The Man Taking on Tesla in the Race for Humanoid Robots w/ Brett Adcock | EP #116

29:05And we're trying to push the boundaries now on how to push that work as hard and as far as possible in the robotics space, which we've just been starting in the last several months. I imagine there's a lot of benefits for them as well. I mean, mention in body AI, there's some theories that say, listen, we're not going to get at the AGI, unless we can embody AIs to understand, sort of, the embodiment gives them an understanding of the universe and allows them to explore. And then there's the other idea that we're going to hit a data wall to get the AGI, and humanoid robots are a means for collecting a lot of data to help shape future models.

29:51Can you speak to that a little bit? I think it's becoming more and more clear that some level of output actions in planning is important to take kind of our next step in intelligence. What we're trying to do here is we're trying to help complete that last leg of actions and and reasoning that we're seeing from some of the best world models that we're helping to work on here internally. And then in the day, if you can talk to a robot and it can output actions, like useful actions in the real world, that would just be an incredible technology for the world that we're trying to work on here, whether you want to call that advanced AI, AI, whatever.

30:44like that's just such an important focal point for us to try to get to is how do we output intelligent actions into the world and do useful things. So most of our focus on the AI side is on that topic and how to make that as scalable as possible and generalizable. I mean I have to imagine you know GPT 40 and or or the multimodal versions of GPT -4 are key for you. Having figure C and be able to understand was that a relatively, when you started that didn't exist, the multimodal models weren't there. Was that the conversation you had going on with Sam? Was he sort of like,

31:31was he part of your inner conversation of what's gonna be possible in the future? I think one of the biggest breakthroughs we have is we have like with LLIMs and VLLIMs more specifically we've had this semantic grounding that's occurred in robotics. We have the world's knowledge in some way like a analogy but in a zip file for the robot to access and understand. And that bridge from robot to human was never really here before. If you want to talk to a ton of his car and say, drop me off on the curb over there and on the right, there was no real semantic bridge for that in the world. And arguably we have that now.

32:17We have like, thus semantic bridge has been built in the world. And what we're really lacking is this like, like, I guess, reasoning and planning and perhaps actions from that system to for us to provide useful work in a robot. And so I would say you've had this like unbelievable technology has been opened up and we're seeing some really cool stuff right now in the world, like, you know, with different technologies all the world, and AI, but like one of the things people are not talking a lot about is like, what does this mean for robotics? This means like a robot knows everything you're saying and knows what you mean.

32:51And we have all this grounded in human level data. Meaning we have all this like semantic world's knowledge is written by humans for humans, which is an incredible ability for a humanoid robot that looks like a human to really have really high efficiency transfer rates. So we like the way humans open up jars is pretty similar to robots, our humanoid robots open jar. So, like the affordances are roughly, you know, really high affordance levels as it relates to humans' work. So, this is like unlocks like an ability for humanoid robots to really tackle like general robotics. Like, how do we talk to robots and how do we output actions that everything a human can do?

33:39And there's, there seems to be at a point in time where we can really try to see if we can crack that. So you mean that we're going to see the normal course of interactions with robots be like you speak to a human? It's like, can you please go grab that for me? And it says, what do you want me to grab? And you say that thing over there and you point and it looks. And it understands the stapler, the bottle of water. And so it has got contextual knowledge and geometric and what you call positional knowledge. How far is that? Yes, this is all realistic today. It's beyond even all those things. In the neural net weights, lives the material of the plastic bottle, roughly mass characteristics, and friction characteristics, and how it will feel to grab and all of this is in the weights.

34:38I work on longevity because I want to see as much of the stuff that I possibly can. Yeah, what kind of hours do you work? Because I know your passion, Brett, I know your dedication to this and you are a kidney candy store. And you have to balance families well, what's your work week like? Yeah, I would say I work almost like basically almost seven days a week. There's really no time I'm like really not working except with home like with the you know, wife, wife and my kids. Yeah. And then how old are your kids? Do they get what you're doing it? We just had a family day at figure. So you know, my daughter's five and like was like, in my son's two.

35:23So we had the robots walking around. All the kids were so excited. It was really cool. Yeah, they get it. They talk about dad building robots all the time. and I think it's like, yeah, I think we actually had probably 50 kids here last week watching all of them like see the robots and touch them and yeah, it was pretty special. You know, I think about that. My boys are, I have two boys who are 11, 13, and I can't wait to have them see what you're building here. And I think about the fact that their future and your kids' future are going to be, you know, if your numbers are correct and I think that they are, they will be more prevalent than cars are out there.

36:21Right? There's like a million cars, like a million other vehicles, and we could see, you know, of five to 10, I'm sorry, a billion cars out there, a billion other vehicles. And so it's gonna be a very different future for them. I asked a friend of mine, what's it gonna be like when you're seeing humanoid robots walking around the street all the time in five years? And his answer was interesting. He said, it's gonna seem normal. Yeah, can I tell you something? We, yeah, a bunch of people, you know, we have a lot of folks here, I've been around robots for a long time, they're like, Like, as soon as we do something, everybody's gonna be in awe, and then nobody's gonna care anymore.

