Thinking on AI, the Future of Work and the Commoditiziation of High(er) Skilled Labor

Kahran Tavya: Seth Godin talks about second wave of automation

Divya: It's interesting to think about when humans say, oh, the jobs are gonna go to the robots. But, like, is that the kind of job that you want to be doing?

Kahran: Correct.

Divya: If somebody could replace a part of my job, please replace it, because if it's replaceable, then it's okay. Just replace it. Hi, I'm m Tavya.

Kahran: Hi, I'm Kahran.

Divya: And this is thinking on thinking.

Kahran: So the notion that, Seth Godin talks about to create a second, I was like, wait, who was it? is that there? And he uses mechanical Turk as an example, which is, Amazon service that lets you break up work into really small segments, and then people can go out and do that work for. For your company or whatever project that's outsourcing the work. And the thing he's talking about is how some jobs are going in that way. That would take. Maybe if you're doing a translation, you would need a highly skilled translator to be able to translate a document. or using the example of Wikipedia, you needed all these researchers, I think. Encyclopedia Britannica had 120 full time staff. If something like billions of dollars of man hours went into creating this encyclopedia, and then Wikipedia managed to carve the work into small chunks that then people could just do. So now you could disperse it out and, you know, millions of people could do small amounts of it. Similarly with mechanical Turk, right. You can get something translated, you're paying by the minute, or maybe by the word or something, right. But you're being able to do something that was previously only available to. I'm sorry. That type of work previously only went to people who were highly skilled. So I was thinking about it in other contexts. Like, if you thought about Uber and like, taxi driving at the system level, then you have transport. as a service was previously something that was kind of done by a, ah, regulated. A highly regulated environment. Right. Like, taxis are highly regulated, and that was because they were trying to make sure that people were safe. Right. I think. I mean, I'm speculating a little bit here, right. But now we're able to deliver that with technology that allows for people to kind of be safe. and, whatever solves the kind of like, needs that were leading to that regulated environment. but similarly, you're able to kind of break up the problem of having a system wide transportation solution into small, discrete chunks that then can be done by a lot of different people. So with like an uber sort of service, like, people can. Individual people can help solve the kind of transportation service, at a city level, so, as I was thinking about that, then, I was just trying to think about, like, what are the other things? What are the other waves that, Or other industries that this wave will now come into?

Divya: So I have actually some ideas around that.

Art market was slightly under disruption because of nfts. com

Kahran: Please.

Divya: So, like, a couple of weeks back, maybe last month or so, I got the invite to Dali too, which is the AI generating, AI image.

Kahran: Yeah.

Divya: I will be talking about, like, you tell it that. You tell it that a man dressed as an avocado at a football game. And it'll give you four images of exactly that. And you could tell it, oh, 50 mm lens, close up, short background, blah, blah, blah. And it'll like, exactly make that thing. It's really quite incredible. And now because of that, so, of course, the art market was slightly under disruption because of nfts. For whatever their work, it was already under consideration of. Okay, where is the future of art? And where is the future of visual stuff going? And now because of AI generated art and how interesting it is, like, there is this almost surrealist quality to almost all of it. Yeah, right. if you see that or, there is another software. I don't remember what it's called. Like, that's also another one of, Wait, let me just check the name.

Kahran: Well, I got served ads this morning for something called Jasper that apparently will write your marketing copy for you.

Divya: Oh, wow.

Kahran: I was getting ads on Instagram for.

Divya: Yeah, the GPT-3. Is that right? Like that. That has been, Yesterday I was talking to, like, I had a call with somebody who is doing, who is working with like, SMBs for their local SEOs, like, and stuff. And they were talking about how a lot of companies, like, especially the marketing companies who do national, or global level SEO, now they will use GPT-3 to generate content for website blogs.

Kahran: Wow.

Divya: And it's so interesting, right? Like, you don't have to pay anyone. You can just use that like an AI to generate this thing. I had another friend who was, who like this couple, they were developing, software which could do, text to voice very, very well. And so well that you could give it a script and you could give it the intonation and all of those things that you want and it would convert it so you could have podcasts episode, even if you didn't want your own voice. Like, you could choose the accent and you could choose the gender and the age and like, you know, all of these specifications and it's so interesting. Yeah. Stable diffusion is the other software that I was talking about.

Kahran: I was reading about it a couple.

