Humane Pins and your own ChatGPT

25:27Yeah. Right, but then there's the Google logo sitting right in their press release, and I don't know why. Then they claim they have their own platform called MyAI. Maybe it is all just chat GBT. And it's all cloud-based, right? Yeah, which I think is also why it burns through batteries. Yeah, well, it burns through batteries, and it also—that just cuts off a huge user base. There's a ton of people who can't have ready access to the cloud all the time, right? Like, internet is inconsistent. They weren't going to buy it anyway. Yeah, you think the people in rural America are— Are you saying they were not watching the Caperni show to see it?

26:00Like, oh, my God. It's running on the T-Mobile network. Those people were already left out.

26:08It's just like even that stuff, right? Primarily the ChatGPT API. That's what we think it is. But we don't know. No one's actually used it to figure out its limits. No one's actually used it to figure out if when you ask it for directions, it falls back to Google Maps or something. or Bing Maps or whatever acquisition like Microsoft made through the Nokia deal 500 years ago that they forgot about that still runs and somehow makes$5 million, which is the true story of many Microsoft acquisitions. And because it doesn't have a screen, you, the user, will never know. Like you will ask it a question and it can never tell you how it arrived at the answer, which has been a problem with voice assistants in general for a very long time.

26:49It could list sources, but still, like, Like, I don't, we'll talk more about this later. I don't trust ChatGPT for like information because it hallucinates. And so if you're thinking day, like tactically, this is my new phone. This is how I'm going to interact with the internet. The main way that you're doing that, you cannot inherently trust what is being said back to you. How is that a replacement for the devices we already have? It doesn't make sense. Yeah, there's this interesting moment in the demo video that they put out, which everyone should watch. It's wild. Neil has suggested to me it was to watch it at 2x speed at which it seems like normal speed which is good advice.

27:24But there's this one moment in it where he holds up I think a book to the pin and says how much does this cost online? He's like demonstrating the shopping features and he's like how much does this cost online? And it responds it costs$28 online and he says buy it. And that's the end of the demo. And it's like I have a million questions between here and there. Like there's only one price on the whole internet for this thing? Where did that price come from? Where am I buying it from? How fast is it going to ship? Where is it going to ship? Which credit card did it use? Like Amazon has gone through this problem where if you just say to your Echo speaker, buy toilet paper, it turns out that actually takes you through a flow that you don't understand and can go wrong in a thousand ways and people don't do it as a result.

28:07And the idea that Humane is going for is to abstract even more of that away and to just say all you have to do is just describe to this thing what you want to accomplish in the world and it will do it for you. And I think that's a super interesting and completely impossible thesis. The price online is$28. It's like, is it Amazon? Do you think Amazon is the only store? Who knows? eBay? Right. Who knows? Maybe it's an artisanal. And it's like, by doing this, am I signing up for whoever Humane has decided to, like, rev share, partner with in order to be able to do this? Should I be able to choose?

28:44And at that point, if I'm choosing, what are we accomplishing by abstracting all of this away? There's just no version of this that feels good in the way that you kind of want it. And by the way, the book is an easy example in the sort of medium scheme of things because the book has a barcode and maybe it's scanning the barcode. Maybe it can see the title. There's structured data on and around the book. There's another part of that demo where he just holds up a handful of almonds and he says, how much protein is in these almonds? and it just tells him it's 15 grams. And it's like, did it count the almonds?

29:23And I pointed out, it must be counting the almonds. And Liam said, well, technically you'd want to weigh the almonds. Right? Because you're inside of that. ChatGPT is just making up grams per almond average and then spitting out a number, amount of protein. Using a dubiously data collected source that could be accurate or could not be accurate. And you would not know. And it's like all you've made is like a very complicated lose it and you don't know if it's counting the almonds accurately. That's weird. Yeah. It's not good. There's no structured data about the almonds. I mean there are a million holes you can poke in this thing and I think we're doing that appropriately.

30:05I think the direction Humane is trying to go here is really interesting. And I want to have competing visions for where computing is going in the world because everyone else, meta, et cetera, Apple, they think the future of computing is more visuals in your eyes, right? Like literally augmenting over the world. And Humane is saying, no, we think actually computing should be recessed more into the background and you should interact through voice. And it's more of that Spike Jonze, her future. And I think those different directions need to exist. People need to be building those different directions.

30:36It just feels very early. This feels like General Magic or Magic Leap. You know, there's a ton of examples of companies that were just too early to the tech. And I think this is the case here. No, I'll make a big distinction between those things because we – now we're old and we know a bunch of ex-General Magic people. They were very sincere. Tony Fidel, friend of the show. He was part of the General Magic story. He financed the documentary, a self-serving documentary, but you will – because they all come out as heroes like in the end. But it's a fun watch, and I encourage you to go watch General Magic documentary.

31:08They were nerds, like huge nerds. And they did not think they were cool. And they thought they were inventing the future. Fine. But at the end of the day, they were like silly nerds. They didn't put their pin on like fashion models in Paris Fashion Week? They weren't. None of that stuff was happening. No Caprotti show. Magic Leap, huge nerds, obviously, took themselves so seriously. Well, they blew Beyonce away. Right. They had other people down there for demos. I think David got a demo once. Didn't you go down there at some point, David? I got a demo. It wasn't – I didn't get the full treatment, but I did get to try it at one point.

31:46Yeah. They said things like we're hacking the GPU of the mind, which is an incredible phrase. They promised to reinvent all of entertainment, all of technology. that like this thing was in they got so far ahead of the technology with the hype and the seriousness that there was no way the product could have lived up to their own ideas general magic was early but they weren't they knew that there was a thing that they were trying to get to that was far away from where they were and they were honest about it i think and if that's the spectrum of companies that are too early like sincere nerds trying to build the future and they're aware of their limitations and kind of silly and honest about it.

32:29And we've hyped ourselves so much that nothing we invent can meet the standard. Like, Humane is beyond that. Do you think it's beyond magically? I think it's beyond magically. Wow. But it actually is. Because they're out there. They have their employees, ex-Apple people, in the press today saying they had retired from Apple and they joined Humane because they were disgusted with having invented the iPhone and destroyed society. And Humane is their chance to like reset. It's like, you guys, I don't even know if it's counting the almonds, you know, what bookstore, you know, it's like, that's crazy.

33:09And if you're, if you're positioning is that this will rescue us from the phone, but you've made a phone nothing there's there's no product that can fill the hole that they are drawing right and that is the issue i think with the the product itself in many big ways but also like how they're putting it out into market but this is also this is just the world we live in now i mean i think about like carl pay and nothing like he's running around being like there are no interesting consumer electronics companies left and it's like well that just isn't true, but you have to like, you have to tell this big grand story about yourself and how you're changing the world.

33:50And it's like, we still live in the, we're making the world a better place thing, which you would think Silicon Valley would have killed like by just being on that TV show, but it didn't. And so I think honestly, like the, the single worst thing that Apple has done to tech culture is convince everybody that this is the right way to talk about your products, that everything has to be finished. Everything has to appear as if it was inevitable and just sort of rose out of the ground, this perfect finished product that we have solved every problem that has ever existed in the world. And there are no companies anymore who are making things being like, we made this.

34:24We don't know if it's any good. Do you want it? Ironically, open AI is probably the closest thing to that right now. But it's just not how this works anymore. And the playbook worked so well for Apple that if you're a company who raised a lot of money and would like to make a lot of money, you just follow the Apple playbook. And I think the thing we've learned over and over is that it doesn't work for anybody other than Apple. I don't think this is fair because there are a lot of interesting consumer tech companies that aren't doing this. They're not based in Silicon Valley. They're not in this weird investor ecosystem, right, where there's almost this demand from their investors to ship something and to change the world at the same time, right?

35:03There's plenty of weird tablets in China. China's doing a lot of cool consumer tech gadgets and not promising to change the world with their Steam Deck knockoff. It would be amazing if Apple was like, we made these GamerLights. Yeah. They're awesome. They're just cool. Yeah. They're never going to do that. But if you go back to like the launch of the original iPhone, Steve Dobbs, he didn't say it's going to change the world. He was like, here are three things it can do at once. Well, no. He did start with today Apple reinvents the phone. But that's not – but that's different. With his arms outstretched.

35:34I would argue that's different than what Humane's doing, than what Magic Leap did, which is like it's the hubris, right? And Apple has hubris on – they have tons of it. And Steve Jobs was the king of hubris. But like you go back to that original iPhone, there was like an almost weird humility with the hubris, right? Where it was like we actually were just focusing on what it does. And what it did was such an aha moment where everybody in the room was like, oh, my God, these three things together in a mobile form factor, yes. You didn't have to over-explain it. Whereas like humane, they can't explain it because it's incomprehensible.

