Speaker 1- Welcome to Figure. We make humanoid robots here. We design 'em, we build 'em, we test 'em, all here. This is the secret room that nobody's allowed to come in. I'm gonna show you every robot we've ever built. So here we lost like a left knee, and you can see the robot's kinda like hobbling on the left leg. So we have a robot doing burpees here. This is where we manufacture Figure 3 robots. - Wow. - And this was the first car in the world built by a humanoid robot that we're aware of. We have a robot here that is designed to tidy the house. These robots are running purely autonomously from a onboard AI policy called Helix. Give it a push.
Speaker 2- Oh gosh. Okay. I feel bad.
Speaker 3Welcome.
Speaker 2- Hi. - How you doing?
Speaker 3- Good, how are you?
Speaker 2- Good to see ya. - I'm excited.
Speaker 3- Yeah. Welcome to Figure.
Speaker 2- Thanks. So where are we right now? What is this? This is the headquarters?
Speaker 1- This is a robot campus.
Speaker 2- Robot campus?
Speaker 1- Robot campus. We make robots here.
Speaker 2- Humanoid robots?
Speaker 1- We make humanoid robots here. - Okay. - We design 'em, we build 'em, we test 'em. All here.
Speaker 2- So for people who don't know what a humanoid robot, you will see in a second, 'cause this is kinda freaky, but can you explain what they are?
Speaker 1- Yeah. Our goal is to build advanced AI that we can put into a general purpose humanoid body. A humanoid's basically just like a robot with a human form. So we have arms, hands, head, feet, legs. We can basically do everything a human can in the world with one piece of hardware. So yeah, our goal is to be able to go out and basically design and ship humanoid robots in the world that can do everything from housework, dishes, laundry, to manufacturing, healthcare, just basically as much things in the world as possible that we can go out and ship robots to.
Speaker 2- And we see this on the screen.
Speaker 1- Oh yeah, that's our figure.
Speaker 2- That's your hype machine screen?
Speaker 1- Yeah, exactly. That's our latest generation robot, figure three, doing a little like a concierge job.
Speaker 2- Cool. Okay, so what are we gonna see today? What's the plan?
Speaker 1- Okay, come on in. I wanna show you some of my robots. - Okay. - I think first here is we have some robots that we basically have constantly running around the office 24/7, talking to humans, greeting people, just basically doing useful work. These robots basically can run fully autonomously without any humans, and they can automatically dock themselves and charge. And then once they're fully charged, basically come off and be able to do useful work. So these robots that are docking here are charging through the feet. So we have a wireless charging stand, like similar to how you would charge a phone, like inductively. - Okay. - The robots can charge through their feet at two kilowatts. So basically the battery lasts about four to five hours, then we can charge basically for an hour. And go back and do work again. So the robots, we don't need to do anything. We don't need to plug them in. They can basically auto charge themselves and just do 24/7 operations.
Speaker 2- Is the role for this one right here, just for the docking exercise? Do these ones roam around?
Speaker 1- These ones roam around. - Okay. - Yeah, yeah. They're just docking here just to charge, and then they'll be out doing useful work all day.
Speaker 2- Okay. And I saw these ones as well. Is one of these special?
Speaker 1- Yeah, so this robot right here, the one, the American flag was actually at the White House last week.
Speaker 2- How did that happen?
Speaker 1- We got a call asking to be basically the first humanoid group basically ever to put robots in the White House. And so last week we basically had the, you know, first humanoid robots in history there doing stuff, talking, greeting folks. We were basically at a very special event with the first lady and it went really well.
Speaker 2- It's exciting.
Speaker 1- Yeah. - So this is our corporate headquarters. We have four buildings on campus. So here we do a lot of basic engineering design work.
Speaker 2- How many people work out here?
Speaker 1- We have about 500, a little over 500 people.
Speaker 2- How many are in the total company?
Speaker 1- Five, oh, sorry. There's about 250, 300 people here and we have 500 in the company.
Speaker 4- Oh, wow.
Speaker 1- Yeah, yeah. I would say most all of it's engineering and then we've been growing out manufacturing, supply chain, some of the areas to basically, how do we make more robots pretty aggressively?
Speaker 5And how many robots do you have here? How many are there?
Speaker 1- A few hundred right now.
Speaker 5- Okay, are they outpacing the humans or?
Speaker 1- My goal for this building is I want more robots, humanoid robots and humans walking around. And they don't necessarily need to walk. They could be sitting or talking. Yeah. So basically we have, we basically do a lot of basic hardware and software validation testing here.
Speaker 4- Okay.
Speaker 1- So like testing for burn-in, durability, basically like any new software or hardware gets validated through this facility.
Speaker 2- Mm-hmm.
Speaker 1- Yeah. So you have robots doing all kinds of crazy stuff here.
Speaker 2- That is definitely some yoga.
Speaker 1- Robot getting on the ground and getting back up again. We're basically trying to stress test the robots to try to find any potential failures. - Okay. - Before like any new hardware or software could get released. So if we have a new camera, a new type of like say structure or anything else, we'll test it here before it goes out.
Speaker 2- How many potential body movements can they do?
Speaker 1- Okay. So this is kind of a crazy. So the robots basically made up about 40 like motors. - Okay. - Every motor can spin like 360 degrees, like all the way around. So the mathematically like how many states it could be in, like body positions. - Yeah. - Is 360 to the power of 40. - What? - Yeah. It's more body positions than atoms in the universe. Which is crazy.
Speaker 2- Potentially.
Speaker 1- Yeah, I've done the math. It's for sure. - You've done the math? - Yeah, it's for sure. - Okay. - So it's basically like the difficulty here is like, how do you control it? You can't write code to make this work. So all of our robots here run on a neural network we call Helix. It's a vision language action model we designed here internally to tell the robot what to go do, to stay balanced, like how to move its joints basically, from pixels, from camera space.
Speaker 2- And so that team that works on Helix is in the same building?
Speaker 1- They're in the same building.
Speaker 2- Okay. - Right here.
Speaker 1- Yeah, they're phenomenal. We basically have a large scale data collection efforts that's going on, and then we train our own models here internally, and then we test them all here as well. It's not just like for balancing and being able to have stability, which we need to have like human-like stability. It's for how do we know what to go do? How do we take in prompts from humans and say like vision from the cameras? And how do we output every single joint, including where the body's positioned and feet and hands to do stuff? So we have to be folding laundry, doing manufacturing and logistics that you'll see here later today. And we have to do that like a few hundred times a second from camera images. And on a neural network that runs on board the robot. So it's a really hard project. - Wow. - Yeah. So yeah, all the base here are running some sort of test for durability or reliability testing.
Speaker 2- Why do some have different suits?
Speaker 1- Oh, we outfit them. - You do? - They have, all the outfits are, are fabric, like a human, like clothes. - Yeah. - And they all have different clothes, which is cool 'cause we can like, you can accessorize the robots how you want it. Our clients can have different like outfits that show like the client logos and colors. - Workwear. - Yeah, workwear.
Speaker 2- This is a new level of merch.
Speaker 1- Yeah, it's a new level of merch. - Okay. - It's also nice because if like things get dirty or they rip or whatever, we can basically easily replace it without a technician. - Yeah. - Yeah. All the robots have a little zipper on the back. I'm sure you see. We can basically just unzip it and basically, take it off and put something new on. - Okay. - Yeah. Also, it's just really cool.
Speaker 2- It is pretty cool. Their shoes look like real sneakers.
Speaker 1- They're high tops.
Speaker 2- They're high tops.
Speaker 1- They're high top sneakers. And yeah, they look, don't they look awesome?
Speaker 2- Yeah, I mean, they look like human bodies.
Speaker 1- But getting to the lab to this level of like, like infrastructure to be able to run 'em like this every single day is like actually quite difficult. And then we need to be very diligent about when we find issues, like how to track 'em, how to do fall analysis really quickly, and then how to solve 'em. And then how to solve 'em across the global fleet, or basically our whole fleet, wherever they're at.
