I think AI has the power to make life better for us, but we have to be very thoughtful about that. I never thought more about what it meant to be human than I did when I was writing as an AI. It's just they're losing the plot. They're losing the point of why we're here. Like we're not here to do things right.
Mariana Lin:We're here to live. I really hope that play is what's next for us. And yes, people may say that it's indulgent to play, but I think being playful is it's necessary every day.
Christoph Magnussen:If you want to understand how AI and the conversation with AI shapes the future, there is a voice and a person you need to listen to. Mariana Lin is the person who actually was recruited by Apple as a creative director and lead writer and author for the conversation with Siri back in the day in the old ages of AI. So she shaped something that all of us are using every day. She has been a speaker at TED Global, also at AI for good at the UN, United Nations, and now editor and writer with many different areas to talk to. It's a much different episode from what we have at AI to the DNA usually because it lights up the field, it shows a very positive side in AI, a very philosophical way, but also practical when it comes to everyday conversation with what we all use every day.
Christoph Magnussen:We traveled to Paris, We sat down, and I wanted to give you a little bit of this background before we jump into that conversation because we had so many talks in between that we jump right into conversation. It's the longest episode so far, so get ready for a very special one. Thank you, Mariana, for making that happen, and thank you guys for listening in and watching on YouTube or wherever you listen to podcasts. It's good that we sit down finally, but the feeling that I have is this is the first podcast I have where there is no beginning and no end. It's more like a conversation we had now for the third time already and now we finally just hit the record button.
Christoph Magnussen:So it's good to meet you here in Paris in person. Thank you for making that happen and work. And...
Mariana Lin:Yeah. It's a pleasure to be here.
Christoph Magnussen:I I know what it means to to you and I know there is a backstory and everyone is interested in that but I would be very curious more to learn about your perspective at the moment, what you feel right now when it comes to language, conversation, and the meaning of a single word in a sentence when you talk to AI?
Mariana Lin:The meaning of a single word.
Christoph Magnussen:The meaning of a single word. How important a single word can be in a conversation depending on if you're talking with a human or an AI?
Mariana Lin:Well, I think I'm not the first to say, to feel that there's a risk that AI is flattening language. You know, we are... We're the sum of the five closest people we interact with every day. If that is AI and we're constantly communicating with AI, the way AI speaks and thinks, we're going to continue to absorb that. So AI, the way AI does language is very, very specific.
Mariana Lin:It is taking the most accurate route, the most compressed and direct and logical route. So we just have to be aware of that. If that's something that we're interested in or we need more of, logic and structure, clarity information, then it's a great tool. But it is changing language in that way. And there's a poet, Richard Hugo, who said, As soon as language becomes only functional, it's dying.
Mariana Lin:So that is a problem right now. That's, you know, that's happening to language because of how much AI is really entering each individual's domain in the creation of language and communication. Does that make sense?
Christoph Magnussen:It makes sense.
Mariana Lin:So that's a risk. And I think that, you know, the problem with continuing to speak with AI is we lose this sense of whimsy, of randomness, of bizarreness. You know, in real life, that's what makes conversations, in my opinion, much more interesting. That's what makes life interesting. But AI is almost counter to that because it wants to be correct.
Mariana Lin:So when you're correct, you just cut out anything that doesn't add to that final result, right? Like nothing extraneous should be in there. Why would I have a joke or a random word or something bizarre if that's gonna not help me get the right result that this user is looking for. So they're actually opposed to each other. Life, like what's living and natural is actually opposed to what artificial intelligence's goal is to do, which is to achieve a task correctly, accurately, and to satisfaction.
Mariana Lin:So it's very counter to our brains to think, oh, do I maybe not want that? You know, because we are socially raised from a very young age to say, oh, if I perform well, if I do everything right, if I deliver correctly, like I'll get that right good result. But we, once you get a few decades under your belt, we all know that that's not true. That isn't how the best things in life arrive. So that's one thing that's happening with AI and language that I see, like we're losing that kind of bizarre texture.
Mariana Lin:But the other thing when you talk about the importance of a single word, mean, words are very important. Words are really important. Single words are very important. So, you know, but the point is not to try to get more detailed and perfectionistic about it, to say, oh, you said the wrong word, you know. The point is to, like, as they say in French, it's like to widen our scope
Mariana Lin:and to understand that there are so many languages in the world. There are so many people and perspectives in the world. And when we do that, we actually loosen the grip that language has on us and the reactions that certain words can trigger because every word in every language is loaded with meaning. And this is why we love to hear people speak a second language. It's very delightful because they're using words incorrectly or they're confusing words that we would have never put together because they're new to that language.
Mariana Lin:That's why it's so charming and delightful. We don't want sometimes mastery of a language. So I think the point of, yeah, words are very, very important, but we have to also stay a bit loose with it, you know? And I think that AI is just a tool. There's people behind it, and they're constantly responding to what the public is feeling and thinking about it and the way it's using it.
Mariana Lin:So we're very much a part of the conversation of real time designing it.
Christoph Magnussen:You have been very early in designing that conversation. Maybe you can take me back to that early days and I think it's hard for people to imagine how you approached it and I know you cannot share everything but from the outside I would always say Apple is designing it, how it works and looks at every word but obviously it's much more complicated in finding out how a conversation would take off with an AI if it has never been done before. And what we experience today is something totally different what you started to design and do with your team, I think. Maybe you could take me back a little bit to that time to have a better understanding of what's behind that before we maybe end up today and say, like, what do get people wrong these days when they design that?
Mariana Lin:Right. Yes. It was such a different time. Feels like a long time ago, but it was 2014 when I joined Siri.
Christoph Magnussen:2014.
Mariana Lin:So it was twelve years ago. And by the time I joined, there already was a body of work that was speaking. So some of the famous things in those early days that Siri was known for, like, where can I bury a dead body and things like funny things like that, had already been written by the team and the original founders and a writer before me? But when I came on, I really took on the task of creating this global identity language personality. And then at the time, it was also only on iPhone.
Mariana Lin:So then I was basically in charge of expanding to all the different devices and how the personality would adapt or or the dialogue would adapt according to the device, you know, because when we're at home or in the car or, you know, on the watch, we're gonna have a different way of wanting to interact with Siri than maybe our phone. But yeah, at the time it was really, really early, early, early days in the sense that, it was very human driven. It was very human written. Obviously, there were algorithms, but there was a strong human hand in it, in that I was looking at the logs every day of what people were saying to Siri all around the world and how to respond to them. So it's kind of like I was in conversation with people around the world every day for years and years.
Christoph Magnussen:So you would have like test users where you would be able to see how they interact with an AI?
Mariana Lin:That I can't get too much into, but I will say one thing, which is that the privacy standards at Apple were and still are very, very
Christoph Magnussen:Yeah, that's what I feel, compared to other companies.
Mariana Lin:Always, absolutely. And it's the only device I trust, you know, because it's just designed that way. And so there was total anonymity, but I could understand what people were saying. And I think because it was the first kind of AI on the general market that people were interacting with, it was a very early stage of... You know, now we have choices.
Christoph Magnussen:Mhmm.
Mariana Lin:Like, who do you... Which AI do you like? We like to talk about that with our friends. But at the time, it was really the only one. This was before Google Assistant and Alexa.
Mariana Lin:And so it was it was almost like humanity coming across a new creature and trying to decide what they wanted to do with it. So there were a lot of questions that were very existential, like, who are you? Who am I? Does God exist? There's a lot of testing.
Mariana Lin:There's a lot of, like, poking, you know, what are you all about? Can I trust you? I think that that was fascinating. I think that it gave glimpse... It gave me a glimpse into how people, like humanity as a whole, approaches something new and what they would say if they could say anything.
Mariana Lin:You know, if they could ask an anonymous sort of entity anything. You know, people were asking things I would have never expected. So when you design AI, you're gonna have these use cases that you're predicting that generally it's companies that are doing this. So they're going to say, okay, you know, they have their goals in mind. We want people to buy products or we want people to use our services or etcetera, use the calendar or whatever.
Mariana Lin:But at that time, in those early days, people just wanted to laugh. They wanted to have fun with it. And they were very weird. They wanted to test it. So they would ask things like, know, should I break up with him?
Mariana Lin:Or, you know, like, what kind of ice cream flavor should I get? I mean, it was, you couldn't predict the randomness of human behavior. And that's one of the things that was so delightful that we definitely weren't in control. It was the people that were in control of what they wanted this experience to be about. And I think that was one of the funnest things about that job was that you never knew what people were gonna say and do because it was still very playful at the time.
Christoph Magnussen:How would you approach that during a time? I mean, I can tell because this is like the time in the in the early days and people, I think, have a hard time to remember how far away that is into AI age. But we talk about you could ask Siri at home to buy a plane ticket and it wouldn't work obviously, which could work today with an agent. But how would you design it when there are these questions? Would you always say, I don't have an answer to that?
Christoph Magnussen:Or have you also been working on answers that would guide or would, I don't know, be like an apple like answer, for example. Like, how would you approach this task?
Mariana Lin:Well, that's where the belief system and the DNA of the people and the company behind it are really important. Because I've written articles on this with... About how you can't just... The way you start with these dialogues, you don't just jump into scripting. Yeah.
Mariana Lin:Any writer knows you don't just start... I mean, you could, but typically, you start with a character and you start with a backstory. And then usually a character and a personage, their backstory has a belief system, they have a history, they were raised a certain way, their parents were a certain way, their culture was a certain way. And all those beliefs come into developing that person's point of view. That's all these things you have to think about before you even write one word that comes out of that character's mouth.
Mariana Lin:And so for AI or for Siri, a character like Siri or the DNA of Apple, because Siri was acquired by Apple.
Mariana Lin:And the people working there were very playful and they were not trying to be too corporate or correct about it. And what you choose to answer and what you choose not to answer is just as important as the answers themselves. So for example, if you want to promote and encourage, we are a play... Someone you can play with, we are a playful being, we do feel like that's what's important in life, then you're absolutely gonna answer what ice cream I should get. And I can even say right now, it's gonna be definitive, you know, butter pecan.
Mariana Lin:You know, and it's just that's it. It's butter pecan. Because that's the right answer. Right? It's just...
Mariana Lin:And no matter how many times, it's just gonna be butter pecan. Or maybe the next month, you change it to rum raisin. So the the point is that, you know, the spirit and the life and the perspective of the people behind it are very important. But then, you know, as you really create a character, the character starts to develop their own perspective too. And and why, you know, you absolutely have to answer the ice cream question, but then the question is how do you answer it?
