Interview Transcript
Interviewer: So to begin with, we’d just like to ask for your full name, your date and place of birth, and anything you’d like to share about your parents, your family, where you come from, earlier lifetimes of things.
Ludovic Righetti: Okay. Yeah, so my name is Ludwig Rigetti. I’m French. I was born in 1981, so a while back. Yeah, I grew up close to the Swiss border, actually, close to Geneva, but on the French side. Yeah, then that’s it.
Interviewer: What about early education, so college, grad school, and mentors that were important to you?
Ludovic Righetti: So I actually went to school in Switzerland, at EPFL in Lausanne, which is actually relatively close to where I grew up. And yeah, so at the time, there was no bachelor or master’s degree, so I have an engineering diploma, maybe one of the last engineering diploma you could get there, because now they have a bachelor or master’s system. And in computer science, and here’s the fun fact, I mean, now I’m a professor in electrical engineering and in mechanical engineering, and I know none of those. So it tells you how interdisciplinary robotics is. Yeah, so I studied there, I gave my engineering diploma, and then I got my PhD there as well. Yeah. After that, I went to California, the U.S.C., to do a postdoc with Steph Anchella. And then after that, briefly I went to the Max Planck Institute for Intelligence Systems in Tübingen, in Germany. And I led a group, a research group there for a few years, and then I moved to NYU. And then I’ve been a professor at NYU for, I think, eight years now. Yeah.
Interviewer: So on your professional, your work experience, you already mentioned some of the places that you’ve worked, but are there any colleagues or anything about your departments that you’d like to share?
Ludovic Righetti: In what sense?
Interviewer: Just over the… Early things that… Yeah, anything significant or that has been helpful or useful or meaningful in some way, and it’s important to your career trajectory overall, like the department you’re in, people you’ve worked with? Just in terms of that professional background kind of information.
Ludovic Righetti: Oh, I mean, yeah, I mean, I have several many collaborators that are very important to me. Obviously, now I’ve met a lot of friends. It’s funny, the first time I came to IKRA, I didn’t know anyone. It was also much… It was about 10 times smaller, and it was already overwhelming. I cannot imagine young students how they feel right now. And then we… Yeah, so I met a lot of… It’s funny, I met a lot of friends, SPU students, now they’re all professors, which is kind of a cool thing to see. So a significant thing, I don’t know. I mean, yeah, my colleagues in general, my colleagues at NYU, also non-roboticists, right? My colleagues in general, it’s been great to work in… So it’s… Yeah, I don’t know. I don’t have anything very profound to say.
Interviewer: Well, then maybe I can ask you a little bit about how you first got interested in robotics proper, given this kind of background that you’ve described. Was there any kind of event, any media, books, movies, people who led you to get interested in the first place?
Ludovic Righetti: So actually, yeah, it’s interesting because I didn’t know I was going to work in robotics. When I was younger, my dad brought a computer at home. You know, at the time, there was not that many computers. It was in the 1980s. And I got really, really excited by the computer, so I was spending all my time. What computer was it? I think it was an Amstrad, which I think was a French company. And… Did you have a Minitel? We had the Minitel as well, yeah. But the Minitel, you’re not allowed to use it too much because it was very expensive. But I was really the precursor to the internet somehow, right? It was very funny, yeah. Not a bad story, I was really playing with this computer, and I really wanted to study computer science. And I didn’t really care about robots at all. And then I went to EPFL, and then I… Sorry, in computer science. And that’s what I was saying, that I ended up doing other things. And I remember because I visited EPFL before deciding where to go for college. And I saw there was labs having robots, small Kepler robots, they were like these small mobile robots, front robots, that were doing things. And I thought they were so cool. So I got really, really excited about that. I don’t remember which… I think it was the labs of Professor Nikhou and the labs of Professor Florano that were showing that. What about when was this? It was in 2000… No, 1999, or 2000, something like that, yeah. That’s when I graduated high school. And I really liked it. And then I was not sure what I wanted to do, so I was really interested in computer architecture. So I specialised in building robotics and computer architecture, and then I didn’t know where… I knew I wanted to become a scientist. I was not completely sure what I wanted to work on. And then I was looking for a master’s thesis project, and there was the… Haoke expert, who became my thesis supervisor, just arrived at the ADPFL, and there was two really cool labs. There was this lab and the lab of Odbiyar, who is now the president of Ras, right? She was… They had these really cool humanoid robots videos. They were doing really cool stuff, both of them with humanoid robots, and I was like, “This is what I wanted to do.” And so then I knew what I wanted to do. So it came very late, at the end of my college studies.
