Phones don't have lights

The Vergecast

The episode explores Meta's recent Connect keynote, focusing on the launch of the Muse AI agent and its implications for consumer privacy. T

Key takeaways

• Meta's Muse AI agent is highly effective as a consumer product but creates significant privacy risks by requiring access to Gmail and Instagram.
• The hosts criticize Meta's 'phones don't have a light' defense regarding the privacy concerns of camera-equipped glasses.

Main topics

• Meta Connect and Muse AI
• Privacy in AI Agents

Notable quotes

Phones don't have a light.

Conclusion

While Meta is successfully productizing AI agents, the company faces a

Transcript preview

Speaker 7 (0:00) Support for the show comes from Sonos. Most of us here at The Verge are big movie buffs. When watching at home, I don't want that experience to be any less immersive than it would be in a theater. And sound plays a huge role in that. The Verge has actually published a few pieces over the years about how much sound matters to a movie. From headphones to soundbars, a Sonos system pulls you in. So far, you might actually think the phone call is coming from inside the house. Visit Sonos.com. and build your system. Sonos. Love the sound, feel the system. Speaker 8 (0:38) Hello, and welcome to the Bridgecast, Speaker 7 (0:40) the flagship Speaker 8 (0:40) podcast of AI mascots. A little bit of butt. That's right. I'm your friend, Eli Patel. I'm back after two weeks away. That's what you were expecting. Jake Castronakis is here. So Speaker 4 (0:50) glad you're back for delightful intros like these, Eli. Speaker 8 (0:54) David Amell is here. What's up, man? I Speaker 4 (0:55) haven't left yet. Speaker 8 (0:57) We're never letting Speaker 4 (0:58) you leave. You Speaker 8 (0:59) understand this, right? We'll see. We'll see. The people demand that you stay in that studio and never go anywhere else. We're going to figure it out. Speaker 5 (1:05) I live here now. Okay. Speaker 8 (1:07) All right. There is a lot going on this week. There's Google Books. There's MetaConnect. Jake, what's on the menu? Speaker 4 (1:12) Yeah, it was a busy week for gadgets, Google Books. We're going to finish up with Brendan Carr as a dummy, naturally, now that you're back, the long way to return. We've got a lightning round. We've got to start with meta. So Wednesday night, MetaConnect, Mark Zuckerberg takes to the stage. This is their VR and glasses show, most recently. Actually, if you watched it this week, it turned into The Muse Show, right? But not the band. If only, if only they had brought on the band. A little more fun. That, yeah, the branding, I think, would have gotten a little conflated with the sexy, what is it? Bear? I don't know what it is. Speaker 8 (1:52) It's called the Jolly. The mascot's name is Jolly. Speaker 4 (1:56) That will fix their privacy. We're going Speaker 8 (1:57) to come back to Jolly, I Speaker 4 (1:59) assure you. I don't Speaker 8 (2:02) want Speaker 4 (2:02) to come back to Jolly. It became a show about Muse. Muse was everywhere in everything. I think this is fascinating, right? This is Meta's new AI agent. It's like three weeks old at this point. Two or three weeks old. Have you guys been using this yet? Have you instructed your Muse to shop for Speaker 5 (2:17) you yet, David? No. I do not know if this makes me a bad tech journalist, and it probably does. I do not feel comfortable with meta having my information. So I have not used it yet. This Speaker 8 (2:29) is the problem with Muse, right? Straightforwardly, every piece of coverage we do, all the comments, even we had an editorial meeting about it the other day. I was like, everyone needs to at least use the thing. And half of our editors were like, I don't, I won't. I was like, you have to. So I did, I gave it one thing. I gave it my personal spam Gmail. Okay. Right. Which is like. Speaker 7 (2:51) If Speaker 8 (2:52) Meta can look at however many Home Depot e-receipts it wants to, that's fine. And there was some boring thing. It's like we returned something overseas and there was like a tax form. And I was like, can you just figure this out? And it did it. Right? Like I bought a thing from overseas and we had to send it back and all this stuff. And I was like, I do not have the time or inclination to do this. And I just asked Muse to do it and it did it. And it did it by connecting to my Gmail and reading my email, which I don't love. And I don't love that Meta now has access to that. There are a lot of promises around Meta's privacy and end-to-end encryption with the virtual machine. And they hired Marlon Moxie Spike from Signal to build it. And Mark Zuckerberg keeps saying his name like a talisman. You Speaker 7 (3:41) know? Speaker 8 (3:42) Like in interviews, it's like, well, Moxie Marlon Spike is here. Nothing can possibly go wrong. But it is shockingly useful. And it's a... It's a good product. There's a lot of it that is like OpenClaw Part 2. Yeah. Like people have dug into it and a lot of the... The file Speaker 5 (3:57) names are exactly the same. Like the structure of the files are effectively OpenClaw. And one of the developers of Muse came on and said, we are heavily inspired by what Speaker 4 (4:07) Peter Steinberg Speaker 5 (4:08) did. Speaker 4 (4:08) By the files we