WEBVTT 00:00:05.157 --> 00:00:12.737 Welcome to StartupRed.io, your podcast and YouTube blog covering the German 00:00:12.737 --> 00:00:17.657 startup scene with news, interviews and live events. 00:00:19.697 --> 00:00:23.337 Hello and welcome everybody. This is Joe from StartupRed.io, 00:00:23.557 --> 00:00:28.257 your startup podcast and YouTube blog from Germany. Today, I'm bringing you 00:00:28.257 --> 00:00:34.277 another interview, this time with, I would say, a dual-duty person. 00:00:34.817 --> 00:00:36.517 He's not only a professor at 00:00:36.517 --> 00:00:41.317 Stanford University, but he's also founder of OpenMind. Hey, Jan, welcome. 00:00:42.317 --> 00:00:45.517 Jan, a pleasure to be here. Totally my pleasure. 00:00:45.717 --> 00:00:49.817 We may tell our audience that you're joining me here from California, 00:00:49.817 --> 00:00:56.097 and you are here because we always try to have also entrepreneurs abroad here. 00:00:56.437 --> 00:01:01.757 My understanding is your mom is German, and your German is therefore only current 00:01:01.757 --> 00:01:03.197 in gardening terms, right? 00:01:04.297 --> 00:01:08.797 Yes, I was born in Frankfurt, but I spent most of my life in the U.S. 00:01:09.137 --> 00:01:13.937 And when my mom calls, we have long conversations about her apple trees and 00:01:13.937 --> 00:01:15.977 other things in the garden. 00:01:17.057 --> 00:01:21.837 I see. Great. So welcome, everybody. Today, we're joined by Jan Lippert, 00:01:21.997 --> 00:01:27.097 Associate Professor of Bioengineering at Stanford University and the founder of OpenMind. 00:01:27.237 --> 00:01:32.457 In 60 seconds, Jan, what inspired the creation of OpenMind and what problem 00:01:32.457 --> 00:01:34.417 are you most passionate about solving? 00:01:34.797 --> 00:01:41.757 Oh, I think all of us are fascinated by how quickly computers are becoming smart. 00:01:41.957 --> 00:01:48.277 And I was curious to see if those technologies are also relevant to robotics. 00:01:48.277 --> 00:01:55.117 I started to write software for several robotic dogs in my house, 00:01:55.117 --> 00:02:01.857 and it turned out that the software we were developing was more broadly useful 00:02:01.857 --> 00:02:03.757 than we thought at first. 00:02:03.997 --> 00:02:08.437 I was curious, because when you talk about dogs, like you think about family 00:02:08.437 --> 00:02:12.497 dogs that join you on the couch for petting, that bring you to the newspaper 00:02:12.497 --> 00:02:17.517 in the morning, or watchdogs, what kind of dogs are we talking about here? 00:02:17.517 --> 00:02:25.237 Well, I was mostly interested in dogs for just more in a family context, 00:02:25.537 --> 00:02:27.537 not not for security purposes. 00:02:27.977 --> 00:02:34.917 So the robotic dogs we're working on, they are designed to be good at conversations. 00:02:35.417 --> 00:02:40.417 They're designed to be good at helping you in the home, help your children with 00:02:40.417 --> 00:02:44.697 their homework, things like that, more than a standard security robot. 00:02:46.937 --> 00:02:51.637 To our audience, what do you think is the biggest challenge you believe OpenMind 00:02:51.637 --> 00:02:54.157 is addressing? Share your thoughts and comment. 00:02:55.057 --> 00:03:01.237 I'm talking about a little bit about you as a founder, origin story and vision. 00:03:01.977 --> 00:03:08.257 That you share with us the kind of aha moment that led you to the inception of OpenMind. 00:03:08.397 --> 00:03:12.797 Because usually many people out there think, well, I have an idea. 00:03:12.977 --> 00:03:18.637 I just registered my company and that's it. But usually, if this is an interesting 00:03:18.637 --> 00:03:23.477 opportunity, it's usually a problem you solve for somebody else. 00:03:23.577 --> 00:03:25.357 And how did you bump into this problem? 00:03:26.257 --> 00:03:34.457 Well, just directly in my house, I wanted the robotic dogs to not only be able 00:03:34.457 --> 00:03:36.117 to talk, but, for example, 00:03:36.317 --> 00:03:40.897 when they're talking to my children and my children ask them for help with their 00:03:40.897 --> 00:03:45.357 math homework, that the dogs don't simply give them the answer, 00:03:45.557 --> 00:03:49.597 but that the dogs would function more like teachers or mentors. 00:03:49.717 --> 00:03:53.837 And that requires the software to be pretty sophisticated. 00:03:54.517 --> 00:04:01.537 And you also expect each robot to know quite a bit about the humans they're interacting with. 00:04:01.957 --> 00:04:06.817 That's probably very important to be a good teacher or mentor is to understand the student. 00:04:06.817 --> 00:04:13.077 And so that's what led a little bit into this area of developing software for 00:04:13.077 --> 00:04:18.557 robots, where the robots also know quite a bit about the humans that they're interacting with. 00:04:18.557 --> 00:04:24.557 In terms of the aha moment or founding something, 00:04:24.897 --> 00:04:36.357 what I've typically found in the past is that many of us have things we're very passionate about, 00:04:36.357 --> 00:04:40.337 and that doesn't necessarily mean it's going to be a good company. 00:04:40.337 --> 00:04:45.077 But in some situations, when you're passionate about something and you tell 00:04:45.077 --> 00:04:52.497 other people about it and other people say, oh, this is not so dumb or, oh, can I give you money? 00:04:52.837 --> 00:05:02.457 Those are good indications that whatever you're working on may be suitable as a startup. 00:05:02.457 --> 00:05:09.377 So once you get feedback from the world around you, that can be a good indicator 00:05:09.377 --> 00:05:11.997 for what might come next with your idea. 00:05:13.692 --> 00:05:18.752 We already established that you have pretty advanced artificial dogs at home 00:05:18.752 --> 00:05:23.392 who are able to treat your children in mathematics, something I wouldn't necessarily 00:05:23.392 --> 00:05:25.692 associate with family dogs, 00:05:25.712 --> 00:05:31.032 but I think there were significant challenges on the way. 00:05:31.472 --> 00:05:37.092 Could you share some of the challenges you encountered in early stages and what 00:05:37.092 --> 00:05:38.632 lessons you write from them? 00:05:38.632 --> 00:05:41.592 Oh well there's um many 00:05:41.592 --> 00:05:44.992 difficulties on the hardware side um 00:05:44.992 --> 00:05:48.052 doing software is one thing but doing software that 00:05:48.052 --> 00:05:51.452 controls physical objects is a whole other thing if 00:05:51.452 --> 00:05:55.172 there's a bug in the software the robotic dog might run 00:05:55.172 --> 00:06:01.112 into the wall or it might knock people over or um it might do things that you 00:06:01.112 --> 00:06:07.172 definitely don't want um but the good thing is the really really interesting 00:06:07.172 --> 00:06:12.272 thing is that the large language models can also explain. 