36:59It's funny, sometimes that happens. We were walking figure two now around the office, quite frequently. And the first few times, like everybody's just like stopped doing work. They're like fist -pumping from the conference room or whatever, taking photos. We have this really crazy video where everybody is falling around the robot, the first time I walked in on the office with like their phones out. And we do it now pretty regularly, nobody cares, nobody says like, oh, the robot's close to me.

37:29It's pretty unbelievable how used all the stuff we get, right? Yeah, we adapt so quickly. It comes boring. I remember the first time I got one of the first model X's and the door wings come up and everybody snapping photos and looking at it and then it's a nuisance after that. Yeah, we get used to things really quick. I don't know. It's even like using, you know, like large language models and stuff today like chat, GPT. It's just like, I use it pretty frequently during the week and it's just I like, it's totally normal, part of my workflow. Yeah, it is. It becomes just an extension. GPT -5, or have you had conversations about integrating that and its derivatives or the mythical strawberry as it's being unveiled.

38:16Do you get any early views of opening eyes content for a figure? No comment, Peter. No comment. Okay, okay. Can't hurt to ask. Everybody, I want to take a short break from our episode to talk about a company that's very important to me and could actually save your life or the life of someone that you love. Companies called Fountain Life and it's a company I started years ago with Tony Robbins and a group of very talented physicians. You know, most of us don't actually know what's going on inside our body. We're all optimists. Until that day, when you have a pain in your side, you go to the physician and they burn into your room and they say, listen, I'm sorry to tell you this, but you have this stage three or four going on.

38:59And, you know, it didn't start that morning. It probably was a problem that's been going on for some time, but because we never look, we don't find out. So what we built at Fountain Life was the world's most advanced diagnostic centers. We have four across the US today and we're building 20 around the world. These centers give you a full body MRI, a brain, a brain vasculature, an AI -nabled coronary CT looking for soft plaque, dexascan, a grail blood cancer test, a full executive blood workup. It's the most advanced workup you'll ever receive. 150 gigabytes of data that then go to our AIs and our physicians to find any disease at the very beginning When it's solvable you're gonna find out eventually Might as well find out when you can take action found life also has an entire side of therapeutics We look around the world for the most advanced therapeutics that can add 10 20 healthy years to your life and we provide them to you At our centers So if this is of interest to you, please go and check it out.

40:06Go to FountainLife .com -peter. When Tony and I wrote our New York Times bestseller Life Force, we had 30 ,000 people who reached out to us for Fountain Life memberships. If you go to FountainLife .com -peter, we'll put you to the top of the list. Really, it's something that is, for me, one of the most important things I offer my entire family the CEOs of my companies my friends. It's a chance to really add decades onto our healthy life spans. Go to fountainlife .com, backslash, Peter. It's one of the most important things I can offer to you as one of my listeners. All right, let's go back to our episode.

40:47I am curious what you how you think about, about Optimus and Tesla and Elon. I know you have great respect for him as an engineer and entrepreneur. What are your thoughts there? I just like have to be, you know, so inspired by what Elon's done in the last like 20, 30 years. It's been just unbelievable. I think they're doing a great job at Optimus and I think they have a really good engineering team. I think they're making really good progress. And I think it's actually heading in the right direction, like the vector for where we need to go as a society to integrate and build humanoid robots. I think they're following a very good direction.

41:40So I think they're going to be, you know, a really real competitor with us. I think the world needs them out there doing this. And yeah, I hope them, I hope they do really well. I think we're at a point in time where the window is just open now for humanoids to be possible. I don't think it was really possible 10 years ago. It's going to be this a little bit of a race here to get volumes out the door in terms of manufacturing and you get the AI training sets built and deployed on the embodied systems. And yeah, it's gonna be a more important time to make this really work. So yeah, I think overall just like very, I think they'll do very well.

42:28And I'm very glad they're tackling it. Yeah, I think when he walks into a new industry like that, he credentials it in a big way. Did they serve as sort of a stalking horse for you to keep the team? Do you guys like look at him go, my God, how'd they do that? Or hours is better? Is it provide a little bit of that kind of like a gamified motivation for the team or you guys just open loop otherwise? I think we try to like make we try to make all of our decisions from like a first order reasoning like what was the right decision to make if That's like super important for every company to really do is like I think we have our own Mission vision values what we care about very different than any other company and we want to head a certain direction like vector space like that We think it's the right way to go longer term So I think that's like where we really grind ourselves and like the decisions we make.

43:17I think there'll be things that we do a lot differently than other groups. You know, longer term because we're just making these decisions like in a vacuum with knowledge that we have and going through that. Yeah, I think they're a really good competitor. So for sure we want to be we want to be winning and we want to compete, you know, overall we. So yeah, I think our goal is to build the best humanoid robotics company in the world over here at Figure. And I think since the marketplace is arguably near infinite, there's plenty of room for two or three significant players on the planet. I think the way I look at it is we need more like really real players.

