The Vergecast - GPT-4 is coming for your work tools — and your job

Episode Date: March 17, 2023

The Verge's Nilay Patel, Alex Cranz, David Pierce, and James Vincent discuss OpenAI announcing GPT-4, the next generation of its AI language model. Further reading: The night sky is always getting faked Samsung responds to fake Moon controversy Samsung’s fake Moon photos aren’t a giant leap for mobile photography OpenAI announces GPT-4 — the next generation of its AI language model The Bing AI bot has been secretly running GPT-4  OpenAI co-founder on company’s past approach to openly sharing research: ‘We were wrong’  What’s new with GPT-4 — from processing pictures to acing tests Microsoft Business Chat is like the Bing AI bot but as a personal assistant  Microsoft spent hundreds of millions of dollars on a ChatGPT supercomputer  Google announces AI features in Gmail, Docs, and more to rival Microsoft   Google opens up its AI language model PaLM to challenge OpenAI and GPT-3  Google-backed Anthropic launches Claude, an AI chatbot that’s easier to talk to   How Siri, Alexa and Google Assistant Lost the A.I. Race The BlackBerry trailer shows the rise and fall of the keyboard phone Biden administration reportedly demanding that TikTok sell or face a ban T-Mobile is buying Ryan Reynolds’ Mint Mobile for up to $1.35 billion Belkin’s smart home brand Wemo is backing away from Matter Learn more about your ad choices. Visit podcastchoices.com/adchoices

Before we start, I just want to say last week we were at South by Southwest, we did a live episode that you can check out. It was at Slack headquarters, Slack City at South by Southwest. It was very cool, very fun. Thanks again to Slack for having us. But go listen to that. It's in the feed. We talked a lot about the moon.

Support for the show comes from Retool. Too many companies run critical operations on duct taped spreadsheets, Slack workflows, and whatever else they could cobble together. Not because they want to, but because building internal tools means weeks of waiting on someone else's backlog. That's where Retool comes in. Build custom internal tools just by describing what you need. Prompt something like, build me a revenue dashboard on our Salesforce data.

And Retool actually builds it on your company's data and your cloud with Enterprise Security built in. Go to Retool.com slash Vergecast. We all need to retool how we build software. What's up, y'all. I'm Skyler Diggins, seven-time WMBA All-Star, Olympic gold medalist, and mom. And I'm Cassidy Hubbard, host and reporter for nearly 20 years covering the biggest names and stories in sports and mom. And this is Am Mom, a community for athletes, game changers, and moms of all kinds. Dropping May 14th.

Tap in with us. Hello, and welcome to the Vergecast, the flagship podcast of the moon, all of its implications. What if we pivoted this entire show to just like moon crystals and like lunar wisdom? What I'm getting from the internet at this point in time is that might be more lucrative than journalism. First of all, I suspect moon podcasts might be an even more competitive space than business or tech podcast. It's like, all right, guys, we can take it to Jason Calcanus and David Sachs with the moon people. That directory is too loaded. Hi, I'm your friend, Eli.

This is a verge cast. It's actually a podcast about technology. It's just the technology implicates the moon in a very serious way. Alex Cranes is here. I'm the dark side of the moon. There we go. That was a lazy joke. Sorry.

It was not great. Although best-selling album in every format throughout history. Very confusing fact. Good album. Yeah. David Pierce is here. Hi. I'm your friend who over the last five days has been totally radicalized into believing the moon doesn't even exist.

There you go. See, I knew this was going to happen. Yes. It's not real. Verge AI moon reporter James Vincent is here. I am on the moon right now as we speak. Can I tell you my favorite silliest culture war moment is when like a bunch of new right wing nut jobs looked at the reissue of Pink Floyd's Dark Side of the Moon, which very famously

has a rainbow going through a prism. And they're like, this band is woke. Just like the sort of like polite, what are you talking about that occurred after that? It was just a choice, choice moment. Sir, this is a tower. records. Yeah, exactly. You're in a Sam Goody.

All right, let's start quickly with a moon update. So if you listen to our show at South by Southwest, you know that there was some extremely important breaking news about the moon, maybe the most diverged story in the history of the world, which is that Samsung and Huawei and other companies have been processing photos of the moon taken with their smartphones in increasingly aggressive ways. And Samsung in particular, obviously sells a lot of phones in United States. People have been trying the photos, testing them, seeing the limits of the processing, a Reddit user discovered that you could take a photo of a blurry photo of the moon,

and the Samsung system would add detail back into the photo. So impose detail that was not in the original shot. However, it's still the moon. So the philosophical implication of what is the moon, as David said, can lead you to go absolutely haywire. James, Samsung has attempted to explain. You wrote the original story here, so I'm curious what you think the moon is and whether it's real. And two, Samsung has now responded to this story,

with an explanation of what its AI is doing. First of all, it's basically just an English translation of the Korean blog post that already existed with a little bit more detail. And second, does not appear to be all that convincing or actually explanatory. Well, I mean, I've been covering this with another UK-based reporter,

my colleague John Porter, and yeah, we've sort of been having, we've been getting into the weeds, we've been getting into philosophical arguments about what is the difference between taking a photo and creating an image. And I think that, for me, that distinction, you take a photo, you create an image, is sort of the crux of this difficulty about what's happening.

Because basically Samsung have copped to everything that this Reddit post has said, that they're adding in details, essentially. And from their point of view, they're just saying, well, we know people try to take nice photos of the moon. We trained a convolutional neural network, which is sort of quite now an old-fashioned form of machine learning model, but is small enough to run on a phone. We trained it on images of the moon. And when we think you're taking a picture of the moon, we use that to, um, bruised it up and make it look more like the moon.

And, you know, I think this is a fascinating edge case because you really can't do it with many things that aren't the moon. Because the moon is the same. It's the same to everyone. Famously, we only get to see one side of the moon at any one point. Right. The other side is woke as I've been let to be. That's where I'm the, I'm the other side.

Yeah. The other side is one huge pride flag and the government does not want you to see that. But you know, you only ever see one side of the moon. You see it in various rotations, and there is a really interesting thing called the librations of the moon, which is essentially how it waggles on its axes and exposes a bit to the edge anyway. So everyone is always looking at the moon from the earth, which makes it a really easy thing for a neural network to improve and turn it into this nice picture.

And it also happens to be this fantastic marketing gimmick. And, you know, there's these adverts for Samsung's latest phones where they show off you know, the camera doing all these fancy things. And one of them is going to be taking a picture of the moon. And in one of them, you know, they obviously have the Samsung camera versus a telescope and the Samsung camera takes a better photo in the moon and everyone's really impressed. Like, I think it is just this really unique conjunction of technical amenability, of marketing, marketability.

And that exposes what others on our site have written about. Alison Johnson did a fantastic piece about this, talking about how, you know, photography on smartphones is shifting ever more and more towards computational rather than optical data, shall we say. And that is the balance that I think people are trying to understand. And that's the difference for me between taking a photo and creating an image, whether it's more optical or whether it's more computational. The moon happens to be an example where it can shift really far towards the computational. And that is why it has caused this sort of explosion of, not outrage, I'd say, but dismay. And also, who doesn't want to put like a headline that says the moon is face?

I've been wanting to do that ever since I became a journalist. I was very jealous that you were able to publish that headline. And I will say, you know, there are like clips of us on the podcast talking on their platforms. And people are like freaking out. They think we're mad. I am overjoyed about this story. I want to live in this story for the rest of my life.

I will say to complicate your dialectic there, James. Yes. Professional photographers have called it making a photo for a very long time. Right. Because they are deeply aware that their choices in love. lighting in exposure, in treatment. All that stuff are intentional artistic choices that they make along the way. And they're not, you know, just like taking a photo.

They're like making, they're making choices. And they, if you talk to any professional photographers, which you can just find them on TikTok and they will talk at you very loudly about this. They're very careful that what they're doing is making a photo. So that little dichotomy, that little tension in the world of photography has existed for it an extraordinarily long time. what I would ask you is you're right

that the moon is a special case it's all of this work right we're doing all this work here is this whole flowchart from Samsung you can go look at and it's like is it worth it shouldn't you just paste in the moon like it's always the same like why are we doing all the work

and there's something about doing all that work that makes it honest or at least arguably honest as opposed to we notice that you're taking a picture of the moon it turns out the moon is always the same here's a clip art moon. Yeah. Yeah, it's a notion of authenticity, isn't it? And this is what people who have agreed that Samsung is faking these images, I think that is what they are annoyed about. They feel

it is inauthentic. And as you say, you're totally right. Professional photographers have been aware of these choices for ever since cameras were a thing. And I think Samsung's explanation is basically, well, we put so much work into it. It can't be a bad thing. And I think that's not, that's not what people are annoyed about. They're not denying that there's work involved. They're saying that, you know, maybe our society is just really artificial. I don't know. It makes me sound like I'm posting on Facebook again. If I go, you know what, guys, oh, everything's really fake right now. Our smartphone's fake. Isn't the world full of phonies? You know what's real? Moon crystals. 1999. We'll send you a bag every month. Sorry. But that, that, I'm also sorry,

that line you just described is really interesting though, because it's like, if you were to just like take out your Samsung phone and say like Bixby, take a picture of the moon for me, it could do a lot of the same work and figure out where you are, what time of day it is, what the weather is like, and conceivably use relatively similar technology to get you a picture of the moon that would look like the moon as you see it now. That, I think, would feel much worse to people than this, despite being functionally not all that different. So there's something about, like, I'm pointing a camera at it and I'm doing something that feels better. I don't know where that line is, but I feel like that's where this has come up. And there's been this big argument. And I think

everybody is right, which is my favorite part about this, right? Where there's like the proponents of this. No, there are some people who are wrong. And they're in the comments or Instagram rails. Some of them are wrong. But there are people who are like, this is exactly what Samsung should be doing. Why would you want to take a crappy picture of the moon when Samsung is able to give you a better picture of the moon? Totally valid point. I completely agree with it. There are other people who are like, this is not the photo I just took. What the hell is this thing that you're giving me? And I, also think they're right. And it just ends up in this like weird philosophical question, which is I think why this has been so much fun for us. David, here's what I want you to do.

