AI 炒作与历史技术变革：Benedict Evans 访谈

鉴于我们刚才关于基础模型的讨论，你在关注重点上是否应该有所不同？我对 Sambleman 很有共鸣。我是 Redpoint 的 Jacob Efron，这里是 Unsupervised Learning，一档播客，我们探讨 AI 领域最敏锐的思想家对当今生态系统的看法以及未来走向。今天，我们与 Benedict Evans 进行了一场精彩的对话。他是我在这个领域最喜欢的思考者之一。我喜欢阅读他的通讯、演示文稿，并有机会与他坐下来讨论很多事情。我们谈到了 AI 中价值将如何积累，以及他认为基础模型的最终估值可能是什么样子。我们讨论了为什么消费者 AI 应用如今并不真正奏效，以及能力的锯齿状边缘及其对企业整体采用的影响。我们还涉及了他关于工作使用的许多有趣观点，无论这是否是一项赋能技术。能与我最喜欢的思考者之一谈论这些事情，真是太棒了。我想大家会喜欢这次对话。闲话少说，有请 Ben。

Given this discussion we've had on the foundation models, should you be doing anything different in terms of like focus? I have a lot of sympathy for Sambleman. I'm Jacob Efron from Redpoint and this is Unsupervised Learning, a podcast where we probe the sharpest minds in AI about what's happening in the ecosystem today and where the future's headed. Today we had an awesome conversation with Benedict Evans. He's one of my favorite thinkers in the space. I love reading his newsletters, presentations, and had the chance to sit down with him and talk through a bunch of things. We talked about where value will acrew in AI as well as what he thinks the ultimate valuations of the foundation models might look like. We talked about why consumer AI applications aren't really working today and the kind of jagged edge of capabilities and their implications on enterprise adoption overall. And we hit on a lot of his really interesting takes on job usage. Uh whether this is an enabling technology or not. Just awesome to get to talk with one of my favorite thinkers about a bunch of these things. I think folks will really enjoy the conversation. Without further ado, here's Ben.

嗯，非常兴奋能这样做。我觉得你是科技领域最有思想的人之一。我喜欢阅读你做的演示文稿。显然，你对整个生态系统有着巨大的影响。今天，我很兴奋能深入探讨你对当前生态系统的看法、模型实验室的未来、价值积累的地方。但我真正想开始的是当下的炒作，因为我觉得你可能有更温和的看法。说这是像互联网和移动一样大的事情，但可能不像工业革命那样改变文明，这很有趣，但我想从这里开始。你关注什么可能会改变你的想法？因为我相信很多听众会不同意。

Well, very excited to do this. I feel like you're one of the most thoughtful people on tech. I love reading the presentations you do. You've obviously, you know, have a huge influence on the ecosystem at large. And today, I excited to kind of dig into, you know, your opinions on the, you know, ecosystem as we have it today, the future of the model labs, where value accr. But where I really want to start with is the hype in the current moment because I feel like you have maybe a more subdued take. It's funny to call a subdued take being that this is as big a deal as the internet and mobile, but maybe not civilization altering like the industrial revolution, but I want to start there. What do you like paying attention to that might change your mind on that? because I'm sure a lot of our listeners will will disagree.

我不确定你能在多大程度上分析说，嗯，这是否像移动一样大，或者像互联网一样大，或者像计算机一样大？移动是否比 PC 更大？我很难量化这些东西。我想我的出发点是说，看，互联网是一件大事，确实改变了一切，移动和 PC 也是如此。是的，我当然可以说，网络比 PC 更大，因为 PC 大约有 5000 万到 1 亿台设备。你知道，当 Mark Anderson 推出 Netscape 时，世界上大约有 7500 万到 1 亿台 PC，而现在有 50 亿部智能手机。所以显然，随着时间的推移，它变得更大。我想我有点纠结的是，以前从未发生过类似的事情，我们以前也没有技术消灭大量工作，我们正在以前所未有的方式自动化事物。前几天我和一家大公司的某个人交谈，他是个非常聪明的人，拥有博士学位，精通各种计算机科学，他说大公司以前从未处理过人们为自己构建工具和系统的情况。我想，“那么，你听说过影子 IT 吗？你听说过 Excel 吗？”是的，这是不同的，但它总是不同的。我们确实经历了这些大变化，它们总是不同的，而且更大，但这并不意味着你不应该回头看看，说，嗯，过去五次我们拥有改变世界的技术时发生了什么？我认为争论哪个更大或更小没有多大意义。我认为有趣的是说，嗯，过去五次我们遇到类似情况时发生了什么？是的，你可以挥挥手说，“不，这就像电力。”我想，“好吧，就像电力。”那么电力发生了什么？在电力出现之前有一个时代。电力出现之前的人也不笨。人们也在挠头，试图弄清楚该怎么处理它。所以，不要想象以前从未发生过什么，这很有用。坐下来挠头思考，嗯，你能看到什么模式？我上周末写了些关于 token 定价的东西。文章的一个部分是关于移动是如何运作的？半导体是如何运作的？光纤是如何运作的？操作系统是如何运作的？这些都没有预测能力。你不能通过争论它与移动的相似程度来证明这会以某种方式运作。但你应该回头看看移动发生了什么，然后想，好吧，这告诉我们未来可能会发生什么？所以你可以指出半导体，说你有这种不断升级的成本结构，即摩尔定律，你知道，尖端晶圆厂的成本每四年翻一番。

I'm not sure how analytic you can get about saying, well, is this as big as mobile or as big as the internet or as big as computer? Was were mobile bigger than PCs? Like I I I struggle to kind of how you can kind of quantify those things. I think the the place that I come from is just to say, look, the internet was kind of a big deal and did absolutely change everything and so was mobile and so was PCs. And yes, I think you certainly kind of can say, you know, that the web was bigger than PCs cuz PCs was like 50 or 100 million devices. You know, when Mark when Mark Anderson launched Netscape, there were like 75 million 100 million PCs in the world and now there's 5 billion smartphones. So clearly, it's got bigger over time. I suppose the thing I sort of struggle with is that nothing like this has ever happened before and we've not had technology wiping out great sways of jobs before and we're automating stuff in a way that this has never been done before and I kind of I was I was I was talking to somebody the other day at a big company and they were you know very smart guy PhD and all sorts of computer sciency stuff and they said you know big companies have never had to deal with people building tools and systems for themselves before and I thought, "So, have you heard of shadow it? Have you heard of Excel?" Yes, this is different, but it's always different. And we do go through these big changes, and they're always different, and there was bigger, but that doesn't mean you shouldn't kind of go back and look and say, well, what is it that happened the last five times we had a world changing piece of technology? I don't think there's much percentage in arguing about which one is bigger or smaller. I think what's interesting is to say, well, what happened the last five times we had something like this? And yes, you can wave your hands and say, "No, this is like electricity." I'm like, "Okay, fine. It's like electricity." Well, what happened with electricity? And there was a time before electricity. And people before electricity weren't dumb either. And people were kind of scratching their heads and trying to work out what to do with this. And so, it's just useful not to imagine that nothing ever has happened, never happened before. And to kind of sit and scratch your heads and think, well, what are the patterns that you can see here? I wrote something at the end of last week talking about token pricing. And one of the sort of blocks of the essay was well how did mobile work? How did semiconductors work? Um how did fiber work? How did operating systems work? And none of those have predictive power. Like you can't prove that this is going to work in a certain way by arguing about how close it is to mobile. But you should go back and look at what happened at mobile and think okay so what does this tell us about what might happen? So you can point to semiconductors and say you had this escalating cost structure box law that said that you know the cost of a cutting edge fab doubles every four years

而基础模型看起来有点像那样。你可以看看移动，说移动有边际成本，比如增加更多，很多科技界的人不理解这一点。所以移动网络有边际成本。如果你增加流量，你必须建造更多的基站，并在基站上安装更多设备，所以你有边际成本，这看起来很像我们过去 6 个月经历的供应紧张，每个人都买了 iPhone 并开始看 YouTube，网络崩溃了，他们不得不改变定价方案。但这个比较的有趣之处在于，移动数据流量在过去 15 年里增长了大约 1000 到 2000 倍，这是一个万亿美元的行业，他们每年在资本支出上花费 2000 亿美元，但他们不赚钱。

and that's and foundation models look a little bit like that. You can look at mobile and say well mobile had marginal cost like adding more a lot of people in tech don't understand this. So mobile networks have marginal cost. You add have you have if you double the number of traffic you have to build a bunch bunch more base stations and put a bunch more equipment on the base stations and so you have marginal cost and that looks a lot like the supply crunch we've had in the last 6 months where everyone got an iPhone and started watching YouTube and the networks fell down and they had to change pricing street schemes. But the interesting part of that comparison is that mobile data traffic in the last 15 years has gone up by like one or 2,000 times and it's a trillion dollar industry and they spend $200 billion a year on capex and they don't make any money.

我的意思是，他们赚了一点钱，但所有价值都在别人那里，你知道，他们没有做成 Uber，他们没有运营你的银行，他们没有做 YouTube，那都是其他人做的，所有价值都往上层堆。显然，这是关于 AI 的问题之一。你可以指云计算，你可以指操作系统。Sam Altman 说我们会像 Windows 一样，但像移动端 LLM 没有网络效应，所以看起来它们不会像 Windows 那样。重点是，你看，有一堆不同的竞争动态、行业动态，不同层面会发生什么。你能向上竞争吗？你能向下竞争吗？你可以从所有这些中学到东西。

I mean they make a bit of money but like all the value is other people you know they didn't get to do Uber they don't run your banking they don't do YouTube that's all other people and all the value went up stack and obviously that's one of the questions about about about about AI you can point to cloud you can point to operating systems Sam Alman said we're going to be like Windows but like mobile LLMs don't have network effects so it doesn't look like they're going to turn out like Windows the point is to say look there's a bunch of different competitive dynamics industry dynamics what happens at different layers as a stat. Can you compete up? Can you compete down? And you can learn from all of those.

是的。我认为，要弄清楚这一切的长期宏观影响，有一件事很具挑战性，那就是显然这些模型的能力不是静态的，但你指出了这个领域一些重大的未解问题。感觉最大的问题就是，你知道，我们最终在模型方面能达到什么能力？我很好奇，什么会改变你的想法，让你认为这实际上比工业革命更重大？是模型能力吗？你知道，

Yeah. I think one thing that makes, you know, challenging to figure out like the, you know, long-term macro impact of all this is just obviously the capabilities of these models are not static, but you pointed to some of the big outstanding questions in the space. It feels like the biggest is just, you know, what capabilities do we ultimately get to on the model side? I'm curious like what would change your mind that actually this is a bigger deal than something more akin to the industrial revolution? Is it a model capability? you know,

嗯，所以我认为，你知道，你可以提出一个非常明显的反驳论点，那就是指出这些技术是复合的。部分原因就是为什么它发生得更快。嗯，但就像移动互联网不需要等待互联网，对吧？而且，嗯，互联网不需要等待个人电脑，个人电脑不需要等待半导体，半导体不需要等待，我不知道，特种化学品或者，你知道，你认为的使能技术，光刻。嗯，那不需要等待电力。而电力已经有了一个完整的制造业，等等。所以总是有这种复合效应，你知道，我的意思是，说互联网有多大的另一种说法是，互联网比个人电脑大 10 倍。

well, so I think, you know, you could make a very obvious kind of counterargument, which is just to point out that each of these compound. Part of that is that's why it's happening quicker. Um, but like mobile internet didn't need to wait for the internet, right? And, um, the internet didn't need to wait for PCs and PCs didn't need to wait for semiconductors and semiconductors didn't need to wait for, I don't know, specialty chemicals or, you know, whatever you think the enabling technology is, photoiththography. Um, and that didn't need to wait for electricity. and electricity had a whole manufacturing industry that already existed and so on. So there's always this kind of compounding effect and you know I mean another way of saying this is big as the internet is to say well the internet was 10 times bigger than PCs

对。

right

所以也许这比互联网大 10 倍。最大的区别，实际有形的区别，与之前的平台转变相比，是我们不知道这的物理极限。所以我们没有很好的科学理解，为什么这些模型工作得这么好，接下来会发生什么，而相比之下，比如你回到 2010 年，你不知道下一个 iPhone 会是什么，但你知道它不会有视网膜植入物、一年电池续航、飞行能力，并且售价 5 美元。你知道，在 1995 年的互联网，你不知道互联网会如何运作，你可能以为你知道，但你不知道，但你知道个人电脑大约 2000 到 3000 美元，大多数人没有，而且电信公司不会在下个月给每个家庭都拉光纤。你大致知道基本的物理参数，所以这显然是一种概念上的差异。嗯，但我认为然后你可以说，好吧，我们有这些我们不知道的东西。好的。然后你可以把它分解。所以可能有一个二元问题，你知道，有一个都市传说，可能真也可能假，在古巴导弹危机期间，有谣言说导弹已经发射，交易所的每个人开始抛售，一个资深交易员走出去开始买入，他说：“嗯，这是二元的。”你知道，要么谣言是真的，那样我们都死了，要么不是，那样股票就便宜了。你知道，如果这些东西，如果 AGI 真的发生了，ASR，随便你选什么缩写，你的想法概念。如果所有这些都发生了，那我们就有比担心中产阶级失业更大的问题，就像我们都是宠物，好吧。否则，那么好吧，让我们担心，试着弄清楚这对企业软件意味着什么。

and so maybe this is 10 times bigger than the internet. The big like difference like the actual tangible difference to previous platform shifts is that we don't know the physical limits of this. So we do not have a good scientific understanding of why the models work so well and what will happen next which was whereas with like you go back to 2010 you didn't know what the next iPhone would be but you knew it wouldn't have like a retinal implant and a one year battery life and flight and cost $5 you know in the internet like 1995 you didn't know how you had no idea how the internet was going to work you probably thought you did but you didn't but you knew that PCs are like $2 and $3,000 and most people don't have one and telos aren't going to give every household on earth fiber to the next month like you kind of knew the basic kind of physical parameters and so that's clearly a kind of a conceptual difference. Um, and but I think then there's then you can kind of say, okay, we we've got this this stuff we don't know. Fine. Then you can kind of break that apart. So there's there's maybe there's one kind of binary question which is, you know, there's there's like a um there's an urban legend may or may not be true that during the Cuban missile crisis, there's a leg there's a a rumor starts that the missiles have launched and everyone on the stock exchange starts selling and one veteran trader goes out and starts buying and he says, "Well, it's binary." you know, either the rumor is true, in which case we're all dead anyway, or it's not, in which case the stocks are cheap and you know, if this stuff like if you know, if if if AGI really does, you know, get a happen and ASR, whatever, pick your acronym, your thought concept. If all of that happens, then we've got bigger problems than worrying about middle- class unemployment like we're all pets, fine. Um otherwise, then okay, let's worry about let's try and work out what this means for enterprise software.

