The Future of Humanoid Robots With 1X’s Bernt Bornich – Ep. 259

0:00:16 Hello, and welcome to the NVIDIA AI podcast. I’m your host, Noah Kravitz. As we move deeper
0:00:21 into the age of physical AI, advanced humanoid robots are reshaping industries and enhancing
0:00:26 daily life. Advancements in AI models and simulation tools are enabling robots to perform
0:00:32 complex tasks with greater dexterity and efficiency, and foundation models are playing a key role
0:00:37 in generalizing robot tasks. Here with us to pull the curtain back a little more on the
0:00:42 rapidly developing world of humanoid robotics is Bernt Bornick. Bernt is founder and CEO of
0:00:48 1x Technologies, a Silicon Valley company that’s dedicated to building fully autonomous humanoid
0:00:53 robots. Bernt, welcome, and thank you so much for joining the AI podcast. Thank you, Noah.
0:00:57 So, can we start with a little bit about your journey, how you got into the field?
0:01:03 Sure. So, to me, this has kind of been a lifelong journey. So, as a kid, I was the kind of kid
0:01:08 who picked everything apart to figure out what’s inside. I think nothing with motors in it of our
0:01:15 kitchen appliances survived in the early days. And I also just, I was very lucky. I grew up with a
0:01:19 dad that loved building things. We built soapbox cars in the garage. We went all in, like, welded
0:01:27 aluminum frames and chassis. And then I discovered computers, right? And we’re back in, like,
0:01:33 Commodore 64. And then, like, early, I think it was Intel 286, like, the first one. But, like,
0:01:38 the first kind of, like, computer I had my, had my, only for me, right, in my room was, like,
0:01:40 a 486. Okay.
0:01:43 And I really got into programming. Right.
0:01:48 Mainly due to games and just being part of the mod community, making mods for games is a big
0:01:51 thing. One of my childhood heroes is John Carmack. Right.
0:01:57 And I learned writing code by reading his Quake code. And then, you know, at some point it clicks
0:02:02 and you’re just like, wait a minute, I can type code and this thing moves. Right.
0:02:04 And, like, connecting these two realities, right? Yeah.
0:02:08 The digital and the physical is, like, it’s just magical. It’s magic, yeah.
0:02:13 And, yeah, one thing led to another. And very early, I think I was, like, 11, I decided I want
0:02:18 to make humanoid robots. And then I follow the field ever since. Right, right. So, I’m, like,
0:02:21 I’m the luckiest guy alive, right? Yeah, that’s amazing. I get to do my childhood dream. Every day
0:02:25 in the morning, I get up. And, like, even now, I get up and there’s a robot walking around in my
0:02:29 house. Right. And it’s absolutely magical. That’s amazing. So, tell us a little bit about
0:02:34 1x then. What inspired, I mean, you’ve been talking about your journey, inspired by everything
0:02:38 you’ve been doing since you were a kid. But how did the company get started? And maybe you
0:02:42 can talk a little bit about the unique approach your company brings to building humanoids.
0:02:48 Sure. So, I think it comes a lot back to how do you build intelligence, right? So, we talk
0:02:53 a lot about physical intelligence. But to me, it’s intelligence is intelligence. There are physical
0:02:57 aspects and there’s digital aspects. And we started out with the internet and we still do
0:03:04 kind of bootstrap because it’s what we have. But if you want to get to actual AGI, you need
0:03:08 to learn in the real world. Right. And if you want to learn in the real world, the same rules
0:03:13 apply as in any kind of AI model. And there’s almost some poetry in it, right? Because intelligence
0:03:21 comes from diversity. And this is very clear in our digital AI models. You have to have information
0:03:27 from all kinds of different dialogues, histories, whatever it is across the internet. Right. If you
0:03:34 try to create a very good language model to write poems by only reading poems, it’s not going to work.
0:03:39 Right, right. You need like the diversity of everything that is human knowledge and how we think.
0:03:46 Same is true for robots. So, if you have a robot that kind of grows up, lives its life and dies in a
0:03:50 factory work cell, moving something from A to B. Yeah. It’s not going to be very intelligent. Right.
0:03:54 It’s not going to give you a lot of information. Right. So, it actually has to happen among people.
0:04:00 So, the first thing to realize is that for robots to be truly intelligent and also for them to have all
0:04:06 the nuances that we appreciate throughout our life, like being careful around your pet, holding the door
0:04:12 open for someone that’s elderly and generally behaving like we want them to behave. Right.