Speaker 2- How many are you in development testing all at the same time? Like what is the typical, is it, these bays are always active? How does it work?
Speaker 1- Yeah, we basically, so the goal of this lab is to basically do final, final checks for all software. It could be like, that could be embedded software. It could be like a neural network, a helix. It could be firmware on the robot. And then any new hardware changes we have. We need to make sure like those changes are bulletproof before they head out of here. 'Cause it's gonna cause a lot of problems if like we're trying to run a use case for logistics or a home, and the robots are messing up. We're not sure why it's messing up. That's not great for us. So basically here, we're basically doing a ton of testing. We have like test plans laid out every single morning. We're running those down. And when we see any potential falls, we have to go solve it. Then we have to retest those plans. So these robots run in here like all day, every single week. And we run 'em really hard. And the goal is like, the goal is we don't wanna be finding like failures upstream out of this lab. Yeah, so this is like the banner on here on system integration test is trust but verify. - Okay. - So yeah, you basically have to make sure we run down every potential thing that could go wrong before leaving here. This is like a, this is a hard thing. Like the robot has 40 plus moving joints. It's like a walking cell phone, self-driving car. All the supply chain is basically new. We've designed almost all of it. And so it's hard. It's hard to get the system to be really, really reliable. And it's not like if we lose power, like, you know, we're not like statically, stable so like if we lose power the robot falls so we can never lose power we can never lose comms and then we need to be able to balance at all times everywhere we're going even if we're moving the body like your pelvis and hips and everything are moving uh in relation to your like you know as it relates to your hand moving and things like this and head so it's a quite a difficult problem
Speaker 2and so that's why you dock them at 15 minutes or 15 we dock like yeah around 10 to 15 they'll go
Speaker 1to dock and then if we need another robot in we'll sub them in off the dock and you'll see here later in our like logistics and other use cases that need like constant like 24 7 uh tension the robot will undock right before the other robot needs to go like leave and the robot will then uh basically do a quick swap and like within like you know 30 seconds it's now doing work another robot will go in and dock and we'll just run that every four or five hours on repeat 24 7. what's the most
Speaker 2common error that they make we most is software at this point software yeah okay hardware's gotten
Speaker 1really robust i mean the hardware here is like is is great we just like basically it's basically a
Speaker 2software issue and then in terms of the hardware you're manufacturing also on campus um yeah we
Speaker 1manufacture here at baku next door okay yeah we're gonna show you that today okay yeah okay so we have a basically a robot doing burpees here we want to be able to safely for any event get down to the ground um and then we want to be able to safely get up okay uh it's like important in case we're like uh not sure what to do we could be on like a very low battery and we might need to be like safely get down um we could we could have like um and then when you then we want to be able to be able to get off off the ground really easily it's also quite hard yeah kind of a ton of like uh like range of motion in the legs and the hips
Speaker 2to be able to do this kind of maneuver yeah he's gonna need some knee pads is there a reason why the joints are hard and not you don't have soft tissue um yeah the most of the upper body torso
Speaker 1is all soft okay and then we have some uh soft foam underneath the legs and uh in arms right now obviously like the more um the more patty more soft i think is is great it just adds like a different level it has more volume in mass of the robot yeah yeah it makes the robot look bigger
Speaker 2basically some thick robots yeah cool okay all right let's go cool so how often do you come into the office every day and you check in on them like how do you every day they feel like your babies like do you feel like they're for sure a parasocial relationship with them they're like we've like made
Speaker 1these like they're little kids and we have to like get them to do useful things now um i think the good news is like we're at a point where the hardware has gotten like pretty like almost like like very robust yeah like the we can run them all day every day like uh we still we see we still see like hardware failures but it's very few and far between most of our problems now are like uh like as we think about this baby growing up or like software problems or ai problems how do we get the software incredibly stable and how do we get the neural networks to be able to actually do useful things 24 7 without failures like most of our failures today are uh kind of in software land um and speaking of software uh more it's uh one of our favorite things to do is like push robots around okay and so moritz here uh is one of our leads on the basically helix controls team and is gonna maybe you can give him a quick give her a quick 101 of uh kind of the s0 controller we have here and then uh we would love for you to push the robot as hard as you can yeah so i i think what we
Speaker 6did recently switch fully to rl i'm from a model based stack and ranch and this is that that we have all this variation that we can give to the robot when you train it in sim so all edge cases are now known to the controller and gets robusted so what this means for example before a model-based robot stack got it freaked out um you have this very robust uh to external services we push it around uh go ahead to convince yourself really i think this really nicely show
Speaker 2how robust our stack is give it a push oh gosh okay i feel bad that's a little harder okay so they don't know they like what do they do if they get attacked you know how like waymos get attacked people were attacking bird scooters yeah do they have a defense mechanism no defense mechanism none no it's not trained in them they don't see the internet and no harm to humans no
Speaker 1harm to humans no no uh they're here just to help okay yeah they're here to take a take a push too if you need to yeah she you want to get another one in sure yeah it feels really heavy yeah it's uh it's like 135 pounds but it's like it's it's actually really um it's got like you know human level like stability and um as march mentioned like we're we're learning that coverage in a simulator so the whole controller like learns uh how to stay stable like this in like synthetically like in its sim like basically like a video game and uh from there uh the robot learns how to stay balanced uh how to basically like uh how to basically not fall whenever there's certain certain forces and we basically can zero shot it onto this robot meaning like we can just put it right on load it to the computer and we can basically uh and get this level of performance in the controller
Speaker 2what are like again like what are the most common tweaks within that in the software like what are you you're obviously balancing a lot of physical issues there so
Speaker 1how do you tweak that in the models more it's how do we how do we uh how do we get the models
Speaker 6to be able to perform like this um basically i think we spend a lot of time thinking what are all the things that can happen to the robot in the real world and then make them happen in simulation i think that's the yeah we basically have like it's like a physics
Speaker 1simulator so it has like gravity has like friction coefficients uh we want to try to mirror like what forces we're about seeing here so we can run them in sim and if they can run in sim well what we've seen is we basically have a really great sim to real transfer so we can get it from a simulator that shows like you'll see it like in a simulator like video game it can like as it can like stay stable with these forces then we load it to the robot and we see the same in the real world and we have like a basically a very high transfer rate it's interesting i feel like um i feel like
Speaker 2it would be it would be interesting to hear your perspective or differentiation on your robot versus the other humanoid robots and like where this physicality and essentially like the behavioral mechanisms change is some might be more commercially focused these ones are clearly
Speaker 1as humanoid as i've seen yeah we're a pure play humanoid okay uh yeah no uh i think uh a few things uh one is like we need to get the hardware in a really good spot that can do like a lot of what humans can do you really want this um you kind of want this like iphone moment where my iphone has a bunch of different apps and if i wanted to learn something new i just download another app what is something for humanoids with the same humanoid to be able to do like dishes and laundry but also do some like package logistics and health care other stuff that's what humans can do right we're all fairly general purpose and uh so we want one set of hardware that we can amortize over like a lot of different use cases uh so the goal is get the hardware like robust enough to do like most things a human can in terms of range of motion they'll get down get off the ground be able to reach up high be able to reach inside a sink like all these different like things we need to do we also need to carry a decent amount of payloads and we operate at decently fast speeds so we've designed the hardware to be able to do that and then separately on the software on the ai side you really want to be able to design the neural nets so they can basically take a take a task uh and then reason through pixel space like take camera videos of like what's happening and then output what the body should be doing um the the you know the math we were doing before um on like how many states the robot could be in is just it's so high yeah um that like the problem just kind of runs away it's um it's like a curse of dimensionality it's just too hard so you can't solve it with like writing like like lines of code like attritionally like in robotics and uh even like three or four years ago here a figure like we would solve the same controller here in code we'd have like hundreds of thousands of lines of c plus plus to figure out how to solve this like inverted pendulum math of how to like like stay balanced here um and what we found is it just doesn't scale it doesn't work you can't like really um in your head as humans uh work across many different humans to code all this stuff into the robot that you think it could encounter in the world it's just too difficult uh so we transitioned only like purely to the neural networks uh with helix our neural network like uh model for those i know you mentioned this