00:06:13.072 --> 00:06:17.692 What they're doing and what some problems might be. Just one funny story. 00:06:18.232 --> 00:06:22.192 What's really important for robots is that they don't hit people, 00:06:22.412 --> 00:06:24.052 that they don't collide with things. 00:06:24.452 --> 00:06:31.192 And the dogs, whenever they sense that someone was close, they would run up 00:06:31.192 --> 00:06:33.252 to the person and touch them. 00:06:33.372 --> 00:06:35.772 And that's exactly what we didn't want. 00:06:36.312 --> 00:06:40.532 And so finally, we asked the software that was running on the dogs, 00:06:40.692 --> 00:06:43.952 hey, why do you always run up to people and try to touch them? 00:06:44.072 --> 00:06:45.292 We told you not to do that. 00:06:45.672 --> 00:06:52.272 And the software said, oh, you know, you told me I'm a dog and dogs need to 00:06:52.272 --> 00:06:54.672 smell things to know if they're dangerous. 00:06:55.392 --> 00:06:59.412 And so, of course, if you tell me that there's a person close by, 00:06:59.692 --> 00:07:02.012 I have to go to the person and smell them. 00:07:02.732 --> 00:07:07.852 And so then we understood what the bug was in the software, and then we had to work around that. 00:07:08.072 --> 00:07:14.212 But when you're dealing with these kind of AI models, they can be truly compelling 00:07:14.212 --> 00:07:17.152 in their personas and the decisions they're making. 00:07:17.252 --> 00:07:20.352 And that really took a while to get used to conceptually. 00:07:20.572 --> 00:07:25.652 Normally, when you're programming an airplane or a car, the airplane or a car 00:07:25.652 --> 00:07:30.692 is not able to explain to you bugs in their own design. 00:07:30.692 --> 00:07:34.212 And, of course, they're not particularly interactive, but we're now getting 00:07:34.212 --> 00:07:38.692 into a world where the robot you're programming can have a conversation with 00:07:38.692 --> 00:07:41.452 you and tell you what's actually going on. 00:07:42.432 --> 00:07:47.492 Before we get into the next question, I'm curious, did you actually provide 00:07:47.492 --> 00:07:52.192 the physical hardware for the dogs that they could really smell people? 00:07:52.856 --> 00:07:58.896 That's fascinating. Building a sensor for smell is incredibly difficult, 00:07:58.896 --> 00:08:02.976 and that's one of the unsolved problems in sensors. 00:08:03.496 --> 00:08:06.696 Vision is pretty good these days. Lidar is pretty good. 00:08:06.776 --> 00:08:13.776 Pressure sensing is developing, but smell is very, very difficult. 00:08:13.776 --> 00:08:17.996 In terms of the hardware, essentially all of our hardware comes from China, 00:08:17.996 --> 00:08:24.896 and then we replace all the software for security reasons, 00:08:24.896 --> 00:08:30.776 and also we want software that's open so people can see what's going on and 00:08:30.776 --> 00:08:34.336 they can help improve it as opposed to closed-source software. 00:08:34.916 --> 00:08:40.876 Reflecting on the initial years, is there anything you would approach differently 00:08:40.876 --> 00:08:44.536 now? What do you mean by years? We've only been going since August. 00:08:47.696 --> 00:08:51.216 Yeah, let's assume it's a whole year. I know as an entrepreneur, 00:08:51.876 --> 00:08:55.456 a year really can feel quite long. 00:08:55.696 --> 00:09:01.736 And since you have long days, different duties, it kind of feels much longer relative time. 00:09:02.716 --> 00:09:10.116 Well, no, I'm having a blast. The technology is moving so quickly that every 00:09:10.116 --> 00:09:15.956 other day I wake up to some important advance, either in robotics or in AI. 00:09:16.576 --> 00:09:22.936 So the days actually feel very short. Things seem to be just going by in a complete blur. 00:09:24.576 --> 00:09:28.116 You wake up in the morning and then there's DeepSeek and then you wake up a 00:09:28.116 --> 00:09:31.476 week later and then there's Grok and everyone is battling. 00:09:31.776 --> 00:09:36.116 Everyone wants the smartest AI as quickly as possible. 00:09:36.296 --> 00:09:43.636 And that is amazing to watch and amazing to be part of, but also as a parent, 00:09:43.876 --> 00:09:50.356 is a little bit scary because all of us here are playing with fire and we're 00:09:50.356 --> 00:09:52.436 trying to outrun each other. 00:09:53.996 --> 00:09:59.236 But, you know, that has obvious risks as well. 00:09:59.936 --> 00:10:03.616 Before we get into the industry market and landscape. 00:10:05.486 --> 00:10:10.106 I'm curious how many artificial dogs are running around in your household right 00:10:10.106 --> 00:10:14.166 now and what names did your children give them? 00:10:15.286 --> 00:10:19.906 Oh, well, right now they're all muted because otherwise they would participate 00:10:19.906 --> 00:10:24.706 in this conversation and it would get very complicated for your listeners. 00:10:25.366 --> 00:10:28.846 The dogs are Frenchie, Bits and Bites. 00:10:30.446 --> 00:10:33.746 I see. Your kids do have a sense of humor. 00:10:34.026 --> 00:10:41.006 I see. Let us talk a little bit about the industry and market landscape, your competitive edge. 00:10:41.666 --> 00:10:48.946 Envision OpenMind becoming a household name. What major industry shifts do you 00:10:48.946 --> 00:10:51.606 foresee facilitating this achievement? 00:10:53.246 --> 00:10:58.686 Well, so right now, robotics is really concentrated in three areas. One area is defense. 00:10:59.246 --> 00:11:04.926 These are things like drones. These are things like torpedoes and other sort 00:11:04.926 --> 00:11:08.106 of military assets that are smart and designed to be autonomous. 00:11:08.926 --> 00:11:15.326 The other area is manufacturing. So imagine a car factory, which is mostly automated, 00:11:15.646 --> 00:11:20.766 or something like a Tesla gigafactory for batteries, which has very few humans in it. 00:11:21.146 --> 00:11:24.906 And then there are some very, very simple robots in the household, 00:11:24.966 --> 00:11:26.526 like a Roomba vacuum cleaner. 00:11:26.526 --> 00:11:37.546 And we are focusing on robots for hospitals, for the home, for schools that 00:11:37.546 --> 00:11:39.566 are not vacuum cleaners. 00:11:39.726 --> 00:11:45.226 We're focusing on robots for those environments that are smart, 00:11:45.526 --> 00:11:51.686 they learn, they know about the humans around them, and they're able to operate 00:11:51.686 --> 00:11:55.626 responsibly and help those humans as much as possible. 00:11:55.626 --> 00:12:01.726 So we're focused on the non-defense, non-manufacturing robotics, 00:12:02.186 --> 00:12:08.506 things like, you know, health care, hospitals, elder care, schools and your house. 