This Humanoid Robot Just Got Human-Level Hands. Here's the Man Behind It | 1X, Bernt Børnich

Intro

00:00So I think accepting failure and having that as part of your culture is incredibly important because if if not you're not innovating right. Most of the things you do that has never been done before will just be plain out wrong. In hindsight it might not just be plain wrong.

00:16It might be like borderline stupid. You're like how did we ever think this was going to work? But that's that's just the nature of the game. That's that's how innovation works. I think the only thing that's not acceptable from a culture point of view is you didn't really try.

00:27you didn't give it everything you had or you didn't actually reflect over why did it fail so you can learn. It is how we make progress. Can you do this under pressure? Can you keep this culture where failure is okay even when there's pressure?

00:42And if you manage to do that then you get great innovation. Focus on the things that make you happy. It's going to be a long journey and if you're not having fun you're not going to make it. There's this notion of like being a founder is like grind, grind, grind, and it is and it's going to be a lot of grind and it's going to be a lot of dark days, but there has to be also a lot of fun to make it worth it.

01:00The best cheat code I have is pick a problem that really excites people. We do humanoid robots. It's really freaking cool. [Music] Thanks, Neil. So, my name is Bar Borick. I'm the founder of OnX and we make humanoid robots for the whole. We thought about it for a long time.

01:29We just got really annoyed that everyone was posting robotic videos online with like 4x, 5x, 6x speed because robots generally don't move naturally, don't don't move at human speeds. We just thought it was really funny like let's make sure like in every frame there's a 1x.

01:43And our mission is to create an abundance of labor through these intelligent machines. How life will be when you actually have a humanoid in your home when you can really focus on enjoying the time we spend together, right? And the things that make us human [Music]

When a Humanoid Robot Becomes Your Roommate

02:03when I was a small kid. I got a hold of my first computer. Kind of like clicked, right? And that was really magical to me. You can write some code and you can have something else move in the physical world. You can combine these two. As I kind of started looking at everything around me and like how how things that move very efficiently get things done and like how much the physical world matters.

02:21I decided I want to make humanoid robots. I got really inspired by Honda Asimo back in the day. Most magical thing about Asimo was the interactions with people. It wasn't an industrial automation system. This was actually kind of straight out of Star Wars, right?

02:34It's this human robot that can walk around, run around, hand you a bottle of water and make sure that all these chores we don't want to do every day. Like we don't need to do them. We can have robots for this, right? That's really what also motivated me when I started 1x.

02:47Just look at why did Asimo actually fail. So if you look at Asimo, they followed what I would call like the classical robotics regime, the same as we see mostly in factories and work cells of how you build this. And it's not really inspired by nature and how we humans are built and how we humans move.

03:03This made it very hard for them to actually be able to do useful things outside of the lab because the system they built really had to make a lot of assumptions about the environment which doesn't work in the real world, right? Because in this creative chaos around us in everyday life that is where so much of our intelligence comes from and our lessons learned.

03:19You cannot really make that many assumptions. Things change all the time. We have to make it safe. We make have to make it affordable. But it can't be a toy. It actually has to do the work. It has to be very useful if we want these machines to be truly intelligent.

03:31Also, if we want these machines to behave in a manner where they're aligned with us, they have to live and learn among us. So, 1X was started about 10 years ago now. Really, it started out from the perspective of how can we make humanoid robots that can actually have a real impact on the world.

03:48Norway was very good in the beginning. We had very little attention. We were all by ourselves. It was just a group of people coming together, spending every waking hour figuring out how to solve the solve the problem. We moved a lot of talent over to Norway.

04:02The field at that point was very small. Thinking about like who was like really the best people in humanoid robotics back in like 2015 2016, right? There wasn't that many people. I remember still when like on the humanoid robotics conference we can like fit everyone in one room.

04:13This created this kind of like very exciting ecosystem where everyone was basically almost living together and just living and breathing this problem really became this great group of friends that just went on this adventure together. We were very early in this space.

04:26There weren't that many believers. It's always challenging to raise money. The space isn't really something people believe in yet. Around like 2017, I would say, is where we struggled the most. We actually had fundraising locked in. We were one signature away from maring money and then COVID hit.

04:42And that's probably the most painful thing I've done as uh as a founder. So, we went down to half the number of people to get through CO because of course in the beginning of CO there was no raising money. Everyone was just sitting on the fence figuring out how what happens now.

04:55And of course these were like people who had moved to Norway to be part of this like giving up everything they had and like they want to be part of this dream and they were all in right and having to let go of these people are of course extremely extremely painful.

05:09There are two ways we lose right either we lose velocity or we run out of money and those are the two things we need to make sure never happens. The first year was really about proving out can this be done? Really lying the foundations for a new paradigm in how we design robots.

05:25We decided to go in a new direction, right? A different paradigm in robotics. It's not that no one has ever done tendon drive systems before or like cable drives, but no one has really worked deeply enough and long enough on the problem to make it work.

