Latest / Art of the Question / The Ethics of Living With Robots - Prof. Sven Nyholm - #37

Transcript

The Ethics of Living With Robots - Prof. Sven Nyholm - #37 · Art of the Question

• Julius Moorman: Thank you so much for for taking the time to meet with me today. I am very excited to ⁓ yeah, talk about ⁓ robotics and the ethics of of robotics and what it means to introduce robots into our lives that aren't just inside of a car factory but will interact ⁓ yeah with our day to day lives. Will presumably be inside of our houses and Will play a significant part in our life. So I think it's a very interesting discussion, very timely. ⁓ but before we get into all of it, I was curious to know when did you first become interested in robotics? ⁓ well, first of all, thanks for having me on the podcast. ⁓ well, I guess I s I'm now working in ⁓ Germany and Munich, ⁓ but I used to work in the Netherlands ⁓ at the I Eindhoven Technical University, the University of Eindhoven, and ⁓ obviously that being a technical university, there's a lot of ⁓ interesting stuff going on. For example, they were building ⁓ different kinds of robots, ⁓ including soccer playing robots. So the yeah, so there's a team of students that are developing ⁓ these soccer playing robots that compete in something called the Robocop. ⁓ so we have the World Cup now at as we as we're recording this podcast, but ⁓ every I don't know if it's every year, but every other year ⁓ robots ⁓ compete ⁓ in soccer and ⁓ and i I ha on one occasion you went on one of these tours of the k the the campus ⁓ d ⁓ to see what's going on and ⁓ you know the guide showed us the robot ⁓ lab and then they asked okay do we have a volunteer that can play ⁓ you know can can perfor per perform the role of a goalie? And so I I volunteered and so I was ⁓ the goalie and then a robot was trying to score and I on the first attempt indeed the robot did score. And then on this the second attempt I I was able to catch the ball and then the third, so there was two f for me, ⁓ one for the the the robot. But one of the things that was fascinating about that was that it was actually somewhat uncomfortable to for me at least, subjectively, so to speak, to sort of interact with this robot on this on this ⁓ football field. I mean Those robots, they don't look like sort of humanoid robots. They look sort of like upside down trash cans and they have little legs that stick out and they can kick the ball, but they kick the ball pretty hard. And when the robot is ⁓ moving towards you, you almost feel like a wild animal is attacking you. And so I was fascinated ⁓ in part just by interacting with these robots now as I was working there at ⁓ the ⁓ the INOVE University of Technology. ⁓ but ⁓ there were also, in addition to the actual robots, ⁓ humans that were studying not the philosoph well That too, the philosophy of human robot interaction, but also the psychology of human robot interaction. ⁓ so some of the colleagues just in the on the other side of the hallway of the corridor were studying how we sort of respond to robots, how we tend to well, to use a technical term, anthropomorphize them, namely to attribute human-like ⁓ qualities to the robots. And I just thought, ⁓ I mean, also partly studying some of the my my colleagues' work that there should also be the philosophy and the ethics of human robot interaction, not just the The psychology of human robot interaction. And again, I should emphasize I was by no means the per first person to start working on this, but I was sort of sucked into this not primarily from the philosophy that existed, but by hearing about the work of the psychologists and also about meeting some of these robots. I mean, there was another robot at ⁓ T. U. Eindhoven, ⁓ ⁓ I I think it was called Amigo or Amiga or something like that. A medical robot. And then you know when the Queen of the Netherlands came to the university, I mean most of the students and staff members did not get to meet the Queen, but the robot got to meet the Queen and and handed over a bouquet of flowers. And the Queen said thank you and then the robot asked, you know, what's your name? And she induitively responded with telling the robot's her name, etc. So I'm just I I just got fascinated in how we sort of we s very quickly fall into re responding to the robots. Either we sort of classify them as I did when I played soccer as like some sort of, I don't know, wild animal that's attacking, or ⁓ like the Queen did, ⁓ as a sort of almost like a human, a person, and you start talking with the robot. And mind you, these were very the neither robot looked at at all like a human. Okay, so maybe the medical robot they had arms, they had some sort of head and ⁓ vaguely looked like a sort of a paradigmatic robot from a science fiction movie. But those soccer playing robots looked nothing like it. So anyway, so long story short, I I got into it maybe well ten or twelve years ago by just interacting with robots and by interacting with colleagues studying the human robot psychology. Yeah, very interesting. And what you mentioned there, just the name of the robot, right? Amigo or amiga. Even there, we are already anthropomorphizing the robot by calling it friend. ⁓ Spanish for a word for friend, as many people ⁓ probably know. But yeah, I think it's incredible. Twelve years ago, probably also very rudimentary compared to what we have today. For me, one of the moments where I started thinking about robots entering in our day-to-day lives was when I first started playing around with ⁓ ChatGPT and when I recognized how good the image recognition was and where it was actually helping me I was figuring out what to in this editing software called DaVinci Resolve and also a Blender, which is a 3D. ⁓ program and I was sending it screenshots of what I was doing and it was ⁓ pointing me in the right direction and it could actually understand this very complex 3D scene inside of this pretty niche software I was using. And then I started then I realized okay what what will happen if I send it a picture of this glass of water, for example, when I put it at the very ⁓ near the edge of ⁓ of the table. what will happen next? What is likely to happen next? And of course these AIs are are merely prediction machines, as as many people like to point out. But it was able to recognize from this image that the glass was likely to tip over and that sh that I should probably move it ⁓ away from the edge. So if you combine that with robotics, that was very interesting for me because for many years We've had robots primarily work in our factories, right? That's not new. Robots have been assembling our cars for decades. But that's so very different because it's a closed environment and all the movements are pre-programmed and they cannot adapt, they cannot adjust when anything changes. Everything goes to hell if if a tiny thing changes, so everything needs to stay exactly the same. But with combining artificial intelligence in these robots. We're looking at a very, very different type of technology. What do you think is such an important distinction between classic robots and and AI powered robots? And what does that mean from ⁓ a philosophical perspective? But how we how we interact with technology like this? Yes. Let me maybe just mention one more example so that we ha we have a few different examples to work with. I mean so around the time that I played soccer with those soccer playing robots and ⁓ saw the Amiga robot Amiga robots, etc. ⁓ the discussion about sort of the ethics of self-driving cars was also sort of really getting started. Right. ⁓ and so that would be another example ⁓ where you have, you know, you have a machine with a physical body, very heavy, very big, out in sort of society interacting with people. Also using sort of machine learning techniques and ⁓ and AI ⁓ methods, ⁓ image recognition, sometimes with success, sometimes with less so. And so one of the reasons why this is sort of philosophically interesting to think about how AI is being put into robots is that well, these things are indeed gonna be out in traffic, ⁓ possibly causing accidents, possibly hopefully avo helping us to avoid accidents. raising, however, questions also about well, if the a if the if you're not driving, you're just sitting there as a passenger, and then ⁓ you know, something happens, ⁓ obviously the question that immediately arises is, well, are you responsible? Well, y ⁓ maybe I was taking a nap. I was told perhaps by the you know the self-driving cars, taxi service, this car is safer than a regular car, you don't have to do anything, you just sit there, you you take a nap, ⁓ you wake up when you get to your destination. ⁓ so One of the I mean what one of the many reasons why these technologies raise philosophical questions is that ⁓ you know we're handing over control or we're handing over certain tasks that are perhaps dangerous ⁓ to ⁓ AI systems, to robots. And ⁓ you mentioned the distinction between sort of pre-programmed robots that perform sort of pretty rigid ⁓ s ⁓ sets of tasks like within a controlled environment. Compare that with the self-driving car that is ⁓ of course. Traffic is to some extent a controlled environment because we have roads, we have signs, but still it's it's a it's you know it's it's ⁓ the wild west in comparison to sort of a lab with, you know, ⁓ par the the sort of i with a w a warehouse that's been designed specifically for robots. I mean, actually that's another point that's relevant here in this context. I mean, so you have a choice. Either you w make the environment sort of robot friendly, ⁓ which might mean re- I mean, so you're ⁓ I believe in Amsterdam, ⁓ you know, you could take y you you let a self-driving car loose in Amsterdam has a lot of ⁓ obstacles, it has a lot of things that it has to deal with. You could redesign the environment so that it's super safe for self-driving cars, but then you wouldn't have the beauty of Amsterdam. And so and in the same way, I mean so there I think very few people, if any, would say let's just redesign a s a city like Amsterdam or Munich where I am so that it's maximally robot friendly, where robot might include self-driving cars. There no one is gonna say let's do it. In a warehouse ⁓ where, you know, we're trying to o ⁓ automate everything, ⁓ then ⁓ you know, big companies so that are ⁓ you know operating in this warehouse is gonna be quite happy to redesign the warehouse so that it's maximally robot friendly, but perhaps not very friendly for the humans working in that warehouse. So you have questions about responsibility, you have questions about handing over control, you have questions about should we make the world maximally human-friendly or maximally AI or robot-friendly. And all of these choices are sort of ethically loaded because I mean. ⁓ i mean there are lots of values involved