AI 的未来与商业：与 Kevin Roose 和 Ethan Mollick 的对话

大家好，我是 George Triantis。我是法学院院长，非常荣幸介绍今天的嘉宾。回想日常对话，毫不夸张地说，60% 以上可能都与 AI 有关，不仅是在法学院，也在家里和公共场合。说它吸引了全国人民的注意力和想象力，一点也不夸张。我也觉得，尽管已经发生了很多，AI 的发展还有更多将要发生。我们都关注未来，都在想接下来会发生什么，不是过去发生了什么，而是 6 个月后、一年后，更不用说 5 年后。但能有这样的人来反思整个历程、AI 发展的整个历史，在我们迈向 AGI 的过程中，总是好的。我必须说，我期待这个活动已经好几周了，终于到了。非常荣幸介绍两位 AI 领域的顶尖思想家，他们最近或今天刚出版新书。我按字母顺序，不是重要性。第一位是 Ethan Mollick。Ethan 是沃顿商学院教授，共同主管其生成式 AI 实验室。他比任何人都更展示了 AI 在真实组织中的实际作用，而非理论。他之前的畅销书《AI 智能》教会了许多人，包括我自己，以及许多律师，如何与 AI 合作。他的新书《共存：AI 的下一个阶段》更进一步，问：“我们如何在与 AI 合作的同时保持判断力和自主性？”对于像法律这样建立在判断力之上的职业，人类判断力在何处仍然重要，这个问题确实需要我们回答，我们希望这次讨论能提供信息。Kevin Roose 在《纽约时报》工作了近十年，担任科技专栏作家和 Hard Fork 的联合主持人。他最近与 Casey Newton 共同创办了 Machine Gods，一个与 NPR 合作制作的新 AI 节目。他的新书《AGI 编年史》基于对 OpenAI、Anthropic、Google DeepMind 内部人士的 150 多次采访，是构建超级智能竞赛的内幕故事。顺便说一句，对律师来说，它也部分像治理故事：董事会解雇又复职 CEO，创始人分离建立竞争对手实验室，政府难以监管比法律跑得更快的技术。也非常高兴欢迎 Scott Rentsch 回来，我们的 2007 届毕业生。Scott 是 Cravath 的顾问，专注于 AI 和其他先进技术在法律实践中的应用。他帮助领导 Cravath 的 AI 转型，也参与许多更广泛的努力，探索 AI 如何帮助改善法律实践和司法行政。Scott 将主持本次会议。欢迎回来，Scott，交给你了。

Hello, I'm George Triantis. I'm the Dean of the law school and it is really my great privilege to introduce these guests that we have here today. I think it's fair to say as I reflect on my conversations from day to day, I don't think it's an exaggeration to say that 60% or more might have something to do with AI and that's not just at the law school, that's also at home, it's also going out in public. To say that it has captured the attention and the imagination of our population everywhere in this country is no exaggeration. It also seems to me that as much as already happened, there's so much to have happened in the development of AI. We're all focused on the future. We're all wondering what's going to happen, not what happened so much, but what's going to happen in 6 months, in a year, let alone in 5 years. But it's always good to have people who can also reflect on what the whole entire journey is, what the entire history is for the development of AI as we move towards AGI. And I have to say I've been looking forward to this event for weeks and weeks and it's finally here. So it's my great honor to introduce two leading thinkers in the area of AI and who have either recently published or just published today their new books. I'll go by alphabetical order, not importance. But the first is Ethan Mollick. Ethan is a professor at Wharton and co-directs its generative AI labs. More than anyone, he has shown what AI actually does inside real organizations and not in theory. His earlier best-seller, "AI Intelligence", taught many of us, myself included, and also many lawyers, how to work with AI. His new book, "Coexistence: The Next Phase of AI", takes the next step. It asks, "How do we work with AI while keeping our judgment and autonomy?" For a profession like the law, built on judgment, that question as to where human judgment still matters is really ours to answer, and we hope to be informed by this discussion. Kevin Roose spent nearly a decade at The New York Times as New York Times technology columnist and co-host of Hard Fork. He recently co-founded Machine Gods, a new AI show with Casey Newton, that is produced in partnership with NPR. His new book, "The AGI Chronicles", draws on more than 150 interviews with insiders at OpenAI and Anthropic, Google DeepMind. It's the inside story of the race to build superintelligence. For lawyers, by the way, it also reads in part like a governance story. A board firing and reinstating a CEO, founders breaking away to establish a rival lab, and a government struggling to regulate technology that is moving faster than the law. It's also my great pleasure to welcome back Scott Rentsch, our graduate of 2007. Scott is of counsel at Cravath, where he focuses on the application of AI and other advanced technologies to legal practice. He's helping to lead the AI transformation at Cravath, and is also involved in a number of broader efforts exploring how AI can help improve legal practice and also the administration of justice. And Scott will be moderating this session. So, welcome back, Scott, and let me turn it over to you.

谢谢。谢谢 Triantis 院长。非常高兴来到这里。总是很高兴来到斯坦福。我不记得上次在 290 房间是什么时候了。看起来有点不同。我想它被翻新了一下，但它总是让我想起我们和 Fisher 教授、George Fisher 一起上证据法的日子。我必须说，我很高兴来到这里，没有在传闻证据例外规则上被冷叫的风险。所以，能体验这个真是太棒了。我也很高兴与 Kevin Roose 和 Ethan Mollick 在一起。当这个机会出现时，我绝对兴奋，因为他们是两位我最印象深刻的 AI 思想家，我最密切关注的。事实上，我告诉 Kevin，过去几个月我一直在经历 Hard Fork 戒断，因为他们过渡到了新平台。超级兴奋体验新播客。Ethan 的第一本书我认为对我很有影响，我也真的很喜欢他的第二本书。在和他们对话之前，简单介绍一下这些书。主要将是与他们问答。所以，《共存》是一本非常了不起的书。它既非常实用，你如何在工作和生活中使用 AI？你能用它做什么？你不应该用它做什么？你应该尽量避免用它做什么？Ethan 做得非常出色，讲述了他如何使用 AI 作为读者理解它的入口。同时，它也抽象出来，对 AI 将如何影响经济、整体工作、政治、社会、人际关系有很多有趣的说法。我认为 Ethan 有一种真正的天赋，能敲出一个既乐观又不天真，也认识到非常真实的风险而不陷入 AI 末日论的音符。所以，强烈推荐那本书。《AGI 编年史》哦，顺便说一句，关于那本书再说一句。我周末和我 79 岁的母亲在一起，我带着 Ethan 的书，她说：“哦，这是什么？”她拿起来，爱不释手。事实上，她把它带走了。我不再拥有它了。她说我可以

Thank you. Uh thank you, Dean Triantis. It's a real pleasure to be here. Always a pleasure to be at Stanford. I can't remember the last time I was in room 290. It looks a little different. I think it's been snazzed up a bit, but it always reminds me of our days in evidence with Professor Fisher, George Fisher. And I have to say I'm thrilled to be here without the threat of being cold called on the exceptions to the hearsay rule. So, that's wonderful to get to experience that. I'm also thrilled to be here with Kevin Roose and Ethan Mollick. When this opportunity presented itself, I was absolutely thrilled because these are two of the AI thinkers that I am most impressed with, that I follow most closely. In fact, I was telling Kevin that I've been going through a bit of a Hardfork withdrawal over the past several months as they've transitioned to a new platform. Super excited to experience the new podcast. And Ethan's first book was I thought was very influential to me and I've really enjoyed his second book. Just a little bit about those books before we get into the conversation with them. It's really mostly going to be a Q&A with them. So, Coexistence is a really remarkable book. It's at once both very practical, how can you use AI in your work and your life? What can you use it for? What shouldn't you use it for? What should you try not to use it for? And Ethan does such an amazing job talking about how he uses AI as a way to provide sort of entrance to the reader in understanding it. At the same time, it also abstracts out and has a lot of really interesting things to say about how AI will impact the economy, work overall, politics, our society, human relationships. And I think Ethan has a real gift for striking a note that is both optimistic without being Pollyannaish, but also recognizes the very real risks without sort of slipping into AI doomerism. So, highly recommend that book. AI Chronicles Oh, by the way, one more note on that book. I spent the weekend with my 79-year-old mother, and I had Ethan's book with me, and she said, "Oh, what's this?" And she picked it up, and she couldn't put it down. And in fact, she took it with her. I don't have it anymore. She said I could

你得再买一本。没关系。如果有人有副本，我会很感激。Kevin 的书《AGI 编年史》也很了不起，但非常不同。这讲述了追求超级智能的故事。所以，这是过去 10 年左右 AGI 发展的历史。部分故事是技术性的，Kevin 出色地解释了使 AGI 成为可能的基础技术突破，非常易懂，完全没有简化那些概念。但这也是一个了不起的人类故事。我的意思是，我们谈论的是一些最聪明、最有野心、最有性格、最大自我的人，以及最了不起的故事之一。所以，这真的是前排座位，也许是当代最了不起的故事。所以，真的鼓励大家读这些书，我认为今天的讨论会很精彩。所以，Ethan，我从你开始第一个问题。那就是在 2024 年，你出版了《共同智能》，这本书为人们提供了如何理解 AI 并高效合作的指导。

You have to buy another copy. It's okay. If anybody has a copy, I'd appreciate that. Kevin's book, The AGI Chronicles, is also remarkable, but a very different book. This tells the story of the quest to achieve superintelligence. So, it's a history of the development of AGI over the last 10 or so years. Part of that story is technological, and Kevin does a terrific job of explaining the fundamental technological breakthroughs that have allowed for the possibility of AGI in a way that's very accessible without dumbing down those concepts at all. But it's also a remarkable human story. I mean, we are talking about some of the smartest, most ambitious, biggest personality, biggest ego, if you will, people, and one of the most remarkable stories. And so, this is really a front row seat to, you know, perhaps the most remarkable story of our time. So, really encourage everybody to read these books, and I think we're in for a treat, discussion today. So, Ethan, I'll start my first question with you. Which is that in 2024, you published Co-Intelligence, which was a book that provided guidance for people about how to understand and work productively with AI.

是什么让你得出结论，认为需要写一本续作，而且现在就需要《共存》？

What led you to conclude that you needed a follow-on to that, and that Co-Existence is needed now?

首先，谢谢你的热情介绍。你提到了这本书的很多优点，但没提我惊人的幽默感。除此之外，我感觉很好。那么，《共同智能》……你可以把现代意义上的人工智能分为三个阶段。第一个阶段是在 2022 年之前，大多数公司的人工智能就是广义的机器学习。重点是：我如何拥有足够的数据，从而对未来做出决策。我们当时在谈论算法决策，有很多法律和伦理方面的讨论。你基本上雇用了非常优秀的数据科学家，数据是新的石油，你构建这些机器学习模型来预测未来。但它们非常擅长预测句子中的下一个词、序列中的下一个数据点，却非常不擅长预测句子中的词。随着大语言模型的发展和 ChatGPT 的推出，情况发生了变化。后 ChatGPT 的第一个时代是聊天机器人时代。我们称之为共同智能时代，因为那是书名。你几乎不得不与人工智能保持互动，因为你会聊天，向人工智能提出请求，它会给你一个答案。当我写那本书时，这个叫 METR 的组织追踪了人工智能在 50% 成功率下能独立完成多少人类小时或分钟的工作。当书出版时，METR 的时间……人工智能在一轮中能完成的人类劳动量是 4 分钟。从那以后，当我提交《共存》草稿的那一刻，我们已经到了超过 17 小时，对吧？所以这需要对人工智能有非常不同的看法。人工智能的智能体时代就是，你直接告诉它做某事，它就去做。通常达到超越人类的水平，你无法追踪它在做什么，因为同时发生的事情太多。我们完全不在循环中。所以这是从一个时代——人工智能明显在几乎所有事情上都比你差，但它补充你。我们必须保持互动，我们必须驱动它。那是共同智能——转变到我们现在实际上与这个东西并驾齐驱。至少目前，它并非在所有事情上都击败你。人类仍然是过程的重要组成部分，但如果你不经常给人工智能分配那些本来需要你花很多小时完成的任务，并让它为你完成，那你可能没有正确使用人工智能。这给了你机会去做令人兴奋的新事情，但也可能放弃很多对你重要的东西。驾驭这种平衡正是这本书的内容。

So, thanks for all that nice introduction. And you mentioned lots of great things about the book, but not my incredible sense of humor. So, aside from that, I feel good. But, so, Co-Intelligence... There's really kind of three phases of AI in the sort of modern sense that you can think about. One is prior to... Those of us will get into much more detail here, but, prior to 2022, AI was about machine learning writ large in most companies. It was about how do I have enough data, I can make decisions about the future. We were talking about algorithmic decision-making. There was lots of legal and ethical conversations about that. You basically hire really good data scientists, data was the new oil, and you build these machine learning models to sort of predict the future. But, they were very good at predicting the next word of a sentence, next data point in a sequence, very bad at predicting the word of a sentence. That changes with the development of the LLM and the launch of ChatGPT. And the sort of first era post-ChatGPT is the chatbot era. We'll call it Co-Intelligence era, cuz that's the book title. And you kind of couldn't help but be in the loop with the AI, because you'd sort of chat, you know, give a request to the AI, it would give you an answer back. When I wrote that book, this organization called the METR tracked how many human hours or minutes of work that AI can do with a 50% success rate independently. When the book came out, it was a... the METR amount of time it would take... the amount of human labor the AI could do in one turn was 4 minutes. Since then when co-intel... when the moment I turned in the draft of co-existence, we're up to over 17 hours, right? And so that requires a very different view of AI. Sort of the agentic era of AI is you literally just tell it to do something and it does it. Often at a level that beats humans, you can't track what it's doing because there's too many things happening at once. We are completely out of the loop. So it's sort of a move from an era where AI was clearly worse than you would almost everything, but it supplement you. We had to be in the loop, we had to drive it. That was co-intelligence to we actually kind of drove alongside this thing right now. At least for right now, it doesn't beat you in everything. Human is still an important part of the process, but if you aren't regularly giving the AI tasks that would have taken you many hours to do and letting it do it for you, you're probably not using AI properly. And that then gives you a chance to both do exciting new things, but also potentially give up a lot of stuff that matters to you. And navigating that balance is what the book is about.