Divya: Of days ago, and it's very interesting, like, how all of these softwares are coming up, especially as it almost, to me, feels like there was a point of time, in the last decade, where everybody was like, Google Maps sucks. And then it transitioned into. Google Maps is the only way people transport around the city. Like, they do not know how else to get around. Right. Yeah. It's such a. It was such a natural transition. You don't even realize when you stopped, like, looking at actual physical maps. Or, like, I remember when I. When we had first moved to Delhi in, like, 2009, and my mom and dad used to have, like, a physical map when they were roaming around in the city. Right. Like, my mom would be telling my dad which road to take and which turn to take and all of that, because she had an actual map of the location.

Kahran: Yeah.

Divya: And now you can't even imagine that. Like, I don't know if there would be anyone who would have purchased a map of Bangalore in the last five years because a map of Bangalore exists on your phone.

Kahran: Yeah. And I think, to me, it's also interesting that now the next wave of that has started. Like. Like, for example, my dad was driving his boat, last week, and my mom and I were talking about how when he started driving a boat, like, ten years ago, we. There was always charts, right? And the charts were a big thing of it because you had to understand the tides and you had to understand, basically, you need to understand the tides and then how it might affect where you could anchor and what, the level of the water. it's just, it's all on the gps, and it's a really nice system. It has two screens that you can look at two different parts. Like, you can zoom in on one and have zoomed out on the other. It's great, right?

Apple's new watch can be a dive computer, and they've partnered

And then similarly, I was watching Apple's, product announcement yesterday, and I noticed that with their new Apple watch, they're really going after some of the more or extreme, use cases. So the new Apple Watch can be a dive computer, and they've partnered with this diving company to build a dive computer app. it's just, it's interesting to me now how we're starting to see these kind of mainstream technologies really go after the really, really specific use cases that used to be tackled by really only specialized equipment. And now it's like, oh, well, why not, right? Like, you can just. If you can do it, like, why not be able to go after and show that this, this tool can also be used for this market. It's really interesting, I was telling Goro, because I'm like, yeah, I don't like that I have a dive computer, right? Like, I use it twice a year. A dive computer is just a fancy name, you know, it's a dive watch, right? Like, they just call them dive computers or something.

Divya: Okay.

Kahran: And the main thing it does is, like, when you're scuba diving, it tells you, what depth you went to. And then based on what depth, how long you need to spend at, ah, lower depths to help, reduce the nitrogen in your, in your system before you come up. The danger is that you, if you come up too fast, the nitrogen. Right? And it's like, well, it's. Again, it was something that, like, when I learned to scuba dive in, like, 2000, I learned how to do it on paper and how you would calculate, like, oh, yes, if I go to this level, I need to wait at this level for five minutes before I come to the surface. and then, you know, then you get to the point where you used a dive computer, but now it's like, well, I could just have my apple watch. Why should I spend dollar 250 on this kind of specialized piece of equipment? yeah, I don't know. I don't know if that's. To me, it feels like it's a similar technology wave or it's part of the same technology wave.

Divya: Yeah.

Kahran: but it's a, it's a definitely, it's a different side of it than the kind of, like, labor replacement side.

Divya: Yeah. But like, to me, it's almost like, I think the place where I was coming from was more. If we look at this GPS or the, dive computer use case that you're talking about, right. Like, in front of our eyes, like, we are living through a revolution, almost. Like, world is changing and we can't. I can't really pinpoint the day when I, like, you know, switched to Google Maps as my source of truth about traveling in a city. But that is the case right now. And when I look at, GPT-3 or when I look at AI generated art, that's what I feel. I just feel like, oh, you won't need somebody to generate actual artworks for you because this is how you can instantly generate artworks. Like, you don't need an illustrator. And if this is where they are today, where they would be in five years, we know that AI grows exponentially.

Kahran: Yeah, no, I was hanging out with some of, my dad's friends who work at Microsoft last week, and it was just really interesting. So, you know, there's a tool called Copilot. Have you come across this and it writes code for you when you're, it's, it's basically a pair programmer. That's an AI.

Divya: Yes, yes, yes. My sister was telling me about this, and she was like, it is insanely amazing.

Kahran: At least I haven't seen it myself. But it's both, like, it will do syntax for you if it's a new programming language. But you can also comment, like, I need a function for this, and it will go right up on the screen.

Divya: Yes. Yeah, so that, yeah, my sister was telling me about how, like, I think in a unity program that she wrote, she used it, and it was very interesting. She just like, she told it to do something and it just wrote a piece of code that did exactly that thing. And it's insane how awesome that is.

Kahran: So apparently on GitHub projects where copilot is being used, it's writing 40% of the code on average. This is what we were hearing last week.

Divya: Yeah.

