 Hello everyone. Thank you so much for having me here. It's a great pleasure to be here speaking to such a distinguished audience and many friends. I have a very personal attachment to this place. I spent 18 years here at Carnegie Mellon University. A lot of the robots that I built and worked on are over here, including that one, Chimp. And it's just an honor and a pleasure to be here speaking to you all. Today I want to talk a little bit, not just about what we did, but actually about why we did it and why it's important. I've had the pleasure of being a roboticist for the past 20 years, done academic research, started my own company, working at a big company now. And I think that the RCTA provided such unique and interesting opportunities for me as a researcher for my own personal growth and for the growth of my own team. I think it's a valuable opportunity at the end of 10 years to share some of that. So I work on manipulation, robotic manipulation. That's all I do. That's literally the only thing I know how to do, and I've been doing that for about 20 years now. And it's really really simple. It's just about picking stuff up and putting it back down and doing it over and over again. I have no idea why it's hard. I remember talking to my grandmother 15 years ago about what I was doing and there was a coffee mug right in front of us and she asked me this question, but it's right there. Just go pick it up like that. It's just right in front of you. I still don't have a good answer to my grandmother other than complex configuration spaces, motion planning, dynamics, all of that. But there is a fundamental truth to grandma's question of like, it's right there. So that's a question that I've been trying to answer all through my career. Enabling robots to get out of the factory floors and into people's homes and into the real world to be able to manipulate objects as effortlessly as we are able to do in in clutter and under uncertainty. So that they can stand side by side beside us and actually perform complex manipulation tasks without being a crutch or without being a teleoperated system. And that that's really the focus of what I do. So why is manipulation hard? I think there's so many reasons and so many people who have many thoughts on it, and I think all of those thoughts are very true. But I think fundamentally for me, manipulation is so different from just navigation or going from A to B because it fundamentally involves not just going from A to B and B to C, but actually doing something there. It involves physically interacting with the world. And I think that interaction piece, that interaction component is what makes manipulation so hard. Natural language has been mastered, but dialogue is still hard. So even when you add just one other human being, there's a common editorial explosion in the number of things that you need to interact about. And so anytime you involve interaction between agents, whether it's two human agents like Nick will talk about or whether it is an agent and the world, it makes everything so incredibly complicated. But yet we're able to do this effortlessly. Like today this morning you woke up and performed incredible feats of manipulation. You made breakfast, you manipulated deformable objects like milk and were able to pour yourself a glass of milk, something that is still remarkable and hard for robots. So I want to talk a little bit about sort of the three spaces that I worked on myself. One is academia, the other is industry, and the third I want to talk about our CTA. So academic robotics is fun. A lot of us have been working on it. This is my robot Herb. This is actually a commercial we shot for Oreo. And this is Herb trying to pick up an Oreo and separating the cookie from the cream of it. It's a highly complex task. There's true complexity in the task that was involved. The Oreo is the smallest object and the most delicate object that we had to manipulate. And we are able to pick it via a lot of the algorithms that we use by the fingernails. So it's a hard problem. It's hard even for my kids. That's hard for me. So the complexity of the problem is incredibly high. But I think I want to collage two things that make this problem easy, right? One of them is the high degree of control that we have over the environment we place. We actually went through about 40 or 50 different Oreo types before we found the right Oreo that works really well. The failures were delicious. And also the robustness is incredibly low. I'll be the first one to say this works about like two out of 10 times. Super lucky. But that's the point of academic research is to be at the forefront and show the world what's possible and to enable us to dream about what we can actually make happen. So high complexity, which is a huge positive, but also a large degree of control over the world, over the environment and robustness, right, being very vocally self critical here. On the other end, you know, there's industry. So this is a video from Varsha Gray, my company. And here you can notice that the amount of degree of robustness that you need to be able to solve these problems is incredibly high, right? You're working in a factory floor. You're working in industry. You need to really make these things happen. But the complexity is is low. You know, we're not trying to separate Oreos. We're just trying to pick stuff up and put it back down. But the most important piece is the degree of control that we have over the environment is so high. We control the lighting in this particular site to the tens of lumens, right? We have light meters everywhere. We have calibration rigs. All of those things that you see out there is a carefully orchestrated dance. I was at a BMW factory out in Munich, and it was one of the most gorgeous places I've ever been, you know, sheet metal on one side and robots, BMW cars come out on the other side. But still, everything is incredibly well orchestrated, so we have great control over it. So where is the RCTA in this spectrum, right? I think RCTA really touches upon all the complexities that are critical for solving manipulation problems. There's a high amount of complexity in what we need to do. Here we're talking about being able to deal with unstructured, uncertain objects like tree branches, very, very hard to model. There's very, very low degree of control over the environment. The world is what it is. Actually, oftentimes the world is adversarial to us. We're not just in a stochastic environment. We're in a fundamentally adversarial environment where people are trying to make this robot not get to its destination. And we demand an incredible high degree of robustness. This lives depend on the success of the system. So it's incredibly important that the system be robust, reliable, and be a trustworthy companion and associate rather than something that is a crutch for you. So I think that's what I'm really excited about and that's what I'm really proud of for what we've been able to achieve. To you, you look at this video and you think, oh, I can go and move this tree branch early. But we have like 200,000 years of human instinct that has enabled us to be able to deal with the lack of structure that's present in the world. And a lot of what we're doing out here in the RCTA is to take that instinct and put that into robots. That's that robots can work with us and alongside us to be able to perform complex manipulation tasks with the narandas. Thank you.