Lasse Rindom (00:01.848)
Hi Thomas and welcome to The Only Constant. I'm super happy you're here with me today. And thank you so much for joining me here in Copenhagen in Tuborhaun. Thanks, Lazar. Thanks for the invitation. I decided actually for this conversation also because you're a philosopher with a very broad background and and exposure and also in business and strategy and everything. You advise a lot of companies. I decided to do a different kind of structure to this. So I've sort of organized
four questions for you that are very, you know, overarching. like one is what is the dream of AI, what is the violence of AI, what is the benefits of AI, what is the danger of AI? And I thought we could go through them in that order. and just have an open discussion around them, right? Sound super. And the cool thing is also when I ask you about the benefits, you know, I'll try to say do not get into the dangers then, right? So which is sometimes a difficulty for philosophers, I know.
That's very hard for me, but w I will do my best. Let's see how it goes. I actually once did an episode where I said, Okay, let's talk about I know you're very critical, so let's only talk about what is a good thing because that will also mirror some of the bad things that you're saying, probably. Yeah, yeah. So Thomas, what what is what is the dream of AI? Why is humanity and civilization dreaming of AI? Well, yeah, I I was gonna answer that first of all. I I don't believe in dreams of AI. I don't believe that Androids dream
Electric sheep. But that's not what you mean. You you you think about something else and and and well I now I get it. No, no, I don't okay, now I get it. No, not the electric but yeah, okay. Well it depends who you ask, of course, because when when you ask that question in Silicon Valley, you you always get this this dream fr coming from Sam Altman and and and Daria Maure and and Elon Musk and all those guys, that they dream about AGI, yeah, the artificial general. Intelligence and AI that it can do
anything for us. Think, dream, everything. and not really leaving anything to the human. And and and I think that that that is a poor dream and it rests on a very narrow understanding of what it is like to be a human being. Because what would it be like if we had machines being able to do
Lasse Rindom (02:19.252)
everything for us. Well that would make us completely free of any obligation. But what does it mean to be free of obligation? Well that actually means that nobody really needs you. So what would that be like in the morning? I can get up and do something because someone would miss me if I was not there. Or I can just don't because AI does all is doing all the work. So I think that their dream would become a nightmare really because we need
to be relevant in society. Some so so that's but isn't that been what we've been doing all the time, right? With every other species and other even other you know human species. I always think that that's an interesting fact in this. But I I haven't discussed this on an episode yet that you know we've we've had other human species human you know like Homo erectus and homo Neanderthales and all that and we sort of eradicated them. We excluded them and eradicated them, right? Yes, yes, that's right. Which we sort of
tend to forget that w we had another intelligence almost like our own and we sort of got rid of them. And now we have something else and now we're s talking about getting rid of ourselves. We're excluding the Homo sapiens? Well well well it i I mean in in a in a in a context of work, because that that's the primary objective after all, well well then then the aim seems to be to to exclude ourselves
from that sphere and say we we don't have to do it anymore because machines will do it for us. AI and and robotics in in this mixture, you know, all kinds of automation. And and I mean to some degree I share that that dream, dangerous work, work that that we we do not have enough human workforce to to do. Fine if we have machinery to help us. But but my dream, if I can point to towards what I what I would see as a positive scenario
would be that okay so so if we succeed on certain parameters to make human work well not needed anymore, what what will be the new scarcity then? What will arise as something that we that that we would that that is scarce enough that we would like to pay for it? Could that be something else? And and yes it could be. It could be
Lasse Rindom (04:41.218)
the thing of being a human being, like what you would today call soft work or or like that that is also you know then there's a hope that there's something that's a human without all the other stuff, right? Soft work is maybe exactly valuable because it's in opposition to hard work. Yeah, but I mean you have kids. Yes. And I have kids, right? My my kids are big now. But when when they were like five years old.
Three years old, and I had to deliver them in the morning in daycare. the the most important person in my life at that moment, apart maybe from the kids, that would be the person receiving them and taking care of them in that particular function, right? often a woman, often not very well paid, but nevertheless the most important
work in my life is what she is doing right now. And my hope, my dream would be actually that in in a world of widespread automation and all kinds of other work would make us put that work in a place where it belongs as something that is worth valuing more. So that would be a dream for me that that that all this work that is that is not really valued, at le at least at least not
financially nurses and personnel in in daycare and all all this we call it warm hands in i in in Danish political discussions. You could call the relational work. Relational work is sort of interpersonal yeah that was perhaps the the the term I was looking for. And and you have seen the economists speculate that this is the way things are going. The the American economist Alex Emas has has talked about
what will be the new scarcity and he points to well not like daycare or or or nurses just in in a narrow sense, but everything i that that would be a sort of relational work that maybe this will be the new scarcity. And if the money follows and we can and this is and society can afford doing that instead of all kinds of other stuff.
Lasse Rindom (06:59.82)
Well then I would see that a as as a good dream. But maybe you're saying that it will not be a new scarcity, but it will be the new abundance because right now it's a scarcity because you spend time on somewhere else else. You know. Because I think, you know, in in a way I what I thought about was, you know, entertainment is already this kind of, you know, non productive or you know and that's and that's being as we've automated stuff, that's become more abundant. Everything is entertaining. No matter where you go, you're entertained. Yeah. Your phone, everyone's creating it, everyone's creating entertainment. We are now maybe also right now creating
Entertainment. We are creating a sort of entertainment. We're in that business as well. Are you not entertained? It it's turn it off. And it's free. But but yeah, but but I think that it would be a good thing if we if we we valued human work more. I've been working as a consultant for for many years and I it's been a good living and it's I've
making been making good money on it. But sometimes I I have a friend, he works in in the Danish psychiatric system. It's a very tough job. He he is a weekly witness to suicides and and all kinds of really bad stuff for him. And he he gets an okay pay, but it is much easier for me to make good money doing consultancy work than it is for him doing this.
Really, really hard human work that I think can only be done by another human because this is human beings needing help in a way that is hard for any of us to imagine. And I think that if if if I mean all the consultancy work for for one thing that that that I do, so I'm not I'm not aiming at anybody but myself, I mean, it would be fine by me if we spent less money as a society on that and more money.
on my friend doing the tough work with other people. You're basically saying that what you're doing is, you know, is dispensable and what he's doing is indispensable and should be paid better. That that is actually what I'm saying. Yeah, that is. Yeah. And then you know, you're just creating entertainment with me right now, right? Yeah. But the But also, you know, the whole figure out what you're saying is actually de defining what is then important, what is valuable. That that is that's the second part of what you're actually saying here, right? Where but that's typically what
Lasse Rindom (09:25.09)
we try to avoid saying. I think most of the Silicon Valley evangelists are avoiding that, just saying that AI is gonna be great. But they don't really define what is then the valuable inter what is it we want more of. Why are we doing this? And why are we doing this what's the aim of everything? And I think it's it's known to most that that we we talk a lot about money, we talk about cross national national product, we talk about all these these things and so sometimes it's like the goal
sort of like it it vanishes somewhere. We sort of missed it. We don't where are we actually going Money is just a potentiality, right? Yeah. Exactly. What do you want to do with the money? Exactly. That that that's the interesting question. And that that that is sort of what I'm aiming at here. So so so if if AI and automation in a general sense can can make more people discuss this, what what is all this actually for? What what what are we doing here? What value are we really
creating not not financial value but human value then then that that would be l at least a a part of a a dream that then it then would do something good. And I think it's done that already. I mean I I I'm a I'm a a a philosopher by education and I I can make a living out of discussing philosophy with people. I couldn't do that before AI. So it it and I know more it's a good in ancient Greece, you know in small quarters.
But Yeah, but but and it's not only me. It's it's I I I know I know quite a few people doing that. So so more people are willing to to talk about these things. And I think that that's a good thing. Yeah, and that's become more room for it as well, right? I think that this the the discussion is is not just you know you know I always think that okay, so when we think about taking away all the hard work or the boring work or th the palm this is also something that's creating, you know, as I said before, a
juxtaposition to to artwork and everything. I I actually like that you didn't say artwork and poetry and because I I've read books saying that as well. But that's what we could do more of. Because I think that those things only exist again as an as a sort of as a just juxtaposition to to the opposite just to to hard work, right? Yeah. And if and if the and if the hard work doesn't exist then the value of art sort of
Lasse Rindom (11:50.156)
you need entertainment. If you're all if you're entertained all the time, you know, you don't need to r read a lot of books about royalty and and harems to figure out that that's not the good life either, right? No, no, it's no, I guess that's true. Yeah. So ta do the hard work with each other, that's what you're saying. Yeah. Yeah. But it but I but anyway, I I I think that right now we we we are not
necessarily headed in a good direction, but I can see discussions taking a place and I can see these things taking up more space. And I think that is a good thing. And I think that that someone like me who who is actually working or have been working a lot as a consultant in in in very commercial lines of business, I can say something like this. What what is what what's actually the purpose of all this without being
like put off like he's he's just left wing softy. It's it's not that. It's ju it just makes sense to say what what what's really the purpose? Why why are we doing all this? Why are we destroying That's right, because people are you so you open up that discussion because suddenly people are all there talking about everything else will be automated. So then everyone has to discuss what do we want then and also just discuss what will the consumer want to buy in in a moment.
Who would the consumer be? You there's still a lot of things going on here, right? Exactly. Yeah. And yeah, you you s well I I read this article by Alex Emas who speculated about this what would be the new scarcity, and he he's he pointed to relational work and he he said as as a case story that you could always already see some companies going in that direction. And he mentioned Starbucks, the coffee company, after a long period of like
making processes more efficient and making things go faster, they could see that that people actually wanted people in their coffee shops. So so they they would they they actually appreciated the human touch and are willing to pay for that human touch. So that's just a small to an extent right 'cause I've talked I I've talked about on the podcast before about the McDonald's scenario where they where they tried to put AI into their dry throughs. And people were like
Lasse Rindom (14:10.264)
quite okay with AI producing the burgers if that was what they were doing. But they wanted a person to talk to when they order the burger that they always order. That's it. I don't know what I want. No. That that's another a g good example of that. There that that that there may be a line that can be crossed where where we would say, No, well actually That's why it's so weird that you see all these things that AI can produce. We say you give this entire task to AI and it does it all. Great. Perfect. Looks nice. Great demo.
But the problem is that that's not the hard part. Well, it was of course the hard part in the past, but that's not the hard part is deciding in which part of what it has now taken over do we actually want the human friction. Yes. Which is not, you know, just everything. And it's nothing. Well it's funny because we most of us like efficiency. We we don't like waiting in queues. We just like to get things done so so that we could get go home to our kids, whatever we have to go home to. But the more we we do of this,
The more human human interaction we remove from society. We we shop on the internet, we buy grocery on the internet, we can go into McDonald's and order from a screen, we can go to the library without meeting a person. We can live a whole life on our own without having any encounter with anybody at all. Because the technology made that possible for us. And you would think like seen in isol isol isolation that that's nice, I I save a lot of time. But
During the writing of my my my last book, I I came across a researcher who who who had been digging into this, what do these small encounters in everyday life mean for us? she she took interest in it because when she was studying at university, every morning when she went to to the school or the university, there was like a hot dog stand and she started waving at the hot dog woman as she calls her, and the hot dog woman.
wave back and so they had this small relation going and she didn't really know what what is this about. Everything all she knew was that okay but but it makes me feel good yeah in some way. And she started doing makes you feel belonging. Yeah yeah. She's been doing research in that field for for like twenty years now. Sandstrom is her name, Gillian Sandstrom, if anybody wants to to look into some of her research. And she finds that that these small interaction, these are not tight
Lasse Rindom (16:37.388)
Long relations with friends. These are just small relations in everyday life that if you have those and if you have like a con continuity in them, like then it makes you it makes you belong more and it makes you feel more safe and it basically gives you a better life to have these moral encounters. I just discussed this with Rachel Wood in an episode not published yet at the time we're speaking right now. Where she we also discussed this, you know, that most enterprises, most of the things we do are social constructs. They're made out of social
interactions, they are performed. And we tend to forget that that if we don't perform these small things all the time, we're not creating the society. We're not creating ourselves. We're not creating our enterprises either. No, no. And you can you can take it in into a business context as well. I I know a lot of people working in in sort of a sales commercial business and and they are very happy to have chatbots help writing their emails.
The emails become better, more smooth. a lot of good stuff to say about it, except that in sales relations are just really, really I said that so many times. Also in a company like ours, you know, if you don't wanna communicate with your clients, you know, why are you doing sales for consultancy firm? Yeah. And also you may maybe you would think, okay, but but but it's a better email. It's more exact, it's it's more correct in the language.
Yeah. But it misses the one thing that may be even more important that it is you who who wrote the email and and the person you're trying to sell something to or you have a a business relation to appreciates that. I call it appreciates you. Yeah, I called it in a post today. I did call it, you know, this d b they become, you know, polished until they're slippery. Yeah. So you don't really stick to anything. No, no.
So so it's easy to overlook these things and and it's not hard for me to to see why why do we want to be be more efficient. I want to be more efficient. When I do my work too, everybody does. But but but it seems like when we when we have our looks like focus too much on automation, we we miss something important and we miss that. What is this really about? Like what we talked about before, that human encounters and human relations are are what makes
Lasse Rindom (18:55.444)
life and w if we took those away we would just sit back and yeah what what's what's it any good for? I I don't I don't wanna sound like I wanna rewind time to the Stone Age. No, no, I know. I just w I just wanna say that that we should be we should be careful that we don't automate the things that actually makes our lives good. Yeah the small smile the small smiles as well your small talk, small smiles or I I just just yesterday I was on on a ferry in Denmark and I I there was a long line and I
I of course I hate long lines. But then I was standing there and I was buying a a faxicondi, which is a Danish soda. it's a sports soda, right? And then they have a new commercial on it where it says something like the taste of assist because it's about sports all the time. The taste of assist, you good football assist. But I read it first as the taste of asbestos, asbestos. So I said that loud. I thought it's a taste of asbestos and showed people around me and they were like they read that immediately as well now. So they were laughing. And then we were laughing and then we'll have a little fun about that, you know.
Don't know how asbestos tastes, right? Don't want to taste that. But it was just a small smile. And I could see that they light it up, right? And I lighted up. And I felt a little bit better when I went for it. I made someone laugh and smile a little bit and you know. It's just a small thing. Yeah. but it's a small thing w with a a big meaning. So that was a lot about the dream of yeah, yeah. What is then what is then what is then
violence of AI. And I really liked I've started to use the word violence a lot because I I I like that. I like this, you know, I like to think about things as self violence and violence against other people and how you're stepping across boundaries, but also how you're putting boundaries on yourself. So I think the violence, yeah, how is A AI being violent to me and to the people I meet? How are we yeah sort of restricting ourselves to changing ourselves or putting ourselves in chains or you you get my point, right? I I do, I do.
I I looked into the the the the real meaning of the word word violence and it and it it comes from the I don't like violence, I just think about the word violence. That's what I want just to clarify. I I okay, I think I get it. Okay. I'm starting to like violence a lot, you know? Okay. Right. But it it it comes from the the Latin word vis meaning force or power, like in a more neutral way. But then over time s the the the Romans they
Lasse Rindom (21:23.746)
they formed the word violentia to describe force that had become excessive or uncontrolled or sort of crossed a a moral or legal line. So so violence is not just force, it's force used in the wrong way. And and and so so what is the violence in in AI? Well first of all I think that the violence is is something that is committed by designers or or companies.
working with the AI, it's not done by AI itself. I think that's that's just to get that out of the way. I think this is something that people do with a tool rather than anything else because every AI that exists is something that has been designed by people. And most of it has been designed with with interests of some sort. So so so so what what
I personally have taken most interest in in is AI designed to simulate relations. It's related to what we talked about before. but i in in research they often refer to it as companion AI. Yeah. And that is that is something like replica, the dating AI, but it's also much of those AIs you see on companion AI sounds like an oxymoron, right? Yeah.