44:07I think Tesla's a really real player. Like I mean, you know, highly capitalized, great engineering team, heading the right direction, like long term, moving fast through those high -rate iteration cycles and milestones that are needed for engineering teams that really prove out the product is getting commercially viable. And what we're lacking in humanoid space is that like we have a lot of groups that have been around for a long time around the hoop. And you know, it does not look like there could be a lot of players at all win. It looks like there'll be very few that make it through this like chasm like crossing the chasm here that's needed for getting to some market and I think Mini will die is trying to you know trying to bridge this.

44:49I hope that we live at figure but like we have like 14 peaks to go climb now. We have like a a very hard road ahead to go from where at now which is like two -year -old company it's like a commercially viable real business and that's what we need to go do. I like we have like Yeah, senior to single to two years to go prove that and get into market so like that's We're charging forward as hard as we can to ship our product into customers and make it useful Yeah, I think you know think most people probably don't realize is your only two years old right and what you put together In terms of team and I think this last round of financing if I looked at what you know if you would said To the average person average technologist or venture capitalist what's the advantage that Elon has for for Optimus, he has a manufacturer, he has a car company there that could use it, he's got Compute, he's got Capital.

45:39I think this last round gave you parity, if not in some places. It's like a such a stupid argument though, because like you could say that about every company that's in why they can't be disrupted, you could have said that, those are all Tesla's weaknesses 20 years ago when they didn't have any of those, none of them. It's already making electric cars and somehow they won. So that's not really what happens in real life. That's not those attributes, do you not set the winner? I'm not telling you that's the winner. I'm just saying that those attributes were advantageous. And now you've got the capital, you've got the compute with OpenAI, and you've got BMW and so forth.

46:20I think all the building blocks that we need to do to build a healthy company are all starting to show up. And that's good because we need, like those are like requirements that are needed in our long -term roadmap. And we just got to put all those pieces together very intelligently. And then not die into market and make a... Not die is a great part of the business plan. Can we talk about China? Because I find that absolutely fascinating. So China succeeded on the backs of low -cost labor. And that's going away. We hit COVID and no one wants to manufacture it on the shipping costs. And so I'm tracking the robot companies coming out of China because I think that road for so many reasons like their aging population, their one child, you know, per family policy, all of that and trying to actually maintain a manufacturer based all requires robotics.

47:14I see Unitry in a few others. What do you think of the Chinese robots? Are there any that seem like good competition I just got back from China a couple of months ago and it was one of the best visits I've done in my career, to be honest. Went to a bunch of kind of manufacturing focus companies in mainland China and it was just, it was crazy. We were touring one facility and I was like, what's that written on the wall over there? And they're like, oh, that's just our motto for this building. I'm like, what does it say? They're like, if you're having a bad day, just work harder. And I was like, these guys are animals over here.

47:54They're just trying to build and ship. The work ethic was really high. The sheer will of the country to try to be, of be like competes and win is really high. And I was floored. I was like, man, you have, you have like, many times before. I've not been there, I would call it a lot. And specifically, it's been a lot of time touring, kind of what high rate manufacturing processes look like out there. And it was shocking. I think we have in the humanoid space, we have like figure and optimist here, outside of China. I do think China is the next group of folks that are going to be really competitive long -term in the human rights space.

48:50I think they have to be. I used to go to China every year and bring a group of abundance members with me and we'd go and visit all the top tech companies. And I remember the motto there was 996, like a great lifestyle was 9am to 9pm, 6 days a week. That was the work ethic. And everybody, China had the reputation Asian has been copycats, but you know, sure, they copy a lot of things, but they also did a lot of authentic new development work there. Is that your experience as well? My experience has been that those folks out there want to win. They want to do it at the lowest cost. They want to do it at the highest rate.

49:34And they'll stop it, nothing to try to be number one. And that's, you know, think about what startups are. It's like it boils down to like those key principles you start a company with that are successful. You have nothing. And it's just sheer willpower to get there and to go win. And they have that in spades in China. And they do not have the resources we do here in the States. Is that, you know, like we said before, does that really matter? That you have like all these things? or is that is at ultimate crutch. And so I think there were gonna be some unbelievable robotics companies come out of China just because of the sheer number of projects being worked on and with the will that I saw out there, is second to none.

50:25Yeah, I mean, I think in the same way that Israel developed an amazing defense industry because they had to survive. I think China and Japan and South Korea with dwindling populations and aging populations are going to need to develop an incredible Robotics industry to survive and thrive and maintain a GDP Yeah, so you you take archer public Which was an incredible success and congratulations on that And after starting that and running that for a number of years You break away to found figure and if I'm correct you know you made a commitment you put a significant amount of capital on the table to start the company which you were able to do because of your previous exits from from veterinary and from from Archer but I think when I when I first heard your presentation and I brought it to my venture fund what I was so impressed by which clinched the deal in my mind was the team you built.