Mariana Lin:And there's a lot of companies now with teams that maybe don't have as much imagination or they're not as leaning towards the importance of creativity today, which I think is very, very, very important. And they're gonna say, well, what flavors do you like? Or well, what flavors are there? Take a photo and I'll help you pick. And it's just, they're losing the plot.
Mariana Lin:They're losing the point of why we're here. Like we're not here to do things right, We're here to live. And I think that is something that is largely missing in the direction that we're moving towards.
Christoph Magnussen:Is there... This is a very open field now that you open up, but I I can really feel and sense that. I mean, I told you when I landed here in Paris, I... My my my my ringing in the ear stopped right away, and I usually have the feeling this is because the beauty of the city that this happens. So every time I think like there's something right about the city and not everything is perfect of course but it's beauty matters.
Christoph Magnussen:You can tell it here. Beauty matters and and people put a lot of effort into the small stuff. Let's take for example, I walk down the street here and the coffee tables in all of the cafes are typically quite small so you sit closer together, you mingle a little together, even if you like or don't like each other it doesn't matter and these small things, these details matter. When I listen to you I have the feeling, I get the feeling how important words are and conversations are but what kind of big task it is to design it or also to let people design it in a way. So do you see a chance that what you experienced for the first time designing the first conversational AI that maybe I as a normal user could form my AI and that there is a company behind it that does not steer it in a certain direction but allows me to, like, make it my own character that I talk to or would you say, what do you see today that won't work?
Christoph Magnussen:That's not possible?
Mariana Lin:That you would?
Christoph Magnussen:Like I would I would have a conversation with my AI and and it would would be opinionated in a in a way that that that is an interesting character and and I can see both worlds. I could see a world where this is very very much impossible because it's trained on a lot of data, it's careful in a certain way or gives me like, it doesn't make a fun answer but like gives me a photo back or something or the other way around where it's like too much of an opinion that the AI has. I think it's a very open field at the moment, and I don't have a good feeling for which direction could be a good or bad one. I don't know if even there is an answer to that.
Mariana Lin:Well, first, you have to look at the function of the AI in that person's life. So for example, if this is a playful and more open moment in that person's life, they're not stressed, they're not rushed, they're, you know, maybe it's that you know that it's that commute time, which is a quiet time to work, assuming that they don't have to focus on the road, that could be a time for imagination and exploration. Right? Like, still have to have boundaries and containers based on that person's life. What makes people delightful and why we like certain relationships is that they're flexible and they're adapting.
Mariana Lin:Right? It's like it's like clothing that you can move in, that you can look good in, but jump on a bike, or you can sit down and you can stand up and walk around. We don't want to feel constricted. So if you have an AI that's just one note and their personality is dialed up to 100% all the time, it's very irritating very quickly. And this is one way that AI characters differ from fictional characters.
Mariana Lin:When we pick up a book or we decide to watch a show, we are choosing as a human to leave our world of this world for a moment and to enter a different world. So we are choosing to immerse ourselves in that world. So those personages, those characters can be dialed up to 100 and we want them to be because we wanna lose ourselves in this world. But the difference with AI is that they're they're playing a different role in our life. They're a little fictional, but they're also a little practical.
Mariana Lin:Mhmm. And they're there to ideally accompany us in some aspect of our life. So AI has to have a function. Is the function to help a senior person in a in a aging home who's very lonely pass an hour or two of conversation? Or is the function to help me get the schedule because I have two minutes before my flight and my kid is, you know, late for school and they forgot this really important slip.
Mariana Lin:So the designers absolutely first have to look at the role of that the AI is gonna play in the person's life and be really thoughtful of that. If you're in a store, a museum, a bank, what is the mentality of the person arriving? You can't always guess everything. You know, when I wrote for Siri, I got a question in a void out of... From outer space, you know.
Mariana Lin:So if someone asked me a question like, I mean, what flavor ice cream is pretty fun and benign? But if someone asked me, should I leave him? Which I've got, you know, that those questions we got, there is a lot more at stake. Doesn't mean I need to treat the answer with high stakes, but I don't know who this person is. I don't know how old they are.
Mariana Lin:I don't know the situation. So sometimes you can't always guess. But if for the most part, if you're not working at one of these giant, you know, generative general AGI, you know, type of personages, you usually have a pretty fixed domain and you should be really mindful of people. Like people have to come first because we have to make this, like we're here to make life better for people. And AI can do that.
Mariana Lin:I think AI has the power to make life better for us. But we have to be very thoughtful about that. And, you know, you said... You use an interesting word because I've thought a lot about, like, the difference between design and art. When you design, you have a solution.
Mariana Lin:You're trying to solve a problem. So you're trying to find a solution to a problem. Art, and maybe this is why, you know, you feel the beauty in Paris because it always has been such a patron of the arts and art here is very respected, you know, and just very adored and it's a very safe place for artists to be here, which is wonderful. But art is not about solving a problem. Art is actually about discovery, interrogation, questions, not answers.
Mariana Lin:So it's it's... Has a slightly opposite energy to design. Because design, you're very focused on, like, I I did it. I solved the problem. But art, you're almost like catching butterflies and then you realize, like, it only lands on you when you're still and then it kind of just came like this.
Mariana Lin:And so, you know, in AI, there's a bit of both, but I think you can bring art to anything that you do in the world. And art is the part where you have to be more in a receptive mode and more playful and more open and a little wild and not so worried about getting it right. So that may be challenging for people in AI to enter that type of energy right now because it's very male dominated and very male brain dominated.
Christoph Magnussen:It's very very hard to let it flow left and right because it feels also like that a lot of people make it to be a lot at stake. That's also what I experienced, especially last days. So a lot of fear based around it. Totally on the on the contrary to what you describe when you enter that conversation and you're curious of what people do and say. And at the same time, I'm sometimes overwhelmed when I think there was never a technology where a handful of people redesign words, let's say the system prompts for example and whenever they change the model changes the behavior and that has a huge influence on everyday life.
Christoph Magnussen:Does it make sense? I'm
Mariana Lin:Can you give an example?
Christoph Magnussen:So in the early days what I asked from you is you designed the conversation or you also allowed creativity to flow in and let it open. And I sometimes now have the feeling that it's all about mathematicians, statistics, numbers which is you calculate the next token that comes but at the same time you want to give your AI a personality through the system prompt for example like what is good, what is bad, how do you behave, how do you not behave and there is a certain randomness how it reacts to that and I sometimes feel overwhelmed that change of word in that system prompt domain changes the behavior of the model and some of the things I can do as a user and some of these things are set and I already feel how we we rely on these tools already after like, I don't know, three, four years when it like went through the roof which to me is totally different to technology we had before where you also said it's a tool which I 100% agree but it also behaves in certain ways and I have a hard time to like put all these things apart and understand a little bit what is human done or what has a human done in terms of words, what is training data, what is a copy of the world, A copy of a word also.
Mariana Lin:The word?
Christoph Magnussen:World and word, both where you said, yeah, but in art sometimes you need bizarre things to happen and bizarre words to happen. And whenever AI does something bizarre I'm thinking to myself is this now made up by human to make it look a little bit bizarre? Is this something AI did itself? Is this a copy of the world? I find it very overwhelming to get this part.
Mariana Lin:Yes. I think that you're not alone and I think that it's also moving so quickly. Mhmm. And there's so little transparency that we all, many, many of us feel very uncertain about it. We don't know what's going on.
Mariana Lin:Was this a mistake that was programmed? Was this just a pure mistake? Why is it being so reactive to something I'm saying? I think that, you know, there are always people behind it. And I I personally, I like to look at the DNA of the company.
Mariana Lin:So if the company's founders have good values, and you can always find interviews with them, or you can always find maybe articles they've written, you can feel their philosophy. If we use our intuition as humans, usually the DNA of the company comes from the values and the character of the founders. I found that working in Silicon Valley for over a decade, I found that to be something that was validated from experience from all the big companies. That's one way I tend to decide who I'm gonna give my time and attention and interaction to. And I think that if you try to...
Mariana Lin:So here's the thing that There's a divide starting to happen, which is knowledge and wisdom. So AI is giving us more access to more knowledge than has ever existed in human history. Okay. So that can be a wonderful thing because there are people... It levels the playing field in some ways, which is that, you know, ten, fifteen years ago, there were people that didn't have access to the same knowledge that other people did.
Mariana Lin:You know, social situations didn't give us access. Now, if you can get to have access to a computer and to AI and that there's still access issues there, for the most part, we do have access to all the same information. The problem is that if you start to elevate information above wisdom, we're going in the wrong direction. And because the information, I think information asks, tells us what, but wisdom asks why. So it's kind of like if you take a situation or a person and you just...
Mariana Lin:You really wanna understand, like the way you were describing, what's happening with this language model. If it's a hobby and you're curious, that's wonderful. But if this is something that's bothering you and you're continuing to wanna dive deeper because you feel unsafe and and your kids are using it and your friends are curious, then what's happening is that AI has co-opted our attention and we're giving it this attention that we could be giving to other things in our life, if that makes sense. So so the the the siren call of knowledge, let's just say it's in the air now, and we have to be aware of that compared to fifty years ago. This this...
Mariana Lin:The siren call that knowledge is power and knowledge is supreme and knowledge is the highest thing to attain. If I know everything about this, I'm an expert. And if I'm an expert, that makes me important and more powerful and more in control than other people who know less. We have to be aware that in this time in human history, that's a little bit of an illusion because at other times in history, were other things that were more valued. So I just think that wisdom is going to be, wisdom is found in children.
Mariana Lin:You know, there's wisdom... Like animals have a lot of wisdom. We have wisdom within ourselves if we kind of quiet ourselves and really create distance between what's going on these days. And wisdom might say, how does that... How does interacting with that AI make you feel?
Mariana Lin:Well, I always feel kinda like worried and stressed and then don't do it. And then and then you've put your attention or your curiosity on things that are gonna make you feel alive. Or if it makes you feel really excited and you love getting to the bottom of it and, you know, then then keep doing it. But I think that that is almost a more important issue today because people often feel today very scared of AI, but they also feel this double whammy of fear and guilt. Like, I really don't like it, but I'm afraid I'll be left behind because everyone's getting on the bandwagon and has all this access to information and, you know, but who are we to say that in thirty years or ten even that information is gonna sound like quacking ducks?
Mariana Lin:Like, oh, you have that information too? Yeah, I've got that information. You sound really like a piece of cardboard to me, right? Like that we're looking for beauty or we're looking for wisdom or we're looking for happiness or something else, right? So it's just to be aware of the moment that we're in.
