Episode #23: Can Apple win the AI race?

And everybody, welcome back to the big cheese AI podcast. I'm the world's second best moderator joined by Sean highs, Jacob Wise and Brandon Corbin. We are the big cheese AI team, usual suspects here on the podcast. Today we are talking about Apple. Apple wins again. Everybody ring the bell. Well, maybe that's that's the question. So what's the Apple and AI? Yeah, has it happened? Is Siri not going to be dumb anymore? Are we going to finally miss out, unfortunately, on the amazing Brandon Rants on how bad Siri is because AI is now a part of Apple. Brandon, what's going on? Yeah, so so we don't have anything right now, right? So okay, let's actually backtrack a little bit, because there are things that Apple's been doing in the AI space before, even really the whole AI thing was a thing, right? So your, your photos, photos they've done a lot of AI imagery work there. And for the longest time, I mean, it's been four years, probably five years that you've been able to do the face detection, right, within all of your photos within the app. Yeah, if you wanted to, for example, and I, I always encourage people, they're like, Oh, I can't find that picture. I'm like, just search for it. Like, what do you mean? Like, Oh, it's your, he's like, Oh, I caught this fish last week. Well, then search fish. And it'll pull up every single picture that you've ever taken that has a fish in it. So they've definitely been doing a lot of work, kind of behind the scenes and subtly, but not being like, Hey, we're AI. And that seems to be kind of apples. And I just the thing though, it's like, half the crap that people are selling as AI, Apple just calls like a feature product. Well, this is exactly what they did with like the vision pro. They called it something else than what it was. It's not augmented virtual reality or augmented reality. It was it was it was a spatial UI or something like that computing. And it's like, yeah, they take the time to think about things at a whole nother level before they even consider going to market. And they also watch the market exist for a while before they they'll put a product out there. But I mean, if you have a trillion dollars and you can attract the best engineers in the world and you have a proven process built by, you know, Johnny Ives and Steve Jobs, you know, whatever, like you're gonna, it's just a different everything they do is just built differently. The salary, the salaries that these folks are making, it's kind of interesting. So Elon's been bitching that open AI has been recruiting X AI employees or no Tesla AI employees been pulling them over because they're offering like 900 G a year, right? Like, could you imagine making 900 G just on a salary? Just on a salary. Yeah, again, like, so I mean, you're your taxes are covered, your insurance is paid and you're getting your I wonder how HR feels when they get like the two week going to New Zealand notification from that from the 900 G guy. He's like, any like clocks out at like 330 every year. I run the TTO. It's like, I'm heading to the golf course. It's like, no, you're not you're working 900 hours a week. Yeah, I couldn't, I couldn't imagine. I mean, imagine opening up your paycheck and what 900,000 a year would be what like an eight seven, would that be 70 grand? A fucking paycheck by a week. Who would sign up an employment agreement? I mean, I guess that's how professional athletes are. But like, that's not the right business structure to make that much money in, right? Like you're paying income taxes, just straight down the tax on it. Okay, 900,000 a year, assuming 25% of that's taken out for taxes, some way shape or form. Yeah, your monthly, your monthly income is $52,000. Oh, yeah, their, their taxes are more than that. The taxes would be like 39%. Oh, yeah, probably. But still, not quite in that tax bracket yet. Sorry, I'm making 900 grand. So, so it's, it's interesting that there's, but no question, Apple again, Apple has all the money they could pay all these people or whatnot. They've been in the space. So we, we saw last week, we talked a little bit about some of the things that Siri can do, like again, with the sleep analysis and all that. So they have these like very small subtle integrations of AI into the platform. But I think that like what we were just talking about with realm, which is R E A L M. Is it real L M real? Or we call it realm? Well, let's, let's, but what you just said, take everyone needs to understand why this is why Apple wins is they have the touch points. They have the device. Yeah. They have your photos. They have the, they have, they have this, they're, they're, they're just, they're, they're building these features on that aren't even, they're not even calling AI that are just a, that are, it's going to be AI, but you won't know it. It's just going to seem really smart. And that's why when we talk about realm, why, why it's so smart. Yeah. So, but Siri, Siri right now is an idiot. Like really? Oh, really? You know this, right? We all know this. Like if you try to ask you anything beyond just like what's the weather today or what's on my calendar. So I was like, well, I can send it to your iPhone if you open it up or whatever the hell it is. Right? That's the one that I'm the most excited for is that we actually get a real L L M powering Siri that we can have these real kind of conversations, just very much like what you have with, with chat GBT's kind of interface. I think Google's doing this as well. And I think, well, my, my point was, yes, Apple's winning a lot. Their apps are better. I mean, the marketplace is better. And it's going to be easier for one of the examples I was watching a video on nobody knows what's going, what it's going to be yet. But one of the examples was I have 10 apps open. I've got Spotify and Amazon and text message. My friends says, Hey, can you buy X, Y and Z on Amazon or, or, you know, get one of these you, you're talking about getting a mic. No, I would never just be like a Siri. Order what Sean just asked me to order if it's a porn or whatever dollar mic. But if it was something innocuous, like I could see myself doing something like that, where it gets the ball rolling gets to Amazon and then removes the need for me to copy and paste. And Google's opportunity opportunity is more for like home automation. I have a ton of automated lights and things like that. And it's all using like that matter protocol now. But so eventually Apple can kind of connect to those things as well. But I want to be able to say, Hey, Google, turn on Netflix and replay the next episode of Brooklyn 99. Because right now I have four different ways to turn