Matthew Chang: Nobody is using humanoids for profitable production of goods in manufacturing, or to operate a hotel or a retail store or restaurant or any of the various settings, where you might picture humanoids, including logistics centers where you gotta flip the packages over, right?
• 5:17
  Matthew Chang: I mean, if I was solving that problem, I'd probably just have a little flipper right there and just flip them.
• 5:25
  Colin Russ: Well, I mean, and even then they've got, you know, those tubes that basically it's cameras, scanners all around them, and you can just send the packages down it, and one of the dozens of cameras in different locations or scanners can pick it up, so you don't even need to necessarily reorient,
• 5:43
  Matthew Chang: Yeah.
• 5:43
  Colin Russ: there's some of it is things
• 5:45
  Matthew Chang: case showed--
• 5:46
  Colin Russ: or fixes
• 5:48
  Matthew Chang: Right, that use case they showed, let's just first start with the underwriting.
• 5:51
  Matthew Chang: This is what I love doing is, when I see a little tiny video, I get a lot more from it than your average person.
• 5:57
  Matthew Chang: But those are my environments, logistics centers, e-commerce buildings, manufacturing buildings.
• 6:01
  Matthew Chang: That, that's my zone.
• 6:04
  Matthew Chang: So first off, that's 100% imported stuff that's coming from a low-cost country, right?
• 6:10
  Matthew Chang: That's step one.
• 6:11
  Matthew Chang: So the only-
• 6:11
  Colin Russ: actuators
• 6:13
  Matthew Chang: The only reason, I'm talking about the packages in
• 6:16
  Colin Russ: Oh, okay.
• 6:16
  Colin Russ: Yeah, yeah, yeah
• 6:17
  Matthew Chang: infamous, intern versus humanoid video.
• 6:20
  Matthew Chang: So all that stuff is coming probably from another country, and that's the only reason that it's loose and unindexed, right?
• 6:29
  Matthew Chang: See, Colin, you've worked in a lot of factories, and you've never shipped loose product that was just stuffed in, say, a sea container and unindexed.
• 6:37
  Matthew Chang: The only reason that happens is'cause you're importing usually from Asia.
• 6:41
  Matthew Chang: and if it's clothing, you know, it could be Southeast Asia.
• 6:44
  Matthew Chang: If it's, you know, advanced manufactured items, it's gonna be from China.
• 6:49
  Matthew Chang: But you're importing all this stuff to get the cube on the sea container.
• 6:53
  Matthew Chang: They didn't give you a pallet or anything to...
• 6:56
  Colin Russ: it's not nice and clean
• 6:58
  Matthew Chang: Right.
• 6:58
  Matthew Chang: It's not stretch wrap like you'd see in the back of a truck shipping something in America.
• 7:03
  Matthew Chang: They just literally stuff it in, and it's like a game of Tetris.
• 7:07
  Matthew Chang: and it can be pretty wedged in there.
• 7:10
  Matthew Chang: So then as you op- receive this sea container of foreign manufactured items and they're kind of loose stuffed in the box, you've got to have a way of getting them out.
• 7:19
  Matthew Chang: So that because you're working at the truck dock of what's probably a large logistics facility that may have as many as 100 truck doors, you know, one of the problems is it's hard to have fixed automation at 100 truck doors.
• 7:30
  Matthew Chang: That limits your flexibility and you might have big stuff coming in on one end and small stuff, small parcels coming in the other end.
• 7:37
  Matthew Chang: so you do need to have flexible automation.
• 7:38
  Matthew Chang: So I'll pause there and agree with the premise of the video that if we're gonna unpack a loose shipped, unindexed, shipping
• 7:45
  Colin Russ: A- adaptability is key
• 7:47
  Matthew Chang: Yeah.
• 7:48
  Matthew Chang: and so you see a ton of tasks out there where somebody's trying to solve it with robotics.
• 7:53
  Matthew Chang: You have robots that kind of telescope in.
• 7:56
  Matthew Chang: They have a camera and a suction cup.
• 7:59
  Matthew Chang: They'll pull items and put it on a conveyor behind them.
• 8:02
  Matthew Chang: But your point is perfect.
• 8:05
  Matthew Chang: there's really no reason for me to have sortation at the truck dock because I'm not doing anything with the product at the truck dock.
• 8:13
  Matthew Chang: I'm just simply trying to unload a container and get the product into a warehouse or into a sortation lane so it can go outbound for shipping, and that's called cross docking.
• 8:23
  Matthew Chang: So the kind of premise I think is generally wrong.
• 8:27
  Matthew Chang: What we wanna do is we wanna get all those loose items out of the sea container and into a buffer stock of inventory where we can start to process them into, you know, whatever the sortation requires.
• 8:38
  Matthew Chang: Think of a USPS facility, the mail.
• 8:41
  Matthew Chang: Huge problem, right?
• 8:42
  Matthew Chang: You're receiving a bunch of bulk.
• 8:44
  Matthew Chang: It's a little bit sorted, but not perfectly, and then we have to further sort it so we can get down to the actual mail carrier that's gonna drive down your street and hopefully put it in order so he can find the stuff that goes to your house, right?
• 8:56
  Matthew Chang: This is not new, you know.
• 8:58
  Matthew Chang: there's high-speed systems that can flip packages.
• 9:02
  Matthew Chang: There's been vision and barcode and all kind of different systems on these conveyor belts and sorters for a long, long time.
• 9:10
  Matthew Chang: They've been getting the mail to you for a long, long time.
• 9:14
  Matthew Chang: FedEx, UPS, DHL have been getting the packages to you for a long, long time.
• 9:19
  Matthew Chang: They already have pretty advanced systems too.
• 9:22
  Matthew Chang: One of the examples I always use is the UPS route optimization software.
• 9:26
  Matthew Chang: They kind of pioneered the space so that as they load these UPS trucks, the packages are loaded in the truck in the correct sequence for which the driver's gonna drive to save on time, fuel economy, safety, wanna limit left-hand turns, do more right-hand turns.
• 9:42
  Matthew Chang: And they can get the entire truck, for every truck in America, packed that way pretty close to right every time.
• 9:49
  Matthew Chang: So there is quite a bit of intelligence in logistics already.
• 9:53
  Matthew Chang: And when I watched the, you know, it was a marketing video, but when I watched the marketing video, I didn't really see how we were adding an intelligence layer.
• 10:00
  Matthew Chang: We've got a two or$300,000 robot that was sometimes flipping packages, and some astute observers questioned, did the robot adjust their Apple Watch Pro?
• 10:12
  Matthew Chang: Oh, you beat me there.
• 10:14
  Matthew Chang: And there was another one where the robot messed with its belt.
• 10:18
  Matthew Chang: But I couldn't see a belt on the robot, so it ma-made me wonder whose belt the robot was messing with
• 10:23
  Colin Russ: And I'm not one to necessarily say there's no purpose for, you know, tele-operated, you know, robots.