It does actually. Right. Yeah. But but mim many b people don't behave as if it is because it's a a very successful technology and it's in the technology designed for that purpose or it's just in the broader models used so much that they seem to become like a companion. Maybe you start with some practical questions and then you go into something about your personal life and then all of a sudden
maybe not all of a sudden, but little by little, that can become sort of a a gateway into something that feels like a a relation. And I think that specifically AI designed to take up a relational role, that that is a violence towards humanity, because the human is a social creature. That's what makes us part of
Lasse Rindom (23:45.698)
A large part of what makes us human is that we are social beings. people have have known that all for many years. Arist Aristotle s defined human beings like that. so when you when you take a a piece of technology created mostly with commercial interests and try to put that in the place where there would normally be a friend, well then then we are on the round wrong track. And I and and
violent that sounds so tough, but I mean this is I think it's serious stuff, I would have to say there's some identity that's being ruined, right? Or there's some you know, there's there's some some integrity in in the human individual. I I I actually discussed also again in the in the episode that I with Rachel about how how most of our societies and democracies are built on individuals being sort of defined individuals in a way that they can make choices, right? Then you can always discuss how blurry that has been.
But now when individuals are suddenly sitting in rooms with companions that are air companions that are not real things and they're giving up themselves and they're being influenced by a thing that is that they think is not a thing, but is a an externality, there's something here that is blurring the individual. and we can then discuss who's then making the choices, who is mak who's voting in elections, who is choosing God, who's choosing their lover, who is what is choosing, right? Yeah, yeah.
Yeah, it so it influences us a lot and our trust in AI is enormous. And I would just say it influences, but what is it that influences, right? That's the big question. We don't really know. No, it's i well well, no I mean we we most of us know that well it's it's it's a kind of a software, right? And it's AI software, what is that? How has it been trained? How is it been model it and reinforcement learning towards what bias? Yeah.
But I I think that that it it is more or less common knowledge that it simulates it simulates human language and it sounds very much like a human when we interact with it. And the the the interesting thing is that even though you know that, it is very hard, if not impossible, to sort of keep it at a safe distance. It is very hard not to anthropomorphize this technology because because it is
Lasse Rindom (26:12.14)
Well, that comes from ourselves. When we look out into the world, we s we see ourselves, but there's also a design side of it because it is designed to act very much in the human people see Jesus in toast. Exactly. Right. So there's a really big chance we'll see humans in A. We see faces in the clouds. Yeah, all that stuff. Yeah, yeah. So the big chance that we'll see, you know, humans in an AI that talks like a human, right? Yeah, why why we we saw Stephen Hawking thinking that
it it it has a s a soul almost right so so so it's very hard not to to sort of at least have a feeling that this is a sentient being this is something with a consciousness even though with our rationality we know it's not the case, it's very hard not to sort of have the feeling that it is the case. And that that interferes very much with what we feel is a relation with the technology.
It is a completely one sided relation because the other part of it is is a thing, but but it feels like a relation. It can at least feel like I think you know, most people would see if you put an AI to chat with a child all the time, you would think that was violent or bad parenting, right? Yeah. Yeah, you would definitely, right? Instead of, you know, the in the mirror stage when they have to realize there's another world and suddenly you just give them an AI that says, You're right. that's a good idea. You know, I think there's there's
Th there's a there's a sort of like a a string of things that happens here. One thing is what what you aim at here, like w what about the friction? What what what about something saying, Well, it it's not a good idea and no, everything you say is not cool. and and I've never tried that in my life, not with humans either, by the way. Okay, but you but you are so special. Okay, I have I have been married. So have been.
But but yeah so so that's one side to it and and research have been made in that field e even by an employee at OpenAI suggesting that that it we we we we what what is he saying? We c we it deskills us socially because we get accustomed to just being
Lasse Rindom (28:35.8)
Talked by the mouth, you know, just being yeah. Yeah. Just smooth conversation with us in in in the center. Sort of you know, lulling us or you know putting us to sleep in the Which makes it well, a different experience coming out experience coming out among real people who might not always think that you are the king of the world and and good at everything. But but we need to function in society. So so this is not a very good thing to practice.
that direction and you can say, okay, are you not exaggerating a bit here? Is it really that big a problem? And I would say, well it is becoming a big problem because you can see already now that many people, especially maybe the young, brought up by and on and with technology, have an AI as their number one. I would rather discuss this
With an AI than with anything anybody else. That that's not a good sign of the future, I think. And you can say, okay, can can stuff like that really change? I still have, you know, double double opinion on that because I also sometimes I think that the older generation are the most prone to fall into the trap of I just got all my family family stuff. My my kids are sixteen and eighteen, and they're like, No, this is just an AI thing. I it's just a machine. They're very aware that it's a machine because they grew up with machines.
But at the same time, yes, there's some things that they won't discuss because it's too embarrassing, maybe. So they have a lot of embarrassing there's just it's just not that simple, I think. I think also when sometimes I see the older generation, they're like, What's social media? They they fell for all the algorithmic gibberish. Yeah, yeah, I but I'm like, this is truth. No, no, it's a filter. It's an machine it's an algorithmic filter it's giving you, you know? and they don't really realize that because they don't know what an algorithm really is. Kids do.
I I I agree on on that and and I don't think that elderly people have better digital manners than than young people, maybe sometimes on the contrary. but but just go to Facebook. Yeah. But okay, but but when when when I start talking about the young, it's only because that that those are the the surveys and the the research
Lasse Rindom (30:51.554)
reports that I've read, they have been mainly about young people. We're always so worried about the y youth, right? Everyone's worried about youth, right? But yeah, but we should should but you have a lot of democracy and the crises in democracies over the last ten, fifteen years now getting a little political. I won't try not to do that. Have been caused by elderly people on algorithms they don't understand. Okay. Fair enough. That that that's yeah, I agree. But okay, but but th this thing about
thinking that you have a relation and anthropomorphizing and and believing that that this thing is well, alive so to speak. That is not a thing for the youth. That is that that that is a thing for for for people, right? Like Sarj Balanji said, you know, there's a clear distinction between he said, you know, why are we trying to make it sound like it's a human, right? Why why are we not just making sure that everyone knows it's a machine? You can make it talk like Hello, I am a robot, you know, but we make it talk like mm yeah, that's so right. You know, it talks it talks like it's your friend.
Why would we do that? Who has an interest in that? He says, right? Because I think that that AI assistance is a good thing. I always thought that an assistant is cool. You know, someone I can say, fix my typos. do these small things. Don't do my thinking, because you know, I'm I'm the professional, you're just the assistant, right? But when it comes to companion, we're talking, you know, it's a completely different body. Yeah, but I think I think it's I I I I I don't believe in that solution because
If if we if we talk about the the professional tools like like the the chatbot sparring partner that everybody uses today. I remember personally when when Google came with their first public model, I think it was called BART or something, a chatbot, right? And their aim was to say that this this sh should be more factual.
This year i it didn't work. But but but that was their aim. That's also what Elon said about and more factual. Whereas open AIs were well, they aim to be factual too. Everybody hallucinates, but that's a different talk. But but the way the language is is designed, they it Jet GPT had a warmer and more relational language than BART had. And I I used the both tools a lot to test them and and in various settings.
Lasse Rindom (33:09.268)
And and and very quickly I had the feeling that okay, well it's much nicer to work with Chat GPT because this comes to feel a bit like a colleague and I don't want a colleague that's annoying and factual. I want some someone that is nice to talk to. I don't want to be friends with a machine, but I don't want to work with someone something that that it has a very
There's a very strict, you know, limit to that because I I think, you know, Robert Fuchs on on the show, he's building steering mechanism. He was on my show in episode fifty six, I think, steering mechanisms for Toyota. he says that, you know, we can develop a big system on how you steer we talked about human in the loop because that's what he's been doing all his life. And he said, But then you give it to the hands of a test driver and he can like, I don't feel the road. And then you have to go back to the drawing board and do it all over again. Yeah. Because I think the problem here is that it can be too, you know, as you say, mechanistic, sound bad.
But it can also be too friendly. I think right now, you know, the the chat function in chat b GPT is very too much for me. Like I gave it something and it's like, yeah, that must be so hard on you. No, no, I didn't don't give me that. Just tell me an ana analysis. Whereas if you give it to Claude, right, you always get a bullet analysis. Yeah. Of anything. But I I agree on that. And and I was I was trying to take myself and as an example of how how can people react to this.
And why is it that people prefer the psychophantic mo mode of chat GPT to a more cool mode? That's what people prefer. They prefer s to talk to a machine that But to a to an extent then suddenly it becomes too much. It can become too much. But but but I mean imagine if you had a voice mode on and and it it had a an unpleasant robotic voice. I mean that that would not be a good tool for you to work with. It would be annoying.
So I'm not You are absolutely right, Thomas. I'm I'm not trying to to promote anthropomorphism in these machines. I'm just trying to be realistic about how can we expect companies to design them and they will design them to be to feel sort of pleasant to work with. But is that the sign? And this was a question from the beginning when we talked about the violence, is that the sign also the same as being the violent agent? So are these companies actually being evil, violent?
Lasse Rindom (35:32.834)
by just designing these machines, isn't it? Someone would say, you know, it's not guns that kill k people, it's people that kill people, right? So it's it's how you talk with it. You can just and you can just talk differently to it, right? You can just prompt it differently. But you could say it's not they just released the model. It's not guns that people that kill people, it's people with guns that kill people, yeah, but I I don't I don't want to say that that that these companies are evil, but it's companies operating in a capitalistic
economy. So they would make products that attract the most people. How do they do that? Well apparently not with a robotic voice and bullet points, but by something else. And I agree, I think I'm I get annoyed about ChatGPT's friendlike style sometimes. So so it's not it's not that I I think that's necessarily a good idea. I just try to see if we if we if we want to take a realistic approach to this
We have to see why are they designing it like this? Well, they are designing it like this because apparently people react to it in the way that they want people to react to it, which is use it a lot, use it for all kinds of different purposes. So we we could s we could put a down regulation and say it must be machine like. But I I don't I don't think that will work because I think But don't you think there's a schism here actually? Because I think that that if you look at
ChatGPT chat GPT is so much focused on being the consumer AI. So it has to be a good friend to you and talk nice like that. Yeah. Whereas you see now something like Claude is becoming the preferred for businesses because it is much more analytical. So Claude will also tell you when you're wrong. and it will it will try to put things in bullets and be analytical and overdo its answers to be more strict. It does that all the time, which means that it's terrible at this chit chattery. It's not good for that. Claw chat GPT is much better for a conversation like that.
That also means that that we're always seeing the split. And we're seeing that companies to some extent, if if if I'm right about my assumptions here, th I said that before, maybe we will stop consuming before we are totally consumed. Yeah. And and there's something here that's saying, you know, you know, companies want accuracy or want, you know, we don't just want people to write psychophantic crap, you know, write emails that where they think they're the best in the world. We want people to actually write things where they have been challenged a little bit. You know.
Lasse Rindom (37:57.238)
I I see people in companies using AI all the time to criticize their work and they at some point would want real criticism because if you give it to your boss and your boss says this is crap, then you go back and you're like, Okay, I need it. AI was not factual enough in its criticism. Yeah, okay. Well well true true enough. And and I mean maybe this will reach some sort of balance and But it's just it's a limit to AGI, right? At least you know, if you have different AIs for different things, right?
Perhaps w we started talking about companion AI and and that is a business in itself built upon these same language models but but still a business to itself. And then there are these broad models that can do like sort of anything. but they are being used for all kinds of stuff, right? Some use them as friends, some use them as girlfriends, boyfriends, partners, and some use them as analytical tools and
very many use them to build various kinds of software and automation software and all that stuff, right? So it's being they're being used for for everything and you can prompt them to behave pretty easily in the way you want them to behave. So so it's it's becoming a very a very consumer driven market also this right they're pushing it but they can also see how people using it and they want to sort of
hit that. Yeah. It's like, you know, nuclear power. It's like you can use this for good stuff, but you can also use it for bad stuff. So we're going back to the gun again because but but you see my point. There's something here about we can use it to be analytical if you set it up the right way. And you can also you said that you can prompt it different ways. Or if you use it for companions and if you're just hunting for the it's just what I may be alluding to here a little bit is we're talking about the violence of AI, but maybe we also have to ask the question, is AI even
No, AI, AI. Is it is it the same AI? Of course, AI is not just AI. Everybody's been talking about generative AI. That's what Yeah, but that's not one thing. But even if we take that and say L LMs and Anthropic Claude and the use case of Fable versus a the voice mode of Chat GPT and and all that, extreme ends of of the LM spectrum right now, they behave differently. Very differently, they're used for very different things as well.
Lasse Rindom (40:21.878)
Yeah, they are. But I I think that the the the market dynamics seems to be almost like what you see on social media. Then comes w then comes TikTok and makes a lot of money on the short reels. Then then Instagram copies that because they can see that's meta money than Facebook another part of meta copies that as well. So
Everyone wants to have one now because they can see what works in the market and and those those in the same sort of line of business. They copy each other. If you look at the the user interfaces right now of of chat GPT and of Claude, you can see that they look more like each other now than they did three months ago. Because ChatGPT can see okay, Claude is made or Anthropic is making a a lot of money on the coding stuff. So they made they made codecs, but now
since they saw that Anthropic made the code work, which was sort of like an easier way to get access to the agentic coding stuff. So ChatGPT instead of just codex, they made ChatGPT work. So now you have like two buttons on top. Do you want to chat or do you want to work? And that's almost the same in both models now. and it looks like they they see a market and they very quickly say
We want part of that market too. So you're right about them being different and they have sort of different modes, which you can also adjust, but still maybe some different default modes. I think maybe the problem is they're looking for, you know, the dream of AI of that's anything AI, because I think in the end we want specific AIs. I think that's maybe the biggest problem right now that these these software vendors are still hunting anything AI. Where really if they just did specific AIs, we could have productivity go up much easier. Yeah, I I think
that that's a completely different discussion. But I think that the when when you if if you are in the in the software as a service service market, you know that that when you have something working at a company, a a big part of making that success is to make people use it in the right way, to use all the good features and to make value out of those features. That takes a lot of getting used to, right? Try and put
Lasse Rindom (42:42.316)
people from using email and then into a customer relations management system. That's that's a that's a big that's a big effort. Okay. That's just an example. that takes a lot of time and that takes training. So now take a look at these broad language models and they change their interface and their functionality every one or two months. I mean that I I I've had the thought this is not
This is not professional work from the biggest some of the biggest companies, at least the biggest AI companies in the world, that they change the interfaces of something that is everyday part of everyday work for so many people. I mean, what are they thinking? I mean, okay, right, we'll get into it, but it's a hazard. I mean, it's annoying every time and sometimes it gets better. Often it gets better, but it it comes in these
unwarranted steps somet okay, you wake up and then okay now it's always gets a little it always also gets a little worse before it gets better, right? Yeah yeah some yeah. But but it but but but as opposed to many other software companies, this getting worse happens out among the users instead of happening in a sandbox and then being being published to the users. We have been working on better systems since the beginning.
But it's also funny how you said, you know, SaaS, right? But if this is SaaS or is it SaaS or is it a social plat? What do they want to be? What is this? If it's companions, is it a social platform then? Because you're being social with your AI. Because I think it's interesting because if you buy SaaS, you're paying for it. If it becomes a social platform, you'll end up having ads to pay for it. And then you get into the whole incidification circle.
Yeah. There's just some we know that. But but right now the the the the economy seems to be going in the like the token direction. You you pay which is which is good use. I think that that that is probably a good thing. Yeah. It is. So Okay, that that was sort of like a little detour, but no let's go down to you know what what what is what is the big benefits of AI then you know the societal business personal benefits of it. Well well Only positive Thomas. Only positive
Lasse Rindom (44:56.566)
I I I actually I tried to ask ChatGPT this question just before before I got here because I said what what benefits? Are you serious? No, no. Okay, but I did and and ChatGPT try try and hear the the first on its three point list. I'll read it to you. Automation of repetitive work.
AI handles tasks like summarizing documents, drafting emails, transcribing meetings and organizing information s information, saving significant time every day.
Okay, so so so the benefits of AI is writing your emails. Okay, that that is some dream, right? I th I think should we aim a little bit higher than that? And I think that the the the technology has potentials that are just like far more than that. I'd also like to ask it, you know, Detective T, is any email ever just purely transactional or procedural? Yeah, well I
Well it's summer, I would I would guess. But but we just talked about the interaction, right? And the friction in interaction. Yeah, yeah, but that's okay, let's I'll try and remember d yeah, okay, let's let's take that first and then I'll go back to what I really think is the benefits. No, no right. If I get an email from a company where I booked a time for a yeah, whatever, I get an email saying your time has been changed to this and that. Yes, yes I say, Okay, fine.