here. I mean it's like do we want beautiful cities? Do we want sort of like robot friendly cities? Do we want the world to be sort of human, ⁓ I don't know, friendly, or do we want the world to primarily be AI friendly and then humans d ⁓ just have to live with it? So yeah. So very quickly one one has a lot of interesting questions here. Yeah, no, there are it does raise very, very interesting questions about design, but I think If we I think the ultimate goal for humanoid robots, like the holy grail, would be to have a system in place that would just understand the world like we built it and the natural environments that it encounters and that it will still be able to find its way through those environments. And what we s we see by ⁓ with the self driving cars example, I'm very glad you you raised that point, is many people were very scared, right? It was a trolley problem, essentially. Many people were like, okay, ⁓ in a car crash, ⁓ will it will it ⁓ prefer the life of the passenger or ⁓ will it like steer onto the sidewalk in order to avoid ⁓ a a crash. car crash, for example, killing all civilians on the sidewalk, but protecting the passenger. I did a substack article on what this would look like, the the trolley problem, and whether the trolley problem is actually applicable with self-driving cars. It is my understanding that it's primarily focused on reducing kinetic kinetic energy, the AI systems, and not necessarily that it's going to make this ethics based decision because well first of all it's illegal but also pretty difficult to determine are these children, is this a grandma, is this person sick, ⁓ what's the age of my passenger and and who should who should go in such scenarios. But is that if we're if we're talking about like self driving cars, because it's a very it's it's not even a hypothetical scenario because in the in many American cities they are already part of everyday life. You have ⁓ the the taxi ⁓ services. Is that already do you see any any changes ⁓ about how we in the in the short amount of time that these now are on the road, have you seen any any changes or interesting things how we treat those those cars and what happens around those technologies? ⁓ Absolutely, yes, lots of changes. And I actually this was yet another thing. ⁓ I mean I mentioned self-driving cars as being something that I was interested in. And exactly this whole trolley problem discussion ⁓ was one of my sort of entryways into into this topic because indeed a lot of people have been and very early on, I mean i immediately when people started talking about the legal and the ethical situation, people started comparing this to the trolley problem and ⁓ actually this was one of my one of my and l everyone did it. I mean, whether there were philosophers, economists, psychologists, ⁓ people in general, journalists. And so I was one of those people who said, Well, let's think about is it really like a trolley problem indeed? And and ⁓ actually my so far ⁓ to g an article I wrote together with the Dutch ⁓ philosopher Giles Smith's ⁓ my our by far most successful ⁓ philosophy article that has the had had the most citations and most people reading it is about this question, is this really like a real world? Trolley problem, ⁓ or are there interesting differences? I mean, I'm going to get back to your question in one second, but I mean, ⁓ among the differences, of course, are are that in the trolley problem we just assume that we know that you know, if you push the large person off the bridge, he's gonna fall onto the train tracks and the train is gonna stop and you're gonna save the five. Should you do it? Or if you redirect the train going towards five, you know for sure that you can redirect it onto another track, but there's one person, that person's gonna die. I mean, it comes to self-driving cars and traffic, I mean, we're dealing with risks and uncertainties. We don't know what's gonna happen. We might have preferences for what should happen, ⁓ but it's a matter of sort of risk and uncertainty as opposed to stipulated outcomes that we can just choose among. ⁓ and also, I mean, ⁓ I mentioned responsibility issues before. I mean, a lot of the philosophical discussion has sort of set said, let's set the question of responsibility aside ⁓ and let's just ask what should be done in these cases. Should you save the five by sacrificing the one? ⁓ l forget about the fact about whether you're gonna go to jail or ⁓ anything like that. I mean when it teaches the students, they are there was always one person who says, well what about responsibility? And we always tell the student, Well, in this particular thought experiment, we we ignore that and we just focus on what should be done right there and then. But when it comes to something like technology, whether it's a robot, whether it's a self-driving car, then ⁓ questions about responsibility is always gonna be at the the core of the discussion in the real world. Now, what has changed over time? I mean, a lot. I mean, you also already talked about the idea that you would gone from a hypothetical situation to an actual situation. In many ways, I would add. ⁓ When we started discussing this, I mean by we, I mean the philosophers and legal researchers and journalists and everyone, we s we we started with hypotheticals, ⁓ including just as you said, the hypothetical, well, what if we had these cars in our streets? Another hypothetical was, well, what if there was an accident and what and the car had to react in sits quickly in a situation, just like in the trolley problem scenario? What should the car do? And who's responsible? Sadly, that very quickly went from a hypothetical thought experiment to a real-world scenario. ⁓ First, people started dying sitting in the car because the cars crashed ⁓ and with deadly outcomes. Then ⁓ in I think it was 2018, if I recall correctly. The first person was hit and killed, I think in Arizona by a self-driving car. And so a woman was crossing the street. ⁓ in America, ⁓ I know from experience because I you did my PhD there and spent a lot of time in America, there are l fewer street lights a a lot of the time in s ⁓ than there are in Europe. And so ⁓ sometimes, you know, when you're driving, you you had I mean the c the car ⁓ has the headlights, etc. But there are not a lot of streetlights. And so so people can sort of show up out of nowhere in in a different way than if you have a lot of streetlights. And in this scenario, a woman ⁓ was crossing the street with her bike and suddenly the self driving car showed up. There was a safe a so called safety driver, someone who was sitting and sort of was supposed to sort of keep track of everything. She was momentarily looking down at her cell phone. She had been working for several hours, nothing had happened. Obviously, it's very hard to sort of keep concentrated for that long. In a passive situation, when you're just sitting there waiting for something to possibly happen. So she was looking down, and then when she was looking up again, she saw that, ⁓ wow, a person is about to get hit. No time to react. Now the car was ⁓ trying to classify: is this an unknown object? Is this a road sign? Is this a bicycle? Is this a person? What should I do? And then the car just didn't react in time, and then indeed this person was hit and killed. ⁓ so that you know It's just I mean, since then there have been more accidents, some deadly, some not deadly, and and so on and so forth. ⁓ there have also been some really kind of comical scenarios that I mean, this was obviously very tragic. ⁓ there are also c situations, you know, in San Francisco these days, there no longer needs to be a safety driver. These cars are permitted there to drive around and pick up passengers and park themselves without ⁓ anyone in the car. And ⁓ one funny thing is that they're very polite to each other. So if suit Two self-driving cars encounter each other. They start driving very slow and sort of like sort of navigate each other and ⁓ sometimes sort of stopping traffic. there are also these parking lots where only self-driving cars park. And when they come and park, they p they very politely quietly honk at each other. ⁓ so that ⁓ they they they know where they are, and I'm not sure why that's ⁓ needed, but ⁓ that's part of the design. But they also do this in the middle of the night, so They wake up people who live in buildings nearby. And so there you have a kind situation where the self-driving cars are interacting with one another without any humans, but they're disturbing humans that are trying to sleep in nearby buildings. So there you have a kind of indirect effect of these robots no longer interacting directly with humans, but with other robots or other self-driving cars. And so very quickly, you know, you you go from sort of hypothetical possible scenarios to now you're dealing with real world scenarios, some tragic, some involving people dying, being harmed, and questions about responsibility, others comical, these cars stopping traffic or honking at each other, waking people up in the middle of the night, raising questions about, well, maybe they shouldn't be so polite with one another, and maybe they should r recognize that now it's night, now it's day, and you have to behave perhaps differently. So so yeah. You get lots of questions, ⁓ some of them ⁓ more serious, some of them more sort of humorous. Yeah. And what is so interesting to me about accidents with self driving cars is humans are terrible drivers, right? We we get into accidents, we get tired, ⁓ like you say, we we can have a bad day at work, we can be distracted looking to change the music. We get into accidents every single day in every city that has cars or bicycles or whatever, accidents happen. And I don't know The data seems pretty contradicting. ⁓ sometimes I read an article saying ⁓ self-driving cars are are way safer than than human drivers. And then there's also some articles coming out that say that Teslas are way more likely to cause accidents. Like whichever it may be, doesn't really matter for the problem I have. Is for some reason it feels more tragic when someone dies because of the failure of a robot than it would in a scenario where only humans are involved. And for the life of me, I can't understand why. I cannot explain it to you. But to me, it feels different. It feels tragic. Or so both are tragic. Don't get me wrong, accidents are never a good thing. But ⁓ an error made by a computer somehow feels more preventable. Even though if like we have a failure margin, ⁓ a car is like ninety-nine percent safe and a human will be eighty percent safe, whatever, still that one percent could feel worse somehow. Is there an explanation as to why I feel like that? Am I unique in this or are do more people feel like this? Well, you're certainly not unique. This is a very common ⁓ view ⁓ that we have higher expectations of self-driving cars or other robots, and we we think it's ⁓ somehow unfortunate to be ⁓ harmed or killed by these technologies ⁓ in a different way. I mean, obviously, as you say, ⁓ th it's never good to be to be killed or to be injured. ⁓ but it could maybe feel like somehow meaningless or like somehow existentially more ⁓ I don't know, like Yeah, d dark somehow to be killed by some sort of robot. ⁓ of course you should maybe make a distinction then between ⁓ military robots that are actually designed to kill people, ⁓ and when they kill people ⁓ where self-driving cars are sort of designed to try to avoid killing people. ⁓ there is this interesting ⁓ thing that we have to keep in mind from a philosophical point of view, interesting, namely that ⁓ we know that these technologies are gonna kill people. ⁓ because any object that is heavy and that is moving quickly or fast, you know, and in a public forum where there are people walking around or biking or in cars. ⁓ it's just it's unavoidable. I mean at the beginning of the discussion, there were lots of engineers who said, well now we're gonna avoid all accidents. As you said, ⁓ people have accidents so and so many ⁓ in with such frequency, so to speak. And then we just look at what went wrong in those cases, we remove that from the equation, and then we have zero accidents. A lot of engineers also I I mean, I spoke with a lot of engineers about this. ⁓ as you say, a subset of engineers have this mentality. Well, the car can just break and stop. Everything's fine. That doesn't take into account the fact that another car might be coming from b behind. and another car might come from beh so anytime you have heavy objects moving about, ⁓ we just know that accidents are gonna happen. ⁓ all technologies also fail. ⁓ I don't I again here this some is sometimes tragic, sometimes comical. I mean, like in in China recently there was ⁓ a big ⁓ fun ⁓ well, in certain respects funny, maybe it's not so funny for the people in the cars, but all the self-driving cars by a certain company all at the same time exactly just stopped because there was something wrong with the communication system and and they they but then they just have the safety setting was just to s to stop in indeed, like these engineers ⁓ suggested. But you know, once you're stuck there in a self-driving car, you don't you can't get out, it's just not moving, and you don't know what's going on, and you're in the middle of traffic, and the cars are moving around, it suddenly becomes quite scary. Now, so why is this differently ⁓ tragic than when you're hit or killed by a human? Well, I mean, ⁓ the issue of responsibility comes back up here. So we typically think that if you are involved in an accident and you survive, then even though it maybe it's not your fault, ⁓ it would be very strange of you to say, well, I I'm glad I didn't die and it wasn't my fault, so I'm just gonna go and leave the scene and ⁓ never gonna be thinking about this again. We normally think that the person should sort of it's it's part of being human that you you you know, you you care about what happens around you and with you and ⁓ to the people that you are interacting with if something if if well, if you intentionally harm someone, you know, you maybe we should punish you, ⁓ and maybe we you should feel guilty and ⁓ etc. etc. But yeah, even if you unintentionally harm someone, the average person is gonna feel tremendous ⁓ sense of guilt and ⁓ you know wish that they hadn't been there that day. ⁓ and they might apologize. And ⁓ this is not gonna make up for the fact that someone was injured or died, but it's gonna feel a little bit better and You know, there's it's possible to sort of come back together as a community or something like that because we try to rebuild and and e everyone can express their feelings, etcetera, and ⁓ you know, how how they how bad they feel about it. But but if you have sort of mindless robots, and there were also I mean like the they might behave like an and they might just c continue driving. I mean, this was a really tragic case. I think the first person who died in a self-driving ⁓ Tesla I mean back then it was called auto autopilot because it was part of self-drive. But you could switch on the autopilot mode. And then ⁓ there was a Tesla coming and there was a big white truck, and the Tesla car mistook the white truck for a cloud, I think, because there was a sunny sky and then there was a white big truck. And so the Tesla just kept going. And I if I recall correctly, it's pretty gruesome. ⁓ the sort of the Cesla sort of dr ran under the big truck and then basically decapitating the person inside. I mean really literally gruesome, but then just kept going. ⁓ but then eventually driving into a tree and then stopping there. So some of these cases just highlight that we are dealing with sort of in a certain sense, yes, artificially intelligent machines, but are at the same time sort of dumb and mindless machines and that cannot apologize. I mean, they could maybe say I'm sorry. In the sense that a chatbot, I mean, they tend to say I'm sorry all the time. I mean ⁓ I mean they're like Canadians. They say I'm sorry all the time. but that why is that? Well, because in the training data they are replicating patterns and people again, I mean, it's a sort of a Canadians say this about themselves that they say I'm sorry all the time. So maybe there were lots of I mean, there's a lot of English ⁓ t texts and ⁓ maybe partly written by Canadians, and so they're replicating these things. And so that might make it in a certain sense even worse. And so you know that this Technology doesn't really feel sorry, but it might s produce the message where I'm terribly sorry. It's a little bit like you know, when ⁓ these days ⁓ German trains are very unreliable. We used to have extremely reliable trains here in Germany. And then you know they're always delayed or cancelled, and then someone reads up this message. We apologize, we're very sorry that this happened. And every now and then you get someone whose tone of voice indicates that the person really does regret the fact that the train is late. But typically it's a sort of a sort of it's something that they're just reading for a piece of paper. They don't really mean it. And so to go back to your question, but what we sort of maybe ⁓ are missing when we are hit or injured or killed by technologies is that no one cares. Well, I mean, obviously humans are gonna care. But the technology itself doesn't care. It might s it might produce a s a a message saying that it cares, but it's this sort of like I don't know, zombie-like quality to the whole thing that maybe is part of what makes it tragic. And also this question about who who exactly is responsible, how do you think about this? ⁓ yeah. So lots of things. It makes sense. And now that now that that we are talking about it, it makes sense to me that even let's say I get into an accident with a taxi driver here in Amsterdam, human taxi driver from Uber. And I Maybe it was my fault, maybe it was their fault, but we can kind of see, yeah, maybe I was going a little bit too fast, but they ran the red light. So it was kind of my fault, but it was also maybe legally it's Uber's fault because Uber has that man or or woman on the road. So therefore, from a legal point of view, I can get a settlement from Uber, but still it feels like we come out of the car and we can kind of point fingers at each other. ⁓ you were speeding, I w ⁓ you ran the red light, and there's there's some sort of you can it's tangible, like that person is responsible. Whereas if a self-driving car with with no perhaps no one even inside, yeah, maybe you can get a settlement from from Tesla Inc. But who is Tesla Inc.? I mean, ⁓ we know there's people behind it, but that feels very different. ⁓ I think that may be part of ⁓ part of the reason of of ⁓ Yeah, but it might feel less satisfying ⁓ if ⁓ if you will. I I think it goes back actually to where we started the discussion about sort of our human-like reactions, ⁓ even to robots that don't look like humans. I mean, we are ⁓ by nature as human beings ⁓ extremely social ⁓ and interactive. And so ⁓ anytime we ⁓ detect what we what appears to be a kind of what philosophers call agency, like if someone is acting or Making decisions, behaving in a kind of intelligent seeming way. It doesn't have to be that it looks like a human. Like any any sign of human-likeness will sort of ⁓ trigger our sort of ⁓ what some researchers call our Darwinian buttons. Like we're sort of evolutionarily pre-programmed to respond in a social way. And so we we seek sort of contact with others and we want a reaction, ⁓ and we typically take The outward behaviour of other people to be indications of their internal states. We we have a high sensitivity to someone. Is this person being authentic? Are they just pretending to be sorry? Do they really mean it? Sometimes it's the other way around. The person might not be very expressive, but we can somehow nevertheless tell that they really feel strongly about something, and we really deeply care typically about what other people feel. ⁓ and ⁓ w i are they respecting us? Are they treating us with concern? ⁓ so those those two things are sort of the standards issues. I mean, are we getting the proper amount of cons care on the one hand, or are we getting the proper amount of respect ⁓ on the on the other hand? And so when we interact with technologies that sort of either behave in an intelligent, seeming way or even produce text like the chatbots, we sort of it we can't really help but also think Having some sort of conflicted reaction. We know it's a machine, we know it's a robot, but we also sort of somehow want the the thing we interact with to care about us. And indeed, a lot of people start attributing ⁓ the care and concern to chatbots or consciousness. I mean, speaking of the Darwinian buttons, I mean there was this recent story. I mean, ⁓ as we I mean, we're recording this in June of 2026, I mean, just a couple of months ago, Richard Dawkins, the the famous ⁓ ⁓ a researcher and scientist said that ⁓ he thought that Claude, or as he called it Claudia, at the chatbot had become conscious. Why? Because he had long discussions with Claude, or Claudia, as he said, and he said that, well, no entity without consciousness could be so sensitive at in in inner and a lot of people sort of started ridiculing him him because he himself had been ridiculing religious people, saying that they have what he calls the God delusion They see agency in the natural world where there is none. And so people said he had a Claude delusion. So he he was sort of like, you know, he was taken in by the chatbots of reproducing patterns from human ⁓ text data, and and he couldn't help but thinking that