05:37It's not easy to catch up because you need to sink a lot of work into this to make it work. It's also pretty exciting because it gives you a real moat.

Why Building Humanoids Is a Whole New Level

05:45If there is a lot of energy when I move, then when I step on the ground, there will be a huge impact and this will disturb me. It will disturb the ground. Like, it's not a good idea. So, you want to make sure there's as little energy as in this as possible.

05:56It actually just comes back to kinetic energy. We have a very good intuition for this. We learned this in school. If a car moves twice as fast, it's not twice as dangerous, it's four times as dangerous because it falls to square. And this is also true for robotics.

06:11So if you think about the traditional industrial robots, they typically have gears that are about 100 to1 gear ratio. So like if your arm is moving like this, something inside here is spinning 100 times faster. And there's just an enormous amount of energy in that rotation.

06:24So you can think about it, something here is spinning at 20,000 RPM and then when your arm hits something, this needs to immediately stop. There's no way it can immediately stop. It's going like at a blazing speed, right? It can't immediately stop.

06:36And everything we do when we interact with the world is collisions, right? Whether we're taking a step or whether I'm just touching my my watch or whether I'm picking something up, it's all collisions. And the way this is typically solved in factories is that you know exactly where things are.

06:50So you will see the robot move and it will kind of stop just before it touches the world because it needs to touch the world very slowly. And this works amazingly in factories and has kind of been a groundwork for like robotics working well over the last 60 years.

07:03But if you're in a home or in a garden or whatever, you don't have a calibrated factory. So you don't know exactly when to stop and that's why we need these very low energy systems that can be safe both with respect to people but also with respect to itself and the world.

07:17You don't want your robot to damage your furniture, right? And of course you don't want the robot to hurt you. But if robots are going to be able to live and learn among us, they need to be able to explore in the world. They need to be able to learn through trial and error.

07:27And that means you just you need to be very low energy. You need to be soft. You need to be compliant. And humans are just an amazing example of this. When we did the early first deployments years ago, right? And we tried this out in homes.

07:38The thing that really struck us very early was how almost everything you do is social. So even just getting something in the fridge is a social act because likely there's someone in the kitchen. And now you need to clearly kind of communicate your intent.

07:54I'm going to go to the fridge and open it. Make sure you're not in the way. Make sure you do this in a safe manner. And you can't really separate these two problems. Like any kind of labor that you do among people is kind of social labor. Really, intelligence comes from diversity.

08:06That was a great insight that we had pretty early on. When you flip the light switch in the morning, there's light. And if not, you're pretty annoyed because humanity has mastered energy for practical purposes in everyday life. It's just abundant.

08:19And this same thing is going to now happen to physical labor. And this is really needed because there's not enough people getting born. We don't have enough people to take care of our elderly. And we see like prices of goods and services increasing.

08:32Everything is inherently limited by our ability to effectively serve labor. The question then becomes how do we get there as quickly as possible. So most of the human robotics companies you see today they are more component integrators buying off the shelf and integrating into a system.

08:46And this in itself is pretty hard. But the journey we set out on here with the tenant drives and our unique motors and everything else. This means that we have to do everything oursel because these components don't exist. So we spent the last 10 years right building not only the foundational technology but also actually the machines that can build these components and automation equipment and everything needed to build a factory.

09:11It's going to be extremely challenging because manufacturing always has but we've done a very good job in simplifying the product. Make sure you don't have anything that requires special alloys. Make sure your product is very light so you don't need that much material.

09:25Just simplify, simplify, simplify. Minimize part count through this. Reduce this from something that has the complexity of a car to something that gets closer to having the complexity of let's say like some kind of electrical appliance in your house, right?

09:38And the way we do that is just building it all oursel having our engineers sit in the factory make sure design, manufacturing, automation, everyone's in one room, really create a very, very efficient process. This is back in 2018. I was at a stage

In This Industry, Failure Comes First

09:56and this was the old robot. He was one of the first prototypes we had and we were opening this health conference where we're talking about robots in healthcare in the long term. And I'm standing on the stage together with my robot and exactly well timed as I say safe, the robot decides to accelerate backwards, hit the wall and then face plant next to me.

10:16Bring with it like all the balloons, those in the back and just it just looks like it had a really rough night like sleeping in the balloons. I spent a lot of these things up through the years. things don't always go the way you planned. So, I think accepting failure and having that as part of your culture is incredibly important because if if not, you're not innovating, right?

10:37Most of the things you do that has never been done before will just be plain out wrong. In hindsight, it might not just be plain wrong. It might be like borderline stupid. You're like, how did we ever think this was going to work? As you get more knowledge, but that's that's just the nature of the game.

10:52That's that's how innovation works. So, I think first of all, like foster a culture where failure is okay. I think the only thing that's not acceptable from a culture point of view is you didn't really try. You didn't give it everything you had or you didn't actually reflect over why did it fail so you can learn.