on Netflix on my my TV for some dumb reason. I have a Roku, a Chromecast off my Samsung TV. And then I could cast it via my actual TV. No Apple TV. No, no, I had it. And then I was like, I'm paying like 10 grand a month in service now. Here's one of my questions to you guys from a development side of things. One thing that makes Apple so amazing is that everyone uses their devices, right? So they're like Sean said, their distributions incredible. And when we start to look at, you know, people, individuals using it makes all sense in the world. That's kind of all the examples you go to businesses. And I mean, we've built products with Apple. We've tried to get, you know, the app approved, and they can slap like regulations on you. And they grab percentages of your revenue, if you're, you know, creating an app for the app store. I mean, if this concept of like, Oh, they're going to take over a chat GPT. Well, I think there's probably an advantage to still using a chat GPT for what they can offer in terms of like the open source and more customization of their API. Because I imagine if you're using like Apple's API, you're not going to be able to like create these like platforms. Yeah, I don't I don't think that's the market they're going to get into though is like they don't give a crap about serving up the general AI API market and trying to like build apps on top of their thing. I mean, have they ever even had an API? No, I mean, we'll try to have your tried to use the revenue cycle API's for it's garbage. Yeah. API's just aren't there. They're just like, man, no, but let's let's just talk about Apple success has always been around the fact that they don't have to build stuff for other people and have them screw it up. Yeah. So like their their software their operating systems were always built to run on their stuff or as something that they have approved. Yeah. So they're not like, Oh, Windows puts out a, you know, a version of of their operating system. And Dell comes out with some underpowered, you know, laptop that's running Intel's on blah, blah, blah, right. And the next thing you know that that somebody's having a bad user experience on Windows. Yeah, right. Apple's always been like, no, this is our stack, our our operating system that we built that you're going to have a a an experience on our device. Well, except for John Scully, when he was the CEO and he came in, he's like, yeah, we're going to we're going to let people white label them. And so you had a handful of manufacturers out there that we're building Mac clones. Really? Oh, yeah. This would have been in the 90s. So probably like 96, 97. Not jobs, not. Well, so as soon as jobs came in, he's like, fuck that. And he killed it. He killed the entire deal. But there were there were, and I, and I want to say it's power PC, but that's not it. That was what the old chipset was. But a power computing might have been the company, but there was like two or three different companies that could make basically Apple clones. And you could them buy them and whatnot. But again, it was the same exact problem. Those guys cheap out on certain things, the experience is bad. And that just tarnishes the overall brand. So as soon as jobs came in, killed the whole, the whole thing of the clones. I do have a bone to pick with M1 and M2s. Sometimes it just doesn't work, right? Because it's a different architecture. And right. Like that's. Yeah, that was, I was surprised that they did that, but obviously it was for a reason. Yeah. Well, they're, so you got to give them props that they're willing to commit, right? They're willing to be like, all right, we're going all in on this and everybody's sorry about just like we were, I was just talking about this with an engineer this morning that I was meeting with, he was on the product side on like the actual physical mechanical engineering. And we were talking about NASA in the 60s. And they, you know, it was like a, like a large percentage of the budget for the United States government was just given to the Apollo space program. And so they just literally dumped all this money every single year. And everyone knew that to go to the moon. And you look at it and you're like, that's kind of how Apple is. Like, Apple dedicates so much investment into what they, what they output that it's almost like, it's the, it's the, it's the only, it's like the brand that people just straight rely on, right? And you, and you, and you, and they, they have to produce, right? It's like, they have a trillion dollars and they put the ridiculous amount of it into R&D. And so they're, they're so well positioned to sit back, wait, watch people build watches, watch people build tablets, watch people build AI, and then eventually come into the game. And that's where you kind of get into everybody sitting there waiting, right? And now we get this paper that comes out that they released with their, their, Kevin, I can't even pronounce these guys. It's Kevin, Gue, Kenton, Lee, Zora, Tung, blah, blah, blah, like five engineers come out with this, this paper that Apple released and it's about, about retrieval augmented language model pre-training. Right. And if you read the paper, it's actually written not like an engineer would read, write a paper. I think they're running these things through chat. And in, of course, you can kind of see an angle where, you know, Apple likes to get patents for things so that they can only do them. You can see that they're basically saying that everybody, everything that everybody everybody's built so far is going to be completely useless once we come out with our thing. And the difference between what they are building is this inputless self-aware AI model. That's doing what you're talking about. It's going to basically be on your screen on your, it's going to ship with your computer. It's going to be on your screen. It's going to be seeing what you're doing. It's going to be hearing what's going on in the background as well. And it's going to be able to read and react. Right. That's incredible. I mean, that actually, going back to last week, because we were talking about like wearables, I mean, it was two weeks ago. And I was like, you know, it seems like the dream would be if it could just track you automatically, and you wouldn't have to, that's exactly what it is. That's what this is. Yeah. Yeah. Well, I think that the big key here is the context. Like one of the examples I saw online was like, hey, you do a Google search and you see that there's a pizza shop and you're like, well, okay, now I have to like scroll and find the one that's the best rating that's the closest to me and click the call button or see what their opening