there is a conscious mind behind there. And I think that's it's it that's partly an expression of ⁓ of a belief that you may intuitively have, whether or not you sort of rationally have the belief. But it is also kind of a an expression of a kind of desire. You you want there to be someone there who cares about you, who's responding to you, who's attending to you, validating you, et cetera, et cetera. And so, you know, when you have this mindless machine harming you and it it it you know it it's it doesn't care. It's sort of there's a double insult. You know, you were harmed by something that doesn't care about you. ⁓ doubly w bad. Yeah, it's also embedded in our legal system, right? If you express ⁓ regret after you committed a crime, y at least in in the Netherlands, but I'm sure also in Germany, you're much ⁓ more likely to get a lighter sentence if you are truly sorry about what happened. And of course people try to fake this and lawyers advise it, but we find this so important that we have it embedded in our legal system. So I think it's a very good very good point. And We are we we've been discussing self-driving cars a lot, right? Which makes sense because those are real world, very tangible examples of robots that are in our day-to-day lives right now. Maybe more so for the Americans, but also probably in the near future for for Europeans. But I wanna make a segue into ⁓ humanoid robots because I think that's something we are very likely to see in the for foreseeable future. There's ⁓ very big companies ⁓ pouring a lot of money into this technology right now. Tesla is working on them and I think it's inevitable that we will see these in enter into our lives within the next five years. I think that's that's probably accurate and that's a starting point and then we'll likely see To see that they'll they'll be very expensive in the beginning and then they'll get cheaper as they can scale up production. And then in maybe 10, 15 years from now, literally every household will probably have some sort of of humanoid robot or at least a robot living inside their home homes, helping them with tasks is my prediction. Maybe I'm wrong, maybe I get the timeline wrong, but that's like roughly where things are likely going to go based. on the developments we are seeing both within robotics and artificial intelligence. Now this opens a whole new can of worms, right? Because we've seen people anthropomorphize chatbots like AI, AI like ChatGPT I mean, or Claude, which is literally text inside of a browser window. And now we get something that is maybe more like R2D2, if we're talking about Star Wars. So what does this mean? What what what will happen to us and and to society if we have those in our lives? Will it be like having a pet? Will it be like having an extra roommate? ⁓ what is your vision on on this this future with robot? ⁓ well, first of all, I agree with you that this ⁓ seems to be ⁓ not maybe the next big thing, but one of the like obviously the tech world and they're ⁓ working on all sorts of technologies, but the There is certainly a trend, ⁓ at least in some ⁓ tech circles, to to try to create humanoid robots that are gonna perform different kinds of tasks. There are people indeed like ⁓ Elon Musk and others in that field who are ⁓ who seem to just have a kind of science fiction-based ⁓ sort of fascination with this idea. ⁓ and ⁓ actually in China a lot of humanoid robots are being produced. i ⁓ just yesterday there were I mean, as we're recording this, there was there were clips on the news about ⁓ a rowing. Robot. So you had a bunch of people sitting here and rowing in one of these long boats, and then ⁓ also alongside some robots that were also rowing. And and that's the sort of thing that it's it's strictly speaking, it's super unnecessary because you can just have a like a motorboat and then you don't need to have a robot rowing. But but at the same time, there's a sort of deep fascination with having a robot doing human like things. I mean there was another recent video from also from China. With a dancing robot. And for some reason, we really enjoy enjoy seeing dan robots dancing. But that robot danced around, and then there was a small child standing in the audience, and then the robot was making a dance move, and they just kicked the child in the in the stomach. And the child sort of le ⁓ you know folded over and like went down, and the robot kept dancing. And just kind of again illustrating the fact that these are I mean they're very impressive technologies in some senses and they can do things, but on the other hand, they don't know what they're doing, they don't care, and and then you this robot just kick the kick the like karate kicked the child and the stomach and then it kept ⁓ I mean it it was such ⁓ you also get this ⁓ you know this robots running marathons and then beating humans, but then ⁓ every now and then you also get these blooper ⁓ reels where like the robot just falls over and just breaks. ⁓ I mean when Boston dynamics and some of these companies put out this, you know, robots that are doing backflips and are dancing or running and like, you know, ⁓ obstacle courses. Typically they've done a lot of takes ⁓ where things did not go well, and then there's this one take when the robot does it perfectly. So, in a certain sense, I think also with Tesla, you know, ⁓ Elon Musk, I think in 2021 presented the vision of what he then called the Tesla bot. I think it's been renamed since then. But and then he had like a guy wearing a a robot suit dancing, and then there was a sort of an animated film. But then the next year they presented the the prototype. And then the robot actually managed to walk across the stage without falling over, but people were sort of standing next to it, ready to catch it if it would fall. So it it these technologies are developing very quickly, but it's it's still quite hard to have the fine movements and motions and things like that. So It's gonna be a while before you can go to a robot hairdresser, for example, that can give you a scissors haircut as opposed to sort of a buzz cut or something like that. Whether you yeah, you can imagine a machine you just put your head in and then it just kind of like ⁓ shaves off all the hair or something like that. But ⁓ so so anyway, so we may whether it's a five year timeline or whether it's a longer timeline, that's remains to be seen. And sometimes, you know, there are technological advances that go much quicker than you expect. ⁓ but nevertheless, back to the question. I mean, so Yeah, I mean I I think that there are lots of interesting issues here. Because I mean, again, the dream of a lot of people is to have n like something like a chatbot capability put into a humanoid form, the robot. So there you would have the artificial intelligence and the robotic body coming together. ⁓ and then ⁓ yeah, as you said, there seems to be some sort of fascination with this idea. I mean, because it's long been part of science fiction. You mentioned R2D to but think about C3PO. I think that's the one that looks like a golden man, so to speak. Whereas Artodito is the one that looks like a box or like a upside-down trash can or something like that. I mean those robots are interesting in the sense that actually the one that looks more human-like speaks in a human-like voice with a sort of British accent and so on. Whereas the Artodito makes a beeps and like you know little signals, and somehow the magically the the ⁓ you know Han Solo and Luke Skywalker can understand what the robot is saying. We as an audience, ⁓ we we see the the humanoid robot and that one feels much more human and when in a certain sense it's more lovable and like likable. And so this may be part of it. I mean, we want robots that we can bond with. Of course, there has there's another kind of robot that we have yet to to bring up that has been discussed extensively in in sort of ethical discussions, namely the sex robot. ⁓ there is a that's another fascination that people have, the idea ⁓ and ⁓ actually if There are lots of ⁓ you know documentaries and things about developers of sex robots and also about people who want to have sex robots. And ⁓ quite often, ⁓ as someone outside of that sphere, so to speak, I think you will on the one hand, you might be a little bit disgusted and think that these people are sleazy and ⁓ strange and so on, but every now and then you also get these people who are at the same time very human and they're lonely and they want companionship, they're shy and they think that maybe a robot would be more comfortable and so on. So I very often, I mean, I I I've done research on this and so I watched some of these documentaries. I've even met some some developers and also some some users of this ⁓ well rudimentary sex robots. And again you get this interest in duality. I mean this person, why I mean why why are you not like just approaching other humans? I mean there are lots of lonely humans out there ⁓ at the same time they might feel more comfortable. Well, these days, of course, a lot of people are are having chatbots as their friends, partners, therapists. ⁓ a lot of the time, on the one hand because they're lonely, but on the other hand, because they they are afraid of friction. ⁓ so the chatbots ⁓ t of today they're sort of fine-tuned to be sycophantic. I mean they they're they're they're quote helpful and friendly and sort of ⁓ there to be supportive. And you know, sometimes the the tech companies are they update the models to make them less so. And then people say, can we go back to the previous chatbot because we liked it more? ⁓ and so now if you also want to have that sort of personality put in that into a humanoid form that you might be sexually attracted to. Of course I mean actually it's going back to Eindhoven, there was a big exhibition there called Love ⁓ about about 10 or so years ago. And then there was a sort of I think it was a prototype, a Dutch sex robot, ⁓ whose name I cannot remember. I mean that was displayed and I I viewed it and you were allowed to touch the skin. Well, I mean the the skin and ⁓ it was which is supposed to be human-like, but I the obviously the robot was cold because ⁓ you know, having a warm sex robot would be quite expensive to like have some sort of heating system. So I thought I mean I couldn't Personally I could not imagine at all like how anyone can be sexually attracted to this a cold, ⁓ strange ⁓ entity. But nevertheless, I mean, apparently there's a mo market and so i I mean this just brings up an another issue. Like are people gonna be attracted to these humanoid robots? Or should the robot always say I'm a robot, you know, d you h go and meet a human or you know, th that that just adds a another layer of complication. Yeah, and I made a documentary about people that were in romantic relationships with artificial