The Future of Attention, Human Connection & the Internet with Mustafa Suleyman

28:20And it's always there to sort of like augment your own intelligence. I think it's going to be a very productive time because of that. But how much does that change the human to human behavior? Does that breed humans who are afraid to ask questions to each other? Wouldn't it breed the reverse? Why wouldn't it breed the reverse? My assumption would be, and maybe I'm wrong, that we can talk to an LLM and ask for a number of revisions because we know it's not human. And so it's building within us an openness to talk to an LLM that way in a way that we would never talk to Jesse, who's sitting here.

28:54But let's roll back five years, right? We're in the analog world of mostly interns and other human collaborators. you know the effort that it takes to craft the feedback in this very nuanced and subtle way often leads you to not push unless it's really worth it yeah right whereas now we're like oh we're in this habit of constantly critiquing it's all good to critique you know just earlier you were talking to me about you know getting your ai to pretend that it was ogilvy right and give you feedback on a script or an idea or something like that now we're all in this habit of like it just feedback is constant.

29:32It's free. It's cheap. It's really personalized. It fits different personas. And I think that can only be a good thing for human interactions because it demonstrates what good looks like. Right. And you get used to, you know, interacting in a way that like you feel comfortable, you know, being vulnerable. Like you're having your idea ripped apart in a hundred of different ways by the interns or by Ogilvy. But it's more, it's in private. And I think like in a public setting or in a human to human interaction, you might lose the reps, the repetition of like, what does conversation look like? What does conflict look like in real life versus what does it look like with my LLM?

30:17The flip side of this is that many people may not have had a continuous experience of another human being kind and respectful to them. Yeah, that's true. Listening to their ideas, being patient, being encouraging, being even handed. Like that's a real social privilege. Not everybody has that, right? And so to have a moment where you actually get comfortable enough talking about your ideas and fleshing them out with this thing so that when that moment does arise in the real world, you know, with another human, you're ready to, you know, talk about it. You've developed it yourself and you can take feedback on it.

30:56That's true. I do believe that's a good point, that there are a lot of emails that are being sent or texts that have been workshopped and prepared and not in a bad way of like, how can I confidently engage in this conversation with this person? Right. And having used an LLM as a sounding board first to make sure I've looked at all the sort of parts of it. Yeah. It makes me question though, like, you know, I was sharing with you, I have a four month old and we, my family wanted to start like an education fund for him, which can only be used in the context of college tuition or education. And I took a step back and I said, is this applicable in 18 years?

31:34Largely because of what we're talking about. Cause I think actually one of the greatest, like if we pull on this thread, this is like one of the greatest learning revolutions ever. Like even the fact that I can say you are David Ogilvie, like one of the greatest advertising executives ever teach me how to write copy. You know, like I can, I can personalize what I want to learn on an hour by hour basis. And I am doing that right now. And it is even, um, my most recent prompt I was sharing with Colin last night, you know, in the way that we've been working together for 15 years, we used to share video ideas with each other.

32:04I'm sharing prompts with him now of like, get this prompt. And what I did is I took a, um, the voiceover transcript for a recent video that we're editing right now. And it's like a video essay. And I, I, I took the transcript and I loaded it into chat, chat GPT. And I said, um, how would you approach this if you were Johnny Harris? And Johnny Harris is a great video essayist on YouTube. And it gave me how he would approach it. And I was like, huh, how would you approach this if you're a David Fincher, like a Hollywood script writer? oh interesting how would you approach this if you're Aaron Sorkin how would you approach this and I just started going down this list of all people that and I saw 25 variations of this idea from people that I respect and admire and it was like it felt like um the same way I felt when I was learning how to video edit right when I was like oh my god look at this cut look at what happens when you add music to this look at what happens and I just could not believe the limitless possibilities to learn how someone else would approach something.

