Naval Ravikant 揭秘 All-In 播客成功秘诀

干得好，Naval。你太棒了。也许我应该在节目里说，但那真的是我录过的最有趣的播客。

Great job, Naval. You rocked it. Maybe I should have said this on air, but that was literally the most fun podcast I've ever recorded.

哦，这是播出的吗？剪进去。对，放到节目里。

Oh, that's on air? Cut that in. Yeah, put it in the show.

我本来有个理论解释你为什么是第一，但现在我明白了。真正的原因是什么？你认识我们很久了。

I had my theory on why you were number one, but now I have the realization. What's the actual reason? You know us for a long time.

是啊，你的理论是什么？现实又是什么？

Yeah, what was your theory? What's the reality?

我的理论是，我上播客的问题通常是和我聊天的人没那么有趣。他们只是问同样的问题，敷衍了事，而且并不真的感兴趣。我们并不是在进行同级别的真正对话。所以这就是为什么我想做 Airchat 和 Clubhouse 这类东西，因为你可以真正地对话。

My theory was that my problem with going on podcasts is usually the person I'm talking to is not that interesting. They're just asking the same questions, they're dialing it in, and they're not that interested. It's not like we're having a peer-level actual conversation. So that's why I wanted to do Airchat and Clubhouse and things like that, because you can actually have a conversation.

我明白了。对。而你们非常独特的一点是，你们有四个人，其中至少三个是聪明的。我开玩笑的。你怎么能这么说？SX 都不在这儿听你说。太刻薄了。最好的。这三个很聪明，而且你们都相处得很好，可以进行持续的对话。这成功率非常高。通常在一个播客里你只有一个有趣的人，而现在你有三个，也许四个。对吧？所以对我来说这就是原因。顺便说一句，你正在和第四个人说话。我们不知道，她永远神秘。四个人里，对吧？问题是，如果你把人们聚在一起聊天，两个人是好的对话，三个人可能四个是上限。这就是为什么在餐厅的餐桌上，四人桌，对吧？你不会安排五个人或六个人，因为那样对话就会分裂成多个。所以你们有四个能聊天的人，对吧？我以为这就是秘密，但还有另一个秘密。另一个秘密是你们玩得很开心。你们互相插话，互相取笑，你们真的在享受。所以这就是为什么我说这是我参加过的最有趣的播客。这也是为什么你随时欢迎回来。

I see. Right. And what you guys have very uniquely is four people, of whom at least three are intelligent. I'm kidding. How could you say that? SX isn't even here to hear you say that. That is so cold. Best right. These three are intelligent, and all of you get along, and you can have an ongoing conversation. That's a very high hit rate. Normally in a podcast you only get one interesting person, and now you've got three, maybe four. Right? Okay, so that to me was why. And by the way, you're talking to number four. We don't know, she remains mysterious forever. Of the four, right? The problem is like, if you get people together to talk, two is a good conversation, three possibly four is the max. That's why at a dinner table in a restaurant, four top, right? You don't do five or six because then it splits into multiple conversations. So you had four people who were capable of talking, right? I thought that was the secret, but there's another secret. The other secret is you guys are having fun. You're talking over each other, you're making fun of each other, you're actually having fun. So that's why I'm saying this is the most fun podcast I've ever been on. And that's why you'll be welcome back anytime.

不，都欢迎回来。是的，当然。三个聪明人。我简直不敢相信你会这么说 Sacks，他甚至不在场为自己辩护。抱歉，David。让赢家继续。我们向粉丝开源，他们简直疯了。好了，欢迎大家回到世界第一播客。今天非常兴奋。再次回来，你的科学苏丹，David Friedberg。你在那儿干什么呢，Friedberg？背景里是什么？

No, all welcome back. Yes, absolutely. In three smart guys. I can't even believe you'd say that about Sacks, he's not even here to defend himself. Sorry, David. Let your winners ride. We open sources to the fans and they've just gone crazy with it. All right, everybody welcome back to the number one podcast in the world. We're really excited today. Back again, your sultan of science, David Friedberg. What do you got going on there, Friedberg? What's in the background?

我以前在我的 Macintosh LC 上玩过很多叫《模拟地球》的游戏，很久以前了。说得通。是啊，说得通。当然还有我们的主席。你小时候玩什么游戏，J？其实我有点好奇，你玩过电子游戏吗？

I used to play a lot of a game called SimEarth on my Macintosh LC way back in the day. That tracks. Yeah, that tracks. And of course with us again, your chairman. What games did you play growing up, J? Actually, I'm kind of curious, did you ever play video games?

比如说 Andrea Allison，呃，Susan。我的意思是，有很多可爱的女孩。我出去约会女孩，Friedberg。是啊，我没有在我的 Apple II 上玩《文明》。让我找一张那些照片……哇，别给我惹麻烦，伙计。80 年代在布鲁克林对我很好。《拒绝》电子游戏。是的，你有三条命：被拒绝，被拒绝，被拒绝。这是个数字游戏，Chamath，你知道，你很清楚。这是个数字游戏。是啊。Nick，来吧，把 Rico Suave 调出来。哦不，这是什么？不是玩电子游戏，我在这儿。不，80 年代。那很胖，那很胖。J，你想让 Nick 帮你叔叔吗？是啊，他在这儿大杀四方。帮你叔叔弄 VIN。Jock C，你知道他在那儿卖什么吗？零食，是啊。前后都卖，对吧？正确。还有举重，牛肉干。请找到我的莱昂纳多·迪卡普里奥照片，把我那张胖下巴的照片换掉。谢谢。哦天哪，我那时候很胖，伙计。多了 40 磅比我重多了……这不是开玩笑，不是开玩笑。40 磅很多。我有很多很棒的 emo 照片。我为你骄傲。谢谢你，老兄。谢谢你，老兄。如果你想要好的……你能把介绍做完吗，这样我们就可以开始了？快点。你好吗，兄弟？你好吗，主席，独裁者？你还好吗？好了，好了。我们今天非常兴奋。今天，在 All-In 播客上首次，AngelList 的铁拳，早期的禅宗大师。他真有口才。他是书呆子的巫师。请欢迎我的哥们，Namaste Naval。你好吗？

Let's say Andrea Allison, uh, Susan. I mean, it was like a lot of cute girls. I was out dating girls, Friedberg. Yeah, I was not on my Apple II playing Civilization. Let me find one of those pictures of who... Whoa, don't get me in trouble, man. The 80s were good to me in Brooklyn. Rejection the video game. Yes, you have three lives: rejected, rejected, rejected. It's a numbers game, Chamath, as you know, as you well know. It is a numbers game. Yeah. Nick, go ahead, pull up Rico Suave here. Oh no, what is this one? Instead of playing video games, here I am. No, the 80s. That's fat, that's fat. J, you want to get Nick to help out your uncle? Yeah, here he is out slaying. Help out your uncle with the VIN. Jock C, you know what he was slinging in there? Snacks, yeah. Pre- and post-asm, right? Correct. And weightlifting, beef jerky. Go find my Leonardo DiCaprio picture, please, and replace my fat jow picture with that. Thank you. Oh God, I was fat, man. Plus 40 lbs is a lot heavier than I... It's no joke, it's no joke. 40 lbs is a lot. There's so many great emo photos of me. I'm proud of you. Thank you, my man. Thank you, my man. If you want a good... Can you get through the intros, please, so we can start? Come on, quick. How you doing, brother? How you doing, chairman, dictator? You good? All right, all right. We're really excited today. Today, for the first time on the All-In podcast, the Iron Fist of AngelList, the Zen master of the early stage. He has such a way with words. He's the sorcerer of nerds. Please welcome my guy, Namaste Naval. How you doing?

介绍回来了。这是我得到过的最好的介绍。我没想到你能做到。太棒了。那就是你的超能力。锁定它。放弃风险投资。就做这个。当然。实际上，你知道吗？有趣的是，世界第一播客，就像有人说的。我的意思是，那是我在显化的。它越来越近了。我们一直在前十名。所以我的意思是，周末对 All-In 来说很好。这一期会达到第一。这一期会病毒式传播。我觉得可以。如果你有一些非常精辟的见解，我们可能会直接登顶。我只需要做个降神会，它就会病毒式传播。哦不，哦不。你会把你的心送给我们吗？我的心与你同在。我的心……我在这儿以心结束。我不把它送出去。我把它留在这里。我把双手放在心上，稳稳地握住。我把它收住。它在向你发送，只是不明确。

The intros are back. That is the best intro I've ever gotten. I didn't think you could do that. That was amazing. That's your superpower right there. Lock it in. Quit venture capital. Just do that. Absolutely. That's actually, you know what? Interestingly, number one podcast in the world, like someone said. I mean, that's what I'm manifesting. It's getting close. We've been in the top 10. So I mean, the weekends are good for All-In. This one will hit number one. This one will go viral. I think it could. If you have some really great pithy insights, we might go right to the top. I just got to do a séance and it'll go viral. Oh no, oh no. Are you gonna send us your heart? My heart goes out to you. My heart... I end here at the heart. I don't send it out. I keep it right here. I put both hands on the heart and I hold it nice and steady. I hold it in. It's sending out to you, but just not explicitly.

好了，对于那些不认识的人，Naval 曾经是个企业家。他干得不错，也被揍过，然后他创办了 Venture Hacks，开始给人们发邮件说，你知道，20、15 年前，也许是 15 年，硅谷有一些交易。他到处跑，开始写 5 万、10 万美元的支票。他打出了一堆本垒打，把 Venture Hacks 变成了 AngelList。然后他投资了大量优秀的初创公司。也许给我们一些最成功的案例。

All right, for those of you who don't know, Naval was an entrepreneur. He kicked a bit of ass, he got his ass kicked, and then he started Venture Hacks and he started emailing folks and saying, you know, 20, 15 years ago, maybe 15, here are some deals in Silicon Valley. He went around, he started writing 50k, 100k checks. He hit a bunch of home runs and he turned Venture Hacks into AngelList. And then he has invested in a ton of great startups. Maybe give us some of the greatest hits there.

Dev，是啊。Twitter、Uber、Notion，还有很多其他的。Postmates，你，我。很多独角兽，很多后起之秀。我的意思是，实际上现在有很多交易。但老实说，我并不一定为成为投资者而自豪。投资者对我来说是副业，是爱好。所以我做初创公司。

Dev, yeah. Twitter, Uber, Notion, bunch of others. Postmates, you, me. A lot of unicorns, bunch of up-and-coming. I mean, it's actually a lot of deals at this point. But honestly, I'm not necessarily proud of being an investor. Investor to me is a side job, it's a hobby. So I do startups.

你怎么定义自己？

How do you define yourself?

我不定义。我的意思是，我想这些天我会说更像是建造东西。你知道，每个所谓的职业都是一场进化，对吧？你们所有人都是独立的，你们做自己最感兴趣的事，对吧？这就是赚钱的意义，这样你就可以做自己想做的事。所以这些天我真的很喜欢构建和打磨产品。我最近建了一个叫 Airchat 的东西。它有点没成功。我仍然为我所建造的感到自豪，并且与一个不可思议的团队合作。现在我正在构建一个新产品，这次我要进入硬件领域。我只是在建造一个我真正想要的东西。我不是完全自己做，Naval。部分地我会带投资者一起。上次他们拿回了钱。之前他们赚了钱。下次希望他们会赚很多钱。带朋友一起是好事。

I don't. I mean, I guess these days I would say more like building things. You know, every so-called career is an evolution, right? And all of you guys are independent and you kind of do what you're most interested in, right? That's the point of making money, so you can just do what you want. So these days I'm really into building and crafting products. So I built one recently called Airchat. It kind of didn't work. I'm still proud of what I built and got to work with an incredible team. And now I'm building a new product, and this time I'm going into hardware. And I'm just building something that I really want. I'm not doing it all yourself, Naval. Partially I bring investors along. Last time they got their money back. Previous times they've made money. Next time hopefully they'll make a lot of money. It's good to bring your friends along.

老实说，我喜欢你说的你爱那个产品但它没成功。没有足够多的人这么说。

I'll be honest, I love that you said you love the product but it didn't work. Not enough people say that.

是啊，我知道。我建了一个我爱的、我自豪的产品，但它没有火起来。而且它是一个社交产品，所以它必须火起来才能成功。所以我为团队找到了好归宿，他们都得到了报酬。

Yeah, I know. I built a product that I loved, that I was proud of, but it didn't catch fire. And it was a social product, so it had to catch fire for it to work. So I found the team great homes, they all got paid.

我带来的投资者拿回了他们的钱，我学到了很多，正在把这些经验用到新项目上。但新项目难得多，是硬件和软件的结合。你在 2024 年和 2025 年的构建过程中学到了什么以前不知道的东西？

The investors that I brought in got their money back and I learned a ton which I'm leveraging into the new thing. But the new thing is much harder. The new thing is hardware and software. And what did you learn building in 2024 and 2025 that you didn't know maybe before then?

主要是手艺本身，逐像素设计软件产品并发布的手艺。我想我学到的主要是，我真的很喜欢构建产品，并且想做一些更难、更真实的东西。和很多人一样，我也受到埃隆的启发，以及他所做的所有了不起的工作。所以我不想做容易的事，我想做困难而有趣的事，我想承担更多的技术风险，而不是市场风险。这是经典的 VC 经验，对吧？你想构建的东西是，如果人们理解了，如果你能交付，你知道人们会想要它，只是很难构建；而不是你建好了却不知道人们是否想要。这就是一个教训。

The main thing was actually just the craft, the craft of pixel by pixel designing a software product and launching it. I guess the main thing I took away that was a learning was that I really enjoyed building products and that I wanted to build something even harder and something even more real. And I think like a lot of us, I'm inspired by Elon and all the incredible work he's done. So I don't want to build things that are easy. I want to build things that are hard and interesting, and I want to take on more technical risk and less market risk. This is the classic VC learning, right? Which is you want to build something that if people get it, if you can deliver it, you know people will want it, and it's just hard to build, as opposed to you build it and you don't know if they want it. So that's a learning.

Airchat 很有趣。对于那些不知道的人来说，它有点像社交媒体网络，你可以提问，然后人们可以回应，就像基于音频的 Twitter。你会说这是最好的描述方式吗？音频 Twitter，异步，AI 转录和各种 AI 让它更容易，翻译。真正的好方法是尝试让播客类型的对话更容易被所有人接受，因为老实说，我不去播客的原因之一就是我不喜欢被中介，可以这么说，对吧？你坐在那里，有人采访你，然后你来我往，重复同样的事情。我只想和人交谈，我想要对等的关系，就像你们在这里运行的那样。

Airchat was a lot of fun. For those of you who don't know, it was kind of like a social media network where you could ask a question and then people could respond, and it was like an audio-based Twitter. Would you say that was the best way to describe it? Audio Twitter, asynchronous, AI transcripts and all kinds of AI to make it easier for you, translation. Really good way to try to make podcasting-type conversations more accessible to everybody, because honestly one of the reasons I don't go on podcasts is I don't like being intermediated, so to speak, right? Where you sit there and someone interviews you and then you go back and forth and you go through the same old things. I just want to talk to people. I want peer relationships, kind of like you guys have running here.

Naval，你经历那个阶段时发生了什么？有一段时间，你似乎经历了一些事情，然后你就知道了答案。你非常脚踏实地。不是说你现在不脚踏实地，但你发帖和写作没那么活跃了。但有一段时间，我想我们所有人都觉得，‘对，Naval 怎么想？真的吗？哦，好吧，这很有用。’我想那大概是十几岁末，二十岁出头。Jason，如果我说错了日期你可以纠正我。但就是那个时候，这些 Naval 主义思想和哲学真正开始了。我想人们对你思考问题的方式非常尊重。我只是好奇，你当时经历了什么？

Naval, what happened when you went through that phase? There was a period where it just seemed like something had gone on in your life and you just knew the answers. You were just so grounded. It's not to say that you're not grounded now, but you're less active posting and writing. But there was this period where I think all of us were like, 'Right, what does Naval think? Oh really? Oh okay, that's useful.' I would say it would be like the late teens, the early 20s. Jason, you can correct me if I'm getting the dates wrong. But it's been that moment where like these Naval isms and this sort of philosophy really started. I think people had a tremendous respect for how you were thinking about things. I'm just curious, like what were you going through in that moment?

是的，是的，是的，没错。不，非常有洞察力。是的，2010 年代。我从 2007 年就开始用 Twitter 了，因为我是早期投资者，但我从来没有真正发过推文。我没有被推荐，没有粉丝。我只是做通常的技术人员的事情，互相交谈。然后我在 2010 年创办了 AngelList。最初关于匹配投资者和初创公司的想法没有规模化。它只是一个邮件列表，早期爆炸式增长，但后来就没有规模化，所以我们没有业务。我试图弄清楚业务。同时，我收到了美国证券交易委员会的一封信，说‘哦，你在充当无牌经纪交易商。’我说，‘什么？我没赚钱，我没拿任何东西，这只是公共服务。’但即使那样，他们也在追查我。所以我当时压力很大。我还从投资者那里筹集了一大笔钱，所以我处于人生中压力非常大的时期。现在回想起来，我为那件事感到压力几乎有点可笑，但在当时感觉非常真实。一切的重担都压在我肩上：期望、人、钱、监管机构。我最终去了华盛顿，推动了法律修改，使我们做的事情合法化，这讽刺地促成了许多其他事情，比如 ICO、孵化器日和演示日。但在那个过程中，我处于人生中压力非常大的时期，我开始发推文，记录我正在经历的事情，我产生的任何领悟。只有在压力下，你才会被迫成长。所以无论我经历了什么样的内心成长，我都开始发推文，没想太多。它混合了三样我经常思考的东西：一是热爱科学，我是个业余爱好者，热爱物理学，就这样吧。我喜欢大量阅读哲学并深入思考。还有我喜欢赚钱，对吧？真理、爱和金钱，这是我 Twitter 简介上的玩笑。这三样东西是我不断回归的。所以我开始发关于所有这些的推文。我认为在那之前，人们期望像我这样的人应该只谈论金钱，待在自己的领域里。人们一直很谨慎。所以我认为这三者的结合引起了人们的注意，因为每个人都会思考一切。在现实生活中，我们不会只待在自己的领域里。我们处理人际关系，处理我们与宇宙的关系，处理我们认为是真实的东西，处理科学以及我们如何做决定、如何解决问题。我们还处理日常的物质生活：如何与配偶、女友、妻子或丈夫相处，如何赚钱，如何与孩子相处。所以我只是发关于一切的推文。我对一切都感兴趣。我发推文。其中很多最好的东西只是给自己的笔记。比如，‘嘿，别忘了这个。’‘如何致富’还记得那个吗？‘如何致富’那个帖子。超级病毒式传播。是的，砰，是的，是的，是的。我认为那仍然是 Twitter 上传播最广的帖子。我喜欢永恒的东西。我喜欢哲学。我喜欢在未来仍然适用的东西。我喜欢思想的复利，如果你愿意这么说的话。显然最近 X 变得如此令人上瘾，我们每天都在看，埃隆建立了完美的‘为你推荐’页面。他为书呆子们建立了 TikTok，我们都在上面。但通常我尽量忽略新闻。显然去年事情变得真实，我们都不得不关注新闻。但我只是喜欢发永恒的推文。我不知道，我的意思是人们有时会注意。他们喜欢我写的东西，有时他们会对我‘Naval’一下。但是的，‘如何致富’那个帖子是一个大事件。

Yeah, yeah, yeah, that's right. No, very insightful. Yeah, 2010s. I've been on Twitter since 2007 because I was an early investor, but I never really tweeted. I didn't get featured. I had no audience. I was just doing the usual techie guy thing, talking to each other. And then I started AngelList in 2010. The original thing about matching investors to startups didn't scale. It was just an email list that exploded early on but then just didn't scale, so we didn't have a business. And I was trying to figure out the business. And at the same time, I got a letter from the Securities and Exchange Commission saying, 'Oh, you're acting as an unlicensed broker-dealer.' And I'm like, 'What? I'm not making any money. I'm not taking anything. It's just a public service.' But even then, they were coming after me. So I wasn't. And I'd raised a bunch of money from investors, so I was in a very high stress period of my life. Now looking back, it's almost comical that I was stressed over that, but at the time it all felt very real. The weight of everything was on my shoulders: expectations, people, money, regulators. And I eventually went to DC and got the law changed to legalize what we do, which ironically enabled a whole bunch of other things like ICOs and incubator days and demo days. But in that process, I was in a very high stress period of my life, and I just started tweeting whatever I was going through, whatever realizations I was having. It's only in stress that you are forced to grow. And so whatever internal growth I was going through, I just started tweeting it, not thinking much of it. And it was a mix of three things that I kind of always run through: one is I love science, you know, I'm an amateur, love physics, let's just leave it at that. I love reading a lot of philosophy and thinking deeply about it. And I like making money, right? Truth, Love, and Money, that's my joke on my Twitter bio. Those are the three things that I keep coming back to. And so I just started tweeting about all of them. And I think before that, the expectation was that someone like me should just be talking about money, stay in your lane. And people had been playing it very safe. And so I think the combination of the three sort of caught people's attention, because every person thinks about everything. We don't just stay in our lane in real life. We're dealing with our relationships, we're dealing with our relationship with the universe, we're dealing with what we know to be true, with science and how we make decisions and how we figure things out. And we're also dealing with practical everyday material things: how to deal with our spouses or girlfriends or wives or husbands, and how to make money, and how to deal with our children. So I'm just tweeting about everything. I just got interested in everything. I'm tweeting about it. And a lot of it, my best stuff was just notes to self. It's like, 'Hey, don't forget this.' 'How to Get Rich' remember that one? 'How to Get Rich' that thread. Super viral. Yeah, bang, yeah, yeah, yeah. I think that is still the most viral thread ever on Twitter. I like timeless things. I like philosophy. I like things that still apply in the future. I like compound interests, if you will, in ideas. Obviously recently X has become so addictive that we're all checking it every day, and Elon's built the perfect 'for you' page. He's built TikTok for nerds, and we're all on it. But normally I try to ignore the news. Obviously last year things got real, we all had to pay attention to the news. But I just like to tweet timeless things. I don't know, I mean people pay attention sometimes. They like what I write, sometimes they go 'Naval' on me. But yeah, the 'How to Get Rich' thread was a big one.