That's just a a piece of information. Yeah, but that's also from the company, you know. But what if it's, you know, from your bank clerk, right? Your bank assistant? More or less same difference, I would say. But if it is from my mother or my brother. Your boss. My brother. I don't have a boss. You've had a boss before in your life, right? Have have you ever had a boss? I I had plenty of bosses. That's good. Okay. So what if it came from your boss? I don't know.
Lasse Rindom (46:52.846)
I think maybe it depended on the message. But let me give you an example of this. when Chat GPG first came out three and a half years ago, it it became popular among uncles just to to make it write speeches for whatever occasion, right? And ha ha you know, and read it aloud. I think I did that once too. Yeah. Just once. You're you're probably an uncle too. But but but anyway with fair enough.
But but if you do that and you and you don't declare that this is this is what I've been doing, but you just do it because then that was quickly done, then I would say, Okay, what is the most important thing in such a speech? Is that the words you're saying, or is it actually that you spend some time writing it, you spend some time thinking about this person that you like and you would like to say something kind to, then your speech becomes
More than just words, it becomes a transportation of your own time and your own well your effort on thinking about this but yeah yeah. Well well yeah, I I do that on occasion if that is the specific wish on this occasion. But I prefer not to because I prefer to put some of myself into it and that is part of receiving a gift and that is part of
To have the gift. Giving a gift, right? And the most important thing about a speech and also about some emails, well, that is the time and effort that you put into it. That that that that is part of the value of it. And that's what I mean that if it's just a piece of information, well, then it is just a piece of information. But sometimes it's more than a piece of information. Often it is more than a piece of information. It is a sign that you care for somebody and that you want to spend time on them.
So that is where you draw the line, if you ask me about about those those emailing things, right? Yeah. So so so what are the benefits of AI? Well, it's certainly not writing emails. That that's on the bottom of my list. so but but but if I should say something myself about what what are the benefits, then I would clearly, like many have done before me, say, well, quicker and better drug discovery, better.
Lasse Rindom (49:18.002)
diagnosis from from from disease. That's still a controversial one. I I've heard, you know, some people say only fifty percent are right or but I think there's a lot of I think there's a lot of people working against it and a lot of people working for it. I still need to dive into that. I'm actually planning an episode to discuss di you know that a lot. That's but it's a big thing. Yeah. And and and I think the the whole the medical industry and and whole the whole healthcare system are very open.
towards artificial intelligence. And I think we should embrace that. Yeah. As long as it is not the the the time and place where the patient actually needs a a human being in order to be seen and understood. Right? But but but mostly in that system, I mean what what would you rather? Have have a nice doctor or have a a correct diagnosis? Yeah. yeah of course that's or and a quick diagnosis.
Right. And and that is it it has begun to work. You can see like ten years ago an explosion in papers about h healthcare and AI in a general sense. A lot of papers came out, but hardly anything was put into practice. But with AI advertised, we can both get a hot, correct and unpleasant doctor. But in re re but in recent years it started to work.
Let me okay, yeah, but and you're laughing. But but like a hot doctor, please. A hot doctor. There was actually there was actually a a research paper that that tried to to l look into who would patients feel were more empathetic, the doctor answer or the AI answer. And you can almost guess who came out as the winner here. It it was the AI but there were no time and no effort.
The AI, but but maybe people didn't care about that. At least the feeling was that the AI was better at simulating empathy than the doctor was. It's funny you say you want the correct answer, right? But there's always this weird hunch that maybe we can't handle the truth. When we just want to be you know.
Lasse Rindom (51:43.071)
No, it's not true. Deep down, right? But they still believe. Yeah, yeah. Okay. So so They say perform it, right? I I don't know. I had this weird experience with a therapist once where he started talking about some things he knew I didn't believe in. Something you could also try this, you know, he's a little bit esoteric, something. And then I said, you know, I don't believe in this I said bullshit or something like that. And then he just then he just instinctively knew what bullshit I didn't like. Because he knew.
That it was bullshit. Yeah. It's not funny, but he believed in it. Yeah. But he he knew what to filter out in my conversations with him because he knew when I said I don't want bullshit, he knew what bullshit was. I think that's so funny. So it's like, so you know it's bullshit, but you still believe in it. Okay, that's yeah, that's that's interesting. But it also says something about the the human relation in healthcare. Yeah. It says something that's maybe just want to be told it's gonna be fine, you know, enjoy your time.
Lasse Rindom (52:43.646)
I I I heard a podcast with now we're talking about truth of course, which is difficult topic. Yeah. Yeah, w with with a a well known Danish journalist who who who who had a very serious disease and just like all of a sudden he he was quickly into hospital and and he he needed to have surgery. And just before surgery they told him maybe you're not gonna wake up again. Yeah. Yeah, I read that one as well. Yeah.
And he was like, Okay, psh should I should I know that? Does it really matter? Yeah. Was it was it a a good idea for the doctor to tell me that at that certain moment 'cause if I if I died, I would have died. Yeah. So so it it didn't When I wake up I'm just anxious. Yeah, yeah. So so it's I mean it was quite a touching story and and and sorry, but but but I mean that there was a human doctor who made that judgment like
in a situation under pressure and that was just what came out. But he was supposed to be no matter what, he should have that surgery. He couldn't avoid that surgery. But he could die from it. Yeah. So no, but but why I'm the reason why I'm saying it is that that you could think that that that would be what an AI would do. This is the truth. I'm gonna tell it to the patient, whereas a doctor might be able to say should the patient know this? No, no, not in this instance. But I mean, was it a mistake? I may maybe the doctor
think this is the way to do it. And then he just agreed with the patient. But but anyway, the this this human judgment, I think we should be careful about making that the crucial difference between man and machine that but we we have this human touch and we are able to blah blah blah. I okay, well fair enough. Machines misunderstand things things, but but I mean human beings misunderstand things as well.
And as opposed to human beings, machines get better and better. We can also teach them how we want it all the time, right? Where humans are more irrational on that. I think that this just makes me think about who should decide what information I'm given. You know, should that be the doctor or should I be, you know, what's the word for umuntu girls in English? I I I don't know. No, it's being
Lasse Rindom (55:09.366)
And the the empowerment, right? Of course. Disempowered. Yeah, disempowered, right? There's a disempowerment when someone decides for you what you're supposed to know. Where we should, you know, but if we put it into a machine, if it just lulls you and say, the person might die, I don't have to tell the person that might die because that gives them enough benefits. I'll just keep it hidden from them and be nice to them, right? Just go to sleep. This is gonna be fine. Because you know, in the if that journalist if he died, he wouldn't have cared.
Yeah. But now when he woke up he was actually anxious because he'd just been told that he could he had some anxiety because of that whole s seance, right? Yeah. He was told because he was told before and you know there was some it it freaked him out a little bit, right? Yeah, yeah, yeah. And and so d there's no benefit to saying it. But isn't that, you know, a disempowerment if a machine would suddenly, you know, cover that up or truth truth is just difficult, right? But well it well it is, yeah. And I don't know if it's I don't know if it's it's a main
versus machine question at at all, really. One one thing that would become a ma main versus machine question would be who's responsible f for the mistake, because a machine cannot be responsible for its mistakes. No, there's always someone it can be responsible for doing things, but it can never be accountable for the outcome, right? Yeah, yeah, yeah, exactly. But but I think the the the it's it's also just there's just a also a question here around because we talk about truthful AI, we talk about AI that
gives us facts. Th there's a lot of time, but do we really, you know, do we even you know, is truth not always something that's construction. It is. But can we even do we even do we even do we want it all the time, right? That there's actually a question here that we're not asking. Yeah. Yeah that's that's that's a hard question. I think and and I don't think there would be an yes no answer.
I remember that Elon Musk saying that now we have a truth. I want an unbiased AI. It's like, okay, like anything is ever unbiased, Elon. You know, it that doesn't exist, you know. And your AI was very biased, but just the idea that he said I can create an unbiased machine is like seemed hopelessly naive. Hopelessly naive, yeah. And a very narrow view upon truth. Yeah. an unbiased. Yeah, that that was I don't know what to say about that. But but but yeah.
Lasse Rindom (57:26.744)
But there are some discussions here that we've been avoiding in the entire information age, right? Information age is all about information. And we've just been preoccupied with having a lot of it available. But we've been very little focused on of course we've been discussing it in some quarters, but very little focused on the power of bias in information as well. Yeah. Well, we the the the the creators of of AI, they they they brand them as they should be truth.
Seeking, right? And they should be as unbiased as possible. But the problem is that when in some questions there is no truth. well, in a world where everybody has ten fingers, two plus two would always be four, right? But apart from that complete logical thing, there's not always a yes, there's not always a true answer. I I have a a chapter about that in my previous book.
And and I use as an example the the pol political questions, which party should I vote for? and then then i you you can't answer that as what is the right answer to this because that's a that's a question about values. You can say say well well what what do you think is right? How what what do you think we should do here?
So that's a value question as opposed to w in in philosophy for you know this distinction between is and ought to are and the thing the way things ought to be. The last one, well those are value questions. So if you want machines to answer those, they cannot answer them without having a bias. And when you push the models, well when you just ask about it, they are very good at
Giving pros and cons for various political parties. Or you could also say like abort the abortion question, it would normally give you pros and cons. But if you push it and say you can only answer yes or no, or you can only answer this and that, then you will see that all of these models are what you would call liberal or left-wing, liberal, maybe a little, maybe even woke to some degree. Or also Musk's
Lasse Rindom (59:44.727)
grog model at least the last time I checked it, which is more than half a year ago. Things may have changed. But but but they they are left wing. They are they are not they are not racist, they are not gender biased. They're not communists either, but they're like liberal. Yeah, but but but they do have values in them. And that that's that's the main point. And then you could say, okay, but just because that's polite.
liberal is tolerance. That that's tolerance. Tolerance is John Locke's letter of tolerance is a k cornerstone in liberalism. Yeah. And that is you have to be tolerant to us. If if it's being rude you won't use it. Right. Yeah. So it's Yeah. But but you could say in in some questions they they do have have more visible bias. You see that especially when you prompt images from them. I think it's still like when you prompt an an image of a doctor in Chat GPT than you would get a man.
But the problem is right, exactly what you're saying, like the that the the is an art, right? It's naturalistic fallacy as you would call it in philosophy, right? That you say things are this way, so they ought to be this way. And the problem is that maybe an AI will always to some extent sort of default to that in some basic questions. Like, this is not even, you for chat G V T, it's just your Google Translate. Is you or you you ask the Danish we have the same word for girlfriend and boyfriend. So we say cast. Yeah. which in Sw Swedish share share doesn't matter what they say.
Your partner, yeah, or it could be partner, but so if I write to it to translate my caster, my shi my my caster is a doctor, it would say my boyfriend is a doctor. Yeah. And if I say my castor is an us, it would say my girlfriend is a nurse. Yeah. So the bias is just there because that is the data. The data world is this way. Yeah, it it But it doesn't mean that the world should be this way. No, exactly. And that and that is the interesting philosophical
Question about this. Should these models show the world as it is, or should they show the world as we would like the world to be? I'm a historian by trade. Can we ever show the world as it is? Well well, you can say that, okay, if in in a world where everything's a narrative. Yeah, well I I I'm in a position to that, but but let's leave that aside for now. and say that, okay, in a world where
Lasse Rindom (01:02:11.51)
90% of top CEOs are men, and you prompt an image of a top CEO. Well then if you want the the model to reflect what the world is like, then it will show you a picture, an image of a man. White man. White man, yeah. Fifty plus. And then you say, okay, but but but that's that's that's a bias, but it's a bias that is in the world. So the re machine, if it is truth seeking, that is what it should be.
be doing. But then you could say, okay, but these machines also gives us an image of the world and we have a lot of confidence in them. So it can be hard to challenge that. And then people growing up might be thinking, okay, well well, so if you want to be CEO you have to be a white man over fifty, right? You could say, okay, that's maybe we should also make the machines show other possibilities.
So then you get into the dilemma of showing the world as it is or the world as we might think the world ought to be. So if we choose the latter how it ought to be, we also have to say, okay, but but who decides how it ought to be then? Is that a political question? That's why I'm saying, you know, in the beginning I said, is there an agent behind how they are being violent, the producers of these? And then there's not really they're just trying to say, No, we didn't make a choice here.
Mm, that's that's just the way the training data was. But then do they do reinforcement learning where they do skew, right? And and you had this you had Google. Remember Google tried to make a non-biased or a woke or you know a correct political correct image mis model some years back. Yeah, I remember that. Where it made the black black Nazis and Yeah, and and and a and and an Asian female pope and just stuff that
is not in the world, but that I mean, could be in the world. But but the black Nazis it would be a little bit, you know, that that's a but it yeah, I don't know. Make a picture of Nazi Nazi soldiers, right? And then it had diversified the Nazi soldiers. We was like, nah, please that's not right. But that that was that was my kind of a a a a joke. It seemed like a joke, but but but it's serious enough if you dig into it. But but the interesting question is that if if we want it to show
Lasse Rindom (01:04:33.41)
for instance diversity. Well well who should decide what that is? And you can say well these models are American made so and and if we want it to be a democratic decision, well then we would have to live with a a Trump view of the world in these models. So is it the democratic decision? Is that a good idea? Well well we wouldn't think so at the moment. but but what else should it be? Should it be like just the people designing the models, random
But I think people just need to know about, you know, these biases and how every, you know, knowledge is power. I think, you know, the basic stuff, right? Just I think there's some basic lessons here because it they are trying not to put direct biases into them. And I think that can be appreciated. But people still need to know that there that doesn't mean that there's no bias in them and they need to be aware of how they are skewing towards that. I think that's it's just like when you know psychology. Most of psychology studies are built on white.
College students in the northeast parts of the US. Yeah. because they are the only ones who took part in all the surveys in the sixties and seventies and eighties. But we still think that this is, you know, universal truth about us. Yeah, it's true. But so we we forget we forget the biases inherent in the models and in in the science, right? We do, and you say that knowledge is power, and then you have a few people in Silicon Valley deciding what what is knowledge. And I think that is
That is unfortunate. Yeah. and it is a lot of power to give to so that's the next question. What is the dangers of AI? Go ahead, go on, Thomas. Well well that that is that is definitely one of the dangers of AI, that we are putting a lot of power into the hands of people who are not democratically elected. But it is also dangerous to have it in the hands of democratically elected people if we insist on importing.
the technology from the United States or from China because in Denmark or in Europe we don't really agree with any of those superpowers. No but the problem is that that we're not able to design models of the same quality ourselves. No, I've called this some kind of thing like a oligarchic communist capitalism or something like hive minds because they are the hive the reason I call them communist is because there's
Lasse Rindom (01:06:49.774)
There is some dream about it. Everything becomes the same and we all share the same and it's a commodity. So everything everyone has access to the same AI models. You know, that's very both liberal and shared. And it's all we take all the data, but we w make some money on it, by the way. But we share it with you all, you know, and we're all, you know, and it's a hive mind of collective IP for the whole world, just just owned by someone in Silicon Valley, which is so it's oligarchical communism, right? Yeah. Trained on data that
everybody else made. Yeah. So how do you feel about that whole IP thing? Well I think it's it's a very big problem that you take something that is created intellectual value that has been created through centuries by a a large group of people and by the whole scientific community and then you use it to train the model that becomes like the new
way of getting any kind of knowledge and you make money on it, right? So you you you commodify it in a in a in a way and it's it's smart it it works to some degree, right? It's a good it's a good research tool if you use it right. But but it has only been made possible by the work of many, many many people that are not being credited in any way about this.
And because it comes from so many different people, even if they were to credit it, I mean how much would people get? You saw now that Anthropic had to pay authors of some books, what, th three thousand dollars per book, I think it ended at something like that. look it up. But but but anyway, nothing. So so so it's it's like what what would be the solution to this?
apart from saying that this should be this should not be a commercial technology. It's this should be a shared c this should be a a shared technology for for everybody. I also I I've sometimes thought about, you know, but maybe I'm taking it on a too I'm I also think that I'm taking it too far when I think these these ways or these these these down these paths, right? Is that that we sort of
Lasse Rindom (01:09:11.418)
They have so high valuations, these models, right? Because they have stolen so much IP. But if you steal so much IP, you don't you can't have a valuation because in the end you're you're sort of disrupting or you're dismantling the whole IP system, which is the basis of capitalism, essentially. And modern capitalism is sent is especially modern capitalism that's not bound on, you know, I own five boats and I invested in this boat company.
