(Preview) Doom Debates Go Mainstream, AI Religion and the Economic Future, Several Vectors of the China Question

Welcome and podcast origin story

0:00Hello, and welcome to a free preview of Sharp Tech.

0:09Hello, and welcome back to another episode of Sharp Tech. I'm Andrew Sharp, and on the other line, Ben Thompson. Ben, how you doing? Just happy to be with you here at the end of the world, Andrew. One of our last pods before the sky falls on all of us. Here we go. Or the sky net, I should say, falls on all of us. The sky net falls. Oh, there you go. That's a good one. No, I thought, well, we still have time. Get some takes off, you know?

0:39Well, we always want to get some takes off, but we, I, conceived of Sharp Tech. You were this, you know, brilliant host, my favorite podcast, Greatest of All Talk. And the sort of normie newbie take to Greatest of All Talk was, you know, the person doesn't know anything about podcasting. They like Ben Goliver, right? He's very funny, very cutting, you know, sounds like he knows stuff, tends a lot of games. The podcast, Sophisticate, appreciates Andrew Sharp. Not only surprisingly good takes, which you wouldn't expect,

1:12but also the ability to set Ben up, sort of be the foil, be made fun of, all these sorts of things. Draw out his bizarre tendencies, sure. Absolutely. And I'm like, I want that. This is actually, I'm not, this is not a joke. I'm like, Andrew is amazing at hosting. And you know what? I like cooking. I like being on someone else's podcast. So my next podcast is not me hosting. It's me permanent guesting. So that is, I think people probably know that. That's sort of the broad origin story.

1:45Permanent guest here on Sharp Tech. Exactly. Yes. But before we go. Okay. I want to put the host jacket back on. Oh boy. For a day.

Some of all fears article overview

1:57So I did not write a big article this week. I wrote a response, sort of initial response to Dario Amadei's Pacing the Frontier piece. There is an aspect of all week. I feel like I've already written about all of this. I think if you've been listening to Sharp Tech, you were very early to almost all the arguments that have been going on this week. We've discussed almost all of them from some angle or another. Probably the most pertinent I've written is Anthropics Alignment, where they both believe it. And also everything's great for them.

2:28And it's just an amazing magic trick that they get to pull off. So we will get to that. I will be incapable of keeping my mouth shut for most of the podcast. But I'd say the official article for the Chucky Post Bundle this week is you, Some of All Fears. S-O-M-E. Great title. How'd you come up with it? I was sitting on a plane coming back from a weekend in Arizona, reading all the doomers on Twitter. And I knew that I was going to write about it. And so rather than write about it on the plane, I thought of potential titles on the plane and came up with Some of All Fears.

3:05There we go. It's great. I love it. Well, give me the overview. I think you really framed this as this is by someone who kind of knows about tech normie perspective. But I don't think that's it. Okay. What I think it is, is someone who pays attention to tech who does not live in San Francisco. That's my high level takeaway. Is that a way to frame it? Yeah. I mean, that's the thing. It's sort of a misnomer to call myself a normie at this point because I'm immersed in this world every week with the podcast.

3:37So I'm familiar with all the characters. I'm familiar with the terminology. All the preparation. Everything is Andrew. That's that is the host job is more difficult than it appears. Right. So I'm surveying the news and trying to keep track of everything. And I'll be honest. I was moved to write this article because. Which, by the way, is coming out the same time as the podcast. So the link's in the show notes. I would encourage you to go read it. I thought it was really good. Doesn't mean I stand by everything in it. Yeah, Andrew is his own voice.

4:08Thank you for that clarification. Which gets me in trouble, by the way, sometimes. Well, so I was moved to write this because I was away with my wife. It was our 10 year wedding anniversary. And so we're at like a resort in Arizona. Thank you so much. We had a delightful weekend. And Saturday afternoon, I got a text message from a friend who sent a link to a long New York Times op-ed. The headline of the op-ed was, this is really bad. And it was like 3,000 words of AI doomerism.

4:40And my friend, who's a smart guy, he works in the defense industrial complex in D.C. And is very accomplished. And is not a rube, per se. And he sent this to our group chat and was like, read this and worry. And I'm sitting there by the pool and I don't even have the energy to respond to all the different arguments in this New York Times article. It's like when a Normie wants to come up to you to talk basketball. It's like, I'm just like not even going to get into it.

