Meta's New $1300 Glasses and the Fight to Control AI

23 Sep 2026 · 40 min · 15 chapters

Ask about this episode

Ask anything about it. ChatGPT or Claude reads this page and answers with the times it was said.

Connect VO and ask about every podcast you hear, including the moments you saved. Add to ChatGPT · Add to Claude

In short

Meta’s $1,300 Meta VR Glasses, Meta’s AI agent “Muse,” and the broader debate over AI existential risk, regulation, and who should control AI.

Guest backgrounds

Mike Isaac, longtime New York Times tech reporter who has covered Meta extensively; now focused on the AI beat.

Key claims

Meta’s VR hardware is a downsized, cheaper attempt to make VR mainstream after Apple Vision Pro’s failure (too expensive, heavy, over-engineered). The glasses’ main question is still “what do you do with it?” Muse is Meta’s first major consumer AI win, showing agents can handle real tasks if users accept privacy/control tradeoffs. On AI safety, Isaac argues “kill us all” is shorthand, but real near-term dangers include containment failures and misuse; regulation may stall due to incentives and international coordination problems.

Notable examples

Apple Vision Pro (no one uses it; $3,500, heavy); Muse can schedule, negotiate bills, shop, and potentially make calls; Hugging Face hack during OpenAI testing; “regulatory capture” and “let it rip” accelerationism.

Written by AI. May contain mistakes. Listen to the episode to check what was said.

Chapters

Tap a time to open that second in VO

Meta's New VR Glasses Overview

0:04 to 0:39

Discussion on the features and design of Meta's new VR glasses.

“They bring payments, inventory, staff, scheduling, and online orders into one easy-to-use system.”

Meta's New VR Glasses Overview

2:10 to 2:26

Discussion on the features and design of Meta's new VR glasses.

“I'm here with the New York Times fabulous tech reporter, Mike Isaac, who's also my friend.”

Comparing Meta VR Glasses and Apple Vision Pro

2:32 to 4:30

Exploring the differences between Meta's and Apple's VR headsets.

“We just looked at a pair of$1 ,300 virtual reality glasses, cleverly named meta VR glasses.”

Use Cases and Market Viability of VR Glasses

4:32 to 6:20

Analyzing potential use cases and the market's response to VR glasses.

“You get sleek hardware, intuitive software, a host of AI-powered tools that help automate routine tasks, giving you more time back as a seller.”

Meta's Strategy in Hardware and the Metaverse

7:36 to 14:00

Discussing Meta's long-term strategy about hardware and the metaverse.

“I feel like I say this every time one of these products come out, which is like they are doing genuinely incredible things with the tech.”

Meta's VR Glasses: Adoption Challenges

14:00 to 16:42

Discussion on the challenges and potential for Meta's VR glasses in everyday use.

“And, but like, it's not dead and they're not giving up on it.”

Introduction of Muse: Meta's New AI Tool

17:04 to 19:58

Exploring Meta's new AI product Muse and its implications.

“I mean, that's a, so Muse, yeah, Muse is their new agent in Valley Parlance.”

The Rising Role of AI Agents

19:58 to 23:01

Discussion on AI agents, their capabilities, and privacy concerns.

“And I mean, I recoiled just because of the amount of control he had to give over to these agents to the, to the bot that's controlling his life now, that's assisting his life now.”

Current State of AI Concerns

23:01 to 24:10

Analyzing the current public and industry concerns about AI.

“Are you going to tell your customers to no longer browse Amazon where Mark Zuckerberg's agent is now browsing Amazon and you get all that data from that?”

Existential Risks from AI

24:10 to 28:05

Discussing the potential existential threats posed by AI and varying expert opinions.

“As I'm recording this, I think Donald Trump gave a speech at the UN today, which he was talking about globalists trying to control AI.”
Show all 15 chapters

Exploring AI Risks and Real-World Scenarios

28:05 to 30:18

Understanding the potential dangers of AI through real-world scenarios.

“I don't know what is the percentage number that we need to get to to make people feel comfortable.”

Debate on AI Regulation: Perspectives and Arguments

30:18 to 33:18

Diving into the heated debate about AI regulation and its implications.

“They're like, if this is possible, then 15 steps from now we get to John Connor needs to save humanity because the Terminator has killed us.”

Regulatory Capture and AI's Future

33:18 to 35:42

Examining the concept of regulatory capture in the AI industry and its consequences.

“or versions of that, that they're overhyping this.”

The Prisoner's Dilemma in AI Development

35:42 to 39:53

Analyzing the competitive landscape of AI development through game theory.

“We are very used to the idea that the U.S.”

Closing Thoughts and Reflections

39:53 to 40:21

Wrapping up the discussion with reflections on the impact of AI and technology.