0:04:16 They have to live and learn among us. Right. So, that means robots have to be safe. And that I think is the
0:04:22 thing we’ve spent the most time on in Onex from day one, 10 years ago, is how do you build a system
0:04:27 that is as capable as a human? So, it needs to be very strong, it needs to be fast, and it needs to generally
0:04:33 be very dexterous, but still as safe as a human. Right. And this is kind of counterintuitive. Okay.
0:04:38 Because we think about robots as this big, stiff, heavy, dangerous things, right? At least for me,
0:04:42 with a robotics background, right? Right. Because you’re used to the industrial robots. Right. I was
0:04:45 going to say, I just see the videos of the robot dogs running around and I’m like, oh, they’re agile
0:04:50 there, but yeah. They are, but like industrial robots are typically extremely dangerous, right? It’s very
0:04:54 high energy devices. Right. And if you get hit by a robot, it’s really dangerous. We even have a word
0:04:59 for it, right? In robotics, we call anything that’s not planned for a collision. Well, it turns out that
0:05:04 humans collide like many times every minute, right? It’s all we do because this world around us is
0:05:09 structured. Yeah. It’s incredibly hard to plan for when you will actually hit the world. Right. So
0:05:14 you need to build robots that like, if we go running because we’re in a hurry and we crash around the
0:05:18 corner, then might be awkward, might be painful, but it’s not dangerous. Right. That’s what we want
0:05:23 to build. Right. Okay. So we set out to do that. I think we’ve gotten very far. We have a robot now
0:05:29 that’s just 30 kilos or about 66 pounds. Wow. Okay. It’s as strong as an adult human. So it can
0:05:34 squat or deadlift like 150 pounds. So that’s roughly actually human athlete level in power
0:05:39 to weight. Okay. But also the energy in the system is just incredibly low when moving because we have
0:05:44 these tendon drives where we’re pulling tendons, loosely inspired by muscle, instead of using the
0:05:48 more classical, heavily geared industrial approaches. Okay. And that helps with efficiency?
0:05:54 It helps with efficiency, but most of all, it helps with being soft and compliant.
0:05:59 Oh, okay. Right, right. It helps with trying not to nerd out too much on the math here, but I have to.
0:06:06 No case. So this principle is actually very simple. It is when a robot moves its joint, then of course,
0:06:10 the joint itself rotates, right? Right. To move the arm. Right. So let’s say that that rotates at a
0:06:15 certain velocity, a certain speed, and then inside you have a gear. So now let’s say the gear ratio is
0:06:20 100 to 1. Okay. Then the components inside, they’re also at 100 times faster than the joint. Right.
0:06:25 Energy in a system is the velocity squared. Right. This is why you learn in school that
0:06:29 a car that goes twice as fast is not twice as dangerous, it’s four times as dangerous. Right,
0:06:33 right, right. So you can just think about it. If you see something spinning at 20,000 RPM,
0:06:38 even if it’s pretty light, and I ask you to stick your fingers inside, your entire body is going to go,
0:06:41 no, no, no, no, no, I don’t want to do that. Because you have this intuition, right?
0:06:45 There’s a lot of energy in that. And that is true. So if you look at these systems,
0:06:51 for a typical robot, it actually has about 10 times the amount of energy due to the drivetrain.
0:06:54 Right. Okay. And a nice analogy is just think about a kettlebell. Right.
0:07:02 So it’s kind of like you attach a 10, 15 kilogram or like 30 pound kettlebell to the wrist of the robot,
0:07:07 and then you go and do everything you’re supposed to do. That’s like the effective mass just or weight
0:07:10 just projected from the drivetrain. Right. Okay.
0:07:11 And that’s what we get rid of. Right.
0:07:16 Right. So it’s not just that the tendons are soft. They’re not that soft. It’s that you don’t actually
0:07:22 have these high gear ratios. And that means that your system is super dynamic and agile because
0:07:23 this is how nature works. Yeah.
0:07:27 We generally have like 0.7 to 1.2 in ratio. Okay.
0:07:30 That’s like the animal kingdom where the animals are agile. Right.
0:07:33 Right. But also it’s what makes us safe. Right.
0:07:36 Right. And we’ll get back to this when we talk about how we learn.
0:07:37 Mm-hmm.
