本尼迪克特·埃文斯：OpenAI 的护城河问题与软件的未来

Turbo Pascal。

Turbo Pascal.

我不知道。

I don't know.

那不是我不清楚。

Wasn't that I don't know.

不管当时说的是什么编程工具，你都应该自己编写库存管理软件。

Whatever the say programming stuff would have been, and you should program your own inventory management software.

或者你应该买一个数据库程序，然后在数据库程序里自己开发库存管理软件。

Or you should buy like a database program, and you should make your own inventory management software in the database program.

这两种思维方式都不对。

And neither of those were the right way of thinking about this.

事实上，我找到了一篇1980年《纽约时报》关于VisiCalc和电子表格的精彩引述。

In fact, I found this wonderful quote from the New York Times from 1980 talking about VisiCalc and spreadsheets.

有位男士（当然，是个男士）说：‘我以前用机器语言编写折现现金流模型，要花二十个小时。’

And there's some this this guy says, of course, it was a guy who says, well, I used to code DCF models in machine code, and it would take me twenty hours.

现在有了VisiCalc，我只需要十五分钟就能完成。

And now with Visicality, I can do it in fifteen minutes.

所以，下次当你听到软件开发者说AI是完全不同的东西，以前从来没有人这样抽象过软件时，你就这么告诉他。

So tell that next time you hear a software developer saying, like, AI is a completely different thing and nobody has ever abstracted software like this before.

是的。

Like, yeah.

我们已经这样做了三十年、五十年了。

We've been doing this for thirty years, fifty years.

所以，这个问题是，你在哪里创建应用程序和使用场景？

So this is a sort of question is, like, where do you create the applications and the use cases?

谁来做这件事？

Who does it?

这离原始的聊天机器人有多近？

How close does that go to the raw chatbot?

聊天机器人在多高的层级上完成这些工作？

How up the how far up the stack does the chatbot do it?

所以现在你知道了，ChatGPT应用曾经失败了两三次。

So now, you know, it was it was ChatGPT apps, which failed, like, two or three times.

当然，还有OpenClaw的技能，或者不管OpenClaw怎么称呼它们。

And, of course, it's OpenClaw skills as well or whatever OpenClaw calls them.

你通常更看好Anthropic吗？

Are you generally more bullish on Anthropic?

没怎么觉得。

Not really.

我的意思是，这周他们风头正劲。

I mean, this week, they've got all the fire.

他们掌握了所有的资源，不管那叫什么。

They've got all the the juice, whatever the word is.

我不知道。

I don't know.

这周，下周。

This week this week next week.

下周又会是别的东西。

Next week, it'll be something else.

OpenClaw 对我来说挺有意思的，因为它看起来很像桌面版 Linux。

OpenClaw is kind of interesting to me because it looks a lot kind of like Desktop Linux.

也就是说，你之前不得不使用这些大公司提供的庞大而封闭的系统，根本无法接触到底层硬件。

In that, like, suddenly, you you've been having to use this monolithic thing from these big companies where you couldn't really get your hands on the metal.

你只能拿到 API 密钥，输入一些内容，然后它们告诉你能做什么。

You were just like you had to get your API key and you type stuff in and they tell you what you could do.

而现在，你可以亲手自己搭建它。

And now you can, like, build it yourself with your own hands.

这就像是七十年代的自制计算机俱乐部。

It's like the homebrew compute club in the seventies.

你可以自己动手做出它。

And you can make it yourself.

当然，另一面是，它还需要再打磨一下，就能面向所有人了——而这，当然，Linux 用户已经说了三十年了。

And then the other side of this, of course, is and it just needs to be a little bit more polished and then it'll be ready for everybody, which, of course, Linux people have been saying for thirty years.

他们根本意识不到，这么做从根上就违背了这套东西的整个文化内核。

And like they don't understand like the whole culture of it is just wrong to do that.

另外，关于OpenClaw还有一个值得玩味的点：它能解释清楚为什么谷歌和苹果一直没推出这类东西。

And, you know, you ask it to tie you know, what the other interesting thing about about OpenClaw is it shows you why Google and Apple haven't shipped this.

比如让系统帮我整理收件箱。

Like tidy up my inbox.

好的。

Okay.

我把你所有的消息都删了。

I deleted all your messages.

挺好。

Great.

不客气。

You're welcome.

对。

Yeah.

所以在中国，OpenClaw 引发了巨大的热潮，而我显然并不了解中国的情况。

So there's this huge craze around OpenClaw in China, which I obviously, I don't speak to China.

这本身就很有趣。

It's interesting in its own right.

但它也凸显了人们积压了多少热情，有多少可能性，但讽刺的是，真正把这一切变成现实却异常困难。

But it it gets to this sense of how much pent up enthusiasm there is how many possibilities there are, but also in paradoxically how hard it is actually to make a real make a real thing out of this.

AI 助手的有趣之处在于，这确实是一个绝佳的使用场景。

This is the kind of the funny thing about AI assistants is like this is really a great use case.

只不过，真正实现它时，要确保它不会删除你的收件箱，难度要大得多。

Except it's much much much harder to actually do it when and be sure that it won't delete your inbox.

是的。

Yep.

而且你提到的这一点，还有 Anthropic，我猜你并不认同人们会用云代码自己开发软件这个观点。

And to the point that you're just making and and also anthropic, presumably, you're not in the camp of believing that, people are gonna be building their own software with cloud code.

这很有趣。

It's funny.

这简直就像一个稻草人，但事实上，确实有一些真实、理性、有意识的人在说这些话。

This is like, it's almost like a straw man, except there are, like, presumably real, rational, sentient people who actually say this stuff.

我不觉得解释为什么人们不会自己开发Stripe有什么特别有趣的。

I don't think it's particularly interesting to explain why people aren't gonna code their own Stripe.

我觉得更有趣的是，我所思考的软件更广泛的分类，即你有大型企业级ERP系统，以及他们所谓的记录系统。

I think what's more interesting is to kind of do what I think about a broader taxonomy of software, which is that you have, like, the big iron enterprise ERPs and and so what they call systems of record.

比如，你需要为五万人和每年十亿笔交易，以相同的方式存储相同的数据。

So, like, you need to store the same data in the same way for 50,000 people and a billion transactions a year.

不行。

No.

你不可能让《办公室》里的大卫·布伦特自己去开发这样的系统。

You can't have David Brent from the office, like, making his own.

然后你还有那些任务和工作流程，像SAP这样的单一系统太过僵化。

Then you have but then you have tasks and workflows where that single system, like, it SAP, is too too inflexible.

因此你把它拆解开来，变成专门的、垂直的工具，这就是垂直工具。

And so you break it out, unbundle it into something dedicated and vertical, you get vertical tool.

因此，如今典型的大型美国公司拥有四到五百个垂直SaaS应用，具体数字因统计方式而异。

And so this is why the typical big US company today has, depending on your numbers, like four to 500 vertical SaaS apps.