5:10But I was so annoyed and I've been so annoyed by the tenor of this conversation over the last 10 days that I had to just sort of walk through why I find myself skeptical of this specific set of people and the claims that they've been making about AI literally for a decade and certainly over the last several years. No, longer than that. And it's actually a critical aspect of the season.

COVID analogies and natural immunity

5:41This is why I'm a bad host. This is why I'm a permanent guest. An important thing to understand is there's aspects about which this movement has been right about a lot of things. And they wouldn't be compelling and effective if that weren't the case. That's always the case, by the way. This goes back, you know, the COVID analogy is going to come up again and again for good reason. But one of the reasons why I was so disgusted by the crackdown, by the censorship crackdown and by the blatant statements of untruths or allowing untruths to propagate because they were quote unquote useful.

6:15Like, remember the whole thing about masks, like, oh, we needed them to be available for emergency personnel. We want people to use them. But then now we want people to use them. It's like whatever whatever the sort of veracity of that. Yeah. The problem with lying slash whatever it might be is you give credence to people who tell the truth. Now, those people who tell the truth might also be crazy, right? You can say there's this childlike view of the world that some people tell all truth and some people tell all lies.

6:51And not only is that absolutely not the case because no one can know everything. So definitionally, at some point in your life, you're going to say something that's not true. But also the reality is people have different motivations, incentives, views of the world, interpret things differently, will use things to their end. So if you create a situation where you're allowing it to propagate the idea that say you don't have any increased resistance to COVID after you've already had it, right?

7:22The idea that you need a vaccination post already having COVID is scientific nonsense. Now, you can understand societally why they want to say, no, everyone has to get the vaccine. And it's useful for people to believe that. Exactly. That was the methodology at the time. The problem is it's a lie. If you had COVID, you developed like natural – the idea like natural immunity was like coded for like a few years as this far right-wing wacko position. Sure. When it's like the most fundamental truth of like immunology.

7:56It is still coded that way. Like people talk about Aaron Rodgers. I felt it right there. I said natural immunity and immediately all my like antenna went up like, oh, I have to cover myself. I have to backtrack. It's nuts. It's absolutely nuts. Aaron Rodgers is still stigmatized to this day. And if you go back and listen to what he was saying, it was just objectively correct. And so all of this is still in the water that we're all swimming in. Well, I love Aaron Rodgers. I'm not going to – I don't remember everything that he said.

8:26That's actually – that's a very good qualification. All right. But I have enjoyed him cooking ESPN over the last few weeks. The last few weeks. So, yes. Not a broad endorsement. He was on with McAfee recently. The reality is, as a Packers fan, Aaron Rodgers cooking ESPN over COVID takes is all we have right now. So, whatever. I'm with you. Let's do it. Fair enough. Fair enough. But, sorry, before I let you go back to talking about your article, which is supposed to be the opening of this podcast, which you didn't know about. I sprung this on you, by the way. That's right. Literally 30 seconds before we came on.

8:57When you make natural immunity as being a kooky sort of thing, then people who are kooks who say natural immunity is real get credence. Sure. And they get credence. Then all their other views also get credence because they were right about this thing that was different than the mainstream. And then, well, maybe they're right about all these other things. It's like, no, actually, they're not if we had just been honest in the mainstream, then we would have had much more like resist, much more immunity to the kooks.

9:32So, anyhow, you go back to the EAs or not. EAs is a catch all charm. There's actually a bunch of different. There's the rationalists. There's EAs. There's a whole like Harry Potter fan. Like, I don't know if you know that Harry Potter does figure actually quite prominently into all this. I became aware of it against my will over the last 10 days. I actually think that would be a good punishment for whoever finishes last place in your fantasy football league is make them read all of the less wrong posts, this Harry Potter situation, some good podcasts about AI agent welfare and the philosophical.

10:12Harry Potter and the methods of rationality. It's like an extension of the Harry Potter universe. And it's like 600,000 words. So, yes, I am vaguely aware. Can't say I've read all 600,000 words. Yeah, I've read zero of them. And that will continue to be the case. Although, I'm a pretty big Harry Potter fan, to be honest. I was pretty late to it. Me too. Because I was older. But I enjoyed them. I think I remember twice. It's a millennial tentpole. So, I'm obligated to be a fan. I'm not a millennial. So, it took me a while to get to it. But as we've established.