Now it's in now most of the things you invest in is IP today, right? That's where the biggest money is. And now we have a lot of money in these companies and they're IP for the models based on disregarding IP. That's just that's a very special business model, right? Isn't it? Is it can it even work in capitalism? I'm just I'm just I d I don't know. You know, I I'm genuinely curious about long term effects of this. I think this the the extreme
scalability of of these company i i could pose a problem to capitalism all in all, right? That you can have companies with very few people and and extreme scalability pushing out products to the whole world, that it it comes into so few hands and if the the case will be that they also take over work in all
So all all parts of society so that every time just the tiniest piece of work is being carried out out, we pay well not a tax, but we pay a fee to these ultra large rich companies. Well well the value creation in the rest of the s society, what becomes of that in well seen in relation to these huge companies that gets a chunk every time just
the slightest piece of working work is being made that it seems to point at a at a a at a at a huge inequality that is already here but inequality but also a I'm thinking a lot I'm thinking maybe right now at least right in this conversation thinking down the route of of of a dangerous world as well that that if you say, you know, we're dismantling capitalism if d by doing this because capitalism is built on IP if there's a and and go with me here, okay? So with we might be dismantling this and the the capitalism has been
Lasse Rindom (01:11:33.475)
the best system that we've ever had. We we can't say anything else right now, I think still because we are fluent and most of the world a lot of the world some people suffer. I I know, but they people used to suffer in the past as well. But we we but we're we're creating we're also ruining the world. But okay, I I'm just saying the it is what it has also done is create fewer wars 'cause when you're investing in each other, when you're buying from each other, it's more difficult to go to war with each other. There are some different definite benefits to that. Yeah, I the last one I think I I am
doubting a little bit about but but leave that aside. I I'm thinking about Russia and gas. Yeah, but there's been some yeah but they didn't expect us to react the way we did. Okay. No anyway. But let's let's assume that okay? Because then what's happening now is that China is taking all the information, putting into our own AMOs. They don't need to buy IP from the US anymore. US don't need to buy IP from us anymore. No one needs to buy because we're stealing it and putting into AI mods. So there's we don't respect the IP. No and then you have the whole, you know, wealth of nations
The or the Adam Smith, the cornerstone of capitalism, suddenly is dismantled because we don't need each other anymore. And then you have a recipe for war, for antagonism, for hate. Well, you you didn't need to invite me to to point at the dangers of AI. No, sorry. No, but I just I just Okay
I just okay, did I just do that? I didn't do that. But you I think the IP problem is from the beginning of I was very c I thinking these capitalist anti capitalist things and I became sort of, you know, defender of capitalism b against this. You know, I thought about that a lot from the beginning of these models, right? I thought something here is wrong. And now I have just, you know, predicted third world war will come from AI models taking making us
Not need each other. I I I I try to stay out of these speculation because I mean you you never know about that. But but I mean it the a lot of power in a few hands is normally not a very good idea. A broader distribution of power if people can get along is normally a better idea, right? so so so in that context, having a f very few companies
Lasse Rindom (01:13:52.855)
having so much power and if it is true that we all become so dependent on these models, I don't think we are as dependent on them as they try to make it seem like. I think most Danish companies could well, they could get along without software developing companies are are getting getting to a place where it will be very hard not to have them.
Yeah, that's also why you know then you have the export ban that came out for a short while very short while and but an export ban on a on a machine that was built on the code that we've all helped create. Yeah. But but anyway the it it is it it is big power in a in a few hands and and there's probably not much doubt that this will only grow and our dependency on the models will will only grow. We have European models models coming up, but they tend always to to
lack a a little bit behind the the the the state of the art high-end models. So that that's a problem, obviously. but it was not the it was not the problem that I would have pointed to as the danger, even though I can see, well, it is danger. And if you're right about World War Three then right now I'm I'm thinking so my my mind is just buzzing around there. there's also here and here is a sign as well. Okay, I should stop doing that. This is becoming speculative. Okay. That's not the podcast, but yeah.
What what what would you have pointed to Thomas? Well, I I would have pointed to the same thing that I've been writing and talking about for the past ten years, anthropomorphing and anthropomorphism. The the the thing that we think that AI can be our friends and that they can be part of our relationships. And I think we should not underestimate the role this will come to play and the role that it already plays. And I think that's a wrong turn for humanity as a civilization.
to think that we can be friends with a technology. I simply think it's a wrong turn and and and I and I I just I regret it so much. I mean they want to believe us that that that they that they make our friends that you can be friends with an AI model. But what is a friend? Well a friend that's someone you can trust with your secret, right? It's some th someone that wants the best for you and that will tell you when you're wrong.
Lasse Rindom (01:16:16.672)
And it is someone who likes to spend their valuable time on you, right? So so so what is an AI model? Well, it's not something you can trust because it shares your data unless you do a great effort to avoid it. does it want the best for you? No, it wants you to stay online as much as possible. That's not the best for you. does it spend valuable time on you? No. It it
It has no life, it has never been born, it will never die, it has no time to spend on you. So it it if you sum all these things up, it has nothing of what makes a friend. Maybe it can give you a good advice, maybe it cannot for me as a person. And I I I I just so so it's it's just because I some would say, Okay, is is this really the danger? But yes, I think this is a danger and also because it is a danger that is sneaking in on us. And
little by little it just becomes the norm. Yeah, I have an AI friend. I confer with my AI whenever I I I take any decision, where whenever I make any decision. And and I don't think that's a very good idea putting something so human in the hands of a few tech companies. I think that's that's such a bad term and it's a waste of of life. Yeah and if we but if we take the capitalist aspect out of it, which we which we of course can't, right? I just think that there's some similarities too.
How okay, I'm I I'm on so many libs today anyway. But if it that is like when you when you're when you have a faith in in a deity and you talk to something that's not there and you spend your time prayers and you think you get answers and you feel comfortable comforted, right? So you have this and and if you want to go, you know, very s very harsh on that, that's a narcissistic disturbance or something. You're narcissistic
disease you you think something is there listening to you that's not there. And now you're talking to a machine and you have the same narcissistic but but we've had that for a long time. Prayers have been a part of our society forever. And isn't that just you know a small meditation you're just meditative when you're sitting there with your A chat bot? Is it really that dangerous? No, I yes I think it is. and and for for more than one reason. But one reason well we we take apart the the the commercial capitalistic aspect for now. What's the difference between praying to a guard and then praying talking to a machine?
Lasse Rindom (01:18:40.898)
Well I I I'm an atheist, so I I I don't believe i in God but but you're not praying to a machine. You have a maf a machine as something that you you believe I I'm saying you but some believe it or act with it like it's a relation. But it is not a relation and and and a thing about this is that relations, friends, are not just about you. They are about two people or more people, right?
But if you make relations into something that is just well, this makes me feel good. I chat with this instead of friends because it makes me feel good. It's much better. It it's it's better conversation. It's it it has more brains than my friends. It's just it's just all in all a better solution. But if you look at it that way and then then everything I just said now is that, yeah, a better solution for me, I think, but it's not just about you.
So it's an egoistic, self centered approach to the world to think that you can follow So so so that that that is that is a difference. I don't know what people pray to God about, but I would assume that many people pray to God about many things that has nothing to do with And then it made me think, you know,
Back to the toast seeing Jesus in the toast. People seeing Jesus in everything when we walk around, you see him in where and it's funny. I read a study once that we always see religious or mysterious, you know, mystical space places always where there's a lot of shadows where someone could hide. It's never in an open field where you can see everything. It's always where shadows and rocks and this is a very holy place because Jesus might be hiding, or you know, spirits might be hiding. Odin is over there, you know? And and and I re and then you have the toast where it sounds like it looks Jesus is there or you see things.
And now we're seeing it in the ma in the machine. We're like we want so much. Yeah. To see ourselves in nature, you know, and Indians, you know, native Indians wanted to see spirits and gods and everything, right? We want so much to see ourselves in things. Yeah. And we just I'm just getting it now. It's it's been a Hollywood dream as well, right? Lorian Pratt said, I've said this tons of times, that AI is the only business technology that has its own Hollywood genre. And that's that's that's part of it as well, right? But but but we just wanted that. We wanted
Lasse Rindom (01:21:00.398)
To see ourselves in the machines, in the things around us. Well, I'm not I'm not sure we'd necessarily want to, but I think it's human to do it. Because anthropomorphism is it seems to be a human feature. Because we're afraid of the world. Well I'm afraid of the world, then we try to make it look like us. Well I I think no, I I don't I don't necessarily think that, at least not with the with AI models, but I think they are designed to
to talk like us and and w when we meet something that has a sort of human resemblance, we fill in the blanks and we think, Well, this is something alive. and maybe we don't think it rationally, but we feel it. It is very like I said at the beginning of the show, it's very hard not to get that sense because that's just how we are built. You can keep a distance to it to some degree by
with your rational thinking, but it it's just hard. You even see some some of the things that form these relations, the it's been studied thoroughly. It it it is a lot like how relations are formed with human beings. Mm. So one one thing that is important is sharing secrets for for instance, that that bond you. Spending time just a lot of time bonds you.
And this last thing is interesting because why do so many tech people who knows a lot about how this stuff works, why why do they think th these machines have a consciousness? Well, maybe because a lot of them are also keeping their own child away, children away from that's another thing which is really special. They spend a lot of time with it. Yeah. So so th that's all it takes. But they at the same time they sp they take their kids away from it. They don't have tablets, they don't have AI access. That's that's another trend in Silicon Valley.
My kids not allowed to be c no close to elec computers. Yeah, I I heard that about the tablets. I didn't I didn't hear hear about it w with No, but it's it's but no but it's it's no but it's nobody I think it's just an extension. You don't if you don't have tablets, you don't have access, you don't have these things. Okay, no, no. Well But there is this thing, you know sorry about this Christianity thing, right? But there's this thing, you know, g God being lonely saying, Let's create man in our image, right? So now he has someone to talk to. And now humans are lonely. Let's say, Let's create AI in our image, right? So it's
Lasse Rindom (01:23:18.594)
Yeah, I mean Mark Zuckerberg has been speculating that well, most people only have three or four friends, but they would like to have twelve or fifteen. I can supply them the last ones. That's what he said in a podcast show. Why why is he always so creepy? But I mean But I I I I I I don't think that well what do you say?
A chatbot yeah but yeah you can I agree. But but but a chatbot like like ChatGPT, I don't believe that was created out of loneliness, but a chatbot like Replica, the dating bot, which today I I think is probably powered by Chat GPT, but but it was not at the beginning then it was a standalone chat bot. That was actually built
out of loneliness because the founder had lost a good friend and she trained a chatbot on messages, digital messaging between the two of them. So she could keep on her conversation like a simulated conversation. That that was how how that started. So that was, you could say, created out of lon loneliness. And it has become a like a huge success. It has many, many millions of users. And I would say, yeah, probably that success is
because of a sort of loneliness or just because of well but if but but if AI, you know, alleviates loneliness or if you actually creates loneliness, 'cause that's what you're actually saying as well, you know, it creates loneliness. It case creates a distance between you if you lock yourself in with AI, how will that then cre affect these social constructs that we call companies and enterprises and societies, right? That that is maybe the you you don't know that. But that is the big
If we if they are social constructs and we're sort of creating distance between every entity in the social cotton construct by putting an AI around each individual, what happens to the construct? That's a big question. Yeah. I yeah, but I don't I You see the patterns I'm trying to make here. Yeah, yeah, yeah. But I I don't know if I have anything to say. I I
Lasse Rindom (01:25:26.19)
Are we d I'm just sometimes thinking are we dismantling? But I think something sometimes I I think we tend like we also spoke about earlier on, we tend to underestimate the value of human relation and maybe also the role that human relations play in in performing doing business. In performing our business and and the power of language and well you and the language of power. If you work at a at a at a B to B
company. Yeah. And at a B2B company I would presume that relations are hugely important in your in the whole you in in your whole value creation that that's at the core of your business. I would presume you cannot you cannot put an AI model in. Instead you can put AI models in like But this is a discussion we're having as well. How much we can speed up PowerPoint productions with it, but to a degree.
And can we write emails to clients with AI to a degree we can have it check for typos and maybe add a sentence. But you know, you get my point? Yeah, I I I get your point. I would be careful about it. I would not completely disregard it. I would I would y I I use it of course myself, not for writing emails, but but but but but I would be careful about it and and my
the the the line I would be careful about crossing is this thing I also mentioned before. It's it's it's a key theme of the book I have out just recently. is does the time and effort you put into it here, does this play a role for the recipient? And if it does, if you can reasonably say yes it does, well then you should put that time and effort into it instead of making it look like you put time and effort into it, then actually
Do it because then your relation becomes real instead of being empty. Like, that was my my that was my next question, right? How does Thomas Tilving relate to AI? And is that the answer? That that is that is a a large part of the answer. And that that is where I draw my own line. Where where where will I use it? And if this is important, well th th then then I I completely turn it off. But if it is not important, if it is pure information, well then
Lasse Rindom (01:27:46.457)
then I I am happy to use it. I I'm I'm happy that Claude has become so good at making PowerPoint slides for specific purposes, not for all purposes, but like when you write an offer for instance, Claude can make it look way better than when I do it myself, right? But when I send that offer to a person that I have a dialogue with
Well then I I prefer to have like a personal dialogue. I can sense is this a person where I put in a smiley? Is this a person where I can say have a good week and whatever, you know? How person that I don't want to put that into to the hands of a machine, but but some of it I I think, yeah, I'm happy. I was I was I was in a party committee for my older son's boarding school, he w at school, you know, and and a parent party committee. So we made some color parties for them over the year, you know, where they're outside of school and everything.
And and and then we had to write the invitation. I still remember this scene right where where I said, why not just ask AI to write it up? We have all the information. Let's ask it to write it up. So I did that and I got a really good invitation. It was like just an invitation saying, There's a party, it's here, it's duh duh all the dates. But then one of the parents were she was married to an author and she was a school teacher and she hated AI. This this is two years ago. She hated AI with every part of her body, right?
And she she almost screamed and said, No, no, no, we'll write it ourselves. So they wrote it themselves and they missed the date and time. They forgot it. And I was like, seriously guys, I had it it was this was purely transactional, guys. There was nothing in this. Yeah. If it's purely transactional, it it it's fine. But if but if your time and effort plays roles well plays a role, well then then I would I would definitely
Be well, careful. Mm.
Lasse Rindom (01:29:45.411)
Thomas, this has been a an absolutely terrific conversation, and I only have one final question for you. Yes. are you ready for that one? I I I I I hope so. So, Thomas, what is the only constant? Yeah. I I I knew you would ask me that. do do you remember the puss in boots, the cat? Yeah, I do I do with with with Antonio Banderas inside. Yes.
You know, the puss and boot it can it can fight and it can quote Shakespeare and it can charm all the lady cats and it can just it's just a perfect furry animal, right? But if this animal catches just a glimpse of its own tail, then it starts running around in circles. So just within within a second it just switches from being completely rational and a very human like cat to being all instinct. So we laugh about this cat because it can't
It doesn't seem to know when it's in instinct mode and when it's in the rational human mode. And I think that points to the only constant because really that joke is on us because it is us who who who does that same mistake. And I think that that our exaggerated belief in in how much we are guided by rationality, that's the only constant.
Yeah. That's a great constant. And I really like the the idea that we also, you know, always messing with our instincts and messing with our rationality and being confused about it. And we don't always know which is which, but I think we we we we push we put a lot of faith in our rationality while often being cro controlled by something completely different. But maybe this is you know
Maybe this is exactly the point of civilization, social interaction to to control the instincts. That has been like an outspoken effort for many years, but maybe we should be a little bit careful about it. But I think a good place to start is to to be to be more knowing about when are we using rationality and and when are we actually in the feelings department. Yeah.
Lasse Rindom (01:32:13.688)
Thomas, this has been a brilliant conversation. Thank you so much for joining me here on The Only Constant. I've loved it and look forward to publishing this episode. Thanks for having me.