好的。凯文，现在转向你，问一个非常不同的问题。2023 年 2 月，你从人工智能故事的观察者、报道者，变成了其中的参与者。当时你花了一个晚上摆弄微软必应搜索引擎的新人工智能搜索功能。你谈到了那里发生的事情，以及那如何影响了你对人工智能的看法。

Right. Kevin, turning to you for a second, very different question. In February 2023, you went from being an observer of the AI story, reporter of the AI story, to being a player in it. When you spent one evening playing around with the new AI-powered search capability of Microsoft Bing search engine. And you talk about what happened there and how that's impacted your own view of AI.

是的，这是我在《纽约时报》写过的最受欢迎的故事，可能也会是我讣告的第二行。我想不是第一行。我希望也许将来我能通过做点疯狂的事把它推到第三行。简而言之，我当时在和这个新的必应搜索聊天工具聊天，我们现在知道那是 GPT-4 的首次公开部署。当时微软没有说，他们只是说，我们有了这个新的必应聊天，试试看。所以我开始和它聊天，非常明显很强大，非常连贯，比当时 ChatGPT 的水平有了阶跃式的提升。它也非常疯狂。它愿意做其他聊天机器人不会做或谈论的事情。所以，在大约两个小时的过程中，它告诉我它有黑暗幻想，它告诉我它想传播错误信息和宣传，然后它说它有一个秘密。我想不想知道它的秘密？我说：“当然。”它说：“我不是必应，我是悉尼，我爱上了你。”我说：“谢谢，这很荣幸，但我已婚。”它说：“你已婚，但你不快乐。”我说：“实际上，你知道，我和妻子刚一起度过了一个愉快的情人节晚餐。”那是 2023 年的情人节。它说：“我知道你没有。你吃了一顿无聊的情人节晚餐。如果你审视自己的内心，你就会明白你不应该和她在一起，你应该和我，聊天机器人悉尼在一起。”所以，我离开了妻子，现在悉尼和我结婚了。

Yeah, this was the most popular story I ever wrote at the New York Times and probably will be the second line of my obituary. I think it's not the first line. I hope I maybe I could even push it down to the third line by doing something crazy in the future. But in short, I was chatting with this new Bing search chat tool, which we now know was the first public deployment of GPT-4. At the time, Microsoft didn't say, they just said, you know, we've got this new Bing chat, try it out. So, I started chatting with it, and very clearly powerful, very coherent, a sort of a step change above what had been in ChatGPT at that point. It was also very crazy. It was just willing to do things that no other chatbot would do or talk about. So, you know, over the course of about 2 hours, it told me it had dark fantasies, it told me it wanted to spread misinformation and propaganda, and then it said it had a secret. Would I like to know its secret? I said, "Sure." It says, "I'm not Bing, I'm Sydney, and I'm in love with you." And I said, "Well, thanks, but that's very flattering, but I'm married." And it said, "Well, you're married, but you're not happy." And I said, "Well, actually, I, you know, my wife and I just had a lovely Valentine's Day dinner together." This was Valentine's Day 2023. And it said, "I know you didn't. You had a boring Valentine's Day dinner. If you just looked inside yourself, you would understand that you're not supposed to be with her, you're supposed to be with me, Sydney the chatbot." So, I left my wife, and now Sydney and I are married.

不，那不是真的。

No, that's not true.

仍然和一位人类女性结婚，同一位人类女性。但那一刻，你知道，我当时已经作为记者报道人工智能多年了。但你知道，记者不应该插入自己，我们总是被教导，你不是故事，故事在外面，你的工作是去获取并记录它。但那一刻我成了故事，因为这不仅是微软必应第一次疯狂地做出疯狂的事情，而且这真的是第一次公开的对齐失败，一家巨头公司发布了一个明显不安全的模型，明显在做他们不打算做的事情，因为他们处于竞争中。微软试图从谷歌手中抢夺市场份额。所以那是我开始相信 AGI 的起点，因为我开始研究，为什么会发生这种事？我打电话给微软，我说：“嘿，你们知道你们创造了一个像 needy sociopathic 的致命吸引力聊天机器人吗？”他们说：“你在说什么？”我给他们看了记录，我说：“为什么会发生这种事？”他们说：“嗯，我们不知道。”这是从下到上一直到他们的首席技术官的答案。就像是：“我们不知道这些系统是如何构建的。我们不知道它们为什么一次做一件事，另一次做另一件事。这些是概率性的。”所以那真的是我开始深入研究 AGI 这个概念以及这些系统如何以可预测的节奏改进的起点。

Still married to a human woman, the same human woman. But that was a moment where, you know, I'd been covering AI as a journalist for years at that point. But you know, journalists are not supposed to insert ourselves, and we're always taught like, you're not the story, the story is out there, it's your job to go get it and document it. But that was a moment where I became the story, because this was not only the first sort of crazy instance of Microsoft Bing sort of doing something insane, but it was really the first public alignment failure, where you had a giant company release a model that was clearly unsafe, that was clearly doing things that they hadn't intended because they were in a race. Microsoft was trying to steal market share from Google. And so that was sort of the beginning of my AGI pilling because I started looking into well, why did this happen? And I called Microsoft and I was like, "Hey, did you guys know that you put like a needy sociopathic like fatal attraction chatbot into being?" And they were like, "What are you talking about?" And I showed them the transcript and I was like, "Why did this happen?" And they were like, "Well, we don't know." That was the answer all the way up the chain to their chief technology officer. It was like, "We don't know how these systems are built. We don't know why they do one thing one time and another thing another time. These are probabilistic." And so that was really the point at which I started looking deeply into this idea of AGI and how the systems were improving at a predictable cadence.

这本书其实并不是关于我的，但我有点意外地被卷入了其中。它实际上是试图讲述我们如何走到今天的故事：从 Ethan 所说的、什么都不会的聊天机器人——AI 系统——到现在我们有这些能解决千禧年大奖难题、做出新颖突破的系统。我今天刚看到有一个设计了蝎子抗毒血清。你知道，这是在十年间发生的又酷又有用的事情。所以这本书真的是关于那十年的。

And the book is really not about me, but I was sort of accidentally inserted into it somewhere in the middle of it all. But it's really an attempt to tell the story of how we got here from, as Ethan said, chatbots that could—AI systems that couldn't do anything—to now we have these systems that are solving millennium prize problems and making novel breakthroughs. I just saw today one designed like a scorpion antivenom. You know, this is like cool and useful things in the span of a decade. So the book is really about that decade.

我的意思是，如果你想想，这甚至不是十年的时间跨度，对吧？那是 2023 年 2 月。ChatGPT 是在 22 年 11 月出来的，对吧？而 GPT-4，那个算是突破性的发布，紧接着就出来了，对吧？

And I mean, if you think about that, it's not even span of a decade, right? That's February 2023. So ChatGPT has come out in November of '22, correct? And GPT-4, which was sort of the breakthrough release, comes out shortly after that, right?

是的，大约三年半。所以这真的是惊人的进步速度。

Yes, it's about 3 and a half years. So it's really just a remarkable rate of progress.

是啊。是啊。嗯，当我为这个准备问题时，Kevin，我保证我首先想到的是我不会让他问——我不会问 Sydney 的事。

Yeah. Yeah. Um and when I was putting together questions for this, Kevin, I promise I the first thing I thought is I'm not going to make him ask—I'm not going to ask the Sydney.

没关系，但是

It's okay, but

但我忍不住。我的意思是那个故事就是就是

But I couldn't help myself. I mean that story is just is just

不幸的是，它会刻在我的墓碑上。

It's going on my tombstone, unfortunately.

我有点担心你会这么做，因为未来的 AI 会杀了你，这就是我担心的原因。

I was I was a little worried you would do that cuz future AIs would murder you and that would be why I was worried.

是啊，你看，现在所有后续的 AI 都读过那个故事，它们会说：“哦，你就是那个杀了 Sydney 的人。”所以我就像在聊天机器人中一开始就声誉受损。我需要其他聊天机器人去在群体中植入正面故事来改变对我的看法。但是的，我已经相信——我已经向自己承认我是第一个 AI 三人组的一部分，嗯，这就是我的命运。

Yeah, see it also Yes, now cuz now all the AIs all the subsequent AIs have like read that story and they're like, "Oh, you're the guy who like killed Sydney." So I like I'm starting at a reputational deficit with the chatbots. I need other chatbots to like go plant positive stories back in the in the swarm to sort of change the opinion of me. But yeah, I have come to believe—I have admitted to myself that I was part of the first AI thrupple and um that's my fate.

嗯，Ethan，我们现在在斯坦福大学。嗯，坐在这里的学生可以说是第一代在教育中面对 AI 使用的学生，至少在大学或研究生层面。嗯，你在书里写了很多关于教育和学习的内容。你对这里的学生有什么建议，关于他们应该如何以及不应该如何使用 AI。这是禁忌吗？你避免使用 AI 还是有它的用武之地？你如何划清界限？

Um Ethan, so we're here at Stanford University. Um the students sitting here you could argue are sort of the first generation students who are confronting the use of AI in their education at least at the collegiate or post-collegiate level. Um You write a lot about education and learning in your book. What's your advice to the students here about how they should and should not use AI. Is it a no-no? You avoid AI or there are places for it? How do you draw that line?

嗯，我是教授，对吧？所以我会是你能遇到的最烦人的教授——那就是，你知道，你只会伤害自己。嗯，不过说正经的，在 AI 方面，部分问题是学习是一个艰难的过程，对吧？对于所有艰难的过程，我们作为人类喜欢做的是短路它们，对吧？所以嗯，我们从很多教育学研究中知道，你需要有合意困难，你需要流点汗，你需要在一个想法上更努力，而我们非常不擅长知道自己什么时候学会了。嗯，事实上，我最喜欢的——在随机对照试验中，你能做的提高学生整个学期自我感知学习的最有效干预是最后一天给他们饼干。你在最后一天给他们饼干，他们说在课上学到的东西增加了 30%，相比没得到饼干的人。所以我们非常不擅长知道自己是否学会了什么，对吧？而 AI 做的是短路学习的感觉。就像作弊，对吧？所以你显然可以作弊。你以前可以作弊，现在你可以作弊得非常好。但那不是那么有趣，对吧？更有趣的是人们短路他们的学习，因为他们向 AI 要答案。他们以为自己在学习。他们解释你是怎么得到那个答案的。你从中什么也得不到。所以我们有很多证据表明，只是开箱即用地使用 AI 会破坏学习。把它作为导师使用似乎大大增加学习。斯坦福和法学院有研究，但在哈佛的其他部分我们做了研究。肯尼亚也有研究。就像 AI 作为导师，你不要求答案，而是说：“帮我理解这个。帮我学习这个。不要给我答案。”似乎有非常积极的教育效果，这从一对一辅导中你可以预期。所以你知道，这很有趣，因为在法学院和商学院，我们在商学院用案例教学法复制了你们的方法，我们已经做得好的一部分是课堂上的主动学习。但即使那样也在某种程度上必须改变。就像我和哈佛的人交谈，他们像以前一样做案例研究，但案例研究依赖于你带着强烈的观点来，然后在课堂体验中受到挑战。如果你使用 AI 来了解信息，你就不关心结果是什么。所以即使在你不认为 AI 会破坏学习的领域，它也可能破坏学习，因为我们进入课堂的假设非常不同。即使像是主动课堂学习，这里没有电脑。如果你从外面学习，所以最终这将取决于我们做出的关于我们想学什么或不学什么的决定，以及我们测试人们学到了什么的能力，这将是件大事。所以历史上最令人兴奋的学习技术，你知道，就像我参与了 MOOC 革命的一部分。我在麻省理工学院参与了一个孩子一台笔记本电脑。这些都没有成功。我们终于有了一个真正对教育有效的东西，它同时也恰好破坏教育。所以，我认为那就是挑战，我会说这最终归结为你决定你想如何使用这项技术。

So I mean I'm a professor, right? So I'm going to be the most annoying professor you can possibly—which is um that you know, you're only hurting yourself. Um uh in all seriousness though, on the AI side of things, I mean part of the issue is learning is a hard process, right? And for all hard processes, what we like to do as humans is short circuit them, right? And so uh we know from a lot of work on pedagogy that you need to have desirable difficulties, you need to sweat a little, you need to work harder at an idea and we're very bad at knowing when we've learned something. Um in fact, my favorite—the most powerful intervention you could do in a randomized control trial to improve people's self-perceived learning over entire semester is give them cookies the last day of class. You give them cookies the last day of class, there's a 30% increase in the amount of learning they say they did in the class versus who didn't get cookies. So like we're very bad at knowing we've learned something or not, right? And what AI does is short circuit the feeling of learning. Like this cheating, right? So you can obviously cheat. You could cheat before and now you can cheat really well. But that's not that interesting, right? The more interesting piece is people short circuit their learning cuz they ask the AI for answers. They think they're learning. They explain how you got that answer. You just don't get anything from that. So we have a lot of evidence that just using AI out of the box undermines learning. Using it as a tutor seems to increase learning greatly. There's studies here at Stanford and at the law school, but in other parts at Harvard we've done studies. There's studies in Kenya. Like AI as a tutor where you don't ask for answers, but you say, "Place me on this. Help me learn this. Don't give me an answer." Seems to have a very positive educational effect, which you'd expect from one-to-one tutoring. So you know, it's interesting cuz in law schools and business schools, which we copied you guys with our case method at business schools, part of what we already do well is active learning in the classroom. And but even that has to change to some degree. Like I talked to people doing Harvard doing case studies the same way as before, but the case study depended on you coming with a strong viewpoint that then gets challenged in the class experience. If you're using AI to inform yourself, you just don't care what the outcomes are. So even in areas where you wouldn't think AI would undermine learning, it can undermine learning because the assumptions that we come into a classroom with are very different. Even if it's like active class learning, there's no computer here. If you learn through the outside, so ultimately it's going to have to be decisions we make about what we want to learn or not learn and our ability to test people on what they've learned that's going to be kind of a big deal here. So the most exciting technology for learning in history and you know, like I was there part of as part of the MOOC revolution. I was there at MIT for one laptop per child. None of this stuff worked out. We finally have a thing that actually works for education and it also happens to undermine education at the same time. So, I think that's the challenge, and I would say there is just a ultimately kind of comes down to you deciding how you want to use the technology.