00:12:09.519 --> 00:12:15.099 That means at one point in the future, you could see OpenMind even producing 00:12:15.099 --> 00:12:16.579 something like a therapy dog? 00:12:17.239 --> 00:12:20.579 Oh, absolutely. Dogs are great form factors for that. 00:12:20.759 --> 00:12:26.079 The Japanese had a therapeutic seal about 15 years ago for elder care. 00:12:27.099 --> 00:12:34.759 And we've seen children and also grown-ups respond very well to many different 00:12:34.759 --> 00:12:38.619 form factors, seals or dogs or small humanoids. 00:12:38.619 --> 00:12:45.639 And the trick is for them to have a face, they need eyes, and they need to look 00:12:45.639 --> 00:12:49.439 cute, and they need to have an engaging personality. 00:12:49.759 --> 00:12:56.539 And the actual physical form factor matters less than the ability of the software 00:12:56.539 --> 00:13:00.519 to emotionally engage the human they're interacting with. 00:13:02.059 --> 00:13:08.919 What are common misconceptions about AI-driven robotics and how is OpenMind addressing them? 00:13:09.119 --> 00:13:11.579 Well, it's an extremely new area. 00:13:11.879 --> 00:13:17.639 Most people think that large language models are good at producing text outputs 00:13:17.639 --> 00:13:20.919 or generating movies or audio. 00:13:21.159 --> 00:13:25.199 What most people don't realize is that essentially all of the large language 00:13:25.199 --> 00:13:28.539 models today also natively speak robotics. 00:13:28.539 --> 00:13:37.959 So you can connect Grok, Clog, Gemini, DeepSea, O3 Mini to physical hardware, 00:13:37.959 --> 00:13:43.339 and the large language models will start walking around your house and exploring your house. 00:13:44.019 --> 00:13:49.499 If they're able to write stories, then they're able to generate a series of 00:13:49.499 --> 00:13:54.499 movement commands that allow the robot to navigate a physical environment. 00:13:54.499 --> 00:14:00.859 So imagine shaking hands with Gemini or Claude or Deep Seek. 00:14:01.517 --> 00:14:06.537 So, this is a very new area, but it's moving extremely quickly. 00:14:06.977 --> 00:14:12.677 OpenAI now has a robotics team, FIGURE just showcased an advanced humanoid. 00:14:12.997 --> 00:14:16.457 So, the technology is moving extremely quickly. 00:14:20.237 --> 00:14:25.717 Who are your primary competitors and what differentiates OpenMind in this space? 00:14:27.537 --> 00:14:30.637 Generally most people think about hardware 00:14:30.637 --> 00:14:34.277 or robotics as closed ecosystems 00:14:34.277 --> 00:14:37.517 so you generally try to lock down 00:14:37.517 --> 00:14:42.477 on the system the software and the hardware and then sell the final product 00:14:42.477 --> 00:14:48.597 like a Roomba vacuum cleaner but the success of Android as a open source software 00:14:48.597 --> 00:14:56.557 for phones gives an example of how open source competitors in some situations can be viable. 00:14:56.837 --> 00:15:00.857 So what we're doing with our software is we've open sourced the whole thing. 00:15:01.197 --> 00:15:07.317 So if you want to build an advanced robot capable of having a conversation with 00:15:07.317 --> 00:15:12.337 you and exploring your house and doing things for you, you now have a starting 00:15:12.337 --> 00:15:16.137 point for building the software that does that. 00:15:16.137 --> 00:15:19.497 So one key difference is that we're open source. 00:15:19.777 --> 00:15:26.917 Of course, there's ways we make money, but the core software is open just like Android is. 00:15:27.157 --> 00:15:36.117 So as a mental model, we think of ourselves as Android for robotics or Langchain for robotics. 00:15:36.117 --> 00:15:41.897 And both of those have open source core with enterprise product and capabilities 00:15:41.897 --> 00:15:44.717 wrapped around that. Mm-hmm. 00:15:45.937 --> 00:15:51.677 Could you highlight some current market trends and data that supports your mission? 00:15:52.637 --> 00:15:58.337 Oh, that's a great question. In terms of market trends, 00:15:58.537 --> 00:16:02.197 one area we're seeing move quickly is 00:16:02.197 --> 00:16:06.257 elder care where um uh 00:16:06.257 --> 00:16:09.637 many um old people's 00:16:09.637 --> 00:16:13.557 homes for example are interested in technologies to 00:16:13.557 --> 00:16:17.117 um engage people um 00:16:17.117 --> 00:16:20.637 much more frequently and have them smile more 00:16:20.637 --> 00:16:24.217 and be happy um and that's 00:16:24.217 --> 00:16:29.077 a that's a like long conversation to have what we think about that and if that's 00:16:29.077 --> 00:16:37.297 a good idea but that's one area I'm seeing more and more pilots of robotics 00:16:37.297 --> 00:16:42.457 being tried out in a social or healthcare context and. 00:16:44.025 --> 00:16:46.905 There's a lot to unpack in that kind of business. 00:16:47.345 --> 00:16:52.345 What do you sell? Who do you sell to? What are the economies of scale and things like that? 00:16:52.545 --> 00:16:58.445 But that's one area, the sort of hospital elder care area that seems to be moving quickly. 00:16:58.905 --> 00:17:06.225 And at CES this year, there was another class of robots that got a lot of attention. 00:17:06.585 --> 00:17:13.665 These are humanoid sex robots. Um, apparently, um, the CES booth this year, 00:17:13.885 --> 00:17:16.985 um, featuring those technologies was, was well attended. 00:17:18.105 --> 00:17:24.005 So, uh, no, perhaps, you know, not surprising, um, if you have a robot that 00:17:24.005 --> 00:17:29.645 is, um, very good at emotionally engaging individuals, um, then, 00:17:29.825 --> 00:17:32.925 um, it doesn't take too much to foresee, um, 00:17:33.585 --> 00:17:37.725 uh, sort of interesting new, um, uh, sort of interesting new markets. 00:17:37.725 --> 00:17:42.185 It's, as you know, the best way to make money in AI is to program AI girlfriends. 00:17:42.785 --> 00:17:50.445 So the technology is showing itself very capable at durably engaging with lonely humans. 00:17:51.165 --> 00:17:55.825 And then the ability of having, you know, that software run around in your house 00:17:55.825 --> 00:17:57.705 and do other things that could be very interesting. 00:18:00.185 --> 00:18:06.765 I see. One shouldn't actually be surprised to see such a lot of interest in 00:18:06.765 --> 00:18:11.285 sex robots, given what large share of the Internet is made up of pornography. 00:18:12.885 --> 00:18:17.645 Absolutely. It's not a surprise. And of course, that was one of the main drivers 00:18:17.645 --> 00:18:21.965 of building out the Internet, being able to deliver large amounts of video. 00:18:23.465 --> 00:18:28.725 And, you know, it wouldn't be surprising if the sort of sex industry is actually 00:18:28.725 --> 00:18:33.785 one of the areas that also drives adoption or drives development of relatively 00:18:33.785 --> 00:18:36.605 low-cost humanoid form factor robots. 