现在人们见到你时，因为炒作和现实之间的差距，会不会有问题？现在有些不协调，因为如果人们吸收了这些内容，他们期望看到某种尤达大师漂浮在空中，你明白我的意思吗？

Is it problematic when people now meet you because the hype versus the reality? There's like it's discordant now because people if they absorb this content they expect to see some Yoda floating in the air, you know what I mean?

是的，没错。就像你们很多人已经戒酒了，但我过去偶尔喜欢喝杯酒。有一次我去见一位信息记者，那时我还经常见记者，她说，‘我们在哪里见面？’我说，‘哦，我们在葡萄酒商店见面，喝一杯。’她说，‘什么？你喝酒？’

Yes, yeah. Like many of you have stopped drinking, but I used to like have the occasional glass of wine. And there was a moment there where I went and met with an information reporter back when I used to meet with reporters, and she said, 'Where are we going to meet?' So I said, 'Oh, let's meet at the wine merchant and we'll get a glass.' She's like, 'What? You drink?'

你最近和蒂姆·费里斯做了一期关于育儿的播客。我很喜欢，还买了这个人的书。简单介绍一下这个育儿理念吧。

You recently did a podcast with Tim Ferriss on parenting. I love this and I bought the book from this guy. Just give a brief overview of this philosophy of parenting.

哦，我没听这期。我得记下来。跟我们说说吧。你会喜欢的。这很打动我，但有点疯狂。我是大卫·多伊奇的超级粉丝。我觉得他基本上是现世最聪明的人。他是一位研究量子计算的科学家，写了几本好书。内容是关于我们今天最伟大理论的交汇：认识论、知识理论、进化论、量子物理和计算。他写了《无穷的开始》和《真实世界的脉络》。我和他相处过，做过播客，雇佣并与他身边的人共事。他的框架让我更聪明。随着年龄增长，我们的大脑变慢，但智慧应该替代原始智力。他开创的一个理念叫“认真对待孩子”。意思是应该像对待成年人一样认真对待孩子。给他们和成年人同样的自由。如果你不会那样对配偶说话，或强迫配偶做事，也不要强迫孩子。我们只是通过潜在的暴力威胁来恐吓孩子。这让我产生共鸣，因为我成长得非常自由。我父亲不在身边，母亲没空管我们，所以我从五岁就开始自己做决定。九岁起，我就开始指挥别人。我一直在在家教育自己的孩子，所以这个理念很合拍。我找到了一个叫亚伦·斯图普尔的人，他非常擅长阐述这个理念。他几乎百分之百地践行它。他的孩子可以随便吃冰淇淋和士力架，随便玩 iPad，不想上学就不上，想穿什么就穿什么。一切都像和室友或成年人一样，通过讨论、协商、解释。这有点疯狂和极端，但我的家庭生活也朝这个方向走。我是个非常自由的人：没有办公室，没有日程表，不送孩子上学，尽量不强迫他们。

Oh, I didn't listen to this. I have to write this down. Tell us this. You're going to love this. This spoke to me, but it was a little crazy. So I'm a big fan of David Deutsch. I think he's basically the smartest living human. He's a scientist who worked on quantum computation and has written a couple of great books. It's about the intersection of the greatest theories we have today: epistemology, the theory of knowledge, evolution, quantum physics, and computation. He wrote 'The Beginning of Infinity' and 'The Fabric of Reality'. I've spent time with him, done podcasts, hired and worked with people around him. His framework has made me smarter. As we age, our brains slow down, but wisdom should substitute for raw horsepower. One thing he pioneered is called 'Taking Children Seriously'. It's the idea that you should take your children seriously, like adults. Give them the same freedom you would give an adult. If you wouldn't speak that way with your spouse or force them to do something, don't force a child. It's only through the latent threat of physical violence that we intimidate children. This resonated with me because I grew up very free. My father wasn't around, my mother didn't have bandwidth to watch us, so I made my own decisions from age five. From age nine, I was telling everyone what to do. I've been homeschooling my own kids, so the philosophy resonated. I found this guy Aaron Stupple on a chat, and he was an incredible expositor of the philosophy. He lives his life with it, 99% as extreme as one can go. His kids can eat all the ice cream and Snickers bars they want, play on the iPad all they want, don't have to go to school if they don't feel like it, dress how they want. Everything is a discussion, negotiation, explanation, just like with a roommate or adult. It's kind of insane and extreme, but I live my own home life in that direction. I'm a very free person: no office, no calendar, I don't send my kids to school, I try not to coerce them.

那么你的孩子，如果晚上 9 点想吃热狗，你会同意吗？两天前我订了一个热狗——其实是另一个牌子——我们晚上 9 点都吃了冰淇淋。他们是快乐的孩子。他们说‘爸爸，我想玩 iPad，玩堡垒之夜，别管我，我想什么时候睡就什么时候睡。’你就说‘好吧’。我最大的孩子可能一天玩九小时 iPad。那你的另一个孩子会因为懒得去厕所而尿裤子吗？他们不会，因为他们不喜欢尿裤子。我明白，但这里有个程度问题。你的观点是百分之百允许，没有任何评判？

So your kids, if they wanted to eat a hog dog at 9 PM, you'd be okay? Two nights ago I ordered a hog dog—actually it was a different brand—and we all ate ice cream at 9 PM. They're happy kids. They say 'Dad, I want to be on my iPad, playing Fortnite, leave me alone, I'll go to sleep when I want.' You're like 'okay.' My oldest probably plays iPad nine hours a day. So then your other kid pees in their pants because they're too lazy to walk to the bathroom? They don't do that because they don't like peeing their pants. I understand, but there's a spectrum. Your point of view is 100% allowed with no judgments?

不，我不是那个意思。那是亚伦的做法。我的规则有点不同。他们每天必须做一小时数学或编程，外加两小时阅读。一旦完成，他们就是自由人。其他一切都是协商。我们必须说服他们——这是说服，甚至不是协商。而且那一小时数学和两小时阅读——实际上数学可能只有 15 到 30 分钟，运气好才有一小时，阅读半小时到两小时。

No, that's not where I am. That's where Aaron is. My rules are a little different. They have to do one hour of math or programming plus two hours of reading every single day. The moment they've done that, they're free creatures. Everything else is a negotiation. We have to persuade them—it's a persuasion, not even a negotiation. And even the hour of math and two hours of reading—really you get 15 to 30 minutes of math, maybe an hour if you're lucky, and half an hour to two hours of reading.

那你觉得这样做的长期后果是什么？还有，如果他们做出不好的决定，比如晚上 9 点吃冰淇淋，健康方面会有什么后果？

And what do you think the long-term consequences of that are? And also, what about health consequences if they make bad decisions, like the ice cream at 9 PM?

你心里是怎么处理这件事的？我觉得无论什么年龄、人生什么阶段，你都在和自己的习惯作斗争。我们所有人还是会吃东西感到内疚，或者想吃不该吃的东西。我们一直在调整饮食。孩子也一样。我家老大上次没吃冰淇淋，说想吃得健康点，因为我终于说服了他。小一点的孩子会吃，但限量。中间那个有时会吃一点。如果他们提要求，你会允许但引导他们，比如'嘿，这不是我会做的选择，但如果你想要，我会试试。'你也要小心别吓到他们，别太专横，让他们觉得爸爸控制欲太强。

How do you manage that in your mind? I think whatever age you're at, you're still struggling with your own habits. All of us still eat food and feel guilty or want to eat something we shouldn't. We're always evolving our diets. Kids are the same. My oldest passed on the ice cream last time and said he wants to eat healthier because I finally got through to him. My younger kids will eat it but limited amounts. My middle kid sometimes eats some. If they say something, you enable it but guide them, like, 'Hey, this isn't the choice I would make, but if you want it, I'll try it.' You also have to be careful not to intimidate them or be so overbearing that they view Dad as controlling.

我觉得这很有意思。你觉得这些孩子长大成人后会变成什么样？你的愿景是什么？

I find this fascinating. What do you think happens to these kids when they're fully formed adults? What is that vision?

我尽量不去想。他们会成为他们自己。我小时候基本是想做什么就做什么。我宁愿他们有自主权，而不是完全按我的意愿发展，因为自主权是最难的——掌控自己的生活，自己做决定。我希望他们快乐。我的家庭非常幸福。

I try not to. They're going to be who they're going to be. I grew up kind of doing what I wanted. I would rather they have agency than turn out exactly the way I want, because agency is the hardest thing—having control over your own life, making your own decisions. I want them to be happy. I have a very happy household.

柏拉图的目标是什么？幸福（Eudaimonia），对吧？快乐而充实的人生。这是你希望他们拥有的吗？

What is Plato's goal? Eudaimonia, right? The happy, fulfilled life. Is that what you want for them?

我其实不期望他们什么。我只希望他们自由，成为最好的自己。

I don't really want anything for them. I just want them to be free and their best selves.

天哪，我想……沙马特在担心细节。他好像有 17 个孩子了。但我喜欢这次采访，因为那人说了一个很有意思的观点：他们迟早要自己做决定，学会权衡利弊——吃东西、玩 iPad。你越早让他们带着知识、学会提问、自主做决定，他们就越有安全感。我觉得这很迷人。我喜欢因果关系，尤其是对青少年。现在我有青少年了，让他们明白'如果你不做作业，就会有问题，然后你得解决它'是好事。我喜欢问'你的计划是什么？'无论他们遇到什么问题——8 岁还是 15 岁——我就说'你打算怎么解决？'然后听他们的计划，如果想一起讨论就告诉我。我觉得这是个非常有趣的讨论。总的来说，我的孩子们很快乐，家庭很幸福，大家相处融洽，彼此相爱。有些孩子远超同龄人，没有人在重要方面落后，没有人看起来不健康，没有异常饮食习惯，也没有异常行为。一切都很好，自我纠正。

God, I want to... Shamet is worrying about details. He's got like 17 kids now. But I love this interview because the guy made a really interesting point: they're going to have to make these decisions at some point, learn the pros and cons, the upside and downside of all these things—eating, iPad. The quicker you get them to have agency to make decisions for themselves with knowledge and to ask questions, the more secure they'll be. I found it fascinating. I like cause and effect, especially in teenagers. Now that I have a teenager, it's good for them to learn: if you don't do your homework, you have a problem, and then you solve it. I like to present it as 'What's your plan?' Anytime they have a problem—8-year-old, 15-year-old—I just say, 'What's your plan to solve this?' Then I hear their plan and let me know if you want to brainstorm. I thought it was a very interesting discussion. Overall, my kids are very happy, the household is happy, everybody gets along, everybody loves each other. Some are way ahead of their peers, nobody is behind in anything that matters, nobody seems unhealthy, no aberrant eating habits, no aberrant behavior. It's all good. Self-correcting.

我非常担心 iPad 这件事。我看到孩子们用 iPad，除非他们在做互动项目，否则如果只是看……说这话的人背景是电子游戏主题，他可能从小不停打游戏，每天在屏幕前待 9 小时。是啊，一回事。自己拿起手机却对孩子说'不许用 iPad'，这很虚伪。我从小不停打游戏，直到几年前。我不是批评 iPad。我四岁就开始用电脑，所以我理解。但我不能连续看 30 分钟的节目，接着再看，然后坐两小时一直看。我是在互动、在构建东西，这不一样。

I worry a lot about this iPad situation. I see my kids on an iPad, and unless they're doing an interactive project, if they end up watching... Says the guy who has a video game theme in the background, who probably grew up playing video games nonstop and spends 9 hours a day on his screen. Yeah, it's the same thing. There's a hypocrisy to picking up your phone and saying to your kid, 'No, you can't use your iPad.' I grew up playing video games non-stop until a few years ago. I'm not criticizing the iPad. I was on a computer since I was four, so I get it. But I didn't have the ability to play a 30-minute show, then the next, then sit for two hours watching shows. I was interacting, building stuff, which was different.

我们以前会控制他们看 YouTube，现在不了。我唯一的要求是看 YouTube 时打开字幕，这样有助于阅读，他们学阅读更快。好建议。我家一个孩子特别喜欢 YouTube，另外两个已经过了那个劲。他们用 YouTube 主要是查他们喜欢的游戏的视频，想玩得更好。

We used to control their YouTube access, although now we don't. The only thing I ask is that they put on captions when watching YouTube, so it helps their reading. They learn to read fast. Good tip. One of my kids is really into YouTube; the other two got over it. To the extent they use YouTube, it's mostly to look up videos on their favorite games to get better.

好了，我们继续。大卫·萨克斯也在。大卫，说说你的育儿哲学。

Alright, let's keep moving. We have David Sachs with us. David, give us your philosophy of parenting.

下一个议程。我们来谈谈真正的问题。这个节目不是育儿节目。我问大卫'你的育儿哲学是什么？'他说'哦，我四年前给他们设了信托基金。'所以他搞定了。信托设好了，万事大吉。搞定。你准备好了。告诉我效果如何。

Next item on the docket. Let's talk about some real issues. This show is not a parenting show. I asked David, 'What's your parenting philosophy?' He said, 'Oh, I set up their trust four years ago.' So he's done. Trust is set up, everything's good. Check. You're all set. Let me know how it works out.

好了，说到锻炼，我们有一位副总统，他不疯疯癫癫，而且真正理解 AI。JD·万斯发表了一篇精彩的演讲。我亲自看了。周二他在巴黎 AI 行动峰会上谈到了 AI。他做了一个 15 分钟的精彩演讲。他谈到了过度监管 AI 以及美国主导 AI 的意图。我们正好请到了 AI 沙皇纳瓦尔。在我深入细节之前，萨克斯，这篇演讲有很多措辞和想法，我以前听过。也许告诉我们这是怎么形成的，你有多自豪。

Alright, speaking of working out, we've got a vice president who isn't cuckoo for Cocoa Puffs and actually understands AI. JD Vance gave a great speech. I watched it myself. He talked about AI in Paris on Tuesday at the AI Action Summit. He gave a 15-minute banger of a speech. He talked about over-regulating AI and America's intention to dominate this. We happen to have with us Naval, the Zar of AI. Before I go into details, Sax, this speech had a lot of verbage, ideas I've heard before. Maybe tell us how it all came together and how proud you are.

有一位懂 AI 的副总统真是令人震惊。他能就热门话题可信地发言。这对美国来说是一个了不起的时刻。

Having a vice president who understands AI is mind-blowing. He can speak credibly on a topical topic. This was an awesome moment for America.

你在暗示什么，JC？是你和他一起打磨的，还是他聪明？两者都有？

What are you implying, JC? That you workshopped it with him, or that he's smart? Both?

副总统写了演讲稿，或者至少指导了全部内容。想法来自他。我不会居功。

The vice president wrote the speech or at least directed all of it. The ideas came from him. I'm not going to take any credit whatsoever.

好吧，它很到位。也许你可以谈谈。我同意它很到位。我认为这是一篇精心构思、表达出色的演讲。他提出了特朗普政府 AI 方针的四个要点。第一：他将确保美国 AI 继续保持黄金标准。太棒了。第二：他说政府……

Okay, well it was on point. Maybe you could talk about it. I agree it was on point. I think it was a very well-crafted and well-delivered speech. He made four main points about the Trump administration's approach to AI. Point one: he's going to ensure that American AI continues to be the gold standard. Fantastic. Check two: he says that the administration...

副总统在巴黎发表了演讲，提出了四个关键点：过度监管可能扼杀 AI；AI 必须保持不受意识形态偏见影响；白宫将维持有利于工人的 AI 增长路径；美国必须赢得 AI 竞赛。你有什么看法？

The Vice President gave a speech in Paris with four key points: excessive regulation could kill AI; AI must remain free from ideological bias; the White House will maintain a pro-worker growth path for AI; and the US must win the AI race. What are your thoughts?

我认为副总统的演讲令人耳目一新。他一开始就说要谈论 AI 机遇而非 AI 安全，这就像一记当头棒喝。之前的会议都聚焦安全——布莱切利园、欧盟 AI 法规、拜登的行政令。媒体也关注风险。所以他强调机遇是非常必要的纠正。然后他说美国必须赢得 AI 竞赛，在芯片、模型和应用上占据主导，释放能源，支持企业而非过度监管。他警告了监管俘获，并说减少监管可以创造公平竞争环境。他点名批评欧盟的《数字服务法》是对美国公司的速度陷阱，并敦促他们成为好伙伴。他还含蓄地对比了用 AI 进行控制的敌对国。总的来说，非常令人印象深刻——礼貌、乐观、前瞻，就像企业家或聪明投资者会说的话。

I think the Vice President's speech was a breath of fresh air. He started by saying he was there to talk about AI opportunity, not AI safety, which was a shot across the bow. Previous conferences were all about safety—Bletchley Park, EU AI regulation, Biden's executive order. The media also focuses on risks. So his emphasis on opportunity was a much-needed corrective. He then said the US must win the AI race, dominate in chips, models, and applications, unlock energy, and be supportive rather than overregulating. He warned about regulatory capture and said less regulation can create a level playing field. He called out the EU's Digital Services Act as a speed trap for American companies and urged them to be good partners. He also implicitly contrasted with adversarial countries using AI for control. Overall, it was very impressive—polite, optimistic, forward-looking, like what an entrepreneur or smart investor would say.

你看了演讲吗？你对 JD·万斯代表美国、追求美国例外主义有什么总体看法？

Did you see the speech? What are your general thoughts on JD Vance representing America and wanting American exceptionalism?

非常令人惊讶，非常令人印象深刻。我认为他礼貌、乐观、前瞻。我同意美国应该赢，我们不应该过度监管。我不是 AI 末日论者——我不认为 AI 会毁灭世界。那是另一个话题。有一种宗教不断说世界会终结：气候变化、AI、小行星、新冠。它吸引注意力。在 AI 的案例中，由于激励偏差，这种说法被过度渲染。领先的公司想抽掉梯子。他们真心相信安全风险，但他们是出于动机去相信。这是一个奇怪的立场：他们说 AI 太危险了，不应该允许开源，但又没危险到私人公司不能拥有。如果真是曼哈顿计划，你不会想让一家公司拥有核武器。山姆·奥特曼曾著名地说 AI 将捕获所有未来价值的光锥。如果那是真的，开源 AI 和小型科技公司就非常重要。问题在于训练模型高度集中，受益于超级计算机集群。所以还不清楚。

Very surprising, very impressive. I thought he was polite, optimistic, and forward-looking. I agree that America should win and we should not overregulate. I'm not an AI doomer—I don't think AI will end the world. That's a separate conversation. There's a religion that keeps saying the world will end: climate change, AI, asteroids, COVID. It fixates attention. In AI's case, it's overplayed due to incentive bias. Companies that are ahead want to pull up the ladder. They genuinely believe in safety risks, but they're motivated to believe. It's a weird position: they say it's so dangerous that open source shouldn't be allowed, but not so dangerous that a private company can't own it. If it were truly like the Manhattan Project, you wouldn't want one company to own nuclear weapons. Sam Altman famously said AI will capture the light cone of all future value. If that's true, open-source AI and small tech matter a lot. The problem is training models is highly centralized, benefiting from supercomputer clusters. So it's not clear.