Interviewer: Were those small robots that you saw the first robots you had ever seen in person?
Ludovic Righetti: Yeah, so that’s when I visited… So I visited the school, and I saw the robots when we were visiting the labs.
Interviewer: That was the very first robot you’d ever seen in person?
Ludovic Righetti: That’s the very first robot I’ve seen, yeah.
Interviewer: Do you feel like those robots in particular, or that experience had an influence on the way that you viewed robotics going forward?
Ludovic Righetti: I don’t know. I didn’t think about it. I just thought there was really cool technology, but it didn’t really change much about… You know, there were like round robots on wheels, so they look more like toys than… You know, they don’t look like the humanoids that you see today in Adikra, right? Yeah. The humanoids I saw on the video really… I thought that was really impressive. I really remember the first time I saw that, I’d say, “Wow, this is really, really cool.” Yeah.
Interviewer: What was the very first robotics project that you ended up working on?
Ludovic Righetti: So my first class project I worked on, we had to control a wheel robot to go to a labyrinth, and it was terrible at that. So a robot was not very good, which I guess is good. He was not a good predictor of my care. But, you know, we didn’t do great in this project. It was OK, but it was not very good. So that’s the first robot I worked with. And, yeah, we didn’t really know how to do many things. So, yeah. But the real project I started to work on was during my math thesis, and then I started to work with things for humanoid robots and stuff like that.
Interviewer: So this is early 2000s?
Ludovic Righetti: In 2000, that would be 2002, 2003. Yeah. Like 2004 for the math thesis. Yeah.
Interviewer: Following that, have there been any significant robotics projects over the trajectory of your career? So, and would you be willing to sort of describe any of those projects that have been significant, or robots in particular, that you’ve worked on or with that have been important?
Ludovic Righetti: Yeah. So during my PhD thesis, I was lucky to be part of this big European project called the Robot Cub Project, which was the project that designed the iCub. I guess you know the iCub is a childlike humanoid robot, and the idea was to develop a humanoid robot that would be able to behave like a child. And so it was a very ambitious project. There was a neuroscientist involved, there was a psychologist, there was a lot of roboticists involved. There was a really, really big deal in Europe. So I got involved in that, because my funding was coming from there. And so yeah, I worked with the very, very first version of the iCub. We were working on this thing to… We wanted to get the iCub to crawl on the ground like a child. Yeah. It was not very good. But knowing that I know now, we probably did differently, but it did some things. But that was really interesting, because that’s the moment we had to learn about all the low-level control, all the mechanical aspects that I didn’t know very much about. So that’s really during my thesis that I really learned all the how to do control and how to really understand mechanics and physics and all these things that you don’t learn in computer science. So yeah. And after that, there’s another… So this was really, really inspiring. The next big project that I worked on was during my postdoc. I was part of the ARM project, which was a project funded by DARPA. And actually, it’s interesting, because since then, I do not take defense funding, but we can talk about this later. But for this specific project, I was part of this team at USC that was led by Stefan Schall and Gaurav Sukatme. And we were a team of, I don’t know, we were maybe six or seven students and postdocs working on that. And we spent… It was a competition, right? So we had to do manipulation tasks, grasping tasks. And we were competing with, I think, six or seven other teams in the US. And that’s really when I learned… I would say that’s where I learned robotics, because I had… With a few people, we designed everything from the low-level system to how to talk to the robot, to the low-level control system, to the planning algorithms, to the vision algorithms. So we did everything. And that’s really when I guess I became a real roboticist, I would say, maybe. And that was really… Yeah, it was nice. The platform worked really well. It was really hard to get it to work almost all the time, but at the end, it worked for the specific tasks we had to do. It was extremely reliable. But then that’s where I learned how difficult it is to make a robotic system really reliable. So that was a big, big deal for me. Yeah.
Interviewer: I’ll probably come back to that in terms of questions about challenges in a moment. But before we get there, are there any robotics projects that… You’ve been at NYU for eight years, that you’ve been working on in that period of time that you’d like to talk about?
