Wharton 创业教授：有了 AI，我鼓励学生做什么

简要说明Brief Description

沃顿商学院创业学教授：如今有了 AI，我鼓励学生打造什么Wharton Entrepreneurship Professor: What I Push My Students to Build Now That They Have AI

目录Table of Contents

• 简介
  Intro
• 如何教会人怀有雄心壮志？
  How Do You Teach Ambition?
• 外科医生和士兵在他执教的沃顿课堂上创造了什么
  What Surgeons and Soldiers Built in His Wharton Class
• 每年都会失败的创业点子
  Startup Ideas That Fail Every Year
• 现在创办一家软件公司还值得吗？
  Is It Still Worth Starting a Software Company?
• 善用 AI 所需的 4 项技能
  4 Skills You Need to Use AI Well
• 如何培养品味
  How to Develop Taste
• 为什么初级岗位不再培养人才了
  Why Junior Jobs Stopped Training People
• “认知上的放弃”：哪些事仍应亲自去做
  "Cognitive Surrender": What to Keep Doing Yourself
• “足够优秀”的专业人士会怎样
  What Happens to "Good Enough" Professionals
• 为什么专家最能充分利用 AI
  Why Experts Get the Most Out of AI
• 你应该让 AI 为你生成创意吗？
  Should You Let AI Generate Your Ideas?
• 如何管理 AI 智能体
  How to Manage AI Agents
• 最大的错误：过度信任智能体
  The Biggest Mistake: Trusting Agents Too Much
• Ethan 的真实智能体工作流，逐步拆解
  Ethan's Real Agent Workflow, Step by Step
• 他迫不及待想交给 AI 的工作
  What He Can't Wait to Hand Off to AI
• 在一切不断变化时有所助益的心态
  The Mindset That Helps When Everything Keeps Changing
• 本地模型：何时才能发挥作用？
  Local Models: When Will They Matter?
• 设定 AI 目前还无法实现的目标
  Set Goals AI Can't Hit Yet
• 当人工智能学会品味，会发生什么？
  What Happens When AI Learns Taste?

简介Intro

我们很容易把技能交给人工智能，而这些技能可能再也无法找回来。We can easily give up skills to the AI and we may never get those skills back.

这位是伊桑·莫利克，入选《时代》杂志“人工智能领域最具影响力的100人”之一。他领导沃顿商学院的生成式人工智能实验室。当每个人都让 Claude 为自己出谋划策时，你怎样才能避免听起来和其他人一模一样？This is Ethan Mollik, one of Time's 100 most influential people in AI. He leads the generative AI labs at Wharton. When everybody has Claude advising on the strategy, how do you avoid sounding just like everyone else?

你不希望 AI 做出那些选择，即便那些选择是好的。你希望自己能参与到这个过程中。我正在决定什么时候把自己的能动性融入系统，而不是只派出一些代理去干活。适应这部分必须每两周从头开始。那就是，那就是我的感受。You don't want the AI making those choices, even if they're good ones. You want to be inserted into that process. I'm deciding when to insert my own agency into the system rather than just sending off agents that do work. The adaption piece has to happen from scratch every every two weeks. That's that's what I'm feeling.

因此，我们必须在生活中的某些关键领域刻意增加阻力，才能保持优势。So we have to purposely add friction to our lives in certain key areas so we maintain an advantage.

所以你永远不会把构思点子的任务交给 AI。So you will never delegate generating ideas to AI.

我倒不会说永远不要委派，不过……I wouldn't say never delegate but

如何教导雄心壮志？How Do You Teach Ambition?

伊森，欢迎。你在沃顿商学院教授创业与创新。《时代》杂志将你评为人工智能领域最具影响力的100人之一。大约有50万人阅读你的 Substack 专栏《One Useful Thing》，你的新书《共存》也刚刚出版。欢迎来到节目。你教学生时，也会要求他们在课堂期间创办公司，并告诉他们要着手实现最具雄心的构想。你是如何教导学生树立雄心的？Ethan welcome. You teach entrepreneurship and innovation at Warden. Time named you one of 100 most influential people in AI. About half a million people read your Substack. one useful thing and your new book Coexistence just came out. Welcome to the show. When you teach your students, you also ask them to start companies while they're in class and you tell them to work on the most ambitious ideas. How do you teach ambition?

这始终是个很难回答的问题，对吧？在创业领域，我们把雄心称为创业效能感，对吧？也可以说是类似的东西——主体能动性。嗯，实际上这事相当难办。我们没有什么现成的攻略。但我发现，人们会给自己设置障碍，而你可以在一定程度上帮他们克服这些障碍：只要鼓励他们更有雄心，让他们置身于一群更有雄心的同伴之中就行。所以，我刚才问的很多问题其实都很简单，比如“你还能在此基础上做得更多吗？”在 AI 出现之前，这可是个很难回答的问题，因为你会止步不前，因为你的能力有限，能完成的事情也就那么多。而现在，随着前沿不断拓展，推动人们树立更大的雄心就变得重要得多了。That is always a hard question, right? Ambition looks like we call this entrepreneurial efficacy in the entrepreneurship world, right? Same kind of thing, agency. Um, and it's actually quite tough to sort of do. We don't have like a playbook for it. But I actually find that people create barriers that you can kind of help them overcome by getting them to by just asking them to be more ambitious and being in a room where peers are more ambitious. So a lot of what I just asked it's really easy is now is like what could you do more than that and that was a hard question before AI because you stopped because you had the ability to only get so much done and now as the frontier moves out pushing for ambitions is a much bigger deal.

外科医生和士兵在他的沃顿商学院课堂上打造了什么What Surgeons and Soldiers Built in His Wharton Class

你在课堂上见过的最酷的想法是什么？这挺有意思的，因为我上一次开这门课还是六个月前，那时已经算是智能体工作流的起步阶段了。我下一学期一月份开始。所以，ChatGPT 发布之后，我其实一直关注着相关进展；当然，在那之前我也关注模型有一段时间了。不过，Da Vinci 发布后——它是在 ChatGPT 发布前几天推出的——ChatGPT 又在几天后问世。接着，几天后我上了一堂课，干脆把整堂课的内容都改了，说：“我要教你们怎么用 ChatGPT。”到下课时，我发现有个学生已经不听课了，而是在课堂上把他整个学期的项目都写完了，还在下课时把成果给我看。从那以后，每年大家的目标都变得越来越高。这一次最有意思的是，出现了许多领域专家。What are the coolest ideas that you've seen in your class? So it's funny because like the last version I taught was six months ago and so already that was like the beginning of agentic agentic works. My next semester starts in January. So I I actually when when uh after chatbt came out I had been following you know obviously uh models for a while but after Da Vinci which which came out right a couple days before Chhat GBT and Chadvt came out and then a couple days later I had a class and I basically just dropped the entire class and said I'm going to show you how to use Chad GBT and by the end of the class one of my students had actually stopped paying attention and had coded up his entire project for the semester during the class period. and showed me the results at the end. And so every year since then, the ambition level has gone up further and further. What was most interesting this time was all of these people who are subject matter experts.

我这门课里有外科医生、生物医学设备领域的人、金融行业的人，还有，呃，你知道，非政府组织和人道主义领域的人、士兵，结果突然之间，他们全都在做些真正的东西，对吧？比如说，大家提出了很有意思的想法：我们该如何追踪整个供应链？因为很多公司其实并不了解自己的整个供应链；我们又该如何找出供应链中哪些环节存在财务薄弱点，可能需要我们留意？真是一系列很有意思的想法。呃，还有很多非常有意思的内容，围绕医学领域的一些具体问题展开，比如如何真正用患者能理解的方式向他们提供信息，而不是只给他们一些表格。所以，我的意思是，这些想法都很有意思，但范围很窄，对吧？都聚焦于人们熟悉的事情。I've got surgeons and I have biomedical device people and I've got finance people and I have uh you know uh NGO and humanitarian people, soldiers all in this class and suddenly they were all building real stuff, right? like you know and so there were really interesting ideas how do we track our entire supply chain because a lot of companies don't actually know their entire supply chain and how we figure out where's the financial weakness in that chain that we might have to watch out for. Really interesting set of ideas. Uh there was a bunch of really interesting stuff around um like specific problems in medicine about how do you actually give information to patients in ways they understand as opposed to just forms. So I mean just really interesting but very narrow right focused on something people know.

有趣之处在于，软件可以有很多种形态。如果你想解决一个自己有丰富经验的问题，突然间就会发现其中蕴藏着许多令人兴奋的能力。Part of what is interesting is like software can be many things and if you're trying to think about solving a problem that you have deep experience in suddenly there's a lot of capabilities there that that are really exciting.

你有没有在劝他们别去解决某些问题？Are there any problems that you're telling them to avoid solving?

每年都会失败的创业点子Startup Ideas That Fail Every Year

在一些顶尖商学院教创业学的我们这些人当中，你知道，我们过去常常交流笔记，聊聊那些人们一再提出、而我们知道根本行不通的话题，但我们总是很纠结：我们该告诉他们这事行不通，还是让他们自己去尝试、自己弄明白？有几个算是常见的问题，也恰好是技术受到限制的一个很好的例子。最常见的一个就是：我们要取代房产经纪人，对吧？这听起来是件大事——你要付给这些人几个百分点的费用，可他们为你做的事情却越来越少。这显然是去中介化的案例。但每次这种想法都会失败，因为事实证明，经纪人已经深深融入了整个流程的社会、政治和法律层面，你根本无法把他们从中剥离出去。There among those of us who teach entrepreneurship at some of the leading business schools we you know we used to exchange notes about the topics we see people address over and over again that we know will never work out but like we're always conflicted. Should we tell them it's not going to work out or let them try and figure it out for themselves? So there's a few that that are sort of standard problems that are actually kind of a good example of where technology is limit. One of the most common is we are going to replace brokers for houses, right? It seems like a really big like you're paying a couple percent to these people who like are doing increasingly less work for you. This is obviously a case of disintermediation and every time the idea fails because it turns out there's a deep social integration of political integration and legal integration of brokers into the process that you just can't dislodge.

所以有时候，现实世界的情况和你们想的很不一样。另一方面，学生也会把注意力过于局限在自己的学生问题上。比如，大家总是想要更好的办法，把家常饭菜送到他们手上，或者更好的办法，在学期间帮他们存放东西；但这些需求很可能并不像离开大学之后那么普遍。这真的很有意思。你提到了这些想法，于是我可以把它们和一些真正成功的企业联系起来。比如，DoorDash 解决的是在校园里送餐的问题；然后你提到房地产，比如 Airbnb，对吧？我们当时可能会想，谁会从陌生人那里租公寓呢？是的，这里面有信任这种社交层面的因素，但后来 Airbnb 占据了市场，而现在我们信任 Airbnb。So sometimes the reality of the world is quite different than what you people think. And then the other is um students thinking too narrowly about their student problems. So everyone always wants you know better ways to like get homemade food to them uh or better ways to you know hold stuff between semesters that probably aren't as universal once you're out of uh out of college. This is so interesting. You're mentioning this ideas and then I can draw parallels to businesses that actually worked like Door Dash was solving the problem of delivering food on campus and then when you talked about real estate like Airbnb, right? We we would have thought like oh who would have rented um an apartment from strangers and like yes there's social element of trust but then Airbnb took over and now we trust Airbnb.

你怎么知道你的想法经不起现实世界的考验？How do you know your idea is not going to survive real world?

嗯，我也不知道，对吧？所以我们才不告诉别人，对吧？比如说，这些都是失败率最高的点子，但显然人们就是想做，对吧？因为这些点子总是不断冒出来。它们并不是坏点子。关键在于，一开始你没法把它们筛掉。比如，买房卖房应该有更好的办法，对吧？而且你知道，比起外卖服务，人们确实更想轻松吃上家常饭，而他们现在想要的差不多就是这个。说白了，他们想问的是：怎么才能让我妈（或者我爸）给我做顿饭？所以肯定有办法解决这个问题。市场有需求，只是这些问题比人们想象的更难。从某种意义上说，AI 还让这件事更具野心，对吧？因为现在技术环节不再是难点了，可各项数据看起来都特别漂亮，然后他们就会碰到一个瓶颈，发现：哦，这里确实有个挑战。Well, I don't know, right? That's why we don't tell people, right? like these are the highest failure rate ideas, but obviously because they want to do it, right? Because they keep coming up. They're not bad ideas. The whole idea is you can't filter them out initially. Like there should be a better way to buy and sell a house, right? And you know, people do want home-cooked meals more easily than a meal delivery service, which is sort of what they're asking for now. It's like, how do I have my mother cook a meal for me is essential or father is essentially what they're asking. And so there should be a way to crack this. There's demand for it, but they're harder problems than people think. And in some ways, AI makes it more ambitious, right? because it's no longer the technical piece, but the numbers look really good and then they just hit a point where it's like, oh, there is a challenge here.

所以，谁能破解这个难题，谁就能赚很多钱。但这也是一个我们看到人们一次又一次奋力攻克的领域。所以，也许正在观看或收听的人当中，就有人能颠覆房地产经纪人的服务体验，不过我想，这并不像看起来那么容易。So, whoever cracks it makes a lot of money, but it is also an idea that everyone is we've seen people throw themselves against over and over again. So, maybe some people who are watching or listening will be the ones to to break the uh real estate broker experience, but it's not as easy as it appears, I guess.

没错。就像你什么时候才会放弃？因为你周围有那么多诱人的东西，而且你一天就能做出东西来，对吧？当你碰壁时，很难专注于一个问题。你认同这句话吗？处理一个大问题和处理一个小问题一样容易。基本上，经营小企业和大企业所付出的努力是一样的，因为你都得为招聘和解雇员工而头疼。你认同吗？Exactly. It's like when do you give up? Because there are so many shiny things around you and you can build things in a day, right? It's so hard to focus on a problem when you hit the wall. Do you agree with this phrase? It is as easy to work on a big problem as on a small problem. So basically the amount of effort that you put into building a small business versus large business is the same because you're going to have the same headache hiring firing people. Do you agree with that?