Kahran: And then my dad's friend was also talking about how you could, because his daughter's very into manga, and she was like, you know, I really want to write one someday, but I don't think I could do the drawing. And it's like, well, this is a different world. Right? Like, now, if you can write it, like, something could generate the art for you, which is just an interesting, they're interesting use cases. It kind of expands, it reduces the kind of technical barrier to do something. which I think is interesting. I think we'll see more waves of, like, innovation, maybe even of kinds of art or people who feel enabled to be artists because they don't have the kind of barriers.

Divya: Yeah.

Did Instagram commoditize photography by making it more accessible, right

I also like, there's so much interesting stuff within this to talk about because, like, one could think about all of, all of the cultural implications of, somebody would say commoditizing it, but making it more accessible is also a fair way of saying it. Right? Like, I mean, did Instagram commoditize photography? Sure. But, like, actually good photographers didn't become, like, professional photographers didn't become worse because of Instagram. I would say that the, like, the average person became a better photographer because they started taking so many photos and they started improving.

Kahran: Yeah.

Divya: Like, people 20 years ago probably wouldn't be so good at photography.

Kahran: I definitely, on an app, I mean, I wouldn't say it's so much m well, I'd be interested in your thoughts. But I wouldn't say it's so much the, output, which I guess is what I would call Instagram and more the evolution in the input tools. Right. Just the cameras have gotten better and the software to support the cameras has gotten better. So now, like, people, you don't have to understand bokeh to be able to take like a cool, you know, I think, what is the term depth effect, right. Like, right. And like, and there's all these people who never would have, like, my brother in law, like, takes amazing photos with his iPhone. and he just would have never put in the time to kind of like, like, learn about photography in that way. But now he doesn't have to. Right. He just knows how to do a few of the things on his phone and he can focus more on, like, framing and he, and I think, absolutely right. He's become much better at framing because of, I guess you're right, Instagram. Right. Because you start to see, so much, ah, of photos in a certain style that's kind of interesting. I wonder if that also pushes styles towards a certain kind of, like, place.

Divya: So this is, I promise this is related. Okay.

The more words a child's parent speaks to them, the faster it learns

so do you know how, like, when they do studies of how kids learn to. The more a kid's parent speaks to them, like, the more words a kid's parent speaks to them, the more faster the child learns the language and the better the child gets at the language.

Kahran: Interesting.

Divya: I, know, did you know this?

Kahran: A little bit of, like the related. So one of my best friends teaches kindergarten. And so one of the predictors for her, for success is how many children, how many words children know coming into kindergarten. Right. And so, like, one of the big things she does is try to get them, you know, to first grade, like, just having expanded there, because those are the good predictors for future success, which, I think is similar to what you're like, a similar body of research to what you're talking about.

Divya: It is similar, but not exactly. So this is more, I think this was done more for the class differences between different people. But basically, generally you'll find that, people who are economically at a lower level, they generally would have single parent, that parent is probably working three jobs, probably not talking so much to the child. So the child starts off on a back foot almost because the parent has not spoken so much to it. On the other hand, when a child has a lot of people speaking to them, they just pick up the language so much faster. Interesting.

Kahran: And because I guess you also have more variation. You get to see more people speaking the same words in different ways.

Divya: Yeah. Because, like, ultimately, it is a neural network trying to learn and make sense of things. Right? So similar to that, I think that, like, that is how all human brain learn. Right? Like, everything that a human brain is trying to learn, they learn it in that particular way, which means that if you are shown a lot of, quote, unquote, good images and you're given easy access to tools to take those images and you can see what you are taking and compare it to the stuff, it becomes a lot easier. And that's why I said Instagram made it a lot easier. It's not just the cameras. It's not just the fact that iPhone had, like, you know, really good camera. And, like, nowadays the phones have amazing cameras. It's also that people see photos and they know, okay, this is what good photography looks like.

Kahran: Yeah, that's interesting, because if I think about, like, the Facebook era of, like, the late two thousands, it was a lot more crazy photos. Right? Like, people would like, hm. You know, it'd be like, they'd be, like, doing crazy pictures or like, they would be, like, throwing, like, you know, it wasn't as much just beautiful photos. And I think part of that is no, like, Instagram knows what I like, so I get to see a lot more beautiful photos. but I do think you're also. What you're saying is true, right. Because it gave people more of a place for feedback for seeing similar photos, for seeing, like, this is the way that you could have taken it. And then you start to. Yeah, you just start to learn and think.

Divya: You start to have a vocabulary for what this thing could be, should be.