是的。嗯，那些是那种

Yeah. Um, those are kind of the

我喜欢这个比较。就像你甚至不，是的，就像现在模型进步的形状有点不可知。有一些我们担心的形状，以及它对其他一切的影响。如果是那样的话，有更大的鱼要炸，几乎就像你不会，你可能不会在快速变化的背景下写企业软件。

I love that comparison. It's like you're not even Yeah, it's like it's kind of unknowable right now what the shape of model progress looks like. There's some all the shape that we're worrying about and the implications it would have for everything else. There's so much bigger fish to fry if that's the case that it's almost like you wouldn't you probably wouldn't be writing about enterprise software in the uh in the context of which there's a rapid

完全正确。就像嗯，是的，完全正确。你知道，你要么可以担心那个。很多这样的对话最终都在寻找隐喻。就像，你知道，就像核武器。像这个，像那个。很多这样的对话最终听起来像凌晨 2 点，如果你是一个有点醉的哲学研究生，你会说的话，比如“嘿，老兄，你有没有想过，也许我们也没有结构理解？”就像“好吧，太好了，谢谢，我对此无能为力，我无法分析。”这就像 dart 谬误，dart 思想，你可以从第一原理推导出上帝的本质，不，你不能，我们不知道，我们会发现的。与此同时，嗯，让我们构建一些东西，弄清楚什么有效。嗯，我的意思是，你回到这里有一点回声。你回去，你读过 John Perry Barlow 的《网络空间独立宣言》吗？

Exactly. It's like um Yeah, exactly. You know, you either you can worry about that. A lot of these conversations end up in Hunt for Metaphors. It's like, you know, it's like nuclear weapons. It's like this, it's like that. A lot of these conversations also end up sounding like this the stuff you say at 2 o'clock in the morning if you're a slightly drunk philosoph philosophy grad student like hey man have you thought that like m maybe we don't have structural understanding either like well great thank you I can't do anything with that I can't analyze that this is like the dart fallacy dart thought that you could like deduce the nature of god from first principles like no you can't we don't know we'll find out meanwhile um let's build some stuff and work out what works um I mean you get back to there's a little bit of an echo of this. I you go back to you ever read the the John Perry Barlow Declaration of Independence of Cyerspace?

不，我没有。

No, I haven't.

天哪，你太年轻了。

God, you're so young.

所以 John Perry Barlow 有点像 Grateful Dead 的作词人，我想。所以我太年轻，不太了解这个，但不管怎样。他写了这个非常诗意的作品，就像“你们这些钢铁和石头巨人，你们在互联网上没有主权。”人们有点认为这会结束战争。

So John Perry Barlo is like one of the Louiswises for the Grateful Dead, I think. So I'm too young to really know this, but anyway. So he wrote this beautifully poetic thing and it's like you giants of of steel and iron and stone. You have no sovereignty on the internet. And people kind of thought this was going to end war.

是的。

Yeah.

因为现在每个人都会互相理解。所以你必须记住，互联网周围也有大量这种狂热的千禧年乌托邦主义。有些是真的，有些不是。与此同时，让我们构建一些软件，弄清楚人们用它做什么。嗯，然后你就像，然后你把整个论点放在一边，对吧，OpenAI 看起来有点像 Netscape。聊天机器人是，这是一个网页浏览器，就像，再说一次，你不想对隐喻太死板，但你有这个东西，它很惊人，但它不工作。我认为这是一种更有趣的思考我们现在所处位置的方式，回到像 90 年代中期尝试使用互联网，或者尝试做移动互联网，更不用说 2010 年，2002 年。它有点工作，显然非常酷和惊人，它将改变一切。

Because everyone would understand each other now. And so you have to remember there was an awful lot of this kind of wildeyed millinarian utopianism around the internet too. Some of which was true, some of which wasn't. Meanwhile, let's build some software and work out what people do with it. Um, and that gets you like then so you set that whole argument aside right open air kind of looks like Netscape. Chat bots are this is this a web browser like and again you don't want to be too rigid on the metaphor but you have this stuff that's amazing and it doesn't work. And I think this is kind of a more interesting way of thinking about where we are now that go back to like trying to use the internet in the mid '90s or trying to do mobile internet, never mind 2010, 2002. And it kind of sort of works and it's obviously really cool and amazing and it's going to change everything.

我有一个非常清晰的记忆，1999 年我和一家移动软件公司开会，我带着 Palm 5，还有一部诺基亚手机，有 GPRS 和红外端口。我们试图通过红外把 Palm 5 连接到诺基亚，再通过 GPRS 上网，我们坐在那儿说：“天哪，这太酷了。”但实际上根本行不通，而且真正实现还要再等 10 年。显然，这次不是 10 年，但那种时刻就是：这里的构建模块是什么？价值在哪里？杠杆点是什么？什么样的体验能让数十亿人每天使用，而不是 1000 万人整天用，10 亿人每月用一两次——我们现在大概就是这样。我有点夸张。现在大约有 10% 到 15% 的人是日活用户。即使是日活，你也就是一天用一两次，不是一直用。还有 20% 到 40% 的人是周活或月活用户。你从社交产品知道，周活用户意味着什么。如果你一周只用一两次，那你显然不是硅谷梦想中那种“改变计算方式”的用户。

And I have this very vivid memory of sitting in a meeting with a mobile software company in like 1999 and I had my Palm 5 and I had a Nokia phone that had GPRS and an infrared port. And so we were trying to connect my Palm 5 to the internet over GPS over infrared to this Nokia and we're sitting there going, "Oh my god, this is so cool." And like but nobody that didn't actually work and no one is actually going to do that for another 10 years. Obviously, this isn't another 10 years, but it is that sort of moment of like what are the building blocks here? Where's the value capture? Um what are the points of leverage? What are the experiences that people are actually going to use to get billions of people to use this every day as opposed to 10 million people using it all day and a billion people using it once or twice a month, which is kind of where we are now. I'm exaggerating a bit. We've got like kind of like 10 15% of people are daily active users right now. And even daily active is you're still you're using this once or twice a day. You're not using it all the time. And then there's another 20 or 30% 40% of people who are like weekly active users or monthly active users. And you remember from social like weekly active user was So like if you're using this once or twice a week, you're clearly not in the dream of Silicon Valley that this changes computing.

嗯。

Yeah.

它是一个有用的新工具，有时有用。我们如何弥合这个差距？我觉得这是一个更有趣的分化：软件开发领域，这东西毫无疑问有产品市场契合度，它确实改变了软件开发的一切。然后有一群像我们这样的人，知识工作者，为自己工作，灵活自由，有很多事情在忙，你会大量使用它。这让我想起很多使用 Nation 的人。然后还有其他人，他们说：“嗯，有点用，挺酷的，但不会一直用。”那这种情况怎么改变？

It's a useful new tool. sometimes and how do we bridge that and I think this is kind of a more interesting split like you've got software development where this has product market fit unquestionably yeah proof works changes everything in software development fine and then we can talk about that then you've got like a bunch of people who are kind of like us like they're knowledge workers who are work for themselves are flexible free phone got lots of stuff going on and you're using this a lot and I think this reminds me quite a lot of people who use nation and then you've got all these other people who are like, "Yeah, it's kind of useful. Yeah, that was cool, but aren't using it all the time." And how does that change?

嗯。

Yeah.

我常用来打比方的是：想象你是一个会计，在 70 年代末第一次看到电子表格软件，这改变了你的人生。现在想象你是一个律师，看到它：哦，挺酷的，下周我可以用它来做时间表，但那不是我每天的工作。我觉得这就是分化。我们录制前也聊过，身处硅谷泡沫里。如果你在硅谷泡沫里，你有五台 Mac mini 组成集群，你整天拿着打开的笔记本电脑走来走去，让 Claude Code 继续运行，那你不是普通用户。普通用户是 1992 年用 Telnet 上网的人，或者 1994 年用拨号上网用 Netscape 的人，甚至更早用 Mosaic 的人——你并没有得到真正的体验。

The analogy I I always use here is imagine you're an accountant seeing the first software spreadsheets in the late '7s. This is life-changing. Now, imagine you're a lawyer seeing it. Okay, that's cool. Yeah, I could use that for my time sheet next week, but that's not what I do every day. And I think that's kind of the split and that's we're talking before we started recording about like being in the Silicon Valley bubble. If you're in the valley bubble and you've got five Mac minis in a cluster and you know you're leaving you're walking around all day holding your laptop open so you know Claude code can carry on running you're not really you're not the normal user here. the normal user you're you're the guy using Telnet to connect to the internet connecting to TNET over using TNET on the internet in 1992 or you're using dial up on Netscape one in 1994 was it Mosaic even before that in 1994 you're not like getting the real experience

可能最难的部分是搞清楚——我觉得这对当前模型能力来说很有道理。你大概可以看两年前或一年前，你会说：“天哪，这东西在消费者领域有产品市场契合度，但在其他地方都不太管用。”然后编码开始起作用了。我不确定你怎么看实验室里最接近研究的人似乎都对持续的能力提升超级乐观，这甚至可能开始改变一些现状。所以我觉得这里有一个问题，这是一个 UX、界面、部署类的问题：我们想要自动化的那些问题，有多少你很难向一个真人解释清楚？换个说法：假设模型能力没问题，假设你想用 Claude 或 ChatGPT 做某件事，但它做不了，或者你想不出怎么告诉它去做。如果另一端是一个真人，那能行吗？你要解释你想要什么有多难？那会是什么样子？那是不是正确的做法？

probably the hardest part of the whole thing is figuring out like just you know I feel like that makes a ton of sense for current model capabilities and it's like you probably could have looked two years ago or a year ago and you would have been like god this stuff has product market fit in cons you know some product market fit in consumer but doesn't really work anywhere else and then the coding stuff starts to work. I don't know what you make of the fact that the people at the labs who are closest to the research all seem super bullish on you know kind of continued capability improvement that might even start to change some of this. So I think one of the questions here and this is like a UX interface deployment kind of question is um how many of the problems that we might want to automate would you struggle to explain to an actual person? Let me put this another way like presuming the model capability presuming this here's this thing you want to do with cloud chat GPT and it kind of can't do it or you can't work out how to tell it to do it. Would that work if you had a person at the other end? How hard would it be for you to explain what you want? What would that be like? Would that be the right way of doing it?

嗯，我既不是律师也不是会计。我是一个奇怪的异类，因为我为自己工作，没有人为我工作。部分原因是，我没有很多重复性任务，因为我安排工作方式时避免了很多重复性任务，因为我做不了，因为我没有一群实习生。所以现在我有了 Claude、Copilot 和 ChatGPT，我有了这个东西，但我实际上没有任务可以交给它。我想说的是，一方面是模型能力，另一方面是你能否识别并描述出你可以交给它的任务。

Um, I mean, I'm like, you know, I'm not neither the lawyer nor the accountant. I'm like a weird outlier in that I work for myself and I don't have anybody working for me. And partly because of that, I don't have lots of repetitive tasks because I've structured what I do so that I don't have lots of repetitive tasks cuz I can't do them because I don't have a bunch of interns. So now I get Claude and Copilot and Chat GBT and now I've got a bunch of of now I've got this thing but I don't actually have the tasks to give to it. What I'm kind of getting at is is there is the model capability and then there is there is your ability to isolate and describe the things that you could give to it.

嗯。

Yeah.

这是一个不同的问题。我认为这在语音和聊天中尤其突出。但很难意识到你有这个问题，也很难意识到你可以把这个问题交给系统。每一项新技术，创造只有新东西才能做到的新事物，比用新技术自动化更多旧事物更难、更耗时。就像每一项新技术一样，你一开始是拿你已经在做的事情，用新技术做得更多更快，强迫新技术适应旧方式。而想出全新的东西需要更长时间。

And that's a different problem. This is I think is particularly a problem with voice and chat. Um, but it's quite hard to work out that you have that problem and to work out that you could give that problem to the system. With every new technology, it's harder and it takes longer to make the new things that you could only do with the new thing as opposed to automate doing more of the old thing. Like with every new technology, you start by to taking the thing you're already doing and doing it more and faster with the old techn with a new technology and forcing the new technology to fit the old way of doing it. And it takes longer to think of the completely new thing.

是的，只有这样才能实现。而且通常全新的东西不是自动化你已经在做的一堆事情，而是别的东西。

Yeah, that's only possible because of this. And generally the completely new thing isn't automating a whole bunch of stuff you already doing. It's something else.

我在 5 月发布的演示文稿中做过一个分析，现在感觉像是上辈子的事了。狗年，是的，狗年，确切地说，AI 年。那就是，20 世纪美国会计师的人数是一条直线上升。即使科技行业不断自动化，比如打孔卡、加法机、大型机等等，会计师的数量还在增加。对此有两种说法。一种是那个陈词滥调，去年大家都去查维基百科的“杰文斯悖论”。它基本上是价格弹性。如果你让做事更便宜，你是用更少的钱做同样的工作，还是用同样的钱做更多的工作，或者用更多的钱做更多的工作，因为你可能有不同的投资回报率。但我不认为这就是原因。

One of the sort of analyses I did in the presentation I published in May now, which is feels like a lifetime ago. Dogs. Yeah, dog years. Exactly. AI years. Um was you know number of people working as accountants in the USA in the course of the 20th century is a straight line up and to the right. Even as the tech industry progressively automates over and over again like punch cards and adding machines and mainframes and all the stuff that the tech industry's done, number of accountants keeps going up. And there's two things you can say about this. One of them is is now this this this cliche, the Jeans paradox that everyone rushed to look up on Wikipedia last year. Um, which basically is price elasticity. Like if you make it cheaper to do stuff, do you do the same work for less money or do you do more work for the same money or more work for more money because you might have a different ROI. Um, but I don't think that's what that is.

我不认为如果你现在在普华永道待了五年，你会比 1970 年在普华永道待五年时做的事情更多。当时还不叫普华永道，叫普华永道或其他什么。总之，你现在做的不是当时做的事情的更多版本。你当时做的事情现在每周只需十秒钟。

I don't think that if you are five years in at PwC now, you're doing exactly what you'd have been doing if you had been five years in at PwC in 1970, but more. It wasn't called PwC then, anyway. Price Waterhouse or something else. Anyway, you're not doing what you were doing then but more. You're doing the stuff that you were doing then is now ten seconds once a week.

是的。但这是他的问题。只是让模型变得更好。好吧。即使你有一个真正是 AGI 的模型，或者不管 AGI 是什么意思，现在人们试图重新定义两年前还管用的 AGI 概念，但好吧。假设你有一个真正像人一样的模型。那意味着什么？它如何融入组织？这并不像看起来那么明显和容易回答。如果你拿会计或咨询之类的行业来说，你知道，我想也许一个反方论点会是，过去你有工具能做某人可能拥有的一大套技能中的某个特定部分。那很好。所以人们找到了其他技能去做。你可以想象，而且似乎前沿模型可能相当接近，你知道，像普通大学毕业生那样擅长一系列技能。比如，一个从未在咨询公司工作过的人，我们拿这个来说。那是我大学毕业后的第一份工作。所以我认为担忧会是，任何可能被重新发明成 22 岁无经验的人能做的工作，模型可能开箱即用。

Yeah. But this is his problem. It's just making the model better. Fine. Okay. Even if you had a model that actually was AGI or whatever AGI means and now people are trying to redefine AGI stuff that was working two years ago, but fine. Presume you have a model that is actually a person. What does that mean? How would what do you where does that fit? What do you do with that? How does that fit into an organization? And it's not as necessarily as obvious and easy to answer that as it seems. If you take like a accounting or consulting or something, you know, and you think about like I I think maybe a counterargument to this would be you had tools in the past that could do one specific part of a of a large set of skills that someone might do. And cool. So people, you know, found other skills that they could go do. You could imagine uh and certainly it seems like pretty models maybe decently close to being, you know, as good at at a set of skills as like your average college grad would be. Um for, you know, who has never worked in a consulting firm, we'll take that. That was my first job out of college. So I think the the the concern would be anything that like might be reinvented into you know that a 22-year-old with no experience in that job might do the model could just do out of the box.

是的。所以这样来框定这个问题：我们一直在自动化越来越高级的人类功能。我们从 19 世纪开始，人类作为负重牲畜。我们自动化了腿。

Yeah. So the way to frame this is like we've been automating higher and higher level human functions. So we started in the 19th century with human beings as beasts of burden. We automate legs.

嗯。

Yep.

然后我们自动化了手臂和手指，我们剩下的只有大脑。但感觉你已经逐步自动化了一切，一直到栈顶。

And then we automate arms and we automate fingers and all we have left is our brains. But it feels like you've progressively automated everything all the way up to the top of the stack.