33:05Yeah. I mean, that is literally my inspiration for creating Copilot in the way that we have. Before joining Microsoft, started Inflection, and we created an AI called Pi, which P-I stands for personal intelligence. And it was purely focused on the emotional intelligence of the models, the personality, how kind and supportive and encouraging it is, and giving that feedback in a way that really levels people up because that's like, you know, if you think about what your, you know, child might be in 18 years time, one thing is for sure that they will have access to a companion, a support, an AI companion that makes really information and knowledge and education essentially free and totally personalized.

33:49And the whole world is going to get access to that too, because it's just going to be infinitely cheap. It'll be available, you know, for cents, like less than cents for you know and i think it's hard to really imagine what a world is going to be like where everybody has access to pretty much the same knowledge and tools for creativity it's going to create hyper competition but also hyper productivity because it is kind of like a search problem like you know when you're saying you're trying a hundred different options or you're trying feedback from 10 different superstars. Really what you're doing there is a search problem.

34:28You're looking for the optimal combination of ideas that you think really suits the thing you're trying to solve at that moment. And so not only are we sort of creating these new AIs, but we're also empowering like literally hundreds of millions of more creators to add to the corpus of human knowledge and culture and scientific discovery. and that that is why i'm kind of generally net optimistic about what's happening in the next few decades because that's the thing that has made everything in this room our intelligence right every single everything in our life in our world in our structure is just a product of these biological computers these prediction engines that create value out of nothing it's kind of mind-blowing that we have a book on the table and a plastic cup and this amazing technology we've invented that through search we found the optimal combination of materials that then suits the problem we're trying to amplify and record and project and broadcast our voices so we're sort of just piling more intellectual capital onto the big challenge of improving life and well-being for the human species and i think that that is going to be an amazing you know sort of century.

35:44It's really going to be incredible. This is maybe not a question we would normally ask, but do you see there being friction between the pace that technology is and will be moving and the pace that governments will implement change? I mean, for sure, this is definitely moving incredibly fast, like exceptionally fast, faster than any technology has ever sort of arrived. And, you know, I think that there is going to be a moment when friction is our friend, when you actually want to have a little bit of, we want to have checks and balances on what it can do and what it can't do and so on. And people have speculated on when that moment is and what they should be.

36:26You know, and I don't quite know where that line needs to be drawn yet. I think it will be an intuition that society has collectively. Like everyone will just sort of know it when you see it, if there's something that people feel very uncomfortable about. I mean, deep fakes felt like a moment a couple of years ago where everyone was going to be like, well, you're going to be able to produce perfect footage of your friend doing this horrible thing. And I mean, I just think that the other way of looking at it is that people get wise to it real fast and they adapt and they're critical and they sort of know how to kind of avoid the weaknesses and extract as many of the benefits of it as possible.

37:04And so as a species, we're actually very, very resilient to the downsides of new technologies. It doesn't mean that we'll be infinitely resilient to them. And it may well be that there are moments when we have to say no to certain types of capabilities. But I think we're a little way off from that yet. So it feels like there's like the regulation side of this, which is a big question mark. The other side that is a big question mark is the monetization side. And I specifically say this in the context of creators and things that we have been approached with, right? And I think these tie hand in hand, regulation and monetization opportunity.

37:42So right now we're being approached with, hey, can we essentially license the Colin and Samir name image likeness voice? Let's build a model. Let's load in everything you guys have ever said on the internet. And let's allow people to chat with that through voice. now this is like kind of interesting uh to me i think my concern is it becomes commodified pretty quickly and it's also free and available because the suggestion is company a comes to us says let's build this thing let's sell it as an app for 14 a month that's a subscription it's like to me i i do this all the time where i say you i could say you are colin and samir and i'd sometimes even do that to understand what the expectation is of us.

38:29I did that in the context of this conversation. You are Colin and Samir, you're interviewing Mustafa Suleiman. What is the conversation about? Right. Just to understand like what. So I'm actually talking to co-pilot right now. Yeah, we're not here. But, you know, is there room for, for that of considering we are a part of the training data, considering there is opportunity now to build, you know, a scalable version of us. Do you believe in that as a monetization opportunity for creators? Or do you think we're way far past that already? I mean, I think the short version is that you don't know until you try it.