AI doomerism and LLM limitations

10:42So, the thing to understand about the AI doomerism is it goes back a long time. Like a few decades. I think really in earnest, like two decades. And the reason this is pertinent is no one in this movement for Shaw, for the large language model. Yeah. Like that this is going to be the transformer. It's going to be the means by which AI comes along. And the reason why it's notable is the biggest failure mode of the large language model is the fact it's so obsequious.

11:18It just gives you what you want. Right? Like if you, it's, you know, it's funny. Like I... Love all the vibe going off. All this sort of things. And I like know enough to be dangerous. I know what sort of database, like I want to use for a particular sort of thing. Like I know particular things as far as security goes. Like I know the pieces, all the pieces of an app and how they work together. And the problem is at the end of the day, though, the deeper I get and make things more and more complex, the more I do feel the fact that I don't actually code. Okay.

11:48And feeling the edges of... Look, this was a foreseeable problem. I could have really used a coder that pushed back on me on this particular point. But you, stupid LLM, were like, yep, got it. Let's go do it. It's like, no, we should not have gotten to do that. Now we have to go back and we have to undo all this sort of stuff. And I have to use up another one of my precious open AI reuse tokens that... Reusage tokens that resets my usage. I used them all. I'm on the last one. Again, it's not to say that LLM-based AI isn't potentially dangerous.

12:21Again, the cybersecurity things, we've been... We've been talking about it for three months. Yeah. Yeah. But there's been a lot of retrofitting of, I have these pre-existing beliefs about AI existentialism. And I'm going to fit this technology onto that and have a deep-rooted, ironically enough, because our whole movement is about defeating confirmation biases, I'm going to fall so deep

12:51into confirmation bias hell that every single thing these things do that is problematic is going to fit onto my pre-existing doomer narrative. And as opposed to there's these issues that are coming along, let's extrapolate those issues and say, oh, that could end up very badly. And that's just a real fundamental distinction. And this call-out about a longstanding movement, it actually is important to say it's a 20 to

13:2230-year-old movement. It's not a five to 10-year-old movement because that is one of the fundamental critiques of what's going on here. And you mentioned the confirmation bias angle in this article. And I think it's really important context for people to have when they're evaluating these claims. And one of the things that I have been frustrated by over the last 10 days is just how much deference has been given to Dario and the leaders in the frontier AI community who all have these sort

13:56of pre-existing biases and blind spots, certainly geopolitical blind spots, because they have been leading on the technological side. And so you have people throughout politics who say, well, these are the experts and they're concerned about this. So now we all have to be concerned about this. And it's worth looking closer at what some of their other claims have been and how they arrived at some of these conclusions and how much overlap there is between the auditors

14:27and the people in the labs and everything. It all ties to this insular community. And again, it's just a reason for me to be a little skeptical. Which, to be clear, has been right about like 85% of things, right? So just in defense of the sort of doomers, they've been right about a lot. And again, I also believe they are right about there being significant issues and challenges, particularly in the cyber arena, about these sorts of things.

14:57And in the very fullness of time in hundreds of years when robots do everything, we can get in the dorm room and imagine like nanobots. Sure. Right. But even then, the one thing I do sort of regret is like, I do think there's a very clear line between digital and physical, but the digital is so enmeshed in the physical that like, again, I've mentioned this in the podcast last week. I can imagine there is AI capable of like shutting down the power grid. I wouldn't be surprised if that's the case. And you sort of like, I thought your framing of this was fair. Like these companies have air gap systems, all these sorts of things.

15:30And like, you go back to like Iran and like the Stuxnet sort of thing. Yeah. Like there is a will, there's a way. And one of the things about LLM, first of all, there's a will and they have a lot of will like China is already, according to various intelligence agencies, China has successfully hacked our infrastructure grid today, pre AI. So there's always going to be a will and there's always going to be a way that does not lead to extinction risk though. And that's where I draw the line.

16:00And this, it has always come back to the extinction risk. And like Dario has said, there's a 10 to 25% chance that this has a catastrophic impact on human civilization. And I just don't think that that is a productive way to talk about this. There's supposed to be 25% of jobs eliminated by now. Yeah, totally. Like, or sorry, I think it was end of 2026. So we still have three months to wipe out 25% of jobs. Um, so his, his phrase was soon and it has not happened soon.