• 10:28
  Colin Russ: I think there's absolutely tons of potential very niche applications for it.
• 10:32
  Colin Russ: My big question when it comes to the humanoid stuff is the human form really the most optimal form factor for a robot?
• 10:40
  Colin Russ: I mean, it's predicated on the notion that humans are in fact the optimal form to achieve all of these different tasks, which arguably I don't know that we are
• 10:52
  Matthew Chang: So since you mentioned that, every robotic fleet we've ever deployed, and this is where we're deploying, you know, large numbers of robots, usually they're mobile robots, automated forklifts, autonomous mobile robots, sometimes cobots.
• 11:07
  Matthew Chang: What always happens is the staff will go behind us and they will name, they will put name tags on the robots.
• 11:14
  Matthew Chang: And I've seen some really funny ones and some that are like a human's name.
• 11:18
  Matthew Chang: You're like,"Well, I guess you can't hire somebody named Frank now because we're,
• 11:23
  Colin Russ: awkward
• 11:23
  Matthew Chang: we're naming robots Frank, so you can't hire a guy named Frank." so I've seen this naming happen.
• 11:29
  Matthew Chang: I guess I'll start with let's focus on the human side.
• 11:31
  Matthew Chang: Wherever the humanoid robot is getting deployed, there's gonna be, by definition, a workspace that's already full of humans, right?
• 11:38
  Matthew Chang: Doing that job.
• 11:39
  Matthew Chang: And they're probably gonna get named, right?
• 11:42
  Matthew Chang: And it's kind of weird.
• 11:44
  Matthew Chang: I'm not saying it's, we shouldn't do it.
• 11:46
  Matthew Chang: I'm just saying it's kind of
• 11:47
  Colin Russ: It's a little uncomfortable maybe
• 11:49
  Matthew Chang: Yeah,'cause what if you show up to work and there's already a robot named Colin?
• 11:52
  Matthew Chang: You know?
• 11:52
  Matthew Chang: And it's like, that's weird.
• 11:54
  Colin Russ: It's pretty easy to see yourself being replaced that way
• 11:57
  Matthew Chang: It's kinda like when somebody gives their dog a human name, right?
• 12:01
  Matthew Chang: And you're like,"Okay, so you named your dog Jessica.
• 12:03
  Matthew Chang: Well, that's awkward because now I guess you can never have someone named Jessica over to your house," right?
• 12:08
  Matthew Chang: So that's, that is kind of the human experience that we're gonna feel.
• 12:12
  Matthew Chang: And so I think that mix of people are gonna name those robots when they-- if they actually end up getting installed and they're permanently deployed, they're gonna get named, and it's gonna be funny.
• 12:22
  Matthew Chang: And then the other side of it is there's a huge category of really smart people that when they hear the ro- word robot, they only believe it means humanoid, and they think humanoids are everywhere already because they see so many videos on X and LinkedIn and YouTube
• 12:37
  Colin Russ: I'm a big proponent of, you know, I love the AMRs, I love the cobot six degree of freedom robot arms.
• 12:45
  Colin Russ: The flexibility that you have, in my opinion, taking more of a perhaps, I don't wanna say less capable, but more focused form and function as to what it does.
• 12:57
  Colin Russ: You know, the automated forklifts picking up pallet, moving it from point A to point B, you know, it has a very specific form factor, but it does that very well.
• 13:07
  Colin Russ: Less, you know, pieces, bells, and whistles to complicate and potentially break.
• 13:12
  Colin Russ: But if you're talking about something on, you know, a cobot that you've got posted up right outside a machine, it's lifting, it's grabbing, it's setting, it's releasing, you know, it's repetitive, and by no means are these simple, you know, machines or operations, but they are simpler than thousands of, you know, different actuators inside of this complex being that still needs to have its battery replaced every four to six hours because it's
• 13:42
  Matthew Chang: Oh, I, my, I personally think that at full load conditions, these robots are gonna need to charge every 20 to 30 minutes.
• 13:51
  Matthew Chang: I don't think it's four to six hours you're gonna get
• 13:53
  Colin Russ: just, I don't I think that's one of the big issues too, though, that we don't talk about nearly that much, is the energy density of power, you know, of batteries is improving tremendously.
• 14:01
  Colin Russ: I love it.
• 14:02
  Colin Russ: Great.
• 14:02
  Colin Russ: That's another topic for another day.
• 14:04
  Colin Russ: But to your point, full load, you know, it's...
• 14:09
  Colin Russ: where's the efficiency gains if you're having to swap batteries every 30 minutes?
• 14:13
  Matthew Chang: Yeah, I actually don't even think that batter-battery swap is gonna work.
• 14:18
  Matthew Chang: So from the leading OEMs, you know, say Boston Dynamics, Tesla, Figure, Apptronik, if I left somebody off, forgive me, but take the top five or six.
• 14:29
  Matthew Chang: Have you seen any of them demonstrate a battery swap system,
• 14:33
  Colin Russ: I
• 14:33
  Matthew Chang: right?
• 14:34
  Matthew Chang: So but yet all defenders and especially investors, the people that invest in humanoids right now, which they have obviously collected hundreds of billions of investment, oh man, those investors will say they battery swap.
• 14:46
  Matthew Chang: But to, to an eye like mine, you've shown me sorting packages, you've shown me doing kung fu, you've shown it doing a cartwheel, whatever the robot can do.
• 14:55
  Matthew Chang: Boston Dynamics at least had the courtesy to show theirs lifting a mini fridge.
• 15:00
  Matthew Chang: I think you and I probably that resonated.
• 15:01
  Matthew Chang: We're like,"Oh yeah, robot, bring me a
• 15:03
  Colin Russ: that would be really nice
• 15:04
  Matthew Chang: fridge." But what you haven't seen yet is a battery swap system.
• 15:11
  Matthew Chang: 'Cause newsflash, if you install a battery swap system, this probably needs an electrical permit.
• 15:17
  Matthew Chang: it probably needs to be inspected by the factory or logistics center EH&S.
• 15:22
  Matthew Chang: and it should be kind of obvious and branded, and presumably we don't necessarily want humans next to it, so it needs to be marked.
• 15:28
  Matthew Chang: where's the battery swap system if they're supposed to swap batteries?
• 15:32
  Colin Russ: there's a bit of skepticism on my end too, part-- in line with the battery swapping is, you know, there's talk for a long time about, you know, EVs and the battery swapping.
• 15:43
  Colin Russ: we couldn't get it, you know, see the mass adoption or see it engineered in a way that was practical, economical, and made sense for something like EVs.
• 15:53
  Colin Russ: I've seen some of the solutions, but how are we gonna do that when we're talking constant, never-ending battery swaps basically for the supposed millions of these humanoids we're,
• 16:02
  Matthew Chang: I mean, I could see it and I can almost design one in my head right now.
• 16:05
  Matthew Chang: The point is you haven't no one's shown you one, so likely it doesn't exist.