downloaded. Yeah, exactly. It seems like OpenClaw. Speaker 8 (4:12) Yeah, that's Speaker 7 (4:12) the Speaker 8 (4:12) way it goes. The thing about it, though, is that it is a product. Like, Meta is a good consumer product company. They have the chops. They hired a bunch of designers. It works. Like, you can show it to a normal person, and it works as a consumer product, not like OpenClaw, which is a little bit terrifying, or any of the other AI products, really, are kind of rough around the edges chatbots. Muse is just a fun consumer product you can use. And the thing that struck me about it is I don't want it to do anything more than look at my spam Gmail. Right. Like that's as far as I'm willing to go. Jake, have you used it? Yeah, Speaker 4 (4:48) I've been poking around, but that is exactly it. Right. Like it is it's sort of like baby's first open claw or something like it's it's it's very friendly and cuddly. And it's like I can do anything. Just give me access. And you're like, oh, cool. I mean, I guess I'll just give you like my Instagram access because like probably you're tapping into that anyway. But then there's a ton of stuff it can do and can access. And every single time you ask it to do something, it's like, oh, if you want me to email that daily brief that I'm making for you based on the fact that I'm tapping into the API of Blue Sky to scan for posts that you asked about, I will need access to your Gmail account. And I'm like, man, like it just went off and did that. And it's able to build this brief for me. and paste it into its own chat. But if I wanted to email it to me in my inbox, I need to give it Gmail access. I don't necessarily Speaker 8 (5:40) want to give it Gmail access. Yeah. Speaker 4 (5:42) And that is just the problem with all of these AI tools. So Speaker 8 (5:45) do you believe, I saw an analyst writing, these downloads indicate that metas... privacy terms and conditions and its promises about secure virtual machines are resonating with consumers. I was like, I don't know what you're talking about. Speaker 5 (6:00) Consumers do not know what a virtual machine is, I Speaker 8 (6:03) guarantee you. I mean, this is the problem for Meta. And we should come back to Muse in detail because it is their next platform. It's their next bet. We can talk about Alexander Wang, the head of Meta Superintelligence Labs, just making that mascot sexy, which I think is hilarious and also very disturbing. But the privacy problem for Meta is super real. And it maybe got worse because of the glasses, right? They're putting cameras on everyone's faces and everyone's calling them pervert glasses. And the reputation that Meta is not good at privacy is everywhere. And it feels to me anyway, like it hangs like a dark cloud over Muse. Like, why would you give this company even more of your data? Speaker 5 (6:42) Yeah. Yeah, I think that this company has like over time used a lot of your data to make you. even more sad. So people don't really want to be like, have access to my Gmail, have access to my text messages. There's a level of trusting that meta will eventually in shitify and all of their products will just sort of extract more from you. So at least from my perspective, and again, I probably should try this, but the red flags that are going up in my head are just like, now I see the horizon where it's good for now. And then at what point does it start to become addictive or it starts like harvesting my data and serving me even more ads? The reason I don't want to meta like meta Raven displays is because I don't want ads like injected into my eyeballs, which is the long horizon of what I see with all meta products. And that's that's sort of the problem. Speaker 8 (7:33) So my feeling here is that they know this is a problem and they're just going to horsepower through it. Speaker 5 (7:39) Yeah. Yeah. Speaker 8 (7:40) They don't have a strategy besides publishing white papers about. secure virtual machines. They're just going to do it. Part of my feeling here is based on interviews Mark Zuckerberg is doing. He just did a big interview with Alex Heath. He did a big interview with Joanna Stern, both former Verge people. You can check out, but you can never leave. And with Joanna, he specifically said, most people think of Meta as a sharing company because we started as Facebook, but we also make WhatsApp, which is encrypted, and many more products we make will be private and encrypted in the future. They won't just be for sharing. I don't know if I buy this. Like WhatsApp is also for sharing, like in a very real way. And like Meta's business is advertising and there's no business model that they have that isn't advertising. Right. So they're going to have to come up with something else to prove that that's the case. And then the second thing, and David, this is just me wanting to prompt another camera crash out with you, which is my favorite thing to do, is their response to. How do you feel about people calling them pervert glasses is to just say it has a light. Like, that's what he's got. You can watch it. He just said it to Alex. We have that clip here. Speaker 6 (8:47) We designed the glasses from the beginning with these privacy considerations in mind, and we built the light into it. I mean, phones don't have a