Ludovic Righetti: Yeah, so obviously I’ve worked on a lot of other projects. I was thinking of the early projects in my care. One thing that… Which I think was… So, OK, we’ve done… Let me think about… So there’s one project that I’m quite proud of, and it really… I would say I played a tiny part in it. It’s… We started to… I was working with someone called Felix Kreminger, who was an engineer at the Max Planck Institute, and he’s a really, really brilliant engineer that knew how to do everything. And he was really interested in… We started to play with this… The motors that we use on quadrotors, and say, “Oh, we can use that to do some force control, and we can maybe…” So we bought some of them, and he started to do… So he’s the person who knew how to do the mechatronics. And then I put a student on… We built a small leg. For he built a small leg, I didn’t do anything. And then I put a student to work with the leg, and start to get the jump and do some control with that, and do some learning with that. And this kind of thing became… At the beginning, it was just to kind of play with something, so everything was 3D printed. And then he ended up… We built a quality pad, and some colleagues got excited. And it kind of became a big thing where it became an open source platform. It’s called… We started… Well, it’s called the Open Dynamic Robot Initiative. We made this several versions of this quality pad. Then there’s colleagues from France who got involved from the last NRS, who also did amazing contributions on the electronics and all of the things. So suddenly we had lots of people working on that, and it became this open source project that many teams made copies of. I still use these legs in my lab in the class that I teach. So we have a teaching lab, and people can use one leg to learn how to do kinematics and dynamics and things like that. And then we build a biped. One of the postdocs that was in my lab at the time was I wanted to build a biped, so that was called Bolt. So Magit Khadiv is the name of the person who did that. And so, yeah, this took a huge… This became much bigger than… At the beginning we wanted to play with something, and thanks to all these great people, as I said, I had a very… Basically I founded a few students to play with it, but that’s pretty much what I did. This became quite a big deal, and it was really exciting. And it really helped doing a lot of really cool experiments because at the time you couldn’t buy quality pads or bipeds. They were all crazy expensive. Most of them you just didn’t have access to, right? Because many companies had really nice quality pads or humanoids, but they would not sell them, right? So you had no way to do anything. So this also unlocked a lot of research that I was able to do. Now nowadays you have access to quality pads and bipeds, so things are a bit different, but 10 years ago there was an education.
Interviewer: So how would you characterize some of the main research questions that you’ve been interested in over the course of your career, and have any of those questions or core interests changed or evolved over time?
Ludovic Righetti: Yeah. That’s interesting. So when I started in my PhD thesis, I did a lot of… OK. So I’ve always been interested in understanding how to get robots to move. So I’m interested in movement. That’s something that I find fascinating. And I started my PhD, but I was in a lab that was very interested in animal motion. And my advisor, RK, was working a lot with neuroscientists to try to really… He’s really at the frontier between neuroscience and robotics, which is very, very special. And I was really interested in saying, “Oh, we can control robots the way animals behave.” Right? And so at the beginning, I worked a lot on a bio-inspired design. So we used a very abstract model of a spinal cord. We called central pattern generators, and we’re using that to generate movement. And then I got really excited by more optimization algorithms, but always about movement. So one of the things… So here’s maybe to explain. Maybe I don’t care so much about applications. I mean, I care. I think it’s important to have robots that are useful. But the thing that fascinates me is to understand things. So I would like to understand how do you build… If you had to build an algorithm, what is a good algorithm to get a robot to do anything that you want it to do? Right? And it’s a very difficult scientific question. And that’s the thing that gets me excited and say, “Oh, how do we do that?” It seems very difficult. And because it’s really difficult, it’s really exciting. And animals and humans are so good at moving that… I would love to understand how we could find algorithms that could do the same. Obviously, it translates to a lot of very useful applications, but the main scientific driver is really this question, right? How do we do that? Because it seems impossible. So, yeah.
Interviewer: So returning to that question of challenges, and you’ve just been highlighting the motivation of these questions, being that these are really difficult questions to sort of approach and answer. Would you describe some of the major challenges that you’ve been reckoning with or that you’ve experienced or even overcome? And it would be interesting to know if they’re technical, if they’re social, if they are institutional in nature, so anything along those lines?
Ludovic Righetti: In terms of challenges? Yes. So I’ve been quite lucky in terms of institutions, I would say, because I was lucky to work in very good labs that would add very healthy, well-functioning, very supportive supervisors. Raina Auke and Stefan, were all incredibly helpful, always willing to buy into whatever stupid idea at the time. So that was great, right? And I know it’s not always the case, right? So I never had any issue from that point of view. Institutions have always been sufficiently lucky to get funding for the kind of things I wanted to do. So, yeah, from that point of view, I know it’s not usually always the case, right? So things have been nice. Scientifically, it’s… Yeah, it’s… I don’t know. These problems are difficult. We see a lot of robust, a lot of things going on, but we… You know, it’s very hard to know what is the… There’s many things we can try, and it’s… How do you explain that? Scientific, I go back and forth between really excited and thinking we’re going to solve everything, to realizing nothing is working, and it’s really difficult to… You know, also, right now, we’re in this big hype where, you know, there’s a lot of media attention, there’s a lot of cool demos that you see online, but there’s a big gap between reality. So whether a robot can really reliably do 100% of the time, and what you see in a one-shot demo that you never know how it’s been done, you don’t know how many hundred times they have to do it until it looks good. So if you remove the hype, you realize that we’re still a long way to go to get robots to be really reliable, to work in an unstructured environment, and to be able to interact with their environment or with people. So it feels that we’re still very far from that. So to me, these are huge challenges, and I don’t know how to predict the future, but I would say it still looks very challenging despite all the progress that has happened this past 20 years.