I want you to make a volumetric scan of your new baby, feed it into Blender and be like, what I need you to do is age this baby over time and then just take photos of the 3D model of your baby. Okay. Like we're that close. Like that's kind of what we're talking. Well, yeah. The moon's a little more static. than David's baby, presumably. At some point, we're just, we're watching Will Smith, the AI generated young Will Smith instead of Will Smith.

Like, we're right there. I mean, that movie sucked, by the way. It just wanted to be very clear. It sucked really, really badly. But Emma Roth did this really great piece this week about astrophotographers and like their reaction to the moon photo. And they were like, yeah, we do that all the time. Like, in astrophotography, they're like, yeah, we take like crummy pictures.

And then we go into Photoshop. We edit the hell out of them. And then you're like, oh, wow, the stars are there. And that is very normal for us. And so they're like, yeah, this is also normal. Samsung's doing it. Thanks, Samsung. I think starting from the assumption that bad pictures are bad and you should make them

not as bad is both not a terrible assumption and also what every camera manufacturer is doing. Like, Google's out here removing things from the background of your photos. Like, you know what Google can do is just remove the moon. I can take picture of the moon and then be like, nope, no moon. Bye-bye, moon. And it's like this is what everybody is doing to different degrees. They're just saying, you don't intend to take bad photos.

So let's help you take good ones wherever we can. And where that line goes wrong, I don't know. But I honestly think for most people is like way down the road where it's like, my picture is not blurry. I'm a happy man. I don't want one quick point because I agree with that totally. Like what is the point of photography? What is the point of hardware?

As you say, David, it's to help people take good photos. And I think this is something that AI is really as needs. Eli said, is going to confuse more and more as it gets more and more accessible and as it's used into more products. Because, you know, I saw this startup the other day, which was like, are you bored of taking bad headshots of yourself for your startup? Give us your photos. We put them in an AI image generator and we give you your perfect headshot photo. In two minutes from now, they're just going to be making David's baby photos. In a business suit. Little business man. We love that. This is what AI is for is hustle

culturing your way through the internet. The moon rocks are just 1999. They're going to re-energize your energy and change your vibes. That's at the birch.com now. Yeah. The moon. All right. I will say this. Our producer has now changed the words.

We got it, fam. Let's move on from normal to bold to red to underlined. Look, I could, I can live in the moon story forever. But there was some actual AI news this week that it was not about the moon, sadly, or happily, depending on your point of view. GPT4 was released from OpenAI. Google announced AI 2. and Gmail and docs, Microsoft, its own AI tools and office.

Let's start with GPT4, which frankly the things people have done with it in just the handful of days that it has been publicly available are kind of mind-blowing. James, what's going on here? So this has been awaited and hyped and ruined for a long time. GPT3 came out in, oh gosh, now, 2022. 3.5 has been powering CHAPGT. So people have been waiting for this, and there's been a lot of pressure on opening out about what they're going to do with this.

They have been, how's the rollout being? They've given it to a lot of people already, I feel. They've already got it in a lot of products. I think this is the point at which, you know, you look back at the first GPT paper, which was in 2018. Obviously, it's a research paper. GPT4 is straight out of the gate. It's a product.

It is something that is being used by Stripe. It's being used by Morgan Stanley. It's being used by duolingo. It's being used by all these companies already. So I think the excitement with GPT4 is, is that it is a fully fledged product. There are all these interesting novelties. You know, you load up Twitter and you cannot escape some thread of someone saying,

here's why GPT-4 is going to be the most revolutionary thing since the steam engine, since fire, since whatever. And a lot of those are truly very impressive. But we've seen in the past with technologies like this, that consistency, that truthfulness continue to be big problems for these language models. And obviously, we also have the Microsoft News, which we'll get to later, that they're putting GPT4, they're using it through. at their office suite in order to help with the sort of drudgery of office tasks. So yeah, it's a huge

story. It's been everywhere, but people are still really trying to assess how much of a revolution this is. The big thing that it does that earlier iterations didn't do is it can understand, it can process images as well as text. So you can show it a visual input of some sort, whether that's a meme or a diagram, and it will be able to answer questions for you. You know, one demo of this was you show it a picture of the inside of your fridge. It recognizes what's inside there. and gives you recipes to make. However, that is still, that's not a skill, a functionality that is widely available yet. So far, we've only got the chat, input and output really available to the public.

So yeah, there's a lot of stuff that we're going to have to process as this sort of rolls out and people test it more. One of the things that is weird to me about this rollout, and I'm curious how you've processed it, James, is that on the one hand, like you said, this is the most producty thing Open AI has ever done. To wild degrees, right? They're charging money for these things now. They're closing their research wings and not showing people a lot less about what they're doing, which we should get to in a minute. But then at the same time, Sam Altman, the CEO of Open AI, has been out here sandbagging GPT4 for

like months. In every interview, he's like, this thing isn't going to be as good as you think. We haven't solved everything. He keeps saying like it's less impressive the longer you use it, which is just a deeply hilarious thing to say about your own product. Like, what is Open AI actually? think of this, do you think? So I think Altman is speaking when he's been in these, it's

really interesting to look at what he says depending on who his audience is going to be. And I think when he's speaking the, when he's saying these things about dumbing down people's expectations, he's often speaking to not the mainstream crowd per se, but he's speaking to a sort of hardcore of AI researchers within Silicon Valley who are very concerned about existential risk, for example. You know, these are people who think that the current

path we're on is that AI is going to become you know, this out-of-control entity that acts on its own instructions, acts under its own regards, and starts messing up things on the internet. So he's definitely speaking to them. He's saying, look, guys, it's not at that point you need to calm down. So he's trying to convince them that it's not the end of the world, like literally the end of the world. Yes. Okay. But he's doing this rhetorical trick.

Oh, my God, I should have looked this up beforehand. Lytosis, I think it's called, where you bring up something in order to say that it's not a thing. It's like I wouldn't, I would never dream of talking about my opponent's lexory. I would never even accuse him of, you know, visiting brothels or whatever it is. And Altman is sort of doing that with AI. He's saying, oh, don't not even worry about this being the world destroying super intelligent computer that's going to take over the world. We would never, we have that so under control.

Don't even think about it. And then suddenly everyone's going, oh, my God, they're building a super intelligent computer. Like, I think that's what he's doing as well, is that he is very clever. using people's cultural expectations of AI in order to gin up excitement. And if you look at the trajectory that Open AI has been on under his command, it's worked fantastically. They are rolling in money. They are beating Google.

They are, you know, Microsoft is working for them in some ways, you know, in terms of the Azure supercomputer that they've built for them. Like Open AI are from a sort of business competition standpoint, doing incredibly well. And a lot of that is to do with how Altman masters that sort of rhetoric in the and field. But he's also doing it to like cover up the fact that it is kind of dangerous. People do have concerns and he hasn't addressed any of the facts of like how it can be a major tool for

misinformation, right? Like he's doing successful in the business stuff at the genuinely the potential cost of like major parts of our society. Sure. And then the evidence of that, Alex, I completely agree. The evidence of that is that they have closed down the research function. It's yeah. The open in open AI is a misnomer now. They've closed it down. And James, they have told you that their previous approach to building in the open was, I think the word they actually used, we were wrong. Yeah, flat out wrong was the quote I got from Ilya Satskiva, who is chief scientist co-founder. So he was one of the original seven or eight figures, along with Altman, along with Elon Musk, who's obviously, he's no longer connected with the company who founded Open AI. And again, they are speaking to this idea of AI safety and AI risk.

And, you know, the tricky thing about this is, well, not the tricky. think the fascinating thing about it is that everyone disagrees. Some people think you should have open AI systems because only the community can truly stress test these things. And I think there's, obviously, we look at what happened to the Bing chatbot. Microsoft obviously rushed that out. The most stress tested. But it got stress tested by the entire internet over the space of a couple of weeks. And they, you know, in a way that was very bad for Microsoft to put out such a easily breakable product. But I bet Microsoft was also pretty happy with the fact that they got all this, fascinating and useful data about how to improve it.

So that is one from no one ever talking about Bing to a lot of people being like, I think Bing is in love with me. Yeah. Which is an incredible journey for any brand to go on. Yeah, they got a lot out of it. And, you know, some people think that opening eyes research should be open for the same reason. Others think that as Satskiva told me that because it is now getting potentially dangerous, that means it should be closed off.

However, I'm on a stress that he gave two reasons. He gave one was the business reason in that we don't want our rivals to copy us. And the second was the safety reason in that we think this could be a threat to society. And he said right now the business reason is foremost. So I think he didn't say this. I should have asked a little bit more directly. But I think Microsoft has probably had a word. I think at this point the amount of money that Microsoft has behind this product and the amount of which their brand is currently tied to this AI and what it can do, you know, we have their new announcement for them today.

I think someone said, Nadella said, you know, we can't be given these secrets away to Google when we are suddenly in a chance to overturn their position in a lot of dominant, in a lot of markets. So a lot of people are incredibly mad at open AI because they were, you know, to use the Star Wars mean they were the chosen one. They were supposed to save AI research by making sure that everything would be developed for the benefit of humanity. That's like part of their original mission statement. And now they are another court. actor like anyone else. And if you're really worried about AI risk, you might now see them as someone who is accelerating AI risk because they are developing it based on business interests, not on

ethical and safety interests. So it's a big trouble. So maybe the most prominent critic there, Elon, who, a surprise, Elon has an opinion about something online. Yeah. But Elon, initial investor in Open AI, and I think his tweet was very curious how my $100 million investment into an Open. I project has turned into a $30 billion for-profit company, which Elon baggage aside, that is a very direct statement of the problem. Yes. Right. Here is this thing that was started as a nonprofit to reduce this risk and has now become

a very profitable private company. Increasing the risk. Yes. Yeah. The tricky thing is that the non-profit still exists. Open AIA has an incredibly bizarre corporate structure, which I'm probably not fully qualified to explain all the details of. But essentially it has a non-profit controlling entity and then a capped profit entity that makes all the money.