我的意思是，你们投资了Frame.io，对吧。

I mean, you guys invested in frame.io, I think.

是的。

Yeah.

对。

Yeah.

Frame.io正在将Google Sheets和Dropbox拆分出来。

And frame.io is unbundling Google Sheets and Dropbox.

嗯。

Mhmm.

理论上，你可以看出人们当时就是在这么做。

Like, theoretically, you can tell that's what people were doing.

他们当时用Google Sheets、Dropbox和电子邮件来管理视频工作流程。

They were managing video workflows in Google Sheets and Dropbox and email.

它说，不行。

And it says, no.

我们要把它变成一个专用工具。

We're going to turn this into a dedicated tool.

我不知道。

And I don't know.

理论上，你本可以在Oracle里做，但不行。

Theoretically, you could have done that in Oracle, but, like, no.

所以你有了专门的垂直工具，也有通用的横向工具。

So you've got the, like, the vertical tool that's specialized, and you've got the horizontal general purpose tool.

还有一种中间情况，就是使用Google Sheets和邮件，或者使用Excel、导出CSV文件之类的，这几乎可以被称为临时软件——因为这个流程并不是以相同的方式、相同的时间反复进行，也不需要被记录、追踪、满足合规性等要求。

And then you've got this middle case of using Google Sheets and email or using Excel or exporting a CSV or something where you could almost kind of call it like improvised software, where it's not a process that's being done over and over again in the same way, the same time, and has to be recorded and tracked and has to have compliance and everything else.

而且这种需求还不够多、不够大，或者还没人意识到它足够大，值得开发一个专用工具。

And it's not kind of enough of it or big enough or no one's realized it's big enough to make a dedicated tool.

所以你只能用Excel、Tableau、Google Sheets或邮件之类的东西来处理。

And so you're doing it in Excel or Tableau or Google Sheets or email or something.

现在，一方面，AI编程意味着编程变得便宜和简单得多，同时也意味着过去无法用软件实现的许多事情现在都可以做到了。

And now on the one side, AI coding means coding is way cheaper and easier and also means that there's a whole bunch of stuff that you couldn't do in software that now you can.

因此，软件的数量会大大增加，这将覆盖更多以前未被自动化的用例，要么是因为这些用例太小，要么是因为过去根本无法用软件实现这些自动化。

And so there will be way more software, and that will pick up many more of those use cases, either that weren't automated before, either because they were too small or because you could actually couldn't automate that thing with software before.

现在有了AI，你可以做到了。

Now with with AI, you can.

你觉得可能会有

You think there could be

一个与之相关的概念，叫做临时性软件吗？

a a related concept of ephemeral software?

嗯，这将是另一半内容：你现在扩展了这种‘临时性软件’的范畴。

Well, this is this is gonna be the other half of it is now you've got you expand this category of, like, improvised software.

以前，你可能会导出CSV文件，然后用Excel处理，或者如果你是那种人，也可能写个Perl脚本。

And so where something it might have been that you do it as get the CSV and do it in Excel and or maybe a Perl script if you're that kind of person.

现在，你可能会直接让ChatGPT帮你完成。

Now maybe you'll ask ChatGPT to do it for you.

帮我分析一下这个。

Analyze this for me.

帮我做一下这个。

Do this for me.

而且很多时候，这些是你以前在 Excel 里根本做不到的事情。

And and then often, it's like stuff that you couldn't have done in Excel before.

你知道，看看那些 PDF 文件。

You know, it's like, look at all those PDFs.

看看那些 PowerPoint，再看看这个 PowerPoint。

Look at all those PowerPoints, and then look at this PowerPoint.

所有这些 PowerPoint 里，有没有我们在这个里面没提到的内容？

Is there something in all of those PowerPoints that we haven't said in this one?

这种问题的答案通常是模糊的，但刚好适合 LLM 处理，不过如果有 PDF 就不行了，因为你读不了 PDF。

Which is the kind of thing that's got the sort of fuzzy answer that it's kind of right that works very well with an LLM, except it wouldn't work if there are PDFs because you can't read PDFs.

所以你得告诉人们，你读不了 PDF。

So you have to tell people you can't read PDFs.

所以这种中间地带的感觉，也许我的分类法有误，但我认为这样看待问题更有意思，而不是简单地说：哦，每个人都会用自己的方式用AI生成代码。

So that sense of, like, this sort of middle ground, and maybe my taxonomy is wrong, but I think that's more interesting as a way of looking than just saying, oh, everyone will just vibe code their own their own thing.

不。

No.

没人会用AI生成自己的ERP系统或自己的Frame.io，但他们可能会问Anthropic、Gemini或者ChatGPT：你能帮我做这件事吗？

No one will vibe code their own ERP or their own frame.io, but they may ask Anthropic or Gemini or or or ChatGPT, can you do this thing for me?

这种方式类似于他们过去用Excel来完成任务。

In a way that's sort of analogous to they might have done it in Excel.

那你觉得软件和SaaS会变成什么样？

So what do you think happens to software than than SaaS?

会涌现出更多的软件。

Way more software.

更多的软件。

More software.

但软件作为一个独立类别，在股票市场上的地位会怎样？

But software as an independent category in in stock markets.

再拆解一下，因为从某个层面来说，如果编写软件变得便宜且容易得多，那么软件的数量自然会增加。

Again, unpick that because on one level, like, okay, there will be if it's much cheaper and easier to write software, there'll be more software.

那谁来完成这些工作呢？

Who will be doing that?

谁会是那个既理解软件运作原理，又深入思考并真正理解问题的人呢？

Who under is is a combination of who understands how software works and who has really thought carefully and understood the problem.

Frame.io 并不是由某位在苏荷区视频制作公司工作的普通人开发出来的，因为他们的思维方式根本不是这样的。

And frame.io was not made by some guy or some woman working at a video production suite somewhere in Soho because that's not how they think.

他们不是搞软件的人。

They're not software people.

他们也不是产品专家。

They're also not product people.

真正要弄清楚代码该做什么、问题是什么、该如何解决，这需要一种不同的能力。

You know, there's a different skill to actually work out what should the code be doing and what is the problem and how would you work this out.

因此，软件的数量会大幅增加，既有那些只需用 AI 来创建但无需 AI 运行的软件，也有那些必须借助 AI 才能实现新功能的软件。

So there will be way more software, both stuff that doesn't need AI except to create it and stuff that needs to use AI to do the new thing.

一些现有的软件将会消失，比如那些大型专家系统，因为大语言模型能做得更好。

Some incumbent software will go away, like it will like big expert systems where an LLM can do that better.

这很简单，但在我看来，这类似于SaaS的重复，SaaS意味着软件数量增加了十倍，甚至可能二十倍。

It's easy just as but there's also this seems to me like an analogy of SaaS or a repeat of SaaS in that SaaS meant that you had an order of magnitude, probably two orders of magnitude more software.