51:31I mean, it was an extraordinary group of engineers from the top AI and robotics companies or tech companies out there. I can use, where to advice for founders who are starting the company. I mean, you're a technical founder as well, which is important, right? But how did you recruit your team?

51:57So my belief is that in order to ship a really good high quality product, you need the world's best team there to go do that. Especially against the difficulty level we have a figure of succeeding like the odds of success are always pretty low. So you need to give it everything you got. You need the best team here. You know, on -site every day. You know, work hard. You know, you need to have like a high -functioning workaholism bit. So I spent the first year basically trying to map out what the organization needs to look like in terms of skills and what is the ultimate org chart need to do to support a really high -functioning team to build a product.

52:43And I spent the first year just basically head hunting all that whole team by hand. And I was cold emailing and calling, draft of the off letter, would give the off letter, go do dinners, try to close on board, you know, 30, 60, 90 day on boarding, bring them in, lead the engineering decisions and direction, and you know, work on those projects with those teams. like I think, you know, it's been a huge payoff because we've gotten like now is point where I think we have like one of the better teams ever built in the world for this. And it's like snowballing, we're able to track like really high quality talent over all our disciplines.

53:20But in the early days, in the early days, Brett, I mean, how did you get that first dozen? Was it your conviction? Was it, you know, your capital commitment to it? I mean, because in one way you had, you had Tesla bot, I don't know if it's called Optimus back then, you know, and Elon tends to suck the oxygen out of a room on the stuff that he does. You know, convincing people to jump where they were and join you, was it, how did you do that? Because that GG2 is really important. Yeah, I mean, the pitch for early figure was we're working on humanoid, we have this big mission to advance human capabilities with AI.

54:02We believe in AI first in the market. We believe a vertically integrated approach to hardware design is important. We're living in the largest town or working in the largest town in the world. It's under half a GDP as human labor. I'll fund the first several years. There's no funding risk in the near term.

54:22And it's my second time going around building hardware and growing hardware teams. So having done a decently well at archer, getting back and doing it from scratch, did it even better this time in terms of like setting the right direction and mission vision values of the team. And then I just spent a lot of time with those folks saying, you know, come here, let's go out and build this commercial. Let's build the best organization we possibly can from early days. The early folks got all like founding, you know, founding member stock, which also was helpful and beneficial. Got good salaries. I was, you know, funding myself.

54:53So it was like, you know, it was being a part of like a from scratch startup, but like almost like cushioned by my ability to self -fund it for multiple years. Good equity, you You know, by the time we had five or six people in the team, they were all superstars. And then that was like more than the risking for the next five or six folks that were joining. So I was able to find folks that really believed in the space and believed in me. A lot of folks that used to work with me in my last two companies, a lot of folks that were new as well. I had my first two people were folks that I've worked with that, like my first employee at Veteri, my early employees at Archer that I worked with, they were like, so it was three of us, day one, basically, that have all worked together.

55:35Added some more folks in my ultimate archer, added some more folks in Boston and Amics and other organizations out really great. Soon enough, we had a dozen people that were superstars and they're disciplined. And then, you know, we walked our robot within 12 months of incorporating the company. So we worked really hard and got out of the gates pretty fast. And, you know, now it looks like, okay, it looks great, then, but I was in a we work phone booth for making cold calls, trying to convince and talking to, you know, wives and husbands and telling them to convince their, you know, it was a fouls to join.

56:12And it was hard. Yeah, not going to lie. I bet what did you give the the chances of success back then in the early days or to get to the point where you are now? I mean, if you had to give a thing back then, you'd be like, yeah. Yeah, I would not have ever assumed that we would be out of place now. We've really You know, we I think there's the like you know, I look at everything as like how well as the product doing is really still like the the roadmap into commercialization We're still you know, we're not like we're not successful like we're not even like shipping Yeah, I think that but like you know, we've been around for I don't know like 20 Like you know a little over like a little over two years or something like that unbelievable, like the hardware that's doing the AI systems and working on ex -canilles was policies.

57:01Like the, the, the, look at the humanoid now versus like 10 years, it's unbelievable. And we don't ever walk into the office and like things go backwards. Like they always go forwards. Like we're not like, you know, sometimes we have like a couple of days where like, you know, a couple of robots break and we don't make progress or fix it or whatever else, or a lot of pro, meaningful progress. But most weeks we're like making pretty decent like if folks are gone for a week they come back and be like this is a whole new company it feels like So like you know every every week goes by or making pretty substantial Impact so I would say like we far exceeded my My early impressions of where we'd be out two years from now.