light. Speaker 8 (8:56) Okay. Phones don't have a light is... Okay, Speaker 4 (9:01) I see Speaker 5 (9:01) where he's coming from, but yeah. Speaker 4 (9:02) I like that he's just like, I just thought of this. Like, it's just... occurred to me. He's just saying the flash does not always go off when you take a picture. Speaker 5 (9:09) Sure. Speaker 8 (9:10) Not everything is a flash photo. Yeah. Phones don't have a light feels like meta, like meta has to hold on to that as tightly as they can now. Yeah. That's what they got. You're mad about these glasses. Everyone's calling them pervert glasses. And the only meaningful response is phones don't have a light. And. I mean, along the code, we don't have to go into it, but my immediate thought here was, well, you have to pull your phone out of your pocket, hold it up in someone's face. And honestly, maybe phones should have a light is a good response to this. But Joanna, several weeks later, asked a similar question and actually did the phone pulling out of the pocket move. We've got that here too. Let's watch this. Speaker 6 (9:47) I mean, this is the only device that I'm aware of that. it tells you when you're taking a photo or recording a video. I mean, people do that with phones all the time. And this is, we're just like, you Speaker 2 (9:57) know, this sort of you're holding your phone up, right? They're not as... Sure. Speaker 6 (10:01) But we kind of took this into account from day one. You know, the very first pair of glasses that we shipped, we're like, this is a thing where we want to lead on. We know that people are going to have questions about this, and we want to make sure that we're handling this responsibly from the beginning. So we built the camera light in. So Speaker 8 (10:17) it's that sure, right? Sure. It's very dismissive of the fact that you have to take your phone out of your pocket and hold it in someone's face. Sure, whatever. But it doesn't have a light. David, I know you've been thinking about this and there's all kinds of rules about photos in different countries. Yeah. Do you buy phones don't have a light as a response to privacy concern about the glasses? Speaker 5 (10:36) No, because the entire point of the glasses is to blend into your face. And much like when you're using your phone and it is obvious to other people that you're not paying attention to them because your attention is being directed in a specific vector towards your phone down at your at your phone. Generally, when people are recording on their Rayman glasses, they just click it really quickly and then they kind of go about their day. And the light is not the most obvious thing in the world. You know, like you mentioned the other countries in Japan. It's not a law. But they have an agreement among all of these cell phone manufacturers that the phones must emit a shutter sound every time you take a photo. And it's very loud. They try to make it very obvious that people can hear around them, see around them when they're being photographed. The light to me is not enough. You know, there are apps now that are like, oh, my God, someone has metaglasses around you. But I Speaker 4 (11:27) don't know. It doesn't it doesn't seem like enough to me. This argument would be so much stronger if it was a prominent light. It is not. It's a little pinpoint that they're doing some work to try to prevent people from covering it up or obfuscating it. But even in the best situation, unless it is dark, like this is a very easy to miss light, right? Like, except that the light is something. Speaker 4 (11:55) More lights. Like, light up the front of those glasses, Speaker 8 (11:57) right? Yeah, the whole, like, Kanye style, like, 2000s Kanye stuff. Yeah, yeah. Speaker 4 (12:02) Own it. New Year's glasses. I think, like, fundamentally, there is a light just doesn't work because the light is barely there. Like, it just, at its most basic argument, even if you accept that a light should work, should be enough. Yeah. Like... That light Speaker 5 (12:17) is not it. Well, and Joanna asks Mark Zuckerberg in the interview around the audio glasses, which we will probably talk about, which are just only recording audio. She says, well, why isn't there a light on those glasses? And he says, well, the lights for the camera and there's no camera. She's like, yeah, but you're recording people. And he goes. the lights for a camera. Yeah. Like, oh God, didn't expect it. Audio is a totally different thing apparently. I Speaker 8 (12:43) very much recommend everyone go watch the full Joanna interview. She makes him uncomfortable several times. I'm still stuck on phones don't have a light because it feels, what's the phrase? It feels like a thought terminating cliche. Speaker 7 (12:56) Like you Speaker 8 (12:56) just say this and no further questions, your honor. We're done here. Phones don't have a light. It feels like Meta's approach to their privacy problems is to come up with a cliche and bash you over the head with it. Right? It's the same as Muse not being allowed to go surf Amazon.com. And they're going to get to, well, it's the user's agent. So you're blocking the user. And it's like, well, this is a totally new way for computers to use the internet. Might it be worthwhile to consider how it works? Or are we just going to