Interviewer: So let’s talk about funding, then. You said that you haven’t had a… You’ve been lucky in terms of funding, but you also raised a point about DARPA funding, and I’d be happy if you could expand on both of those points.
Ludovic Righetti: Sure. So, yeah, so I’ve been lucky because I got funding both… I work in Europe, I got funding in Europe, a lot of funding to be able to have a group. I like to have a small group because I like to do science, and I don’t want to have to manage… I’m not a good manager, I think I’m a better scientist, so I like to have a small group so I can really keep doing the science myself as well. So this actually is not too complicated to sustain, and I’ve been successful enough in the US when I work with the federal agencies, but as well with companies, so things have been going well. So I work on this DARPA project, which was great, and I have nothing to complain about, but I think at that time I started to worry about the use of military applications of robotics.
Interviewer: Was this around 2000… 2009-10? So very much in the middle of robotics beginning to be used in existing ways.
Ludovic Righetti: Yeah, so that’s the time where… With a friend of mine, Kong Pham, we started to discuss about this problem. So he was the person who raised the issue and the conference at Humanoids, and I don’t remember when, around that time, 2010. And then we organized a workshop on military applications and dual-use applications of robotics. The first time we wanted to organize the workshop was actually rejected, so we kind of organized a counter-workshop in the hotel next door, and then the next year we organized actually the workshop at the Cryo-Riros, I don’t remember, and we invited people from NGOs, there was a lot of people working on hard drones, or changing policies at the international level, how they are a challenge for peace and security, and things like that. All the things that I didn’t really know, understand at the time, right? And that was really interesting, and then I got to work with a lot of international organizations on peace and security and autonomous weapon systems. At the time, that’s when several NGOs started the campaign to stop killer robots, which is a campaign to put regulations on the use of, basically on letting machines to algorithmically decide how to target and kill people, right? And I guess, like most people, I want to live in a world where technology is used for peaceful applications, and I personally think it’s important to have a separation between what is civilian applications and what are non-civilian applications. So that’s one reason, so my involvement with international organizations working on that makes it that I don’t want to have conflict of interest, right? So if I start to work also for military founders or for different applications in general, there’s potentially a conflict of interest. The second thing is that I profoundly think that technology needs to be, a civilian technology needs to remain civilian, and it needs to be dedicated to peaceful application. I’m a profoundly pacifist, and I think it’s really important. And the last thing, because a lot of fundamental research is funded in the United States by the defense, so you could argue that a lot of this research is not completely dual use, but there’s also a problem of public relations. So we learned our personal scientists to, you know, a company that funds you, a person that funds you. They can use what you do to reframe themselves, right? If you go on the website of a defense agency, you have no chance to know what they’re actually making weapons, right? They will tell you that they invented the internet, right? But they don’t tell you that this is the cost that corresponds to 0.1% of the actual money that they spend every year, right? Because the rest of the money is made to do patriot missiles and stuff like that. So I don’t want my, me as a scientist, I think I don’t want my image to be used to, you know, I don’t know, I see that in English, greenwash or, you know, other things. So these are the three main reasons why I don’t do that.
Interviewer: So it sounds like part of the story there is this conflict, if conflict is the right word for it, between even leaving aside the military for a moment. It sounds like a lot of your funding over the course of your career has been public. Yeah. Okay, so you have public funding for all of the things that you have developed, all of the research you’ve done. And part of the concern here is saying, okay, well, we’re getting all this public funding. But then, you know, I work with a commercial company where I work with the DOD, and they get to pretend that they’re the ones who develop this thing when, in fact, like, there’s this whole foundation of public funding and public money that wasn’t private or wasn’t the DOD that actually is responsible for this. Is that an accurate way of understanding that last point?