And that is controlled by the charter of Open AI. And they have this big promise in the non-profits charter that as soon as they think they are, anyone in the world is two years away from developing a super intelligent AGI. They will stop all business work that they're doing and work towards helping that project be launched safely. Now, in one way, it's obviously very admirable. some people see it as very admiral to be like, yes, okay, we'll give that up in order to have the safer thing first. On the other hand, it's also, for lots of critics, complete bullshit. Who has a definition of what a super intelligent AI is? They can change that whenever they like. There is, you know, if the Open AI has proved anything, it's that they're extremely malleable.

And if they've proved what their goal is, at the moment, is making money for Microsoft. That's it. Well, and if you're OpenAI, the simplest business decision now would be to decide that you are the one who is two years away. And like, oh, no, we happen to win this race. I guess we'll take all of the money. And what's been weird to me about this is, on the one hand, it does totally fly in the face of everything that Open AI has ever said about what it wants to be as an organization. But on the other hand, they've been kind of telegraphing this for a long time. And my running theory for the last, like, six months has been that this happened much, much, much faster than I. anybody at OpenAI thought. And that somewhere in the last like 12 months, the executive team there

and frankly, like the whole tech industry went like, oh my God, this is, this is here. Like, this is real now we can do this. And somebody at Open AI went like, oh, no, we're, we're screwed if we don't figure out how to do this. And somebody at Microsoft, potentially including somebody like Sachi Nadella, was like, we need to pull as much of this under our own auspices as possible. And I don't know. I just can't stop thinking everybody thought this was going to be like a 2027 problem. and then turned around in like September and was like, oh, shit, we should probably start doing this now. I think that's a yes or no. In our blessed summer of NFTs, like the burbles that from Microsoft in particular, the burbles that that was nonsense and AI would be the thing were already there, right, in the background. Sure.

But Google's been saying AI was going to be there for 10 years. Like, this is, everybody knew it was coming. I just don't think anybody knew it was here. This is the thing I say about Twitter all time. Saying that current administration on Twitter is bad does not mean that previous administration was good. Google is just like lost, right? And like Google's errors along the way to not let people play with the product or demonstrate it convincingly have created a perception that is very far behind, whether or not that is true. Right. Like they're giving away one of their models, but the,

the good one, Lambda that the guy thought was alive. We haven't really seen it in action. It's going to get rolled out into some of these products. James, I'm very curious, you know, if the Facebook model was leaked, you can just like torrent to it. weird things have started happening with that model. That has implications for safety. I think that's going to be the first kind of big test case of what happens when these models are just literally out in the open. And in the meantime, I don't want to lose sight of this.

The things we have seen from GPT4 this week are legitimately amazing. Yeah. Right? I took a photo of a hand-drawn sketch of a website and it coded the website for me and it works is amazing. I saw one person feed into it. Code me a game where the right side. is Pong and I control the paddle.

And the left side is the game of life. Yep. You know, where the bits and bobs like the pixels like breed and kill each other. It worked. It just worked. I can't do that. That is a capability I didn't have and now I can just will it into existence by telling a robot to make it for me. I've seen people fully code Swift UI iOS apps.

Did you guys see the one where the guy built an app that would recommend five movies to him every day? Yeah. Yeah. That to me, it's such a, for whatever reason for me, like all of the chat GPT stuff has just been like a movie recommendation service. Like it's all neat and whatever. But I can say like, I like heist movies and I have Amazon Prime and Netflix. What movies should I watch tonight? And it like does it successfully and it's incredible. And this guy was just he basically with some

back and forth with GPT4 built an app that every day recommends five movies, pulls in trailers, pulls in information. And it just does it. No coding, no nothing. And then at least according to his thread basically like copied and pasted it and submitted it to the app store. It's nuts. Right. So there's there's something underlying that that is important, which is he, He knew how to make an iOS. And so the thing writing the code, like chatGPD couldn't do the work, as James is pointing out. Like it can't, it might be able to deduce the nuclear codes, but it can't go push the buttons for you. And Sam Altman is like, and we'll never let it do that, wink, wink.

This is like, yeah, I can write a bunch of swift code because there's a bunch of swift code in the internet for it to go look at. But it can't actually put it into X code. It can't actually hit compile. I can't tell you if it's going to work or not, right? This is like the examples here are a human being working with that system almost disappear. Yeah. Right. And like going back and forth together in a way that even the previous GPT 3.5 iteration in chat GPT was not up to that standard.

So that's really impressive. But the idea that now it's a business and the GPT4 announcement came with a lot of sort of like API customer announcements, as you were saying, James. Now other businesses can depend on this to run their business. That's the turn. And that's where I think Casey and Zoe had it in Platformer this week, that Microsoft is like scaling down its ethical AI research investment because they're the people who say no. And now you're like, well, there's money here.

So we're going to start saying yes at a higher rate. Yeah. So the turn here, it seems to me, isn't that chat GPT got better, the AIs got better. It's that we realized we could make money. We figured out how to profit off of them. And we found that the profit was better than the potential moral ramifications of releasing this thing that has like is a major misinformation tool. Wait, James, do you think that's accurate to say? Is that accurate?

Yeah, James, I mean, call me bullshit if it's bullshit. But that's like my read. I think that is a hundred percent part of the dynamic. And I think that is why this lack of information that OpenAI released for GPT4 in particular has stung people because they were founded as a nonprofit where they were supposed to be above these corporate motivations. and obviously they proved that they're not. Their reason for getting into bed with Microsoft was that they needed a huge amount of computing power in order to create these systems. And Microsoft or some other corporate partner was necessary to provide that.

And I speak to a lot of people in the AI policy world. And this is like one of the big things they're really worried about is that only corporations can build these systems. And corporations have very different incentives, have incentives that are not necessarily aligned with the rest of society. And that is not a controversial statement. You look at any, I don't know, the history of any chemical manufacturer, DuPont or whoever it is, or any industrial supplier, they will always make decisions where, or consistently make decisions

where people's safety is just not first and foremost. So, Alex, I think you're completely right that this is part of what is happening now. The interesting, or the thing that makes this so naughty is that you have people who believe, not that this is, say, a minor hazard to society, misinformation or propaganda or spam, but it is an existential threat to society that it's going to turn everything into paper clips and it's going to turn us all into gray goo. It's very difficult to pull apart those two strands of thinking and to know what is motivating whom, because a lot of people in Silicon Valley who are building this stuff, you know, you look at surveys and they're like, yeah, I fully believe

this is a huge threat to humanity. And then you kind of go, well, why are you building it then, man? And they say because I think I can help it be less of a threat. And that's very real as well. So pulling apart what is a corporate motivation and what is an existential, a philosophical motivation, very, very difficult. Yeah. I think that the thing that gets me right now is we're so on the verge of there being a robot internet where robots are making SEO spam for other robots to read and turn into affiliate

links. And then a human internet where real people are. just writing fanfic for each other or whatever. Yeah. This was another GPT4 experiment where a guy was like, he started a conversation with it and was like, I've got a budget of $100. I want you to make me money. Tell me what to do and I will do it.

And he ended up, the GPG4 came up with this idea of starting an affiliate link website for green gadgets. Beautiful. And so it, but it coded the website for him. It decided what gadgets to put in there. And it is, this is happening as we speak now.

So I don't know how far it's doing. But it started making a little trickle of money. I'm pretty sure. But also, it started making money because people started investing in it because everyone knew that the thread was going viral on Twitter. So it was bound to get some click through. So it became this sort of like, it ruined the experiment immediately. It was not a, it's very good.

An isolated experiment. But, you know, yeah, that is one way. And we saw this news with LinkedIn this week that they're now saying GPT4 will write your user bio for you. Yeah. And LinkedIn already has prompts, like AI written prompts. and then people answer it and it generates AI written articles, what I don't know, how to be a work fluencer or whatever it is on LinkedIn.

Yeah. There's a whole universe of robot internet to come, right? A lot of AI moons. A lot of AI moons. And it's like, if you're shopping for a desk, is it, and you know what the answer is, like you have to like slice your way through the AI chum that is designed to convince Google its people and land at like, you just buy this desk or whatever.

And I think that there's like a parallel human internet that will be created out of all this. This is just my, this is my theory. Will it just be Reddit? Yeah, it's like Reddit. Like it's there's a, there's another kind of social networking that is to come. Yeah. In particular on the internet that where people, people are pretty good at spotting the chat, like the chat GPT output. They're better than the AI systems are detecting it.

And they are certainly good at spotting boring stuff, which is the vast majority of what is produced. Yes. And I, you just like, that's my optimistic take on it is that, Eventually, it'll be more profitable for the robots to just be robots at each other. We'll just like hang out somewhere. We're just going to be in a V-bulletin that hasn't been updated since 2004, and that'll be fine. That's also a really good transition into the next thing we should talk about, which is Microsoft and Google. We should take a break first.

Yeah. We should. Let's take a break and we'll come back. We'll talk about Microsoft and Google. And then there's a very funny tidbit about Apple that we should get to as well. We'll be right back. We're going to talk about some actual tools here. Support for this show comes from Shopify.

Every thriving successful business has to start somewhere. A good place to start is a relatively simple question. What if, given the right tools, I've really put my all into this. One tool that can help grow your sprouting business to new heights is Shopify. Millions of businesses around the world rely on Shopify for e-commerce. They offer a host of helpful tools you can take advantage of, from payment processing to analytics to website design. Their design studio includes hundreds of templates to help you create

the exact website you've been envisioning for your business. If you're wondering, what if I need help, then no worries, because you're never left to fend for yourself. Shopify's award-winning customer support is available 24-7. It's time to turn those what-ifs into a thriving business with Shopify today. Sign up for your $1 per month trial today at Shopify.com slash vergecast. Go to Shopify.com slash vergecast. That's Shopify.com

slash vergecast Support for the show comes from Upwork. The days of doing it all, all by yourself, are over. There's no romance and burning out while you're trying to scale. Instead, you can check out Upwork.

Upwork helps grow your business by giving you fast access to specialize talent across more than 125 categories so you can fill skill gaps, launch projects faster, and scale without committing, to full-time headcount.

And finding the right talent is easy. You can browse profiles, review pass work, and get help scoping the role so you can get started quickly. Seriously, you could connect with the right freelancer in just a few hours, especially when you sign up with Business Plus.