一些现有企业因此被彻底击垮。

And some incumbents got completely screwed by this.

许多以前无法自动化的任务因为SaaS而实现了自动化，因为SaaS让进入市场变得便宜得多、容易得多，也更容易解决这些问题。

And loads of tasks that you couldn't have automated before got automated because SaaS made it much cheaper and much easier to go to market and easier to unlock these problems.

所以，正如我所说，当我们还在使用大型机时，大公司大概只有五款软件。

And so, as I said, you know, when we went for mainframes, like the big company had what, like, five pieces of software.

而在本地部署时代，你有几十款软件。

And with on prem, you've got dozens of pieces of software.

而在云时代，你有数百款软件。

And with cloud, you have hundreds of pieces of software.

因此，你应该直接假设，有了这项技术，软件的数量将会多得多。

And so you should just sort of presume that with this, you'll have, like, way more software.

你还会看到更多事情被自动化，而无需专门配备工具，也不必闭上眼睛去仔细梳理工作流程中的每一步、它们应该如何运作以及你该如何处理。

You'll also have way more stuff being automated where you don't need to dedicate a tool, where you don't need to really shut your eyes and work out every step of the workflow in the process and how they should work and what you should do with it.

这就像一种临时即兴的做法，你只需让模型去完成它。

It is just like a one off improvised thing where you'll get the the model to do it.

因此，会有更多事情通过软件来完成。

So you'll have way more stuff being done in software.

我的意思是，去年突然间，每个人都去维基百科上搜索了杰文斯悖论。

I mean, this is, you know, people last year suddenly everyone discovered, looked up the Jevons Paradox in Wikipedia.

假装他们早就知道这个概念了

Pretended that they knew all along what

是什么。

it is.

是的。

Yeah.

挺有趣的，其实我确实对这个有所了解，但更多是因为我对工业史之类的东西感兴趣。

Funny if I actually kind of did know about this, but more just because I'm sort of interested in industrial history or something.

我依稀记得曾经听说过这个。

Like, I vaguely remember hearing about it.

但当我重新审视时，我觉得这其实就是价格弹性。

But it and I looked at this again, I thought, but this is just price elasticity.

你真正谈论的不过就是这个。

That's all you're really talking about.

如果你让某件事变得更便宜、更容易，你可能会用更少的钱做同样的事，或者用同样的钱做更多的事，又或者如果你的回报率完全不同，你可能会投入更多资金去做更多事——这正是你在金融服务领域所看到的情况。

If you make it cheaper and easier to do something, you might do the same thing for less money or you might do more for the same amount of money or you might do do more with more money if you have a completely different ROI, which is exactly what you see in, like, financial services.

电子表格并没有导致金融行业从业人员数量的减少。

Spreadsheets did not result in a collapse in the number of people working in finance.

恰恰相反。

They quite the opposite.

现在因为可以做以前无法做到的大量新事情，金融行业的人反而更多了。

You have way more people in finance because now it's possible to do all this more all this new stuff that you couldn't have done before.

我以前呢，你知道的，如果做一次DCF要花一周时间，那你能做一次就不错了。

Me back when, you know, if it took you a week to do a DCF, like, many do you did a DCF, that was it.

你就做完了。

You were done.

如果花十分钟就能做二十个DCF，那你就会做多得多的DCF。

If it, you know, if it takes you ten minutes to do 20 DCFs, then you do way more DCFs.

所以，当然，这正是我想说的，关于采用过程的不连续性与模型本身行为的不连续性之间的区别。

And so then the of course, you know, it's kind of my point about the jagginess of adoption versus the jagginess of the what the models do.

但几乎成了一个陈词滥调，人们总在谈论马克·安德森说的‘软件正在吞噬世界’，比如优步和爱彼迎。

The the but of almost like a cliche that people always talk about, you know, the the Mark Andreessen software is eating the world thing of Uber and Airbnb.

优步并没有向出租车公司销售软件。

That Uber didn't sell software to taxi companies.

爱彼迎也没有向酒店销售软件。

Airbnb doesn't sell software to hotels.

它们改变了‘酒店’这个词的含义。

They change what you mean when you say hotel.

好吧。

Fine.

看看市场份额就知道了。

Take a look at the market share.

在很多城市，优步彻底击垮了传统出租车行业，同时还挖掘出了新的市场需求。

And so many cities, Uber basically demolish the taxi business and also unlocks UK demand.

就拿纽约的出租车和优步来说，我记得优步的日均订单量大概是原有黄色出租车的两到三倍，而黄色出租车的订单量则下降了四分之三左右，大致是这样的数字。

So taxis, Uber in New York, I think Uber rides per day is like triple double or triple what yellow cabs were, and yellow cabs are down by three quarters or something, but ballpark numbers.

再看看酒店行业的情况。

Look at hotels.

好的。

Okay.

酒店行业或许还在保持增长，

Hotels have maybe hotels are still growing.

只是增长速度可能慢了一点。

Maybe they grew a bit slower.

也说不定不是这样呢。

Maybe not.

也许吧，也许不是。

Maybe maybe not.

还有Airbnb，但主要是补充性的。

And Airbnb, but was mostly additive.

如果你深入探究，答案是：我记得曾在社交媒体上看到有人说，本尼迪克特的问题在于他总说答案是‘视情况而定’。

And you dig into that and the answer is why somebody I remember I saw saw somebody on social media saying the problem with Benedict is he always says the answer is it it depends.

干得漂亮。

It's like, well done.

谢谢你的关注。

Thank you for paying attention.

是的。

Yes.

视情况而定。

It depends.

那么，软件会彻底改变酒店和出租车行业吗？

Well, is software going to completely change hotels and taxis?

怎么会是呢。

How yes.

多少呢？

How much?

嗯，这些完全是两码事。

Well, those are completely different things.

是的。

Yep.

比如我未婚妻去美国中西部某个城市出差，晚上九点才抵达，第二天早上八点就有客户会议，她需要健身房、客房服务、冰箱和浴缸。

So like my fiance goes on, you know, business trip to like some Midwest American city and she lands at 09:00 at night and she's got a client meeting at 08:00 the next morning and she wants a gym and, you know, room service and a fridge and a bath.

她肯定不会选择住Airbnb。

And, you know, she's not gonna go and stay in an Airbnb.

她绝对不可能去住Airbnb，完全没有这种可能性。

Like, absolutely zero negative possibility that she's gonna go and stay in an Airbnb.

而商务旅行占了整个旅行和酒店业务的一半。

And travel is business travel is half the travel business, hotel business.

可以随意多举一些例子。

Proliferate examples as often as you like.

为什么互联网对消费电子产品销售的影响比对高端时尚销售的影响更大？

Why did the Internet have a bigger impact on selling consumer electronics than selling high fashion?

嗯，这要看情况。

Well, it it depends.