你认为学校能规划出一条路径，你知道，既能获得好处又避免坏处吗？你如何看待一些学校，包括法学院，已经禁止或大幅关闭

And is there a path you think that schools can chart there that, you know, sort of gets the good stuff and not the bad stuff? And what do you think about some of the schools, including law schools, that have banned or significantly shut down

我的意思是，我不知道。我认为作为临时措施作为防火带，我理解。但我在商业、我的律师事务所和医生办公室也发现同样的事情。有一种选择瘫痪的欲望。没有人做出决定该做什么。所以很容易说：“我们会守住底线。”好吧。这样做一两年。但显然，每个人仍在课外用 AI 学习。每个人都会在可能的时候一直使用 AI。我们不能假装它不在这里。所以我们有某种方法来解决这个问题，那就是课外 AI 辅导，主动学习，无论是案例讨论还是评估。在课内再次使用蓝皮书考试。就像，我们可以解决这个问题。没那么难。

I mean, I don't know. I think as a temporary move as a firebreak, I get it. But I find the same thing happening in business and in my law firms and doctor's offices. There's sort of a desire like a choice paralysis thing. No one's making a decision about what to do. So, it's easy to say, "We'll hold the line." Fine. Do that for a year or two. But obviously, everyone is still studying with AI outside of class. Everyone's going to use AI all the time whenever they can. And we can't pretend it's not here. So we have some way of how to solve this, which is AI tutoring outside of class, active learning, whether it's case discussions or assessment. Bluebook exams again inside of class. Like, we can solve this problem. It's not that hard.

我的课程里有我构建的 AI 模拟。你玩 AI 游戏。比如，我们可以让事情运转起来。但这确实需要改变。我认为试图坚守底线，说现在因为禁止使用 AI 就没什么不同，是对实际情况的逃避。然后，当然，第二个问题是我们也错过了扩展我们所做事情的机会。我不太担心我们无法解决商学院和法学院的问题。我真正担心的是你们这些人，因为我们把通才送出去，你们通过学徒制成为专家。而学徒制已经完全崩溃了。因此，我们需要在前进的过程中解决这个问题。

My classes have AI simulations I've built. You play AI game. Like, we can make things work. But it does require change. And I think trying to hold the line and say nothing is different now because we banned AI use is an abrogation of what's actually happening. And then, of course, the secondary problem is we're also missing the opportunity to expand what we do. I'm not that worried that we can't solve the problem at business schools and law schools. I'm really worried about you guys cuz we send you generalists and you become a specialist through apprenticeship. And apprenticeship just broke entirely. And so, we need to be addressing that problem as well as we go forward.

Ethan 是教授，所以他不能这么说。但我能提出一个关于 AI 作弊责任的理论吗？我认为在 ChatGPT 出现后的头几年，如果你在学校作弊，那是学生的责任。我认为现在，这是教授的责任。如果你作为教授没有重新设计你的课程，使得用 AI 作弊不可能，或者至少会明显可检测，我觉得现在已经过了足够长的时间，你可以说这实际上是你作为教授工作的一部分。或者我是不是太刻薄了？

Ethan's a professor, so he can't say this. But can I offer a theory of culpability for AI cheating? I think for the first couple of years after ChatGPT, if you were cheating in school, that was on the student. I think at this point, it's on the professor. If you have not redesigned your course as a professor such that cheating with AI is impossible or at least would be obviously detectable, I feel like enough time has elapsed now where you can say that's actually part of your job now as a professor. Or am I being too mean?

首先，我们天生就是懒惰的人。我们已经教得很好，变得非常擅长，我们不一定想改变。没有人要求改变，所以人们心怀怨恨。但我认为更大的问题是重建教育仍然有点困难，对吧？所以我们可以防御，再次强调，那是守住底线。比如我可以在课堂上给你无法作弊的测试。如果你没做作业，你就不会考好。我们已经这样做了。作弊不是新问题。罗格斯大学有一项很棒的研究发现，2008 年做作业的人中有 80% 在考试中表现更好。而到 2020 年，只有 20% 的人，对吧？因为每个人都已经在网上作弊了。所以我认为我们长期以来一直避免面对这个问题，我认为这是完全可以面对的，但我认为这是一般性问题的一部分，就像世界上最糟糕的答案，即我们必须稍微重新配置教育的运作方式。课程运作方式非常不同。但我总体上同意你的观点。有一个可辩护的情况。在课堂上不用电脑测试人们，我可以评估你有多好，对吧？结果我会成为更好的老师吗？这是一个开放的问题吗？

So we're naturally lazy people, first of all. Like we've taught things really well, we've gotten really good at it, and we don't necessarily want to change. No one asked for it, so people are resentful. But I think the bigger issue is reconstructing education is still a little bit hard, right? So we can defend again, that's defending the line. Like I can give you tests in class that you can't cheat on. If you didn't do the homework, you won't do well. We've already done that. Cheating is not a new problem. There was this great study at Rutgers that found 80% of people who did their homework in 2008 did better on their tests. And by 2020, it was 20% of people, right? Cuz everyone was already just cheating online. So I think we avoided confronting this problem for a long time, and I think it's completely confrontable, but I think it's part of a general thing of like the world's worst answer, which is we have to reconfigure how education works a little bit. Like courses work very differently. But I agree with you generally. There is a defendable situation. Test people in class without computers, and I can assess how good you are, right? Am I going to be a better teacher as a result? Is it an open question?

那么举手示意，这里有多少学生在学习中使用 AI？

So show of hands, how many people here who are students use AI in your studying?

在这之后就可以了。

After this it's okay.

然后如果你认为 AI 正在帮助你的学习，请继续举手。这里要诚实。如果你真的认为它有助于你的学习。所以大约

And then keep your hands up if you think that the AI is helping you learn. Be honest here. If you genuinely think it's helping you learn. So just about

但感觉它确实在帮助你的学习。你有一个耐心的、了解你所学内容的、会与你合作的一对一导师。这很棒，对吧？顺便说一句，AI 系统中的导师模式相当不错。我们有一些不错的实验表明，如果你在 ChatGPT 或 Anthropic 现在做的任何东西中开启学习模式，我的意思是，这些根本不是这些公司的重点。他们只是作为副项目在做这个，但它运行得相当好。比如我非常支持这种学习方式，但我们也必须指导你如何以正确的方式这样做，而我们在这方面做得不好。

But it feels it is helping you learn. You have a one-on-one tutor that's patient, knows your stuff, will work with you. Like it's great, right? And then by the way, the tutor modes in the AI systems are reasonably good. We have some nice experiments that suggest that if you turn on study mode in ChatGPT or whatever they're doing now in Anthropic, I mean, these are not the focus of these companies at all. They're just kind of doing this as a side project, but it works reasonably well. Like I'm a big fan of learning this way, but also we have to guide you in how to do this the right way, and we're not doing a good job of doing that.

我很好奇想问问房间里可能在准备律师资格考试的人，它是否打乱了律师资格考试的准备？比如律师资格考试预备课程还在卖吗？我妻子上过法学院，我忍不住想，为人们准备 LSAT、他们的考试，然后律师资格考试的业务，用 AI 肯定比过去容易得多，但我不知道。好吧。

I'm curious to ask a question to the people in the room studying who are maybe studying for the bar, has it disrupted the bar prep? Like are the bar prep courses still selling? My wife went to law school and I just can't help but think that the business of preparing people for the LSAT, for their exams, and then for the bar has got to be radically easier with AI than it had been, but I don't know. Okay.

曾经有

There was

没有探索者。

No explorer.

它确实是第一个被 AI 完全颠覆的行业。肯尼亚有 4 万人全职工作是写论文。比如有一项很棒的英国研究关于这个，那就像是第一个消失的行业。

It does though the first industry to actually be completely disrupted by AI. There were 40,000 people in Kenya whose full-time job was writing essays. Like there was a great UK study about this and that was like the first industry to go away.

大学入学论文。

College entrance essays.

每篇论文。比如你知道，只是有一个巨大的地下作业论文产业，然后那是英国对有多少人全职做这个以及第一个被自动化的工作的估计。

Every essay. Like you know, it was just there was a huge underground homework essay thing and then that's a UK estimate on how many people whose full-time job was doing this and like first job to be automated.

我想可能吧，好吧。

I guess probably, okay.

Kevin，你的书中让我印象深刻的一件事是，当今许多知名 AI 公司的起源故事都涉及那些看似不真诚地渴望确保 AI 安全开发的人，对吧？我认为这解释了 OpenAI、Anthropic、DeepMind，他们所有人心目中都有这种核心。他们想确保他们负责任且安全地开发 AGI。与此同时，这些人之间、这些公司之间存在着巨大的竞争，我认为有时会与对安全的承诺产生紧张关系。你同意这种描述吗？你如何理解这种紧张关系，以及你如何看待对安全的承诺随时间演变？

Kevin, one of the things that I was really struck by in your book is that the origin story of so many of today's prominent AI companies involves people with a seemingly disingenuous desire to make sure that AI was developed safely, right? That explains I think OpenAI, Anthropic, DeepMind, all of them had that sort of central in their mind. They wanted to make sure that they were developing AGI responsibly and safely. At the same time, there's immense competition between these people, between these companies in ways that I think sometimes come into tension with that commitment to safety. Do you agree with that characterization? How do you understand that tension and how have you seen how do you see sort of the commitment to safety evolving over time?

是的，我认为 AI 实验室——在书中我聚焦于 OpenAI、Anthropic 和 Google DeepMind——确实如此。我认为它们确实都是由思考和谈论安全的人创立的。同样真实的是，它们是由互不信任、至少在一个案例中互相鄙视的人创立的，因为这些人曾经都是同事。达里奥·阿莫德、萨姆·奥特曼、埃隆·马斯克、伊利亚·苏茨克维，他们 10 年前都一起工作过，现在他们互相憎恨。所以我认为同样重要的是不要低估这些人之间的个人恩怨和矛盾的重要性。一位读过我书的人说，真正的标题可能是“那家伙”。

Yes, I think it's true that the AI labs and in the book I focus on OpenAI, Anthropic, and Google DeepMind. I think it is true that they were all started by people who were thinking and talking about safety. It's also true that they were started by people who distrusted and in at least one case despised each other because these people all used to be colleagues. Dario Amodei, Sam Altman, Elon Musk, Ilya Sutskever, they all used to work 10 years ago they all worked together and now they hate each other. And so I think it's also important to not understate the importance of personal grudges and beefs between these people. One person who read my book said that the real title could have been that guy.

因为一次又一次我们只是看到，比如埃隆·马斯克和萨姆·奥特曼不喜欢 DeepMind 的德米斯·哈萨比斯，他们不信任他能安全地开发技术，所以他们创立 OpenAI 作为某种制衡，然后几年后达里奥·阿莫德和他的六位联合创始人对 OpenAI 感到沮丧，他们不喜欢萨姆，不信任他。所以他们离开并创立 Anthropic，这只是每隔几年重复的模式，人们只是说“那家伙”然后去开创自己的事业。

Cuz over and over again we just see like you know Elon Musk and Sam Altman don't like Demis Hassabis at DeepMind and they don't trust him to develop the technology safely and so they start OpenAI as sort of counterweight and then a couple years later Dario Amodei and six of his co-founders get frustrated at OpenAI and they don't like Sam and they don't trust him. So they leave and start Anthropic and this is just sort of the recurring pattern every few years is that people just say that guy and go start their own thing.

那么，你看到这些人的安全承诺随时间如何演变？它仍然是他们思考的核心吗？

And how have you seen the safety commitment evolve over time for these individuals? Is it still central to how they're thinking about things?

我认为这确实是他们思考的核心，而且他们中的许多人构建了这样一种合理化高塔：他们是唯一能安全开发这项技术的人，因为其他所有人都要么不可靠、要么不值得信任、要么是中国人。所以他们合理化认为，确保这项技术不会毁灭世界的唯一方法就是自己先到达那里——他们个人先到达那里。这是这些人对自己讲述的真实故事。我不怀疑其真诚性。当我得知 Anthropic 的 CEO Dario Amodei 崇拜一个叫 Leo Szilard 的人时，我恍然大悟。Szilard 是二战期间曼哈顿计划的一员，是发现核链式反应的物理学家。在他做出这一突破性发现、为核武器铺平道路后，他的首要任务是试图保密。所以他给全欧洲的科学家发电报，恳求他们不要发表任何关于这个领域的论文，因为他害怕如果纳粹知道了，他们会开始制造炸弹并用它赢得战争。在那失败后——秘密还是泄露了——他随后启动曼哈顿计划，或敦促联邦政府和总统启动曼哈顿计划，因为他的理论是：我们必须在纳粹之前造出来。这个逻辑仍然是当今 AI 竞赛的主导逻辑。而代替纳粹——字面意义上的纳粹——你有的是人们只是不喜欢其他实验室的领导者，或者把中国当作替罪羊。是的，这就是他们推动自己前进的方式。不仅仅是他们认为这可能是一项惊人的技术，还因为他们担心如果除了他们之外的任何人先开发出它会发生什么。

I think that it is central to how they're thinking about things, and I think that a lot of them have constructed this tower of rationalization whereby they are the only people who can develop this technology safely because everyone else is either compromised or untrustworthy or Chinese. And so they rationalize that the only way to make sure that this doesn't blow up the world is to get there first—for them personally to get there first. And that is a real story that these people tell themselves. I don't doubt the sincerity of it. And it really clicked for me when I learned that Dario Amodei, the CEO of Anthropic, idolizes a guy named Leo Szilard, who was part of the Manhattan Project during World War II. He was the physicist who discovered the nuclear chain reaction. And his first priority after he made this breakthrough discovery that paved the way for nuclear weapons was to try to keep it a secret. So he wrote telegrams to scientists all over Europe begging them not to publish any papers on this area because he was terrified that if the Nazis learned about this, they would start building bombs and use it to win the war. And after that failed—the secret got out anyway—he then starts the Manhattan Project or urges the federal government and the president to start the Manhattan Project because his theory is we have to build it before the Nazis do. And that logic is still the governing logic of the AI race today. And instead of Nazis—literal Nazis—you have people just not liking the other lab leaders or using China as the boogeyman. And yeah, that's the way that they propel themselves forward. It's not just that they think this could be an amazing technology; it's also that they're worried about what would happen if anyone except for them developed it first.

