Linus Ekenstam: The Future with AI (Transcript)

Aqeel: Hey everyone. Today we're here with Linus Ekenstam. Please yell at me if I'm pronouncing that. But yeah, Linus, is an awesome guy. He's an AI enthusiast who describes himself as an AI gardener. He's got a background in design. He's previously worked at Typeform, Flodesk, and Thingtesting.

Aqeel: He's a creator of Inside My Head. Which is a new subset publication by him where he is writing about AI tech, design, building digital products. There's a lot of great material in there recently about his awesome content that's got him a lot of Twitter followers regarding prompt engineering, where he has actually grown a good massive following.

Aqeel: Over 60, almost 70,000 followers on Twitter. And he's involved in the development of Bedtime Story AI and We Are Konch, which we'll get into later. And he's been sharing great thoughts on AI, showing amazing creative output and lots of engagement. So yeah, thanks Linus for joining

Linus: us. Thank you. What an introduction.

Linus: I feel honored.

Aqeel: Of course, man. I guess I'm just curious, How did you even become interested in AI? I guess you had more of a design background and something really sparked your passion, which it seems like you're bringing your design kind of brain into your work nowadays.

Linus: We'll have to rewind a bit I think in naturally I've been an early adopter for my entire life.

Linus: I think, four years old I was trying to get my mom and dad to buy me the latest tech and they never bought it. So then something went wrong when I got older and I got my own money and I could kind of get into the latest and the greatest. So it started basically, with me just liking new tech.

Linus: I think my first foray into startups was back in the days when Justin Kan started Justin.tv. I was one of the first people to stream together with him and I, and that's so long ago. I was literally just a kid.

Linus: But I never really capitalized on that opportunity. It was a lot of fun and I learned a lot, but kind of just threw over my head. From then to now was like, I've been around kind of every major technology jump. I tried to social hack my way into getting a Facebook account here in Sweden when Facebook was only .edu , in the US managed to get an M.I.T. Email from a friend.

Linus: Used start to sign up to Facebook when basically there were only college students using Facebook. I've been kind of interested in AI since someone showed me GPT for the first time with GPT-2 or 3 for the first time. And before that I just.

Linus: Did not like comprehend the change that we were gonna see. I had friends doing thesis on their master studies, they were just writing a bunch from AI and all of them kind of say , this is 10 years out. I'm like, okay, cool , but what does it actually do? Right? And they're like, no one could really answer.

Linus: Here we are now. The question is no longer, what can it do? The question is, what can't it do? We're basically just scratching the surface. We're just getting started. The more people that will, get into this, the crazier it will get.

Linus: How I got started with AI is basically that I just gravitates towards new things. I kind of like diving head first into these things, not being afraid of failing. I think just having a very playful mindset. And to be frank, I was not very old.

Linus: The .com crash, or the 1999 internet hysteria, I don't remember much of that, if I look back, but I, I do remember what it meant from me and my friends . And since then I've only witnessed one similar instance where there's like an 'aha' moment for everyone, and that was kind of the iPhone or like past the iPhone when App Store was released.

Linus: That was kind of like, okay, this is gonna change everything. And now I'm seeing the same thing with AI, it's the rebirth of the internet. It's like it's rewriting the loss of internet period. It's just so hard to comprehend. And I think people, maybe people that are close to the action in sf, they see this as well.

Aqeel: I completely agree with that. There's been this, like quoted , the dinner table lately of. You know, we're overestimating the near term hype, but underestimating the long term. I don't think we're necessarily underestimating or overestimating the near term because knowledge work is disrupted overnight now.

Aqeel: And you really just have more decision making criteria than it is having to switch your cognitive gear as a human, or even biomechanically operating a keyboard to write out your essays, write out your email. Make the thing, transmit the information, packet it, or expand it, query it, then go back and forth, and you're just focused on clarification and decision making, which is all we've ever really needed as managers for anything.

Aqeel: And so I think those folks were who were set for white collar work, that mentality. There's unlimited leverage points. There's a lot of this whole stuff of oh, no one's really gonna use the internet.

Aqeel: No one's gonna even pay with credit card no one's gonna get in a taxi with a stranger. All these things are horrible until suddenly people look at you silly if you're not doing that, right? They look at you as some sort of foreign person, right? 'Cause like you're not doing it.

Aqeel: Let's go there on like a sociology kind of perspective. Like how do you see people at scale on the planet and society

Linus: as a whole? This is a deep question. I think we don't know yet. I don't know.