Their AI-powered shortlisting pairs you with the top 1% of talent in under six hours. No endless searcher required. You can visit upwork.com right now to post your job for free. That's Upwork. Upwork.com to connect with top talent ready to help your business grow.

That's UPWRK.com. Upwork.com. We are back. In addition to the GAPT4 news this week, Microsoft announced direct integration into office products. Google announced the same. Let's start with Microsoft just because that is GPD4. By the silliest mystery in tech was solved this week, when Microsoft announced Bing, we kept asking them if it was GPT4.

and they kept intimating that version numbers were not the right way to think about it. And Open AI would make a big decision about version. And then now it's just GPT4 and they copped to it. Cool mystery in Microsoft. But James, walk us through what's happening with the Office products and GPT4. Well, I mean, I don't want to say it because everyone's going to say it, but it's a souped up clippy, right? It is a little assistant called Co-Pilot that has a nice little logo that lives in your sidebar

of whatever office app you're using. And you can give it instructions. And I think this is actually a pretty good product and actually could be very useful. Because a lot of the stuff it's doing is tasks that in my experience, these bots actually can do quite good at. And it's things like summarizing documents. It's saying, you know, you've got a lot of meetings here and we're going to pull out the three key points. You have a presentation and we'll turn it from a word document into a slideshow. Now, of course, the proof is in the pudding.

We don't know how good this stuff is going to be until we can say. start using it ourselves and we haven't been able to do that yet. We will very soon. But I think in theory, this is the sort of limited task that these systems can be quite good at. I think one of the big mistakes that we've seen in the tech world with language models is thinking they'll be good at search. But hallucination, making up information is such a bad problem when you point something at such a large array of information that we're never going to get over that. Well, not never, but it's a tricky one. Whereas if you just go, if it's quite a short document,

it can fit inside the context window of the language model, which is the amount of information it can sort of hold in its short-term memory, as it were, then yeah, it's much less likely to make mistakes, and it probably will do some useful drudgery for you. So I think this is actually potentially really useful. It's funny because hallucination isn't a problem in search,

but it's not for your average corporate PowerPoint. It's like hallucinating a good idea for me that I can tell to my boss and like Clippy will just do it. But I feel like the thing about that, and I think the constraint that James you're talking about is really important. And is the thing that we're seeing, right? Like, it seems like the first killer app, or at least the first profitable one, is going to be feed this AI, some subset of our corporate data and have it make sense of it, right? Like, who should I invite to this meeting? What should my email say?

What is this threat about? Where does this information live? It's like, that is the kind of thing because it's constrained, because it's a relatively small amount of data. it's just not getting enough information from which to make a lot of really aggressive, spectacular mistakes. And they're going to be easier to catch because it is in theory information that everybody already knows. It's just doing a better job of distilling it and organizing it for you.

The thing that is amazing to me about this is like, A, I think this is going to be like a truly magical thing in Excel, which is the kind of application that can do almost anything and you kind of know what you want it to do, but you don't know how to do it. And just being able to tell a thing like, hey, the thing that's in here, can you put it in all these other places too, is going to put like a bunch of TikTokers out of business immediately. But the other thing is like, I just can't stop thinking about this as like all of these companies telling on themselves

where it's just like, do you know how many people's job it is in the world to take a word document and turn it into a 10th slide PowerPoint presentation? Like a lot of people do that more or less for a living or write summaries of email threads for the person that they work for. Like so much of modern work is this stuff. I mean, you know my dream, David. My dream is to arrive at a point in my career where I no longer use software at work.

And if I can just get there, I will know I have succeeded. And if that requires using invisible software, just like commanding an AI to print out my emails for me and summarize them. Printed summaries of Slack 10 times a day for an email. But as long as I'm not using a graphical user interface, I will have succeeded. Like, I'm not using some piece of enterprise software. And I'm just like yelling at a robot. That's fine. It's totally fine.

The thing that you're saying, though, Microsoft has been at this forever. This is where Clippy came from. Yeah. They've always known the products are too complicated. They have not been able to build a natural user interface. I mean, this is like 90s, Microsoft dreams. And you can, you know, Nadella is a life for Microsoft. He grew up in like the cloud of Bill Gates dreams that Clippy would like do it for you.

And he could compete with Steve Jobs. And now he's here. And he's like, here's what you do. You tell Excel to act like in a. accountant and develop a business model based on these five inputs that you have. And Excel will generate an Excel spreadsheet for you. And just like GPT writing Swift code to make an iOS app, Excel is a computer. Like people have built Doom inside of Excel. It is a programming environment. So you've now built a natural user interface for one of the most powerful programming environments has ever

shipped at scale. And that's rad. And then over here, you've got people be like, I need to make a PowerPoint presentation about the Battle of 1812 for my eighth grade class. And maybe along the way we've destroyed the American educational system. Like, who knows? But the Excel thing is really cool, so we're definitely going to do that. It's basically right. No, I think we're in this place now where, like, we've talked a lot about kind of AI writing in general.

And to me, I think the sort of generative writing thing is the thing that seems to click for most people first, where it's like, oh, I can have this, write me a poem in the style of my favorite poet or like write my. my eighth grade essay for me. But I think the thing that has jumped out to me is, and especially this goes back to Microsoft too, is like, you know what's super boring is most work emails?

And like, we get really precious about writing because it's like what we do for a living. Most people just need to like make a marketing brief that doesn't get them fired and need to like put all the information into a document. And I feel like all so much of this stuff is so low stakes and so straightforward that the idea of having my AI

summarize and write an email, for me. Like there was one thing in Google's marketing materials where it was like it was a whole thing and you you typed into the Google AI, I'm on it. And it wrote this like flowery three paragraph email about like, I will prepare this information for you ahead of our next meeting. Thank you all so much for your contributions to. And it's like on the one hand that makes me want to die. On the other hand, that's how people talk. I'm telling you man, that's the robot internet. That's no idea. You don't understand. That thing where we are now sending AI generated emails to other people's AI. to decide whether that should result in a calendar invite so humans can talk to each other

is just going to silo itself away. And then my dream of never using software at work will come true. Yeah. No, I think that thing is a victory in more places than I have given it credit for. And then where you cap that before it becomes disastrous in other places is really complicated. But like, if I never have to write an email again, like that's a gigantic win in my whole life. I can just type to Microsoft like, no, I hate your guts. email me again and then it can just send a nice email to someone telling them to

never like or you just don't do it so this we should talk about google too because google's role all this stuff too and they're bringing it to gmail which is a direct application of this they're bringing it to docs uh doc's got a pretty spiffy user interface update this week you can by the way see how just a little competition is amazing in these markets like no one had thought about what google docs had looked like for a decade microsoft was like, there's going to be a clipy in it. And Google's like, here's this redesign. Here it is. It looks cooler now. Here's all these new features for it. It's going to be better now. I digress. But they're rolling it out in Gmail. They're rolling out in docs. We have not seen a power of their

model. And James, let's start with this. They Google opened up one of its models, Palm, but it has not let us anyone really see even what its other model Bard can do. They have a lot of models. I don't know what the difference is between them, but what do you think of them opening up this one. Well, opening up Palm, so they have a couple of big ones, Palm and Lambda. BARD is going to be built on Lambda. Palm is, you know, another state of the yard. It's 540 billion parameters, I believe. That is more of a competition with OpenAI's enterprise business, which is selling their API for various flavors of the GPT series and making, giving businesses the opportunity to build their own stuff on that. Google has basically said, we want to get in on the

that business as well for startups to be building stuff on our language models. Here is the API and here is also some, they are going to release some additional software that is going to make processing that sort of information and integrating that easier. However, as with the announcements they made for workplace, it is, you know, they said they're going to do it and there was this thing where the news went live and then someone said, well, yeah, but if you click on the blog post and where it links to the Palm API, it links to another blog post that links back to the original blog post. And there's nothing that says how much this costs or how you sign up for it. And I saw people getting very annoyed that like Google had announced like a little labyrinth for you to click

through, but no actual product. That's the robot internet. I'm telling you. It is. But the thing that Palm, I don't think it's necessarily a significant thing. I think it's big for their, you know, it'll help their cloud business. It may win some stuff from Open AI. But I think, think the consumer facing stuff is where all this is at the moment. I want to sort of stress that not only do we have Microsoft writing assistants coming very soon, Google writing assistance coming very soon, they're in Slack. Salesforce put them in Slack about generating and summarizing meetings. There are dozens of startups who have AI keyboards for your phone while you download them and it's a button that shows up next to the emoji selector and it will write what it thinks

is a response to what you just got sent. So you don't even need to switch to a browser to get chat GPT's input. You can do it direct from the keyboard in your phone. So I don't know. I feel weird about this stuff because in a way we've been introducing this technology for a while with stuff like Gmail auto replies, smart replies.

But I feel we're now at the point in 23 where there's so much of this that's going to be released simultaneously that it is going to have a real tangible difference on how we communicate one another. And it's sort of similar to what you. You were saying, Eli, like, about how it might start, you know, people might reconsider whether that email was worth sending or what it's going to do. Maybe people will pick up the phone more.

Maybe that will be the thing instead. We'll start because if you want to guarantee that an AI didn't write what you just sent to your colleague, send a voice note. Well, that's super easy to fake too. That can be fake too. But it can't be fake with this much expression. That's what I will say for the moment. That's true.

My favorite genre of AI TikTok right now is there's a creator. who just has Barack Obama, Joe Biden, and Donald Trump arguing about sound mixing and whether some beats are good. And it is the funniest shit I have ever. I cannot get enough of it. Like, hearing Donald Trump say to Joe Biden, those beats are whack is one of the funniest things you can possibly imagine. And you're like, well, deep bakes are a real problem. I'm worried about the elections. The stability of world peace is at stake.

But this was extremely funny. And I honestly wish it was real. I just keep coming back to this. Like the Google stuff, I'll just read the list here. Soon you will be able to draft, reply, summarize, prioritize your Gmail, brainstorm proofread, write and rewrite and docs, auto-generated images, audio and video and slides, raw data to insights and analysis via auto-completion, formula generation in sheets. That one seems very dangerous.