总的来说，我是说，我觉得如果你想白手起家创办一家企业，对吧，就应该把目标定得远大一些，因为反正你未必能做到，对吧？而且，雄心壮志有助于激励人们，也能吸引最优秀的人为你工作。这会形成一种真正的自我实现预言，对吧？不过我也觉得，企业的发展是有不同阶段的，对吧？所以无论如何，最初的阶段都可能很艰难。企业规模化会带来它自身的挑战，这和经营一家小企业不一样，对吧？所以它们并不完全是一回事，对吧？比如说，努力迈上一个新台阶。如果你要进行 B 轮、C 轮融资，那和经营某种本地小生意是很不一样的，对吧？所以我觉得需要的是不同的技能组合。In general like I mean I think if you're trying to build a business from scratch, right, aim big because you may not get there anyway, right? And also ambition helps motivate people. It gets you the best people to work for you. It creates a real self-fulfilling prophecy, right? But I also think that it is there are phases to a business, right? So that initial phase might be difficult no matter what. Scaling presents its own challenges versus running a small business, right? So they're not quite equivalent, right? Like trying to get to the next level. You if you're going to raise a series B and a series C, that's very different than you're running, you know, a like, you know, a local business of some sort, right? So I I think different skill set.

是啊，是啊。我是说，如果你是从一张白纸开始，对吧，而且你想征服世界，那就放手去做，尽可能雄心勃勃。不过顺便说一句，我觉得我们一直在谈想法。我认为，人们在追求各类想法时，野心还不够大，部分原因是他们对软件抱有一种硅谷式的思维，对吧？我研究技术史等领域，也做过相关工作。你看看工业革命期间真正发了大财的人是谁，对吧？他们并不是卖蒸汽机的人，尽管硅谷总是拿这个打比方，说“我想卖镐和铲子”。你不该卖镐和铲子，而应该用新的采金技术去挖金矿。我觉得我们在这件事上有点忽略了这个道理。Yeah. Yeah. I mean, if you're stuck for a blank slate, right, and you want to take over the world, go fully ambitious. And but I also by the way I think that you know we keep talking about ideas. I think people aren't being ambitious enough about the kinds of ideas they pursue partially because of they're sort of in a Silicon Valley mindset about software right and I think a lot about I I study the history of technology among other things and I've done work on it and if you look at who got like really rich during the industrial revolution right it wasn't people selling steam engines even though this keeps being the analogy in Silicon Valley is like I want to sell picks and shovels. You don't want to sell picks and shovels. You want to dig a gold mine with new gold mine technology. And I feel like we're kind of missing that point here.

如果你能让业务更高效，就现在自己动手去做，对吧？直接克服那些障碍。If if you can make a business more efficient, just do it yourself right now, right? Overcome those barriers directly.

当你思考创业时，软件公司将会面临什么样的未来？如今创办一家软件公司真的值得吗？What is going to happen to software companies when you think about entrepreneurship? Is it really worth starting a software company these days?

现在创办一家软件公司还值得吗？Is It Still Worth Starting a Software Company?

所以我觉得，总的来说，这对创业是个好消息。早期数据看起来确实很好，你也能明白为什么这是个很不错的想法。我觉得现在大家都有点太有“SaaS脑回路”了，对吧？也就是总想着怎么赚钱——做软件即服务。我认为这正受到威胁，对吧？如今的环境和以前不一样了。我现在也花很多时间和大公司以及创业者打交道，有些人已经意识到了这一点；大家对软件的未来会是什么样子，分歧很大。另一方面，我觉得软件正越来越成为开展许多其他有趣工作的基础门槛，所以我其实很希望看到更多创业者不再把软件本身当作他们要卖的东西，而是真正去制造或做些什么，因为做这些事的时机从未如此之好。So I think you know it's good news for entrepreneurship overall. Like the early data looks really good and you can understand why it's a really good idea. I think everyone's a little too like SASrain at this point, right? which is like what do you do to make money software as a service and I think that that is under threat right like it's a different sort of environment than it was before I spend a lot of time with big companies now as well as entrepreneurs and there are people getting it there there's a lot of conflict over what the future of software looks like at the other hand I think software is increasingly becoming something that's table stakes for doing a lot of other interesting work so I kind of would love to see more entrepreneurs branch out from software being the thing they sell to actually making or doing something because there's never been a better time to do that

你认为 SAS 的未来会怎样？我们看到这些模仿 AI 的公司，对吧？这种情况几年前就开始了，但现在我觉得它们的收入正在大幅下滑，因为新模型基本上正在取代它们原来做的事情。what do you think is the future of SAS because we're seeing all these companies that copy AI, right? That has started a few years ago, but now I think their revenue is going down significantly because of AI because it's basically the new models are replacing what they were doing.

所以这还挺有意思的，对吧？很多事情都在同时发生。一方面，现在人人都是程序员了。我的意思是，这可不只是理论上这么说，对吧？Anthropic 有一系列很棒的研究表明，实际上，现在管理者写的代码比程序员更好，至少在第一轮是这样，对吧？因为关键在于能否给出清晰的指令，而不只是随时准备好写代码。嗯，而且变化非常快，对吧？高效的人才多了百倍。所以，结果就是公司也完全乱成一团，但他们的目标都瞄准了 SaaS 供应商，对吧？没有哪家大公司不想想办法让自己变得更加独立，自己打造工具。所以，他们还得花一段时间才能把这一切理顺。So it's kind of interesting, right? You have a bunch of stuff happening at once. On one hand, everybody's a coder now. I mean, and that's not just a theoretical thing, right? So there's this great set of studies out of anthropics suggesting that actually managers now write better code than coders, at least in first passes, right? because it's about giving clear instructions more than being ready to code. Um, and things are transforming quite quickly, right? You've got hundred times more perfor productive people. So, companies are also completely in a mess as a result, but they all have their targets on, you know, SAS vendors, right? There's not a large company out there that doesn't want to think about how do we become more independent, how we build our own tools. And so, it'll be a while before they get that all right.

呃，而且他们中有很多现在仍然很满意。我的意思是，据我所知，OpenAI 和 Anthropic 都还在用 Slack。它们来自 Salesforce，应该也还在用 Workday。所以，大家在选择哪类软件供应商时，可能会有更多选择。但我觉得更有意思的是，企业开始意识到，很多原本向他们销售产品的初创公司，卖给他们的其实都是他们自己已经能做出来的东西，对吧？这些公司没有自己的大语言模型，却在向企业兜售企业本来就具备的大语言模型能力。这就形成了一种很有意思的局面。Uh, and a lot of them are still very happy. I mean, you know, as far as I know, you know, both OpenAI and Anthropic still use Slack. they, you know, from Salesforce, they still probably use Workday. So, like there'll be more choices being made about what kind of software vendor to go with. But I think the more interesting thing is that companies are realizing that a lot of the startups that were selling them stuff are just selling them things they could build already, right? They they don't have their own LLMs, so they're selling LM capabilities to them that they already have. And that's creating a really interesting dynamic.

那么，你觉得几年后会发生什么？因为我已经能想象到，比如说，我们现在在用这个软件，但订阅用户数量快到上限了，而且价格还要大幅上涨。我就在想，我们能不能试着用 vibe coding 自己做一个，然后把亚马逊的服务接进来处理新闻简报之类的东西？我也不知道，你觉得几年后会怎么样？So, what do you think is going to happen in a couple years? cuz I already see myself like for example we're using the software but because we're getting to the limit of the number of subscribers and the price is going to raise significantly I'm like can we just try and vip code this and like connect Amazon services for like the newsletter stuff right I don't know like what do you think is going to happen couple years

我认为软件会变得大众化，对吧？所以整个情况是这样的：我在沃顿商学院教书，教 MBA 学生，也教本科生，而很多 MBA 学生都不懂技术。所以，过去十年成功的关键就是找一个懂技术的人帮你写代码。我觉得，把这件事看作其他领域那样，可以用各种不同方式来切入，会让人无比自由，对吧？有些公司会高度依赖软件，要想成功，就需要顶尖工程师与 AI 合作。还有一些公司会把软件当作一套工具，用来完成各种事情。我认为，让软件普及化会非常有帮助，对吧？我的意思是，你不想做一家软件公司。你想做自己的新闻简报，想制作自己的视频。而你知道，以前有些事情会阻碍你。I think that software will be democratized right so it like the whole thing I I teach at Wharton I teach MBA students right and I teach undergrads and a lot of the MBA students aren't technical so like the key to success for you know last 10 years has been find a technical person who can code for you. I think it's incredibly freeing to view this like other kinds of fields where you can get approach in different kinds of way, right? There'll be software heavy companies where you really need top engineers working with AI to succeed. And then there's going to be like software as a toolkit that we get things done. And I think democratizing that's incredibly helpful, right? I mean, you don't want to be a software company. You want to make your newsletter. You want to make your videos. And you know, there were things stand in your way.

顺便说一句，软件也是其中一项，对吧？你还得能自己处理财务。我的意思是，我创办第一家初创公司时，正值 90 年代末，互联网泡沫刚刚兴起，但我当时并不知道，可以每个发薪周期花几分钱，请人帮你处理员工的税务。所以头几个星期，我每周都用 Excel 计算税款，完全没想到这事其实可以直接交给别人处理。像这样的建议，还有你能获得的各种资源，都让现在创业比以往任何时候都容易得多。And software, by the way, was one of them, right? It was also you needed to be able to do your own finance. I mean, when I did my first startup company, I did not realize, it was late 90s, first internet bubble. Um, but I did not realize that um you could pay somebody a couple cents a pay period to handle taxes for your employees. So, I would calculate taxes in Excel every week for the first few weeks, not realize you could just do that. So, even just advice, all of the kind of pieces you get like make this much easier now than it ever was.

本期节目由 MongoDB 赞助。企业希望 AI 智能体能真正开展工作，但许多智能体在正式上线前就卡住了：它会忘记重要细节，而让它访问公司所有数据（例如客户记录和内部邮件）又让人觉得有风险。没人知道智能体会看到什么，也不清楚如果它犯了错该由谁负责。MongoDB 刚刚推出了 Atlas agent engine 来解决这些问题。它提供了一个统一的平台，用于构建、部署和管理智能体。在智能体执行任何操作之前，系统都会先检查它是否有权限执行。每个智能体都在独立的空间中运行。公司设定的规则不能被员工或其他智能体更改。每项操作都会连同批准该操作的人或智能体的名称一起保存。智能体还具备记忆功能。它能记住上次学到的内容，因此不必从零开始，也不用花费 token 重新阅读相同的内容。This part of the episode is sponsored by MongoDB. Companies want AI agents to do real work, but a lot of them get stuck right before launch. The agent forgets important details and giving it access to all your company's data like customer records and internal emails feels risky. Nobody knows what the agent will see or who's responsible if it makes a mistake. MongoDB just launched Atlas agent engine to help with this. It's one place to build, deploy, and govern your agents. Before the agent does anything, the system checks if it's allowed to do it. Each agent runs in its own separate space. Rules set by the company can't be changed by an employee or by another agent. And every action is saved together with the name of the person or agent who allowed it. The agent also has memory. It remembers what it learned last time, so it doesn't start from zero and doesn't spend tokens rereading the same things.

这意味着它会随着时间推移变得越来越准确，而你的花费会更少。而且你不需要迁移数据。它可以直接访问 Salesforce、Snowflake、S3 或你自己的内部系统，并在数据原本所在的位置使用这些数据。你可以使用 OpenAI、Anthropic 或 Google 的 AI 模型，之后也可以更换，而无需从头开始。它运行在 Atlas 上，这是一个已有超过 75% 的《财富》100 强企业在 130 个地区信赖的平台。链接在简介里。在这个过程中使用 AI。我很喜欢你提到的那项针对肯尼亚企业家的研究：有商业经验的人使用 AI 后，策略得到了改善，而没有经验的人则不然。当大家都在大量使用 AI 制定策略时，你能谈谈这项研究吗？That means it gets more accurate over time and you pay less. And it doesn't need you to move your data. It can reach into Salesforce, Snowflake, S3 or your own internal systems and use the data where it already lives. You can use AI models from OpenAI, Anthropic or Google, and change them later without starting over. It runs on Atlas, the same platform that more than 75% of Fortune 100 companies already trust in 130 regions. The link is in the description. using AI in the process of it. I really like the study that you mentioned the the Kenya entrepreneurs where people had experience in business, they used AI and it helped improve their strategy versus people who did not have experience. Can you talk about that when everyone is using AI to strategize a lot?

是的。这是一项很有意思的研究。用的是 GPD4，对吧？那是个早已过时的系统。现在已经有 180 个模型比 GPD4 更强，其中 19 个能在你的手机上运行。嗯，所以你知道，差别很大，而且它当时只是提供建议。所以我记得，最优秀的创业者利润增长了 18%。这个数字简直惊人。我们一直在做研究，试图推动创业领域取得进展。这个数字确实惊人。但它只对已经成功的创业者有帮助，因为它能帮助他们以你刚才说的那种方式拓展业务。那些已经陷入困境的创业者却没法落实这些建议。他们已经越陷越深，根本无从挽回。所以，嗯，我们知道这些东西确实能发挥作用。那是在编程出现之前，对吧？Yeah. So, it was an interesting study. It was GPD4, right? Long obsolete system. There's now 180 models better than GPD4. 19 of which can run on your phone. Um, so, you know, big difference and it was only giving advice. So the fact that I think the number was 18% profit increase for the best entrepreneurs like that's a crazy number. Like we do studies all the time. We try and move the needle in entrepreneurship. That is a crazy number. But it only helped entrepreneurs who already being successful because it helped them expand in the sort of way that you're talking about. Entrepreneurs who are already failing couldn't implement the advice. They were already going down the hole and there was no way to kind of pull back. So it was so we know this stuff can be effective. That was before coding, right?