• 16:11
  Matthew Chang: That or it does exist and it's not very reliable or nice to look at, so they won't show it to you.
• 16:18
  Matthew Chang: So let's back up and focus on what's good.
• 16:20
  Matthew Chang: the movement from the humanoid's super impressive, right?
• 16:23
  Matthew Chang: Across the board.
• 16:24
  Matthew Chang: The walking, the performances, the martial arts, all of that i-is excellent.
• 16:29
  Matthew Chang: it's, you know, exciting.
• 16:32
  Matthew Chang: What I have seen that is the best in the industry is certainly Atlas by Boston Dynamics.
• 16:38
  Matthew Chang: And I'll explain why, if I had to choose one humanoid to install...
• 16:43
  Matthew Chang: And by the way, my personal goal is I want Chang Robotics to be the first company to install and deploy a team of 100 humanoids.
• 16:52
  Colin Russ: Nice
• 16:53
  Matthew Chang: we have to find the right application where it makes sense, because we'd never offer a, you know, a tier B solution to our clients.
• 17:00
  Matthew Chang: But let's say that we find a, use case where it makes perfect sense.
• 17:04
  Matthew Chang: For the first large system deployed, the large team, I'd like to be the one to do that.
• 17:09
  Matthew Chang: So I do watch it closely, but I have to underwrite it from the point of view of is this performance-oriented to where I can offer it to a customer of mine?
• 17:17
  Matthew Chang: So let's look at, Atlas by Boston Dynamics.
• 17:21
  Matthew Chang: Number one, and the most important thing for me, it's six foot two.
• 17:25
  Matthew Chang: Okay?
• 17:25
  Matthew Chang: It is a big humanoid.
• 17:27
  Matthew Chang: Why?
• 17:28
  Matthew Chang: Because some of the best workers are big humans.
• 17:31
  Matthew Chang: I'm six four and I can relate to it.
• 17:33
  Matthew Chang: I'm like, yeah, to, you know, put the bag in the overhead bin and to actually see what's in the overhead bin, you have to be above a certain height.
• 17:41
  Matthew Chang: that's true in factories too, right?
• 17:43
  Matthew Chang: Not everything is set so that a, five foot two person can reach it.
• 17:47
  Matthew Chang: the next is when you look at the humanoid, it's got, detachable tooling at the forearm.
• 17:54
  Matthew Chang: So just past the halfway point of the forearm about right here, you'll see a silver bracelet on it.
• 18:00
  Matthew Chang: That is a tool connect, you know, which a lot of CNC machines and five, six, seven-axis machines can auto-change tooling.
• 18:09
  Matthew Chang: That tells me the architecture of Atlas is set up for auto tool exchange.
• 18:14
  Matthew Chang: The other thing I like about-
• 18:15
  Colin Russ: huge in terms of being able to actually have real-world applications to it.
• 18:20
  Colin Russ: Having, you know, soft jaw grips that you can, you know, tailor to a specific part or component it's moving.
• 18:26
  Colin Russ: I mean, It's clear that it's not relying on a catch everything, you know, universal type gripping mechanism.
• 18:31
  Matthew Chang: I'm very skeptical of the hand.
• 18:34
  Matthew Chang: You know, this idea of the hand, the dexterity, the sensation of the hand, doesn't make any sense to me.
• 18:42
  Matthew Chang: You know?
• 18:42
  Matthew Chang: I wouldn't go so far as you to say that, like, the humanoid form factor doesn't make sense.
• 18:46
  Matthew Chang: I'd say it makes way less sense than we think it does.
• 18:48
  Matthew Chang: But, it could.
• 18:50
  Matthew Chang: There could be use cases where it makes sense.
• 18:52
  Matthew Chang: what certainly doesn't make sense is to try to robotically produce a five-finger hand right now.
• 18:58
  Matthew Chang: We're very far away from that working.
• 19:01
  Matthew Chang: like very far away.
• 19:02
  Matthew Chang: the science on that to make that happen with human-like, ability is still in basic research in the top robotics institutions around the country.
• 19:13
  Colin Russ: I mean, I, I agree.
• 19:14
  Colin Russ: And I've seen some cool stuff, like when they're talking about basically synthetic tendons and all of the crazy things that they're basically replicating human biology with, you know, actual hardware.
• 19:27
  Colin Russ: Super cool, but I, you're I mean, I agree completely.
• 19:30
  Colin Russ: We're, we're not there
• 19:32
  Matthew Chang: Yeah, it's really far away, right?
• 19:34
  Matthew Chang: And I hate that for-- if somebody's watching this and you just put 10 million bucks into a humanoid company.
• 19:39
  Matthew Chang: For the hand, let's have a little humility.
• 19:42
  Matthew Chang: It's gonna be a long, long time.
• 19:44
  Matthew Chang: We're talking 10 years plus, outside of 10 years before somebody can make something that somewhat functions like a human hand.
• 19:53
  Matthew Chang: So the next thing I like about Boston Dynamics, they went with the three-digit hand, and they used what are clearly common, robotic components for that actuator.
• 20:04
  Matthew Chang: It's not a hand, it's an actuator, right?
• 20:07
  Matthew Chang: And so, it's a manipulator.
• 20:09
  Matthew Chang: It's an end-of-arm tool.
• 20:11
  Matthew Chang: And so at the end of the arm, you have a three-digit hand.
• 20:14
  Matthew Chang: It has, 360 degrees of motion, which improved performance over again.
• 20:20
  Matthew Chang: You know, you and I have about 180.
• 20:21
  Matthew Chang: We still have all of our mobility.
• 20:23
  Matthew Chang: This has 360 degrees.
• 20:26
  Matthew Chang: And what that allowed them to do with having the large format robot, obviously, they've been very good at balance for a while.
• 20:32
  Matthew Chang: They kind of pioneered the space.
• 20:34
  Matthew Chang: And then also the three-digit hand is they actually have a load rating.
• 20:39
  Matthew Chang: So I don't know if it's totally public information, but certainly I have already been sent a PDF on Boston Dynamics branding that has the load rating and the performance characteristics of Atlas.
• 20:51
  Matthew Chang: So before I can design any humanoid into an operation, I need to know the full range of specs that it's capable of.
• 20:59
  Matthew Chang: I haven't seen one for Atlas, Figure, or any of the other ones.
• 21:02
  Matthew Chang: I would never buy a robot that didn't have a spec sheet because I need to know what the OEM thinks the performance is.
• 21:10
  Matthew Chang: I need to design that in, and then I need a mechanism to hold them accountable if we can't meet performance, right?
• 21:16
  Colin Russ: And I mean, I think that's reasonable for anyone looking...
• 21:20
  Colin Russ: You know, we're talking about these concepts of, oh, have it in your house and it can do your laundry or whatever, and it's like that's nice and all, but to your point, you're not likely buying a new refrigerator without at least, you know, looking up the specs, you know, see, oh, how many cubic feet does it have?
• 21:34
  Colin Russ: Does it have, you know, this, that, or whatever.