任何去中心化的模型如何竞争？所以对我来说，真正的问题归结为：如何在推动 AI 前进的同时，不让极少数玩家控制整个领域？我们曾以为最初的 OpenAI 就是解决方案，它是一个非营利组织，本应为人类服务。但现在，为了激励团队和筹集资金，他们不得不至少将部分私有化。虽然我不清楚为什么他们需要将整个公司私有化，比如为什么要买断非营利部分。你可以保留非营利部分，同时设立私有部分用于激励。但我认为真正的挑战是如何防止 AI 自然走向中心化，因为底层的所有经济和技术本质上都是中心化的。如果你真的认为你要创造神，你愿意让一个实体牵着神的 leash 吗？对我来说，这才是真正的恐惧。我不害怕 AI，我害怕的是控制 AI 的极少数人为了我们好而对我们做些什么，因为历史总是如此。

How can any decentralized model compete? So to me, the real issue boils down to: how do you push AI forward while not having just a very small number of players control the entire thing? We thought we had that solution with the original OpenAI, which was a nonprofit supposed to do for humanity. But now, because they want to incentivize the team and raise money, they have to privatize at least part of it. Although it's not clear to me why they need to privatize the whole thing, like why they need to buy out the nonprofit portion. You could leave a nonprofit portion and have a private portion for incentives. But I think the real challenge is how to keep AI from naturally centralizing, because all the economics and technology underneath are centralizing in nature. If you really think you're going to create God, do you want to put God on a leash with one entity controlling God? That to me is the real fear. I'm not scared of AI; I'm scared of what a very small number of people who control AI do to the rest of us for our own good, because that's how it always works.

说得好。或许应该采用希腊模式，有很多神和英雄。Freeberg，你听了 JD Vance 的演讲吧？你对过度监管有什么看法？也许结合 Naval 的观点，一个人拥有这一切与开源相比如何？

Well said. Probably should go with the Greek model, having many gods and heroes as well. Freeberg, you heard the JD Vance speech, I assume. What are your thoughts on overregulation and, maybe to Naval's point, one person owning this versus open source?

我认为现在存在一种关于社会平衡的大定义，我称之为技术乐观主义和技术悲观主义。一般来说，人们会落入这两个阵营之一。技术乐观主义者认为，通过 AI、自动化、生物工程、制造业、半导体、量子计算、核能等加速成果，将带来一个富足的时代，因为技术创造了杠杆。技术会让东西更便宜、通缩，并给每个人更多，所以它创造了富足。挑战在于，已经拥有很多的人更担心下行风险，而不是渴望上行收益。所以技术悲观主义者通常是欧盟和美国大部分地区，担心失去 X、失去工作、失去这个那个。而像中国和印度这样的国家，更热衷于为民众创造财富、杠杆和富足的机会。欧盟的人均 GDP 是每年 6 万美元，美国是 8.2 万，但印度是 2500 美元，中国是 1.26 万美元。这些国家有更大的动力去实现上行，而美国和欧盟更担心下行。所以这是一场非常艰难的社会斗争。我认为随着时间的推移，那些拥抱这些技术的政府、国家和社会体系将变得更加资本主义，需要更少的政府控制和干预来创造就业、经济、支付等。而更技术悲观主义的国家将不幸地发现自己要求更多的政府控制、市场干预、政府创造就业、支付给人民，实际上由政府运行经济。我个人的观点是，我强烈支持技术加速，因为在人类历史上几乎每一个案例中，当新技术出现时，我们大多假设它在今天或昨天的框架下运作。汽车出现时，没有人预见到每个美国人都会拥有一辆，从而需要机械师、汽车经销商、道路等新行业。今天我们很难坐在这里说 AI 会摧毁工作，并预测它会创造什么。我认为我们很可能在给出的任何估计中都是错的。我认为最被低估的领域是那些今天看似不可行但 AI 可以解锁的大型技术项目，比如海洋居住、在水下建造城市、在月球上建造城市、在火星上建造城市或寻找新的居住地。这些在技术上今天很难实现，但 AI 可以解锁新的行业来实现这些转变。所以我认为当我们试图将技术视为去年或上个世纪的移植时，我们就大错特错了，然后我们因为担心失去现有的一切而成为技术悲观主义者。

I think there's this kind of big definition of social balance right now on what I would call techno-optimism and techno-pessimism. Generally, people sort of fall into one of those two camps. Techno-optimists are folks that believe accelerating outcomes with AI, automation, bioengineering, manufacturing, semiconductors, quantum computing, nuclear energy, etc., will usher in an era of abundance by creating leverage, which is what technology gives us. Technology will make things cheaper, deflationary, and give everyone more, so it creates abundance. The challenge is that people who already have a lot worry more about exposure to the downside than they desire the upside. So techno-pessimists are generally like the EU and large parts of the United States, worried about the loss of X, loss of jobs, loss of this and that. Whereas countries like China and India are more excited about the opportunity to create wealth, leverage, and abundance for their people. GDP per capita in the EU is $60,000 a year, in the US $82,000, but in India it's $2,500 and in China $12,600. There's a greater incentive in those countries to manifest upside than for the US and EU, who are more worried about downside. So it's a very difficult social battle underway. I do think over time, those governments and countries and social systems that embrace these technologies are going to become more capitalist and require less government control and intervention in job creation, economy, payments to people, etc. The countries that are more techno-pessimistic will unfortunately find themselves asking for greater government control, intervention in markets, government creating jobs, making payments to people, effectively running the economy. My personal view is that I'm a very strong advocate for technology acceleration because in nearly every case in human history, when a new technology emerged, we largely assumed it worked in the framework of today or yesterday. The automobile came along, and no one envisioned that everyone in the US would own one, requiring new industries like mechanics, car dealerships, roads, etc. It's very hard for us to sit here today and say AI will destroy jobs and predict what it will create. I think we're very likely wrong in whatever estimations we give. The area I think is most underestimated is large technical projects that seem infeasible today but AI can unlock, like habitation in the oceans, creating cities underwater, on the moon, on Mars, or finding new places to live. Those are technically very hard to pull off today, but AI can unlock new industries to enable those transitions. So I think we really get it wrong when we try to assume technology as a transplant for last year or last century, and then we become techno-pessimists because we're worried about losing what we have.

你是技术悲观主义者还是乐观主义者？因为你在节目中经常提到下行风险，但你每天都在以非常乐观的方式工作，为人们培育更好的草莓和土豆。所以你有点……

Are you a techno-pessimist or an optimist? Because you bring up the downside an awful lot here on the program, but you are working every day in a very optimistic way to breed better strawberries and potatoes for folks. So you're a little bit of a...

不，我完全没有技术悲观主义。我只是试图指出为什么另一方会那样行事。明白吗？好吧，放在完整的背景下，我想强调的是，我认为那个框架是错误的。我认为那个框架、那种运作方式，制造了对下行风险的担忧，从而产生了像我们在欧盟看到的那种监管。结果，如果欧盟和美国走那条路，中国的 GDP 将增长，而它们将停滞。这就是我对将要发生的事情的评估或看法。

No, I have no techno-pessimism whatsoever. I try to point out why the other side is acting the way they are. Got it? Okay, putting it in full context, what I'm trying to highlight is that I think that framework is wrong. I think that frame, the way of operating, creates this fear about worrying about downside, which creates regulation like we see in the EU. As a result, China's GDP will scale while the EU and US will stagnate if that's where they go. That's my assessment or opinion on what will happen.

Chath，你想为我们总结一下吗？你对 JD 有什么看法？我给你两个。好吧，第一个是，我认为这是一个非常有趣的时刻，我称之为两位副总统的故事。在拜登政府早期，卡玛拉被派去处理当时同样重要的问题，即非法移民，她去了墨西哥和危地马拉。所以这里实际上有一个非常有趣的 A/B 测试。两位副总统都在处理当时极其重要的问题。我认为 JD 专注、精准、有抱负。即使是那些非常支持卡玛拉的媒体，也找不到很多正面的话来说她，反馈是她漫无边际、回避问题，没有很好地回答被问到的问题。这很有趣，因为这在某种程度上是接下来四年和她的竞选活动的缩影，老实说，你可以从那个窗口看到……

Chath, you want to wrap this up for us? What are your thoughts on JD? I'll give you two. Okay, the first is I would say this is a really interesting moment where I would call this the tale of two vice presidents. Very early in the Biden administration, Kamala was dispatched on an equally important topic at that time, which was illegal immigration, and she went to Mexico and Guatemala. So you actually have a really interesting A/B test here. You have both vice presidents dealing with what were in that moment incredibly important issues. I think that JD was focused, precise, ambitious. Even the part of the press that was very supportive of Kamala couldn't find a lot of positive things to say about her, and the feedback was that she was meandering, ducking questions, and didn't answer the questions she was asked very well. It's so interesting because it's a bit of a microcosm of what happened over these next four years and her campaign, quite honestly, which you could have taken that window of the...

反馈对她来说不幸地一直非常一致。这是我观察到的一点，因为我听了他的演讲，也听了她的，那一刻我心想，哇，完全是两个不同的人。

Feedback and unfortunately for her it just continued to be very consistent. So that was one observation I had because I heard him give the speech, I heard her, and I had this kind of moment where I was like, wow, two totally different people.

第二点是关于 JD 讲话的实质。我在 Tucker 节目上说过，我会把这一切简化成一个非常基本的框架：如果你希望一个国家繁荣，它需要经济优势和军事优势。没有这两样，社会就会崩溃。而支撑这两者的唯一东西就是技术优势。我们今天就能看到这一点。

The second is on the substance of what JD said. I said this on Tucker and I'll just simplify all of this into a very basic framework: if you want a country to thrive, it needs to have economic supremacy and it needs to have military supremacy. In the absence of those two things, societies crumble. And the only thing that underpins those two things is technological supremacy. And we see this today.

所以周四，微软发生了什么？他们与美国陆军签订了一份 240 亿美元的合同，要交付某个尖端玩意儿，但他们意识到自己无法交付。那他们做了什么？他们去找了 Anduril。为什么找 Anduril？因为 Anduril 拥有实际执行的技术优势。

So on Thursday, what happened with Microsoft? They had a $24 billion contract with the United States Army to deliver some whizbang thing, and they realized that they couldn't deliver it. So what did they do? They went to Anduril. Now why did they go to Anduril? Because Anduril has the technological supremacy to actually execute.

几周前，我们看到中国在 DeepSeek 方面尝试技术优势。所以我认为这是一场非常简单的生存之战：那些能够驾驭和管理技术优势事物的人将会获胜，这将推动经济活力和军事优势，从而创造安全、强大的社会。就是这样。

A few weeks ago, we saw some attempts at technological supremacy from the Chinese with respect to DeepSeek. So I think that this is a very simple existential battle: those who can harness and govern the things that are technologically superior will win, and it will drive economic vibrancy and military supremacy, which then creates safe, strong societies. That's it.

所以从这个角度看，JD 说得很好。他看到了全局，而不是只见树木不见森林。他说出了我认为需要说的话，并提醒大家，你要么上船，要么下船。我认为这非常好。

So from that perspective, JD nailed it. He saw the forest from the trees. He said exactly what I think needed to be said and put folks on notice that you're either on the ship or you're off the ship. And I think that that was really good.

是的，还有一个小插曲，Sachs，如果你愿意的话，我很想和你聊聊。那就是所谓的 MAGA 1.0、MAGA 2.0、科技派与像我们这样的 MAGA 派，以及核心 MAGA 派之间的内战。我们可以把推文调出来。用 JD 自己的话说，他一直在与人互动，很明显这些推文是他自己写的。与其他政客和本届政府的一个显著区别是：他们直接告诉你他们的想法。推文是：‘我会试着写点东西来详细讨论这个问题。’这是 JD 万斯的推文。但我认为这场内战被夸大了。虽然确实，我称之为 MAGA 的民众与科技派之间存在一些真正的分歧。但简而言之，总的来说，我不喜欢用廉价劳动力取代美国劳动力。我对移民和外包的看法源于此。我喜欢增长和生产力提升，这影响了我对科技和监管的看法。具体到 AI，风险要么是被夸大了（正如你所说，Naval），要么是难以避免。我真正担心的一个例子是消费者欺诈。这是担心安全的合理理由。但另一个问题更严重：如果某个纯粹的国家在 AI 上领先美国 6 个月。我会再试着多说一些。而 JD 直接触及了，我认为，本届政府正在处理且已经处理过的最具争议的话题之一，即移民和科技，因为这两件事是相互关联的。如果我们失去数百万司机的工作，这在未来 10 年内会发生，就像我们失去数百万收银员的工作一样，那将影响我们国家和许多选民对边境和移民的看法。如果我们因为 AI 和自动驾驶汽车而失去数百万个工作岗位，我们可能无法让那么多人移民到这里。你对他直接讨论这个问题有什么看法，Sachs？

Yeah, and there was a little secondary conversation that emerged, Sachs, that I would love to engage you with if you're willing. Which is this civil war, quote unquote, between maybe MAGA 1.0, MAGA 2.0, techies versus the MAGA party like ourselves, and maybe the core MAGA folks. We can pull up the tweet here. In JD's own words, and he's been engaging people, it's very clear that he's writing these tweets. A distinct difference between other politicians and this administration: they just tell you what they think. Here it is: 'I'll try and write something to address this in detail.' This is JD Vance's tweet. But I think this civil war is overstated. Though yes, there are some real divergences between the populace I would describe as MAGA and the techies. But briefly, in general, I dislike substituting American labor for cheap labor. My views on immigration and offshoring flow from this. I like growth and productivity gains, and this informs my view on tech and regulation. When it comes to AI specifically, the risks are number one, overstated (to your point, Naval), or two, difficult to avoid. One of my any real concerns, for instance, is about consumer fraud. That's a valid reason to worry about safety. But the other problem is much worse: if a pure nation is 6 months ahead of the US on AI. Again, I'll try and say more. And this is JD going right at, I think, one of the more controversial topics that the administration is dealing with and has dealt with when it comes to immigration and tech, because these two things are dovetailing each other. If we lose millions of driver jobs, which we will in the next 10 years, just like we lost millions of cashier jobs, well, that's going to impact how our nation and many of the voters look at the border and immigration. We might not be able to let as many people immigrate here if we're losing millions of jobs to AI and self-driving cars. What are your thoughts on him engaging this directly, Sachs?

他提出的第一点是关于工资压力，对吧？当你开放边境，或者开放美国市场给那些由廉价劳动力在外国生产的产品，这些劳动力不受同样的标准、同样的最低工资、同样的工会规则或同样的安全标准约束，并且拥有巨大的成本优势时，你就给美国工人带来了工资压力。他反对这一点。我认为这一点很重要，因为媒体或新自由主义者喜欢把这种论点描绘成 MAGA 对无限移民的抵制是基于仇外心理之类的东西。不，这是基于实实在在的、餐桌上的问题：如果你有这种荒谬的开放边境政策，它必然会给金字塔底层的人带来巨大的工资压力。所以我认为 JD 是在提出这个论点。

Well, the first point he's making there is about wage pressure, right? Which is when you throw open our borders or you throw open American markets to products that can be made in foreign countries by much cheaper labor that's not held to the same standards, the same minimum wage, or the same union rules, or the same safety standards that American labor is, and has a huge cost advantage, then you're creating wage pressure for American workers. And he's opposed to that. And I think that is an important point because I think the way that the media or neoliberals like to portray this argument is that somehow MAGA's resistance to unlimited immigration is somehow based on xenophobia or something like that. No, it's based on bread and butter, kitchen table issues, which is if you have this ridiculous open border policy, it's inevitably going to create a lot of wage pressure for people at the bottom of the pyramid. So I think JD is making that argument.

但是，这是第二点，他说我不反对生产力增长。所以技术是好的，因为它使我们所有的工人能够提高生产力，这应该会带来更高的工资，因为工人可以生产更多，如果他们拥有更多提高生产力的工具，他们的劳动价值就会上升。所以这里没有矛盾。我认为他在解释为什么没有矛盾。

But, and this is point two, he's saying I'm not against productivity growth. So technology is good because it enables all of our workers to improve their productivity, and that should result in better wages because workers can produce more, the value of their labor goes up if they have more tools to be productive. So there's no contradiction there. And I think he's explaining why there isn't a contradiction.

我想补充一点，他在那条推文中没有提到这一点，但我想补充的是，过去 30 年来我们面临的一个问题是，美国的生产力增长巨大，但劳动力却无法从中获益。所有这些好处基本上都流向了资本或公司。我认为一个很大的原因是我们实行了基本上不受限制的移民政策。所以我认为，如果你限制移民，如果你停止非法移民，那么劳动力或许能够获得更多生产力增长的好处，这将是一件好事。这将更公平地分配生产力和技术带来的收益，我认为这将有助于平息你看到的科技界与美国其他地区（至少是腹地）之间日益增长的冲突。

A point I would add, he doesn't make this point in that tweet, but I would add is that one of the problems that we've had over the last, I don't know, 30 years is that we have had tremendous productivity growth in the US, but labor has not been able to capture it. All that benefit has basically gone to capital or to companies. And I think a big part of the reason why is because we've had this largely unrestricted immigration policy. So I think if you were to tamp down on immigration, if you were to stop the illegal immigration, then labor might be able to capture more of the benefits of productivity growth, and that would be a good thing. It'd be a more equitable distribution of the gains from productivity and from technology, and that I think would help tamp down this growing conflict that you see between technologists and the rest of the country, or certainly the heartland of the country.

Naval，这是一个……好吧，你想补充什么吗？David，抱歉。

Naval, this is a... okay, you want to add anything else? David, sorry.

嗯，我认为他在那条推文中提出的最后一点是，他谈到我们生活在一个有其他竞争国家的世界里，具体来说他没有提到中国，但他说我们有一个纯粹的竞争对手，如果他们最终在 AI 上领先我们六个月而不是落后六个月，世界将会糟糕得多。这是一个非常重要的观点，需要牢记。我认为整个巴黎 AI 峰会都是在认识到这一点的背景下举行的，因为就在几周前，我们看到了 DeepSeek，很明显中国并不比我们落后一年；他们紧追不舍，可能只落后我们几个月。所以如果我们用不必要的监管束缚自己，如果我们让我们的 AI 公司更难竞争，那并不意味着中国会效仿我们；他们会利用这一点并获胜。

Well, I think just the final point he makes in that tweet is that he talks about how we live in a world in which there are other countries that are competitive, and specifically he doesn't mention China, but he says we have a pure competitor, and it's going to be a much worse world if they end up being six months ahead of us on AI rather than six months behind. That is a really important point to keep in mind. I think that the whole Paris AI Summit took place against the backdrop of this recognition, because just a few weeks ago we had DeepSeek, and it's really clear that China is not a year behind us; they're hot on our heels, or only maybe months behind us. So if we hobble ourselves with unnecessary regulations, if we make it more difficult for our AI companies to compete, that doesn't mean that China's going to follow suit and copy us; they're going to take advantage of that fact and they're going to win.

好了，Naval，这似乎是我们这个时代的主要问题之一。现在这个播客的五个人中有四个是移民。所以我们有这个惊人的……在美国，这是一个由移民为移民建立的国家。你认为面对工作岗位流失，现在应该改变这一点吗？

All right, Naval, this seems to be one of the main issues of our time. Four of the five people on this podcast right now are immigrants. So we have this amazing... in America, this is a country built by immigrants for immigrants. Do you think that should change now in the face of job displacement?