0:07:40 But needless to say, we need to be safe to be among people.
0:07:41 Of course.
0:07:45 But also we need to be safe to be able to explore. That’s how humans learn. We need
0:07:49 to be able to interact with the world and fail. And then the world should be okay. And we should
0:07:51 be okay. And we can try again. Yeah.
0:07:54 And this has been like a canonical problem in robotics. Yeah.
0:07:59 Because robots are stiff and high energy. So you can never fail. Of course, we fail all the time.
0:08:01 Right. That’s how we learn. So that’s one. Sorry. Okay.
0:08:07 That’s a long answer, but that’s the first one. And the second one is affordability. If you have a
0:08:11 system that costs hundreds of thousands of dollars, you can to some extent defend that if it’s in a
0:08:15 factory working 24/7, it’s still a relatively low hourly rate. Right.
0:08:21 But it’s not going to get you to hundreds of millions of robots in homes so that you can get to actual true
0:08:27 intelligence. So those are the two main things we focused on early on. It’s just needs to be safe,
0:08:31 needs to be very affordable. And then, of course, it needs to get the job done. So it needs to
0:08:34 to be able to do the things. So should we talk about learning?
0:08:36 We could talk a bit about learning. Let’s talk about learning.
0:08:40 I’ve been hearing about foundational models. I’ve been hearing about reinforcement learning.
0:08:46 I’ve been seeing demos of robots that learn by, you know, a human doing a task. And then whether
0:08:51 it’s through cameras or physical sensors, the robot learns from that. What, in your view,
0:08:56 are kind of the most significant advancements in robot learning? I mean, recently, but you can go
0:08:59 back because you’ve been doing this for years now. But, you know, what do you think are some of the
0:09:02 milestones that really changed the way that we develop robots?
0:09:06 Right. So if we do this from the AI side first, because we already talked a bit about how we
0:09:07 develop the robots. Right.
0:09:11 The field has like a history of we started with robots, right? In AI. Right.
0:09:18 If you go back even to early days of OpenAI and like their Rubrics Cube and also their gym,
0:09:26 right? And all of these agentic behaviors with RL, because you get pretty far on this. And it’s a very
0:09:31 interesting way to approach intelligence. Right. I would say that in general, all of the early models out of
0:09:37 deep mind, they’re actually closer, in my opinion, to AGI than a lot of the things we have today,
0:09:40 like in the approach that they’re taking. Okay. Can you dig into that a little bit?
0:09:46 Well, it’s just about how we learn, right? How do you create reasoning tokens and like these long-term
0:09:51 deep reasoning tokens, right? And you can write them out or try to bootstrap them or you can have
0:09:56 them naturally emerge from kind of like whatever reward you’re tuning with respect to how you survive
0:10:00 in the world, right? And say like for humanity, it’s the ultimate one has been basically survival,
0:10:05 right? Right. And I think there’s still this kind of open question of exactly how the reasoning
0:10:13 emerge naturally just from data. And there seems to be a lot more tangible proof there. But then the
0:10:18 realization is we don’t have the data, right? So where do we have the data? Well, we have the internet.
0:10:22 So now if we have the internet, let’s use that data. And then turns out that you can get really
0:10:28 freaking far, right? It works surprisingly well. So that’s really where I think like the last few
0:10:34 years have been. And some canonical, very important problems have to a large degree been solved, right?
0:10:39 Now, what has not been solved is things like spatial reasoning. We say that these language models have
0:10:44 a world understanding and they’re kind of like a world model, but they still catastrophically fail on
0:10:49 very simple things. Like just if I move one meter forward, will I collide, right? It’s 50-50 where it
0:10:56 gets it right. So it doesn’t understand physics that well. And in general, you can clearly see that like
0:11:01 the data we have does not contain necessarily disinformation about the world. What’s also
0:11:08 extremely interesting is that all of these different approaches are merging. And this kind of all always
0:11:13 happens once you go through a full cycle. And we see now in LLMs that they’re using unsupervised
0:11:18 learning, supervised learning, reinforcement learning, and like everything comes together to
0:11:23 solve the problem, right? It’s just the question is mainly in which order do you do this? And a very
0:11:30 interesting part of this is what we call grounding or verifiable truth for your model. So think about
0:11:35 when you have a reasoning model, it’s pretty good at math, actually starting to be amazing at math,
0:11:39 and the same at coding because you can verify whether or not your solution is correct.