hours are. Or you could just say like, hey, from the results, find the one that's closest to me that has the best ratings that's open tonight at seven o'clock and give them a call real quick. And then it can like do that for you because it has context of the screen. I mean, maybe that's a bad example, but it's still interesting. It's a technology. Yeah. Right. That's the difference between what Apple does and what other people don't is like chat GBT implemented an interface to interact with their model that is the best one. But Apple's thinking about it as a complete as a this is a we're building a new technology here every time they always try to act like they did that. But they're the applications of these things they don't even they're not even professing to say yet. I think that the one they put in the paper that is like like the example is like you're looking at a document and it and it's going to know how to summarize the document. And it's supposed to all be some of the tech that we don't know. It's too early to tell yet, but it's supposed to be on device. Yes, it's on device technology. Are they going to have I missed? Is there going to be a dedicated chip that's just a AI chip or a battery just going to be completely drained? I think it'll be I think it'll be targeted on the neural engine part of their SOC. Okay. Right. So much that they all have that neural engine. And now you're talking about a decentralized computing infrastructure for AI, which is probably needed. Yeah. Right. So that it's not just all going to one place, you know. But have you guys been hearing that they've been talking to Google about potentially? Yeah. Which is kind of an interesting I'm wondering if they're going to have a general applicability of this technology and then figure out where their API's need to connect. So if you go back to before the iPhone, right? When they so they did a partnership with singular and so singular. Remember singular? Yeah. So Steve Jobs brings out the CEO of Singular and we're going to be partnering to build the singular phone. It's going to have access to the iTunes and have access to your music and all this stuff. The whole thing was awful. Like the entire experience was absolute dog shit. And then the next year they basically killed the whole deal and they release all their own phone, right? So that's it feels like we're kind of repeating this story, where it's like, hey, Google, we need you for some some things that we're not going to basically be building our own LLM to be at the end of the search engine for Safari by default. You know, that's a good question. They pay a lot of money for Google to be able to be on the Apple phone. She wrote. Yeah. So that may be a part of it. The article that I read was like literally like the second sentence was Apple and Google's long standing partnership. Yeah. I think that they're kind of buddies. Well, they pay Apple a shit ton of money though. It's like that's the thing is like they want to be. Yeah. They want to be the search engine of the iPhone, right? Because then you're just getting all that traffic. Now that being said though, I I kind of feel like it's Google. It's it's Google. I know but for whatever reason, I thought I remember hearing that Apple's looking to buy another search engine. And I hope it asks Jeeves. Did you ever okay so this so you used to be able to go and say is Jeeves gay? It would be like I prefer Jovial. That was the way it answered. That was like one of the best Easter eggs that I had ever seen because a lot of people apparently were asking if Jeeves is gay. And so now he would reply that I prefer Jovial. I don't like that's awesome. Now they're kind of like privacy concerns that we should be thinking about when it comes to Well, that's what I like it. So so with Apple's whole concept of like doing a lot of things on device, having a lot of end-end encryption, you know, that's kind of their their stick, right? Like where Google everything has to be readable so they can give you the ads. Google's kind of like the antithesis of that saying, you know what, we don't want to have anything to you. So if these large language models and realm, which is what they're there, I've seen a couple different names, but realm is kind of the name that everybody's referring to. If that's truly just running on device, then that just gives you that I mean, that's no different than a llama or the ones that were running just on our computers to be able to have like inference happening, you know, just locally. Now, no question in my mind though, when they start releasing these features that it's only going to be available for the iPhone 16. Yeah, the next iPhone 17, you know, and all that that they're gonna read. Yeah, they're going to use this. They're absolutely going to use this as means of basically saying, I'm sorry, your 15 is not strong enough to be able to handle large language models. So you don't have that access to that feature that they'll use AI, the AI features as a means to basically sell more hardware. I think the general applicability of this AI is something that has never been done before. If I'm a multi Apple device user, if I have a watch, if I have this, I have my iPad, and they pretty much every digital interaction that I have is occurring on this device. And if it's screen aware, it's going to basically have a database of everything that I'm doing, and it's going to be able to read and react to that. So if it might be able to figure out if I'm in a bad mood, if I'm not even working today, it might be able to be like, Hey, Brandon, or you in a funk, because you haven't moved, you've all you've been on TikTok for the last 12 hours. Right. Like, hey, hey, you might want to, you know, how it's like, hey, you can still do it, bud, go to the goals, those rings, it's going to be like, stop drinking, you idiot. Now that's just this whole concept, I mean, how many companies is Apple going to kill with that? I mean, how many apps have it like all these applications that now the core iPhone can do? That's, but that's, that's their thing, right? They go and they release an app store, they find a couple, they sit back and they look like this app is blowing up. Everybody's using it. They're like, that'd be pretty good to have as a feature within iOS. And then they just go and they send their developers making 500 G a year, and they go and they build a better version of it. And the app developers are just like, what happened? Yeah, so they'll do it again. And they'll just kill all those apps that were just special. So it's safe to say that we actually don't, no one actually really knows what Apple's doing with AI. We just know Apple's doing