57:40Yeah two years ago. I'm curious. What did you think would be really hard? That turned out to be easy And nothing easy, but it was easier than you expected. Well, I thought, maybe like the opposite I'll do. I thought like we'd be able to procure a lot of supply chain to go build the robot. Like electronics boards, motors, actuator systems, battery pack systems, cameras, like lights, like speakers. Like we like speakers on figure two. I thought I'd be going to Ali Baba I go to Amazon by screen, screens. Like, screen speakers, lights. These things have to be about the shelf, right? Like, you have to go buy those things and they come in from Amazon and you put them on a prototype.

58:27Like, that's like what, that should be that easy. It's still not that easy. Like, we're making custom speakers. Like, how is it possible that we're making custom speakers in writing custom firmware for some of those areas? Like, how is that even, it doesn't make any sense? that's even possible. So going into a little naive of like these sensors for torque cells or whatever, like would be procured and would come in. I wouldn't have to do it. Like we make our torque cells, like four cells here, which is like a flexure, the board, they need to be gauged, they need to be calibrated. They need to be tested.

59:04Integrated. There's like software and firmware on those on there, like, like, and like that's got to work then at really high rates. It's like, it just still boggles my mind. There's no mature supply chain across all this stuff. It's just unbelievable. And I'm assuming that that's not your first choice that you'd rather buy from a reliable supplier versus vertically and vertically and read everything. Everybody would always choose to buy something if it's easy to procure and not to only vendor in town. Everybody would always choose to buy. Nobody in the right mind would ever want to build in that case.

59:38It's just an enormous effort in burden to maintain it, and to QA it, and to fix bugs, and to pay human salary, and manage humans. It's just like, it's just a hard thing to do. And, you know, you'll get through the human phase pretty soon. Yeah. What percentage of the robot is manufacturing house, do you think, like rough orders? Is it like 75%, 90 % by whatever metric you want? Wait, or a cost? I mean, we probably look at how much of this is like we're designing ourselves.

1:00:14I don't know the exact number of top ahead, but I have to think it's probably like 70, 80 % of things we're designing at this point for the robot. I mean, there is an advantage of control and quality and pulling out overhead or margins out of the vendors. I remember I was visiting Elon shop in the early days and he was in this conversation by some supplier, because a lot of these suppliers are all defense aerospace and they're just extraordinarily expensive. And he was like screw them, we'll just make it ourselves was the attitude. I don't know, that's what we, like we've been around for two years.

1:00:54So most of the engineering decisions around the build materials and the supply chain would have been let's get speed by getting parts in the robot, get robots to AI, you know, AI engineers, controls engineers, let's get work done. So we would have made those decisions quite quickly to outsource that work to get robots up and running faster. So in the I think longer term, we would have worked the supply chain to own it better, to have less risk to reduce the margin profile across all the supply chain. But there's just like, yeah. I think we thought that was easy. It became like a bloody hell. So what's the flip side of this?

1:01:39The thing that you thought was going to be hard and turn up be easy. Would it maybe be the AI integration? I mean, when you started, did you imagine you're going to have to build out your own AI models? We still do a lot of our own AI models here, which is like maybe not super well known. But we do. We have a whole AI team. we do most of the lot of work ourselves. I would say across this, we leverage some group, like OpenAI I really work with on new models and stuff and use their VLM for some things. So I think that's been really helpful. I think AI systems is probably the one that's been like, wow, this is really working.

1:02:25And it's, you know, you think about this split of like, what percentage of the things are you going to have to hard code and code to go do, which is like, you know, more classical controls and heuristics. What percentage of things can we do through neural networks? And I would, I would have thought, going into this, I guess my thoughts two years ago is like, you know, a large percentage of the stack would be written in heuristics and code, called like 95, 90%. And then as we are able to, you know, write neural nets for those that are at equivalent performance, we will just take those at level of like code C++ and remove it with the neural net.

1:03:03And it's kind of been flipped. All the new iOS is working for like a small team with like, you know, we don't have like a ton of tooling and infrastructure and things, almost like right off the bat with like everything, like just like using slam or perception systems and object detectors and planning and speech -to -speech reasoning, everything in the stack, like the high and low levels have been working incredibly well. And so, yeah, I think it's pretty magical. I just like, it would take us such a long time to code all this stuff. Did you see the movie Oppenheimer? If you did, did you know that besides building the atomic bomb at Los Alamos National Labs, that they spent billions on bio -defense weapons, the ability to accurately detect viruses and microbes by reading their RNA?

1:03:54Well, a company called Viome exclusively licensed since the technology from Los Alamos Labs to build a platform that can measure your microbiome and the RNA in your blood. Now, biome has a product that I've personally used for years called full body intelligence, which collects a few drops of your blood, spit and stool, and can tell you so much about your health. They've tested over 700 ,000 individuals and used their AI models to deliver members critical health guidance. Like, what foods you should eat, what foods you shouldn't eat, as well as your supplements and probiotics, your biological age and other deep health insights.