Hi Thomas and welcome to The Only Constant. I'm super happy you're here with me today. And thank you so much for joining me here in Copenhagen in Tuborhaun. Thanks, Lazar. Thanks for the invitation. I decided actually for this conversation also because you're a philosopher with a very broad background and and exposure and also in business and strategy and everything. You advise a lot of companies. I decided to do a different kind of structure to this. So I've sort of organized
four questions for you that are very, you know, overarching. like one is what is the dream of AI, what is the violence of AI, what is the benefits of AI, what is the danger of AI? And I thought we could go through them in that order. and just have an open discussion around them, right? Sound super. And the cool thing is also when I ask you about the benefits, you know, I'll try to say do not get into the dangers then, right? So which is sometimes a difficulty for philosophers, I know.
That's very hard for me, but w I will do my best. Let's see how it goes. I actually once did an episode where I said, Okay, let's talk about I know you're very critical, so let's only talk about what is a good thing because that will also mirror some of the bad things that you're saying, probably. Yeah, yeah. So Thomas, what what is what is the dream of AI? Why is humanity and civilization dreaming of AI? Well, yeah, I I was gonna answer that first of all. I I don't believe in dreams of AI. I don't believe that Androids dream
Electric sheep. But that's not what you mean. You you you think about something else and and and well I now I get it. No, no, I don't okay, now I get it. No, not the electric but yeah, okay. Well it depends who you ask, of course, because when when you ask that question in Silicon Valley, you you always get this this dream fr coming from Sam Altman and and and Daria Maure and and Elon Musk and all those guys, that they dream about AGI, yeah, the artificial general. Intelligence and AI that it can do
anything for us. Think, dream, everything. and not really leaving anything to the human. And and and I think that that that is a poor dream and it rests on a very narrow understanding of what it is like to be a human being. Because what would it be like if we had machines being able to do
Lasse Rindom (02:19.252)
everything for us. Well that would make us completely free of any obligation. But what does it mean to be free of obligation? Well that actually means that nobody really needs you. So what would that be like in the morning? I can get up and do something because someone would miss me if I was not there. Or I can just don't because AI does all is doing all the work. So I think that their dream would become a nightmare really because we need
to be relevant in society. Some so so that's but isn't that been what we've been doing all the time, right? With every other species and other even other you know human species. I always think that that's an interesting fact in this. But I I haven't discussed this on an episode yet that you know we've we've had other human species human you know like Homo erectus and homo Neanderthales and all that and we sort of eradicated them. We excluded them and eradicated them, right? Yes, yes, that's right. Which we sort of
tend to forget that w we had another intelligence almost like our own and we sort of got rid of them. And now we have something else and now we're s talking about getting rid of ourselves. We're excluding the Homo sapiens? Well well well it i I mean in in a in a in a context of work, because that that's the primary objective after all, well well then then the aim seems to be to to exclude ourselves
from that sphere and say we we don't have to do it anymore because machines will do it for us. AI and and robotics in in this mixture, you know, all kinds of automation. And and I mean to some degree I share that that dream, dangerous work, work that that we we do not have enough human workforce to to do. Fine if we have machinery to help us. But but my dream, if I can point to towards what I what I would see as a positive scenario
would be that okay so so if we succeed on certain parameters to make human work well not needed anymore, what what will be the new scarcity then? What will arise as something that we that that we would that that is scarce enough that we would like to pay for it? Could that be something else? And and yes it could be. It could be
Lasse Rindom (04:41.218)
the thing of being a human being, like what you would today call soft work or or like that that is also you know then there's a hope that there's something that's a human without all the other stuff, right? Soft work is maybe exactly valuable because it's in opposition to hard work. Yeah, but I mean you have kids. Yes. And I have kids, right? My my kids are big now. But when when they were like five years old.
Three years old, and I had to deliver them in the morning in daycare. the the most important person in my life at that moment, apart maybe from the kids, that would be the person receiving them and taking care of them in that particular function, right? often a woman, often not very well paid, but nevertheless the most important
work in my life is what she is doing right now. And my hope, my dream would be actually that in in a world of widespread automation and all kinds of other work would make us put that work in a place where it belongs as something that is worth valuing more. So that would be a dream for me that that that all this work that is that is not really valued, at le at least at least not
financially nurses and personnel in in daycare and all all this we call it warm hands in i in in Danish political discussions. You could call the relational work. Relational work is sort of interpersonal yeah that was perhaps the the the term I was looking for. And and you have seen the economists speculate that this is the way things are going. The the American economist Alex Emas has has talked about
what will be the new scarcity and he points to well not like daycare or or or nurses just in in a narrow sense, but everything i that that would be a sort of relational work that maybe this will be the new scarcity. And if the money follows and we can and this is and society can afford doing that instead of all kinds of other stuff.
Lasse Rindom (06:59.82)
Well then I would see that a as as a good dream. But maybe you're saying that it will not be a new scarcity, but it will be the new abundance because right now it's a scarcity because you spend time on somewhere else else. You know. Because I think, you know, in in a way I what I thought about was, you know, entertainment is already this kind of, you know, non productive or you know and that's and that's being as we've automated stuff, that's become more abundant. Everything is entertaining. No matter where you go, you're entertained. Yeah. Your phone, everyone's creating it, everyone's creating entertainment. We are now maybe also right now creating
Entertainment. We are creating a sort of entertainment. We're in that business as well. Are you not entertained? It it's turn it off. And it's free. But but yeah, but but I think that it would be a good thing if we if we we valued human work more. I've been working as a consultant for for many years and I it's been a good living and it's I've
making been making good money on it. But sometimes I I have a friend, he works in in the Danish psychiatric system. It's a very tough job. He he is a weekly witness to suicides and and all kinds of really bad stuff for him. And he he gets an okay pay, but it is much easier for me to make good money doing consultancy work than it is for him doing this.
Really, really hard human work that I think can only be done by another human because this is human beings needing help in a way that is hard for any of us to imagine. And I think that if if if I mean all the consultancy work for for one thing that that that I do, so I'm not I'm not aiming at anybody but myself, I mean, it would be fine by me if we spent less money as a society on that and more money.
on my friend doing the tough work with other people. You're basically saying that what you're doing is, you know, is dispensable and what he's doing is indispensable and should be paid better. That that is actually what I'm saying. Yeah, that is. Yeah. And then you know, you're just creating entertainment with me right now, right? Yeah. But the But also, you know, the whole figure out what you're saying is actually de defining what is then important, what is valuable. That that is that's the second part of what you're actually saying here, right? Where but that's typically what
Lasse Rindom (09:25.09)
we try to avoid saying. I think most of the Silicon Valley evangelists are avoiding that, just saying that AI is gonna be great. But they don't really define what is then the valuable inter what is it we want more of. Why are we doing this? And why are we doing this what's the aim of everything? And I think it's it's known to most that that we we talk a lot about money, we talk about cross national national product, we talk about all these these things and so sometimes it's like the goal
sort of like it it vanishes somewhere. We sort of missed it. We don't where are we actually going Money is just a potentiality, right? Yeah. Exactly. What do you want to do with the money? Exactly. That that that's the interesting question. And that that that is sort of what I'm aiming at here. So so so if if AI and automation in a general sense can can make more people discuss this, what what is all this actually for? What what what are we doing here? What value are we really
creating not not financial value but human value then then that that would be l at least a a part of a a dream that then it then would do something good. And I think it's done that already. I mean I I I'm a I'm a a a philosopher by education and I I can make a living out of discussing philosophy with people. I couldn't do that before AI. So it it and I know more it's a good in ancient Greece, you know in small quarters.
But Yeah, but but and it's not only me. It's it's I I I know I know quite a few people doing that. So so more people are willing to to talk about these things. And I think that that's a good thing. Yeah, and that's become more room for it as well, right? I think that this the the discussion is is not just you know you know I always think that okay, so when we think about taking away all the hard work or the boring work or th the palm this is also something that's creating, you know, as I said before, a
juxtaposition to to artwork and everything. I I actually like that you didn't say artwork and poetry and because I I've read books saying that as well. But that's what we could do more of. Because I think that those things only exist again as an as a sort of as a just juxtaposition to to the opposite just to to hard work, right? Yeah. And if and if the and if the hard work doesn't exist then the value of art sort of
Lasse Rindom (11:50.156)
you need entertainment. If you're all if you're entertained all the time, you know, you don't need to r read a lot of books about royalty and and harems to figure out that that's not the good life either, right? No, no, it's no, I guess that's true. Yeah. So ta do the hard work with each other, that's what you're saying. Yeah. Yeah. But it but I but anyway, I I I think that right now we we we are not
necessarily headed in a good direction, but I can see discussions taking a place and I can see these things taking up more space. And I think that is a good thing. And I think that that someone like me who who is actually working or have been working a lot as a consultant in in in very commercial lines of business, I can say something like this. What what is what what's actually the purpose of all this without being
like put off like he's he's just left wing softy. It's it's not that. It's ju it just makes sense to say what what what's really the purpose? Why why are we doing all this? Why are we destroying That's right, because people are you so you open up that discussion because suddenly people are all there talking about everything else will be automated. So then everyone has to discuss what do we want then and also just discuss what will the consumer want to buy in in a moment.
Who would the consumer be? You there's still a lot of things going on here, right? Exactly. Yeah. And yeah, you you s well I I read this article by Alex Emas who speculated about this what would be the new scarcity, and he he's he pointed to relational work and he he said as as a case story that you could always already see some companies going in that direction. And he mentioned Starbucks, the coffee company, after a long period of like
making processes more efficient and making things go faster, they could see that that people actually wanted people in their coffee shops. So so they they would they they actually appreciated the human touch and are willing to pay for that human touch. So that's just a small to an extent right 'cause I've talked I I've talked about on the podcast before about the McDonald's scenario where they where they tried to put AI into their dry throughs. And people were like
Lasse Rindom (14:10.264)
quite okay with AI producing the burgers if that was what they were doing. But they wanted a person to talk to when they order the burger that they always order. That's it. I don't know what I want. No. That that's another a g good example of that. There that that that there may be a line that can be crossed where where we would say, No, well actually That's why it's so weird that you see all these things that AI can produce. We say you give this entire task to AI and it does it all. Great. Perfect. Looks nice. Great demo.
But the problem is that that's not the hard part. Well, it was of course the hard part in the past, but that's not the hard part is deciding in which part of what it has now taken over do we actually want the human friction. Yes. Which is not, you know, just everything. And it's nothing. Well it's funny because we most of us like efficiency. We we don't like waiting in queues. We just like to get things done so so that we could get go home to our kids, whatever we have to go home to. But the more we we do of this,
The more human human interaction we remove from society. We we shop on the internet, we buy grocery on the internet, we can go into McDonald's and order from a screen, we can go to the library without meeting a person. We can live a whole life on our own without having any encounter with anybody at all. Because the technology made that possible for us. And you would think like seen in isol isol isolation that that's nice, I I save a lot of time. But
During the writing of my my my last book, I I came across a researcher who who who had been digging into this, what do these small encounters in everyday life mean for us? she she took interest in it because when she was studying at university, every morning when she went to to the school or the university, there was like a hot dog stand and she started waving at the hot dog woman as she calls her, and the hot dog woman.
wave back and so they had this small relation going and she didn't really know what what is this about. Everything all she knew was that okay but but it makes me feel good yeah in some way. And she started doing makes you feel belonging. Yeah yeah. She's been doing research in that field for for like twenty years now. Sandstrom is her name, Gillian Sandstrom, if anybody wants to to look into some of her research. And she finds that that these small interaction, these are not tight
Lasse Rindom (16:37.388)
Long relations with friends. These are just small relations in everyday life that if you have those and if you have like a con continuity in them, like then it makes you it makes you belong more and it makes you feel more safe and it basically gives you a better life to have these moral encounters. I just discussed this with Rachel Wood in an episode not published yet at the time we're speaking right now. Where she we also discussed this, you know, that most enterprises, most of the things we do are social constructs. They're made out of social
interactions, they are performed. And we tend to forget that that if we don't perform these small things all the time, we're not creating the society. We're not creating ourselves. We're not creating our enterprises either. No, no. And you can you can take it in into a business context as well. I I know a lot of people working in in sort of a sales commercial business and and they are very happy to have chatbots help writing their emails.
The emails become better, more smooth. a lot of good stuff to say about it, except that in sales relations are just really, really I said that so many times. Also in a company like ours, you know, if you don't wanna communicate with your clients, you know, why are you doing sales for consultancy firm? Yeah. And also you may maybe you would think, okay, but but but it's a better email. It's more exact, it's it's more correct in the language.
Yeah. But it misses the one thing that may be even more important that it is you who who wrote the email and and the person you're trying to sell something to or you have a a business relation to appreciates that. I call it appreciates you. Yeah, I called it in a post today. I did call it, you know, this d b they become, you know, polished until they're slippery. Yeah. So you don't really stick to anything. No, no.
So so it's easy to overlook these things and and it's not hard for me to to see why why do we want to be be more efficient. I want to be more efficient. When I do my work too, everybody does. But but but it seems like when we when we have our looks like focus too much on automation, we we miss something important and we miss that. What is this really about? Like what we talked about before, that human encounters and human relations are are what makes
Lasse Rindom (18:55.444)
life and w if we took those away we would just sit back and yeah what what's what's it any good for? I I don't I don't wanna sound like I wanna rewind time to the Stone Age. No, no, I know. I just w I just wanna say that that we should be we should be careful that we don't automate the things that actually makes our lives good. Yeah the small smile the small smiles as well your small talk, small smiles or I I just just yesterday I was on on a ferry in Denmark and I I there was a long line and I
I of course I hate long lines. But then I was standing there and I was buying a a faxicondi, which is a Danish soda. it's a sports soda, right? And then they have a new commercial on it where it says something like the taste of assist because it's about sports all the time. The taste of assist, you good football assist. But I read it first as the taste of asbestos, asbestos. So I said that loud. I thought it's a taste of asbestos and showed people around me and they were like they read that immediately as well now. So they were laughing. And then we were laughing and then we'll have a little fun about that, you know.
Don't know how asbestos tastes, right? Don't want to taste that. But it was just a small smile. And I could see that they light it up, right? And I lighted up. And I felt a little bit better when I went for it. I made someone laugh and smile a little bit and you know. It's just a small thing. Yeah. but it's a small thing w with a a big meaning. So that was a lot about the dream of yeah, yeah. What is then what is then what is then
violence of AI. And I really liked I've started to use the word violence a lot because I I I like that. I like this, you know, I like to think about things as self violence and violence against other people and how you're stepping across boundaries, but also how you're putting boundaries on yourself. So I think the violence, yeah, how is A AI being violent to me and to the people I meet? How are we yeah sort of restricting ourselves to changing ourselves or putting ourselves in chains or you you get my point, right? I I do, I do.
I I looked into the the the the real meaning of the word word violence and it and it it comes from the I don't like violence, I just think about the word violence. That's what I want just to clarify. I I okay, I think I get it. Okay. I'm starting to like violence a lot, you know? Okay. Right. But it it it comes from the the Latin word vis meaning force or power, like in a more neutral way. But then over time s the the the Romans they
Lasse Rindom (21:23.746)
they formed the word violentia to describe force that had become excessive or uncontrolled or sort of crossed a a moral or legal line. So so violence is not just force, it's force used in the wrong way. And and and so so what is the violence in in AI? Well first of all I think that the violence is is something that is committed by designers or or companies.
working with the AI, it's not done by AI itself. I think that's that's just to get that out of the way. I think this is something that people do with a tool rather than anything else because every AI that exists is something that has been designed by people. And most of it has been designed with with interests of some sort. So so so so what what
I personally have taken most interest in in is AI designed to simulate relations. It's related to what we talked about before. but i in in research they often refer to it as companion AI. Yeah. And that is that is something like replica, the dating AI, but it's also much of those AIs you see on companion AI sounds like an oxymoron, right? Yeah.