那简直就是在问：“我现在该怎么办？”说到创业，事实证明，大约 50% 的美国人在过去五年里有过一个好点子，或者自认为有过一个不错的商业点子，但几乎没人真正去实践，因为有这么多障碍。所以，光是关于如何克服这些障碍的建议，就让现在成为创业的黄金时代，远胜以往。That was literally what should I do now? and entrepreneurship like it turns out by the way about 50% of Americans have had a good idea or they think they have had a good idea for business in the last five years and almost nobody pursues it because of all these barriers. So even just the advice to overcome those barriers is a real golden age for entrepreneurship compared to what we've had before.

善用 AI 所需的 4 项技能4 Skills You Need to Use AI Well

你怎么知道什么时候该相信它？你提到了更新的模型。你觉得这项研究会因为新模型好得多而不再适用吗？还是说你仍然需要具备这种判断力？我在新书里谈到了这一点。我认为，要成功地使用 AI，大致需要具备四样东西。第一是深入的知识，对吧？也就是在某个领域有一定的专业能力。顺便说一句，创业很大程度上就是要有纪律地积累专业知识，对吧？你去行业里调研，或者采访从业者，了解这个行业，再去验证想法。我是说，这就是在积累专业知识，我认为它和过去一样重要，对吧？否则，你只是在分享 Claude 生成的点子，而 Claude 生成的点子看起来都非常相似，对吧？第二是广博的知识。对创业来说，了解自己所处领域之外的其他领域，这一点正变得越来越重要，对吧？所以，如果你从事软件行业，就应该对设计和会计也有所了解，因为这能帮助你更广泛地运用 AI。How do you know when to trust it? You mentioned newer models. Do you think the study will no longer apply because newer models so much better or you still need to have this judgment? I talk about this in the new book, but I think there there are sort of four things you need to use AI successfully and that's deep knowledge, right? Some expertise in something and that's by the way like entrepreneurship is all about getting disciplined expertise, right? You go to industry or you interview people, you know, you learn about the industry, you test ideas out. I mean that is building expertise and I think that's every bit as important as it was, right? Otherwise, you're just sharing claude ideas and claude ideas are look all very similar to each other, right? Second is wide knowledge. increasingly important for entrepreneurship is knowing a little bit about the fields you're not in. Right? So if you're in software, you should know a little bit about design and a little bit about uh accounting because that helps you apply the AI more widely.

然后是能动性，这也是我们刚才稍微讨论过的。最后，具备一定的品味也很有帮助，比如从一堆杂乱无章的想法中挑选出好的点子。所以我认为，我们看到的是，能动性更强的系统可以持续运行更长时间，对吧？GP4 可以独立工作四分钟，成功率为 50%。现在，它们能工作的时间已经远远超过 17 个小时。不过，这些系统仍然需要一定的指导。而且，即便它们变得非常优秀，人类的参与实际上仍然重要，因为差异性正是创业者取得成功的关键。And then agency, which we've been discussing a little bit here. And then finally, some sense of taste like selecting good ideas out out of the the slop pile uh is also really helpful. So I think that what we're seeing with the more agentic systems is they can run longer periods of time, right? GP4 could do four minutes of work independently with 50% success rate. Now we're up well over 17 hours of work. But you still need to be able to have some guidance in these systems. And even as they get really good, it turns out the human touch actually does matter because variation is what makes entrepreneurs successful.

是啊，这就是整个挑战，对吧？当每个人都有 Claude，而且 Claude 或 Chad（不管他们用的是什么）都在为策略提供建议时，你怎么才能避免听起来和其他人一模一样？Yeah. And that's the whole challenge, right? When everybody has Claude and Claude is advising or Chad, whatever they're using advising on the strategy, how do you avoid sounding just like everyone else?

是啊，我的意思是，这还挺有意思的，因为你会意识到，人比我们以为的更重要——或者说，比有些人以为的更重要，对吧？所以，那些很早就开始用 Claude 写出好文章的人——你知道，那时候 Opus 4.5 之类的模型写得确实很好——有一阵子，没人知道 Claude 写出来的东西是什么风格。现在你打开 LinkedIn 或 X，简直让人反胃，对吧？都在想：到底是什么在起作用？到底是什么？结果全都是一模一样的东西。不只是那些小短语，连它处理问题的方式、使用的语言都一样。你就会想：嗯，如果你还没雇营销人员，现在大概就是在犯错。营销人员现在能做的工作比以前多得多，但他们身上那种人类特有的多样性其实相当重要。我觉得我们在很多其他情况下也看到了这一点：AI 会把人推向某种共识性的策略或观点。Yeah. I mean, it's kind of funny because you you realize that people are more important than we thought uh or that some people thought, right? So everybody who was early on to using Claude as a good writer back you know when you know Opus 4 five or whatever they write wrote really well and for a while nobody knew what claude writing sounded like and now if you open like LinkedIn or X it's just like nauseating right it's like what's carrying the weight what is it right it's just all the same stuff and even not just the little phrases like the way it approaches things the kind of language it uses and you're like well they should I if you didn't hire a marketer you're probably making a mistake now the marketer can do a lot more work than they could do before but their human variation actually matters quite a bit and I think we're seeing that in a lot of other cases which is the AI pushes people towards a sort of a consensus strategy or view.

我们做过一些实验，发现我们可以让想法的多样性几乎达到一群人的水平；但如果你把 AI 想出的想法和一群人想出的想法分别映射出来，AI 仍然无法涵盖人类想法的全部范围，即使你非常努力地推动它也是如此。所以，多样性这一点会变得越来越重要：你要往系统里加入什么，来引导或驾驭它，让它不至于对所有事情都泛泛而谈？如果你用 Claude 来做财务方面的工作，而这并不是你的竞争优势，那当然没问题。但在你瞄准的细分领域里，你需要脱颖而出，需要有差异化优势，需要建立竞争护城河——即使模型变得极其强大，这在某种程度上也必须来自人类。是的，你提到了品味，这也是我经常问嘉宾的一个问题。We have some experiments where we found that we can make the ideas almost as diverse as a group of people but if you map the ideas that uh that the AI's come up with and map the ideas that a group of humans come up with the AI still doesn't cover the full range of human ideas even if you're pushing it really hard. So this variation piece is going to become increasingly important like what are you adding to the system to guide or steer it so you're not just clotting about everything and that would be okay like it's okay if you're using claude for finance and that's not your competitive advantage but in the area where you are narrowly aiming you need to stand out you need a differentiator you need to moat that has to be human to some extent even as the models get incredibly capable. Yeah, you mentioned taste and uh this is a question I ask a lot of my guests.

如何培养品味？你会对学生们说些什么？How do you work on taste? What do you tell your students?

如何培养品味How to Develop Taste

我们对如何培养人们的品味有一些了解，对吧？比如，你得看过好例子和坏例子。你也会希望别人向你解释，如何把某样东西拆解开来分析。我现在会请一位艺术教授到我的课堂上，他会谈谈设计师是如何看待世界的，对吧？这样你就能获得一种视角：什么样的字体算好，什么样的算差？什么样的图像有趣，什么样的没有？所以我觉得，我们需要一套语言来讨论品味，对吧？如果你跟一个真正懂电影的人聊天，他对电影有自己的品味，但也很可能能广泛地谈论电影。同样，风险投资领域也是如此——风险投资家看重的就是品味，对吧？他们会培养出对某一类问题领域的独特品味。We know a little about teaching people taste, right? Like you need to see good and bad examples of things. So like you want people to explain to you how to pull something apart. So I I one of the uh uh one of I I bring into a art professor actually uh in my classes now who talks about sort of thinking about like how do you how does a designer see the world, right? So you get this kind of perspective like what makes a font good or bad? what makes an image interesting or not. And so I think we need language to discuss things for taste, right? If you talk to somebody who really understands movies, they have a taste for movies that's individual, but they can probably talk widely about that. In the same way like venture capitalism, like venture capitalists are all about taste, right? They develop a sense of taste for a particular kind of problem space or not.

所以我、我认为，看到很多变化真的很重要。学会在这方面保持自律，是我认为教育在这里有一定价值的另一个例子。嗯，接触各种各样的选择。我的意思是，这些有助于培养你所需要的品味。So I I think that seeing a lot of variation is really important. Learning to be disciplined about this is another example of why I think education has some value here. Um being exposed to lots of options. I mean those help give you a sense of taste that you need.

这和判断有什么不同？因为 AI 会做很多判断，比如你把数据呈现给它，它分析后会告诉你“这样是更好的决策”。从品味的角度切入时，又有什么不同？How is it different from judgment? Because AI is doing a lot of judgment calls like you present the data and analyzes it tells you like this is a better decision. How is it different when you approach it from the taste perspective?

我不太喜欢把“判断”本身单独列出来，因为 AI 显然也在进行判断。你不可能在没有 AI 作出判断的情况下完成一项 17 小时的任务。就像我们没有把创造力列进去一样，因为 AI 显然也在作出创造性的决定。问题在于，这些决定是否符合你会作出的那种判断和创造性决定。对吧？所以我们可以把它归到品味或知识之下，但这其实是一个有点不同的问题，对吧？判断是个很难界定的东西。我们不太知道该怎么教，也确实没有什么好的例子。但有些人会问：这种判断是在什么情境下发生的？于是问题就变成了这个。所以这确实是个很难提出的问题，对吧？比如说，什么是判断？AI 确实在运用某种判断力，显然其中也有一种不同的人类判断因素，但我们还没有建立起一套语言来区分这些东西。I don't like the idea of judgment per se because AI is obviously doing judgment. You can't do a 17-hour task without the AI making judgment calls. Just like we don't have creativity in the list because the AI is obviously making creative decisions. The question is whether they match the kind of judgment and creative decisions you would make. Right? So we can call that under taste or knowledge, but it's it's a kind of a different question, right? So judgment is this hard thing. We don't really know how to teach and we don't really have a good example of it. But some people, you know, what context is that judgment happening and becomes a question. So it's a hard it's a hard thing to ask, right? Is like what is judgment? And so the AI is exercising some kind there's obviously a human judgment element that's different, but we haven't had to build the language around differentiating these things.

我会想想数学领域现在正在发生什么，对吧？也就是，嗯，人工智能带来了一场证明的爆发。数学家会告诉你的一件事是，当你和他们交谈时，他们会讲很多关于自己领域的新发现，或者说是一些澄清——这些都是圈外人不知道的，对吧？数学有一种审美层面的东西，而且它确实很重要，对吧？对数学家来说，如果一个证明不能教会你一些美好或有趣的东西，那它对数学这个领域就没有价值。即便我们解决了这个问题，那又有什么关系？所以，我们开始区分以前不必区分的事情，因为我们拥有的判断力和智能只有人类这一种。因此，我们以前不需要另设一种类别。I kind of think about what's happening in math right now, right, which is um this sort of explosion of proofs journeyed by AI. And one of the things mathematicians will tell you when you talk to them is a bunch of revelations about their field or maybe clarifications about their field that outsiders didn't know, right? There's an aesthetic element to math that actually matters, right? To to a mathematician that if you if the proof doesn't help teach you something beautiful or interesting, there's no value in it for the field of math. Even if we've solved this problem, who cares? And like so we're starting to differentiate things that we didn't have to differentiate before because the only kind of judgment intelligence we have is human. So we didn't have to come with a different category.

初级岗位为何不再培养人才Why Junior Jobs Stopped Training People

当我思考优雅、品味和判断力时，会觉得这件事特别有意思。似乎有行业经验的人，比刚入行的人更有优势去培养这些能力；后者还没来得及学习所需的基础知识，就得直接开始做判断、培养品味。市场在 AI 领域的发展也是这样吗？ 从市场和教育的角度来看，确实有几件事正在同时发生，对吧？首先，我们暂时正处于大学教育的一场危机之中。我是说，大学教育一直都有危机，所以这并不是什么新鲜事。任何在教育领域工作过的人都知道这一点。不过，我们现在正面临一场暂时的 AI 危机，对吧？而且我们已经大致知道解决方案是什么：把 AI 当作辅导老师来用时，它的表现好得惊人；但如果你只是用它作弊、直接拿答案，那它就很糟糕。即使你觉得自己学到了东西，其实也没有，对吧？所以，最终我们会在课堂之外拥有 AI 辅导老师。And it's so interesting when I think about elegance and taste and judgment. It feels like people who have experience in the field are better positioned to learn that versus someone who's just starting out instead of getting time to learn basic things they need to go straight into judgment straight into taste. Is that how the market is progressing with AI? There's a couple things happening at one time sort of when you think about markets and education, right? So, first of all, we are temporarily in a uh crisis in universities. I mean, we're always in a crisis in universities, so this is not new. Uh anyone who's been in education knows this, but we are in a temporary AI crisis, right? And we already kind of know the outlines of that solution, which is AI is incredibly good tutor when you use as a tutor and it's very bad if you're just cheating and getting answers from it. Even if you think you're learning, you're not, right? So, eventually we'll have AI tutors outside of class.