Linus: If we put it this way, we'll just go back again to the .com crash for like 1999, 2000. No one was on the internet. If we compare relative numbers, we have 5 billion people on the internet today and 1999, I don't know the numbers, but it wasn't very much. Then information traveled very slow 'cause you were like at a terminal taxes internet.

Linus: By that I mean you had a personal computer, right? It was either stationary in your home, some people had laptops, but it was the main way to access the internet. From a fixed computer. Nowadays, we have smartphones, we have internet everywhere, 5 billion people of us and all these new technologies that they're coming, they're not like hardware bound.

Linus: I can do AI stuff on a $100 phone. I don't even need an iPhone. Everything is in the cloud. If it took a while to get people comfortable paying with credit cards, people still are not comfortable paying with credit cards in some parts of the world. They're gonna get a rude awakening because this is traveling like wildfire.

Linus: Had a really interesting call with an ex-colleague of mine today and he lives in Vietnam. He has a lot of office worker friends that are working regular jobs but in office buildings.

Linus: He is trying to like, explain to them the benefits of using LLMs for example. Most of them are afraid 'cause they don't even know where to start and they think it's like frightening. Some of them are starting to like find something in the workflow to kind of improve and then, a few days later, they're just mind blown.

Linus: They're like, I don't have to work anymore. I reduced my working hours by half. What am I gonna do? I can't tell my boss. The more people they're gonna get that aha moment, the crazier it's gonna be. And it's really hard to figure out the implications, like what happened with Industrial Revolution, if we talk about that.

Linus: Many people lost their jobs. I'm not saying this 'cause I'm an alarmist, I'm not saying this 'cause I'm not positive about AI. I'm super positive about it. But there is a realness to what implications does technology like this have on a societal level? I think it's very important to talk about, not that I have any, I don't have any solutions.

Aqeel: Yeah. No, I don't think we don't know all the data points to solve for problems that don't exist yet either.

Aqeel: Do you see the evolution of people's human relationships with AI as we continue to evolve and go?

Linus: One thing is very sure.

Linus: Within a very short period of time, we'll all have our own personal trained language model running locally on an edge device on our phone, wherever it's gonna know everything about us. We're not talking sci-fi here, not 10 years out, not even five years out.

Linus: I would be very surprised if we don't have this within the next 24 months from a big player. It is just gonna get omnipresent. It's very difficult to comprehend. You're seeing stuff now already, people running things on edge devices. Right? There's someone just the other day, Alpaca from Stanford?

Linus: Someone is able to get GPT 3.5 performance. You know, 600 bucks, like, come on. Yeah, yeah.

Aqeel: Right. Just way

Linus: above. And even if it's fluffy. Even if it's fluffy, right. Under this circumstances and you can't do these things, whatever. It doesn't matter. The point is proven, the trajectory is very clear.

Linus: Sometimes it's just hard to compute. It's just hard to realize where things are going and the fact that we're kind of like in the middle of it. Everyone that's in this space is in a massive bubble 'cause just try to poke a hole in it and just bounces back. People are not really recept for it yet.

Linus: I think we're gonna see a weird shift probably after summer if we're talking about like predictions in that sense. Imagine now Google pushing their stuff and Microsoft's pushing their stuff. Adobe's pushing their stuff. Workforce people like regular Janes and Joe's are gonna start seeing these things in the products that they use day.

Linus: I think we're gonna see after summer, after everyone's been on holidays like Northern Hemisphere, things are going to look slightly different. We're gonna start to see a bit more of a craze from the General Janes and Joe's.

Aqeel: So you're a father. One thing on the world of building a digital me, I think it's very important.

Aqeel: There is a human cycle of whatever x number of years it takes to develop a sense of self-awareness, personality and values of good and wrong. We're augmenting or even outsourcing a lot of aspects of ourselves. Especially when it comes to communication and sort of underlying ability to process whatever mushrooms going on.

Aqeel: What world do you really do you see for the coming generation, right? There's all these studies right now about the iPad generation or they're using phones before they can speak and then you have all this dopamine spike in games, right.

Aqeel: As a father, I'm wondering what you're thinking about, in terms of implications when you say that if everyone's got a digital me. What does that mean for child development and

Linus: psychology? This is a also a very good question. It's also a very big subject. It's something that we speak.