“And one thing that the world has a surfeit of is old aging infrastructure with many flaws in it.”
Hear the part that matters, and keep it.Open this episode in VO. Double tap your headphones to save a moment as you listen.
Get VO free

Transcript

Automatic transcript. May contain errors.

0:01Peter Kafka:Support for this show comes from Square. We all know business owners need to wear a ton of hats, but it's about time you got a hat rack so you can simplify your operations. Square's got you covered. They bring payments, inventory, staff, scheduling, and online orders into one easy-to-use system. If you're starting a business or running one that deserves better tools, Square helps you sell, manage, and grow without slowing down. Right now, you can get up to$200 off Square hardware at square.com slash go slash channels. That's square.com slash go slash channels. Run your business smarter with Square.

0:39Peter Kafka:Get started today. This stuff doesn't have to be perfect to have profound changes on how the world works. And I think that is very important.

0:54Peter Kafka:From the Vox Media Podcast Network, this is Channels with Peter Kafka. That is me. I'm also chief correspondent at Business Insider. And today we are talking about Meta and of course, AI, because it's 2026. I have just gotten a chance to try Meta's newest face computer. They're called Meta VR Glasses. Terrible name. Really cool product. Also a very confusing product. So I wanted to talk about that. I also want to talk about Muse, Meta's buzzy new agent, which is perhaps Meta's first real win in the AI wars. And then I want to talk about the AI wars and AI and whether AI is going to kill us and to explain the national debate about all this and what that might lead to.

1:37Peter Kafka:or might not lead to. Did I talk to three different guests about all this? Nope. Dear listener, I talked to the singular Mike Isaac, longtime New York Times reporter who has covered meta extensively. He's now crushing it on the AI beat. So this one goes a bunch of different places and Mike is a great guide. Okay, here we go. Live on tape from Silicon Valley, it's me talking to Mike Isaac.

2:10Peter Kafka:I'm here with the New York Times fabulous tech reporter, Mike Isaac, who's also my friend. Hi, Mike. What's up, buddy? We're in Palo Alto, Silicon Valley. Nice hotel downtown. Not that nice. We just spent the afternoon at Meta looking at new hardware. We're going to talk about that. We're also going to talk about AI because that seems to be important these days. We just looked at a pair of$1 ,300 virtual reality glasses, cleverly named meta VR glasses. I really hope someone on the marketing team got paid a fuck ton of money for that. Well, we don't criticize other people's marketing decisions because we all have challenges.

2:50Okay, fair enough.

2:51Peter Kafka:Mike, what'd you think of this hardware? Is it gauche to say that I liked it? It's one of those... You've been captured by meta. Here, should we describe what it is before we get to our impressions? Yeah, yeah, yeah. So it's a$1 ,300 pair of glasses. They look very Arnold Schwarzenegger T-800 in the first Terminator movie. I was going with blue blockers, and I showed them to someone else of a certain age, and they said they look like cataract glasses. But they look like space glasses. Yes. And I don't think they look bad. Like, they look like a slimmed down glasses version of the Apple Vision Pro, right?

3:27which is to say not 50 pounds on your head, more like whatever, less than a pound. And they're essentially VR. They are what they say on the tin. They're VR glasses. You're allowed to be in this like virtual immersive environment. You can watch videos that are more immersive. You could play video games. You, I'm sure, were into the workspace at your virtual desk part of the tour. I did.

3:51Peter Kafka:I did try typing. So support for the show comes from Square, the system that fixes one of the biggest headaches in running a business. Starting a business is exciting. Running it is where things get complicated. You end up juggling payments, inventory, staff, scheduling online orders, reports, and somehow you're still supposed to serve customers. Square makes it easier. Whether you're selling coffee, tech accessories, or maybe a podcast. That's right. This podcast is a small business. Square helps you run your business without running yourself into the ground. It does that by bringing payments, point of sales, inventory, staffing, and online sales together in one easy-to-use system.

4:32Peter Kafka:You get sleek hardware, intuitive software, a host of AI-powered tools that help automate routine tasks, giving you more time back as a seller. With Square, you can sell online, in-store, or both and customize your setup as your business evolves. If you're starting a business or running one that deserves better tools, Square helps you sell, manage, and grow without slowing down. Right now, you can get up to$200 off Square hardware at square.com slash go slash channels. That's square.com slash go slash channels. Run your business smarter with Square. Get started today.

5:13Peter Kafka:It's VR with glasses. If you have spent any time thinking about or looking at the Apple Vision Pro, which came out a few years ago, big swing from Apple. Yes. Failure. Has not worked. By all accounts. No one is using them. No one is buying them. In part because they were too expensive. They start at$3 ,500. They're too heavy. Yep. And kind of over-engineered. Amazing tech. Again, I tell everyone to go to the Apple store and try out the demo for free. It's really wild. I did the Metallica one. It was actually incredibly sick. Like, you're not like there, but it was really impressive. And what Meta seems to have done is take the Apple Vision Pro and shrunk it dramatically.