These models are not good at math, like generally. And I would not trust my business to the AI and sheet. sheets. I would not trust your business to sheets. Like, generate new backgrounds and notes and meat, auto summarizes meat, which many, there are many startups that are doing that for Zoom already.

And then this one is just very ominous. Enable workflows for getting things done in chat. And it's like, there you go. That's the one where it just, it just lays off 10,000 employees on Google chat for you. It's like, we need to boost our margins,

just hit the button and the, and Lambda will reduce your workforce. It's actually going to force everybody back into the office. That's, that's what it does. Just like, here's our new workflow for getting things done. You're coming back to the office.

It just drives you there itself. I just, I can't, I can't shake this idea that like, most of this stuff doesn't need to be human. Like, the thing where you should pick up the phone and call is like, maybe what we're actually going to realize is that the email your boss sends you

that is like, where is that thing? And then you send them the thing. Like, actually never needed to have two humans involved in it at all. That, like, it turns out that a vast, majority of the stuff that we do at work is not human. It's like transactional nonsense.

I'm 100% setting up a serious shortcut that asks you for a thing like three times a day. Can you send me that thing? Yeah, just, just every and like you can start asking Bing and Google Docs. Where's that thing? And I like, I do think the things were like there were, there have been the people who are getting called out for using chat GPT to like write notes when they're leaving or notes like apologizing for layoffs. And there is this really complicated thing where it's like all of these things we've been able to assume are like written by a person. And now that you can't, the question of like what actually needs to be human, I think is way less than everything, but it's something. And I don't, I don't know where that lands. But to me, it's like if we can

solve this problem of transactional commerce that we all do with each other at work all day, gigantic victory. Even if it means less human to human contact, I'm into it. This sounds like it's going to increase it. Sounds like this is just going to make that more common because everybody's going to be like, oh, I'll just rely on the AI to send the email. And then everybody's just sending AI emails and not actually getting any playing Crusader Kings all day and said. And then we stop emailing. Hooray. Right. And then you have a company full of robots talking each other. Yeah. And like one human one monopoly guy with a monocle. I'm just collecting the profits from his affiliate website that is only traffic by fraud bots. And everyone else is on like a V bulletin from

2004. Yeah. Like there's a thing like that. I don't want to sound too much like the boss because as this group will tell you, I hate meanings and process more than anyone at the entire company. It's true. And it is my goal very much to never use software again. But the business that you're talking about, David, is like, that's how you make friends at work, right?

Like this thing, we're like sending each other emails. They're going to meetings. Or like, just like, be useful to someone else is what builds or they trust. So you can like do the next thing together. And that's actually the thing.

Like the combo platter of, I have five remote jobs, most of which are being performed by AI. And I'm just like a brain and a vass. waiting for the one thing that requires, like, a human to do is just a weird outcome of all of this. And, like, maybe it's not, you know, chat GPT turns us on to paper clips. I mean, just like, we willingly turn ourselves into paper clips while, like, business is done by robots. Yeah, I mean, I would, I would quibble with the idea that those are the only two options.

That either I have to keep sending email or I become a paperclip. Look, those are your two choices. Do you want the job or not? Eli, what if the outcome of this is we get three martini lunches back? Like, we can all go drink for several hours in the middle of the day. Four day week. Four day week. Yeah, yeah.

I don't know. And I guess part of the end of this comes back to like, where is that line? Right. And like what is what of work is required to be human is not really a question we've ever had to answer? Because at some functional level, it has all had to be human, even taking your word doc and turning into a PowerPoint. But that is like massively going away very quickly.

I don't agree with that. I think we've been careening towards this for a very long. time. Like we were all having to take those courses. I say that maybe we weren't. James, I don't know how they do things in the UK. But we had to like take a course in high school. Yeah, three martini lunches. Yeah, three martini lunches in like first grade. But you had to take like a course in school where you had to learn how to use Word. You had to learn how to use Excel. You had to learn how to use all of these things and all of the tools. And then you would go to your job. And like I think we've been, we've been at that place where we're interacting with the computers, where we're not interacting with

people for a very long time. And when where there's a lot of processes, there's, a lot of busy work around like the actual jobs. And what we've seen again and again and again is that when we create these things, we don't get more martini lunches, we get people being paid less because, oh, robots can do this so you don't need to be as knowledgeable. And your skills no longer are as valuable. So like,

I'm sure this will create other new forms of jobs, right? Like, Nilai, I think you were talking when we were at Slack the other day, about how there's AI prompt engineers who are going and figuring out how to do these engineering. favorite new job. Stuff and stuff. So, like, it's going to, we're going to see other things come out of this that are really, really good. But I do not think this is suddenly going to give everybody their time back.

I think it's just going to reallocate their time in hopefully less miserable ways. Right. So David's positive case there is that we'll reallocate our times towards more useful, creative work. And three martini watches. My three martini lines, like being creative or like, yeah, making better product ideas, like, whatever it is. And this sort of drudgery of operating the company will go away. My negative case is there are no jobs left. The drudgery is like a necessary component of working together.

And if you remove that, then everyone's just showing up sauce on three martinis being like, I have a great idea. I have no idea how to execute it because I've never actually done any work. And that's just like a weird place to be. Yeah. And then you kind of get this like automated LinkedIn affiliate SEO scheme. This is like the real economy. It's like a house of cards.

That's, I think that's the negative case. It's obviously somewhere in the middle. By the way, you definitely started out being like Google is adding a chatbot to Gmail and we ended with it's either paper clips or gray goo and you just got to pick. That's like, that's just the natural trajectory. Anytime you get AI into the conversation, you just spin it there immediately. You can't help it. Yeah. Like we as people can't help it.

Wait, before we switch away from this, James, can I ask you? I want to follow up on one thing that you said a few minutes ago. Sure. I want to know if I heard you correctly. I think what you said is all of the heat in AI right now is around consumer stuff and like regular people stuff. Do you actually think that's true? Like is that going to be the most interesting space? Yeah, because I think the research side, we have just not caught up with this at all.

Like there is so much research at the moment that we haven't yet productized that a lot of attention is going into that. If you go on the subreddit's for like our machine learning, what you will see right now is a lot of people who work on natural. language processing, this is the field that underlies language models, in universities going, well, our job is useless now. Literally, no, I mean, I need to write a story about this because there are a lot of people in academia despairing because what they do has now been outclassed by GPT4. And they're now going, why would we ever bother to continue this research? Because this huge corporate commercial lab has done it better than us. So I think we're just catching up with this stuff. There is obviously, there is more interesting stuff coming down the pike.

Audio generation is one, video generation is another. That's really going to mess stuff up. Or it's really going to, really going to cause a lot of weird stuff. But I think we kind of have enough to be going on with in order to sort of integrate all this new technology into society. I want to say, I sort of out of everyone here, I am nearer to Alex's stance on this in terms of what the effects of this productivity software will be on our lives.

Because I think, has there ever been a year in my life when I got less email than the previous year. No. I think people just create this work for themselves when it's not very necessary. And that is unfortunately how a lot of, you know, the corporate life is structured. So I sort of agree that people will use this to be careless about messages they send to people rather than be careful. But it's a balance. We'll work it up.

Well, you just got to find something harder. You know, like my work Gmail is a wasteland, you know, but there are people get, through to me on channels that are not nearly as automated. And I think that's just, like, fascinating. Like, in a deep level, sending me an email is at once the most personal and, like, highly valued piece of communication that anyone can do. And it's the most versatile in many ways. It is also the, I am just inclined to ignore the vast majority of it,

because that's what the flood of automated email messages that I get has trained me to do. Yes. And like the tricks that people play, I don't know if this has. happens to other people. It certainly happens to us as journalists. The tricks people play to get our attention in email, like the robots have picked up on them and like, I just get calendar invites. Yeah. Or like emails that are like, I'm following up. Like you haven't responded to my last email. And it works like a lot of the time. I'm like, did I, are they mad? No, they're just lying. The robots just lying to me. That's the one I forget. Anytime somebody says I've been following up, you didn't respond to

my last one. I'm like, that's a lie. I never looked at your first one. I don't care. about any of your emails delete i don't care about anyone how could i care about you find a better way to communicate with me goodbye i want to end on this note uh so we talked about microsoft workplace software we talked about uh google workplace software where open eyes going with the chat stuff this is obviously for better or worse paper clips or goo this is the future of most interfaces to software right we can see the sort of natural language component has just it's taken off and like everyone's imagination The companies are racing towards it. The people, consumers, like, know about it.

We're seeing people use it in ways they're unexpected. Joe Biden has strong thoughts about audio mixing. Apple is, like, woefully behind. And so the Times this week had a piece about Siri, Alexa, and Google Assistant, how they're all about a decade old. They were supposed to do this. In 2016, we ran an entire piece about chatbots in the future of the internet. Like, we lived through this cycle before the technology was ready,

or that this technology really existed in this form. And the way they architected, protected this technology, did not allow it to become this. And really, it's the part where I can respond to any query, being creative. But here's the line that jumped out to me from this piece, which is good and worth reading, because it's just a good recap of what went wrong. At Apple's headquarters last month, the company held its annual AI summit, an internal event for employees to learn about its large language model and other AI tools,

so two people who have reached on the program. Many engineers, including members of the Siri team, have been testing language-generating concepts every week. So we're all at, oh man, Google's so far behind because they haven't just let anyone talk to Bard or whatever. And Apple's just still testing it and like Siri's still a joke. It's like not very useful at all. James, do you think this is just a matter of replacing the Siri back end with some LLM and Apple sprinting ahead or is do you have to do something more more drastic to compete? Well, I mean, I think I disagree a little bit with the premise of your question, your question because you're saying that we're now secured that this is the future of interface,

that it is going to be conversational. I don't think we've yet worked that out for sure. I think this year is going to be the year of testing and it's going to be the year of norms changing and we're going to work out where we're going to have to find out because we're not using it yet. It looks very good. It might work. It also might still be kind of awkward and clumsy to use And it might be the fact that these natural language interfaces are only used in very specific things. And actually, people still want to click around. They still want to tap menus. They still want to get that precision of control and that speed of control.