Kahran: When we started this conversation, I was kind of wondering about, like, you know, is it semi skilled work that's gonna be replaced by. By this kind of, like, wave of automation? And now, like, this, as we've been talking, it's like, it feels more like. It's almost like there's potential for new kinds of art, or at least like, reducing technical burden to art. I don't know. Do you?

Divya: So I think that the, metaphor that comes closest to my mind is maybe like 40 years ago, if you wanted to play music, you had to actually learn how to play an instrument. If you couldn't play a guitar, you could not make music, you could not play music. But if you today own an iPad or an iPhone, you have garageband. If you own a computer, you can find some free software that will let you, quote, unquote, make music.

Kahran: There's no equivalent though, in music yet, right. Of these kind of AI generation softwares that we've been talking about that exist.

Divya: For imagery and sort of, I don't know how much, I think in. So there is this thing called meta sounds that Unreal has released with their newest version. Know if it is procedural generation or if it is AI generation, I'm not sure which one it is.

Kahran: What is the difference?

Divya: But music is also happening.

Kahran: What's the difference between this?

Divya: So AI, go ahead. AI is like, you just give, generally a generation is like, you just give it a bunch of keywords and it'll make something procedural is you'll tell it the rules that it has to follow.

Kahran: Oh, I see, I see. So one would be like, make something like Beethoven, but rap. right. And then it would figure something out. And then the other would be like, I need something in this certain key, like with this, like, I don't know, cadence.

Divya: Right, okay, I see, yeah. Or even like more detail where like, okay, this is what the drum beat should be. Like, it should have this many instruments and stuff like that. I'm sure that like, people are doing it for music as well. There is no reason why you couldn't.

Kahran: Do it because there's so much data in the public domain. I was thinking, right, like, there's so many songs that are out there interesting.

Divya: And also, like, there is just so much possibility as well.

Kahran: Yeah.

Divya: Right.

What makes AI art separate from regular art, right?

So one of the interesting debates that has been happening has been, what makes AI art separate from regular art, right? Like, if I draw something versus I ask the AI to make something, what's the difference? like, recently, one of my friends, she decided to publish a book of poetry. And she just went on, Dali, because she had an invite and she chose a cover from there. Like she, you know, gave it a bunch of keywords, and she decided to pick a cover from there instead of asking someone to make it. Right. so I've just been thinking about, like, how do you distinguish if somebody else had, like, drawn it for her?

Kahran: Well, what's the line I'm right in remembering, right, that most of the historical artists, they worked with people they did, right? So it's not like. Yeah, yeah. Like they always had someone who actually was executing the painting for them. Or even like, I know there's a very famous glass maker, out of Seattle, named Chihuly, Dale Chihuly. And like, I met some people who work in his studio, and they're like, I mean, he's older. He's like, I think he's blind in one or both eyes. Right. It's like he can't be doing that much glass work anymore. And so a lot of the stuff that's coming out of his studio is being made by his, assistants. But, you know, everyone still calls it Chihuly art.

Divya: yeah, it's like Sistine Chapel was not actually hand painted by Michelangelo.

Kahran: Yeah. So I don't know. To me, that feels like it's just a set of tools. You know, it's not necessarily, I'm sure there was, like, in other eras where people are like, no, I don't know what came first. If it was like, oil, and I probably was like, oil, acrylic, and then tempura. Right. Would kind of like. And I'm sure that there were similar sort of reactions where people were like, oh, how could you paint with tempura? Only real painting is oil painting?

Divya: I would imagine so. I would also think that, like, this might not be true, but this has been my observation that the appetite for entertainment has increased over the last few years. Like, there are a lot more careers for entertainers.

Kahran: I think it's a reaction to what we were actually talking about a few weeks ago, but that there's not a good. That getting into a flow state in your relaxation time is, takes work. And we are, as a society, we've created so many things that don't, don't easily let you go into a flow state during a relaxation time. Or let me. Let me say that a little differently. There's lots of, things that you. That are slightly satisfying, and then you don't end up going into find things that are more deeply satisfying because you're being slightly satisfied enough.

Divya: interesting. Right, right.

Kahran: So I wonder if that's creating more of this deeper demand where it's like, people feel like there's unfulfilled and now you're willing to go and spend things on because they're not being. I'm not satisfied by spending, you know, 25 minutes on Instagram every day, so. But I still have that need of, like, I wanted something to entertain me. I wanted to see stuff that is visually exciting, and now I'm, like, going out and, like, filling it with something else. Like, you keep eating, like, a booze, booges, or little appetizers, and then you're like, damn it. At some point, I need a meal.