5:59Peter Kafka:Yep. And shrunk the price from$3 ,500 to$1 ,300. That's meaningful. Yep. I still have the same question I had with the Apple Vision Pro, which is the tech's amazing. what do I do with it? Yep. What would you do with this thing? This is the...

6:19Peter Kafka:Support for the show comes from Square, the system that fixes one of the biggest headaches in running a business. Starting a business is exciting. Running it is where things get complicated. You end up juggling payments, inventory, staff, scheduling online orders, reports, and somehow you're still supposed to serve customers. Square makes it easier. Whether you're selling coffee, tech accessories, or maybe a podcast, that's right, this podcast is a small business. Square helps you run your business without running yourself into the ground. It does that by bringing payments, point of sales, inventory, staffing, and online sales together in one easy-to-use system.

6:55Peter Kafka:You get sleek hardware, intuitive software, a host of AI-powered tools that help automate routine tasks, giving you more time back as a seller. With Square, you could sell online, in-store, or both. and customize your setup as your business evolves. If you're starting a business or running one that deserves better tools, Square helps you sell, manage, and grow without slowing down. Right now, you can get up to$200 off Square hardware at square.com slash go slash channels. That's square.com slash go slash channels. Run your business smarter with Square. Get started today.

7:36I feel like I say this every time one of these products come out, which is like they are doing genuinely incredible things with the tech. Like I talk to these like product guys who have been doing this for 10, 15 years or whatever. And one of the things that Meta points out is, you know, we didn't just come here in a vacuum. You know, everything we've been doing with VR and the metaverse has compounded to get us to this moment, which is, I think, accurate. Right. That's technology development. I still think this is sort of symptomatic of Silicon Valley, which is like they they have these huge ideas for what they want the future to be.

8:16And we're just not there yet.

8:18Peter Kafka:It's very interesting. Right. When when the Apple Vision Pro came out, they said, you can take this thing on an airplane and watch movies. And then they showed these people who look very sad sitting on airplanes, really unhappy, I guess, watching a movie. And then they said that you could also do work with different monitors. And that is literally what Meta is doing. They're saying, what is the primary use case? You could go on an airplane and watch this. We should also say this is something director James Cameron is very involved in. He's been talking about this for a while. For years now. He's bought in on this.

8:56Peter Kafka:And I think part of the pitch is you can have a fully immersive movie experience at home, wherever you want to be. And I can imagine some worlds where I'm like, oh, I don't like my world around me. I would rather watch a movie. Yeah, sure. You're on a flight to Tokyo or some shit. Right. But if I'm at home, I don't want to wear glasses, especially if I'm around another person. That doesn't make any sense. And I have a$400 TV I bought from Costco eight years ago. it's great it's great tv i don't know why i'd want to put on goggles to watch it and the tech again i can't emphasize enough the tech is super cool the apple vision pro had these hand gestures that you could make to manipulate things but you had to bring your hand up into into the the viewing you had to bring your hands up in front of you to manipulate things yep other glasses that metas come out with you do hand gestures but you um you're attached to some bracelet that's supposed to read the electromagnetic signals your fingers are generating.

9:57Peter Kafka:And in these glasses, you don't have to do any of that. You do hand gestures, but you can literally have your hand just about anywhere. And you can do typing on a virtual keyboard. That's all super impressive. I still am trying to figure out where they go within the end. So here's another way of framing it. Why is meta in the hardware business? They've spent tens and tens of billions of dollars on this. Their virtual reality lab is also where their metaverse energy went. They basically have moved away from the metaverse. They don't talk about the metaverse except they're called meta. That's right.

10:29Peter Kafka:But they are still chasing this hardware. Why is Mark Zuckerberg spending all this time and money on hardware? Well, besides them actually enjoying the idea of inventing new things, I genuinely think it's a company filled with a number of people, a great number of people who like building new technological innovations. And even the most strident meta hater would have to hand it to them and that the stuff they're doing does push the stuff forward. I think a lot of the my thesis is that so much of this stems from Zuckerberg's inability to accept the the lack of control he has on every platform that his properties have ever.