Ludovic Righetti: No, it’s not that an agency will pretend that they did the work and not you, right? They will say they did the work. They say they funded you. They say this is what we do. This is what you need to give us funding because we help create, discover new things that are helpful to everyone. But if you look at the, actually, it’s public, right? You could look at the actual appropriation for the, for military agencies. It’s public in the U.S. I think it’s reasonably public in France, but I don’t know in other countries. The actual amount of money that is spent to do fundamental research is a microscopic fraction of what is actually used. Most of the R&D money is used to develop weapons, right? But everything that you would see on their website is this tiny fraction. So it’s really used to change the image. It’s not that they pretend that they did take my work and pretend that it’s their work, but it’s more that they would use really cool, you know, beneficial technology and use it to hide everything else that has been developed, right? So that’s one aspect. That’s not the only aspect, but that’s one aspect. The dual use aspect is also very important because a lot of technology we develop, it’s very easy to weaponize a robot. And it really worries me. I’m really, really worried about these things because I really, we don’t, we live in a world with this becoming less secure and less peaceful, right? And making it easier to, and I’m not talking even about governance, right? We can imagine any type of person can use robots to do, and this worries me, and I think we need to be very careful. Because technology can be very helpful, but we need to be very, very careful in how we do it. And that’s why I think we need to have a clear separation between what is civilian use and making regulations to make sure that civilian things remain civilian and what is not civilian. Yeah.
Interviewer: So I guess I can follow up by asking a little bit about how funding and funding structures have affected your career in a kind of concrete, you know, historical sense. But also it sounds as though, I don’t want to put words in your mouth, but maybe you can talk a little bit about the relationship between funding and incentives or funding and the way that the course of research goes in relation to funding and funders.
Ludovic Righetti: So, I mean, to be honest, the funders I’ve had so far from companies and from the federal government, so I’ve been funded by the National Science Foundation a lot. I’ve been funded by the European Commission a lot. These are the, in my career, but I also receive money from different companies. It’s always been, you know, first they promote open with science, right? They promote open research, so I’ve always been able to do exactly what I want. So from that point of view, it’s been quite incredible. I think these are amazing agencies, actually. So I think I’ve been very lucky. We’ve been very lucky to be able to do the science that we want, ask the questions that we’re interested in asking. The companies I work with so far, they’ve been very open to doing fundamental science, and they want to do things that align with the kind of goals that I have. So that’s been very, yeah, so that’s been great, actually. The funding agencies have been, I don’t know, incredibly supportive. You know, you can complain about the review process, the fact that the success rates are not very high, which is true, right? I wish there was much more funding in fundamental science, but once you get the funding, it’s incredible because they understand what fundamental science should be, so it’s pretty nice.
Interviewer: So I’d like to return to this in a moment, because I know you’ve also been involved with questions of responsible AI, and maybe we can talk about that in a moment. But before then, in terms of thinking about the trajectory of your own career, have there been any tools, you know, whether it’s hardware or software, that have emerged, that have really had a significant impact on your work, in whatever respect? And you know, new possibilities or advancing things in a particular way?
Ludovic Righetti: Yeah, so, well, obviously, we have now robots that are available that you can get at the not-crazy-expensive price that are really opening incredible possibilities. So this is really, really exciting. I was very lucky to have access to humanoids early on in my career, which I understand that many, you know, I’m not that many people had access to that. So they really allowed me to do certain type of research that maybe was not possible in another lab or things like that. So I have access to really cool robots. One thing that really changed my career is the moment when I realized we got access to optimizers, so solvers to solve optimization problems. And that’s when I started to, me and a lot of colleagues, everybody started to do optimization in the loop for doing control. And to me, that was really a huge revolution of the kind of things you can actually do. So having a lot of compute, so very good computers, and having reliable solvers that you could use and be very creative in how you want to use them and do very complicated things with that was really a game changer. So a lot of software actually, I’ve been fundamental in my work in terms of libraries that can do numerics, libraries that can do optimization, that I would never been able to write myself right. Also now we have like incredible machine learning libraries. So all of these tools make our life, it opens so many possibilities, right? So then you can rethink all problems and say, oh, I can take this whole problem, but I can do so much more with that. So then I can do crazy things, right? So that’s been really, really exciting. Yeah.
Interviewer: So have you found machine learning or LLMs in general having a significant impact of late or has that not been particularly significant to the work that you’ve been working on?
Ludovic Righetti: No, no, machine learning. I mean, I’ve always worked with machine learning from the very beginning, but that, you know, from, you know, 15 years ago, what we could do with machine learning was very primitive compared to today. But no, no, no, I’ve always been. But now today we have tools that allow us to do much, much more. It’s really exciting. I’ve not worked with LLM so much, but I’ve worked with foundation models, with vision models and things like that. And they allow us to do things that are incredible. So, for example, I was recently we now we’re able to in Lego Commotion, right? We took us a lot of time to understand how to add vision into the control system. And now with the with, you know, reinforcement learning and then good neural network architecture, we’re able to actually add vision in the control. And this is to me, that’s incredible because we had no clue how to do that 10 years ago. So, yeah, this is this is really exciting.