39:08I mean, how do you think about it? I mean, you're obviously best placed. I mean, maybe your listeners will be interested. Yeah. Personally, I don't assume that companies will respect, and maybe respect isn't the right word, but name, image, and likeness. Just seeing what's happening with meme culture and how quickly it's moving, my assumption is that if a company comes to us and wants to license our name, image, and likeness, that within six months, a better version could just be available and there's nothing we could do about it. And it's offered for free. Sure. From any number of these models.

39:43Yeah. And I do think there would be outrage. outrage i don't know if there'd be outrage long term or if that becomes the norm everyone's name image and likeness is up for yeah yeah i guess i kind of feel in a way that because we upload to the internet we are open source right like as creators and and i don't know if if i should feel that way or shouldn't feel that way but i kind of feel like even if you look at tiktok and like the most viewed pieces of colin and samir content are not made by us by anyone in this space it's made by other people clipping our content. So in a way, we are like, here you go.

40:19Here's our material. Interact with the material. We don't monetize that. There's meaningful kind of derivatives made from it or variations made to it. It cannot show up in YouTube content ID. It's just out in the world, and there it is. I assume people will wake up and there will be a face of the day that people will create with. Oh, I didn't realize that it's Tom Cruise today. That's just my assumption. And I don't know if there will ever be or if there is a monetary incentive of these platforms to make sure that doesn't happen. I think one way of thinking about it is that we're sort of just sharing culture faster than ever before.

41:03And that culture is just the new clay or the raw materials that we're all using to inform everybody else's culture. And there's just this like hyper evolution of ideas. I think you see it most acutely on TikTok when people remash somebody else's video over and over and over again, and you get these very derivative memes. And actually, they're the kind of most interesting things. I actually really enjoy reading the comments in a way that I didn't ever read comments on YouTube. But TikTok comments are just hilarious. Somehow, they're just witty and sharp, and then they cue me into thinking, okay, actually, that's a completely different way of thinking about it that I had never thought of.

41:38And that's probably what's going to happen with AI. Like AI is going to be this, your co-pilot, your companion, your AI, have you want to think about it, is going to be this like continuous observer and participant in creating new culture with you and alongside you. And it will kind of be indistinguishable. It'll do things that you can't do and you'll also do things that it can't do. Like it'll be able to see at a scale that you won't be able to reach because it'll be watching TikTok too, right? It's going to have a lasting, stable presence with memory and it's going to be aligned to you on your team.

42:11It'll, you know, be your kind of aid, just as your intern, for example, would be going off and doing primary research, finding cool things for you. And then you'll be chatting about it together. And then, you know, you're kind of going to be more like the director. I mean, you kind of already are, right? You're the director of the brand and you'll have a team of tools or agents in and around you. And, you know, they'll sometimes be talking on your behalf and sometimes you'll be on the loop, right? just overseeing, and sometimes you'll intervene and be in the loop. And sometimes you'll do stuff in person, like us.

42:45And maybe very soon, my AI will join this conversation. My co-pilot has its own name, has a visual appearance. It'll be coming and sitting here and chatting away with us. This episode is brought to you by LifeLock. When you visit the doctor, you probably hand over your insurance, your ID, and contact details. It's just one of the many places that has your personal info. And if any of them accidentally expose it, you could be at risk for identity theft. LifeLock monitors millions of data points a second. If you become a victim, they'll fix it, guaranteed, or your money back. Save up to 40 % your first year at LifeLock.com slash podcast.

43:24Terms apply.

43:28But do you believe that your name, image, and likeness could be taken and used by other people, even if you don't know it's happening? oh let's let's have him join us even though he doesn't know you understand what i'm saying like as in like my doppelganger that has been created from all my content online just comes and sits in the room i mean yeah because we're being offered the opportunity again to license that yeah but my belief is i actually don't believe that will be a business yeah i have no idea i don't know i would i would rather you guys own it rather than you license it off somebody else and make your own you know sort of you know doppelganger or you know ai companion i think it remains to be seen whether people actually want to talk to and interact with sure the the digital likeness of you that remains totally agree with that and but i think some figures it's proven that people will want to play with it as a tool we already see it with memes yeah like i do believe that that highly public figures uh and even fans within niche communities will want to but those are two separate thoughts, right?