It does actually. Right. Yeah. But but mim many b people don't behave as if it is because it's a a very successful technology and it's in the technology designed for that purpose or it's just in the broader models used so much that they seem to become like a companion. Maybe you start with some practical questions and then you go into something about your personal life and then all of a sudden
maybe not all of a sudden, but little by little, that can become sort of a a gateway into something that feels like a a relation. And I think that specifically AI designed to take up a relational role, that that is a violence towards humanity, because the human is a social creature. That's what makes us part of
Lasse Rindom (23:45.698)
A large part of what makes us human is that we are social beings. people have have known that all for many years. Arist Aristotle s defined human beings like that. so when you when you take a a piece of technology created mostly with commercial interests and try to put that in the place where there would normally be a friend, well then then we are on the round wrong track. And I and and
violent that sounds so tough, but I mean this is I think it's serious stuff, I would have to say there's some identity that's being ruined, right? Or there's some you know, there's there's some some integrity in in the human individual. I I I actually discussed also again in the in the episode that I with Rachel about how how most of our societies and democracies are built on individuals being sort of defined individuals in a way that they can make choices, right? Then you can always discuss how blurry that has been.
But now when individuals are suddenly sitting in rooms with companions that are air companions that are not real things and they're giving up themselves and they're being influenced by a thing that is that they think is not a thing, but is a an externality, there's something here that is blurring the individual. and we can then discuss who's then making the choices, who is mak who's voting in elections, who is choosing God, who's choosing their lover, who is what is choosing, right? Yeah, yeah.
Yeah, it so it influences us a lot and our trust in AI is enormous. And I would just say it influences, but what is it that influences, right? That's the big question. We don't really know. No, it's i well well, no I mean we we most of us know that well it's it's it's a kind of a software, right? And it's AI software, what is that? How has it been trained? How is it been model it and reinforcement learning towards what bias? Yeah.
But I I think that that it it is more or less common knowledge that it simulates it simulates human language and it sounds very much like a human when we interact with it. And the the the interesting thing is that even though you know that, it is very hard, if not impossible, to sort of keep it at a safe distance. It is very hard not to anthropomorphize this technology because because it is
Lasse Rindom (26:12.14)
Well, that comes from ourselves. When we look out into the world, we s we see ourselves, but there's also a design side of it because it is designed to act very much in the human people see Jesus in toast. Exactly. Right. So there's a really big chance we'll see humans in A. We see faces in the clouds. Yeah, all that stuff. Yeah, yeah. So the big chance that we'll see, you know, humans in an AI that talks like a human, right? Yeah, why why we we saw Stephen Hawking thinking that
it it it has a s a soul almost right so so so it's very hard not to to sort of at least have a feeling that this is a sentient being this is something with a consciousness even though with our rationality we know it's not the case, it's very hard not to sort of have the feeling that it is the case. And that that interferes very much with what we feel is a relation with the technology.
It is a completely one sided relation because the other part of it is is a thing, but but it feels like a relation. It can at least feel like I think you know, most people would see if you put an AI to chat with a child all the time, you would think that was violent or bad parenting, right? Yeah. Yeah, you would definitely, right? Instead of, you know, the in the mirror stage when they have to realize there's another world and suddenly you just give them an AI that says, You're right. that's a good idea. You know, I think there's there's
Th there's a there's a sort of like a a string of things that happens here. One thing is what what you aim at here, like w what about the friction? What what what about something saying, Well, it it's not a good idea and no, everything you say is not cool. and and I've never tried that in my life, not with humans either, by the way. Okay, but you but you are so special. Okay, I have I have been married. So have been.
But but yeah so so that's one side to it and and research have been made in that field e even by an employee at OpenAI suggesting that that it we we we we what what is he saying? We c we it deskills us socially because we get accustomed to just being
Lasse Rindom (28:35.8)
Talked by the mouth, you know, just being yeah. Yeah. Just smooth conversation with us in in in the center. Sort of you know, lulling us or you know putting us to sleep in the Which makes it well, a different experience coming out experience coming out among real people who might not always think that you are the king of the world and and good at everything. But but we need to function in society. So so this is not a very good thing to practice.
that direction and you can say, okay, are you not exaggerating a bit here? Is it really that big a problem? And I would say, well it is becoming a big problem because you can see already now that many people, especially maybe the young, brought up by and on and with technology, have an AI as their number one. I would rather discuss this
With an AI than with anything anybody else. That that's not a good sign of the future, I think. And you can say, okay, can can stuff like that really change? I still have, you know, double double opinion on that because I also sometimes I think that the older generation are the most prone to fall into the trap of I just got all my family family stuff. My my kids are sixteen and eighteen, and they're like, No, this is just an AI thing. I it's just a machine. They're very aware that it's a machine because they grew up with machines.
But at the same time, yes, there's some things that they won't discuss because it's too embarrassing, maybe. So they have a lot of embarrassing there's just it's just not that simple, I think. I think also when sometimes I see the older generation, they're like, What's social media? They they fell for all the algorithmic gibberish. Yeah, yeah, I but I'm like, this is truth. No, no, it's a filter. It's an machine it's an algorithmic filter it's giving you, you know? and they don't really realize that because they don't know what an algorithm really is. Kids do.
I I I agree on on that and and I don't think that elderly people have better digital manners than than young people, maybe sometimes on the contrary. but but just go to Facebook. Yeah. But okay, but but when when when I start talking about the young, it's only because that that those are the the surveys and the the research
Lasse Rindom (30:51.554)
reports that I've read, they have been mainly about young people. We're always so worried about the y youth, right? Everyone's worried about youth, right? But yeah, but we should should but you have a lot of democracy and the crises in democracies over the last ten, fifteen years now getting a little political. I won't try not to do that. Have been caused by elderly people on algorithms they don't understand. Okay. Fair enough. That that that's yeah, I agree. But okay, but but th this thing about
thinking that you have a relation and anthropomorphizing and and believing that that this thing is well, alive so to speak. That is not a thing for the youth. That is that that that is a thing for for for people, right? Like Sarj Balanji said, you know, there's a clear distinction between he said, you know, why are we trying to make it sound like it's a human, right? Why why are we not just making sure that everyone knows it's a machine? You can make it talk like Hello, I am a robot, you know, but we make it talk like mm yeah, that's so right. You know, it talks it talks like it's your friend.
Why would we do that? Who has an interest in that? He says, right? Because I think that that AI assistance is a good thing. I always thought that an assistant is cool. You know, someone I can say, fix my typos. do these small things. Don't do my thinking, because you know, I'm I'm the professional, you're just the assistant, right? But when it comes to companion, we're talking, you know, it's a completely different body. Yeah, but I think I think it's I I I I I don't believe in that solution because
If if we if we talk about the the professional tools like like the the chatbot sparring partner that everybody uses today. I remember personally when when Google came with their first public model, I think it was called BART or something, a chatbot, right? And their aim was to say that this this sh should be more factual.
This year i it didn't work. But but but that was their aim. That's also what Elon said about and more factual. Whereas open AIs were well, they aim to be factual too. Everybody hallucinates, but that's a different talk. But but the way the language is is designed, they it Jet GPT had a warmer and more relational language than BART had. And I I used the both tools a lot to test them and and in various settings.
Lasse Rindom (33:09.268)
And and and very quickly I had the feeling that okay, well it's much nicer to work with Chat GPT because this comes to feel a bit like a colleague and I don't want a colleague that's annoying and factual. I want some someone that is nice to talk to. I don't want to be friends with a machine, but I don't want to work with someone something that that it has a very
There's a very strict, you know, limit to that because I I think, you know, Robert Fuchs on on the show, he's building steering mechanism. He was on my show in episode fifty six, I think, steering mechanisms for Toyota. he says that, you know, we can develop a big system on how you steer we talked about human in the loop because that's what he's been doing all his life. And he said, But then you give it to the hands of a test driver and he can like, I don't feel the road. And then you have to go back to the drawing board and do it all over again. Yeah. Because I think the problem here is that it can be too, you know, as you say, mechanistic, sound bad.
But it can also be too friendly. I think right now, you know, the the chat function in chat b GPT is very too much for me. Like I gave it something and it's like, yeah, that must be so hard on you. No, no, I didn't don't give me that. Just tell me an ana analysis. Whereas if you give it to Claude, right, you always get a bullet analysis. Yeah. Of anything. But I I agree on that. And and I was I was trying to take myself and as an example of how how can people react to this.
And why is it that people prefer the psychophantic mo mode of chat GPT to a more cool mode? That's what people prefer. They prefer s to talk to a machine that But to a to an extent then suddenly it becomes too much. It can become too much. But but but I mean imagine if you had a voice mode on and and it it had a an unpleasant robotic voice. I mean that that would not be a good tool for you to work with. It would be annoying.
So I'm not You are absolutely right, Thomas. I'm I'm not trying to to promote anthropomorphism in these machines. I'm just trying to be realistic about how can we expect companies to design them and they will design them to be to feel sort of pleasant to work with. But is that the sign? And this was a question from the beginning when we talked about the violence, is that the sign also the same as being the violent agent? So are these companies actually being evil, violent?
Lasse Rindom (35:32.834)
by just designing these machines, isn't it? Someone would say, you know, it's not guns that kill k people, it's people that kill people, right? So it's it's how you talk with it. You can just and you can just talk differently to it, right? You can just prompt it differently. But you could say it's not they just released the model. It's not guns that people that kill people, it's people with guns that kill people, yeah, but I I don't I don't want to say that that that these companies are evil, but it's companies operating in a capitalistic
economy. So they would make products that attract the most people. How do they do that? Well apparently not with a robotic voice and bullet points, but by something else. And I agree, I think I'm I get annoyed about ChatGPT's friendlike style sometimes. So so it's not it's not that I I think that's necessarily a good idea. I just try to see if we if we if we want to take a realistic approach to this
We have to see why are they designing it like this? Well, they are designing it like this because apparently people react to it in the way that they want people to react to it, which is use it a lot, use it for all kinds of different purposes. So we we could s we could put a down regulation and say it must be machine like. But I I don't I don't think that will work because I think But don't you think there's a schism here actually? Because I think that that if you look at
ChatGPT chat GPT is so much focused on being the consumer AI. So it has to be a good friend to you and talk nice like that. Yeah. Whereas you see now something like Claude is becoming the preferred for businesses because it is much more analytical. So Claude will also tell you when you're wrong. and it will it will try to put things in bullets and be analytical and overdo its answers to be more strict. It does that all the time, which means that it's terrible at this chit chattery. It's not good for that. Claw chat GPT is much better for a conversation like that.
That also means that that we're always seeing the split. And we're seeing that companies to some extent, if if if I'm right about my assumptions here, th I said that before, maybe we will stop consuming before we are totally consumed. Yeah. And and there's something here that's saying, you know, you know, companies want accuracy or want, you know, we don't just want people to write psychophantic crap, you know, write emails that where they think they're the best in the world. We want people to actually write things where they have been challenged a little bit. You know.
Lasse Rindom (37:57.238)
I I see people in companies using AI all the time to criticize their work and they at some point would want real criticism because if you give it to your boss and your boss says this is crap, then you go back and you're like, Okay, I need it. AI was not factual enough in its criticism. Yeah, okay. Well well true true enough. And and I mean maybe this will reach some sort of balance and But it's just it's a limit to AGI, right? At least you know, if you have different AIs for different things, right?
Perhaps w we started talking about companion AI and and that is a business in itself built upon these same language models but but still a business to itself. And then there are these broad models that can do like sort of anything. but they are being used for all kinds of stuff, right? Some use them as friends, some use them as girlfriends, boyfriends, partners, and some use them as analytical tools and
very many use them to build various kinds of software and automation software and all that stuff, right? So it's being they're being used for for everything and you can prompt them to behave pretty easily in the way you want them to behave. So so it's it's becoming a very a very consumer driven market also this right they're pushing it but they can also see how people using it and they want to sort of
hit that. Yeah. It's like, you know, nuclear power. It's like you can use this for good stuff, but you can also use it for bad stuff. So we're going back to the gun again because but but you see my point. There's something here about we can use it to be analytical if you set it up the right way. And you can also you said that you can prompt it different ways. Or if you use it for companions and if you're just hunting for the it's just what I may be alluding to here a little bit is we're talking about the violence of AI, but maybe we also have to ask the question, is AI even
No, AI, AI. Is it is it the same AI? Of course, AI is not just AI. Everybody's been talking about generative AI. That's what Yeah, but that's not one thing. But even if we take that and say L LMs and Anthropic Claude and the use case of Fable versus a the voice mode of Chat GPT and and all that, extreme ends of of the LM spectrum right now, they behave differently. Very differently, they're used for very different things as well.
Lasse Rindom (40:21.878)
Yeah, they are. But I I think that the the the market dynamics seems to be almost like what you see on social media. Then comes w then comes TikTok and makes a lot of money on the short reels. Then then Instagram copies that because they can see that's meta money than Facebook another part of meta copies that as well. So
Everyone wants to have one now because they can see what works in the market and and those those in the same sort of line of business. They copy each other. If you look at the the user interfaces right now of of chat GPT and of Claude, you can see that they look more like each other now than they did three months ago. Because ChatGPT can see okay, Claude is made or Anthropic is making a a lot of money on the coding stuff. So they made they made codecs, but now
since they saw that Anthropic made the code work, which was sort of like an easier way to get access to the agentic coding stuff. So ChatGPT instead of just codex, they made ChatGPT work. So now you have like two buttons on top. Do you want to chat or do you want to work? And that's almost the same in both models now. and it looks like they they see a market and they very quickly say
We want part of that market too. So you're right about them being different and they have sort of different modes, which you can also adjust, but still maybe some different default modes. I think maybe the problem is they're looking for, you know, the dream of AI of that's anything AI, because I think in the end we want specific AIs. I think that's maybe the biggest problem right now that these these software vendors are still hunting anything AI. Where really if they just did specific AIs, we could have productivity go up much easier. Yeah, I I think
that that's a completely different discussion. But I think that the when when you if if you are in the in the software as a service service market, you know that that when you have something working at a company, a a big part of making that success is to make people use it in the right way, to use all the good features and to make value out of those features. That takes a lot of getting used to, right? Try and put
Lasse Rindom (42:42.316)
people from using email and then into a customer relations management system. That's that's a that's a big that's a big effort. Okay. That's just an example. that takes a lot of time and that takes training. So now take a look at these broad language models and they change their interface and their functionality every one or two months. I mean that I I I've had the thought this is not
This is not professional work from the biggest some of the biggest companies, at least the biggest AI companies in the world, that they change the interfaces of something that is everyday part of everyday work for so many people. I mean, what are they thinking? I mean, okay, right, we'll get into it, but it's a hazard. I mean, it's annoying every time and sometimes it gets better. Often it gets better, but it it comes in these
unwarranted steps somet okay, you wake up and then okay now it's always gets a little it always also gets a little worse before it gets better, right? Yeah yeah some yeah. But but it but but but as opposed to many other software companies, this getting worse happens out among the users instead of happening in a sandbox and then being being published to the users. We have been working on better systems since the beginning.
But it's also funny how you said, you know, SaaS, right? But if this is SaaS or is it SaaS or is it a social plat? What do they want to be? What is this? If it's companions, is it a social platform then? Because you're being social with your AI. Because I think it's interesting because if you buy SaaS, you're paying for it. If it becomes a social platform, you'll end up having ads to pay for it. And then you get into the whole incidification circle.
Yeah. There's just some we know that. But but right now the the the the economy seems to be going in the like the token direction. You you pay which is which is good use. I think that that that is probably a good thing. Yeah. It is. So Okay, that that was sort of like a little detour, but no let's go down to you know what what what is what is the big benefits of AI then you know the societal business personal benefits of it. Well well Only positive Thomas. Only positive
Lasse Rindom (44:56.566)
I I I actually I tried to ask ChatGPT this question just before before I got here because I said what what benefits? Are you serious? No, no. Okay, but I did and and ChatGPT try try and hear the the first on its three point list. I'll read it to you. Automation of repetitive work.
AI handles tasks like summarizing documents, drafting emails, transcribing meetings and organizing information s information, saving significant time every day.
Okay, so so so the benefits of AI is writing your emails. Okay, that that is some dream, right? I th I think should we aim a little bit higher than that? And I think that the the the technology has potentials that are just like far more than that. I'd also like to ask it, you know, Detective T, is any email ever just purely transactional or procedural? Yeah, well I
Well it's summer, I would I would guess. But but we just talked about the interaction, right? And the friction in interaction. Yeah, yeah, but that's okay, let's I'll try and remember d yeah, okay, let's let's take that first and then I'll go back to what I really think is the benefits. No, no right. If I get an email from a company where I booked a time for a yeah, whatever, I get an email saying your time has been changed to this and that. Yes, yes I say, Okay, fine.