所以，这就是我对人们试图给职业打分、评估AI暴露程度这个问题的疑虑——你大致方向上可能是对的，比如，直觉上确实觉得某个职业比另一个职业更容易受到AI影响。

So this is the problem I have with people kind of trying to score professions, AI exposure, because you're kind of directionally right, probably, like, yes, it feels intuitively like that profession is more exposed to AI than this profession.

我们讨论的是GDP估值或类似的标准。

We're talking about GDP val or that kind of benchmarks.

所有这些因素，大概方向上都是对的。

All of that kind of stuff, it's sort of probably directionally right.

但这种方向上的正确性，就像你在1997年对互联网所做的分析一样，也只是方向正确。

But it's directionally right in the same way that an analysis you did in 1997 about the Internet would have been directionally right.

我的意思是，其中大部分内容可能差不多都是对的。

I mean, most of it would probably be more or less true.

你想把所有数字都去掉，因为你可以告诉自己，这个是96.5，那个是78。

You wanna take all the numbers off because it's it could tell yourself that that one is 96.5 and that one is 78.

这简直荒谬。

It's just ludicrous.

但你不可能通过这种分析想到Uber。

But you would not have got Uber from that analysis.

你会说，出租车司机。

You would have said, well, taxi drivers.

那互联网会如何改变这一点呢？

Well, how would the Internet change that?

显然，这根本不会受影响。

Obviously, won't touch that one at all.

你会说报纸。

You would have said newspapers.

也许吧

Maybe

嗯，回头来看，我们确实又回到了这个问题，但当时报纸行业看到这种情况，心想：这简直太棒了。

Well, there's a we even in again, in hindsight, but at the time, newspapers looked to this and thought, well, you know, this is gonna be great.

我的意思是，还记得美国在线收购时代华纳的交易吗？

I mean, remember the AOL Time Warner deal?

他们为什么让美国在线收购时代华纳？

What did they why did the AOL buy Time Warner?

就像一堆现在已经不复存在的杂志一样。

Like, a whole bunch of magazines that don't exist anymore.

我的意思是，时代华纳也是，但人们根本没意识到互联网会对传媒行业造成怎样的影响。

I mean, Warner too, but, like, you know, people didn't really understand that what it was that the Internet would do to the media business.

而在其中，互联网对地方报纸的影响，显然和对迪士尼的影响完全不同。

And then within that, clearly, impact on, like, regional newspapers was completely different to the impact on Disney.

迪士尼倒是没事。

Disney's fine.

地方报纸却消失了。

Regional newspapers disappeared.

所以你直接介入了，事后回头看，你可以说，真正发生的是你把实体资产和底层产品分离开了。

And so you go right in, and in hindsight, you can say, well, what really happened was you unbundled physical assets from the underlying product.

如果你的竞争力依赖于拥有实体资产，而这个实体资产突然变得无关紧要，那么你的整个商业模式就彻底崩塌了。

And if your defensibility was based on owning a physical asset and that physical asset suddenly stopped mattering mattering, then your whole business business model has just exploded.

你可以用这个视角来看 Airbnb 和 Uber，但当然，这两家公司的结果却大不相同。

And you could apply that lens to Airbnb and also to Uber, but then, of course, those two turned out very differently.

我认为另一方面，如果你把 Uber 的例子延伸到现在，你可以做这样的评估，比如说，健身教练们其实还好。

I think the other the other side to this, I mean, you you make the the pull the Uber example through to today, you know, you can do that evaluation and say, well, like fitness instructors are fine.

好的。

Okay.

你有没有看过？我把手机放在房间里，对准自己，打开带摄像头的 AI。

Have you seen so I put my phone in my room and pointed at me and turn on the AI with the camera.

我为什么需要，比如说，我不知道。

Why do I need, like I don't know.

也许那样行不通。

Maybe maybe that won't work.

也许会的。

Maybe it will.

但你无法在如此细粒度的层面知道这些事情。

But you can't you can't know those things at that level of granularity.

我认为这正是问题所在。

I think that's kind of the problem.

你能做的只是提出一些思想实验，比如说，这里有一种方法，你可以用它来审视这个领域，并问自己：这个领域是否存在一个问题？

What you can do is just kind of point to thought experiments, I suppose, and say, well, here is a way of of here's a test you can apply to look at this field and ask yourself, well, is that does that field have a question?

这个领域存在问题吗？

Does that field have a problem?

但你是否认为，类似于互联网发生的情况，我们需要先经历一段破坏阶段，才能弄清楚这一点？

But do you think that similarly to a lot

了互联网发生的事情，我们需要

of what happened with the Internet, we need to

先经历一段破坏阶段，才能弄清楚这一点？

go through a phase of destruction first before we figure that out?

那么，这些公司中有许多会倒闭吗？

Well, will a lot of these companies go bust?

是的。

Yes.

你知道，我想我上次可能提到过，这本关于泡沫历史的经典著作名为《这次不同了》，这个书名有双重含义：一方面，在泡沫中，人们会说，这次不一样了。

You know, there's I think I may have mentioned last time, this is a classic book on the history of bubbles, and the title is This Time It's Different, which has a double meaning because, a, in a bubble, people say, it's no.

这次不一样。

It's different.

这不是泡沫，但他们错了。

It's not a bubble, and they're wrong.

但另一方面，当他们说这次不一样时，某种程度上他们又是对的。

But also that when they say it's different, they're kind of right.

比如，互联网泡沫与以往的每一次泡沫都不同，而如今发生的事情显然也与互联网泡沫不同。

Like the .com bubble was different from every previous bubble and what's going on now clearly is different from the .com bubble.

比如，我们现在没有大量没有盈利的消费公司上市。

Like, we don't have loads of IPOs of consumer companies with no profits.

我们根本没有IPO。

Like, we don't have any IPOs.

这并不是由散户在公开或私人市场股票中的投机行为推动的。

It's not being driven by retail speculation in private in in public market stocks.

除了其他因素外，它也不是由风险投资资助的。

It's not being funded by venture capital apart from anything else.

但它仍然可能是一个泡沫，这只是个无关紧要的观察。

But it can still be a bubble, just as an uninteresting observation.

我认为，你知道，我们是不是有太多过度投资？

I think the the you know, do we have a lot of over investment?

是的，当然有。

Yes, of course.

这些投资中有一部分最终会不会没有回报？

Will a bunch of this end up not producing a return?

是的，当然会。

Like, yes, of course.

这就像这些东西的运作方式。

That's like how this stuff works.

地面上的焦痕在哪里？

Where were the smoking holes in the ground?

要更难说清楚一些。

A bit more difficult to tell.

我的意思是，你知道，有些人显然可以看看近处的乌云，或者看看甲骨文，说杠杆往往不会有好结果。

I mean, you know, there are people who have, you know obviously, you can kind of, you know, look at the near clouds or look at Oracle and say, you know, leverage doesn't tend to end well.

但我已经二十年没当过股票分析师了。

But, you know, hey, I haven't been an equity analyst for twenty years.

我不确定。

I don't know.

这里没有任何投资建议。

No investment advice here.

