Latest / Automated with Brian Heater / Daniele Pucci on Why Humanoid Robots Need a Sense of Touch

Transcript

Daniele Pucci on Why Humanoid Robots Need a Sense of Touch · Automated with Brian Heater

• 0:00
  SPEAKER_01: Our mission is to amplify humans because we see a future of that kind.
• 0:05
  SPEAKER_01: We think that humans won't disappear and we will work for that future to happen.
• 0:12
  SPEAKER_01: The first thing that I thought when I saw iCub, which was the human robot of the Taliban Technology, I thought, let's make it fly.
• 0:21
  SPEAKER_01: We decided to stop that project because we really want to be focused on the product development for measurable adult value for customers.
• 0:29
  SPEAKER_01: If you look at the numbers, more or less 70% of the workers stop working because of muscleskeleton diseases.
• 0:35
  SPEAKER_01: So the idea was since this kind of diseases really happen at the manual work level, can we envisage a machine to help humans and be coordinated with them?
• 0:47
  SPEAKER_02: There's really no scenario in which this robot can operate in a cage because it always has to operate directly with a person.
• 0:56
  SPEAKER_01: If you want to scale, then you cannot simply confine all these robots without humans.
• 1:02
  SPEAKER_01: But this presents a huge safety problem.
• 1:05
  SPEAKER_01: It is, in order to enable this new society of humans and human robots, clearly the regulatory frameworks have to evolve.
• 1:14
  SPEAKER_02: I guess sort of the next logical question that a lot of people would ask, you know, especially as we're looking at a lot of the other people who are trying to build these industrial robots, is why not build a robot that would just replace that physical labor outright?
• 1:27
  SPEAKER_00: Yeah, I mean, uh welcome to Automated.
• 1:43
  SPEAKER_02: My name is Brian Heater.
• 1:44
  SPEAKER_02: I am the managing editor at the Association for Advancing Automation.
• 1:48
  SPEAKER_02: Daniela Pucci is the founder and CEO of Generative Bionics and a longtime professor at IIT in Genova.
• 1:57
  SPEAKER_02: I actually spoke to him for uh the automated newsletter a while back around some of the company's AMD news.
• 2:03
  SPEAKER_02: And I told producer Christina that we had to have him on the pod to discuss what is an utterly fascinating backstory.
• 2:11
  SPEAKER_02: So thanks so much to him and uh Generative Bionics for setting up this conversation.
• 2:15
  SPEAKER_02: Thanks to you as always for tuning in.
• 2:17
  SPEAKER_02: Don't forget to like, subscribe, review, uh, check out our newsletter and news site over at automated.fm, and please enjoy this conversation with Daniela.
• 2:29
  SPEAKER_02: Hey folks, I wanted to tell you about something that we've been working on here at A3 EditorialHQ, which is actually a series of uh work from home settings decentralized on the web 3.0 blockchain.
• 2:41
  SPEAKER_02: I'm excited to report that in addition to our recently launched automated.fm news page, we now have a news center for all of A3's editorial content that includes automated, Aaron's new business of automation, all of the podcasts, the market intelligence reports.
• 2:59
  SPEAKER_02: We've grown this team really fast, and frankly, it's useful for us to have all of this information in one place.
• 3:05
  SPEAKER_02: Check it out over at automationnewscenter.com.
• 3:10
  SPEAKER_02: Bookmark the site.
• 3:11
  SPEAKER_02: Uh, we're continuing to roll out new features throughout the end of the year.
• 3:14
  SPEAKER_02: Thank you for reading.
• 3:17
  SPEAKER_02: I I'm sure you're familiar with the the turns of phrase that we have here.
• 3:22
  SPEAKER_02: Uh we say when something is like not overly complex, we say it's it's not like rocket science, but your work is uh and has been rocket science.
• 3:32
  SPEAKER_02: I mean, that's sort of your background, right?
• 3:34
  SPEAKER_02: I mean, you are effectively a rocket scientist.
• 3:37
  SPEAKER_01: Yeah, I was actually.
• 3:38
  SPEAKER_01: So I've been working a lot on Vega back at the time, which is the launcher of the satellites of Isa.
• 3:45
  SPEAKER_01: And then after more or less 10 years, uh then I moved to flying humanoid robots.
• 3:50
  SPEAKER_01: So I used that background then also to make flying humanoid robots.
• 3:54
  SPEAKER_01: Uh and it was a kind of fun.
• 3:56
  SPEAKER_02: Yeah, it's a it's a fascinating background and a very uh uh strange and and interesting uh intersection between technologies.
• 4:04
  SPEAKER_02: But um so I mean, really though, that was your pardon the turn of phrase, but that was your initial trajectory.
• 4:12
  SPEAKER_02: Um you were studying these um, I guess like non-linear navigation systems, and that's that's what you studied, that's what you went to school for?
• 4:24
  SPEAKER_01: Yeah, so basically, I mean, I graduated back in Rome uh in so-called system engineering and control.
• 4:30
  SPEAKER_01: Uh and I mean uh in part I mean, as part of that uh uh of that journey, I basically uh studied a lot on first of all flying you flying robots and uh uh flying machines.
• 4:46
  SPEAKER_01: Then I ended up to the Vega Launcher for the ISA.
• 4:50
  SPEAKER_01: And then I also uh studied extensively nuclear fusion as part of my background.
• 4:56
  SPEAKER_02: That's what does, yeah.
• 4:58
  SPEAKER_01: And so basically, because I mean uh the idea back at that time was you know uh the trajectory, the vision was at the end of the day, we you moved you you might you may put some nuclear fusion inside robots.
• 5:09
  SPEAKER_01: So that that was my long-term vision that didn't happen yet.
• 5:13
  SPEAKER_01: And then I I kept working on basically during my PhD on flight dynamics and control aerodynamics and also nuclear fusion at the same time for four years in France.
• 5:23
  SPEAKER_01: And then I ended up at the Italian University of Technology, working for 10 years on human robotics and physical AI, and then CEO of General D Bionics.
• 5:31
  SPEAKER_02: Well, one you said didn't happen yet, so is that something you still it is still kind of on the back burner, still in the back of your mind, that this uh nuclear fusion?
• 5:40
  SPEAKER_01: Yeah, I mean, uh uh energy is going to be an issue, okay, for any uh physical device system.
• 5:47
  SPEAKER_01: Uh for sure.
• 5:48
  SPEAKER_01: I mean, uh that's an issue, right?