0:11:39 Right.
0:11:43 And as long as you can do this, you can apply reinforcement learning really well. Now,
0:11:49 robots are kind of like the ultimate tool to verify truth because ground truth is literally
0:11:54 the ground, right? So like when you have robots in the real world, anything is verifiable. Some of it
0:12:01 might be quite convoluted to get to an answer, but everything is verifiable. That’s how we ground our
0:12:07 reality, right? We have the real world. And I think this is going to be where the biggest impact from
0:12:13 humanoids is going to be. It will progress AI in general, not just physical intelligence. Because
0:12:18 back to what I said, in my mind, they are the same, right? They’re different aspects of the same. So to
0:12:23 more directly answer your question though, I think right now in robotics, the way we approach this is
0:12:28 we have reinforcement learning in the bottom where we learn how to handle our body and how to handle
0:12:36 our own dexterity. And this works pretty well because we can do this on relatively simple tasks,
0:12:41 right? So like if you think about walking, running, just squatting down or sitting in a couch and getting
0:12:46 up or like all of these kinds of things, then the object interactions are pretty simple, right? The ground
0:12:51 is basically roughly flat. Maybe it’s not completely flat. You can model this a bit not too flat, but
0:12:57 it’s pretty simple compared to like you’re peeling a shrimp, which is very complicated with respect to
0:13:02 like how does this object deform? What’s the different forces as is going on? And like we’re nowhere near
0:13:07 being able to simulate that. Yeah. My favorite example here actually is I also do a lot of mechanical
0:13:10 engineering, of course, because we’re making robots. Right, sure. And if I have something simple,
0:13:15 as simple as I’m going to bolt two things together, then I go and run a simulation on this on a small
0:13:22 supercomputer overnight. And what we’re simulating is like 0.1 second of a load interaction on this
0:13:26 bolt circle. Right, right. And then I still go and test it in the lab because it’s probably not
0:13:31 that accurate. Right. So that’s the actual accuracy we have of simulation. So if you want to simulate
0:13:35 complex things, it doesn’t work. But if you want to simulate relatively simple things, it works really
0:13:39 well. And that’s what we leverage currently in reinforcement learning. Right. So that means the robots can learn
0:13:45 to sit, to squat, to run, to walk, to support its own body in a good manner. It can also learn some dexterity
0:13:52 for like simple objects. So like how do you pick up all kinds of different shapes and handle them with
0:13:57 like in-hand dexterity. Here’s extremely powerful because you can do this completely autonomously with
0:14:02 no people involved. But in the end to do this, you now have to move to the next paradigm, which is how
0:14:05 do you do the same thing in the real world. And that’s something that I’m very interested in these
0:14:10 days and we’re working a lot on, which is learning in the real world. Yeah. And also with RL.
0:14:17 So I call this kind of like broadly learning from failure. Okay. And what it is is you have a model
0:14:22 which involves a good amount of reinforcement learning in the bottom to learn how to handle
0:14:27 its own dynamics. Then you have some expert demonstrations from humans, which is typically
0:14:31 teleoperation data. You of course also have all of the internet because that’s how you get a better
0:14:33 understanding of a lot of things around us. Right.
0:14:35 And the data is there. Right.
0:14:40 And then this allows you to get to where you have a task that is non-zero in success. Right.
0:14:45 So you tell the robot to do something and the robot is able to sometimes succeed. Right.
0:14:49 And at this point, you can actually close the loop and you can say like, hey, now I only need to know
0:14:52 whether it succeeded or failed, which is a way simpler problem. Right.
0:14:58 And then the robot can learn. But again, this requires you to have a robot that can fail. Right.
0:14:59 And it can fail gracefully. Yeah. Right.
0:15:04 And once you can do that, you can just think about a robot opening the fridge and it’s like,
0:15:08 it’s hand slips on the handle, tries again, but oh, it moved a bit because the fridge is stuck in
0:15:12 vacuum. So like it had to like shuffle, it feeds the feet a bit to not fall over. Right.
0:15:15 And like it grabs again and like, oh, there, I opened the fridge. Yeah.
0:15:19 So now you have two examples of what not to do and one example of what to do. And as you keep iterating
0:15:22 on this, you can kind of like self-improve the system. Right.
0:15:25 Which is incredibly interesting because teleoperation does not scale. Right.