但你知道，你只需确定地回过头去看看互联网和移动时代，列出所有没成功的东西。

But, you know, you just deterministically, you should you need to go back and look at the Internet and look at mobile and make a list of all the stuff that didn't work.

你知道的，所有那些曾经备受关注、令人兴奋、酷炫有趣的东西，那些缩写、公司、概念创意，最后都没成功。

You know, all the stuff that was a big deal and really exciting and cool and interesting, all the acronyms and companies, concept ideas that didn't work.

当然，现在人们正在研究的很多东西也最终会失败。

Well, of course, a bunch of stuff that people are working on now is going to not work.

这本来就是世界的运行方式。

Just that's just how the world works.

这就是创新的运作方式。

That's how innovation works.

会有很多创造性的尝试，但其中大部分最终都不会成为主流。

There'll be a whole bunch of creative creation and a lot of it won't end up being the thing.

我认为亨特·沃克曾经说过，硅谷的钱就像那种小橡胶弹跳球。

I think Hunter Walk once said that money in Silicon Valley is like one of those little rubber ball bouncing balls.

你把它扔进房间，它就会弹来弹去，总会有一些人因为球恰好砸到他们而发财。

Like, you throw it into the room and it just and, like, there'll be a bunch of people to get rich for, like, because the ball hit them at the right time.

就像那个在Meta收购WhatsApp前两天加入的人。

It's like, you know, the person who joined WhatsApp two days before Meta bought it.

干得好。

Well done.

不错的人。

Nice guy.

但是

But

为了深入分析一下，你提到我们现在是97分。

And to unpack some of this, so you mentioned we're 97.

所以我们在理解这件事上是97分，但我们在预测接下来会有99分出现这件事上，也同样是97分吗？

So we're 97 in terms of figuring this out, but are we also 97 in terms of, like, oh, there's 99 coming, you think?

你知道的，你没法预测泡沫。

You know, you know, you can't call bubbles.

这就像一个笑话，说的是那位经济学家，他成功预测了过去五次衰退中的十次。

You know, it's a joke about the economist who successfully predicted, you know, 10 of the last five recessions.

是的。

Yep.

你知道，你无法预测时机。

You know, you can't call the timing.

如果你能预测，那我们就在另一个宇宙了。

If you could, like, we'd be in a different universe.

如果现在这不是一个泡沫，那它迟早会成为。

If this is not a bubble now, it will be.

这些公司中的一些最终会出问题，你可以大致指出一些可能发生的地方，当然，很多人都知道人们感到不安的地方，但你不知道具体是哪一个，也不知道时机。

Some of these companies will end up, like and and you can kind of point to some of the places where that might happen and, obviously, that, you know, many people know where the where people are kind of nervous, but you don't know which and you don't know the timing.

是的。

Yep.

你所知道的是，你面对的是一项极具影响力的技术，而这正是驱动所有这些超额收益和所有这些不确定性的根源。

What you do know is you've got this kind of hugely consequential technology, and that's what drives this all of all of this alpha and all of this uncertainty.

很好。

Great.

回到大型科技公司，比如你刚刚提到的甲骨文，这难道不既是一种赌博，也是一种合理的策略，让甲骨文这样的公司利用其传统业务转型为新事物吗？

Going back to Big Tech, perhaps with with Oracle that you just mentioned, isn't that kind of both a gamble, also kind of rationale for a company like Oracle to be leveraging an old business to turn it into something new?

呃，你懂的，总体来说所有人，最重要的是，所有人都在做理性的决策。

Well, you know, everyone in in in generally and above all, everybody's a rational actor.

几乎所有人都会根据自身的处境做出理性的选择。

Almost everyone's a rational actor given their situation.

如果你是山姆·奥尔特曼，你手里的技术就已经沦为普通商品了。

If you're Sam Altman, you've got a commodity technology.

你要和那些手握庞大 legacy 现金流的对手竞争。

You've got you're competing with people who have giant legacy cash flows.

你没有自己的基础设施，也没有真正的差异化优势，却拥有极高的用户认知度。

You don't have your own infrastructure, don't really have any differentiation, but you've got massive mindshare.

那你会怎么做呢？

So what do you do?

那你就会想方设法把这种认知度兑换成硬资产，还会努力通过造势让预言自我成真。

Well, you try and swap that for hard assets And you try and talk your way into a self fulfilling prophecy.

你会努力把这种认知度兑换成硬资产，想方设法让那9亿活跃度不高的周活用户转化为更具实际价值的资产。

You try and swap that for hard assets and you try and turn those lightly engaged 900,000,000 weekly active users into something something more tangible.

如果你是拉里·埃里森，你就拥有一个已经持续衰退二十五年的、能产生大量现金的遗留业务。

If you are Larry Ellison, like, you've got this very cash generative legacy business that's been in structural decline for twenty five years.

大多数参加YCombinator的人根本没听说过甲骨文，几乎是字面意义上的没听过。

Most people going through YC have never heard of Oracle, like almost literally.

自1998年以来，就没人邀请你参加过任何派对。

No one has invited you to a party since like 1998.

那么，当你面对这样一个

So what do you do when here is this

机会时，你会怎么做？

through?

英伟达。

NVIDIA.

你用双手紧紧抓住它，一路突破重围。

You grab onto it with both hands and you burn your way through.

英伟达也是如此。

The same with NVIDIA.

我的意思是，我还没查看过，也没更新过这里的数字，但我觉得去年第三季度，英伟达的十二个月自由现金流已经超过700亿美元了。

I mean, I haven't looked at I haven't updated my number here, but, like, I think q three last year, I think NVIDIA had something over $70,000,000,000 of trading twelve month free cash flow.

所以他们给台积电送钱都来不及。

So they can't give the money to TSMC fast enough.

台积电收钱都收不过来。

TSMC won't take it fast enough.

台积电说，哥们，这可是个周期性行业。

And TSMC is it says, like, dude, this is a cyclical industry.

回家吧。

Go home.

不行。

No.

我们今年不会把产能扩大三倍。

We're not gonna triple our capacity this year.

ASML也来不及发货。

And ASML can't ship the stuff fast enough either.

那么，你会怎么处理这708亿、900亿美元呢？

So what would you do with that $70.80, $90,000,000,000?

你可以把它用在两笔投资上，或者用于建设基础设施、提升市场地位、扩大市场份额、壮大你的生态系统。

You know, put it in two bills or put it into building infrastructure, building market position, building market share, building up your ecosystem.

包括循环交易吗？

Including in circular deals?

这确实是供应商行为。

It is vendor yeah.

你知道吗，我年纪够大，记得这曾经被称为供应商融资。

You know, I'm old enough to remember this being called this is when this is called vendor financing.

对吧？

Right?

供应商融资，只要你披露清楚，不隐瞒资金的真实去向，原则上并没有什么问题。

Vendor financing is, you know, it's a it's as long as you're disclosing it and you're not lying about where the money's going from, there's nothing wrong with that in principle.