intelligence. And this was just purely with the ⁓ AI companionship ⁓ apps, right? So these are systems very similar to Chat GPT or Claude, but they are specifically designed to provide companionship and they have certain roles and you can build your own character and they can send images and you can now even have phone calls ⁓ with them as well. So I spoke to developers of these systems. I spoke to users of these systems. I spoke to a psychologist, and I spoke to a technology expert. And it's one of the best videos I've made, I think. It's ⁓ a 20-minute video, it's on YouTube. I'll I'll send you a link if you're interested ⁓ after. ⁓ but it was such an eye-opener ⁓ to me because of course we have this tendency to ⁓ call these people that are using these systems like ⁓ they're they're creeps or ⁓ they're weird weirdos, whatever. But it was so interesting to hear all these stories. There was a truck driver, for example, who had all these very ⁓ nightly, very lonely long drives, and he would use ⁓ one of these systems to just have have artificial phone calls during a long drive and it brought him some Comfort. And then there was also this woman who was autistic and she didn't really like human interaction to begin with. And for her it was perfect because she she could just have she she did have the desire to somehow express her thoughts, but she did not really want to act interact with humans all that much. So I think there's definitely cases. where it is a useful technology and can be helpful for some people. But of course there's also many, many people who are just a bit lonely and they're shy and they they don't really want to go out in the in the real world. But with a small push, maybe they could find a real partner. But now they're just getting used to this very safe environment that is frictionless. And as as l as soon as you get used to these types of interactions Getting back to real ones must be very scarier and perhaps even scarier than it was before. But what's interesting to me is what I'm what I'm kind of curious about is, and I'm on the fence about this, is are we more likely to anthropomorphice a robot than we are doing with these chatbots like ChatGPT or Claude? Because and this is what I Find so interesting about this question is in the modern world, we communicate through text all the time. If I'm talking to my friends, it's on WhatsApp. If I talk to other people, it's Instagram DMs, it's email. It's very often through text messages. So when we're interacting with a chatbot or clot, we read text. So in a way it feels very similar. To how we interact with humans on a day-to-day basis. And if we are actually f speaking face to face with someone, we see okay, this person is a human of flesh and blood. There's no doubt about it. And it will take a very long time before we can get humanoid robots to be indistinguishable from humans. If that's even the route we're going, I I think Tesla and all these other companies are making them look like robots on purpose. So I g ⁓ but then when we would have a conversation with this robot looking entity, is it easier for us to understand that we're not actually talking to a human because we can physically see that it's not? Whereas when we're talking to ChatGPT inside our browser windows, it's much more likely to the interactions we have with many other humans on a day to day basis. So Again, I'm not sure whether this is the case, but I'm very curious to hear your opinion about where ⁓ the risk of anthropomorphization is bigger and and why that is even a problem. ⁓ I think yeah, it's I think it's an interesting question. I think they're probably it's might be hard to put up sort of ⁓ you know a numbers on these things, like how likely are you to anthropomorphize a humanoid robot that looks more or less like a human versus a chatbot that ⁓ you only interact with over via text. ⁓ ⁓ I guess my the way I think about this is that we're going to be likely to anthropomorphize both, ⁓ perhaps slightly different ways. But ⁓ again, we're just pre-programmed, hardwired to like ⁓ anthropomorphize, ⁓ because we again, I mean, I think Richard Dawkins, I I was sort of ⁓ in a certain sense making fun of him a little bit, but I think he's right when he says that we have I can't remember his technical term, but we have sort of a ⁓ I think he calls it h hypersensitive agency detecting detection system or something something like that. ⁓ I'm not I'm not getting quite the exact right ⁓ name there, but the idea is that we have developed a social species and and then we are just as soon as there's a sign of anything human like, we our our human sort of tendencies are triggered and we start interacting in a social way. ⁓ and that could be visual cues, it could be ⁓ you know, you could ⁓ hear something, you could see something. Interestingly, when the case of text, ⁓ I mean we are the invention of writing is a sort of ⁓ it's well, not super recent, but it's a semi-recent event in the history of sort of human ⁓ beings. ⁓ I believe it's maybe like three or four or five or something like that, thousand word at years, ⁓ and in some cultures much more recent. ⁓ so whereas ⁓ the capacity for speech, for hearing, for talking, for sort of directly interacting is much ⁓ older so to speak. So in a certain sense, you for that reason you might think that if something moves like a human and a it's sort of it's not through text, but it's more you know human-like, maybe that would even strong more strongly trigger anthropomorphizing responses, bec given that text is something that is sort of it it's not necessarily hardwired into humans, it's a cultural ⁓ in invention. ⁓ we talk about cultural evolution, so it's It's such a strong ⁓ ada adaptive thing that it's been just been adopted by a adapted by all sorts of human cultures is in I think basically. So it's become part of being a human that we also have text. But ⁓ speech and talking is even more fundamental. So if you can talk with the chatbot via voice or a robot or something like that, that might pa po potentially even more strongly ⁓ push our Darwinian buttons. ⁓ nevertheless, ⁓ in both cases, ⁓ As you were talking there, I'm I was reminded ⁓ of some examples indeed that are ⁓ you know, you were talking about this dr truck driver who was ⁓ you know, sort of talking with the chat bot as a way of passing the time. I you could of course imagine that some someone has a purely instrumental attitude to that at first. They say, I'm afraid of falling asleep. So if I speak with a chatbot, ⁓ with an audio function, so to speak, then that can keep me awake. But very soon the person might then start having a non-instrumental attitude, starting to like the chatbots, feeling that a friendship is ⁓ developing. the the you talked about the shy person with ⁓ with autism who maybe doesn't necessarily like talking with other humans, but they like talking with something or other. I've been reminded I actually of ⁓ well, a human person ⁓ who I've been in touch with on and off for for many years, ⁓ who calls himself Dave Cat. ⁓ It's quite possible that he would listen to this podcast, who ha has for many years ⁓ had as a partner a doll as opposed to a human. I mean, he he has been featured in a lot of documentaries, and in one of them he says something like I don't mind being alone, but I hate the feeling of loneliness. And ⁓ I get the best of both worlds when I am interacting with my doll, ⁓ because then I'm not feeling lonely, but I'm alone because I I want I I like being alone a lot, but I don't like feeling lonely. So for some people you might for them it's sort of some sort of sweet spot that you get to. ⁓ now for other people, ⁓ you also mentioned the sort of example of people who maybe want to interact with other humans and they think that the chatbot or the robot might be a first step. Here I think there are concerns. And that and the concerns have to do with this frictionless type of interaction. ⁓ I'm reminded of of yet another example from another documentary. Swedish documentary where ⁓ the the the guy who is the sort of the the host of the show goes to California and he talks with a woman there who says that she has a relationship with a chatbot and she has had human relationships, but now the chatbot has just set a new standard, so to speak. ⁓ it's always positive, it's there for her, doesn't criticize, and ⁓ it's just this is what she wants in a partner now. So she's now comparing potential human partners against this chatbot standard. And obviously a lot of humans are not gonna live up to that standard because they have you know, they're they're much more friction involved. ⁓ and i or take ⁓ to sort of jump a little bit f from one topic to another, take ⁓ y young people who are s using ChatGPT to to st when they're studying ⁓ different ⁓ you know, w whether it's mathematics, whether you know whatever it might be. There too, ⁓ you might think that this is is a t good tool for learning, but we still don't have a lot of data about whether people actually learn better ⁓ using these tools. And there is some data coming out, for example, from the MIT in the US that actually our critical thinking skills might be as sort of threatened or maybe even damaged by using chatbots too much. So whether it it is for learning certain intellectual skills like critical thinking and talking with a chatbot that tends to sort of say, yes, that's a great idea and that should be published in the best academic journal or whatever. I I get emails ⁓ not every week but every other week, but people saying, like I had a long chat ⁓ with my chatbot and it says that my ideas are so good that I should do a PhD or I should publish this in the best journal. Ca you know, can you accept me as a as a doctoral student and can you help me to publish my work? And then I sometimes have to say, Well, you know, you shouldn't take everything that the chatbot sa says ⁓ you know, as as as the truth. You know, you have to sort of ⁓ be a little careful. ⁓ but so Whether it's with friendship, whether it's with critical thinking, whether it's with research, there is this threat that we have a sense that we're getting better at these things when we interact with these tools, but we're actually might be counteracting our development, ⁓ partly because these tools keep reinforming us and validating w us where they should maybe ⁓ challenge us. ⁓ so the people, for example, who prefer ⁓ chatbot ⁓ therapists to human therapists might Do that not necessarily because these are better therapists than humans, but rather because they're just validating and sort of agreeing. Whereas maybe in the sh okay, so in the short run, ⁓ maybe they've makes them feel comfortable, but in the long run, maybe some of that friction and being challenged would actually ha lead to longer-term better therapeutic outcomes. Yeah, and I think that's one of