So I sort of disagree that this is necessarily the future of computer interfaces. Whether Apple can catch up, yeah. I think they can't. I mean, like, you know, the work that they've done some really interesting small bits of work around this, like creating a customer. version of stable diffusion and AI image generator to work directly on Apple Silicon, we can't yet get Apple, we can't yet, sorry, get AI language models working on a phone locally, right? They're still too big. They need to be compressed a little bit more in size. When that is available, that's, I think, when Apple can make that switch. Because right now,

it all needs to be sent off to the cloud. And if Apple is going to stick with everything it has in terms of marketing about privacy, keeping it local, under your control, they're going to want to do this stuff locally. So I think they will be able to catch up when this becomes a possibility, but actually, iOS is their biggest platform. It's their most important one. That's where they need to get this stuff working. And I don't think it's, it's not technically possible just yet to get it working locally. So I don't think they've yet lost a lot of ground necessarily. They will be able to catch up, I think. Yeah, I would argue one of the biggest mistakes that, you know, Alexa, Siri and Google Assistant have in common was betting everything on voice as a primary thing.

One, because voice is really hard. Speech to text is really hard. Text to speech is really hard. Natural language processing is really hard. So like just starting with a chat bot is actually like a necessarily easier computational problem to solve. But then also all these companies said voice is going to be everything you're going to live your life with microphones and speakers. And like that turned out not to be at all even remotely true for all the reasons you're talking. about James. Like people like to click. People like to look. And when it's like, okay, I have a, I'm going to

yell at my screen, but then it's going to show me something on it. Like that doesn't actually make any sense. So what chat pots did is just like take a UI affordance that people understand and use all the time and give it back to them. And I think like Satchanadella is quoted in this time's piece. He said something and they quote him in this time's piece. He says something to the effect of voice assistants are dumb as rocks, which is like a really easy thing to say when your company made Cortana. like congratulations. But I don't, I don't agree with that assessment. It's just that we, all these companies did them wrong.

Yeah. But I don't think that fundamentally the science underneath them has gotten drastically better over time. And I think it is much more of an interface problem for a lot of these companies than a tech problem at this point. Let me just call out that in the height of the Siri versus Google Assistant Wars, our dear friend Paul Miller, who used to be a co-host on the show, complained vociferously that you couldn't just type to these assistants.

And no one listened to him. So yeah, they should have that. Yeah. He called that one. But I don't think it is just a matter of, uh, the underlying technology or they built it wrong, right? If you swap out the model and let Siri answer any question, you do still have the do you type to it or do you talk to it problem?

And I would wager that if you asked most people, okay, you can either type in a command to your phone or you can look at it and say, take a photo of the moon and it'll just do it for you. Most people are like, I just want to, I just want to tell it things. right. I've got this Star Trek communicator around me all the time. I just want to tell it to do stuff. And that's what Siri was supposed to do. That was the

promise of it. And now you potentially have to swap out its entire architecture and build a model that runs locally because you're Apple and all this stuff. But you can see how you can glue those concepts together. And you might end up actually being it like for the first time in quite a while being able to rethink what the

interface of a phone is. I'll bring up another Vurchast favorite, Bixby the dog Samsung's assistant. Where the promise was you would be able to change the settings on the phone. Yeah. Right. You would be able to say, hey, like, I don't want to get text messages between 8 midnight tonight. Just do it.

And Bixby would, like, adjust the do not disturb settings on the phone. That did not work. I think we made Bixby more famous than Samsung ever managed to. I'll just take credit for that. It's dog wearing shoes. But if it could, that is actually a good idea, right? And like, it's these models that would actually enable something like that to happen in some way. And that's the convergence.

And I think it's very funny that Apple's privacy stance in many ways prevents it from happening right now, right? They've got to increase the processing power of the phone to make that happen locally. And I want to say that that is a use case that GPT4 specifically enables because it can process image alongside text input. It's perfectly positioned to have that sort of control over the operating system. that is also what a lot of the AI safety researchers worry most about, because at that point, you are no longer constraining the systems activities to a little box, to generating text. You are now saying, you can take control of my computer and you can make actions on my behalf, whether that is buying something on the internet, whether that is selling something.

And I think that is where the safety problems really come back into it again, because you're like, okay, how do I limit, how do I, how do I balance making this thing respond to my commands in a way that doesn't have the same problems as past voice interfaces, which is when you tell it to do something and it doesn't know how, how do you balance giving it that control with not letting it become out of control as well? So, yeah, that's another problem, but that is something that is technically feasible on the near horizon and is going to change very soon. And I will say that in the commercial context, already deployed at the scale of

multi-billion dollar companies, right? Like, UiPath is a robotic process automation company that basically just builds computer vision systems, AI systems that use, like, Windows 3.1 machines to do hospital billing on their behalf because it's easier to have AI control your Windows 3.1 system than it has to replace your Windows 3.1 system. And so in another context, Daniel Dines, the CEO of UiPath was on Decoder. Go listen to that episode. Because we talk about at what point are you just, is it just robots talking about? talking to robots. And he was like, I don't know, making a lot of money. But he was very thoughtful. Like he knows he's building kind of a weird closed loop. And so it's already happening there. And you can see how this brings it to this context outside of the constraint of the person who's

interested in using this is a hospital insurance administrator. And that's usually one thing that's sideways. Hospital insurance administrators, very safe. They should have all the AI tools. All right. We are cruising our way to a two-hour show, which David promised me that we would not do. James, thank you so much for coming on. We've got to take a break. There's a bunch of stuff in the lightning round that we should get to when we come back. But James, thank you so much.

My pleasure. My pleasure. Support for this show comes from Whatnot. Whether you're selling online or out of a storefront, you already know the challenge. You're simply hoping for people to find your listing

or waiting for them to walk in. But Whatnot flips that. They say they're the live shopping marketplace where you can shop, sell, and connect around the things you love. On What Not, you go live and sell directly to people in real time. They see what you've got, ask questions, and buy. And they keep coming back.

Whether it's beauty, collectibles, electronics, luxury fashion, and yes, even cookies, sellers are building real thriving businesses. And for a limited time, What Not says they'll match your first $150 sold in the first month. You can visit What Not, dot com slash sell to start selling. That's W-H-A-T-N-O-T dot com slash sell. What-N-O-D-com slash sell. Support for the show comes from Anthropic.

Not every question has an easy answer. And the ones that are really worth asking usually come with a healthy mix of inspiration and backpedaling. aha moments and quiet meditation. When you're working through one of those problems, you want a partner to bounce ideas off of and figure out where the deeper issue lies. That's where Claude can help.

Claude is the AI for minds that don't stop at good enough. It's the collaborator that actually understands your entire workflow and thinks with you, whether you're debugging code at midnight or strategizing your next business move. Claude extends your thinking to tackle the problems that matter. Plus, Claude's research capabilities go deeper than basic search. It can have comprehensive, reliable analysis with proper citations, turning hours of research into minutes.

Ready to tackle bigger problems? Get started with Claude today at cloud. com.com. That's clod.aI. slash vergecast. And check out Claude Pro, which includes access to all the features mentioned in today's episode. Claude.AI slash birchcast. We're back.

We've got a lightning round. Honestly, I can just talk about the trailer for the Blackberry movie for the next 20 minutes if you want me to. 45 minutes. Go. Let's talk about it. So there's a new movie called Blackberry. I've seen it. Do I tell you this?

I've seen it. When did you see it? I watched it on Monday in my hotel room at South by Southwest. We're going to have more to talk about it on this show at some point in the next few weeks. I also interviewed Matt Johnson, the director, about the movie. We had a very good time. We're going to run that on this. How are you telling me this now?

We should have called me right away. I like to bring surprises to the Verge casting. Oh my God. Is it great? It's great. It's, it's Liam has also seen it. Liam, I don't know if you have thoughts, but I, most tech movies are bad. I would say is like my general stance on tech movies.

This one is not bad. It's like weird and funny and enjoyable and also lands in sort of an unusual place about Blackberry. It's like it doesn't just tell the like Apple showed up. with the iPhone and a bunch of Canadian idiots didn't see it coming and goodbye Blackberry. The story, like, is more complicated than that and lands that way. And it's a very good movie. Everybody should see it. It's based on a book that I loved when I came out called Losing the Signal, the untold story behind the extraordinary rise and spectacular fall of Blackberry. Excellent book title. I feel like that title really set the stage, like the formula for almost

every textbook title to come. That book was great when it came out in, like, for 2012, 13. We had a big feature that we re-promoted in 2012 BlackBerry. The wave of nostalgia that I felt, do you remember the BlackBerry storm came out? We all thought they were going to take it to the iPhone. Did we? These are like our early days. We were children. My favorite, though, was the storm came out and it was like, this is the time, this is coming.

And then my distinct memory is the first time, I think this happened to everyone. You held a Blackberry storm in your hand for one second. And you went, oh, nope, never mind. This ain't it. Like, the whole thing clicks. The whole screen clicked. They thought that was the big innovation. Because they wanted to bring the keyboard back and it was a full stretcher.

Yeah. But it was, I think the, in particular, the Blackberry devices and their sort of failed attempts to compete with the iPhone, which is not what this movie is about. But that for me was when I realized it was not the hardware that was important. The hardware is important. If you listen to the show, you know, I think hardware is very important. But it was like they didn't change the software. They thought that people just wanted a touchscreen. Yeah.

Yeah. And no one actually cared about the touch screen. And that was, which is a very formative moment in this trailer, just watching this trailer. If you're a longtime Birchcast fan, go watch this trailer. And it will just bring you back

to that entire moment, which was super excited. It's like where we came from. That's enough nostalgia. But go watch the trailer. And then David, when we're done with this, you tell me everything about this movie

and great detail. The movie comes out in what May. So it's coming out soon. Early May. If you're in Canada, it comes out before then. Of course. Yeah, we're going to have,

we interviewed Matt. We're going to have him on the show. I think sometime in it. April. You know, the United States, it comes out May exclusively on Verizon. And if you, if you pay to see it, you will eventually get your money back. That's the worst joke I've ever told on the show. Bad joke to it. Extremely strange bit of news from TikTok. So I think a couple of weeks that we talked about the House passed a bill that would allow the Biden administration to ban TikTok.