毁灭，我知道你一直在密切关注自动驾驶。我们俩都对这感兴趣。这些年来，你知道，解决方案是什么？如果我们将会看到大量工作岗位被取代——某些工作肯定会这样，我们都心知肚明——我们是否应该关闭边境，不让下一个 Naval、Chamath、Max 和 Friedberg 进入这个国家？

Destruction which I know you've been tracking self-driving pretty acutely. We both have an interest there. I think over the years, you know, what's the solution here? If we're going to see a bunch of job displacement, which will happen for certain jobs, we all kind of know that. Should we shut the border and not let the next Naval, Chamath, Max, and Friedberg into the country?

嗯，我先声明一下我的立场。我是第一代移民，9 岁时来到这里，确切说是我的父母带我来的，后来我归化成了公民。所以显然我支持一定程度的移民。但话说回来，我已经同化了。我首先认为自己是个美国人，我流着红白蓝的血。我信仰《权利法案》和宪法，第一、第二、第四修正案等等。每年 7 月 4 日我都会站出来，故意在推特上捍卫第二修正案，这时我一半的粉丝就会发疯，你知道，因为我不该这么做。我应该是个好移民，对吧，带着一套连贯的左派政策、全球主义政策。所以我认为，合法的高技能移民，并给予同化的空间和时间，是合理的。你希望把最优秀最聪明的人才吸引到世界上最自由的国家，来建设技术，推动文明前进。正如 Chth 所说，经济实力和军事实力都源于技术。事实上，连文化也源于技术。看看避孕药对文化做了什么，汽车对文化做了什么，广播和电视对文化做了什么，还有互联网。所以技术驱动一切。如果你看财富，财富是一系列你能影响的物理转变，是资本和知识的结合。其中更大的投入是知识。所以美国成为了知识创造的家园，得益于引进了最优秀最聪明的人才。你甚至可以争论说 DeepSeek：我们失去它的部分原因就是那些孩子在美国学习，但我们却把他们送回了家，对吧？

Well, let me declare my biases up front. I'm a first-generation immigrant. I moved here when I was 9 years old, rather my parents did, and then I naturalized as a citizen. So obviously I'm in favor of some level of immigration. That said, I'm assimilated. I consider myself an American first and foremost. I bleed red, white, and blue. I believe in the Bill of Rights and the Constitution, first, second, fourth, and all the proper amendments. I get up there every July 4th and I deliberately defend the Second Amendment on Twitter, at which point half my followers go bananas, you know, because I'm not supposed to. I'm supposed to be a good immigrant, right, and carry the usual set of coherent leftist policies, globalist policies. So I think that legal, high-skill immigration with room and time for assimilation makes sense. You want to have a brain drain of the best and brightest coming to the freest country in the world to build technology and to help civilization move forward. And as Chth was saying, economic power and military power are downstream of technology. In fact, even culture is downstream of technology. Look at what the birth control pill did to culture, or what the automobile did to culture, or what radio and television did to culture, and then the internet. So technology drives everything. And if you look at wealth, wealth is a set of physical transformations that you can affect, and that's a combination of capital and knowledge. The bigger input to that is knowledge. And so the US has become the home of knowledge creation thanks to bringing in the best and brightest. You could even argue DeepSeek: part of the reason why we lost that is because a bunch of those kids studied in the US but then we sent them back home, right?

这真的准确吗？他们……是的，是的，是的，其中一些确实如此。哦天哪，这就像证据 A。哇，我不知道。

Is that actually accurate? That they were... yeah, yeah, yeah, some a few of them really. Oh my God, that's like exhibit A. Wow, I didn't know that.

所以我认为你绝对必须把熟练的、同化的移民分开，这是一个小群体，而且必须两者兼备：他们必须熟练，而且必须成为美国人。那个誓言不是无意义的，对吧？它必须有意义。所以熟练的、同化的移民，你必须把它和开放边境分开——谁都可以溜进来，随便进。后一部分毫无意义。如果拜登政府只允许智商 150 的人入境，我们现在就不会有这场辩论。我们之所以有这场辩论，是因为他们只是开放了边境，让数百万人涌入。把合法和非法移民混为一谈对他们有利。所以每次你都会说，‘嗯，我们不能只是开放边境’，然后说‘那 Elon 呢？这个呢？’他们就会列出来。如果他们只让 Elon、Jensen 和 Friedberg 进来，我们今天就不会有同样的对话。开放边境和工资压制之间的相关性是不可辩驳的。我们知道那些数据。而且我认为民主党，无论出于什么逻辑，犯了一个巨大的错误，基本上削弱了他们的核心群体。

So I think you absolutely have to split skilled, assimilated immigration, which is a small set, and it has to be both: they have to be skilled and they have to become Americans. That oath is not meaningless, right? It has to mean something. So skilled, assimilated immigration, you have to separate that from just open borders, whoever can wander in, just come on in. That latter part makes no sense. If the Biden administration had only been letting in people with 150 IQs, we wouldn't have this debate right now. The reason why we're having this debate is because they just opened the border and let millions and millions of people in. It was to their advantage to conflate legal and illegal immigration. So every time you'd be like, 'Well, we can't just open the border,' say 'What about Elon? What about this?' and they would just parade. If they were just letting Elon and Jensen and Friedberg in, we wouldn't be having the same conversation today. The correlation between open borders and wage suppression is irrefutable. We know that data. And I think that the Democrats, for whatever logic, committed an incredible error in basically undermining their core cohort.

我想回到你刚才说的，因为我认为这非常重要。现在有一种新的政治算法，我同意你的看法。我认为未来的三个群体是：轻资产的工薪和中产阶级，这是第一群体，大概有 1 到 1.5 亿人；然后是爱国的企业主；再然后是创新领袖。就是这三个。我认为 MAGA 做对的地方是，他们找到了这三个群体的交集。所以每次你看到这种左对右的二分法，它完全被误判了，对我们很多人来说听起来不和谐，因为我们没有人这样认同自己，对吧？我认为这是一个非常重要的观察，因为我们采取的政策需要反映这三个群体。这三者之间的共同点是什么？在这一点上，Naval 是对的：这三个群体不会对一种非常有针对性的、极其有用的合法移民形式——让非常非常聪明的人同意同化并成为美国的一部分——说太多不对的地方。

I want to go back to what you said because I think it's super important. There is a new political calculus on the field, and I agree with you. I think that the three cohorts of the future are: the asset-light working and middle class, that's cohort number one, there are probably 100 to 150 million of those folks; then there are patriotic business owners; and then there are leaders in innovation. Those are the three. And I think that what MAGA gets right is they found the middle ground that intersects those three cohorts of people. And so every time you see this sort of left versus right dichotomy, it's totally miscast and it sounds discordant to so many of us because that's not how any of us identify, right? And I think that that's a very important observation because the policies that we adopt will need to reflect those three cohorts. What is the common ground amongst those three? And on that point, Naval is right: there's not a lot that those three would say is wrong with a very targeted form of extremely useful legal immigration of very, very, very smart people who agree to assimilate and be a part of America.

我的意思是，我很高兴你这么说。就像，我记得小时候我父母会假装他们在斯里兰卡，有时我会非常沮丧。我说，‘如果你想在斯里兰卡，就回斯里兰卡去。我想成为加拿大人，’因为对我来说交朋友更容易，生活更容易。我尽了最大努力。我想成为加拿大人。然后 25 年前我搬到美国时，我想成为美国人。我现在感觉自己是美国人，我为自己是美国人而自豪。我认为这就是你想要的：你想要拥抱它的人。这不意味着我们不能偶尔穿上纱丽表演，但关键是，你相信什么，你的忠诚在哪里？

I mean, I'm so glad you said it the way you said it. Like, I remember growing up where my parents would try to pretend that they were in Sri Lanka, and sometimes I would get so frustrated. I'm like, 'If you want to be in Sri Lanka, go back to Sri Lanka. I want to be Canadian,' because it was easier for me to make friends, it was easier for me to have a life. I was trying my best. I wanted to be Canadian. And then when I moved to the United States 25 years ago, I wanted to be American. And I feel that I'm American now, and I'm proud to be an American. And I think that's what you want: you want people that embrace it. It doesn't mean that we can't dress up in a show of saris every now and then, but the point is, like, what do you believe and where is your loyalty?

Friedberg，我们曾经有一个大熔炉的概念，即同化，那是一件好事。然后它变成了文化挪用。我们在这里有点右转了。你对此持什么立场：招募最优秀最聪明的人，强迫他们同化，确保他们认同？Jason，找到那些在乎这里的人。

Friedberg, we used to have this concept of a melting pot, of assimilation, and that was a good thing. Then it became cultural appropriation. We kind of made a right turn here. Where do you stand on this: recruiting the best and brightest and forcing them to assimilate, making sure that they're down? Jason, find the people that care to be here.

是的，让我……我拒绝这个前提。整个对话……等等，等等，等一下。听着，我是第一代美国人，5 岁时搬到这里，10 岁时成为公民。是的，我完全是美国人，那是我唯一忠诚的国家。但我拒绝的前提是，不知何故，AI 对话导致了移民对话，因为数百万工作岗位将会消失。我们不知道那是否也是真的。我同意你做了一个巨大的假设，相信工作的末日，那不是……那些是事实，不是证据。我认为有任何工作被 AI 取代了吗？现实点吧。我们已经有 AI 两年半了，我认为它很棒，但到目前为止它只是一个更好的搜索引擎，帮助高中生作弊写论文。我的意思是，你不相信自动驾驶正在到来？等一下，Sacks。你不相信数百万……但等等，那些司机工作在 10 年前甚至不存在。Uber 出现了，创造了所有这些司机工作，DoorDash 创造了所有这些司机工作。所以技术做了什么？是的，技术摧毁了工作，但它用更好的机会取代了它们。然后要么你自己去抓住那个机会，要么一个企业家会出现，创造一些东西让你抓住那些机会。

Yeah, let me... I reject the premise. This whole conversation... wait, wait, hold on. Look, I'm a first-generation American who moved here when I was five and became a citizen when I was 10. And yes, I'm fully American, and that's the only country I have any loyalty to. But the premise that I reject here is that somehow an AI conversation leads to an immigration conversation because millions of jobs are going to be lost. We don't know that that's also true. I agree you're making a huge assumption buying into the doomsday of jobs that is not... those are fact not evidence. I think any jobs have been lost by AI? Let's be real. We've had AI for two and a half years, and I think it's great, but so far it's a better search engine and it helps high school kids cheat on their essays. I mean, you don't believe that self-driving is coming? Hold on a second, Sacks. You don't believe that millions... but hold on, those driver jobs weren't even there 10 years ago. Uber came along and created all these driver jobs, DoorDash created all these driver jobs. So what technology does? Yes, technology destroys jobs, but it replaces them with opportunities that are even better. And then either you can go capture that opportunity yourself, or an entrepreneur will come along and create something that allows you to capture those.

AI 是一种生产力工具。它提高了工人的生产力，让他们能做更多创造性工作，减少重复性劳动。因此，它让他们变得更有价值。是的，这需要一些再培训，但不多。这些是自然语言计算机；你可以用简单的英语和它们交谈，它们也用简单的英语回应你。但我认为 David 完全正确。我认为我们会看到 AI 创造就业的速度与破坏就业的速度一样快，甚至更快。你在互联网上就看到了这一点。比如 YouTube 出现后，看看所有这些 YouTube 主播和网红，这以前根本不是一份工作。新工作，实际上是新机遇，因为“工作”这个词用错了。“工作”意味着别人必须给我一份工作，像是施舍，是零和游戏。忘掉这些吧。这是机遇。疫情之后，看看有多少人通过在家工作、以各种你甚至无法理解的神秘方式在互联网上赚钱。

AI is a productivity tool. It increases the productivity of a worker, allowing them to do more creative work and less repetitive work. As such, it makes them more valuable. Yes, there is some retraining involved, but not a lot. These are natural language computers; you can talk to them in plain English and they talk back to you in plain English. But I think David is absolutely right. I think we will see job creation by AI that will be as fast or faster than job destruction. You saw this even with the internet. Like YouTube came along, look at all these YouTube streamers and influencers that didn't used to be a job. New jobs, really opportunities, because 'job' is a wrong word. 'Job' implies someone else has to give it to me, and it's sort of handed out, a zero-sum game. Forget all that. It's opportunities. After COVID, look at how many people are making money by working from home in mysterious little ways on the internet that you can't even quite grasp.

我是这样分类的。每当出现新技术，你都会获得生产力提升。你会遇到一些工作颠覆，意味着你工作的某部分可能会消失，但随后你会得到新的部分，希望是更高级、更有趣的。然后还有一些工作流失。我只是认为第三类会遵循历史趋势，即前两类总是更大，最终你会获得更多的净生产力和净财富创造。迄今为止，我们没有看到任何证据表明情况不会如此。诚然，AI 即将变得更强大。今年你会看到一波被称为智能体的新产品。智能体式产品能为你做更多事情。但目前还没有证据表明这些东西会完全无人监督并取代人们的工作。所以我认为我们必须观察这项技术如何发展。我认为所谓的欧洲方法的一个错误是，假设你可以完美准确地预测未来，准确到你可以今天制定法规来避免未来所有风险。而我们目前还了解得不够，无法做到这一点。那是确定性的问题。

Here's the way I categorize it. Whenever you have a new technology, you get productivity gains. You get some job disruption, meaning that part of your job may go away, but then you get other parts that are new and hopefully more elevated, more interesting. And then there is some job loss. I just think that the third category will follow the historical trend, which is that the first two categories are always bigger, and you end up with more net productivity and more net wealth creation. And we've seen no evidence to date that that's not going to be the case. Now it's true that AI is about to get more powerful. You're going to see a whole new wave of what are called agents this year. Agentic products are able to do more for you. But there's no evidence yet that those things are going to be completely unsupervised and replace people's jobs. So I think that we have to see how this technology evolves. And I think one of the mistakes of, let's call it the European approach, is assuming that you can predict the future with perfect accuracy, with such good accuracy that you can create regulations today that are going to avoid all these risks in the future. And we just don't know enough yet to be able to do that. That's level of certainty.

我同意你的看法。那些宣扬这种观点的公司，正如 Naval 所说，那些有经济利益的公司，至少是为了说服下一个增量投资者这可能成真，因为他们想声称所有资金都应该流向他们，这样他们就能攫取所有经济收益。这就是我们目前所处的周期阶段。所以如果你真的对这些反应进行分层，有小型 AI 初创公司相信生产力飞跃即将到来，繁荣即将来临。所有人都在旁观。然后有几家公司有极大的既得利益成为守门人，因为他们需要筹集下一个 300 或 400 亿美元，试图让人们相信这是真的。如果你从这个角度看，你是对的。我们还没有取得任何成就来证明这将是灾难性的。相反，历史会告诉你，它很可能像过去一样，总体上是生产性的，并丰富社会。

I agree with you. And the companies that are promulgating that view, as Naval said, those that have an economic vested interest in at least convincing the next incremental investor that this could be true, because they want to make the claim that all the money should go to them so they can Hoover up all the economic gains. And that's the part of the cycle we're in. So if you actually stratify these reactions, there's the small startup companies in AI that believe there's a productivity leap to be had and that there's going to be prosperity. Everybody on the sidelines watching. And then a few companies that have an extremely vested interest in them being a gatekeeper because they need to raise the next 30 or 40 billion dollars, trying to convince people that that's true. And if you view it through that lens, you're right. We have not accomplished anything yet that proves that this is going to be cataclysmically bad. And if anything, right now history would tell you it's probably going to be like the past, which is generally productive and enriches society.

是的，回到 JD 的演讲，那是我们的起点。我认为那是一个典型的美国式演讲，他说我们应该对这里的机遇持乐观态度，我认为这基本上是正确的。我们想要引领，想要利用这一点，不想束缚它。我们甚至还不完全知道它会是什么。我们将以工人为中心，我们希望支持工人。我认为如果对工人有不利影响，我们可以在未来缓解这些问题。但现在说我们知道该怎么做还为时过早。这更多是关于价值观的声明。

Yeah, and just to bring it back to JD's speech, which is where we started, I think it was a quintessentially American speech in the sense that he said we should be optimistic about the opportunities here, which I think is basically right. And we want to lead, we want to take advantage of this, we don't want to hobble it. We don't even fully know what it's going to be yet. We are going to center workers, we want to be pro-worker. And I think that if there are downsides for workers, then we can mitigate those things in the future. But it's too early to say that we know what the program should be. It's more about a statement of values at this point.

Freeberg，你觉得现在说还为时过早吗？考虑到乐观主义者以及所有这些正在被创造的机器人，我们在自动驾驶中看到的，你谈到过 Waymo 的加速发展，实际上要说我们不会看到数百万个工作岗位和数百万人被这些工作取代。你怎么看，Freeberg？我很好奇你的想法，因为那是反对的观点。

Do you think it's too early, Freeberg? Given optimists and all these robots being created, what we're seeing in self-driving, you've talked about the ramp-up with Waymo, to actually say we will not see millions of jobs and millions of people get displaced from those jobs. What do you think, Freeberg? I'm curious your thoughts because that is the counterargument.

我在工作场所的经验是，AI 工具正在做分析师或知识工作者过去需要很多小时才能完成的事情，现在只需几分钟。这并不意味着他们一天剩下的时间无所事事。对我们的业务以及像我们这样可以利用 AI 工具的其他企业来说，好处是这些人现在可以做更多事情。因此，我们组织的吞吐量、生产力提高了，我们现在可以更快地创造更多东西。所以无论我的公司生产什么产品，我们现在可以更快地制造更多东西，我们可以做更多的开发。你在实地看到了，对吧？在 Hollow？我也在实地看到了。我不认为这种关于 AI 对就业有多糟糕的移植是恰当的框架，它更多是关于生产力的加速。这就是为什么我回到人均 GDP 和 GDP 增长的观点。那些有兴趣加速产出、加速生产力、制造和销售能力的国家、社会、地区或行业，将迅速拥抱这些工具，因为它们允许用更少的资源做更多的事情。我认为这就是我在实地真正看到的。然后我要说的第二点是我之前提到的，我将以第三点结束。我认为我们大大低估了将会急剧、戏剧性出现的新行业。将会有很多我们现在没有深入思考的新事物，我们可以再花两个小时进行头脑风暴，讨论 AI 在大型项目方面解锁了什么，这些项目传统上或通常因为技术可行性限制而被搁置。这包括从加速新半导体技术到量子计算、能源系统、交通、居住等等。随着这些工具上线，每个行业都有可能发生各种转型，这将催生疯狂的新行业。最重要的一点是第三点，即我们不知道失业和就业创造之间的重叠（如果有的话）。因此，这些新技术影响和创造新市场的速度，但我认为 Naval 是对的。我认为在资本主义和自由社会中，资本和人们会涌向 AI 带来的新机遇，这些机遇的增长速度比旧泡沫的破灭更快。所以如果其他行业的就业需求出现通缩效应，我认为损失的速度会慢于利用创造新事物的热潮。

My experience in the workplace is that AI tools that are doing things that an analyst or knowledge worker was doing with many hours in the past is allowing them to do something in minutes. That doesn't mean that they spend the rest of the day doing nothing. What's great for our business and for other businesses like ours that can leverage AI tools is that those individuals can now do more. And so our throughput, our productivity as an organization has gone up, and we can now create more things faster. So whatever the product is that my company makes, we can now make more things more quickly, we can do more development. You're seeing it on the ground, correct? At a hollow? And I'm seeing it on the ground. And I don't think that this transplantation of how bad AI will be for jobs is the right framing, as much as it is about an acceleration of productivity. And this is why I go back to the point about GDP per capita and GDP growth. Countries, societies, areas that are interested, or industries that are interested in accelerating output, in accelerating productivity, the ability to make stuff and sell stuff, are going to rapidly embrace these tools because it allows them to do more with less. And I think that's what I really see on the ground. And then the second point I'll make is the one that I mentioned earlier, and I'll wrap up with a third point. Which is I think we're underestimating the new industries that will emerge drastically, dramatically. There is going to be so much new stuff that we are not really thinking deeply about right now that we could do a whole other two-hour brainstorming session on what AI unlocks in terms of large-scale projects that are traditionally or typically today held back because of the constraints on the technical feasibility of these projects. And that ranges from accelerating new semiconductor technology to quantum computing, to energy systems, to transportation, to habitation, etc., etc. There's all sorts of transformations in every industry that are possible as these tools come online, and that will spurn insane new industries. The most important point is the third one, which is we don't know the overlap of job loss and job creation if there is one. And so the rate at which these new technologies impact and create new markets, but I think Naval is right. I think that what happens in capitalism and in free societies is that capital and people rush to fill the hole of new opportunities that emerge because of AI, and that those grow more quickly than the old bubbles deflate. So if there's a deflationary effect in terms of job need in other industries, I think the loss will happen slower than the rush to take advantage of creating new things will happen.