Speaker 2a couple times but this will be like a general audience for those that don't know what a neural network is can you just explain that a little bit further yeah we basically use like an ai policy
Speaker 1that we've trained here with like data we trained like a transformer transformer policy to basically output uh a certain type of action space basically we trained an ai policy to do this work of what coda used to do and learn this uh so now uh now basically we we run inference on board the robot uh across the policy that we call helix it's a ai it's an ai model that we designed here internally and that policy is outputting what the robot should go do similar to how you'll talk to like an llm and you'll ask it what to go do uh it'll just it'll it'll do inference and output like basically like a next token prediction for words uh we do the same thing here but for uh like a physical physical physical humanoid and we'll output things like where to put the wrist head uh torso every joint will get an output from the neural network like what to go do and we'll do that anywhere between 50 and 100 like 200 milliseconds uh so basically like 50 to 200 times a second the neural network uh computer is then telling all the joints like what to go what to go do then every joint level our motor controllers are outputting torque on on where to send basically uh like where to position the motors right wow so yeah so like basically like you basically we like uh you can have like two paths here like two two you basically can code your way out of this and i think that's a full dead end or you can run uh you can ai first strategy in the market and that's what we do here at figure uh so i think you would do a lot of work and i think that's what we do here at figure uh so i think you would do a lot of work and i think that's what we do here at figure uh so i think you would do a lot of work at figure i think we probably have i think this is probably the best human or hardware in the world to do general purpose work and secondly all the work that we're doing is all like neural network based or ai based we don't code any of this work anymore uh so like you'll see some use cases today that we do both for the home and for uh in like like cases of a commercial market those are all run by our helix neural network which is hard like we have to we have to kind of tell the whole body how to reason over like camera frames and then how to like position itself fast and dynamic so you'll see in cases like logistics these packages are like moving they're moving while we're grabbing them because they're plastic we have to reason over all of that in real time and be able to do like you know if we like uh if things move in real time we have to basically a closed loop react to it that's like really hard we've been spending like the greater of like last two years trying to solve this problem and i think uh i think we probably have like some of the best ai policies
Speaker 2for humanoids on the planet today sorcery is brought to you by brex the financial stack trusted
Speaker 7by more than 30 000 companies including one in three venture-backed startups in the u.s nearly 40 of startups bail because they run out of cash brex is literally built to help founders avoid that unlike traditional banks that let your money sit idle chipping away at it with fees brex's designs help you spend smarter and move faster their all-in-one solution combines checking treasury and fdic protection into one powerful account you can send and receive money globally at lightning speeds get 20 times the standard fdic coverage through their partner banks and even high yield from day one with same day and even same hour look at it all it's a great way to build your business and it's a great way to access your funds anytime companies like scale ai doordash service titan hymns anthropic flexport robin hood and plaid trust and use brex start today at brex.com slash sorcery that's b-r-e-x.com slash sorcery turing is training the next generation of ai with tasks that require real expertise and real world judgment that's why companies like nvidia anthropic salesforce and gemini partner with turing turing builds realistic reinforcement learning environments and data systems based on real operational traces the kind of infrastructure frontier labs need to train super intelligence visit turing.com slash s-o-u-r-c-e-r-y another thing i want to show you
Speaker 1is um uh this is like um you know we just walked out of system integration and test the lab we're trying to find all these different failure cases yeah one of our failure cases is for a humanoid like we were balancing so if we lose power we're going to lose power we're going to lose power uh like the robot falls uh it's the same for like losing power or like losing a motor in the leg
Speaker 2do they need to be connected to wi-fi too uh we what are the what are all the integration these
Speaker 1these robots have 5g and wi-fi and bluetooth uh but we do not need to be connected to wi-fi to do work like these robots out here if they lost like internet connection or network they can basically continue to do work we run helix on board uh so they're actually loaded into gpus on board and memory and we run inference on board meaning if we um if we lose internet connection we can still do housework and let's say we lose internet connection we can still do housework and let's say we lose internet connection we can still do housework and let's say we lose internet connection like like humans are yeah i mean maybe i have a hard time like doing work when i like losing our connection but most humans can do most work um so uh so another thing that we uh we like we're working on solving that i'm like excited to show you here is um what if you lose like what if you lose communications with like all any of the 40 joints or what if we lose power and upper body's kind of fine if you lose a wrist or elbow like it's like you're not you're not going to fall at the very least um but falling is like a it's terrible event for us we don't ever want to fall we actually have an initiative internally called never fall it's like we never ever want to fall even uh and you know we will fall but we don't ever want to tolerate it um the hardest problem here uh for the controller is like what if we lose like a like a like an ankle a knee or hip what if you just like lost a knee like right um you know normally for human noise you just like you just literally fall over like you can't really balance we feel like lose a knee uh we've uh we've been working on a project we call uh vulcan here internally um that basically allows us to lose a single or even multiple joints and we can't really balance it we can't really lose a knee and we can't really balance it we can't really balance the joints and the legs and still not fall so what we're going to do here is more it's can you show her what would happen if we lost like a left knee so here we have um a view of like all the different joints on the robot green means like we have comms and power to them and the robots like uh can basically communicate and it's fine and red here will mean we'll lose certain communications with the robot so here we lost like a left knee and you can see the robot's kind of like hobbling on the left leg yeah so right now the knee is basically we lost we lost we lost power communication
Speaker 2is now to the knee the knee's locked locked got a locked knee so we velocity locked the knee and we can basically hobble around i thought it was going to be a little bit more dramatic honestly
Speaker 1it's not bad right it's not that bad it's unbelievable uh that we can even do this kind of work we're doing this also in inside of a reinforcement learned neural net controller so the same stuff that morris talked about earlier of us learning in simulation the robot learned in simulation how to move the body here uh to extend it lost different joints uh this is uh i watched the robot like a few minutes ago and i was like what's going on with this robot and i'm like oh A few weeks ago, we were doing work on this logistics use case. And it's like months ago, we'd lose a knee, the robot would just fall. Now, the robot loses a knee. It can either continue to do work or can just hobble off. It just limps. Yeah, it can basically ask its buddy in the back, say, I need another robot to come fill in. And the robot comes in and fills in and continues to do work. And we basically don't lose any time.
Speaker 2That's pretty impressive.
Speaker 1Yeah, it's really impressive. I think this is kudos to the controls team. I think we have one of the best controls team in the world. About a year ago, we made a strong pivot away from code into neural networks. And I think the team has probably shown, I think, some of the best neural networks for control in the world on humanoids.
Speaker 2What happens if it loses its knee while it's bent?
Speaker 1It'll basically do a velocity lock on that knee and should be able to hobble around. It obviously depends if we're in a full squat down. That might be really tough. It depends what state it's in. But in most cases... We can basically recover from this at this point and survive. You want to build a real time operating system that is basically like fault free.
Speaker 2So this team here is all focused on trying to figure out all of the potential faults and risks.
Speaker 1This team here is responsible for building our controller, which is responsible for how do we move the entire, like all the joints on the robot, keep balance and ultimately end up doing tasks.
Speaker 2How many different teams are in the company and like how do you portion out who works on what and who gets integrated?