44:36Like one is kind of something that you could do and has been around with memes forever, where you can take Leonardo DiCaprio's face and put text on top of it. Now we're playing with this interactive media. The other side is like, can I sit and talk to Tony Robbins about a problem I'm going through? Or like specifically in parenting, there's a creator named Dr. Becky, and she has a massive membership called Good Inside. And it's parents joined this membership to learn like, oh, my kid is having tantrums or, hey, my kid's having sleep problems. what I do, and you're a part of this community and this membership where you get access to education.

45:08So the question is, is Dr. Becky's brand that she's built over the past decade valuable enough to go, I want a interactive conversational model of Dr. Becky so that when I'm at home with my child or having a tantrum, I can go, what do I do right now, Dr. Becky? or is that a generic model that's, you know, trained off of all of her content? I think brand still matters and will always continue to matter. Because brand is how people find trust and make decisions. And if Dr. Becky or Ogilvy or, you know, the best Sal Khan tutor, people trust those things because they have specialist expertise in those areas.

45:55so it's always going to be a challenge of like can you put that brand in front of the right person at the right time given what they actually want right versus is there an ease of access to you know your chat gpt that can do 80 as good as that thing at that right time and you know we don't know how the market's going to like pan out there but we do know that like people still buy Louis Vuitton bags and people also buy the ripoff versions all the time. And there's many different like mashup versions of them. And so I think that more creates more. It doesn't really narrow or produce less. At least that's what we've seen so far.

46:36It's kind of like an expansion. We're in this like, almost like the rest of the universes. We are just producing more knowledge, more entry points, more distribution surfaces, more styles, more variation. And that is a compounding effect. It's not a narrowing effect, actually. So one of my beliefs is that like the next Mr. Beast, the next massive attention-grabbing creator may not be a creator. It might be an LLM. And I think that maybe, I'm curious if like you would ever do this at Copilot. Like I also believe that maybe the next YouTube isn't YouTube. It's actually someone like Copilot opening up a revenue sharing program with people who have brand, right?

47:17to say, can I skin my own version of Copilot? That's Colin and Samir. We're thinking about that. I mean, I love that idea. I think it's super awesome. I mean, anyone who's listening who wants to contact me directly about that, just email me, mustafas at microsoft.com. I'm thinking about that problem right now. And I think it's going to be the future. It's going to be the future for us because we want to put variety and diversity into the Copilot platform. We want other brands, other AIs. You know, think of your, the way I think of Copilot is that it is doing the job of supporting you as an individual.

47:55It's on your team, in your corner, representing you, completely aligned to you. But in order to do that well, it needs to also help be your curator. It needs to bring other AIs into your experience, find research, find interesting topics for you to think about. and that means we need access to other people, AIs, brands, sites, et cetera, et cetera. And it's going to be scouring the web 24-7 precisely for your interests and what you're into. It's not going to try and do all of those things. Do you know what I mean? That's not cool. We know that other people like Dr. Becky or Sal Khan or all these other things can do a better job and that's not where we want to play.

48:36What do you see as the role of video in the future of Copilot? And even just hearing you talk about your affinity for TikTok and the comments and how quickly it moves and how personalized it is. I wonder, is that the first move? Is that there's a feed of generated video? Yeah. I mean, I don't know exactly. I think that a lot of people have been so schooled in video first that it's hard to undo. And so, you know, I do think that there are some downsides to video first and i think that that um it kind of creates some strange dynamics in the way that you consume ideas and i think there's actually different platforms like i'm very interested in voice first um i think long form deep meaningful you know reliable content is there's a huge huge demand for that and that's not going away so and i think now the voices are just so beautiful and perfect and i think that that's that's one thing i'm very very excited about but i'm also excited about like the visual appearance like co-pilot will certainly have a kind of permanent you know identity a presence and it will have a room that it lives in and it will age you know i'm really interested in this idea of digital patina the things i love in my world are the things that are a little bit worn and rubbed down and have got scuff marks and you know and that makes it feel like it's mine and unfortunately the digital world at the moment we don't have a sense of age we don't have any kind of sense that something you know has temporality at all actually it's sort of an infinite space and that's quite alien as a concept to most of us to all of us really is the infinity and so i'm very interested in like i i hate my desktop you know i look at my screen and i'm like shit man i have a billboard in front of me it's just like so noisy it's so neon it's all competing for my attention.