Linus: We've talked a lot about this at home now for the past few months. First we have the whole epiphany about, I'm used to Google, I've Googled my entire adult life. Yeah, which essentially is like a librarian. I'm going to this archaic website. I'm typing, typing in a string that I compose myself, trying to figure out what it is that I'm looking for, and I'm getting this registry of data and then I have to manually sift through it and there's paid stuff there and there's not paid stuff there.

Linus: And I find a website and it's bloated. I find another website and it's bloated, and then there's some viruses. Finally, I'll get what I'm looking for. I need to synthesize that and grab that. That has been state of the art, how we find information on the internet. That's my experience with Google for my most of my entire life.

Linus: Then we look at my daughter, I have a three-year-old daughter and a one-year-old daughter. And my three-year-old, her entire experience with Google is a voice in our house. It's an omni present voice that she can tell to do anything. Turn on and off the lights, play some music for her, start the bedtime routine.

Linus: Tell her what time it is, what weather it is. Call grandma. It's an omnipresent voice for her. Google has nothing to do with a search box on a website. She doesn't even know what that is. This has been her experience of like the internet from like age zero to three and now enter LLMs and we're using midjourney together,

Linus: me and her. She asked for a unicorn ice cream and we create a unicorn ice cream on demand. That's her experience of the technology as a three-year-old, where my experience with technology as a three year old was fucking Tetris. If we think about rate of change itself being exponential, I went from Tetris to LLMS.

Linus: She will go from LLMs to God knows what, it's just insane. In a very short amount of time, she will have seen more innovation, more progress than I've seen in my entire lifetime. It's difficult for general folks to kind of get this. They see, they apply their own, they did what the generation before us.

Linus: They apply their thinking of technological advancements onto the next generation. And it's a false dichotomy. It doesn't work that way, especially not with tech. I don't even know what was announced today, I haven't checked, but something tells me that it was like two nanometer processors or something like that, which is just like ridiculous in itself.

Linus: It's really hard to tell where we're going because like we can't even phantom the shame. Like we can't understand. It's hard for our minds to comprehend, what happens.

Linus: My older sister, she has like an iPad generation pair of kids, right. I have two nieces and they're kind of the iPad generation. I don't know what my kids are cuz they're certainly not gonna be the iPad generation.

Linus: They're gonna be something else, but I don't know really what it is. Because for them, embodied computer or embodied intelligence, the synthesized intelligence is gonna be everywhere and the way they interact with it will be primarily in natural ways. Voice or natural language, which we never did.

Linus: Like we, we never interacted with computers that way. Not, not on our day-to-day. So for us, it's gonna be very strange. We're the mouse and keyboard generation, and then we have some kids that are like the iPad touch generation. And I think the next generation is just gonna be completely natural. It's just gonna be like, I'm speaking to my computer, or I'm like, I'm, I'm doing something.

Linus: Maybe it's brain waves, who knows? And neural links gets their shit together. And it sounds sci-fi, but I don't think it is. I think it's like very real. Why would someone that's born today use a keyboard and mouse? Like, gimme one good reason. It's like, I don't, I don't see it. that's exactly

Aqeel: right.

Aqeel: The world's gonna be, it's already changing, right? We're already seeing the shifts. You're right, like things with hardware comes next, things with like bio. Bio data comes next. We'll, we'll see. Actually it's not comes next. All these things are happening in

Linus: parallels.

Linus: It's happening in parallel and convergence is coming next. I think if we just comprehend the fact that everything is happening at the same time, like all the fields are advancing at the same. More, more or less. At the same pace. At the same time. Yeah. It's all major technology fields, so just like running in their own tracks.

Linus: Some of them are like lucky to. At some points, but I think what we're, what's happening right now is like the, the great, the great race. I foresee that the outcome of that great race is some sort of mega convergence where we're making breakthroughs on multiple pathways at the same time.

Linus: And it's gonna be really difficult for society to kind of comprehend that convergence. The same with LLMs. It's not an overnight success. They've been working on this hard problem for years, like decades. If we look at biomed or we look at something else and they're also working on hard problems for years, like decades at some point, it's just gonna be very natural that there is actually a big breakthrough, right?

Linus: And what's scary this time around is that there seems to be a lot of things that have been riding kind of the same. The same wave of like increase of compute, right? So we have increase of compute and the rate of increase of compute has pushed certain things forward. We see that with kind of Nvidia taking the lead on like how through even software that we've been keeping track of more slow, even though physical silicon really hasn't kept up to speed, right?