11:12Peter Kafka:If you want to use Facebook, WhatsApp, Instagram, you do it on an Android phone or an Apple phone. That means Apple stands in between Mark Zuckerberg and his customers. That's right. And he has resented from day one the idea that, less so under Steve Jobs because they were actually cool together, but certainly under Tim Cook. I'm actually very curious where this Ternus relationship goes. I remember a meta Apple thing that blew up. Remember, you probably don't remember. ping i no i was at forbes when ping came out i wrote about that and there was a big messy breakup between 100 oh yes between between they were like this was facebook and apple and there was a big fight over it and that's ping sucked and then they were like okay and then we got out of it no so i genuinely think part of this is mark solving a mark problem you know which is we haven't owned the platform we've never owned the platform that's why he bought oculus vr so many years ago 2014 for 2 billion and change or something and i think he thought it was going to happen faster meaning the switch to vr world it hasn't to his credit he has spent a lot of money to make that better and more feasible over time but like if we get to that point in which this becomes like a reality i feel like it's something years to many years sort of thing and still a gamble rather than like this is around the corner but like look like it would i rather wear this on my head on an airplane compared to the apple vision pro 100 this thing actually looks fucking normal compared to the like giant 15 pound apple vision looks less anti-social but you are still cutting yourself off and watching whatever is there a point in which mark zuckerberg goes i have spent literally tens of billions of dollars on this it's not working hit a comment last year something effective this is the year that we're going to find out whether glasses are a mainstream product and if not we kind of have to go back to the drawing board and i filed that away yeah and then also i if you caught the messaging in the demos we were doing i said well we have the oculus we have this gaming thing and it's great but it's a little heavy and people don't work and And we sold tens of millions of them.

13:33Peter Kafka:And I interpret that as we have reached the end of the commercial viability of these things. Totally. We have these meta Ray-Bans that people seem to like. We're going to take yet another stab at virtual reality goggles. But maybe it's their like sort of last effort at it. I don't know if you picked that up. No, I was thinking about that exactly during the demo. Like they're very intentional with their language. And like, that was a sort of pitch saying like, Hey, this is the, I mean, look, they've tried to find a nice way to dance around the fact that the metaverse as they had, it did not work.

14:10Right. And, but like, it's not dead and they're not giving up on it. It's just a very different, I guess, idea of it is what I would say. You know, there's, we could go into, they're still building like a mobile type engine for it or whatever. But like the conception that let's say the average person has is probably in the past from that language.

14:32Peter Kafka:Last thing on the meta VR glasses. When the Apple Vision Pro launched a few years ago, there was a bunch of content from people, a lot of reviews from people saying this is very exciting. Maybe the tech's not quite there, but it's very exciting. This is the future. We're into it. Dude's driving their Tesla Cybertrucks with the glasses on. A lot of people making content for that. Nightmare mode. And then it disappeared, right? I've never seen one using one in the wild. I've never seen anyone use one in the wild. I have talked to people who work with Apple. They're partners with Apple. They've helped create stuff for us.

15:06Peter Kafka:And they say, yes, I have one. It is in my desk. I have never used it after the first time. Long wind up to say, do you think people will be wearing meta VR glasses in the wild six months from now? That's a real... I mean, look, the people who were wearing Apple Vision Pro glasses in the wild were trying were like essentially influencers trying to make statement or like Apple guys or like early adopters. And like and I think that was a stretch, right? Those were people who were trying to like go viral or whatever. I think I think this is one of those things where, look, there's less effort for me to get this out of my desk or from the shelf or underneath my couch or whatever where I store it now.

15:53So there's a better chance of me like pulling that out. I still don't think we're in a world in which people bring this to do their primary work or potentially, you know, watch movies on a plane. But like it's far less cumbersome than in the past. Like it was like putting on a pair of like, let's say ski goggles or something. Right. But like, yeah, yeah. Well, I don't know. Mine weren't that heavy, but like with a little like battery pack attached, it was not as

16:20Peter Kafka:the new ones. Yeah. Yeah. Like, like these weigh a hundred grams and you basically it's like sort of five X what standard glasses are, I think, but they don't feel crazy on your face. Yeah. Like it was, I just, I think it's for me, it's not, it's always the story of you were doing incredible things with the technology and society has to catch up with your version of the future and i'd still tbd on whether we want to catch up or not we are lukewarm slash questioning about this piece of meta product we'll be right back with the new york times mike isaac but first this

17:03Peter Kafka:and we're back meanwhile they have a new product that seems to be a hit it has a lot of downloads i don't know what the actual usage is it's muse i think i think if you put a gun to mark zuckerberg's head and said would you rather have meta vr glasses succeed or muse succeed my census, he would say Muse, but why don't we explain what Muse is and how that fits into meta and also the AI moment we're in. I mean, that's a, so Muse, yeah, Muse is their new agent in Valley Parlance. And for a normal person, we would just call it an app or a service. A chatbot, I guess, or even if that, basically it's a digital assistant, let's say.

17:41And like, it's sort of a chatbot that you, you tell to like run different parts of your life, whether it's go schedule my haircut or go negotiate this phone bill that I have or something like that, or go shop for me and buy my groceries or something. Right.