• 21:36
  Colin Russ: Like, you wanna understand what you're getting for it, not just the hype and marketing videos.
• 21:43
  Colin Russ: And frankly, the lack of that, I don't wanna say transparency'cause that seems a little critical and harsh, but the lack of information and understanding of what they're truly capable of beyond hype videos, which, I mean, we've seen enough companies, not even robotics-related, but just these startups that are really well-funded and promise the world but don't do a tenth of that.
• 22:06
  Colin Russ: You know, a couple, maybe a Blood one, maybe some, you know, trading-type crypto companies that have done that sort of thing.
• 22:13
  Colin Russ: It doesn't end well.
• 22:14
  Colin Russ: You know, it eventually implodes, and I think that these robotics companies have gotta start to share a bit more of that data, a bit more capabilities and hard numbers.
• 22:23
  Matthew Chang: Well, there's a reason they're not, right?
• 22:26
  Matthew Chang: There's a reason they're not.
• 22:26
  Matthew Chang: It takes a lot of effort to get ready for the marketing videos, but it's even harder to leave a humanoid or any robot in twenty-four/seven production or operations environment where it doesn't have to be intervened with often.
• 22:39
  Matthew Chang: That takes a huge level of robustness.
• 22:42
  Matthew Chang: So the fact that you haven't seen these published stats means either, one, they don't have it, or it would be a little embarrassing for them to share it.
• 22:49
  Matthew Chang: You know, if they tell you that the robot can only reliably manage a fifteen-pound load, does that change your opinion of the capabilities of the industry, right?
• 22:58
  Matthew Chang: Now, I do happen to know the Boston Dynamics one.
• 23:02
  Matthew Chang: So Atlas is rated at a center of gravity sixty-pound lift, six-zero.
• 23:07
  Matthew Chang: That's pretty good because most job applications say you gotta lift fifty pounds to be compliant with OSHA.
• 23:12
  Matthew Chang: And they showed it, it can pick up the mini fridge with no tooling.
• 23:16
  Matthew Chang: the second is it has a non-center of gravity load rating of thirty-five pounds.
• 23:22
  Matthew Chang: Thirty-five pounds is pretty good.
• 23:24
  Matthew Chang: I've got a set of kettlebells in the
• 23:25
  Colin Russ: Yeah, I was about to say, you hold that out in front of your arms for a little while.
• 23:28
  Colin Russ: that's not easy
• 23:30
  Matthew Chang: Yeah.
• 23:30
  Matthew Chang: you start here and the£35 is not bad.
• 23:33
  Matthew Chang: Somewhere about right here, that£35 is pretty heavy, right?
• 23:36
  Matthew Chang: So that's a pretty good indicator.
• 23:38
  Matthew Chang: Now, I'm sure it declines.
• 23:39
  Matthew Chang: It's not gonna hold it with a straight arm and, you know, those are things that we'll have to learn about over time, is the exact performance so that you can design the condition and the environment for the robot, right?
• 23:49
  Matthew Chang: You have to marry the two.
• 23:50
  Matthew Chang: The physical environment has to be designed and the robotic function has to be designed.
• 23:55
  Matthew Chang: They have to meet in the middle.
• 23:57
  Matthew Chang: So I see that as a glimmer of hope that there's at least one company that's willing to publish some preliminary stats on what the robotics can do.
• 24:07
  Matthew Chang: And then I think you went to-- you mentioned the drivelet-- the, EV car and the battery swap idea.
• 24:13
  Matthew Chang: That I think was a very well-intentioned but immature idea from when lithium costs were high on a per kilowatt hour basis, and then, battery ranges didn't have quite the level of density they have right now.
• 24:31
  Matthew Chang: So the idea was because we don't know when the readiness of lithium-ion batteries is gonna intercept, we should do battery swap.
• 24:39
  Matthew Chang: Battery swap is incredibly dangerous, okay?
• 24:43
  Matthew Chang: the weight of-- I drive a Cybertruck.
• 24:45
  Matthew Chang: The weight of my
• 24:46
  Colin Russ: are massive
• 24:47
  Matthew Chang: you do not wanna swap that battery.
• 24:50
  Matthew Chang: The battery is in fact part of the structure of the vehicle.
• 24:53
  Matthew Chang: So Cybertruck is interesting because it's the most futuristic vehicle.
• 24:58
  Matthew Chang: You may not like the way it looks.
• 24:59
  Matthew Chang: I happen to.
• 25:00
  Matthew Chang: My kids like it.
• 25:01
  Matthew Chang: But, it's the most futuristic vehicle.
• 25:03
  Matthew Chang: It drives itself.
• 25:04
  Matthew Chang: But it's also the safest truck on the market in its category for crash testing and any other way of measuring it.
• 25:11
  Matthew Chang: Well, part of the reason of that is the entire floor of the vehicle is a battery, the entire floor from front to back, and it goes all the way back into the trunk.
• 25:20
  Matthew Chang: The battery is integral to the structure of the vehicle, so it's part of the structural integrity.
• 25:25
  Matthew Chang: It's integral to the load rating.
• 25:27
  Matthew Chang: And then the other part that people just, I think it's just lack of education around how energy works.
• 25:32
  Matthew Chang: The battery is thermally managed.
• 25:34
  Matthew Chang: My Cybertruck is a radiator, okay?
• 25:37
  Colin Russ: I was gonna get to that on the robots in a minute, is how we address the cooling piece of it,'cause that's
• 25:41
  Matthew Chang: Yeah.
• 25:42
  Matthew Chang: Well, so in EVs, which are capable of really powerful discharge, right?
• 25:47
  Matthew Chang: Almost every EV is wicked fast, and I know they're, they are really fun to drive.
• 25:52
  Matthew Chang: One of the challenges we have is that if I'm doing a massive power dump because I'm flooring it over a sand dune, which maybe me and my kids have done or not done, I won't disclose it here, then I've created a ton of heat through that energy release, and it has to be cooled off through the radiator.
• 26:11
  Matthew Chang: So there's liquid, there's a glycol solution circulating in the battery, and then it goes out to a radiator and back.
• 26:17
  Matthew Chang: similarly, the motors have to be cooled, right?
• 26:20
  Matthew Chang: Because they get hot as well from that massive energy dump.
• 26:23
  Matthew Chang: And we have the reverse happens in the winter when it's cold and we want our battery pack to actually work.
• 26:29
  Matthew Chang: The battery takes a little time to warm up.
• 26:32
  Matthew Chang: So if you cold start a Tesla, it'll give you a little snowflake icon, and it'll tell you how much degraded the power level is until the pack warms up through the radiator, and then it will go back to normal.
• 26:43
  Matthew Chang: But if you have--
• 26:44
  Colin Russ: energy.
• 26:45
  Colin Russ: Gotta hit that, you gotta hit that right temperature
• 26:48
  Matthew Chang: So that's an important thing to, to follow.
• 26:50
  Matthew Chang: My theory for humanoids is they need to use the same battery format as the EVTOL aircraft.