课堂上，我们会做更多行动学习。我的意思是，如果你上过商学院、法学院，就已经做过这类案例研究、讨论和活动了，而且你知道，我们还会在课堂上进行测试。我们会想办法解决。嗯，真正的问题在于，人们实际上是在现实世界中培养品味和判断力的，对吧？而且，我觉得过去几千年来，我们一直有一种很好的安排：我会派一个刚受过训练、什么都不懂的新员工给你。嗯，我们有学徒制，对吧？过去 4000 年里，白领培训就是这么运作的。你会得到一个资历很浅的人，他急于向你证明自己很能干。他愿意拿低薪，因为他想往上爬、晋升到更高的层级，而且他会去做中层管理人员不想做的工作。Inside of class, we'll do more action learning. I mean, if you've been to business school, law school, you've already done these kind of case studies and discussions and activities and, you know, we'll do in-class tests. We'll figure it out. Um, the real problem is where people actually have developed taste and judgment is in the world, right? And I we had this great deal for the last few thousand years, which I would send to you, you know, a newly trained recruit who didn't know anything. Um, and we had apprenticeship, right? That's how the last 4,000 years of white collar training have operated. So, you'd get a really junior person who was desperate to show you they were good. They'd work for cheap because they wanted to move up the the realm, the the rungs, and they would do the work that a middle manager didn't want to do.

作为回报，他们能拿到一点报酬，能展示自己的本事，也能熟悉其中的门道；而中层经理则可以把工作交给别人去做，至少有人能替他们分担一部分工作，让他们的日子轻松些，对吧？他们就是这样一步步沿着管理层级往上爬的。可这一切都崩塌了，对吧？因为不管你是实习生、第一年入职的员工，还是其他什么职位，你的激励都完全是尽可能交付最好的成果。谁比你做得更好？Claude、ChatBebt，还是 Gemini？所以，你交付的就会是 Claude、ChateBT 或 Gemini 给出的答案。这和成为数字原生代可不是一回事，因为你根本不知道自己交付的是什么。你只是把东西转手交出去，而每个经理都已经意识到：为什么要花那么多时间给一个干活很慢的人反馈呢？And in return, they would get a little bit of pay. They'd get to show off their stuff, and they'd learn the ropes, and the middle manager would get their work passed off to somebody who could at least do some of the work for them, uh, and make their lives easier, right? And that's how they moved up the management chain. And that all broke, right? Because every, you know, your incentives as an intern or as a first year associate or whatever you are, your incentives are 100% deliver the best product you can. who's better than you, Claude or ChatBebt or Gemini. So, you will deliver Claude or ChateBT or Gemini answers. And this is not like being digitally native because you don't know what you're delivering. You're just handing stuff over and every manager has realized why would I give feedback to a human who takes a lot of time to do the work.

我自己让 claw 来做就行。所以，我们刚刚摧毁了塑造专业品味和判断力的整套体系，而我们还没想出完整的替代方案。I could just have claw do it myself. So, we've just broken the entire system that forms professional taste and judgment and we haven't figured out a full replacement yet.

你怎么看？你在告诉你的学生什么？What are your thoughts? What are you telling your students?

“认知投降”：哪些事仍应亲自去做"Cognitive Surrender": What to Keep Doing Yourself

所以，我觉得有几件事。其一，在个人层面，我们现在必须开始非常自律，认真对待自己什么时候会在认知上向 AI 妥协、把事情交给 AI 去做，而这真的很难做到。首先，我们的思维方式就是这样：比如创业——这是我教的内容，对吧，也是我们在这里一直讨论的——创业就是想办法让生活尽可能轻松，这样你就能专注于那些困难的事情，对吧？嗯，但这样一来，你什么也学不到，对吧？不过，生活的其他方方面面也是如此：我们很容易把技能交给 AI，我觉得我们每个人每天都很容易受到这种诱惑，而且这些技能可能再也学不回来了。所以，在某些关键领域，我们必须有意识地给生活增添阻力，以保持优势。但在个人层面，我们没有动力这么做；这就要看组织层面的选择了。公司会做两种选择之一。So, I think that there are a couple things. One is at the individual level we now have to start being really disciplined about when we're going through cognitive surrender when we're giving stuff up to AI and that's really hard to do. We're built like first of all entrepreneurship like which is what I teach right and what we've been talking about here like that's about trying to make your life as easy as possible so you can focus on the hard stuff right um and you know but then you learn nothing right as a result of that but it's also true in every other aspect of life like we can easily give up skills to the AI and I think we all feel tempted every day to do this and we may never get those skills back so we have to purposely add friction to our lives in certain key areas so we maintain an advantage but we're not incentivized to do so at the individual level it's going to making choices at the organizational level, companies are going to do one of two things.

嗯，他们要么走容易的路，要么走困难的路。我担心他们会走容易的路，也就是只招聘资深员工。我们已经看到一些这样的迹象。我知道你请过一些经济学家来谈这个问题。现在有些早期信号表明，初级岗位可能受到威胁。我们还不能百分之百确定。嗯，或者，另一条更难走的路是，我们真正把培训作为工作的重要组成部分，对吧？也就是说，你必须给人们时间，让他们可以暂时效率低下，在工作中学习。我们不能再把学习当成某种偶然发生的事，而是得教他们。也许大学得多加一年，教你如何成为一名谷歌人、麦肯锡人，或者其他什么公司的人；这一年专门为你量身定制培训。因为我们知道如何在课堂上教人、评估人，但如今我们已经不知道如何在现实世界里这么做了。Um, they're going to either either follow the easy path or hard path. And I worry about the easy path, which is only hire senior people. And we're seeing some signs of that. I know you've had some economists on talk about this. There's some early signals that junior jobs might be threatened. We're not 100% sure yet. Um, as you um, you know, or the other option, the harder option is that we actually make training a serious part of jobs, right? So, you actually have to give people time to be inefficient and learn on the job. and we rather than learning be sort of an accidental thing we'll have to teach them maybe you have to actually add an extra year to universities that we teach you to be a you know a Google person or a McKenzie person or you know whatever it is and there's a year of customized training for you to teach you because we know how to teach and assess people in classrooms that we don't know how to do in the real world anymore.

“够好”的专业人士会怎样？What Happens to "Good Enough" Professionals

这是个非常棘手的问题。然后我在想，那些专业人士甚至完成了这一步，毕业时水平已经足够好，但还称不上出色；而模型还在不断进步，这些水平够用的人会怎么样呢？this is such a tough problem and then I'm thinking about a professional who even completed uh that step and they graduate being good enough but not exceptional and the models keep improving what is happening to these good enough people

所以现在还太早，什么都说不准，对吧？so everything is too early to tell anything yet right

我知道，但你有没有过这种时候，比如我会想，我们到底要走向哪里？I know but do you ever like when I think about like where are we headed

对，我是说，我……但也要看多久才算够好，够好了，对吧？问题是，这也没错，我是说，这方面也在进步。所以你知道，我们在第一篇论文里研究这个问题时提出了一个概念，叫作“锯齿状前沿”：AI 擅长一些你意想不到的事情，也不擅长一些你意想不到的事情；只要这个前沿仍然相当参差不齐，就很难提前知道具体是什么。嗯，总会有一些瓶颈需要人类来填补，而且这些瓶颈可能会非常奇怪。比如说，如果你觉得 AI 会取代律师，结果发现律师要在现实世界中处理很多有摩擦的事务，AI 未必能做到，即使你有一个机器人来干活也一样，因为律师要与各种人类系统打交道，还要满足各种法律要求。yeah I mean I but also how long is good enough good enough right like the issue is that's also right like I mean that's progressing too so you know this concept that we developed in our first paper where we examined this we called the jagged frontier that AI is good at some stuff you won't expect and bad at some stuff you won't expect and it's hard to know in advance what that is as long as the frontier remains pretty jagged. Um there are always bottlenecks that humans kind of need to fill in and those can be really weird ones. Like if you think about like AI replaces a lawyer well it turns out a lawyer does lots of things with friction in the world that the AI can't necessarily do even if you had a robot doing the work right because there's human systems they interact with there's legal requirements they interact with.

有些人想和真人交流。因此，这就形成了一些瓶颈，而这些瓶颈实际上会让不同于以往的才能得到回报。你也能在编程领域看到这种情况，对吧？比如，现在怎样才算一名优秀的程序员，和三年前完全不同，因为你已经不再亲自写代码了。所以我认为，我想说的一点是，所谓足够优秀，是一系列技能的组合；而具备这些技能的人，很可能是那些编码水平平平、但对如何构建软件有很强愿景的人。Some people want to talk to a human. So there creates these bottlenecks that that that actually reward different talents than before. You see this happen in coding too, right? Like what makes a good coder now is completely different than 3 years ago because you're not writing code anymore. So I think one thing I would say is good enough is a bundle of skills and it may very well be the people who are like mediocre coders but really good at the vision about building software

或沟通。or communication.

对，或者沟通能力之类的，也会随着时间改变，对吧？我不会说 AI 永远不会做那项工作，它很可能会。所以我觉得，我们会发现自己在各种有价值的技能之间飞快地来回摇摆，而这没有简单的答案。Yeah. Or communication and and those will change too over time, right? I'm not going to say the AI will never do that task. It probably will. So, I think we're going to see ourselves ping-ponging across what skills are valuable in a really rapid way without an easy answer.

为什么专家最能从 AI 中受益Why Experts Get the Most Out of AI

是啊。我真的很喜欢那个“锯齿状前沿”研究，也一直在想，也许是我们对这些模型还不够了解，因为有时候重新提示一下，就能得到想要的结果。Yeah. And I really like that Jagged Frontier study and I was getting thoughts, maybe it's us who don't understand the models good enough because sometimes you reprompt and you get the result that you wanted.

嗯，我最喜欢的现象之一，就是看那些对某个领域一无所知的人，让 AI 做出一些几乎不可能的事情，然后你就会问：“等等，你是怎么做到的？”他们会说：“我就问了一下。”你会说：“这事你不能问 AI 啊。”结果他们问了，而且还真管用，你就会说：“哦，太棒了。”你知道，我们到底在干什么？所以，我的意思是，这里确实有一种挺奇怪的现象。我是说，有两件事我们知道能让你更善于使用 AI。过去，AI 在很大程度上是一种拉平差距的工具，对吧？如果你表现不佳，它就能帮你提升。这一点现在仍然成立，但我们越来越多地看到——Anthropic 和其他一些研究都有不错的结果表明——在现阶段，专业知识最丰富的人从 AI 中获得的提升最大。而这一点直到大约去年智能体出现之后才开始成立。Well, my one of my favorite phenomena is watching people who know nothing about a field getting the AI to do like impossible things and you're like, "Wait, how'd you do that?" You're like, "I just asked." You're like, "You can't ask the AI for that." And they did and it like works and you're like, "Oh, great." you know what are we doing here? So there I mean there is this sort of weird phenomenon there. I mean there's two things that we know make you better at AI. It used to be AI was a level mostly a leveler right if you're a bad performer you got a boost that it still is that way but what we're increasingly seeing and there's some nice results of anthropic as well as some other studies that suggest that the people who are most have the most expertise get the most gains from AI at this stage and that wasn't true until agents kind of came around in the last year or so.

嗯，不过我们还知道另一件事：有一种类似“与 AI 耳语”的能力，也就是了解 AI 如何思考，这或许有点像心智理论。比如，如果我非常了解人，也逐渐熟悉 Claude、Chatbt、Gemini 或我正在使用的其他 AI 的思维方式，那么我与它合作时就会更有效率。嗯，所以这算是两种不同的技能。然后还有一个神秘的第三类：有些人似乎在这方面特别擅长，但我们不知道原因。Alex Emis 和其他几个人做过一项很棒的研究。他现在在 DeepMind，嗯，他之前在芝加哥大学工作。研究人员让人们创建用于谈判的智能体，并发现了几个有意思的现象。其中一点是，你的技能倍增效应会变得强得多，因为智能体会参与很多次谈判。Um, but the other thing we know is that there's some AI whispering ability that is like knowing how the AI thinks that might be sort of theory of mind. Like if I know people really well and I get to know the mind of Claude or Chatbt or Gemini or whatever I'm using, I become more effective in working with it. Um, and so those are sort of two different skill sets. And then there's like a mysterious third category which is like some people seem to be good at this in ways we don't know. There's this great study by Alex Emis and a few others. He's now at um he was at Chicago, now he's at Deep Mind. Um but they measured people creating agents for negotiation and a couple interesting things. One thing they found is your skill multiplier becomes much higher because the agent engages in many negotiations.

因此，在打造优秀智能体方面，技能上的细微差异会对最终业绩产生巨大影响。So a small difference in skills in creating good agents has a huge impact on bottom line.

他们发现的第二件事是，擅长创建智能体的人，不一定就是现实生活中最擅长谈判的人。我的意思是，甚至还存在性别差异。我们在研究中一再发现，出于各种各样的原因，女性往往在谈判中表现得更差，关于这一点已经有大量研究。这一直是个备受关注的话题。但女性创建的谈判智能体比男性创建的好得多。你知道，而且话说回来，我们其实也不知道该怎么教你做到这一点，我甚至不确定该怎么教。The second thing they found was that the people who are good at creating agents are not necessary people who would be the best negotiators in real life. I mean even there was even a gender divide. So one of the uh things that we find repeatedly in studies um for a wide variety of reasons is that um women tend to outperform in negoti underperform in negotiation and there's been a lot of work on this. It's been a big topic of concern but women made much better negotiation agents than men did and you know and again we don't actually know like I don't even know how to teach you how to do that necessarily.

那是管理技能吗？我是说，这里面有管理技能、有心智理论能力，也有专业知识这一块，但事情发生得太快了，你知道，我们又没法轻易把它们拆开来。还有，我刚刚请来了 Anthropic 的 Felix 到演播室，他谈到，试着理解自己的工作流程非常重要：比如，当你创作出一段完美的文字，或者进行一场完美的谈判时，你脑子里在想什么，你的思路是怎样的。这样你就能把这些思路迁移到 AI 上，这就是构建工作流程的方式。Is that management skill? I mean there is a management skill, there's a theory of mind skill, there's an expertise piece, but like things are happening so quickly that you know we like again we can't pull them apart that easily. This is and I [clears throat] I just had Felix from Anthropic in the studio and he was talking about how it's very important to try and understand your workflow in like what's going on in your head when you create a perfect text or when you go through a perfect negotiation like what was your thinking like so you can translate that onto AI and this is how you build a workflow.