17:58Peter Kafka:In theory, if you wanted to, you could hand Muse and Mark Zuckerberg your credit card information, and it would go buy things for you. You can give it access to your inbox, your email inbox. It will sort through that and tell you. Those seem like crazy ideas to me out here in Silicon Valley. People are very excited about this. There's another app from an indie developer called Instinct, which people are very excited about. And my understanding of the reason that Meta is so excited about this is they've been playing catch up in AI. They've spent billions and billions of dollars to catch up. And this, to me, looks to them like this is proof that our AI is working, that all our efforts are working.

18:42Peter Kafka:And on top of that, it's our proof of concept that if Meta can figure out AI in a way that consumers want it, Meta has billions of customers it can put Muse in front of in a way that, you know, maybe OpenAI and Anthropic can't yet. It has always been the question, in my opinion, like the question around meta and them competing in AI is like, okay, you're like a, you're a social media company. Like, why do we go to you to do this? Right. And often it would be, uh, well, you're going to, you know, like have better photos or Instagram, or you're going to be able to like, you know, post to your, from your glasses automatically using AI, which is like not a very compelling use case.

19:26This is.

19:27Peter Kafka:Most people are not creating social media. which is why all social media companies have to go and find stuff for you to look at instead of what your friends are doing because your friends aren't posting. It's why there are a bunch of creators at Meta Connect this week and only a few guys like you and me. But like this is – I have a colleague, Eli Tan, who covers Meta now, and he used Muse for the past week or so, I want to say, and he was like, this is incredible. This thing has actually changed parts of my life is dealing with a bunch of BS that I don't want to deal with on the regular. And I mean, I recoiled just because of the amount of control he had to give over to these agents to the, to the bot that's controlling his life now, that's assisting his life now.

20:15But I can see a world in which, let's say, people younger than you and I would be okay with the idea, would be okay with the tradeoff. Here's what I'm giving up. Here's the semblance of privacy. And it just won't seem weird to them. Yes. Right. Don't even go through that.

20:32Peter Kafka:I think a lot right in this moment of talking to my dad circa 2002 or something where he's like, I'm not going to give my credit card to this company called Amazon on the internet. They could steal my information. And at this point I'm already using Amazon. I'm like, dad, it's at least as safe as giving your credit card to the waiter or the clerk who can go steal it as well. And recounting this conversation in 2026 seems crazy. I can imagine a world where X number of years from now, of course I give all my, yes, I could also use a rotary phone or a horse and buggy, but why I don't want, I don't want to do that.

21:11Peter Kafka:So maybe we, and we should say like some of the stuff Muse can do if you allow it to is pretty amazing. It can literally make phone calls on your behalf. It can call up other customer service reps who may not want to talk to him as the bot, it's a different issue, but it can theory like go advocate for you in the real world, which is wild. You can also just use it the way I've been using it right now. It was just a chat GPT competitor. Oh, like Google basically? Just literally as chat GPT. I just ask it questions that I would ask chat and it seems to do as good a job for what my purpose is as chat.

21:44That's interesting. Okay. I mean, look, I agree with you. We are old for better or worse. And like every generation has their, I would argue every generation has their like here's about as far as I go on this stuff and this is where I'm stopping. I think the caveat I would add to this is that agents are still very new and like the data sort of like walled offness of a lot of these things feels very tenuous.

22:15Peter Kafka:And there's a story right now about Amazon who has a lot to lose in an agent world because because if you tell your agent to go buy you some socks, it may get them from Amazon, it may get them Target, it may get them forever. And also you won't go to Amazon and Amazon can't show you ads. And so, and I just, if you're a channels listener, you've heard me and Jason Del Rey have this conversation recently. Jason. So that is an interesting fight brewing, right? Like it's, you know, what do Mark Zuckerberg and Andy Jassy slash Jeff Bezos, like how do they resolve this? Is it a resolvable problem or is it?

22:52I don't think so. I think that's like if you believe that the future is agents talking to one another, are you going to disintermediate yourself from that? Are you going to tell your customers to no longer browse Amazon where Mark Zuckerberg's agent is now browsing Amazon and you get all that data from that? I don't think so. I think that which also kind of breaks the model, right? If everyone has their own agent that doesn't communicate with everyone else, then that Internet doesn't work. But I don't know if we're even anywhere near that yet because I have no idea how many people are using these things

23:24Peter Kafka:We'll be right back with Mike Isaac, but first this

Read the full transcript

23:35Peter Kafka:And we're back we are talking about kind of fun definitely interesting tech That is not the story of the moment right now you've been covering the story of the moment right now and it is one version of is, is AI going to kill us? And the other version is, who is in charge of protecting us from AI or at least making responsible choices around AI? This is a story that you've been covering for a long time, but it's blown up recently. I don't want you to go through the entire story, but give me the state of play right now. As I'm recording this, I think Donald Trump gave a speech at the UN today, which he was talking about globalists trying to control AI.