但你已经逐步自动化了一切，一直到栈顶。我不认为这是一个不可证伪的陈述，而且我认为问题在于你当时可能说过类似的话，因为当时你并没有看到所有其他发生的事，那些你当时看不到的、结果证明无法自动化的事情。当然我们不知道这些东西会走向何方，能自动化什么，不能自动化什么。所以正如我所说，有一个不可证伪的陈述，唯一证伪它的方法就是回来等五年看看会发生什么。也许我们确实得到了一个真正能做人类能做的任何事情的东西。显然那样一切都会改变，但现在我们还没有。我们拥有的东西非常参差不齐。

But like you've progressively automated everything all the way up to the top of the stack. I don't think that's kind of an unfalsifiable statement and I think the problem is you could have said that in because you kind of didn't there all this other stuff happened that you couldn't see then that it turned out you couldn't automate and of course we don't know where this stuff is going to go and what it will be able to automate and won't automate. So as I said you know there's a there's like an unfalsifiable statement that says like you know the only way you can falsify it by coming back and kind of waiting for 5 years and seeing what happens. Like it may be yes that we get a thing that actually can do anything that people can do by definition. Well clearly then everything changes but right now we don't have that. We have something that's very jagged.

是的。

Yeah.

我们不知道它是否会一直发展到那一步。我们当然不能仅仅假设，嗯，这就是让我对这个图表抓狂的地方，比如 AI 能做需要人们 17 小时的事情。

And we don't know whether it would go all the way to that. And we certainly can't just presume well I mean this is the thing that drives me crazy about this chart of like you know the AI can do something that takes people 17 hours.

是的。但也有需要人们五分钟但它做不到的事情。它不是那样的线性。智能不是那样的线性。所以，你知道，也许它确实能发展到做任何人可能做的任何事情，但目前它做不到。我们不知道何时或是否会做到。回顾所有其他时候，你自动化了你 25 岁员工做的所有事情，他们现在在做什么，这如何改变了？我的意思是，这就像回到 19 世纪。这是劳动总量谬误，你总是能看到会消失的工作，但你不知道新工作。是的。新工作将需要完全不同的技能，这些技能你没想到存在或你需要或想要。这些都是不可证伪的陈述。

Yeah. But there's also stuff that takes people five minutes that it can't do. That's not it's not linear like that. Intelligence isn't linear like that. So, you know, it may be that yes, this goes all the way to do stuff that everything that anybody could possibly do, but at the moment it can't. We don't know when or if it will be able to do that. Go back and look at all the other times that you automated absolutely everything that your 25-year-old was doing and what is it they're doing they're doing now and how did that change? I mean, this is like, you know, the going back to the 19th century. This is the lump of labor fallacy that you can always see the jobs that will go away and you don't know the new jobs. Yeah. And the new jobs will be doing things with completely different kind of skills that it didn't occur to you existed or that you needed or wanted. These are all unfalsifiable statements.

是的。但我想鉴于所有这些，你认为这些实验室的负责人如此多地谈论失业是否不合理？你知道，这似乎是一个概率分布，关于这些模型会变得多好是未知的。在概率分布的某些部分，这些东西确实很重要，对吧？

Yeah. But I guess given all that, do you think it's irrational that like the heads of these labs are are, you know, talking so much about job loss, you know, it seems like it's a probability distribution of kind of an unknowable of how much better these models are going to get. And there's some part of the probability distribution where uh this stuff does really matter, right?

是的。我的意思是，我们在这里深入探讨逻辑谬误。所以杰夫·辛顿，大约 10 年前，说停止培训放射科医生。问题是他并不真正理解放射科医生做什么。所以有一个狭隘的问题，实际上机器学习做不到他认为它能做的事情。而更广泛的问题是，那也不是放射科医生实际做的事情。

Yeah. I mean, we we get deep into logical fallacies here. Um so you've got Jeff Hinton, whatever it was 10 years ago, saying stop radio training radiologists. And the problem here is he doesn't actually understand what radiologists do. So there's a narrow problem is actually machine learning couldn't do what he thought it could do. And the broader problem is but that isn't what radiologists did anyway.

是的。你知道，我的意思是，你从咨询业开始，客户通常不是买一份 PowerPoint。PowerPoint 可能是项目的具体表达，但那不是他们付钱给你的原因。他们买的是各种其他东西，会计或律师也是如此。我在这里想知道的是消费者剩余。或者企业版的消费者剩余是什么。所以为了论证，比如在 1980 年，你的律师派律师助理去地下室图书馆。

Yeah. And you know, I mean, you you you started in consulting, like the client generally isn't buying a PowerPoint. The PowerPoint might be the tangible expression of the project, but that isn't what they're paying you for. Um, they're buying all sorts of other stuff and the same thing as an accountant or the same thing as a lawyer. The the thing I wonder here is about consumer surplus. So, or whatever the enterprise version of consumer surplus is. So for the sake of argument like in 1980 you know your lawyer goes and sends paralegal to the basement into the library.

嗯。

Y

他们一天后带着两个判例回来。很好。今天，5 年前，你的会计师或你的助理会去数据库，找到 20 个判例，对方也一样。对吧？所以客户以相同的小时数支付相同的金额。你提交相同页数的文件。你的文件胜诉的可能性与当时完全相同，你得到的结果也差不多。消费者剩余就像你以相同的价格做了更多的事情来交付相同的结果。你知道，这是极端情况，这是 iPhone。你知道，iPhone 是 1 万美元，10 万美元的消费电子产品消失了。所以肯定会有一些情况是这样。我认为也许还有另一种看待这个问题的方式，那就是想想这将会有多大的变化。所以有一张我用了很久的图表。你知道，马克·安德森喜欢“软件正在吞噬世界”的说法，Uber 不向出租车公司出售软件，Airbnb 不向酒店出售软件。非常重要的观察。好吧。Uber 对出租车做了什么，与 Airbnb 对酒店做了什么相比？结果你去看看，比如纽约。有点模糊，因为这取决于城市，但在纽约，出租车市场下降了，我忘了数字，四分之三。Uber 就这样消失了。

and they come back after a day with two precedents. Great. Today 5 years ago your accountant your your your associate would have gone to the database and they had found 20 precedents and so did the opposition. Right? And so the client gets build the same amount of money for the same number of hours. You deliver the same number of pages in the filing. your filing has exactly the same likelihood of winning that it did then and you get kind of the same result you would have got then consume a surplus like you you've done kind of a bunch more stuff for the same price to deliver the same result which is you know this is the extreme this is the iPhone you know the iPhone is $10,000 $100,000 of consumer electronics disappeared so there will certainly be some cases where that's what's happened I think there's another way of of looking at this here maybe which is just to think about how massively variable this will be so there's a chart I' I've used for a while. You know the Mark Andre loves software is eating the world thing that Uber doesn't sell software to taxi companies and Airbnb doesn't sell software to hotels. Really important observation. Fine. What did Uber do to taxis and compare that to what Airbnb did to hotels? Turns out you go and look in the like say look in New York. It's a bit fuzzy because it depends by the city but in New York taxi market is down by I forget the number three quarters. Uber is gone like this.

结果是市场变得更大，参与的人更多，市场形态也变了。每位司机的平均接单量等指标都不同了。但关键是，它释放了所有这些新需求，同时摧毁了黄色出租车行业。

And the result is there's a bigger market with more people and I mean the markets looks different. The average number of rides per driver is different and so on. But the point is it unlocked all this new demand and demolished the yellow car business.

对。

Yeah.

再看 Airbnb，可能稍微减缓了酒店业的增长。

Now look at Airbnb maybe slowed down the growth of hotels a bit.

嗯。

Yeah.

它的规模大概是酒店市场的 10%、5%、10% 或 15%。原因在于，酒店和出租车是不同的，对吧？

It's maybe 10, 5, 10, 15% the size of the hotel market. And the reasons why are well hotels and cabs are different, right?

对。

Right.

酒店业务的一半是商务出行，还有会议业务。

Half of hotel business is business. There's a conference business.

只要模型的能力是这种参差不齐的，有些地方行，有些地方不行，而且使用场景也是参差不齐的，其中一些还涉及物理世界，而不仅仅是处理信息。

As long as the models have jagged capabilities like this where they work some other the models are jagged and then the use cases are jagged and some of the use cases are physical and not just about processing information.

所以几周前我写了一篇文章，试图预测对就业的影响，因为显然这会影响就业，但试图做那种雷达图或星图，然后说“啊，你看，Opus 8.6 能做第一年律所律师助理工作的 93%”，这是妄想。因为你没法那样衡量律师助理的工作，也没法衡量模型能不能做到。

So I wrote an essay a couple of weeks ago about like trying to predict job impact because like clearly this will have impact on employment but trying to do these like star charts of radar charts of which ones where you say ah well you know opus 8.6 six can do 93% of what a first year law firm associate is doing. This is a delusional statement because you can't measure what the law firm associate is doing like that. You also can't measure whether the model can do it.

这实际上是“系统内专家谬误”，就像人们以为可以衡量如何区分猫和狗一样，你犯的是同样的错误。你无法衡量律师助理的工作。

This is actually the expert in the system fallacy where you people kind of thought that you could measure how you recognize a cat from a dog and you're saying it's the same mistake. You can't measure what a law associate does.

我要说的重点是，首先，会计工作已经变了，这是我们之前提到的。其次，有些行业你的分析会说完全不受影响，但会被其他东西摧毁，比如新闻业，对吧？

The point I was making was like first of all like the accountancy job changed which is the comment we made earlier. Secondly, there are industries that your analysis would say are completely unaffected that will get demolished by something else like journalism, right?

对。

Yeah.

举个例子，互联网并没有改变记者的本质，但它摧毁了本地广告业务，而这根本不在你的分析里，对吧？

For the sake of argument, the internet didn't change what it was to be a journalist at all, but it demolished the local advertising business, which wasn't in your analysis at all, right?

对。

Right.

Uber 也一样。你不会去看智能手机。没有人——我是说，我当时在移动行业，我们一直在谈论位置，但没人看到那个机会，对吧？

And the same thing for Uber. Like, you wouldn't have looked at the smartphone. Nobody was I mean, I was in the mobile business. We were talking about location all the time. Nobody saw that opportunity, right?

对。

Right.

所以你回去检验这些分析，说“这个工作暴露了多少，那个工作暴露了多少”，这完全是胡说。

And so you kind of go and back test these analysis where you say, 'Well, this job is exposed this much and this job is exposed this much.' It's bullshit.

你知道，这就像那个笑话：物理学家预测哪匹马会赢得比赛，他们说“首先，我们假设马是一个完美的球体”。这很棒，如果你把它定义成你能理解的东西，那你就能理解它。但它不是那样，它是别的东西，对吧？

You know, it's like the joke about the physicists who are predicting which horse is going to win a race. And they say, 'First, we're going to presume that the horse is a perfect sphere.' You know, it's like great. If you presume that if you define this into something that you can understand, then you can understand it. But it isn't that. It's something else, right?

对。

Right.

我想这确实是个问题：是啊，是否还有一些能力，比如初级员工有而模型没有的？

And I guess it's really this question of like, yeah, are there still capabilities that like, you know, entry-level workers will have that models don't?

对。

Yeah.

你知道，你遇到过安瑟伦的证明吗？安瑟伦是中世纪的神学家，他说：第一命题，上帝的定义是，没有什么能比上帝更伟大。这听起来合理。第二命题，存在的事物比不存在的事物更伟大。好，因此上帝存在。30 秒后，房间里每个人都说，这也能证明任何东西。比如，存在的马比不存在的马更好，独角兽存在比不存在更好。你不能这样证明。但人们争论了一千年才弄清楚为什么这是错的。

So you know, do you ever come across Anselm's proof? Anselm is this medieval theologian and Anselm says okay first proposition God by definition is that by which nothing could possibly be greater than God. Okay that seems reasonable. Second proposition something that exists is greater than something that doesn't exist. Okay. Therefore God exists. And 30 seconds later everyone in the room says and some dude that's cuz like you could prove anything like that. Like, well, a horse that exists is better than a horse that doesn't exist. Like, it would be better if a unicorn existed than not exist. Like, you can't prove stuff like that. But it took like a thousand years of people arguing about this to work out why that was wrong.

对。

Yeah.

我提到这个的原因是，如果你把 AGI 定义为必然发生且必然杀死我们所有人，那么 AGI 就必然会杀死我们所有人。很好，但你什么也没证明。

And the reason I mention it is like if you define AGI as necessarily inevitable and necessarily going to kill us all, then AGI is necessarily going to kill us all. Great. But you haven't proved anything.

对。

Yeah.

那么在此期间你关注什么呢？你知道，显然感觉几乎总是有一群人，你知道模型能做更多事情，但“我仍然不能做 X 或 Y”。我非常欣赏你的一点是，你对此有细致入微的看法，你根据现有事实来判断。那么，你密切关注哪些事情，或者哪些事情会让你开始调整你的看法？

So then what do you pay attention to like in the interim to you know obviously it feels like almost uh there there's always a set of people that you know the models can do something more and it's like well I still can't do X or Y like one thing I really appreciate about you is like you have a nuanced perspective on this you kind of take the facts that are out there. What like you know what are the things you're paying close attention to or would start to adjust your view on on some of this.

我觉得要拆解一下，有哪些有趣的问题？首先，是关于资本的讨论。芯片领域在发生什么？数据中心在发生什么？芯片何时能赶上供需？比如，SK 海力士的 CEO 前几天说，他认为供需失衡会持续到 2030 年。我当时想，他当然会这么说，不是吗？

I think there's that like kind of pulling apart like what are the sort of interesting sets of questions? Firstly there is like a capital conversation. So, what's going on in chips? What goes on in data centers? When do chips catch up with supply and demand? Like, what's his CEO of SK Hynix the other day said he reckons, you know, supply and demand is out of whack until 2030. Well, I was like, well, he would say that, wouldn't he?

生意嘛。

Business.

对，他生意好。所以有所有这些事情，数据中心、通电时间、电网部署等等。

Yeah, he's got a good business. Um, so there's all of that stuff, data centers, time to power, grid deployment, etc.

其次，有一个在我看来不可否认的事实：现在大约有 3 到 6 家公司在做模型，而且它们都差不多。你知道，这周有人领先，下个月有人领先，但 SpaceX 能完全失败后直接跳回排行榜几乎顶部，这对 SpaceX 来说是一个非常负面的信号。

Secondly, there is the sort of fairly what seems to me really incontestable fact that right now there's something between three and six companies making models and they're all kind of the same. And you know, someone's ahead this week, someone's ahead next month, you know, but the fact that SpaceX is like a paradoxical statement like the fact that SpaceX managed to jump straight back up to almost at the top of the leaderboards after completely flunking out is a really negative signal for SpaceX.

这是个矛盾的说法吗？因为在我看来，如果你愿意花几十亿美元并雇佣合适的人，这并不难。你知道，这些东西仍然存在。似乎还没有根本性的进入壁垒，就像搜索、社交或移动领域那样的进入壁垒。

Is that a paradoxical statement? Because what that says to me is not very hard if you're willing to spend a couple of billion dollars and hire the right people. You know, this stuff remains. There do not appear there are not yet fundamental barriers to entry. Barriers to entry in the sense that there were for search or social or mobile.

资本，你知道，我的意思是，在这一点上，只有这么多公司能承担训练成本。

Capital, you know, I mean at this point like there's only so many companies that can uh can bear the cost to train.

嗯，有，但不是两家，对吧？

Well, there are, but it's not two, right?

对。

Right.

但它会继续上升，对吧？

But like it's going to keep going up, right?

嗯，这就是我之前提到的半导体类比。半导体的情况是，每一代都更难、更贵，所以我们从几十家前沿公司缩减到只有一家，然后两家落后一步。这可能就是它的演变方式。我们还没到那一步。

Well, so this was my semiconductor comparison that we made earlier. What happened with semiconductors is it got harder and more expensive with each generation and so we shrank from dozens of companies at the cutting edge just to just one and then two one step behind. That may be how this evolves. We're not there yet.

对。

Yeah.

当然，如果曲线放缓，那么其他一切都会改变。

Of course if the curve slows down then everything else changes.

是的，算力是这些业务的主要护城河。

Yes, compute is like the major moat of these business.