0:15:31 Teleoperation is extremely interesting to bootstrap your way into non-zero chance of success.
0:15:34 But you’re not going to be teleoperating millions of robots. No. Right.
0:15:39 So it is just a tool. I think it has a bad reputation. Everyone’s always saying like,
0:15:43 oh, is it teleoperated? Is it AI? And I understand that because it’s interesting to see how far have
0:15:47 we managed to push the AI. But I actually look at it the other way around. If you can teleoperate it,
0:15:50 we can do it with AI. That’s just a question of data. Yeah.
0:15:57 And what really is exciting about this is how can you take all of that data and create this data
0:16:01 flywheel that just keeps running. And my dream, of course, right, is that you could come back in a
0:16:04 million years and the system is just better. Right. Right.
0:16:06 And we’re not quite there yet. Yeah.
0:16:07 But like that’s where it needs to go. Yeah.
0:16:13 I want to shift gears a little bit and ask you kind of about the societal aspects of robotics and
0:16:19 humanoid robotics. There’s been, I mean, even going back a couple of years when LLMs and LLM chatbots
0:16:22 came out and everybody started talking about, well, what does this mean for work? What does this mean
0:16:29 for my job as a writer or an accountant or a lawyer or what have you? How do you think about humanoid
0:16:33 robots from the human perspective? Right. Because we’ve been talking about how the robots learn and
0:16:37 how important it is for them to be out in society, picking up those nuances. What’s been your experience
0:16:43 and how do you think about how humans react to humanoids? And specifically in the context of work,
0:16:47 where robots are coming in and, you know, depending on the job, paying the situation,
0:16:51 I would imagine there’s a range of reactions from folks, but there’s got to be a little trepidation,
0:16:54 right? Yeah. Yeah. So let me start with what we’re actually doing, right? Yeah.
0:17:00 So right now, 2025 for us is about consumer. And there’s going to be a lot of people this year
0:17:06 with neos in their homes. And I can’t wait to give people back 2.3 hours, roughly. That’s the average
0:17:12 of work you do at home every day to just keep your house tidy and like all of the chores around the
0:17:16 home. Right. Which people generally do not want to do, right? Yeah. You’re finally back home from
0:17:20 work and you’d like to spend time with your family and now you need to do the laundry. Right. So there,
0:17:25 we’re not so far at least meeting any challenges. That’s mainly people being very happy. Right. Right.
0:17:30 But it points to a broader thing in society, right? Which is, what do we actually want to spend our
0:17:37 time on? What makes us happy? And I think, first of all, we need to ensure that we can keep the
0:17:42 standard of living we currently have and preferably grow it, right? Yes. And we’re having this enormous
0:17:46 shortage of worker coming because we’re not having enough kids. Right. I think it’s going to be a long
0:17:53 time until we kind of run out of things to do, even if we change nothing. Yeah. But I think also,
0:17:58 it’s a great opportunity for us to be able to focus on really what matters and like what makes us human.
0:18:04 Right. So I have this personal experience on this many years ago. It was back early in college and I
0:18:11 had a job just going from house to house for elderly and cleaning. Okay. Just as a side gig. And so many
0:18:16 of the people I came to, they’d already cleaned, even though like they were very old and frail because this
0:18:21 was their 20 minutes a week where they got someone to talk to them. Right. So they would just like have
0:18:25 coffee and chocolate ready and like, Hey, sit down and talk with me instead. Yeah. Yeah. I mean,
0:18:31 it’s obvious that that job should have been done by a machine so that people could spend time on
0:18:36 interpersonal things. Right. Right. And this is just one example of how I think like how we view
0:18:44 the value of humans will change throughout society once we start moving towards an actual true abundance
0:18:48 of labor. Right. Because that is what we’re talking about here. We’re talking about the same kind of
0:18:53 shift that we saw with electricity a couple of hundred years ago. Right. Where you move from,
0:18:58 this is just never going to like humanity is never going to master energy. Right. That sounds like
0:19:02 implausible to like you flip the light switch in the morning and you’re annoyed if there’s no light. Right.
0:19:07 Right. The same is going to happen with labor. But I think it is really going to enable us to
0:19:13 reassess what makes us human and what makes us valuable. And I don’t think it is basically reducing
0:19:18 people in quote to robots in like repeating certain labor so that we can get products and goods and
0:19:24 services. And I think so far up through history, we have never run out of human creativity. I think
0:19:27 it’s Jensen who says this, right? It’s a good way of saying it. Yeah. Yeah. Yeah.