bash people over the head until we get our way? And I don't think that is the right approach to privacy. We actually have a big piece on the site right now from Janice Rose. I think Jake wrote this headline. It's very good. Everything is spying on you and there's no opting out. We are in an era of passive surveillance and it's a hijack of the law, which says you don't have an expectation of privacy in public. But that law was written 10 million years ago before cameras existed. Speaker 8 (13:53) We're not ready for this. Like the legal framework of privacy has nothing to do with anyone can quietly record you at all times. And I think that the social norms are actually way ahead of the legal norms here, where people are saying about flock cameras and ring doorbells and metaglasses. We don't like this. And while phones don't have a light is absolutely not the rule. correct response. And I don't think it's going to buy them the privacy reputation they need to get me to put my credit card into Muse. And this just feels like the circle Meta is going to be in for quite a long time. Speaker 4 (14:30) Meta has learned all the wrong lessons from their privacy scandals, right? They went, oh, when we apologize, everyone's still mad at us. And so let's just plow ahead, right? It doesn't matter. Let's just go for it. They did not try the thing where they respect what people ask. Like that, that is, I think what we wanted from them. And instead they're like, they're selling. I don't know what to tell you. Like someone's going to do it. May as well be us. Which in this case, Muse is by, you know, their account, a success, right? Has a million Speaker 5 (15:02) users or so. And we can get into whether or not that's actually a success, but it's a success for them, you know? So they're not going to Speaker 4 (15:10) apologize for something that's working for them. For sure. They need this to work too, right? Like their reputation in AI has been kind of disastrous. Zuckerberg went out and spent an enormous amount of money to hire Alexander Wang to reboot their AI system. And they've had basically nothing up until this month when they launched Muse, which admittedly is off to a very strong start. Right. I think Facebook, sorry, I think Meta can scale just about anything pretty quickly just due to the size of their network. But clearly they did put together the right elements of a consumer product to make something click here. And Speaker 8 (15:46) a lot of it is like hard-coded prompts. And we know they're hard-coded because people are getting used to send it, their file system and pulling out the hard-coded prompts. But, you know, the hard-coded prompt of like, go through my receipts and find returns, like. They're good ideas. They're what you would want from a consumer product. And this is really basic, make the product fun and easy to use from the jump design that Meta's good at, right? What is the onboarding flow? What are the first few things it's going to suggest? There's an ideas feed. These, again, are very basic product design ideas in consumer products, but Meta's good at them already. The numbers question is really interesting. The idea that it's a big success comes from the fact that it's at the top of the App Store. It has been for several days. It's early numbers outpaced chat GPT, which is a fascinating thing we can dig into. And, you know, there aren't huge scaled consumer AI products. Everyone's been waiting for one. I think there's some thirst to prove that an agent like this will be the future, right? This is the future of all computer interaction. You got a little fuzzy guy who's going to go do stuff for you. I don't know about those numbers, actually. First, David Pierce wrote an entire story for us about what it means to be at the top of the app store, and the answer is not a lot. So you can just set that one aside. Second, the early comparison to ChatGPT kind of forgets the fact that ChatGPT was a little bit of a slow burn, right? OpenAI put this thing out as a research preview. It took off like wildfire in a small community, but it really accelerated a little bit later. So these early comparisons are... You just have to be careful. And then... I think most importantly, Casey Newton has been pointing out that success and failure is very relative here. Like Threads had 100 million users because Meta is good at scaling a social graph really fast. And everyone still thinks Threads is a flop. And it's way bigger than X actually now. And so what does success here actually mean? For an AI product, it... For from meta, I think it is like Speaker 5 (17:40) I think it is actually just Speaker 4 (17:41) good vibes. Like I think that is success for meta right now. They need to be in the conversation and they are in the conversation. And I think a very major way. And you see that from Zuckerberg's presentation at Connect, where it is so fascinating to me. Right. Imagine Muse is delayed by three weeks. The presentation the other night would have been we have some new glasses and there are new styles. And their battery life is a little bit longer. And we would have been like, cool. All right. Sure. Yeah. You didn't change the light at all. But instead it became. Speaker 8 (18:13) We Speaker 4 (18:13) probably would Speaker 8 (18:14) have