只要 Scaling 继续，你就应该预期前沿公司的数量会减少。

As long as scaling continues then you should sort of expect the number of cutting edge companies to shrink down.

然后还有一个问题：你需要处于什么位置，在帕累托曲线上？

Then there's a question of well what where do you need to be on what is it the Pareto curve?

对，有多少使用场景需要绝对前沿？或者换句话说，有多少使用场景需要存在，并且有投资回报率？

Yeah, like how many use cases need to be absolutely the cutting edge? How many or put another way, how many use cases need to be there, have an ROI to be there.

嗯，曲线的另一端是，比如手机上的语音输入是免费的。是的。那中间是什么？比如亚马逊，在这条曲线上，亚马逊需要做多少才能从推荐或评论摘要等中获得多少投资回报率？所以这条曲线上会有人。而且大概只要越便宜，就越商品化，然后你就会得到像动态实时竞价跨 30 个神经核这样的东西。

Um, the other extreme, the end of the curve is like dictation works for free on your phone. Yep. And what's in the middle? Like does Amazon where on that curve does Amazon need to do be to do how much to get what ROI on recommendations or review summarization or whatever. And so there's there will be people all the way along that curve. And presumably as long as you the further the cheaper you get the more commoditized it gets and you get to like you know you've got dynamic real-time bidding across 30 neocads.

嗯，那曲线头部有多少东西需要那个？嗯，头部能保持多少价值，竞争力有多强？当然，你知道，可能有一篇论文说，嘿，你猜怎么着，你可以用同样的价格得到一个 10 倍大的模型。所以可能即使一切都在 Scaling，价格也会崩溃。是的，对。我们不知道。嗯，但那么模型能捕获多少价值？这就是我们之前讨论的关于如何向模型解释的问题。我想到的是，我每周写一份通讯，昨晚我写了一篇专栏，所以记忆犹新，就是大多数人都是三件事。大多数人不是工具构建者。大多数人看不到工具要解决的问题。而且即使前两者成立，大多数人也没有能力构建工具。所以我解释一下我的意思。比如想想大多数企业软件，一半的企业软件，你看着它，你不明白。客户看着它，他们也不明白。你看不出那是个问题。你看不出我们为什么需要它。我们不明白为什么要那样做。想想我们每天使用的多少工具，第一次看到时你会想，这有什么意义？

Um and so how much stuff is at the head of the curve that needs that? Um how much value how competitive is the head of the curve remain? And of course you know there may be a paper that says hey guess what you can have a model that's 10x the size for the price. So it may be that you get some price collapse even as everything scales. Yeah, right. We don't know. Um, but then like how much of the value can the model capture? And this is kind of what we were talking about earlier about the the sort of how do you explain it to the model question. And the thing I was thinking about I I do a weekly newsletter and I wrote a column about this last night so it's sort of fresh in my mind is like most people are three things. Most people aren't tool builders. Most people don't see the problem that the tool would be solving. And most people aren't in a position to build the tool even if the other two weren't true. And so I kind of explain what I mean by that. Like if you think about like most enterprise soft half of enterprise software you look at it and you don't get it. The customer looks at it and they don't get it. You don't see why that's a problem. You don't see why we need that. We don't understand why we would do that. Think about how many tools we all use every day where the first time you saw it you thought what's the point of that.

是的。所以现在你有了工作，你每天都有那个问题。很多时候你看不到你有那个问题。然后即使你看到了，大多数人，比如成为一个真正优秀的销售人员的技能，与擅长设计一个新的销售支持软件是完全不同的技能。成为一个真正优秀的视频编辑的技能，与擅长设计一种全新的视频编辑方式是完全不同的。我的意思是，做那件事的人可能确实对视频编辑了解很多，但大多数人不是那个人。嗯，所以真正弄清楚软件应该是什么、工具应该是什么、应该怎么做、正确的结构和工作流程是什么、网络应该如何运作的技能。想想你看过的每一个伟大的软件推介。他们就像，“哦，哇。那真是一种聪明的做法。”

Yeah. So now you you've got the job, you've got that problem every day. Very often you don't see that you've got that problem. And then even if you see it, most people the skill of being like a really great salesperson is completely different to the skill of being really good at designing a new piece of sales enablement software. The skill of being a really good video editor is completely different to the skill of being really good at designing a completely different way to edit video. I mean the person who is who does do that probably is does know a lot about video editing but most people aren't that person. Um, so the skill of actually working out what the software should be, what the tools should be, how it should be done, what the right structures and workflows, and how the network should work. Think about every great software pitch you've ever seen. They're like, "Oh, wow. That's a really clever way of doing it."

是的，

Yeah,

我永远不会想到那个。然而你却期望大公司后台的普通中层管理者，如果给他们 Claude，就能在今天下午想出这个。然后第三部分，很多时候是受监管的数据，即使不是，你知道有 1500 人接触这些数据，它需要进入记录系统，而且涉及金钱。所以你不能让随便什么人构建工具，替换 SAP 或替换这些系统，所以一切都变得更加模糊。所以有一个即兴的空间，CSV、Excel、Tableau、Perl 等等，然后你有大型横向系统，然后每个大公司内部有 4 或 500 个 SaaS 应用，AI 有点重新洗牌所有这些。所以如果你是普华永道，你每年雇佣一千名毕业生，一万名毕业生，你使用软件。如果你是红点，你每年可能雇佣几个人，你有 Google Sheets。是的。然后中间还有东西。然后有一个点，你说，“嘿，也许我们应该用 Notion，或者也许我们应该构建这个软件。也许我们应该得到这个东西。”然后那个可怕的时刻，你看到，嘿，也许我们应该用 Workday。嗯，这就像 AGI 的终极测试。嗯，如果它能成功导航 Workday，而且这些都不是二元的，所以然后好吧，会有某个用 AI 构建的新 SaaS 应用，也许它用 AI 做以前做不到的事情，但也许不是，而且现在你也可以在 SAP 里做到。

I would never have thought of that. And yet you're expecting like random middle managers in the back office of big companies to dream that up like this afternoon if you give them claud. And then like the third piece is very often it's regulated data and even if it's not it's you know there's 1500 people who are touching this data and it needs to go into a system of record and there's money attached to it. So you can't like have random people just like building tools and like replacing SAP or replacing these systems and so everything gets more fuzzy. So like there's a sort of there's like an improvised space of CSVs and Excel and Tableau and Pearl and so on and then you've got your big eye on horizontal systems and then you've got 4 or 500 SAS apps inside every big company and AI kind of shuffles all of those around. And so like if you're PWC and you hire a thousand graduates a year, 10,000 graduates a year, you use software. If you're Redpoint and you maybe hire a couple of people every year, you've got Google Sheets. Yeah. And then there's something in between. And then there's a point where you say, "Hey, maybe we should use notion or may maybe we should build this software. Maybe we should get this thing." And then this horrible moment where you see, hey, maybe we should use workday. Um, which is like the ultimate test for AGI. Um, if it can navigate workday successfully and that none of that is binary like and so then okay, there will be this new SAS app that somebody built using AI and maybe or maybe not it does stuff using AI that you couldn't do before, but also you'll be able to do that in SAP now.

是的。而且你也能让 Excel 做到，也许你还能用 CoWork 做到。那你选哪个？嗯，以前你是怎么选择买软件、用 Excel 还是用 SAP 的？

Yeah. And also you'll be able to get Excel to do that and also maybe you'll be able to use co-work to do that. And so which do you choose? Well, how did you choose whether to buy software or use Excel or use SAP before?

是的，同样的问题。

Yeah, same question.

是的。我觉得你的论点有两部分，你知道，我想你公开说过，基础模型，我的意思是我甚至不认为它们不一定有价值。更多是它们不会捕获这个生态系统的大部分价值。我仍然不确定你是否认为它们长期来看，比如 Anthropic 和 OpenAI 上市后会怎样。

Yeah. I feel like there's two parts of your argument like you know I think you've you know publicly said you you know the foundation models are you know I mean I don't even take it as like not necessarily valuable. It's more just like will not capture most of the value of this uh of this ecosystem. I I I'm I'm still unsure whether you think they'll actually be, you know, whether these long-term as as entropic and open go public will be.

所以我的意思是，你知道，再次，不通过类比论证，我会指出类比。台积电在半导体前沿拥有垄断地位。你不会为台积电写应用。

So I mean, you know, again, without arguing by analogy, I will point to analogy. TSMC has a monopoly on the cutting edge of semis. You don't write apps for TSMC.

是的。但它是世界十大公司之一。

Yeah. One of the 10 biggest companies in the world though.

是的。但他们去年的净收入是苹果的一半。这是一家伟大的公司。非常赚钱的公司。Uber 不为台积电写代码。你在技术栈中有多层，不同的抽象层，不同的构建方式。而为企业市场构建创意软件的正确人选，与构建其他东西的正确人选不同。你知道，这就是为什么我们有技术栈。这就是为什么 AWS 没有拥有整个科技行业。这就是为什么你 iPhone 上的每个应用都不是苹果制造的。

Yes. But their net income last year was half of Apples. It's a great company. Very profitable company. Uber does not write out for TSMC. You have a multiple layers in the stack, different abstraction layers, different ways of building that. And the right people to build an enterprise go to market for creativity software are different people to the right people to build. You know, this is why we have a stack. This is why AWS doesn't own the entire tech industry. It's why every app on your iPhone isn't made by Apple.

但我认为显然，如果你是 AI 领域的 AWS，而且机会像看起来那么巨大，那仍然是一项相当有价值的业务。

but I think obviously if you're the AWS of, you know, the AI space and the opportunity is as massive as it appears, that's still quite a valuable business.

是的，伟大的业务。

Yeah, great business.

是的，这就是关键。Windows 是一项伟大的业务。台积电是一项伟大的业务。AWS 是一项伟大的业务。它们都没有拥有整个技术栈一直到顶端。现在，你可以转过身说，“是的，但是 Benedict，你可以直接去 Claude 说，发明 10 个新的企业软件应用，然后去构建它们，然后去弄清楚正确的构建方式，你可以构建整个东西。”然后，我们又回到了 Anel。你回到，好吧，如果我把它定义为能做绝对一切的东西，它能做绝对一切吗？

Yeah, this is the thing. Windows is a great business. TSMC is a great business. AWS is a great business. None of them own the whole stack all the way up to the top. Now, you can turn around and say, "Yes, but Benedict, like you'll just be able to go to Claude and say, invent 10 new enterprise software applications and then go build them and then go and work out the right way of building the you could build the whole thing." And again, we're back to Anel. you're back to, well, if I define this as something that can do absolutely everything, can it do absolutely everything?

是的。好的。但这就是问题。你不能通过定义来回避问题而赢得争论。

Yes. Okay. But that's the question. You can't you can't win the argument by defining it away.

你怎么看待编码领域发生的事情？

What do you make of what's happened in the coding space?

有趣的是，显然软件工作正在爆炸式增长。

It's funny to see obviously software jobs exploding.

是的。

Yeah.

我更倾向于认为，模型提供商在与应用公司竞争时表现得相当有效。

I'm more just from the model providers, you know, competing pretty effectively against application companies.

嗯，我觉得想想 SaaS 的情况很有意思。在 SaaS 时代，软件的数量可能增加了一个数量级，甚至几个数量级。在 PC 时代，典型的大公司在数据中心里大概有十几个或二十几个应用。而现在，典型的美国大公司有 4 到 500 个 SaaS 应用，加上过去 30 年积累的几千个遗留系统。那只是市场推广的结果。这意味着你可以自动化那些以前不值得用软件来做的事情。所以至少我们应该假设，构建新软件会变得更便宜、更快。此外，它还能做以前软件完全做不到的一整类事情。因此，软件会更多。

Well, so I think it's kind of interesting to think about what happened with SaaS here. Whereas with SaaS, we went from probably an order of magnitude, maybe several orders of magnitude increase in the amount of software that there was. And so, you know, the PC era, the typical big company has, I don't know, a dozen, couple of dozen apps in the data center. And now the typical big American company has 4 to 500 SaaS applications plus several thousand legacy starts they've accumulated over the last 30 years. And that was just go to market. And so that meant you could automate all of these things that you could not have justified getting on software for before. And so at a minimum we should presume, okay, it's going to be way cheaper and way quicker to build new software. Plus, it can do this whole class of thing that your software couldn't do at all before. Therefore, there will be more software.

是的。其中一些现有企业会衰落，一些会无法完成转型。一些庞大复杂的专家系统会被非常简单的机器学习系统自动化掉。有些东西会被捆绑或拆分。这就是 PeopleSoft 和 Siebel Systems 的遭遇。那将是那种平台转移。有人会死。好吧。第二个问题是，我之前提出的观点是，这意味着软件会变少还是变多？有多少东西你不再需要 SaaS 应用了？你会在 Excel 里用 Copilot 做，或者你在……

Yep. And some of that, some of the incumbents will fall away. Some of them will fail to make that jump. Some big complex expert systems will get automated away by some very simple machine learning system. Some stuff will get bundled unbundled. This is what happened to PeopleSoft? What happened to Siebel Systems? Like it will be that kind of platform shift. People die. Fine. Second question is, was the point I was making earlier about does this mean you have less or more software? How much stuff you no longer need the SaaS app? You'll just do it inside Excel with Copilot or you'll do it inside...

我的意思是，几乎可以肯定我们都同意软件会更多，但我想我在 Claude Code 上试图表达的观点是，这是模型提供商进入应用层并有效获胜的一个例子。

I mean, I think almost certainly we'd all agree there's more software, but I guess that maybe the point I was trying to make on Claude Code is like that is an example of the model providers moving into the application layer and winning it pretty effectively.

是的。我的意思是，这很有趣。这就像所有关于什么应该集成到 Mac OS 或 Windows 中的讨论。所以，我多年前写过这方面的东西，比如回到 80 年代中期，电子表格不能打印，或者特别不能做图表。

Yeah. I mean, I suppose this is funny. This is like all the conversations about, you know, what should be integrated in Mac OS or what should be integrated in Windows. So, I wrote stuff about this years ago that like you go back to the mid 80s and spreadsheets didn't do printing or didn't do charts particularly didn't do charts.

是的。

Yeah.

所以，图表是单独的程序。事实上，你可以回到 90 年代初，找到 30 或 40 个拼写检查应用的对比测试。工作方式是，你在 Word 或 WordPerfect 中保存文档，然后打开拼写检查应用，这些应用每个要 200 到 300 美元，它们会有一个表格，比如是否支持 Word 6，还是只支持 Word 5，是否支持 WordPerfect 等等，然后你运行它。当然，随着时间的推移，这被集成进去，变成了波浪红线。打印当然也是，那是一整件事，被集成到操作系统中，所有 DOS 应用都认为它们应该做打印，然后它被集成到操作系统中。所以有一个问题，什么自然地位于底层，什么应该自然集成到你的使能层，什么应该是独立的应用。这当然是关于 IE 浏览器的反垄断争论，但今天如果你买了一部 iPhone 而没有浏览器，那会很疯狂。所以它随时间变化，所以有一个问题，什么自然集成到最低层的基板，什么自然拆分成独立的东西，在那里你真正需要深入垂直领域知识并推向市场，以及其他一切构建垂直软件应用的东西，这和“为什么那不是 AWS 的一部分”是同一个问题。

So, charts were a separate program. In fact, you can go back to the late early 90s and you can find these group tests of 30 or 40 spellcheck apps. So the way it worked is you write save your document in your word or WordPerfect or whatever and then you open the spellcheck app and the spellcheck apps were like $200 and $300 each and they would have this grid of like do they support Word 6 or do they only support Word 5 and do they support WordPerfect and so on and you would run that and of course over time that gets integrated and then it becomes wavy red lines and printing of course was like that was a whole thing that got integrated into the operating system and the word all the DOS apps thought they should do printing and then it gets integrated in the operating system. So there is this question of what sits naturally at what should what naturally gets integrated down into your enabling layer and what should be a separate application. This is of course the antitrust argument about Internet Explorer but today like if you bought an iPhone and it didn't have a web browser that would be insane. So it changes over time and so there is this question well what naturally gets integrated into the lowest level substrates and what naturally gets unbundled into separate things where you really need to understand have deep vertical domain knowledge and go to market and everything else that builds a vertical software app and this is the same question of well why isn't that part of AWS.