it. Right. They have a lot of. So yeah, pretty much everything that we're going to be talking about today is all just rumor, right? Like, or news that we're hearing about, because there's realm, there's their image, their image model that they did, which was like for interacting with and editing like images on the fly. So I think we'll see that being incorporated with an iPhoto. It wouldn't surprise me in the least if Apple is going to be treating AI at a much more use case focus versus open AI, which wants to try to kind of come up with like a generalized model that can do all the things that we're going to see health kit is going to have AI incorporated into it. No reason why you wouldn't, right? Like all that data is there. I should be able to ask be like, Hey, what things do I do that increase my resting heart rate, right? And it should have access to that same with the wallet, right? There's no question that we're going to have a finance AI that will be based on the wallet data that we have and that Apple probably will make it so you can start incorporating all of your other accounts and just basically to be able to get to here's my entire because finance is one that you and I've been talking about a lot about doing an AI and finance podcast. That could be really interesting. But yeah, I think Apple could just absolutely clean clock there. Well, that's the other thing they talked about in that paper is they they say that this the model the underlying model of this that they're using or that they built is is faster than chat GPT. Yeah, and more accurate and it's multi modal and it's on device and it's on and it's on device. So like, which is crazy, they're making some pretty insane claims. Yeah, they're saying it's better faster. Right. And I mean, if you think about like when chat GPT is running on the hardware that chat GPT runs on would just blow your mind, right? Like the cost of its ridiculous. So to say that they're getting even anything remotely close to a chat GPT for or sorry, GPT for experience on device only is insane. Well, and just compare this to the Apple Vision Pro. So was we can all agree probably that the strap on headset VR market wasn't necessarily like this life changing thing, right? Like, were you guys out there? Like the next ones to buy that? Any of you guys have one before? I did buy a meta. You did? Yeah, I had one. You had one. I played some games on it for a couple of weeks. No, it was the game was like a shooting game. It's awesome. I only played the game for about three or four weeks. And now it collected. I think I actually true story. It was in my trunk for so long. And I was driving through the mics car wash. And I was just like, it's a bit in my car. Everyone to take it out. I looked over to the mics car wash guy said, dude, do you want a meta quest? No, you didn't. He said, I'd love one. I said, I'm opening my trunk. Grab it. No, I swear to you. I gave it to him. That's what happened to my meta quest. But like, so, so, so when you look at Apple last year, it was all it was mostly about this new product launch that they were going to do from their stocks perspective. So if you're looking at this from an investment state, I would say this is beyond that that times 100. Yeah, because maybe a thousand. Yeah, because it's going to be incorporated in everything. So it'll be in the quest or in the in the headset. We're not talking about the touch bar here. Yeah. We're talking. We're talking about a completely transformational piece of technology that's that's going to take a already mainstream technology in terms of LLM's. Yeah. And bringing it to a new generation of devices. I'm gonna I'm buying. Isn't this like the rabbit? What was that one? Oh, a rabbit or isn't this rabbit or one, but on your phone? Because we talked about how great that was that it's like trying to connect all these apps together. But it's still fundamentally like, it's not your phone. Well, any guy with a, you know, a few hundred thousand, a few million exactly about a trillion. I haven't talked to anybody who's got one of the actual rabbits. Yeah, but it's the same thing, right? Like what would stop Apple from saying, all right, we're gonna build, we're gonna train an LLM on the UI of the apps that would exist within our our ecosystem. I think we talked about this early, early on too, as like, this is probably where this is going. But it's just funny that like, it does always amaze me like Google is, you're talking about Apple's position. I think Google is also positioned to do a lot of these things. But it does always surprise me how Apple gets to these things. And it's like, all right, what's going on over at Google? I agree. I think it's just the difference in the companies. So I had two points. I'll say this point a second. The first point was, or question to you guys, was Google and Apple are both in the race. There is this concept that's brewing around legislation regarding AI. I would have to imagine that both of those companies are going, once they get big, put their foot in the race, or whatever, that they're going to be basically dictating what that legislation has to say about a question there. Like, if there's anything there at all, like, do you think they're going to say, hey, it needs to be like, because Apple's a big one, privacy, anything. It's not going to happen this year, because it's an election year. They just want to talk about wedge issues. Like the more lasting Trump or Biden want to do is look like they're stopping capitalism. Right. But I will say, when it comes to Apple and Google, I think the fundamental difference between Apple and Google is Apple looks to things from a hardware perspective, and Google looks at things from a software perspective. Google got into the phone game because they had to, not because they knew anything about making phones. Apple's got this. They're so much more primed to build the first mainstream device that ships with AI on it. Whether that's a good thing, whether that'll work, I don't know. But there's also a partnership there where Apple doesn't necessarily lead with, right? So they might need that API or that software. So I think it's just the rich getting richer. I mean, they're all getting 900 years. Again, we talked about this a lot early on, which is like the people who are best positioned. Open AI, great, they're first mover and all that. But like, there's two big companies, three big companies that have a lot of users. A lot of users already. They're going to win. It's a foregone conclusion. It's just, in what capacity, how are they going to cut up that pie? And what's that in-store going to be? I mean, from the way that I think about, you