• 5:50
  SPEAKER_01: So we have to somehow uh re-envisage how we'll basically power up all these systems from training to deployment to the operations, and also how basically we will get rid.
• 6:02
  SPEAKER_01: So we have to also think about the death of the physical devices also in the shape of humid robotics.
• 6:08
  SPEAKER_01: So energy, okay.
• 6:10
  SPEAKER_01: Uh, in the beginning and at the end of the device, it is an issue that anyway is in the back of my mind.
• 6:17
  SPEAKER_01: Nuclear fusion might be one source of that, of that.
• 6:20
  SPEAKER_01: Unfortunately, uh, right now it's not that easy to build the nuclear fusion machines.
• 6:24
  SPEAKER_01: There are some technologies that want to somehow minutarize that, but uh well, we have to wait a bit before putting them into new metrobics.
• 6:33
  SPEAKER_02: Yeah, I think the reason there effectively like the reason why I made the the the the nuclear fusion joke uh a little bit earlier is because that that's not just like a casual aside that most people make.
• 6:44
  SPEAKER_02: That's not something that people just you know dip their toe in to the water, but you um decided that I mean you were interested in robotics, you thought, hey, this is uh a good potential energy source, and then you really just launched headlong into that?
• 7:03
  SPEAKER_01: Yeah, I mean, uh back at the time, I mean I came from the experience of the Vega launcher uh and the basically, you know, the launcher for satellites.
• 7:12
  SPEAKER_01: And I was about I was very curious on how to somehow these systems alternatively, so in other ways.
• 7:20
  SPEAKER_01: And uh that's why uh I started being really uh curious uh about nuclear fusion, and then uh robotics has always been uh uh in the back of my mind, and uh the the trajectory of my journey then at the end I was really envisaging to put these uh sources of power inside uh the physical device systems, but uh well that is not that easy, and uh but that was somehow uh the wheel when I was uh a little bit younger.
• 7:48
  SPEAKER_02: Yeah, I mean I would hazard to say that nothing that you've chosen is is especially easy.
• 7:56
  SPEAKER_01: Uh no, no, no, no, no, no.
• 7:58
  SPEAKER_01: I I agree with you, okay.
• 7:59
  SPEAKER_01: But uh I mean after the uh after my PhD in nuclear fusion and uh aerodynamics and phytodynamics, uh when I started working on human robotics at uh so I came back from France to Italy and I started working on human robotics.
• 8:13
  SPEAKER_01: I the first thing that I thought when I saw iCAB, which was basically the human robot of the Italian technology, I thought, let's make it fly.
• 8:25
  SPEAKER_01: That was the first thought that came through my mind.
• 8:28
  SPEAKER_02: Yeah, yeah.
• 8:29
  SPEAKER_02: Uh I mean I I it strikes me too that sort of looking through this, the the looking at this journey that you've had, that I think uh one of the I I I have this issue too, you know, that um I I my my mind works very quickly, and oftentimes that is a positive, but that can also be uh a net negative in that it can make it hard to sort of like focus to pick a path and to sort of see that thing to its end.
• 8:53
  SPEAKER_02: And and maybe that that maybe over the course of of your uh long career maybe has been something that has been, I don't know, maybe maybe a bit of uh a bit difficult to really sort of like find a focus and and just like continue on that specific path for an extended period of time.
• 9:15
  SPEAKER_01: Uh I mean uh let's say that focus back focus has never been a problem really.
• 9:23
  SPEAKER_01: When I start, I really deep dive.
• 9:27
  SPEAKER_01: Okay, so but I have to start.
• 9:29
  SPEAKER_01: Okay, but if I start, if I get people get engaged, and if I have a mission, uh that is usually also the other issues I have.
• 9:36
  SPEAKER_01: So I have to have impact, otherwise things really don't get uh too exciting for me.
• 9:42
  SPEAKER_01: And uh so once I have the the impact um uh in the horizon uh and then uh I get the the vision, then I get the focus, and uh for three, four years I stick to the plan.
• 9:55
  SPEAKER_02: iCub starts as a project in 2004, is that right?
• 9:59
  SPEAKER_01: Yeah, with a European project called the Robocub back at the time.
• 10:03
  SPEAKER_02: So what is the state of the art of humanoid robotics in 2004?
• 10:12
  SPEAKER_01: In the beginning, really, the idea was trying to understand the mechanisms of intelligence.
• 10:19
  SPEAKER_01: So uh there were many people who were who basically were trying to uh work on the so-called developmental robotics and trying to understand how evolution was able to drive uh the uh intelligence.
• 10:34
  SPEAKER_01: So the role of, for instance, skin, the role of sensors, uh, the role also of vision, and sometimes right now we are uh uh really stuck to these uh RGBD cameras where we have frequency, but uh, for instance, the human retina works uh in a completely different way, uh, it works with a variation of our image.
• 10:55
  SPEAKER_01: So uh people have been working a lot on trying to understand how the evolution shaped technology in order to make human intelligent and how somehow uh this uh was possible at the theoretical level.
• 11:11
  SPEAKER_01: Uh so in the beginning, uh uh the human robotics was really viewed as a platform to test scientific hypotheses.
• 11:21
  SPEAKER_01: So instead of uh somehow um having uh let's say it's difficult to somehow have uh uh a theory on psychology and then uh having tested it on humans.
• 11:32
  SPEAKER_01: So instead of uh having tested it on humans, let's test it on uh um human robotics.
• 11:38
  SPEAKER_01: And give me um give me this sorry, just real quick, yeah, give me an example of uh example.
• 11:44
  SPEAKER_01: Yeah, uh concrete example.
• 11:45
  SPEAKER_01: So the uh the so-called uh minimization of jerk.
• 11:49
  SPEAKER_01: What is the jerk in uh uh in robotics?
• 11:52
  SPEAKER_01: Is the uh let's say the variation of the acceleration.
• 11:56
  SPEAKER_01: So there is this principle that humans minimize the jerk, okay, which is somehow related to energy, but it's not exactly energy, okay.
• 12:03
  SPEAKER_01: Uh and then experimental studies somehow were pointing out that humans minimize this quantity while they move.
• 12:11
  SPEAKER_01: How can you then verify this?
• 12:14
  SPEAKER_01: You implement the same principle uh onto a human robot, and then suddenly a human robot is going to be uh to start moving uh as smoothly as a human.
• 12:23
  SPEAKER_01: Okay.