36:08It's an idea in search of a product, right? And I think that's the key difference here between like the early Apple where Apple is now and humane, I think. So here's a question I have for you, Alex. Just thinking about the cycle OpenAI is in. Brother Spirit, the Buddhist monk, introduced them to Mark Benioff in 2018. The connection, by the way, made through their acupuncturist. This is all very important detail. 2018, that's a long time ago to start working on this pin. OpenAI was nowhere in 2018, right? The first glimmers of the thing were happening. They were a research project. Yeah, they were maybe 60 people at that point.

36:53Yeah. And they were still kind of focused on not being a commercial company. So if you're making this product in 2018 and you have this vision, what is the technology that you're basing this on? What have they been working on? Oh, I know what happened. There's a pivot that's contained in the story of this product where they were like, you're going to talk to ChatGPT and that's the thing that's going to sell it, right? Because that's what we're talking about today. In 2018, you could not begin work on that product. Well, ChatGPT didn't exist. ChatGPT barely existed a year ago. This was totally going to be like an Android product until OpenAI popped off, right?

37:30Like that had to have been what happened. They were building their own little Android dumb phone because it's effectively a dumb phone, right? And then OpenAI popped off and they were like, oh, finally we have something cool to attach this to, to ship it. So we're not just like we did a phone, but it's a pen and not really a good phone. Yeah. I don't know. Yeah, there's just – I'm thinking about this chronology and it's like in 2018 when you set off to make this product, you cannot assume that there will be an LLM explosion in mid-2023. No, but you could assume – That recontextualizes this whole product.

38:04But you could assume Google was doing – Google was doing a lot at that point, right? They were doing a lot about visual identity, like being able to visually ID things with a camera. They were doing a lot about being able to just talk and have those natural conversations with Google Assistant. Like, all of that was happening in 2018. So that's why I'm fully convinced that at some point we're going to find out that this was originally an Android product. I mean, it's an Android product now. Like, the thing runs Android. I think you're exactly right. I think if you rewind five years, they're looking at voice assistants and Google Lens and some of these other, like, what can we do without putting a screen and a bunch of apps on your smartphone?

38:45which in the abstract like heath has been saying is a really interesting and valuable question like this is stuff we should be talking about because the idea that the like platonic ideal of how we interact with technology is like a grid of icons on a screen is just not true and we we should be exploring all these things but i do think it's very clear that at some point along the way they went from we're gonna have to cobble together a bunch of different things and probably invent a bunch of technology ourselves to just, this is now, we are an interface company on top of ChatGPT, which is functionally what this is.

39:20This is a, this is a, it's a revolutionary input device on par with the, with the keyboard and mouse for ChatGPT. Like that's what they're trying to build. They make a big song and dance about all the other stuff they can do, but this is functionally a way to input things to ChatGPT. That is what this device exists to do. It's a real digital crown moment. You know, any minute Apple's going to do a partnership with one of these companies because they can't build it by themselves because of all the privacy reasons, and they will say, if you want to use ChatGPT, your Apple Watch can do it. And that will be the end of the AI pin.

39:55Oh, I will take the opposite side of that bet. I'll bet money that they will not. Because of the privacy reason you cited, they will not do this as an external partnership. No, but we're watching this in the trial right now where Tim Cook and Sundar Pichai are in meetings discussing their search deal. And Apple's like, your services can begin where our limitations end. So you can, they could say. Well, there's an important fact there, which is that Google is paying Apple on the order of$20 billion a year. So yes, if someone is paying you, you know, 20 % of your profit margin, yeah, sure. You can have deep partnerships.

40:34You don't think OpenAI and Microsoft are going to throw that money down? Like, that's what I mean. the second Apple gets over it and there's a few ways for them to get over it. They can invent a privacy-focused LLM very hard inside your values. You could take the money. It's true. Or you can do shortcuts with David's favorite. Love shortcuts. But like all I'm saying is there's a big action button on the side of my watch and there's a lot of different ways for that thing to trigger talk to chat GPT. Yeah. Right. There's many methods by which that button can do the thing. And at that moment, the value of$700 plus another$24 of service fees a month begins to slide.

41:19If I hold my phone just right, I can do the shortcuts method with ChatGPT and the Ray-Bans. Yeah. Like – And they're adding – so Kranz, they're adding early next year. Zuckerberg was talking to me about this when I interviewed him ahead of Kinect. They're going to do a software update for the Ray-Bans where it's visual recognition. So in the interview, he was like, you'll just look at the camera and say, what is that camera? Where can I buy it? And he didn't say buy it. He said, where can I buy it? And so that's a much more, I think, compelling because it's on you, right? And cheaper. And cheaper.

41:52No subscription. And like if you go to the beach, you don't have to worry about where you're going to pin it. It turns out your head is on you all the time. Who knew? And you can have stuff on your head all the time, no matter what you're wearing. It's actually a really cool human feature. Okay, this brings me to the last thing I want to talk about in this sequence, which is a graph that's in my head at all times. Is there a y-axis? There's a y, and I'm going to label the graph for you. So pull over in your car, and I want you to imagine x and y-axis. Okay. And I've worked this out, because I was on Threads last night, and people really helped me figure this out.

42:25This is the formula of wearable product success. Yeah, I saw this on Threads. Okay, so it's X and Y axis. It doesn't matter which side's which. The Y axis, let's call it, is amount I have to care about this shit. Right. Times face multiple. What's a face multiple? Is it on my face? Okay, okay. And is it a VR headset? So is it on my face and it's like a cool pair of sunglasses? Low face multiple. Is it on my face and it's the Meta Quest Pro? High. Enormous face multiple. Yeah. If it's not on your face, on your wrist, no face multiple. Okay. Got me? Got it. See it. Okay. So that's just one variable.

43:11Yeah. And you've got to decide for yourself in the glasses. How much do I have to care about it is, is this a tiny little computer? Do I got a software update, plug it into battery, charge it? Does it beep? Computer stuff. Yeah. And you've got to decide. Is it weird to get the photos off it? Yeah. All in the first factor. Yeah. So fiddliness times is it face? That's one axis. Okay. And then the x-axis is how useful is it? Okay. Okay. So if x is greater than or equal to y, so there's a line at 45 degrees. Oh, God. Then it's good. And if it's lower, you're duped. So like regular glasses, right? Alex, you're wearing some glasses?

43:55Yeah. You got to think about them never. Right. Low number times on your face, low number because you have cool glasses. Value, very high. So you're like historic success. Oh, okay. See, I didn't do well at math in school, but I think I'm tracking with you. You see what I'm saying? No, Nilay also is describing a graph that has like six or seven different axes. No, no, no. It's just one. It's just two numbers. There's algebra involved. Is fiddly bullshit greater than or less than how useful is this? I'm going to make a GPT for this where you can just. and I'm just saying the only real variable here is face yeah okay it's really just face is it on your face and then it's gotta be super valuable right right Apple Watch a lot of fiddly bits at the beginning low value not on your face not on your face but it still didn't do a lot at the beginning you could like send flirty heartbeats to people who maybe didn't want them right and they figured it out and they brought it up above the curve and now it's actually kind of less fiddly too yeah success okay MetaQuest Pro fiddliest product in the world gigantic face multiple no value no value not a success they won't even talk about it anymore you see what I'm saying so humane AI pin very fiddly no face multiple unclear value big question mark yeah Vision Pro enormous face multiple on the order of they put another face on your face right yeah huge problem there very fiddly we think or perhaps not because it's kind of standalone.

45:28Unclear fiddliness. Unclear value. X. It's an X. The value is X. Who knows? I'm just telling you this works every time. Wow. Someone's going to animate this graph and it's going to be amazing, but I promise you if you just work this out for any product, you get there. So the meta glasses, people like them, right? Yeah. So you do it. They look cool. Low face multiple. Eh, fiddly. Liam complimented mine today before he saw the lenses. Before he knew. Yeah. So medium face multiple because people are like, oh, you've been watching me, you creep. That's a problem. But they look cool. Yeah. And then they're fun.

46:04Right. Because they have good music. They've got great call mics. You should watch that video with Becca. They're useful. They're useful. Right on the line of success. Okay. My challenge to any Vergecast listener right now who's playing with chat GPT's GPT builder is to make a GPT based on what Neelai just said. Send it to us. And it's going to be amazing. fiddliness times face multiple has to be less than or equal to how useful is it? What is this bot called? Neelize gadget or Neelize wearable value indicator? This is a horrible name. The theory of wearable bullshit. Okay, there we go. That's the episode title.

46:45Yeah, it needs a name. I'm just telling you, I've been working this out on the show forever, and last night it finally clicked when I looked at the humane pin, where I was like, oh, it's not on your face. So that changes the whole equation. But then it is all the other things that it is. I was disgusted by my work on the iPhone. So instead I made a little laser projector that you have to look at your hand. There's no Instagram. You can't do Instagram with lasers. It just doesn't work. This thing is real silly. They should have let us and other reviewers actually into it, and they should have been less self-serious about it is what I'll say.