答案是……

And the answer is well...

我的意思是，也许稍微反驳一下，比如在编码的情况下，模型在某种程度上就是产品，对吧？所以你在模型和模型的实际使用之间有一个反馈循环，然后使底层产品变得更好。而 AWS 就像计算本身，无论它用于制药还是杂货店，使用并不会从根本上改变产品。计算就是计算。在这种情况下，其他参与者很难，而且一直很难。我认为 Windsor 和 Cursor 都得出结论，我们需要拥有底层模型层才能在这个领域拥有有效的产品，而仅仅作为上层应用，我们会被彻底打败。

I mean maybe to push back for a second like in the case of coding like the model to some extent is the product right and so that feedback loop that you have between the model the real world usage of the model you know the then makes the underlying product better you know whereas like AWS is just like the compute itself whether it's used in pharma or whether it's used in you know uh a grocery store um it's not like that usage uh fundamentally changes the the product in any way um computes compute you know in this case it would be very hard to and it has been very hard for other players and I think both Windsor cursor they've all concluded like god we need to own the underlying model layer to have an effective product in this space and ultimately just as an application on top we're going to get totally hosed.

是的，但这不就是我们与超大规模厂商的对话吗？哪些东西应该自然成为平台的一部分，哪些部分是专门的垂直工具？哪些部分自然属于 Windows？

Yeah, but isn't that the conversation that we had with the hyperscalers? Which things should you naturally be part of the platform and which parts are dedicated vertical tools? Which parts are naturally part of Windows?

是的。

Yeah.

哪些部分如果属于操作系统会更好，哪些不是。

And which should are naturally better if they're part of the operating system and which parts aren't.

到目前为止，在语言模型领域，我们看到的唯一一个大市场实际上证明对模型提供商来说是非常好的。我认为有一个有趣的问题，你可以问一些关于为什么编码会成功的具体原因。

The one big market we've had so far in LM world has actually proven to be very good for the model providers to provide. I think there's an interesting, you know, there's a couple of like things you can ask about the specificity of why is it that it's coding that worked?

是的。

Yeah.

是因为这些人都是程序员吗？就像，我们解决了眼前的问题？是因为 LLM 和代码之间有某种特殊的东西使它特别有效？是因为你看到眼前的问题然后说这就是你要解决的？是的。还是说有什么特别之处？是我关于工具制造者的观点吗？是工具制造者正在为他们的任务构建工具吗？你知道工具制造者是开发者的唯一地方是开发工具。所以我不知道。是这样吗？我不认为我们有一个好的答案。也许我很想知道。是否有某种理论论证说代码特别适合 LLM？

Is it just that it's that all these people are coders? Is it like, well, we fix the problem that we see in front of us? Is it that there's something specific about LLMs and code that makes it work particularly well? Is it that you see the problem in front of you and say that's what you work on? Yeah. Versus is there something specific about this? Is it my point about tool builders? Is it that the tool builders are building the tool for their task? You know the one place where the tool builders are developers is development tools. So I don't know. Is it that? And I don't think we have a good answer. Maybe I'd love to know. Is there some theoretical argument that says that code is uniquely well suited to LLMs?

100% 归功于数学。你知道，非常可验证。你可以运行它们数百万次。

100% go to math. You know, very verifiable. You can run them millions of times.

嗯，这是另一点。这是另一点，如果我们回到关于自动化的对话，你在哪里拥有足够的训练数据，在哪里验证是可扩展的？你越深入复杂的咨询项目，就越难进行可扩展、可重复的验证。

Well, this is another point. This is another point is where, if we're going back to the conversation about automation, where do you have enough training data and where is validation scalable? And the more that you get into like complex consulting projects, the harder it is to have scalable, repeatable validation.

是的。

Yeah.

因为你离具体、明确和客观的东西越来越远。

Because you're getting a long way away from something that's tangible and specific and objective.

我认为这正是很多人一直在说的，尤其是最近 Neolab 的这股热潮，人们纷纷出来说：“天哪，肯定有一种更数据高效的方式来训练模型。”最终，如果我们做代码的方式是唯一的方式，也许我们最终会达到目标，整个经济就只是数据标注员。随着时间的推移，你可以做足够多的数据标注来达到目标，但相比更算法高效的方式，这感觉会是一场真正的苦战。

And I think this is exactly what a lot of folks have been saying, and with this whole Neolab push lately, people spinning out and saying, 'God, there's got to be a more data-efficient way to train models.' Ultimately, if the way we did code is the only way to do everything else, maybe we get there over time, and the whole economy is just data labelers. Over time, you can do enough data labeling to get there, but it feels like that will be a real slog relative to something more algorithmically efficient.

那么换个方向，既然我们不知道模型会如何运作，有哪些地方在概念上值得怀疑它们能否奏效？

So pushing in a different direction, given that we don't know how the models will work, what are places where it's conceptually questionable whether they could work?

是的。其中一个是在关于没有训练数据、数据是隐性的、是隐性知识，甚至很难解释如何做或为什么做的讨论中，这直接回到了机器学习的起点。很难解释那些无法转化为统计问题的地方，对吧？哪里很难从逻辑问题转化为统计问题？其次，哪里你不想要平均值？因为原则上，所有这些做的事情都是问“大多数可能会怎么做”。在软件中，你说的是大多数可能会如何构建一个满足这些标准的应用，而你想要的正是大多数人可能采用的方式。你越是进入“不，我不想要大多数人可能的方式，我要做点不一样的”，就越会触及模型的概念性问题。我过去常常这样想：如果你回到像 AlphaGo 这样的东西，AlphaGo 能做出没人做过的动作，但它有可扩展的验证系统，可以检查动作是否获胜。而如果你要创造一种全新的企业战略，你怎么检查它是否与任何人以前做过的完全不同？你怎么检查它是否有意义？

Yeah. One of them is in the conversation about where there isn't training data, where the data is implicit, where it's implicit knowledge, maybe even where it's hard to explain how you do it or why you do it, which comes right back to the beginning of machine learning. It's hard to explain where you can't turn it into a statistics problem, right? Where is it hard to turn this from a logic problem into a statistics problem? And secondly, where do you not want the average? Because in principle, what all of these things are doing is asking how would most people probably do this. In software, you are saying how would most people probably build an app that does X, Y, and Z with these criteria, and you want the way most people would probably do it. The more you get into 'No, I kind of don't want it the way most people would probably do it. I'm going to do something different,' the more you get into a kind of conceptual question for the models. The way I used to think about this was that if you go back to something like AlphaGo, AlphaGo can do moves that no one's done before, but it's got a scalable verifying system. It can check whether the moves win. Whereas if you were to create a completely new enterprise strategy, how would you check that it was completely different from what anyone had ever done before? How would you check whether that made any sense?

对吧？今天人们都在创建这些非常基础的强化学习环境，对吧？就像“嘿，你通过 Salesforce 应用正确更新了吗？”然后用它来训练模型。很明显，随着时间的推移，很多这类问题的唯一解决方案将需要是模拟，对吧？你实际上是在模拟复杂的商业情境或经济。

Right? And it's all people creating these very basic RL environments today, right? Where it's like, 'Hey, did you go through the Salesforce app and update things correctly?' and use that to train models. It feels very clear that over time, the only solution to a lot of this stuff is going to need to be in simulation, right? And you kind of are simulating complex business situations or economies.

是的。所以在这里随便想想，有一种愚蠢的批评，说模型做不到。有一种短期的批评，说模型做不到。长期的问题是，你能获得足够的训练数据让模型做到，以及足够的反馈循环让模型做到。第三步是，是的，但你想要训练数据的产物吗？你想要平均值吗？然后你可以把它推向不同的地方。我的意思是，我上次演讲最后有类似的幻灯片。我在想有消费者版和企业版。消费者版是第一步：这是一件外套的图片。我在哪里能买到？

Yeah. So just thinking out loud here, there's the dumb criticism, which is the model can't do that. There's a short-term criticism which says the model can't do that. The longer-term question is can you get enough training data for the model to be able to do that, and enough feedback loops for the model to be able to do that. The third step is yes, but do you want the product of the training data? Do you want the average? And then you can push this in different places. I mean, I have the sort of imagine if slides towards the end of my last presentation. I was thinking there's a consumer version and an enterprise version. Consumer version is step one: here's a picture of a coat. Where can I buy that?

好的。这在 10 年前会是科幻小说，现在这是一个引擎，现在这是一个产品问题。是的，那会奏效。外套可能有点难，但原则上会奏效。这是一个花瓶，我在哪里能买到？这是一盏灯，我在哪里能买到？第二步：给我 10 个类似外套的选择，并解释我该如何选择。好的，这现在可以解决了。第三步：去我的 Instagram 看看，推荐一件符合我风格的外套，能提升我的形象，能给我一个我可能喜欢的彻底改变。这现在可能也能做到了。

Okay. That wouldn't have 10 years ago been science fiction. Now that's an engine. Now that's a product problem. Yep, that would work. Coats might be hard, but yeah, in principle that will work. Here's a vase. Where can I buy this? Here's a lamp. Where can I buy this? Step two: give me 10 options for a coat like that and explain how I would choose. Okay, that's now solvable. Step three: go look on my Instagram and suggest a coat that matches my look, that would improve my look, that would give me a radical change that I might like. That can probably do now.

是的。但你越接近不是显而易见而是好的东西，就越难做到。企业版会是这样的：第一步是听我们与客户的 Zoom 录音，告诉我最近几个月是否有新的担忧出现。

Yeah. But the further you get to not what's obvious but good, the harder it is to do that. The enterprise version of this would be something like step one is listen to our Zoom recordings with clients and tell me if there are new concerns emerging in the last couple of months.

是的。

Yeah.

第二步或第三步是去查看我们的应用遥测数据、Salesforce 里的所有内容以及我们所有的客户电话，然后做竞争性市场回顾，再发起用户调查，自动化问题，确定用户列表，召开电话会议，完成所有通话，综合结果，再全部汇总起来，然后告诉我为什么我需要改变我的流失率。我如何改变我的流失率？顺便说一句，这是一个你可能需要站在定价曲线前沿的例子。这不是一个商品问题。这也是公司里随便一个人无法在 Claude 中快速搞定的。你需要一个月的项目才能启动并运行起来。但这些都是你越沿着这条曲线前进，你越不是在做“这是任何人都会告诉我的”，你越是走向差异。为了更简单一点，你可以让机器学习，让 AI 音乐模型给你做更多糟糕的爵士乐或糟糕的流行音乐。你可以让它做出更多听起来像 Taylor Swift 的东西。向任何正在听的 Taylor Swift 粉丝道歉。你可以做更多 Taylor Swift。没问题。但想象一下发明朋克。

Step two or three would be go look at our app telemetry and everything in Salesforce and all of our client calls, and then do a competitive market review, and then spin up a user survey, automate the questions, work out the list of users, run the conference calls, do all the calls, synthesize the results, pull it all back together again, and then tell me why do I need to change my churn. How do I change my churn? Now, that's an example, incidentally, of something where you probably do need to be at the head of that pricing curve. It's not a commodity question. It's also something that one random person in the company isn't going to be able to spin up in Claude. You're going to need a one-month project just to kick that off to get that going. But these are sort of the further you get along that curve, the more you're not doing 'this is what anyone would tell me' and the more you're getting to variance. To make that much simpler, you can get a machine learning, you can get an AI music model to make you a lot more bad jazz or bad pop music. You can get it to make you more stuff that sounds like Taylor Swift. Apologies to any Taylor Swift fans listening. You can make more Taylor Swift. Fine. But imagine inventing punk.

是的。

Yeah.

想象一下发明嘻哈。

Imagine inventing hip-hop.

对吧？

Right?

而且企业里 99% 的事情可能不是发明嘻哈。就像你刚才举的例子，我认为你今天可能可以用 LMS 做到。人们正在做；今天还不完美，但你有 AI 面试官，而且有足够的迭代，感觉还不错。

And come from 99% of things in enterprise, probably not inventing hip-hop. Like that example you just gave, I think you probably could do it with LMS today. And people are doing it; it's not perfect today, but you have your AI interviewers and there's enough iteration on it that it doesn't feel.

关于 AI 面试官，有趣的是，我前段时间和贝恩的某个人聊过，他谈到将其用于面试和合成客户。这里有一个常见模式，完全题外话。人们在谈论 AI 对某个行业意味着什么时，有一个常见模式。如果你不在那个行业，你会认为你会用它做这个。而行业内的人会说：“不，不，不，我们用它做这个。这才是我们能做的。”所以在广告业，有整个关于你可以用它制作广告的说法。而 Martin Sorrell 说：“不，不，不，不，不。你用来自动化所有无聊的后台琐事。”是的，你自动化传真。我们发给 Google 预订广告的传真，几乎就是字面意思。

The funny thing about AI interviewers is I was talking to somebody from Bain a while ago who was talking about using this for interviews and for synthetic customers. There's a common pattern, complete sidebar here. There's a common pattern in people talking about what AI means for X, Y, Z industry. If you're not in that industry, you'd think you do this with it. And the people in the industry are like, 'No, no, no, we do this. This is the thing we can do with it.' So in advertising, there's a whole thing of you can make the ads with it. And Martin Sorrell is like, 'No, no, no, no, no. You use it to automate all the really boring back office crap.' Yeah, you automate the faxes. The faxes we're sending to Google to book the ads, almost literally.

咨询行业也一样，我们用这个来做 PPT 吗？嗯，有点。但我们真正用它来做的，是自动化研究、客户研究，对吧？

And the same thing in consulting like do we use this to make PowerPoints? Well, kind of. But what we really use it for is to automate the research, customer research, right?

因为现在你可以做那个查询、那个调查，一天就能完成，而不是一周。

Because now you can do that query, you can do that survey in a day instead of a week.

对，这就是我说的消费者剩余。同样的花费、同样的调查，但你交付的时间从一周缩短到一天。

Yeah. Which is my consumer surplus point. You do the same spend, same survey, but you deliver it in a week instead of in a day instead of a week.

对，有道理。

Yeah, it makes sense.

我的意思是，鉴于我们刚才讨论的基础模型，如果你在经营其中一家公司，你会不会在重点或他们正在做的事情上有所不同？

I mean, I guess given this discussion we've had on the foundation models, like if you were running one of these companies, like would you be doing anything different in terms of like focus or you know what they're

我对此非常同情 Sam Altman。嗯，这可能是一个独特且有争议的说法。嗯，你要同时处理很多事情。

I have a lot of sympathy for Sam Altman here. Um, that may be a unique and controversial statement. Um, you got a lot of plates to juggle.

对。

Yeah.

因为你有很多事情要处理——我的意思是，你可以列个清单。实际上，你可以让 Claude 帮你列出所有问题。你要做设备吗？哎呀，抱歉。好，我们继续。我现在要花很多时间和律师通电话。嗯，你怎么看——当然 OpenAI 和 Anthropic 没有基础设施。所以有个基础设施问题。我们怎么获得基础设施？我们怎么为基础设施融资？嗯，我们自己做芯片吗？是的。好。那要花多长时间？你知道，你得一路解决技术栈的各个层面。就像你在发明创造。就像你是比尔·盖茨，你得在 1985 年或 1980 年发明并建造个人电脑、企业软件和宽带网络。你得——你知道，技术栈的很多部分都完全不清楚，而且同时发生。有一种——嗯，显然这就是你要做的。所以，你可以看到两种策略，你可以看到不同公司的策略。如果你是现有企业，你会试图把新事物变成一项功能。对。

Because you have all—I mean you can make your list. You could, in fact, you could ask Claude to make you a list of what are all the problems. Do you make a device? Oops. Um, sorry about that. Okay, let's go. I'm going to spend a lot of time on calls with lawyers now. Um, how do you think about—of course OpenAI and Anthropic don't have infrastructure. So there's an infrastructure question. How do we get infrastructure? How do we get the funding for the infrastructure? Um, do we make our own chips? Yes. Okay. How long is that going to take? You know, work your way all the way through those layers of the stack. It's like you're trying to invent. It's like you're Bill Gates and you've got to invent, you've got to build PCs and enterprise software and broadband networks in like 1985 or 1980. You've got to—you know, so many different parts of that stack are completely unclear and all happening at once. There's a sort of—well, obviously this is what you would do. And so, you can see the two strategies, you can see the strategies across companies. So if you're an incumbent, you try and make the new thing a feature. Yeah.