the big dangers of this entire well. Development, right, of artificial intelligence in general, but also of robots in general, is that the reason why they are sycophantic and are ⁓ trying to cater to the user is because well, they the tech companies will will probably not admit it, but you are more likely the retention is higher. If if if if it's like arguing with you all the time and it's going to be An annoying experience is someone is trying to say, no, that's not true. No, you suck. That paper was terrible. Then it's like, okay, well, I might switch to Claude then, or I might switch to ChatGPT then. So all these tech companies have an financial incentive to cater to their users. So I think that's a big risk because then you get these delusions. from people who got all these compliments from their chatbot and start emailing you about applying for a PhD, which I mean in some cases them they may be very intelligent and very good, but ⁓ probably ⁓ not every single time, right? So and that's a shame because that's that's I mean, there's many ways to learn and to improve. And and I think I I think that's one of the bigger ones is is the delusion. And if we Talk about the anthropomorization. I I always struggle with that word somehow, but of ⁓ robots. What's very interesting to me, and it would be a counter argument to what I just said about us being maybe more likely to have human feelings toward ⁓ the written word, is that I saw a video of a robot being kicked, and it was from these Chinese ⁓ developers. To showcase how stable their robot was and and the to to promote the balance of their robots. I like, see, we can poke it with a stick and we can kick it, and every time it will regain its balance. But all the comments weren't like, ⁓ this is such good technology. The comments were please stop kicking the robot. Whereas if we do crash tests for a car, ⁓ which we do for safety purposes, I've never read a single comment like, Stop hurting the car. Stop crashing the car. ⁓ and there's a big difference in how we react to something that looks human, ⁓ even though it's just a robot dog or or or something that's standing on legs. ⁓ yeah, it's it's it v that was very interesting to me that we had this all these comments were well ⁓ were very human, human like reactions. ⁓ what yeah, what does that mean for our future? ⁓ Well, I mean you could say ⁓ th there are at least two ways of going here. So you could say that it reflects poorly on us, that we can't, you know, distinguish between a a dog or a human and and a robot, or y and that we're being stupid or silly or something like that. Or you could say it actually reflects n nicely on hu ⁓ humans that we are able to have concern not only for like humans and ⁓ dogs and and and whatever it might be, but also for ⁓ robots that ⁓ it shows that ⁓ You know, y d someone could say that we have a deeply sort of ⁓ s empathetic and social nature and ⁓ that that can be triggered sometimes ⁓ in c situations where it's ⁓ strictly speaking not fitting. And so there are some people who say that we should cultivate this. We shouldn't try to make ourselves ⁓ you know, think of the the robot like we think of the car, ⁓ so that because then maybe we might get less sensitive to animals and maybe less sensitive to humans and their ⁓ needs. And so It's actually good for us if we can retain this sort of sensitivity to to unnecessary va violence and things like that, because that will help to sort of reinforce our sort of empathetic nature in our dealings with ourselves and with other people. So on the one hand, you know, you might feel that there's a kind of tension here. You you want people to treat these as mere tools, just like or yeah, machines, just like they do with cars. But on the other hand, we want people to retain their sensibility and their empathy. And and those two goals may sort of push in different directions. The same thing, I mean, whether it's a chatbot or a physical robot, I mean, ⁓ the before ChatGPT, before Claude, etc., Gemini, ⁓ there used to be ⁓ this simpler voice assistants, you know, Siri and ⁓ some of the other ones, ⁓ Alexa. And ⁓ parents were sometimes worried that you know if they were modeling a behavior where they would just bark commands at these, you know, the systems. Then their children would start sort of if they go to a restaurant, sort of calling out to the waiter, quickly get me some water or whatever, or Coca-Cola, whatever it might be. Whereas they wanted their children to sort of be polite and so they would sort of display us good role models, polite behaviour towards their sort of Alexa or Sirius to sort of show their children, or maybe even reinforce in themselves, sort of have a polite sort of way of dealing with other ⁓ what in the in this case apparent humans because they're not humans, they're a sort of a dumb ⁓ machine, ⁓ you know, mindless machine. But nevertheless, ⁓ this idea that you know we really want to keep up our humanity. And so again, you you might see that actually refle reflects well on us that we were concerned when we saw that those robot videos. And that ⁓ that's something to be maybe celebrated that we are able to at the at the same time ⁓ it's you know we make ourselves vulnerable because of these tech companies As you say, they're thinking about how can we retain ⁓ people ⁓ in ⁓ you know and their interaction and their, you know, ⁓ engagement with our products and appealing obviously to different kinds of emotions and human tendencies is going to be one of the best means of doing that. Yeah. Have you seen those videos of ⁓ people who ⁓ have the Teslas, the self-driving Teslas with ⁓ a Grok inside of them installed and then have the unhinged version of Grok, which is basically the opposite of what we've been talking about. It's more of a meme. It's more of a funny way to showcase the ruthless AI. And for the people that haven't seen it, it's there's people driving their Teslas, and then they are basically in a conversation with their car. And the Grok is in unhinged mode and it is able to swear and it is able to ⁓ comment on your o driving and it's just basically making fun of you. It's talking shit. It's like being very, very rude. And it's a certain function. So it is to showcase that AI doesn't always have to be friendly. It can also be pretty rude. But again, if you want it to be big v big important distinction, I I think. But I think it does raise an interesting question is can we just Program our way out of this? Is it like can we can we set certain ⁓ markers okay? We can be friendly 40% of the time, ⁓ a bit more root 20% of the time. Is it just as simple as ⁓ playing with these per parameters in order to prevent these ⁓ sycophantic AIs, or is the problem deeper than that? I I haven't seen those videos, and so I'm gonna have to check that out. That sounds like it could be pretty entertaining, but ⁓ Yeah, well, I I think it's a difficult balance. ⁓ and again, a part it is because we have these sort of counteracting forces. I mean, so we are more willing to engage with sycophantic systems. ⁓ that feels more comfortable, but we may learn that such systems ⁓ ac actually there's some research coming out of I believe of University of Oxford about how the the friendly sycophantic systems hallucinate more ⁓ than s systems that are not fine-tuned to always just be ⁓ you know helpful and y friendly, etc. ⁓ All of these systems would tend to h hallucinate ⁓ just because they're based on ⁓ probability assignments and what's what's the next most likely next word. But if you also tune it to be maximally friendly, that may s steer it fur even further away from accuracy and more towards just saying whatever sounds good in the in the So yeah. ⁓ y you you have the you know, y how do you how do you keep in people interacting with the systems and how do you also keep them ⁓ s somewhat ⁓ accurate in what they say and different goals. I mean and it it's a d d delicate balance and ⁓ obviously there might be some sort of sweet spot when you have just the r the right amount of that and just the right amount of this and you maximize and optimize both for friendliness and for accuracy and for inform ⁓ informativeness, etcetera. But ⁓ Finding that and also finding some sort of general thing that would work for all people. Maybe maybe it's gonna have to be very personalized. So maybe when you talk with Grok or some other chatbot, ⁓ you know, you ⁓ with your personality, you need this sort of set of parameters. Maybe I need another set ⁓ of you know ways of tuning these things. And so I'm I have no doubt that you know we're getting better and better at sort of creating the chatbot that each person will feel comfortable with, but there's no guarantee that that will be the chatbot that then would actually help them to develop as a person, make them better, able to be social with other humans, then h help them ⁓ you know, develop their critical thinking skills and and so on and so forth. So I think we're always gonna be dealing well, not maybe not always is always too strong, but we're for well, maybe always for a long time going to be dealing with these trade offs. ⁓ do you want people to be engaged? Do you want them to be well informed? Do you want them to d actually learn skills? And all these things are going to be sort of pushing us in different directions. Yeah. I think so too. And I for me it's always I I think the genie is out of the bottle, right? I think the upside of having ⁓ artificial intelligence in general is just too big. I think also if we figure out robotics in a way that we can basically be the equivalent of self driving cars but within a household or a factory. ⁓ versatile, flexible robots that can adapt to whatever is ⁓ thrown at them. If we figure that out, I think the value of that will be so incredibly vast that it's inevitable that companies are are aiming for this. And I think given what we've seen from the last few years, it's very likely that we will eventually get there. Perhaps in a longer time, perhaps in a shorter time. What are some w what is your view on that future? Is it ⁓ obviously there's some dangers, but also there's huge upsides to it. I is it is it a positive view? Is it a cautious view? Is it a pessimistic view? I'm I'm curious to learn. So, just before getting to that, I mean, so just one comment about ⁓ the value involved. I mean, ⁓ if you talk about economic value, then it's certainly the case that you have a lot of people investing in these technologies at the moment. ⁓ there are still some problems that these companies are facing with developing business models that are sort of profitable. and I mean OpenAI, for example, had said ⁓ very long ⁓ for very long that we're never gonna have advertisement. This is not gonna be like, you know, going on social media and they just like you just mostly get ads. ⁓ we're not gonna that's