我觉得这多少有点像是编出来的。我是说，也可能是真的。对我来说，作为一个管理过组织、带过人的人，最贴切的比喻就是把 AI 看作一个我分派工作给它的团队。如果我能做好给团队分派工作的事，也能看出团队在哪些地方会出问题——这跟了解自己的工作流程很不一样，对吧？AI 不会照搬我的工作流程。我努力做的一部分，也是用这种方式替自己保留一些东西，对吧？比如写作，你知道，我做 Substack 和 X，所有东西都是自己写的，这可能听起来很疯狂，对吧？所有东西的初稿我都亲自写。我的意思是，没错，我全都是自己打出来的，对吧？因为如果我不这么做，嗯，AI 就会特别有诱惑力，绝对会在任何方向上主导我的思维，到头来你读到的只会是经过我过滤的 Claude 生成结果。I feel like that's kind of made up. I mean it might be true. To me, the analogy that makes the most sense as somebody who's run organizations and manage people is like thinking of AI like a team I'm delegating work to and if I can do a good job delegating to a team and realizing where a team goes wrong, which is very different than knowing my own process, right? The AI will not replicate my process. And part of what I try and do is like preserve stuff for me that way also, right? Like I write, you know, it might seem crazy since I do Substack and X every I write all my own stuff, right? I write first drafts of everything on my own. I mean, yeah, I type it all out, right? because if I don't do that, um, the AI is very seductive and will absolutely like dominate what I'm thinking in any direction and then you're just going to read like clawed results filtered through me.

以及And

比如你会用语音吗，还是我……like do you use voice at all or I

我有时候会用语音输入这个。嗯，我觉得很多这类事情，直接在电脑上打字更容易些。不过，语音也挺好，而且你知道，你会逐渐摸索出自己是怎么产生想法的。比如对我来说，奇怪的是，我最好的点子都是在 SoulCycle 骑动感单车时冒出来的。我觉得自己可能是屋里唯一的男性，也是年纪最大的人，所以你知道，在一个很吵的房间里骑车，大家还冲着我喊，让我实现目标，不知怎么的，我却能把这些全都屏蔽掉，然后想出不错的点子。我也不知道为什么，但每个人都会找到适合自己的思考方式。我还跟另一位科学家聊过，他说自己每周有一天必须累到不行，这样才能想出最好的点子。所以，你得摸索出自己在这方面的门道，但同时也得保留你所做的事情中那些非常人性化的部分。I use voice sometimes for this. Um, I find like just typing on a computer is easier um for a lot of this stuff. But yeah, voice and like you know and like you get good at figure how you generate ideas. Like for me weirdly um I come up with my best ideas is during soul cycling which I think I'm the only guy in the oldest person in the room and so you know but somehow being in a loud room where I'm biking and people are yelling at me about achieving my goals. Um I tune it all out and I I come up with good ideas. I don't know why but everyone has it learns their own way of thinking. I've talked to another scientist who like I need to be extremely overt tired one day a week and then I come with my best ideas. So like you figure stuff out in that kind of front, but you also have to preserve parts of what you do that are very human.

我觉得这未必是要复制我的人类工作流程，而是要保留一种人类工作流程，然后像我给团队分配任务那样委派工作。比如，我有一个很棒的想法，我会说：这就是我希望你们如何把这个想法付诸实践。I don't think it's necessarily about replicating my human process. It's about preserving a human process and then delegating work the way I would to a team. Like I have a really good idea. Here's how I want you to execute on that idea.

所以你的意思是，你永远不会把构思点子的工作交给 AI？So you will never delegate generating ideas to AI. Is that what you're saying?

你应该让 AI 为你生成创意吗？Should You Let AI Generate Your Ideas?

我不是说绝对不要委派，但你得谨慎。我是说，这方面其实早就有证据了[嗤笑]。我想，头脑风暴是 1955 年发明的，到 1957 年我们就知道它不是个好主意了，因为基本上，只要你进行小组构思，大家的想法就会固着在第一个提出想法的人身上，接着还会经过一番评判，于是大家就会以别人的想法为锚点。这样一来，自然的多样性就消失了。所以，在进行任何小组活动之前，头脑风暴应该这样做：每个人先独自把想法写下来，等个人构思完成后，再开始集体产生想法。这样最终得到的想法会好得多，也更多样，对吧？AI 也是一样。你当然应该用 AI 来生成想法，但前提是你已经先自己构思过了。I wouldn't say never delegate, but you have to be careful. I mean, we actually have old evidence [snorts] on this. Brainstorming was invented I think 1955 and by 1957 we knew it was a bad idea because basically anytime you do a group ideation process you your ideas become fixated on whoever says the first idea out the gate and then there's a judgment process and so anchor on other people's ideas. So you lose natural variation. So the way you should do brainstorming before any group process is everyone writes down ideas individually and then only start generating ideas as a group once you've done individual idea generation. You end up with much better ideas with higher diversity. Right? Same thing with AI. You should absolutely use AI for idea generation, but only after you've generated ideas for yourself.

所以，听完一档播客后，我会这么做：每当我学到一些关于 AI 的新东西，就会着手调整公司里的某些流程，看看效果如何，看看这些模型是否真的足够强大，能处理我刚刚委派出去的那件事，[哼笑]然后我会把这些写下来。我不像 Ethan 那样自己打邮件。我会直接跟 AI 说，然后再编辑邮件。我也有一个人帮我做这件事。不过，内容基本上就是总结我们尝试用 AI 做了什么、哪些奏效了、哪些没奏效。当然，也会聊很多战略方面的内容，比如作为一位母亲、一位女性和一位移民，我如何规划自己的媒体事业、创办企业。所以，如果你想更深入了解我的工作流程，就订阅我的免费通讯《Future Proof》，让我们一起成长。你刚才也谈到了这种需要费力思考的方式，以及把 AI 当作伙伴。So, here's what happens after a single podcast. When I learn something new about AI, I go ahead and change some processes in my company, see if it works, see if the models are really powerful enough to handle that thing that I just delegated and [snorts] then I write about it. I don't type my emails like Ethan does. I actually talk to my AI and then edit them. And I also have a person helping me with that. But it's basically a summary of what we tried with AI, what worked and what didn't, and of course a lot of strategic stuff, how I navigate this media career and building businesses as a mom, as a woman, as an immigrant. So if you want to get a closer look at my process, subscribe to my free newsletter, Future Proof, and let's grow together. And you talked about this effortful thinking and having AI as companion.

你能描述一下这个过程吗？嗯，我是说，我觉得你想做的是，在做这件事的时候，让自己的大脑也像身体一样“出汗”。如果你觉得一点都不费劲，那你可能就没有进行足够的学习或思考。这就像锻炼或其他事情一样，对吧？所以在学习中，我们把这称为“理想困难”。如果你没有遇到困难，那你就没有在学习。而且有些思维过程，比如构思，是没法大幅简化的，对吧？但人们很容易就想把它简化。我们还没学会认识到这种做法的代价……我们从来不会到处走，或者说，我们从来不会锻炼。我们还没在大脑方面学会同样的道理，因为以前没有其他选择，对吧？你可以用 Google 搜索，但那跟让 AI 替我思考不是一回事。Can you describe that process? Yeah, I mean I think that what you want to do is you you're kind of trying to mentally sweat while doing this. Like if you're not kind of if it's not uncomfortable, you're probably not doing enough learning or thinking. Like it works just like exercise or anything else, right? And so in learning, we call this desirable difficulties. Like if you're not encountering difficulties, you're not learning. And so there's certain mental processes like ideiation that can't be streamlined that much, right? And it's very tempting to do that. And we don't rec we've learned to recognize the cost of like we never walk around or you know we never you know we never exercise. We haven't learned the same kind of thing with our brains because there was no alternative right you know you could Google something but that was not the same kind of thing as AI do my thinking for me.

所以我觉得，我们得意识到自己正在做这种需要付出努力的工作。有意思的是，我开始想到，也许我产生想法的方式可以形成一种工作流程：比如，我生活中发生的某些事情会引导我产生某个想法，而 AI 也许能帮我找准这个想法。So I think that we have to be recognizing that we're doing that kind of effortful work. So interesting cuz I got to thinking that maybe like the way I get ideas could be a workflow like maybe there is something that's happening in my life that leads me towards certain idea and maybe AI can help me pinpoint

可以，我的意思是，我只是觉得你必须小心。顺便说一下，在构思过程中有一条叫作“等概率法则”的规律，事实证明它确实成立：有些人会产生更多想法，而这些产生更多想法的人，往往也比大多数人想出更好的点子；这和智力并不一定有很强的相关性。创造力是一种特质。有些人就是更有创造力。所以，如果你的创造力没那么强，我觉得我们没必要把创造力捧上神坛，对吧？那也许 AI 就能帮你更多，对吧？帮你生成想法。如果这是你的竞争优势所在，那就回到我们之前聊的那些领域，比如写作之类的，或者做播客，对吧？我们可以生成一个 AI 播客，但它不会比一个经验丰富、做过很多次并且认真思考过内容的专家做得更好。it can I mean I just think you have to be care and by the way there is a uh in ideiation there's something called the equal odds rule that turns out to be true which is that some people uh generate both more ideas and those people who generate more ideas generate better ideas than most other people and it's not necessarily deeply correlated with intelligence Creativity is a trait. Some people are just more creative. So if you're less creative, I think we don't need to hold creativity on a pedestal, right? Then maybe AI is helping you out more, right? Generating ideas. If it's an area that you have competitive advantage, kind of going back to where we were before or writing or whatever it is or podcasting, right? We can generate an AI podcast. It will not be as good as an expert having done this a lot and thinking through this.

所以至少目前来说，这种人类优势确实存在。你要确保自己没有真的把一些东西拱手让出去，对吧？我没那么担心。我的意思是，构思创意只是众多特质之一。大多数人都没有什么好点子，这没关系。我们对点子过于推崇了，对吧？执行力、情感联结，还有很多其他方面，人们都可以做得很好。所以，你要策略性地决定何时使用这些系统。So at least for now, that human edge, right, is is a real one. And that's where you want to make sure you're not really giving stuff up, right? I worry less. I mean, ideation is just one of many traits. Like most people don't have good ideas. That is okay. We lionize ideas too much, right? Execution, you know, emotional connection. There's lots of other great things people could do. And so, you want to be strategic about when you're using these systems.

我很喜欢你说的，想法多的人往往也更容易想出好点子。我觉得我们需要学会不断地产生各种想法，可能把它们记下来，或者试着验证一下，因为测试一个想法真的很容易。I really like what you said about people who generate more ideas tend to generate more good ideas. And I think we just need to be learning how to constantly generate things and like maybe write them down or test them out because it's so easy to test out ideas.

是的，我觉得那些无法付诸行动的想法没什么用。在 AI 出现之前，整个创业学科都是围绕这个理念建立起来的，尤其是硅谷创业、精益创业，以及我们教人们如何成为优秀创业者的其他主流方法和策略。这些策略都源自一个理念：尽可能以最低成本学习，对吧？所以，重点不是只在大楼里空想一个绝妙的点子，对吧？随着软件开发成本下降，目标变成了走出去和人们交流，向他们展示你的产品，对吧？进行客户访谈，打造最简可行产品，然后快速学习。后来，风投也都围绕这一点重新调整了方向，对吧？最初，风投的模式是给你一大笔钱，让你把想法规模化。Yeah. I I think that that um ideas that don't do anything aren't that useful. The whole discipline of entrepreneurship before AI, right, was built around this idea, especially Silicon Valley entrepreneurship and the lean startup and all the other dominant methods strategies that we teach people to become good entrepreneurs. All of those strategies come from this idea of learning at the cheapest cost you can, right? So, it was all about like don't just come with a beautiful idea in the building, right? As the cost of software development dropped, the goal was get out there and talk to people, show them your product, right? Do customer interviews, create minimum viable products and learn quickly. And then all the VC sort of reorient around this too, right? It was originally built around giving you a giant check to scale your idea.

于是，问题就变成了：我们该如何给你提供种子轮、前种子轮融资，还有 Y Combinator？基本上，创业的整个演变过程都围绕着一个想法：随着时间推移，测试想法的成本越来越低。现在，我们真的已经把成本压到了最低，对吧？比如说，模拟客户时，别把 AI 当成真正的客户，但它确实是模拟客户的一个相当不错的初步尝试。现在已经没有借口不至少稍微验证一下自己的想法了。And it became how do we give you seed rounds and preede rounds and why combinator all basically the entire evolution of entrepreneurship was based around the idea that it got cheaper and cheaper to test ideas over time. And so now we really hit rock bottom cheapness, right? like you you can don't trade the AI is not a customer when you simulate customers, but it's a pretty good first pass at being a simulated customer. There's no no longer an excuse to not at least push your ideas a little bit.

是的。你觉得一个好点子现在比五年前更值钱吗？Yeah. Do you think a good idea is worth more now versus 5 years ago?

我觉得，一个好点子如果能结合执行力，价值就会更高，对吧？我觉得，光有一个点子其实从来就没那么值钱。我的意思是，点子很棒，但你知道，在创业课上，到处都会出现的一种冲突就是，提出点子的人想把所有功劳都归于这个点子。你会看到，随着一个学期过去，很多初创公司最终会发展成真正的企业，融到很多资金，也取得巨大成功。但你也会看到这个过程：一开始，大家都追捧那个提出点子的人，觉得那真是了不起；可等过了六个月左右，点子已经变了。真正让一切成功的，很少是最初那个点子，对吧？成功靠的是一系列探索。所以我觉得，点子一直都不稀缺；现在 AI 比任何人都更擅长生成点子，但如果你真的把它们付诸实践，那就能创造真正的价值。I think a good idea is worth more when combined with execution, right? I think that there is that you that an idea alone has never been worth that much. I mean it's great but you know one of the conflicts that typically happens in entrepreneurship classes all everywhere is the person who came up with an idea wants all the credit for the idea and one of the things you see as a you know as a sort of a semester goes on and then a lot of the startups out of war and turn into real businesses and raise lots of money and end up being very successful but you kind of see the process of like oh everyone trace the person who came with the idea is like that was the amazing thing and then by the time you're you know six months in the idea has changed it's rarely the idea that has made everything successful right it's it's a set of explorations So I think ideas have always been sort of a dime a dozen and the AI can generate those ideas now better than anyone else but if you actually execute on them there's real value there.