Speaker 4 (18:14) talked Speaker 8 (18:14) about the VR glasses a little bit. Speaker 4 (18:15) The VR glasses are quite impressive. But instead they, it felt to me like that show was retooled to put Muse through everything. And most of the Muse stuff was forward looking. It is coming to the glasses. It is getting computer use. It is getting an email address. Yeah. Soon. Right. They see that this is. high potential for them, but they have to get there. They're still building this. Every Speaker 5 (18:38) AI company dreams of releasing an agent like this, to be honest. Like all of these companies, yeah, they have apps on your phone, but they're not as integrated as something like news is, right? You have companies like Poke from the interaction company, which I believe got acquired. That lived in your iMessage. And that was their whole thing is like, treat this like a friend that you're texting and it can do things for you, but it needs access to all this stuff. That was from a company that nobody really knew about that people were talking about on Twitter. But when you have meta and you have the scalability of injecting ads for your product into all of the other products that three to four billion people use every single day, you're going to get something out of that. Even if you scrape 1 % of all your users, that's a huge amount of people. Speaker 8 (19:20) All right. So here's my question for you. I've been thinking about this a lot. There's a reason people are excited about a consumer AI agent, right? This is the promise of all of this technology. That you will have a robot butler that will just do stuff for you. Maybe it lives in your iMessage or maybe like OpenClock and live in your WhatsApp or Telegram or whatever. How do you think the people at Humane felt when they saw Mark hold up the little news charm? Because this is the whole thing, right? You're going to have a not phone device with its own modem and talking to the cell network doing stuff for you. And it, it'll be a little personality. Well, or the people at Rabbit. Speaker 5 (19:58) Yeah. Muse works. That's one thing. That's a big differentiator right there. Yeah, probably vindicated, I suppose, but also they're owned by HP now, so what are you going to do? Speaker 4 (20:11) Yeah, go ahead. When they release this thing, so they have this little puck, the Muse charm, supposed to come out later this year. I feel like this is going to be very revealing about the quality and capabilities of Muse. Right? Because we saw the Rabbit R1 and we're like, this hardware looks beautiful and then it can't do anything. And right now, Muse is fantastic in a browser tab that I can go to every now and then. Will it be good as a standalone device that I'm paying for 5G service for every month? Can it actually accomplish that much? Is it just like a cute little Tamagotchi? Speaker 8 (20:45) Yeah. Yeah. And like, how much will AT &T make me pay for this? Are they going to consider it a smartwatch at $10 a month or a phone at $40? I have a lot of questions about that, but it just occurs to me, all of Meta's scandals aside, and there's a reason we led with scandals and privacy issues and phones don't have a light, which I find infuriating. If you set all that aside, which I appreciate that not everybody can set all that aside, they did it. They did the thing that everyone is trying to do. They built a consumer agent. It doesn't even appear to be running like a frontier level model, right? They took all the technology that exists. They wrapped it up in a good consumer experience and they built the thing that Siri should have always been or Google should have been shipping for five years now or Humane wanted to build or Rabbit promised. Remember at CS they said they had the large action model and really it was like absolutely brain dead automations running on a Spotify server in the cloud. Like they built the thing that this industry has been promising to make. in one way or another for quite a long time. I actually do think that that is remarkable. Whether or not it's hacked together or in the back end, it is a bunch of bash scripts running on other people's models, which it might be. There's something there where they actually productized the promise this industry has been talking about for a minute. And now I think a lot of people see what it's supposed to look like, right? The gap between vision and reality has been vast. And I think Muse has in fact closed it. Yeah, I have a Speaker 5 (22:15) question for you guys though. Like, yeah. If I'm going to be paying, you know, $5 a month for an internet connection on this thing or whatever. Speaker 8 (22:21) $5 a month is such a pipe dream. Okay, $20 a month for an internet connection on my little muse Speaker 5 (22:27) puck. A lot of these models, I just don't even really know what to do with. Like we can test the edge cases of them. You can say, go in my Gmail and delete all of my Home Depot, you know, receipts or whatever. How often do I want to do something like that? A lot of these agents, to me, it just feels like it's a solution to a problem I don't currently have. And I would love to find out about what these problems that I don't even know about are, I suppose. But up until now, none of these companies have really given me a reason to