1:04:30And the results of the recommendations are nothing short of stellar. You know, as reported in the American Journal of Lifestyle Medicine after just six months of following biomes recommendations, members reported the following, a 36 % reduction in depression, a 40 % reduction in anxiety, a 30 % reduction in diabetes, and a 48 % reduction in IBS. Listen, I've been using Viome for three years. I know that my oral and gut health is one of my highest priorities. Best of all, Viome is affordable, which is part of my mission to democratize health. If you want to join me on this journey, go to Viome .com slash Peter.

1:05:09I've asked Naveen Jane, a friend of mine, who's the founder and CEO of Viome to give my listeners a special discount. You'll find it at Viome .com slash Peter. I remember you described into me the time when figure one put the cure cup and made you a cup of coffee and you were like we're trying to code this and then we just had them watch somebody do it like a dozen times and they're able to model it. I'm curious you know is the way you're interacting with figure two like hey watch me pick this thing up and put it over here and now you do it. Is that conversational, you know, and the visual models, is that where the state of the practice is?

1:05:50Yeah, we're like literally talking about to go do something and it does it. So what's like the most interesting thing you've had figure do for you? I mean, we can literally talk to our robot and it can like do tasks right now, which is like unbelievable. I mean, like the in -state for this is you really want the default UI to be speech. You just want to talk to the robot. I want to, when I'm like, like, another robot. Yeah. Like, another way to think about it is like, I'm on the robot, I'm like, you know, one way to do is like, I pull my phone out or my laptop out and I'm like, you know, opening the terminal and like, trying to get a command to the robot to go do something.

1:06:32And the robot's like, standing right there listening. And I'm just like, I just, it just like wants to be talked to. You just want to just be like put all this stuff away and it's be like figure two, you know go do this and We're doing that now, and it's just magical. It's it's it's um It's really clear that the default UI for humaner robot is gonna be speech. Yeah, you'll just want speech speech reasoning to get really high and You ultimately at the end state want to talk to robot and have it be able to Learn as this in the environment through vision and sensors and you wanted to get better over time and it certainly seems like they were heading in that direction.

1:07:12Yeah, I remember we think robotics used to like move the hand for them and say do this again, but yeah, I mean speaking to it and showing it what you want if it doesn't understand. Can we talk about robotic safety? I mean, where are you on safety? And how do you make sure there's no like like Aaron Firmware upgrade, it turns robots into, you know, overpowered slayers. You know, I mean, there's enough dystopian Hollywood movies that I want to, you know, bring them forward. But how do you think about safety? How important is that where does Asimov's laws come into your mind and practice? There's like many ways to think of those.

1:07:57We have like the, like this system safety engineering side of like making the robot like actually really safe when it walks around the facility and does things with next to humans. There's a whole architecture. Architecture is actually designed from top, like the bottoms up to make that a safe, safe system that ultimately we can put a safe hardware system into the world and it will basically react like we would want it to in all these different various conditions. And I think there's just one piece, we need to save hardware around humans. There's another piece of cyber security and other things that we don't want anybody to have root access to the robot and be able to take command of robots and to do potentially bad, delicious things with them.

1:08:43And then you have other things like what happens on set of AGI? How does that really impact the safety of the robot? So I think there's like many things we're doing here as it relates to like the low level like read only, like firmware on the robot is really like what it like, what is the lowest level code sitting at the robot that can't be overwritten, say, so that more can like the, you know, three laws. I mean, it's pretty cool. You get to think about, you know, the laws of robotics you want to instill in your robots. Yeah, I mean, a lot of this stuff is like, if we don't think about this now, we have to do like full architecture rebuilds, which affect the system quite a lot.

1:09:22So we kind of have to start thinking about all this stuff now to design it. So I think it helped me understand this. So right now, can a robot mechanically have enough velocity and torque to harm you? Because a lot of the sort of automotive robots, you know, they put cages around them and so forth. And there have been designs where it's like, okay, I'm just going to make the robot. So it can't build enough velocity. It can't outrun you. You can't, you know, sort of you put that kind of a constriction on it. I mean, the robots like 150 pounds, you know, it certainly has enough like gravitational potential energy, no matter what, to hurt you so that fell off of much of stairs and was falling on you, like it certainly could hurt you.

1:10:06So like, no matter what, I think whether, because it has, even has like, you know, torque sensing applied so it can't hurt you next to you, there's certainly a scenario where it could be harmful, no matter what. So I think, yes, this is for sure a system that can be harmful to humans if not designed properly. Or it has like a certain episode or like fault on the robot that somehow is next to human and hurts them. So, this is something that has to be done very thoughtfully from the beginning as a system safety architecture. And then we have to gradually prove out performance of that CC system over time through an incremental approach.