16:31Um, I think he may have given himself until the end of the decade and look, that may happen. Um, and there are definitely areas where AI poses real concerning questions, but I think talking about these existential risks kind of confounds any reasonable conversations about how government may or should respond to any of those things. And right now I think it's so early that it's impossible to respond coherently to any of

17:02the downsides that people have been worried about. And then that's where the last 10 days, it's all gone to a pretty crazy place. Like we finished on the last podcast talking about P doom. And I said to you, after we finished recording, I don't know how much more there is to say about this topic. And unfortunately, after we finished, like the world just went crazy and it became the biggest story in the world over the last seven days. Well, I think that's actually the exact issue. And it's also why I do find it a little tiresome is by virtue of it being unfalsifiable, it

17:38can be used endlessly to whatever, yeah, whatever one wants. And that's how you see this emerging coalition between this P doom movements between, unfortunately, like I don't like to get partisan on our shows, but you know, this is funny where I get the I get the AI sort of like, you know, I do the run through and again, usually not particularly helpful. And I think they, the AIs always want me to soften stuff. Oh, they hate it when I say stuff there, because they're like, where's the evidence? Like, these are the, these are the, I mean, to your point, you talk about very smart people.

18:13I'm sure your friend's a nice guy. Hopefully he doesn't listen to the podcast. Sounds like a bit of a midwit. I mean, like there's, there's a real, there's a real, like this take it back to like sports analytics, right? Like at the end of the day, you can get a lot of traction by looking at the numbers and like sounding smart, all these sorts of things. No comment on grace of all talk. We don't bend us unless it's a podcast, so we can, we can, but there's, there's some, the beautiful thing about basketball is there's, there's something else. There's something that is not tangible.

18:45There's intangible. That's the word I'm looking for. That goes into actual winning and yeah, Derek Moray, to your point about biggest loser moments. Daryl, love Daryl. You called him Derek. Oh no, I said Daryl. I said Daryl. I think Daryl probably does listen to the podcast. Hi, Daryl. Sorry about that. We love Daryl. You can spread, send a spreadsheet about 81 missed calls in a series and you can say, what are the chances that we missed 27 straight threes? And I can sit here and just infuriate you by saying there's a reason that it happened.

19:15And I can't put my finger on what it is, right? But there's a reason those rockets never made up, but by the way, all things considered was they are underrated. We're a great team. They ran up against one of the greatest teams of all time. And this is why I was so devastated. Built around a fraud. That was the ultimate problem for Daryl, but it is what it is. But you always go back to Steve jobs and making fun of complaining about technology, being obsessed with feeds and speeds. And there are things that can't be measured.

19:47And there obviously are. And a consequence of increased transparency, of increased tracking, of increased data is when there was no data, just tracking a little bit gave you this big advantage. But there is this false sense of certainty that arises from more data. Just because you have more data doesn't mean you have all the data. And tech has always been susceptible to falling in love with data.

20:18Data, definitionally, is backwards looking. Like, the LLMs want me to cite something when I'm saying something that's happening right now or prediction of the future because that's how they work. They are consensus. Like, LLMs are literally the midwit at the center of the bell curve. Like, that's what they are. It is a probability distribution. They're going to pull from the center of the distribution. They're not going to pull from the edges. And why? If you pull from the edges, you might pull from the crazy side as much as you pull from the brilliant side. You can get things wrong and you're sort of engineered to be accurate.

20:50So, I said in that article, like, this is clearly, like, half of the political divide. And it didn't like that because it's like, you're going to make people upset. I'm like, but look, that's literally what's happening. And the reason it's happening is because this is unfalsifiable. It's something that can't be proven. The overall precept of actually government needs to do more. There needs to be this entire bureaucracy around controlling this. There needs to be people that have jobs to monitor, to do things.

21:21Sorry, it's just a reality. There is a one side of the political divide that leans towards more government, that leans towards more bureaucracy. And this is not a commentary about whether that's good or bad. I'm obviously fairly skeptical of that and concerned about its impact on long-term innovation and all those sorts of things. But we can't just – I have to point that out. It's just a reality. Not only that, it's going to be a reality that informs how policy is crafted here.

21:54And I do have some insight because I'm in Washington, D.C. and have a bunch of friends who are in politics. And I remember – I think it was like nine months ago – I met with somebody who is consulting with a potential 2028 presidential candidate. And she was asking me, like, what should our AI policy be? I don't know what we should be pushing. And I think the entire Democratic Party has been sort of casting about looking for what to say.