• 12:24
  SPEAKER_01: And uh or other uh, let's say other examples, uh, there are studies that show that the role of our eyes, no, sometimes we are tempted to copy humans, but it's really right now measured the fact that if you're interacting with someone at work and the gaze is pointing uh to directions that you are not really confident with, then the safety decreases and your distraction increases.
• 12:53
  SPEAKER_01: So uh, in order to test that hypothesis in human-robot interaction uh uh experiments, you can basically test them onto the humid robotics.
• 13:04
  SPEAKER_01: Uh so basically, back the time, uh humid robotics iCUB, uh more precisely, uh, was really developed uh for somehow as a testing platform for neuroscience.
• 13:15
  SPEAKER_02: And um if if I understood correctly, so iCub uh predates you?
• 13:20
  SPEAKER_02: You you arrived, and iCub was already a work in progress at the time?
• 13:24
  SPEAKER_01: Yes, because I mean I arrived at IAT actually on 2013.
• 13:29
  SPEAKER_02: So the quite a bit then it had already been worked on.
• 13:35
  SPEAKER_01: Yeah, so the f so just tell a little bit of background.
• 13:37
  SPEAKER_01: The project started in 2004, Robocab.
• 13:40
  SPEAKER_01: The first iCub was produced in 2009.
• 13:43
  SPEAKER_01: Uh uh back at the time the project was led by Julius Andin and Giorgio Metta, and then uh it was becoming uh a reference platform for basically Europe, but also worldwide.
• 13:53
  SPEAKER_01: This open source platform where scientists could contribute openly to a shared knowledge which could describe how humans develop intelligence.
• 14:05
  SPEAKER_01: Uh, this is really fascinating, right?
• 14:07
  SPEAKER_01: Uh so that's how basically the uh that's how basically the the iCub uh started, and that that was the mission of ICUB.
• 14:16
  SPEAKER_01: And then all the success of the story, I got the lead of the project later on.
• 14:22
  SPEAKER_01: I got the the lead then of the pro of the success, uh somehow successive uh iterations uh of the successors of iCUBE.
• 14:31
  SPEAKER_01: Uh this is what I got the lead of.
• 14:33
  SPEAKER_02: So having um having approached it from that standpoint of uh of evolution and really sort of pulling apart, it sounds like pulling apart um effectively why parts of the human body evolved the way they did and and why um certain uh elements are are advantageous.
• 14:57
  SPEAKER_02: Um from a uh I guess you know what they call first principles perspective, um is there value in starting with the humanoid form factor, moving on from there?
• 15:10
  SPEAKER_02: Or is there a way in which, you know, in which effectively like these single-purpose robots that have been in the world for a long time have kind of uh evolved into humanoids?
• 15:25
  SPEAKER_01: Yeah, I mean you're you're somehow touching topics, so which are at the intersection between uh uh the so-called convergent evolution in biology and evolution, evolutionary robotics.
• 15:37
  SPEAKER_01: Okay.
• 15:38
  SPEAKER_02: Yeah, I I love convergent.
• 15:39
  SPEAKER_02: Convergent evolution is great, right?
• 15:40
  SPEAKER_02: It's uh the convergent evolution is like why the platypus has a duckbill.
• 15:44
  SPEAKER_01: Yeah.
• 15:45
  SPEAKER_01: And uh uh the the concept of form follows function, which I think that then is related to human robotics, okay.
• 15:52
  SPEAKER_01: Uh, somehow it's been characterizing uh a lot the uh the evolution uh also of robots.
• 15:59
  SPEAKER_01: So at the end of the day, uh if you look at these uh theories, uh most of them characterize the evolution of humans uh by basically a couple of things.
• 16:08
  SPEAKER_01: The first one is minimization of energy.
• 16:10
  SPEAKER_01: So we've been trying to minimize our energy throughout our evolution, okay, one generation after the other.
• 16:16
  SPEAKER_01: And basically, we wanted to maximize uh the probability that our kids basically uh survived.
• 16:22
  SPEAKER_01: So it was really a matter of survivability, okay.
• 16:26
  SPEAKER_01: Uh so these are basically okay, by synthesizing the two main uh uh the two main drives for the evolution of humans.
• 16:36
  SPEAKER_01: And this, however, it's important.
• 16:38
  SPEAKER_01: Why is it important?
• 16:39
  SPEAKER_01: So it is important because sometimes making a human robot exactly as a human is not that smart, okay, because humans uh convert there because of certain reasons.
• 16:50
  SPEAKER_01: Okay.
• 16:50
  SPEAKER_01: Yeah.
• 16:51
  SPEAKER_01: And if I if a human robot has to be fast, okay, uh, well, uh, you know, uh, you don't actually need all the characteristics of human.
• 17:00
  SPEAKER_01: And there is an impact actually on the development of our products, of Jersey, of Jersey robotics uh products, that is going to there, there is a huge impact on how we see the product development strategy.
• 17:11
  SPEAKER_02: Yeah, well, so okay, so this is, I mean, just to kind of put a bow on this particular part of the conversation, um, this is very interesting.
• 17:22
  SPEAKER_02: Uh speaking of survivability or or why humans effectively evolved like bipedal locomotion, I think is is you know, sort of in a sense what you're talking about.
• 17:32
  SPEAKER_02: So conservation of energy, okay, that's something that is valuable for robots, right?
• 17:38
  SPEAKER_02: I mean, conserving energy, using less energy to do certain actions, that's valuable.
• 17:44
  SPEAKER_02: They have they have batteries on board, but survivability for offspring isn't necessarily something that translates to robots.
• 17:52
  SPEAKER_01: Not yet.
• 17:54
  SPEAKER_01: Okay.
• 17:55
  SPEAKER_01: Although we know that, you know, uh at a certain point, if we keep copying uh uh humans by using human data, then uh somehow this kind of uh behaviors will emerge.
• 18:09
  SPEAKER_01: Uh and that's why we need also guardrails at the physical levels, which are much more important than the uh, let's say LLMs-like guardrails, the impact of physical devices that move around.
• 18:22
  SPEAKER_01: If they feel to be threat, uh if they feel a threat, then uh you know that you have to deal with that in the proper way.
• 18:29
  SPEAKER_02: Yeah, well, so so so let me back up then.
• 18:31
  SPEAKER_02: So let me let me take me back to, I don't know, let's go as far back as 2013, and you know, you were very interested in robotics and you had been for a long time.
• 18:41
  SPEAKER_02: Um you saw what robot robots were out in the world, you had a notion of, you know, to a certain extent what robots would look like going forward.