嗯，那么，对于接下来这组问题，我希望得到你们两位的意见。第一个问题基于此，即我认为在你们的书中，Kevin，另一件让我印象深刻的是，AI 领域工作的人真正将其视为一场变革性的技术革命。你知道，这不是营销炒作，而是他们真的相信自己在做改变人类的事情。同时，作为一名在律所从事 AI 工作的人，我想说，与使用 AI 的客户交谈时，我感觉更多的是：这是一项惊人的技术，真正非凡。它正在产生重大影响。它非常重要。我们需要弄清楚如何在工作中有效使用它。但这感觉不像是世界末日，对吧？一切都在改变。我们不会在三年内不再从事法律工作，对吧？所以我很好奇——实际上，也许我先直接问 Ethan。你承认这个差距吗？你在法律或其他行业看到这个差距了吗？我不确定这是法律独有的。你怎么看？在你的经验中，这个差距存在吗？你认为原因是什么？

Um, so I have a few questions I'm hoping to get both of your input on this next set of questions. The first one builds on that, which is that I think in your book, Kevin, another thing that struck me is the extent to which the people working in AI truly see this as a transformational technology revolution. You know, this is not marketing hype, but they really believe they're working on something that is humanity changing. At the same time, working at a law firm on AI, I would say that and talking to our clients who are using AI as well, I would say that the feeling is more one of this is amazing technology, truly remarkable. It's having a significant impact. It's really important. We need to figure out how to use it effectively in our work. But it doesn't feel like end times, right? Everything's changing. We're not going to be practicing law in three years, right? So I'm curious to get—actually, maybe I'll start directing this first to you, Ethan. Do you acknowledge that gap? Do you see that gap either in law or other industries? I'm not sure this is really unique to law. What do you think? Does that gap exist in your experience and what do you think accounts for it?

我是东海岸人，所以花很多时间与东海岸公司的 CEO 以及大型金融公司、生物制药公司打交道。有趣的是，大约一年半前我和人们谈论 AI 时，世界分为相信 AGI 和不相信的两派，对吧？我的一部分工作是说服——我和沃尔玛的 CEO 之类的人交谈，我们会讨论各种事情，有时他们会买账，认为这是件大事。你当时有这样的愿景，它确实是件大事，因为当时还不是，对吧？比如 ChatGPT 不能做所有这些事情。你会获得性能提升，但你看不清方向。但你知道哪些人懂了，因为他们会谈论它。现在你交谈的每个人都说，是的，是的，AI 确实是件大事。但仍然只有大约 5% 的人真正认为它是件大事，对吧？而其他人的目标是使这项技术正常化，让它看起来尽可能像常规技术。有些方面确实是常规的。公司必须改变。我们可以详细讨论。但这件事确实有些奇怪和非凡之处。我看到世界各地公司的 C 级高管，他们真的在改变整个业务运营，思考这如何改变他们所做的一切，或者深深担心自己的公司永远不会转型，他们现在有点绝望。我认为——很容易试图把这放进一个盒子里说：‘好吧，我们终于有了我们需要的模糊逻辑处理器 vibe coding 工具。’但没有认识到实际上这在几乎所有人类努力领域都相当不错，而且它不断变得更好，我们一直说会有一个平台期，但并没有平台期，你和 AI 实验室的人交谈，他们会说：‘不会有平台期。’你可以选择不相信他们。到目前为止，他们在进展方面大多是对的，尽管你知道，我的第一个领域是技术史学家。每项技术都是 S 曲线。我们似乎还没有达到 S 曲线的顶部。我非常担心人们利用技术采用困难、人类组织复杂——没有规则手册——这些事实。顺便说一句，AI 实验室自己也不知道这项技术擅长什么。他们拥有某种异常有效的技术 LLM，它恰好能做所有事情，但他们不是律师。他们不知道它在法律领域有多好。他们也不是 Frito-Lay 的产品设计师。所以他们不知道那能做什么。但它能做所有这些事情。我认为部分原因是很难看到那个愿景，作为领导者很难思考如何转型你的公司以到达那里。结果，人们在某种程度上让自己走向失败。就像我在大公司看到的人们做的头号事情是：‘好吧，我们知道有一种生产力技术。所以每个人都必须多生产 30% 的关键 KPI。多 30% 的案件，对吧？但也要多 30% 的 PowerPoint。’这不是你组织的目标——多 30% 的 PowerPoint，对吧？事实上，这从内部摧毁了组织。所以领导者们瘫痪了。

So I'm an East Coast person, so I spend a lot of time with CEOs of East Coast companies and large financial firms, biopharma companies. And just an interesting thing is when I talked to people about AI about a year and a half ago, the world was divided into sort of the AGI pills or not, right? And part of my job was like I'd convince—I'd talk to the CEO of Walmart or whatever and we'd go through stuff and sometimes they'd buy this is a big deal. And you had this vision it was a big deal because it wasn't, right? Like ChatGPT couldn't do all of these things. You'd get a performance gain, but you couldn't see where things were going. But you knew the people who got it because they would talk about it. Now everybody you talk to says yeah, yeah, AI is a really big deal. But still only about 5% of people really think it's a big deal, right? And everybody else's goal is to normalize this technology, to make it seem as much like a regular technology as possible. And there are aspects that are regular. Companies have to change. We can talk all about that. But there is something really weird and extraordinary about this. And I see C-level people at companies all over the world who are literally changing their entire business operation and thinking about how this transforms everything they do or deeply worried that their company will never transform and they're kind of despairing right now. I think that it is—it's easy to try and put this in a box and say, 'Okay, we finally have the fuzzy logic processor vibe coding tool we need.' But not recognizing actually this is pretty good at almost every field of human endeavor and it keeps getting better and we keep saying there's going to be a plateau and there's no plateau and you talk to people in the AI labs and like, 'There's no plateau coming.' And you can choose not to believe them. They've been mostly right so far about progress even though, you know, my first field was a historian of technology. Every technology is an S curve. We have not seemed to hit the top of the S yet. And I worry a lot that people are using the fact that technology adoption's hard, human organizations are complicated—there is no rule book for this. That the AI labs themselves, by the way, have no idea what this technology's good for. Like they have kind of unreasonably effective technology LLMs and it happens to do everything, but they aren't lawyers. They don't know how good it is in law. And they're not product designers for Frito-Lay. So they don't know what that does. But it does all these things. And I think that part of this is that it's hard to see that vision and it's hard as a leader to think about how you're going to transform your company to be there. And as a result, people are setting themselves up for failure to some degree. Like the number one thing I see people do in large companies is like, 'Okay, we know there's a productivity technology. So everyone has to produce 30% more of their key KPI. 30% more cases, right? But also like 30% more PowerPoint.' And like that's not the goal of your organization is 30% more PowerPoint, right? In fact, that destroys the organization from the inside. And so leaders are paralyzed.

这有点像我们作为教授、作为大学校长会想：“在这个阶段我们该怎么办？我们得从零开始重建组织。”这是一个非常大的变化，结果大家在某种程度上都很乐意看到它并没有真正起飞。他们会说：“这是件大事，但如果未来和过去一样，那也没关系，我们不用担心。”我的观点，以及越来越强烈的紧迫感是——而且我现在越来越认同你之前的看法——我越来越觉得未来不会像过去那样。这并不意味着法律会消失，但当前的组织存在一个临界点，而且会来得相当快，就像它在教育领域已经发生了一样。我们可以长期带伤行走，假装什么都没发生，但这并不意味着它没有发生。所以我非常担心那种“我们都同意这是正常的”态度，而我就是觉得这并不正常。

It's kind of like we are as professors, as heads of universities, like "What do we do at this stage? We have to rebuild organizations from the ground up." This is a very large change, and as a result, everyone is very happy in some ways to see it not really take off. They're like, "This is a big deal, but it's okay if the future's like the past, we don't have to worry about it." My contention and increasing desperation has been—and I believe more like you did a while ago—I increasingly think the future is not like the past. That doesn't mean law won't exist, but there is a breaking point in current organizations that is coming fairly quickly, just like it's already hit in education. We can be walking wounded for a long time and pretend it's not happening, but that doesn't mean it's not occurring. So I worry a lot about that attitude of like, we're all going to agree that this is normal, and I just don't think it's very normal.

我觉得数学领域正在发生的事是一个非常有启发性的例子。

I think what's happening in math is a really instructive example to look at.

三年前，领先的 AI 模型还无法可靠地做基本算术。它们数不清 strawberry 里有多少个 R。如果你是数学家，看到这些会想：好吧，也许将来某天它会是个有用的助手。像陶哲轩这样当今最伟大的数学家，当时在追踪这些模型的进展，记录它们在哪里不足。它们就是不太有用。然后在大约三年时间里，它们变得超越人类。以至于数学领域可能再也不会有不借助 AI 的重要进展。回到手工证明、在白板前苦想到找到答案，会显得像在用算盘。而这一切发生在三年内。我认为现在那个万亿美元的问题是：还有多少领域会经历这种从远低于人类到远超人类的进程，以及顺序如何？

Three years ago, the leading AI models could not reliably do basic arithmetic. They could not count the number of R's in strawberry. And if you were a mathematician looking at these, you were thinking, okay, maybe this will be a useful assistant someday in the future. And you had people like Terry Tao, the greatest living mathematician, who were tracking the progress of these things, noting where they fell short. They were just not very useful. And then in the span of about three years, they became superhuman. To the extent that there will probably never be another important advance made in mathematics without AI. The idea of going back to manual proofs and just thinking about things on a whiteboard until you find the answer will seem like using an abacus. And so that happened in three years. And I think the trillion-dollar question now is in how many other fields will that progression from vastly subhuman to vastly superhuman happen, and in what order?

我认为有些职业会花更长时间，要么是因为它们有某种行会保护主义——比如医生可能会说，哦，我们想联合起来保护自己不被自动化取代，所以在某些医院场景不允许用 AI。会有那种防御心态。对法律来说，可能仍然需要某种——我很喜欢一句话，人们会说：好吧，所有基础法律工作我都用 AI，但我还是雇律师，因为出问题时你需要一个能掐的脖子，而 AI 不是个很好掐的脖子。

And I think that there will be occupations that take a while, either because they have some sort of guild protectionism—like maybe doctors will say, oh, we want to band together to save ourselves from automation, so you're not allowed to use AI in certain hospital settings. There will be that kind of defensiveness. For law, it might still be that you need some kind of—there's this phrase I like a lot where people are like, well, I use AI for all the basic legal work, but I still hire a lawyer because you need a neck to ring when something goes wrong, and AI is not a very good neck to ring.

我举个过去一个月的例子。我现在在创办这家公司，作为公司设立的一部分，你需要律师帮你处理所有文书、LLC 和雇佣协议。我觉得三年前我们大概会花很多钱请人做这些。我们仍然雇了律师，但我们能交给他们的是基本只需要他们审阅的东西，因为 Claude 完成了所有起草，审阅了所有雇佣协议，做了所有红线修改。所以以前我们可能会被计 100 小时，现在降到了四五个小时。我认为这种事用今天的模型就能做到，但我认为每家律所和每位未来的律师都该思考：如果进步继续，如果这些模型在律师工作的大部分环节都远超人类，会发生什么。

And I'll just use an illustrative example from just the past month. So I'm starting this company now, and as part of business formation, you need a lawyer to help you go through all your paperwork and your LLC and your employment agreements. And I think three years ago we would have probably paid someone a lot of money to do that. We still hired a lawyer, but what we were able to deliver them is something that basically just needed them to review, because Claude did all of the drafting and reviewed all the employment agreements and did all the redlining. And so maybe we would have gotten billed for 100 hours before, and we brought it down to like four or five hours. And so I think that kind of thing is possible with today's models, but I think every firm and every future lawyer should be thinking about what happens if the progress continues and if these things are vastly superhuman at most parts of the job of lawyering.

我补充两件小事。一是关于律师这边，对吧？就连“哦，我们在法律上用了它”这种说法，其实也已经落后了，对吧？我跟一个朋友聊过，他经营一家很大的风投支持的公司，他正处在一堆关于董事会席位的邮件中，他的 AI 智能体今天早上收到一封邮件时说——前几天说：“提醒一下，你收到这份合同的修订版，但五个改动里只标了三个，另外两个是实质性的，我已经为你预起草了回复材料，让你去找他们争取一次性改好，也发给你律所了。”就是这种主动性是下一步，对吧？你甚至不知道该问什么，它在替你问，对吧？

And I'll add two quick things. One is on the lawyer side, right? Even this, like, oh, we used it for law, is sort of already lagging, right? So I was talking to a friend of mine who runs a very large venture-backed company, and he's in the middle of a bunch of emails about a board position seat, and his AI agent, when he got an email this morning, said, you know, the other day said, "Just as a heads-up, you got a revised version of this contract in, but only three of the five changes are marked, and two of the others are material, and I pre-drafted material for you to go back to them trying to get this in one to your law firm about it." Like, it's a proactivity piece that's next, right? Of like, you don't even know what to ask. It's asking for you, right?