22:24These aren't first principle takes. It's a – there's a very big issue that happens to have one side of the divide that in broad strokes aligns with our view of the world. Like, that's what's happening. Like, I think it makes all kinds of sense politically, to be clear. And how – exactly. How do we stake out a position that will be popular, people will rally around? And I think that's what's happening here. And government has always been incentivized to assert control over AI. And that may happen at some point over the next couple of years.

22:57And I think all of it has accelerated over the last two weeks here.

Listener questions on AI benchmarks

23:01Should we get into some questions? Because we did have – I know. I know there's much to get into here. And I feel like I'm doing a bad job. So I need to take over. No, you're good. You're good. Phillip says, Andrew and Ben, it was hard not to chuckle this weekend thinking about Ben being happy in 2023 that he could write about AI instead of antitrust regulation, only to be condemned to now writing about AI regulation for the next 10 years. Scary Dario voice, if we even live that long.

23:33With all of the talk of scary AI, I thought it would be fun to lighten the mood and share our dumbest personal benchmarks for when AI has arrived. Here are mine. Phillip's observation cuts so deeply. I just tried to waste as much time of this podcast pretending to be the host so you didn't have to read this out loud because it hurts. It's true, Phil. Before digging in, yeah, you were like the guy at the end of Shawshank Redemption. And now you're going right back to prison here for the next 10 years.

24:04You know what? It actually is interesting. I am.

24:10I finally got a good night C plus. A few weeks of not. I woke up with a bunch of ideas. Funny how that works. But there was an aspect this week where Phillip was on it. I'm like, I'm in the muck here. This is not interesting. I've already said my piece. This is so tiring. And I do think that I have a perspective that I'm sort of working on in my head. Okay. Which means it will probably never show up, but we'll see what happens. But there's also an aspect to this that it's not like what it was.

24:42The problem with five years ago is there was nothing cool happening, right? Like there's room to like actually write about what you can do and what's possible. And like there is still room for thinking about businesses and what the impacts are. And, you know, there is a bit where this week, you know, I wrote about like Dreamforce of all things, right? Like, which, you know, Mark Benioff is a great character. So they're actually quite entertaining because he is an absolute character. But I also found it very satisfying and I think pertinent and useful.

25:14The question of AI on large enterprises is actually a totally open question. There are a lot of uncertainties here beyond the political. And I think I'm okay with continuing to, like, I feel like I said my piece on Monday and the other two days are about other stuff. And that's, I think, how I will continue to operate. And that's good for everyone. And there will be new products that are thrilling. I think it is a good distinction between where we were four or five years ago and where we are today.

25:45Because four or five years ago, it was the end of the beginning. And there were these sort of fixtures in tech and these tentpole products that everything organized around. And there's a lot of greenfield over the next five to ten years here. And that will keep things interesting. Every single large and all those tentpoles that seem so certain, like, there's legitimate, like, maybe five percent, ten percent, like, threats, right? It's so sobering note from Philip, but it's not so bad.

26:17There you go. The magic of a good night's sleep filled with optimism and enthusiasm for the future. Well, Philip's nominations for AGI Benchmarks. HBO Max moves me to the next episode automatically when I log back in instead of picking up in the credits of the episode I watched the night before. The footnote links work in the Substack app. And then three, NBA teams stop hiring Doc Rivers to be a coach. Look, Philip, okay? Look, I was – you brought me down.

26:48I was taking myself back up. I had to relive to Philip Riverson. Brings you right back down again. The most – or Doc Rivers. The most – Philip Riverson. Philip Riverson. Great quarterback. Great dad, apparently. Busy dad. The most devastating meme I saw of the Packers are mysteriously – they've lost – was like 77 to 5 in their last two fourth quarters or something like that. Or 47 to 5. And in between that – or 47 to 6. Between that, they extended their head coach. It's baffling.

27:18It's unbelievable to me. Someone posted a meme of the Packers sideline instead of Matt LaFleur. It was Doc Rivers. Oh, boy. You do have football Doc Rivers out there. Oh, you deserve better. Pretty, pretty rough. That image haunts me. Talk about cutting to your core. That's a tough beat.

Software experiences and family tech support

27:41But the other thing to Philip's point, I was thinking about this week. I think my mom – I was talking to my mom to set up a new phone. I think there was – there was something else I was doing. The extent to which software just sucks is really underrated. And honestly, this is part of tech's problem. Like, we endure it. We put up with it because the benefits are so large. But we underrate and underappreciate just how awful the experience is.