Speaker 1So this is part of our AI team, it's one of our biggest teams here. We have like multiple different groups inside the AI division. We also have a team that does hardware. So they design motors, batteries, wire harness, structure, joints, kinematics, like basically a wide range of stuff. We have a platform software team that does all of our embedded software, firmware. We also have electronics team that basically builds like all the PCBs and electronics work that we do here internally. We design almost, there's over 100 PCBs that we design here on the robot that we do on that team. Things like our motor controllers, all this different type of work. We have a system integration testing that you saw today that helps basically make sure we ship like really good robots out to the world. We also have a team that does like basically design, like industrial design, which we're going to show you towards the end of this tour. How do we design something that's like, how do we design something that's like people really want? And I think we have a pretty high design standard here of designing something that is a really delightful piece of technology. We also have a teams that also work on system design. And thermals, the robot produces a lot of heat while it's moving around and how do we get that heat out of the robot and how do we design for it here? We also have fabrication teams that like make fabrication prototypes. We have a BACU team for manufacturing, supply chain team, facilities, and then we have like all of our business operations side. So maybe like a dozen teams here internally that are needed to do this. It's basically the same teams that you need to build robots, so like any kind of robots. When I, you know, at Archer, when I was building an aircraft, we had like, it's like a flying robot, so you have like the same type of stuff. You have like electric motors, batteries, control software, embedded systems, and sensors. So we have basically teams around all that, and then obviously on the AI side is a big focus here internally. Do you want to see some stuff in the home? Yeah, let's go. All right, let's go. Thanks Moritz.
Speaker 2Thank you. All right, so. I was hoping it would be more dramatic, but that's okay. Losing me? Yeah.
Speaker 1Yeah. It's like, honestly, it's a really hard engineering problem that we're so proud of internally. We have a large initiative, like one of the teams that we have here is like a never fall team, and it's a team that basically predicts any potential faults in design around them. This is a case where like, I think you're always going to have a period of time where you're going to lose a lower body, like motor actuator, then how do we survive this? Right. Okay. Outside of stuff we're doing on the commercial market, one of our big focuses here is how do we ship a general product? We ship a general purpose robot to do things like in the home. So, all right, come over here. We have a robot here that is designed to tidy the house. So clean up any, clean up the table, like basically like put away all the different cups, clean the toys up, tidy the couch. Like, you know, things, my house is kind of careless.
Speaker 2Oh, it's spraying and cleaning the table.
Speaker 1Oh, yeah. Don't you want, we all need this. Yeah. Yeah. Yeah.
Speaker 2Yeah. Yeah.
Speaker 1It was a little sassy. It was. So what's cool here is the robot is running an onboard Helix 2 neural network to be able to do all this work, to be able to do all this cleaning. So it's basically just taking a prompt, which is like clean the living room, and it's basically reasoning through what to do from its cameras. And it's ultimately telling the whole body what to do from a neural network, from an AI policy.
Speaker 2How many hours has this been tracked? How many hours has this been trained on? Doesn't it take millions of hours to train?
Speaker 1We've probably had in Helix, like the, like the biggest training we've ever had, like like in like kind of base pre and mid training uh helix models that we're working on now have um maybe like maybe like a little under a million hours of total data um and then uh we also do uh from mid training and pre-training we do post training uh here which probably is uh i would say low like basically thousands of hours okay that are running here uh the goal for us is design a single kind of neural network uh platform that can basically do things like tidy tidiest living room
Speaker 2but also do things like logistics and this isn't tele-operated this is not tele-operated there are rumors that these are tele-operated for sure not teleop you don't have a secret room back here
Speaker 1who's in there no secret room these robots are running purely autonomously from on board ai policy
Speaker 2called helix running on board the robot in the torso and so this robot's job is to clean this room up and does this robot do this over and over again each day or do they take turns like how do you any robot in the fleet can do any
Speaker 1of this work here okay so it all connects yeah it all connects we run a single uh neural network that can basically run it's the goal it's the reason why human is so great it's like uh this same humanoid can go over and do like logistics and healthcare and manufacturing uh or do dishes like just like we can right um so we like our our uh our platform here is to basically run a single neural network that we call helix on board the robot that can multitask between different things
Speaker 2when there is in people's homes will that data still be trained and will it stay local or will it be spread
Speaker 1throughout the network yeah we basically uh uh we need data like basically the biggest blocker for us now of going from where we're at today to like like large-scale deployment is data we need like an enormous amount of data uh we need to pull a lot of our resources further into pre-training for the helix team and we just need a lot of diverse really high quality data across the world this means like data in the home means maybe data more in the commercial market so we have like two efforts going on here one is like basically a large-scale data collection effort that we're doing now at the company and two is when we deploy robots we do want to be collecting data and we do want to be training on that and we do want to be sending that out up training on it into a central training training jobs and then we want to be software updating the robots with the latest
Speaker 2neural network uh weights does it get anonymized how do you deal with the privacy oh yeah we like
Speaker 1fully want to anonymize all of it uh there's a lot of data we really don't care about most of the data what do we uh what's the robot from a state perspective like uh like seen and how do we use that to basically train the robot to uh be more like to generalize better in the future at
Speaker 2those different areas are you sticking mostly to the us now because if you go to europe there's
Speaker 1obviously most of all a lot of most of all of our work today is in the us okay yeah we do want to be
Speaker 2global though yeah europe's a little tricky yeah yeah what are your plans like how do you skirt
Speaker 1around their data privacy uh you basically we got to play by the rules in europe yeah i think um our hypothesis here is that we're we're missing a lot of data we're missing a lot of data we're missing a certain set of data that we're like we're collecting now that will allow the robot to generalize in almost every condition it sees like we're kind of going off and doing the same things every day we're doing dishes and laundry and tidying the home like the same stuff we're seeing we're kind of doing the same movements like grabbing some off the ground putting away or pulling a like the dishwasher open like it's kind of the same stuff uh our hypothesis now is that we will see enormous amount of positive transfer from the data collection efforts we're doing now into pre-training across like basically any environment in the world um i mean that'll be like somewhat um you know we'll we'll approach that somewhat like um it'll take i would say over time a large amount of data to find every out of distribution like basically be able to do everything possible in the world but we think there's a path to do this and what is the price
Speaker 2point differentiation for the in-home robot versus commercial uh the in-home like we we have
Speaker 1we're not selling right now to the home we want we want to sell here uh like in the near term and we want to sell the robot like for like hundreds of dollars a month as like somewhere like a car lease um maybe like four or five six hundred bucks a month
Speaker 2how are you thinking of deploying them in homes like do you have to see if people have enough room like in new york city i can't imagine these little apartments uh they're just like it takes a dock
Speaker 1it's like two feet by two feet uh you can plug it in a wall outlet it'll go to a stock and charge and then throughout the day it'll just go off and do work whatever you want to want to do like for me i wanted to do like the laundry probably almost every day dishes every day and tidy the house multiple times a day
Speaker 2do they have a distinctive diet what do they eat they eat nothing they're key they just work 24 7. just constantly intermittent fasting they're in this like they're
Speaker 1in this like uh they're in this like purgatory state of just working 24 7 for us their lives wow yeah fun yeah okay i want to next go show you how we make them okay over at baku yeah great let's go it's uh you know like kind of like hq but baku it's the hq for bots oh it's bot q yeah bot q i
Speaker 2thought you were saying baku