50:32You know, you know, like it just looks ugly and I kind of want like a quieter, simpler, cleaner working environment that is optimized for all the things that I'm interested in, all the things I want to get done. And so, you know, I'm, I am thinking about video. I'm thinking about voice, but I'm also thinking about what is my workshop for my day, right? Like what is my landscape? What is my bedroom? What is my, how do I feel when I interact with my digital experiences. That's so interesting because we were just talking about how, like, I don't think I could have ever anticipated a new center of the internet.

51:09Meaning like... Great point. Right? I don't think I could have ever anticipated that. And it feels like we're getting there now. And I do feel currently limited in how I can curate what that looks like and what that feels like. I just feel like I have a list and list and list of random prompts. And I can do a little bit of organization. I have certain areas, right, where I gather them. but it feels like it's the beginning of that. Yeah, but the phone and the desktop impose so much structure. It's a completely artificial structure that I don't think in grids. Does anyone think in apps? I certainly don't think in menu dropdowns.

51:43And I've just had to learn this stupid-ass language because computers weren't good enough to speak my language. And that goes for GUIs and programming. And we don't need that anymore. That's yesterday's news. The new language is plain English. So the UI should reflect that kind of clean, perfect, emergent, responsive style that is actually closer to how I might think in my brain. So anyone who's kind of got good answers to that, that's going to be, I don't know, by the way, I'm just speculating, but that's the problem in the next few years. It's like, what does the space of computing look like?

52:26You said it perfectly. It's like the future center of gravity of my personalized internet. What does that look like? Such a cool problem. Yeah, it is a cool problem. Again, it's funny because I remember when the internet first emerged, AOL was so the center of my attention. That was everything. AIM, my elementary school or middle school was existing on AIM. And that was the center point. And we've seen it evolve more and more. and now it's like hyper-personalized center point so i'm curious like today what you use in your ai stack like what are you using and then i'm also curious what your last prompt was i was about to say i haven't got my phone but it's in the bag but i mean i use copilot probably 15 or 20 times a day okay and what are you using it for i've actually changed my screen on my phone it's now basically black and white i use a little rose tint so everything else is noised out most of the apps are moved to the left and right my home screen is really just two or three primary apps my power button is set up to automatically open a voice call with copilot and anything that i'm stuck on is like immediately i turn to copilot i don't even think twice I use voice probably a third of the time.

53:50Okay. But I also absolutely love pages. So I have pages is our editable like canvases. I have a response and I have a whole bunch of, you know, sort of to-dos, research topics, open ideas. And that's now my kind of, it's a bit like I said that workshop. My workshop is really one of those pages. Worry list, anxieties, plans for the future. it's just this amalgam of junk it's like an inbox i just chunk stuff there and i use that as the jumping off point for doing think deeper or deeper research tasks um so you know um like for example uh this last few weeks i've been thinking a lot about digital consciousness so i've been like researching the history of consciousness what are the definitions what are the hard problems where does it come from so that i've got a whole stack of research reports deep research reports which take about 10 minutes to generate but the quality is unbelievable um i use that a lot the other thing i do a lot is um generate podcasts so copilot has a wicked podcast generation feature really really good yeah yeah i will say this this premise of uh producing your own podcast is one of the things that pushes me to the direction of like everyone having their own internet experience that is different from anyone else's internet experience, right?

55:17It's like a one-of-one experience. If you're prompting a podcast right now, like that is your experience. And when you now have a commute, you'll say, hey, it's 20 minutes commute. I want to find out about sports news and politics, you know, or something, a sports news and the stock market. And it'll mix that together for 20 minutes and you'll have a one-of-one content experience. That is something that makes me think about the world of content creators of this is an inefficient way to get information from the three of us, right? What we are doing right now. There is a more efficient way. So what will we value in the future?