但这是一种杠杆。

But it's leverage.

而且，你知道，当一切都在上涨时，杠杆总是没问题的。

And, you know, leverage is always fine when stuff is going up.

目前我们有着大量不同类型的杠杆，无论是特殊目的实体，还是循环收入，或者所有这些类似的东西。

And we've got an awful lot of different kinds of leverage going on at the moment, whether that's SPVs or, you know, circular revenue or all this kind of stuff.

这全是杠杆。

It's all leverage.

我的意思是，这总能奏效，直到它不再有效。

I mean, that always works until it stops working.

当它不再有效时，你就遇到问题了。

And then when it stops working, then you've got a problem.

同样地，从互联网泡沫和崩盘与今天的比较来看，你是否从这一切最终对经济有利的股息概念中获得了一些安慰？

Still in the same vein of comparison between the .com bubble and .com crash and and today, do you take some comfort in the concept of a piece dividends from all of this that ends up working out for the economy?

这是与每一次其他平台转型截然不同的一点。

One of the really basic ways this is different from every other platform shift.

我花了很多时间说，嗯，这就是平台转型的工作方式，以及过去五次发生了什么。

A lot of I spent a lot of time saying, well, this is kind of how platform shifts work, and this is what happened the last five times.

这一点与其他所有平台变革截然不同的是，过去我们清楚地知道物理限制在哪里。

The one way this is unquestionably different is that with all the other platform shifts, knew what the physical limits of the silence were.

我们知道它是如何运作的。

We knew how it worked.

我们知道接下来一个月可能发生什么。

We knew it could happen what could happen, like, next month.

你可能不知道iPhone三或四代会是什么样子，但你知道它不会飞，也不会有一年的电池寿命。

So you didn't know what, like, the iPhone three or four would be, but you knew it wouldn't fly and it wouldn't have a one year battery life.

你确实不知道1997年互联网会如何发展，但你知道电信公司不可能在下周就给全世界每个人提供光纤互联网。

You didn't really know how the Internet would evolve in 1997, but you knew that telcos wouldn't give everybody on the world fiber Internet next week.

而对于大语言模型，我们并不清楚它可能演进的物理极限在哪里。

Whereas with LLMs, we kind of don't know the physical limits of how this could evolve.

因此，也许下周就会有一篇论文表明，你只需1%的算力就能获得差不多的结果。

And so it may be that, you know, next week we have a paper that means that you can get more or less the same results for 1% of compute.

你知道，这可能是个荒谬的说法，但也可能是10%的算力，而我们并不知道。

You know, maybe that's a silly statement, but then but it might be 10 of the compute, and we don't know that.

而你当时根本一无所知。

Whereas you did know absolutely no.

没有人会发表论文说，可以用芯片上1%的晶体管实现同等的计算能力。

No one was gonna publish a paper that said you could get, like, the same compute with 1% of the transistors on a chip.

但我们并不了解这些物理限制，因此也无法预测可能导致价格崩溃的可能与不可能情况。

But we don't know those physical limits, and so we don't know the parameters of what could and couldn't happen to cause a pricing collapse.

到目前为止，发生的主要情况恰恰相反：我们不断发明出使用10倍更多token的方法。

What's happened mostly so far is the other way is we keep inventing ways of using 10 x more tokens.

于是你有了推理模型，接着又有了智能代理。

So you get reasoning models and then you get agents.

这听起来很棒。

It's like, great.

我们现在使用的token量已经是以前的100倍了。

We're using a 100 times more tokens now.

有趣的是，预测token的使用量，对我来说，就像回到九十年代末预测带宽消耗一样。

The funny thing is, like, predicting token usage, to me, it's like it's like looking again, it's like being in the late nineties and looking at bandwidth consumption.

你知道，这其实非常对称，你拥有的就是这样。

You know, it's a very I mean, it's symmetric that you've got.

但想象一下，那是2003年，我们说YouTube的带宽使用量每个月翻一番。

But, you know, imagine it's like 2,003 and we're saying, you know, YouTube bandwidth use is doubling every month.

好吧。

Okay.

嗯，这听起来还不错，我想。

Well, that sounds good, I guess.

但背后有五六个不同的乘数在推动这一增长，而你根本不清楚它们具体是什么。

But there's like five or six different multipliers that are producing that and you don't really know what any of them are.

AI的资本支出现在也是同样的情况。

The same thing now with AI CapEx.

所以，使用量在上升，使用量还在持续增长，而且还有多个乘数在推高使用量，比如更多人使用、人们使用得更频繁。

So, like, the usage is going up and the usage is going up by and there's multipliers that drive the usage up, which is more people using it, people using it more.

人们进行推理、处理视频和图像、使用智能代理、企业应用、人们用它来编程、全天候运行。

People doing reasoning, people doing video and imaging, people using agents, corporations using this, people using it for coding, people using it running all day.

那你说tokens（令牌量）在增长是什么意思？

So what do you mean by tokens are going up?

然后还有效率提升的因素。

And then you've got the efficiency gain.

我记得萨提亚·纳德拉说过，推理成本每三个月就会减半。

So I think Satya said Satya Nadella said that, like, we inference cost halves every three months.

好的。

K.

没错。

Great.

所以这些因素会拉低成本。

So that's pushing it down.

但紧接着你就要为下一代模型投入资源，而每一代新模型的规模都更大，成本也更高。

But then you're chasing the next model and the next model is always bigger and more expensive.

而且如果新一代模型问世，你现在持有的模型不出六个月就会被淘汰。

And, like, your model that you have is gonna be irrelevant in six months if the next model comes.

那么，你要追多久的前沿技术？

So how long are you chasing the frontier?

同时，今年——我们一开始就说过了——但今年，Meta表示将把超过50%的营收用于资本支出。

And meanwhile, like this year and we've said this at the beginning, but like this year, Meta will spend over says it will spend over 50% of revenue on on CapEx.

不是50%的利润或现金流，而是50%的营收用于资本支出。

Not 50% of profits or cash 50% of revenue on CapEx.

所以谷歌和微软也不会落后太多。

So Google and Microsoft are not far behind that.

但他们不可能再把这数字翻倍。

But they can't double that again.

他们不可能明年把100%的营收都花在资本支出上。

They can't spend a 100% of revenue on CapEx next year.

他们已经出去借钱了，而且还有很多租赁方面的支出。

And they've already gone out and started borrowing money, and then there's a whole bunch of stuff in leases.

所以，如果你把租赁和其他相关支出都算上，实际的资本支出数字比那个还要高出好几个百分点。

And so, like, the actual CapEx number is double digit percentage numbers higher than that if you look at the leases and and and the and and all of that kind of stuff.

所以在某个时刻，会出现一种财务上的引力点，那就是它们无法再以那样的速度持续增长。

So there's a sort of financial gravity point at a certain point, which is they can't keep increasing at that rate.