Speaker 1isn't that bocce no anyways this is robot q this is robot quarters um one thing that's really cool is on the way is uh we we we uh we work at bmw and last year we had we deployed uh for six months robots on the uh basically body shop factory line to build cars yeah uh and this was the first build this entire car no not the entire thing but we helped build this car okay uh this is an x3 uh that we helped basically uh like we basically helped the robot helped assemble it and this was the first car in the world built by a humanoid robot that we're aware of and i straight off the assembly line straight off assembly i actually i bought the first four okay we have three here on campus and i have one at my house and uh yeah it's like a collector's item now that's exciting yeah it's pretty pretty interesting all right um so we have uh this is kind of our campus here we have four buildings um we're gonna go through our manufacturing site which is bot q and then we also have a site uh up here that i'll show you uh called the grid okay and the goal of that facility is to run robots like just like we would at our client sites could be in the home and also in the commercial side and 24 7 operations why is it called the grid um it's a kind of a nod to like a sci-fi movie uh and uh i don't know you have a lot of inspiration from sci-fi movies so what's your
Speaker 2i don't know you have a lot of inspiration from sci-fi movies a sci-fi geek it's a total sci-fi
Speaker 1geek i've seen every sci-fi movie what's your favorite um probably contact jodie foster the alien thing yeah it's kind of embarrassing to say but why uh i don't know i just the contact's amazing if you like it don't be embarrassed yeah um but like i'll just i'll watch i'm a sci-fi junkie i'll watch any sci-fi um yeah so uh so that's the grid up here um we'll run robots in that facility 24 7 and the goal of that is uh last line of defense before we send out any code to our customers so you don't want robots uh you know you don't want you don't want like robots having any problems we want to run them close enough to like heavy operations that we would see out in the real world and so we have a whole facility dedicated to basically running robots as hard as possible in 24 7. we we run on holidays weekends two to three in the morning they just run all day every day have any escaped we run on holidays weekends two to three in the morning they just run all day every day have any escaped we run on holidays weekends two to three in the morning they just run all day every day have any escaped no we've like we've had one almost escaped really no nothing's nothing's escaped do you do
Speaker 2geofence them within properties like yeah we do that when you set them up we we track them obviously
Speaker 1this is like for us right now it's like a these are like high ip uh like very complicated hardware we don't want to get stolen or out the wrong hands so uh so we track it has that happened before are people stealing your ip um we do a lot of work on security internally here um so we haven't had any known ip thefts at the company okay yeah all right welcome to baku all right now oh my god now welcome to baku um so this is where we uh this is where we manufacture uh figure three robots wow um so we uh we do everything from build heads batteries legs arms fingers thumbs hands uh and we basically do all testing here before we box the robots out or they walk next door what does the box look like the box yeah um we'll show you we have one over there in a minute really um and right now we basically if we need them at headquarters or the grid they basically just walk over where do you store them um at the at the office or client sites yeah on on the docks they basically dock at night time or whenever they're not needed um okay we're gonna show you some of the manufacturing lines so we start first uh with basically head and uh and battery uh and we do some like electronics uh like basically quality and eol checking uh eol is end of line so we make sure we want every single subsystem to go through a bit pretty like pretty crazy test here's our here's our headline a rack of heads uh rack of heads so um here here's a head our heads have um basically uh bluetooth like wi-fi 5g they have camera systems on on board uh we have lights uh we have thermal systems uh we have an imu so basically the head is like basically a lot of sensors in here um the heads go through a pretty uh like a rigorous test uh here um uh that we've designed uh internally for end of line testing so heads here go through uh well first is we basically are flashing software here under the head for the first time uh all the firmware it's going through a calibration process for the cameras uh and then we're basically making sure their head is in a nominal state to basically put onto the robot so does it getting are we getting signals out of it uh does it have any does it have any issues at all or any errors if it does we'll try to triage it if it doesn't we'll end up putting it on the robot this is like when it's first born yeah it's like uh it's kind of just like just raw hardware and it goes in here and it comes out with software and comes out with all the checks that we can use it with
Speaker 2Wow. Yeah. All right. How often do you walk through every part of the campus?
Speaker 1Every day. Every day? Every day. So here is our battery line. So we have battery cells that come in, and then we basically do cell testing and basically voltage balancing. So here we're basically checking every single cell against the data sheet. We're also checking voltage, and we're balancing out the packs. So if there's any kind of voltage differential, we're basically making sure that all the packs basically have somewhat of a somewhat balanced voltage.
Speaker 2Where do you get this machine from?
Speaker 1We designed this machine.
Speaker 2Oh.
Speaker 1Yeah. This was custom designed for figure here, for battery. And then we go through a process for potting, like wire bonding, and there's some polyurethane we put inside the battery pack for thermal runaway. And then at the end, pops out basically a pretty heavy 2.25 kilowatt hour battery pack. Yeah. I don't know if you want to try it. It's really quite heavy.
Speaker 2Oh.
Speaker 1Yeah.
Speaker 2No, I don't think it's... Wait, I want to try.
Speaker 1You got it? Okay.
Speaker 2Oh, no. That is actually really heavy.
Speaker 1Yeah. So this battery will basically go right into the torso. It's one of the heaviest components we have. Yeah.
Speaker 2How do you... I mean, I know all of this is stabilized, but is it better that it's one piece in the torso versus distributed across the board? Yeah.
Speaker 1It's way better if it's one piece. The battery pack has... Just even for safety, we have a lot of thermal runaway properties inside the pack that we've designed here internally to make sure... In the worst case, you basically want to say like, okay, if a cell or a battery cell is going to thermal runaway, you never want that to ever propagate outside the pack. So you want it to contain it to the battery system itself. So we have basically a structural system and also basically a thermal runaway venting process that we've designed internally to basically allow for the battery to be extremely safe. The requirement is you want no flame to ever exit the pack. You don't want a robot on fire or something like that out in the world. So we've designed the
Speaker 2right safety systems. How many of them?
Speaker 1We've never had a robot ever have... Catch fire?
Speaker 2No. No.
Speaker 1And then all of our figure threes are designed in a way that basically will prevent the robot from ever catching fire. Okay. So that actually was a pretty crazy hard engineering feat that we designed here internally. Also, the pack is structural, take loads. So in case we fall, even on like sharp objects or corners and things like this, like we can never propagate inside the pack, the cells itself, meaning you don't want anything to kind of like kind of like send the battery cells itself into thermal runaway conditions. Have you had any supply chain risks,
Speaker 2whether it's with China or other countries? Is that why you do everything here?
Speaker 1We do most of the manufacturing here because the product is so new and it needs to be really controlled. And we also think about IP as a really important here. We don't want any of the technologies to be stolen. Yeah. And this is just hard too. Like we'd be able to put this thing together through a basically brand new supply chain that we had to design and get it in and make it work. It's like, it's non-trivial. Like you see how much testing we do at headquarters for all this work, testing we do at the grid. We'll do a ton of testing. We'll show you here today. It's enormous. And the product is, we're kind of like early in the humanoid kind of like chapter book. So like cars have been around for over a century. We kind of...
Speaker 2The company's been around for four years, no?
Speaker 1Yeah. Not even four years yet. Yeah. That's... And then humanoids are like really early in that whole process.
Speaker 2How did you... We'll talk about this more in the long form, but like in terms of getting this up to scale so fast, like you have now created a humanoid robot. This is one of the most complex robot, like, I don't know, engineering problems ever. So like, how did you get up to speed so fast?
Speaker 1My company before this designed like flying robots at Archer. And it's got the same properties. We have a battery pack, but instead of like a two kilowatt hours, it's 160 kilowatt hours and it's distributed. We have electric motors, we have control software, we have embedded systems and sensors. That's a robot. Archer's aircrafts, like my aircrafts there at like say midnight are like highly over-actuated. All the propellers have variable pitch. The front leading edge actuators tilt 90 degrees. You have the flaps on the wings and tail all move. You have like basically 24 degrees of freedom. You have about over a little over 40 here on the robot. So... So in some ways, like similar enough systems here. And then, you know, when I started FIGURE, we were like, we have a very crisp and clear vision for the, how to think about the product and engineering roadmap. And we just like went like, went really hard building a team to 40 and putting the right resources in place for us to design stuff really fast. We had the, we'll show you here next visit, but we have our FIGURE 1 robot. It's kind of gnarly. It's got wires everywhere. It's our first generation. We had that walking before we were a year old. And we think it's probably one of the fastest times in human history. So it was just like a, you know, we were like laser focused, pedal to the metal, trying to get this thing to work.
Speaker 4Yeah.
Speaker 1Yeah. Okay. Let me show you some more stuff. So we have, we have a bunch of different lines here that helps build pelvises, install battery, compute arms, legs. Here we're installing the lower leg.