Divya: That's a very interesting way to put it. I was thinking that, like, it's just overall our, demand from our time has increased so much that we often are doing more than one thing. Like, you know, I'm driving, but I'm also listening to audiobooks, or I'm listening to podcasts, or I'm like, you know, doing my chores, and then I'm listening to podcasts. Or I would often be drawing and watching a YouTube video on the side or something like that, right? Like, my attention is divided. So, which means that even if I'm doing, let's say, 8 hours of work, I am simultaneously also overlapping 5 hours of entertainment with that same time.

Kahran: That's interesting because I wonder about. So if you compared to, a much earlier time, right? Like, if, you know, we were at a time where, like, you were still, looking for securing your, like, shelter or something, right. Then it would be much more encompassing to be like, oh, you know, I need to figure out where I'm going to sleep tonight, right? Like, you know, there's not as much time for. Or there's not as much brain space, for non essential activities.

Divya: or maybe there is. Like, I don't, I don't. I mean, that's a separate point. I don't know how I feel about assigning, intent. That's very fair to how, like, foraging societies live.

Kahran: Yeah, that's very fair. Fine. So maybe this is a bad example to go down, but the point I was, I was trying to get at with the example was that I wonder about how much it's that, that you're being able to use, like, the system one versus system two part of your brain, right? So you're being able to kind of do it almost on autopilot. These things, because of you. You've done them before. You've watched YouTube before, right? Like, it's not gonna be so startling to you. For example, like, m normally when I'm talking to you, when we, you know, when we have our calls and stuff, I might do something else at the same time, right? Where, like, just. Cause I'll have a head of thought or I'll have seen something. And so I'm like, yes, I'm. So we're still talking, but I'll be doing something at the same time. But when I'm recording a podcast with you, it's not something super familiar with me, so it requires a lot more of my intention. I have to, like, you know, am I speaking at a, like, staying close to the microphone? Am I speaking at a reasonably level volume? And so, like, more of those kind of, like, there's more processes involved in doing something new because it's novel. Whereas I think a lot of the things are not novel anymore. Right. Because we effectively do the same 25 things, you know, over the course of a week.

Divya: Interesting.

This is also making me think that as some jobs get democratized, others might not

This is also making me think that like, as some of these jobs get democratized.

Kahran: Yeah.

Divya: Not necessarily replaced because I don't know if the demand is outpacing the supply or not. Like, it could very well, like, instagrammer wasn't a job ten years ago. YouTubers wasn't a job. Ten YouTuber wasn't a job ten years ago. Which, like, celebrity wasn't a job ten.

Kahran: Years ago, like, well, but I think most, all of those were jobs. Just the path to becoming, successful in that career was not something so democratized. Right. It wasn't Instagram or.

Divya: No, but, no, but like, I. People still do make shows. I think that more movies and more shows come out even on a Hollywood level these days. Like, that market has also increased, but all of the tech entertainment is completely a brand new market.

Kahran: right.

Divya: Like, it's not like conventional tv has gone away.

Kahran: Yeah. I mean, if you look at the, like, ratings for tv shows, you know, nobody watches the way they used to. Right. Like, I think Grey's Anatomy, its second season or something. I remember looking this up. it was, I think 30 million people watched like the primetime. And that was a time when the United States population was like 350 million, right. It was like 10% of the country was watching. Wow. And now it's like 2 million is like a prime time, like success.

Divya: That is a very interesting point.

Kahran: There's just more options, you know, it was.

Divya: Okay. I'll take your point. That is fair. I was just thinking more from the perspective of there are more shows being made, there are more movies being released, and there are all of these other people, and everybody is spending much more time on entertainment than they used to before.

Kahran: Yeah.

Divya: So maybe the supply, like, despite the fact, or maybe because of the fact that it's more democratized, there is more supply and there is more demand.

Kahran: Interesting. I'm also wondering a little bit like, I'll draw a few blatant generalizations about history here, but I think, the printing press probably, or the printing press obviously made distribution of longer form things easier. Right. And I imagine things like autocorrect and like all these little tools, ah, editing tools just made it easier for you to write longer form things. You didn't have to focus as much on the individual words, individual sentences, so I wonder if these kind of supporting technologies we were talking about for generating, copy or for generating images, we'll start to see more, composite, creations or just like. Like, elaborate creations. And I wonder if that will start. If we'll start to see that into the entertainment side as well. Cause I don't think we'll get to a point that will be AI generated tv shows really fast. Right. But we might have more AI generated pieces of them. And I wonder if that might be an interesting. I don't know. I wonder if there will be new genres that will start to pop up as we start to see more of that. Like, today, I think.