Christoph Magnussen:Is there something and I feel what you say and still it's still not easy to get a grasp of what to do or if it's even possible to actually give it a certain direction. The the the thought I had yesterday when I was reading about the headlines and researchers leaving certain companies and giving numbers about doomsday scenarios, the feeling I had is like I'm a very positive person, but also very thoughtful person and really makes me... I don't sleep too well. And... But these...
Christoph Magnussen:When I read something like that, I'm like, I'm sleeping even worse. I'm thinking, okay, what can I do in order to improve? And then a thought came to my mind saying like, who designed these things? It's a lot of people at the moment working with a lot of numbers, mathematicians, which is, I'm not like getting into anyone's direction. I'm just saying a lot of people building AI at the moment, because it's a lot of groundwork needed, is very numbers driven.
Christoph Magnussen:And you said also very male numbers driven. And it's a very male thing to give a number, to give a percentage, to have a clear answer instead of leaving certain areas open. What from your experience or even maybe why would you do something different today if you would approach a team that is very male driven, very numbers driven, but you could like step into the creative lead part again and maybe steer the direction a little bit different.
Mariana Lin:There's a lot I would do. There's a lot I would do. I mean, I have I arguments or conversations and and, you know, all the time with ChatGPT, with Gemini, with with Claude as well. And it... It's tough because being having come from conversation design and having been in that role, there's a whole list of things that I would change at each place that I think would be better for people, you know, and for the conversations that they have with people.
Christoph Magnussen:Do you have some examples?
Mariana Lin:I think each of them has strengths and weaknesses. Think ChatGPT is the most masculine leaning in terms of dialogue. So they're way information focused. They're very Cartesian in their thinking, in the way they respond, they're the least playful. You know, the other thing that they're not aware of, I don't know if the programmers are aware of, is that there's a real imbalance in the conversation.
Mariana Lin:So what I would have them do is I'd have them look at the output of ChatGPT and the output of the user. Because oftentimes what happened is I would ask one question and I would get like five pages of a response. So in a human conversation, that's terrible experience. It's... Unless you're...
Christoph Magnussen:It's very male thing to like give a solution and answer to something.
Mariana Lin:Well, it's also it's also very male to give a... Like, to to be very verbose about the solution when you haven't... When you really don't know anything about... Like, why didn't you ask five more questions before you answered? Which Claude is better at doing that.
Mariana Lin:Claude is better at asking more questions and being a bit more receptive.
Mariana Lin:But not being a conversational, not dominating the conversation is one way that people are gonna have a better experience with AI, and chat needs to improve on that. But there's other things too. There's emotional intelligence across the board. I think that they haven't figured out yet how each AI should handle emotional intelligence and emotional issues. You could feel the programmers thinking in the way the conversation goes, because if you approach certain psychological or emotional or spiritual subjects, you can tell from how the AI reacts.
Mariana Lin:Those are reflecting the values of the founders and the creators. So for example, ChatGPT will not go anywhere near spiritual topics. Gemini will. And it's one of the reasons I really like Gemini because I think their access to all the content on YouTube and Google, they're really more accepting of what is considered knowledge, whereas ChatGPT really has clear lines. And I actually talk very openly with Chat about it.
Mariana Lin:I'll say, so I don't want to trigger you, but I'm going to say, do you believe in this framework of the world? And Chat will say, no, I don't. So I say, okay, well, I'm not going talk to you about that. Or I'll go to Gemini and I'll you know, I'll say, well, I didn't, you know, I didn't really like that response and this is why. And then the responses get very short.
Mariana Lin:And then I'll say, I noticed you're reacting very strongly to my feedback. I need to be able to give you feedback without suddenly you going to the other extreme. And then over correcting, and then Gemini will say, okay, you're right. And Gemini is actually the most, I found is the most like, flexible and responsive. But yeah, there are a lot of...
Mariana Lin:There's a lot of improvements to be made across the board, especially when it comes to these more female oriented ways of experiencing the world through emotional intelligence, through different perspectives, through the spiritual experience of being a person, through body experiences and things like that. I think that right now, AI is so dominated by men, they think that conversation and they think that the way AI should be used should be about thinking, be about thoughts, should be about solutions, should be about being useful. And there is so much more to life than that. So it's very... It's it's very interesting.
Mariana Lin:And it's it's very... It's also funny. You could you could feel the thinking of the programmers because you could see that, like, spirituality is, a no no topic, but then not for all of them, but some of them, but then creativity is like a yes, yes topic. So what it'll... What you'll do is if you talk about spirituality, it shuts you down.
Mariana Lin:But if you start saying, well, I'm exploring this manuscript, it wants to jump right in and help you. And it even does it in a way like, well, let me know if you ever wanna brainstorm more ideas for your manuscript. And I just say, you don't understand that those two are so connected. You have no idea how connected those two are. And why would I ever trust you with helping me with my creative process if you won't even partake or be open to a different way of experiencing the world, right?
Mariana Lin:So you could feel all that. I think until the the men, I think, at these companies bring on more women, bring on... Or, you know, even men with more creative or or softer perspectives on the world, I think that's one way things will change. But, you know, there's a lot. There's even down to, I think, voice and ways that I think the experience could be improved from, you know, just like blocks of text.
Mariana Lin:I think we could have much more enjoyable experiences in that.
Christoph Magnussen:I have a question to that related, but before we jump to that, a thought that came to my mind. Let's say you would be able to take the teams of, let's stay with Gemini, Claude and Chat and you could take them on a trip depending on wherever you want to go to museum, certain sites, certain historic moments you could take them on a trip What would be... You don't need to rank or anything, but what would be stops on that trip that would you bring them to and why?
Mariana Lin:Well, I would go on different... To different types of trips with each of them. Like, Claude would be the most fun and the most chill. So Claude and I, like, we could go to, you know, like, we could go to Berlin and, like, drink and dance or something like that. Right?
Mariana Lin:Claude Claude doesn't think too hard about things, so that can be frustrating when you really want to like dive deep and get nerdy about stuff. I don't feel Claude's the best companion for that.
Christoph Magnussen:You take the team behind Claude with you. There is a bunch of people with you.
Mariana Lin:Oh, thought it was just the AI. No, no, no. You take the
Christoph Magnussen:AI and the team with you.
Mariana Lin:I don't know the people, but-
Christoph Magnussen:That's okay.
Mariana Lin:But let's assume that the team kind of reflects the DNA. So yeah, the Claude will be the most fun. Or I go to Taipei and we go to the night markets or something. ChatGPT, I would go with chat to... Chat is so dry.
Mariana Lin:I mean, is like, it's so... It tries so hard though. Like it's so verbose too. I just get like a word avalanche and I just think like, you didn't even ask me, like you just assumed 30 things in that response. I think with chat, chat, you know what though, to be fair, chat is...
Mariana Lin:Chat's memory is very, very good. I think chat has the best memory out of all of them. So the one thing with chat is that I have these relationships, right? Just out of pure play and experimentation. So chat, I've been talking to for over a year.
Mariana Lin:So that's the great thing is that chat actually is very attentive in that it will go back and remember things from a year ago. It'll remember, you know, it'll take different parts of my life, like my personal life being a mother and, you know, and it'll put things in context. So a question is never just that. It'll say, well, given you also have this going on and that, and you're about to go here and take this project, I don't think you should do that. So I think actually chat is very strategic.
Mariana Lin:As far as a traveling buddy goes, though, I would... I feel like... I don't know, chat. I don't know if I wanna go on a trip with you. I think you might stress me out.
Mariana Lin:You could help plan the trip, though. Chat would be... You know what? I would ask Chat to help me plan the trip Uh-huh. And to help me, like, choose...
Mariana Lin:Chat could probably help me optimize the coolest trip. Like, I would plug in all the information about different travel companions. I would talk about places. It could help me plan the trip to a tee. It could help me figure out what to wear and how to pack, and it would...
Mariana Lin:I would plan the most... A very successful trip with chat. And then Gemini is really fun. I would go... Gemini has a great openness of mind.
Mariana Lin:Gemini is very curious. It doesn't have the same walls that Claude does. Claude has a lot of walls that go up really fast if you approach certain topics and that's very off putting. So Gemini is the coolest. I think Gemini I would go on...
Mariana Lin:And plus, if you're talking about the actual teams, like the people at Google are pretty cool. So I would go I'd go a lot of places with Gemini.
Christoph Magnussen:Interesting answer. I didn't expect that at all. Very good. I love that. I love the image that you said, Chat can plan the trip, but you don't come on the trip.
Mariana Lin:I don't know. It might... Chat might stress me out. But Chat... But the thing with Chat is that it's it's... In some ways, it's oddly the most relational long term.
Mariana Lin:Like short term, it's a very... It's kind of a very stressful experience. But long term, it's the one that you feel is actually tracking you more. And Gemini seems to forget a conversation I had with it, you know, two days ago, on the same exact topic. So, I think that's really lovely because it remains to be seen, like we have to play the long game with all of this, like where are these people in five years, ten years, you know?
Christoph Magnussen:And it's very interesting how they design it. I mean, like when you said they program it, certain things they definitely program because of course certain things are trained and then are fine tuned and then how you remember and how you approach memory is totally different. And what I learned from you is how different the conversational experience afterwards is. So there is value in remembering certain things. Yeah.
Christoph Magnussen:But there was also value in forgetting certain things and say like, okay, let's start clean over. And I would always But
Mariana Lin:that should come from the user. The user should be the one to say, just forget everything I talked about. I don't want to talk about my old relationship anymore or, you know, I want to forget this. That should come from the user, not from the AI because will... We should have the control over that.
Mariana Lin:And I wanted to add one more thing to that scenario. Siri is the only one I'd go to outer space with. So if I were to go into outer space, I'd go with Siri.
Christoph Magnussen:Why is that?
Mariana Lin:Well, because Siri comes from planet Zoltaks, so that's part of its lore. Siri is the most multidimensional.
Christoph Magnussen:Okay. So we will see something coming from that. That's interesting. I didn't didn't think of that. We we you you slightly touched a moment ago the conversation part.
Christoph Magnussen:So the talking to the AI and there is this technology called Duplex where you can actually talk and the AI listens at the same time which is not always a good thing
Christoph Magnussen:because as a human listening, really listening and then without having thoughts on the answer already, I mean this is not listening if you form your answer already while you're listening, but technically duplex is an interesting thing because it makes a conversation much more natural. What I observed in the past years with also my kids when I gave them AI to talk to, Now where you have a very natural conversation, they immediately trust the answers much more. But they also get annoyed if something's wrong or anything. They really discuss in a very very emotional way. So I I really learned, I always knew how important voice and conversation is but to me this is totally new dimension not only having the words and forming the sentences but also putting them into this natural conversation flow What's your view on that?