Speaker 2Is there a reason there are humans installing the leg?
Speaker 1Well, at some point we will have robots doing all of this work.
Speaker 2Don't let them hear that.
Speaker 1And we're putting more and more automation in the lines now. We will be shipping robots, our humanoid robots into the production lines here this year. And yeah, now we're doing like the lower leg assembly for this robot. And at some point today it will go through some testing. We'll show you in a minute and we'll basically walk over to headquarters and start basically helping us either do like AI development or doing use case testing for our customers. Pretty cool, huh?
Speaker 2It's pretty cool. Yeah. Yeah. It's so crazy seeing them get assembled. Are you worried they're going to become sentient?
Speaker 1Um, I, I think we'll, I think it will be okay. Um, these things will get really smart. Like they're, they're able to like do what humans can physically. And I think the neural network technologies we're designing are, we're trying to give like human common sense to all the robots. So in some way, uh, I think we'll get to at, or even beyond human level intelligence in these systems. It actually might be the case that we, we get to artificial general intelligence first in these embodiments. Really? Yeah. Why? Because we're able to like this interaction data of like touching the world, um, and seeing what happens through trial and error is like, um, it basically, uh, most of human intelligence, uh, is built this way. And I think this is the last missing piece to get the true AGI is this like real world interaction. Yeah. Yeah. Yeah. Yeah. Yeah. Yeah. It's a real world interaction with our environments. Okay. Come with me over here. I'm gonna show you some of the testing we work. So the robots are basically built, uh, starting with the pelvis, uh, torso, head, arms, uh, legs, and hands. And then we basically want to, we want to go through like a, basically a very, uh, strenuous testing process to make sure everything is like working nominally before we send it out. So we don't want any like loose cables or bad parts or bad communication. So we send the robot through what we call a final EOL test or end of line test. This is where the robots go through all the final checks and they basically go through also a burn-in on, on, on these lines. What do you mean by burn-in? They'll run for several hours and we'll basically make sure there's nothing that basically, uh, no, no issues pop up over those few hours before we, before we send the, send the robots out. So, uh, so here we have, um, these bays that are running a combination of burn-in testing and, uh, and end of line checks that we've designed here internally. So we basically go through a process for the robots, basically self, like trying to understand itself, like do, is anything seem wrong? If it is, we will, uh, we'll flag it. If it's not, basically we go through a process where we basically do like a, basically a bunch of checks and burn-in, make sure the robot's in good, good condition. Um, if they pass here, we basically, uh, we'll, we'll walk over to headquarters and if they fail here, we need to go fix it and understand why. Why do these ones have vests on them? Um, when the robots are getting brought up, we basically, uh, we, I think we hold it through a gantry system on the back and they have their vests on for that system. If you see like the robots are like just been born and they're waking up, they're saying, you know, they look at it, they look at their hands, they start calibrating itself visually and, um, trying to make sure everything's in a healthy state.
Speaker 2Of this campus, what is your favorite part? I think Baku is one of my favorite parts here,
Speaker 1like being able to build robots. Um, in March we had record, um, you know, we, we, we had record manufacturing. We, we, we, we made more robots in March than we had ever in our entire lifetime combined. Um, it's just cool to see us being able to do this and then get them out the door. Um, so I would say, um, this is probably one of my favorite places in the, in the campus. I think maybe my other favorite place that will show, um, some point here is like the robots doing really useful work 24/7, um, on our either commercial customers or in the home. Uh, that stuff is just amazing. Like, cause what we're here to do is we're here to like basically build like, uh, human little robots. Um, we're here to like basically build like, uh, human little human-like intelligence in the world. And to see robots working 24/7, being able to do things like humans can is like so special. It's like such a hard thing to do and be able to see us doing it, like, uh, at these levels of reliability is like, it's awesome. Um, so I think it was probably a
Speaker 2couple of my favorite places on campus. If you didn't have your job, what job would you want?
Speaker 1Here? Yeah. I think, um, I think there's a few things I really like. I like the engineering design process of how do you think about clean sheet and proving this system? To be like more reliable, cheaper, lower in mass and, um, overall a better functioning robot that can do more of what humans can with less complexity. I think that, that job, uh, across the hardware engineering and software engineering design org is like, I spend a lot of time in that leading engineering here. And it's just a really fun and I think very hard problem. Every choice you make to try to make the robot better for thermals or lower mass or lower weight makes something else on the other side worse. If you're trying to make the robot lighter, it means it probably can't hold as much weight then. If you can't hold as much weight, the customers are like, well, if you can't carry 30-pound boxes around, I can't use you on this assembly line. So I think there's a lot of interesting and very hard problems to solve there. I think second is, how do we get neural networks to run on robots and generalize at scale on the Helix team? And that really is, at this point, a data and generalization issue. And that's a really hard, fun problem. Manufacturing is another one of, how do we manufacture robots at scale? How do we continue to get robots in the manufacturing process to build themselves? And how do we get them off the lines into the real world as fast as possible? At some point here, it'll just be full lights out manufacturing. We'll have robots only building robots and sending out to the world. Robots will be getting into boxes themselves. Other robots will be boxing them up. And we'll be shipping them out to customers.
Speaker 2Sounds a bit sentient.
Speaker 1It's a bit sentient, yeah. And so that's another area where it's extremely interesting. I think the last piece is, we have a whole commercial operations team. And we want to get robots out to the world at scale and make them really useful. We did this with BMW last year. And we're doing it with more customers this year for figure three. And it's a really cool problem because it's really hard. Robots need to get in the environment safely. They basically can almost never fault. And when they do fault, they need to understand that and self-correct. And then we need to be able to do useful human work at human performance. So our comparison is, what does a human do today in terms of speeds and accuracies and then reliability? So that's a hard bar to hit. So those are all-- kind of like gigantic problems to go solve that I think if I was here, I'd want to spend all my time on those.
Speaker 2Cool.
Speaker 1Yeah. Awesome. OK. Well, you want to see the design studio?
Speaker 2Yeah, let's go.
Speaker 3Let's go. What do you think of this?
Speaker 2It's really fun. It's really fun. I have the coolest job because I just get to visit people's factories all day.
Speaker 4Yeah.
Speaker 2And everybody's building something different. I was at Applied Intuition earlier this week. I was then at Skydio. I've been-- I've been to Archer. I've been to Anderle. We have just been looking through every door.
Speaker 1That's awesome. Yeah.
Speaker 2VCX by Fundrise, the public ticker for private tech, allowing investors of all sizes to invest in venture capital. View the portfolio at getvcx.com.
Speaker 7That's getvcx.com. Some of you may not have heard this yet. But our sponsor, Public, just launched something called Generated Assets.
Speaker 2And it brings AI into investing in a way I've honestly never seen before. Here's how it works. You type in an idea like AI-powered supply chain companies with positive free cash flow or defense tech
Speaker 7companies growing revenue over 25% year over year. Public's AI then dispatches a swarm of agents that scan every single US stock, evaluates them, and instantly builds a custom index around your thesis. What really stands out is how clearly it explains why each stock is included. And before you even think about it, you can even backtest your idea against the S&P 500. So you're making decisions with real context, not just guessing. And beyond generated assets, Public lets you invest in stocks, bonds, options, crypto, all in one place. They'll even give you an uncapped 1% match when you transfer your investments over from another platform. If you want to build a portfolio that actually reflects your thesis, visit public.com/sorcery. Paid for by Public Investing. Full disclosures in the description. Enterprise AI runs on Merge. The AI infra platform for integrations, agent tooling, and model orchestration. So your teams ship product, not plumbing. Mistrawl, Dropbox, and Drada already trust Merge in production. Start building at merge.dev. Founders scale faster on Deal. Set up payroll for any country in minutes. Hire anyone anywhere. Get visas handled fast and get back to building. Visit deal.com/sorcery. That's D-E-E-L.com/sorcery.