28:11What a grind of an experience it is for normal people. It's brutal. Yeah. Dude, it's brutal for me. And, like, I actually – I don't want to give it away because it's part of my overall idea of the take that I have. But there's a good chance that the software that – as we think about software – well, it's – you know, people ask, where's all the great AI software apps? There's a bunch of great AI software apps that are running a directory. Like, there's, like, literally making people better, which I will never sell in a million years because I don't want that responsibility.

28:43Yeah. And – but I can do that because it's not like I have to make the money back that it costs to build them, right? Like, the – it's just – it solves a need. It fills it. There's this really, really fundamental shift in what does it actually mean to democratize something. There is an analogy to blogging and newspapers and all those sorts of things. And it cannot come soon enough because, I'm sorry, the vast majority of stuff that even – that I deal with is just awful. And to see my poor mom, it's, like, it's –

29:14Oh, I can't even imagine. Yeah. I am – generally, I'm not as proficient as you, but I'm, like, fairly comfortable with computers and software. And I find myself tearing my hair out all the time. So, the Substack app – Like, there's this preciousness about, like, oh, our code is, like, what AI slop. It's, like, dude, your slop's not that good. The Substack app, honestly, makes me appreciate how hard it must be to build a functional app because it malfunctions in all kinds of maddening ways.

29:49And I love Substack as a platform, and I love a bunch of writers on there. So, it'll hit me with these notifications, and then I'll go to open the app, and it just doesn't work half the time. So, that's a great call out. If it doesn't work – no, it's, like, the self-driving car thing. It's, like, a self-driving car gets in an accident. It's, like, it's headline news. It's, like, people get in accidents all the freaking time. This is, like, the AI slop software thing. It's, like, oh, that's AI slop software. It's, like, yeah. Yeah, and it's a way smaller in volume than all the human slop software we've been putting up for years and years.

30:20So, sorry. I just made a lot of people really mad. I actually love you and appreciate you, summer developers. We absolutely love you. We do and appreciate how difficult it is. But, oh, my word. I, like, I just – I felt bad. Like, it's, like, this stuff – oh, actually, there's a funny anecdote for this. So, my parents – my mom was trying to set up a new YouTube premium plan. Okay. This is actually independent of her new phone. And I think what happened, I'm suspecting, is my dad joined it while he was at my house. He was on a different IP address.

30:50The long and short of it is my dad is definitely part of her family. My mom definitely has a YouTube premium plan. And YouTube is insistent my dad does not have YouTube premium. Oh, God. And so, we're, like, we're contacting them. And what's funny about the support, which I'm pretty sure is a real person, is the degree to which they were insistent that they told me, well, that's because your parents don't live together. And I'm, like, sir, I have 50 years of evidence to the contrary, that they do, in fact, live together.

31:23Maybe there's things I don't know about that's very possible. Yeah. But YouTube knows, but Ben doesn't. The Google Panopticon. Great. And the funny thing is I kind of understood, like, how – I'm sure that's what happened. I remember when my dad was doing it, like, oh, it's sort of quick for me. But the assumption on their part and the way it was stated was it was stated matter-of-factly. That's because they don't live together. Yeah. And I'm, like, they do live together. It's, anyhow. All part of the adventure, you know? Yeah, it's –

31:54Life's rich pageant as we live in the future here. Yes. Well, that's a good nomination. Let's understand that my parents do, in fact, live together. Maybe that's my measurement of AGI. Yeah, well, and every time I update my Apple software, my computing experience gets worse, which is not how it's supposed to work. So I'd love to solve that as well. Oh, it's rough. It's a new iPhone week, which is exciting and actually terrible because – Again, you just have to sort of grind through it. In any event, we could keep it moving and go to Thomas.

32:25We'll go rapid fire with some of the responses to your article.

Closing thoughts and preview wrap

32:28Thomas says – These are all, like, two paragraphs long. Well, Ben and Andrew, all of the AI conversations have been insane recently, but I'm a little confused on one specific thing. Why can't Anthropic or OpenAI unilaterally slow down development for safety? All right, and that is the end of the free preview. If you'd like to hear more from Ben and I, there are links to subscribe in the show notes, or you can also go to sharptech.fm. Either option will get you access to a personalized feed that has all the shows we do every week,

33:00plus lots more great content from Stratechery and the Stratechery Plus bundle. Check it out, and if you've got feedback, please email us at email at sharptech.fm. Thank you. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye.

33:31Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye. Bye.