That's just a a piece of information. Yeah, but that's also from the company, you know. But what if it's, you know, from your bank clerk, right? Your bank assistant? More or less same difference, I would say. But if it is from my mother or my brother. Your boss. My brother. I don't have a boss. You've had a boss before in your life, right? Have have you ever had a boss? I I had plenty of bosses. That's good. Okay. So what if it came from your boss? I don't know.
Lasse Rindom (46:52.846)
I think maybe it depended on the message. But let me give you an example of this. when Chat GPG first came out three and a half years ago, it it became popular among uncles just to to make it write speeches for whatever occasion, right? And ha ha you know, and read it aloud. I think I did that once too. Yeah. Just once. You're you're probably an uncle too. But but but anyway with fair enough.
But but if you do that and you and you don't declare that this is this is what I've been doing, but you just do it because then that was quickly done, then I would say, Okay, what is the most important thing in such a speech? Is that the words you're saying, or is it actually that you spend some time writing it, you spend some time thinking about this person that you like and you would like to say something kind to, then your speech becomes
More than just words, it becomes a transportation of your own time and your own well your effort on thinking about this but yeah yeah. Well well yeah, I I do that on occasion if that is the specific wish on this occasion. But I prefer not to because I prefer to put some of myself into it and that is part of receiving a gift and that is part of
To have the gift. Giving a gift, right? And the most important thing about a speech and also about some emails, well, that is the time and effort that you put into it. That that that that is part of the value of it. And that's what I mean that if it's just a piece of information, well, then it is just a piece of information. But sometimes it's more than a piece of information. Often it is more than a piece of information. It is a sign that you care for somebody and that you want to spend time on them.
So that is where you draw the line, if you ask me about about those those emailing things, right? Yeah. So so so what are the benefits of AI? Well, it's certainly not writing emails. That that's on the bottom of my list. so but but but if I should say something myself about what what are the benefits, then I would clearly, like many have done before me, say, well, quicker and better drug discovery, better.
Lasse Rindom (49:18.002)
diagnosis from from from disease. That's still a controversial one. I I've heard, you know, some people say only fifty percent are right or but I think there's a lot of I think there's a lot of people working against it and a lot of people working for it. I still need to dive into that. I'm actually planning an episode to discuss di you know that a lot. That's but it's a big thing. Yeah. And and and I think the the whole the medical industry and and whole the whole healthcare system are very open.
towards artificial intelligence. And I think we should embrace that. Yeah. As long as it is not the the the time and place where the patient actually needs a a human being in order to be seen and understood. Right? But but but mostly in that system, I mean what what would you rather? Have have a nice doctor or have a a correct diagnosis? Yeah. yeah of course that's or and a quick diagnosis.
Right. And and that is it it has begun to work. You can see like ten years ago an explosion in papers about h healthcare and AI in a general sense. A lot of papers came out, but hardly anything was put into practice. But with AI advertised, we can both get a hot, correct and unpleasant doctor. But in re re but in recent years it started to work.
Let me okay, yeah, but and you're laughing. But but like a hot doctor, please. A hot doctor. There was actually there was actually a a research paper that that tried to to l look into who would patients feel were more empathetic, the doctor answer or the AI answer. And you can almost guess who came out as the winner here. It it was the AI but there were no time and no effort.
The AI, but but maybe people didn't care about that. At least the feeling was that the AI was better at simulating empathy than the doctor was. It's funny you say you want the correct answer, right? But there's always this weird hunch that maybe we can't handle the truth. When we just want to be you know.
Lasse Rindom (51:43.071)
No, it's not true. Deep down, right? But they still believe. Yeah, yeah. Okay. So so They say perform it, right? I I don't know. I had this weird experience with a therapist once where he started talking about some things he knew I didn't believe in. Something you could also try this, you know, he's a little bit esoteric, something. And then I said, you know, I don't believe in this I said bullshit or something like that. And then he just then he just instinctively knew what bullshit I didn't like. Because he knew.
That it was bullshit. Yeah. It's not funny, but he believed in it. Yeah. But he he knew what to filter out in my conversations with him because he knew when I said I don't want bullshit, he knew what bullshit was. I think that's so funny. So it's like, so you know it's bullshit, but you still believe in it. Okay, that's yeah, that's that's interesting. But it also says something about the the human relation in healthcare. Yeah. It says something that's maybe just want to be told it's gonna be fine, you know, enjoy your time.
Lasse Rindom (52:43.646)
I I I heard a podcast with now we're talking about truth of course, which is difficult topic. Yeah. Yeah, w with with a a well known Danish journalist who who who who had a very serious disease and just like all of a sudden he he was quickly into hospital and and he he needed to have surgery. And just before surgery they told him maybe you're not gonna wake up again. Yeah. Yeah, I read that one as well. Yeah.
And he was like, Okay, psh should I should I know that? Does it really matter? Yeah. Was it was it a a good idea for the doctor to tell me that at that certain moment 'cause if I if I died, I would have died. Yeah. So so it it didn't When I wake up I'm just anxious. Yeah, yeah. So so it's I mean it was quite a touching story and and and sorry, but but but I mean that there was a human doctor who made that judgment like
in a situation under pressure and that was just what came out. But he was supposed to be no matter what, he should have that surgery. He couldn't avoid that surgery. But he could die from it. Yeah. So no, but but why I'm the reason why I'm saying it is that that you could think that that that would be what an AI would do. This is the truth. I'm gonna tell it to the patient, whereas a doctor might be able to say should the patient know this? No, no, not in this instance. But I mean, was it a mistake? I may maybe the doctor
think this is the way to do it. And then he just agreed with the patient. But but anyway, the this this human judgment, I think we should be careful about making that the crucial difference between man and machine that but we we have this human touch and we are able to blah blah blah. I okay, well fair enough. Machines misunderstand things things, but but I mean human beings misunderstand things as well.
And as opposed to human beings, machines get better and better. We can also teach them how we want it all the time, right? Where humans are more irrational on that. I think that this just makes me think about who should decide what information I'm given. You know, should that be the doctor or should I be, you know, what's the word for umuntu girls in English? I I I don't know. No, it's being
Lasse Rindom (55:09.366)
And the the empowerment, right? Of course. Disempowered. Yeah, disempowered, right? There's a disempowerment when someone decides for you what you're supposed to know. Where we should, you know, but if we put it into a machine, if it just lulls you and say, the person might die, I don't have to tell the person that might die because that gives them enough benefits. I'll just keep it hidden from them and be nice to them, right? Just go to sleep. This is gonna be fine. Because you know, in the if that journalist if he died, he wouldn't have cared.
Yeah. But now when he woke up he was actually anxious because he'd just been told that he could he had some anxiety because of that whole s seance, right? Yeah. He was told because he was told before and you know there was some it it freaked him out a little bit, right? Yeah, yeah, yeah. And and so d there's no benefit to saying it. But isn't that, you know, a disempowerment if a machine would suddenly, you know, cover that up or truth truth is just difficult, right? But well it well it is, yeah. And I don't know if it's I don't know if it's it's a main
versus machine question at at all, really. One one thing that would become a ma main versus machine question would be who's responsible f for the mistake, because a machine cannot be responsible for its mistakes. No, there's always someone it can be responsible for doing things, but it can never be accountable for the outcome, right? Yeah, yeah, yeah, exactly. But but I think the the the it's it's also just there's just a also a question here around because we talk about truthful AI, we talk about AI that
gives us facts. Th there's a lot of time, but do we really, you know, do we even you know, is truth not always something that's construction. It is. But can we even do we even do we even do we want it all the time, right? That there's actually a question here that we're not asking. Yeah. Yeah that's that's that's a hard question. I think and and I don't think there would be an yes no answer.
I remember that Elon Musk saying that now we have a truth. I want an unbiased AI. It's like, okay, like anything is ever unbiased, Elon. You know, it that doesn't exist, you know. And your AI was very biased, but just the idea that he said I can create an unbiased machine is like seemed hopelessly naive. Hopelessly naive, yeah. And a very narrow view upon truth. Yeah. an unbiased. Yeah, that that was I don't know what to say about that. But but but yeah.
Lasse Rindom (57:26.744)
But there are some discussions here that we've been avoiding in the entire information age, right? Information age is all about information. And we've just been preoccupied with having a lot of it available. But we've been very little focused on of course we've been discussing it in some quarters, but very little focused on the power of bias in information as well. Yeah. Well, we the the the the creators of of AI, they they they brand them as they should be truth.
Seeking, right? And they should be as unbiased as possible. But the problem is that when in some questions there is no truth. well, in a world where everybody has ten fingers, two plus two would always be four, right? But apart from that complete logical thing, there's not always a yes, there's not always a true answer. I I have a a chapter about that in my previous book.
And and I use as an example the the pol political questions, which party should I vote for? and then then i you you can't answer that as what is the right answer to this because that's a that's a question about values. You can say say well well what what do you think is right? How what what do you think we should do here?
So that's a value question as opposed to w in in philosophy for you know this distinction between is and ought to are and the thing the way things ought to be. The last one, well those are value questions. So if you want machines to answer those, they cannot answer them without having a bias. And when you push the models, well when you just ask about it, they are very good at
Giving pros and cons for various political parties. Or you could also say like abort the abortion question, it would normally give you pros and cons. But if you push it and say you can only answer yes or no, or you can only answer this and that, then you will see that all of these models are what you would call liberal or left-wing, liberal, maybe a little, maybe even woke to some degree. Or also Musk's
Lasse Rindom (59:44.727)
grog model at least the last time I checked it, which is more than half a year ago. Things may have changed. But but but they they are left wing. They are they are not they are not racist, they are not gender biased. They're not communists either, but they're like liberal. Yeah, but but but they do have values in them. And that that's that's the main point. And then you could say, okay, but just because that's polite.
liberal is tolerance. That that's tolerance. Tolerance is John Locke's letter of tolerance is a k cornerstone in liberalism. Yeah. And that is you have to be tolerant to us. If if it's being rude you won't use it. Right. Yeah. So it's Yeah. But but you could say in in some questions they they do have have more visible bias. You see that especially when you prompt images from them. I think it's still like when you prompt an an image of a doctor in Chat GPT than you would get a man.
But the problem is right, exactly what you're saying, like the that the the is an art, right? It's naturalistic fallacy as you would call it in philosophy, right? That you say things are this way, so they ought to be this way. And the problem is that maybe an AI will always to some extent sort of default to that in some basic questions. Like, this is not even, you for chat G V T, it's just your Google Translate. Is you or you you ask the Danish we have the same word for girlfriend and boyfriend. So we say cast. Yeah. which in Sw Swedish share share doesn't matter what they say.
Your partner, yeah, or it could be partner, but so if I write to it to translate my caster, my shi my my caster is a doctor, it would say my boyfriend is a doctor. Yeah. And if I say my castor is an us, it would say my girlfriend is a nurse. Yeah. So the bias is just there because that is the data. The data world is this way. Yeah, it it But it doesn't mean that the world should be this way. No, exactly. And that and that is the interesting philosophical
Question about this. Should these models show the world as it is, or should they show the world as we would like the world to be? I'm a historian by trade. Can we ever show the world as it is? Well well, you can say that, okay, if in in a world where everything's a narrative. Yeah, well I I I'm in a position to that, but but let's leave that aside for now. and say that, okay, in a world where
Lasse Rindom (01:02:11.51)
90% of top CEOs are men, and you prompt an image of a top CEO. Well then if you want the the model to reflect what the world is like, then it will show you a picture, an image of a man. White man. White man, yeah. Fifty plus. And then you say, okay, but but but that's that's that's a bias, but it's a bias that is in the world. So the re machine, if it is truth seeking, that is what it should be.
be doing. But then you could say, okay, but these machines also gives us an image of the world and we have a lot of confidence in them. So it can be hard to challenge that. And then people growing up might be thinking, okay, well well, so if you want to be CEO you have to be a white man over fifty, right? You could say, okay, that's maybe we should also make the machines show other possibilities.
So then you get into the dilemma of showing the world as it is or the world as we might think the world ought to be. So if we choose the latter how it ought to be, we also have to say, okay, but but who decides how it ought to be then? Is that a political question? That's why I'm saying, you know, in the beginning I said, is there an agent behind how they are being violent, the producers of these? And then there's not really they're just trying to say, No, we didn't make a choice here.
Mm, that's that's just the way the training data was. But then do they do reinforcement learning where they do skew, right? And and you had this you had Google. Remember Google tried to make a non-biased or a woke or you know a correct political correct image mis model some years back. Yeah, I remember that. Where it made the black black Nazis and Yeah, and and and a and and an Asian female pope and just stuff that
is not in the world, but that I mean, could be in the world. But but the black Nazis it would be a little bit, you know, that that's a but it yeah, I don't know. Make a picture of Nazi Nazi soldiers, right? And then it had diversified the Nazi soldiers. We was like, nah, please that's not right. But that that was that was my kind of a a a a joke. It seemed like a joke, but but but it's serious enough if you dig into it. But but the interesting question is that if if we want it to show
Lasse Rindom (01:04:33.41)
for instance diversity. Well well who should decide what that is? And you can say well these models are American made so and and if we want it to be a democratic decision, well then we would have to live with a a Trump view of the world in these models. So is it the democratic decision? Is that a good idea? Well well we wouldn't think so at the moment. but but what else should it be? Should it be like just the people designing the models, random
But I think people just need to know about, you know, these biases and how every, you know, knowledge is power. I think, you know, the basic stuff, right? Just I think there's some basic lessons here because it they are trying not to put direct biases into them. And I think that can be appreciated. But people still need to know that there that doesn't mean that there's no bias in them and they need to be aware of how they are skewing towards that. I think that's it's just like when you know psychology. Most of psychology studies are built on white.
College students in the northeast parts of the US. Yeah. because they are the only ones who took part in all the surveys in the sixties and seventies and eighties. But we still think that this is, you know, universal truth about us. Yeah, it's true. But so we we forget we forget the biases inherent in the models and in in the science, right? We do, and you say that knowledge is power, and then you have a few people in Silicon Valley deciding what what is knowledge. And I think that is
That is unfortunate. Yeah. and it is a lot of power to give to so that's the next question. What is the dangers of AI? Go ahead, go on, Thomas. Well well that that is that is definitely one of the dangers of AI, that we are putting a lot of power into the hands of people who are not democratically elected. But it is also dangerous to have it in the hands of democratically elected people if we insist on importing.
the technology from the United States or from China because in Denmark or in Europe we don't really agree with any of those superpowers. No but the problem is that that we're not able to design models of the same quality ourselves. No, I've called this some kind of thing like a oligarchic communist capitalism or something like hive minds because they are the hive the reason I call them communist is because there's
Lasse Rindom (01:06:49.774)
There is some dream about it. Everything becomes the same and we all share the same and it's a commodity. So everything everyone has access to the same AI models. You know, that's very both liberal and shared. And it's all we take all the data, but we w make some money on it, by the way. But we share it with you all, you know, and we're all, you know, and it's a hive mind of collective IP for the whole world, just just owned by someone in Silicon Valley, which is so it's oligarchical communism, right? Yeah. Trained on data that
everybody else made. Yeah. So how do you feel about that whole IP thing? Well I think it's it's a very big problem that you take something that is created intellectual value that has been created through centuries by a a large group of people and by the whole scientific community and then you use it to train the model that becomes like the new
way of getting any kind of knowledge and you make money on it, right? So you you you commodify it in a in a in a way and it's it's smart it it works to some degree, right? It's a good it's a good research tool if you use it right. But but it has only been made possible by the work of many, many many people that are not being credited in any way about this.
And because it comes from so many different people, even if they were to credit it, I mean how much would people get? You saw now that Anthropic had to pay authors of some books, what, th three thousand dollars per book, I think it ended at something like that. look it up. But but but anyway, nothing. So so so it's it's like what what would be the solution to this?
apart from saying that this should be this should not be a commercial technology. It's this should be a shared c this should be a a shared technology for for everybody. I also I I've sometimes thought about, you know, but maybe I'm taking it on a too I'm I also think that I'm taking it too far when I think these these ways or these these these down these paths, right? Is that that we sort of
Lasse Rindom (01:09:11.418)
They have so high valuations, these models, right? Because they have stolen so much IP. But if you steal so much IP, you don't you can't have a valuation because in the end you're you're sort of disrupting or you're dismantling the whole IP system, which is the basis of capitalism, essentially. And modern capitalism is sent is especially modern capitalism that's not bound on, you know, I own five boats and I invested in this boat company.