• 18:51
  SPEAKER_02: Um based on all of your interactions specifically with the humanoid form factor, uh, did you anticipate that humanoids at some point were going to become a popular, more ubiquitous form factor?
• 19:06
  SPEAKER_01: So uh basically uh back at that time, it was still a hard scientific problem.
• 19:14
  SPEAKER_01: Okay, from 2013 to 2020, to be more precise.
• 19:19
  SPEAKER_01: The fundamental technology, because here people do not talk about uh basically the muscles of uh human robots that actually was what enabled all this acceleration.
• 19:30
  SPEAKER_01: Okay, you see all these robots jumping around because there has been one single evolution at the mechatronics level that then enabled the reinforcement learning to be deployed at zero shot, and then the robots started basically jumping and dancing.
• 19:44
  SPEAKER_01: Okay, so the the um the role of the body, okay, is uh I mean overlooked on all this story, and that's why generally bionics where we say we have to use uh the concept of body as compute, you cannot detach the brain from the body, these two things are attached.
• 20:04
  SPEAKER_01: You cannot treat the eye separately from the body, didn't work for humans, it's not gonna work for uh human robotics.
• 20:11
  SPEAKER_01: Okay, however, back at the time, old technology, old models, old gearboxes, and then it was really difficult to make uh human robots move and basically uh have a degree of robustness in terms of perturbations.
• 20:26
  SPEAKER_01: You push them, they fell.
• 20:28
  SPEAKER_01: The DARPA robotic challenge basically showed all these complexities, it was around 2015-2016, and it was really an art problem.
• 20:36
  SPEAKER_01: Then in 2020, things started changing.
• 20:39
  SPEAKER_01: Okay, we had a certain number of evolutions from the mechatronics point of view and from the learning point of view.
• 20:45
  SPEAKER_01: Uh, on the one hand, we understood how to basically have uh, let's say, models more efficient uh for human robotics, and this enabled algorithms in order to launch the training and the learning of the motor control intelligence of human roads.
• 21:01
  SPEAKER_02: So to I I guess to break that down, so more efficient meaning that it uses less compute and less battery, and then that can be devoted to the the compute required for the algorithms?
• 21:12
  SPEAKER_01: So more efficient actually means that the effect of friction, okay.
• 21:19
  SPEAKER_01: Okay.
• 21:20
  SPEAKER_02: Yeah, yeah, yeah.
• 21:20
  SPEAKER_01: So you're talking from like a physics standpoint, from the physics standpoint, okay.
• 21:26
  SPEAKER_01: And uh and also we understood how to to do motors with higher uh power density.
• 21:32
  SPEAKER_01: Okay, it's like a battery with uh more capacity, but the motors more powerful, okay, and with less friction, so they were more efficient.
• 21:41
  SPEAKER_01: You put some current in and you got more energy out, okay.
• 21:45
  SPEAKER_01: To be uh to use simple words.
• 21:48
  SPEAKER_01: So these are the two things that we understood uh back at the time.
• 21:52
  SPEAKER_01: And uh uh the European network somehow was didn't actually understand, in my opinion, didn't Actually, understand uh profoundly uh how the entire field could be accelerated.
• 22:05
  SPEAKER_01: And uh actually back at the time uh I knew that some some of my friends and colleagues in the United States were moving to uh uh to some companies, big companies.
• 22:17
  SPEAKER_01: Uh one of these was uh figure, okay.
• 22:19
  SPEAKER_01: Jerry Pratt was moving to Figur as CTO, I know Jerry very well.
• 22:23
  SPEAKER_01: And then also the Chinese were investing a lot of money onto uh onto human robotics.
• 22:28
  SPEAKER_01: Uh they some who wanted to hire us.
• 22:31
  SPEAKER_01: And then we said uh okay, so if we do not stop doing prototypes and start accelerating towards products, basically the entire work that we have been doing for 15 years will get obsolete in 24 months because here the entire field is accelerating.
• 22:46
  SPEAKER_01: And that's why we said let's start General G Bionics as an industrial spin out because here the exercise is completely different.
• 22:53
  SPEAKER_01: We have to stop doing prototypes and we have to start industrializing um robots.
• 22:58
  SPEAKER_02: When when was that decision made?
• 22:59
  SPEAKER_02: Was that made how close to 2020 was that that you decided to really start commercializing?
• 23:03
  SPEAKER_01: Uh it was uh in between uh 2020.
• 23:06
  SPEAKER_01: So I started realizing in 2021 uh decision taking between 2022 and 2023.
• 23:14
  SPEAKER_02: Okay, got got it.
• 23:15
  SPEAKER_02: And and sorry, just to back up a second, I'm I'm curious because this this is a real this is a very interesting point, because right now I I hear a lot of people something that's obvious, I think, to anybody who's followed the space now because they've heard this a million times.
• 23:29
  SPEAKER_02: Um, when we're talking about Morvek's paradox and um, you know, uh the the like relative ease with which a robot can like do a backflip or all of these other things versus um that that takes for granted what you're talking about, right?
• 23:45
  SPEAKER_02: We're saying, oh, it's easy for a robot to do a backflip, but that takes for granted what you're talking about, which is the hardware, which is and and and um the power, the strength of the actuators, all of these factors.
• 23:58
  SPEAKER_02: What's your sense of why that hasn't been a part of the conversation in the same way?
• 24:44
  SPEAKER_01: Uh yeah, I mean, um because uh somehow uh we ask that this is also uh a problem that has to be dealt with at the trust level.
• 24:54
  SPEAKER_01: Uh we're gonna touch a little bit on it.
• 24:57
  SPEAKER_01: But we see a human robot and we somehow expect uh similar performance uh of human beings because of the same uh form factor.
• 25:07
  SPEAKER_01: Okay, so we somehow um uh implicitly give to the body uh a um a property simply because it looks uh it looks like something that we know.
• 25:19
  SPEAKER_01: Yeah.
• 25:19
  SPEAKER_01: Okay, and there is going to be a problem uh because uh uh I mean uh if we do not tune also the uh expected performance with the real performance on how these robots look like, then we are going to create an except an acceptability problem here.
• 25:40
  SPEAKER_01: And uh because uh we may expect some properties of a robot, and then if they do not meet that those properties, we are going to decrease our level of trust.
• 25:51
  SPEAKER_01: And uh that is, you know, uh in this transition moment, these kind of things cannot be overlooked.