know, Sessam Altman who came through Y Combinator, right? Yeah. When do I come in? What's the end game in Y Combinator business? Continue to own Stripe. Okay. My answer was going to be acquisition, right? Sam Altman's the type of guy that to me seems like he would exit that business, right? And maybe that, you know, it's just going to happen pretty quickly. I think when it all comes down to it, like it's like the race is going to come down to just a few horses and, you know, is it going to be an AT&T Verizon thing or is it just going to be, I mean, it could be an Apple Samsung game. So Samsung, I mean, Samsung's doing some really interesting stuff. They're already adding a lot of AI into their products. I mean, what are we on the Samsung Galaxy 34? I know. Does anybody have a Galaxy? So I've got like, I think I've got a Galaxy 20 that I use just for more for testing. I've got like a 10 that I use for the old, old school Android testing. I've got one somewhere around here. So I mean, again, I've been an Apple person since I was like a wee lad. Like I was eight years old and I'm like, this is amazing. Like, I don't have to do DOS. This is fucking great. I love DOS. Eat a dick. I used to change around my RAM so I could play like Carmen Sandiega and chop it on me. Oh, dude, what a great, so again, tons of DOS games. Like you guys just won't appreciate it that to launching an app or you had to be like, I forget even what the command was open. Exactly. Yeah, whatever not. And it was, yeah, it was great. And the Carmen Sandiega, Indiana Jones had a budget like that game was good. The fate of Atlantis. Oh, yeah. Did you guys have Dick Tracy at all? Oh, hell, yeah, I didn't have that. You could grill him. You could grill him. Right. Yeah. But where in the world in Carmen Sandiega, if you haven't played same with Oregon, Oregon. But here's a gem. If you really want to get Carmen Sandiego, where in space is Carmen Sandiega is one of the greatest games ever. You learned so much about space. That's cool. Yeah. So here's another learning. Now, so the story was is that so my dad owned a print shop in 1985, '84, late '84, '85 ended up buying one of the first Macs for the preprint department. And so he dropped me off or I went with him one weekend and he's like, check out this Mac. I get in and I find Mac paint. And I'm just like, I'm eight years old. And I'm just like, this is the most amazing. My mind is blown. I'm making fly years. I'm coming up with all sorts of shit. I'm printing things out. And it was absolutely the most transformational thing as a young lad. So he would drop me off on the weekends and I would go and I would just play on these computers and do a bunch of stuff. But in the back of some magazine, I found someone selling software for Mac and for Mac NPCs. And it was like, I forget the name of it, but it was some horror game. And I'm like, all right. And I managed to get 20 bucks, put it in an envelope, mail it out, end up like, you know, a year later or whatever it was. I get some disc back. Dude, I put it in and it's a role playing game. But you can like, it was the most horrific, like no eight year old should have had their hands on this. Now you know, like, you're such a fucker. My wife would be like, yeah, that that tracks. Also, what? So you just mailed someone cash. Oh, yeah, that was the way it was. Well, like back at Boys Life magazine, you wanted that throwing star. Yeah, you had to send out like literally a cat. You just had to send money in the mail. And you, it was always a 50 50 shot if they were just going to screw you or not. And yeah, every once in a while I did, I ended up getting some throwing stars in the back of Boys, Boys Life magazine, took it to like fifth grade this last year, fifth grade, got caught by the principal with my throwing stars. He's like, I'm taking these like, if you want them back, you can have your parents come get them. I'm like, well, I'm never seeing those things again. Sorry, sidetracked to Brandon's youth you were talking about. For that, couldn't tell you. Round. So I guess I guess I have a different question. If I could take us on a sidetrack here, 23 episodes in talking about all things AI, right? Health and AI, data privacy. I mean, you name it. We've probably talked about some essence of the beginning. We talked about the fundamentals of AI. And we've been doing this for, I don't know, almost four months and what in your guys's opinions is happening in the AI space and what practically should people be looking at to just learn, just get everybody up to speed on what your perspective is on AI. Over the last four years, how has it changed? I guess would probably be the question, right? Yeah. Like, how is this evolving? Well, if you see an in-game in this. First of all, I would say all the startups that are AI focused, that money is drying up. And if there's more like expectation of like, okay, there was a lot of hope and a lot of just dreams out there, you know, six months ago, a year ago, now it's like, all right, how are you going to make money? And it's becoming more clear or clear than ever to me that the big players are the ones that are going to produce the LLMs. I know there are a few open source ones that are doing pretty well. Yeah. And there's some profit and markets there, but it's the big players that are going to win the kind of big game, long game there. The smaller people have to figure out how they're going to participate in that ecosystem and not generate their own thing. It's just, I don't think it's realistic, right? I think the thing I'm seeing that I've learned is the model is going to become eventually a commodity. And it's the implementation and access to that model that will be what you buy. Right. Because again, if I can chat with Apple's realm or whatever it is, do I ever really need to go to chat GPT? Right? Like if it's just available, right on my device, on all my devices, on my computer, on my laptop, on my iPad, do I need to interact with some of those others? Same with like the original big cheese app that we built, right? There's no question that these things are just going to continue to get more focused and you're going to have these people that are just going to be available to you. Do you need co-pilots? Do you need that when it's all just built directly into the operating system? And yeah, so again, I think the big tech players always just end up winning in the long run. Of course, there's going to be your stripes or your unicorns that come out that are just taking a different approach that the moons and the stars line up. They end up raising, you know, 80 million bucks and just build something amazing. But a lot of times, yeah, I think