That obviously created leverage for the Biden administration. Now the Biden administration. saying, look, you're either sell it or we're going to use this power that we have. And TikTok, this reporting is weird. So it first broke in the Wall Street Journal and then it was confirmed by the New York Times, but TikTok won't say it out loud. Like they said it to those two newspapers and they will not respond to any comments from anyone else. The White House isn't talking. But you can see what is happening here.

There's so much bipartisan concern about the Chinese ownership of TikTok that the Biden administration was given this leverage and now they are using it. And the funny thing is like, this is exactly what we're. what Trump wanted. He was like mad at TikTok because the rumors about his rally not selling out or whatever. And he's like, I'm going to ban it. And they try to ban it. And there's this whole thing where Microsoft almost bought it. And they were going to sell it to Oracle. And then we back that off because the process there was bad, right? The procedure of the government being mad at you and then you have to destroy your business was bad. Right. The president is mad. Now you have to sell your business. It's

like not a good operating principle for our government. But the outcome that we're about to land on is the same, right? which is we now have done all this process and there's this whole plan and there's Project Texas and Oracle data centers and we're still like, yeah, that's weird. You've got to sell.

And I bet Oracle's applying it anyway. I would agree. They certainly seem to be the leader in the clubhouse to be the ones who buy it at this point. But it also seems like if we've gotten to this point, is there anything left

for TikTok to do to convince anybody that it is above board? I should say, I sincerely have no idea. Like, there are lots of very smart people who argue in both directions. TikTok has gone way out of its way for years to talk about where it stores its data and how the corporate governance works and all this stuff. I don't know.

And I'm not convinced that anybody does. But it does seem like we are past the point where, like, the CEO of TikTok is coming to D.C. to testify to Congress next week. And I literally cannot imagine what he would say that would convince anybody of anything at this point. I don't think he can say anything because the general consensus in, like, the intelligence community. is that if your company is based in China, you are a Chinese company and they have access to all of your data, regardless of where you put it and everything else, because that's how China works. It doesn't matter if you're Apo, it doesn't matter if you're bite dance, parent company of TikTok. If you are based, if you're Huawei, if you're based in China, China has its fingers all over your data, all over your infrastructure. And that's like the general consensus of the majority of the intelligence community. So if that doesn't change, around the world, not just in the U.S. Yeah, around the world, not just in the U.S. I've spoken with experts like in Canada about this as well. And if they're all saying that, then I think it's a really difficult thing for for TikTok to do to convince you because how you can

say it all day long. Oh, definitely I do not give my data to China. But there's no proof of that. And there's no way you can prove that. It's just a big broken trust issue. And it's larger, it's part of that larger problem of nobody trusts China. Right. So one notable exception to that assessment, which is 100% accurate, is Apple. Apple does a lot of business in China. They have, right, they've turned over the ICloud servers in China to a state controlled entity. Yes.

They obviously manufacture the phones there. And they seem to just, it never comes up. Like, it just, they're not based in China, though. So they, they have part of their company is in China. And that part of the company is very tightly bound with China. But the American company is not based. Like, it's an American company. It's the same with like AMD and stuff.

Like, there's a lot of companies that do significant business in China. And we tend to give them a pass because, oh, you're based in the U.S. Like, it's a little bit of xenophobia that drives all of this. And it is chin to the intelligence like conversation. It's definitely xenophobia is part of this, right? But yeah, I just don't see how anybody can kind of get out of that. If you're a Chinese-based company, you're always going to be fighting this battle. Well, and it is true, Nelai, that if you play the argument that is being used against TikTok all the way out, why shouldn't we be. looking at the companies who manufacture things at Foxcon. Like, remember that big Bloomberg story about the tiny chips that were being embedded in things that turned out to probably not be true? Like, God only knows.

That story. One of the, one of the truly weirdest moments in tech reporting, Bloomberg drops this big business week cover story. And every company that never likes to talk about anything comes back, comes out and is like, nope, not that. But like, if you're going to scrutinize these things based on the idea that if you have any connection to the Chinese government, there is probably a way for the Chinese government to influence your business and your company and your users, like, you're going to,

you're kind of right that there is no end to that. That eventually consumes the entire tech industry. Well, I think, I mean, that's one of the reasons we're seeing a lot of tech companies say we're moving out of China. We're looking to India to manufacture iPhones and stuff. Everybody wants to show. Looking, being the operative word. Looking, right? They're going to be looking for years.

Everybody's really interested in the idea. Yeah. They're just hoping this all blows over. Like, I think that's part of it. But at the same time, there is a, there is definitely like a functional difference between Apple doing business in China. and bite dance being based in China. Oh, I don't disagree. I'm just saying that functional difference is kind of not well defined.

Well, one can be kidnapped by the government and held until they pay taxes. Tim Cook, I think, would have a more challenging. Apple owns how many seats permanently and however many flights to China every day. That's true. The Chinese government can destroy Apple's business tomorrow. Or Tesla. Tesla is a good example of this, too. Right.

Everybody has talked about, like, the way that Elon Musk is running Twitter. could change the way that Tesla operates in China. It's like all this stuff just is tangled. And it like I don't know how you pick which one to be mad about at some point. So we'll see. I just, this was supposed to lightning around and we're like deep into international relations. Sorry.

But we'll see and I think this ban again, it will just, it will set off a conversation on where that line is. And right now it feels like we know, right? Some things are owned by Chinese companies and something's run by American companies. But if you're like, this threat is so big that we will shut

down a thing on the scale of TikTok because there's political will on both sides. And also, what politician does not want to go home in an election year and tell parents we shut down TikTok? So I think there's quite a lot to come here. But we'll see how this hearing goes next week. McKenna is going to be there. I think that we're sending back up with her to make a video about the hearing. Stay tuned. Can we talk about better TikTok news really fast before we get off of TikTok? Yeah. TikTok launched my favorite thing to exist on a social media platform in a very long time, which it's calling refresh. And the idea is basically you can go into your TikTok settings

and you can say all the things that you have learned about me that govern what I see on my For You page. Never mind. Let's just start from scratch and build this up again. And I want this on every single algorithmic feed that exists. Like my YouTube has been absolutely overrun by people telling me what gadgets to put on my desk because of like one story I wrote

and one admittedly like months long obsession that I had. And my TikTok for you page is only, like I'm not kidding, only videos of Maddie Healy, the lead singer of the 1975 singing, O'Caroline on Saturday Night Live. Like two out of three videos are different angles of that same thing. And I just want a thing that is like, I get it. This was great. We had a good run. Let's start over.

And like, thank you TikTok for saying, letting me pull myself out of this hell that I have brought myself into. But don't you want to be able to save your old algorithm? That's the one thing that this feature needs. That's a good idea. Just save whoever I was two years ago, and I'm going to start over to be a new person. That's fun. That's like on Spotify, I start a new playlist on the first day of every month.

And then every song I come across that I just kind of sort of like, I just dump in there. Oh, that's cool. And so I can go back to like my, you know, July 2017 playlist and it just like instantly transport me back. That would be so fun to do on these services. It's like, who was I when this was my algorithm? It was just new girl clips and office bloopers. Exactly.

The depths of my algorithm in, like, lockdown. Dark. No good. How to make sourdose starter was the long run that I had. Yeah. It was just a real. Anyway, that's my only request for TikTok. Let me save the algorithm.

We talked a bunch about net neutrality last week, the FCC, Gigi Sown. We kind of has a story this week. Biden's just, like, running out of time to add a fifth commissioner to the FCC so they can bring back net neutrality. There's actually a lot of action on that front. right with these state bills that would block like in Texas a state bill that would block any abortion information on the internet in Texas crazy stuff it would be good to have a federal non neutrality law at this time Biden's just fully running at time because it was disaster at juju-sone we've got more on that story to come and then on sort of the same note

uh Ryan Reynolds owns mint mobile which is a T-Mobile and v-no so they just resell T-Mobile's network but for cheaper prices and funnier advertising T-Mobile's just going to buy it the number is for up to 1.35 billion, depending on some escalator clauses. On the one hand, you could argue this doesn't reduce competition because it's all just T-Mobile and you're all just running on T-ball on the end. On another hand, you're like, oh, there was a challenger brand to AT&T, right? And now they're just going to go away. And it's like, we get ourselves to have any competition in mobile service in this country.

Like, whenever there's the barest whiff of it, we're like, no, let somebody buy it. Liam, I need you to turn on your camera and unmute your microphone. Liam is a longtime mint mobile customer. I need to know your feelings, Liam. It's time. This is your moment. You've been preparing for this all week. Fine.

I'm not camera ready. I've been a Mint Mobile customer for eight years. As someone who used to be in the freelance world, it's an especially attractive service because you can pay by the year. And that's one of the few companies I can think of that, let you do that. I've also been shocked at how good the service is on Mint.

It's better than it should be. Sure, when I go visit my folks in Atlanta, Georgia, it's sometimes flaky, but for the most part, I like to say you get like 90% of the coverage for like 20% of the cost of going to one of the three big ones. So kind of furious at Ryan Reynolds. Like most people, I mean, like most people in this country, I think he's just a lovable guy. How could you not like Ryan Reynolds? But now if Ryan Reynolds were involved in a car crash, I wouldn't be that upset about it.

I'll say that. Wow. Ice cold. You know what? Ryan Reynolds would appreciate that joke. I've got to be honest with you. I think he'd be fine with it. He's going to make a commercial about that joke.

We are supposed to have four national wireless carriers in this country. One of them is so irrelevant. Systems have been crashed for weeks. Yeah. Is Dish still down? I should check. We are the only site that covers it. That's how relevant this company is.

It's like negative competition. and then even the little MV&O competitor that managed to make a name for itself is not getting swallowed up. We're just horrible at this. Americans pay more money for slower speeds anywhere else. Wait, can I read you just in the spirit of, I'm just going to do this every week now. I'm going to read you the notice at the top of Dish's website. It says, thank you for your interest in Dish TV. Your needs are important to us.