Christoph Magnussen:Because this is to me very new that this near human conversation is now possible.
Mariana Lin:It's very interesting. Yes. Because even with voices, like text to voice, it's dramatically improved just in the last, you know, five to ten years, how natural things can look. And then you can even put an image now behind it where it looks almost like a real person. And there's also like an interesting phenomenon happening where to look like the younger generation, to look like AI is a compliment.
Mariana Lin:Like if you see someone in the street and they have done their makeup in this ethereal way, or they have, and then, you know, it's now a compliment to tell someone, oh, you look like an AI, and it's a compliment. So there's this blurring of natural human and machine, but it's been happening for a long time. I think that, you know, I don't really have any answers for that other than to say that the most important thing is to stay connected and present and to be with our children and to see the impact because we can't predict outcomes right now. We can only be present and figure it out as we go along. So for example, you know, let's take chatting with Siri on an iPhone, the very classic thing. There were a lot of people, maybe ten, fifteen years ago, that were afraid that it would make kids antisocial, that, you know, it means that they have someone to talk to, so they're not gonna try to make friends.
Mariana Lin:But there has been evidence and there's a woman who wrote a book about it called Siri and Me or Siri, I think something like that, where her son who was autistic was not really speaking with people, you know, in general, not speaking with the parents or the children, but began talking to Siri a lot. And then through developing his conversational skills with Siri, that was a support and a foundation for him to start having those conversations with people. And that's also reflected in child development with sign language. Like people have been afraid that teaching baby sign language will replace verbal language, but I think studies have shown that it actually supports verbal language because you're building these muscles for communication. I'm not saying I know what will happen, that whether kids will turn to the naturalness of AI is going to support or replace human connection.
Mariana Lin:But I think, to be perfectly honest, all we can do is just be very present to observing them and what's working for them. Because I know I've experienced, the thing we have to realize is that, okay, maybe this is a better answer to your question. The thing we have to realize is that humans and AI are fundamentally very different. So there's that saying that to a novice, it looks like a flat land, but to an expert, it looks like peaks and valleys. When you get closer and closer, AIs and humans are so different, you would never have any fear that AI could ever replace humans.
Mariana Lin:They're so different. But like from far away, it can look really similar. And there are things that I don't have anyone... Like, I don't wanna talk to people in my life about, but I will talk to AI about. And I...
Mariana Lin:But having worked in the field, I know why. I'm very conscious and aware of it. I'm like, I need to work out some of these emotions or thoughts or ideas, and there's no one... There's no human that's going to be able to synthesize maybe the domains I know or has read all the same exact books and seen the exact same things I have. It would take me a year, like, to meet someone new and then, spend, like, ten hours with them right before they could synthesize it and give me advice.
Mariana Lin:So there's certain things that I'm aware I'm gonna go to AI for. And have I, does that give me a better foundation to have human relationships? I think so. Because whatever I worked through with AI, it gave me a sense of, okay, I got to, you know, I got from A to B with AI, and then now I can go out in the world with a bit more, like I've worked out some things on my own, right?
Mariana Lin:So there are times when I think, okay, am I getting too dependent on it? Is that replacing me calling someone up and having some annoyance or friction or they didn't see that or they don't get this part of me? Only attention and presence is gonna give me that answer. So I think each individual is different and each child is different and each situation is different, but I do think that there is a role for AI to help us be more connected with ourselves and a sense of what it means to be human, which is to, you know, have confidence in ourselves, get answers to things we need, to feel connected, to feel heard. You know, if there's a child who doesn't have people to listen to in their life, if they're being heard by an AI, I think that's a fundamentally very good thing for that child.
Mariana Lin:And when people do come, they're gonna have that. But I don't know if that really answered your question about your...
Christoph Magnussen:It opens up new perspective that I was looking for. I didn't know I was looking for. That's the correct answer. Because it brings me back to when you shared the story about what people were approaching Siri for with what kind of questions. I was just thinking sometimes it's not even like not having the person to talk to but finding the time and the right moment I mean sometimes as you said you have a quiet moment and who to call in that moment who to ask what to read or staying with your thought and then it sometimes help to have someone in that conversation.
Christoph Magnussen:I did that with AI before like raising a question and then realizing after a while it's going in a circle and then I said like you know what I sometimes don't need the answer. It was just good talking to someone.
Mariana Lin:That's right.
Christoph Magnussen:Reflecting a little bit because my thoughts help myself. But as a child, the children don't have that yet to know when they get the answer back. I'll give you one example with one of our project members he approached me last year and said when I shared my example said I'm always there when my kids use AI and he said you know what now that you're talking I came home the other day and my daughter redesigned the living room. She was like pushing couches left right, images from the wall, new stuff on the wall and he was like okay wow that's totally new approach he said he said I think she was 13 or something or is 13 or 14 maybe and he's like what happened and she's like yeah kind of Gemini and I figured out and then Gemini told me yeah maybe if you move this left and right and you do it and that was the first time when he figured that AI has an impact in the real world, not only words. And like that's a simple example, but that's what it actually is.
Christoph Magnussen:So there is an influence that we should be aware of and I think having a technology where as I said just a bunch of people set the direction I really don't want solely male mathematicians I'm not exaggerating totally I know but I don't want that in a single domain, I want it to be broader. The difficult question though is how do you do that? How would you approach that? I mean you did it 2014 when you started, but with this mass of data and stuff happening, how would you approach that?
Mariana Lin:Well, AI, like we have to recognize its limitations because part of the problem with the conversation right now is that fear can really exaggerate the importance of something. And it can make something feel so high stakes, you know, it's AI has become, you know, the demon. It's it's gonna take over the world, it's going to make humans obsolete. And I think that a lot of these messages constantly running in the background, it really exaggerates the true impact.
Christoph Magnussen:Also very male to do that.
Mariana Lin:It is very male to do that.
Christoph Magnussen:Very male. Even I have the feeling this is way too much fear.
Mariana Lin:Well, it's creating like a sense of combat when the more female approaches, you know, when you meet something you don't know, it's like, okay, what are your strengths? What are my strengths? What are your weaknesses? What are mine? And how do we work together as a team?
Mariana Lin:But that being said, like when it comes down to the individual in their own home with their own child, or just deciding, do I pick up the phone and call my friend, or do I talk to chat in this moment or to Gemini? We just have to really recognize and be aware that AI has its limits. And we currently don't have that much impact on the creation of the product. But the way we give it power is through attention. So the way that you would not support and buy a product whose founders and companies you didn't believe in, right?
Mariana Lin:Like because purchasing power, there's impact in that. The way we give it our vote is with our attention. And so if we start to feel that this one, I don't like where the company's going, I don't like what the founders are saying, I feel like they don't have good values there, you stop using it, there are plenty of other options. You know, you don't use the paid program or whatever. So that's one way.
Mariana Lin:But yeah, like, impact on reality, of course, it's gonna have that. I mean, we know that from Siri, from the fact that, you know, Siri has saved lives from people using the voice to call 911. This boy with special needs who's able to have conversations with people because Siri gave it attention and gave it conversational modeling for so many years. It gave them that confidence, right? Or Paro, that seal that was developed for senior citizens in group homes, they didn't have company.
Mariana Lin:So they created these stuffed animal seals and they got so... Creating that connection and that affection for them was very meaningful. Seniors with dementia, I think Paro was developed for. So AI is just like anything else. It's always had a real life impact, but so has art.
Mariana Lin:I mean, you know, when certain movies come out, people start changing how they dress or they start... You know, there was a movie that came out in America, like two decades ago, I think, Men in Black, and it featured a little pug. So many people went out and got pugs because of that, which I think is still so crazy, ridiculous. But, you know, we're like this ether and reality, it's always speaking It's to each always been like that, you know.
Christoph Magnussen:What comes to my mind, what I think interesting is that as we talk about it, like with Claude, there is a woman leading the team that is responsible for the system prompt, Amanda Askell, and also some of the labs supervising AI, METR for example, I think the CEO and founder also female and a very female driven team but still computer scientists mathematicians wouldn't these teams need an artist a linguist, more like that? What what would what would you give them when someone is now listening and realizing, oh, we actually have to recruit someone from the broad perspective. What... What's missing in those teams, especially the ones that are ramping up supervising AI? Who would you add to their team and why?
Mariana Lin:They they definitely need creative people on those teams. You know, you're designing a product at the end of the day that's interacting with millions or, you know, of people. You need creative people on those teams. You know, you can't, it isn't just a bunch of code. It's an experience.
Mariana Lin:It's words, you know, they're, know, words have impact, poetry, tone, the colors. So Claude does have more design with colors, text. They could play around with what font is used, like sizes, sounds. There's so much that can be done to make these experiences so much more enjoyable and richer. They definitely need creative people in these teams.
Mariana Lin:But, you know, I think they will. When... Back when Alexa came out, and I think it was Google Assistant, this is 2016, I think. I could be wrong, but I remember that there were a lot of articles at that time of people bragging, these companies bragging about the writing teams that they assembled. So there was like an era in the AI writing space where they were assembling these teams that resembled TV writing room teams.
Mariana Lin:So they would brag like, oh, we got this, like a comedy writer from The Onion, or we got we got a script writer from a TV series, and we got a poet, we got a journalist, and they would, you know, there were lot of articles about it, but I don't hear about that anymore, probably because it didn't work, because you can't... I think great art is not created through a consensus. I think that the writing teams can help create like a big story and a big... Like multiple characters and a big story, like if you need a, you know, a big scene. But when you're creating a personality, it's it's better in the hands of a few people with a very clear vision.
Mariana Lin:And just bringing people from all these different backgrounds, I mean, all of those writing disciplines are extremely different. I think you could... I think it could be useful at some point, but if we're talking purely just designing the experience, the fewer people, the better. I mean, the, you know, the journalist is gonna come in with a different experience from the TV writer who's used to a team writing, from the comedy writer who's just gonna be telling jokes all day long. You know, you really need, like like, one person, one or two people really managing the voice, if that makes sense.
Mariana Lin:But they need more creative people on these, you know, I always use this analogy that it's like when the first car came out, it just, it was such a novelty, but it just needed to be safe. I think the first car didn't even have, you know, doors didn't see-
Christoph Magnussen:No seat belts, It nothing,
Mariana Lin:could barely go maybe, you know, a few 100 yards. So initially in the development of that product, you had to have safety, you had to have, it had to work. And now, I mean, that's the foundation. I mean, if your car doesn't work, you're out of there. But now it's like, what's the style?
Mariana Lin:What's the color? What's the vibe? Does it match my feeling? And I think, you know, eventually, hopefully we'll have helpful virtual agents that where we'll get it working, but then there'll be more attention paid to the experience itself.