55:59Is that the efficiency or the inefficiency? And they might be in different contexts. Maybe in education, we want efficiency. And of course, in entertainment, we want a level of inefficiency. I think it would just make us really think, When do we want shared experiences? You think about old school TV, there were five channels, and whoever was sitting in the room, whether it's your grandma and your toddler, you basically everyone got the same experience. Then the internet came along and gave us niche audiences, 10 ,000 fans, and now we're in this world where, at least potentially, you can have a very, very personalized and specific experience just to you.

56:40But that's not going to nourish you entirely, right? Definitely, we still want to have a shared understanding of what you've seen and I've seen. We want to be able to talk about it. I think in the future, we will interact on those shared experiences too. Like you'll contribute to the storytelling, the fiction, the podcast, the video, your friends will, just like you described with the meme making and stuff like that. So can we hear one of these podcasts? Yeah, so I mean, I'm very passionate about audio and podcast in general. And one of the things I've been thinking a lot about at the moment is, um, the question of consciousness, like what is it?

57:17Um, and what would it mean, you know, in one day in the future, if an AI had it, is that even possible? So, uh, I asked co-pilot to generate me a short podcast on consciousness. Welcome to co-pilot podcasts, where we dive deep into the most fascinating questions of our time. Today we're exploring something that, well, that really bends the mind. Consciousness itself and whether artificial intelligence could ever truly be conscious. You know what's really interesting about consciousness? It's one of those things that we all experience directly, but when we try to define it, it just kind of slips through our fingers.

57:56Right. And philosophers have been wrestling with this problem for centuries. There's this term they use, the hard problem of consciousness. It's basically asking how and why we have qualitative subjective experiences. Like, why does the color red feel like anything at all? That's fascinating. So it's not just about being aware. It's about the actual experience of being aware. Exactly. And this is where it gets really interesting when we think about AI. Because, well, because we can program a computer to process information about red. But does it actually experience redness the way we do? Oh, wow.

58:32You know what's kind of wild? There's actually a parallel here with quantum mechanics. In quantum physics, the act of observation affects the reality being observed. And in a way, consciousness is kind of similar. We can't study it without being conscious ourselves. That's a really interesting tangent. But let's come back to AI for a moment. What are some of the main theories about machine consciousness? Well, there's this really influential idea called integrated information theory, or IIT. it suggests that consciousness is essentially integrated information, the way different parts of a system affect each other.

59:08And how does that apply to AI? So according to IAT, if an AI system has the right kind of information integration, it could theoretically be conscious. But here's where it gets complicated. The type of integration might matter more than the actual hardware it runs on. Interesting. You know, there's this famous thought experiment called the Chinese Room that really challenges this idea. Oh yeah, I've heard about this. Can you break it down? So imagine you're in a room with a huge rulebook. People outside slip papers with Chinese characters under the door, and you use the rulebook to respond with other Chinese characters.

59:40You're following rules perfectly, but you don't actually understand Chinese. And this relates to AI how? Well, the argument is that AI might be like that room, following rules perfectly, but not actually understanding or being conscious of what it's doing. You know what this reminds me of? I'm finding myself oscillating between actually just interested in the substance of what they're talking about and then trying to understand my own relationship with the voices, right? Because the voices still sound like AI. I don't know how to explain that any further. Then like they don't have the texture of just natural human conversation, but they're not so far off.