0:19:33 We will find new ways to do or new things to do and how to add value to each other. And I think there
0:19:37 will be higher value than what we have today. But always when this happens, of course, there’s a
0:19:43 transition period. And that transition period can be painful. Right. And I think that’s a big one that
0:19:47 we as a society need to discuss and come up with good solutions to because any kind of like large
0:19:52 technological shift will mean that people need to change. You need to adapt. You need to adapt. Yeah.
0:19:57 Adapt is a good word. Yeah. When is NIO shipping? You mentioned NIOs being out in people’s homes,
0:20:02 doing the labor. I’m going to be kind to my team and not say exactly when it’s shipping. Yeah. But
0:20:08 it is shipping this year. Sure. And so when you talk about and can you talk about the things that
0:20:14 NIO will be able to do in homes? Sure. So, I mean, I can go through the things that we’re doing right
0:20:19 now to be very concrete and then I can talk about what we plan to do. Perfect. So today what works
0:20:24 really well is vacuuming. And you can see that if you go to TTC to our booth, you can see the robot
0:20:28 vacuuming. It’s actually very good because you can move things around. Right. Right. Et cetera,
0:20:34 et cetera. It can tidy. It can do the laundry. It can fold. It can fold. Not everything a hundred
0:20:38 percent yet. Okay. Okay. But we’re getting there. I’m pretty confident we can do that and do a good
0:20:43 job on that. Folding is one of these very interesting things, by the way, where when you, let me first
0:20:48 preface this. Everything I talk about now are things that we are doing partly with AI and partly with
0:20:54 Telia. Okay. And the plan is to make it autonomous, but we’re going to do that together with you guys as
0:20:58 customers. And I want to talk about that. Okay. But this is basically what is the platform capable
0:21:03 of doing, right? And folding is just this very interesting thing that in classical robotics is
0:21:08 almost impossible to solve because it’s like a deformable object that every time you put the
0:21:13 shirt down, it looks completely different. Right. And for some reason, this just works with AI.
0:21:19 It’s actually not so hard to automate folding. And I can’t actually tell you why. I’m not quite sure,
0:21:22 but like, that’s one of the things we found out pretty early that like, you know what,
0:21:27 it actually works really well to automate. Amazing. Yeah. Yeah. But it is classically
0:21:32 seen as extremely hard. Right. Folding today, common cold tomorrow. Yeah. So, but then of course,
0:21:37 I’m also very excited about just like having something around the house that you can interact
0:21:42 with and I can help with small things. I had this magical experience the other day in my home where
0:21:49 I was just sitting there. I had, I was actually a candidate for a job coming over and in my house and
0:21:54 we were just sitting there talking and the robot is going around doing his things. Right. So,
0:21:59 it’s like wiping the counter, I think at this point, and then my food comes and I just tell the robot,
0:22:03 like, Hey, Neo, can you get the door? That’s my food. And Neo goes over and opens the door,
0:22:09 gets the food package from the delivery guy. He was kind of shocked and then goes over and puts it on
0:22:14 the counter for me. Right. And you could say like, Oh, that’s not that valuable. It’s kind of gimmicky,
0:22:18 but the sum of all of these things. Right. Is extremely valuable. Right.
0:22:22 So I’m also very excited to see how people will use it because I think this, this will evolve over
0:22:28 time. And we can talk about like the canonical tasks, like laundry, vacuuming, cleaning, tidying,
0:22:30 like these kinds of things. And they’re all there, but yeah.
0:22:33 Yeah. No, I was thinking to your point earlier though, that, you know, during this,
0:22:37 when the robot’s cleaning the counter and it goes against the food, you don’t have to break
0:22:39 your conversation with the candidate, right? No.
0:22:43 You’re spending that time as you were saying, well, what’s really important to us? I want to
0:22:47 connect with this person and see, are we going to work together? Do we have things in common,
0:22:51 et cetera? Yeah. That’s, that’s fantastic. And it also points to how like almost everything we do
0:22:57 that is physical labor is social. Like whenever you do something, you are kind of also navigating
0:23:01 a social situation. Even if I just say, Hey Neo, can you get me a Coke in the fridge? Maybe someone’s
0:23:06 in the kitchen making food, right? So you’re navigating this social situation to do this. And this is not
0:23:09 just through in the home. It is also through all out through enterprise, right? Right.