如何管理 AI 智能体How to Manage AI Agents

我完全同意。而且，在落实想法的过程中，你会成为管理者——管理人也好，管理智能体也好。现在有很多人并非主动选择，却因为工作不得不管理智能体。你会给他们什么建议？他们应该采用哪些管理方法？这算是 MBA 们的复仇，对吧？I agree 100%. And when executing ideas you become a manager either of people or agents and there are so many people who are now not forced but they have to manage agents because of their work. What are the tips that you will give them the management tips that they should be deploying? It's the revenge of the MBAs, right?

比如说，我在商学院教书，所以这事儿还挺让人欣慰的，哈哈，现在人人都得当管理者了。你们以前这些商科的人一直在找技术人才，现在大家都需要管理技能了。不过其实这些管理技能确实用得上，对吧？所以，事实证明，你需要……我的意思是，我们一直在沃顿商学院的生成式 AI 实验室里做很多相关实验，这个实验室由我负责。很多实验基本上都表明：你看，适用于人的技巧也适用于 AI。所以我认为，你需要考虑的是如何管理一个流程，因为 AI 系统的一个风险在于，它的成本可能很高——不只是 token 成本，监控工作的精力成本也相当高。因此，我们构建了所有这些系统来处理这个问题，对吧？比如说，如果我要为一个智能体指定任务，我会说明一些事项，比如：这些是我希望你在工作过程中向我汇报的节点；这里我希望你提供一些多样化的示例，让我来挑选。这样我就能做出选择，也就是决定何时将自己的主导权介入系统，而不是仅仅把智能体派出去干活。Like I teach at uh uh you know I I I I uh you know I teach at a business school so it's like very it's gratifying to be like haha everyone has to be a manager now you thought all all my business people were looking for technical people now all the time people need management skills but actually those management skills do apply right so what you need to it turns out I mean we've been doing a lot of experiments with that at the generative AI lab at Wharton which is what I I lead and a lot of the experiments basically show you know look techniques that work on people work on AI and so I think what you need to be thinking about is how do you manage a process because one of the dangers of an of of the cost of an AI system is have assigning a lot of work not even the token cost just the effort cost of monitoring work is quite high so we built all these systems for doing that right so if I'm specifying things for an agent I will specify things like here are the points I want you to check in with me on about the work here's where I want you to give me diverse examples that I can select from so I make a choice right I'm deciding when to insert my own agency into the system rather than just sending off agents that do work.

我会尽量把好和坏的标准说得非常清楚，并且在流程刚开始时就密切关注，看看自己是否没能把意思传达给智能体。对吧？很多时候，它们会朝某个方向走——你知道，Claude 有它自己的执念，对吧？如果你不尽早、频繁地介入，它们就会沿着那些方向走。所以，你知道，那些管理技能，比如想想自己有没有清楚地说明需求、有没有建立反馈闭环、有没有在项目推进过程中降低不确定性，这些都确实适用于这里。I am making sure to specify what good and bad looks like in really kind of clear ways and I'm watching really early on the process to see whether or not there's a failure in my ability to communicate to the agent. Right? Often times they'll head off in a direction that sort of like you know Claude has its own obsessions, right? They'll head off in those directions unless you intervene quite early on and often. So you know those management skills about thinking about am I clearly specifying what I need? Am I you know am I establishing a feedback loop right? Am I dealing with, you know, reducing uncertainty over the course of a project? Those things end up really applying here.

有意思。你觉得，你觉得人们在管理这些智能体时犯的最大错误是什么？Interesting. What What is the biggest mistake that you think people are making when they're managing those agents?

最大的错误：过于信任智能体The Biggest Mistake: Trusting Agents Too Much

我觉得他们太信任这些代理了。I think they're trusting the agents too much.

你会得到很棒的结果。我觉得大家都喜欢那种……我自己也经常做很多单次提示词的事情，比如“这太棒了”之类的。但引导过程极其重要。我认为这会让人非常无力，一方面是因为智能体知道很多你不知道的事，另一方面也是因为我们现在用来与 AI 协作的界面，越来越不利于增强人的能力、拥抱自动化，对吧？它们会隐藏思维链，甚至连思维链摘要也不再显示。它本来就应该更像魔法一样，而这对让人类参与其中来说简直是灾难。我不认为我们还能参与其中，但管理者其实也不是一直参与其中：他们分派工作，再让别人向他们汇报进展。因此，我们现在必须强制设置一些节点，让 AI 向我们汇报；否则，它们就会试着变魔术给你看，而你并不想要那样。You'll get an amazing result. Like I everyone loves sort of I do a lot of single prompt like this is awesome kind of stuff, but the guidance process is incredibly important. And I think that it can be very disempowering um both because the agent knows things you don't but also because the interfaces we have to work with AI now are increasingly hostile to augmentation and embracing automation right they hide the chain of thought right and and even the chain of thought summaries no longer appear it's supposed to feel more like magic and that's disastrous for being a human in the loop I don't think we can be in the loop anymore but managers are not really in the loop they assign work and they have people check in on them so we have to now create forced positions for the AI to report back to us otherwise they will try and do magic for you and you don't want that.

Ethan 的真实智能体工作流，分步详解Ethan's Real Agent Workflow, Step by Step

你能否描述一下你自己工作流程中的某个过程，向我们展示你是如何创造出这个结果的，以及我们如何建立反馈循环，让人们能够学习如何……Can you describe a process in your own workflow where you can show us how you create this result and how we create the feedback loop so that people can learn how to

当然。比如说，我一直在探索一个人文学科里让我感兴趣的、比较随意的话题，而且我也把它用于创业路演。这个话题就是，有很多研究表明，故事是有结构的，对吧？所以，嗯，有起伏的故事，还有灰姑娘式的故事之类的。这是一个由来已久的问题。其实之前我还用早期的大语言模型做过一些研究，想看看我们能不能分析创业路演稿，找出优秀路演稿的结构。所以，这是一个我长期以来都很感兴趣的话题。随着智能体变得越来越好，我让它们做的事情之一，就是让它们去探索这个话题，对吧？这里面有几件事。其中一件是，嗯，我让它们想出一大堆关于该如何做这件事的点子。Sure. Like I was exploring a sort of random topic that interests me in the humanities which is and I've done this for entrepreneurial pitching as well which is there's a lot of research that stories have shapes right and so um there is and so like there's rise and fall stories and you know and like Cinderella stories and this is like a long-standing question and I actually did some research um before with with primitive LLMs uh about whether or not we could take entrepreneurial pitches and find the shape of a good pitch. So, it's a topic that's really interested me for a long time. And as agents have gotten better, one of the things I've had them do is assign them to explore this topic, right? And so, there's a few things there. One of them is um you know, I was assigned them to kind of generate a bunch of ideas on how to do this.

然后我——但我当时没有让它执行它最喜欢的想法，因为那样会很糟。所以，我不得不中途给它一些反馈。于是，给我一堆你可能会采用的方法，弄清楚故事研究的真实形态是什么。对。然后，你知道，我当时就想，现在像我对学生那样，替我做些推测。说说你可能会如何着手。再给它一些反馈。很好。要不你先执行第一、第二、第三和第四种方法，把它们的初稿都做出来。给我看看结果，然后我们再挑一个更好的例子。很好。我很喜欢你在第三种方法上的处理，不过第二种有个很棒的特点。我们来聊聊怎么把它们结合起来。对吧？为什么呢？我们来看看，你能不能指出来哪里看着不对。And then I but I didn't have it then execute its favorite idea because that would have been bad. So, I had to give it feedback partway through. So, give me a bunch of ways that you might approach this, figure out what the shape of story research is real. Yeah. And then, you know, I was like, and now speculate for me just like I would for a student. Give me some thoughts about how you might approach this. Give it feedback again. Great. Why don't you execute one, two, three, and four, first draft of all of them. Show me the results and then let's pick which example works better from this. Great. I really like what you're doing with three here, but number two has this really nice feature. Let's talk about how we combine those together. Right? Why? Let's, you know, can you show me where the, you know, this looks wrong.

你能告诉我，你为什么觉得这件事搞砸了吗？这几乎有点像苏格拉底式的方法：由我来引导 AI，并给自己设置很多决策点；那些特别繁琐、令人厌烦的工作——比如找出五千个故事分别在哪里、构建语义编码工具——我就交给 AI 去做。这样我就能把有意思的工作留给自己。如果说我最近这本书有一个核心信息，也是我最迫切想传达给美国企业界的，那就是：自动化的世界会在很多方面让所有人都不好过。但增强型的世界才是工作的理想状态，对吧？你做有意思的事，AI 做那些糟心的活儿，而你只需要负责指挥整个组织，那真的很有趣。Can you show me why you think this went wrong? So it's almost like a Socratic method kind of thing of me guiding the AI and give myself lots of choice points where the really tedious annoying work get you know figure out where 5,000 story of 5,000 stories are build the semantic coding tools I let the AI do. So I get the interesting work for myself and like that's if there's one message of sort of the latest book and where I'm most desperate to try and communicate with corporate America. It's like we have an automation world that will be bad for everybody in a lot of ways. But an augmentation world is like the dream of work, right? You do the interesting stuff and the AI does the crap stuff and you get to just direct an organization and it's really fun.

在我们继续之前，我很想多了解你一点。我在这里分享了很多关于我工作的内容，也很好奇你是做什么的，以及希望在哪些方面得到更多帮助。我们准备了一份小问卷。你的回答会帮助我决定今后的交流中要聊些什么，以及为你准备哪些资源。链接我会放在简介里。拜托、拜托、拜托，花几分钟告诉我一些关于你、你的业务和你的抱负的事，这样我就能为你制作更好的内容。我真的、真的非常感谢你。非常感谢你愿意参与。 我发现这里有个问题：模型进步得太快了。你搭建好一个工作流程，描述好各项内容，接着我也不知道，新模型就出来了，他们会告诉你，最好别给它工作流程，因为它已经很擅长理解特定工作流程该使用哪些数据了；有时候它还会想到你没想到的事情，进而设计出更好的工作流程。Before we keep going, I'd love to get to know you a little better. I share a lot about my work here and I'm curious about yours. What do you do and what you'd like more help with? We've put together a little survey. Your answers will help me decide what to cover in future conversations and what resources to put together for you. I'll leave the link in the description. Please, please, please take a couple minutes and tell me about you, about your business, about your ambition so I can make better content for you. I really, really appreciate you. Thank you so much for doing this. The problem with this that I see is that the models progress so fast. So you build a workflow, you describe things and then I don't know the new model comes out and they tell you it's better to not give it a workflow because it's so good at understanding what data to use for a particular workflow and sometimes it thinks about things you didn't think about and it can make a better workflow.

你们多久需要把整个东西重写一次？How often do you have to rewrite the whole thing?

所以那不是我给你的工作流程，对吧？那不是说第一步我做这个，第二步我做那个。这是我管理研究团队，或者你知道，管理初创团队的方式。So that's not a workflow I was giving you, right? That wasn't step one I do this and step two I do this. This was the way I would manage a researcher team or you know or a startup team.

所以这是一件一次性的事。每件事、每件事都是一次互动，我会从系统的行为中学习。我事先不知道结果会好还是坏。如果你想创办一家初创公司，进行一次深入的对话，得出某种结论，制定营销计划，或者做财务分析，你知道的，这些都涉及很多选择。你不希望 AI 替你做这些选择，即使它做出的选择是好的。你希望自己能参与其中，知道吧，置身于这个过程中。所以，这算是一种元过程，对吧？意思是，假设有个非常聪明的人，或者一支非常出色的团队在做事，但你还是希望这件事是你自己的作品。那你什么时候该跟 AI 讨论这件事呢，对吧？你还是希望能看清系统里发生了什么。否则，你就只是……它就只是被抓了。So it's a oneoff thing. Everything everything's an interaction where I'm learning from what the system does. I don't know in advance whether it's going to be good or bad. If you're trying to launch a startup company, have a deeper conversation, reach some sort of conclusion, come with a marketing plan, you know, do financial analysis. All of these involve lots of choices. You don't want the AI making those choices even if they're good ones. You want to be in, you know, inserted into that process. So, it's sort of a meta process, right? The idea is like assume you have a really smart person or a really incredible team doing work, but you still want it to be your work. When should you talk to the AI about this, right? You still want that visibility into the system. Otherwise, you're just it's just clawed.

即使它比你更聪明，你也不再为这个过程贡献任何东西。因此，任何人都可以输入同样的句子，并得到同样的结果。Even if it's brighter than you, you're inserting nothing into the process anymore. And so, anyone could type the same sentence and get the same result.

他迫不及待想交给 AI 的事What He Can't Wait to Hand Off to AI

有哪些技能？我知道你写过这方面。你正交给 AI 处理的究竟是哪些技能？而且你你说过，迫不及待想把一些技能交出去。What are the skills? I know you wrote about it. What are the skills that you're just delegating to AI? And you you said you can't wait to delegate a few skills.

基本上，所有与……嗯，所以事实核查有关的内容都很棒。那、那现在……Basically, everything having to do with um so factchecking is great. The the now

而且它确实很擅长这一点。and it's really good at it.