We appreciate your patience at this time while our teams are working hard to update our full website and get services back up to help you. No one gives a shit. Can you imagine if AT&T was down for this long? Right? Like there'd be hearings. Oh, man. Like, no one cares. This is supposed to be the fourth competitor to AT&T, Mobile, and Verizon in this country.

And now they're, anyway, whatever. I'm going to just set up my own renegade network, like in palm of the volume. The FCC is going to have to drive around in vans trying to find me. $19. $19 a month, you also get a bag of moon crystals. Okay. So the other thing I want to talk about, just while we're doing, like, lightning around stuff here is Emmett Shear, the CEO of Twitch, just suddenly and summarily

resigned from Twitch, just like dropped the mic in a blog post and left. This doesn't seem to be quite as sort of earth-shattering a thing as when Susan Wojewski left YouTube, whatever that was a few weeks ago now. But this is still a big moment. He's been running Twitch for a long time up through the Amazon acquisition. Twitch is huge. He is like, he is the face of Twitch in every meaningful way. He was running it before it was Twitch when it was still Justin.tv. Yeah, that's right. He's been doing this for ages. What was it? 16 years, I think he said. Yeah. And they sold it to Amazon for less than a billion dollars in 2014, which is just amazing to think about it in the context of all the deals now. And so it's not like

a buyout situation. Yeah. Right. Like usually those are three, four, five year deals. This is, he's been at it. I'm very curious. I mean, he basically, he wrote the thing where he's like, the company's my kid. And now it's time for the kid to go to college, which people sail. a lot. I mean, you look at this and like chat GPD could have written this, right? Like, it's not that far off. We'll get more color. There's a new CEO at Twitch, Dan Clancy. Obviously, this thing is still owned by Amazon. I don't think there's going to be any meaningful disruption there. But another sign of the time, right? That's Susan Wojcicki leaving YouTube, Emmett leaving Twitch. That generation of tech leaders is moving on. Uh, okay, real lightning

round stuff. The real lightning round. Now the fake lightning round, which was not Swift at all, is over. The real one. now onto the real lightning eye. Alex Heath interview, the co-president of Spotify, Gustav Soderstrom, on Decoder. He said, Spotify, hi-fi is still coming at some point. And then Chris Walsh did a little bit of extra reporting. The thing has been ready. Spotify employees have had hi-fi for a long time. It's Apple and Amazon blew up the entire cost structure of lost list and hi-fi music by giving away for free. So Spotify can't figure out

how to charge for it. So it's ready. The business model is not ready. Amazing. You just described Spotify. Basically. Wemo, the Belkin Smart Home brand. They announced a bunch of Matter stuff last year at CES, this year waiting for more. They're done. We have a little scoop.

They're done. Alex, this is yesterday. A freelancer hit me up and said, hey, I think Wimo just accidentally announced to me that they're done with Matter for the time being. And so they said they confirmed they are taking a step back. They are not planning to update anything that. already out there to support matter. They're not planning to release anything on matter. They just kind of want to wait and see. And the general consensus from folks we spoke with like all around is it probably is something that's a lot of people in the industry are worried about, which is

becoming a commodity and not have anything special, having to be being belkin and having to compete with like these tiny, tiny companies making smart light switches. So they're like, Ours isn't going to be any different from theirs. Why should we even be in this business? Step back. Because you could make one that's better and works better. I guess they don't. I guess they disagree.

If anyone from Balkan is listening to this, I just, I believe in you. And the garbage ones will be garbage and fall off the network and restart and sometimes just die. And you can make a good one that you charge slightly more money for. I think, didn't you just describe in you? Every platform and protocol war, there has. ever been. It's like everybody's like, what if we build open things? And somebody's like, I don't really feel like competing with that. I'm going to build my own thing. Yeah. I mean, this has been the story of the smart home, right? Is they've all been these

weird little ecosystems. I will say that I own a bunch of Wimo stuff. I have a much of I home stuff, whatever. Are we losing anything? Do people love these products? I don't. But Belkin, I believe in you. Yeah. Just look at me. I believe in you. Just right in his eyes. You can make good stuff that people will pay slightly more money for. You can do it. Okay. That's it. I hope the CEO of Buckin is listening.

We're in the midst of what I would say is a personally devastating football news cycle for me. But YouTube TV has the rights to Sunday ticket, and they've been promising multi-view for sports for a long time. They've started testing it. You can watch four games at once. You can see how this is just going to be for Sunday ticket when it happens. It's for a specific set of things in a specific set of ways.

You don't get to really choose it. it's like the least multi-view you could do while still being multi-view. And also, I don't know if you saw, but while we were recording this, YouTube TV announced it's raising its prices another $8 to $72.99. So it's just cable now. It's just, it's just full cable. And this is the world in which we live. That's so funny.

That's so sad. And I'm going to pay for it because they have Sunday ticket. Yep. That's how they get you. That's going to work. David, you looked at the Camel app, which I think I agreed with you. reading your piece. I thought continuity cam on the Mac would get rid of it. This is Apple let's use an iPhone as webcam, but you say the new one's pretty good. It's just,

it's power user continuity camera basically, right? Like the thing that Apple likes to do is try to make everything works sort of magically and simply. And what that means is not giving you any options or any controls. It either just kind of works or it doesn't. And continuity camera switch credit works more than it doesn't, but you don't get a lot of control over how it looks or the lighting or something like that. And Camo just jumps in and is like, oh, would you like to choose which lens you use and how much you zoom and how much exposure you have and what the lighting looks like and you can do all of that. I'm using it right now. So if you're watching on YouTube, you can be the judge of how it looks. But it's great. It's expensive. It's 40 bucks a year or

$80 once, which is too much money for what it is. Like if you're not a person who records and publishes video all the time, it's probably overkill. But it really, like for the purpose it does serve, it does it really, really well. I'm a fan. Do I get a free upgrade? I bought it. I bought the lifetime last time. Do we get an upgrade or you get it by it again? I'm pretty sure if you bought the lifetime when you get free upgrades. Oh, that's nice. I bought a one year subscription that I laps and I got it cheaper. But I think if you buy the lifetime, if you don't, that feels like bad advertising on their part.

I don't know. I only use it when I have to like be on television on the road. And it is fine. But it's also because I refuse to upgrade my macOS. I'm on every other year's cycle. And I'm sorry, this year isn't it? So Camel Camad is. All right.

The last story. Another one of the, I would say, top 10 Verge ideas, verge stories that we've encountered recently. The CEO of Match Group, which is the company that owns every dating app in the world, except for Bumble. Yeah, that's like barely an exaggeration. And I think they're suing Bumble.

So it's going great. But they own Tinder, they own... Okay, Cupid, I think. Match. They own OKCupid. We cover them a lot. When Ashley Carmen is here, we did a lot of coverage of this one company, how big it is.

But they've got a new CEO of Bernard Kim, and he was at a summit, and he's an ex-gaming CEO. And he was talking about why it's worth money spending. He was talking about why it's worth it to spend money on Tinder. And he was like, well, people spend a lot of money in games. And here's the quote. He says, no one plays these games forever. After a certain point, people churn out of a game experience. I've personally spent $50,000 in three months on Clash of Clans.

And I look at back in that with a lot of shame. I'm like, what did I really get out of that experience? Nothing other than like a really amazing wall, which is not cool. And then he's like, but I paid for Tinder and now I have the life. Which in a generation of CEOs that might be mobile gaming whales is an incredible thing to think about. Wow. And that's like, you would think when somebody says something like $50,000 in three months on Classical Clans is like a super exaggeration. not an clearly not an exaggeration like my man clearly spent 50,000 dollars just he spent the money this is

the follow of quote now look at what can come out of tinder you meeting your wife that's an measurable reward something that will last a lifetime something that will lead to unbelievable happiness maybe sometimes despair but at least you're feeling something wow this is just evil shit I've ever heard in my This is at a conference? But it's wrapped up in this dude being a clash of clans whale. He's Apple services line. 30% of that clash of clans money went to Apple services revenue.

The best part about this story is that Mitchell kept reducing the denominator of the map. So he started with $50,000 in three years. And then he's like, it's $16,666 a month. And then he's like, it's $5.55 a day. But you know it is happiness worth $550 to $5 a day, Eli. Look, what can happen? You can get a wife.

You don't spend it on the game. You spend it on the wife. He invested it wrong. I'm telling you, if you know that the CEO of your company is a clash of plans whale, please reach out to us. I would just like to compile a list. We'll never even publish it. We just want to have it.

It's like, no wonder all these people were in NFTs. This dude's like, I got to sell this wall and there's no one to sell it to. My $50,000 wall. That's really good. Okay. I promise you that I have not spent any money of Clash of Clines. I did once spend some money in Candy Crush and I felt such overwhelming shame that I had cheated my family out of what was rightfully theirs that I never played the game again. So you did even worse.

You spent the money and got no happiness. At least he spent money. I was like, I already got a wife. It's her money. He got a nice wall out of it. He got nothing. That's horrible. All right, we are so far over. Thanks to James for being on. Thanks to all of you came in sauce at Southwest, Southwest. That was really fun to meet folks. This week, we got solo acts. This is a good one. Marco Arment, the developer of Overcast is going to be on. We've been wanting to do some kind of podcast with him forever. So I'm excited about that one. And then what's on the Wednesday show, David? We're going to talk about iPads. Dan Sefer wrote two consecutive really good things about iPads. So we're going to talk about Dan's feelings about iPads, which long time listeners.

will know he has many of. We're going to talk about ATSC3, just so Cranza will shut up about it for once. That promises to be the longest segment we've ever done. And then Mitchell is coming on to do a really cool, fun, special thing that we're not going to spoil just yet. But it's going to be great. Yeah, very exciting. All right. That's it.

That's for Chast. Buckle. And that's a wrap for Vergecast this week. Thanks for listening. If you enjoy the show, subscribe in the podcast app of your choice or tell a friend. You can send us feedback at vergecast at theverge.com. This show is produced by me, Liam James, and our senior audio director, Andrew Marino. This episode was edited and mixed by Amanda Rose Smith.

Our editorial director is Brooke Minters, and our executive producer is Eleanor Donovan. The Verge cast is a production of The Verge and Box Media Podcast Network. And that's it. We'll see you next week.