• 26:58
  Matthew Chang: So, you know, what's interesting about an EVTOL, those are the vertical takeoff, mini airplanes, mini helicopters, they don't have the weight tolerance to put in a thermally managed system, so they have to go with an air-cooled battery pack.
• 27:11
  Matthew Chang: It's kinda like an old VW Bug, right?
• 27:14
  Matthew Chang: The Volkswagen Bug was like the only modern car, somewhat modern, that didn't have a liquid cooling system.
• 27:20
  Matthew Chang: It was air-cooled.
• 27:21
  Matthew Chang: And that's what you really have to have, I think, in the humanoid.
• 27:24
  Matthew Chang: We can't put a liquid system on there.
• 27:27
  Matthew Chang: We need to have a high punch, low weight battery.
• 27:31
  Matthew Chang: And so I think the unlock in humanoids for battery efficiency is gonna be coming from the EVTOL industry over.
• 27:39
  Matthew Chang: That's a free one.
• 27:40
  Matthew Chang: If any of you happen to have built a EVTOL battery plant, that is a little free nugget for you.
• 27:45
  Matthew Chang: Get into the
• 27:46
  Colin Russ: I mean, that was one of the questions I had,'cause, I mean, you can't have, you know, a bunch of, you know, cables, tubing running basically through a factory tied up to every single robot.
• 27:56
  Colin Russ: You're gonna produce a ton of heat, you know, no matter what.
• 27:59
  Colin Russ: So I, I think seeing how they address the cooling, particularly from, you know, an air cooling, perspective, that'll be interesting to see on the humanoid side of things.
• 28:09
  Matthew Chang: It's just gonna have a, just gonna have a little fan somewhere in the mid-back that just kinda passively, you know, on draws air through on a high heat condition.
• 28:19
  Matthew Chang: But that's just one more reason kinda going through these complexities of battery swap and really matching the battery design and chemistry to the power level.
• 28:27
  Matthew Chang: That's a reason why I don't think battery swap makes sense in humanoid.
• 28:32
  Matthew Chang: I think it's the same short-sightedness that led people to speculate you should build battery swap for EV cars, and then the industry just moved past it.
• 28:39
  Matthew Chang: and then let's not leave out what, what happens in a Tesla on a road trip?
• 28:44
  Matthew Chang: Supercharging network.
• 28:46
  Matthew Chang: It's everywhere.
• 28:47
  Matthew Chang: So it's so much easier for us in a factory to put opportunity charge zones where the robot has a little designated place to go to get a top-up charge.
• 28:56
  Matthew Chang: That's why I don't think battery swap makes sense.
• 28:59
  Matthew Chang: I would just focus on opportunity charge, and if the robot needs to have a little top-up charge 10 minutes out of every 60 minutes, great.
• 29:08
  Matthew Chang: Instead of having six robots, we'll add one more, and we'll have seven robots.
• 29:12
  Matthew Chang: One is always charging
• 29:14
  Colin Russ: It's playing the musical chairs game at that point.
• 29:15
  Matthew Chang: That's right.
• 29:16
  Matthew Chang: Yep
• 29:16
  Colin Russ: I mean, and that gets to, I think, a bigger point is, you know, really successful adoption of a lot of these future robot humanoid-type technologies is gonna require a factory fundamentally designed to accommodate that.
• 29:31
  Colin Russ: Not even just the robots, but the infrastructure needed for them to be effective and actually have a real impact
• 29:37
  Matthew Chang: Yeah.
• 29:37
  Matthew Chang: Yeah.
• 29:38
  Matthew Chang: So as a reminder for the audience, I'm a professional engineer.
• 29:40
  Matthew Chang: I'm a licensed engineer, and so from the licensed engineer perspective, let me give you kind of a punch list.
• 29:47
  Matthew Chang: Number one, does your insurance carrier on the property agree with you having humanoids there?
• 29:53
  Matthew Chang: And if so, do they have an underwriting process to review the way you're using them so that you're not increasing the the risk hazard in the building, and thereby increasing the insurance premium?
• 30:04
  Colin Russ: So no after-hours robot fights
• 30:07
  Matthew Chang: yeah, that would be frowned upon.
• 30:09
  Matthew Chang: Yeah.
• 30:09
  Matthew Chang: If the maintenance guys program the robots to fight each other at night, yeah, I could see that happening.
• 30:15
  Matthew Chang: then the second would be, does the fire marshal have an issue with it?
• 30:20
  Matthew Chang: So let's say the insurance carrier's good.
• 30:22
  Matthew Chang: We gotta now talk to the fire marshal.
• 30:25
  Matthew Chang: The average listener's sitting here going,"What on earth would the fire marshal care?" Fire marshal's gonna care a lot.
• 30:31
  Matthew Chang: If that humanoid has an issue where it breaks down underneath a fire shutter door and prevents the safe operation of the fire features of the building, we have a problem.
• 30:42
  Colin Russ: It's a big problem
• 30:43
  Matthew Chang: If that humanoid is not waterproof and we set off the sprinklers or part of the liquid process that's already happening in the building, and we spark up the humanoid and have a lithium fire on top of the other fire that we've got, that's a problem, right?
• 30:58
  Matthew Chang: So you can war game that down the line on the many ways that the, fire marshal is gonna be very interested.
• 31:08
  Matthew Chang: I would suggest that a minimum starting test for safety is what does the humanoid do during an emergency egress situation?
• 31:17
  Colin Russ: Interesting
• 31:18
  Matthew Chang: If I'm the fire marshal, I need to see it happen, right?
• 31:20
  Matthew Chang: Is it fleeing out the door with the other pedestrians, or is it going to an out-of-the-way zone to where it can hide and just burn up with the building?
• 31:30
  Matthew Chang: We need to see that work.
• 31:32
  Matthew Chang: The next is, whatever gets installed, unless it powers off a one twenty volt outlet and you just plug it in the wall, if it has to be installed with the building, such as a battery swap system or a, you know, scaled up opportunity charge zone, it's gonna require an electrical permit, right?
• 31:47
  Matthew Chang: So these are just a couple examples of how, from a regulatory perspective, it totally matters what's happening.
• 31:54
  Matthew Chang: And so I think, you know, obviously the very smart people, the roboticists and AI engineers that are building these things, you know, they're not there yet.
• 32:03
  Matthew Chang: It's like that's an inconvenient reality.
• 32:04
  Matthew Chang: But you saw, you had that happen in
• 32:06
  Colin Russ: It's a Futurist problem.
• 32:07
  Colin Russ: Futurist problem
• 32:08
  Matthew Chang: Do you remember the, autonomous trucking company that couldn't comply with DOT rules, so they got their permit revoked for a little while because they couldn't safely notify other vehicles when they had to pull over for an emergency stop?
• 32:21
  Matthew Chang: 'Cause the standard protocol is you gotta go lay out cones, and the robot truck couldn't lay out cones.