47:18I'm excited to play with one. I think we're all excited about new gadgets all the time. I want to see how the speakers work. The personic speaker. Yeah, I want you and I both to have one and see if we can hear each other's. That's going to be the worst episode of the show. The worst episode. I'm so sorry. Just two people listening to a title four feet apart. I mean, I will say, again, to go back to the meta things on that one, I've seen a bunch of people be like, oh, the personic speaker. This is the thing that the smart glasses have been doing for a while. This personal directed audio that you can hear and other people don't is like relatively proven technology.

47:52The Ray-Bans do that. Yeah. I do this, and I hear so well with the Ray-Bans. Yeah. It's the projector. I sit on the train like this. Yeah. The projector is what I want to try. I want to – I'm going to a concert Friday. I wish I had this so that during the concert I could just hold my hand out and it just like Matrix displays. Are you going to have a concert that sucks? Is anyone else going to be there? No. I just want to do it to get the reaction around me, you know? It's just like your – Alex is in the pit. Yeah.

48:20I just assume it's very bright. Yeah. I brought my own laser show. Yeah, there you go. That would actually be sick. Okay, here's the killer app for the Humane AI pin. Personal laser show. Oh, yeah. Imagine Coldplay, and instead of the bracelets they give you, it's just the pins. They all... Yeah, this is a good time to note that Yugo Labs had a board NFT party, and they bought the wrong UV lights, and they burned everyone's eyes. Maybe don't do the laser show thing. Just an idea. All right, we got to take a break. That's a true story, by the way. It's heartbreaking. Check your UV lights. We got to take a break.

48:55We'll be right back to talk about OpenAI.com.

49:01Support for this show comes from Shopify. One of the most exciting parts of starting a business is when the sales start coming in. If your business is looking to start selling online, Shopify has everything you need to create your online store. Shopify is the commerce platform behind millions of businesses around the world and 10 % of all e-commerce in the U.S. from household names like Mattel and Gymshark to brands just getting started. With hundreds of ready-to-use templates, Shopify helps you build a beautiful online store to match your brand style. And Shopify is packed with helpful AI tools that write product descriptions, page headlines, and even enhance your product photography.

49:40And best yet, Shopify is your commerce expert with world-class expertise in everything from managing inventory to international shipping to processing, returns, and beyond. Turn those dreams into and give them the best shot at success with Shopify. Sign up for your$1 per month trial and start selling today at shopify.com slash vergecast. Go to shopify.com slash vergecast. Shopify.com slash vergecast. Support for the show comes from Crucible Moments, a podcast from Sequoia Capital. Sometimes the difference between success and failure comes down to one chance encounter. Or following a counterintuitive instinct.

50:25Or ignoring conventional wisdom to make a bold decision. Like when the founders at Palo Alto Networks wanted to redefine cybersecurity for the modern age. Everybody thought we were crazy. Nobody would use the cloud for cybersecurity. And when I hear something like that, I say, of course I'm going to do it, you know. If everybody believes it shouldn't be done or it cannot be done, then, you know, it's a good reason to do it. or when mobile gaming giant Supercell could only rewrite the rules of the industry after failure in the company's formative stages. Many of the best things we've learned have actually come through failures.

50:57These are Crucible Moments, turning points in a company's journey that made them what they are today. Hosted by Sequoia Capitals, Rule Off Botha, Crucible Moments is back for a new season with stories from Zipline, Stripe, Palo Alto Networks, Supercell, and more. Subscribe to Season 3 of Crucible Moments. New episodes are out now, and you can catch up on Seasons 1 and 2 at cruciblemoments.com, on YouTube, or wherever you get your podcasts. Listen to Crucible Moments today.

51:33All right, we're back. Less hardware this time, more software. But the future of software in a very exciting way. Alex, you went to OpenAI Dev Day this week. They announced a bunch of stuff. Take us through it. Yeah, so this was in downtown San Francisco Monday, and this was OpenAI's first big conference. And when I was going to it, realizing that ChatGPT is not even a year old at this point, blew my mind. And then you get there, and they have about 1 ,000 developers there. And it just had this energy of—I'll get into the announcements, but I think it's important to contextualize the energy. It felt like the early Apple events.

52:12It felt like this buzz that, you know, we had that a little bit. I think, you know, David and Eli, you were there too for the Vision Pro unveiling. That had a special kind of feel because, you know, people had high hopes for Apple finally doing a big new thing. But, like, this just felt like something I haven't experienced since, like, early Macworlds. And you get in and, I mean, they did a really nice keynote. It was about an hour and a half's worth of content packed into, like, a tight 45 minutes. with like real live demos. Like there was a moment where they were showing off like a Zapier integration where you could like have a bot you make, you know, text someone or send them a Slack based on what you tell it.

52:53And like, they're doing it on stage live. They're building the thing in like a minute. And then like, it sends the ping to Sam and he's on the other side of the stage and he just holds his phone up and it's like, yep, got it. Like, I really don't like the direction the industry is going thanks to Apple where it's these highly produced pre-recorded keynotes that are really just like becoming marketing videos. I think we saw that with the last Mac event, especially. So OpenA, it felt like this returned to form. It was like, wow, this feels good. In terms of what they announced, it was a lot. I mean, I think like the most immediate thing that the developers were like freaking out about and like Sam had to tell everyone to like stop clapping because they kept clapping over and over was just like making GPT cheaper to use as a developer.

53:37So it's like three times cheaper. They've got GTP4 Turbo, which apparently people really want larger context windows, which is the ability to upload more text tokens to these models and do more with them. So I think the limit was around like 20-ish K max before, and now it's like 128 plus K context window. So you could put a book in this thing and have it do all kinds of things. People got really excited about that. And then a bunch of like kind of more developer-focused stuff, right? It's a developer conference. The biggest thing, though, was this new AI platform. I mean, really what OpenAI wants to become, and I wrote about this in Command Line this week, is the App Store of AI.

54:18And so they've got this new GPT platform where with no coding required, you can build your own GPT, upload custom knowledge, and do a bunch of what is essentially like abstracting away prompt engineering, right? and making it very like tailored to a specific use case and do all that in like honestly a matter of minutes. I got a live demo of one of these things being made and you can do it in under five minutes and then just publish it to the web. Eventually, and when I say eventually, I think just in like a few weeks, OpenAI is gonna have this store, the GPT store. And that's where things get really interesting, I think, because you're seeing OpenAI, not only is it becoming a huge developer platform, an external platform that powers other things, They have over 2 million developers in under a year, which is just like wild.

55:06But they've got this massive growing consumer business. They have 100 million weekly users of ChatGPT, which is like for a product that is less than a year old is just absolutely insane. That means their monthlies are probably on the order of half a billion. And so what they're trying to do is like aggregate that audience and also become a consumer internet platform and have people making GPTs, no coding required, that they then surface to people through ChatGPT and through the store. And so the store is coming very soon, but they showed a preview of it. And I think that's where things get really interesting.

55:42Is that 100 million, is that GPT Plus the paid thing or just? That was just ChatGPT users. They didn't give a plus number. Very important for the custom GPTs. You have to be a plus subscriber to access and build the custom GPTs. So this is all feeding that loop of like OpenAI. I think the information reported they're already on like a$1 billion a year revenue run rate with the consumer subscription business. They really want to have more reasons to pay for ChatGPT and an app store and custom GPTs. And again, abstracting away all that prompt engineering they think is going to be a key unlock there.

56:18So like some of the examples were like – and it's multimodal, so you can use Dolly. So it's like an interior design GPT that's like specifically for helping you figure out how to design your home or a cooking one. Or, you know, theoretically here at The Verge, we could upload all Verge content ever made into a GPT and it could replace search on our website. That's an idea, right? And OpenAI, which we should get into this next, OpenAI will be compensating creators of these GPTs based on their usage, which I think is a very interesting distinction from other platforms. I think two things that kind of – I read your piece and it was great.

56:56The two things that stuck out to me were, one, the compensation and how they're planning to do that because that seems like a whole mess. And also just how quickly he seemed to gloss over the hallucination problem. Yeah, just like any normal story. Like you're releasing an enormous app store. Walmart or Target, you're like, most of these products are lying. Like an enormous lying, enormous lies store, basically. Yeah, just any normal store. Yeah. You can go into a Walmart or a Target. Well, yes. But the difference with the custom GPTs is you can upload custom knowledge. So like on stage, Sam made a bot and he used to be the head of Y Combinator, the startup incubator.

57:32And he was like, I have wanted to have a bot that just tells people what I would tell startup founders, the advice I would give them for like a decade. He just built it on stage in like three minutes. And he uploaded transcripts of every talk he's given on these topics. Right. So, yes, it's still a problem. Like hallucination obviously is still a problem. He talked about this more after the keynote in a press thing I went to. But the fact that you can augment it with custom knowledge makes it more domain-specific. And so it's hopefully going to alleviate a lot of that. Doesn't that kind of just make it like – I think the no-code part is probably the coolest part of this.