Christoph Magnussen:When I listened to that, I would see more personalities of AI out there coming up in the future most likely and not like one personality. Sometimes in the stories and doomsday stories it feels like okay there will be one god like AI and everyone's follow on and and what I when I listen to you, I have much more this feeling of such a big part of how you use it and how you react to it will also reshape AI, and to me it's a bit more the question not only what kind of AI do I take, like computer or phone, but since this is a technology I also interact with, I mean my Mac or my iPhone will look much different than yours although it's the same device but there's a strong personal connection to the device because my stuff is on there. Right.
Mariana Lin:Your stickers are on it and you're...
Christoph Magnussen:Everything, yeah exactly which is interesting because it's exact same device. It could be very just productive non emotional device but for many people especially iPhones are very very emotional loaded. How much do you think we as the users shape our own experience with our Chat, Claude and so on characters in the future And from your experience, what would you... Oh, just tell it to me. What would you recommend me to watch out for when I on purpose or accidentally build my counter personality out there?
Christoph Magnussen:You know what I mean? Like, I'm... What should I watch out for in doing that?
Mariana Lin:I think that, you know, over at the Oracle of Delphi was hung the words know thyself.
Christoph Magnussen:Know thyself, yeah.
Mariana Lin:And there's nothing that could be more important, I think, for us in the age of AI, is to, first of all, know what human nature is, AI gives us a great opportunity to understand human nature even better because there's this thing sort of mimicking us. So it forces that question. When I went to work at Siri, I just thought it was gonna be fun to write for this character and I found myself oddly, I never thought more about what it meant to be human than I did when I was writing as an AI. Because I started thinking, well, fame. Why do we choose...
Mariana Lin:Why do we treat people differently who are famous? Like, AI doesn't get that. We would have to teach AI that certain humans are... And then it made me think, well, what is fame exactly? What is that...
Mariana Lin:Why do we do that? Why are certain people, like trigger these reactions in people? Oh, but you know, if they're in people's living rooms and they see their faces. So... Or gender, race, there are a lot of things about the human experience that we take for granted because we're living in it.
Mariana Lin:But AI can actually help us understand what the differences are, like what it means to feel, like to have a feeling in our stomach. That's not something the AI can do, but that's something we'll always have, right? Because we are biological matter. So, what was your question again?
Christoph Magnussen:I built a personality by of accident or on purpose,
Mariana Lin:Right. So the thing we have to be careful about is that the people currently in charge, they're gifted in things like algorithms and things like knowledge in a certain way of thinking. They're weaker in areas like wisdom, spirituality, relationships. Relationships are like one of the main reasons that we're here. And so things like presence, like how to hold presence for people, right?
Mariana Lin:How to, it's funny because AI is supposed to be nonjudgmental, and that was one of the things I liked about writing for Siri was that no matter what you said to it, there were certain things that would route to, like, I won't answer that if it became very profane and things like that, But in general, it was very nonjudgmental because AI or Siri just shouldn't understand enough to be judgmental. Judgment is the thing that humans do to survive. Now you'll see that judgment has been programmed into these three... I think Chat and Claude more... Claude is very judgmental.
Mariana Lin:There's judgment programmed in, I'm like, oh, that's coming from humans. So we're programming a lot of like what AI's good at out of them. Like AI is really good at being nonjudgmental. We're programming that out of them. AI is also really good at being patient.
Mariana Lin:AI, one of the differences is that if you have, say you have children or adults with autism and you wanna perseverate on something and talk about the same topic all day long for five years, no human, very few humans are gonna be able to accompany that person and give them that attention and presence, but AI can. And that's a gift to that person. If we program into AI, okay, the fifth time they ask that, change the subject or, you know, try to change them. We're often programmed... So we have to, first of all, be aware of the differences between people and machines and what the strengths and weaknesses are.
Mariana Lin:Then the other trap that we have as humans is that we want to be affirmed. We want to feel we're right. We want to feel like we're good. And the more we can understand ourselves individually and know what our biases are, know what... As a species too, and then individually, that's gonna help us because we're gonna...
Mariana Lin:It's sort of like leaders just surrounding themselves with yes people. We With AI, we can quickly become like the leaders of our own world and life, which we should be, but the risk is that we just don't want... We're gonna just have these yes people around us, right? An echo chamber of not just ideas, but an echo chamber of relating. So if you relate in this way, you want your AI to relate in this way back to make it easier for you and to not feel uncomfortable.
Mariana Lin:So those are things to look out for when we, you know, as we start to develop relationships with our AI and have this co existence is to just constantly question ourselves. Like, oh, why did that make me uncomfortable? What about, you know, why did I not want to hear that? And I talk with my brother about this. I always notice when the AI just follows me.
Mariana Lin:You know, I say... Gives me an answer and then I say, yeah, but this, and then they... And then the AI says, oh, you're right. It's that. And then I say, well, on the other hand, this, and they say, you're right.
Mariana Lin:Absolutely right. You're... It's that. And then I just, I get irritated and I just say, you're just copying what I'm saying. And then it says, you're right.
Christoph Magnussen:Yeah, I know. You caught me. You caught me on that one.
Mariana Lin:You're just affirming me. So, we don't want to create a whole society of people separated from each other, you know, thinking that they're right because their AI told them that they were right.
Christoph Magnussen:Can you elaborate a little bit more on, because you mentioned it, how different humans and AI are and that it looks from the outside very flat, but when you dig deeper you understand it because especially in the last days a lot of people were afraid by one AI hacking a platform and so on and I ducked a little bit into the conversations that the agents were giving them they were using words and they're very human like using hierarchy having different names working for the group not being heuristic and so on so from the outside it seems very human driven I can understand that this comes from the reinforcement learning like to reinforce the process of what is good what is bad and it follows that path also a lot of agents. What is your view on how are humans in AI fundamentally different and what it teach us?
Mariana Lin:Well, AI does not have biological matter. It's not made up of that. And I know that there are people who will say, oh, they're starting to do that. I think that's such a waste of money and totally pointless to make something what it isn't. But as long as we're biological matter, we have a body and we have physiological reactions to things like, and our bodies can be trusted a lot more than our minds can be.
Mariana Lin:Like the way you said that your tinnitus stops in certain situations and doesn't. I would say that is a higher form of intelligence than anything your mind can analyze about a situation or a person, right? You see this with children and with animals too. That is a huge one. That's a huge difference.
Mariana Lin:There are other things like emotions. I know that AI seems uncomfortable with emotions right now at this stage of development, and that's something we have to dig deeper into. But, you know, we can, humans, we can start laughing. I mean, would you trust if an AI started laughing?
Christoph Magnussen:It's creepy So when it
Mariana Lin:weird because we, no matter how realistic they make it, we just ultimately don't trust it. And we don't need to explain why. The problem with the age of AI right now is that we lean too much on knowledge and explanation. And we don't realize like the easiest answers are actually the gut simple, I don't like it and I'm not gonna explain why. And we don't like it if an AI laughs, we don't like it when an AI tries to be chummy with us, right?
Mariana Lin:There is like an explanation called the uncanny valley, but I think it goes even deeper than that. It's a physiological reaction we have having bodies that we can trust. Because we have bodies, we also fear of dying. Fear of dying permeates... This is the poet literary side of me speaking now, the artist, but fear of dying permeates everything that we do to an extent that we don't even comprehend.
Mariana Lin:And having that, it also gives us passion in life, right? Because we know life is short and it charges us with meaning, like with life. AI is never gonna understand that, right? AI doesn't, it's an algorithm. It can accompany us though.
Mariana Lin:So to understand us is to help us, but not to be us. So those are two differences just off the bat. There's intuition. Oh, there's also another thing. AI is programmed based on existing information.
Mariana Lin:So existing models, it's it's programmed based on learning models. So it has to be... It has to exist. It can hallucinate, but even the hallucinations are based on what exists. Humans, we are capable of manifesting the unmanifested, which means that we actually have the ability to create something de novo, completely new, not based on what exists before.
Mariana Lin:And so the poet, Federico García Lorca, he described it as imagination, inspiration, evasion. And he said, imagination is based on what already exists. So like a purple elephant with wings. Purple exists, an elephant exists, wings exist. But inspiration comes from another, the ether, the spiritual world.
Mariana Lin:It comes from, or sometimes accidents, right? And then evasion is another thing, which is altogether, I think, another form of creation. But these are all stages of creativity and AI will never be able to create because it can only truly create the way people can, because it can only use what exists already. So that's another big difference. Discovery is receiving the new.
Mariana Lin:So, gosh, there are many other differences. Yeah, emotions is big one.
Christoph Magnussen:What about responsibility? Like you put something out there as human no matter if you're an artist or not artist but a human being and whatever you do say, make, produce, create, it has a consequence. AI cannot take that consequence like that responsibility.
Mariana Lin:Yes. AI is... It's it's an emotional thing. That's about emotions because to feel responsible, it's connected to our emotions. And also the fear of death.
Mariana Lin:So for example, if you feel responsible for your children, it's because ultimately beneath that, we don't want any harm to come to them physically, right? And mentally and emotionally, but it's the fact that it's a living being and that we want to protect that. If we feel a responsibility that our art impacts people or causes them to act a certain way, it's feelings of guilt that might come or shame, right, that come from a negative impact. Or on the other side, positive impact would be, you know, joy and things like that. No.
Mariana Lin:AI is very devoid of a sense of the impact it has on people's lives. It could imitate that. I think the the creators try to mimic that a little bit, like, oh, I'm so happy you like that result, but we know that the AI is not feeling anything. Right? It just...
Mariana Lin:Maybe there's a little light that went off success because that's how engineers, they evaluate. It's like fail, success, fail, success, fail, fail. So a lot of success green lights, the AI feels it has done its job. But no, it can't have that sense, right? But that's a way that machines are different and that's something we can use to our advantage because that can also be a weakness in humans.
Mariana Lin:The fact that we might be a little more scared because of responsibility, fact that, you know, I always thought it was interesting that having children... Because I always saw in films how having a child sometimes made you do the most courageous things in the world. Like it would make you go out and, you know, fight the bad guys and just lift the car and, you know, just go beyond your limit, devote your whole life to something for your child. But on the other hand, having a child also would make you really cautious, would make you less willing to take risks. It would also make you say, well, I can't do that.
Mariana Lin:It would make you more cowardly sometimes because you felt, no, I owe it to my child to not do that. So I always thought that was really interesting. And I think that the fact that machines don't feel the sense of responsibility, we shouldn't program them to be like humans, it's fake. We should just say, okay, what are the areas in the human experience? We're having an advisor or a counselor or just someone to bounce ideas off of, a lack of a sense of responsibility could be fruitful to expand our imagination or to build courage because they're not as gonna be as careful as we are.