Interviewer: So, thus far in your career, have there been any lines of questioning or research or projects that you would have liked to have pursued, but haven’t had the opportunity to so far?
Ludovic Righetti: Yeah, I think things are left by the wayside or or whatever. I mean, I guess so, right, because we have, you know, there’s only 24 hours in a day. And since I really care about my personal life, there’s much less that I can dedicate to my work. So, so, yeah, of course, I don’t I cannot say all of them, but there’s a lot of things that we should try.
Interviewer: Nothing that stands out or?
Ludovic Righetti: No, I mean, you know, I mean, yeah, I mean, it’s always you can always say, you know, I just say that, you know, I wish I had enough time to also do something else than robotics. Oh, I could maybe do other types of science, right? We have a lot of computational tools that we develop and say, oh, maybe you could use that to work with people who try to understand human movement, right? Or we can maybe we could use that to use it. So there’s a lot of things we could potentially do. But there’s no first I don’t have the expertise and I cannot take two years off to learn this expertise to do that. So sometimes we can do it with like collaborate with with people that are doing in other fields and say, oh, we can use our computational tools to solve other problems. But, you know, I wish I have more time to do much more of that because, you know, a lot of algorithms we develop, they are not just specific robots, but the decision making algorithms. So you can use them for other types of decision making process, the physical process, right? So human movement, animal movement, any other machines movement. So, you know, you say, oh, we could try all of these things, but there’s no enough time to do it.
Interviewer: Is there anything that you’ve developed or worked on that you would see as your most important or one of your most important contributions to robotics so far?
Ludovic Righetti: I don’t know if I’ve made any very important contributions to robotics, but I think we saw the project that was talking about this open source project. I think it had an impact because it created a community of people who contributed and build these robots and it helps doing research. We build a few. Recently, we published a solver for doing non-linear body creative control that works reasonably well and have several collaborators that are using it. So I guess it’s having some impact. Maybe the things that, you know, the few papers I’m really proud of are weirdly the least innovative papers. So we have, you know, there’s a lot of algorithms, people do a lot of different things and there’s a moment when you realize that all these things are connected and you can write a paper that say, oh, you know, it turns out that things are much simpler than we thought. And you can actually let’s not reinvent anything new. So one of the latest papers we wrote is about this solver we’re describing. We realized that a lot of the important ideas that we needed already existed 20 years ago. And then it was a matter to connect the dots, go to another field, right, because it’s not robotics. People who wrote about it, you understand that, connect the dots with what we were doing, then write the solver. And then it’s great because you feel you understand everything, which is a really good sentiment. Then you write your paper, you sand it, and the reviewers say, well, there’s nothing new. And say, yeah, that’s the point. There is nothing new. But this is the solution to the problem. And now we’re done. And then you say, look, this is the best solver that exists. It’s really fast. And so it’s really, I don’t know, it makes me very happy when I understand something. So I don’t know if it’s important for the field, but it makes me happy, which I guess is the most important thing for me.
Interviewer: As you’ve been working in the field, have you observed anything that you would say has been a significant change in robotics over the course of your career? So I guess both, if you could reflect on anything that in the course of your career you see as having changed or been an interesting development or not. And then on the other hand, where do you see things going in the next 10 or 20 years? Are there any particular developments or lines of flight that you see emerging right now that are going to be important?