1:10:47So like our first robots in our in the work cells in the workforce will be isolated from humans Um if humans enter those work cells we will shut the robot off uh over time we will go from there to like fully um next to humans um you know collaboratively And that will be like a gradual phase 3 best guess is that two years five years? Um I actually don't like we're mostly concerned on trying to get the performance up of the system in the first few years in the reliability. That's probably the hardest thing we have to do. Like getting, you wanna go in to see a robot like working full, whether it's like, you know, like you'll see it like a light curtain like an artificial cage around the robot.

1:11:31I think solving like a collaborative robot next to all humans is like super solvable. I think the hardest hill we have now to climb, which is harder than that, is getting the robot to do end -to -end work every day without like failure. Like, you know, it makes like local failures. Like sometimes I miss like I miss an object but I go and re grab it. So we can like we can fail locally but not globally. We need to like we have certain performance outputs. Like let's say we go into a warehouse. They're gonna have certain amount of output per day of packages or whatever they're gonna do of SKUs. They'll need to go hit.

1:12:03If it's a bunch of robots doing that, they'll still have those same performance goals. So we know if we miss a package, we just gotta deliver it the right way on time. So what I guess what I'm trying to say is, yes, I think we'll solve this decade humanoid around humans and interacting with them closely. But before you ever see that, you'll see, it's like the analogy of the Guamma, right? You saw in Wemmo in certain permitting areas of San Francisco before you saw in other cities. Sure. And you'll be sure with the practice drivers in the seat, you know, watching everything. Yeah. You've seen it with everything, even like autopilot, right?

1:12:41Like you've seen it on highways doing it well as they're beta testing the new software updates for other types of concept operations. So like I think for us, we're just, you know, we're at this period now where we need to prove that it can work and do useful work, even on a confined perspective. And then over time, like yeah, we got to build in the right system safety certification almost to like make sure it's ubiquitous. It can walk around humans, it can like give items to humans, It can, you know, and we'll have certain safety precautions we do next to humans to make sure we're not Operating at full torques and full speeds next to humans.

1:13:16Do you think as the mob's laws would should actually be incorporated? You know, they're pretty fundamental. You know don't harm a human or do something that causes a human to be harmed or by you know not do something that by by By not taking action causes of human to be harmed. I mean, it sounds pretty fundamental, but But it sounds like it's the AI layer. Yeah, there's certainly a select number of really important do not override read only instructions that need to live on the robot that can never be altered. That's like for sure. And then do you imagine figure, I mean, so one of the biggest advantages of having a robotic work force is that they can all learn when one robot learns a task, they all know the task.

1:14:10And that requires sort of a central control. Is that going to be, you know, is that like one central control for a BMW plant or is that figure on a global level, where there's a sort of mission control that is watching all robots and learning from them? Yeah, you, sorry, for us, we would want robots as a fleet doing continuous learning and continuous training on that data set. So it'll be a situation where we have millions, if not billions of robots on the planet, hopefully someday that are all continuously learning as a group. We're doing offline training on that and then the robots are basically getting smarter collectively.

1:14:57It's like a collective intelligence. That's for sure what's happening in the direction we're heading towards. What's so powerful about this is humans, my kids, they learn how to walk. They learn how to walk, mostly failing, learning what not to do in some ways. They learn new things. That takes a lot of time. But once we know things like really well, we really don't forget, rarely do we forget how to walk or a certain things. And that's happened throughout history, but most time we really don't forget. So for robots, one of the biggest advantages we have is once one robot learns a certain task, every robot in the fleet will know this.

1:15:41And so it'll be very exponential in terms of the amount of new, almost like the matrix. We will teach robots will learn to human demonstrations and through reasoning, like how to do something. And once it's been able to demonstrate that several times, and we've been able to, you know, like close the loop on that, like, okay, that's worked really well, this reward system. Every robot in the fleet will know this. That's why I think, you know, future surgeons will best be done, surgery will best be done by robots. you know, when robot is seen millions of different surgeries and can be the best and reliable surgeon out there.

1:16:24Can we wrap up a conversation and just talk about jobs? Because I have to imagine that people are still fearful about robots taking their jobs. We have like really good feedback so far. our goals to really be able to do a lot of the jobs that are not desirable by humans. I mean, you mentioned earlier in this podcast, we had like 8 million US jobs that people just don't want to do. We want to do those jobs. And we want to do the things right now that are really harmful and dangerous for humans to do. And a lot of those jobs have very high unemployment rates, very high, very low retention rates.

1:17:17And we want to start trying to do those and try to help automate. And we've been doing that as a world for several centuries. My family is farmers, right? At some point we have to roll as farmers. I grew up in a 90 % of the world was farmers. Yeah, like everybody was farming and now, you know, like if you want to know 1 % you know not very many so like you know I grew up on a farm like that's like you know And nobody's mad that we're all not farming like you know what I mean like I'm like you know I'm not out there being like 80 % of everybody should be farming right now like that wouldn't be great right for all like plowing fields and harvesting corn and soybeans like that's not that I don't think that would have been productive state for humanity in 2024 Do you remember when the Fukushima nuclear reaction reactor and then DARPA had the DARPA robotics challenge?