Divya: I mean, I'm also.

Kahran: Go ahead.

Divya: Sorry.

Kahran: Okay, well. Well, I just, like, today, I think there's this profusion of reality tv, right. In different forms. There's reality tv for baking and reality tv for, you know, obviously for dating and, like, you know, a hundred other different forms. I wonder if we'll start to. And why is that? I mean, it's. It's cheap, it's unscripted. You get to people, like, to see people like them. Like, it has all these ways that it resonates with people. But I wonder if we might start to see that kind of art form, particularly being, like, elevated in a certain way, because it's able to, You're able to offload some of the burden of the entertainment to the technology, and that's being fully carried by the people themselves.

Divya: No, that's a very interesting thought.

Divya: I wonder if that's a million dollar idea.

Kahran: AI generated reality show, AI supported or something. Yeah.

Divya: no, that sounds like such an incredible idea. We should totally sell it to Netflix. We should be like, we originally came up with this thing, and you should totally do this.

Kahran: Don't you have some contacts with Netflix and. Yeah, maybe we can make it happen.

Divya: It would be really entertaining, though.

Kahran: Yeah, I am.

I'm curious why more people are not out there developing music models

I'm also pretty interested by this music one. I'm, like, curious why more people are not out there. Because I feel like you also have good. Because there's rankings, right? There's billboard rankings, and there's, like, purchase data, and there's, listen data. You would be able to actually get feedback into the model, which doesn't always.

Divya: Exist, but you wouldn't. But you wouldn't be able to use most of the music that is out there. Right? Like, popular music. And all you wouldn't be able to use for training your models.

Kahran: Why not? Doesn't it all relieve copyright after, like, 50 years or 99 years. It's like one of those.

Divya: Yeah, but, like, there was no music at that time. I mean, there was, like, there was no recording.

Kahran: There was no recording in 2023. There was music in the 1970s.

Divya: No, 100 years. Right. 1923.

Kahran: You would have to look, I really. I'm not sure if it's 100 years. It might be.

Divya: It is 100 years.

Kahran: Is it really a hundred years? Okay. Okay.

Divya: It is hundreds.

Kahran: Well, the twenties still had music, but. Yeah, that was a big bandaid. I don't know what was happening in all parts of the world.

Divya: yeah, like, most of the times, like, they had music, but there weren't recordings which lasted till now, first of all. And second, like, even if you have open source music, there are a lot of places where they do have open source music. Like, I think the volume that these models need is insane, which is also super interesting.

So I was watching this video about this person trying to train an AI model

So I was watching this video about this person who tried to train an AI model to recognize faces.

Kahran: Okay.

Divya: And, to recognize k pop idols specifically. And the interesting thing is that most of these models are based off of white people faces.

Kahran: Got it.

Divya: And so even despite her trying to train it, the model just would not recognize the people. Like, it just kept calling everyone the same person.

Kahran: Wow, that is so interesting.

Divya: So, like, as we go into, you know, GPT generated text, or we go into dali generated artworks. What kind of stuff? Like, whose jobs get replaced and whose culture gets replaced?

Kahran: Yeah, because, like, I know the dali one. I was reading about how, if you type like a beautiful woman, it tends to show an unclothed woman, because that's how western media portrays women on the Internet.

Divya: Oh, wow.

Kahran: Isn't that crazy?

Divya: Oh, wow.

Kahran: And it's like, you don't want to train on the Internet, but you do have to train on the Internet, because the Internet is the best data source to train from. But then the Internet is very biased in certain ways. Yeah.

Divya: It'S crazy. Wow. That is insane.

Kahran: so I think. I think because of stuff like that is showing up so early on, I feel like it will push us to be at least from, like. Like, wiping out, kind of like creating a general Internet culture. I feel like that is less likely to happen because people are already starting to see this issues with the training sets. I don't know, speculating.

Divya: Very true. And, I mean, I would assume that most of my friends who are from, like, southeast asian or south asian countries, there is. There used to be, like, especially when you were younger, there used to be this glorification of american pop culture.

Kahran: Yeah.

Divya: Like, american music is the best, and american movies are the best, and american shows are the best. But, like, if I see just the general Internet culture these days, it's not so much like, western, like, glorification of the west. There is some elements of west which are there, but, like, there are other elements where because of the abundance of content online and because of the abundance of just access, yeah. People are able to just say, no, I want my voice, and I want representation. Like, it. It's just interesting to. I don't know how we arrived here, but, like, the way you said it, it just made me think about all of that.

