The Robotics Stack
00:08:13 · Xara
[00:08:13] S2: It's less than $0.5. So dollar roughly... You're looking at 177 meter dollar. It's already 46 kilogram, but Unitree is only 1.3 meter. It's only 130 centimeter. It's very, very short. So, obviously, because the whole thing is metallic, it's it's very, very heavy. Right? So third thing is that we observe the doll market is that it's very oily because usually the oil would come out of silicone. Right? If you use the silicone doll for extended period of time, the oil actually comes out of it. It's very disgusting. So when user interact with it, the oil might stick to human... The user's skin. So observe it that this is a huge market. We decided to build an Intimacy Robot. You might wonder how do we do it? What's the technology? What's the software behind it that makes it possible? Let me show you a few things. Once we decided to build an Intimacy Robot, we actually found a real point star, put her in a motion capture suit. So motion capture, as you know, is very commonly seen in animations and in robotics. We actually capture 160 different sex move, and we wanna make sure we have any data from any sex positions to any other sex positions. Because we need to make sure the robots can change from any positions to any other positions, so we recorded a lot of data. And in simulation, using reinforcement learning and imitation learning, we teach Unitree how to have sex in simulations. This is our last accent. And as you see, the Unitree is doing all kind of fancy stuff. And once it's ready, we implement onto Unitree platform. And on top of it, we add a voice agent so that you can voice control it. Hey, baby. What's going on? Hey, baby. Just waiting for you. What do you wanna do, gorgeous? Come to my bedroom. Okay. I'm walking to the bedroom now, my love. So as you see, the whole thing can be voice control. I asked you to come to my bedroom. It's able to use internal navigations using SLAM. It's gonna navigate inside my house. It's gonna recognize the bed, and it's gonna climb onto the bed. It shakes its butt, lays down, and flips itself, and so on and so forth. The whole thing was done with voice control. So, basically, you can use voice command to ask it to go from any positions to any other positions, stably on bed. To say reinforcement learning, we're able to control the robots to do any move from any position to any other position.
00:10:07 · Xara, Leo
But, obviously, we cannot sell Unitree. Right? [00:10:10] S1: Does it run on device or you rely on the cloud to do that? [00:10:15] S2: Everything was run on the edge. We actually have a one unified policy in order for the robots to do any positions from any to any. Let me show you a few things. That's a sexy dance. At least we try. It's a lap dance. Obviously, it's one of the first robotic lap dance. So, obviously, it's an interesting dance, but we cannot sell Unitree because Unitree has three issues. Number one is that Unitree is also very, very heavy. It's 40 kilograms. Right? It's not a very light machine. Second thing is it's only 1.3 meter, which is a 130 centimeter. It's too short for any kind of intimacy robotics. It's a children size, which is not very good. Third thing is that who would have desire? Who would get interested in a metallic or steel looking robots? Right? So that's why we decided to build our own hardware. We call it Somnia. Somnia is actually one of the first intimacy robots on the planet right now. Somnia, in Latin, it means dream. That's why we wanna become people's dream companion. Our whole company is called Somnia Lab. We are a lab. We are a company that builds dream. And let me show you a few very crazy things. So this is this is our own hardware. It's 175 centimeters, 1.75 meters, but it's only 20 kilogram. Let's see it doing a doggy positions. So as you see, obviously, it's a robot with the black hat and white all all this kind of stuff. But if we move forward a little bit, it could look very, very natural.
00:10:15 · Xara, Oskar
[00:10:15] S2: Everything was run on the edge. We actually have a one unified policy in order for the robots to do any positions from any to any. Let me show you a few things. That's a sexy dance. At least we try. It's a lap dance. Obviously, it's one of the first robotic lap dance. So, obviously, it's an interesting dance, but we cannot sell Unitree because Unitree has three issues. Number one is that Unitree is also very, very heavy. It's 40 kilograms. Right? It's not a very light machine. Second thing is it's only 1.3 meter, which is a 130 centimeter. It's too short for any kind of intimacy robotics. It's a children size, which is not very good. Third thing is that who would have desire? Who would get interested in a metallic or steel looking robots? Right? So that's why we decided to build our own hardware. We call it Somnia. Somnia is actually one of the first intimacy robots on the planet right now. Somnia, in Latin, it means dream. That's why we wanna become people's dream companion. Our whole company is called Somnia Lab. We are a lab. We are a company that builds dream. And let me show you a few very crazy things. So this is this is our own hardware. It's 175 centimeters, 1.75 meters, but it's only 20 kilogram. Let's see it doing a doggy positions. So as you see, obviously, it's a robot with the black hat and white all all this kind of stuff. But if we move forward a little bit, it could look very, very natural. [00:11:25] S0: And that's the intern there, or that's you?