0:23:14 Whether you’re in like a retail or a hotel, or even most factories, there’s a lot of people.
0:23:19 There’s other people. Sure. Yeah. So, so this is why the home and consumer has to happen first
0:23:23 before we go into all of these other markets. Right. But back to how this works, where I think
0:23:28 it’s quite important, is that if you are one of the people who buy, and I hope you will, right,
0:23:35 buys a Neo in 2025, we are selling almost more a journey than a destination. So it’ll be,
0:23:38 do you want to be one of the first ones to ever have this at home? Right. It’s going to be a
0:23:42 transformational product, right? And it’s something, at least me and I know many others
0:23:46 have dreamt of all their life. Yeah. And you get to be part of this journey together with us. Right.
0:23:52 Where we develop this towards an actual abundance of labor. And not everything will work day one,
0:23:56 but it will get better every day. It’s going to be a lot of fun. And we’re going to treat you
0:24:00 really well. And we’re going to respect your privacy. And we can talk about that too. I think that’s a big
0:24:05 topic and really make this into a journey that you will never forget. And it’s almost
0:24:09 like adopt a Neo, right? And I’m not saying it’s not going to be useful because it’s already very
0:24:13 useful. So it will be useful and it will reduce the amount of work you do in your home every day,
0:24:17 but it’s not going to do everything for you. Right. That’s not day one. Yeah.
0:24:25 And some will be autonomous. Some will be tally up and sometimes it’ll fail. And that’s something you
0:24:30 have to be okay with. And I think expectation management here is extremely important, right?
0:24:37 So because we cannot do this alone. We have to do this together with all of you out there and kind of
0:24:41 collaborate, right? On how do we teach these robots how to behave in society.
0:24:48 Are there things that the end user and I’m thinking of the consumer setting will be able to do to sort
0:24:53 of intentionally try to teach the robot things as opposed to, you know, just interacting with the
0:24:58 robots learning, but like, will the user have the ability, I don’t know, almost like, I was going to say
0:25:03 like a dev kit, but that’s not quite right. No, I get what you mean. Let’s say to be very specific,
0:25:08 right. Can our customers fine tune models? Okay. Perfect. What is it? Most likely the answer to that
0:25:13 is yes. Okay. We have some stuff that we’re very proud of on that and we’ll see when we roll it out.
0:25:20 Okay. I think it’s going to be incredibly important to enable people to use this general robot kind of
0:25:25 almost in any manner you want, right? Sure. Because the creativity of the community is going to far
0:25:31 outweigh like the creativity of the company. Right. And just putting it into a few homes, we’ve learned so
0:25:35 much. Yeah. I can imagine. Like maybe if we’re very good, like half of our assumptions are
0:25:39 correct. The other half is just completely wrong. And that’s just the way it is. Right. And then you
0:25:45 iterate. Before we wrap up, I want to ask you about the world model challenge, and maybe you can speak
0:25:51 a little bit about the importance of world models when it comes to robotics and robots learning. Yes.
0:25:57 So this one is a bit harder. So, you know, when we talk about AI, it’s very easy to explain like,
0:26:03 hey, the robot learns something through demonstration. Then it learns something that keeps improving it
0:26:07 through RL. And it gets really good at getting you a Coke in the fridge. The world model is a bit more
0:26:13 abstract, but it’s very connected to how this happens. So to me, what a world model is, is to a large
0:26:21 part the ability to predict the future. Right. So given that I know my own state, right? So I know
0:26:25 like where are my fingers where I’m like, I basically have the positions of my joints and everything. I have
0:26:30 the forces I’m interacting with on the world. Like if I’m holding something, I know how heavy it is. I know
0:26:37 what I see. I know what I hear. I know what I feel. Now, if I take an action, what will happen? And you can picture
0:26:43 this as like a tree of probabilities, right? Like all of this, like for simulating forward in time,
0:26:48 what is the potential things that can happen? Now, this is very interesting because it allows you to
0:26:54 do things like validation and figure out like, am I doing the right thing? And that’s kind of like the
0:27:00 first way you use world models practically is to be able to help the robot improve because you know what
0:27:05 might or might not succeed and whether it succeeded. But I’m just validating where your model is better
0:27:09 than the previous model. But where it gets really interesting to me is that this actually makes you
0:27:17 able to search backwards from the goal. So if I have this perfect simulator of the world, and this