比如，明年不可能达到每年两万亿美元的支出，你说是吧。

Like, it can't be it can't be $2,000,000,000,000 a year next year, you say.

这可能是我们唯一真正的财务限制，一个实实在在的硬性极限。

That's maybe that's the only, like, financial limit, like, actual hard limit we have.

因此，你实际上面临一个边界，那就是你能持续建设这些东西多久。

And so you kind of got this envelope of how long can you keep building this stuff out.

难道只是我们在建造工厂，然后最终会回到我之前提到的寡头垄断问题吗？

Is it just that we're building the factory and you're going to get to a point it's back to my point about oligopolies we talked about a while ago.

我们会不会最终进入一种状态，只有四家公司，每家每年花费2500亿到5000亿美元来建设和维护这些东西？

Like, are we going to wake are we going to get to an end state where there are, say, four companies that each spend 250 billions, $500,000,000,000 a year each on building and maintaining this stuff.

还有，这些芯片的使用寿命有多长？

How also the what's the lifespan of the chips?

你是觉得会是这样，还是会变成那样？

Like, you know, is is it gonna look like this or is it gonna do like that?

我们最终会达到一个点，整个行业每年在基础设施上的投入达到一到两万亿美元，但几乎只获得边际成本回报。

And we get to get to a point where the industry collectively spends 1 to $2,000,000,000,000 a year on infrastructure every year and basically makes marginal cost on that.

嗯。

Mhmm.

那之后会发生什么？

Then what happens on top?

然后你就进入了这些非常模糊的总可用市场（TAM）问题，比如萨姆·阿尔特曼和黄仁勋会说，TAM就是全球GDP。

And then you get to these kind of very vague kind of TAM questions where, you know, Sam Altman and Jensen say, like, the TAM is global GDP.

不对。

And no.

不对。

No.

不对。

No.

这甚至超过了全球GDP。

It's more than global GDP.

全球GBT是什么？

Global GBT is what?

7080万亿美元？

$7,080,000,000,000,000 dollars?

差不多是这个数。

Something like that.

是的。

Yep.

所以我们将会让GDP翻倍。

So we're gonna double GDP.

比如，其中的5%归萨姆，另外5%归詹森，还有一些会归到SML。

Like, and we're 5% of that will go to Sam, and the other 5% will go to Jensen, and some of it will go to SML.

太好了。

Great.

谢谢。

Thank you.

这些数字中存在一个很大的谬误，其中很大一部分显然不只是软件的总可用市场，而是整个经济的服务部分。

This like a big fallacy that's built into those numbers, which, a lot of it is obviously, it's not just the software TAM, but it's the entire economy's services.

我认为，把AI的收费和人类员工的工资定在一个水平上，这在我看来并不合理。

And I think this idea that, you could be charging for AI at the same price as you would charge for a human worker, like, doesn't seem right to me.

不。

No.

我的一个微观观察是，这又回到了GDP评估的问题，比如那位纽约金融教授试图通过将出租车市场的总可用市场作为Uber的总可用市场来估算Uber的价值，但这是不对的。

I mean, a sort of a micro observation is that, you know, it's back to the the GDP eval thing of, like, the the finance New York finance professor who tried to estimate the TAM for Uber by saying he said he'd value Uber because this is the TAM for taxis and say that's Uber's TAM, which is no.

那不是Uber的总可用市场。

That's not Uber's TAM.

Uber的总可用市场完全是另一回事。

Uber's TAM is something completely different.

你得考虑到广告，数字广告的平均总可用市场必须包括零售租金，还必须包括物流费用。

You get to something like advertising, the average the TAM for digital advertising has to include retail rents, and and it has to include shipping.

它基本上还必须包括零售商的利润，而零售商的利润占零售额的50%。

And it basically has to include retailer margin, is 50% of retail.

所以数字广告的总可用市场并不是广告本身。

So the TAM for advertising for digital advertising is not advertising.

它是别的东西。

It's something else.

而且，是的，你知道，GDP增长就是GDP增长。

And, yes, you know, GDP growth is GDP growth.

你得非常天真才会认为，我们会从目前的低个位数GDP增长跃升到两位数的GDP增长。

You would have to be quite wild eyed to think that, you know, we're gonna go from whatever it is, low single digit GDP growth to double digits GDP growth.

那完全是另一个话题了。

That would be, like, different conversation.

关于人力的问题，你看，回到工业革命时期的学术讨论，自动化只会发生在劳动力昂贵的地方。

The human labor thing, I mean, look, you know, go right back to, like, academic discussions of the industrial revolution, you know, automation happens in places where labor is expensive.

因为如果劳动力便宜，干嘛要费这个劲？

Because if labor is cheap, why bother?

几周前我读了一本书，讲的是上世纪三十年代意大利南部的情况。

I was reading a piece a book a couple of weeks ago about Rome about Italy in the thirties in the South of somebody in the South Of Italy in the thirties.

整个省好像只有一辆车。

It's like there's one car in the whole province.

所以，你为什么不干脆弄辆车来做这件事呢？

So, well, why didn't you, like, you get a car to do that?

因为人们是自由的。

Well, because people are free.

你跟一个人，跟一个农民说，他们还是农民。

You just tell a person you tell a peasant and say, they're still peasants.

你让一个农民走八个小时到那个镇上取东西，再走八个小时回来，他们就会照做。

You tell a peasant to walk at eight hours to that town and get something and walk eight hours back and they do it.

而且你甚至都不用付他们钱。

And you don't even have to pay them.

那为什么要买车呢？

So why would you buy a car?

所以，这就像经济学大一学生对这个问题的最基本回答：自动化必须相对于人力有投资回报率。

So that's like the very basic first year economic student answer to this is, you know, the automation has to have an ROI relative to a person.

然后你回到价格弹性的观点。

Then you get back to the price elasticity point.

对我来说，再强调一下，准备一张有三个要点的幻灯片。

And to me, I mean, again, have a slide with three bullet points.

我可能会说，本尼迪克特，你习惯用幻灯片表达，这大概没错。

And I might say, Benedict, you talk in slides, which is is probably true.

用幻灯片思考。

Think in slides.

所以第一步是：如果你让某件事变得更便宜、更容易，你是用更少的钱做同样的事，还是用更多的钱做更多的事，或者服务更多的人？

And so step one is, okay, if you make it cheaper and easier to do a given thing, do you do the same thing for less money, or do you do more for more money or more people?

第二步是：拥有一整栋楼的人来做这件事，是你的护城河吗？

Step two is, was having a building full of people doing that your moat?

因此，这里人们常举的例子是医疗保险。

So that the example people kick around here is health insurance.

我认为有人说过，争论的边际成本现在是零。

I think somebody said the marginal cost of arguing is now zero.

所以，如果你的业务建立在让你所做的事情变得极其痛苦和困难的基础上，而突然间它不再痛苦和困难了，那这还是你的进入壁垒吗？

So if your business is based on making it really painful and difficult to do what you do, and suddenly it's not painful and difficult, was that your barrier to entry?