Linus: We're finding ways to just make sure more slow, still, still dictates the terms. So, because we don't even know what happens when we, when we crack the nut, on the next generation of computers. Right. What's gonna happen when we got qubits running around? We don't even know.

Linus: No one knows. It's really hard to comprehend order of magnitude improves certain things that we thought was gonna get stuck at like one nanometer, we don't know. I'm not a futurist, by the way. I'm just a, a guy that loves technology and I kind of, I try to be close to the action whether there's something new happening.

Linus: If you talk to futurists, they will be like, yeah, all this sci-fi stuff will get real because like we made sci-fi and now everyone is racing to try to get sci-fi into reality. And I think there's something true in that, right? A lot to unpack from a, a simple question about the next generation of kids.

Linus: No,

Aqeel: no, no. Of course, of course. Cause you think about everything. I can imagine as a parent that's just already like a pre prompted thing in your head.

Linus: How am I gonna deal with this? Right? Right. What are gonna be my teenager's requests and 10 years from now we're like, when are they gonna come home and be like,

Linus: someone stole my, I don't even know what it's,

Aqeel: well, yeah. Mean we're, like you said, we might be transcending physical laws at that point, right? Like, we saw this one

Aqeel: a company called Humane, right. Where there's like a code in. Yeah. Like, that's freaky to me, right?

Linus: There's no device anymore.

Linus: If anyone from Humane is viewing this or listening to this, please reach out to my inbox. It's open.

Aqeel: I wanted to ask about dynamic prompting. it's very good. I think it's what people need to know about and this will be a good thing for my follow up question to you, which is kinda alluded to with your friend, your colleague before you talked to Jay on the phone about how he's impacted his fellow coworkers and colleagues and office workers.

Aqeel: Dynamic promptings, it's a very important skill. Curious about like how you first discovered it and then, you know, you I I'll explain real quick for the audience.

Aqeel: One sentence about it is that the technique involves breaking down a text prompt to multiple sections or tags. Which can be easily modified or manipulated to create different variations of the same generated output. And you can use all the tools, particularly you allude to a lot of it for image generation, a stable diffusion type technologies, Midjourney, Dall E. Stable diffusion itself.

Aqeel: Yeah, I'm curious to, hear your explanation on it and then, anything you might have to share about it. I'm more curious as well of what are common challenges first time users might be facing when trying to drag this out, or even folks who are using it who claim to be kind of like AI tool enthusiasts, but don't understand the implications you're mentioning here.

Linus: Let's just start with like dynamic prompting. I think like by no means I'm not an inventor of this technique. This has been around like, I think the way of writing, Aquarius. So stable diffusion, particularly like this has been around since the introduction of stable diffusion.

Linus: There's a few of us doing various things within this field. Some is calling it dynamic prompting, some is just calling it prompting 'cause a string of tags and then you'll be smart about how you order your tags and then they become interchangeable objects, right?

Linus: Yeah. If like, I'm standing on the street and then like, okay, standing, could be sitting, could be laying like whatever, you know, there's nothing strange to that. It's more about like maybe the framework of how, how we're prompting. Funny thing is Midjourney v5 dropped last week and they kind of completely changed how

Linus: the prompting engine works for V5, so now it's back to more natural language, which like they might have had early on. So dynamic prompting or this way to tag separate your prompts, are not working really great in V5 from Midjourney, for example. Then we don't know, Midjourney V6 is in the pipeline, and it's supposed to drop within the next two months or three months.

Linus: So maybe someone from the journey listens to this and they go like, well, it's gonna happen earlier, but what's interesting though is if we look at that, like the way the prompting this way is more like, you have an idea and you're trying to like segment your idea down into tags, and this fits our brain really well.

Linus: If we're not used to like Google search or like, you know, separate things in boxes and, and build out a prompt, then that gets interchangeable. But what's interesting with the change they did with V5 and going with the natural language approach where it's like you're describing the image more vividly.

Linus: So like, you know, there's a man on a bench, he's reading a newspaper, the morning sun, and there's a mist in the, in the park. Most people when they've actually tried Midjourney or Lexi or, or stable Diffusion Dall-E, this is the way that they think they should write a prompt, and most times it fails because it's not, this is not how these models work, or at least not the text interpretation part of the models.