58:09But doesn't this just make it like no-code chatbots like you had on AIM 20 years ago? Like if you can upload – like for the same Altman one in particular, okay, that's really, really cool. Somebody could have done that 20 years ago. It just would have taken them quite a bit of time, and now it should take you. You have the sort of like Eliza problem. Right. Right, where it's like there's some canned responses. This one is, like, going to talk to you. It's going to be more natural. I think the fidelity of the conversation is higher. The thing that's really interesting to me is what you're saying, Alex, the compensation model.

58:39Yeah. So OpenAI is, what,$20 a month for a plus? Yeah, yeah. And then they're probably going to, what, just share out some of that to the bot builders. That's the Spotify model, which is historically doesn't make anyone rich except Spotify. Well, Spotify, important distinction, there's this whole thing called the labels, right, that OpenAI doesn't have to deal with because they don't pay for a lot of their content, which is something that Sam did not really want to talk about. But yes, I actually think it's closer to YouTube because the way they're going to do it initially is they're going to share some percentage.

59:14We don't know. We don't know if it's fixed. We don't know if it fluctuates based on the category, but they're going to share a percentage of the companies. So OpenAI's ChadGPT subscription revenue with GPT creators. And it's based on usage. And Altman was very kind of high level about this on stage. I think they're going to share more in the next few weeks. But at this press Q &A I went to with After, and I wrote about this in Command Line, he elaborated and said there's going to be category-specific bonuses. and that down the road they're open to actually letting you subscribe directly to custom GPTs, letting GPT creators charge one-off fees for them.

59:53But it sounds like a lot of it is going to get figured out in like the first few months of the store. They're rolling out with this initial kind of mushy rev share thing. But it is YouTube-esque. It's not Apple-esque because there isn't going to be like a transactional piece to this. Have they asked ChatGPT itself how it thinks it should handle the red models? I'm sure they have. Analogies are difficult because it's weird and it's mushy. YouTube, most YouTubers are getting a cut of the ads. There's a sort of like infinite amount of money coming into YouTube or at least an uncapped amount of advertising money coming into YouTube and that goes out.

1:00:33Right. Spotify is like you pay Spotify premium and I know my ever increasing amount of money I pay for Spotify premium. I listen to a song. There's some formula and that goes to the label and the artist and it's pennies. And this feels more like that, right? I pay chat GPT 20 bucks a month. I go use a bunch of GPTs. My usage gets divvied up into fractions and goes to those creators. and that they got it. The reason I'm harping on this is because of what David said earlier, which is the entire game is the app store for AI. And you cannot distribute this anywhere because the second you start doing a better revenue model, which is just pay for the apps or pay for a subscription or have transactions, Apple is going to sit there and take 30 % away from you.

1:01:18And Google is going to take 30 % away from you. Apple already is taking 30 % of ChatGPT Plus revenue on iOS, right? Because it's a subscription that goes down. Well, if you sign up on the phone. Right, if you sign up on the phone. So, yeah, and this is where I kind of like ended with command line this week is like, I would love to know what Apple thinks about all this because Apple, they don't like stores within apps. You can ask Mark Zuckerberg about that one. You can ask a lot of people about that one, actually. A lot of people have tried versions of this. The only one that's gotten it through for reasons that are too complex to get into is WeChat in China.

1:01:49But I think they're going to run into a massive wall there eventually, which is Apple's take rate and their historical hatred of an app trying to recreate any kind of a digital store experience. And potentially having their own AI, right? Right. And wanting their own AI to be their primary AI in the device. Because once the AI is looking at books and just buying them for you, that's a whole other world of transactions that Apple can take money out of. And they're going to want to be their primary AI there. So just to bring this together, you can see why people want to build a not phone. Why Sam Altman is investing in 45 different attempts at this and might be working with Johnny Ive at some attempt at this.

1:02:34there's a massive blocker to mobile innovation here which is that 30 % cut and I just don't have any sense of how anyone will actually get around it because building the humane AI pin or the rewind pendant or whatever Johnny Ive comes up with on the side the phone is still gonna be the primary device yeah well the the other part of it is like this is not all that different from what Amazon and Google were trying to do with skills for their voice assistant. Like we've been down this road where Amazon was like, okay, you're going to interact with Alexa, but we're also going to offer you lots of ways to interact with other data and other services.

1:03:14Skills is going to be a huge business. That got past Apple. It was fine. It just turns out nobody wanted that because it has this heinous discovery problem. The interface is actually not that good for it in most ways because like the thing where you have to essentially say a wake word to get to the thing and then another wake word to get to the app that you want and then talk to it in a very specific way all to route right back to you like that flow sucks and so i think what chat gpt is going to have to figure out is how to like again we're talking about abstracting all of this away and by virtue of making you choose things you lose that abstraction in a really helpful way like chat gpt already has these plugins yes the plugin system sucks those are replacing it's a perfectly useful thing.

1:03:59Right. But I think that is one way to go down that road that doesn't work because you can't find the stuff. You have to like turn it on every time you want to use it. It just doesn't work. And I think figuring out how to do this thing they're trying to do in a way that doesn't cause those exact same problems is going to be really hard, especially because OpenAI is trying to have its cake and eat it too because it also wants to be the interface to all of it. It does. Yeah. It wants to be the underlying infrastructure and the most powerful, important product. And at some point, doing both of those becomes really, really, really hard to do.

1:04:35This is what Altman said in the Q &A I was at after his keynote. I thought it was really interesting because it gets at that tension, David, that you talk about where OpenAI is building a platform very clearly. They're continuing to add things into ChatGPT, add different modalities, different data sets. At the same time, they're saying, come build on our platform. We'll share revenue with you. Come make your own GPTs, et cetera. And he got asked about this, and he said very clearly, he was like, don't build a thin wrapper on top of OpenAI. We are planning to build the obvious features that you would expect for a robust platform over time, and there's enormous value to build.

1:05:13So that's him just being straight up like, we will probably Sherlock you. Yeah. And I see a lot of parallels here to the early App Store. We're going to build the app. And I think that's okay. Like, you know, the early app store had a bunch of flashlight apps, right? Eventually that just became a feature on the phone. And it's a better feature as a result. You can go in the app store right now and download a home interior app that is a thin wrapper on Dolly. Yeah. Like we're painting our house and I wanted an app that would just like put paint colors on a wall. And like half of them are now just Dolly wrappers.

1:05:49And they're bananas. Like it's like don't do this. I don't recommend them. Your house starts to look real wild after a minute. But there's obviously a market for that thing to occur right now. And maybe one day OpenAI will build that exact thing because that's a useful case for Dolly. But in the meantime, they're just watching a bunch of these companies try and fail. And that's useful to them. And in the short term, useful companies are making some money. I just keep thinking about all the scummy apps that are going to come to this thing, right? Like there's the obvious ones where I as an adult call it scummy.

1:06:25A lot of our younger listeners would probably love it. It's like one that will just write your papers for you, right? That's a thing ChatGPD can already do. Somebody will probably make a really, really good version of it using this. And a lot of people will probably be very upset about that and demand for it to be pulled down because it's helping a bunch of kids cheat. There's ones where like, okay, you can have an, you know, input some actress that you really want to talk to. And then everybody's like, oh, yeah, I can go talk to that hot actress. And it's like, oh, no. You can already do that, Alex.

1:06:55I mean, there's like replica character AI. But that's going to happen now on open AI's forum, right? Like they're going to have to deal with that. Oh, someone's going to build that GPT. Right. If you build a GPT of me, it's over. So that is like – The law will come down. So, Neelai, like we are – there's a lot of stuff on the internet that could like be used to inform a chatbot about us because we write on the internet. Like that is a perfect example. someone could just make a Neelai bot and it's everything Neelai's written on The Verge and you ask Neelai questions about copyright, you know, forever, right?

1:07:25Yeah. And the Neelai bot is like, this is copyright infringement. But you – Well, we should – This is a little bit of a tangent, but like they – another thing OpenAI announced this week is a copyright shield, which is a program where they will pay for any legal fees incurred by their developers for copyright claims of use of the technology. So theoretically, someone builds a Neelai bot. They scrape the verge for all Neelai's stories. They put it in the bot. Vox Media goes, hey, you stole our IP and you're making money off of this because OpenAI is paying you. We're going to sue you. And OpenAI is saying now we will fight all those legal things that happen.

1:08:02So, like, what do you think about that, Neelai? That's a bad answer to this problem. We should all just, like, I just want to acknowledge that is a bad answer. That's how OpenAI goes bankrupt. Real fast. Well, no, there are also – there are so many interesting ways of thinking about this content. Like, Alex, your paper example is a really interesting one because, like, you put together all the announcements that came out this week. And what's going to happen now is you're going to be able to upload your history textbook to a GPT and you're going to be able to write papers from it. That is – it now has a big enough context window and, like, they're using this thing.