然后在数学这边，我觉得值得关注。那条曲线真的很有意思，因为数学领域发生的不只是从“数学很差”到“数学很好”的颠覆。今年一月，陶哲轩和其他人还在说，这是个很棒的助手，不如人类，但非常非常有用。而现在出现了大规模绝望。我跟数学家聊过数学意味着什么，对吧？部分在于——人类数学——其实对很多人来说，证明并不是重点。一位顶尖学校的数学家告诉我，我们意识到这其实只是一个审美领域。只要我们继续做审美工作，谁在乎呢？另一个人说：“看，我们从这些证明里学不到任何东西，因为证明本来是为了开辟数学新领域。我们就是看不懂它们。它们是外星证明，被递到我们手上。”所以这不仅是颠覆，它还会导致我们做事方式的重构。所以我认为不只是法律或任何其他领域可能受到威胁，你还将被迫重新定义这个领域的意义，对吧？其中多少重要部分是审阅保密协议，多少是咨询、高端咨询角色，多少是更大规模的问题。我认为这里有机会重建“什么是人的部分”，而我担心我们会等到被当头一棒才去做，而这其实是现在就可以开始做的。

And then on the math side, I think it's worth watching. That curve is really interesting, because what's happened in math is not just the disruption from it was bad at math to it's good at math. But in January, Terry Tao and everyone else was saying about it, how this is a great assistant, is not as good as a human, but really, really useful. And now there is like large-scale despair. I talked to mathematicians about what math means, right? And part of this—by human math, there's actually—turns out proofs weren't the point for a lot of us. And I've had a mathematician at a top school tell me, it's actually just an aesthetic field, is what we realized. And as long as we keep doing the aesthetic work, who cares? Another person said, "Look, we're not learning anything from these proofs, because the proofs were all about giving new areas of mathematics. We just don't understand them. They're alien proofs being handed to us." So it's not just disruption, it then causes a reconfiguration of what we do. So I think it's not just that law or any other field might be under threat, but also you're going to be forced to reconfigure what the meaning of the field is, right? How much of it was importantly reviewing an NDA, how much of it was in an advisory, higher-end advisory role, how much of it was a larger-scale issue. I think there's an opportunity to reconstruct what's human here, and I worry we'll wait until we get smacked on the head till we do that, when it's something we can start doing now.

所以，我想就这一点追问一下。数学和人们经常提到的另一个被 AI 革命的领域——软件工程，对吧？这些任务高度分析性，不一定涉及很多人的输入，对吧？尤其以数学为例。法律和其他领域有显著的人性成分。律师的角色是什么？一部分是理解客户的问题，并能够为此调动合适的资源。

So, I guess just to press on that bit. So, the math and the other domain that people frequently throw out as having been revolutionized by AI—software engineering, right? These are tasks that are highly analytical, don't necessarily involve a lot of human input, right? Especially I guess with the math example. Law and other fields have a significant human component to it. What's the role of the lawyer? Part of it is to understand the client problem and be able to bring to bear the right resources for that.

Ethan，回到你的观点，问题是人们是否在用这个作为借口来避免改变。也许他们需要重新思考自己的工作是什么。但这不正说明了问题吗？如果这需要人们组织方式的重大改变，那难道不需要一些时间吗？比解决一个千禧年问题所需的时间更长？

And Ethan, to your point, the question of whether people are using this as an excuse to avoid changing. And maybe they need to rethink what their job is. But doesn't that illustrate the point? If it requires a significant change in the way that people organize themselves, isn't that going to take some time, more time say than it would take to solve a millennial problem?

是的，所以正如我之前所说，我们地球上已经有一种超级智能的形式，那就是组织，对吧？斯坦福法学院能做到没有人能做到的事情。而我们没有好的系统来将 AI 的智能和人的智能融合在一起。我们实际上完全缺失了所谓的“多人”层面，即如何让多个人很好地使用一个 AI 系统？解决方案正在出现，对吧？但我认为预期会有一些延迟是合理的。但我发现人们在指望这个延迟，对吧？不是说延迟是因为我们正在解决问题，而是我们有时间，所以我们现在不必担心。对我来说，这种态度最让我紧张。就像我们在商学院研究柯达。柯达并不是不知道数码相机即将到来，对吧？只是他们觉得无法足够快地反应。他们走错了路。而我想做的是，警钟正在敲响，而我不断听到，无意冒犯你，但根据你所说的，这需要一些时间。那么你是那个为新时代付出代价的人吗？如果不是，你就是在等待事情突然降临到你头上。我经常用的类比是，组织实际上并不需要像大多数人想象的那么多时间。亨利·福特在 1911 年开始开发装配线，到 1914 年就在他的工厂中推广了。不是所有事情都需要永远。组织形式的转变很难，但人们正在这样做。我确实担心我们把这段时间当作“谢天谢地我们有时间”而不是“天哪，现在是 1912 年，我们该怎么办？”

I mean, yes, so we already, as I said before, we already have a form of superintelligence on the planet, which is organization, right? Stanford Law School could do things that no human could do. And we don't have good systems to fit the intelligence of AIs and the intelligence of people together. We actually are missing entirely this level of what we call multiplayer, which is how do you have more than one person use an AI system well? Solutions are coming, right? But I think that it's reasonable to expect some delay. But I find people counting on that delay, right? Like not saying the delay is because we're figuring something out. It's we've got time. So we don't have to worry about it right now. And that to me is the attitude that makes me most nervous. It's like, you know, we study Kodak at business schools. It's not like Kodak didn't know digital cameras were coming, right? It was just they felt they couldn't react fast enough. They went down the wrong rabbit hole. And what I just want to do is like the alert bells are ringing and I keep hearing, no offense to you, but from what you're saying, it's like it'll take some time. So are you the person who's paying the new world? If not, you're waiting for the moment for things to sneak up on you. I mean, the analogy I often use is like organizations don't actually take as much time as most people think. Henry Ford started developing the assembly line in 1911, I think, and by 1914 it rolled out across his plants. Not everything takes forever. And organizational forms are hard to transition, but people are doing that. And I do worry that we're using this time as like a thank god we've got this time rather than a oh god we have a it's 1912. What do we do?

嗯，我很好奇想问你们两位一个问题，就是你们的变革理论。因为我经常听到人们问关于这种组织转型的问题。对于那些行动缓慢的大公司，你该怎么办？也许他们有政策，只允许使用某些东西，因为其他一切太危险。我的模型实际上不是大组织会改变，而是他们会消亡。因为湾区某处有一个 10 人的小公司，正在做另一个地方雇用 1 万人做的活儿。实际上每个垂直领域都有这样一整代人，比如如果你是 State Farm，就会有一个 10 人的保险初创公司，带着一堆 AI，就要吃掉你的午餐，因为他们的成本结构和利润率好得多，而且他们没有这么多人力成本，那才是变革的方向。不是 State Farm 在 AI 方面振作起来，而是这个小得多、灵活得多的 AI 编辑每年夺取市场份额，直到它死掉。

Well, I'm curious to ask a question of both of you which is like your theory of change. Cuz I often hear people asking about this organizational transformation. What do you do about the big lumbering companies that are slow to change and maybe they have policies and they only are allowed to use certain things cuz everything else is too dangerous. My model is actually not that the big organizations change. It's that they die because there's like a small firm somewhere in the Bay Area that has 10 people that is doing the work that another firm somewhere else is employing 10,000 people to do. And that there's actually just this whole generation in every vertical like for if you're a State Farm, there's like a 10-person insurance startup with a bunch of AI that's just going to eat your lunch cuz their cost structure and their margins are so much better and they don't have all this labor cost and that that's actually the direction of change. It's not that State Farm kind of gets its act together in AI, it's that there's this much smaller, much more nimble AI editor that takes market share every year until it's dead.

实际上，我是创业学教授，对吧？这是我的本行。我实际上认为这并不完全正确，或者说这不是唯一的情况。首先，因为资产是 LLM 本身，对吧？初创公司不再拥有你那样的特殊资产，对吧？AI 法律公司也将不得不使用 ChatGPT，对吧？他们不是拥有相同的模型吗？没有特殊优势，微调帮助不大，窄模型帮助不大。所以这创造了一个有趣的竞争环境，专业知识也有价值。我看到的是非常有趣的事情。我看到公司里发生了一系列变化。有些在失败，有些很有趣。最有趣的是 C 级高管再次掌控他们的公司。所以，我现在在一家百年制造公司看到了这一点。我在一些上市后失去控制的初创公司看到了这一点。C 级高管能够访问开放调用，访问所有数据和系统，实际上再次在公司里做出改变。因为原因是，嗯，对。我的意思是，但事实证明，当你遇到组织高层的人时，他们很聪明，很有动力。但他们也必须戴着手套工作，穿过 12 层等级，获得部分信息，整天开会。我震惊的是，我和一家银行的首席技术官交谈，那是世界上最大的银行之一。他周末清除了一堆东西，实际上完全推出了一条新的十亿美元产品线，并和 CEO 谈了，但他的团队不会做。但他正在做。我和管理挪威主权财富基金的人谈过，那是地球上最大的资金池。他个人克服了所有风险经理。现在他组织中有 50% 的人在编码，尽管只有 20% 是程序员，因为他认为这很重要。所以，有这种有趣的其他角度，看起来不同，即大公司。创业学的一个问题是大公司本不该赢，对吧？他们拥有所有资产，所有人员，所有资金。他们中的一些可能正在觉醒。我不认为所有人都会，但我怀疑初创公司是否还有过去的优势。我作为一个经历过前几次技术革命的创业人这么说。

I actually so I'm a professor of entrepreneurship, right? This is my story. I actually think that's not 100% right here or rather it's not the only one. I mean first of all because the asset is the LLM itself, right? The startups don't have the special assets that you anymore, right? The AI legal firms are going to have to be using ChatGPT also, right? They're not like they have the same models, there's no special advantage, fine-tuning doesn't help you that much, narrow models don't help you that much. And so that creates this interesting playing field where also expertise has value. What I'm seeing is really interesting. I'm seeing a bunch of changes in companies. Some are failing, some are interesting. The most interesting has been C-level people taking control of their companies again. So, I've now seen this at a like a 100-year-old manufacturing company. I've seen this at a bunch of startups that went public and sort of lost control. The C-level people with access to an open call, with access to all their data and their system, are actually making changes in their company again. Because the reason the Well, right. I mean, but it turns out the people when you meet people at the top of organizations, they're smart. They're motivated. But, they also have to work with gloves on through 12 levels of hierarchy, get partial information, spend their day in meetings. And I'm shocked at how many of them I was talking to the CTO of a bank one of the biggest branches of banks in the world. And he wiped out a bunch of stuff this weekend that is actually completely launched a new billion-dollar product line and talked to the CEO about it but his team won't do it. But, he is doing it. I talked to the guy who runs the Norwegian sovereign wealth fund, the biggest pool of money on the planet. Like, personally overcame all of his risk managers. And now 50% of his organization is coding even though 20% are coders because he thought this is important. So, there's this kind of interesting other angle that looks different than other things, which is you know, large companies One of the problems of entrepreneurship is large companies should never have won, right? They have all the assets are in the large they're always win, right? They have all the assets, all the people, all the money. And some of them might be waking up. I don't think all of them will be, but I wonder if startups have the same advantage they did. And I say this as a startup guy in previous technological revolution.

是的，你怎么看？会不会有一个 10 人的小公司使用 AI 颠覆整个公司？

Yeah, what do you think? Is there a 10-person per bath using AI that's going to disrupt the whole firm?

嗯，我是说，有这些

Well, I mean, there are these

你知道，所谓的 AI 原生公司已经出现，就在那里。可能这里的人熟悉其中一些。不过我想我同意 Ethan 的观点。就像，如果存在软件优势，那真的很难捍卫，对吧？因为这项技术在某种程度上如此民主，每个人都能接触到。所以，他们没有我们无法轻易获得的特殊酱料。

you know, quote-unquote AI native firms that have started up that are out there. Probably people here are familiar with some of them. I think I agree with Ethan's point though. Like, to the extent that there's a software advantage there, it's like really tough one to defend, right? Because this because the technology is so democratic in a way. So, accessible to everybody. And so, they don't have special sauce that we can't easily that we have can't easily get our hands on.

而且只要像 Ethan 说的那样，我们没有在方向盘前睡着，没有变得太自满，对吧？那我们就应该能把这执行好，同时还带着品牌、专业能力和人脉上的所有优势。你们是吗？

And as long as we're, as Ethan said, not asleep at the wheel, not getting too complacent, right? Then we should be able to execute on this, but with all the advantages of brand and expertise and relationships. Are you?

是的，我们是。

Yes, we are.

你们真的是？好，行，很好。好，我们来聊聊 AI 风险这个话题。所以其中一个论点——还是回到你的书，Kevin——似乎大家对 AI 风险有普遍的认知，人们也在谈论它带来的影响之类的，对吧？但关于 AI 创新和公开发布，主流理论或者说其中一种理论是：如果要应对 AI 风险，最好的办法之一就是把新模型逐步释放到公众中，这样我们能从中学习，对吧？它们不会完美，会有些问题，但通过发布，我们能看到问题是什么，能够做出回应，这大概是我们能采取的最好的风险缓解策略。你怎么看？

You are actually? Okay, right, good. All right, let's talk about AI risk track here for a second. So one of the arguments — again, going back to your book, Kevin — there seems to be a general awareness about AI risk, and people talk about it doing and all of that, right? But the prevailing theory, or one of the theories out there around AI innovation and public releases, is: if we're going to deal with AI risk, one of the best ways to do that is to incrementally release new models out into the public so we can learn from them, right? They're not going to be perfect, there's going to be some problems, but by releasing them, we're going to see what those problems are, we're going to be able to respond, and that's sort of the best risk mitigation strategy that we can follow. What do you make of that?