Kahran: Yeah, I mean, I think I sent you a voice note a few a week or so ago, but I had this kind of interesting notion. I wonder if, like, because we generate so much content in our lives, right? Like, I'll probably have taken, I think, 30,000 photos at this point in my life. So probably by the time I die, it'll be, I don't know, 200,000 photos, right? And I was just thinking about, like, I wonder if there could be, like, a nonprofit or something that people could donate their digital content to that could be trying to create less biased training sets. Because I think part of what you're saying is true, right? Like, people, this stuff isn't copyright. Certain stuff is in the public domain, and there are biases that exist because of that. But, like, I have a very close friend, Shweta, who, we started a couple of companies together, and she and I have talked a lot about how she grew up in this part of Bangalore called Malaysia. And Malaysia Rum has all these old stories that are just mostly captured in people, right? people know the stories of those areas, but there are photos, there are places, whatnot, but there's no real, like, she was like, I would love to, like, create a place that could be a repository for these stories or a repository for this content and data. I think that even just trying to create the places that could be, like. Like, repositories, even if we don't know what to curate from them yet, I'm very confident our abilities to curate are going to get. Keep getting better and better with these technologies. But I feel like we're starting. We don't necessarily. We haven't. We haven't put enough focus in saying, what is the source data going to be? What's the training data going to be? And, you know, how can we? There's certain places where I think, like, people still have to add value. and I think stuff like Malaysia, like this example of Malaysia, of creating stories around local areas. it's hard today because that kind of data, it's getting lost in. I think there's so much content out there. What content actually relates to this specific place? Can we create something that is from this kind of place or style?

Divya: Interesting. This is also making me think of, like, as you were talking about curation, right? Especially of the stories and local stuff. It's like, it's almost like art direction and curation and bringing taste into the picture becomes the role of the human element and the technology, like, does the rest of the work, almost. That's really what am I trying to convey and why am I trying to convey it? Are the questions that the human needs to answer. And then the generation ends up being on the shoulders of the AI.

Kahran: Interesting.

Karen: It's interesting to think about when humans say robots will replace jobs

That kind of reminds me about how, you and I were talking to, one, of my friends who's a conservator a couple of weeks ago, and, it was interesting to hear about how exhibits happen in museums. It does seem like it's basically like there's one curator or someone who kind of has a very controlling ability and they have a vision for it. And there's a lot of people who execute their vision, right? There's a lot of people whose job it is to run around, around and, you know, make this floor into that. yeah, so it's not that different today, right? Like, you know, maybe the AI would replace all of those, like, you know, ten people whose job it is to take that curator's vision and, execute it in reality.

Divya: It's also interesting to, like, as you said that, I was also thinking, it's interesting to think about when humans say, oh, the jobs are gonna go to the robots, but like, is that the kind of job that you want to be doing?

Kahran: Correct.

Divya: so my brother, sister and I, we were talking about, some of these stuff recently. Like AI generated art, a generated code, and because we are artists or developers, where does that leave us in some ways, right? Like, people around us are having those conversations, and it's like, if somebody could replace a part of my job, please replace it, because if it's replaceable, then it's okay, just replace it. I would rather do the stuff that is irreplaceable, but I don't know if that's coming from, like, a position of extreme privilege in some sense or not.

Kahran: I think a lot of things get standardized over time, right? And as they get standardized, I don't know whether to say it's like the fun goes out of it or the art goes out of it. But, like, for example, like, there was a time, that I kind of enjoyed writing Facebook ads, right? And it was about, I guess maybe like seven, eight years ago. And it was just like, it was fun, right? Because you could kind of, like, think about it. There was like, there was a lot more just like, I don't know, I would say kind of like art to it, right? But now, now it's like, what you should do is you should be just like, creating. Like, so you have a certain set of images, you have a certain set of copy, and you create all the iterations of them, right? See which ones work. And then you. And then once you know, like, what is the kind of direction for the imagery? Then you try another set of five images with the copy, right? And you keep just winning and steadily iterating in a way that's right for your audience. that's just not as fun. It's a different kind of skill set. It's a different. It's much more of, like, I have a good friend who I play games with, and we play this game called Stellaris, and he loves to call it, that. It's a spreadsheet game, right? What is it doing? It's a fancy UI for spreadsheets. You're just moving stuff around in spreadsheets. So I think jobs that are spreadsheet jobs, right, are just not that fun at some point. nothing wrong with accountants.