Speaker 1And if you're new to the show, make sure you hit the subscribe button and hit the bell button so you get notified every time we upload a new video. And if you're new to the show, make sure you hit the subscribe button and hit the bell button so you don't miss out on any of our latest updates. And don't forget to hit the bell button if you want to get notified every time we upload a new video. And if you're new to the show, make sure you hit the subscribe button if you want to get notified every time we upload a new video. And if you're new to the show, make sure you hit the subscribe
Speaker 2button if you want to get notified every time we upload a new video. And if you're new to the show, make sure you hit the subscribe
Speaker 1button if you want to get notified every time we upload a new video. no like uh we're like we like it looks like we're uh maybe we're manufacturing in here and running robots so it's it's uh no it's fine for us and like you it has like like uh some industrial grit here you know i mean it feels uh like we're like we're kind of like builders and makers so it's uh it's kind of nice it's kind of like reflects a lot about like kind of who we are uh on the team and it's just like my happy place but i've been looking at this for five years we finally we finally got it and uh it's great and i can just walk to manufacturing i can walk to the grid i can walk to building four i can walk here to headquarters and i can just be here with my team and pop in where i need uh to solve like any like whatever the biggest problem the day is
Speaker 2is most of your talent down in south bay or do they all commute um most of the talents here and
Speaker 1we have a shuttle for folks in the city that want to come down so maybe like i don't know five or ten percent of the of you know folks here working like live in the city and commute down and then i think maybe most majority of the rest all live kind of pretty local within like 20 or 30 minutes from the office yeah luckily
Speaker 2the weather is nice here today the weather is freezing and cold and rainy yesterday yeah yeah
Speaker 1the weather is pretty much the same here every day okay let's oh my god we're back we're back so fast how did you come up with the logo okay the logo is uh i've got a couple uh things that are cool here it's like one is like uh it kind of like looks like um like basically how the robot steps and tracking its steps and feet yeah and then two it's like a little f for figure
Speaker 2so this is the fancy secret room now this is a secret room that nobody's allowed to come in
Speaker 1oh uh this is our design studio so i'm going to show you every robot we've ever built wow yeah so we started in 2022 and uh the goal was like how do we get humanoid robots to our ai and software team as fast as possible uh so we designed figure one here this is our first generation robot uh
Speaker 2pretty cyberpunk and how much did this one cost to make and develop going down the line
Speaker 1oh wow um this one was like uh built to be expensive and move extremely fast uh in terms of building it so this was like hundreds of thousands of dollars uh and the robots we have now are are like uh uh you know well under a hundred thousand dollars each um and uh so yeah this was very expensive mostly expensive because we cnc manufactured the entire thing uh like basically the way we made all the metal parts was like extremely high precision like i think like formula one race car type type type stuff um we had this walk-in within the first year uh and we did we did a lot of the early ai stuff here that we kind of proved out the company was it was great um and then we moved on to figure two which is which is here uh some improvements we did is we moved the battery uh that was on a backpack into the torso yeah and then this one had like a basically relatively small computer compared to this we basically tripled the compute so we basically doubled the battery pack uh tripled the compute uh we have new camera sensors in the head and the pelvis in the back uh we had our next generation hands and uh we basically wired the whole robot internally um and we also designed uh designed the structure similar to aircraft aircrafts take loads to the aircraft skin and so we designed the basically the structure is an exoskeleton so all the load pass uh for the structures is exterior of the exterior surface we made about i think like 50 of these and uh we just recently we're just we're just we just recently tired in about a month or two uh and then we've moved now to generation three uh this is our figure three robot those are all through the same with different outfits and um the robot here we like reduced we reduced the weight we made it skinnier but also keep the same power and torques and speeds it's got ozempic yeah exactly uh i think it looks better like this
Speaker 2looks a little too like uh kind of too little robot-y yeah robot less robot too much robot
Speaker 1exactly like uh so we slimmed everything down we soft wrapped it it's got a layer of foam like on the shoulders and tor in the chest so it's soft we have our next we have our newest generation hands on this which have camera tactile sensors uh it's basically able to grasp items much much easier um we reduced the cost by about 90 between these two really yeah yeah what was the major cost there um we didn't care like we've optimized these first few generations for speed we didn't care as much about cost so that was
Speaker 2like the biggest like just like misconception of designing it initially with speed yeah like
Speaker 1i like we were sitting here in 2022 we're like um our software folks need a humanoid to do testing and do ai work or whatever it is on it and so and there's like there wasn't at the time uh still really isn't a good humanoid robot to go by to help us speed us up so we had to go build it so it's like even if it's expensive let's get stuff to the team as fast as possible to start getting like the basically start working on the development process for commercialization and same with figure two the goal was like like um we had a lot of problems with reliability here that we needed to clean up we had like wires poking out and all kinds of issues here with this robot um it was kind of it was kind of faulting every few hours uh so we just had a ton of reliability issues we needed to clean up because we're kind of known though because we're moving really really really fast so i had a figure two was just like way too expensive like too hard to manufacture at scale so figure three was like how do we like reduce the cost by like almost an order of magnitude how do we make a lot of them and how to make it like closer to what we think the ideal outcome is for like every robot in the world um so these robots uh basically have the ability to take these clothes on and off so we have different like different different types of um accessories we can put on them uh shoes gloves
Speaker 2uh like yeah fabrics how often are you thinking about the next version like what do you want the next version to be and like do you start thinking about it while you're building this one or after it like at what point of the production stage do you start the next iteration we we are now building
Speaker 1out almost every year and uh we'll have uh you know we were like we're like late stage now in the design process for figure four and um there are some changes that we like we want to put into the iterate this is like a cell it's like this is like iphone sale like iphones you know what i mean like everyone's getting better apple yeah exactly uh so every word we're trying to get better so there are some things we just didn't get to and have enough time to work on on figure three that we want to de-risk we put into four uh and there are also some things we've learned from like operating figure three now that we're like man this is like could be way better if we did this or that that we're putting into figure four as well and then we want to keep reducing costs and making it easier to manufacture to every step so we're looking at figure four we're like uh figure three we're like oh man like some of these things are like kind of hard to manufacture at baku um or like how are we going to get out of a box or how are we going to get like a new user in a home like really easily so like taking all those collective learnings and we're putting them into the engineering design efforts and we go through many different like basically gating processes for that like starting with like an architecture review of what the system should look like in high level like you know like level zero requirements all the way through to detail design which we're we're now in for figure uh figure four what's kind of crazy here is um i thought at some point we'd like we would saturate out like an iphone you know iphone like doesn't really change much anymore i thought figure three at this point was like this is like a this is like your best human robot in the world and it's every robot is going to be like it's going to be better but not by much what's going to happen here is you're going to have figure one kind of a you know to figure two as like a step up and figure two to figure three at a step up figure four will be the biggest step up we've ever made by far we'll have it out here at one point and you'll be like oh my gosh it's just like radically different and um so we obviously can't talk too much we can't talk about anything basically on what we're doing there but like uh we are just so early we're like almost in like flip phones and now we're entering like iphone one moment i think maybe figure four will be our first like iphone one moment for this where it's just like radically different and probably like for me i think it's like almost the perfect humanoid robot i can think of i'm sure there's things on five and six as we iterate through it'll be even better um but we've learned a lot like here's a couple examples we have um here's some parts we have on the table for uh this is our generation of hands starting with our first generation hand to our current generation hand today we've gone through
Speaker 2like five uh versions of this well they're kind of like same size as my hand yeah uh one thing we