it's just going to be, you know, these big companies who have the most money that they can build these models and train these models and they're going to be great, then no startup's going to be able to compete. Now, having said that, I want to actually contradict myself a little bit because I think there's like the large language model game where there's only the big players will survive. And then what about this? We haven't really talked a lot about the small language model or something that's way hyper focused on trained data, essentially just training your own model for a specific use case. And I think that's probably the next chapter that we're going to start to see everyone's familiarize. I don't think though that it will be as much of a global topic because it's very much so like what prompt privacy is doing. So prompt privacy's Publius model is very specifically trained on being able to interact with enterprise data. So it's like you can't go and ask it, you know, tell me about how to build a nuclear bomb because it's just not trained on that. But if you can say here's a document that I have, tell me what type of document it is, what's relevant about this kind of contract, what are all the attributes and blah, blah, blah. So they have built their own foundational model that's very specific. So I think that's where it goes. So to jump in, go ahead, Jacob. No, I was just going to say, I think what we'll see from that is an extreme productivity jump at the corporate corporate level. Think about like, if you have to ask legal for a document, no, yes, you asked Sally for a document. Sally doesn't know if legal, legally, she's allowed to give you that document. So she has to go to legal. There's a four or five day wait period there. Well, now if you have it all, I mean, it's going to be a lot of work to get there. But if you have it all, all these systems intertwined, it's 10 seconds you can have that document, not 10 days or four days or whatever. Yeah, I think from buy side of things, I agree with all the points, of course. So from the top, yes, I think that AI was VC's tech industry had just plummeted from 2022 or 2021, 2022 started to go down 2023. It's been a was a wash. 2024 is not looking any better. And AI was like the saving grace to pull interest rates back what they need to be. VC's can go back and raise money from LPs. So that happened. And then so a bunch of money gets invested in AI companies. I think that to your guys' point, yes. So the reason that people are going to talk about it on privacy is the same reason people don't talk about a jobber, right? Because it's a very it's very beneficial, but it's not a big deal. That's not something you're going to bring up unless we talk about like something that big cheese pulled on weekly. So yeah, I agree with that. Like if we hone in the models, and this goes back and every talk with Aubrey, hone in the models specifically to one organization, that's probably the next version of the SaaS company in terms of like really leveraging AI and growing and generating good revenue. It seems to me, though, that there might be the winner of that of that next chapter, if that chapter is the case, is the is the vendor of the databases of the data. The person selling the data to all the companies that need to train a model for their specific niche, person selling that one, what are they selling? They're selling the easy access because they it's the endpoints. The endpoints are important. And then and that's you know, just it's like a Trojan horse for your for your next thing. That's why that's why the grammarly thing. I keep going back to that. Grammarly has been advertising like unbelievable bonanza on the NCAA tournament. And why? It's because they now have an AI offering on their product and somebody just gave them a bunch of cash. And they're like, yeah, we already have like a billion, a billion users and where I've got Chrome extension installed on them. Let's get them to pay more money. Yeah. You know, let's let's go big for me that. So, so those types of companies, the grammarlies, the ones that kind of came out pre AI that offer solutions that are still AI driven. If I was their executives, I would be shitting my pants, right? Because everybody's gone and for now, because like before building a Grammarly was hard, you know, right? Now building a Grammarly competitor is as much as literally taken a single string and passed it. Well, then I think there's two contexts. We talk about this. We talk about this is who's the overall winner? And who's who's going to get a piece of the pie? Right. I got a question. Go for it. Got a question. And I have an idea. So, if my going back to my little theory is and I'll wrap it up. But my brain is now just going on a tangent here. So, could you create a company? And this is this, I need to know if this is even feasible from training a model's perspective. So, if you created a company that the job of that company is we went out and we partnered with businesses and through some sort of a partnership and engagement or say, hey, look, we'll be the vendor of your data for potential companies that want to train their models on your data. Right. It's all private with privatize, whatever. But whenever we sell that, you get a cut of it. Right. And so, could there be a business that goes out, partners with all these large organizations, it becomes the vendor of making those businesses data accessible because the likelihood that that business is going to be like from a bureaucracy's perspective, like a lily is going to be like, oh, yeah, we're going to create a marketplace. We're getting a marketplace for the data. Well, you go back to Aubrey's initial conversation, which is it's really hard to go in there as a vendor and unlock these large organizations data sets. That's why I liked your idea about just like literally the database company. Like the company that's literally they're outsourcing their SQL server too. You know what I mean? It's latching on that technology. So, that's it. That even makes sense to be possible. No, I think it does. And so, it's funny, and I've talked about it a little bit with prompt privacy, is the cognitive storage engine that they've been building, which is a hardware play that basically can go in, you add your Google service accounts, you add all of your different service accounts, and it can go in and personate each individual that works in the company, starts ingesting all of the data. I finally got access to the API this week to start doing, and I'm able to literally search my Gmail, like all my emails that exist and anything that's on Google Drive, anything that exists. I want that so bad. You know, you