所以这就是我们看到 Google 把整个东西撒得到处都是，你看到的就是这样，而且可能大部分是成功的——如果你是微软，嗯，我们没有这个能力，所以我们会买、租，随便，把它铺得到处都是。企业 AI 部署的步骤之一是：第一步部署 Copilot，第二步——哎呀——第三步做点别的。嗯，如果你是苹果，显然两年前他们严重失误了。如果你用过新版 Siri 的测试版，你会发现它和 6 到 9 个月前的 ChatGPT 一样好，这是一个消费者部署和用例的故事。嗯，如果你是 OpenAI 或 Anthropic，你没有传统业务。你没有办法把它变成一项功能。你没有分销渠道。你没有基础设施。你拥有的是前沿科学，但你得构建其他所有东西。所以我认为 OpenAI 去年下半年做的是：我们能在商品化的基础模型之上构建什么？是浏览器吗？是社交视频应用吗？是应用商店吗？是另一个应用商店吗？是第三个应用商店吗？是购物吗？是广告吗？什么都有。如果你是 Anthropic，钱更少，更专注，更担心邪恶的事情。好吧，随便。更专注于编码。偶然发现编码是有效的，对吧？好。现在其他人都转向编码。就像《巨蟒与圣杯》里《布莱恩的一生》中的场景，你知道，大家都在向编码扔石头。嗯，但那是下一个大事件吗？我不知道。

And so that's what we saw Google spray the whole thing over everything is what you see it with and probably mostly successfully is what you—if you're Microsoft, well we don't have this capability so we'll buy it, rent it, whatever, spread it all over everything. One of the steps in enterprise AI deployment was like step one is deploy co-pilot, step two—oops—step three let's do something else. Um, if you're Apple, clearly they fumbled massively two years ago. If you've played with a beta of the new Siri, it's like this is as good as ChatGPT was 6 months, 9 months ago and it's a consumer deployment and use case story there. Um, if you are OpenAI or Anthropic, you don't have a legacy business. You don't have a way to make it a feature. You don't have distribution. You don't have infrastructure. What you do have is cutting edge science, but you've kind of got to build all of the rest. And so I think what OpenAI did second half of last year was what can we build on top of a commodity foundation model? Is it a browser? Is it a social video app? Is it an app store? Is it another app store? Is it a third app store? Is it shopping? Is it ads? Everything. If you're Anthropic, less money, more focus, more worried about evil stuff. Fine, whatever. Much more narrowly focused on coding. Stumble into coding is something that works, right? Okay. Everyone else now brushes towards coding. Like it's like the scene in Monty Python with, you know, in life of Brian with Java, you know, like everyone's throwing rocks at coding, coding. Um, but is that the next thing? I don't know.

我的意思是，关于编码，我们还可以说的一点是，编码实际上是一个相当小的行业。我的意思是，每当我指出这一点时，科技界的人都会很生气。有数百万人编写软件。好吧。想象一下，如果我们有一个真正有效的用例，有十亿人想用——比如我们无法——基础设施根本无法支持。我的意思是，这是与以往每一次平台转变的一个有趣区别：它更贵而不是更便宜。它有边际成本，所以你不能做一个消费者应用，获得 1 亿用户，然后再想出收入模式。

I mean, part of the other thing we—one could have said about coding is coding is actually a pretty small industry. I mean, if I—I mean it's weird whenever I point this out, people in tech get really upset with me. There's millions of people writing software. Fine. Imagine if we had a use case that really, really worked that like a billion people wanted to do—like we couldn't—the infrastructure couldn't possibly support it. I mean this is one of the interesting differences from every other previous platform shift is it's more expensive rather than cheaper. It has marginal cost and so you can't make a consumer app and get 100 million users and then work out the revenue model.

这就是为什么都是企业级而不是消费者级。嗯，但如果你在经营一家模型公司，我的意思是，显然 OpenAI 有各种奇怪的政治事情发生，但——

Which is why it's all enterprise as opposed to all consumer. Um, but if you're running a model company, I mean obviously OpenAI had all sorts of weird political stuff going on but—

就像你在引导你的研究人员，你在引导资本，你在引导基础设施——你在各个方向赶猫。你得弄清楚：好吧，当每 4 到 6 周就有另一个模型登上排行榜榜首，而且完全没有迹象表明你能做什么来获得可持续的竞争优势时，通往可持续竞争优势的道路是什么？我们这里有网络效应的路径吗？我们能在其上构建产品层，让每个人都必须使用我们的产品，即使底层的模型看起来没什么不同吗？嗯，我们能将模型聚焦于一项能力，并试图在该能力上领先吗？所以你可以说 Claude 做的是专注于让编码变得非常非常好。嗯，也许他们在那方面有竞争优势。嗯，但你还得不断意识到，每 3 到 6 个月一切都在变化。没有人真正知道三四年后这会是什么样子，也没有人真正知道构建块是什么。

Like you're shepherding your researchers, you're shepherding capital, you're shepherding infrastructure—you're herding cats in all sorts of different directions. You've got to work out: well, what is the path to sustainable competitive differentiation when every six to four to six weeks there's another model at the top of the leaderboard and there's absolutely no sign of what you would do that would mean you would have sustainable competitive differentiation. Do we have a path to network effects here? Do we—can we build product layers on top of this that mean everyone has to use our product even though the model underneath doesn't seem to be very different? Um, can we focus our model in on one capability and try and pull ahead in that capability? So you could argue what Claude has done is focused on getting coding really, really good. Um, and maybe they have a competitive lead there. Um, but you also have to be continually conscious that everything is changing every 3 to 6 months. No one really knows what this is going to look like in three or four years time and no one really knows what the building blocks are.

那么，考虑到这一切的不确定性，如果你在经营 OpenAI，你认为进行所有这些实验有意义吗？

So do you think then running all these experiments makes sense if you're OpenAI, just given how uncertain this all is?

我认为——嗯，问题是两种策略都有道理。我的意思是，如果你是 OpenAI，你有这么多资本，尝试所有这些事情是有道理的。如果你是 Anthropic，你资本较少，更担心奇点和邪恶的事情发生，那么更合理地就是专注于编码。你知道，我们不知道编码会成功。我的意思是，我们有点——也许事后你可以说显然它会成功，但显然今年年初它从“有点用”变成了“真的有用”。

I think—well, the problem is both strategies make sense. I mean, I think if you're OpenAI and you've got all this capital, it kind of made sense to try all of that stuff. If you're Anthropic and you had less capital and you're more worried about like takeoff and evil stuff happening, then it made more sense just to focus in narrowly on coding. You know, we didn't know coding was going to work. I mean, we kind of—maybe you could have said in hindsight obviously it was going to work, but clearly it flipped at the beginning of this year from kind of working to actually working.

嗯，我们之前聊过这个话题，但那是必然的吗？是那回事吗？但你又会回到这个问题。最近几天让我对 OpenAI 着迷的是新应用的发布。如果它告诉你一件事，那就是写代码不是难的部分，因为它简直是一团糟，混乱不堪。有个奇怪的现象，我们要给每个人 Copilot，还有 Work，而 Work 和 Chat 不同，它叫 Chat 但 Chat 是隐藏的，当你用 Chat 时，它出现在一个奇怪的弹窗里，但你不能在这个里用那个，然后你又把它藏起来。然后这个发生了，这个在云端能用，那个不能用。我就想，真的，你是怎么发布那个的？那是怎么发生的？

Um, we've had this conversation before, but was that inevitable? Was that the thing? But you come back to this question. The thing that's fascinated me about OpenAI in the last couple of days is the new app launches. If it tells you one thing, it's that writing the code isn't the hard part, because it's such a disastrous, chaotic, confused mess. There's this weird thing where we're going to give everybody Copilot, and also Work, and Work is different from Chat, and it's called Chat but Chat's hidden, and when you do Chat, it appears in this weird popup, but you can't use this in that, and then you hide it. And then this happens, and this works on cloud, and that doesn't work. I'm like, really, how did you ship that? How did that happen?

从这一切退一步看，在微软、谷歌和 Meta 等公司获得赢家通吃的效应之前，他们必须靠执行来达到那个位置。

There's a step back from all of this, which is that before Microsoft and Google and Meta and so on had the winner-takes-all effects, they had to execute their way into that.

是的。

Yeah.

我的意思是，你有点知道赢家通吃的效应会完全明显。还记得人们以为社交会在不同国家获胜，比如 Bebo 会在欧洲赢，Facebook 会在美国赢，还有 Uber 会赢，比如拼车会是一个城市一个城市地赢。

I mean, you kind of knew that the winner-takes-all effects would be completely obvious. Remember when people thought that maybe social would win in different countries, like Bebo would win in Europe and Facebook would win in America, and like Uber would win it, like ride-sharing was going to be city by city.

所以你知道网络效应是什么，但你不确定。但你也必须靠执行来获得它。记住 MySpace 是先来的，记住当 Facebook 出现时，人们，谷歌做了 Wave、Buzz 和其他东西。而且谷歌拥有所有数据。所以有一个执行的故事。你可以对网络效应和商品化等非常确定，但那里也有一个执行的故事。你实际上必须让事情发生。就像马克说的，历史不是物质的人。历史不是人。所以你可以标记这整个关于历史如此运作的事情，这些是不可避免的力量，然后他说，是的，但历史不是人，所以实际上必须有人去做。

And so you knew what the network effect was, but you weren't sure. But also you had to execute your way into getting it. Remember that MySpace was there first, and remember that when Facebook was happening, people, Google did Wave, Buzz, other things. And Google had all the data. So there's an execution story. You can get very deterministic about network effects and commoditization and so on, but there is an execution story there as well. You actually have to make the thing happen. It's like Mark said, history is not a material person. History isn't a person. So you can mark this whole thing about history works like this and these are inevitable forces, and then he says, yeah, but history isn't a person, so somebody actually has to do it.

是的，所以你必须真正让那件事发生。

Yeah, so you have to actually make that thing happen.

是的，不，我喜欢那个。我的意思是，我想在 OpenAI 方面，我认为他们显然有所有这些元效应进来，我想你谈过他们有点试图运行类似的剧本。

Yeah, no, I love that. I mean, I guess on the OpenAI side, I think they obviously had all these meta effects come in, and I think you've talked about they kind of tried to run a similar playbook.

是的，感觉有点货物崇拜。

Yeah, it felt a bit cargo culty.

你知道那里出了什么问题吗？你知道什么是货物崇拜吗？我一直不确定这个引用。

Do you know what went wrong there? Do you know what a cargo cult is? I was never sure of this reference.

你知道，这是我说过很多次的话，我实际上不知道它的起源是什么。

You know, it's something I've said a bunch, and I don't actually think I have any idea what the origin is.

这是二战时期的事。所以，太平洋战争，太平洋上有所有这些热带岛屿，那里的人们每六个月或每年才看到一艘欧洲船，但这些人都没有接触过现代大规模生产的工业化社会。然后美国人来了，他们建了空军基地、基地、船、港口和码头，有所有这些东西，特别是美国有无限的钱，所以他们把所有物资都分发出去。像罐头食品、酱料，还有你想要的一切。然后战争结束，美国人就离开了。太平洋某处有个岛，有大约 1 万架飞机和卡车之类的东西被扔进泻湖，因为把它们运回美国报废根本不值得。所以，我只是稍微夸张一点，有数百辆崭新的卡车和飞机被扔在这个泻湖里。所以重点是，突然机场空了，不再有飞机来了。那你怎么办？好吧，你走进控制塔，你说话，你排队游行，你在旗杆上升旗，因为那是美国人做的，然后飞机就来了。我的意思是，这绝对是真的。这是 40 年代末和 50 年代的现象。你实际上不明白是什么导致货物来的。

It's a World War II thing. So, war in the Pacific, there are all these tropical islands in the Pacific where people have seen a European ship every six months or something, or every year, but these are not people who are exposed to modern mass-produced industrialized society. And the Americans arrive, and they build an air base and a base and a ship and a port and a pier, and there's all this stuff, and America in particular has infinite money, so all the supplies they hand all the stuff out. And like canned food and like source and like everything you want. And then the war ends, and the Americans just leave. There's this island somewhere in the Pacific that has like 10,000 airplanes and trucks and stuff just dumped into a lagoon, because it just wasn't worth shipping it all back to America and scrapping it. So there's like, I'm only slightly exaggerating, there's like hundreds of brand new trucks and aircraft dumped in this lagoon. So the point is, suddenly the airfield's empty, and there's no more planes coming. So what do you do? Well, you go up into the control tower and you speak, and you line up in parade, and you run a flag up the flagpole, because that's what the Americans did, and the plane came. I mean, this is absolutely true. This is a phenomenon in the late '40s and '50s. You didn't actually understand what it was that was causing the cargo to come.

是的。

Yeah.

所以你模仿形式。

So you imitate the forms.

有趣。我又陷入兔子洞了，但重点是你在 2010 年加入 Meta，那是在它有了产品市场契合之后很久。现在你在……那你怎么办？好吧，你做一个应用商店，你建立一个广告业务，你建立一个电子商务的东西，你知道如何执行所有这些剧本，但你不在知道什么有效的阶段。而你真正想要的是那些在 Meta 还只是哈佛时就在那里的人。是的，不是那些在 50 亿人使用它时在 Meta 的人。

Interesting. And I've gone down a rabbit hole again, but the point is you joined Meta in 2010, way after it had product-market fit. And now you're at... So what do you do? Well, you make an app store, you build an ad business, you build an e-commerce thing, and you know how to do all of those playbooks, but you're not at the part where you know what works. And what you actually want is the people who were at Meta when it was just Harvard. Yeah, not the people who are at Meta when five billion people were using it.

显然，在消费者方面，我觉得你谈过这些产品如今的使用相对较浅，全面来看。我认为你之前讨论过的难点之一是，如果你在这些实验室之一的产品团队，你有点在等待模型能力来弄清楚要构建什么。所以这是一种艰难的产品构建方式。同时，我无法判断我们是否只差几个模型能力就能实现一套完全不同的消费者使用。我的意思是，杀手级应用总是感觉像计算机使用。如果你真的让计算机使用工作，人们会不会更多地使用这些产品？

And obviously, on the consumer side, I feel like you've talked about how these products have some relatively shallow usage today, across the board. I think one of the hard parts that you've discussed before is if you're at one of these labs and you're on the product team, you're kind of waiting for the model capabilities to figure out what to go build. And so it's a hard way to build product. And then at the same time, I can't tell whether we're just a few model capabilities away from a completely different set of consumer use. I mean, the killer ones always felt like computer use. If you really got computer use to work, would people just use these products way more?