1:18:10Have you ever seen those videos where the robot was just laughably capable of opening a door or climbing steps? I mean, I think that's the, is the defense department and the government a client or a near client? How do you feel about plugging into that world? Yeah, we won't do anything defense related at all. It's like Interim made a FESO online. We think this civilian market is just orders of magnitude bigger than defense. And it's not in our interest to build a war machine of any kind, even not a kinetic war machine. So we won't even have conversations. we will take phone calls with anybody in this area.

1:19:01Fascinating. How do you feel about civil police and security? Right now, now, like we're not touching anything. We don't want to give any requirements that are needed to have the robot produce harm. I respect that. You know, to like any humans. Like our goal is to do work. Like we want to do work. We think that frees, we think that significantly helps the economy, helps lower goods and service prices and do good work for the world. And I think it's much needed. Like I think, so we're putting all our eggs in that basket right now. What industry do you see next? I mean, you basically have, I mean, it's every industry, but where do you think you want to play?

1:19:48I mean, you're probably getting a ton of solicitations, right? Yeah, we have a lot of taking it a step at a time. I mean, these companies we're talking to, our massive. We could ship thousands of robots in these groups. So for us, we'd rather work with only a couple of groups right now and do that really well than opening up the hundreds of groups at the moment. But at some point in the coming years, we'll sell to anybody. And we're starting our production line next year. So we'll start with a few as we have engineering really close with those customers and making sure that the product really works well.

1:20:28There'll be a bunch of bugs and process improvements we need to make to make those well oiled machine for the market. We want to do those with those customers now. And as we get to some level of product maturity, we'll branch out to more and more customers. And something that we're going to spend more and more time on is also robots in the home. and when can I own a robot in my home? Because I'd like to put my order in the now. Every six months that goes by, I go to a couple of folks here and I say the timeline for us getting in home is accelerating. And it, one of the things that will be really helpful for us getting into the workforce, it'll really help us get system reliability up, safety up and the cost down and volumes up for manufacturing.

1:21:14It'll really help us in the home. because you really want, like you'll probably have like an order of magnitude pricing collapse going into the home from the workforce. Sure. And you really need economies of scale to get there and you really need a safety certified system in the home. So like you need a really safe product in the home. So we are using the workforce in a lot of ways to boost that vision for us. And but I think from a performance perspective, I think we'll start doing early work in the home and then like, you know, I think the home for us is an area where, yeah, I think it's accelerating in my mind every six months that we...

1:21:56Okay, but give me a guess, is it three years, five years, eight years? I would say within the next three years, we'll definitely have robots pounding in the homes. Nice, okay, I want to volunteer early on. I'll pay.

1:22:10I'm sorry, give me the bugs worked out, understanding how the system architectures all work. But I'm interested in seeing what problems we face that we're not prepared for, that are limiting our ability to get in the home long term as well. Amazing. So last question for you, Brett, manufacturing. You scaling up, you're building a manufacturing plan for a figure. Yeah, we're actually starting our production line next year. We're going to do that. We're going to do it here in California, like close to engineering. So we really work out the kinks on more of a traditional pilot manufacturing line or production line.

1:22:48And then from there, we'll start announcing our intentions for like, like, high rate manufacturing. But our production line is being, we're starting to design it as we speak. We'll have robots coming off the production line next year. And you're scaling it for what kind of volume you think. and we'll start with hundreds of robots and then thousands. Like we really want to get the, what's more important for us than just having like, you know, here's 3 ,000, 4 ,000 robots out, is getting the process really dialed in for how to do that, and then also making sure those robots that we're producing are working really well.

1:23:24There's a situation where anybody easily can get into where we have like too many robots that don't work well, and you're basically in the second recursive loop fixing them, I'm always having them down. Like volume doesn't really help. So you really need the system reliability to be at a certain point. And the production, like the performance of those robots then into the use cases to be pretty high. I do not feel like building thousands of robots is it seems super tractable to go do. Now building like hundreds of thousands of robots has a different story than building millions is a different story.

1:23:57But building like 2000, 5 ,000, 10 ,000 robots It's just, at this point, we're building one a week, so it might seem, and then we'll start building one a day, and then we'll start building multiples a day here next 12 months. So I think, but it seems like pretty straightforward path. I mean, we make cell phones almost by hand in the world to make a few billion a year. Like, this is more complex than a cell phone, but far less complex than a car. So, yeah, I feel like the path to making thousands in the near -term is just not super difficult. What's the difficult part is making those thousands to a point where they're really useful and really work.

1:24:36That's the name of the game for me. Brett Adcock, I'm a huge fan of what you built and excited for figure two's roll out into the world. And actually what I saw with figure three is absolutely gorgeous. So excited for that as well. Thank you for all the hard work you're doing. and you know, I do think this is a way of uplifting humanity and creating an age of abundance, so much appreciate. Let's do it Peter. See ya. See ya pal.