Linus: Hmm. So it's interesting that Midjourney's kind of taken this approach where they're like, tag separating things or like just broken down list of words, let's try to use natural language instead. It's gonna be interesting to see cuz it's a bit harder to work with. It's a bit harder to kind of find if you, a specifically from a creative standpoint, if you're kind of like looking to create a specific set of images with a specific set of like attributes and

Linus: you are going from having a really nicely chained command to be like, all right, I need to like actually change the wording here to be like more fluid, more natural. It's gonna be a bit harder to work with. My thesis on all of this, when it comes to prompt engineering, which is a term that's ridiculously funny in itself, it's that we're gonna get rid of all of this crap and they're gonna be like, probably like precursors or like natural language precursors. Much like when we're writing code, none of us, I may, maybe some that listens to this, they might write assembly code, or very low level code close to the silicone, but most people don't. Right? We're quoting very abstract layer, very, very high up in the stack.

Linus: So right now, if we look at how we're integrating or how we're communicating with AI, we're very close to the AI. We're as close as a user can get at this point without getting into the model and the pre-training and fine tuning. So we're kind of at assembly level. If we're just making a comparison or analogy between like normal coding or programming.

Linus: So I think that we'll see universal layers where like there's one layer where I can just attach what I want and it can string down to all the potential models. Or like, you know, maybe I'm running stable diffusion or I'm running stable diffusion Dall-E and Midjourney and I want to be able to write one prompt, but then it adapts to the different platforms, so the different models to get exactly what I want from all of them.

Linus: I think there are people working on this already. More people should work on this to kind of get the ball rolling on this. Cuz like, if we wanna lower the the barrier of entry, and this is by the way, just like touching on that. Now we're segueing into that. Like, one of the key reasons why I'm sharing as much as I'm sharing on Twitter and why I'm writing my newsletter is that.

Linus: When I got my aha moment, I'm like, shit, this is the first time ever. I'm gonna be probably obsolete very quick and a machine's gonna take my job. And my response to that was not to like run and hide or put my head in the sand. It was just like, I need to learn as much I can as fast as possible and then share that with as many people as possible to make sure that,

Linus: the rift between the people that don't know the skills or don't know how to use AI, and the people that do know AI, that divide doesn't get too long in terms of time. Because if that time, you know, time delay between those two cohorts, that gets too long. We're gonna get more societal risk if we can keep that tight and get people moving along with technology as it advances, we're gonna get less societal problems.

Linus: So, I think one part of that is like lowering the barrier of entry and it's already super low. Like you can run all these things on your device, your phone, in a browser, whatever. Most of these things just run. However, if it's cumbersome to use, if it's complicated to use, like Midjourney is in Discord.

Linus: It's the biggest hurdle to use Midjourney is learning Discord, right? If we can get the barrier to entry very, very low, we are gonna see mass adoption really quick and we saw that which had GPT, right? It's fucking massive. Like whoever took the decision at OpenAI, kudos like, of course, of course.

Linus: This is the interface, right? Everyone can relate to it. It feels like magic and there's no questions asked. You put some, some stuff in, you get something out, right? The playground was jarring for most people for like over one and a half years. I looked at the Playground like the first time back in 2021 and I'm like, holy shit.

Linus: What's this? It was all these parameters. What can I do with this? I don't, I I don't wanna touch it. It's like close to lid on my laptop, moving on.

Aqeel: That's interesting. Well, I mean, on, on that note, I guess my follow up question goes to like what you were just saying, you want the information accessible for everybody and once they have that aha moment, I think it's very, addictive is the wrong word, but I guess people can obsessively

Aqeel: not have to sleep anymore for a while just to ingest cuz we really are on the tides of the future, right? Yeah. And when Emory has the moment of like, wow, every conventional thing I thought about my work and how I relate to output and I relate to productivity is gone. Right? Education, studying, from just planning things out to creativity in itself is just completely rewired the brain.

Aqeel: I'm curious about, when you're seeing knowledge workers or when you're talking to folks, you're talking to friends, family, et cetera. What are the things that really get people to come to that aha moment and what sort of patterns do you see when people are beginning to adapt AI's first time users?

Aqeel: They've been following tech for years. They've maybe heard of artificial intelligence or algorithms and machine learning, but they're kind of buzzwordy to them. Some folks are still thinking that, you know, machine learning AI is just a lot of data and analysts

Linus: so what I've seen and, when the simple thing is just like people are doing the basic stuff they're going to chat GPT and start talking to it, or asking it to, to like, make this shorter, make this longer. Like, write this for me.