00:11:27 · Xara
[00:11:27] S2: Well, actually, I did the whole shooting myself. You know? I'm the I'm the founder of companies, so I gotta do all the shooting. So I I actually took the responsibility and I was interacting with it and actually videotape the whole thing. And here's our robots from another perspective. Let me show you. This is a robot moving in in a front and a back. So on top of the robots, we actually attached the skin. The video on the right was shoot by me too, and I did it on the... My old office. I was having my, phone pointing at the robot's butt. The whole time, I was laughing very hard. I was smiling really hard because when the robot was on top of me riding me, I think I saw the future. Right? We believe that if the robot is beautiful, is a nice body shape, humans are gonna interact with robots. So after a while, I took out my legs, and I asked the robots to move into a darker positions by voice control. It moves down. It it start moving its ass faster and faster. This is from a first person perspective. And in order to make the experience interesting, we actually have sensor in the robot's private parts. And you can probably see some of the technology that's involved. In fact, the muscle layer is three d printed lattice structure. You think three d printed technology, we're able to make the muscle part of the robots very, very light. That's why the robot's only 20 kilograms. Second thing is that using three d printing technology, basically, you can customize the robot size. If you want a bigger breast, bigger butt, you can change to anything that you want. Third thing is that on top of the robotic platform, there's actually camera inside her eyes. So as I move, it actually follows me and tracks me. This is the part that I like so much. When I move, it actually looks at me deep in the eyes. For this short period of time, I really do feel connected to a robot even though I know it's it's all powered by AI, by robotics algorithm, but it's a very interesting and intimate moments I share with the robots. Obviously, we cannot just tell this black hat, this kind of robotic face. That's why we're building a whole humanoid looking face. So this was our first generation face. It can only blink his eyes. And here's our second generation face when it's getting better. The second generations, not only can it blink his eyes, it also can move his mouth.
00:13:21 · Xara, Leo
So we believe the whole thing has to be a package. It has to have a human body, a human like So initially, when we first build a face, the robots without a skin can only just flex his eyes. Later on, we've improved on a face and we built a, 13 degrees freedom face, which is slight improvement with the skin on. Later on, we actually improved further on on the newer version of the robotic phase, and we built a 30 degrees of freedom phase. As you can see, it can map to our facial expression in real time. [00:13:46] S1: So in a way, you could imagine a business in the future where these robots are teleoperated. That's why.
00:13:46 · Leo, Xara
[00:13:46] S1: So in a way, you could imagine a business in the future where these robots are teleoperated. That's why. [00:13:53] S2: Yes. That's why... Let me show you another demo. The whole thing can be teleoperated. You might have a human, let's say, somewhere other parts of the world, let's say, France, and you can teleoperate a robot to China. That's... It's possible. Well... The one of the main reason... Let me go back to the prior questions that you asked him, why do we do this and why we started? Because there are so much interesting things happening in robotic space right now, but we haven't seen a real business model where robots can become a common thing. Right? You see a lot of people try to implement humanoid robots in assembly lines in factory. You see a lot of people who try to build humanoid robot for the household. For example, Figure or even Tesla Optimus. But we haven't seen a real use case or applications where robots become useful in people's life. Fundamentally, because there's not enough breakthrough in the robotic brain part. Robots are not smart enough, are not robust and intelligent enough to do all kind of household tasks. The reason why we choose intimacy is that we think it's easier to optimize for. So every single algorithm, every single software we built in the hardware, it's optimized for intimacy and companionship. And it's a much more vertical domain of robotics versus any other regards. Because at this moment, we don't really do any kind of general purpose robots. We only wanna focus on intimacy.