1:08:37It's like the AI term for it is rag, which is basically like by doing this custom knowledge, you're able to reduce it all the way down just to the stuff that you know that it knows, which means it hallucinates less, which is actually a really important part of this. It's going to write you a good paper based on your history textbook. And that is just – it's all sitting there. And that's a mess in so many ways. And for OpenAI to just be like, don't worry about it. We have lawyers is not a good answer to the problem. So Sam's exact wording on this issue was we don't like to do things that are illegal.

1:09:10Congrats, Sam. Cool, Sam. Same, Sam. Same. But like, I mean, they clearly think all this is under fair use. And there's a lot, you know, you all have talked about on the show a lot of, you know, there's a lot of lawsuits right now that are trying to, you know, they are going to figure this out, whether this is fair use. This is happening while all this is going on. Can I color that in for you? Yeah. So the Copyright Office took comments about whether it's fair use, and all the big companies filed comments. Oh, yeah. That's good. And we have some of them on site. And the one that just stuck out was Andreessen Horowitz, a big Silicon Valley firm.

1:09:42Their comment was, look, companies have invested billions and billions of dollars into this technology, assuming it's fair use. So it's fair use. No, I mean, they were clear. They were like, if this is not fair use, like, these investments go to zero, basically. It was like the whole structure of all this is predicated on the idea that we can have this for free. Yeah. I know a lot of media executives. They don't agree with that. Yeah.

1:10:12Not a little bit. We had Barack Obama on Decoder this week. I asked him this question like, you're an author. Do you think this is for you? Because he really wants to talk about AI. And he was like, he's a politician. He's very good at this. He's like, leave me out of it. I've sold enough books. Like you watch him do it. He's like, no, no, no, no, no. I refuse to engage this question. And then he said, this is a speed bump. The revenue will get figured out. And so I think there's a real tension in the industry. There's an expectation from the people who make the IP that this is, yep, a speed bump, that the money will sort itself out, the creators will get paid.

1:10:46and there's the expectation from the tech industry that they're going to get to do a Google and just take it. And like Google's history is built on taking it and they were the scrappy upstart when they did that. When Google scraped all the books on the internet and made Google Books, they got sued, but they showed up with their like big colorful logo and their propeller hats. Literally that image of Google protected it from these lawsuits. When Viacom sued Google over YouTube for just ingesting huge amounts of Viacom television shows onto YouTube. YouTube was like, but it's so cool. Like, everyone loves it.

1:11:25You can just watch South Park for free. And Viacom was like, yeah, that's the problem. They won. And they won running away because Viacom was perceived as evil. And Google, again, stupid logo, slides in the office, beanie hat, and it was really useful. That's part of the story. The rest of the story is that Google had a reasonable defense, which was to say, we're not the ones uploading it. And it turns out there's like a large part of tech policy that is based on that specific idea. No, they, I mean, their other defense was, and we built you content ID. We built you a private copyright enforcement system that is unaccountable to everyone and pisses everybody off every day, right?

1:12:06That's the actual, that's the thing that they did to get out of that jam. Google got so much out of this in a lot of ways by returning value to people. Like Google made the case to Viacom that more people are watching Viacom stuff because they're finding it on YouTube. And whether that's true or not, it's a case to make. And it's done the same thing with websites, right? Like there have been all these issues for years about like media organizations being like, how are we supposed to feel about the fact that there is a constantly updated version of our entire website sitting on Google servers? And that's what they're crawling and serving, not our website.

1:12:37And what Google says is we send you traffic. And there's like there is some set of trades that you're willing to make there. And chat GPT so far has not returned any value to the people whose stuff that it is taking. It's not directing traffic. They're not getting money. It's just like a nifty thing that other people get to do. And I think without if it can't figure out how to make that turn back to like here is what it's worth for you. I think it's much bigger than a speed bump. Yeah, no, it will bring the whole thing down. Like it is the end of the industry if this isn't fair use. And the thing for all of these companies is you cannot rely on the court system to be reliable when it comes to fair use.

1:13:19Like it is not a stable decision-making framework. It is a coin flip. Like every time. It is a coin flip even in just like recent memory. Like I bring up this example all the time, but I'll do it again because it's useful. The estate of Marvin Gaye was not happy. No, it just comes up. Drink. Somebody owes me$10. I'm just telling you, the Blurred Lines case is going to go down as like the thing that upends maybe all of technology. It did bring us Emily Ratajkowski. Many things happened because of this song. They, Robin Thicke and Frel, sued Marvin Gaye's estate preemptively because they thought it was so clear cut.

1:13:56They hadn't stolen a note. It was just a vibe. And they lost. That is a totally unpredictable result. And then the estate of Marvin Gaye in Bolden goes and sues Ed Sheeran. Ed Sheeran shows up in court all floppy hair and says oh just the music and he wins and he actually took the chords he played guitar in court incredible Ed Sheeran impression by the way you could use some accent work but yeah it was alright he's just a floppy guy he's just a floppy guy but that is a straight coin flip if I gave you the facts of that outside of the actual characters and said these two didn't take any of the chords or the music they took a vibe and this guy took the actual cords.

1:14:38Who should win, who should lose, right? And it's the exact opposite of what you expect. And if you are OpenAI or Google or any one of these other folks, you're staring at a problem where it is so unpredictable that you should go and make some deals just to bring some semblance of order to this marketplace. But to what David's point is, no one knows what that deal is worth. The second you make one deal, everyone else is going to want a deal. the second you make a deal you have communicated to the world you think deals might be necessary and everyone shows up with their hands out like this is classic like rocked in a hard place situation for all these companies and if they end up going to court and fighting this out no one knows if they're going to get to the right answer on either side a couple weeks ago on Decoder we had a record label executive and she said look the music industry is like this you sue each other in the morning and you go to dinner at night and that's how it works.

1:15:35And like, you just have to deal with it. And they're going to have to get good at that too because that's how we do this. And I don't think Sam is ready for that. They are not thinking about it at all. It didn't come up on like the stage. It didn't really come up afterwards in the conversations I was hearing. They're very focused on what is the value we can give to developers and users. And the data that is powering all of this is like something they do not want to talk about because of what we've all been saying. But I mean, I do want to just like, to take devil's advocate here, Like, yes, like there's a huge copyright question here.

1:16:07Generally, I feel like with these new things, the Viacom YouTube example is a great example where better product experiences tend to win in the long run, right? So – and if I'm thinking about a GPT for The Verge, unless I'm in kind of like discovery mode where I just want to literally surf our beautiful site and look at things and I have no idea what I'm wanting, I actually would love the idea of just going, what should I know about the new iPhone? Is it good or not? And a bot just – it surfaces our article. It tells me. It says this is the score. This is the history of our coverage. Here's a link to the video or just watch the video there.

1:16:44Like that's a pretty good experience. There's a lot of like finding things on the internet that like that time collapses immediately when you're using an interface like this. And that's going to, I think, provide value. And so if you believe that that's valuable, and I think a lot of people will, and I think we're going to start seeing a lot of really interesting GPTs that are doing very dubious, you know, fair use things but are showing like, hey, this interface is actually better for just finding information. I do think that idea may win out in the long run regardless of what happens in the short term with the legal system.

1:17:17It just feels like they're moving way too fast. They are, yeah. Between the lying thing and the fact that they haven't figured out whether or not they're wholesale ripping off a whole bunch of creators, like, what are you doing? What do you mean ripping off? What do you mean? What we just talked about. Like the payout structure? Yeah, a lot of people straight up accuse them of – if they're ingesting their stuff without paying them, that's ripping them off. Oh, here. I'll just give you – let me give you just another example of this. Outside of Chashu Petit, I think about this with TikTok all the time.

1:17:47I keep a list of TikToks that I think should be media PhDs. It's great. One day I'll publish the whole list. And there's one I keep coming back to. Hagerty Media, which is a great car YouTube channel. I love it. They did a stop motion video of an engine being disassembled. Cool. This took forever. Like just making a video of an engine being disassembled such that anyone can understand what's happening takes a long time with people. To do it in stop motion is like a – just obviously massive investment. And so it instantly got cut up into 10 ,000 different TikTok channels that repost it into 10 ,000 different clips because it looks cool as hell.

1:18:26And if you're running one of these TikTok channels and just trying to get views so you can like send people to your TikTok shop, this is great. Should any of that exist? Media PhD. okay here's the real thing that i think about all the time the comments on all of the tiktoks that have freebooted this video are this took so long this must have been so hard to do this is so cool i'm so glad you spent all this time doing this and it's like not like everyone can see the value and the difficulty of the work and no one not not a cent of that value is going back to the person that made the video. Right.

1:19:03And no one's confused that this was like a hard thing to do. TikTok is making the most money in that. The freebooting channel is making some money in that. The users are getting a ton of value, right? Because they're seeing a video they wouldn't have ever seen. And the poor person who like had to move the bolt like a sixteenth of an inch and then take another frame gets nothing. Yeah. Yeah. And like I keep thinking, like I came up as a guy who when I was my short stint as a lawyer, I defended the kids who used Kazaa. And I'm sitting here being like, am I the bad guy now? Like, what is going on here?