Linus: I'd say transformation or even summarization is one of those things where like you paste a website or, or an article and you ask for a summary and then they go, shit, okay, that was cool. That's like, what else can it do? And then you go like, well, ask it about any subject, any subject matter.

Linus: Ask it to explain it to you as a five year old. And they go like, okay, I'll try that. And they do that, and then it's like, Oh, that's cool. And I'm like, now ask it to describe it to you as a PhD student. And they do that and they go like, holy fuck. Right? What? What just happened? Like when they can like string things together, it is like, okay, it can do this.

Linus: Okay. That's very okay. You know, that was nice trick. And then you ask the AI to be more of verbose, and they go like, oh, okay. And then you go, wait. Now actually you know what we're gonna do? We're actually gonna ask if we're contradicting opinions, we're gonna ask you to take on the lens to someone else.

Linus: So like to try to poke hole in your ideas. And then that's when stuff gets really kind of like, Like, oh shit. You know, like, why am I needed? You know, that, that is kinda like that. That's when you get people scared. And then, they kinda like go into their shell, right? But then you go like, well, now figure out how to use this to your advantage.

Linus: And then that's when people, most people get excited, I think. Another thing that I witness is kind of like when you ask people to tell you- Usually I do this, I asked a friend, like two weeks ago, and he's, he's a CMO for like relative big sports brand. I'm like, so are you guys using, you know, GPT products yet?

Linus: Like anything in work? And he's like, ah, we played with it. Like, you know, when everyone else played with it back in December. And I'm like, okay, you haven't found a use case for it. And I'm like, oh, what's your, What's your biggest, like, time sync at work? Like in terms of comms, right? He's like, oh, I'm writing emails to all these salespeople that, you know, cold emails me.

Linus: Some of them are not cold emailing, but they're just emailing. I was like, why don't you just make it a template and I showed him like one of my templates. It knows my style of writing, and I can just like be polite, paste like a part of whatever email I get and I get like a nice response back and he is like, holy fuck, you just saved me like two hours a day.

Linus: And I'm like, there you go. Like two hours a day times, you know, 280 days a year of working that has a lot of time saved. Right and you don't have to be worried or you're stressed about answering those fucking salespeople. So it's a bit funny, right? When you, when you can show people, like in their workflows, like normal day-today things? Yeah.

Linus: That most people get, like really like, oh shit. That's why I think like Google and Microsoft, they're gonna, they're gonna kill in this space, right? They're gonna make some really great products. That's funny. I think it's hilarious. And I'm, I'm feeling sorry for like, you know, Jasper Copy AI.

Linus: They're gonna have to figure out their modes. They're gonna have to figure out how to stay relevant, right? It's the fringes that's gonna hurt, right? The people that have been in the industry and worked work for 30 years, they're gonna have a hard time adopting to the new tech people that are going to law school today or starting law school today.

Linus: I dunno what's going through their minds, but like, what's the likelihood that they're gonna be able to like pay back their student debt? I have no idea, man. But it's to, from what I'm seeing, is pretty low. So, and everyone in the middle is going like divided into two camps.

Linus: Either you are with AI or you're against AI. But I think we need to like really come together in the middle and be like, how do we make sure most people benefit from this? Like, we can never get like a hundred percent contribution to everyone, but like, we need to get 99.9% of humanity along the right.

Linus: Otherwise we're just gonna create a lot of problems. Right. Yeah, I think that's like, this has been a lot of big thinking, but all this stuff goes through my mind when I share like a small little tweet about, you know, some damn announcement from some company doing some cool stuff in AI. I'm like, oh shit.

Linus: This is like, the underlying stuff is really like, yeah, it's, it's crazy when you think about it. just the implications,

Aqeel: the mass adoption. I had a feeling there's all these joke going on right now where I, with all these new startups happening, but. Microsoft and Google have the world's data, the distribution, the infrastructure.

Aqeel: These big players I think are just gonna keep squeezing out the runner ups and well, it, it's gonna be a very interesting world. Interesting is probably the wrong word for it, but yeah. With Google has the Enterprise Suites tools rolling out right now with the wait list announcements for the Bard Technologies?

Aqeel: Yeah. I mean, it's, it's happening I think. To your point about, as the world's waking up, having the aha moments, I'd argue these companies had the aha moment not too long ago. And now you overnight see these giants who've been sort of dormant for a little while actually use that marine power into production, right?