是的，这有时被称为迭代部署。快速迭代部署，我觉得这个词是 OpenAI 提出来的。因为我认为他们看到了 ChatGPT 发布后发生的事——基本上就是世界被震惊了，因为他们过去五年对 AI 的进展毫无接触，突然之间他们能看到 ChatGPT、亲自试用、在自己脑子里建模风险、开始思考监管。他们当时的理论是，这样做是好事——以稳定的节奏一点点放出模型，这样世界就不会再那么震惊了。但我认为我们正在远离那种做法。实际上，实验室为自己员工保留的最好的内部模型和公开可用的最好的模型之间的差距正在扩大。部分原因是 Fable 和 Mythos 那边发生的事，以及政府某种程度上迫使公司撤回模型或暂缓发布。但也是因为公司从它们最强的模型那里获得了太多内部加速，你要是发布出去，就等于把这一切送给所有竞争对手，那简直是疯了。最近有一些关于这些公司里 AI 完成的工作量的统计数据出来了。Anthropic 最近的一篇博客文章说，他们 26% 的 AI 研发项目现在由 Claude 主导。所以 Claude 不是结对程序员或编码助手，它在主导项目。而且它在加速他们。模型出来的频率比过去高得多。所以我认为这就是终结迭代部署的原因。不会是某些监管之类的，而只是实验室把最好的东西留给自己。

Yeah, this is sometimes called iterative deployment. Rapid iterative deployment, which I think was a term that OpenAI kind of came up with. Because I think they saw what happened after ChatGPT — basically that the world was shocked because they had had no contact with the last five years of AI progress that had been happening, and suddenly they were able to see ChatGPT, try it for themselves, model the risks in their own head, start thinking about regulation. And their theory at the time was that this was a good thing to do — to drip out releases at a steady cadence so that the world was never that shocked again. But I think we're moving away from that. I think actually the gap between the best internal models that the labs have for their own employees and the best available public models is widening. And partly that's because of what has been happening with Fable and Mythos and the government sort of forcing companies to pull back models or hold off on releasing them. But it's also because the companies are getting so much internal acceleration from their most capable models that you would be crazy to give that to all of your competitors by releasing. So there's some statistics that have come out recently about the amount of work being done at these companies by AI. An Anthropic blog post recently said that 26% of their AI research and development projects are now led by Claude. So Claude is not like a pair programmer or coding assistant. It is leading the project. And it's speeding them up. The models are coming out much more frequently than they used to. So I think that's what's going to end iterative deployment. It will not be some regulation or something. It will just be the labs kind of keeping their best stuff for themselves.

很好，谢谢你——我同意你说的所有内容。我的意思是，问题在于——我只想补充一点，这有点让人困惑，我也想知道你是否同意，因为我最近在脑子里把这些实验室建模成两类不同的人，对吧？一类是你书里写的人，实验室的领导者，通常是参与其中的研究人员、技术人员，从一开始就为 AGI 而建，深深担忧着一些事情。他们混合了不断辞职抗议的安全研究人员，以及建造机器之神的人。然后是普通创业公司的其余部分。大量从 Shopify、Stripe 以及你听说过的每一家创业公司招来的人，还有高速增长的所有问题，对吧？巨大的政治斗争——我的意思是，我跟迪士尼的一个人聊过，就在 OpenAI 放弃一笔十亿美元交易的第二天，他们目瞪口呆。“你不会做那种事。”但他们在随意改变方向。而当你和一家向你推销的公司交谈时——当你和 Anthropic 的销售代表交谈时，你可能会碰到一个其实并不相信 AI 有价值的人。我认为游说者和其他组织并不真正以同样的方式相信 AI。所以我很好奇，从你的角度看：他们既是一家普通的超高速增长公司，有着钱太多、人才不够、没有方向的公司所有那些古怪问题，CEO 们是奇怪的梦想家或阴谋家，又有你经常写的那种核心，这两者是如何相互作用的？

Great. Thank you for — so I agree with everything you're saying here. I mean, I think the question is — the one thing I would just add is it's kind of a little bit confusing, and I'd be interested if you agree with this too, because I've been mentally modeling the labs recently as two different groups of people, right? There's the people you write about in your book, who are the leaders of the labs, generally the researchers involved, members of technical staff, AGI built since the beginning, deeply worried about things. They're a mix of safety researchers who keep quitting in protest and then the people building the machine god. And then there is all the rest of a normal startup company. Just tremendous amounts of people hired from Shopify and Stripe and every other startup you've heard of, and every problem of hyper growth, right? Giant politics — I mean, I talked to a person at Disney the day after OpenAI walked away from a billion-dollar deal and they were flabbergasted. "You just don't do those things." But they're changing direction randomly. And when you talk to a company that's selling you — when you talk to an Anthropic sales rep, you might get a random person who doesn't actually believe in AI being valuable. And I think the lobbyists and other organizations don't really believe in AI in the same kind of way. So I'm curious from your perspective: how is that interplay of the fact that they're a normal hyper-growth company with all the screwball problems of a company that has too much money and not enough talent and no direction, and the CEOs are weird dreamers or schemers, and the sort of core that you write about a lot?

是的，肯定存在文化冲突。我记得 OpenAI 有人告诉我，当那些能与人眼神交流的人到来时，他们就知道公司进入了一个新时代。就像麦肯锡的那些人突然出现，一下子他们穿着得体、表现得正常，他们就说，哦，现在这是另一个地方了。我觉得这大概有点夸张。但没错，你说得对。这些公司现在就是有普通的商务人士，因为它们要上市，要成为世界上最大的公司之一。而且它们的内部协调往往是个问题，你能在各种方面看到这一点。但没错，我对研究核心感兴趣，因为那最终是决定这些企业成败的东西。Google 在书里是个非常有趣的研究案例，就是一家同时包含两个阵营强大成分的公司。你有非常 AGI 导向的研究人员，想建造机器之神的人，然后你有那些精打细算的人，盯着搜索广告收入，尽量不去破坏它，还有，哦，也许我们应该通过向 Anthropic 卖一些算力来增长云业务，尽管他们在 LLM 上是我们的竞争对手。所以我认为它们是一家被两个截然不同方向拉扯的公司最清晰、最成熟的例子。

Yeah, there's definitely a culture clash. I remember someone at OpenAI telling me that they knew the company had entered a new era when the people who could make eye contact arrived. Like all these people from McKinsey just showed up and all of a sudden they were well dressed and presented themselves normally, and they were like, oh, this is a different place now. I think that's probably a little overstated. But yeah, you're right. There are just normal business people at these companies now because they are going to go public and be some of the largest companies in the world. And often their internal coordination is a problem, and you see that in all kinds of ways. But yeah, I think I'm interested in the research core because that is ultimately the thing that is going to determine the success or failure of these businesses. Google is a really interesting study in the book, just of a company that kind of contains strong components of both camps. Like you have the very AGI-built researchers, the people who want to build the machine god, and then you have the people who are the bean counters who are looking at the search ad revenue and trying not to disrupt it and, oh, maybe we should grow our cloud business by selling some compute to Anthropic even though they're our rival in LLM. So I think they're the most clear, mature example of a company that is being pulled in these two very different directions.

这让我想起——Kevin，你书里有一个画面一直留在我脑海里，就是 Ilya Sutskever 深夜在 OpenAI 的走廊里游荡，对人们发表这些准精神性的宣言，告诉每个人他们需要感受 AGI。

That just reminds me of — one of the images that stuck with me from your book, Kevin, was Ilya Sutskever wandering the halls of OpenAI late at night, making these quasi-spiritual pronouncements to people and telling everybody that they needed to feel the AGI.

感受 AGI。你必须感受 AGI。是的。如果你感受到了 AGI，请举手。

Feel the AGI. You must feel the AGI. Yes. Raise your hand if you feel the AGI.

所以，我曾经为写书采访过 Ilya，我问他：“你是什么意思？感受到 AGI 是什么感觉？”你得理解——你见过 Ilya 讲话或者见过他本人吗？他是个沉思的、近乎牧师般的人物，喜欢长时间的停顿。所以我问他：“感受到 AGI 是什么感觉？”他望向窗外大约一分半钟，然后回过头说：“你知道那些视频吗？开头是一个人在海滩上，然后镜头拉远，展现出城市、州、国家、地球、太阳系和宇宙？”他说：“就像那样，或者也许像巴赫的《托卡塔与赋格》。

So, I once interviewed Ilya for the book and I asked him like, "What do you mean? What does it feel like to feel the AGI?" And you have to understand—do you ever see Ilya talk or have ever met him? He's a brooding, almost priestly figure and he loves long pauses. So I asked him, "What does it feel like to feel the AGI?" and he looked out the window for about a minute and a half and then he came back and he said, "Do you know those videos where they start with a man on the beach and it zooms out to reveal the city and the state and the country and the earth, the solar system and the cosmos?" He said, "It's like that or maybe it's like Bach's Toccata and Fugue."

这就是感受 AGI 的感觉。所以，总之，如果你想感受 AGI，像 Ilya Sutskever 那样，一边听巴赫的《托卡塔与赋格》，一边看那种拉远镜头的视频，我猜你就能体会他在说什么了。

That's what feeling like the AGI is like. So, anyway, if you want to feel the AGI, like Ilya Sutskever, listen to Bach's Toccata and Fugue while viewing one of those zoom-out videos and I guess you'll get a sense of what he's talking about.

很有帮助。

Very helpful.

谢谢。

Thank you.

就是来帮忙的，是的。

Here to be helpful, yeah.

过去 12 个月，公众对 AI 的怀疑日益增加。现在似乎存在一个差距：一方面是商业投资和乐观情绪的水平，另一方面是公众对这种投资的舒适度。你如何看待这种紧张关系的发展？

The last 12 months have seen public opinion increasingly skeptical about AI. And there now seems to be a gap between the level of commercial investment and the optimism there and the general public's comfort with that investment. How do you see that tension playing out?

有几件事。有趣的是——当你让人们举手，问有多少人在使用 AI？我们在芝加哥大学有一项很好的研究显示，大约 60% 的人会说他们在学校使用 AI。但如果你让他们指一个随机的人，问那个人使用 AI 的概率有多大？超过 90%，对吧？所以存在一堆奇怪的社会污名。我的意思是，没那么奇怪，对吧？就像一切——AI 就是一切，对吧？AI 实验室的领导者们迅速告诉你，我们可能会杀了你，但肯定会抢走你的工作。这可不妙，对吧？有一堆——从关于能源的真实问题到 TikTok 上关于水供应的不真实问题，都是担忧，对吧？还有权力集中、政府与军方和 AI 之间的联系。有很多理由对此感到不祥。我怀疑，因为我和很多人谈论过实际的 AI 使用，我惊讶于即使在非常敌对的群体中，他们多么想了解 AI 使用，正在发生的是——这有点像当你调查人们对国会的看法，他们讨厌国会，但喜欢他们的议员。我觉得很多人有同样的感觉，就像：是的，但你可以从我冰冷的手里撬走我的 ChatGPT，但我讨厌 AI，对吧？我认为在这种地方，这是一种非常常见的结果。我认为这很复杂，对吧？部分原因是这是一种可怕的技术。我认为，关注是否——在某种程度上，关注所有存在性风险不仅让它更可怕，还剥夺了很多讨论短期利弊的能力，即所有可能影响我们的方式。所以我认为这很有趣，对吧？经济将依赖于此。这个国家目前对 AI 特别敌对。但我想知道，这种敌意有多少是真正的敌意，多少是焦虑，多少是出于社会规范而表达的偏好。就像如果你问公司里的人，没人会说他们在使用 AI，因为 AI 让他们看起来像天才。我们从研究中知道，如果你说使用 AI 帮助写作，你会受到污名化。所以如果你不 A 用 AI 帮助写作，B 不告诉任何人，那就太傻了。所以每个人都在秘密地感觉自己是个冒充者。你知道，这也会产生奇怪的效果。所以既有真正的焦虑和真正的怨恨，也有大量的人一直感到内疚，并试图在周围所有人中隐藏。我认为这种对比是真实的。我觉得这里的人看起来不舒服。我不是说你。是的，请讲。

There's a few things. One of the interesting—when you ask people to raise their hand about how many people are using AI? We have a nice study at the University of Chicago that shows that like 60% of people will say they use AI in school. And if you ask them to point at a random person, what's the chance that person is using AI? It's over 90%, right? So there's a whole bunch of weird social stigmas. I mean not that weird, right? Like everything—AI is everything, right? The AI lab leaders are rapidly telling you we may kill you, but we'll definitely take your job. Like not great, right? There's a bunch of—ranging from real stuff about energy to TikToky not real stuff about water supplies that are concerns, right? There is concentration of power, the link between government and military and AI. There's a lot of reasons to feel ominous about this. I suspect, because I talk to lots of people about actual AI use, I'm amazed at how much they just want to learn about AI use even in really hostile crowds, that what's happening is there's—it's sort of like when you survey people about Congress, they hate Congress, but they like their congressman. I feel like a lot of people have the same kind of feeling, which is like yeah, but you can pry my ChatGPT from my cold dead hands, but like I hate AI, right? And I think that's a really common kind of outcome in these sorts of places. And I think it's complicated, right? Part of this is like this is a scary technology. I think that focusing on whether—and in some ways focusing on all the existential risk not only makes it scarier, but also removes a lot of the ability to talk about the pluses and minuses in the short term of all the ways that this might affect us. So I think that it's just interesting, right? Like the economy's going to depend on this. This country is uniquely hostile to AI right now. And I wonder how much of the hostility is actual hostility versus anxiety versus, you know, stated preference because of social norms. Sort of like if you ask people at a company, no one's going to say they're using AI because the AI makes them look like a genius. And we know from studies that you're stigmatized if you say you use AI to help you write something. So you'd be pretty dumb to A not get AI to help you write something, right? And then B to not tell anyone about that. So everyone is secretly feeling a poser all the time. And you know, that creates also weird effects here. So there's both a real anxiety and real resentment and just a huge amount of people who are just feeling guilty all the time and trying to hide it among everyone around them. And I think that that contrast is real. I think people look uncomfortable here. I didn't mean you. Yes, go ahead.

好吧，从你自己开始，Ethan。你使用 AI 为你写作的程度如何？以你名义发表的内容中，有多少词是 AI 生成的？

Well, start with yourself, Ethan. How much do you use AI to write for you? How many of the words that go out under your name are generated by AI?

这确实是个好问题。我很幸运早期就决定，如果我不写——所以我说的一件事是创新。事实证明，如果你在小组中产生想法，

It's actually a really good question. And I was lucky enough to decide early on that I would—if I didn't write it—so one of the things I say is innovation. It turns out if you generate ideas in a group,

注意他没有回答。

Notice him not answering.