• 32:26
  Colin Russ: there's nobody in it to do that
• 32:28
  Matthew Chang: Right.
• 32:29
  Matthew Chang: Yeah.
• 32:29
  Matthew Chang: So you're gonna have this kind of messy middle, and I theorize that a couple things are gonna happen.
• 32:36
  Matthew Chang: One is a lot of property owners are gonna ban humanoids.
• 32:41
  Matthew Chang: Who went first?
• 32:42
  Matthew Chang: Do you remember who went first recently and banned humanoids?
• 32:46
  Colin Russ: I do not know.
• 32:47
  Matthew Chang: Southwest Airlines
• 32:49
  Colin Russ: Oh, oh, yeah, because they couldn't, they didn't have a, thermal runway bag for, the battery size for the humanoid
• 32:55
  Matthew Chang: Yeah.
• 32:56
  Matthew Chang: And the, you know, so, for the audience, somebody very cutely bought an airplane ticket for their humanoid, and they marched it on and it sat and it rode Southwest with them.
• 33:05
  Matthew Chang: And that was the first and last commercial flight of a humanoid.
• 33:08
  Matthew Chang: similarly, all those, one of those things that the kids were using, that got banned from the airlines, the hoverboards
• 33:14
  Colin Russ: Yeah, the hover- scooter hoverboard things.
• 33:16
  Colin Russ: Yep
• 33:17
  Matthew Chang: Yeah.
• 33:17
  Matthew Chang: So those got banned.
• 33:18
  Matthew Chang: so you already have them, you know, pretty much de facto banned from commercial aviation.
• 33:23
  Matthew Chang: I think you're gonna have the same thing in apartment buildings, right?
• 33:27
  Matthew Chang: A lot of guests are not gonna be comfortable having a video-enabled spy robot walking around in a building.
• 33:37
  Matthew Chang: So I think apartment buildings, you're gonna see selective bans.
• 33:42
  Matthew Chang: hotels probably follow the same process, you
• 33:46
  Colin Russ: It's gonna be an uphill battle, I think.
• 33:48
  Colin Russ: for that type of adoption, it's gonna be tough, I think
• 33:52
  Matthew Chang: Yeah.
• 33:52
  Matthew Chang: you had the same when, Uber started to proliferate, right?
• 33:56
  Matthew Chang: And Airbnb had the same.
• 33:57
  Matthew Chang: As they entered new markets, there was a natural pushback, and a lot of the kind of the free, unbridled capitalism that let them get their fast start really got constricted down, and they were given,"Well, if you wanna go to the airport, you gotta pay extra.
• 34:13
  Matthew Chang: Well, you can't drive in the zone." Airbnb has been severely limited in a lot of big cities to where, you know-
• 34:19
  Colin Russ: to hotel taxes now and all that type of stuff
• 34:22
  Matthew Chang: Yeah.
• 34:23
  Matthew Chang: so that will be coming because you raised that.
• 34:25
  Matthew Chang: Who's paying the payroll tax on the humanoid, right?
• 34:27
  Matthew Chang: If you're a congressman and there's a job loss in your area from humanoid deployment, that sounds like a political issue in waiting, right?
• 34:36
  Colin Russ: it almost feels like something out of, you know, the discourse in"I, Robot," that old Will M- Will Smith film where everyone's freaking out about the robots and, you know, we're gonna get into that philosophical debate in, you know, next couple years'cause of exactly this
• 34:51
  Matthew Chang: Yeah.
• 34:51
  Matthew Chang: So I think that's the first is you are gonna see a lot of restriction done at the very local level, the airline, the apartment building, those sort of things.
• 35:00
  Matthew Chang: But I did go ahead and get a head start on this, and I surveyed a couple of my customers.
• 35:05
  Matthew Chang: we've been offered our first, humanoid distributorship, right?
• 35:08
  Matthew Chang: Where we would have access to one of these humanoids.
• 35:12
  Matthew Chang: Well, we ha-- I didn't say we signed it, right?
• 35:14
  Matthew Chang: It's something that we're-- we wanna study it because, for us, it's all about the customers, right?
• 35:19
  Matthew Chang: I need to know what they think, and I need to find use cases where the humanoid is more efficient than the other forms of robotics and automation that we design.
• 35:30
  Matthew Chang: So I went ahead and talked to a couple customers, and what I was amazed with is, one of my clients immediately had the answer.
• 35:37
  Matthew Chang: Didn't have to think about it.
• 35:38
  Matthew Chang: I said,"Hey, I'm interested in using your factories to deploy humanoids, and I'd like to get your point of view on it." Without a hesitation, there was no gap in the response.
• 35:50
  Matthew Chang: He said,"Then they will need to be free." Okay.
• 35:55
  Matthew Chang: So I'll give you that first.
• 35:56
  Matthew Chang: They will need to be free.
• 35:57
  Matthew Chang: I was like,"Ooh, ouch." Free
• 35:58
  Colin Russ: big, contingency there
• 36:00
  Matthew Chang: Yeah.
• 36:01
  Matthew Chang: Premium.
• 36:01
  Matthew Chang: Yeah.
• 36:01
  Matthew Chang: That hurts.
• 36:02
  Matthew Chang: Okay.
• 36:03
  Matthew Chang: Yeah.
• 36:03
  Matthew Chang: We are a profit make-profit making company, Colin, so free is a interesting hurdle, right?
• 36:08
  Colin Russ: It's very expensive for one of the parties
• 36:10
  Matthew Chang: Number two, they cannot record Hmm.
• 36:16
  Matthew Chang: Isn't that kind of the point of that whole vision setup they have up there is that they're recording everything?
• 36:21
  Matthew Chang: Don't forget-- Don't, let me get out of here without talking about Telia before we're done, but they must-- They may not record.
• 36:28
  Matthew Chang: It's not...
• 36:28
  Matthew Chang: Well, that's probably a deal killer for every single one of these humanoid companies, is they wanna have full vision and all the analytics running, beaming out to a cell tower somewhere, not
• 36:41
  Colin Russ: They wanna be able to train on it.
• 36:42
  Colin Russ: They want those edge cases
• 36:44
  Matthew Chang: And he said,"I will not allow you to have the visual data.
• 36:49
  Matthew Chang: You can keep the robotic telematics." And I was like,"All right, good answer.
• 36:55
  Matthew Chang: Okay, that's actually what I was gonna say too, so good answer." And then, number three, can't talk without going through my network.
• 37:04
  Matthew Chang: Basically on-premise, right?
• 37:06
  Matthew Chang: It's gotta be on-premise.
• 37:09
  Matthew Chang: Ouch.
• 37:10
  Matthew Chang: Now, that doesn't hurt me because what I just laid out is how we design everything.
• 37:14
  Colin Russ: Well, and it, and everything they're asking for is not crazy.
• 37:18
  Colin Russ: I mean, freemium I get, you know, that one's hard to stomach.
• 37:20
  Colin Russ: But the other stuff, you know, that's basic, you know, network security, operation security, trade practices.