Speaker 1have never shown is our first generation hand really yeah why is that because it's like a why um it was really difficult from an ip perspective and engineering perspective to go build um and two is we think um we learned a lot about it and we learned like things like what are good and good and bad in this case we felt like we had um uh learned a lot about the hand of like why it's probably not the right direction so our first generation hand you can see here is a tendon driven hand yeah we designed all the motors and actuators here ourselves uh even the gearboxes uh basically the the rationale here is that a human hands like this are most of our motors are in our forearm and we're going to be using a lot of the motors and actuators here arm and they're basically we're like uh little like tendons basically driving all the fingers and um so our first generation hand is like how do we get a really dexterous hand built which is like really good for intelligence and ai and uh and then uh ultimately i can do a lot of things that human can and how do we mimic the biological kind of architecture of a human and um so i was like this is going to be great we're going to put motors in here they're going to be really powerful they're going to drive a really high degree of freedom hand and end up becoming uh ended up becoming like the wrong hand and we're going to be really powerful they're going to drive a really high degree of freedom hand and end up becoming like the wrong engineering choice and we end up pivoting away from it really early most of the wrist motors are also in here so on figure one i don't know if you noticed but the wrists look crazy and the reason for this is that we pivoted away really early away from this tendon driven hand right we had to figure out a way to get like new motors in for the wrist so uh instead of waiting like four months to redesign those from scratch um i took the motors from the feet so we have three foot motors here in the forearm and it's just like this frankenstein forearm and it's like it bends like in mid like you know the the instead of bending here the wrist it bends like halfway through the forearm which is just like really weird and i was like uh i was so ashamed i'm like we're gonna get this thing out it's like you know at the time i was like this is incredible did you sell any of these we didn't sell them we just used them internally but we showed it and it was like this big like forearm i was like everybody's gonna notice this big forum it seems so weird yeah and i don't think i've ever had a single person in like three years ask why the forum was like this really everyone was just like it's not a robot not a single one questions yeah so we uh so we ended up pivoting away from like the tendon driven hand to our current generations of hands and um we we've learned a lot but like i um yeah we ultimately i think are building like some of the best hand technology in the world we recently unveiled our um high degree of freedom uh hand as a teaser our next generation hand um we're going to show you how to do that in just a couple of minutes so i'm going to show you how to do that in just a couple of minutes and i'm going to show you how to do that in just a couple of minutes and i'm going to show you how to do that in just a couple of minutes and i'm going to show you how to do that in just a couple of minutes and i'm going to show you how to do that in just a couple of minutes and i'm going to show you how to do that in just a couple of minutes and i'm going to show you how to do that in just a couple of minutes and i'm going to show you how to do that in just a couple of minutes and this is really important not just for like being able to dexterous tasks but we need to be able to learn passively from humans at scale and if humans can move hands in all these different crazy way we need to be able to map to this uh at test time on the robot so we have um i think this is extremely important to get if we want to solve like uh agi and get to like human intelligence in the physical world like it's all going to start here with the hands for us yeah it's intense
Speaker 2it's complex i did see a couple people walking around the campus in spandex outfits yeah it's
Speaker 1a mandatory outfit they work yeah like you just got to be in spandex so you can't come um yeah so we basically are doing a lot of data collection here where we're trying to i basically do like joint level tracking uh and different type of uh data collection efforts like learning from humans like our training set is like how do we like we're humanoid we need to learn from humans at scale and so we're trying to learn as much as possible about human movements and like image conditioning these policies uh here a figure is that the oddest job um what is that called um what is the oddest job in the office yeah do you think that's odd i think it's kind of cool uncommon yeah it's uncommon um probably the oddest job we have there yeah how does one apply you play on the site you play on the website uh we do like we basically have like data collection folks that we basically are here that we have both here and out in the world doing data collection for helix that's cool yeah yeah i actually think it's a really cool job have you ever tried it uh full spandex yeah i haven't tried the full spandex but i've done every other type of data collection effort here maybe that'll be your next job maybe you should yeah i'm gonna go get in some spandex later and i'll test it out okay so we saw the generation of hands okay so generation hands uh we also have some um some mock-ups for the head and feet i think the feet are actually really interesting here uh this is kind of our first prototype for figure three um it's like basically like uh we really wanted to get a toe in the robot which is important both of like a natural looking gate like as you walk but also like getting off off the ground yeah it's really important or even squatting down when we squat down we're like on our toe box um and it's really difficult with a flat foot um and then basically this is our this is our figure two like foot it's basically just a fixed piece of metal nothing too crazy impressive and then this is our current generation of uh is for figure figure three we basically have a toe we have this um opening here in the foot um some people ask about this this is basically for uh thermal venting as we're charging we're pushing air through the calf and the shin uh through through the foot to cool it down as it's charging because it's got inductive coils on the bottom so these feet are basically uh basically stepping onto uh this basically the charger uh we then initiate charge like wirelessly and the robot can charge at two kilowatts so basically we can charge for an hour by standing there that's really cool that's really cool yeah we can charge like at client sites like this we can walk over uh we can dock to it we can just stand on it um over time we're going to get these uh systems uh even smaller and uh be able to put them you basically be able to put it anywhere and you can plug this into a normal wall outlet for charging are you going to do any brand deals um like foot brand like foot deals uh i would love to do maybe not foot but sneaker deals if like if nike's watching like you're gonna be our next uh yeah nike sneaker shoe dealer yeah i really wanted a high top for figure three i think it's just like so cool yeah it's pretty cool um like our figure two looks like a like a penny loafer and just like you know what i mean can't be having that no we can't have a penny loafer out here it's
Speaker 2like this high top's like kind of made to doing work so of the designs here like you have quite a sleek design it's very futuristic how did how did like how did you get to this point yeah we
Speaker 1have a design we have a industrial design team here internally run by uh david uh he just met and we we have a team that are obsessed with trying to create like uh yeah trying to create like the Um, we, we want to create something that, uh, is really delightful to be around. And, um, it's, it's, it's not just like the way the robot looks or the size of it. It's how it walks and interacts and how its body language and how the human machine interaction is it, does it look at you while it's, while it's talking? Uh, how did we deal with speech? What do we do with like, do we have like three screens in the head? Like, what do we show there? How do we make us like really pleasant to be around?
Speaker 2Are there any, I mean, you love sci-fi. Were there any sci-fi movies where you wanted?
Speaker 1Oh yeah. Like the robot movies.
Speaker 2I robot X Machina. Like which one?
Speaker 1I think a thing we always talk about here is like, there's like two roads for humanoids. There's like a road to head to like down, like the robotic road, which is like I robot. And there's a road to head down for like Westworld. Okay. Also humanoid. Yeah. Uh, where do we go? What do you think?
Speaker 2Um, I mean, I'm a big fan of X Machina.
Speaker 1Okay. So what we go to Westworld.
Speaker 2Yeah.
Speaker 1Yeah. Okay. Let's do it. Okay. Yeah. Okay. We're heading to Westworld. Um, I think lastly is we're like super proud to be on the cover of time magazine this past year, which is really cool. Uh, we had a robot in, in, uh, in a home basically doing like full, like, you know, like doing housework with Helix, um, AI system that we designed here internally.
Speaker 2What's with the dead mouth five.
Speaker 1Oh yeah. We had, uh, we had dead mouse at our holiday party two years ago and he was like, this is insane. We had robots on stage. He's like, I got it. I got to get you guys out to concerts with me and we've got to be like dancing on stage. Yeah. So we opened for him at red rock, uh, end of last year, uh, in Colorado. I don't know if you've been to the concert I flew in for. It was amazing. Just like amazing venue. And we had multiple robots on stage, just dancing and they're all tuned into the music and, um, and they kind of dance with it. And it was just, it was awesome. And we had him also here at our last holiday party in December. Um, and, uh, I don't know, just, we just, we just raged with dead mouse.
Speaker 2Hey, it's Molly. If you enjoy our interviews, check out our newsletter sorcery.vc, where we deliver a once a week top deals and tech
Speaker 7headlines email, and also go deeper on our podcast interviews, subscribe to sorcery today.
Speaker 2And don't forget to subscribe to the podcast on YouTube, Spotify, Apple, or wherever you listen link in description to sign up.
PodcastTranscript AI