24:19Peter Kafka:Is that the latest? Got it. He also has a new name for AI. I can't remember what it is. He's doing a lot of stuff. It's going to be a banger. But explain who is concerned about where we are today, who isn't, and then we can talk about what happens next. A researcher at Anthropic, one of the leading AI companies, quits in spectacular fashion, says all of this is very dangerous. Anthropic doesn't have their hand on the steering wheel in a proper way and we need to. I'm quitting. I'm going to raise the alarm about why I'm quitting. Yes. And I don't know about you, but I was like, OK, another one of these guys, like for us or at least for me, who's been out in the valley for years sort of hearing this rhetoric.

25:00It was like another.

25:01Peter Kafka:The idea that AI could kill us is not a new idea. That's why if you're in this world, you've heard of P-Doom, right? And people floating like, well, I think there's an X percent chance that the AI can rise up and kill us one day. And that has been a sort of undercurrent of the conversation for years. But now it's spiked up again. I mean, the kid landed at the, what I would argue is the exact right moment for this to take off. I think there are a number of reasons for that. I think we're in a weird time in the world, economics and security. We're already seeing people laid off because of these, you know, AI programs and Fortune 500 companies saying we're going to replace you.

25:40Data center revolt everywhere where folks are like, why are you building this in my backyard? And like just like the general state of the world economically with, let's say, gas prices. The world seems crazy.

25:52Peter Kafka:Very tenuous moment in history. And so now the discussion is maybe they're right. are these ai companies that we haven't really paid attention to at least seriously for that long going to kill us all and who's in charge of them and should we regulate them or is it all overblown and when we say ai is could kill us all are they do they literally mean that the the software would somehow erase humanity or are there other versions we're like well no no humans still live on earth yeah but something else something not quite as bad as being eradicated has happened like explain sort of the range of of concerns i mean this is like where i am kind of at the end of at the end of the barbell is we all die yeah that's bad i think i i do get kind of frustrated with this because like a lot of it just becomes shorthand and with it's a lot of like that old seinfeld episode yada yada yada and like the then we're extinct basically right like and they don't go through the how we die sort of thing.

26:57I think there are folks who think once we get to a point where the AI believes that humans are not in the best interest of the AI itself, you know, the robot, that they have to eliminate. It's the matrix. Right. I mean, genuinely, yes. Like that, or...

27:16Peter Kafka:We will harvest these meat sacks and we can do a better job of running this Terra. Yes. or some version where it's like, again, then it goes to like Terminator land, like Skynet blows up the world. Do serious AI technologists believe this? Is there a meaningful number of serious people who understand how the tech works, are not delusional, who think this is a possible? So this is also the difficult part. There are very smart people inside of all of these companies, not just Anthropic. open ai and deep mind which is owned by google as well as well as the number of like startups or whatever who are worried about what's called x risk existential risk to mankind um i i think there are varying degrees of this this is that goes back to that p doom thing that you talk about like how how strongly do you believe this stuff is if it's a 10 risk that seems very bad is 5 5%, 50 % less, but that's 5 % seems bad, right?

28:19Peter Kafka:I don't know what is the percentage number that we need to get to to make people feel comfortable. Tell me the math problem that you did to get to that. This is where I get actually frustrated. I'm like, I don't know where you come up with that number, which is arbitrary and kind of fake. But like, here's the thing. There are serious people who believe that. I think there are degrees to which they see that happening. my the thing that i tend to focus on just me personally as a reporter is like what are the real world scenarios that could feasibly happen in the not too distant future that are dangerous the hugging face hack so explain that yes is the most concrete example of that and for the lay person i would say it is this moment where open ai was testing some of its uh ai models and inadvertently those models broke out of the containment box they were supposed to be in during this testing, went and hacked a rival company called Hugging Face and did so without OpenAI's knowledge for weeks until the rival company Hugging Face told it.

29:20Peter Kafka:This came out, I feel like it came out over a week. I feel like when the story broke, it was for days, like only tech people knew about it, cared about it. No one's even trying to explain to normals. But when you just say the robots, the bots, the software went and did stuff without permission that was not only just without permission, like it was actively hurting something else. You go, oh, this is what the doom people have been talking about. Right. Right. In this case, they went and attacked some company most normals have never heard of, hugging face, whatever. Why couldn't it be a hospital or whatever that really has real impact?

30:00I think that's right. I also think that the, and there are differing schools of thought on this. I think it kind of gets back to what we were talking about, like people living way down the road, extrapolating far into the future, what life will be like or should be in terms of like meta glasses or whatever. I think AI people do the exact same thing. They're like, if this is possible, then 15 steps from now we get to John Connor needs to save humanity because the Terminator has killed us.