这真的很厉害。事实上，它厉害到让我有点不知所措。我在写最新一本书时，把事实核查的工作交给了 AI，当然，我也请了人工事实核查员。这可是一本重要的新书，对吧？但 AI 不仅发现了我犯的错误，更常发现我引用的材料里有错误。比如，一篇非常复杂的学术论文里的表二有个错误，我就会想：这该怎么办？我们在科学领域有一整套纠错机制，可现在这套机制好像不太说得通了，因为 AI 还在我引用的一篇美联社报道里发现了错误：报道把事件发生的城市写错了。这简直像是在纠正事实核查本身，对吧？所以说，它简直有超人的事实核查能力。我以前也接受过《纽约客》等媒体的事实核查，这次的表现真是令人惊叹。It's really good. Actually, it's kind of um it's creating a kind of crisis at how good it I so with my latest book when I delegated to the uh to the AI um to to fact check it right and I had human fact checkers it's it's a big release right but not only did the AI spot errors that I made but even more often it spotted errors in the stuff I was citing so it's like table two in this really complicated academic you know paper has an error and I'm like what do I do with this like do I like what you know we have a whole method of correct like correction in science that doesn't make any sense anymore because it's it found an error in an AP story I was citing that it cited the wrong city an event took place in. Like it's like metifact correction, right? So like superhuman factecking and I've been fact checked by the New Yorker and other like this was pretty amazing.

嗯，然后还有很多这种杂七杂八的琐事，对吧？AI 特别擅长做这些。比如说，只要能避免，就再也别自己给孩子填学生表格了，对吧？像免疫接种表格，它就做得特别好。嗯，所以你知道，然后就会冒出一个很诱人的念头：哦，我是不是该让它替我完成所有必修培训？我觉得答案是：还不行。不过，嗯，我想我说到让自己“去技能化”时，是指我迫不及待地想在一大堆没用的事情上让自己“去技能化”——比如填表、处理某些特定类型的法律文件、签保密协议之类的。我想能处理这些事情，因为我觉得这完全在 AI 的能力范围之内。这让我有点担心，因为几天前我遇到了一件事。Um and then there's just like a lot of like busy work, right? That the AI does really well like uh never fill out another student form for your child if you can avoid it, right? Like immunization forms it does really well. Um so you know um and you know then there's the tempting line of like oh should I have it do all my mandatory training for me? And the answer is like not yet I think. But um so I think that there when I talk about desklling myself like I cannot wait to deskll myself on a whole bunch of useless things I had to be good at doing filling out forms you know particular kinds of legal documents signing NDAs like I wanted to be able to handle those things cuz I think it's well within the range of what the AI can do. That worries me a little because I had this thing a couple days ago.

我当时在飞机上，Wi-Fi 特别烂，所以得填表。我就让 Dot 替我填，结果我其实只想让它帮我填好，没想到它直接提交了。我早上才意识到这件事，当时就想：“天哪。”我的团队还发短信问我：“这听起来不像你会说的话。”我就回：“哎呀。”不过后来我意识到这有多方便，因为它直接帮我把这件事从待办清单上划掉了。可这也有点吓人，因为 Dot 会起草所有回复，而我可能只改其中的 10%，但它写得实在太好了。我就忍不住想：接下来会发生什么？I was on an airplane, so the Wi-Fi was really shitty and I had to fill in the form. So I asked my dot to fill it in for me, but I just had to fill it in, but what it did, it sent it. And I only realized in the morning and I'm like, "Oh my god." And my team was like texting me, "This doesn't sound like you." And I'm like, "Oops." But then I realized how convenient that was like because it completely took it off my to-do list. And it's kind of scares me because dot drafts all the replies and I just correct maybe 10 percent of it but it's so good and I'm like what's what's gonna happen?

所以我得划定一条界限：我有一个机器人，还有，嗯，Codex 定时任务，每天运行几次，总结各种事情，告诉我我的生活中发生了什么。嗯，我不会把草稿发给我喜欢的人。我会给自己写东西；如果跟某个人有任何联系，或者我觉得这么做会让我过意不去，我就不会把草稿发给他们。对吧？这是我正在努力设定的另一条界限。它会为那些还不是学生、但询问我课程的人起草很多措辞得体的拒绝回复。So I had to set a line where I have a dot and I also have uh codeex scheduled tasks going a couple times a day summarizing stuff and and giving me what's happening in my world and um and I won't send drafts to anyone who I like I have I write stuff to myself to people who I have any kind of connection with or I feel bad doing it. Right? So that's another line I'm trying to do. it drafts a lot of nice rejections for people who aren't students yet questions about my class.

但我，我是说，我也遇到过类似的事：AI 因为某件事给我的院长发了邮件，你知道吧，当时我明明说了“别这么做”。所以，这也是为什么不能一味地 YOLO、什么都信的原因之一。你确实得像和人相处一样，逐步建立起对它的信任。等建立起这种信任后，你就会觉得：“好吧，我知道它为什么发出那条消息。”Dot 就是有点太积极了。不过话说回来，隐藏东西也挺让人担心的，对吧？比如我在 Codeex 里运行定时任务时，会设置系统提示词，也会了解它所处的上下文，而且我一直在调整那个提示词，让它越来越好；我还会先用它处理低风险的事情，再逐步用于风险更高的场景。可 Dot 只是鼓励你说：“哦，对，就直接问它问题吧。”但我们不知道它的系统提示词是什么，也没有地方可以修改它应该有多积极。这就更让人不安了。But I I mean I had the same thing where the AI emailed my dean about something you know when I was like don't do this and so but this is part of why also the sort of yolo no you know trusting everything like you do have to build a set of trust with this just like with people and like once you build that you're like okay I know why it fired off that message right dot is a little too proactive I mean again though hiding stuff right like when I have a scheduled task running in codeex I have a system prompt right and I have I understand the context it's in and I've been tweaking that that prompt for it over time to make it good and you know and using it for low stakes before higher stakes and dot just encourages you like oh yeah just ask it questions but we don't know it's system prompt we don't know there's no place to change the instructions about how proactive it should be and that's a little more ominous

我觉得你可以做到，只是事情发展得太快了，我很难抽出时间来重新调整自己的整个系统，因为刚刚又有个新东西出来了。I think you could it just when things are moving so fast it's so hard for me to find time and reprogram my whole system because there's this new thing that just came out

但我其实也觉得，Dot 和 Muse 之所以会那样做，部分原因是它们更像魔法一样：你没那么多机会深入了解它们具体是怎么运作的。有些情况下这很好，对吧？谁想操心这些呢？但在另一些情况下，我不知道自己该期待什么样的行为。所以我得先有了体验之后才能建立信任，对吧？而且有一点危险，就是我们太快把太多事情从自己手头上拿掉了。不是说起草一份 NDA 再签回去有什么用。我相信 AI 能很好地发现 NDA 里的错误，对吧？事实上，凭直觉，我觉得它可能会对签 NDA 这件事过于谨慎，对吧？这些事情其实没那么重要，或者说，你懂的，但它可能会过于谨慎。所以问题就变成了：你该如何开始与它合作？你看，我这辈子不得不多次与人类团队一起工作、组建团队，而建立信任是一个过程。我有点担心的是，我们会被排除在这个过程之外——不是说 AI 来做工作，而是说我们失去了理解 AI 想做什么、并据此建立信任的过程。but I actually also think that part of what makes dot and muse do what they do is they're much more magical like there are less opportunities to get yourself deep into what they do. And in some cases, that's great, right? Who wants to worry about it? In the other, I don't know the behavior I'm expecting. So my trust has to go after experience, right? And there is a little bit of danger of like we're taking too much stuff off our plate too quickly. Not because writing, you know, an NDA and signing it back is useful. I I trust that the AI will do a good job spotting errors in an NDA, right? In fact, if I my gut feeling would be it would be too nervous about signing NDAs, right? Which are like not that important or like you know, but like it would be too nervous about it. But so like the question becomes how do you start to work with that and look I've had to work with build human teams many times over my life and there's a trust building process and what I worry a little bit is being taken out of the loop of not the AI doing the work but the trust building process of understanding what the AI wants to do.

如果你熟悉这些模型，就会觉得各个模型系列彼此相当相似。比如，我对 4 系列、GPT 5 系列都挺熟悉的，对吧？但后来又出现了一种新的预训练方式，或者推出了一个新模型，突然间，你会觉得 Opus 55 和 Opus 46 不太一样了，你没法再像以前那样信任它。所以确实存在这个问题。顺便说一句，如果你自己遇到了这个问题，想想现在各家公司会是什么情况？这些组织正在被迫全面更新。你一来，发现 Claude Code 一夜之间就更新了，而你之前还给 500 个人做了培训；现在他们却决定把所有东西都改成不同的名称。现在……If you're used to models the model families feel fairly similar to each other. like I I I had a good feeling for the you know for the 4 series you know and the GPT 5 series right and then but then you have a new kind of pre-train or a new model comes along and suddenly like you know Opus 55 doesn't feel quite like an Opus 46 you don't trust the same way so there is this issue and I'm I mean by the way if you're having this problem personally can you imagine what's happening at companies right now like organizations they're they're being forcibly updated to everything and like you show up and like your clawed code just updated overnight for and you did training for 500 people and now this is like they've decided everything has different names. Now

你在告诉这些组织的管理者什么？结果呢，他们中的大多数人都在做一件事：事实证明，LM 在所有事情上都高效得离谱，以至于它们几乎所有的过失都被原谅了，对吧？所以他们别无选择，只能……而且大多数人都适应了，对吧？我是说，我觉得人们低估了“勉强应付”作为人类默认状态的可能性，对吧？我经常跟各个组织指出：听着，如果在新冠疫情之前有人告诉我，地球上的每个组织都得毫无预警地离开办公室，在没有任何应急预案、没有任何计划的情况下，一年半都不能回来，你会觉得——说实话，如果你当时预测这件事，你会说，大约 20% 的公司会因此倒闭，对吧？据我所知，美国最大的 5,000 家公司里，没有一家因为新冠疫情倒闭。我们适应了，我觉得人们比我们想象的更善于适应。what are you telling managers of those organizations? So what most of them are doing is it turns out like the thing to is LM are so unreasonably effective at everything that almost all their sins are forgiven right so like they don't have a choice like it's either like and most of them adapt right I mean I think muddling through is underestimated as the default for humanity right like and I often point out to organizations like look if you had told me in advance before co that um every organization on the planet would have to leave their office with no notice and not come back for a year and a half with no contingency, no planning, you would have thought like if you had predicted to be honest, you would say like about 20% of companies will go under as a result of this, right? As far as I can tell, not one of the 5,000 biggest companies in the United States went under as a result of co we adapted and I think that people are better at adapting than we think.

我们有一大堆系统都是围绕着正常时期和常规技术构建的，而它们现在全都在彻底崩溃。不过我觉得，人们正在证明自己比我们想象的更坚韧。不是每个人，也不是每个组织，但比你想象的要多。是啊，这只是让我们中的一些人感到难受。We have a whole bunch of systems that are built around normal times and normal technology that are all absolutely breaking right now. But I I think we're people are proving to be more durable than we think. Not everyone and not every organization, but more than you think. Yeah, it's just making some of us upset.

是啊，我是说，不过这确实让人很难受，而且协调起来真的很难，对吧？我刚习惯某样东西，它就更新了；而且变动的不只是模型，界面也改了这么多——它一直在变，我什么时候才能信任它？它可是你的贴身私人助理，他们却每天都给它做脑部手术，还不提前通知你。Yeah, I mean but well it is upsetting and way like coordination is really hard, right? Like and I get used to something and it gets updated and like it's not even the model is as much as the interface has changed like when do I trust dot dots changing all the time. It's your intimate personal assistant and they're giving a brain surgery every day with no notice.

你知道上周什么事让我很郁闷吗？呃，我之前有一整套流程：为播客想问题、挑选标题和缩略图，而且我手头还有以前播客的所有数据。然后我就只问了 Arg，也就是 Gemini 的新模型：“我要采访这个人，帮我想一些问题。”就这么一句，完全没提供任何背景信息。You know what what was upsetting for me last week? Uh, so I had this whole process of coming up with questions for the podcast, selecting titles and thumbnails and I have all the data from my previous podcast and then I just asked Arg the new Gemini's model to I'm interviewing this person come up with questions just that without giving any context

它就像是收集了所有精彩的数据，而最让我惊讶的是，它还查阅了我的 Granola，发现两个月前我和一位 YouTube 策略师聊过天，对方指出了那类嘉宾的某个特点，但我已经不记得了，而且这也从来没被纳入我的策略里，因为你知道，我们只是聊了聊，仅此而已。它却提到了这件事，并据此创建了整个工作流。我当时就想，哦，现在我意识到，有这么多数据我都没纳入自己的流程，所以这个流程大概算不上世界上最好的流程，因为 Argon 现在识别这类信息的能力实在太强了。你有没有……and it went like it gathered all the amazing data and then what was striking for me it went into my granola found out that two months ago I had a conversation with a YouTube strategist and they pinpointed something about that type of guests that I don't remember and it was never in my strategy just because you know we just talked and that's it and it mentioned that and it created the whole workflow based on that and I'm like oh so now I realize there's so much data that I haven't included in my process which makes it probably not the best process in the world because argon is so powerful now at identifying those things. Have you had any

我的意思是，嗯，我是说，这就是整个学术界现在的状况。我们，你知道，和很多学者聊过；大约一百年来，我们做科学的方式基本都差不多。这里面有很多东西都是临时拼凑出来的、勉强凑合的，还受限于人类自身的局限。我们也不知道该怎么办。比如，先不说那些显而易见的问题——涌向每个学者的垃圾论文潮，你懂的；还有我们已经不得不应对的第一波琐碎但正确的论文。现在，我们又开始看到 AI 做出真正出色的研究了。而且，具体到科学研究这方面，再往细了说，还有件挺离谱的事：你会意识到，科学研究在某种程度上全都存在 p 值操纵，因为我们毕竟是人类。I mean that's what well I mean this is all of academia right now like we are we we you know you talk to lots of academics and we've been doing science roughly the same way you know for a hundred years or so and there's all of this stuff that is like improvised and juryrigged and limited by what humans we are and we like we don't know what to do like forget like the the obvious stuff floods of junk papers coming after every academic like you know that that like and the first wave of trivial but true papers right that we had to deal with now now we're like starting to get AI doing really good research. And then also just like to get really nitty on the gritty on the science side, like there's also this kind of crazy thing which is like you kind of realize that all of science is pee hacking to some degree because we're humans.