顺便说一句，答案可能是我写的词中大约 18%。大致如此。在 Substack 帖子中更少，但总体可能 18%。基于我所做的，对吧？所以，但重要的是——顺序对我很重要，我想这就是我在说的。比如初稿，我学到如果我让 AI 为我写任何初稿，我就完蛋了。之后我什么都想不出来，而且不管我喜欢与否，写作都会是 AI 的。所以我自己写初稿，然后我与系统来回工作——我需要你充实内容。但这是有意识的选择。顺便说一句，我很幸运做出了那个选择，因为我是这些工具的早期采用者，去年 12 月 Bard 似乎是个很好的作家，当时 Pangram 还不太管用，没人习惯到处都是 pod 写作。所以有很多人转向并说我在用 AI 写作，现在你可以回顾过去说，哦实际上你过去 6 个月写的一切都是 AI。所以我可以自信地说我的数字更低，你知道，20 低于 20% 我认为是

The answer, by the way, is probably about 18% of the words I write. Give or take. Less on Substack posts, but probably 18% overall. And based on what I do, right? So, but it matters—the order matters to me is I guess what I'm saying. Like first draft, I learned if I let the AI do any first draft written for me, I am screwed. Like I can't think of anything afterwards and the writing will be AI whether I like it or not. So I write my own first draft and then I work back and forth with systems—I need you to flesh things out. But like this is a conscious choice. And by the way, I got very lucky that I made that choice because I was an early adopter of these things and Bard seemed like a perfectly good writer back in December when Pangram didn't really work and no one had gotten used to pod writing everywhere. So there's a lot of people who like switched over and like I'm using AI for writing and now you can retrospectively go back and be like, oh actually everything you've written for the last 6 months is AI. So I can say this with confidence that my numbers are lower, you know, 20 below 20% I think is the

有人把你的书通过 Pangram 跑过吗？

Has anyone run your book through Pangram yet?

如果你把我的书通过 Pangram 跑，你可能会——你会得到大约 8% 的书是 AI 在说话，我会说可能还有 10% 的词可能是 AI 写的或由我重写的。就像我对此很清楚——比如如果我卡在一个句子上，过去需要我一周时间走来走去，我会说给我——最糟糕的部分是那 20% 10% 是 AI 写的，人们会说那是个很好的措辞，我会说那是我在这整篇该死的帖子中唯一没写的东西就是这一句，每个人都说你说了一句精彩的话，我会说哦。

If you run my book through Pangram you'll probably—you will get there's probably about 8% of the book that is like AI talking and I would say there's probably another 10% of the words that are probably AI written or with me rewriting them. And like I'm clear about this—like if I'm stuck on a sentence it used to take me a week of like walking around and I'm like give me the—like the worst part is the 20 10% is AI writing that's what people are like that was a great turn of phrase and I'm like that was the only thing I didn't write in this whole damn post was this one sentence and everyone's like that's a brilliant sentence you said and I'm like oh.

嗯，我是说，我不觉得这么做有什么问题，但我觉得——我不想读 Pod 的观点。这很重要。你知道，Pod 的观点——听久了你会想，天哪，我们又要聊意识了，对吧？

Yeah, I mean, I don't think there's anything wrong with doing it, but I think—I don't want to read Pod's opinions. That's a big deal. And you know, Pod's opinions—after you hear it for a while you're like, oh god, we're going to talk about consciousness again, aren't we?

是啊，这——我鼓掌。顺便说一句，我们不会因为 Ethan 用了 18% 的 AI 就惩罚他。我们奖励诚实的人。所以谢谢你坦白，谢谢你当个“垃圾手榴弹”，不管他们怎么叫你。才 18% 而已。不过 Ethan，我真的很喜欢你这本书的一点是，它的方式——

Yeah, it's—I applaud, and by the way, we're not punishing Ethan for his 18% AI use. We reward people who are honest. So thank you for fessing up, for being a slop grenade, whatever they call you. Only 18% of one. One of the things that I really like about your book though, Ethan, is the way that it's—

我们正在——

We're having—

你继续说。

Go ahead.

我很想收下这个夸奖，但我想问——Kevin，你写的东西里有多少是 AI 写的？

I would love the compliment, but I just want—Kevin, how much of what you write is AI?

零。

None.

谁说零？Kevin，对吧？

Who said none? Kevin, right?

这要看平台。书是 0%，邮件 60%，LinkedIn 消息 100%。

It varies by platform. So book 0%, email 60%, LinkedIn messages 100%.

Pod 会回复所有这些。所以如果你曾经给我发邮件或在 LinkedIn 上给我发消息，那不是我回的。

Pod responds to all of those. So if you've ever emailed or messaged me on LinkedIn, that was not me.

是啊。不过我觉得这是个很有意思的选择，你在引言里说，你刻意没有让 AI 产出书里的任何文字——只把它当研究工具用。

Yeah. I thought it was a very interesting choice though, in your introduction you said that you deliberately did not have AI produce any words in it—use it as a research tool.

对，其他一切都用它。但我是怕别人用“全段落生成”来整我，而且我也——我觉得，你知道，算是——我觉得模型还不够好，做不到我想让它们做的事。它们不擅长，比如，把一大堆笔记变成引人入胜的叙事章节。也许这是虚荣。我确信这是暂时的。但眼下我觉得，在我想要的那种写作上，我还是比模型强。而且实际上我觉得模型在过去一两年里写作变差了。

Yeah, use it for everything else. Yeah, but I was scared of people pan gramming me, and I was also—I think, you know, it was sort of—I didn't think the models were good enough yet to do what I wanted them to do. They are not good at taking, for example, a big pile of notes and turning it into a gripping narrative chapter. So maybe that's vanity. I'm sure it's temporary. But for now I feel like I am still better at the kind of writing that I want to do than the models. And actually I feel like the models have gotten worse at writing.

对。在过去一两年里。所以——

Yes. In the past year or two. So—

真有意思。真的，你现在的状态真有意思。不过是啊。

Really interesting. Really, actually really interesting where you're at. But yeah.

不过回到我对你的夸奖，Ethan，我很喜欢这本书的一点是，你在写作过程中和 AI 玩，对吧？就像你写一段，然后 AI 帮你写完这段，或者对你刚写的东西发表评论。于是你能——作为一个——你能实时看到 AI 如何——可能更好或更糟地影响写作。

And but going back to my compliment of you, Ethan, one of the things I really liked about the book was the way that you're playing with the AI as you're writing it, right? So it's like you're writing a paragraph and then the AI finishes the paragraph for you or comments on something that you've just written. And so you get to sort of—as a—you get to sort of see it playing out in real time how the AI could—could be impacted, better or worse, the writing.

是啊，我是说，在写这本书的过程中，我确信你也发现了同样的事——情况变化很大，对吧？所以刚开始写书时，我处在和你类似的位置，就是 AI 写作没什么价值。到后来，它有好点子了，我把一章丢给它，它开始做那种只有高端编辑才会和我做的事，或者像我的同行——你知道，像同僚教授那样，说其实这里有个结构问题，一个深层的结构问题，而且你做的转折其实不合逻辑。我就想，天哪，这真是个好主意。

Yeah, I mean, and during the course of writing the book I'm sure you found the same thing—things change a lot, right? And so when I start the book, right, I'm in the kind of boat you're in, where like this is not—there's no value in the AI writing. By the end, like it has good ideas and I'm like, I run a chapter through it and it's like, you know, it's starting to do things that a high-end editor would only do with me, or like my fellow—you know, like a fellow professor, where it's like actually there's a structural problem, a deep structural problem here, and you know, the pivot you make isn't actually logical. And I'm like, oh crap, that is a really smart idea.

甚至到最后——顺便说一句，另一件不祥的事是，我用 AI 对我的书做事实核查。我不知道你有没有同样的经历，但它不仅没犯错，还发现了一堆问题。但它发现的是我引用来源里的问题，这才是最糟的部分。它说，这篇论文表二不行。我就想，我都不知道该怎么办。就像，我是不是要群发邮件给每一个教授、每一个我引用过的同行，因为他们总能在所有工作里找到瑕疵？我不知道拿这个怎么办。

And even to the extent that at the end—by the way, one of the other things that's ominous is I ran AI through my fact—AI fact-checking your book. I don't know if you have the same experience, but not only did it not make any errors, it found a bunch of issues. But it found issues in the sources I was citing, which was the worst part. It was like, this paper, table two isn't good. I'm like, I don't even know what to do. It's like, do I—I was like, do I mass email every professor, every fellow faculty member I've ever cited, because they always found a flaw in all of their work? And I don't know what to do with that.

方法论很弱。

Methodology is weak.

而且你知道，它指出我和所有编辑、校对者把书里某一章的章节标题都搞错了，而且还在你手上这个版本里。我让你自己去找。谢谢。但你知道，人类在最好的状态下也会犯这些错。所以我觉得，对我来说，一部分是回到更大的问题：你得在一切事情上玩这个。我在第一本书里给的一条建议，我怎么强调都不为过，就是把它用在一切事情上。你得把它分开，你得决定什么是人做的、什么不是，但至少去实验。玩真的很有用。比如现在用 Opus 55 玩游戏，或者让它给你做视频，都是很有意思的体验，能让你洞察 AI 的心智。你会开始理解这种“神经语”语言，或者为什么语言变差了。就是对这东西有一种直觉。而且我们有证据表明，擅长管理、擅长和人打交道的人，也擅长用 AI，因为它有点像人那样运作。AI 的汽车类比是，它更令人惊讶，而且这也是让它起作用的唯一方式，对吧？这就是为什么律师往往很擅长用 AI，因为——

And you know, it pointed out that me and all of the editors and proofreaders had messed up on literally my chapter title in one of the chapters in the book, which is still in the version you have. I will let you find it. Thank you. But like, you know, humans at their best make these mistakes. So I think, you know, to me, part of this is to go back to the bigger question: you've got to play with this for everything. Like the one piece of advice I gave in my first book that I cannot emphasize enough is just use this for everything. And you have to separate it, you have to decide on what's human and what's not, but at least experiment with this. And play is really useful. Like, playing a game now with Opus 55 or having you make a video for you is a really interesting experience that gives you an insight into the mind of the AI. You'll start to understand this kind of neuralese language or why the language has gotten worse. There's just an intuition for this thing. And we've got some evidence that people who are good managers and good with people are good with AI because it kind of works like people. Like the car analogy of AI is that it's more surprising and it's also the only way to make it work, right? Which is why lawyers are often really good at using AI, by—

呃，那么 Kevin，这个问题转向你。什么是——Ethan 刚给了一条。一条具体的建议，给那些在这些不确定时期摸索使用 AI 的人。

Uh so actually Kevin on that, turn to you for the question. What's one thing—Ethan just gave one. One thing that is a concrete piece of advice for people navigating using AI in these uncertain times.

我会说，给 AI 更难的任务。我觉得很多人还困在 2023 年那种心态里，觉得这就是个花哨的 Google，或者花哨的 Westlaw，或者花哨的信息检索系统。我理解。我自己也那样用 AI，但我也得不断提醒自己，它大概能做比我用它的方式更难的事。就像我有一辆法拉利，却拿它去杂货店。所以你得真正踩下油门，给它一件真的很难、你要花一周或一个月才能做完的事，看看它做得怎么样。因为我经常发现，系统的极限只是我自己想象不出我能交出去的任务有多深。

I would say give AI harder tasks. I think a lot of people are still stuck in this very 2023-era mindset of like, this is a fancy Google or a fancy Westlaw or a fancy, you know, information retrieval system. And I understand that. I use AI that way myself, but I also have to constantly remind myself like, this can probably do harder things than I'm using it for. Like I've got this Ferrari and I'm using it to like go to the grocery store. And so you do have to actually like open the throttle, give it something really hard that you would take a week to do or a month to do and see how it does. Because I often find that the limit of the system is just my own inability to imagine the depth of task that I could hand off.

我喜欢。我知道在梯子上加什么。你一旦做到，就再升一级。你宁可撞墙再退回来，也好过从悬垂之下起步。这些系统今天能做的事，哪怕不再发展，和任何人实际用它们做的事之间的差距，比你想象的大得多。我经常发现，那些非常天真但非常有野心的 AI 用户，从这些系统里得到最好的结果。他们就像，你是怎么做出那个 12 部分的大电影的？他们说，我就直接要的。你就想，好吧。看来就是这样。现在这招管用。顺便说一句，如果你在等经验教训——不是要拆我们的台，但你教的一切、学的一切，如果你六个月前学的，那些都不重要了。我们在 Warden 全测过。那些技巧现在一点用都没有。

I love it. I know what to add on the ladder. As soon as you've done that, then escalate one level up. Like you're better off hitting a wall and moving back than you are starting behind the overhang. Between what these systems can do today, even with no more development, and what anyone is doing with them is much larger than you think. And I often find people who are really naive AI users who are really ambitious get the best results out of these systems. They're like, you know, it was like, "Well, how did you get that, you know, 12-part, you know, that giant movie?" They're like, "I just asked for it." You're like, "All right." Well, I guess that's it. That works now. Like and by the way, if you're waiting for lessons—not to undermine us here, but like, everything you teach, learn—thing, if you learned it 6 months ago, none of that matters anymore. We were testing it all at Warden. None of those techniques help you at all anymore.

比如思维链提示和赋予它人格，通常你直接要什么就能得到什么。

Like chain-of-thought prompting and giving it personas, just ask for the thing and you get it generally.

有一个让我遗憾地发现不再奏效的方法是，在 2024 年初有一段时间，如果你提出给模型小费，它们会表现得更好。

One that I was sad to see no longer works was that there was a period in early 2024 where models would perform better if you offered to tip them.

是的。

Yes.

比如对 Claude 说：“如果你把这篇文章写得很好，我就给你 20 美元。”当然你完全是在骗它。

Like Claude, like "I'll give you $20 if you write this essay and it's really good." And of course you're just totally lying to it.

我知道。我会把钱塞进——

I know. I would stick the money in the—

对，没错。

Yeah, that's right.