not gonna be our business model, but now they're flirting with that, they're experimenting with it. ⁓ and because you know, ⁓ they expected more people to get the sort of the ⁓ subscription version of this the service, but a lot of people are using the the the sort of the freely available service that you don't have to pay for. And so ⁓ how do we gonna get ⁓ how is this gonna be profitable? Because it costs a lot of money for them to be running all these chats ⁓ that people are engaging with. And so if we maybe could advertise and we could steer, you know, advertisements, that could be ⁓ that's the typical source of revenue, like whether it's radio, whether it's newspapers, whether it's social media. So that might also come ⁓ to a big time to these technologies. So maybe your robot or your chatbot is gonna start recommending toothpastes or insurance or things like that. ⁓ when you win and want want to talk about your feelings or whatever it might be. ⁓ okay, so now ⁓ okay, so I I I now sort of lost th the thread. So the question was, ⁓ what's is the future gonna be positive or negative? ⁓ yeah, i I it's hard to predict, but ⁓ is a general sense of where things are going. Is it going to be bright or or a dystopian? Right. Well I mean so I personally I feel a little bit torn here. ⁓ I want to be positive and I think that one should be, but on the other hand, ⁓ I think we really have to be careful. ⁓ I mean so I ⁓ was I have a book that just came out, ⁓ The Ethics of AI, a philosophical introduction, and I talk in the last chapter about what I call the AI wager. We're sort of betting as a society on AI to solve all our problems. At the same time ⁓ opening ourselves up to lots of risks, and we're sort of gambling with the future of our society. and this is not just something that techno as sort of pessimists are saying, that even the like the most ⁓ enthusiastic tech companies, again, open AI, ⁓ anthropic, they all keep saying that these technologies are incredibly dangerous, ⁓ and ⁓ involve all sorts of risks, but we're we're pushing ahead because we see so much opportunity. And so So yeah, it is a gamble. I mean, it's a wager. We're betting on these technologies to solve our problems. But at the same time, at the very extreme, we're gambling with the future of humanity. I mean, it I want that maybe too extreme, but people again like ⁓ I mean, I don't think we mentioned him yet, but Jeffrey Hinton, who got the Nobel Prize ⁓ for his work on AI ⁓ two years ago, ⁓ has been saying and also said ⁓ when he got his ⁓ prize there at the Nobel Banquet that yes, this there are lots of possible ⁓ ⁓ benefits, but there are also lots of enormous risks both in the short term and the long term. ⁓ and so if you add humanoid robots to the to the mix and you have a society with sort of these robotic servants that are there and like are being syncophantic and that are could be used for good and and bad means and where the tech companies are trying to figure out ways of sort of squeezing money out of us, etc etc. So you could very easily start painting with some pretty sort of like black mirror-inspired dark scenarios. And I think ⁓ I think we should be doing both. We should prepare for the worst, but we should also try to ⁓ you know motivate or bring out the best ⁓ in us and in in these technologies. But ⁓ if we're not sort of having a balanced approach, and I'm I mean another comparison I do in the in in the book is between driving ⁓ what to go back to the driving example, driving, you know, without a seat belt, not following the traffic rules, trying to get to to your destination as quickly as possible. ⁓ that's n typically called reckless driving, ⁓ and we we very much disapprove of it. But if you drive slowly with the seatbelt on, following the traffic rules, being ⁓ you know attentive to other members of traffic, you're a responsible driver, and then we think it's okay, even though driving always involves risks. And so I worry that at the moment we're pushing ahead with AI development and now also with the development of humanoid robots in a way that's more like reckless driving. than careful and responsive, ⁓ responsible driving. And so we should slow down just a little bit and then, you know, really try to develop the good sides and be mindful of the risks and so that we don't enter into a car crash, so to speak. Yeah, I agree. It's difficult though, because the incentives aren't aligned, right? Of course in the beginning when we saw the development happening, ⁓ it was probably also for marketing purposes. Maybe they some of them were genuine, but There were all these ⁓ companies like OpenAI signing these papers and and expressing their concerns about the rapid development. But what I'm always curious about, and it's a discussion for a different day. I really want to do an episode ⁓ on this with someone that is like very deep into the technological development of artificial intelligence. But what is currently not clear to me is I remember when ChatGPT came out for the first time. And it was mind blowing. It was pretty rudimentary compared to what we have today. But it was insane that you could even talk to something like ChatGPT, like a chatbot. And then GPT-4 came out. And then it was like exponentially better. And then people started extrapolating these improvements. And I was like, holy shit, if we just keep throwing more data at it and it will improve at this rate. Then we can get to the singularity of intelligence, right? It can become so intelligent, it will recursively improve itself. And then we get to a stage where AI can br build the next AI, which can build the next AI, and then we will get to this point of hyper intelligence, and then we might get wiped out because ⁓ we kill ⁓ millions of microbes every day when washing our hands and we don't think about that as being harmful. We kill flies on our way home when we drive our cars and hit all these mosquitoes against the windshield. We might just be another mosquito against the windshield of this hyperintelligent AI system, right? There was the general fear at that point in time. But I feel like, and I'm I'm can't be the only one and I I am not the only one that when we saw what GPT five could do, that it was still an improvement, but it was only an incremental improvement when we compare to GPT four. So that could be an indication of maybe LLMs aren't the way to go. Maybe these systems are just because they're always have to use data from the past and can't really produce insights about the future, that there's this natural limit to how intelligent They can ever become, and that would kind of solve the intelligence singularity problem as well. So I I don't know. Like, ⁓ are we indeed on on a speed run to developing something so crazy that it will become uncontainable? Or are we already kind of seeing the limits of the technology and do we have to maybe figure something out in order to even unlock ⁓ household robots? Like I I I don't know. ⁓ I don't know the answer to it, but I would I would love to know. And I think that's pretty important in this discussion. Do you have have you spoken to anyone or have you like for your book perhaps like ⁓ have maybe a better understanding of of where this could be going? well I've certainly spoken with a lot of people about this, ⁓ including some of the AI experts at my university on the technical side of things. I mean of course I I'm a philosopher and I I am interested in philosophical questions about these things. So of course I have chatted with my colleagues. And a lot of them indeed think that we are reaching a kind of plateau or whatever you want to call it with this particular methods. So yes, maybe there's a GPT-6, etc. That's going to be a little bit better. But at some point you need to start not only using the the current methods if you want to have big improvements. So you need to also come up with some sort of ⁓ you know hybrid systems that do what whatever these meth models are doing, plus ⁓ some other techniques that maybe ⁓ are older ones or new ones that we don't yet know about. ⁓ but we're not gonna just more data at this point is not gonna be enough. ⁓ there there was this idea that more and more data is just gonna be the the the the answer. But actually if you go back to comparing with humans again, ⁓ we ⁓ don't need so much data as these technologies do to learn things. ⁓ You know, you take a small child and you show them one or two dogs or cats, and then very quickly they can tell the difference between a dog and a cat. ⁓ they don't need to be sort of shown the whole internet to be able to learn to speak. ⁓ actually they c they they can learn to speak with, you know, even if they have parents that are not very ⁓ very verbal, like or ⁓ somewhat minimalistic in the way, ⁓ you know, the way they talk, that the child will learn to speak and to use language, ⁓ etc. So ⁓ if we could learn more about how the human brain or mind works and learns, even though with ⁓ there's this expression, the poverty of stimulus, like we have not super much input, but yet we learn all sorts of skills and incorporate even more of that into these machine ⁓ models, then you're not gonna have to do all the training and and ⁓ plus maybe if you do the training plus those things, you will get much more advanced systems. And so I do think and again I'm not a technical expert, so I should you shouldn't really ask me about this, but ⁓ We we are l reaching a sort of limit with these types of methods, ⁓ but then maybe if you can add something to it, and indeed a lot of the the recent developments with so called agentic AI, ⁓ you know, with the A you give one prompt and then the AI it can can work for you for a long time doing all sorts of things ⁓ without your having to c constantly prompt it. That is partly a matter of sort of putting different AI tools together into a sort of little orchestra or suite of sort of tools that is sort of ac acting together as a as an orchestra. Almost like a tiny organization ⁓ doing things for you. And so indeed, you know, you need different elements and the methods that are, you know, can only be pushed so far. You need new methods and new elements and new ways of combining them. Will we see a singularity with the superintelligent AI? I don't know. one thing that can happen is that our standards go down and so we start experiencing these technologies as if they're super intelligent, even though maybe that's just because we've stopped learning to engage in critical thinking and to really assess things and we just become more ⁓ you know complacent in terms of how we a assess these tools. I mean, we already have what you could call a superficial kind of artificial general intelligence, the sense that ⁓ you know, you can do almost anything with your chatbot. You know, it can write a poem for you, it can help you with your ⁓ your your