0:27:22 simulator would simulate how to peel a shrimp, right? Because it actually has the data of all the force
0:27:27 interactions and the vision and even the tactile, the audio of like the sound while this happens,
0:27:33 it has all the information. Then that basically becomes a search problem for how to get to an
0:27:39 intelligent policy. And I think this lies a lot closer to how human intelligence works. But the
0:27:43 very interesting thing about world models is that they just scale incredibly well. So if you think
0:27:49 about robotics in general, we don’t necessarily have the model architecture yet that allows you to scale
0:27:56 to like infinite scale. But a world model is just incredibly efficient in taking all of the data
0:28:00 and turning it into something useful. And then our remaining problem is now, okay, you have this
0:28:06 incredibly powerful tool, which can basically simulate the future. How do you use this to do
0:28:12 something very useful on the robotics side and also generally on the intelligence side? And I think
0:28:16 there’s a lot of exciting work there. It’s not fully figured out yet, but there’s a lot of early
0:28:22 applications which are very powerful. And I think a lot more will come in the coming years. But there’s a lot of
0:28:27 different definitions of what a world model is, right? Sure. So that’s what I’m saying. Like to me,
0:28:33 it is basically the ability to simulate any kind of future given the actions you want to take and your
0:28:39 current state. And interestingly enough, if you look at our world model, the data and computer efficiency to,
0:28:44 for example, understand physics is just off the charts, right? Because the robot actually has the data.
0:28:50 And this is something you see is very much missing in what I would call maybe the closest to world models
0:28:55 that you could see publicly today, which is the video generation models, because they are essentially
0:29:00 a subset of a world model. But they really struggle with physics, especially when it gets complicated.
0:29:04 Like how does a human move and interact with the world? They really struggle. This, of course,
0:29:10 is completely different when you have a robot that actually has spatial data. It is learning in 3D.
0:29:15 It is learning directional audio. It’s learning tactile, all of these things, like interaction
0:29:20 forces. Yeah. Oh, it’s fascinating. You weren’t kidding. You have a fun job. This is fascinating
0:29:24 stuff. And I don’t know. I, of course, always start to think about it and then I immediately go to the,
0:29:30 like, robot in my house wrote my job and kind of think about that stuff. But thinking about just the
0:29:33 way you’re approaching these problems and that idea of searching backwards from the goal and just these
0:29:37 different ways you have to think about the problems is, is fascinating.
0:29:42 Bern, before we let you go, for people who want to learn more about any, everything you’ve been
0:29:46 talking about, about the company, about the robots, is the website the best place to go?
0:29:52 Yeah, I’d say so. And then follow us on X. We, uh, pretty present there, but I think we have an
0:29:57 amazing website and it tells you a lot about the product. And then of course, if you have that
0:29:59 opportunity, stop by our booth. You have to see it in person.
0:30:05 You have to see it in person. And there will be a lot of places to see Neo in the coming months.
0:30:06 I can only imagine.
0:30:10 And I can’t wait to let everyone actually experience it because it is a product that’s
0:30:17 very hard to visualize with words and even visualize with video. It’s something you have to touch.
0:30:21 Excellent. Well, I can’t wait to interact with the Neo. It’s going to be an exciting year,
0:30:27 to say the least. Bern Bornik, thank you so much for taking the time to talk with us. And I very much
0:30:30 look forward to following your progress, following the launch of Neo and, uh,
0:30:32 maybe catch up with you again down the road.
0:30:34 Awesome. It was fun.
0:31:02 Thank you so much for listening to me.
0:31:04 Thank you so much for listening to me.
0:31:05 Thank you so much for listening to me.
0:31:06 Thank you so much for listening to me.
0:31:07 Thank you so much for listening to me.
0:31:09 Thank you so much for listening to me.
0:31:10 Thank you so much for listening to me.
0:31:11 Thank you so much for listening to me.
0:31:11 Thank you so much for listening to me.
0:31:13 Thank you so much for listening to me.
0:31:14 Thank you so much for listening to me.
0:31:14 Thank you so much for listening to me.
0:31:15 Thank you so much for listening to me.
0:31:17 Thank you so much for listening to me.
0:31:19 Thank you so much for listening to me.
0:31:23 Thank you so much for listening to me.
0:31:23 Thank you so much for listening to me.
0:31:24 Thank you so much for listening to me.