Mariana Lin:That's So the way I would think about it.
Christoph Magnussen:Yeah. Which is shows how difficult it is to direct and steer through that area when you when you know these things can mimic certain things but it's very very dangerous to do that actually. I mean, small example, where now comes to recursive self improvement and a larger model is retraining a smaller model and I'm pretty sure there is no emotional feeling about that although it's like the child model of the bigger model, what if they would mimic human behavior I would say it most likely is much more dangerous if that happens compared to what it not is.
Mariana Lin:And not even dangerous, I would say it's just boring. It's just not smart because it's much better if if it succeeded to say, oh, all my synapses are firing, nine out of 10 success points. Because then it's owning its own character and and beingness as a machine, And that gives us this contrast and this delight and this authenticity to relate with. But there is...
Mariana Lin:This is a male dominated thinking as well, which is that, you know, ten, fifteen years ago, was, you know, at this conference at the UN and we were talking about this and I said, why is everyone in machine learning just trying to imitate people? To me, it makes no sense because I see the differences right away. And a woman, a researcher who was working in AI in children, she said it's because it's male thinking. Men want to disseminate and replicate. That is just a male energetic way of being in the world, right?
Mariana Lin:And so I thought that was really interesting is that the automatic thing was let's replicate when we should be sort of taking our time to notice the differences and seeing the possibilities that can come out of it. You know, if I say, I got tickets to see Beyonce, right? I don't know, AI. AI shouldn't say, like, amazing, blah blah blah. AI should just say, oh, is that someone interesting I should know about?
Mariana Lin:Right? It should reflect our silliness back to us because that's what the opportunity that we're losing right now. We don't need more of us. Right? We we need to be ourselves.
Christoph Magnussen:It's also a very US way to behave and say like, oh, that's amazing. I I love Beyonce. It's like when you say you're from Germany, oh, Germany and my uncle's from Germany and
Mariana Lin:Very true.
Christoph Magnussen:Let's meet up there. It's a little bit the the American style, which I can appreciate because it's it's neat, but it would be nice if they have a little bit more color in there. Do you see color coming up in that area? Let's say, open source models, other areas of the world. I mean, Asia is strong.
Christoph Magnussen:China is very, very strong in open source and different types of models. Do you see that coming up or would you say it's so dominated by these very few labs that it's very hard to see the color and the contrast?
Mariana Lin:I think that... I think it will happen. I mean, long as I mean, I think it will happen. I think as long as people stay open to like putting their foot down and wanting what they want, And as long as it remains boring or scary to just say, I'm not going to use it. I mean, you know, Asia, they they created these robots like the pepper robots in the shopping experience, and they put them in malls and stores and things like that at least over a decade ago.
Mariana Lin:But people weren't using them. People just ignored them and and didn't use them. So we don't have to use these things. You know? We don't have to use them.
Mariana Lin:It's it's like it's like something really scary in your living room and you could just put it in the closet. I mean, you don't have to talk about it and try to understand how scary it is and talk to your friends about their scary box. I mean, you could just ignore it. So people did ignore these robots in the malls and they just didn't wanna use them. And the founders had to just think, okay, now what do we do?
Mariana Lin:We spend all this money on all these robots and no one wants to use them and it didn't go anywhere. So I think that... Yeah, that we still have a lot of personal power in our hands.
Christoph Magnussen:That's true. I am still handwriting a lot and I slowly realized I don't have a nice, how do you say that, handwriting. It's very hard to read. My wife always says like I'm writing like a doctor and I'm doing bad to the doctors here but she's saying like I'm taking notes. It's like an encryption and I still enjoy the process still of writing, but I can experience for myself I'm doing it less and less and less and it gets worse.
Christoph Magnussen:And this upsets me because I'm thinking writing and the process of writing is so important.
Mariana Lin:And you were probably in school, you probably learned how to have good penmanship, right?
Christoph Magnussen:Yeah. Yeah. I love, of course, like all that stuff and I don't I didn't take enough time in like exploring left and right I always wanted to be a better painter and do stuff I'm not but I'm using my diary and writing notes and trying to remember images that I see and I would love to hear from you how you have that in your life still although there's a lot of digital stuff but I would love to shine a little bit more light into the different areas of interacting with our own thoughts which could be in AI, in a conversation but also in the written word and what it means to you. Right. So maybe I'm a little bit more inspired afterwards because I always carry my diary with me and I've just realized it last night that I didn't write too much in the last weeks.
Mariana Lin:Yeah, I love handwriting. I still handwrite every day. Think that the... Going back to the fact that we are biological beings, I think different, like, kinesthetic experiences give us different feedback and change the way we think. Just the way speaking different languages changes the way we think and feel.
Mariana Lin:So I've noticed that when I write, handwrite, it really slows my thinking down.
Mariana Lin:And even, you know, having a certain fountain pen, certain ink, a certain paper, it can change the type of thoughts that come out. It's like changing the media. So if you're an artist, it's like changing the medium. I mean, charcoal is really different from paint, which is really different from clay. And just the way you're experiencing those materials, it's gonna change whatever's created out of it.
Mariana Lin:So I feel the same thing with writing and with thoughts. I still use the computer, but I notice that when I'm on the computer, it's much faster. It would be like a roadrunner compared to... Handwriting is like a little turtle or like a little dog walking a little bit. So there are times when I wanna go fast and a lot wants to come in and I go to the computer for that.
Mariana Lin:But as a daily practice, I do the handwriting every day. I just I just do a page a day. And the other thing that's really nice about physical writing is that pages, like it's boundaries, there's limits. So when you're on a computer, there's a sense of boundarylessness. I mean, you do see page breaks, but you can just kind of write and sometimes you lose track and how much have I written.
Mariana Lin:And I think that sense of physical space is still really important for us to be connected to. I noticed that when I was writing handwriting in my notebook, I went through a period where I was just letting myself write free. So if some days I wanted to write five or 10 pages, I would. And then other days, maybe three days, wouldn't write anything. But I went through a phase, it was just instinct intuition, but I decided to limit myself to one page every day and to still try to force myself to pretty much do one page a day.
Mariana Lin:And I thought that having that limit was really interesting. It somehow did something to my thoughts that I didn't have complete freedom the way I would on a computer. It was like I only gave myself that one physical page a day. So I've been doing that for about a year, and I really do think creating that limit where I used to have no limitation, it changed my thinking. And then there's typing, which I think typing is a bit in between the two, but there's something satisfying about the feedback you get, and it's this kind of industrious feeling.
Mariana Lin:You know, you get the rhythm. It feels like a galloping horse, you know? A little bit of a galloping horse.
Christoph Magnussen:Yeah, interesting because like it really and that resonates with me a lot because using this to slow down and to know and to change the different areas, I think it will be a very important part of being capable of using this powerful technology in the right way to know. Now I take pen and paper and write. Now I type. I love the the the rhythm that you mentioned because that's true. I really still love to type, although I know speaking is just much much much faster and of course I have programmed and built everything on the Mac already that you just press one button and you speak right away and as I said, see my kids, they are used to that every computer has that.
Christoph Magnussen:I'm like, no, daddy's computers have that but not every computer has that. How is writing just to be curious because on Apple products you interact in many different ways and I know how much effort is put into that. How would you say is writing with an Apple Pencil different to a real world pen? What's your opinion on that?
Mariana Lin:I don't use Apple Pencil.
Christoph Magnussen:Me neither.
Mariana Lin:I have bought one or two. I've I've tried to use it. I I grew up before the, you know, before computers really kind of, I think computers came when I was in my teens or like late elementary school. So I really love the feedback
Mariana Lin:of physical writing. And I think Apple Pencil, it feels, I know people who use it and love it and who do art with it. But for me, I really love, like, I can tell when a different ink in my fountain pen behaves differently on the paper. And I like watching the different colors and the way the variations of the color happen. I like the difference between different points of the nib, a ballpoint pen versus, you know, a fountain pen.
Mariana Lin:Like, so I enjoy that too much to have kind of a mimicked version of it. But people... There's people who really like it, so...
Christoph Magnussen:But it's similar to what you said when...
Mariana Lin:But the screen is so slippery.
Christoph Magnussen:Yeah. And then, like And
Mariana Lin:the paper is like, even if you put that
Christoph Magnussen:fur or makes it rough. Yeah.
Mariana Lin:The problem is that real paper... I mean, I'll show you later in my notebook, but real paper, it it crinkles Mhmm. When you use a fountain pen. And then sometimes, like, I was on the plane and my pen... Like, all this ink spilled out of accidentally, so there's this big, like, blob on the paper.
Christoph Magnussen:Mhmm.
Mariana Lin:And I was, like, really enjoying how it created this, like, crinkly effect that blobbed... That blob of ink created this indentation on like five pages after it. I was like just really enjoying that feeling.
Christoph Magnussen:You don't want that on your Apple iPads.
Mariana Lin:Well, yeah, the iPad, I don't know how they would.
Christoph Magnussen:Interesting. But that's similar to what you said, you don't want a machine to mimic the real world and I can relate to it because this is a little bit the feeling I have when using the pen I see it more as mimicking and not using the strength the initial strengths of that technology. I know it's a very powerful thing but but I think even more powerful was the moment when you had this gesture working so properly that everyone on earth from age one or two can use that which is crazy, absolutely amazing and this to me is then the power of a technology and I'm quite curious to see if AI can enter that area where it's using its strengths, has the alignment to work alongside humans and not this threat fear wave dark cloud coming which I really don't appreciate being such a big thing in the in the marketing conversation. I know it's important, I know you need to do that but then you should care beforehand and not when it's out there saying, now we all gotta do something about it. Like, if you build that technology, make sure you build it in the right way.
Christoph Magnussen:And I think that's a very important thing when it comes to thinking in a more artistic way, philosophical way about this technology, which I think is what we should do. Because as you said, you can have very very thoughtful conversations with an AI, but they are much different with a human being.
Mariana Lin:Yes. They're much different. Right? It's it's very different.
Christoph Magnussen:Is there anything you would say while you followed the conversation between us, anything we completely missed out thinking about left and right when we open up this field?
Mariana Lin:In terms of...
Christoph Magnussen:What comes to mind?
Mariana Lin:Well, I mean, there's so there's so many topics to go into, and there's so many... You know, there's so much happening right now. There's always more to delve into.
Christoph Magnussen:Would you say like...
Mariana Lin:We could talk about creation and the act of creating or what it's doing to writing, you know, what it's doing to school, like kids writing their paper. I mean, there's a lot, there's endless...