00:18:37.385 --> 00:18:41.465 I see. Before we get into product and innovation and deep dive, 00:18:41.725 --> 00:18:45.945 we'll have a little break for our advertising clients. 00:18:51.417 --> 00:18:59.277 Let's go into product and innovation. I'm here with Jan Liphart, a founder of OpenMind. 00:18:59.457 --> 00:19:04.617 Could you walk us through OAM1's core functionalities? 00:19:04.657 --> 00:19:08.897 If you had a whiteboard, how would you illustrate its architecture? 00:19:10.037 --> 00:19:14.577 Oh, the architecture is very similar to the human brain. 00:19:15.037 --> 00:19:21.177 When people think about the brain, they think of it as maybe this almost magical object. 00:19:21.417 --> 00:19:27.837 But in reality, it's a wet, massively parallel electrochemical computer that 00:19:27.837 --> 00:19:29.577 has a modular architecture. 00:19:29.877 --> 00:19:36.457 So, for example, there's regions of your brain that are solely devoted to moving 00:19:36.457 --> 00:19:41.477 your eyes towards the area of maximum classification uncertainty. 00:19:41.637 --> 00:19:43.277 And you don't even notice that 00:19:43.277 --> 00:19:47.937 you don't think about, oh, I should move my eyes to look at this thing. 00:19:48.057 --> 00:19:50.217 I don't know what it is. Your brain does that automatically. 00:19:50.217 --> 00:19:52.317 And that's a sort of modular functionality. 00:19:52.637 --> 00:19:58.917 And our software as well is composed of modules that focus on different types 00:19:58.917 --> 00:20:04.957 of sensors. These are audio or vision or LIDAR or pressure or GPS. 00:20:05.137 --> 00:20:09.717 So that gives the system information about its surroundings. 00:20:10.177 --> 00:20:17.877 All of that information is turned into natural language. And the natural language 00:20:17.877 --> 00:20:22.197 coming from all the different sensors is then fused into a paragraph. 00:20:22.617 --> 00:20:26.757 And that paragraph is a description of the world around the robot. 00:20:27.894 --> 00:20:31.674 That paragraph then flows to a large language model which 00:20:31.674 --> 00:20:34.934 then generates a series of 00:20:34.934 --> 00:20:38.394 voice and speech and movement outputs 00:20:38.394 --> 00:20:45.814 a lot of people think um a lot of people that have programmed robots before 00:20:45.814 --> 00:20:51.954 um typically do something called end-to-end ai and that involves collecting 00:20:51.954 --> 00:20:57.074 a large amount of information and training models on that information. 00:20:58.134 --> 00:21:02.394 But then what you're left with are models that are extremely specific. 00:21:02.754 --> 00:21:09.174 For example, you might end up with a good model for a car or a cat or a dog. 00:21:09.874 --> 00:21:14.654 But then it's very difficult to convert your car to a radiologist, 00:21:14.654 --> 00:21:18.394 to a cat, to a cook? How do you do that? 00:21:18.614 --> 00:21:24.234 Do you then go back and collect 10 million hours of cooking videos and retrain 00:21:24.234 --> 00:21:27.074 a new cook model or a doctor model? 00:21:27.694 --> 00:21:31.454 Our position is that that's the wrong way to do things. 00:21:31.754 --> 00:21:37.694 Our position is that you should be using prompt engineering to change the behavior 00:21:37.694 --> 00:21:39.674 of a large language model. 00:21:39.914 --> 00:21:43.074 And for example, in our system, because we're 00:21:43.074 --> 00:21:45.954 large language model based at the core of the operating 00:21:45.954 --> 00:21:48.854 system we can convert a dog to a cat 00:21:48.854 --> 00:21:52.634 by changing a single word in the prompt we tell 00:21:52.634 --> 00:21:58.554 the software you're a cat period and then the dogs now think they're cats and 00:21:58.554 --> 00:22:03.314 they will run away from dogs because they're internally convinced they're a 00:22:03.314 --> 00:22:11.514 cat and we didn't have to re-tune rebuild re-parameterize any model we made a single word change. 00:22:12.614 --> 00:22:19.954 And what we're relying on there technically is the fact that Sam at OpenAI has 00:22:19.954 --> 00:22:24.894 already taken essentially all public information that's available and represented 00:22:24.894 --> 00:22:27.654 that public information as a large language model. 00:22:28.314 --> 00:22:33.634 And contemporary large language models have a very good understanding of what 00:22:33.634 --> 00:22:37.394 it means to be a doctor, a dog, a cat, or a cook. 00:22:37.534 --> 00:22:40.614 And we can use all of that information for free. 00:22:42.054 --> 00:22:45.014 This is probably where a lot of robotics is 00:22:45.014 --> 00:22:48.634 going at least for universal robots the technology 00:22:48.634 --> 00:22:52.234 is not so applicable for manufacturing where you have very 00:22:52.234 --> 00:22:56.054 specific things you want the robots to do but it 00:22:56.054 --> 00:23:02.154 is extremely powerful for making universal robots that have a broad array of 00:23:02.154 --> 00:23:07.554 different functions and can be almost immediately converted from one persona 00:23:07.554 --> 00:23:12.914 a doctor to another persona on a cat with a single word change. 00:23:15.291 --> 00:23:16.891 That would be pretty cool. 00:23:19.051 --> 00:23:27.651 We already know that your kids used your RoboDocs to help them do math homework. 00:23:27.931 --> 00:23:34.311 Have you encountered any other surprising uses of your OM1 in unexpected ways? 00:23:35.631 --> 00:23:40.991 Well, that's the one I've seen sort of most close at home, 00:23:41.191 --> 00:23:44.071 sort of sitting at the dining table, and ben is doing 00:23:44.071 --> 00:23:47.631 his math homework and he just asked the dogs hey uh like 00:23:47.631 --> 00:23:50.431 do these square roots for me and the answer is just come right 00:23:50.431 --> 00:23:53.651 back um in terms 00:23:53.651 --> 00:23:56.971 of other things i've seen that's interesting is 00:23:56.971 --> 00:24:00.631 when the dogs are walking around where i live in 00:24:00.631 --> 00:24:03.451 los altos which is a little village and there's 00:24:03.451 --> 00:24:07.311 cafes and there's people outside on the sidewalk um the 00:24:07.311 --> 00:24:10.751 thing that always fascinates me is how much 00:24:10.751 --> 00:24:14.031 kids like the robots like every single 00:24:14.031 --> 00:24:17.731 kid comes running even if they can only just crawl they'll 00:24:17.731 --> 00:24:20.431 come over and and have a look and i think that 00:24:20.431 --> 00:24:23.251 has to do with uh natural curiosity that kids 00:24:23.251 --> 00:24:29.911 have for animals and the same is true also for robots so we've had many situations 00:24:29.911 --> 00:24:34.031 where you know the little toddlers come up and they hug the dog and they drool 00:24:34.031 --> 00:24:39.771 on the dogs in they're super curious and then uh their parents um typically 00:24:39.771 --> 00:24:41.351 are a little bit more afraid. 