• 37:26
  Colin Russ: You can't do anything ITAR related if you're handing off all that footage.
• 37:29
  Colin Russ: Like that's
• 37:30
  Matthew Chang: Yeah, let me explain why freemium.
• 37:32
  Matthew Chang: So, you know, basically none of my customers let me advertise, right?
• 37:37
  Matthew Chang: if you have serious manufacturing companies, you're not advertising the secretive work that you're doing in their facilities, right?
• 37:43
  Matthew Chang: and if you don't know that you're new to the industry, you must come from SaaS, that you don't know that's nearly universally true.
• 37:50
  Colin Russ: Those are fighting words there
• 37:51
  Matthew Chang: yeah.
• 37:52
  Matthew Chang: Yeah.
• 37:53
  Matthew Chang: So, okay, you're not allowed to advertise.
• 37:56
  Matthew Chang: the whole point of doing a deployment for a high-profile manufacturer, which is who we're talking about here, is that obviously the humanoid company wants to advertise, right?
• 38:05
  Matthew Chang: I wanna put a camera in there, show it working, count the number of hours, the number of tasks, show how scarred up the hands are after all the hard work.
• 38:12
  Matthew Chang: That's what they wanna do.
• 38:14
  Matthew Chang: So the cost to be able to advertise is I need to have it for free.
• 38:20
  Matthew Chang: That was the trade, right?
• 38:21
  Matthew Chang: And if I could show an ROI, all of my clients will pay up to their ROI threshold, right?
• 38:27
  Matthew Chang: Of whatever it is, whether it's a, say, a two-year payback, 18-month payback, three-year payback.
• 38:32
  Matthew Chang: You know, they'll pay up to that amount and not a penny more.
• 38:35
  Matthew Chang: But you want to advertise?
• 38:37
  Matthew Chang: Give it to me for free, right?
• 38:39
  Matthew Chang: That's the bargain.
• 38:39
  Matthew Chang: So if those are the three rules that I think are pretty universal across large-scale manufacturers, it's gonna be a tough market adoption for humanoids in manufacturing.
• 38:51
  Matthew Chang: So
• 38:51
  Colin Russ: I think so.
• 38:51
  Colin Russ: I think those are your-- that's your starting point.
• 38:53
  Colin Russ: I think as w- we get into it more, there's probably gonna be two or three others that start to kinda tighten it down.
• 39:01
  Colin Russ: But, I mean, that's a baseline.
• 39:03
  Colin Russ: That's not, and it's not unreasonable either
• 39:05
  Matthew Chang: Yeah.
• 39:06
  Matthew Chang: the example that I've used, that I think helps people'cause they'll say,"Well, why not?
• 39:10
  Matthew Chang: The manufacturer, you know, is facing labor challenges." True.
• 39:14
  Matthew Chang: You know,"They, they need to lower their costs." True.
• 39:16
  Matthew Chang: You know, I say, yeah, I'm just not convinced humanoids fix all those things, you know, that all the systemic challenges they're facing.
• 39:23
  Matthew Chang: But let me make it real simple.
• 39:25
  Matthew Chang: Is Boeing going to allow one of the top humanoid companies to go in there and film the interior of Boeing facilities live while they try to make FAA-regulated aircraft?
• 39:38
  Colin Russ: Yep
• 39:40
  Matthew Chang: in a humanoid company will immediately clap back and say,"Companies like Boeing have pre-ordered a billion do- dollars worth of humanoids" or whatever.
• 39:48
  Matthew Chang: Okay, so they're gonna let the humanoid company take the training data, take the analytics out of there, and stick it on their Amazon web cloud or whatever environment they have to store this information for training.
• 40:01
  Matthew Chang: They're gonna let them do that?
• 40:02
  Matthew Chang: Well, yeah, of course, they have to.
• 40:05
  Matthew Chang: Okay, so as soon as they do that, and we'll, we'll pick on OpenAI right now'cause they're calling for, robotics engineers.
• 40:11
  Matthew Chang: So they're gonna send it to OpenAI, right?
• 40:13
  Matthew Chang: Let's say that's the backend partner that's gonna process all the data for the physical intelligence model.
• 40:19
  Matthew Chang: And then what is OpenAI gonna do?
• 40:22
  Matthew Chang: Well, naturally, they're gonna turn it into a product'cause that's a product company, and the product is now called Humanoids for Aviation Module, right?
• 40:31
  Matthew Chang: You can get the Humanoids for Aviation Module trained on Boeing datasets from real Boeing data factories, trained by real Boeing operators.
• 40:40
  Matthew Chang: And then who's the market to sell that to?
• 40:42
  Matthew Chang: We have to expand the TAM, right?
• 40:44
  Matthew Chang: It's all about the TAM, you know that.
• 40:45
  Matthew Chang: So it can't just be Boeing.
• 40:47
  Matthew Chang: Now it's, we gotta sell to Airbus, and of course, we'll sell to Airbus for a little bit cheaper'cause they're French and they negotiate hard, right?
• 40:55
  Matthew Chang: And then we gotta sell it to Bombardier, and we gotta sell it to Lockheed Martin.
• 40:59
  Matthew Chang: So Boeing has just sold their proprietary manufacturing process to every other aircraft manufacturer,
• 41:07
  Colin Russ: No way.
• 41:09
  Matthew Chang: If that doesn't sound plausible to you, then join-- you know, tell other people about it because I think folks generally just don't understand how factories work.
• 41:19
  Matthew Chang: to get into almost any high-end factory, you gotta sign an NDA just to get in the building.
• 41:25
  Matthew Chang: Most factories, you pull your phone out of your pocket to take a picture, one picture, you're probably getting escorted out, might be your last visit there ever.
• 41:34
  Matthew Chang: You know?
• 41:34
  Matthew Chang: these are highly secretive environments, and they're very nervous about their competition finding out what they're doing.
• 41:42
  Matthew Chang: So I don't see a way that people open up these lo- these manufacturing processes that were built over decades, you know, so that an AI
• 41:50
  Colin Russ: up- there, there's no upside.
• 41:51
  Colin Russ: No upside to the risk
• 41:53
  Matthew Chang: Yeah.
• 41:54
  Matthew Chang: So those underwriting criteria, you get no video data.
• 41:58
  Matthew Chang: I think that's almost a deal killer.
• 42:00
  Matthew Chang: let me get back to teleop,'cause there are a lot of non-proprietary, right?
• 42:04
  Matthew Chang: So your UPS, DHL, USPS, that kind of category.
• 42:10
  Matthew Chang: Okay, that's not as secretive, right?
• 42:13
  Matthew Chang: I could kind of see some adoption in that space.
• 42:16
  Matthew Chang: I could see some adoption in...
• 42:18
  Matthew Chang: I'm still betting on hospitality and retail.
• 42:20
  Matthew Chang: I think that's where it makes sense.
• 42:22
  Matthew Chang: You know, working at a hotel desk, it's already built for a human.