just, oh, it's dude, it's mind-blowing. Here, I think there's a couple opportunities. One, there's probably bigger companies. The big corporations are going to do it internally, because it's either too sensitive, too many legal issues. They will just hire it. Eventually, there might be a prompt privacy that gets big enough that they're like, okay, these guys are in the marketplace, we trust them or whatever. Prompt privacy is going to be after the big organizations that don't want to do it internally. But I think there's also a mid-market play where there's the companies that need the people who actually implement the service. I think about all the HubSpot, all the Salesforce service providers, right? Well, why is there a HubSpot service provider? I totally agree with you, and I think that one thing that we do is we look at things like at such like a discrete level, because we get this understanding, and we're like, oh, gosh, well, that's the way to do it. Apple. That's the way I'm going to interact with the AI. But like, there's other problems out there that aren't just your problems. And there are tons of opportunities out there. What's here? Point is like, we are a company of 14 people, and it would be great if we had better docs. We have all sorts of problems. We've got processes. We've got the way the data's put in there. There's someone, there's an opportunity there for like the 100 to 300 employee companies, someone to come in and be like, here's, you were talking about standards and processes of like, this is how your data is going to be input. Stuff like that is going to start to bubble up. And these are just new companies that will exist. And it's a combination of like existing standards and existing processes and the new technology. And you kind of bring those together. I do. I think that like a lot of these enterprises are ultimately going to benefit by just saying, okay, we're just going to align with a specific spec. That's going to be our standard. And now anything that we do from an AI needs to just make sure we're following those specs. But which also has really positive downstream effects. If all your data is now, because you got to do a lot of work, the reason that you don't do it is because it's hard and the juice isn't worth the squeeze. But if it does, all of a sudden, start to make more sense to organize and categorize and do things a certain way, well, the downstream positive effects are that you can now interact or interface with way more softwares or whatever you need to do. Yeah. So I got two things. So we talked about Apple. Their big lagging problem is their lack of a existing, referenceable, stable model. They're talking about this model that they did that's the best, right? But you can switch any model another. What is a realm? We're talking about realm. Yeah. And what does it stand for again? It stands for it stands for retrieval augmented language model pre-training. Most of the time they spend talking about in this paper is the effect that existing models have limited modalities, if you will. They're sure they can do text to this thing. But what about just watching their screen the whole time they're using their computer? But there was a couple different things that came out of my research that I thought was interesting. We talked about their partnership with Google, but there's two companies that are rumored to potentially be a huge part of Apple's future. One is perplexity. So I wanted to get your take on perplexity because there's a specific thing I like about perplexity. Don't let me forget to say that. Well, that is. Go ahead. So it was funny is that it was and it was on April 1st that I saw it that Apple was buying perplexity. And I'm like, that makes all the sense to me in the world. And then I delve deeper and I was like, you're a pro fool! But why does it make sense for Apple to buy perplexity? Because they've already got it. They've got a lot of it figured out. And Apple's sitting on a trillion dollars on the bank, just go buy these companies and just consume them. Yeah, but perplexity has the same problem, not maybe not as deep, but the way they source their data, it's not good for the content producers. But they do source it. So that's the thing right? Because they source it by stealing. I know. But so the difference between a perplexity and a find and chat GBT is that when you search on their system, it'll tell you where they got the info from. And you can link back to it technically. Which I've always appreciated. And I think their CEO is on hard fork a couple weeks ago, but I think that they're looking at different ways to monetize that and to actually pay the people that they're showing the information to. For example, it would make sense for the New York Times. If I have a question in perplexity, it shows me in New York. So let's just so for the people who are listening all the way to the end of this. What's perplexity? It's a new search engine. It's basically an AI powered search engine that you type in your query, and it goes and finds a result searches for the results on the internet. It's not searching NALLM. It's searching for the results, and then consuming that, putting it in a rag. Or for the results back as a response. So it's actually doing what I think we want AI to be doing. It's like, hey, go do the search and then research all the top 10 results in something. I like that kind of stuff. I do too. I do too a lot. I think what was an interesting model was like, okay, if it's showing a bunch of New York Times, if a user goes there and sees 10 New York Times results over time, it's like, hey, this person's high intent. He probably likes the New York Times. Well, inside of the app, they could offer something that's subscribed to the New York Times right now because you keep reading and getting results that are relevant or sourced by them. But of course, that doesn't exist today, and it would only exist for the major players similar to how Google, you can buy things directly from Google products. What did you have to do? You had to put that source code in your website. You had to tie a bunch of things together to make that work, but they're just not big enough, and there's not enough players, I guess. But anyway. If you just go to our app, then the whole web makes sense and system. The other company that is rumored, which I think is maybe a little bit, maybe more far-fetched is anthropic. Mm-hmm. So talk to me about, anthropic is a model producing company. They have a bunch of open AI guys that went off and did their own thing because they were concerned about the usage. Yeah, I do. I think a lot of them came from open AI. Yeah, I mean, as far as model