另一方面，我的赌注大概是这样的：我认为新事物的创造速度会快于旧事物的消失速度。顺便提一句，现在帮人找工作最快的方法，如果你认识正在求职的人，最好的建议就是：‘嘿，去下载 AI 工具，开始跟它们对话，用各种方式使用它们。’然后你几乎可以走进任何领域的任何雇主那里说：‘嘿，我懂 AI’，他们就会当场录用你。

The other side, so my bet is probably on the order of I think new things will be created faster than old things will be lost. I think actually as a quick side note to that, the fastest way to help somebody get a job right now, if you know somebody in the market who's looking for a job, the best thing you can do is say, 'Hey, go download the AI tools and just start talking to them, just start using them in any way.' And then you can walk into any employer in almost any field and say, 'Hey, I understand AI,' and they'll hire you on the spot.

没错。Naval，你我都亲眼目睹了这一切。当年你做 Venture Hacks、我做 Open Angel Forum 的时候，我们每个月要费很大劲才能找到五到十家公司。然后创办公司的成本大幅下降，从五百万降到两百万，再到二十五万，再到十万。我认为我们同时看到了三件事：你会看到所有这些工作因自动化、自动驾驶汽车、收银员等而消失。但我们也看到像谷歌这样的公司团队规模停滞不前；他们不招人，只是让现有臃肿的员工学习工具。但不知道你是否注意到，能够用两三个人将产品推向市场并达到百万美元收入的初创公司数量正在激增。你在创业领域看到了什么？

Exactly. Naval, you and I watched this happen. We had a front row seat to it back in the day when you were doing Venture Hacks and I was doing Open Angel Forum. We had to like fight to find five or 10 companies a month. Then the cost of running these companies went down, they went down massively from five million to start a company to two, then to 250k, then to 100K. I think what we're seeing is like three things concurrently: you're going to see all these jobs go away for automation, self-driving cars, cashier, etc. But we're also going to see static team size at places like Google; they're just not hiring because they're just having the existing bloated employee base learn the tools. But I don't know if you're seeing this, the number of startups able to get a product to market with two or three people and get to a million in revenue is booming. What are you seeing in the startup landscape?

确实如你所说，存在杠杆效应，但与此同时，我认为更有趣的是，新的初创公司得以成立，否则它们根本不可能存在。我上一家创业公司 AirChat 如果没有 AI 就无法存在，因为我们需要转录和翻译。即使我现在正在做的事情，虽然不是 AI 公司，但没有 AI 也无法存在，它依赖 AI。甚至在 AngelList，我们也在大量采用 AI。无论你转向哪里，都是更多的机会、更多的机会、更多的机会。而人们喜欢在 Twitter 或那个前身为 Twitter 的平台上夸大其词，比如‘天哪，我们已经实现了 AGI’，‘天哪，我刚替换了我所有的中级工程师’，‘天哪，我已经停止招聘了’。对我来说，这很愚蠢。只有两种合理的情况：一种是创业者展示一个人或一个女孩，他们借助 AI 疯狂扩张；另一种是那些拥抱 AI 的人说‘我需要招聘，我需要招聘任何会拼写 AI 的人，任何使用过 AI 的人，都请进来，请进来。’再次强调，要看到 AI 不是在夺走工作而是在创造机会，最简单的方法是：去提升你的技能，学一点东西，看一些视频，使用 AI，摆弄它，然后重新申请那份拒绝过你的工作，看看他们如何把你拉进来。

Definitely what you're saying in that there's leverage, but at the same time, I think the more interesting part is that new startups are enabled that could not exist otherwise. My last startup, AirChat, could not have existed without AI because we needed transcription and translation. Even the current thing I'm working on, it's not an AI company, but it cannot exist without AI; it is relying on AI. Even at AngelList, we're significantly adopting AI. Like everywhere you turn, it's more opportunity, more opportunity, more opportunity. And people like to go on Twitter or the artist formerly known as Twitter and basically they like to exaggerate, like 'Oh my God, we've hit AGI,' 'Oh my God, I just replaced all my mid-level engineers,' 'Oh my God, I've stopped hiring.' To me, that's like moronic. The two valid ones are the one entrepreneur shows where there's like one guy or one gal and they're like scaling up like crazy thanks to AI, or there are people who are embracing AI and being like 'I need to hire and I need to hire anyone who can even spell AI, like anyone who's even used AI, just come on in, come on in.' Again, I would say the easiest way to see that AI is not taking jobs but creating opportunities is: go brush up on your skills, learn a little bit, watch a few videos, use the AI, tinker with it, and then go reapply for that job that rejected you and watch how they pull you in.

2023 年，经济学家理查德·鲍德温说：‘AI 不会夺走你的工作，而是会使用 AI 的人夺走你的工作，因为他们比你更懂得如何使用它。’这已经成了一个梗，你在 X 上经常能看到。但我认为这很有道理。你知道，只要你保持适应能力，不断学习，学会利用这些工具，你就会做得更好。如果你把自己与科技隔绝，不去利用它，那才是你把自己置于危险之中。另一种思考方式是，这些是自然语言计算机，所以以前对电脑感到畏惧的人不再需要畏惧了。你不再需要用某种晦涩的语言编程或学习某种深奥的数学才能使用它们；你只需跟它们说话，它们就会回应你。这就是魔法。新的编程语言是英语。

In 2023, an economist named Richard Baldwin said: 'AI won't take your job, it's someone using AI that will take your job because they know how to use it better than you.' And that's kind of become a meme, and you see it floating around X. But I think there's a lot of truth in that. You know, as long as you remain adaptive and you keep learning and you learn how to take advantage of these tools, you should do better. And if you wall yourself off from the technology and don't take advantage of it, that's when you put yourself at risk. Another way to think about it is these are natural language computers, so everyone who's intimidated by computers before should no longer be intimidated. You don't need to program anymore in some esoteric language or learn some obscure mathematics to be able to use these; you can just talk to them and they talk back to you. That's magic. The new programming language is English.

Chat，你想就此总结一下这个机遇/替代/混乱的话题吗？我之前就想说，但我现在越来越不相信你还应该费心去学习很多我们过去视为基础的硬科学和数学了。比如我过去认为每个人都应该去学工程。我现在不像以前那么确信了，因为我过去常说：‘嗯，这很好，第一性原理思维等等，你会接受一套可扩展的工具包训练。’但我不确定这是否正确。我认为你可以使用这些智能体，使用深度研究，它们突然就取代了很多那种技能。那么剩下的是什么？是创造力、判断力、历史、心理学，以及所有其他软技能、领导力、沟通能力，这些让你能够以建设性的方式操控这些模型。因为当你想到能让你得到出色答案的提示工程时，它实际上只是以完全不同的正交和非线性方式思考。所以这是我最后的想法：它确实打开了视野，意味着对于每一个聪明的数学天才，还有许许多多其他高情商的人，突然之间，这个工具把技能从只有高智商的人手中夺走，并说如果你有这些其他技能，现在你可以和我平等竞争。我认为这对很多人来说是一种解放。

Chat, you want to wrap us up here on this opportunity slash displacement slash chaos? I was gonna say this before, but I'm pretty unconvinced anymore that you should bother even learning many of the hard sciences and maths that we used to as underpinnings. Like I used to believe that the right thing to do was for everybody to go into engineering. I'm not necessarily as convinced as I used to be, because I used to say, 'Well, that's great, first principles thinking, etc., etc., and you're going to get trained in a toolkit that will scale.' And I'm not sure that that's true. I think like you can use these agents and you can use deep research and all of a sudden they replace a lot of that skill. So what's left over? It's creativity, it's judgment, it's history, it's psychology, it's all of these other sort of soft skills, leadership, communication that allow you to manipulate these models in constructive ways. Because when you think of the prompt engineering that gets you to great answers, it's actually just thinking in totally different orthogonal ways and nonlinearly. So that's my last thought, which is it does open up the aperture, meaning for every smart mathematical genius, there's many, many, many other people who have high EQ, and all of a sudden this tool actually takes the skill away from the person with just a high IQ and says if you have these other skills, now you can compete with me equally. And I think that that's liberating for a lot of people.

我属于更多机会的阵营。你知道，我目睹了电影行业在数码相机出现时的巨大变化，更多人开始制作纪录片，更多人开始制作独立短片，然后当然是 YouTube 革命，人们开始在 YouTube 上制作视频或像这样的播客。如果你看看特效行业发生了什么，制作一部《星球大战》电影、一部《星球大战》系列、一部漫威系列所需的人手大大减少了。正如我们现在所见，我们可以用更少的人制作《曼达洛人》、《阿索卡》以及所有其他系列，而且它们看起来显然比原版《星球大战》系列甚至前传更好。所以会有更多的机会。我们现在制作更多的电视节目、更多的系列，每一个小角色我们想看的都有了。初创公司也在发生同样的事情。我简直不敢相信现在有一个叫 Slopes 的应用，专门用于滑雪，还有 20 个非常好的冥想应用，以及 10 个非常好的禁食应用。我们正在沿着这条长长的机会尾巴走下去，会有大量百万到千万美元级别的生意等着你。你知道，人们学会了使用这些工具。我喜欢这一点让你心动：你得到一部额外的漫威电影或一部额外的《星球大战》剧集。所以这让你心动。我认为对很多人来说，AI 可能会接管世界，这感觉很好，但我会得到下一部《星球大战》电影，所以我没问题。

I'm in the camp of more opportunity. You know, I got to watch the movie industry a whole bunch when the digital cameras came out, and more people started making documentaries, more people started making independent film shorts, and then of course the YouTube Revolution, people started making videos on YouTube or podcasts like this. And if you look at what happened with the special effects industry as well, we need far fewer people to make a Star Wars movie, to make a Star Wars series, to make a Marvel series. As we've seen now, we can get The Mandalorian and Ahsoka and all these other series with smaller numbers of people, and they look better than obviously the original Star Wars series or even the prequels. So there's going to be so many more opportunities. We're now making more TV shows, more series, everything we wanted to see of every little character. That's the same thing that's happening in startups. I can't believe that there is an app now, Naval, called Slopes just for skiing, and there are 20 really good apps for just meditation, and there are 10 really good ones just for fasting. Like we're going down this long tail of opportunity, and there'll be plenty of million to 10 million businesses for you. You know, people learn to use these tools. I love how that's the thing that tips you over: which one you get an extra Marvel movie or an extra Star Wars show. So that tips you over. I think for a lot of people, it feels great that AI may take over the world, but I'm gonna get a next Star Wars movie, so I'm cool.

是啊，我是说，你不觉得有趣吗？最后一点：你看，在技术乐观主义和技术悲观主义这两类之间做选择，我绝对是乐观主义阵营的。我认为我们也应该如此。但我认为实际上还有第三类，我想提出的是……

Yeah, I mean, are you not entertained? Want a final point on this: look, I mean, given the choice between the two categories of techno-optimist and techno-pessimist, I'm definitely in the optimist camp. And I think we should be. But I think there's actually a third category that I would submit, which is...

技术现实主义认为技术一定会发生。试图阻止它就像命令潮汐停下一样。如果我们不做，别人就会做——中国会做，或者其他人会做。我们最好掌控技术，成为领导者，而不是被动等待它降临到我们头上。我认为这总是对的：企业最好主动出击，引领变革，自我颠覆，而不是等着别人来做。对国家来说也是如此。我在副总统的演讲和推文中看到了技术现实主义的影子：看，人工智能一定会发生，我们不如成为领导者。如果我们不这样做，我们可能会在一个对国家安全、经济和许多方面都有影响的关键领域失利。所以我们不想生活在那样一个世界里。我认为这场辩论很多是学术性的，因为无论你是乐观主义者还是悲观主义者，都只是半杯水的问题。问题只是：它会不会发生？我认为答案是肯定的。那么我们就想控制它。简单来说：我们在这个问题上没有太多选择。

Techno-realism is the view that technology is going to happen. Trying to stop it is like ordering the tides to stop. If we don't do it, somebody else will—China will do it, or somebody else will. It's better for us to be in control of the technology, to be the leader, rather than passively waiting for it to happen to us. I think that's always true: it's better for businesses to be proactive and take the lead, disrupt themselves instead of waiting for someone else to do it. And I think it's better for countries. I did see a hint of this techno-realism idea in the Vice President's speech and his tweet: look, AI is going to happen, we might as well be the leader. If we don't, we could lose in a key category that has implications for national security, our economy, and many things. So that's not a world we want to live in. I think a lot of this debate is academic because whether you're an optimist or pessimist is just glass half empty or half full. The question is just: is it going to happen or not? And I think the answer is yes. So then we want to control it. Let's boil it down: there's not a tremendous amount of choice in this.

我非常同意一点，但想对另一点稍作调整。我同意的一点是，它无论如何都会发生，而 DeepSeek 证明了这一点。你可以切断对他们的芯片供应和人才流动，但他们怎么做？他们只是变得更高效，然后把它出口回给我们。当我们的团队出于安全原因保持闭源时，他们给我们送回了最好的开源模型。DeepSeek 打破了美国在这一领域拥有垄断地位、我们可以放慢列车速度并完全控制它的谬论。DeepSeek 表明：不，如果我们放慢列车速度，他们只会赢。

I would agree heavily with one point and tweak another. The point I agree with is that it's going to happen anyway, and that's what DeepSeek proved. You can turn off the flow of chips to them and the flow of talent, but what do they do? They just get more efficient, and they exported it back to us. They sent us back the best open-source model when our guys were staying closed-source for safety reasons. DeepSeek exploded the fallacy that the US has a monopoly in this category and that somehow we can slow down the train and have total control over it. What DeepSeek showed is: no, if we slow down the train, they're just going to win.

我要稍作调整的一点是，当你说“我们”美国会赢时，问题在于，获胜的最佳方式是尽可能开放、分布、创新。如果这一切最终由一家公司控制，他们的创新速度实际上会比一个动态系统慢。而动态系统本质上将是开放的；它会泄露到中国，泄露到印度。但这些事物具有强大的网络效应。我们对技术了解这一点：几乎所有技术底层都有网络效应。所以即使你开放，你仍然会赢，仍然会控制。看看互联网：互联网就是如此，对吧？互联网是一种基于大量开源技术的开放技术。主导公司是谁？所有主导公司都是美国公司，因为它们处于领先地位。

The part I would tweak a little bit is the idea that we are going to win by 'we' when you say America. The problem is that the best way to win is to be as open, as distributed, as innovative as possible. If this all ends up in the control of one company, they're actually going to be slower to innovate than if there's a dynamic system. And that dynamic system by its nature will be open; it will leak to China, it will leak to India. But these things have powerful network effects. We know this about technology: almost all technologies have network effects underneath. So even if you are open, you're still going to win and you're still going to control. Look at the internet: that was all true for the internet, right? The internet is an open technology based on tons of open source. Who are the dominant companies? All the dominant companies are US companies because they were in the lead.

完全正确。我们拥抱了开放的互联网。那是不一样的。

Exactly right. We embraced the open internet. That was different.

所以这对全人类都有好处。我认为副总统的演讲非常明确：看，我们希望你们加入，我们希望成为好伙伴。然而，经济上和军事上肯定会有赢家。而要成为赢家之一，你必须成为率先实现 AGI 的领导者。

So there will be benefits for all of humanity. And I think the Vice President's speech was really clear: look, we want you guys to be on board, we want to be good partners. However, there are definitely going to be winners economically and militarily. And in order to be one of those winners, you have to be a leader who's going to get to AGI first.

Naval，最终会是开源还是闭源获胜？如果我们五年或十年后坐在这里，看看排名前三的语言模型……我不认为我们知道如何构建 AGI，但那是更长的讨论。五年后谁将拥有最好的模型？

Naval, is it going to be open source or closed source that wins the day? If we're sitting here five or ten years from now and looking at the top three language models... I don't think we know how to build AGI, but that's a much longer discussion. Who's going to have the best model five years from now?

等等，我百分之百同意你。我只是认为这是不同的事情。但我们正在构建的是这些不可思议的自然语言计算机。实际上，David 非常简洁地总结了两个主要用例：搜索和家庭作业，以及文书工作。这真的是文书工作，我们谈论的许多即将消失的工作实际上都是文书工作。它们是文书搬运。这些是虚构的工作。就像联邦政府，我们通过 DOGE 发现，三分之一的人在用勺子挖洞，另外三分之一在填洞。他们填满文书工作，然后把它埋进矿井里。他们把它埋进矿井山。

Hold on, I 100% agree with you. I just think it's a different thing. But what we're building are these incredible natural language computers. And actually, David in a very pithy way summarized the two big use cases: it's search and it's homework, it's paperwork. It's really paperwork, and a lot of these jobs that we're talking about disappearing are actually paperwork jobs. They're paperwork shuffling. These are made-up jobs. Like the federal government, as we're finding out through DOGE, a third of it is like people digging holes with spoons and another third are filling them back up. They're filling up paperwork and then burying it in a mineshaft. They're burying it in a mineshaft mountain.

是的，我认为很多这些虚构的工作——他们得下到矿井里，当有人退休时取出文书工作再带上来。你知道吗？我要给他们一些 U 盘。我们可以用 U 盘提高电梯的吞吐量。那会不可思议。我们发现 DMV 在过去 70 年里一直在管理政府。这是复利效应。这就是实际情况。DMV 在负责。我的意思是，如果世界在核战争中终结，上帝保佑，唯一剩下的将是蟑螂，然后是一堆政府工作文件、TPS 报告，基本上都在矿井里。

Yeah, I think a lot of these made-up jobs—they got to go down the mineshaft to get the paperwork when someone retires and bring it up. You know what? I'm going to get them some thumb drives. We can increase the throughput of the elevator with some thumb drives. It would be incredible. What we found out is the DMV has been running the government for the last 70 years. It's been compounding. That's really what's going on. DMV is in charge. I mean, if the world ends in nuclear war, God forbid, the only things left will be the cockroaches and then a bunch of government work documents, TPS reports, down in the mineshaft basically.

是的，让我们花点时间感谢我们的 Zar。我们想念他。我们希望他能全程参与节目。谢谢你，Zar。谢谢 Zar。再见，各位。想你。我们想你，小家伙。我希望我们能谈谈乌克兰，但不允许。回去工作吧。我们改天喝咖啡时再聊。我会在食堂见你。谢谢邀请。再见，伙计。

Yeah, let's take a moment everybody to thank our Zar. We miss him. We wish he could be here for the whole show. And thank you, Zar. Thank you to the Zar. See you guys. Miss you. We miss you, little buddy. I wish we could talk about Ukraine, but we're not allowed. Get back to work. We'll talk about it another time over coffee. I'll see you in the commissary. Thanks for the invite. Bye, man.

我太兴奋了。Naval 邀请我去军队食堂。我要去小卖部了。不，他没有。是你自己邀请自己的。老实说。

I'm so excited. Naval invited me to go to the military mess. I'm going to be in the commissary. No, he didn't. You invited yourself. Be honest.

是的，我确实邀请了。我把它放在他的日历上了。

I did, yes I did. I put it on his calendar.

为了保持对话进行，让我转入一个非常重要的话题：关税。关键是，尽管互联网是开放的，美国赢得了互联网的很大一部分。许多美国公司赢得了互联网，他们之所以赢，是因为我们以军事术语所说的“最先最多”到达。这很重要，因为许多科技企业底层具有规模经济和网络效应，甚至基本的基于品牌的网络效应。如果你回到 90 年代末、2000 年代初，很少有人会预测到亚马逊最终会基本上拥有所有电子商务。你会认为那会是完全竞争且非常分散的。这同样适用于我们最终如何得到 Uber 作为基本上一家出租车服务，或者我们得到 Airbnb、Meta。只是网络效应、网络效应、网络效应统治世界。但当涉及到关税和贸易时，我们表现得好像网络效应不存在。经典的李嘉图比较优势教条说，你应该生产你最擅长的，我生产我最擅长的，然后我们贸易。然后即使你想对我收取更多费用，如果你想对我运给你的商品征收关税，我仍然应该……

To keep the conversation moving, let me segue into a point that came up that was really important: tariffs. And the point is, even though the internet was open, the US won a lot of the internet. A lot of US companies won the internet, and they won that because we got there the firstest with the mostest, as they say in the military. And that matters because a lot of technology businesses have scale economies and network effects underneath, even basic brand-based network effects. If you go back to the late 90s, early 2000s, very few people would have predicted that we would have ended up with Amazon basically owning all of e-commerce. You would have thought it would have been perfect competition and very spread out. And that applies to how we end up with Uber as basically one taxi service, or we end up with Airbnb, Meta. It's just network effects, network effects, network effects rule the world. But when it comes to tariffs and when it comes to trade, we act like network effects don't exist. The classic Ricardian comparative advantage dogma says that you should produce what you're best at, I produce what I'm best at, and we trade. And then even if you want to charge me more for it, if you want to impose tariffs for me to ship to you, I should still...