• 25:58
  SPEAKER_02: Yeah, okay, so so that's what you're saying, because you know, usually what when I think about um uh like like taking the robot for granted and we have this conversation again about like more box Morvex paradox, it's it's you know, things that like we would like tasks that we would consider simple tasks, folding laundry, like that's been a big one that we've been talking about.
• 26:16
  SPEAKER_02: But I think what you're you're saying is that also applies to these physical actions, these things that require these advanced uh uh physics, these advanced um uh components.
• 26:29
  SPEAKER_01: Yes.
• 26:30
  SPEAKER_01: Uh just to give you a sense of what we're talking about.
• 26:33
  SPEAKER_01: Uh the design of robots, okay, and the clothes of robots, okay.
• 26:37
  SPEAKER_01: So uh they're gonna have implicitly an impact on how we will interact with them.
• 26:42
  SPEAKER_01: Uh, but also there will be a source of intelligence.
• 26:45
  SPEAKER_01: Just to give you an example, uh Geno 1, our human robot that we released, has the skin or body tactile sensors within the exterior design.
• 26:56
  SPEAKER_01: So the design of the robot is no longer only, let's say, a design accessory uh to the machine, it's a source of intelligence.
• 27:05
  SPEAKER_01: And this basically changes abruptly uh anything that we know, anything that we knew about, for instance, automotive design uh that had an impact on the performance, a little bit on aerodynamics, but that is not a source of the intelligence of the machine.
• 27:20
  SPEAKER_01: Okay, and how this uh source of his intelligence is this is distributed will have an impact on the performance of the robot, and then we might also somehow have a feel-off.
• 27:32
  SPEAKER_01: Uh, because if uh I don't know, a robot uh uh is dressed like uh uh I don't know, a welder or a nurse, uh we have uh we have expectations over the machine.
• 27:43
  SPEAKER_02: Yeah.
• 27:44
  SPEAKER_02: Yeah.
• 27:44
  SPEAKER_02: I I as as we're talking about um uh I I I guess aesthetics, you know, and and and th this is uh you know it's uh uh an interesting conversation.
• 27:54
  SPEAKER_02: Um, you know, you mentioned figure before.
• 27:57
  SPEAKER_02: There was a big push, figure optimist, all these kind of big scary industrial robots for for a while there.
• 28:04
  SPEAKER_02: I look at iCub and it looks like a bit it's a big baby doll head, right?
• 28:09
  SPEAKER_02: I mean that's like effectively how I would describe it.
• 28:12
  SPEAKER_02: I'm curious, again, like you know, the project does predate you, but um uh what what was what was behind the design of that robot?
• 28:21
  SPEAKER_01: Yeah, I mean it was basically 2009, right?
• 28:24
  SPEAKER_01: So it's like uh a lot of uh a long time ago.
• 28:29
  SPEAKER_01: Uh but really the design uh uh iCab was basically an equivalent of uh three-year-old uh and uh there are there were reasons uh uh on why it had to somehow uh look like a child, because the the objective uh was to study evolutionary uh robotics, okay, uh what we said before.
• 28:51
  SPEAKER_01: And uh and then the mechanisms uh for developing intelligence uh in those years are very, very important.
• 28:59
  SPEAKER_01: Okay, uh so that's why uh having a small humanoid robot uh would have simplified the process of understanding uh uh the impact of the uh uh algorithms that describe the intelligence of basically a kid uh developing intelligence of the same age.
• 29:14
  SPEAKER_01: Okay, yeah.
• 29:16
  SPEAKER_01: So that was basically one reason why to make a child human robot.
• 29:20
  SPEAKER_01: The uh the somehow the face of ICAV uh uh was an output of uh a long uh long study, the uh Carnival effect, we know the uncanny valley effect.
• 29:32
  SPEAKER_01: So we know that basically if uh you let's say a human robot or in general device uh looks too much as a human, we basically have a pushback feeling.
• 29:44
  SPEAKER_01: Or if it looks like too much as a machine, we have a pushback, so we don't engage with that.
• 29:49
  SPEAKER_01: So there is a certain number, there is uh an uncanny valley uh where you have to get rid of, and the face of iCarb actually was the output of a study that we wanted to avoid that.
• 30:01
  SPEAKER_01: Wanted to have something that you could interact with, but uh you still could recognize it as a as a as a machine.
• 30:08
  SPEAKER_02: At what point does the project really start moving into the space of um you know how how can we start leveraging this form factor for something for something useful?
• 30:22
  SPEAKER_02: How can we start introducing humanoid robots into industry?
• 30:25
  SPEAKER_01: Yeah, so basically iCub from uh 2009 to 2020 was basically a research platform for scientists.
• 30:37
  SPEAKER_01: So an open source platform, the great idea of having this open source platform where anybody uh could contribute to, and it was uh also be it uh distributed around the world.
• 30:48
  SPEAKER_01: We were basically the largest robot manufacturers around the world back at the time.
• 30:52
  SPEAKER_01: So we seemed like more than 60 copies uh back at the time.
• 30:57
  SPEAKER_01: And uh around 2020, however, we had a big collaboration with the Italian Institute for Work Accident and Prevention.
• 31:07
  SPEAKER_01: Okay, and this institute uh basically pointed out that if you look at the numbers, more or less 70% of the workers stop working because of muscles diseases.
• 31:19
  SPEAKER_01: So the manual work uh uh is still stopping uh because of the of the massoskeleton diseases.
• 31:27
  SPEAKER_01: So the idea was since this kind of diseases really happen at the manual work level, uh when uh workers have to move objects around, logistics nowadays, uh, okay, of human robotics, can we envisage a machine to help humans and be coordinated with them in order to reduce the so-called biomechanical risks of getting injured?
• 31:51
  SPEAKER_01: Okay, but that was 2020.
• 31:53
  SPEAKER_01: And uh that's when we actually built Ergo Cub, uh, which was the next iteration of iCub.
• 32:00
  SPEAKER_01: Ergo Cub, Ergo stands for ergonomy.
• 32:02
  SPEAKER_01: The idea was great.
• 32:03
  SPEAKER_01: If we can somehow have uh, on the one hand, a sensorized human, uh, humans are all are is uh I mean humans are always around our stories.
• 32:14
  SPEAKER_01: So if we have a sensorized human, uh then we can basically have human uh measurements like the fatigue connected to the brain of the robot so that he can understand when to intervene or what when not to intervene.
• 32:31
  SPEAKER_01: Okay.