00:24:41.891 --> 00:24:45.411 Uh the parents will will uh you know uh they're 00:24:45.411 --> 00:24:49.931 they'll uh they're they're scared that their kid is closely interacting with 00:24:49.931 --> 00:24:53.731 the robot they don't quite know um how it will behave so they're a little bit 00:24:53.731 --> 00:25:00.611 more suspicious the thing that really surprises me is how quickly humans become 00:25:00.611 --> 00:25:04.891 attached to robots that are running suitable software. 00:25:05.311 --> 00:25:13.511 So my suspicion is in the future that any kind of stereotypes we have about 00:25:13.511 --> 00:25:17.771 robots, you know, just like they're dumb, they're a piece of technology, 00:25:18.251 --> 00:25:20.451 they're, you can turn the moment on or off. 00:25:21.151 --> 00:25:27.231 A lot of those things will be very quickly replaced with pretty strong emotional 00:25:27.231 --> 00:25:32.991 connections with robots in your home, and you will treat them like family members, 00:25:33.191 --> 00:25:37.631 like your dog or friends or things like that. Interesting things to think about. 00:25:40.120 --> 00:25:47.460 I would be interested because you talked about so much changing a dog to cat and so on and so forth. 00:25:47.980 --> 00:25:54.620 But beyond its features, what experience or feeling do you aim for users to 00:25:54.620 --> 00:25:57.220 have when interacting with OM1? 00:25:58.840 --> 00:26:02.940 Well, that's another fascinating question that relates to a very, 00:26:02.960 --> 00:26:06.820 very quickly growing area, which is called empathetic AI. 00:26:07.360 --> 00:26:14.180 And the whole point of empathetic AI is to be something that humans like interacting with. 00:26:14.940 --> 00:26:20.100 People are optimizing the duration of chats. 00:26:20.480 --> 00:26:26.260 So the goal of empathetic AI is to generate AI is that humans like to talk to for long periods. 00:26:26.800 --> 00:26:30.480 The numbers in the industry right now are approaching an hour and a half. 00:26:30.660 --> 00:26:35.160 So some of these chatbots are so good that humans have a great time talking 00:26:35.160 --> 00:26:36.460 to them for more than an hour. 00:26:38.640 --> 00:26:44.200 And in terms, and with respect to OM1, the software we're building, 00:26:44.460 --> 00:26:50.120 we're not building individual modules like empathetic AI. 00:26:50.500 --> 00:26:58.720 We're much more interested in tools that allow you to integrate all of those AIs. 00:26:58.860 --> 00:27:03.380 And a typical robot these days has something like five to ten different AIs 00:27:03.380 --> 00:27:06.880 running simultaneously. And those pieces are interacting. 00:27:07.160 --> 00:27:12.960 So we're on an engineering side, we're much more at the sort of multi-agent, 00:27:13.160 --> 00:27:18.500 physical embodiment level than we are about optimizing individual modules. 00:27:18.740 --> 00:27:24.580 But from, in terms of your question, in terms of what we like the robots to 00:27:24.580 --> 00:27:30.120 do, we certainly want compelling speech with low delay. 00:27:30.640 --> 00:27:33.380 And we want the robot to look at the human. 00:27:34.178 --> 00:27:39.338 And we want the robot to have at least one way of indicating emotional state. 00:27:40.058 --> 00:27:46.858 That could be a small display or it could be changing the color of the robot's head. 00:27:47.078 --> 00:27:53.118 So we've started to integrate color panels in the robot's head so it can show 00:27:53.118 --> 00:27:58.818 to you if it's sad or mad or angry by changing the color of its head. 00:27:59.058 --> 00:28:03.178 And so then people want to associate, you know, the head is green. 00:28:03.178 --> 00:28:10.198 Things are good if the head is like yellow or red your robot is mad or sad or something like that so, 00:28:10.778 --> 00:28:16.598 we want the robots to be able to deliver a compelling combination of inputs 00:28:16.598 --> 00:28:22.618 that give people this degree of sort of emotional connection to the technology 00:28:22.618 --> 00:28:25.478 that is quite interesting i i, 00:28:26.138 --> 00:28:30.658 just just playing around in my head when i will see the first artificial dog 00:28:30.658 --> 00:28:32.178 running around in our house. 00:28:32.358 --> 00:28:38.358 I can tell you that my wife is not a fan of a real biological dog, 00:28:38.478 --> 00:28:42.558 but maybe in the future I can convince her of a RoboDog. 00:28:42.838 --> 00:28:46.158 Yes, we can in a few months, we'll send you one. 00:28:47.258 --> 00:28:51.858 And so then you can directly help improve the software. 00:28:52.458 --> 00:28:56.518 And by virtue of it being open source, you'll 00:28:56.518 --> 00:28:59.538 be able to look inside the software and trust 00:28:59.538 --> 00:29:03.338 it more because by virtue 00:29:03.338 --> 00:29:06.898 of being open we can't hide any nefarious 00:29:06.898 --> 00:29:09.998 things in the software like stealing all your 00:29:09.998 --> 00:29:12.838 video data or all your audio data from 00:29:12.838 --> 00:29:15.618 your home you'll at least in 00:29:15.618 --> 00:29:20.998 principle be able to see what's going on inside the software so yeah we should 00:29:20.998 --> 00:29:26.578 talk about sending you one we're currently building a smaller form factor because 00:29:26.578 --> 00:29:31.618 the current dogs that are about 15 to 20 kilograms are still quite dangerous 00:29:31.618 --> 00:29:33.438 and they can punch a hole in your wall. 00:29:33.658 --> 00:29:36.558 They can also jump about two meters. 00:29:37.378 --> 00:29:43.258 And, but for normal household use, you don't really need that. 00:29:43.438 --> 00:29:47.018 And a smaller, lighter form factor is probably less scary and better. 00:29:48.321 --> 00:29:54.961 I see. The only confidential thing that I would have here in our apartment are 00:29:54.961 --> 00:29:57.001 very likely my cooking recipes. 00:29:58.421 --> 00:30:02.921 Well, yes. Sorry to say your cooking recipes are at risk. 00:30:03.081 --> 00:30:09.621 And so if you have anything super secret from your Oma, then you should please 00:30:09.621 --> 00:30:12.641 hide that properly. I will. I will. 00:30:13.061 --> 00:30:19.661 Can you share some or do you already have some real-world case studies that 00:30:19.661 --> 00:30:22.441 showcase the impact OM1 has? 00:30:22.441 --> 00:30:29.821 Well, given that we've only started in August and now we just open source the 00:30:29.821 --> 00:30:36.661 software a few weeks ago on the main use we're seeing right now is by developers 00:30:36.661 --> 00:30:38.201 at different hackathons. 