保持低关税，因为这样我更有利，你只是便宜卖给我东西，很好。或者如果你想补贴你的人，那也很好，你还是在便宜卖给我东西。但问题是，这不是大多数现代企业的运作方式。大多数现代企业都有网络效应。举个简单的思想实验：假设有两个国家，我是中国，你是美国。我一开始补贴所有具有网络效应的公司和行业。所以我补贴 TikTok，我禁止你的社交媒体，但推广我的。我补贴我的半导体，这些半导体在某些类别中往往赢家通吃，或者我补贴我的无人机，然后比亚迪、自动驾驶等等。当我赢了，我就拥有整个市场，我可以提高价格。如果你试图创办一个竞争对手，那就太晚了。我有网络效应，或者如果我有规模经济，我可以把价格降到零，把你挤出市场，没有人会理智地投资，然后我再把价格涨回来。所以你必须明白，某些行业有滞后效应，或者有网络效应，或者有规模经济，这些都是有趣的行业，都是高利润的业务。在这些行业中，如果有人补贴或对你加征关税来保护他们的行业并让其发展，你必须采取行动。你不能完全退缩。Jamanth，你对关税和网络效应有什么看法？我们确实希望供应链有冗余，所以这里有一些例外。你对这将如何发展有什么想法？因为特朗普每 48 小时就提关税，但似乎没有一个落地，所以我不知道。我仍然坚持我的 72 小时特朗普规则：不管他说什么，等 72 小时，然后看看是否真的发生。你对所有这些关税和关税言论持什么立场？

Keep tariffs down because I'm better off you're just selling me stuff cheaply great or if you want to subsidize your guys great you're selling me stuff cheaply. The problem is that is not how most modern businesses work. Most modern businesses have network effects. As a simple thought experiment, suppose that we have two countries right I'm China you're the US. I start out by subsidizing all of my companies and industries that have network effects. So I'll subsidize TikTok, I'll ban your social media but I'll push mine. I will subsidize my semiconductors which tend to have winner-take-all in certain categories, or I'll subsidize my drones, and then eventually BYD, self-driving, whatever. And then when I win, I own the whole market and I can raise prices. And if you try to start up a competitor, then it's too late. I've got network effects, or if I've got scale economies, I can lower my price to zero, crash you out of business, no one in their right mind will invest, and I'll raise prices right back up. So you have to understand that certain industries have hysteresis or they have network effects or they have economies of scale, and these are all the interesting ones, these are all the high margin businesses. So in those, if somebody is subsidizing or they're raising tariffs against you to protect your industries and let them develop, you do have to do something. You can't just completely back down. What are your thoughts, Jamanth, about tariffs and network effects? It does seem like we do want to have redundancy in supply chain, so there are some exceptions here. Any thoughts on how this might play out? Because yeah, Trump brings up tariffs every 48 hours and then it doesn't seem like any of them land, so I don't know. I'm still on my 72-hour Trump rule, which is whatever he says, wait 72 hours and then maybe see if it actually comes to pass. Where do you stand on all these tariffs and tariff talk?

嗯，我认为关税会是一个问题。它们会来吗？绝对会。具体幅度我不知道。我认为你可以通过预算的立法计划来判断它会有多极端。现在有两条路径。路径一，我认为可能性稍大一点，是参议院通过一个精简的计划，只涉及边境安全和军费开支，然后预算问题再推迟三到四个月。路径二是这个正在众议院推进的“一大美丽法案”，他们提议削减数万亿美元。在这种模式下，你需要以某种方式增加收入，尤其是如果你要减税，唯一的方法可能就是通过关税，或者说一种方法是通过关税。我的诚实意见，Jason，是我们处在一个非常复杂的时刻。我认为参议院的计划实际上更可能也更好，原因是特朗普最好先获得接下来 60 到 90 天的数据。我的意思是，我们真的陷入了困境。我们有持续的通货膨胀，美联储失灵，他们完全失职，而耶伦和拜登做的事情，事后看来极其危险，是他们发行了太多短期债券，总计我们需要在未来 6 到 9 个月内融资 1 万亿美元。所以利率可能会达到 5%、5.25%、5.5%。我认为这非常糟糕，同时还有通胀，同时违约率在上升。所以我认为关税很可能会发生，但我认为如果特朗普有时间看看实际经济状况会怎样，那他将拥有最大的灵活性，而经济状况在三、四、五个月后会更加清晰。所以我几乎认为这个“一大美丽法案”实际上适得其反，因为我不确定我们是否会有所有必要的数据来把它做对。

Well, I think the tariffs will be a thing. Are they coming? Absolutely. The quantum of them I don't know. And I think that the way that you can figure out how extreme it will be, it'll be based on what the legislative plan is for the budget. So there are two paths right now. Path one, which I think is a little bit more likely, is that they're going to pass a slimmed-down plan in the Senate just on border security and military spending, and then they'll kick the can down the road for probably another three or four months on the budget. Path two is this one big beautiful bill that's working its way through the House, and there they're proposing trillions of dollars of cuts. In that mode, you're going to need to raise revenue somehow, and especially if you're giving away tax breaks, the only way to do that is probably through tariffs, or one way to do it is through tariffs. My honest opinion, Jason, is that I think we're in a very complicated moment. I think the Senate plan is actually on the margins more likely and better, and the reason is because I think that Trump is better off getting the next 60 to 90 days of data. I mean, we're in a real pickle here. We have persistent inflation, we have a broken Fed, they are totally asleep at the switch, and the thing that Yellen and Biden did, which in hindsight now was extremely dangerous, is they issued so much short-term paper that in totality we have $1 trillion that we need to finance in the next 6 to 9 months. So it could be the case that we have rates that are like five, five and a quarter, five and a half percent. I think that that's extremely bad at the same time as inflation, at the same time as delinquencies are ticking up. So I think tariffs are probably going to happen, but I think that Trump will have the most flexibility if he has time to see what the actual economic conditions will be, which will be more clear in three, four, five months. And so I almost think this big beautiful bill is actually counterproductive because I'm not sure we're gonna have all the data we need to get it right.

Fred，你对这些关税有什么看法？你一直参与全球市场，尤其是在农产品、小麦、玉米等方面。你认为这里的动态会是什么？还是说这只是虚张声势，是特朗普的一个工具？

Fred, any thoughts on these tariffs? You've been involved in the global marketplace, especially when it comes to produce and wheat and all this corn and everything. What do you think the dynamic here is going to be? Or is it saber rattling and a tool for Trump?

美国农产品出口的最大买家是中国。中国的农产品出口是许多州的主要收入来源、主要收入来源和经济的重要组成部分。因此，就像在特朗普第一个总统任期内一样，很可能会有非常大的转移支付给农民，因为中国很可能会对进口征收关税或完全停止进口采购，这正是第一个总统任期内发生的事情。当他们这样做时，我相信联邦政府向农民提供了超过 200 亿美元的转移支付。这不是一个小数目，也不是一个可以忽略的经济影响，因为随后整个农业经济会产生连锁反应。所以我认为我听到人们谈论的一个关键点是，随着美国最大的农业客户消失，需要采取什么行动来支持农业经济。

The biggest buyer of US agricultural exports is China. Chinese agricultural exports are a major revenue source, major income source, and a major part of the economy for a large number of states. And so there will be, as there was in the first Trump presidency, very likely very large transfer payments made to farmers because China is very likely going to tariff imports or stop making import purchases altogether, which is what happened during the first presidency. When they did that, the federal government, I believe, had transfer payments of north of $20 billion to farmers. This is a not negligible sum, and it's a not negligible economic effect because there's then a rippling effect throughout the agricultural economy. So I think that's one key thing that I've heard folks talk about is the activity that's going to be needed to support the farm economy as the US's biggest agricultural customer disappears.

在 20 世纪初，我们没有所得税，联邦收入几乎完全依赖关税。当关税被削减时，人们预期联邦政府收入会下降，但实际发生的是贸易量上升了。所以较低的关税实际上增加了贸易，扩大了经济规模。这就是许多经济学家的依据：如果我们征收这些关税，实际上会收缩经济，导致贸易减少。但抵消效应在经济学中尚未被检验，即如果我们在实施关税的同时，减少所得税和公司税，通过减税增加资本流动，会发生什么。所以这是一个宏大的经济实验，我认为随着这一切的推进，我们会学到很多。我确实认为最终这些国家会在一定程度上屈服，我们会达成某种谈判解决方案，希望不会对依赖贸易的经济、人民和就业产生太短期的影响。经济感觉处于非常不稳定的状态。对资产持有者来说显然如此，而且上一届政府把它留在了糟糕的状态，我们因为新冠疫情关闭了整个国家一年。那笔账单已经到期，反映在通货膨胀中。我认为关于关税还有几点。首先，这不仅仅是钱的问题，还关乎确保我们有一个拥有好工作的功能性中产阶级。因为如果是一个没有关税的世界，也许所有收益都流向上层阶级和下层阶级，那么当普通人处于这两个极端之一时，你无法拥有一个运作良好的民主制度。所以我认为这是一个问题。另一个是战略产业。如果你看看今天，可能……

In the early 20th century, we didn't have an income tax and the federal revenue was almost entirely dependent on tariffs. When tariffs were cut, there was an expectation that there would be a decline in federal government revenue, but what actually happened is volume went up. So lower tariffs actually increased trade, increased the size of the economies. This is where a lot of economists take their basis in: hey guys, if we do these tariffs, it's actually going to shrink the economy, it's going to cause a reduction in trade. The counterbalancing effect is one that has not been tested in economics, right, which is what's going to happen if simultaneously we reduce the income tax and reduce the corporate income tax and basically increased capital flows through reduced taxation while doing the tariff implementation at the same time. So it's a grand economic experiment, and I think we'll learn a lot about what's going to happen here as this all moves forward. I do think ultimately these countries are going to capitulate to some degree and we're going to end up with some negotiated settlement that's going to hopefully not be too short-term impactful on the economies and the people and the jobs that are dependent on trade. Economy feels like it's in a very precarious place. It does to asset holders obviously, and obviously they've left it in a bad place in the last Administration and we shut down the entire country for a year over COVID. The bill for that has come due and that's reflected in inflation. I think there are a couple other points on tariffs. First is, it's not just about money, it's also about making sure we have a functional middle class with good jobs. Because if you have a non-tariff world, maybe all the gains go to the upper class and an underclass, and then you can't have a functioning democracy when the average person is on one of those two extremes. So I think that's one issue. Another is strategic industries. If you look at it today, probably the...

全球最大的防务承包商是大疆。他们垄断了无人机市场。即使在乌克兰，双方都在从大疆获取无人机零件，只是通过不同的供应链。但乌克兰和俄罗斯的无人机，绝大多数都来自中国，来自大疆。而我们没有这个产业。如果我们现在发生热战，而美国国内没有良好的无人机供应链，我们很可能会输，因为那些东西就是自主子弹。这就是所有战争的特征。我们在买 F-35，而中国人在大规模制造无人机蜂群。所以我们确实需要重新本土化那些关键供应链。什么是无人机供应链？它不只是一个叫无人机的东西，而是电机、半导体、光学、激光，方方面面。所以我认为至少有一些其他好的理由来重新本土化这些产业。我们需要它们。美国非常幸运，因为它几乎是自给自足的。我们拥有所有资源，所有供应。我们可以凭借所有能源成为所有人的上游。我们进口任何能源都是我们自己的选择，而不是因为我们根本上缺乏能源。因为结合所有石油资源、天然气、水力压裂，再加上我们在核裂变和小型反应堆方面所做的工作，我们绝对应该实现能源独立。我们应该全面领先。我们应该有巨大的盈余。嘿，如果你担心几百万 DoorDash 和 Uber 司机因自动化而失业，嘿，会有工厂来制造我们需要的无人机零件。所以我认为人们有很多机会。不同类型的工作之间是有区别的。那些工作是更好的工作——大规模物理制造我们为国家安全和创新所需的困难东西。那些工作比文书工作、为别人写文章，甚至开车要好。

The largest defense contractor in the world is DJI. They got all the drones. Even in Ukraine, both sides are getting all their drone parts from DJI. Now they're getting it through different supply chains and so on, but Ukrainian drones and Russian drones, the vast majority of them are coming through China, through DJI. And we don't have that industry. If we have a kinetic conflict right now and we don't have a good drone supply chain internally in the US, we're probably going to lose because those things are autonomous bullets. That's the feature of all warfare. We're buying F-35s, and the Chinese are building swarms of drones at scale. So we do have to re-on-shore those critical supply chains. And what is a drone supply chain? It's not just a thing called a drone; it's motors, semiconductors, optics, lasers, just everything across the board. So I think there are other good arguments for at least reshoring some of these industries. We need them. And the United States is very lucky in that it's very autarkic. We have all the resources, we have all the supplies. We can be upstream of everybody with all the energy. To the extent we're importing any energy, that is a choice we made, not because fundamentally we lack the energy. Because between all the oil resources, natural gas, fracking, combined with all the work we've done in nuclear fission and small reactors, we should absolutely be energy independent. We should be running the table on it. We should have a massive surplus. And hey, if you're worried about a couple million DoorDash Uber drivers losing their jobs to automation, hey, there's going to be factories to build these parts for these drones that we're going to need. So there's a lot of opportunity, I guess, for people. And there is a difference between different kinds of jobs. Those kinds of jobs are better jobs—building difficult things at scale physically that we need for both national security and for innovation. Those are better jobs than paperwork, writing essays for other people to read, or even driving cars.

好了，听着，我想再讲两个故事。我们有一个非常有趣的版权故事，我想提一下。汤森路透刚刚赢得了美国首例重大 AI 版权案，合理使用在这项判决中发挥了重要作用。这对美国的 AI 公司有巨大影响。显然，OpenAI 和《纽约时报》、Getty Images 对 Stability——我们谈过这些，但已经有一段时间了，因为法律系统需要时间，而且这些案件非常复杂，正如我们讨论过的。汤森路透拥有 Westlaw，如果你不知道的话。它有点像 LexisNexis，是律师用来查找案例等的法律数据库之一。他们有一个付费产品，提供法律判决的摘要和分析。早在 2020 年，也就是 ChatGPT 问世前两年，路透社起诉了一家名为 Ross 的法律研究竞争对手，指控其侵犯版权。Ross 创建了一个 AI 驱动的法律搜索引擎——听起来很棒——但 Ross 曾询问 Westlaw 是否愿意付费许可其内容用于训练。Westlaw 拒绝了。这件事来回折腾，然后 Ross 与一家名为 LegalEase 的公司签署了类似协议。问题是 LegalEase 的数据库只是从一堆 Westlaw 答案中复制粘贴过来的。所以路透社/Westlaw 在 2020 年起诉 Ross，指控该公司对 LegalEase 的直接侵权承担替代责任。非常重要的一点。无论如何，法官最初在合理使用问题上支持 Ross。本周，法官推翻了这一裁决，认定 Ross 负有责任，并指出经过进一步审查，合理使用不适用于此案。这是第一个重大胜利。我们对此进行过辩论，所以这里有一段片段：你知道，你是在 All-In 播客上最先听到的。我想说的是，当你审视合理使用原则时，我对此有丰富的经验。你知道第四个因素测试——我相信你很熟悉——是使用对潜在市场和工作价值的影响。如果你看看那些开始出现的诉讼，Getty 有权基于其图像制作衍生作品。我想我们都会同意：Stable Diffusion，当他们使用开放网络时，这不能成为使用开放网络爬虫来避免从原始所有者那里获得许可的借口。仅仅因为你在技术上可以做到，并不意味着你被允许这样做。事实上，提供这些的开放网络项目明确表示：我们不授予你使用它的权利；你必须去阅读每个网站上的版权法。除此之外，如果有人窃取他人的版权并将其放在开放网络上——这每天都在发生——如果你正在构建这样的衍生作品，你仍然需要去获取许可。所以，我拿了一个俄罗斯的网站，它做了大量版权侵权，然后我将其索引用于我的训练模型，这不是借口。所以我认为这将导致……自由？你能朝我脸上开枪，然后告诉我什么时候结束吗？哦，太好了。我也有同感。同感。现在确实如此。我知道，我也是。是的，好的，好的片段。我们继续吧。好吧，既然这些家伙不关心版权持有人，你怎么看……？

All right, listen, I want to get to two more stories here. We have a really interesting copyright story that I wanted to touch on. Thomson Reuters just won the first major US AI copyright case, and fair use played a major role in this decision. This has huge implications for AI companies here in the United States. Obviously, OpenAI and the New York Times, Getty Images versus Stability—we've talked about these, but it's been a little while because the legal system takes a little bit of time, and these are very complicated cases as we've talked about. Thomson Reuters owns Westlaw, if you don't know. It's kind of like LexisNexis, it's one of the legal databases out there that lawyers use to find cases, etc. And they have a paid product with summaries and analysis of legal decisions. Back in 2020, this is two years before ChatGPT, Reuters sued a legal research competitor called Ross for copyright infringement. Ross had created an AI-powered legal search engine—sounds great—but Ross had asked Westlaw if they would pay a license to its content for training. Westlaw said no. This all went back and forth, and then Ross signed a similar deal with a company called LegalEase. The problem is LegalEase's database was just copied and pasted from a bunch of Westlaw answers. So Reuters/Westlaw sued Ross in 2020, accusing the company of being vicariously liable for LegalEase's direct infringement. Super important point. Anyway, the judge originally favored Ross on fair use. This week, the judge reversed this ruling and found Ross liable, noting that after further review, fair use does not apply in this case. This is the first major win. And we debated this, so here's a clip: you know, you heard it here first on the All-In Pod. What I would say is, you know, when you look at that fair use doctrine, I've got a lot of experience with it. You know the fourth factor test—I'm sure you're well aware of this—is the effect of the use on the potential market and the value of the work. If you look at the lawsuits that are starting to emerge, it is Getty's right to then make derivative products based on their images. I think we would all agree: Stable Diffusion, when they use the open web, that is no excuse to use an open web crawler to avoid getting a license from the original owner of that. Just because you can technically do it doesn't mean you're allowed to do it. In fact, the open web projects that provide these say explicitly: we do not give you the right to use this; you have to then go read the copyright laws on each of those websites. And on top of that, if somebody were to steal the copyrights of other people and put it on the open web—which is happening all day long—you still, if you're building a derivative work like this, you still need to go get it. So it's no excuse that I took some site in Russia that did a bunch of copyright violation and then I indexed them for my training model. So I think this is going to result in... free? Can you shoot me in the face and let me know when this is over? Oh, great. I feel the same way. Same way. Now exactly. I know, me too. Yeah, okay, good, good segment. Let's move on. Well, since these guys don't give a damn about copyright holders, what do you think about...?