• 32:31
  SPEAKER_01: And uh and that's why we started working on wearables in order to measure the fatigue of human beings, which was then later on a fundamental asset because then we were really we had key expertise in digital digitalizing human beings.
• 32:47
  SPEAKER_02: So to clarify, then this isn't uh a robot coming in and replacing a job outright.
• 32:52
  SPEAKER_02: It's when the person like clearly isn't capable of doing all the work themselves, the robot is coming in and assisting them.
• 33:00
  SPEAKER_01: Yes.
• 33:01
  SPEAKER_01: So uh the the scenarios back at the time were two, uh, one in a warehouse, where you had uh worker moving objects, and uh as long as the uh object was not impacting the mass skeleton system was fine.
• 33:15
  SPEAKER_01: But since we could somehow online monitor, but also predicting the future what was the fatigue, if the future fatigue was too high with the eye, then the robot would have intervened and the suit that he was wearing basically would have vibrated in order to stop uh the worker to do the dangerous work.
• 33:34
  SPEAKER_01: Okay, that was in Werazus and then also in hospitals.
• 33:37
  SPEAKER_01: I mean, there is a big issue, uh especially in Europe, concerning the uh musculaton disorders and in general uh problems of nurses uh while moving patients.
• 33:50
  SPEAKER_01: So the so-called uh uh, I mean, moving patients physically is uh inducing a lot of problems for the healthcare system.
• 33:57
  SPEAKER_01: So that that was the other scenario.
• 33:59
  SPEAKER_01: A robot that could help patients in order to be moved uh uh while being on the bed.
• 34:04
  SPEAKER_02: This is this is interesting because I feel like what we're talking about here is a case of convergent evolution, because you effectively end up in a similar place as a lot of these other humanoid robotics companies, but you're approaching it from a very different direction in a lot of respects.
• 34:21
  SPEAKER_01: Yes.
• 34:22
  SPEAKER_01: Uh and we really believe that the next iteration of uh convergent evolution is humans integrated with physical AI, and whose shape, one of those, is for sure human or human robots.
• 34:37
  SPEAKER_01: So we see really in the future a society of humans and human robots that integrate and will have also a new ergonomy.
• 34:48
  SPEAKER_01: So you have to think of the energy expenditure, for instance, of a new society which can be then optimized as a whole.
• 34:57
  SPEAKER_01: You can optimize as a whole a new society which has humans and humanoid robots that work together, and then locally uh you can also optimize how much work human robots should should do in a certain area or not, in order to optimize the energy of this new society.
• 35:15
  SPEAKER_01: I don't know if I was clear or not, but basically, if you zoom out in the future, you will have physical systems in the form of uh human robots of human robots integrated with humans, and this is a completely new society that we are building.
• 35:27
  SPEAKER_02: Yeah, yeah.
• 35:28
  SPEAKER_02: Break down what what integrated means in this respect.
• 35:32
  SPEAKER_01: So integrated means uh you are at work and uh you're lifting weights, and at a certain point uh uh the AI of the robot says, you shouldn't you shouldn't be lifting these weights because otherwise in three months you are getting to a hospital and I'm gonna help you.
• 35:49
  SPEAKER_01: Okay, because you're being monitored.
• 35:52
  SPEAKER_01: Okay.
• 35:52
  SPEAKER_01: And uh and this new AI integrates with your physical system.
• 35:58
  SPEAKER_02: So I I I I guess I guess sort of the next logical question that a lot of people would ask, you know, especially as we're looking at um uh you know, a lot of the other people who are trying to build these industrial robots, is uh why why the robot why not build a robot that would just replace that physical labor outright?
• 36:16
  SPEAKER_01: Yeah, I mean uh there are people who want to do that.
• 36:21
  SPEAKER_01: We think that that is not gonna happen.
• 36:23
  SPEAKER_01: Humans won't disappear.
• 36:25
  SPEAKER_01: If we want these robots to be in, I mean accepted, then they have to work with us.
• 36:30
  SPEAKER_01: And I think that is also a strategic direction that companies take.
• 36:34
  SPEAKER_01: We, Jared D Bionics, decided that our mission is to amplify humans because we see a future of that kind.
• 36:42
  SPEAKER_01: We think that humans want to disappear and we will work for that future to happen.
• 36:50
  SPEAKER_02: Yeah, yeah.
• 36:51
  SPEAKER_02: That that's a okay.
• 36:52
  SPEAKER_02: So so so this this remains uh a core piece of uh of the um of the system, which but but which also means, I mean, this this introduces this is okay, so this is fascinating for for a couple of reasons.
• 37:06
  SPEAKER_02: I'm just gonna spew out uh two two pieces of information, so or two ideas, so so bear with me.
• 37:13
  SPEAKER_02: Uh one of them is that um you're truncating, potentially truncating the timeline because you're not asking a system to be doing all of that hard labor around the clock, right?
• 37:25
  SPEAKER_02: Because it's gonna be supplementing some of the hard labor uh with the person, but you're to a certain extent maybe creating a slightly longer timeline to getting it to market because there's really no scenario in which this robot can operate in a cage because it always has to operate directly with not just like not just near a person, but like with a person.
• 37:52
  SPEAKER_02: With a person.
• 37:54
  SPEAKER_01: Okay.
• 37:54
  SPEAKER_01: Uh we see this system says an evolution of wearables that right now we many of us actually wear.
• 38:00
  SPEAKER_02: But but this presents a huge uh safety problem.
• 38:04
  SPEAKER_01: It is.
• 38:05
  SPEAKER_01: It is presenting, but the serious companies out there are not faking that this is not this is not happening, this is not the this is not a problem.
• 38:14
  SPEAKER_01: Okay.
• 38:15
  SPEAKER_01: Yeah.
• 38:15
  SPEAKER_02: Yeah, they're not they're not fake, they know that, but but but but I think you know that the sort of the tacit understanding right now is that like all of these pilots are happening in cages and then they're gonna work towards that.
• 38:25
  SPEAKER_02: But for you, in order to actually from day one, it's gonna have to be my people.
• 38:32
  SPEAKER_01: If you want to skill, if you want to be serious at skill, then you cannot see simply confine all these robots without humans.
• 38:41
  SPEAKER_01: Okay, uh, you have to think of a new regulation system, and this is something that we are somehow positioning ourselves really strongly on, uh, in order to enable this new society of humans and human robots.