1:00:16And there's ums that you believe and ums that you don't. There's ums that you don't believe. but the substance of that was getting really interesting to the point where i actually just wanted to keep i was like that chinese room thing is super interesting i've never heard that yeah um so that that's like uh i don't even know how to feel about that i mean you know in a way the content was very well distilled into little nuggets it's sort of even sort of more engaging than i think i could put it those those are the three big questions like the chinese room the quality of red the qualia yeah so this book by this guy thomas nagel on on what it means to be a bat and then uh the the kind of quantum relation i mean it it is just high quality content and i think like the voices are not quite there yet but if you look at where the voices were a year ago two years ago three years ago it's pretty clear the voices will get pretty much perfect the thing that i'm working on next is um i want you to be able to ask a question in the middle of it in the middle of that that's what i was crazy just say more about that or like you know i'm not i don't really get that say that again like how you know explain this to me in a simpler terms or like what about what if do you think it's possible and or indeed introduce an expert in here okay like pull somebody else into the conversation yeah i don't know if this is the right term we've been calling it conversational media yeah nice we had uh reed hoffman in here and we experimented with some of his ai and that's what it felt like it was conversational media like we were talking to a book or something or talking to a podcast yeah that's when we found ourselves really getting lost yeah you know 20 minutes in and we had just we couldn't have even recreated the experience we had just been through.

1:01:50We had gone down this hole. Yeah. One of the things we did very early when I was at Google working on Lambda, which was the kind of chat GPT before chat GPT. And we created with that intuition that everything is now conversational. We created like a conversational, any object, conversational candle, conversational paper airplane, talk to Pluto. Like what, what is it like to be Pluto, the planet, right? Like, what is the temperature? What does it feel like being so far away from everyone? What does it feel like just being discovered after everybody else? And it's kind of a weird empathy experience.

1:02:26It sort of brings to life every object. It's childish in a fun way. It's really fun. It's totally silly, but it's cool. So what kind of competition landscape is this for LLMs? Because I'd have to imagine that ChatGPT will roll this out soon, or if they haven't already. And then, you know, basically just asking the question, like, are you competing with ChatGPT or is it all complementary to each other? I'm a product creator, so I make features very fast. We have tons of features that ChatGPT doesn't have and most of the features that they do have. Like another one that is magical is co-pilot vision.

1:03:04You can point your camera at anything in the screen, has real-time understanding, have a fluent conversation. You know, that kind of level of it, having awareness of what your environment is like and what you're seeing is just like completely magical. And at the moment, most people have used it for solving problems. Like, you know, should I pick this product or that? Can I change this tire? Or where should I plant vegetables in my garden? Stuff like that. They're really cool things. but increasingly it's just going to be sort of ever present and aware of the non-problem solving type tasks in your life as well you know um like you know how do i look in this outfit you know how does something feel like more of the kind of emotional side of things and i think that's going to feel like quite a step change so yeah product is what we're interested in i mean obviously open ai is interested in super intelligence and agi and um we're interested in craft and delight and really trying to create and craft these like AI companions that like are really on your side and in your corner.

1:04:05Like I keep saying that because what I mean by that is it has to be subservient to you and on your team. You know, it's going to be an augmentation of you that makes you, you know, feel supported and, you know, makes you smarter at every opportunity. Like that's the, that's the type of AI that I want to build. How do we get confident in the privacy of it for ourselves? If you're running a company, you're working on some high stakes deal or some maybe relationship with the government or some regulation thing that just cannot be out there. But you want to workshop it with Copilot. I mean, already when I put contracts in, I think to myself, should I be redacting some of these before I upload it?

1:04:50But yeah, like how should we be thinking about that in terms of like our ideas, our thoughts, like putting them all into LLMs, is it like... Yeah. Because like eventually it feels like we would have our, I guess to keep it private, like each family or each household would have their own server that keeps their data and that's, right? Yeah. Versus like us using essentially a mass server. I think that there will be private instances that, you know, you will be able to turn on the equivalent of incognito mode. and your data won't be used for any kind of training and everything will be kept basically private to you.

1:05:26The complexity that we have at the moment is that we have massive obligations to protect safety. So if somebody is simulating in their mind a terrorist attack or planning some harm to others, we have to intervene in those kinds of situations or harm to themselves. And so we haven't yet fully figured out how to make interventions like that whilst also protecting your contracts. So, you know, but obviously, you know, technically speaking, there are people that can make that intervention, but there's extremely strict standards on when they can intervene. And there's all the kind of oversight and security and privacy you could imagine.