Aqeel: What happens when they operationalized thousands of engineers are on the planet, who are some of the best people, right? Yeah. That's hell they fired. All the cards on the table, everyone's being operationalized and, and let's keep shipping this into everything we

Linus: can. Shipping it to everything. Like the, the only thing I can see taking some of these software plays out, right. What happens? So we just look at the M1, M2, M3, like look at the M3 ships coming from Apple at some point. Yeah. At what point will system on a ship where like Apples can of silicon be able with their neural engine to run an LLM locally?

Linus: And the the, when we have that at a GPT, I don't know, for whatever level we need to like where they're gonna need some grading at some point. Like people gonna say, I got an LLM level one or level 10, whatever. Like, we're gonna have to figure that out once we have that running locally on your machine with like unlimited.

Linus: Just like the only thing like stopping you is wattage, like how much electricity you have at home and like, given the, the computers basically run forever. Then what happens then? What happens to the Microsofts and the Googles and what's gonna be their mode? Right. And if we look at like, the big models, they're gonna be a handful of them, maybe maybe two handfuls.

Linus: And similar to what we have with the, with the whole kind of hosting situation where we have Azure, AWS, and Google Cloud, like being like the three predominant biggest players on the planet. Let's say we, we just take AI and make, that's the similar case. What's gonna be the use case for like a big ass model that runs in the cloud versus like a nimble scalpel model running on the edge.

Linus: I think there will be like a combination of both of those where you're kind of edge AI, whatever you have running there will solve like any quick inquiries, anything that you need less compute around and it's super secure, super private, and whenever you need something bigger or like the task grows exponentially, then it goes to the cloud.

Linus: This is kind of like, I, I guess like open AI sold for like half the company for 10 billion because they're seeing something similar happening, right? Like they know that there's a time cap on, on capitalizing on this type of technology because it's gonna get near free to use. Wow.

Aqeel: Yeah, that's a very interesting that applications just go on forever right now when it's like the power.

Aqeel: As long just we've had information. Able tools, center finger, inner fingers, and people go on to become entrepreneurs. They do YouTube. They do things like you're doing in terms of just like being active and having a following. And then you have opportunities for your business to do well because of that leverage you've built up and, and attention and credibility.

Aqeel: Yeah. But nobody go beyond the layer of information and then you have power to execute and put out things in the world just through giving

Linus: directions. Let's put it this way. We've had like access to information at our fingertips. Yeah, yeah. We are gonna have access to intelligence at our fingertips.

Linus: Yeah. Right? That's where we're going. Like, and this is the cure. That's the purest thing. This like intelligence. Oh man. And then we can argue what type of intelligence, but it's gonna be like having intelligence at your fingertips.

Linus: This is very interesting. Yeah. I mean, for sure. And I think like, go check out, like for anyone listening should go check out Konch.ai, like if you're into transcription or translating anything. I think, yeah, what we're doing is pretty cool.

Linus: Let's start with Konch. So Konch, like we've been around the block for a while, we're focusing on education and we're doing transcriptions and translations, and creation for anything.

Linus: We're mainly for students in your universities, right? But we do have some, some consumer products as well. And I think the moat that we have and what we're focusing on is like human augmentation. So if you do transcription today, you'll see where there are rates for like a hundred words, for like a thousand words and we're , zero. And we're at zero because we're using humans to augment the last percent. So if you want perfect translations or, or transcriptions, then you know, it's Konch.ai that works. And then, with bedtime Stories, it's more like, it's us fiddling. How can we build a personalization narrative company? What does the narrative company of 2030 looks like?

Linus: I think Bedtime Story AI is like our seedling to that. That's where we we're playing around as like a test bed. I think the personalized narratives or stories will be crazy. My kids are gonna be able to go and say, I want a story about myself doing this.

Linus: Or I want a story about my friends doing that. And they're gonna get it. Like in video form, Netflix is gonna look very different 15 years from now. So I think bedtime stories is us playing around with that, so it's BedtimeStory.ai and you can just generate bedtime stories as text and, and images, but also audio form it's really cool.

Linus: And we have a lot more in the pipeline for that. So it's those two things that I'm focusing my time on.

Aqeel: Amazing man. Well, I'll, I'll be sure to link everything about your work. Perfect. So this has been great Linus.