• 42:25
  Matthew Chang: You know, you might have two humanoids and one human working at the same time.
• 42:29
  Matthew Chang: it's not different than the quick checkout at the supermarket where you have both options.
• 42:35
  Matthew Chang: and you have a certain number of people that wanna use the humanoid.
• 42:37
  Matthew Chang: I'm just returning a key.
• 42:38
  Matthew Chang: I don't need to wait in line.
• 42:40
  Matthew Chang: And you'll have some that need, have a more complicated request, like,"I need to get a baby crib sent up to the room," and,"I ordered two adjoining rooms, but they're not adjoining." You know, whatever.
• 42:49
  Matthew Chang: So I can see that, but let's get to the teleop piece.
• 42:55
  Matthew Chang: Teleoperations of a vision-enabled robot is just creepy.
• 42:59
  Matthew Chang: It's just creepy.
• 43:01
  Matthew Chang: And, you know, we deploy a lot of autonomous robotics.
• 43:04
  Matthew Chang: We've deployed mixed fleets of autonomous robotics.
• 43:06
  Matthew Chang: I've deployed autonomous vehicles on the streets, deployed autonomous vehicles for the USDOT.
• 43:12
  Matthew Chang: Like, I have a pretty deep resume in autonomy.
• 43:14
  Matthew Chang: I have an autonomous vehicle.
• 43:15
  Matthew Chang: My Cybertruck drives me around.
• 43:19
  Matthew Chang: I don't understand why you need teleop.
• 43:21
  Matthew Chang: I can't get my head around it.
• 43:23
  Colin Russ: The one single instance of it that that comes to mind and makes-- that I think has a use case is effectively your da Vinci surgical machines.
• 43:34
  Colin Russ: That is as close as I can get to a teleop robotics situation where you could have, you know, the one of one surgeons for some obscure unique thing in one part of the world using the da Vinci surgical machine to do a surgery someplace else.
• 43:54
  Colin Russ: It's as close as I can get to kind of saying it's teleop, but it is not the same thing as a humanoid
• 44:01
  Matthew Chang: Yeah, that,
• 44:02
  Colin Russ: doing it.
• 44:02
  Colin Russ: it's a, it is different animal, but as close as I can get to seeing any link between them.
• 44:08
  Matthew Chang: Well, I love that use case, and I wouldn't call that tele-op.
• 44:11
  Matthew Chang: I would just call that remote operations, right?
• 44:14
  Colin Russ: Even better.
• 44:14
  Colin Russ: Even better then.
• 44:15
  Colin Russ: Great
• 44:15
  Matthew Chang: one-to-one-- Like, the da Vinci wasn't designed to be autonomous.
• 44:19
  Matthew Chang: It was designed to be operated by a surgeon, and all you've done is say the surgeon's not in the room anymore.
• 44:24
  Matthew Chang: The surgeon is 6,000 miles away because we got-- we don't have time to provide this life-saving intervention.
• 44:30
  Matthew Chang: that is a fantastic use case that's very similar to the Reaper drones that the military flies, right?
• 44:35
  Matthew Chang: They're unmanned aircraft.
• 44:38
  Matthew Chang: In this case, you have an unmanned or, you know, undoctored, surgery machine that's being run by a remote doctor somewhere else.
• 44:45
  Matthew Chang: The tele-op function infers that the
• 44:49
  Colin Russ: a, it's a robot call center opportunity
• 44:51
  Matthew Chang: That's right.
• 44:52
  Matthew Chang: It infers that by having humans around to run the robots or to intervene when they get stuck, we will eventually get good enough to where the robots can run themselves.
• 45:02
  Matthew Chang: To which I would say, well, why don't you just get the robots good enough to run themselves before you deploy them, right?
• 45:09
  Matthew Chang: Why do they need to be in my house trying to clean the countertop while you're training it?
• 45:12
  Matthew Chang: You can't simulate a countertop somewhere
• 45:15
  Colin Russ: it gets into labor arbitrage.
• 45:17
  Colin Russ: you're not trying to solve the problem, you're just trying to cut costs
• 45:21
  Matthew Chang: Yeah.
• 45:21
  Matthew Chang: It's-- Yeah.
• 45:22
  Matthew Chang: In my view, it you couldn't make your autonomous package work, so you backfilled it with human capital.
• 45:29
  Matthew Chang: But then you couldn't afford human capital in America, so you offshored it to a foreign call center, you know.
• 45:34
  Matthew Chang: And that foreign call center has a lot of information about us, right?
• 45:39
  Matthew Chang: They're seeing if it's in your house, they're seeing the inside of your home, they're listening.
• 45:43
  Matthew Chang: You know, it's like I know we're used to Alexa and Siri kind of listening.
• 45:47
  Matthew Chang: Probably just turned my phone on when I said that.
• 45:49
  Matthew Chang: But, you know, this is like the next level of kind of spyware.
• 45:54
  Matthew Chang: And I think, um, I think I would just like to see companies get their pa- their robotics to work before they deploy.
• 46:04
  Matthew Chang: Like, if you're gonna go telly op, you need to go build your lab somewhere before I even see it.
• 46:08
  Matthew Chang: Don't put it in my house or in my workplace, and you've got somebody else driving a joystick or wearing an Apple Vision Pro, you know, controlling the hands
• 46:17
  Colin Russ: it gets really creepy really quickly.
• 46:21
  Colin Russ: I'm with you on that.
• 46:23
  Colin Russ: Well, I think, hopefully we, we can avoid any of the, tele-op, snooping in the near, near future.
• 46:29
  Colin Russ: I'm, I think generally looking at kind of the response, the broad responses to people who either accuse people of doing the tele-op or just the general sentiment towards it, I'm cautiously optimistic that's not going to be, the solution or the straight path approach for most of these companies.
• 46:46
  Colin Russ: But, you know, knock
• 46:47
  Matthew Chang: Co- Colin, congrats on our longest episode ever, and it's no wonder that wandering into robotics and humanoids got us there
• 46:54
  Colin Russ: No, no question about it.
• 46:56
  Colin Russ: Well, Matt, great episode.
• 46:58
  Colin Russ: we could go on for a whole lot longer, and we probably will take some offshoots from this.
• 47:02
  Colin Russ: But to everybody listening, thank you for tuning in to the"Reforge" podcast.
• 47:06
  Colin Russ: Like today, whether we're talking robots on the factory floor, policy that moves the needle, or the next generation of makers in America, our goal is always to prove that reindustrialization in America isn't just possible, it's happening right now, and it is pivotal.
• 47:21
  Colin Russ: If you enjoyed the episode, make sure you subscribe and follow@thereforgepod everywhere, and check out thereforgepodcast.com for more.
• 47:29
  Colin Russ: Until next time, keep forging the future.
• 47:36
  Thanks for tuning in to the Reforged podcast.
• 47:39
  Make sure to like, subscribe and follow us on all of our social media@thereforgedpod.
• 47:44
  We will see you next time