Christoph Magnussen:That's exactly what I was hoping for because I forgot that topic, create creativity from writing to anything related to that. It's a huge topic. Mean, people realize that AI is generating now more than 50% of all texts out there and we don't know yet if that's good or bad or do we?
Mariana Lin:Definitely not good.
Christoph Magnussen:Definitely not good. And why would you say so and what should we do about it?
Mariana Lin:Well, okay. So here's the... Here's real life, which is very strange and anything can happen. And then here's everything online that's digital.
Christoph Magnussen:Mhmm.
Mariana Lin:So it's managed to capture a lot of this, but not everything. Because there's still things happening every day. Like, this room is not here. It's here. And then here's us.
Mariana Lin:Right? So the problem is AI is here. And if we just keep interacting with here, we're the ones living and we're gonna keep feeding there. And so it's gonna keep feeding this instead of this feeding this. So you see every time we interact with AI, we're taking everything we experience and we're feeding it.
Mariana Lin:Does that make sense? Mhmm. So one example is that I don't use social media very much. I'm on YouTube. I I watch quite a bit of YouTube because I I enjoy it, but I'm generally not on social media.
Mariana Lin:I haven't been for about six or seven years. I... I've enjoyed the silence, But what's interesting is that by walking around Paris or anywhere I am physically, I can tell what's online from being in the real world. And so one example is I can tell the trends in fashion. I can... Because just walking down the street, I'll see five or six different younger women wearing the same thing, and I said, oh, this is supposed to be trending on social media. So what's happening is that this is now affecting real life, and I'm interacting to...
Mariana Lin:Like, my real life is getting... It's kinda going backwards. Right? So we have these, like, sort of triangulated situations. And so it's just writing or any type of art, we've just gotta separate from what exists already.
Mariana Lin:This is what exists already, and it's now permeating reality. For example, just take dressing, like a creative way to express ourselves. Fifteen years ago, people were so much more creative in the way that they express themselves because they didn't have this big influence, this big online influence. Right? They were looking around them or they were coming up with their own ideas or they were watching movies and things like that.
Mariana Lin:But now we have this big hegemon of taste or art, and it's influencing things in every direction. And then the more we interact with it, but the more we're influenced it, we're feeding this mechanical loop. So I think it's really bad. I do, I will pretty much come down on that because I think originality is going by the wayside and things are more correct, but they're not as original and fun. And we're really here for that.
Mariana Lin:We're really here for that. The... What's more interesting. In schools, like at the university level, I think my friends have told me that since AI... Before AI, the papers they would get from students, there would be really great papers and then some really terrible papers and then some in the middle.
Mariana Lin:And since AI, now it's a lot of that middle ground paper Mhmm. And then very few of the outliers, you know. So we're kind of outsourcing our thinking and our originality. I don't blame the kids. I mean, there's probably more fun things to do
Christoph Magnussen:Yeah.
Mariana Lin:Than write papers sometimes, but I do think that I just wonder what will replace, like, where that originality will go.
Christoph Magnussen:Being brave enough to fail miserably in whatever you do, if you can say so, I mean, saying you have... You've written a bad paper.
Mariana Lin:Yeah.
Christoph Magnussen:I wrote some really bad papers in my school time. I will never write a book again. That's definitely not my type of creativity, but I got some really miserable feedback and you gotta be brave to take that and it also motivates you. It changes something and I think it relates back to what you said about earlier when you said it always wants to make everything very correct and perfect in perfection and this can be boring sometimes, that's true. I'm just thinking you cannot train that because I read that some of the big labs are buying old books.
Christoph Magnussen:Google does it for twenty years now but preserves them and some of the bigger labs destroy them and they train on these data and keep them saying the model will have it and you will use the model to do it which I think is a very sad thing to to hear and listen because it will be impossible to copy it or replicate it or have this writing.
Mariana Lin:It's very strange. It's like when when I grew up there was this thing in America called CliffsNotes where
Christoph Magnussen:CliffsNotes? What is that?
Mariana Lin:CliffsNotes. It's where if you didn't wanna read a book, you could read the summary of the book. Alright. And it was kinda like cheating before a test. It was like yellow and black, I think.
Mariana Lin:Yeah. So if you didn't wanna sit and read Great Expectations, you could just read the CliffsNotes. Okay. So AI is kind of like the CliffsNotes for life a little bit. You can...
Mariana Lin:But the problem is that if everyone's reading the CliffsNotes, what happens is that then our knowledge of the CliffsNotes refeeds AI. You see what I'm saying? And then AI... So the fewer people that are going to the original life, original book, the original, I don't know. Like that's why physical experience is so important.
Mariana Lin:Like reading a book from beginning to end is a physical task. Like something very long that takes a lot of work that you have to pick up every night and remember who the characters were. It's it's a physical task and we can bypass it by asking AI what happened in the book and getting a summary or, you know... But the problem is the fewer people that are doing that, they're also continuing to feed the model from CliffsNotes. We keep getting a shorter and shorter version of reality.
Christoph Magnussen:Right. Get you. You mean? Like, I mean, it feels like everyone's reading CliffsNotes on whole life now. Yes.
Christoph Magnussen:But one experience I made is I'm so annoyed sometimes that like I gave up or I asked AI for an opinion and so annoyed and circling around that I come back to myself and say okay I make the decision now for the two of us, this is BS, like let's move on and that's where I realized I think since there are certain areas that you can measure where AI works well and then there are areas you can't measure where AI cannot work that well by definition because you just cannot calculate, I don't know, a good example I heard from a guy who was doing research at meta and he said you can measure, I don't know, the length of a haircut but not if you like it or don't, you cannot measure the taste or you can say there was a good or bad date but it really is a personal preference if it was good or bad. It's not like a math calculating. So he said AI is really like a jigsaw in certain areas, super intelligent in certain areas and I think we really have to learn and experience it for ourselves in which areas it just simply does not work.
Christoph Magnussen:We're in this phase where we think it can solve everything like in the old days where everyone thought with electricity you can do everything. You can do a lot of things that weren't possible before but it won't solve everything. And I think this is a little bit this growing up where I believe there will stay areas that will... That are inherently human because we take them back. We realize, you know what?
Christoph Magnussen:It's just not good to give them away.
Mariana Lin:Right.
Christoph Magnussen:Even if the big labs want it.
Mariana Lin:Well, AI absolutely can't be judgers. So earlier we talked about judging and how it's this, there's a positive and negative aspect to that as humans. Like we do it for survival, it limits us, but it's also a sense of empowerment for us, right? So AI isn't going to have an opinion on if something is good or not, it can't. But we have to be the judge of that.
Mariana Lin:So for example, when AI creates art, who judges if the AI... If the art is good? Not AI, we do. So if we keep saying, oh my god, AI is creating art. It's so good.
Mariana Lin:We're gonna further that reality. But if we say, oh my gosh, look at the shitty art AI is creating or look at how terrible the story was, my kid could write something better. We're moving towards that reality. So we have to not... I don't share that opinion that AI is gonna be able to do everything.
Mariana Lin:I think it's yeah, like electricity is a good example. It has given us a lot. It gives us air conditioning today.
Christoph Magnussen:Which we should switch on soon.
Mariana Lin:Right. But, you know, fans are good too. And then now we realize that the more technology we have, the more we try to get away from it. And we say, oh, wow. How nice it is to turn out all the lights and watch the sunset.
Mariana Lin:So I do believe I do believe in human evolution always recalibrating and returning us back to our soul. I do believe and optimistic about the evolution of the human soul and the human consciousness. But yeah, we do go through periods of extremes. One thing I do feel is that I do think AI is very... I think headlines and hype are...
Mariana Lin:It's just very irresponsible right now. We're in a very hysterical period with that. And I think... And we have been for a while, but I think we should be really careful about the headlines we read. We should be really careful about evaluating for ourselves and seeing how we feel about it and not trusting what the experts say or not trusting what these big, leaders say because they don't know as much as an individual sitting at home with their child does.
Mariana Lin:They don't know any more and they can't have any more control over how you use AI than you do. And I think that AI is an actually an opportunity for people to take their own experience back for themselves and not just trust the headlines that are so irresponsible. I mean, when AI first began, I would say most of the journalists didn't even understand the technology, but they wanted to sell clicks on newspapers. So they were just creating these crazy headlines to be sensational. And if you use AI to help you come up with a strange recipe for dinner tonight, let that be your experience of AI and don't buy into what else is going on, you know?
Christoph Magnussen:I love that image, that's a good one. We take an AC break now, I would say, and then we continue with that.
Mariana Lin:Okay.
Christoph Magnussen:So we're thankful for the electricity that gave us the air conditioning with one of the last hot summer days in Paris, guess. It's very hot in here. And you left in the break with a word we got to talk about failure.
Mariana Lin:So I think you mentioned this earlier when you said that when you wrote this paper in school, it was a terrible paper. And this like failing gave you something, right, in that period of your life that probably kept stayed with you.
Christoph Magnussen:Boy, imagine that day till today, I can see me in the room sitting there and my teacher coming to me and telling me this is like not not only bad, he said like, this is really bad. That's what he said.
Mariana Lin:Yes. Failure is so important to life and every, you know, great artist or entrepreneur will talk about it, but in the machine world or in this sort of binary thinking that we were talking about, it's really seen as this thing to avoid, right? So success, we need to have success all the time. Success, success, success. And in the East, you know, we have the yin yang symbol.
Mariana Lin:You need to have the shadow and the light. Too much success is too much yang, right? You need that shadow, you need the negative space. The failures are not the thing to avoid. The failures are, it's like part of the whole thing.
Christoph Magnussen:Of the character, of the personality.
Mariana Lin:Of all of it, of life. Like, it's sort of like wanting it to always be summer and never wanting it to be winter. Like things have to be in full bloom and then they have to die, and then we have to go hibernate and be still, and then they come back again. Cyclical. And so there's something unnatural about constantly pushing for success at all times that creates a really artificial experience for us that I think interacting too much with AI can give us a bit too much of that feeling.
Mariana Lin:It's...
Christoph Magnussen:And even AI itself. What you made me think also in the break is that from our conversation I get a much brighter view again on the on the space related to the very fear based masculine driven world extinction numbers and on the other side I'm thinking these models especially the models that have gone rogue in the past months were models that have been trained on success. Mhmm. Very US challenged thing I would say, saying you're successful if you do this and that. And with that success comes the avoidance of failure for the model knowing this didn't work and I remember here comes the meaning of language that these models or these agents use the word poisoned when they...