• 38:55
  SPEAKER_01: So this is and clearly the regulatory frameworks have to evolve as the robots are doing, as the eye is doing.
• 39:02
  SPEAKER_01: Well, also the regulatory framework has to evolve, but uh we have to be serious on this, okay.
• 39:10
  SPEAKER_01: Uh the certification, the certification processes uh of this system still has to be addressed seriously, uh the new standards have to be put in place.
• 39:20
  SPEAKER_01: We are, I mean, on the edge working on this because compliance and safety is uh I mean, these are two pillars of generative bionics.
• 39:31
  SPEAKER_01: That's why we believe in human robot integration, because in order to scale properly, you have to be safe and you have to interact with humans.
• 39:40
  SPEAKER_01: It's like, I mean, Chat GPT is interacting with humans at the next level, it's of course a physical added interacts with me.
• 39:47
  SPEAKER_02: Yeah, and and and uh we've obviously seen some of the problems that can happen when that and then now we're talking about obviously adding in a physical element with these big with these big heavy machines.
• 39:58
  SPEAKER_02: But uh it what it's interesting because you know our our parent organization deal deals in standards and um you know they're they're US standards, but a lot of them also uh operate to a certain extent internationally as well.
• 40:13
  SPEAKER_02: And um one of the things that I really appreciate about labor in the EU is like, or just regulation in the EU in general, is that it it tends to be stricter in a way that favors um people, laborers, you know, let less so certainly than here, right?
• 40:32
  SPEAKER_02: Like really our laws tend to favor the corporations.
• 40:37
  SPEAKER_02: But what that means for a company that is starting this process in Italy is that you're gonna have a higher bar to clear from day one because you're gonna be operating within the EU.
• 40:52
  SPEAKER_01: I mean, uh our our our objective uh is to be the European champion of physical life for human robotics, but designed to scale globally.
• 41:01
  SPEAKER_01: So we have really uh internally projects uh all over the world.
• 41:07
  SPEAKER_01: So we really somehow are and we are also monitoring uh all the regulations around the world, which is going to be an asset of generative bionics because we can also then suggest uh uh what are the things that matter and what are the things that probably do not matter.
• 41:22
  SPEAKER_01: Okay, so uh we are designed uh from day one to scale globally.
• 41:27
  SPEAKER_01: Uh clearly we uh we were born in Europe, and it means that our first objective uh is indeed to lead the European uh uh market for human robotics, but we also have collaboration world worldwide.
• 41:40
  SPEAKER_01: And this is going to be an asset of the company because we believe that can we can also help at the regulation level in order to blend uh what is important and what is not at the safety, uh at the safety, uh at the safety level in order to have this human robot integration.
• 41:53
  SPEAKER_02: And and and the the sensors, the skin is like that that's the core part of safety, is that right?
• 42:00
  SPEAKER_01: Yeah, so we really believe that uh if you want a human robot to interact with uh uh humans, you have to feel it in a way or another.
• 42:12
  SPEAKER_01: Okay.
• 42:12
  SPEAKER_01: There are many people out there that say no, the interface is only voice.
• 42:16
  SPEAKER_01: Well, if you go to the doctor, I think that it doesn't only listen to you, it also puts on sensors on you.
• 42:21
  SPEAKER_01: Okay, so and it runs some checks.
• 42:24
  SPEAKER_01: So uh robots really have to feel, and the feel the feeling is for sure the sense of touch, because elderly needs physical assistance.
• 42:33
  SPEAKER_01: I mean, uh, what are we talking about here?
• 42:36
  SPEAKER_01: Okay, uh, we have an elderly society that needs help and physical assistance.
• 42:42
  SPEAKER_02: And they're also they're also obviously the most vulnerable.
• 42:45
  SPEAKER_01: They're most vulnerable.
• 42:46
  SPEAKER_01: So these systems have to protect uh this uh uh piece of society.
• 42:51
  SPEAKER_01: And of course on the one hand, uh, the sense of touch uh is going to be a fundamental piece of the story here in order to help this piece of society.
• 43:00
  SPEAKER_01: On the other hand, uh, we have also proximity sensors that can anticipate the touch.
• 43:06
  SPEAKER_01: Okay, that's why Geno One, our uh first humanoid robot, can feel not only the touch but also the distance without looking at it.
• 43:14
  SPEAKER_01: Okay, and that's one of the assets uh in our opinion, in order to integrate humans with humanoid robots.
• 43:21
  SPEAKER_02: Yeah, I I I mean, as you're talking, I know obviously you know you you you were at CES, you you gave that present.
• 43:26
  SPEAKER_02: I know I I know AMD has been a really good partner for you.
• 43:29
  SPEAKER_02: You gave that presentation at their event as well.
• 43:32
  SPEAKER_02: And it sounds like you've been talking about healthcare really since the beginning, but um do you do you you do you get the sense that healthcare is a a further off is further off, that it's just gonna take longer to get there than it will for have the having these robots in like factories and warehouses?
• 43:50
  SPEAKER_01: I mean, uh in my opinion, uh uh robots in the house will take longer because uh the environment is less structured.
• 43:58
  SPEAKER_02: So But is healthcare somewhere in the middle between the two?
• 44:01
  SPEAKER_01: Exactly.
• 44:01
  SPEAKER_01: So in our opinion, net care is something in the middle between uh B2B industry and the house.
• 44:07
  SPEAKER_01: Okay.
• 44:08
  SPEAKER_01: And I mean, most in our opinion, uh or human robots can help also at the entrance of hospitals.
• 44:14
  SPEAKER_01: There are many elderly out there that they don't know where to go when they go to go to the hospital.
• 44:19
  SPEAKER_01: They don't know who to who to ask to.
• 44:21
  SPEAKER_01: Okay, so the interface between the service and the customer in that case is going to be why not a human robot that can help uh at the cognitive level uh these patients in order to simply orient themselves.
• 44:34
  SPEAKER_01: Then there is micrologistics, there is a lot of problems in the hospitals in order to bring uh drugs to.
• 44:39
  SPEAKER_01: The patient.
• 44:40
  SPEAKER_01: Or another example, neurohabilitation.
• 44:43
  SPEAKER_01: I mean, if you have issues at the neurological level, you need someone that can basically help you in order to train back your capabilities.
• 44:54
  SPEAKER_01: So new so neural habilitation is going to be a fundamental direction that humid robots have to deal with.
• 45:00
  SPEAKER_01: Humid robots never get tired by talking.
• 45:04
  SPEAKER_01: Okay.