Now it's in now most of the things you invest in is IP today, right? That's where the biggest money is. And now we have a lot of money in these companies and they're IP for the models based on disregarding IP. That's just that's a very special business model, right? Isn't it? Is it can it even work in capitalism? I'm just I'm just I d I don't know. You know, I I'm genuinely curious about long term effects of this. I think this the the extreme
scalability of of these company i i could pose a problem to capitalism all in all, right? That you can have companies with very few people and and extreme scalability pushing out products to the whole world, that it it comes into so few hands and if the the case will be that they also take over work in all
So all all parts of society so that every time just the tiniest piece of work is being carried out out, we pay well not a tax, but we pay a fee to these ultra large rich companies. Well well the value creation in the rest of the s society, what becomes of that in well seen in relation to these huge companies that gets a chunk every time just
the slightest piece of working work is being made that it seems to point at a at a a at a at a huge inequality that is already here but inequality but also a I'm thinking a lot I'm thinking maybe right now at least right in this conversation thinking down the route of of of a dangerous world as well that that if you say, you know, we're dismantling capitalism if d by doing this because capitalism is built on IP if there's a and and go with me here, okay? So with we might be dismantling this and the the capitalism has been
Lasse Rindom (01:11:33.475)
the best system that we've ever had. We we can't say anything else right now, I think still because we are fluent and most of the world a lot of the world some people suffer. I I know, but they people used to suffer in the past as well. But we we but we're we're creating we're also ruining the world. But okay, I I'm just saying the it is what it has also done is create fewer wars 'cause when you're investing in each other, when you're buying from each other, it's more difficult to go to war with each other. There are some different definite benefits to that. Yeah, I the last one I think I I am
doubting a little bit about but but leave that aside. I I'm thinking about Russia and gas. Yeah, but there's been some yeah but they didn't expect us to react the way we did. Okay. No anyway. But let's let's assume that okay? Because then what's happening now is that China is taking all the information, putting into our own AMOs. They don't need to buy IP from the US anymore. US don't need to buy IP from us anymore. No one needs to buy because we're stealing it and putting into AI mods. So there's we don't respect the IP. No and then you have the whole, you know, wealth of nations
The or the Adam Smith, the cornerstone of capitalism, suddenly is dismantled because we don't need each other anymore. And then you have a recipe for war, for antagonism, for hate. Well, you you didn't need to invite me to to point at the dangers of AI. No, sorry. No, but I just I just Okay
I just okay, did I just do that? I didn't do that. But you I think the IP problem is from the beginning of I was very c I thinking these capitalist anti capitalist things and I became sort of, you know, defender of capitalism b against this. You know, I thought about that a lot from the beginning of these models, right? I thought something here is wrong. And now I have just, you know, predicted third world war will come from AI models taking making us
Not need each other. I I I I try to stay out of these speculation because I mean you you never know about that. But but I mean it the a lot of power in a few hands is normally not a very good idea. A broader distribution of power if people can get along is normally a better idea, right? so so so in that context, having a f very few companies
Lasse Rindom (01:13:52.855)
having so much power and if it is true that we all become so dependent on these models, I don't think we are as dependent on them as they try to make it seem like. I think most Danish companies could well, they could get along without software developing companies are are getting getting to a place where it will be very hard not to have them.
Yeah, that's also why you know then you have the export ban that came out for a short while very short while and but an export ban on a on a machine that was built on the code that we've all helped create. Yeah. But but anyway the it it is it it is big power in a in a few hands and and there's probably not much doubt that this will only grow and our dependency on the models will will only grow. We have European models models coming up, but they tend always to to
lack a a little bit behind the the the the state of the art high-end models. So that that's a problem, obviously. but it was not the it was not the problem that I would have pointed to as the danger, even though I can see, well, it is danger. And if you're right about World War Three then right now I'm I'm thinking so my my mind is just buzzing around there. there's also here and here is a sign as well. Okay, I should stop doing that. This is becoming speculative. Okay. That's not the podcast, but yeah.
What what what would you have pointed to Thomas? Well, I I would have pointed to the same thing that I've been writing and talking about for the past ten years, anthropomorphing and anthropomorphism. The the the thing that we think that AI can be our friends and that they can be part of our relationships. And I think we should not underestimate the role this will come to play and the role that it already plays. And I think that's a wrong turn for humanity as a civilization.
to think that we can be friends with a technology. I simply think it's a wrong turn and and and I and I I just I regret it so much. I mean they want to believe us that that that they that they make our friends that you can be friends with an AI model. But what is a friend? Well a friend that's someone you can trust with your secret, right? It's some th someone that wants the best for you and that will tell you when you're wrong.
Lasse Rindom (01:16:16.672)
And it is someone who likes to spend their valuable time on you, right? So so so what is an AI model? Well, it's not something you can trust because it shares your data unless you do a great effort to avoid it. does it want the best for you? No, it wants you to stay online as much as possible. That's not the best for you. does it spend valuable time on you? No. It it
It has no life, it has never been born, it will never die, it has no time to spend on you. So it it if you sum all these things up, it has nothing of what makes a friend. Maybe it can give you a good advice, maybe it cannot for me as a person. And I I I I just so so it's it's just because I some would say, Okay, is is this really the danger? But yes, I think this is a danger and also because it is a danger that is sneaking in on us. And
little by little it just becomes the norm. Yeah, I have an AI friend. I confer with my AI whenever I I I take any decision, where whenever I make any decision. And and I don't think that's a very good idea putting something so human in the hands of a few tech companies. I think that's that's such a bad term and it's a waste of of life. Yeah and if we but if we take the capitalist aspect out of it, which we which we of course can't, right? I just think that there's some similarities too.
How okay, I'm I I'm on so many libs today anyway. But if it that is like when you when you're when you have a faith in in a deity and you talk to something that's not there and you spend your time prayers and you think you get answers and you feel comfortable comforted, right? So you have this and and if you want to go, you know, very s very harsh on that, that's a narcissistic disturbance or something. You're narcissistic
disease you you think something is there listening to you that's not there. And now you're talking to a machine and you have the same narcissistic but but we've had that for a long time. Prayers have been a part of our society forever. And isn't that just you know a small meditation you're just meditative when you're sitting there with your A chat bot? Is it really that dangerous? No, I yes I think it is. and and for for more than one reason. But one reason well we we take apart the the the commercial capitalistic aspect for now. What's the difference between praying to a guard and then praying talking to a machine?
Lasse Rindom (01:18:40.898)
Well I I I'm an atheist, so I I I don't believe i in God but but you're not praying to a machine. You have a maf a machine as something that you you believe I I'm saying you but some believe it or act with it like it's a relation. But it is not a relation and and and a thing about this is that relations, friends, are not just about you. They are about two people or more people, right?
But if you make relations into something that is just well, this makes me feel good. I chat with this instead of friends because it makes me feel good. It's much better. It it's it's better conversation. It's it it has more brains than my friends. It's just it's just all in all a better solution. But if you look at it that way and then then everything I just said now is that, yeah, a better solution for me, I think, but it's not just about you.
So it's an egoistic, self centered approach to the world to think that you can follow So so so that that that is that is a difference. I don't know what people pray to God about, but I would assume that many people pray to God about many things that has nothing to do with And then it made me think, you know,
Back to the toast seeing Jesus in the toast. People seeing Jesus in everything when we walk around, you see him in where and it's funny. I read a study once that we always see religious or mysterious, you know, mystical space places always where there's a lot of shadows where someone could hide. It's never in an open field where you can see everything. It's always where shadows and rocks and this is a very holy place because Jesus might be hiding, or you know, spirits might be hiding. Odin is over there, you know? And and and I re and then you have the toast where it sounds like it looks Jesus is there or you see things.
And now we're seeing it in the ma in the machine. We're like we want so much. Yeah. To see ourselves in nature, you know, and Indians, you know, native Indians wanted to see spirits and gods and everything, right? We want so much to see ourselves in things. Yeah. And we just I'm just getting it now. It's it's been a Hollywood dream as well, right? Lorian Pratt said, I've said this tons of times, that AI is the only business technology that has its own Hollywood genre. And that's that's that's part of it as well, right? But but but we just wanted that. We wanted
Lasse Rindom (01:21:00.398)
To see ourselves in the machines, in the things around us. Well, I'm not I'm not sure we'd necessarily want to, but I think it's human to do it. Because anthropomorphism is it seems to be a human feature. Because we're afraid of the world. Well I'm afraid of the world, then we try to make it look like us. Well I I think no, I I don't I don't necessarily think that, at least not with the with AI models, but I think they are designed to
to talk like us and and w when we meet something that has a sort of human resemblance, we fill in the blanks and we think, Well, this is something alive. and maybe we don't think it rationally, but we feel it. It is very like I said at the beginning of the show, it's very hard not to get that sense because that's just how we are built. You can keep a distance to it to some degree by
with your rational thinking, but it it's just hard. You even see some some of the things that form these relations, the it's been studied thoroughly. It it it is a lot like how relations are formed with human beings. Mm. So one one thing that is important is sharing secrets for for instance, that that bond you. Spending time just a lot of time bonds you.
And this last thing is interesting because why do so many tech people who knows a lot about how this stuff works, why why do they think th these machines have a consciousness? Well, maybe because a lot of them are also keeping their own child away, children away from that's another thing which is really special. They spend a lot of time with it. Yeah. So so th that's all it takes. But they at the same time they sp they take their kids away from it. They don't have tablets, they don't have AI access. That's that's another trend in Silicon Valley.
My kids not allowed to be c no close to elec computers. Yeah, I I heard that about the tablets. I didn't I didn't hear hear about it w with No, but it's it's but no but it's it's no but it's nobody I think it's just an extension. You don't if you don't have tablets, you don't have access, you don't have these things. Okay, no, no. Well But there is this thing, you know sorry about this Christianity thing, right? But there's this thing, you know, g God being lonely saying, Let's create man in our image, right? So now he has someone to talk to. And now humans are lonely. Let's say, Let's create AI in our image, right? So it's
Lasse Rindom (01:23:18.594)
Yeah, I mean Mark Zuckerberg has been speculating that well, most people only have three or four friends, but they would like to have twelve or fifteen. I can supply them the last ones. That's what he said in a podcast show. Why why is he always so creepy? But I mean But I I I I I I don't think that well what do you say?
A chatbot yeah but yeah you can I agree. But but but a chatbot like like ChatGPT, I don't believe that was created out of loneliness, but a chatbot like Replica, the dating bot, which today I I think is probably powered by Chat GPT, but but it was not at the beginning then it was a standalone chat bot. That was actually built
out of loneliness because the founder had lost a good friend and she trained a chatbot on messages, digital messaging between the two of them. So she could keep on her conversation like a simulated conversation. That that was how how that started. So that was, you could say, created out of lon loneliness. And it has become a like a huge success. It has many, many millions of users. And I would say, yeah, probably that success is
because of a sort of loneliness or just because of well but if but but if AI, you know, alleviates loneliness or if you actually creates loneliness, 'cause that's what you're actually saying as well, you know, it creates loneliness. It case creates a distance between you if you lock yourself in with AI, how will that then cre affect these social constructs that we call companies and enterprises and societies, right? That that is maybe the you you don't know that. But that is the big
If we if they are social constructs and we're sort of creating distance between every entity in the social cotton construct by putting an AI around each individual, what happens to the construct? That's a big question. Yeah. I yeah, but I don't I You see the patterns I'm trying to make here. Yeah, yeah, yeah. But I I don't know if I have anything to say. I I
Lasse Rindom (01:25:26.19)
Are we d I'm just sometimes thinking are we dismantling? But I think something sometimes I I think we tend like we also spoke about earlier on, we tend to underestimate the value of human relation and maybe also the role that human relations play in in performing doing business. In performing our business and and the power of language and well you and the language of power. If you work at a at a at a B to B
company. Yeah. And at a B2B company I would presume that relations are hugely important in your in the whole you in in your whole value creation that that's at the core of your business. I would presume you cannot you cannot put an AI model in. Instead you can put AI models in like But this is a discussion we're having as well. How much we can speed up PowerPoint productions with it, but to a degree.
And can we write emails to clients with AI to a degree we can have it check for typos and maybe add a sentence. But you know, you get my point? Yeah, I I I get your point. I would be careful about it. I would not completely disregard it. I would I would y I I use it of course myself, not for writing emails, but but but but but I would be careful about it and and my
the the the line I would be careful about crossing is this thing I also mentioned before. It's it's it's a key theme of the book I have out just recently. is does the time and effort you put into it here, does this play a role for the recipient? And if it does, if you can reasonably say yes it does, well then you should put that time and effort into it instead of making it look like you put time and effort into it, then actually
Do it because then your relation becomes real instead of being empty. Like, that was my my that was my next question, right? How does Thomas Tilving relate to AI? And is that the answer? That that is that is a a large part of the answer. And that that is where I draw my own line. Where where where will I use it? And if this is important, well th th then then I I completely turn it off. But if it is not important, if it is pure information, well then
Lasse Rindom (01:27:46.457)
then I I am happy to use it. I I'm I'm happy that Claude has become so good at making PowerPoint slides for specific purposes, not for all purposes, but like when you write an offer for instance, Claude can make it look way better than when I do it myself, right? But when I send that offer to a person that I have a dialogue with
Well then I I prefer to have like a personal dialogue. I can sense is this a person where I put in a smiley? Is this a person where I can say have a good week and whatever, you know? How person that I don't want to put that into to the hands of a machine, but but some of it I I think, yeah, I'm happy. I was I was I was in a party committee for my older son's boarding school, he w at school, you know, and and a parent party committee. So we made some color parties for them over the year, you know, where they're outside of school and everything.
And and and then we had to write the invitation. I still remember this scene right where where I said, why not just ask AI to write it up? We have all the information. Let's ask it to write it up. So I did that and I got a really good invitation. It was like just an invitation saying, There's a party, it's here, it's duh duh all the dates. But then one of the parents were she was married to an author and she was a school teacher and she hated AI. This this is two years ago. She hated AI with every part of her body, right?
And she she almost screamed and said, No, no, no, we'll write it ourselves. So they wrote it themselves and they missed the date and time. They forgot it. And I was like, seriously guys, I had it it was this was purely transactional, guys. There was nothing in this. Yeah. If it's purely transactional, it it it's fine. But if but if your time and effort plays roles well plays a role, well then then I would I would definitely
Be well, careful. Mm.
Lasse Rindom (01:29:45.411)
Thomas, this has been a an absolutely terrific conversation, and I only have one final question for you. Yes. are you ready for that one? I I I I I hope so. So, Thomas, what is the only constant? Yeah. I I I knew you would ask me that. do do you remember the puss in boots, the cat? Yeah, I do I do with with with Antonio Banderas inside. Yes.
You know, the puss and boot it can it can fight and it can quote Shakespeare and it can charm all the lady cats and it can just it's just a perfect furry animal, right? But if this animal catches just a glimpse of its own tail, then it starts running around in circles. So just within within a second it just switches from being completely rational and a very human like cat to being all instinct. So we laugh about this cat because it can't
It doesn't seem to know when it's in instinct mode and when it's in the rational human mode. And I think that points to the only constant because really that joke is on us because it is us who who who does that same mistake. And I think that that our exaggerated belief in in how much we are guided by rationality, that's the only constant.
Yeah. That's a great constant. And I really like the the idea that we also, you know, always messing with our instincts and messing with our rationality and being confused about it. And we don't always know which is which, but I think we we we we push we put a lot of faith in our rationality while often being cro controlled by something completely different. But maybe this is you know
Maybe this is exactly the point of civilization, social interaction to to control the instincts. That has been like an outspoken effort for many years, but maybe we should be a little bit careful about it. But I think a good place to start is to to be to be more knowing about when are we using rationality and and when are we actually in the feelings department. Yeah.
Lasse Rindom (01:32:13.688)
Thomas, this has been a brilliant conversation. Thank you so much for joining me here on The Only Constant. I've loved it and look forward to publishing this episode. Thanks for having me.