Ludovic Righetti: Yeah. Okay. That’s a difficult question. So I mean, all these questions are difficult, but the important things I don’t know what is very striking is the size of the community. So my first ICRA, we were 600. I think there were about 600 people that were at ICRA that year. And I always thought it was really huge, right? Right now, at ICRA, we are 7,000 people. So you know the fact that there’s so many people, it means there’s so many ideas. It’s really exciting. So to me, this is maybe a community change. The community has grown a lot. Now we also have lots of robots. So hardware has made a huge progress. We have lots of tools, robots, software to really try ideas quickly and to do very interesting things. So this is quite exciting. This has been, I guess, more than the science, the fact that we have access to these things and we have a big community is changing a lot of things in the field. So this is quite exciting. And obviously, right, having access to optimization was a huge deal when in LEGO Commotion, the ability to do optimization online was a game changer in how we can do things. Machine learning and especially reinforcement learning in LEGO Commotion has been a game changer for everything related to vision, right? All the progress in deep learning has been a game changer. But we kind of have this kind of game changer every few years because suddenly you have a new technology, then everybody gets really excited, then you reflect on it and then you have this cycle. But the growth of the community is really exciting. In terms of the future, you were asking, I have no clue. I think anyone who tries to predict the future usually is wrong, so I will not predict. When I was a kid, people were saying, “Oh, we’re going to have flying cars, right? Do you see any flying cars?” Exactly. So I don’t know. I can see different paths, right? We talked a bit about military applications, right? There’s a problem now. People are using more and more robots on the different… We have two main conflicts that people talk about right now in Gaza and in Ukraine, and there’s been an insane amount of use of AI and robotics in these conflicts, and this is very upsetting to me. I wish robotics was not involved in killing people. So that’s one thing. So I hope that we don’t take this path, and I think a lot of colleagues of mine have the same hope. I think there’s a lot of… Most people on this planet don’t want to be in conflict, so hopefully we don’t take this path. I can see a lot of really nice applications though, right? Because we have opportunities to build robots to go into disasters, right? Disaster relief, getting help to people as fast as you can, right? These are things you can do that it makes a lot of sense to have machines, right? I live in New York City. Infrastructure management is a huge thing, right? How do you automate… It’s very dangerous to do construction. It’s very dangerous to manage the sewers of the city or whatever, right? Can you have machines that can do these things, that can do it better, maybe cheaper, maybe cleaner? So I see a lot of opportunities that are very exciting. What is the future going to look like? I don’t know. I hope it looks good, but I have no clue.
Interviewer: I was going to gently challenge you on that point, given the fact that if you are interested in questions of responsible AI, it does suggest a kind of future orientation, a sort of concern about what might be coming and an interest in taking concrete steps now or making plans now. And I wondered if you could expand on that at all or if you feel like you’ve said enough on that point.
Ludovic Righetti: No, I mean, it’s difficult. As a community, I think we can maybe do more in terms of being involved in the world and not just say we do technology but we don’t know. It doesn’t matter what happens because when I decide to work on a specific problem, I actually make a choice. If I tell you the most important things, I’m going to spend my working life working on self-driving cars. Implicitly means I think this is good for society, right? If I tell you I’m going to spend my working life working on leg robots, implicitly means that I think it’s a good thing for society, right? So we already make a decision that is a societal decision saying, I think this is important. So the rest is probably, I think it’s maybe less important or less interesting to me. So we need to own that and say, okay, so what does it mean? What is the consequence of that? And how do we think about these things, right? Is there a way as a community to set norms or to set educational programs to talk and limit dual use, for example, to limit that students understand what does it mean to develop a technology? If the technology can be weaponized, what can I do to mitigate that? What can I do to prevent certain applications, right? I think there’s a lot of things we can do. We are not very good at education, so we’re interested in at NYU to integrate more responsible AI and robotics education, to teach students tools to understand what can be problems. Because most people are really well-meaning, right? And a lot of firms appear without you knowing or you don’t anticipate what might happen, right? So having the tools to think about it, having the tools to act on it if something happens from an engineering point of view, but also from a policy point of view. How do you talk to your university president or talk to your society or to politicians in your city, right? These are things we need to learn how to do because if you really think robotics is going to have an implication in the world, then we don’t really have a choice, right? If we have that and important technology, then we have a responsibility behind that. How to do it? I don’t know. If I knew how to bring world peace, I would not be sitting here with you.
Interviewer: Well, sort of world peace, have you seen the results of your research applied in real world situations anywhere? You know, commercial or social, whatever?
Ludovic Righetti: Former politicians in my lab told me the other day that in their company they were using the software that we developed and made me very happy. I know that some companies have reused some of the ideas we wrote in papers to reimplement some algorithms to do things like that. So, yeah, this has been some use of the work that we do. Nothing that has revolutionized the world, but that’s normal. But, yeah, some of the things we do have been used in companies. A lot of the code, everything is open, so I don’t know exactly what people do with it. That’s one of the advantages and disadvantages of doing science in an open way, I suppose.
Interviewer: But, yeah, it must feel good to see it.
Ludovic Righetti: It’s nice when people tell you that the views, things that you’ve done is really exciting. I fundamentally think that as a scientist, I want to tell people how to do things and I want to tell this to everyone. And I want to teach people what I know. So being open is very important. That said, I think it’s important to think if there’s dual use issues, right, in peace and security. I work with a colleague of mine at the Stockholm International Peace Research Institute on peace and security risks of open source. And we’ve been thinking about how do you remain as open as you can while limiting misuse of your technology, right? Because I obviously don’t want people to use my technology to kill people. I mean, I guess you understood that. So, you know, how do you put barriers? How do you maybe put things in your code to limit certain things? There’s a lot of things that you educate people to not do that. There’s a lot of work to be done. But, yeah, I think you can have both.