1:19:38Like, that's like one step too far. And that's just with video and everybody can understand, everyone can see it. If you make a GPT and you ingest all of some YouTube creators' videos and you're like, now I can talk to Doug DeMuro and he doesn't get a cent. Like, at some point you have to say this is not like fair. Oh, it's not even – Not fair use or whatever. Right. Like, this is just, like, outrageous. It's not even that the creator doesn't get a cent. It's that in that case, YouTube doesn't get a cent. And guess who has a lot of lawyers who is comfortable suing is Google, right? So, like, I think OpenAI is entering a world of hurt here.

1:20:12Like, I saw, again, I got a demo of the builder. And you would think that the ability to upload custom knowledge, they would have some pretty stringent guidelines on what you can upload. and maybe they scan it before they let you put it in the GPT to make sure it's, you know, not something that is like very blatantly like paid copyrighted material that is like behind a paywall or something like that. It's literally just an upload field. It's just an upload field. You can upload anything. Like there was nothing said about like how they're going to look for this stuff. So I don't think they're ready for what's about to happen, which is that like someone is going to be – it's just obvious.

1:20:48Like you take a paid product online, you pay for it, you scrape it, you put it into a GPT, and you make money from it. And it's like in that flow of events, like the value is just being stolen, right? And like an OpenAI is saying at the same time, guess what? We're going to pay all your legal defenses. Yeah. It's just wild. We should just get OpenAI sued. That's the new Virtu. Well, they're being sued. I mean, they're being – Look, I'm throwing out my wearable graph. The next episode is just us doing dubious things to get open AI soon. No, but wait. They are getting sued by artists like Sarah Silverman and all these people.

1:21:27Yeah. But this next turn where they're offering revenue to people to build things and then protecting them from the claims. I'm going to make a bot that adjusts all of the Game of Thrones books and writes the final two. I think George R.R. Martin at this point would just pay you for that. Yeah. That's my loophole. He should just do that. That's my loophole. George, if you're listening, and I know you are. I got you, George. You can do it. They've expanded the field. The context window is bigger, George. I do think this is, like, a huge problem, but I do want to acknowledge, like, these GPTs, they're cool.

1:22:02They're going to be cool. They're cool as hell. People are going to like using them. The Dolly stuff is wild, like, just having, like, something that can, like, generate. And, like, I have this problem. I don't know about you guys with these chatbots, these, like, all-in-one chatbots. It's, like, it's just a text bar. I don't know what to do with it. I don't know all the things it can do. There's this context problem with these bots, like Bard and all these, where they can do so much that, like, it's overwhelming, frankly, for me. So having, like, a cooking bot that has, like, all the cooking information I would want to, like, make recipes, I would pay separately for that because, like, I actually just want that.

1:22:36I don't want all of it, right? And so I do think it's a really powerful idea, and Meta's doing it. They're going to have a bot creation platform next year. There's already character AI, replica. The industry seems this year to have realized that we are moving on from these, like, one bot to rule them all concept to, like, millions of bots everywhere, right? Yeah. And that's a cool idea. It's just, yeah, there's all this, like, copyright problem. But I think as an idea, it's – There's just existential risk. Actually, the model thing is really interesting. They had Sachin Nadella on stage. Microsoft, they're obviously a huge open-air investor.

1:23:12And part of the argument that they're making is they're all just using one big model, but they're learning to segment it into these use cases. So the big model is getting better overall while people are doing these specialized things with it. And then on the side, Microsoft Research is trying to figure out if you can take small models that can run locally or on these other places and make them fully useful in, like, very narrow contexts. So you see this like kind of parallel innovation pathway where like the big model, maybe you can put guardrails around it and make these GPTs. Or you could have like the little model that's just good at what it does.

1:23:48And there's something – and the funniest thing about I think the little model training that Microsoft is doing is one of the prompts they give it is pretend you're the big model. Which is incredible. Right? It's like don't – it's not impossible. It's like that thing, you know? It's pretty good. But it's funny that we've had this whole conversation about voice assistants and AI and all this stuff. And it's like we've brought up Amazon and Google many times, which have voice assistants, millions of them deployed in homes today and seem not ready for this next turn at all. OpenAI is moving so fast.

1:24:23And that was just the takeaway from being there and, like, talking to developers there and just – I mean, when you think about the fact that ChatGPT is almost a year old and everything they've done in a year and everything they just announced. on stage. And like the demand for it, like there was a massive chat GPT outage this week, right after they announced all this stuff, like their servers are melting. So they have clearly like captured lightning in a bottle. It's their game to lose. I really feel that it's just a matter of the legal ramifications, potentially catching up with them and the diciness of this store concept.

1:24:55And like, if you're doing a usage based system, what's to keep me from buying a bunch of bots to use my bot so that I have the most used bot? Like, how do they know that it's actually real? I guess because it's through the subscription business, but so I guess that is a good limiting factor on that. But, um, there's just a lot of like dicey, like who decided, like I'm already imagining like chat GPT is used to like, uh, influence the next election and they decide to demonetize news. Right. It's like everything old is new again. Right. Like, yeah. Uh, well, that's like, what is, what's going to be the Cocoa Melon of ChatGBT.

1:25:31That's on the good side, right? That's like the biggest YouTube success story ever. Yeah. Enough so that Disney has listed it not owning Cocoa Melon as like a risk. Yeah. Because it, and that's just a YouTube success story. And it's like on the other side is like, what is the horrible white noise playlist on Spotify? Yeah. App, right? That is just like zero value delivered and just sucking money out of the ecosystem until Spotify has to shut it down. Yeah. Yes. And the next turn here, really, and this is what Sam was telling us after the keynote, is he imagines that we will all have our own GPT, which I'm really excited about.

1:26:08So a GPT that is trained on all of your stuff, like the rewind idea, the pendant, but like all of your software, all of your notes. And your GPT engages with other GPTs on your behalf. So it's like bots on bots all the way down talking to each other. So it's like you tell your GPT, I want to go to see a concert and have dinner with these friends on this date, line up my ride, my reservation, the tickets, and it all talks to different bots to do that. Sam thinks that's inevitable. And what really – so OpenAI's stated mission is to build AGI, right? Artificial General Intelligence, which is this like super – it's like godlike AI that can think better than all of us combined.

1:26:49And people in the AI field that I've talked to over the last year think that this next step where it's GPTs representing us engaging on our behalf unsupervised is the environment that is needed to train the models to become a step function even more intelligent. Because you need to have this concept called self-supervised learning where these agents are all interacting and training on their own. That's pretty freaky, right? That's like matrix stuff. Yeah. That's coming in like the next couple of years. I think the world could look a lot different. It's just, I think that, and that was the vibe there, was like we are kind of witnessing a very exciting, but also like very scary next few years here.

1:27:33Yeah. And inside of it is a copyright lawsuit. It's a giant ticking time bomb. It's just Ed Sheeran looking mad at you. All right, we got to take a break. We'll come back. We got a lightning round. We got to get out of here. We'll be right back.

1:27:49Support for this show comes from Zapier. When people talk about AI, they often talk about the future. How will AI affect the economy? What will AI's potential be in the future? Will AI be able to make popular art? But the more important question is, what can AI actually do for your business today? If you're looking for concrete implementation of AI for your business, you can try Zapier. Zapier says they can help you and your team actually reap the benefits of AI, not just talk about it. With Zapier's AI orchestration platform, you can bring the power of AI to your work. Connect top AI models like ChatGPT and Claude to the apps your team already uses, so you can add AI exactly where you need it.

1:28:29Whether that's automatically enriching leads, coaching your sales rep, resolving IT tickets, or more, you can do it all with Zapier. Zapier is for everyone, tech expert or not. No IT bottlenecks, no complexity. No AI hype, just results. Teams have already automated over 300 million AI tasks using Zapier. Join the 3.4 million companies already automating with Zapier and transform how you work with Zapier in AI. Get started for free by visiting zapier.com slash verge. That's Z-A-P-I-E-R dot com slash verge.

1:29:11Support for the show comes from Crucible Moments, a podcast from Sequoia Capital. We've all had pivotal decision points in our lives that, whether we know it or not at the time, changed everything. This is especially true in business. Like, did you know that autonomous drone delivery company Zipline originally produced a robotic toy? Or that Bolt went from an Estonian transportation company to one of the largest rideshare and food delivery platforms in the world? That's what Crucible Moments is all about. Deep diving into make-or-break moments that set the course for some of the most important tech companies of our time.

1:29:44with interviews from some of the key players that made these companies a success. Hosted by Sequoia Capital's managing partner, Rule Off Botha, Crucible Moments is back for a new season with stories of companies as they navigated the most consequential crossroads in their journeys. Hear conversations with leaders at Zipline, Stripe, Palo Alto Networks, Klarna, Supercell, and more. Subscribe to Season 3 of Crucible Moments and catch up on Seasons 1 and 2 at cruciblemoments.com on YouTube or wherever you get your podcasts. Listen to Crucible Moments today.