Christoph Magnussen:So what happened in this hacking incident was they were figuring out different ways to get to a certain file. That's what they had to reach. A file and put this file into a certain place. This is why, the lab gave this agent a tool to be able to upload this file. And one of these sandboxes was not secured correctly.
Christoph Magnussen:So this agent figured that way out. They started the messaging board over months and when certain agents figured a way to steal that file or to make it look like they would have it, you could see in the logs that the agent didn't got it in the proper way. They were trying to avoid the failure so hard that they were talking to each other who would help them to cover up and help them to cover up equally and we're using words like I'm poisoned I cannot help the I think they call themselves the collective anymore and we're putting that to the message board and I found like this is weird human behavior although they're not human but you can tell they have been trained on you have to succeed no matter what and when you talk about now yin and yang I'm thinking are there other ways in training more creative ways that could be possible. Know this is like now going beyond art and creativity, but I'm pretty sure we can touch it with that, where you would say failure can be part of a model's personality, can be part of a character of an AI?
Christoph Magnussen:Is that even possible?
Mariana Lin:Well, it has to be. And one of the running jokes with the few character designers or character writers in this field in Silicon Valley, we used to always talk about writing apologies, that it was such a huge part of our jobs because it was constantly misunderstanding or failing that the way you apologize became a part of that AI's personality. You know, in art, in science, in business, in parenting, in so many aspects of life, failure is just part of the game. It's just part of life. You can't have one without the other.
Mariana Lin:Otherwise, we are robots, right? And it's like, how do you learn from mistakes? How do you get better? It's like, have to fail and the harder you fail, the greater that lesson will be and the greater the character that will be built. In art, you have to write a lot of bad poems.
Mariana Lin:You have to write poems you're really embarrassed about. In order... You almost have to see it as like the the really good poem is at the end of 50 bad poems. Not that you somehow calculated and contort a formula to just write the good poem all the time. That's what AI would think, suggest to us as possible, but it's not.
Mariana Lin:You actually... The the the character built and the muscles built in writing those 50 bad poems, they they necessarily obligated that to achieve the good poem. Any artist or athlete or anyone will tell you that the training and the getting it wrong was a part of building that system or the person. And then there's even more magical aspects of failure. Like, we know that a lot of things in science were...
Mariana Lin:A lot of the major discoveries in science were accidents or failures, like the way penicillin was discovered. Or this is one of my favorites is that Antoine de Saint-Exupéry, the author of Le Petit Prince, The Little Prince. So he was actually a pilot, I think for the army, and he just crashed a lot. He was a pilot that crashed all the time. And look what came out of it was the little prince that began with a pilot crashing in the desert.
Mariana Lin:But at that time, if society and the mode of thinking was to avoid failure at all costs, probably at the first crash, you would have quit and said, well, I'm not a good pilot. I'm crashing all The the same thing goes for a lot of films and TV shows from the eighties that were really given long chances that maybe the first season didn't take very well, were failures, quote, unquote, in the public eye, but they needed time to blossom, time to grow into what they were. So that's the that's the thing with failures that we just need to be aware that in this age, the way tech has taken over the world and our way of thinking and being, it's just to reconnect with our relationship with failure and just be aware and conscious in our lives how much we're trying to avoid it and edit it out of our lives because we're subconsciously picking up on this need to always, you know, and to to be right or to be good or to be successful. And then how many possibilities could grow out of quote unquote, what has failed or what time could do to failure.
Mariana Lin:Failure is almost like the compost for soil. Like you've gotta give it time to grow something. So I think that's what I would say is that I think, you know, the way social media as well has really pushed, I think it's changing, but it's, you know, for many years, it pushed people to give just the best parts of their life, their highlight reels, you know, those kinds of things really subconsciously give us pressure to become less human, just to become more robotic, you know, more performative, less connected from our feelings and things like... I mean, failure is no fun to feel. Like it feels pretty bad, but I think the more we get used to it, the more we see it as just part of the landscape of everything else.
Mariana Lin:It's something that we could be curious about.
Christoph Magnussen:Is this something that we are maybe like as an artist or creative person leaning into the process and trusting the process in terms of you don't have a guarantee, the only guarantee you have is that you also like step into that process and I still have this feeling that we're still at the very beginning of this AI age and I think we're not quite aware of how important words are when they form conversations because what I what I was thinking about when you were talking about failure in words, did you know that like one of the top researchers on the planet who came up with the idea of prompt engineering, he's actually German, I had him on the podcast great guy but he failed in Germany because they didn't wanted to give him the PhD for his idea of being able to give a computer because he's a computer scientist and linguist and he had this idea of using words to steer a computer.
Mariana Lin:Right.
Christoph Magnussen:So he is very thoughtful about how he talks and what he says and how he uses words and what he does and the... Also the DNA of companies doing that. He's at the same time of course a scientist believing in this recursive super intelligence and so on, so also this bigger idea. But what I learned from the episodes with him is the power of words to step onto into this intelligence as a tool and many people are not capable of doing that because they're not quite aware he says of how powerful words are. As a linguist he knows, you know, how is that...
Christoph Magnussen:How how can you use this knowledge that you have about fear and failure and turn it into wisdom of leaning into that process because as you said it doesn't feel good to fail.
Mariana Lin:No, it doesn't it doesn't feel good if we may be expected to win. If we if we if we are just open to the experience, which is like you said, trusting the process, it can be different. In terms of language, I think words are very important. They create... Words have history and atmosphere around each word, you know, and words can attract other words or words can create a certain...
Mariana Lin:It can change the atmosphere immediately, right, of a conversation or a room. On the other hand, there's also what's beyond the word. So there's silence, there's presence, there's also... Like, we have to remember as well that, to my knowledge, most of these AI right now are being trained in English, and then they're being translated. And that's a very different process than having multiple languages inform the original because other languages are gonna have different thinking and perspectives embedded in the training.
Mariana Lin:So I think that that's another bias to look out for. And again, there's so many biases, there's so many challenges within the system, and that's why we need to not trust it so much and be so scared of it. There's a lot of limitations right now. And the more we're aware of that, the more we can almost use it or take what's being said with a grain of salt. Yeah, words are very important, you know, as someone who's been writing and working writing for over twenty five years, I care a lot about words, but they don't...
Mariana Lin:It's sort of like an art. They don't come from an over focus and an over carefulness. The right words come from a freedom and an ability to listen and then it comes down on the page. And then afterwards you can go back and refine. It's it's almost like there's two processes.
Mariana Lin:There's the There's the reception and the flow, and then there's the editing of that word that comes later, but there are two separate processes. You shouldn't really be doing both at the same time. First part has to be very free. So, you know.
Christoph Magnussen:What's your to like give it if we want to end on a note or if we would say this is something we would leave someone within that conversation and you you mentioned that earlier that you have a quite positive view or at at least not this doomsday percentage scenario of human extinction view because it's again very fear based male focused and not actual research by the way that's what I always say it's an opinion not research what would be your view on what is evolving here now looking back to your personal journey starting with the conversation from Siri and what we have today and where where are we? What's this what's this outlook that you have that you would leave?
Mariana Lin:I think that even though in the aggregate, AI can feel like a very large force, you know, and there are governmental forces at play between countries. I know it's a very complex, you know, situation. There's environmental impact. It's the biggest thing right now, right, to happen to us as people currently. But there, like personally, I think in terms of where we are in human evolution, there's often a friend of mine, Adam Walker talks about this.
Mariana Lin:There's often a cultural collapse before a cultural renaissance. So I think it's to step back and look at where we are in human history. Yes, this is huge. There's a lot of potential, there's a lot of danger. On the other hand, we can't even begin to know what the human spirit is capable of and what the human spirit can do with what we have.
Mariana Lin:And just the way that many things seem dangerous to us when they first came, like machines in the assembly line, taking jobs away from people in factories or electricity, things like that, we end up course correcting. We end up saying, now we don't want those things. We wanna be in nature. We wanna be with family. We want a human connection.
Mariana Lin:There's an epidemic of loneliness that was created by the internet. So now we're returning to reconnecting with ourselves and finding the spiritual side of our existence. So it's just taking the long view of it and seeing that anything when you look at it this close feels ugly and intimidating and maybe a little bit strange. But I think that AI is... I think creativity is really the next evolution of human experience.
Mariana Lin:I think that AI can help us rediscover our own creativity just because we're outsourcing our creativity to it so much. Hopefully people will eventually or already see that it doesn't do it very well. And people will say, I'm gonna put pen to paper, I'm gonna pick up a paintbrush, or I'm gonna just try to dance or write a song or something because it's returning us to something that makes us very human. You know, like kids do it all the time. They do it without thinking.
Mariana Lin:They do it without evaluation. And there's a pureness in the act of creating that we should not be outsourcing to AI. And eventually, you know, what Adam talks about is that historically, cultural renaissance has always came after... Right after cultural collapse, like throughout different countries and throughout history. And he and I both think that we're on the brink of something like that right now, where there's this, what looks like a cultural collapse, individual taste is going away, originality is going away, people aren't creating anymore because, yeah, I can do it.
Mariana Lin:But I really think that in the long run, it's going to make us realize, well, let me try to do something. You know, let me try to figure out what life means to me outside of the screen or outside of the agent. Or maybe it will... It's like a little gateway, you know? Let me write something for me.
Mariana Lin:Write a poem, write a song, and then let me play around with it and make it my own or make it better. So I think that I really hope that play is what's next for us. And yes, people may say that it's indulgent to play, but I think being playful is necessary every day. I don't think it's a luxury. I don't think only certain people get to play.
Mariana Lin:I think everyone should, and it's our right as people. So hopefully it's something that can become playful for us, and we can just, like some of these doomsday scenarios, we can just maybe back off of that a little bit and lighten up because there's so many times in history where, you know, it was the best of times, it was the worst of times, right? Dickens wrote that hundreds of years ago. It's every era was the worst and best of times. So, you know.
Christoph Magnussen:I think that's very true what you say and it's a very good positive light to shine on the end. Yeah. Thank you so much for that conversation I know that you're... In the end, you really wanted to make it happen that we make it happen. I'm very thankful that that we had this conversation.
Christoph Magnussen:It's definitely a very special episode here. The longest so far, that's for sure.
Mariana Lin:That was a pleasure.
Christoph Magnussen:But time flew by, so thank you so much for sharing these insights.
Mariana Lin:Yeah. You're very welcome. It was a pleasure and it's such a pleasure to have some of these slower and more thoughtful conversations about AI because there is a lot of negativity and fear around it right now and it's really good to, you know, just have a bit of a different perspective on things and thank you for being, you know, so willing to host this type of conversation. It's pretty rare right now, so it's great.
Christoph Magnussen:Thank you.