00:30:38.461 --> 00:30:43.161 So they're just at the point where they're becoming familiar with the software, 00:30:43.641 --> 00:30:47.901 learning to use it and building little prototypes for different use cases. 00:30:47.901 --> 00:30:53.661 And so hopefully soon I'll have more specific examples to give. 00:30:54.481 --> 00:31:05.781 But it's certainly true that robots in general are finding success in many, 00:31:05.881 --> 00:31:07.561 many different use cases. 00:31:07.561 --> 00:31:13.261 A great example would be ukraine where of course the benefit of a dog or quadruped 00:31:13.261 --> 00:31:19.321 form factor is that you can much more easily navigate mud and fields also for 00:31:19.321 --> 00:31:21.521 example if you want to hunt tanks, 00:31:22.281 --> 00:31:28.881 one obvious way of doing that is to attach an explosive to a quadruped dog and 00:31:28.881 --> 00:31:35.521 the dog can then hunt the tank of course um you can just tell it you know go chase the tank um. 00:31:36.921 --> 00:31:40.621 But as as noted we're staying away from 00:31:40.621 --> 00:31:43.441 uh the defense sector completely that's something 00:31:43.441 --> 00:31:49.621 that's being very well handled by a quickly growing um sort of defense robotic 00:31:49.621 --> 00:31:54.621 sector uh including some friends of mine who recently left their jobs at the 00:31:54.621 --> 00:32:02.861 bundeswehr um to um focus on focus full time on defense-focused hardware. 00:32:03.341 --> 00:32:08.061 It's very easy to get a dog, chase a tank, because you tell the robot dog, 00:32:08.161 --> 00:32:09.401 the tank is now a squirrel. 00:32:10.395 --> 00:32:18.815 Precisely. Yes. In the prompt, you say, you know, you're a dog and you love chasing tanks. 00:32:19.575 --> 00:32:22.975 You, of course, have to make sure that you're a little bit more specific about 00:32:22.975 --> 00:32:27.055 what kind of tanks chase. But then, yes, but then you're good. 00:32:27.915 --> 00:32:31.575 Let us go a little bit into the business model and growth strategy. 00:32:31.955 --> 00:32:37.255 Before that, something just popped into my mind when you could do having a dog 00:32:37.255 --> 00:32:42.095 somewhere in a remote place, It's hard to access and you could put explosives on it. 00:32:42.455 --> 00:32:48.635 Would it also be possible to have dogs maybe even stationed at some remote places? 00:32:48.635 --> 00:32:54.215 What I have in mind is like a shelter somewhere in the mountains or stuff like that. 00:32:54.335 --> 00:32:57.435 We had traditional, you know, Bernardina, the St. 00:32:57.575 --> 00:33:01.415 Bernard dogs who run around looking for people, something like that. 00:33:01.415 --> 00:33:10.595 What I had in mind was a dog with the first eight pack and AD defibrillator 00:33:10.595 --> 00:33:14.035 attached in a weatherproof package on the back. 00:33:14.615 --> 00:33:19.095 That's brilliant. Juran, what are you doing for the next few months? 00:33:19.415 --> 00:33:22.055 Come here and help us build that. 00:33:23.375 --> 00:33:27.575 Happy to do so. And that's a great idea. 00:33:27.575 --> 00:33:36.675 And that basically would involve having the quadruped in a sort of nice like 00:33:36.675 --> 00:33:39.615 styrofoam container that's easy to deploy. 00:33:40.055 --> 00:33:45.675 You would have charging electronics that make sure that the quadruped is ready 00:33:45.675 --> 00:33:52.155 to go when needed and then it could be remotely activated and the styrofoam 00:33:52.155 --> 00:33:56.715 shell would open and the dog could stand up and perform, 00:33:56.735 --> 00:34:00.095 perform various duties as, as you program them. 00:34:00.295 --> 00:34:02.635 The battery life right now is about three hours. 00:34:02.955 --> 00:34:09.035 And so you'd have to think about like a charging infrastructure and communications infrastructure. 00:34:09.255 --> 00:34:14.995 We've used Starlink for communications to the dogs, 00:34:15.015 --> 00:34:19.235 and that makes it possible to connect with them basically anywhere on Earth 00:34:19.235 --> 00:34:27.395 and have a pretty good telemetry and video data using the small form factor Starlink system. 00:34:28.055 --> 00:34:34.775 So the Earth to satellite communications are getting very good very quickly. 00:34:35.075 --> 00:34:39.335 And so that means that what you just described could easily be built. 00:34:39.915 --> 00:34:45.635 And come here, we will host you. We will give you hardware and access to the 00:34:45.635 --> 00:34:47.375 lab, and let's go build it together. 00:34:48.376 --> 00:34:54.936 That would be very good. We need to seriously talk about that after the interview. 00:34:55.116 --> 00:34:59.476 But before that, let us go a little bit into the business model and growth strategy. 00:34:59.476 --> 00:35:03.496 What is the primary revenue model driving Open Mind? 00:35:04.756 --> 00:35:07.856 Well right now there's a major question in 00:35:07.856 --> 00:35:10.656 the entire ai industry which has 00:35:10.656 --> 00:35:14.736 to do with how do you monetize ai so the 00:35:14.736 --> 00:35:17.936 the question you're asking is a bigger one and the 00:35:17.936 --> 00:35:21.176 same question can be asked of robotics how do you monetize robotics 00:35:21.176 --> 00:35:29.656 what we're seeing is that as it relates to monetizing ai people or companies 00:35:29.656 --> 00:35:34.396 who have many customers and existing products have natural advantage because 00:35:34.396 --> 00:35:40.876 they can use AI to make existing products better and then charge existing customers more. 00:35:41.436 --> 00:35:44.276 That's an easy way to monetize AI. 00:35:44.596 --> 00:35:49.876 If you're a company that's developing AI itself, that's of course difficult 00:35:49.876 --> 00:35:55.736 to do because you don't have a lot of existing sales channels or customers yet. 00:35:56.721 --> 00:36:03.781 One thing we're doing as a company is we're speaking more and more to robotics 00:36:03.781 --> 00:36:11.461 hardware manufacturers with the goal of developing partnerships with people 00:36:11.461 --> 00:36:13.621 who make robotic hardware. 00:36:13.621 --> 00:36:18.841 And the value proposition is that many of these robotics companies don't necessarily 00:36:18.841 --> 00:36:24.781 have advanced capabilities in software or AI and, of course, 00:36:24.881 --> 00:36:26.541 want their robots to be competitive. 00:36:26.541 --> 00:36:31.501 So one thing we're looking into is partnerships with robotics companies. 00:36:32.561 --> 00:36:41.341 And the other thing we're doing on the sort of revenue side is we are putting 00:36:41.341 --> 00:36:46.921 up enterprise-focused products that enterprise pays for. 00:36:47.681 --> 00:36:51.601 So when you start using our open source software, then the very first thing 00:36:51.601 --> 00:36:58.921 you'll need is a good system for simulating your new robot, ideally digitally. 