Interviewer: It’s very interesting. The conversation about open source usually doesn’t head in the direction that you’re describing, but I think that’s an important area to raise for precisely the reasons you mentioned. In terms of people using your research, have you seen it applied? You mentioned some companies have been using it. Have you been in any kind of relationship with industry or with commercial actors? Have you done any collaborating with industry or commercial areas?
Ludovic Righetti: So, I have collaborated with companies. So, there’s companies who have funded my research. So, we’ve done a business more like fundamental research. So, as far as I know, they have not used any of the things that we’ve done in their product. But they have funded some fundamental research for things that we are doing. So, what I do is abstract enough that I have not worked with companies that wanted to build a specific product with a specific application in mind. So, this has not happened.
Interviewer: How did that arise? How did that opportunity to work with them happen?
Ludovic Righetti: I don’t know. No, because also I’m really interested in algorithms and things like that. So, we have not spent a lot of time, and maybe we should change that in my lab, but we have not spent a lot of time on very specific applications. One thing we started to work on, which I think is kind of interesting, but it’s not translated to a product or anything, is we’ve used some of the technology we have with our colleagues of mine who build exoskeletons. And we realized that it seems to work really well. So, that’s something maybe that can become an application that might actually be helpful. But this is very preliminary. I’m working with a company who builds exoskeleton for lower limb exoskeletons. So, we are implementing some of our algorithms on their system and trying to see how it works. But we have not… So far, it’s really a basic research. We don’t know if this will translate to practical things. But it would be great.
Interviewer: It’s not sort of a matter of companies reaching out to you, just in terms of how those relationships develop. It’s not always clear or easy to remember.
Ludovic Righetti: Oh, you mean how to develop relationships with companies?
Interviewer: Yeah, how do you get the opportunity to work with them? What does it look like? How does that develop?
Ludovic Righetti: I mean, I guess so far, it’s always been some sort of personal relations, right? Because I know a lot of people who work in companies. We talk a lot. I’m always happy to explain what we do. So, I do that all the time. I explain to people in companies that reach out to me and say, “Hey, we want to do that. How do you do this?” And I’m always happy to share what we know openly. And sometimes these relations become, “Oh, we’re doing that. Maybe we should do it together.” So, it’s always been a bit organic like that.
Interviewer: Okay, and the penultimate question I have for you is whether or not you have any advice or recommendations for students who are interested in getting involved with robotics?
Ludovic Righetti: I don’t know. Study robotics. Have fun. That’s maybe the thing that is… So, the field is competitive. And there’s a lot of people involved. And sometimes I can see that, I guess it can be overwhelming to realize there’s so many things happening that you feel you’re always late to what’s going on. That’s not true. And I think the most important thing is to have fun. Sometimes people forget that. Where scientists, the goal is to try to learn new things and be happy about it. And if you keep doing that, things will be fine.
Interviewer: That’s a great point. There is that concern about being late, that pressure, that feeling that can exist. I think that’s a really good piece of advice. And then the last question I have for you is whether there’s anything that you’d like to share that we haven’t covered?
Ludovic Righetti: That’s a… Yes. Do you have another hour?
Interviewer: Absolutely. Anything that you would like to share? Anything that feels significant or that maybe we skipped over or missed an opportunity to talk about?
Ludovic Righetti: I don’t know. We’ve talked a lot about the things. We haven’t talked so much about education. So maybe… I think it’s really exciting to… So we start to see problems in robotics. And I think… Maybe… Okay. Let’s put it this way. Something when I started to become… When I became a professor at NYU, I didn’t anticipate that I would really love doing is teaching. And actually, I spent the huge part of my career just really doing research and not really share it, teach people about it. And now that I’m actually teaching, it’s very, very nice because you have to realize that you have to actually know. You have to revise what you think you know. And then you realize you don’t know it as well as you hoped you did. So it’s very humbling from that point of view because it’s very… You have to read book, go back to the basics that you thought you know. And it’s really nice to… I don’t know how well I’m doing the job, but when you’re able to convey something and you realize that people are having fun in your class or they’ve learned something new or when someone tells you, “Oh, I got a job because of these projects,” or something like that. It’s maybe the most gratifying thing I do in my job. So I love doing research, but when you have these moments, when someone says, “Oh, I’ve learned something. Thank you.” It’s been incredible. So I encourage everyone. Teaching is really incredible.
Interviewer: Anything else?
Ludovic Righetti: I think I’m done.
Interviewer: Thank you so much. This has been wonderful. Thank you for sharing your time and your experience with us.
Ludovic Righetti: Of course. Thank you so much for talking to me.