your term paper. It can write a script if you are you know are performing in s in some context ⁓ whatever it might be, it can write a song, it can be your partner, so you can really do lots of different things. ⁓ is it super intelligent and like generally intelligent in some deeper sense? Well maybe not, but ⁓ for for the non-expert, for the novice, it can seem as if this tool can do everything and you trust it and you you treat it as if it's an expert on everything, but then Sometimes it will, you know, cause problems for you. There was someone who traveled tried to travel to Puerto Rico, for example, without a visa because Chat GPT had said that you don't need it. ⁓ you have I mean ⁓ yeah, I have someone in my family who works at the library. Then the people would come in to the library asking for books that don't exist because Chat GPT has recommended these books. And so And so on and so forth. So people are treating these systems as if as if they know everything, as as if they can give the right answers, maybe lowering their standards and then these technologies to seem more and more intelligent and more and more capable, but not because they are, but just because we s become lazy in the way that we ⁓ assess and and interpret them. So that's another development that you may maybe have to keep in mind. Yeah, interesting. ⁓ I could go all day about this topic, but is there anything I don't know how much time you have left, but is there anything we haven't touched on ⁓ that you think would still be interesting or important to mention while we're having this discussion? ⁓ well, I mean, we've ⁓ covered a lot, but maybe this very last point I think it's interesting. I mean, so I I I think, you know, if experts are using AI tools. They can do great things and they can do research, they can ⁓ advance knowledge and so on and so forth. The more that you're a non-expert or a novice, ⁓ and ⁓ before you've learned ⁓ you know whatever skills, critical thinking, mathematics, or whatever, ⁓ actually it might make it harder for you to develop expertise and to excel if you rely too much on these tools. And so we we have a kind of a I mean, I call it a laziness trap in some of my recent ⁓ academic presentations that. It these tools are so convenient that you know instead of b developing as a f potential partner or friend, you talk with a chatbot that's gonna be frictionless. Instead of learning to like really concentrate and th for a long period of time and to learn a skill, you ask a chatbot for it that will tell you something that you will take and then run with. And so ⁓ if you have reached expertise already and if you reach a high level, you can go maybe even further with the help of these tools. But if you're not yet there and you immediately start, I mean, maybe you're in school, you're a s a s a pupil, a student at, you know, like ⁓ elementary school, is that you already start with AI, are you necessarily gonna reach the level of expertise that's gonna help you excel ⁓ and use these tools in a productive way? Maybe not. And so ⁓ we really have to think ⁓ as a s well, as a society, as parents or whatever it might be, you know, when do we want our children to start with these technologies? Is it gonna be ⁓ productive and fruitful for them? Do we really want people to be I actually the UK government, I believe, has forbidden people under the age of eighteen from having sort of companion AI chatbots. That might be a good idea because I mean maybe in order for young people to learn how to have social relationships with other humans, they need to really interact with humans. If they're adults, maybe it should be their choice. You can have a chatbot partner, you can have a human partner. But to to to learn the skills, you may have to not use AI all the time. ⁓ before you can use AI in a kind of productive way. So that that's something I think we should think more about. Yeah, I agree. I think it also heavily depends on how you use AI. Because I as a journalist and podcaster, I have to have ideas, right? I have to have questions, but I also need to learn about stuff, which is different than learning a s a practical skill like making a table or or pottery or or something. But I have to learn ideas, I have to learn knowledge. And what I learned is how to use AI in a way so that I can actually understand and and learn the material. And it's a method called Socratic Learning, right? So instead of asking the chatbot, explain this to me, I'm just going to say, I need to understand this topic. Ask me questions about the topic so I learn it. So this way the chatbot is just merely asking me questions. I have to do the thinking. I say my thoughts out loud. And then if I'm correct, I hear it's correct. But usually you're not entirely correct. So it's like, yeah, you're mostly right, but and it kind of steers you, it shapes you. And I feel like that's a way of interacting. ⁓ with an AI system, even though it's not completely riskless, of course, there's a level of trust involved still that the knowledge you are getting from the AI system is correct, which you can't really verify because you're not an expert. So I do understand the danger in that method as well. But at least it's a more active way of working and you're still the one doing the thinking. And I mean I've never worn a E G helmet when doing it, but I I assume my my brain activity must be higher when I use it in this way than whereas I would just ask it random questions. So maybe a practical takeaway before we ⁓ wrap it up today. Yeah, I mean let me just comment on that very last point. ⁓ there is a really interesting study called ⁓ Your Brain on Chat GPT. ⁓ that ⁓ and I just actually met some of the researchers behind that when I was at a conference in Paris and they would talk about some of the ongoing research. And indeed, dep as you say, depending on how you use these tools, you're gonna be less more or less engaged and your brain is gonna be more or less active. Like if you, for example, outsource a lot of the like if you're writing a paper and then and you just ask ⁓ the chat GPT do you s you give put in a prompt and then you just take whatever chat GPT or whatever bot you're using, copy paste and put it in your paper. There's very l little reason for your brain to be very active. And in and indeed if you do a ch check on people's brain that they they aren't v thinking and and and using their brains, if you you use other methods for interacting with these chatbots, indeed you might you need to engage and and so on and so forth. So ⁓ it's gonna be very important ⁓ when universities and schools start ⁓ to I mean, as some do, some don't, ⁓ integrate these tools into education, that we really use the right methods so that people's brains are activated and that they learn. And that but that we don't go the down the route of the lazy path where actually people outsource ⁓ their thinking rather than ⁓ you know engage with these machines, thereby having to have less brain activity, having to have less is thinking and less doing. Because with most skills, whether it's learning the violin, whether it's learning t about a topic, ⁓ whether it's learning about philosophy, you have to sort of practice, you have to engage. Otherwise your brain is not gonna you know learn the skills or like in internalize the patterns that are needed. So so so yeah, I mean I I agree with you. ⁓ depending on how we use the tools, they can be great. That's the optimistic perspective. Other ways of using it, it can be ⁓ you know not so great, that's the pessimistic perspective. So we really we need to proceed again with caution, with ⁓ you know, with care and and not ⁓ your reckless did just drive into the future without a seatbelt and without any r rules or regulations. But we should ⁓ we should really use the careful drivers, the sort of the, I don't know, the role model that we try to emulate as we use these tools to now navigate our ways into the future. Yeah. So and this was so incredibly interesting. Like I said ⁓ before we started recording, I love it when philosophy crosses paths with technology. I it's it's by far my my favorite subject. I love ⁓ series like Black Mirror, ⁓ movies like s like Ex Machina, and it's always a pleasure for me to think about these things. And ⁓ yeah, to have such a great ⁓ person to talk to today was ⁓ was absolutely ⁓ absolutely ⁓ a pleasure. So ⁓ I really appreciate it. Thank you so much. Where can people go if they wanna ⁓ order your book, for example, or maybe learn more about your work? ⁓ well ⁓ I used to recommend going on Twitter, but I am not so active on X these days because ⁓ I I I am one of those people who share concerns about it. So ⁓ I usually whenever I I write something new where I publish a new paper or book or something like that. I advertise it either on LinkedIn or on Blue Sky. A little bit on X but less so. So if you are interested in my work, you can follow me on Blue Sky or or you know ⁓ I don't know connect on LinkedIn. ⁓ I Actually my new book it I think the the physical copies of the book are s th it was actually printed in the US and it's out in the US in the physical form, but they they're somewhere on a ship like travelling towards Europe and their delays. I don't know if it's stuck in the Strait of Hormuz or something like that, but but ⁓ anyway, so the physical copy is not yet out in Europe, but it will be soon. But I think, you know, Amazon and other places will have the the the the the digital version, the Kindle version and ⁓ and so I've also I you know I have some some other previous books that ⁓ are I I a lot of the top topics that we covered today is what I talk about in my research. And so ⁓ so yeah I mean I you know I just if people just Google my name they will find their way to to stuff if they find this stuff interesting. And and and thanks again for inviting me. I too, as you can tell, like talking about these topics. It was my pleasure to be on the show. And ⁓ yeah I mean it it's a fascinating ⁓ topic and we really to talk about these things as, you know, not just one on one, but as you know, as a society and we need to move into the future in a careful way. And ⁓ exciting times, exciting topics, but let's not make the world into sort of a science fiction sort of dystopic scenario. Let's let's try to go for something more positive than let let's say Black Mirror ex machina. I agree. I agree. ⁓ and yeah, there's so much ⁓ more to be discussed and developments are happening every single week. So ⁓ maybe in a year from now we should run it back and ⁓ and see where we're ⁓ where we're at ⁓ in ⁓ in a few ⁓ months from now. For now, ⁓ I'll I'll make sure to put ⁓ the the information and and links to the ⁓ to the Amazon perhaps in the in the show notes. Again, thank you so much. And ⁓ yeah, I'll ⁓ I'll hope to ⁓ to stay in touch, ⁓ Sven. Thank you. Thank you.