11:06If you give it everything you have and you learn from your failures, then you should embrace and celebrate failure, right? It is how we make progress. And this is hard to do actually. It's it sounds like it's the cliche, right? And it sounds like something that should be pretty straightforward.

11:20But of course, we're under a lot of pressure to make this happen and we need to deliver. We have timelines. We have schedules. We need to hit our manufacturing milestones. That's really where the importance of this comes in. Can you do this under pressure?

11:34Can you keep this culture where failure is okay even when there's pressure? And if you manage to do that, then you get great innovation.

How Robots Learn

11:43Let me go through like how does the robot learn in general? How does it actually work? It begs the question, how do you get to a system that's intelligent enough that when you ask it to go and get a Coke in the fridge, it at least manages to do it sometimes?

11:52And we need to bootstrap. So you start with internet data of course because we have a lot of it. Then you have some synthetic simulated data and then you need some robot data and to get that robot data we typically use teleoperation. So that means we have a human that actually embodies the robot and you see through the eyes of a robot and a robot moves like you move and it's a pretty magical experience.

12:12It's kind of like hey my hands are somewhere else and I can do something anywhere in the world and it's a very nice way of transferring knowledge from a human into a machine. And once you have a bit of this, you're now able to do these tasks autonomously.

12:23And from there, you can kind of like iterate on it and learn from the real world. In the end, it's about having enough of these robots out in the real world. Learning from just trying things. You have to be able to experiment. You have some hypothesis about how to do something.

12:36You try. You see how it went, you try again. That's how we learn, right? [Music]

Build Before the Market Exists

12:44Traditionally, scale does not happen first in enterprise. And this might be slightly surprising, but if you look at the history of highly innovative products, they almost never happen in enterprise first. They happen in consumer. And this is just because enterprise is risk averse and there are just too much red tape and barriers, right?

13:04From IT departments to labor unions to risk averse CEOs, it just takes a lot of time. While if you have good product market fit, nothing scales the way consumer adoption does. This is true for a lot of products. There's a very good example of this is chat GPT and the roll out of digital AI.

13:21Open AAI really tried in enterprise for a long time and they couldn't really get it working and then they released chat to see what people do with it. People figure out this incredible diverse set of like tasks they can do with this tool and they bring it to work and now you would think that then finally you succeeded but you don't.

13:37Now enterprise actually says no no you can't use this here. This is new and dangerous. We can't do this. After a while, people actually get so annoyed that they kind of say, "Oh, if I can't use it here, I'll go work somewhere else where I can use it because this tool makes me so productive."

13:50And now you created so much bottom-up pressure that you're forcing adoption. And now you can start top down and do great in enterprise. But you have to create that forced adoption through bottom-up pressure. Juno robotics is no different. If you want to do this in the next few years, not in the next few decades, you have to go through consumers.

14:05You have to find your early adopters and your true believers that can really help you push this forward and be part of this journey. Then of course it will be used in enterprise because it's going to greatly improve your productivity. But it has to happen in consumer first.

14:20If you run a deep tech company, you're

Lessons for Deep Tech Founders

14:22not really sitting that close to your customers in the beginning. You're trying to solve this fundamental problem that if you solve it, your market is there. It's like if you cure cancer, you're not wondering whether or not people will actually buy your product.

14:35And if you figure out a way to take energy and turn it into any kind of labor, any kind of product and service, clearly your market will be there. It's more a question of can you actually solve the problem. And of course, you get into product market fit questions once you start to deploy this and how you make it like gradually useful because the problem you're working on is such a big problem that you can't really just say like I'm going to just sit here in my lab and work on this for 10 years and then I'm going to go out and like launch this product.

15:00First of all, you of course want it to be aligned with humans and how we want our robots to behave around around us. But also, of course, you need to show it's useful. You need to create revenue. You need to build a business because this is a really long journey.

15:14But fundamentally, it is a deep tech type journey where it's more about solving the problem than really ensuring you sit close to your customers. And once you solve the problem, then you can start caring about how all those small details, right?

15:28that takes this from being just a solution technically to something that's like packaged as a great product. We are just starting already this year hopefully we're going to have something that is very useful but in the next few years this will change the way we live.

15:48Focus on the things that make you happy because it's going to be a long journey and if you're not having fun you're not going to make it. I think that that's that's undervalued, right? Because there's there's this notion of like

Build Something That Excites People

16:00being a founder is like grind grind grind and it is and it's going to be a lot of grind and it's going to be a lot of dark days where you just need to like pull it together and just get it done. But there has to be also a lot of fun to make it worth it.

16:11And uh that just means work on the interesting problems, right? The best cheat code I have is pick a problem that really excites people. If you want to find the best people in the world to come work at something, do something that excites people.

16:22We do human robots. It's really freaking cool. And most people come in and they like say hi to a robot and they're like, "Oh man, kind like I want to be part of this. I want to do this." So whatever it should be, right? Do something that matters.

16:34As long as I get to do this, I think I'll be pretty happy. [Music] [Music] la. Wow.