1:30:25We're back. There's quite a lot in this lightning round, but we got to go fast. Karans, why don't you start? Because I think you have the best one. I do have the best one. There's a new Steam Deck. Sean Hollister has reviewed it. He naturally loves it. I think I called him like the day after he got it and I was like, tell me everything. And he sure did. He really likes it. The big things that have changed with this one is there's a new processor. They went from the 7 nanometer Aerith processor, named after the character from Final Fantasy 7, to the 6 nanometer Sephiroth processor. Sames. They've switched to an OLED display.

1:31:09They've switched the battery lice has supposedly increased, which led to an incredible quote from Sean where he said, for example, in 2022, I was able to play the graphically intensive control for just under two hours on my original Steam Deck review unit at 60 frames per second. This week, I played the same game at up to 80 frames per second for two hours and 11 minutes. I mean, more frames. Yeah, yeah. He saw more frames. He does add more stuff to it. It is a pretty measurably big difference in battery life, especially on games that don't use as much of the CPU and the GPU. So, like, those lower-impact games are going to last a lot, lot longer.

1:31:57The high-impact games, those of us who are making our way through Baldur's Gate 3 on the thing, we're not going to see as much success. The big shock for me was that there are no Hall effect joysticks. I just assumed that was going to happen because there's been so much frustration with the drift on the current joysticks. Yeah. They didn't do that. That's a big bummer. But they got rid of the 32 gigabyte EEMC version, which was the one that I bought because you could replace the drive really cheap and easy. They got rid of that one. Now starts at$550 or$549. I'm resisting buying one, but Sean loves it.

1:32:32Oh, and the power button is now orange. Hey, that's pretty good. I love it. Yeah, I'm excited. I would put this out as evidence, as we've been saying, there's more action in gadgets than people ever give any credit to. And the love that our audience, and in particular Sean, has for the Steam Deck is like off the charts. It's an awesome device. And they don't have to do a bunch of really long infomercials for us to get hyped about it. Yeah. You're like, look, you can play Control at 80 frames per second for 11 minutes more. That's all you got to say. Oh, yeah. I just love this thing. They were just like, what are all the things we can do to make this device better?

1:33:07And then they just did all those things. And now here we are. Like, God forbid more companies don't just like make their things better all the time. Like, it just makes me happy. And it was the same thing with the Switch OLED where they're like, we got this thing pretty right the first time. But what if we just did a couple of things better? Wouldn't that be cool? And everybody's like, yes. But it's like most companies are like, okay, this is called the Steam Deck 2. and this time it comes in six pieces and you have to like do backflips while you use it. It's going to be so dope. And Valve instead is just like, no, we like pretty much got this right.

1:33:42We just made it better now. It's cool. It's like, yeah, that's the way. Just make your things better. Yeah, that's all you need to do. I just really appreciate that one of the headline features of this one is that the fan is both larger and quieter, which is like crushing it, guys. No more fan ones. This is some video game stuff here. Nailed it. We did it. I would know that, but I've been trying to scroll this article using Samsung Dex, which is not possible. I didn't realize you were actively scrolling. I thought you were doing – I always do the thing where I, like, fidget and scroll up and down.

1:34:12No, you're just trying to read it. No, I'm just trying to land on a sentence that I can read because this mouse is – Are you just flying top to bottom of the table? Just bouncing up and down. Out of control. Love it. And look, Android as many things as an operating system and Samsung's One UI as many things as a skin on Android. a combination of things that is designed to let you change mouse tracking speed is not on the list of priorities for either of those companies all right david what is your lightning uh mine's very simple it's just that uh youtube in its relentless quest to be tiktok is now adding a for you page to or a for you carousel to people's channels which is basically like personalized recommended videos on people's channels that you land on a very smart be forever interesting to me that the phrase for you has just completely taken over the internet like we've just decided that that is the term for every personalized everything for you just one uh and also youtube really ought to stop trying to be tiktok like just oh my god just stop i think everything is inevitably gonna be tiktok but i i opened my youtube mobile app to the shorts thing And it was just – it was little tiles of shorts filling my entire stream when I opened my YouTube app on my phone.

1:35:27Like, no. No. I still sometimes accidentally scroll it forgetting I'm not in TikTok. Oh, yeah. That's incredible. That and Instagram. I'm like, oh, that's why it's kind of crummy and weird. Yeah. It's like watching the algorithms grow up. Like there are different stages of being toddlers and like TikTok is like running away with the election. But like YouTube is just like, is this anything? Will this make you mad? Will it make you sad? Horny? You know, Instagram is like, here's a Kardashian, you know, like back to basics. Yeah. At least in my feed. They figured out that I like truck jumps. A lot of truck jumps on my Instagram.

1:36:05Like, you know, it's like getting better. Yeah, yeah, yeah. Because I engage with it more. And then TikTok is just like full crazy. I screwed up on TikTok the other day and I thought I had broken my algorithm because it just kept showing me bagpipe videos. No, that's what I mean. TikTok is now at the point where it's in the deepest, darkest recesses of your personality. Oh, yeah. If you've been scrolling it for years and years. It's horrible. It's like beyond truck jumps. It's like deep fried truck memes that like I can't explain to anyone else. There's a video I keep seeing of like a motorized pizza peel that just picks up a pizza and puts it back down.

1:36:39Cool. It was bagpipes. I'm saying it's telling you more about yourself than you need to know. I never. It's in there. It's in there. Five-year-old Alex thought they were cool. I left it at five. Leave me alone, backpives. TikTok has the reset algorithm button, and I think about hitting it all the time. Oh, my God. I should hit it. I think I have to at this point. We should all hit it. Here's my pitch for this whole thing. I've now watched more. Again, we've had a lot of copyright on the show, but I've watched more movies on TikTok recently than anything. And all of these video streaming platforms are trying to get you to engage again.

1:37:16and it's like, Netflix, just make the TikTok. Yeah. Just show me. Just like, let me scroll through it. And I'm like, oh man, yeah, I'll start couples retreat in the middle of it for some reason. And then that'll play on my TV and we'll be done. Yeah. A, I absolutely 100 % guarantee you that's coming. Like there is not one shadow of a doubt in my mind that at some point in the very near future, the Netflix mobile app will have that thing for you. But I also think part of the reason I picked this one as my lightning round thing is because it's just like I've gotten totally obsessed with the idea that scrolling is just like the behavior.

1:37:54And it doesn't matter what you're scrolling. It doesn't matter what you're doing. It's just like people want to scroll. You just like have a moment where you're like, I'm going to get on my phone and just do some scrolling. It's just like a thing people talk about. And the activity is so much more important than what you're doing it with. But it's just like I now think about that all the time is it's just like that is a mode of engaging with technology is I'm just going to scroll. and I think the more that becomes the thing, especially as young people are trained on that as like, I have five minutes, I'm gonna do some scrolling.

1:38:20I just, like, I hear more people say that and you're gonna see it absolutely everywhere and it is gonna like completely change the internet because it's like, instead of, I'm gonna go like look at Reddit or whatever, it's just like, I'm just gonna do some scrolling. And it's just the for you feed of scrolls everywhere. Yeah, it's nuts. And it's just a bunch of algorithms starting at the very beginning of like personalities, Like angry, sad, hungry, horny, and all the way to bagpipes. No. Whatever complex. I would love to see the database field of like bagpipe emotional reactions that TikTok thinks it's getting when it gets a bagpipe engagement.

1:38:59You know like Netflix has those like deeply over-constructed movie categories? Yeah. Bagpipes. It was all because I liked a Friends video. And for some reason it was like I know which part of that we need to pull out for you. bagpipes I'm telling you one of the things I love about these conversations is it's always just people telling on themselves about the stuff that they watch on social media it's so fun you're like aren't you also on bagpipe TikTok and it's like no no one else there are a lot of people on bagpipe TikTok including me now alright here's mine great piece by Jen Tui this week if you're a smart home nerd this will really hit you there's a company called Chamberlain makes all the garage door openers like if you buy a house the chance that you have a LiftMaster garage door opener made by Chamberlain.

1:39:43Very, very high. They, like many bad companies, have smart devices. So you can make the garage door opener smart. They have a proprietary platform called MyQ, which is long since a disaster. MyQ is monetized through partnerships with car companies. So you've got a car with like a Wi-Fi connection in it. You can like push the button cloud service happens your garage door opens whatever man many many many people have hacked into my queue with home bridge and home assistant which are really cool things it's very nerdy but you can have a little tamagotchi of a raspberry pi in your house you have to reset every couple months and it bridges all the things together into home kit that's my home bridge like my ring cameras show up in home kit because of home bridge my queue showed up in home kit for a lot of people because of HomeBridge.