Divya: Yeah. It's also very interesting to think about, like, as people move into this direction, as the culture moves into that direction, how do people perceive their job, right? Like, do we want to be doing repetitive stuff? Do we want to be, like, developing skills which are just repetitive stuff?

Kahran: I think I met some people during my, so far in my career, especially when I was running the company in India, that they kind of wanted you to tell them what to do, you know? and I think maybe it was like that they didn't have the mind space available for work, right then. You know, maybe they had a lot of things going, going on in their personal life. Maybe they were trying to do something else with their life. And this job was kind of just like a way to make men's meat. but it was, it's an interesting thing because I interpreted that the, way I kind of internalized that was I was like, hey, okay, Karen, you have the ability to kind of create opportunity for people that will give them a sense of purpose. Like, what higher calling can there be to that or than that? Right. Like that you're being able to kind of create something that will tell people what to do who want to. Right. Who will tell people what to do, in a way that is at least to the best of your abilities, useful for them, for society, for whatever is helping the greater good? I don't know.

So industrialization made repetitive work happen, correct? Yeah, right. Like, because industries needed repetitive work

So I'm kind of curious of your thoughts if you feel like there will be. And I think as we kind of started talking about something Seth Godin talks about in his book Linchpins, he's saying you can either be a person that people are telling you what to do, or you can be a person who just be a genius at something. He loves to use that word, kind of genius a lot.

Divya: Yeah.

Kahran: I don't know. it's kind of, yeah, like you're saying it's an interesting place. I think it's an interesting place. Even as you, as we kind of think about as entrepreneurs and as we're starting a company, like what will success look like and what does, what are the kind of interim successes? And should we like, think about creating that kind of work or looking at, places where we've created repetitive work and say, hey, you know, is that a success for our business or is that not a success because we're creating jobs that are not good jobs or jobs that we think of as not being good jobs?

Divya: Oh, I just had this thought. So industrialization made repetitive work happen, correct? Yeah, right. Like, because industries needed repetitive work.

Kahran: Yeah.

Divya: And it's almost like, I don't know how right this is because, like, my brain just connected those dots right now. Now. And it's almost like now in the world of extreme personalization, we don't want repetitiveness. So we are almost fighting against that, I guess. Like there is like a tidal wave coming from one side and crashing into what existed long before.

Kahran: Yeah, you're exactly right.

Divya: And it's like people, like, from childhood, people expect to do a certain kind of work, and that's why they are like, you know, sent to and everybody's given the same education because they want to be doing the same work. But like, most of my friends who graduated from college are not doing the kind of jobs that they thought that they would be doing when they graduated ten years ago. Like, many of these jobs didn't exist, like, many of them have constructed their own jobs. What I am doing did not exist in this way that I am doing it ten years ago, definitely not in India. What my brother is doing definitely didn't exist. My sister is making something on VR. VR was not a thing. Like, it's so interesting.

Kahran: Yeah, it's. Yeah, it's really interesting.

Part of why so many startups are failing is because of the education system

I mean, I think an interesting point that maybe we can talk about in a future week is how the education system is. I think part of the reason why so many people have failed, so many startups are failing in going after ed tech and changing the education space. My dad is very deep into the education space, and he's very fond of saying that if you put a teacher, ah, from 150 years ago into a classroom of today, they would have literally no problem. What if you put a doctor from 150 years ago or any other kind of. Right. They would have no idea what they were doing. But a teacher, it's like, it's still. Maybe it's a black. It's a whiteboard instead of a blackboard. Right. But like, And I think it's because of what you were just saying. It's because the system there's. Society is looking to reward people for individuality and for kind of like, shining in their bright personhood or whatever. And our schooling is not about that. Our schooling is about trying to make people standardized. I think part of why there's such a discrepancy and why we're struggling so much in trying to solve this problem is because it's not actually solving education the way it's needed. It's like solving preparation for life. But education is nothing really doing that right now. It's like doing things that are orthogonal but kind of related.

Divya: Very true.

Kahran: We can talk about in a future week.

Divya: We should definitely talk about this next week. This is such an interesting. I feel like this is the first time when we have arrived at another concrete topic from, a topic. I love it.

Kahran: I love it, too.

Divya: Awesome.

Kahran: Is there a good talk?

Divya: Then we should continue. Yeah, this was awesome. Okay. We should talk about it later.

Kahran: We'll talk about it next week. Okay, bye bye.

Divya: Thanks for listening to this episode of thinking on thinking. Our theme music is by Steve Gomes, and you can find a link to it in the show notes.