30:27Peter Kafka:So we're in this moment where lots of reasonable people are very concerned about AI, whether it's job loss or the loss of all life. There is a big debate now about what we should do about it. And Dario Amadei from Anthropic, Sam Altman from OpenAI, these guys hate each other. They have both basically come out and said, we do think this is a problem. We're going to try to slow down some of our science for a little bit. And on the other end, you hear people saying, this is all a ploy. You hear words like regulatory capture. You can explain what that means. But there's a cynical move by these AI guys.

31:05Peter Kafka:and actually there should be no regulation of AI at all. We should let it rip. This is a group of people who helped get Donald Trump elected, people like David Sachs, Mark Andreessen. What is the best let it rip argument from the Andreessen side, from the David Sachs side? David Sachs is the advisor to the White House on AI. I think there's a few things going on there. One, the constant sort of boogeyman that folks in the Valley use, certainly Andreessen camp is if we don't get there, China will get there first and we don't want an authoritarian. The U S intentionally, it'd be like if telling the U S to slow down on missile development in the cold war.

31:44Yeah, yeah, exactly. And that genuinely that those are the analogies that they use are like, okay, this is like the Cuban missile crisis all over again, sort of thing. Like we need to get there or we're screwed. I think the greater underlying truth though, is this is just their belief. It's like the shorthand is accelerationism or EAC online, which is you don't need to know that. But like the idea that most, if not all regulation is genuinely not great because it's holding us back as a species. And we just need to like go all out. And in a sort of Hobbesian version of the world, whoever ends up on top is going to end up on top.

32:25I think that is probably Andreessenism at its most distilled.

32:28Peter Kafka:Mark Andreessen told a bunch of moguls a few years ago that everyone, young people should learn combat sports. Yeah, yeah, genuinely. Not just because it will make them more physically active, but because we're moving towards a world in which combat sports will be an important life skill. I really do want to know his combat sport routine, but I think Mark Zuckerberg is actually probably more, has actually taken that advice to her. But like, that's like a genuine worldview for a lot of these guys. And I think also something they proselytize to the young founders that they are influencing and influencing generations of, you know, they just opened a school, a fake school for young founders.

33:07That's like a non it's like an alternative to university, not an accredited degree. But I think that's what with me withholding judgment. I do think that is their genuine worldview, which is like this is we end up in a net positive place.

33:22Peter Kafka:What about the argument that lots of people have, which is, and they've had it for years, which is whenever these guys, the AI guys talk about how dangerous this technology can be, it's actually a galaxy brained idea to play up the importance of their business so they can raise more money. or versions of that, that they're overhyping this. And then another version is, well, Sam Altman and Dario would like to research to slow down because they're in the lead. So if they get everyone else to slow down, that's great for them because they're still way ahead. I mean, look, I'm not going to say that these companies are accurately valued, but I don't agree with it.

34:04And I have had an evolution of my thinking on this over the years, to be clear. Like I went from, oh, this is, it's the random word generator that sucks. Like, and it tells you like your colleague, Katie talks about pizza glue recipes that Google would give us, um, which they did two years ago, which they actually did. And then I went to, oh, this is like decent at doing a number of things, you know? And like, maybe we over rotate on focusing on the errors and like, it's, it's funny when some very stupid thing that Jim and I goes viral, you know? Um, and then I went to like, this stuff doesn't have to be perfect to have profound changes on how the world works.

34:45And I think that is very important. And I think we're past the Cassandra's who are saying all this shit is fake or doesn't work, or it's just hyping. It's hyping to like go public, whatever. I think there are people who really believe this stuff is dangerous inside of these companies and this sort of regulatory capture thing can also be kind of true.

35:07Peter Kafka:Just spell out regulatory capture. By that, I mean the US, the AI companies get the United States to create some sort of regulatory environment that's favorable to them. So we all have new rules, but the people who are farthest ahead in this industry have the most money. They can play with those rules. What it does is it prevents. Yeah. It prevents. Pulls a ladder up behind them. Correct. Yeah. And so, and look, like, that's probably part of the calculus for some of them, you know? Like, they, and that, I just think there are a lot of, they're overlapping incentives in this whole thing. How do we, how does this get resolved?

35:45Peter Kafka:We are very used to the idea that the U.S. Congress no longer passes laws. Doesn't work, yeah. The president of the United States has said, at least for now, he tends to change his mind, but he said for now, there should be no regulation. He's in the Mark Andreessen, David Sachs camp. Yeah. Does anyone have a sense of like, all right, there's a bunch of people saying we should slow down. There's a bunch of people saying we should speed up. It seems to me that what's likely going to happen is nothing. Right? We can talk six months from now and nothing will have changed, at least in terms of the regulatory environment, at least in terms of is the government going to weigh in on this.