1:40:34MyQ shut off this integration. Not cool. They claim that they're getting DDoS through it, which is like you could have a real integration with HomeKit and everyone shut up. You wouldn't get DDoS. So they, and this is the part that just rankles. They explain their decision by saying, we've made the decision to prevent unauthorized usage of the MyQ ecosystem through third-party apps. It's so that we can continue to provide the best possible experience for our 10 million plus users. We understand this impacts a small percentage of users. But ultimately, this will improve the performance and reliability of MyQ benefiting all of our users.

1:41:07That is the cleanest garage I've ever seen. Yeah, the picture here. So they said this is 2 % of people. So Jen just did the math. You've got 10 million users. You've broken 200 ,000 garage doors, which is just incredible. So now they're freaking out. I will tell you, MyQ has always been a disaster. They've always been pretty user hostile. This is not a secret. It's just what people have. And no one thinks about replacing the garage door opener. So everyone's been hacking around it. You can buy little adapters. I have one called Miros. I think the refurbished ones are called Refos. It's very confusing.

1:41:48They're like anywhere between$30 and$50 depending on sale. You can just wire them into your garage door opener and put them in HomeKit. They work. There's other ones. Jen has a whole list of other alternatives. if you are the sort of person who wants to open your garage door from your phone and I am that sort of person. You are. There are many ways to do this that do not involve the MyQ platform. And our article has irritated them so much that they're issuing corrections and they're trying to explain their decision even more when all they need to do is just either build the HomeKit integration or let the people do the hacks.

1:42:18Yeah, just like say, okay, Home Assistant's fine. I love it when companies are like, it's a tiny percentage of users. You just broke 200 ,000 garage doors. What are you doing? Only 200K. All right. Alex, you got a lightning round? Yeah, it's actually a thing I've been trying. I've been test driving a Rivian R1S for the last week. Yeah. So RJ, the CO, is at Code. And this is not an ad by any means, but I have had fun in it a lot. We took it off-roading in Joshua Tree. And an electric – it's my first time in an electric, like, SUV. and the thinking goes zero to 60 in three seconds and has like 840 horsepower.

1:42:59And it's just like driving a big-ass go-kart through the desert. Yeah, cool. And I like the, I mean, just a lot of the small decisions they've made in terms of the, you know, the OS and the interior of the car. The moment they do a mid-sized one of these, I think it's going to be a really compelling Tesla alternative. But right now it's very like outdoorsy, you know, big, big hulking vehicles. But yeah, had a lot of – Next up, Cybertruck for you? That's the – yeah? I've heard – yeah, I've heard the Rivian described as the – I can't say it on our air, but the version of the Cybertruck that for people who – yeah.

1:43:37All right. Yeah, anyway. Okay, can I say my Cybertruck thing? So I asked for pictures of the Wiper and there have been lots of Cybertrucks everywhere. and people keep sending me not close-ups. They're just like, here's another video of the wiper. And it's like, you're six miles away. So I'm going to be more specific. I need a close-up of the middle of the wiper. The middle, not the arm, not the part at the bottom of the windshield that turns the middle of the wiper blade. Or even better, close-ups all the way up the wiper. Like connecting to the glass. Yes, and I'll be more specific. I have heard rumors of how the actual blade is constructed that I would like to verify.

1:44:18I will not tell you what they are. Please. Because I don't want to put disinformation in the world. I know the Tesla people are crazy. I don't want clips of me saying a thing that's wrong. I'm just saying I've heard some information that I cannot verify about the wiper blade itself. So if you encounter one, just go look at the blade. Take some pictures. Start at the bottom. Click, click, click, click, click, click all the way at the top. See if you notice anything. A thing about the Rivian wipers, they work, and there's two of them. I was about to say, I heard a plural there. Yeah, there's two of them.

1:44:50They cover the entire windshield, and they work. The only thing I don't like is there's not the cleaner fluid. Boring. Why are we making new cars that cost a lot of money that don't have— It doesn't have windshield washers? It doesn't have the washer fluid. I'm driving this thing through the desert. Wait, it's empty, or it just doesn't have it? It doesn't exist. No. You guys, we cannot pick the windshield wiper fight again. We can't. We've been down this road before. Neal is about to say, just put water in it. It'll be fine. We're going to get emails again. There's no tank. There's no tank. He's saying they've obviated the entire problem.

1:45:23It's bizarre. Someone asked me if I was going to ask Obama what windshield wiper fluid is. We have picked a lot of dumb battles on the show, and this is easily the dumbest. And do not just put water in your windshield wiper. Anyway. Neal, the one thing I did want to ask you about Obama was how is the Secret Service? So this was our second round with him. And we were in his offices. So they were chill. Okay. The first round. Did you get a pat down? I did not. We were already all cleared from before. So when we were at Harvard, there was like intense security. And the other protests that day because he was speaking.

1:45:57We're on a college campus. There were dogs. Like the whole thing happened. We hung out with the dogs and we found out he wasn't going to be there. So then we – it was cool. The dogs are cool. They're very friendly. At his offices, we had already been pre-screened and we were on his home turf. So there was a presence. He doesn't not have a presence. And there was some discussion about whether the blinds can be open or closed and whatever. But it was very chill because we were in his space as opposed to when you're not in his – just moving him around is a complicated affair. I would imagine. Which is what we have discovered.

1:46:32Can't go to Cinnabon. Actually, can I tell you my one Obama behind-the-scenes story? and then we can end this episode. In that episode, we talk a lot about free speech on the internet and regulating free speech. And one of the things that's really hard to do is, like, you need some constitutional authority to go ahead and regulate speech. So the one that came up in the episode is the FCC has some authority over broadcast stations, right? So you swear on ABC or CBS or NBC, people can file complaints to the FCC. The FCC can find them because you're using the public airwaves. This is like a real thing.

1:47:08So like radio stations and broadcast TV, there's some basis for the FCC to regulate the content in there. Yeah. And that is wax and waint. This doesn't exist for cable TV, which is why HBO can exist, why cable news is like off the rails, why I can just swear at will on the internet on the show, right? The government has basically no authority because there's no public airwaves. So I ask Obama this question. It's in the episode. He's like, hey, you've seen some hook. So then we get up and we take a group photo because he's an excellent politician and he knows that you should take a group photo so everyone's like happy.

1:47:39And I did the least journalistic thing in my life, which I took a video of him saying hi to Max, which, again, I'm just admitting to you, disclosure, the least journalistic thing I ever did in my life. But my daughter has a video of Obama saying hi to her, which – That's awesome. You can't get that on Cameo. Yeah. Yeah. It's like hard. And then he's leaving and I was like that – this like thing, like this feels like the problem to me. Like everyone wants to do this. Everyone wants to issue some speech regulation, but you got to get over this problem. And he looked at me and goes, yeah, you just need a hook.

1:48:08You just got to figure it out. We'll figure it out. And he like walked out of the room. And I was like, oh, you used to be the most powerful person in the world. Like your brain is just like this problem should be solved. Like figure it out. Yeah. And it was like Kate Cox, the Dakota producer and I, both our minds were just blown because that's just like not – I don't have the ability to just like order someone to solve a constitutional law problem. And he was like, yeah, just, I don't know, shut up. He's just like, what? Just delegate. And it was just like, you don't have that experience very often.

1:48:40Yeah, the charisma, too. Yeah. Yeah, it was a lot. Anyhow, that's it. That's a very cast. Please, if you see a Cybertruck, just all the way up the wiper stock, I'm telling you there's something there. I don't know what it is, but people tell me there's something there I should look at. And make the Neelai wearable index GPT. Yeah, the graph. We need the graph. Yeah. I'm telling you, the Vision Pro, the whole Vision Pro review is just going to be me doing math. But I want it to be an input. I want it to be a GPT where you can say, is this thing, how does this fit into Neelai's graph of wearable awesomeness?

1:49:15Right. Fiddliness times face has to be less than or equal to usefulness. That's the whole equation. You just figured that out for me. All right. That's it. Thanks, everybody. That's Rich Haster. Back on.

1:49:30and that's a wrap for verge cast this week hey we'd love to hear from you give us a call at 866 verge 11 the verge cast is a production of the verge and box media podcast network the show is produced by andrew marino and liam james this episode was mixed and edited by zander adams and that's it we'll see you next week

1:50:00The Limo Emu

1:50:17natural ally, Doug. Uh, Lemu, is that guy with the binoculars watching us? Cut the camera. They see us. Only pay for what you need at LibertyMutual.com. Liberty, Liberty, Liberty, Liberty. Savings vary. Underwritten by Liberty Mutual Insurance Company and affiliates. Excludes Massachusetts. Hey, Ryan Reynolds here for Mint Mobile. You know, one of the perks about having four kids that you know about is actually getting a direct line to the big man up north. And this year, he wants you to know the best gift that you can give someone is the gift of Mint Mobile's unlimited wireless for$15. a month.

1:50:48Now you don't even need to wrap it. Give it a try at mintmobile.com slash switch. Upfront payment of$45 for three month plan equivalent to$15 per month required. New customer offer for first three months only. Speed slow after 35 gigabytes if network's busy. Taxes and fees extra. See mintmobile.com.