• 45:04
  SPEAKER_01: And the let's say the AI can have as context your previous medical histories, okay, so that it can target also specific words that you need to say in order to recover your abilities.
• 45:16
  SPEAKER_01: Okay.
• 45:17
  SPEAKER_01: So this is our journey to be to zoom out.
• 45:20
  SPEAKER_01: Our journey into the healthcare is has been put on the x-axis, and the x-axis is the distance with the patient.
• 45:28
  SPEAKER_01: So we want to start far from the patient, and then we get closer and closer to the physical assistance.
• 45:34
  SPEAKER_02: Well, we are we're coming really close to time, and I know I'm gonna regret not bringing this up.
• 45:39
  SPEAKER_02: So we're gonna close on this.
• 45:40
  SPEAKER_02: We'll bring it full circle back to rocket science.
• 45:43
  SPEAKER_02: Iron Cub 3 uh made its debut, uh launches 20 inches off the ground.
• 45:48
  SPEAKER_02: This is the fork between the the iCub robots.
• 45:52
  SPEAKER_02: This is, I think, your baby, basically iCub with a rocket attached to the back.
• 45:58
  SPEAKER_02: What is the status of that project and why did you take iCub and stick a rocket on it?
• 46:04
  SPEAKER_01: So, I mean, yeah.
• 46:05
  SPEAKER_01: The project uh started back at the time, uh back basically it was 2017.
• 46:10
  SPEAKER_01: Uh the uh the idea was that uh after an earthquake or floods, uh you basically need three capabilities.
• 46:20
  SPEAKER_01: You need a platform that can fly and overcome floods and fire, you have to land, you have to look for survivors, and you have to locomote uh into somehow uh difficult unstructured environments, or legs, and then you need uh hands in order to open wolves, doors, and look for survivors once again.
• 46:37
  SPEAKER_01: So you need these three capabilities, which is flying, uh manipulation, and also locomotion.
• 46:43
  SPEAKER_01: That's why we said, well, uh we had a human robot that can walk and manipulate, let's give them uh jets and we make it fly.
• 46:51
  SPEAKER_01: So that was the rationale of it.
• 46:54
  SPEAKER_01: So it's not easy.
• 46:55
  SPEAKER_01: It was not easy because you have uh fire getting out uh at 800 degrees Celsius supersonic speed.
• 47:01
  SPEAKER_01: So it's a little bit thermodynamics, there is an issue.
• 47:05
  SPEAKER_01: Uh but then uh we we were able to make it uh make it fly.
• 47:09
  SPEAKER_01: After generated bionics born, which we remember it was uh uh basically January 2026 this year, uh we we decided to stop that project because we really want to be focused on the uh product development for uh measurable adult value for customers because we believe that this is going to be the issue that human robotics have to deal with.
• 47:30
  SPEAKER_01: Uh so we said now we stopped that project and we uh stick to the B2B industrial application for human robotics, and that's why, Brian, our strategy was really simple.
• 47:42
  SPEAKER_01: We believe that one human robot does it all, it does that doesn't work.
• 47:46
  SPEAKER_01: I mean, a welder, a welding human robot cannot be the same as uh a nurse, uh human robot, so a human robot in the hospital.
• 47:55
  SPEAKER_01: So with this approach, we said we develop a platform, which is our robot Gene 1, from which we derive application-specific human robots by customizing the AI, because the AI of a robot that does logistics is not the same as a robot that does, I don't know, uh, running.
• 48:13
  SPEAKER_01: And then uh the uh exterior design, the exterior design, the covers of the robot must be different.
• 48:20
  SPEAKER_01: If you are welding, uh you have to somehow uh uh withstand temperatures of a certain kind.
• 48:26
  SPEAKER_01: If you are uh inside the logistics, you do not.
• 48:29
  SPEAKER_01: And then end the factors and uh which means hands and and feet.
• 48:33
  SPEAKER_01: So that's our strategy, and we this we are really sticky to it in order to develop human robots.
• 48:40
  SPEAKER_02: So it's the same platform, but you really you don't believe in a like a truly general purpose robot in the pure sense.
• 48:47
  SPEAKER_01: I mean, uh we believe in multi-purpose, and uh uh on this we believe in it, but we really want to go bottom up.
• 48:57
  SPEAKER_01: We want we don't want to go top down.
• 48:59
  SPEAKER_01: Otherwise, you risk to fall in law with the method because you have this new AI algorithm.
• 49:04
  SPEAKER_01: So we want to solve concrete problems.
• 49:07
  SPEAKER_01: I mean, our mantra was let's try to avoid to solve useless problems.
• 49:11
  SPEAKER_01: Okay, and that's why we start from the use cases uh uh surveillance, patrolling, uh logistics and manufacturing, and FCA defines scenarios uh that basically requires a little bit different human robots that you have to customize, and then but these customizations have to happen at with the same platform.
• 49:32
  SPEAKER_01: Just to make an example, there are many cars out there that have the same platform.
• 49:35
  SPEAKER_01: Issue chassis.
• 49:37
  SPEAKER_01: The chassis.
• 49:38
  SPEAKER_01: Okay.
• 49:38
  SPEAKER_01: So the chassis is the same, but then the dress of the car is different, and that's our approach in order to make it scalable.
• 49:45
  SPEAKER_02: Yeah, that's why you hire the guy from Ferrari to help you build a robot.
• 49:50
  SPEAKER_02: Um yeah.
• 49:52
  SPEAKER_02: So so so so just to clarify, speaking of uh uh bottom up, so so the iron cub project is on hold for now.
• 49:59
  SPEAKER_01: Yes, the iron cub project is uh was an old yeah, since uh since December 2025.
• 50:05
  SPEAKER_02: Got it.
• 50:05
  SPEAKER_02: Well, uh, thank you so much for taking the time.
• 50:07
  SPEAKER_02: And obviously, we got a lot more to talk about, so let's chat again soon.
• 50:11
  SPEAKER_01: Thank you, Brian.
• 50:13
  SPEAKER_02: Thanks so much to Daniela and Generative Bionics.
• 50:16
  SPEAKER_02: Absolutely fascinating conversation.
• 50:18
  SPEAKER_02: Thanks to you, the viewer slash listener, for tuning slash listening on in.
• 50:23
  SPEAKER_02: Please like, subscribe, share, review, check out all the rest of our work over at automated.fm.
• 50:31
  SPEAKER_02: Stick around, and we will be back soon with another episode of Automated.