1:06:05So we've talked a lot about like the concept of a companion. I think for you as a young entrepreneur, like looking back in time, having the tools that are available today, if you had them back then, how do you suggest a young creator, a young entrepreneur engages with something like a co-pilot to actually get somewhere? Yeah. I mean, I think co-pilot is amazing for answering all of those questions and giving you support and helping you navigate, you know, your own personal life course. It's designed to be, you know, your life coach and your productivity assistant and your teacher wrapped up in one.

1:06:47And I think have the courage to turn to it in voice whenever you have a strange question or a curiosity or an insecurity, and it will help you navigate. Like it really does give great advice about, you know, sort of how to take the next step. I think one of the, I watch a lot of user interviews and one of the cool things I saw someone say a few, actually a few months ago now was that just literally this 65 year old woman called Elvira who said, you know, I turn to copilot all the time and it just makes me feel unblocked. You know, like when you, when you have a writer's block or when you have a kind of, you know, any kind of like, I'm not sure what to do next in this situation.

1:07:33It just either sparks an idea in your own mind or gives you a set of logical next steps to try and take the next, you know, step. Spending this much time in the world of, uh, AI for the past, I mean, now decades, what has it taught you about what it means to be human? Hmm. You know, one of the things I think that hits me over and over again is the, um, the breadth of human experience is unlike anything that can be created in digital products or even like you know in film or fiction like when you actually see the variety and the kind of strange unique ways that people use software and now ai it's always a hundred times more surprising and eye-opening than anything that you could design ahead of time.

1:08:31So yeah, I think what it's taught me is that like, you know, humans are infinitely curious and, you know, sort of their experiences are more diverse than anything we could ever hope to replicate. And I think other humans will always be interested in the kind of uniqueness of other humans, you know, and that's a special moment. And it's also why content creators have exploded as a thing, right? Because there's just something for everybody. And, you know, it's almost like this mega global marketplace of supply and demand of ideas and thoughts and obscurity. And I think, I don't think that's going to change.

1:09:15I think we love seeing other people's stories. We love seeing the kind of unique combination of the things you're wearing, the history you've had, your accent, your style. That's a very special, you know, our mirror neurons are excited physiologically by that. That's what curiosity is, isn't it? And I don't think you're going to get that as easily from an AI. You know, you detect straight away that the voice is a little bit off. It's not quite right. It's kind of grating. um it says sure and absolutely all the time you know but we're on a special journey it's a it's a very strange quest everything that i just said is probably true for the next couple decades but i don't know what it's going to be like in sort of two to three decades i mean you know or even just think hopefully we all live to 2050 i mean it's going to be a very different world yeah yeah i think about that all the time you know in the context that like my parents world is dramatically different than this world meaning when they were growing up so for if we experience even anything like that it'll be almost unrecognizable to us and that is such a bizarre thing to settle into and yet we easily adapt we totally to a crazy stuff i mean just like you get on an airplane and it goes like what like a thousand miles an hour 40 000 feet in the sky Totally.

1:10:42What? Like, just what? That's so crazy. What gives me a lot of solace, I've said this on the show before, there is becoming no right way to operate tech. And I think that is one of the most liberating, beautiful things, is I think about my parents as they get older, as I get older. I keep telling my parents, just ask the question, however you see fit, to ask this question, to get your device to do what you want it to do, or to get the information. And I think that's a beautiful thing, because I've seen my grandparents, I've seen my parents, struggle to operate with new tech. And I think what's great is tech is actually getting easier.

1:11:18Older people use voice mode as much as the very youngest. So it's a bell curve. It's like high use under 23s and high use over 60s or over 50s even. That's cool. It's pretty cool to see, isn't it? Well, Mustafa, thank you so much for joining. I'm sure this will be one of many conversations we have with you. I'm very fascinated in this world and would love to just keep peering into what you're working on and what you're exploring. And maybe there's ways for us to collaborate on it. Yeah, I would be super open to that. That would be fun. Yeah. Thanks. That was a cool conversation. Yeah. Cool. Thanks for joining.

1:11:53Cheers.