我不太确定。只要记住有多少人，对多少人来说他们的主要设备是智能手机。大多数人使用电脑的频率不如智能手机。所以对那个有点犹豫。对我来说感觉非常极客。即使不考虑安全性和失败率，它可能意外删除你所有的东西，这又是，你知道，我在用……我的意思是，关于 80 年代初个人电脑的这一点。即使在 90 年代中期、90 年代末，你工作时完全正常，你抬头看屏幕，你意识到它冻结了，什么都没有。你的解决方案是爬到桌子底下，拔掉电脑插头，然后等它重新启动，然后希望你没有丢失超过一个小时的工作。

I'm not sure about that. Just remember how many people, for how many people their main device is a smartphone. Most people don't use a computer as much as they use a smartphone. So a little bit hesitant about that one. It does feel very geeky to me. Even setting aside security and failure rates, and it might accidentally delete all of your stuff, which again is like, you know, I was using... I mean, this point about PCs in the early '80s. It was completely normal that you'd be working, even in the mid-'90s, late '90s. You'd be working, and you look up at your screen, and you realize it's frozen, nothing. And your solution is you crawl under your desk and you unplug your PC, and then you wait for it to turn on, and then you hope that you haven't lost more than an hour of work.

是的。

Yeah.

那只是正常的。完全正常。我是……现在完全一样，比如，你知道，我让 Copilot 清理我的收件箱，它删除了我的收件箱。嘿，它清理干净了。太好了。所以有一个消费者……这里的起飞是，再次，这些是不可证伪的陈述。

That was just normal. Completely normal. I was... And that's exactly the same now, like, you know, I told Copilot to clean out my inbox, and it deleted my inbox. Hey, it's cleaned up. Great. So there's a consumer... The takeoff here is like, again, these are unfalsifiable statements.

是模型会变得更好，还是说问题并不完全在这里？是不是大多数人并不经常有那种能和这个很好契合的事情？先不谈企业那边，比如你坐在 SAP 里，做这件事很麻烦。很好。你不能用 ChatGPT 来做，因为没授权，没接入，不允许。好吧，那是 CIA 的对话。你是消费者。你怎么搞清楚这是怎么回事？你怎么搭桥？你打开 ChatGPT，然后……我知道，我不认为我知道答案。我的意思是，显然我不知道答案，但我很矛盾，因为我记得 90 年代末互联网上的门户网站，你给人一个浏览器，你不能只给他们一个带地址栏和空白屏幕的浏览器。你得帮他们。

Is it just that the models will get better or is it that that's not quite what the problem is? Is it that most people don't very often have the kind of things that mesh very well with this? Set aside the enterprise side of like you're sitting in SAP, it's a pain in the ass to do this thing. Great. You can't use chat GPT for that cuz it's not authorized. It's not plugged in. You're not allowed to do that. Fine. That's that's a CIA conversation. You're a consumer. How do you work out what this is? How do you bridge it? you get you open chat GP and and and this I know I don't think I'm I don't think I know the answer here. I mean obviously I don't know the answer but like I'm I'm ambivalent here because I remember like portals on the internet in the late '9s where you give people a web browser and you can't just give them a browser with a URL bar and a blank screen. You've got to help them.

是的。

Yeah.

最终你不需要帮他们了。其中一部分原因并不是宽带或更好的技术，而是生态系统成熟了，人们养成了新习惯。所以当你现在打开一个聊天机器人，它有一些小方块，比如你可以做这个，做那个，做那个。那是不是就像门户网站一样，手把手地帮人们弄清楚该怎么做？我不知道。你在互联网上做的很多事情都是你已经在做的。比如现在买空调、订假期、买这个东西、找那个东西、看那个视频。这些你本来就知道。就像电子表格一样，你本来就知道电子表格。你本来就知道能做的事情。你用大语言模型做的事情有多少是你已经在做的，然后你意识到，啊，我可以做得更好。这显然是搜索对话和订假期的对话之类的。或者我关于找外套的观点。有多少是另一面，企业家会意识到，啊，不，这才是你该做的。我的意思是，这就是 Flickr 和 Instagram。

And eventually you didn't need to help them. And some of that is that wasn't really about broadband or better tech. that was about the ecosystem kind of maturing and people forming new habits. And so when you see you open a chatbot now and it's got these little tiles of like you could do this, you could do this, you could do this. Is that like the portal then of like the handholding of helping people work out what to do with this? I don't know. An awful lot of what you were doing on the internet was stuff you're already doing. like now buy an air conditioner, book a holiday, buy this thing, find this thing, watch that video. It's stuff you kind of already knew. It's like the spreadsheet point. Like you account, you already knew about spreadsheets. The thing you already knew you could do. How much of what you do with an LLM is stuff that you're already doing and you realize, ah, I could do that much better with this. This is obviously the search conversation and booking a holiday conversation or whatever. Or my my point about find me a coat. How much of it is the other side of that that an entrepreneur is going to realize, ah, no, this is what you do. I mean, this is Flickr and Instagram.

是的。

Yeah.

还有 TikTok。那不像消费者想，哦，我想要一种分享图片的方式。必须有人发明它。那不是消费者有机涌现的行为。不是说你给每个人足够的宽带，他们就会开始分享图片。

And Tik Tok. That's not like a consumer thinking, oh, I want a way to share pictures. Somebody has to invent that. That's not organically emergent behavior out of consumers. It's not like you just give enough everybody enough broadband and they'll start sharing pictures.

嗯，在企业方面，我们显然看到人们这样做，你知道，实验室倾向于这些部署公司，对吧？比如走出去的想法，你觉得那会成功吗？而且……

Well, I'm sure in the enterprise side, we've obviously seen people do this with um you know, the labs lean into these like deployment companies, right? And like the idea of like going out there and like do you think that will be successful? And

所以再说一次，我们有很多不同的方向。企业部署的第一步是给每个人一个副驾驶。第二步是那没用。

And so again, we're going in lots of different directions. Like step one of enterprise deployment was give everyone a co-pilot. Step two was a crap that didn't work.

对一小部分人来说，比如你给他们 Claude，也许很好，比如我妻子整天用 Claude。好吧。但大多数人知道，出于我们讨论过的原因，那没有意义。第二步是试点。每个人都做了一堆试点，部署了大约一半，它们运行良好，但那是一次一个的事情，然后对话变成：就这样吗？感觉不是。我们一次自动化一个痛点。

For like a small portion of people, like if you given them claude, maybe great like my wife lives in Claude. Fine. But most people know for the reasons we've talked about that doesn't make sense. Step two is pilots. and everyone has done now done a whole bunch of pilots and deployed about half of them and they they work fine but that's kind of a one at a time thing and then the conversation becomes like is that it that feels like that's not it well we automated this one pain point one at a time

是的，感觉这不是正确的思考方式。我们实际上如何从结构上改变我们围绕这项新技术做事的方式？我们仍然有点像我们成功地把所有目录做成 PDF 并放到网站上。很好，那有效。有多少人下载了？很多。我们的打印账单是不是崩溃了？我们有了更多客户。很好。好吧，但那可能不是终点。

Yeah feels like that's not the right way to think about this how do we actually change structurally how we do things around this new this new technology we're still sort of at the point of like we successfully PDFed all of our cataloges and put them on our website. Great, that worked. How many people have downloaded them? Loads. What's our printing bills collapsed? We've got way more customers. Great. Okay, but that's probably not the end point.

所以这就是所有这些事情的本质。这也是咨询存在的原因，因为大多数人没有那些技能，他们没有把它设置为一个流程，他们没有那么多闲着没事做的人。嗯，比如说你聘请贝恩、BCG 或麦肯锡，或者你聘请埃森哲、IBM、高知特、德勤、普华永道等等。嗯，或者你聘请 WPP、阳狮等等，或者你聘请爱德曼或其他什么公司来解决这些不同类型的问题，他们来找你，帮你解决。这就是为什么律师，这就是为什么专业服务存在，因为你没有所有这些内部人才。嗯，但弄清楚所有这些重新构想会是什么是一个项目。这是一个大而困难的项目。这就是以前所谓的数字化转型，现在叫企业转型。我总是开玩笑说，如果你说三次“数字化转型”，埃森哲的合伙人就会冒烟出现。这是我在会议上最可靠的笑点，尤其是企业会议。我说完之后，整个会议会停顿 30 秒。嗯，但是，是的，他们拥有大约 80 万人是有原因的，那可能会减半，可能会减半或任何数字。但没有公司会有所有这些技能，只是坐在那里准备重新构想。

And so that's what all of these things are. And this is of course why consulting exists because most people do not have those skills and they don't have it set up as a process and they don't have all those people sitting around not doing anything. Um and say you hire Bane or BCG or McKenzie or you hire Accenture, IBM, Cognizant, Deote, PWC and so on. Um or you hire WPPP publicist and so on or you hire Adelman or whoever it is for these different kinds of problems and they come to you and help you work it out. This is why or lawyers this is this is why professional services exist because you don't have all of that that talent in house. Um but working out what all of that reimagination would be is a project. It's a big difficult project. This is what used to be called digital transformation and now enterprise transformation. I always used to joke that if you say digital transformation three times then a partner from Accenture will appear in a puff of smoke. It's like my most reliable last line at conferences, especially enterprise conferences. The whole conference grinds for a halt for 30 seconds after I say that. Um but like yeah like there's a reason why they've got like 800,000 people and that will probably go to half that will probably half or whatever the number is. But like no company is going to have all of those skills there just sitting around ready to reimagine stuff.

不过，我想到了一个问题，那就是企业软件公司，尤其是垂直企业软件公司所做的，与咨询公司所做的，有一种对称性或重叠性。

I think the one thing that occurred to me there though is that um there's a sort of symmetry or an overlap between what an enterprise software company does, a vertical enterprise software particularly does and what a consulting firm does

那就是战略咨询公司、管理咨询公司，比如贝恩、BCG、麦肯锡等等。他们所做的就是去看看你的公司，然后说：“你这样做，但实际上你可以这样做，那样会好得多。”

In that what they an strategy consulting firm management consulting like Bane BCG Mckenzie and so on. What they're doing is they're kind of going and looking at your company and saying, "You're kind of doing it like this, but actually you could do it like this and that would work way better."

100%。我的意思是，这就是很多垂直 AI 公司今天所做的，对吧？你必须拥有它。他们称之为部署，也就是，你知道，重新发明。

100%. I mean, that's what a lot of these vertical AI companies are doing today, right? You have to have it. They call this deployed, which is, you know, reinvention.

但那是任何垂直软件公司都在做的。是的。

But that's what any vertical software company's doing. Yeah.

嗯，至少一半的公司是这样做的。他们说：“嗯，你以前这样做，我们想出了你可以这样做，那样会好得多。”这也是贝恩、麦肯锡、BCG 所做的，但只是以一种不同的方式，这就是为什么所有这些公司都在抓耳挠腮，试图弄清楚，你知道，我的意思是，你知道，有个笑话：机器学习科学家是住在硅谷的统计学家。肯定有，我还没完全想出正确的等价物，前沿模型部署工程师就像埃森哲的“身体贩子”。

Well, at least half of them, that's what they're doing. They're saying, "Well, you were doing it like this, and we've worked out you could do it like this, and that would work way better." And that's also what Bane does, what Mackenzie does, what BCG does, but just in a kind of a different modality, which is why um all of these companies are sort of sitting and scratching their heads and trying to work out like you know, I mean, you know, there was a joke that a machine learning scientist is a statistician who lives in Silicon Valley and there must be I haven't quite worked out the right equivalent for a forward engine deployed engineer is like an Accenture body shopper

在 OpenAI 找到工作。嗯，但他们有点像在从相反的方向爬山。

Who got a job got a job at at OpenAI. Um, but they're kind of coming at they're climbing the mountain from opposite directions.

这就像，你是想要一家拥有大量前置部署工程师的软件公司，还是想要一家拥有大量软件工程师的咨询公司？因为它们其实是从不同方向做同一件事。我认为这一点基本上很明显，在企业领域，你不能直接把模型交给人们，他们就能完全搞清楚怎么用。一直都需要一个大的翻译层。然后我认为在消费者端，到目前为止肯定也是这样的。而且我觉得这个问题，你知道，就是你之前在 Instagram 上提到的那个点，就是会不会有一些伟大的创业者出现，或者也许在这些公司内部，找到一种方式，稍微有点，你知道，能挖掘出人们内心潜在的、他们想做的行为？

It's like, do you want a software company that's got a bunch of forward-deployed engineers, or do you want a consulting company that's got a bunch of software engineers? Because they're kind of doing the same thing from different directions. I think this point of basically obviously in the enterprise, you know, people can't just give people these models and they figure out exactly how to use them. There's been needed to be this big translation layer. And then I think on the consumer side, it's certainly been that to date. And I think the question, you know, to the point you're making on Instagram earlier, is like, will some great entrepreneurs come along, or maybe within these companies, and find a way that's a little more, you know, that has some latent behavior within people that they want to do?

所有东西实际上都得被发明出来。所有的用例都得被发明。很少有纯粹自发的、草根式的消费者行为。我们手机上的每一个应用，都得有人去想，“嗯，那会是个好主意。”然后所有人看了都说，“这是我听过最蠢的东西。”我的意思是，还记得 Uber 看起来有多蠢吗？还记得 Instacart 看起来有多蠢吗？我当时在 A16Z 的 pitch 现场，就是 Instacart 那次。

Everything has to actually be invented. All of the use cases have to be invented. Very few of them are just purely spontaneous grassroots consumer behavior. All every app on our phones, somebody had to think, "Well, that would be a good idea." And then everybody looked at it and said, "That's the dumbest thing I've ever heard." I mean, remember how dumb Uber looked? Remember how dumb Instacart looked? I was in the pitch at A16Z for Instacart.

就像，嗯，经济上可行，因为这说得通。但人们会这么做吗？我的意思是，你知道那个创始人的故事，他叫什么来着，他没有车。所以他建了一个很简陋的网站，把订单发到他的邮箱，然后他叫个 Uber 去超市采购，再叫个 Uber 送到人家家里。

Like, well, the economics work because this makes sense. Will people do this? I mean, you know the story about what's his name, the founder, that he didn't have a car. And so he set up this barebones website that emailed him orders, and then he would get an Uber and go to a supermarket and do the shop, and get an Uber to people's homes to deliver them.

很难从经济上判断这是否可行。

Tough for the economics to work out whether this would work.

是啊，太搞笑了。

Yeah, that's hilarious.

那么，你怎么看 Sora 以及一些早期的消费者应用尝试？

Well, what do you make of Sora and like some of these early consumer app drives?

所以我认为 Sora 符合消费者社交的经典模式，就是你得想出一些新的机制、一些新的群体层级、一些新的感受方式，来吸引注意力，但它不能只是个噱头。TikTok 以一种形式做到了，Snap 也做到了，再举一个例子，这真的很难。Snap 可能找到了两三个，比如阅后即焚、故事功能等等。但它确实有效。我们可以说，社交在 S 曲线的顶端已经趋于平缓，就像智能手机或 PC 那样，所有东西都被发明出来了。而 Sora 就像，我不知道，就像人们在智能手机上尝试过的那些愚蠢的、奇怪的实验，比如模块化手机，你扣上这个或扣上那个，比如背面加个第二屏幕或耳屏。你可能会觉得，好吧，不，所有社交的东西都做完了。然后当然，Datator 出现了，做了别的事情。但我觉得这就是我把它放进的那个框架，而不是专门放在 AI 的叙事里。那只是又一次尝试做一个酷炫的社交东西，结果发现它不太行。我的意思是，Midjourney 也是一样。你还记得吗，一两年年前，我们都花了一个月玩 Midjourney。

So I think Sora fits into a classic pattern of consumer social, which is you have to come up with some new mechanic, some new piece of mass hierarchy, some new way of feeling that gets attention, but then it has to be more than a gimmick. And TikTok found that in one form, and Snap found that, and name another one. Like it's really hard actually. Snap probably found two or three like disappearing messages and stories and so on. But like it's working. Like we could kind of argue that social has flattened out at the top of the S-curve the way smartphones did or the way PCs did, that like everything's been invented. And like Sora was like, I don't know, it was like all the stupid like the weird experiments people tried with smartphones, you know, like modular smartphones and things like you clip on this or clip on that, like second screen or ear screen on the back. You may like there's a moment where like okay no, all of social stuff has been done. And then of course Datator comes along and does something else. But I think that's the frame that I would fit it into that narrative rather than specifically an AI narrative. It was another attempt at a cool social thing, and it turned out that it kind of didn't work. I mean it's the same thing with Midjourney. You remember like a year or two ago we all spent a month playing with Midjourney.