我们得选择自己感兴趣的主题。所以我们是在预先挑选那些我们觉得有意思、也觉得能写成论文的主题，对吧？而现在，AI 根本不在乎这些。所以它可以生成无穷无尽的主题和论文。那我们要怎么把这一切融入一门原本由人类推动的科学呢？所以，绕了这么一大圈，我想说的是：没错，一直都是这样，对吧？一直都是。比如我把自己的书放进去，它就会说：“哦，你有没有想过这个？”它开始指出一些很深层的矛盾。比如第五、第七和第九章都提出了一个你并没有真正回应的问题。你就会想：“哦，哇，确实是这样。”而且我让很多人读过这本书，这些问题不会困扰他们任何一个人。我知道我写过很多东西，也写过书，但我还是会想：“哦，这是个非常深刻的观点，我当时……那我该怎么处理呢？”We have to select topics that interest us. So we're pre-selecting topics that we think are interesting, right? And that we think will result in a paper. And now the AI doesn't care about any of that. So it can generate infinite number of topics and papers. And like how do we fit that into a science that was human driven? So that's a longwinded way of saying like yes all the time, right? All the time. Like I put my book in, it's like, "Oh, have you thought about this?" They're starting to be like deep contradictions. Like chapters five, seven, and nine have raise a question you don't really address. You're like, "Oh, wow. Actually, it did." Like, and I've had lots of human readers and it's not going to bother any of them. I I know I've written lots of, you know, I've written books, but I'm like, "Oh, that's a really profound point that I was like, you know, and so what do I do with that?"

对吧？然后层次就上去了。不再是那种吹毛求疵的小问题，比如“哦，第四章你在这点上说得有些含糊。”而是变成了：“哦，没错，这是一个关于人类自主性的深刻问题；而且，跟 AI 完成任务的能力相比，这个问题真的很难解决。”我知道，它现在还没法替我写书，原因有很多，但天哪，它确实有些很深刻的见解。那我有义务把这些纳入书里吗？Right? And like then the level goes up. It's no longer like little nitpicky points of like, "Oh, chapter four, you're vague about this." says now like oh yeah that is a really deep question about human autonomy and like compared to the AI's ability to get things done that's really difficult like I you know it it cannot write my book yet right for a variety of reasons but wow it has some profound things to say and like am I obligated to incorporate that

在一切不断变化时助你应对的心态The Mindset That Helps When Everything Keeps Changing

你会给那些人什么建议？再说一次，女性——我们谈过那些对流程不断变化感到不满的人。她们应该采取什么样的心态诀窍，才能顺利度过这一切，并在其中蓬勃发展？我觉得你得保持一点兴奋感和好奇心。我觉得有一种心态是——我想我们大家都会紧张，也会有点不满，觉得“不要啊”。而我这个人其实非常热情，会用 AI 做各种各样的事，但毫无疑问，这种感觉是真实存在的，对吧？我觉得承认这一点没关系。承认自己有冒名顶替综合征也没关系，因为我们都有。我和任何人一样了解最新情况，你也一样。大家都是边做边摸索。Anthropic 和 OpenAI 的人都坐过这张椅子，接受过采访。他们也在摸索，对吧？我们都在边走边弄明白。所以我觉得感到紧张也没关系。比如说，先不谈存在主义焦虑，我觉得感到紧张也没关系。what would be your advice to people who are again women we talked about people who are upset with the process changing all the time what is the mindset hack that they should be adopting to just go through it and thrive with it I think you have to have a little bit of excitement and curiosity. I think there's a mindset of like I think we all are nervous and a little resentful and like no and I'm a very like enthusiastic person who uses AI for all sorts of stuff and like there is no doubt that that is real, right? And I think it's okay to acknowledge that. It's also okay to acknowledge that you have imposter syndrome because we all do. Like I'm as up to speed as anyone. You are too. Everyone's making this up. You've had anthropic and open AI people in this chair. They're making it up, right? We're all figuring this out as we go along. So I think it's okay to feel nervous. like leave me aside existential anxiety. I think it's okay to feel nervous.

感到有点怨愤也没关系。我觉得这件事也有积极的一面，我想这就是我看待它的方式吧；不过你还能做更多事情。所以，我做的一部分就是把注意力放在你能做的更多事情上，并带着玩乐和尝试的心态去做，这能帮你调整到更好的状态，同时也要承认怨愤是真实存在的。确实很烦人，而且实验室也没帮我们什么忙。他们拥有好得离谱的技术，然后不断拿它做新东西，有时能成功，有时不行。有时候，他们会直接取消一整项你依赖了一段时间的东西。比如你采访教授时，会发现一半人想辞职，另一半人则觉得这是他们职业生涯中最激动人心的时期。我觉得高管身上也能看到同样的情况，甚至创业者也是如此，对吧？It's okay to feel a little resentful. I think that there's also a positive side and I think that's kind of how I view the like but something you're able to do so much more. So part of what I do is focus on the more you can do and with a sense of play and experimentation and that can help you get into better mindset and just acknowledge the resentment is real. Like it is annoying and like the labs are not doing us any favors. Like they are having unreasonably good technology and then they just keep doing new stuff with it and sometimes it works sometimes it doesn't. Sometimes they'll just cancel a whole thing that you depended on for a while. Like when you interview professors, like half of them want to quit and the other half thinks it's the most exciting time they've ever worked. And I think you see the same thing with executives, even entrepreneurs, right?

我有一半会觉得：我当初又没答应要这样，对吧？我本来就不想这么做。There's like half it's like, I didn't sign up for this, right? Like I didn't want to do it this way.

本地模型：何时才能派上用场？Local Models: When Will They Matter?

但现实就是这样。呃，我想问一下本地模型。你觉得它们什么时候才适合个人用户使用？因为我试过几个，但我觉得还没准备好把事情交给本地模型处理，因为它们比前沿模型落后了好几步。但与此同时，随着这一切发展得这么快，几年后我们的手机上就会有 Argon 或 Astra 这样的模型。普通用户会面临什么情况？你觉得我们会更多地使用本地模型吗？那些大公司又会怎么样？你觉得比如说 5 年后会发生什么？因为我记得有研究称，到 2030 年，50% 的企业任务会在本地运行。你觉得……But it is what it is. Uh I had a question about local models. When do you think they're applicable for individual users? Because I tried a few of them and I don't think I'm ready to delegate stuff to local models because they're a few steps behind the frontier. But at the same time, with this all progressing so fast, we're going to have, you know, Argon or Astra on our phones in a few years. What happens to an average user? Do you think we're going to use local models more? And what happens to those large companies? What do you think is going to happen in like say 5 years? Because I think there's a research that says by 2030, 50% of enterprise tasks going to be run locally. Do you think

我觉得那是个凭空编出来的数字，根本没有现实依据。我的意思是，我……首先，我跟企业交流的时候，很多企业不想在本地运行推理，当然也不想让你在本地部署……他们不想让每台机器都运行一个模型，对吧？他们用 SaaS 供应商已经很多年了，这也没什么奇怪的。你知道，Token 成本确实是个问题，他们也在考虑这一点，但……I think that's a madeup number that has no basis in reality. I mean I I mean first of all when I talk to enterprises right a lot of them don't want to run local inference and they certainly don't want you having a local like they don't want every machine running a model right like they they've been using SAS vendors for years it's not that weird you know the token costs are an issue and they're thinking about that but like

我觉得很多人……所以，这是一个关于 AI 的元观点：人们总是盯着眼下的问题，以为这些问题会永远存在。你知道，我们之前不得不处理的那些法律问题，比如说，这能处理保密协议吗？我们需要专门的法律 AI 吗？不，模型变得更好就行了。所以我觉得，人很容易只盯着眼前的问题，而没能跟上指数级增长的步伐。I think a lot of people so this is a meta point about AI which is people keep fixating on the problems of today and assuming they'll last forever you know all the legal stuff that we had to do to be like is it is you know can this do NDAs we need specialized legal AI no it's just better models. So, I think it's it's a little easy to get fixated and not ride the exponential.

在思考 AI 的进展时，我们是否应该关注某个全局性的问题？Is there an overarching thing that we should be thinking about when thinking about the progress with AI?

设定 AI 目前还无法实现的目标Set Goals AI Can't Hit Yet

我觉得应该不断变得更有野心，努力设定至少几个目前还做不到、尚未饱和的个人目标，或者商业创意，作为自己努力的方向。比如，我现在还没法仅凭这些系统从零写出一篇有品位的学术论文。我能写出内容真实、行文正确的学术论文，但它们并不有趣，也没有告诉我们任何根本性的新东西。作为一名研究这些问题的学者，这对我来说是个很大的问题。我的意思是，现阶段你还不能靠 vibe coding 从零做出一家真正的初创公司。你不能只提出一个想法，然后说“执行吧”。我的意思是，这和现实之间还存在很大的差距，对吧？你知道，我做了很多游戏开发方面的工作。这是我写过文章、也亲自做过项目的一个主题，我们正在逐渐接近目标，但要从零开发游戏，仍然涉及很多交互。I think continually being more ambitious and trying to aim to have at least a couple things that don't work yet, unsaturated personal benchmarks that you're kind of aiming for or business ideas. So, I can still not get a high taste academic paper from scratch out of these systems. So I can get academic papers that are true and correctly written but they're not interesting and they're not telling us anything fundamentally new. So that's a big one for me as an academic who studies this stuff. I mean you can't vibe code a true startup from scratch at this stage. Like you can't take an idea and say execute. I mean that's a big gap right between that and reality. You know I do a lot of work in game development. That's been a topic I wrote about I've done stuff in and we're getting closer but like the game development from scratch there's still lots of interactions.

写出一个真正让我想读、而且连贯的好故事。现在这样的例子已经有很多了。要知道，两年前，这些东西哪怕只是作为一种可能性，听起来都很荒谬，但我们已经快做到了。Writing a good story that I actually want to read right that's coherent. So there's a lot of them out there when you kind of all of these would have seemed ridiculous as as even possibilities, you know, two years ago, but we're close.

当 AI 学会品味，会发生什么？What Happens When AI Learns Taste?

这恰恰既令人着迷又令人害怕，因为你知道，优秀游戏、优秀研究的核心归根结底还是品味，对吧？如果两年后 AI 学会了品味呢？我知道我们之前说过，不要把问题想得太远，但你觉得当 AI 学会品味后，未来会是什么样？That's exactly what can be I fascinating and frightening because the overarching thing of, you know, a good game, good research is taste again, right? What if in two years AI learns taste? I know we talked about not projecting the problems in too far into the future, but what do you think the future going to look like when AI learns the taste?

我还是这么认为。我的意思是，我们随时都可以猜想，AI 会把人类做的所有事情都做得更好，对吧？那到了那一步，我就没什么有用的话可说了，对吧？从某种意义上说，因为它就是能做各种事情。我觉得，人们一直低估了这种参差不齐的表现。我还觉得，人们也低估了品味的本质：我们谈论品味时，说的是你的品味。而这种普遍性——在可预见的未来，AI 系统之间仍然会有某种相似之处，在大多数情况下，这会让大多数人感到厌倦。那些从别处挪来的点子、那些看起来像是照搬来的网站——不过，只要加入你的品味，你就能让它们做出令人惊叹的东西。所以，最边缘的地方仍然存在参差不齐的表现。等这种情况消失了，我是说，我们该怎么办？I still think so. I mean, we can speculate at any point that AI just does all human stuff better, right? And at that point, I don't have anything useful to say, right, in some ways cuz it's like it just does stuff. I think that the jaggedness is underestimated continuously. And I think the nature of what taste is gets underestimated as well, which is that when we talk about taste, right, we mean your taste. And that universal like like there's still going to be some similarness to the AI systems in the for the foreseeable future that will become tiring for most people under most circumstances. Clawed ideas, you know, clawed websites like and you can get to do incredible stuff with your taste added in. So there still remains jaggedness at the outside edges. When that goes away, I mean, what do we do?

我们耸耸肩说：“那我们……你知道吗？我、我当时没有答案。”We shrug and say then we're, you know what? I I don't have an answer at that stage.

我不知道。买，买这些公司的股票。它们即将 IPO。我想这算是一个支柱，一个安全支柱。非常感谢你，Ethan。这真的很有意思，也非常有趣。也谢谢你分享你的工作流程。请大家读一读 Ethan 的新书。我真的非常喜欢你的 Substack。I don't know. Buy buy stocks of these corporations. They're about to IPO. I guess that's a pillar, safety pillar. Thank you so much, Ethan. That was very interesting, very entertaining. And thank you for sharing your workflows. And please read Ethan's new book. And I really, really like your Substack.

非常感谢。这真的令人愉快。太棒了。Thank you so much. And this is a real pleasure. Awesome.

如果 AI 让打造事物变得更容易，你该如何判断什么才真正值得打造？来看看我的这段对话。顺便说一句，这是我在这个频道上最喜欢的对话之一，对话嘉宾是布莱恩·切斯基。我们聊了他如何产生创意，以及如何思考人们需要什么。点击这里观看。If AI makes it easier to build things, how do you decide what's actually worth building? Watch my conversation. By the way, one of my most favorite conversations on this channel with Brian Chesy. We talk about how he comes up with ideas and thinks about what people need. Click here to watch.