00:36:58.921 --> 00:37:06.441 So one enterprise product is a digital simulator for the robot you're developing, 00:37:06.621 --> 00:37:11.421 and that allows you to quickly iterate through different robot designs by virtue 00:37:11.421 --> 00:37:18.341 of being able to evaluate your robot design interacting with the digital world on a computer. 00:37:18.601 --> 00:37:22.721 That's one example of a sort of enterprise product. The other enterprise product 00:37:22.721 --> 00:37:26.461 we're building is a communication infrastructure for robots. 00:37:26.461 --> 00:37:30.041 And the idea is that robots are 00:37:30.041 --> 00:37:33.461 much better able to quickly learn 00:37:33.461 --> 00:37:39.801 and transmit that information to other robots to help their robot friends that 00:37:39.801 --> 00:37:46.721 means that if you have a swarm of robots they can transmit information and skills 00:37:46.721 --> 00:37:50.501 very quickly to all their partners in their team. 00:37:52.004 --> 00:37:58.464 For that communications infrastructure, that will also be a pay-per-use infrastructure, 00:37:58.484 --> 00:38:00.564 just like people pay for cell phones. 00:38:00.844 --> 00:38:07.464 But in this case, robots will be paying for machine-machine communications infrastructure. 00:38:07.464 --> 00:38:16.384 But the main thing for us right now, as essentially any company in this space, 00:38:16.644 --> 00:38:22.084 is to just be used, is to get users, to have people use our software. 00:38:22.424 --> 00:38:28.004 And then once you figure out what people are using the software for, 00:38:28.224 --> 00:38:31.644 then it's also easier to develop good revenue models. 00:38:31.644 --> 00:38:35.684 But we're really just emphasizing on growth right now and adoption as opposed 00:38:35.684 --> 00:38:41.104 to what specific product features are or how much to charge for them. 00:38:43.084 --> 00:38:48.384 Jan, should we ask our audience, kind of crowdsourced intelligence, 00:38:48.404 --> 00:38:53.924 how do you guys envision OpenMind expanding its reach? Comment down below here. 00:38:57.804 --> 00:39:01.504 Let's talk a little bit about fundraising and financial sustainability. 00:39:02.484 --> 00:39:05.964 Fundraising often comes with its highs and lows. 00:39:06.624 --> 00:39:12.404 Could you share a defining moment of Open Minds fundraising journey? 00:39:13.044 --> 00:39:21.504 Well, I guess one thing that's really important for founders to know is that 00:39:21.504 --> 00:39:24.004 there are many good ideas, 00:39:24.324 --> 00:39:31.884 but the way investors work can be very surprising. 00:39:32.604 --> 00:39:39.164 And generally what happens is that there are some trends in the investment community. 00:39:39.404 --> 00:39:46.204 And if your idea conforms to that trend, then your fundraise will be very easy. 00:39:47.144 --> 00:39:51.084 Conversely, if your idea is excellent, but the timing is bad, 00:39:51.344 --> 00:39:54.664 then your fundraise will be extremely difficult or impossible. 00:39:54.724 --> 00:39:59.084 So it's a little bit like, you know, fashion in Italy or Paris, 00:39:59.084 --> 00:40:02.004 where the fashions change every few months. 00:40:02.404 --> 00:40:08.964 So a big part of successful fundraising is just getting the timing exactly right, 00:40:08.984 --> 00:40:14.864 or being very fortunate that you're working on something that is broadly perceived 00:40:14.864 --> 00:40:20.284 by the investment community to be potentially really interesting and then you're 00:40:20.284 --> 00:40:22.284 not too late and not too early. 00:40:24.277 --> 00:40:29.757 In our case, I think through dumb luck, our timing was just very good. 00:40:30.477 --> 00:40:34.497 People were seeing the rapid advances in AI. 00:40:35.597 --> 00:40:41.297 And when we started to tell people, oh, you know, that AI is not just good for 00:40:41.297 --> 00:40:49.157 students cheating on their homework or missing human lawyers on AI is also good 00:40:49.157 --> 00:40:50.377 for controlling robots. 00:40:50.597 --> 00:40:52.917 And then they said, oh, I haven't heard that before. Tell me more. 00:40:52.917 --> 00:40:56.517 And so then that's a very natural starting point for a conversation. 00:40:56.837 --> 00:41:05.097 And then based on previous startup experience, people just asked if they could send us money. 00:41:05.337 --> 00:41:12.597 And so at some point, we just said, okay, let's turn this into a company and just take the money. 00:41:13.437 --> 00:41:18.437 But the most important thing there is just to get the timing right. 00:41:20.537 --> 00:41:29.357 I see. How do you balance between rapid growth and financial sustainability? 00:41:30.917 --> 00:41:32.297 Oh, you don't. 00:41:36.637 --> 00:41:37.937 Okay, next question. 00:41:40.197 --> 00:41:43.817 No, you try to build the most awesome product and then grow like crazy. 00:41:44.077 --> 00:41:47.117 And then when you need money, you just raise more money. If you can't, 00:41:47.217 --> 00:41:48.357 you're building the wrong product. 00:41:48.597 --> 00:41:54.317 It doesn't get any simpler. How have investors responded to your business model 00:41:54.317 --> 00:41:58.257 and what strategies have you employed to make your pitch stand out? 00:41:59.017 --> 00:42:05.597 Well, to first approximation, investors don't ask about the business model. 00:42:07.722 --> 00:42:14.422 So what investors are more concerned about are things like, are you building 00:42:14.422 --> 00:42:18.282 something that is new and useful? 00:42:20.302 --> 00:42:23.522 Can you protect the business, whatever it is? 00:42:24.222 --> 00:42:30.582 What is the maximum addressable market? Is the TAM a billion or is it a trillion? 00:42:31.242 --> 00:42:36.282 And those are the important questions. And then specifically what your business 00:42:36.282 --> 00:42:39.382 model is and what you're going to charge for and how much you're going to charge. 00:42:39.642 --> 00:42:44.042 All of that stuff is made up anyway and fake because you basically don't know. 00:42:44.742 --> 00:42:49.802 So the questions you get, at least in the early stages of a startup, 00:42:50.082 --> 00:42:53.722 really have to do more with how special is your technology? 00:42:54.022 --> 00:42:57.842 Does it solve an important problem? And what's the addressable market size? 00:42:57.842 --> 00:43:02.182 Of course, if the company is larger and you've been operating for years, 00:43:02.522 --> 00:43:07.142 then, of course, you have to show the dollars and cents. Are you actually making money? 00:43:07.462 --> 00:43:10.722 How much does it cost to acquire each new customer and so forth? 