production goes, the people over at anthropic are producing some of the best. Three Opus is really, really impressive. I mean, again, we all compare everything to like ChatGPT-4, and even Sam Moltman on Lex Friedman's podcast was like, it sucks. He's like, ChatGPT-4 sucks. I'm really hopeful that the next one that we release is going to be as dramatic as going from three, five to four. Considering that right now, I can go to ChatGPT-4 and have a very in-depth conversation about framework-specific code. I'm not even talking about programming language-specific code, Dre. I'm talking about frameworks that exist within programming languages. Really? Oh, yeah. And you can cross the frameworks. You can go to ChatGPT-4 right now and say, I really kind of just want you to talk. Here's my stack. I'm building in next JS and tailwind, and I'm using the 13+, and I might be like, oh, dirt, dirt, dirt. I might not know what that, but they do. Just tell them to search Bing. Yeah. Like that's my hack. Exactly. Okay. Stop acting like you don't know this. Just go search it. You first. And then come back, which it can do. So the thing that perplexity was kind of doing. Yeah. But anyways, I mean, the fact that Sam Moltman says that that that it sucks is insane. So we need to do an episode of ChatGPT-5 rumors, because I know that they're trying to drop this bomb. And maybe that's the way that they can stay afloat, is just by just crushing- Every time, just a new release- Crushing- Oh, yeah, absolutely the best. I mean, if you're Altman, you've been through multiple exits. You've been through multiple wins. Maybe this is the one you stick with through the end. Yeah, but there's no doubt that eventually somebody just buys it, right? Or they end up just getting too big that nobody can really afford to buy it. But again, a Microsoft seems like the most logical acquirer of OpenAI. I mean, they've already put 10 whatever billion dollars into it, where then you could see Amazon who just put like one or two billion into enthropic just says, all right, Skirt, we're just buying them buying them. Two more things that I got because I think this is good. Are we overplaying Apple because, I mean, they're really good at building UI like that people love to use, and they're really good at building core operating system stuff. Is the application of this AI? Are they going to build something that seems like it works really well, but just can't give you the weather on the right side? You know what I mean? Like, are they going to build another like, is it going to be like, it takes Siri for 10 seconds to respond now, or are they going to trip over themselves? I think they absolutely are going to trip over themselves. But I think that Apple, because Apple's obsessive, so they have this obsessive attention to the un-obvious, right, which I think is a very important kind of key thing that most people don't think about that. There's just a lot of subtle things that they put a lot of focus on that are very important to us as humans as we're interacting with technology that other places just don't even think about. And that was Johnny Ives was the obsessive attention to the un-obvious, his whole MO. Now, he's obviously gone doing his love design or whatever the hell that name of that company is. But I think what's going to happen is they're going to roll it out, they're going to stumble, and then they're just going to keep iterating and improving and making it better and better and better, and eventually win, right? Apple Maps is the story. So, if you remember when Apple Maps first came out, people were driving into lakes, like people are driving off of bridges, you know, because like, they're maps, it was just wrong. I mean, it was wrong on like every friggin level. And now, when you go and you compare Apple Maps, even Google Maps, Apple Maps, in my humble opinion, is better, right? It looks better. Because every time you need to make a turn, your Apple watch vibrates. What the hell just happened? Versus versus Waze, which tells you literally like two feet. My wife hates Waze so much. I know. And I just like, I've literally gotten in almost multiple car accidents. Exactly, because of the perspective on the turn. Yeah. You hear or notice that? It's horrible. And Waze was like, I was a Waze guy. So, I was a Waze guy before Google bought Waze. And I'm like, Waze is the absolute best navigation app, because it's redirecting you based on intelligence. But this is what I mean by these companies who are launching these AIs. So again, Waze had a lot of intelligence. And that's really what we're talking about when we talk about AI is just intelligence within software. And Waze really did a great job. Then Google buys them. And then slowly and surely, Apple just keeps iterating, just like the turtis turtle turtle in the hair. Yeah, right. And they just keep iterating, improving, improving, improving. And now I can go and like drop things and like 3D. It's like, I'm playing a fucking video game. And I could see like downtown, I could Google. Some of these Google software applications are the most mature in the entire world and stand the test of time. They do, but then they kind of rest on those laurels, right? And they just be like, yeah, it's working. We're good to go. We don't need any more improvement. Level DB is a great example of that, right? So level DB is this like low level JavaScript database that you can leverage. And so I find it, I'm like, this is amazing. And I look, the last time it was updated, it was like seven years ago. And I'm like, does anybody still you still waiting for everybody? So I'm still waiting for Google Wave 2.0. Oh, right. Hey, by the way, this is a good quote. So this is this is everyone's asking about what the hell is Rome stand for? This is what it means. This is the this is in the abstract of Apple's paper that they just put out. They say, wow, LLMs have been shown to be extremely powerful for a variety of tasks. There are use, there are use in reference resolution, pertific, particularly for non conversational entities and remains underutilized. They're basically saying their technology that they're going to build isn't around a conversation. It's around a context that it's going to be ever present and available for all with the screen and the screen being the screen is the context. And guess what? It's got all of our screens. Exactly. Got all the screens. Yeah. Well, I'm still waiting on a response from the fat API for the old nutrition platform with it all in. Never got a response from their team. I don't think there's a lot of clue. Anyways, guys, this is the biggest AI podcast that you guys so much are making it to the end next week.