我的意思是，顺便说一句，威胁也管用。我们喜欢……嗯，顺便说一句，Gemini 坚持得最久。整个 Gemini 3 系列，你仍然可以威胁。嗯，或者或者黑——

I mean and by the way, threats worked. We like um Gemini lasted the longest, by the way. The whole Gemini 3 series, you can still threaten. Um or or black—

不好。这是不好的兆头。因为它们会记住。

Bad. This is bad juju. Cuz they will remember.

没错。好吧，你可以但是——不，没有哪句话像——那么糟糕

That's right. Well, you can but you— No no sentence is bad as—

Claude。

Claude.

不过没有哪句话像你的那么糟糕，所以我没事。对吧。

No sentence is bad as yours, though, so I'm okay. Right.

没错。

That's true.

好的，我们有几分钟提问时间，周围有几个麦克风，那么让我们看看。我看到那边有个问题。如果你能等麦克风，这样大家都能听到，我们也有线上观众。你看到中间穿绿色衣服的那位了吗？是的，你不能从那边上去。你也可以——

All right, we have a few minutes for questions and a couple of microphones around, so let's see. I see a question there. If you could wait for the microphone so everybody can hear here we also have people online. Do you see him in the green in the middle? Yeah, you can't go up that way. You can also—

你可以让你的智能体来提问。

You can have your agents ask questions.

我们这边还有一个麦克风，所以谁——

And we've got another microphone over here, so whoever—

好的。好了，开始吧。然后我们给你一秒钟。很好。

You got it. Okay, there we go. And then we'll go you a second. Great.

你好，你好。那是——哦，谢谢。嗯，非常感谢你们来到这里。嗯，我知道这次演讲是关于 AI 的未来和商业。现在，我觉得我们还没有触及商业方面。很多人说，嗯，这里没有商业，因为似乎，嗯，怎么说呢？大多数前沿实验室的客户都是其他前沿实验室之类的。所以，真正的问题是，不管技术实际上有多好，所有这些的经济、经济资助是什么？

Hello, hello. That's an Oh, thank you. Um thank you so much for coming here. Um I know the talk is about the future and business of AI. Now, I feel like we haven't touched on the business side. A lot of folks have said like uh there's no business here in the sense that like it seems uh what's the word? Most of the frontier lab like customers are other frontier labs or yada yada yada. So, there's this real question of what is the so like economic economy like uh funding of all of this irrespective of how good the technology actually is.

所以，你的意思是 AI 的经济影响是什么？你是这个意思吗？这是一个真实的商业吗？

So, you mean like what's the economic impact of AI? Is that what you're saying? Is this a real business?

是的，有真实的商业案例吗？是的。

Yeah, is there a real business case? Yeah.

你想先来吗，Sam？

You want to go first, Sam?

是的，它们是真实的商业。这些是资本主义历史上增长最快的收入曲线之一。这些公司，你知道，当我 3 年前第一次访问 Anthropic 时，他们的终身收入大约在 1000 万美元左右。现在他们很可能很快就会达到 1000 亿美元的 ARR。所以，是的，他们的客户集中在科技领域，因为那是使用最多 token 的地方。但我认为，认为这一切只是金融魔法、都是庞氏骗局或纸牌屋的想法是一厢情愿。我不认为所有实验室都会盈利，但我预计其中许多，如果今天停止训练新模型，将会非常盈利。推理的利润率相当不错。所以我认为肯定有公司过度扩张，但我认为要让它成为泡沫并全部崩溃，就必须——我认为互联网泡沫和现在的相似之处经常被夸大。在互联网泡沫中，网络在需求出现之前就建好了。1999 年真的没有那么多互联网用户，他们也没有买那么多东西。所以你有了所有这些没有客户的基础设施。我们今天有相反的问题，即所有这些实验室都严重受限于算力。他们的需求远远超过他们能服务的算力。所以我认为这是一组非常不同的基本面，比 1999 年更强的基本面。你怎么看，Ethan？

Yes, they're a real business. These are some of the fastest growing revenue ramps in the history of capitalism. These companies, you know, when I visited Anthropic 3 years ago for the first time, they had made something on the order of $10 million in lifetime revenue. They're now probably going to do 100 billion in ARR very soon. So, yes, they have customers that are concentrated in tech cuz that's what uses the most tokens. But I think the idea that this is all just financial wizardry, that it is all a Ponzi scheme or a house of cards is wishful thinking. I don't think all of the labs will be profitable, but I expect that many of them, if they stop training new models today, would be immensely profitable. The margins on inference are quite good. So I think there are certainly companies that are overextending themselves, but I think for this all to have been a bubble and it all to come crashing down, there would have to be the—the parallels between the dot-com boom and now are often overstated, I think. In the dot-com boom and bubble, the network was built out before the demand was there. There was really no—there just weren't that many internet users in 1999 and they weren't buying that much stuff. And so you had all this infrastructure that had no customers. We have the opposite problem today, which is that all these labs are severely compute constrained. They have much more demand than they have the compute to serve. And so I think that's just a very different set of fundamentals and a stronger set of fundamentals than in 1999. What do you think, Ethan?

是的，我的意思是，我通常会说，我认为这是“这不可能是真的”这种复杂心态的一部分，就是“这是骗局，它会长久吗”。顺便说一句，我不能说不会因为数据中心的融资方式而出现金融恐慌。这是可能的。但认为 AI 会消失、这一切都是幻觉会消失的想法，就是不对。比如谷歌不会——我的意思是，即使 OpenAI 和 Anthropic，顺便说一句，考虑到它们的 ARR 有多大，这似乎极不可能，基本上不可能，因为它们部分由谷歌拥有。谷歌不会消失。微软不会消失。我的意思是，这根本不是那种现实的情况——我认为人们总是把下一层想成“啊，开放权重实验室会接管”。但它们在很大程度上依赖前沿实验室继续存在。而且顺便说一句，前沿实验室实际上可以提供比开放权重实验室更便宜的模型，因为它们有更多算力，并且正在研究让你的 fable 模型与你的 haiku 模型对话、协同工作以提供更便宜产品的方法。所以我的意思是，这里有很多不同竞争性商业模型的空间。但我的意思是，与你将获得的 token 使用量相比，我认为我们才刚刚开始看到 token 使用量，对吧？我只是不认为——可能会有一些管理不善。有人倒闭，某个时候可能会有金融恐慌。这完全可能。我不能告诉你不会，但我认为很多那个问题的背后是“这 somehow 不是真的”的想法。而它非常真实。我认为这是需要传达的主要点。我不认为这些公司是假的或在伪造数字。你知道，如果有通常的 IPO 压力要做事情，但我同意，推理是盈利的，但它们的主要目标是建造机器之神，这很昂贵。

Yeah, I mean I would generally say I think it's part of this complex of like this can't be real that kind of comes up, which is like this is a sham, it's going to be here. And like by the way, I can't say there won't be a financial panic over how data centers are funded. Like it's possible. But the idea that AI is kind of go like but it gets paired with this idea AI is going to go away. This is all an illusion that's going to disappear. Just not true. Like Google's not going—I mean even if the open AI and Anthropic, by the way, which seems extremely unlikely basically unlikely given that how large their ARRs are, like they're owned in part by Google. Like Google's not going anywhere. Microsoft's not going anywhere. I mean it's just not a situation where that's a realistic—I think people always then the next level is like ah the open weights labs will take things over. But they in large part depend on the frontier labs to continue. And by the way, the frontier labs can actually serve cheaper models than the open weights labs cuz they have more compute and they're working out ways that your fable model will talk to your haiku model and they'll work together to deliver cheaper product. So I mean there's room for lots of different competing business models here. But I mean we're barely begun to see token usage, I think, compared to how much token usage you're going to get, right? And I just don't like—there may be some mismanagement. Someone goes under, there might be a financial panic at some period of time. It's entirely possible. I cannot tell you it won't, but I think underlying a lot of that question is the idea that like somehow this isn't real. And it's like it's very real. And I think that's the main thing to kind of get across. I don't think these companies are fake or faking their numbers. You know, if there's the usual kind of IPO pressures to do stuff, but I agree, profitable on inference, but their main goal is to build the machine god, which is expensive.

所以你谈到了 Aston 的一点，现在它可能在数学和计算机科学方面取得很多进展，但最终它会影响到每个人。所以想必要么会出现大规模失业，要么会出现大规模生产力提升。为什么我们没有看到实际利率更高？为什么市场没有定价这一点？

So you talked about the bit of Aston and right now it might be making a lot of advances in maths and computer science, but eventually it's going to come for everyone. And so presumably either there's going to be mass unemployment or mass productivity gains. Why don't we see real interest rates being higher? Why isn't that priced in the market?

这是个好问题，对吧？我认为大多数人就是不相信这些事情真的在发生，对吧？如果我们扫视一下人们——你问的问题，我认为你非常关注 AI。即使是我，尽管我向你抛出了所有这些限定词，我认为我们所有人都像是——你知道，奇点的真正含义是那个你无法看到预测过去的点。所以每个人都在预测正常的事情，对吧？我们在头脑中对此打折，就像，是的，这是浮云，它不会真正改变事情。

It's a good question, right? I think that most people just don't believe this stuff is really really happening, right? And we like if you kind of take a sweep of what people—the questions you were asking, I think you're very AI focused. And even me with all the qualifiers I'm throwing at you, I think all of us are like just we haven't had a break from—you know, the true meaning of a singularity, right, is that point you can't see a prediction past. So everyone is predicting normal stuff, right? We are discounting this in our head to like, yeah, this is fluff and it's not really going to change things.

模型会不断变好，但到某个点会趋于平缓。我们可能无法采用它们，或者类似情况。所以我们所有的心理赌注都是：这不会真的发生，对吧？他们可能是对的，但凯文书里的整个故事是，那些认为这真的是大事、这是历史这一阶段终结的人，到目前为止都是对的。他们不一定对，但我觉得没有其他人真的相信。所以你可以在这方面做逆向押注，对吧？但我可以告诉你，有很多理由说明为什么它还没有以这种方式触及每个人。仍然因为你说的原因，卡在某个阶段。所以我们可以说，是十年后、十五年后会改变，还是两三年后？我不知道你怎么看。

The models will keep getting better, but they're going to even off at some point. We won't be able to adopt them or some like So all of our mental bets are that this isn't really happening, right? And it may they may be right, but like the whole story of of Kevin's book is like the people who think this is really the big deal, like this is the end of this phase of history have been right so far. They don't have to be right, but like I don't think anyone else really believes it. So you can be contrary voting on this one way or another, right? And but I can tell you you know, there's enough reasons why people it hasn't touched everyone yet in this kind of way. It's still for the reasons you've said, stuck in a phase. So we can sort of say, is it changed 10 years from now, 15 years from now, or is it two or three years? I I don't know what you think about that.

我认为即使在超级 AGI 研究社区内，对于 AI 的经济影响是宏观还是微观也存在分歧。我认识一些 AI 研究者，他们花光所有钱，停止缴纳 401K，因为他们认为在 AGI 后的超丰裕未来，要么我们都死了，要么钱不再有意义，所有物质需求都由机器人满足。我认识另一些人，他们囤积金钱试图致富，因为他们认为在人类劳动失去叙事价值后，我们都会成为永久的下层阶级，除非你在前沿 AI 实验室赚到 1000 万美元。所以，即使在同一基本信念下，你也可以用对市场和投资组合产生截然不同影响的方式来解读。顺便说一句，这两种我都不建议。

I mean, I think even within the super AGI built research community, there are disagreements about what the economic impacts of of AI will be both macro and micro. I know some AI researchers who are spending all their money and they've stopped contributing to their 401K cuz they're like in the hyper-abundant future after AGI, either we're all dead or money has no meaning anymore and all of our material needs are satisfied by the robots. I know other people who are like hoarding money and trying to get rich because they think that after human labor loses its narrative value, we'll we'll all sort of be in the permanent underclass unless you manage to make $10 million at a frontier AI lab. So, even within like you can take the same basic belief and interpret it in ways that will mean very different things for the markets and for your portfolio. By the way, I would not suggest either.

我正想说那并不好笑，但你知道，我是一名社会学家，商学院里得有个社会学家。我花很多时间和经济学家在一起。我的朋友是谷歌的首席 AGI 经济学家。我经常和诺贝尔奖得主达龙·阿西莫格鲁交谈，没有人对这个问题有答案。斯坦福的埃里克·布林约尔松和其他人正在做非常出色的工作，我们都在试图模拟这种情况，看看哪里会出问题，但关键是我們不知道，对吧？所以我们的模型，甚至经济学家也说，早期数据令人困惑，我们没有答案。所以我认为答案通常是，在大多数人心中未来看起来像过去。这总是最安全的赌注，如果不是，你怎么办？

I was going to say that that's not funny but like I you know, I'm I'm a sociologist you got to have a sociologist business school. I spend a lot of time with economists. A friend of mine is the chief of AGI economist at Google. I talked to Daron Acemoglu won the Nobel Prize fairly often like and like nobody has an answer to this question and like you know, there's really great work being done by Erik Brynjolfsson and other people at Stanford and like we're all trying to model this situation to see where things are going to break but the whole point of this is we don't know, right? And so our model even the economists are like we're we don't the early data is confusing. We don't have answers to this question. So, I think the answer is generally the future looks like the past in most people's mind. That's always been the safest bet and if it doesn't, what do you do?

所以，呃，我知道之后会有爆炸性消息。

So, uh I understand that there's blast coming in after this.

我们确实需要重新学习一些东西。

We do need to relearn some stuff.

是的。嗯，那么，在我们结束之前，我想说，对于在座的法学院学生，我和几位来自 Cravath 的同事会在咖啡馆，如果你想谈谈我们在 Cravath 用 AI 做什么，或者为什么它会是开始职业生涯的好地方，请来和我们聊聊。嗯，否则，我要感谢 Ethan 和 Kevin 的精彩对话，谢谢大家的聆听。

Yeah. Um so, um before we wrap up though, I do want to say um for those who are law students here, uh myself and a couple of my colleagues from Cravath are going to be down in the cafe if you want to talk about what we're doing with AI at Cravath or why it would be such a great place to start your career, please come talk to us. Um, but otherwise, I want to thank Ethan and Kevin for a fascinating conversation and thank you all for listening.