我很高兴你在这里，Naval，来真正讨论这些话题。这两个家伙，我要冒个险。我要冒更大的险，说我基本同意你的看法。我认为爬取开放网络、吸走所有数据、为许多用例提供直接替代品，这有点过分，因为你知道，现在你以 AI 模型开始和结束。它甚至不像谷歌那样链接出去，然后你仅仅出于安全原因封闭模型。我认为如果你在开放网络上训练，你的模型应该是开源的。是的，绝对。那会是一件好事。我有一个预测。我认为这一切最终会像 Napster 或 Spotify 案一样。对于不知道的人来说，Spotify 支付，我认为，每美元 65 美分给内容的原始承销商，音乐行业，他们找到了做生意的办法，而 Napster 成了路毙动物。我认为存在非零概率——可能是 5%或 10%——OpenAI 会输掉《纽约时报》的诉讼，而且会输得很惨，还会有禁令。我认为和解可能是这些语言模型，尤其是封闭模型，将不得不支付其收入的一定比例——一半、三分之二——通过协商和解给内容持有人。这可能会使内容行业获得巨大的、巨大的提升和巨大的复兴。我认为问题在于另一边有一个例子，那就是有一家公司为 Oracle 提供技术支持，是一家第三方公司，Oracle 曾多次试图用版权侵权作为部分理由起诉他们，让他们消失……

I'm so glad you're here, Naval, to actually talk about the topics. These two other guys, I'm going to go out on a limb. I'm going to go out on an even thinner limb and say I largely agree with you. I think it's a bit rich to crawl the open web, hoover up all the data, offer direct substitution for a lot of use cases because, you know, now you start and end with the AI model. It's not even like you link out like Google did, and then you just close off the models for safety reasons. I think if you trained on the open web, your model should be open source. Yeah, absolutely. That would be a fine thing. I have a prediction here. I think this is all going to wind up like the Napster or Spotify case. For people who don't know, Spotify pays, I think, 65 cents on the dollar to the original underwriters of that content, the music industry, and they figured out a way to make a business, and Napster is roadkill. I think that there is a nonzero chance—like it might be 5 or 10%—that OpenAI is going to lose the New York Times lawsuit, and they're going to lose it hard, and there are going to be injunctions. And I think the settlement might be that these language models, especially the closed ones, are going to have to pay some percentage in a negotiated settlement of their revenue—half, two-thirds—to the content holders. And this could make the content industry have a massive, massive uplift and a massive resurgence. I think the problem is there's an example on the other side of this, which is that there's a company that provides technical support for Oracle, a third-party company, and Oracle has tried multiple times to sue them into oblivion using copyright infringement as part of the...

我长期关注一只股票，公司名叫 Riny Street。别问我为什么，它一直在我的雷达上。他们输了一场大官司，Oracle 赢了，然后上诉法院又全部撤销了。我为什么提这个？我认为法律界对这些模型的工作原理一无所知，因为你可以找到一个案子判这边赢，另一个案子判那边赢。我觉得任何提起这类诉讼的人都应该把一段视频作为必读材料——Karpathy 刚发布了一个关于 LLM 的深度解析，从头讲解 ChatGPT。在 YouTube 上，三个小时，非常精彩。看完那段视频很难不得出和你们一样的结论。我就说这么多。

Justification and it's been a PA over the stock for a long time. The company's name is Riny Street. Don't ask me why, it's on my radar but I've been looking at it. They lost this huge lawsuit Oracle won, and then it went to appellate court and then it was all vacated. Why am I bringing this up? I think that the legal community has absolutely no idea how these models work, because you can find one case that goes one way and one case that goes the other. What I would say should become standard reading for anybody bringing in any of these lawsuits: there's an incredible video that Karpathy just dropped, that Andrej just dropped, where he does like this deep dive into LLMs and he explains ChatGPT from the ground up. It's on YouTube, it's three hours, it's excellent, and it's very difficult to watch that and not get to the same conclusion that you guys did. I'll just leave it at that.

我倾向于同意这一点。还有一段 Ilya Sutskever 的经典视频，他应该是 OpenAI 的创始首席科学家或 CTO，他谈到这些大语言模型本质上是极端的压缩器。他完全用压缩能力来建模它们，而且是有损压缩。

I tend to agree with this. There's also a good old video by Ilya Sutskever where he was, I believe, the founding Chief Scientist or CTO of OpenAI, and he talks about how these large language models are basically extreme compressors. He models them entirely as their ability to compress, and they're lossy compression.

没错，有损压缩。

Exactly, lossy compression.

没错。当年 Google 因为合理使用被起诉，但他们绕过这个论点的方式是始终链接回你，显示小片段，把流量导给你。这是有损压缩。我现在站在你这边。我讨厌这么说，Jason，我同意你。你是对的，你是对的。

Exactly. And Google got sued for fair use back in the day, but the way they managed to get past the argument was they were always linking back to you, they showed tiny snippets, they sent you the traffic. This is lossy compression. It is absolutely... I'm now on your side. I hate to say this, Jason, I agree with you. You were right, you were right.

这就是我想听到的。这么多年，这就是我想要的。看到那些视频时我直摇头，因为我想，天哪，Jason 是对的。Jason 是对的。哦我的天。

That's all I wanted to hear. All these years, that's all I wanted. Shaking my head when I saw those videos because I was like, oh man, Jason was right. Jason was right. Oh my God.

不，我只是……我经历过太多次了。我认为……你知道，鲁珀特·默多克说过我们应该对 Google 保持强硬，不允许他们在没有许可的情况下索引我们的内容，但 Google 成功应对了，没有让他得逞。我认为现在发生的是，《纽约时报》记得他们当年在片段和 OneBox 上输给了 Google，而且无法让那个精灵回到瓶子里。我认为《纽约时报》意识到这是他们的发财机会。我认为《纽约时报》从 LLM 许可中赚的钱会比从广告或订阅中赚的更多。最终这几乎会重塑商业模式。我认为《纽约时报》的内容对 LLM 来说毫无价值，但那是另一回事。

No, I just... I've been through this so many times. I think this is... you know, Rupert Murdoch said we should hold the line with Google and not allow them to index our content without a license, and Google navigated it successfully and they were able to not get him to stop. I think what's happened now is that the New York Times remembers that they all remember losing their content and these snippets and the one box to Google, and they couldn't get that genie back in the bottle. I think the New York Times realizes this is their payday. I think the New York Times will make more money from licenses from LLMs than they will make from advertising or subscriptions. Eventually this will renew the model almost. I think the New York Times content is worthless to an LLM, but that's a different story.

我认为实际价值……政治原因，随便吧。但我可以作为一个用户告诉你，我喜欢 The Wirecutter。我想你认识 Brian 和 Wirecutter 的所有人。那很公平。是啊，Wirecutter，多好的产品。我以前付费订阅《纽约时报》，现在不付了。主要原因是以前我去 Wirecutter，他们让我买什么我就买什么。现在我去 ChatGPT，我付钱给它，它根据 Wirecutter 告诉我买什么。所以……我已经在付钱了，所以我不再付给《纽约时报》了。

I think the actual value... political reason, whatever. But I can tell you as a user, I loved The Wirecutter. I think you knew Brian and everybody over at The Wirecutter. That was like fair enough. Yeah, Wirecutter, what a great product. I used to pay for the New York Times. I no longer pay for the New York Times. My main reason was I would go to The Wirecutter and I would just buy whatever they told me to buy. Now I go to ChatGPT, which I pay for, it tells me what to buy based on The Wirecutter. So it's... and I'm already paying for it, so I stop paying for it.

我在哲学上不同意你们在这个话题上的所有废话。你们三个都错了，我告诉你们为什么。第一：如果信息在开放的互联网上，我认为它是可访问和可查看的，我把 LLM 或网络爬虫看作是一个正在阅读并能将信息存储在脑子里的人。如果它在开放的地方，如果它在付费墙后面，100%可以。如果它在受密码保护后面……等等，等等，David。那样的话，Google 爬虫能爬取整个网站并在 Google 上提供吗？为什么它们不能那样做？

I philosophically disagree with all of your nonsense on this topic. All three of you are wrong, and I'll tell you why. Number one: if information is out in the open internet, I believe it's accessible and it's viewable, and I view an LLM or a web crawler as basically being a human that's reading and can store information in its brain. If it's out there in the open, if it's behind a paywall, 100%. If it's behind some protected password... wait, wait, wait, David. In that case, can a Google crawler just crawl an entire site and serve it on Google? Why can't they do that?

合理使用是这样的：你不能复制，你不能重复内容，你不能拿内容然后重复它。这是目前法律的写法。但现在我有了一个工具，可以把内容和其他 50 个类似内容混在一起，我可以稍微改几个词，甚至翻译成另一种语言。那么界限在哪里？你知道音乐人 Girl Talk 吗？我们今天真该放一首他的歌。他很有音乐品味。他基本上从流行歌曲中取小样本，然后……他因为同样的问题被起诉了。还有一个人叫 White Panda，也有同样的问题。Ed Sheeran 也因为这个被起诉了。但像 Stack Overflow 和维基百科这样的整个网站现在基本上消失了，因为你可以把它们全部吞下，然后在 ChatGPT 里稍微改改就吐出来。所以我认为合理使用……改变多少才算是微小的改变，这正是问题所在。你改变了多少？是啊。所以这就是问题。这实际上归结为 AGI 问题：这些东西是真的智能，在学习，还是在压缩和重复？这就是问题。

Here's the fair use: the fair use is you cannot copy, you cannot repeat the content, you cannot take the content and repeat it. That is how the law is currently written. But now what I have is a tool that can remix it with 50 other pieces of similar content, and I can change the words slightly and maybe even translate into a different language. So where does it stop? Do you know the musical artist Girl Talk? We should have done a Girl Talk track here today. He's got musical taste, this guy. Oh good, here we go. He basically takes small samples of popular tracks and he made... and he got sued for the same problem. There was another guy named White Panda, I believe, had the same problem. Ed Sheeran got sued for this. Yeah. But their entire sites like Stack Overflow and Wikipedia have basically disappeared now because you can just swallow them all up and you can just spit it all back out in ChatGPT with slight changes. So I think that the fair... how much of a slight change is exactly the right question. How much are you changing? Yeah. So that's the question. And it actually boils down to the AGI question: are these things actually intelligent and are they learning, or are they compressing and regurgitating? That's the question.

我对人类也有同样的疑问。这就是为什么我提到 White Panda、Girl Talk 在音频领域，还有视觉艺术。总有艺术家，即使在古典音乐中……我不知道你们是不是古典音乐爱好者，但有一个演示说明一个作曲家如何从另一个作曲家学习，你实际上可以追踪音乐是站在前人的肩膀上。这在几乎所有艺术形式、几乎所有人类知识和媒体传播中都是如此。这很难搞清楚。

I wonder this about humans. And that's why I bring up the White Panda, the Girl Talk in audio, but also visual art. There were always artists, and even in classical music... I don't know if you guys are classical music people, but there was a demonstration of how one composer learned from the next, and that you can actually track the music as kind of standing on the shoulders of the prior. And the same is true in almost all art forms, in almost all human knowledge and media communication. It's very hard to figure that out.

嗯，完全正确。这很难……这很难搞清楚。这就是为什么我回到：这个问题只有两种稳定的解决方案，而且无论如何都会发生。如果我们不爬取，中国人会爬取。DeepSeek 证明了这一点。所以只有两种稳定的解决方案：要么你付钱给版权持有者，但我认为这行不通，因为中国会有人爬取然后直接转储权重，对吧？他们可以爬取、压缩并转储压缩后的权重。要么如果你爬取，就让它开放，至少为开源做点贡献。对吧？你爬取开放数据，回馈给开源，而那些不想被爬取的人，将不得不花大力气保护他们的数据。现在大家都知道要保护数据了。

Well, that's exactly right. That's the hard... it's very hard to figure that out. Which is why I come back to: there's only one of two stable solutions to this, and it's going to happen anyway. If we don't crawl it, the Chinese will crawl it. DeepSeek proved that. So there's only one of two stable solutions: either you pay the copyright holders, which I actually think doesn't work, and the reason is because someone in China will crawl it and they just dump the weights, right? So they can just crawl and compress and dump the compressed weights. Or if you crawl, make it open, at least contribute something back to open source. Right? You crawled open data, contributed back to open source, and the people who don't want to be crawled, they're going to have to go to a huge length to protect their data. Now everybody knows to protect the data.

是啊，事情正在发生。我身后架子上有一本 Harper Business 出版的书，我因为微软索引它，未来三年能拿到 2500 美元。所以他们出去授权这些东西，他们出价 2500 美元。所以你的书真的……我从 Harper 那里拿到 2500 美元，为期三年，让内容进入 LLM，具体是进入微软。你知道吗？我会签的。我决定这么做是因为我想树立一个先例。也许下次是 1 万美元，也许下次是 250 美元。我不在乎。我只是希望看到人们的内容得到尊重。而且我只是希望 Sam Altman 输掉这场官司。

Yeah, well, thing is happening here. I have a book out from Harper Business on the shelf behind me, and I'm getting $2,500 for the next three years for Microsoft indexing it. So they're going out and they're licensing this stuff, and they're going $2,500. So your book literally... I'm getting $2,500 for three years from Harper to go into an LLM, to go into Microsoft specifically. And you know what? I'm going to sign it. I decided because I just want to set the precedent. Maybe next time it's $10,000, maybe next time it's $250. I don't care. I just want to see people have their content respected. And I'm just hoping that Sam Altman loses this lawsuit.

他们还拿到了禁令。嘿，就因为他把 OpenAI 变成一个封闭的东西，简直是个滑头。我本人挺喜欢 Sam，但我认为他做的事是史上最滑头的操作，完全是为了个人利益。如果他——还有那些谎话，比如‘我没有股权，我只有医保’——不，兄弟，他做这事是为了爱。他怎么说来着？他做这事是为了——‘我为了人类利益而做’。我觉得他拿到了医保。为 OpenAI 辩护的话，他们确实需要筹集大量资金，也需要激励员工，但这不意味着他们需要接管整个公司。非营利部分可以继续保持非营利，获得大部分利益，并担任董事会。然后他可以有一个激励方案，员工也可以有。为什么他们不拿收入的一定比例？我不明白。现在必须用 400 亿美元买断，然后整个东西就消失在一个封闭系统里。这部分对我来说完全说不通。那叫骗局和欺诈。

And they get an injunction against it. Hey, well, just because he's such a weasel in terms of making OpenAI into a closed thing. I mean, I like Sam personally, but I think what he did was the super weasel move of all time for his own personal benefit. If he—and this whole lying like, 'Oh, I have no equity, I get healthcare'—no, bro, he does it for the love. What was the statement? He does it for the—'I do it for the benefit of humanity.' I think he got healthcare. In OpenAI's defense, they do need to raise a lot of money and they need to incentivize their employees, but that doesn't mean they need to take over the whole thing. The nonprofit portion can still stay the nonprofit portion, get the lion's share of the benefits, and be the board. Then he can have an incentive package, and employees can have an incentive package. Why don't they get a percentage of the revenue? I don't understand. It has to be bought out right now for $40 billion, and then the whole thing disappears into a closed system. That part makes no sense to me. That's called a shell game and a scam.

是的，我认为 Sam 和他的团队最好保持非营利部分不变，让一个真正独立的非营利董事会负责，然后为投资者和员工制定一个强有力的激励计划和筹款计划。所以我认为这是可行的。试图全部攫取似乎太过分了，尤其是它建立在 Google 的开放算法、网络上的开放数据以及 Elon 和其他人的非营利资助之上。

Yeah, I think Sam and his team would do better to leave the nonprofit part alone, leave an actual independent nonprofit board in charge, and then have a strong incentive plan and a strong fundraising plan for the investors and the employees. So I think this is workable. It's just trying to grab it all seems way off, especially when it was built on open algorithms from Google, open data from the web, and nonprofit funding from Elon and others.

我的意思是，我们刚刚在这里讨论出了一个多么好的提议。如果他们——他们每年赚 60 亿美元，只拿 10%，每年 6 亿美元，作为奖金呢？他们在亏损，Jason，所以他们必须——好吧，最终，不，但即使是股权——他们可以给建设它的人股权，但仍然可以把它控制在非营利组织手中。我就是不理解这种转换。我的意思是，有过一次董事会政变，对吧？董事会试图解雇他。现在是他小气，对吧？是的，他们会得到独立估值，但我们都知道这个把戏：你雇一个估值专家，他会说你想让他们说的话，然后他们打个勾。如果你要捕获所有未来价值的生命周期，或者建造价值更高的超级智能——这就是为什么 Elon 刚刚花了 1000 亿美元。没错。你说的是监管机构和法律界没有洞察力的事情，因为他们会看到一份公平意见，然后想，‘哦，它写着公平和意见，两个词并排，那一定是公平的’，他们不知道这一切是如何被操纵的。

I mean, what a great proposal we just workshopped here. What if they just—they make six billion a year, just take 10% of it, 600 million every year, and that goes into a bonus? They're losing money, Jason, so they have to—okay, eventually, no, but even equity—they could give equity to the people building it, but they could still leave it in the control of the nonprofit. I just don't understand this conversion. I mean, there was a board coup, right? The board tried to fire him. Now it's his pettiness, right? And yeah, they'll get an independent valuation, but we all know that game: you hire a valuation expert who's going to say what you want them to say, and they'll check a box. If you're going to capture the lifecycle of all future value or build superintelligence that's worth a lot more—that's why Elon just spent $100 billion. Exactly. You're saying the things that the regulators and the legal community have no insight into, because they'll see a fairness opinion and think, 'Oh, it says fairness and opinion, two words side by side, it must be fair,' and they don't know how all of this stuff is gamed.

是的，是的。我有关于 409A 的故事——没错，409A 被操纵了，这些公平意见也被操纵了。但现实是，我认为法律和司法界毫不知情。我的意思是，想象一下，如果你投资的一个创始人——完全虚构的情况——Naval 拿到了一份金额惊人的好条款清单，没有接受，把估值压到一百万以下，给自己一堆股份，然后三个月后接受了。我不知道那叫什么——证券欺诈。

Yeah, yeah. I've got stories about 409A that would—exactly, 409As are gamed, these fairness opinions are gamed. But the reality is, I don't think the legal and judicial community has any idea. I mean, imagine if a founder you invested in—just a total imaginary situation—Naval had a great term sheet at some incredible dollar amount, didn't take it, ran the valuation down to under a million, gave themselves a bunch of shares, and then took it three months later. I don't know what that would be called—securities fraud.

是的，我们来说说你的故事。我有个有趣的——Nick，我们会给你看照片——周一我和 Brian Johnson 吃了一顿有趣的晚餐，就是那个‘不要死’的家伙。他来了我家。他夜间勃起怎么样？我们谈到的是他每晚有三个小时的夜间勃起。哇，看看这个。顺便说一句，首先，我要告诉你我认为——我认为他是——等等，哪个在给他勃起？不，不，不，他测量他的夜间勃起。我认为给了他勃起，或者他——但他说当他开始的时候——顺便说一句，他说他开始时 43 岁，基本上临床肥胖，在这四年里变成了一个标本。他现在每晚有三个小时的夜间勃起。但这还不是有趣的事。顺便说一句，晚餐结束时，他的皮肤——我不确定，因为你在网上看到的照片，但他在现实生活中的皮肤像瓷娃娃一样。我和我妻子都觉得，我们从未见过这样的皮肤，而且非常柔软。等等，等等，哇，哇，哇——你怎么知道皮肤软？你知道，你用手拂过他的前臂什么的，晚上结束时拥抱一下。我告诉你，那家伙的皮肤——有弹性的皮肤，兄弟。这是我一生中摸过的最柔软的皮肤。不管怎样，这不是重点。这是一顿非常迷人的晚餐。他介绍了他的整个方案，但最后，我想是 Alter Networks 的 CEO Nash，他说，‘给我前三样东西。’前三样。在这前三样中，我要归结为第一样，它占了 80% 的 80%——全是关于睡眠。我正要猜睡眠。他介绍了他的夜间例行程序，令人难以置信，而且直截了当。非常简单。就像你如何放松一样。不管怎样，我试着简要解释一下放松过程。这么说吧，因为 Brian 上床睡觉要早得多，所以我们的正常时间——比如说 10 点——你首先需要在三四个小时前停止进食，对吧？我做到了——我 6:30 吃饭，所以我有大约三个小时。你在 9:30 或 10 点上床。你要处理自我对话，对吧？比如，活跃的大脑告诉你所有早上必须解决的事情。说出来，把它放好，说‘我早上会处理这个’，写在日记里。你做任何事都是为了把它放下。你不能看手机——手机必须放在另一个房间，或者你只能关机。然后读一本书，这样你实际上就投入了一些事情。他说他通常在躺下开始这个程序后三到四分钟内入睡。我试了一下——自从周一和他吃饭以来我一直在做。昨晚我在 50 分钟内睡着了。哈！对我来说最难的是把手机放下。我做不到。当然，当然。你呢，Naval？告诉我们你的放松过程。

Yeah, let's wrap on your story. I had an interesting—Nick, we'll show you the photo—I had an interesting dinner on Monday with Brian Johnson, the 'Don't Die' guy. He came over to my house. How's his erection doing overnight? What we talked about is he's got three hours a night of nighttime erections. Wow, look at this. By the way, first of all, I'll tell you I think that—I think that he's—wait, which one of those is giving him the erection? No, no, no, he measures his nighttime erection. I think has given him the erection, or he—but he said that when he started—so by the way, he said he was 43 when he started this thing, he was basically clinically obese, and in these four years has become a specimen. He now has three hours a night of nighttime erections. But that's not the interesting thing. At the end of this dinner, by the way, his skin—I was not sure because when you see the pictures online, but his skin in real life is like a porcelain doll's. Both my wife and I were like, we've never seen skin like this, and it's incredibly soft. Wait, wait, wait, whoa, whoa, whoa—how do you know skin is soft? You know, you brush your hand against his forearm or whatever, you give a hug at the end of the night. I'm telling you, the guy's skin is—had supple skin, bro. It's the softest skin I've ever touched in my life. Anyways, that's not the point. It was a really fascinating dinner. He walked through his whole protocol, but at the end of it, I think it was Nash, the CEO of Alter Networks, he was just like, 'Give me the top three things.' Top three. And of the top three things, what I'll boil it down to is the top one thing, which is like 80% of the 80%—it's all about sleep. I was about to guess sleep. And he walked through his nighttime routine, and it's incredible and it's straightforward. It's really simple. It's like how you do a wind-down. Anyways, I have tried to explain the wind-down briefly. Let's just say that because Brian goes to bed much earlier, so our normal time—let's just say you know 10—you need to first of all stop eating three or four hours before, right? And I do that—I eat at 6:30, so I have about three hours. You're in bed by 9:30 or 10. You deal with the self-talk, right? Like, okay, here's the active mind telling you all the things you have to fix in the morning. Talk it out, put it in its place, say 'I'm gonna deal with this in the morning,' write it down in a journal. You're saying whatever you do so that you put it away. You cannot be on your phone—that's got to be in a different room, or you just got to be able to shut it down. And then read a book so that you're actually just engaged in something. And he said that he typically falls asleep within three to four minutes of getting into bed and starting this. What I tried it—so I've been doing it since I had dinner with him on Monday. Last night I fell asleep within 50 minutes. Hah! The hardest part for me is to put the phone away. I can't do it. Of course, of course. What about you, Naval? Tell us your wind-down.