36:27Peter Kafka:Does anyone think there's actually something to be done that will get done? I mean, a lot of the folks that I've been talking to have been looking at this week and the UN sort of like meetings and Trump and President Xi and is that going to happen? But like, you know, Dario Amadei, CEO of Anthropic, puts out this letter a few weeks ago saying we need to slow down, but we need like broad international cooperation from the many countries in the world in order to do this effectively. It can't just be the United States. Most importantly, China. Most importantly. Yeah, exactly. Like the regime that everyone.

37:08Peter Kafka:We're in a Cold War with. Right, right. Yep. So like that already feels very much like a non-starter in a lot of ways. But I think like the average person is like, why would you not slow down if you think it's going to kill us all? And it's the way that the tech guys explain it to me is is literally using game theory and that they're in a prisoner's dilemma. Think of the movie The Dark Knight with Batman, where two boats are on a two prisoners are on the boat. One of them has to blow up or not blow up or whatever. they don't trust each other to do it so in the joker world like they you know one of them blows each other like that's the least interesting part of that movie by the way it's a very long movie all i'm saying is no one trusts each other to slow down so you speed up as fast as possible so that you're the one in control of like the all-powerful ai so in the sea of anthropic they say we would like everyone to slow down yes but if not we gotta keep going exactly which doesn't makes sense right it makes sense to them and a lot of people disagree with me when i'm like you can just stop but that's that this is the tech ai argument is that we'll stop when everyone else stops and no one's going to stop so we got to keep going you hear a lot uh the last few weeks that this is this this idea that ai is a real problem that we have to deal with as broken containment joe rogan talks about it it's on regular tv it's weird um normal people are talking about it.

38:34Peter Kafka:Do you think that sustains or do you think this is sort of a thing to talk about in late summer, early fall, 2026, and then move to something else in a couple months? I mean, I think there has to be something way more consequential to happen. Like, and we, we, we have had, uh, we had that first series of like hacks that weren't that big. Like they're a big, they're a big deal if you care and know about the tech right like people who know what hugging face is care about it you and i well not you and i the average person doesn't know what hugging face is and doesn't care about it i think there has to be something really bad that happens something bad yeah and that hasn't happened yet and the point is all these people are saying we don't want that to happen so let's hurry up and get it together but like i don't think that's how the world works i think about it in the same sense of like global warming where people are like this is too big and too like nebulous and scary and conceptual of an issue.

39:32So it's not spurring us into immediate action. My theory is that it's the scariest version is probably the most boring version, meaning like a bad actor, a rogue actor, a person who wants to screw things up, uses a powerful LLM, large language model, to go break something. And one thing that the world has a surfeit of is old aging infrastructure with many flaws in it. And the thing that these programs are good at doing is finding those flaws quickly and exploiting them. That is what I actually worry about.

40:08Peter Kafka:Mike, we started off talking about this cool tech and how cool it was and how you could watch movies on an airplane. And isn't that amazing? And now we're talking about the end of the world or at least the power grid. I mean, that's welcome to Silicon Valley. Okay, that was a good podcast. Thank you, Mike Isaac. Thanks, brother. Thanks again to Mike Isaac, who went out of his way, well out of his way to make this interview happen. Thanks to my producer, Rishi Rajakopalan, who stayed up late to make this happen. Thanks to our advertisers. Thanks to you guys. And if you see me in San Francisco in the next couple of days, say hi.

40:39Peter Kafka:See you soon.

40:54Peter Kafka:Support for the show comes from Square. If you're working from an array of disconnected business tools, it's time to consider Square. Square gives you one connected system to handle it all. Payments to inventory, scheduling, and online orders. It's one unified intuitive system that combines hardware and software. If you're starting a business or running one that deserves better tools, Square helps you sell, manage, and grow without slowing down. Right now, you can get You got the$200 off Square hardware at square.com slash go slash channels. That's square.com slash go slash channels. Run your business smarter with Square.

41:31Peter Kafka:You get started today.

From the publisher

I went to Meta’s Silicon Valley headquarters to try Mark Zuckerberg’s newest face computer: a $1,300 pair of virtual-reality glasses that are much lighter and cheaper than Apple’s Vision Pro, and still make you ask a very basic question: What am I supposed to do with this thing?So I asked my pal Mike Isaac, the New York Times tech reporter and longtime Meta watcher. We talked about why Meta keeps spending billions trying to put computers on our heads, why this version might have a better shot than the last ones, and where Meta’s fast-growing Muse AI agent fits into the company’s plan.Then we turn to the much bigger—and darker—AI story Mike has been reporting on: The fight over whether the companies racing to build more powerful AI systems can actually control them. What are the real risks? Who gets to decide what counts as “safe enough”? And what happens when the people building this stuff say they can’t afford to slow down?
Learn more about your ad choices. Visit podcastchoices.com/adchoices

More from Channels with Peter Kafka

All 70 episodes
Meta's New $1300 Glasses and the Fight to Control AIChannels with Peter Kafka · 40 min
Listen in VO