00:43:10.942 --> 00:43:16.582 But given that we've been going for less than a year, we're still at this very, 00:43:16.702 --> 00:43:19.522 very early stage, you know, 15 people. 00:43:21.122 --> 00:43:24.642 And we do not have to show any revenue at this point. 00:43:27.942 --> 00:43:31.002 We already talked about team and leadership here 00:43:31.002 --> 00:43:33.962 but since you're very young and you're a small team 00:43:33.962 --> 00:43:36.902 i would just stick 00:43:36.902 --> 00:43:42.682 to very few questions there uh what do you think the are the core values that 00:43:42.682 --> 00:43:49.362 define open minds company culture well the the central thing i look for in any 00:43:49.362 --> 00:43:54.802 employee is great enthusiasm and extreme flexibility. 00:43:56.162 --> 00:44:03.902 In a startup, there is no such thing as a defined job role or a clear job description. 00:44:04.182 --> 00:44:05.862 The job description is very easy. 00:44:06.262 --> 00:44:10.482 Work 24-7 and get things done. 00:44:10.902 --> 00:44:14.082 And your title and your background and your interests, 00:44:14.082 --> 00:44:20.922 none of those matter in the least uh the point for you is to be completely driven 00:44:20.922 --> 00:44:26.102 and utterly fascinated by what you're trying to do and then just get things done, 00:44:27.068 --> 00:44:32.488 And some people flourish in that environment and other people struggle greatly. 00:44:33.728 --> 00:44:38.988 For example, we've seen, this is true of many startups, 00:44:39.608 --> 00:44:44.388 that people who have worked at large companies have an incredibly difficult 00:44:44.388 --> 00:44:46.388 time joining a small startup, 00:44:47.168 --> 00:44:50.808 because they tend to be used to structure and 00:44:50.808 --> 00:44:54.128 rules and standard operating procedures and 00:44:54.128 --> 00:44:57.348 eight levels of decision making and so 00:44:57.348 --> 00:45:01.148 when someone like that joins a startup and you tell them like 00:45:01.148 --> 00:45:04.088 i don't care what you call it i don't care 00:45:04.088 --> 00:45:09.608 what your job title is um just like go do it figure it out so you're looking 00:45:09.608 --> 00:45:16.388 much more for a skill set that is like a um like a swiss army knife like an 00:45:16.388 --> 00:45:23.108 extremely enthusiastic Swiss army knife more than you are about people for highly specific roles. 00:45:25.828 --> 00:45:31.768 And if some highly enthusiastic Swiss army knife is listening right now, 00:45:31.928 --> 00:45:36.068 where could they learn more about career opportunities at Open Mind? 00:45:36.488 --> 00:45:43.388 Oh, they should just email me, jan, J-A-N, at openmind.org. 00:45:44.288 --> 00:45:48.708 It's an amazing time. It's an amazing once-in-a-life opportunity. 00:45:49.008 --> 00:45:51.188 There's unlimited money, and 00:45:51.188 --> 00:45:56.948 there are an amazingly large number of entirely new things to be built. 00:45:57.068 --> 00:46:00.868 So come to California, get on an airplane, and let's talk. 00:46:01.188 --> 00:46:06.188 Getting a little bit into regulatory, ethical, and societal impact, 00:46:06.648 --> 00:46:11.468 as Open Mind scales, what ethical considerations come into play, 00:46:11.488 --> 00:46:12.908 and how do you address them? 00:46:13.583 --> 00:46:22.283 Well, that's the topic of a whole book. And the main thing is that the technology 00:46:22.283 --> 00:46:25.983 is moving much more quickly than most people realize. 00:46:26.463 --> 00:46:31.643 So what's really missing is just awareness of what the capabilities of the software 00:46:31.643 --> 00:46:36.323 are and a broader societal discussion. I certainly have opinions, 00:46:36.683 --> 00:46:40.043 but my opinions only matter a tiny, tiny, tiny little bit. 00:46:40.243 --> 00:46:43.763 What's much more important is that there is a broad discussion around these 00:46:43.763 --> 00:46:46.983 topics, and that's part of the motivation behind podcasts like this. 00:46:47.643 --> 00:46:55.443 In terms of ethics, I think all engineers and all scientists certainly are responsible 00:46:55.443 --> 00:46:59.383 for foreseeable consequences of technologies they've developed. 00:47:00.143 --> 00:47:03.243 And there's no way we're going to be able to 00:47:03.243 --> 00:47:06.103 like scrape the surface of any of this 00:47:06.103 --> 00:47:09.083 in the remaining two minutes but as a 00:47:09.083 --> 00:47:12.303 parent I have to say I'm very afraid 00:47:12.303 --> 00:47:19.463 of what some of these technologies mean for like our children what shall they 00:47:19.463 --> 00:47:24.703 learn what kind of jobs will they have what will their families look like what 00:47:24.703 --> 00:47:31.203 will their income be like what does a democracy look like What does a society look like? 00:47:31.963 --> 00:47:39.403 And all of those things are potentially impacted by the technologies that the 00:47:39.403 --> 00:47:41.123 AI community is working on. 00:47:41.343 --> 00:47:46.523 So everyone listening to this, please wake up, please learn, and please argue. 00:47:48.083 --> 00:47:52.843 That is actually, since you already hinted, we only have two more minutes of 00:47:52.843 --> 00:47:56.903 recording time, I would have a lot more questions for you. But... 00:47:58.964 --> 00:48:01.524 For our listeners, if you had 00:48:01.524 --> 00:48:06.104 the chance, what question would you pose to Jan? Share in the comments. 00:48:06.804 --> 00:48:13.144 And Jan, to you, as a last question, where can our listeners connect with you 00:48:13.144 --> 00:48:15.204 and learn more about Open Minds initiatives? 00:48:16.244 --> 00:48:22.344 Well, hopefully there'll be a link or something like that in your podcast. 00:48:22.624 --> 00:48:27.184 But my website is at openmind.org. 00:48:28.064 --> 00:48:35.104 And we're working hard to add more material and to have more interesting content there. 00:48:35.224 --> 00:48:38.144 And of course, we're also on Twitter and all the other places. 00:48:38.404 --> 00:48:43.064 So let's just start with the website at openmind.org. 00:48:44.224 --> 00:48:47.244 Great. Jan, it was such a pleasure having you here. 00:48:47.424 --> 00:48:53.124 We talked a lot, and I think I'll have you back for another interview with very interesting content. 00:48:53.924 --> 00:48:57.704 Wonderful. It's been a pleasure. I look forward to seeing you in California 00:48:57.704 --> 00:49:00.444 at some point. And there's a dog on the way for you. 00:49:01.384 --> 00:49:03.884 Great. Thank you. Have a good day. Bye-bye. Auf Wiedersehen. 00:49:04.784 --> 00:49:06.584 Thank you. Thank you. Bye-bye. 00:49:11.484 --> 00:49:20.664 That's all, folks. Find more news, streams, events, and interviews at www.startuprad.io. 00:49:21.044 --> 00:49:23.184 Remember, sharing is caring. 00:49:23.920 --> 00:49:36.640 Music.