哦，是的。我其实很了解 Brian。我开玩笑说我娶了女性版的 Brian Johnson，因为我妻子有他的一些习惯，但她是天然版——没有补品——而且她很极端。我想当 Brian 看到我的 Eight Sleep 睡眠分数时，他震惊了。他就像，‘你会死的。’他说，‘你真的会死。’嗯，你得了 70、80？不，太糟糕了，太可怕了。但是——告诉我们你的数字是多少。你的数字是多少？大概在 30 多、40 多。但你知道，是的，但也是因为我睡得不多——我每晚只睡几个小时——而且我在床上也动来动去，等等。但没关系。我入睡从来没问题。但我要说，Brian 的护肤程序很棒，他的饮食——

Oh, yeah. I know Brian pretty well, actually. And I joke that I'm married to the female Brian Johnson because my wife has some of his routines, but she's like the natural version—no supplements—and she's intense. And I think when Brian saw my sleep score from my Eight Sleep, he was shocked. He was just like, 'You're going to die.' He's like, 'You're literally going to die.' Well, you got 70, 80? No, it's terrible, it's awful. But it's—tell us what your number is. What's your number? It's like in the 30s, 40s. But you know, yeah, but it's also because I don't sleep much—I only sleep a few hours a night—and I also move around a lot in the bed, and so on. But it's fine. I never have trouble falling asleep. But I would say that Brian's skincare routine is amazing, his diet is—

你们都有冥想？固定时间？不，不。我想睡就睡，通常凌晨一两点。

You all have a meditation? You have a set time? No, no. I sleep whenever I feel like, usually around 1 in the morning, two in the morning.

天哪，我 10 点就上床了。对，我需要睡觉。我是个夜猫子。

God damn, I'm in bed by 10. Yeah, I need to sleep. I'm an owl.

但如果你想入睡，我发现的小窍门是：每个人都试过某种冥想套路。就坐在床上冥想，你的大脑会非常讨厌冥想，以至于如果你强迫它在冥想和睡觉之间做选择，你每次都会睡着。好吧，如果你没睡着，那你就冥想，这也很好。所以我喜欢冥想，做身体扫描。

But if you want to fall asleep, the hack I found is everybody has tried some kind of a meditation routine. Just sit in bed and meditate, and your mind will hate meditation so much that if you force it to choose between the fork of meditation and sleeping, you will fall asleep every time. Well, okay, so after, if you don't fall asleep, you'll end up meditating, which is great too. So just, I like the meditation, do the body scan.

这个故事的结尾是：我一个朋友从阿联酋来看我，他周二在这，我跟他讲和布莱恩吃饭的事，他给我讲了个故事，因为他和 UFC 拳手卡比布是朋友。他说，卡比布在家什么都吃，油炸食品、披萨，什么都行，但他坚持训练。我的朋友阿达拉问，你怎么能做到这样，而且不影响身体？卡比布说，我从小就知道，我早上训练后睡 3 小时，晚上睡 10 小时，我从 12、13 岁就开始这样了。

The coda to this story was a friend, a friend of mine came to see me from the UAE, and he was here on Tuesday, and I was telling him about the dinner with Brian, and he told me the story, 'cause he's friends with Khabib, the UFC fighter. And he says, you know, when Khabib goes to his house, he eats anything and everything, fried food, pizza, whatever, but he trains consistently. And my friend Adala says, how are you able to do that and how does it not affect your physiology? He goes, I've learned since I was a kid, I sleep 3 hours after I train in the morning, and I sleep 10 hours at night, and I've done it since I was like 12 or 13 years old.

那睡眠量很大。

That's a lot of sleep.

睡眠量很大。对我来说，直接关联是如果我做了认知上的事情，比如很重的对话之类的。所以晚上不要有沉重的对话，不要有存在主义的对话。然后如果我去负重行军，我在牧场穿上 35 磅的负重背心走路。你晚上睡前做这个？不，不，不。我白天任何时候做都行，通常早上或下午。但负重 35 磅走 1 到 2 英里，不管怎样，它让我全身疲惫，这样我躺下时……

It's a lot of sleep. I, you know, the direct correlation for me is if I do something cognitively, like you know, big heavy duty conversations or whatever. So no heavy conversations at the end of the night, no existential conversations at night. And then if I go rucking, I have, you know, on the ranch I put on a 35lb weight vest, I walk. You do that at night before you go to bed? No, no, no. If I do it anytime during the day, I typically do it in the morning or the afternoon. But the 1 to 2 mile ruck with the 35 lbs, whatever it is, it just tires my whole body out, so that when I do lay down...

这就是你不准备播客的原因吗？你知道，这个播客是世界前十的播客。Chim，你觉得这是偶然吗？Freeberg，你的睡眠习惯是什么？你直接上床？你喜欢热水澡？我给 J Cali 发一张我脚的照片。我等 J 完事。

Is that why you don't prepare for the podcast? You know, I mean, this podcast is the top 10 podcast in the world. Chim, do you think it's an accident? Freeberg, what's your sleep routine? Can you just go to bed? You just like warm bath? And I send J Cali a picture of my feet. I'll wait till J's done.

我确实会泡个舒服的热水澡，你每晚都泡？热水澡？是的，我每晚都泡，还点蜡烛。你是在睡前泡吗？对，我通常哄孩子睡着后泡，然后开始放松准备睡觉。我有时会看电视，但我有个问题，就是睡前看手机太久，这是个错误。

I do take a nice warm bath, but you do it every night? A warm bath? I do, yeah. I do a warm bath every night with candles too. And do you do it right before you go to bed? Yeah, I usually do it after I put the kids down, and I'll basically start to wind down for bed. I do watch TV sometimes, but I do have the problem and the mistake of looking at my phone probably for too long before I turn the lights off.

那你睡觉时间固定吗？通常 11 点到午夜，然后 6:30 起床。老兄，我需要 8 小时，不然就一团糟。我尽量睡 8 小时，但通常只睡 6 到 7 小时。我试图在 11 点到凌晨 1 点之间上床。

So do you have a consistent time where you go to bed or no? Usually 11:00 to midnight, and then up at 6:30. Man, I need eight hours, otherwise I'm a mess. I go to, I'm trying to get eight. I hit between six and seven consistently. I try to go to bed that 11:00 to 1:00 a.m.

我的问题是，如果有工作要做，我就会坐到电脑或笔记本前，晚上例行事务之后一旦开始就停不下来。然后突然就凌晨三点了，我心想，'哦不，我刚才干了什么？' 可我还得六点半起床。所以我确实会这样。昨晚对我来说不寻常，但还挺好笑的。我本来想，'哦，我应该早点睡，' 结果却和孩子们一起吃了冰淇淋。等等，你说的是什么牌子？你换了另一个牌子？我想知道那个牌子。

My problem is if I have work to do, I'll get on the computer or my laptop, and then when I start that after my evening routine, I can't stop. Then all of a sudden it's like 3 in the morning and I'm like, 'Oh no, what did I just do?' And then I still have to get up at 6:30. So that does happen to me. Last night was unusual for me, but it was kind of funny. I thought, 'Oh, I should go to bed early,' but I ended up eating ice cream with the kids late. Wait, what was the brand you said? You went for another brand? I want to know the brand.

我想是 Van Leeuwen 之类的。纽约的，假日曲奇奶油味。天哪，太好吃了。

I think it's Van Leeuwen or something like that. New York and the holiday cookies and cream. Oh my God, so good.

是啊，太好吃了。所以吃完之后，我心想，'哦，可能吃太多了睡不着，最好锻炼一下。' 于是我就做了个壶铃锻炼。

Yeah, it's so good. So after I polished that off, I was like, 'Oh, I probably ate too much to go to bed, so I better work out.' So I did a kettlebell workout.

你听起来像 Mo。你说什么？我这里有八个壶铃，就在我旁边。这叫免费锻炼。

You sound like Mo. What did you say? I have eight kettlebells right here, right next to me. This is called working out free.

然后在我做壶铃手提箱式行走的时候，我还在和一个创业朋友发短信。你可以想象我的锻炼强度有多大。他在新加坡，所以对我来说是半夜，对他来说还早。到了该睡觉的时候，我心想，'好了，现在得去睡了。怎么才能睡着？我身体亢奋，胃里有食物，大脑也亢奋，而且明天还有 All-In 播客。现在几点了？凌晨一点半。最好去睡吧。' 于是我放了一个小灵性视频来平静下来，然后上床，心想，'我绝对睡不着。' 然后我开始冥想，五分钟后我就睡着了。

And while I'm doing my kettlebell suitcase carry, I was texting with an entrepreneur friend. So you can tell how intense my workout was. And he's in Singapore, so it was in the middle of the night for me and early for him. And it was time to go to bed. I was like, 'Okay, now I got to get to bed. How do I get to bed? My body is all amped up, I've got food in my stomach, my brain is all amped up, and All-In podcast is tomorrow. And what time is it? 1:30 in the morning. I better get to bed.' So I put on a little spiritual video to calm me down, and then I got in bed and I was like, 'There's no way I'm falling asleep.' And I started meditating, and five minutes later I was asleep.

你知道吗，达赖喇嘛在他的 YouTube 频道上有一些很棒的两小时讨论。你听个二三十分钟，就会睡着。

You know, the Dalai Lama has these great two-hour discussions on his YouTube channel. You get about 20-30 minutes into that, you will fall asleep.

嗯，我的体会是，大脑会想尽一切办法逃避冥想。

Well, my learning is that the mind will do anything to avoid meditation.

顺便问一下，你们看到所有那些确认了吗？小罗伯特·肯尼迪确认了，布鲁克·罗林斯确认了。顺便说一句，如果你看 Polymarket，几周前它全都预测对了。有一阵子 RFK 跌到 75%，但后来反弹了，然后就定了。你能抓住那个机会吗？你可以赚钱的。而媒体当时说，'他不可能被确认，这不会发生。' 但 Polymarket 知道。真有意思。

By the way, did you guys see all the confirmations? RFK Jr. confirmed, Brooke Rollins confirmed. By the way, if you look at Polymarket, Polymarket had it all right a couple weeks ago. There was a moment where RFK fell to 75%, but then they bounced back and it was done. Could you have snipped that, man? You could have made money. And the media was like, 'No way he's getting confirmed, this is not going to happen.' But Polymarket knows. It's so interesting.

嗯，我看到一条很有见地的推文，我忘了是谁写的，所以抱歉没法注明出处。但那人基本上说：'看，特朗普在众议院和参议院拥有微弱多数，只要共和党人保持一致，他就能得到他想要的一切。所以 MAGA 运动对左派施加的所有压力和愤怒都是毫无意义的。关键在于让右派保持一致。' 所以，正是那些对参议员说'嘿，我要在初选中挑战你'的人，妮可·沙纳汉说'我要在初选中挑战你'，斯科特·普雷斯勒说'我要搬到你的选区'——这些才是推动事情进展、让确认通过的因素。这就是你得到卡什·帕特尔的方式，这就是你得到图尔西·加巴德担任国家情报总监的方式，这就是你得到 RFK 的方式。

Well, I saw a very insightful tweet, and I forget who wrote it, so I'm sorry I can't give credit. But the guy basically said, 'Look, Trump has a narrow majority in the House and the Senate, and he can get everything he wants as long as the Republicans stay in line. So all the pressure and all the anger that the MAGA movement is doing against the left is pointless. It's all about keeping the right wing in line.' So it's all the people saying to the senators, 'Hey, I'm going to primary you,' Nicole Shanahan saying, 'I'm going to primary you,' Scott Pressler saying, 'I'm moving to your district.' That's the stuff that's moving the needle and causing the confirmations to go through. That's how you get Kash Patel, that's how you get Tulsi Gabbard for DNI, that's how you get RFK.

你担心这些人中的任何一个吗？你觉得他们中有没有对你来说太激进的，还是你就喜欢那种'推倒重来，让疯狂的局外人上'的整个做法？

Do you worry about any of these? Do you think any of them are too spicy for your taste, or do you just like the whole 'burn it down, put in the crazy outsiders'?

这个描述太糟糕了。这不公平。我的意思是，老实说，我从没想过会看到这一幕，但我认为在埃隆和萨克斯以及类似的人之间，我们实际上有建设者、实干家、有财务智慧和经济智慧的人在掌权。而且你知道，尽管有这么多疯狂，埃隆做这些不是为了钱。他做是因为他认为这是正确的事。他搬进了罗斯福酒店。

That's such a bad characterization. That's not a fair characterization. I mean, honestly, it's like I never thought I'd see it, but I think between Elon and Sacks and people like that, we actually have builders and doers and financially intelligent people and economically intelligent people in charge. And you know, despite all the craziness, Elon's not doing this for the money. He's doing it because he thinks it's the right thing to do. He moved into the Roosevelt.

我想和许多人一样，我曾经相信'历史的大趋势'这种心态，就是觉得：'好吧，这是不可避免的，事情就是这样。政府总是越来越大，越来越慢，我们只能在他们关闭一切、我们变成欧洲之前努力把事情建成。' 但后来发生的事情是，你知道，凯撒越过了卢比孔河。'历史伟人理论'应验了，我们就生活在那个时代。这对我们所有人都是激励。尽管山姆·奥特曼和埃隆目前有争执，我知道山姆曾经受到埃隆的启发，我想我们所有人都受到埃隆的启发。我的意思是，这个人可以一边玩暗黑破坏神，一边管理 DOGE，同时运营 SpaceX、特斯拉、Boring 和 Neuralink。这简直令人难以置信。这让我们——这就是为什么我现在在做一家硬件公司。这让我想在有生之年做些有用的事。你知道，埃隆总是让我质疑：'我这辈子做的事情足够有用吗？' 这就是为什么我不想做投资人。你知道，彼得·蒂尔，讽刺的是他是个投资人，但他在那方面也很鼓舞人心，因为他会说：'没错，未来不会自己发生。你必须去创造它。' 所以你知道，我们可以去创造未来。我很高兴埃隆、DOGE 和其他人正在创造未来。

I think like many of us, I had bought into the 'great forces of history' mindset, where it's just like, 'Okay, it's inevitable, this is what's happening. Government always gets bigger, always gets slower, and we just have to try and get stuff built before they just shut everything down and we turn into Europe.' But the thing that happened then was, you know, Caesar crossed the Rubicon. The 'great man theory of history' played out, and we're living in that time. And it's an inspiration to all of us. Despite Sam Altman and Elon's current fighting, I know Sam was inspired by Elon at one point, and I think all of us are inspired by Elon. I mean, the guy can be the Diablo player and do DOGE and run SpaceX and Tesla and Boring and Neuralink. I mean, it's incredibly impressive. It makes us—that's why I'm doing a hardware company now. It makes me want to do something useful with my life. You know, Elon always makes me question, 'Am I doing something useful enough with my life?' It's why I don't want to be an investor. You know, Peter Thiel, ironically he's an investor, but he's inspirational that way too, because he's like, 'Yeah, the future doesn't just happen. You have to go make it.' So you know, we get to go make the future. And I'm just glad that Elon and DOGE and others are making the future.

我们这里在搞什么？也许几个月后我会做一个 All-In 播客，但这真的很难。我不确定我能搞定。所以让我试试，让我先确认它是否可行。是和无人机有关吗？是自动驾驶无人机吗？无人机很酷，但不是。

What do we got going on here? Maybe I'll do an All-In podcast in a couple of months, but it's really difficult. I'm not sure I can pull it off. So let me try, let me just make sure it's viable. Is it drone-related? Is it self-driving drone? Drones are cool, but no, it's not.

也许播客应该做天使投资人。哦，是的，绝对。不，不，是联合投资。杰森，就是我们的钱。你在说什么？你知道我是怎么了解联合投资的吗？Naval，我在 AngelList 上做的第一个联合投资，我想仍然是最大的。我不知道，5%。而 Naval 是我在这个 for.com 上的合伙人。我想如果我能做成，你们会喜欢我正在做的事情。我想你们会喜欢的。我很乐意给你们看个演示。告诉我们支票寄到哪里。

Maybe podcast should be an angel investor. Oh yes, absolutely. No, no, syndicate. Jason, just our money. What are you talking about? You know how I learned about syndicates? Naval, the first syndicate I ever did on AngelList, I think is still the biggest. I don't know, 5%. And Naval is my partner on this for.com. I think you'll love what I'm working on if I pull it off. I think you guys will love it. I'd love to show you a demo. Let us know where to send the check.

去买那个黑樱桃片 Van Leeuwen。我爱你们。我们学到了什么？我得走了。好的，大声感谢 Bobby 和 Tulsi。这对美国来说是巨大的胜利。我对他们两个都很兴奋。恭喜。我爱你们。谢谢你们来。让 Bobby 回来。Bobby，四年后再来播客。David Sacks，你的科学苏丹。David Friedberg，主席独裁者。Chamath，P hatita，POA，还有 Namaste。我是世界上最棒的主持人。我们下次 All-In Pod 再见。说再见。让你的赢家继续跑。Rainman David。相反，我们把它开源给粉丝，他们为之疯狂。爱你，Besties 女王。那是我的狗在占用你的车道吗？哦，老兄，它会在我这里见面。我们应该都找个房间，来一场大型狂欢，因为他们都这么没用。就像这种性张力，他们需要以某种方式释放。我们需要得到我们的。我要 all in。

Get that black cherry chip Van Leeuwen. I love you guys. What have we learned? I gotta go. Okay, big shout out to Bobby and to Tulsi. That's a huge, huge win for America. I'm stoked about both of them. Congratulations. I love you. Thanks for coming. Let's get Bobby. Bobby, come back on the pod for the four-year. David Sacks, your Sulon of Science. David Friedberg, the Chairman Dictator. Chamath, P hatita, POA, and Namaste. I am the world's greatest moderator. We'll see you next time on the All-In Pod. Say bye-bye. Let your winners ride. Rainman David. And instead we open source it to the fans and they've just gone crazy with it. Love you, Queen of Besties. Are that's my dog taking your driveways? Oh man, my it will meet me at. We should all just get a room and have one big huge orgy, because they're all this useless. It's like this sexual tension that they just need to release somehow. We need to get mer our. I'm going all in.