现在 Meta 有了一个新的图像模型。我试过了吗？就是图像生成。

And now Meta's got a new image model. Have I tried it? The image generation.

我的意思是，这里还有一点要说，就像无人机和 3D 打印发生的情况。比如某个圣诞节，不管是什么时候，10 年前，我们都在圣诞节收到了无人机，然后 3 天后我们说，“好吧，我现在已经看过我家屋顶了。”这很酷。3D 打印机也一样。“好吧，我做了个小埃菲尔铁塔。”就像 3D 打印没有消费者用例。绘画也没有消费者用例。有爱好者用例。有一大堆人觉得它很酷。但那是短暂的，它不是每个人都有的东西。所以我认为视频创作最终就落到了那个境地。现在可能有人需要把整个事情翻个底朝天。现在比如虚拟试穿。如果你能让虚拟试穿工作起来，那会是图像生成，但不会是 Midjourney。是的。你知道，看看我的 Instagram，去看看这个品牌最新的 lookbook。做 10 个我穿不同衣服的视频。我觉得那可能完全疯狂，但那会是一个产品。是的，Midjourney 不是一个产品，因为你需要用 Discord 登录，但它也不是一个，它有和无人机一样的问题。它没有转化成一个具体的、能解决问题的、你会去做的事情。

I mean, there's another point to make here, which is like what happened with drones and 3D printing. Which is like one Christmas, whatever whenever it was 10 years ago, we all got a drone for Christmas, and then 3 days later we say, "Okay, I've seen the roof of my house now." It's very cool. Same with 3D printers. "Okay, I made a little Eiffel Tower." Like there's no consumer use case for 3D printing. There is no consumer use case for drawing. There's a hobbyist case. There's a bunch of people who think it's really cool. There was this brief, like it's not a thing that everyone has. And so I think that was kind of where video creation ended up. Now somebody may need to pull the whole thing inside out. And now like virtual try-ons. If you could get virtual try-ons working, that would be image generation, but it wouldn't be Midjourney. Yeah. You know, here's look at my Instagram and go look at the latest lookbook from this brand. Make 10 videos of me in different looks. I'm like that's probably completely insane, but like that would be a product. Yeah, Midjourney wasn't a product in the sense that you had to log into it with Discord, but it also wasn't like a, it had the same problem as drones. It's like it didn't translate into a tangible specific thing that you would do that solved a problem.

是的。不，非常有趣。这是我们整个对话中一个共同的主题。我总是喜欢以一个快速问答环节来结束，就是我把一堆我们没讨论过的东西塞到结尾。所以也许开始吧，我很好奇，你知道，我觉得你对这项技术的看法一直相当一致。在过去一年里，你对 AI 的看法有什么改变吗？

Yeah. No, super interesting. That's a common theme throughout our conversation. I always like to end with kind of like a quick fire round where I just stuff a bunch of things into the end that we haven't discussed. And so maybe to start, I'm curious like, you know, I feel like you've been pretty consistent in your view on this technology. What's one thing you've changed your mind on in the last year around AI?

我仍然看到人们犯的基本错误是，没有理解这是一项使能技术，它让很多不同的事情成为可能，而仍然把它看作是一种生成文本和图片的东西，没有理解不，这会让欺诈检测工作得更好，或者这会解决你每天使用的公司内部各种晦涩的后台流程，你从未想过。所以你会得到各种更好的体验和更好的产品，它们不会以 AI 的形式出现在你面前，也不会告诉你这是 ChatGPT。而且你仍然会听到人们说，“但是你知道，我问了它一个问题，答案不太对。”或者，好吧。这就像看着互联网说它有点慢，任何人都能说任何话。就像，是的。

The basic mistake that I still see people making is not to understand that this is an enabling technology that makes lots of different things possible, and still look at it as something that makes text and pictures, and not understand no, this will like make fraud detection work better, or this will solve all sorts of obscure back office processes inside companies you use every day that you've never thought of. And so you'll get all sorts of better experiences and better products that won't come to you looking as AI and won't come to you saying this was ChatGPT. And you still get people kind of saying, "But you know, I asked it a question and the answer wasn't quite right." Or like, great. This is kind of like looking at the internet and saying it's kind of slow and anyone can say anything. Like yes.

是的，但这更像是看 1995 年的互联网，然后说，但是……网上有一段采访，是大卫·鲍伊和英国电视记者杰里米·帕克斯曼在 1994 年左右做的，大卫·鲍伊说这会改变世界，杰里米·帕克斯曼说，但这只是陌生人在聊天室里聊天。这就是错误所在，不是聊天室，而是网络。这里也一样，不是文字和图片，而是使能技术。

Yeah, but well more it's like looking at the internet in like 1995 and saying well but but but like there was an interview you can find online with um David Bowie and a British journalist called TV journalist called Jeremy Paxman in like 1994 or something and David Bowie is saying this is going to change the world and Jeremy Paxman says but it's just like random people talking to each other on chat rooms this is the mistake it's not the chat rooms it's the network and it's the same thing here it's not the text in the pictures it's the enabling technology

是的，但我觉得你写的很多东西都把 AI 正在发生的事情与其他技术革命进行对比，以提供背景。我认为一个有趣的事情是，这些实验室里处于 AI 前沿的许多人都是二十多岁，他们不一定有同样的背景。

Yeah, but I think obviously a lot of what you write about compares what's happening in AI you know to other technological revolutions for context and I think one interesting thing is you know many people at the forefront of AI at these labs are like in their 20s they don't necessarily have the same context around some of these

嗯，他们中有很多人听这个播客。我想，如果你能给他们传递一些关于那些时代的信息，那会是什么呢？

Uh a bunch of them listen to the podcast like I guess if you could impart a few messages to them on you know maybe uh on those eras like what would they be

所以，我觉得我们已经讨论了很多这类事情，比如事情是不确定的，事情会改变，还有所有的缩写词。我演示文稿里的幻灯片是：这里有一大堆 1995 年的缩写词、公司、概念、想法和技术，它们都失败了。2010 年或 2005 年也一样。我们应该假设我们现在正在做的很多事情都不会成功。其中一些你非常清楚。我是说，这就是去年关于 OpenAI 的对话。当然，其中一些不会成功，但值得尝试。我的意思是，就在上周，OpenAI 放弃了网页浏览器。

So I mean I think we've talked about a bunch of this stuff that like you know stuff is unclear and stuff will change and all the acronyms and you the slide in my presentation is here are a whole bunch of acronyms and companies and concepts and ideas and technologies from 1995 that that failed. Same thing from 2010 or 2005. We should just presume a bunch of stuffs that we're working on now won't work. And some of it you are very conscious of. I mean this was what the conversation about open AI last year. Of course some of this wasn't going to work but it was worth trying. I mean I think just like last week Open AI gave up on the web browser.

是的。

Yeah.

我的幻灯片上就有这个。这是会成功的事情吗？可能不会。不，我们应该假设 MCP 会被其他东西取代，或者可能会。所以这是一个要点。只是假设你处于这个彻底不确定的阶段，一切都在翻转和变化。

And I had that on my sliders. Is this a thing that will work? Probably not. No, we should presume MCP will get replaced by something else or may do and so on. So that's kind of one point. just presume you're in this stage of radical uncertainty where everything is turning over and changing.

其次，你知道，我总是展示这张幻灯片，来自高盛 CIO 调查，企业工作流在云上的百分比，数字大约是 30%。是的。

Secondly, like you know, I always have this this slide I show of the so from the Goldman CIO survey of the percentage of enterprise workflows that are in the cloud and the number is like 30%. Yeah.

20 年后，这需要时间，而且困难复杂，人们有其他优先事项。我时不时在社交媒体上做调查。机器学习还是 AI 吗？

After 20 years like it takes time and it's hard and complicated and people have other priorities. I do a poll on social media every now and then. Is machine learning still AI?

有些人非常沮丧，因为他们凭空想出了一个学术定义，关于 AI 是什么。但 AI 就是新的东西。就像技术，一旦存在一段时间，就不再是技术了。嗯，但我当时在想的是，我给一家欧洲水务公司做了这个演示，我做了关于机器学习的演示，最后 CEO 说这非常有趣，也许创新应该成为我们明年的五大优先事项之一。

And some people get very upset like there's this academic definition that they've dreamt up of what AI is. But like AI is whatever's just is new. It's like technology. It's not technology once it's been around for a while. Um but the the point I I was sort of thinking is like um I gave this presentation to I think it was a like a a European water company and I gave the presentation on machine learning and the end the CIA says CI CEO says this is very interesting. Maybe innovation should be one of our top five priorities next year.

如果你在科技行业，这听起来很疯狂。然后你想想，但他们是水务公司。让我们想想他们的优先事项可能是什么。监管、地震、管道里有铅。他们需要拆除华为设备。明年会有重新定价。他们还有其他事情。有干旱。那个水库，大坝在漏水。哦，糟糕，我们需要削减其余预算。所以大多数人都有其他事情在进行。然后可能更实际的事情是，其他人的工作也很难。其他每个人的工作都很难。其他每个行业都有一堆他们试图解决的事情，你可能不知道。

And if you're in tech, that sounds insane. And then you think, but they're a water company. Let's let's think about what their priorities might be. Regulation, earthquake, there's lead in the pipes. They need to rip out the Huawei equipment. There's a repricing coming up next year. Like they've got other stuff. There's a drought. That reservoir, the dam is leaking. Oh crap, we're going to need to like screw the rest of our budget. So like most people have other stuff going on. And then maybe a much more tangible thing, other people's jobs are hard, too. Everybody else's job is hard. Every other industry has a bunch of stuff they're trying to work out that you probably don't know about.

我记得几年前在一个活动上，与现在世界上最大的广告公司之一的负责人交谈。我说，你知道，试图表现得聪明，你担心什么？你担心机器学习吗？这是在 OpenAI 之前。他说，是的，我还担心我们某个国家没有好的创意主管。我担心这个大客户会让我们进行审查。我担心我的 XYZ 主管与另一个人有冲突。每个人都有所有这些其他事情在进行。AI 不是唯一的事情。如果你想想，你知道，谈论以前的教训，你回去想想互联网影响了哪些行业，没有影响哪些行业。如果你是报纸，好吧，你被摧毁了。现在想象你是卡特彼勒。互联网为卡特彼勒改变了什么？好吧，它非常有用。归根结底，你的业务是制造大块金属。

I remember having a conversation with the head of now the head of one of the world's biggest ad agencies at an event years ago. And I said, you know, trying to be clever, like, what are you worrying about? You worried about machine learning? This is before open AI. He said, yeah, I'm also worried that one of our countries doesn't have a good head of creative. I'm worried that this big client is going to be put us up for review. I'm worried that my head of XYZ is in a conflict with this other person. Everyone has all this other stuff going on. And AI isn't the only thing. And if you think about, you know, this is this, you know, talking about previous lessons like you go back and think about what industries the internet affected and didn't affect. If you're the newspaper, fine, you got demolished. Now imagine you're Caterpillar. What did the internet change for Caterpillar? Okay. It was really useful. In the end of the day, your business is making big pieces of metal.

是的。

Yeah.

互联网为骨料公司改变了什么？你知道骨料公司是什么吗？你的工作是从地下挖沙子，放到火车上，然后放到卡车上，然后交给人们。互联网改变了你的行业多少？嗯，最终，改变不大。AI 将改变那个行业多少？取决于物理 AI 的东西效果如何，我猜。

What did the internet change for an aggregates company? Do you even know what an aggregates company is? Your job is digging sand out of the ground, putting it on trains, and then on trucks, and then giving it to people. How much did the internet change your industry? Well, in the end, kind of not very much. How much is AI going to change that industry? Depends how well the physical AI stuff works, I guess.

是的。但你想把类人机器人放进三层高的卡车里，还是卡车会连接 GPS 并自动驾驶？它们可能已经这样了。卡车会实现自动驾驶，没问题。好吧。那实际上改变了你的行业多少？你猜怎么着？有些行业它不会改变任何东西。而且这将是不可预测的。可能会有一些，你知道，你现在想到的一些，你完全错了。会有一堆行业，互联网实际上没有改变任何该死的东西。

Yeah. But do you want to put are you going to put humanoid robots in the threetory high trucks or the trucks going to get connected to GPS and drive themselves around? They probably already are. The trucks will get autonomous fine. Okay. How much does that actually change your industry? Like guess what? There are some industries where it's not going to change anything. And it won't be predictable. There'll be some maybe, you know, there'll be some you think of now where you're completely wrong. There will be a bunch of industries where the internet actually didn't change a damn thing.

好吧，我想确保把最后一句话留给你。我们这里没有谈到的任何东西，你认为我们的听众应该带走，或者你想让他们去哪里？我们肯定会链接到你的演示文稿，但还有其他什么吗？

Well, I want to make sure to leave the last word to you. anything we haven't talked about here uh that you think our listeners should take away or anywhere you want to point them uh we'll certainly link to your presentations but anything else uh

是的，嗯，我父母有很好的 SEO。有一个我最近想到的故事，我没有用在写关于代币的文章里。那是在我工作的银行，我们打算研究一家在互联网上销售电子产品和软件公司的 IPO。他们以前是一家目录公司，当时还不清楚亚马逊会这样做。创始人 16 岁在卧室里起步，他们营业额超过 1 亿英镑，他们将以 7 亿或 8 亿英镑的价格上市，即使在当时这也是疯狂的。所以我们所有人都去伯明翰看它，然后在回来的火车上，我们都在谈论创始人和董事会有点奇怪，以及估值。一位资深银行家，叫大卫·泰特，我想他现在退休了，我应该查一下他。大卫·泰特睁开一只眼睛说，这是一个低利润的转售商，一次性销售，然后又睡着了。

Yeah my well my parents had good SEO um there's a story I was I was thinking about recently which I didn't use writing about tokens which was in like um the bank I worked for was going to look at working on the IPO of a company that sold electronics and software on the internet um they'd previously been a catalog company and it wasn't clear Amazon would do this and the guy had the founder had started in his bedroom age 16 and they like turn over 100 million quid and they were going to flight for 7 or 800 million which was insane even then and so we all go up to Birmingham to see it and then we're on the train back and um we're all talking about the founder and the board's kind of weird and the valuation and um guy veteran banker called David Tate I think he's retired now I should look him up David Tate sort of opens one eye and says um it's a low margin reseller one time sells and goes back to

我的意思是，有时候事情就是那么简单。对我来说，这是一个劳动力卖家。

I mean, sometimes it is kind of simple. It's a labor to me seller.

几点了？

What time?

能和你聊这些话题真是太开心了。非常感谢你抽出时间来参加播客。

This has been so much fun to get to talk to you about all this stuff. I really appreciate taking the time to come on the pod.

当然，谢谢你的邀请。

Sure. Thanks for having me.

我是 Jacob Efron，这里是 Unsupervised Learning 播客。在这个节目里，我会和 AI 领域最聪明的人对话，问他们大量关于模型发展以及这对商业世界意味着什么的问题。如你所见，我对此充满热情。这是我利用晚上和周末做的项目，我的本职工作是 Redpoint 的投资人。我们能请到这些出色的嘉宾，全靠像你这样的听众订阅播客、分享给朋友。正是这些让整个节目得以运转。所以，请考虑这样做。非常感谢你的支持和收听。我们下期再见。

I'm Jacob Efron and this has been Unsupervised Learning, a podcast where I get to talk to the smartest people in AI and ask them tons of questions about what's happening with models and what it means for businesses in the world. As I hope is clear, I have a ton of fun doing this. It's a nights and weekends project in addition to my day job as an investor at Redpoint. But our ability to get these incredible guests on really comes from folks like you subscribing to the podcast, sharing it with friends. It's really what ultimately makes this whole thing work. And so, please consider doing that. And thank you so much for your support and listening. We'll see you next episode.