但对我来说，更有趣的是第三步，那就是有一类事情你根本就无法做到。

But then to me, the interesting is a sort of the third step, which is there's a whole class of stuff that you just couldn't do at all.

你知道，回到杰文斯悖论和蒸汽机的话题。

You know, go back to talking about the Jevons Paradox and steam engines.

想象一下，在1800年，你想从伦敦到苏格兰开一列特快列车。

Imagine if you want to make an express train from, you know, London to Scotland in 1800.

别提你没有钢铁，无法大规模生产铁材，也无法铺设铁轨了。

Like, never mind that you didn't have steel and, you know, iron, couldn't mass produce iron, you couldn't have laid the rails.

但假设你已经能建好整列火车，只是没有蒸汽机。

But presume you could have built the whole thing, but you didn't have a steam engine.

你不可能买上一万匹马，让它们在前面拉火车去苏格兰。

You couldn't, like, buy 10,000 horses and put them on the front and have the horses pull your train to Scotland.

这根本行不通。

That you just it didn't matter.

不管你有多少匹马，这都没用。

That doesn't matter how many horses you have.

你就是做不到。

You can't you just can't do that.

但现在有很多业务，不管你雇多少人，都根本做不到。

But, you know, there's there's whole businesses now where you could not do that with people no matter how many people you had.

如果你用一万人代替计算机，根本不可能运营一个量化基金。

But you couldn't run a quant fund if you had 10,000 people instead of a computer.

即使你能支付那一万人的工资，你也根本做不到。

Even even if you could pay the 10,000 people, you just couldn't do it.

是的。

Mhmm.

多年前，我觉得有个人——我记不清是谁了——说过，当某样东西变得非常便宜，和当它变得真正免费时，是完全不同的，而互联网带宽在某个时刻就发生了这种情况。

Years ago, I think somebody, I can't remember who it was, said, you know, there's a huge difference between when something gets very cheap and when something gets actually free, which is what happened with Internet bandwidth at a certain point.

移动带宽也是如此。

It happened with mobile bandwidth.

而且again，这就像在问：你是在用新方法做旧事，还是在用新方法做全新的事？

And again, it's like, are you doing the old thing with a new thing or are you doing something different with a new thing?

这其实是一种有点啰嗦的方式来回应你的问题。

Which is sort of sort of sort of discursive way of of of talking about your question.

但你知道，我们会让人用AI做那些不会取代人类的工作。

But, you know, we're going to have people do stuff with AI that doesn't replace a person.

它完成的是原本需要一百万人才能完成的任务，这就是所有自动化所发生的情况。

It does a thing that would have needed a million people, which is what happens with all automation.

蒸汽机也是如此。

That's what happens with steam engines.

电力也是如此。

It's what happened with electricity.

飞机、钢铁以及所有其他技术的发展也是如此。

What happened with aircraft and and and steel and everything else.

你确实会看到，你不会去做那些仿佛你拥有一百匹马或一千匹马的事情。

You do see you know, you don't do stuff that says it's as though you had a 100 horses or a thousand horses.

你做的是用再多马也无法完成的事情。

You do a thing you couldn't couldn't have done with horses or no matter how many how many horses you had.

所以，是的。

And so Yes.

但这也正是，你知道的，一直以来的情况，而且我不是一位学术经济学家。

Well but that's also, you know, been the you know, and I'm, you know, I'm not an academic economist.

但你知道，有一句著名的说法是：信息革命无处不在，唯独在生产率统计数据中看不到。

But, you know, there's a famous quote that you can see the information revolution everywhere except in the productivity statistics.

于是，人们就开始争论：生产率统计数据里到底包含了什么？

And so then you get these arguments about, well, what's in the productivity statistics?

你知道，从购买几十台价值数千美元的设备，转向购买这种新设备，这会导致GDP下降。

And, you know, this, you know, switching from buying 40 devices that cost thousands of dollars to buying this is a decline in GDP.

在所有这些方面，我们都面临着测量和指标的问题。

We use this sort of there's a measurement problem and a metrics problem in all of those things.

但归根结底，你会用人工智能取代麦肯锡吗？

But, you know, to the core of this, are you going to replace McKinsey with AI?

这就像说，你会用电子表格取代KPMG或普华永道吗？

Well, that's like saying, are you going to replace KPMG or PwC with spreadsheets?

这种想法根本就不太对。

Kind of not the right way of thinking about it.

因为一方面，你会遇到价格弹性的问题。

Because on the one hand, you get that price elasticity.

另一方面，如果你问这个问题，其实就是告诉我你根本不知道麦肯锡到底做什么。

On the other hand, if you ask that question, you kinda just told me you don't actually know really know anything about what McKinsey does.

我犯了个错误，回复了LinkedIn上一个讨论这个问题的人。

I I made the mistake of replying to somebody on LinkedIn who was talking about this.

他们说，但你看，上市公司咨询公司已经崩盘了，你知道的，但它们都是合伙制企业。

And they said, but look, publicly listed consultancies have already crashed, you know, but but but they're all partnerships.

看啊，博思艾伦已经崩盘了。

Look, Booze Allen has crashed.

好吧。

It's like, okay.

你需要我解释一下为什么博思艾伦的股价自特朗普就职以来暴跌吗？

Do you need me to explain why Booz Allen share price Booz Allen Booz's share price has collapsed since Trump was sworn in?

因为这跟人工智能无关。

Because it's not to do with AI.

这是另一个完全不同的话题。

It's a whole other conversation.

我本来想深入研究一下这个会计问题。

I I wanted to do digging into this accounting thing.

过去自动化对会计行业产生了什么影响？

What what did automation do to accounting in the past?

我决定，也许我该去查查审计费用的数据。

I decided it well, maybe I should go and look for audit costs.

结果发现，你可以获取自巴恩扎克泄密事件以来的审计费用平均数据，各种数据都显示，自21世纪初以来基本保持平稳。

And it turns out that either you can get a chart of average order data, all sorts of data for audit costs since the Barnzalk leak, which has basically been flat since about since the early two thousands.

尽管软件领域发生了所有变化，平均审计费用却没有任何改变。

Despite everything that's happened in software, average audit cost hasn't changed.

但后来，我决定再深入一点，查了七十年代和八十年代审计费用的变化情况。

But then then I kind of well, I can dig a bit further and got into what happened to audit costs in the seventies and eighties.

是的。

Yeah.

你找到的都是些长达五十页的学术报告，里面提到了各种影响审计费用和审计工作变化的因素。

And what you get is, you know, all these like 50 page academic reports that mention all sorts of stuff that changed audit costs and what was going on in audit.

但这些报告里没有一个提到计算机。

And then none of them mentioned computers.

有七百件其他事情正在改变审计的方式，但没有一件是真正跟计算机有关的。

There's, like, the 700 other things going on that are changing how audit works, none of which are actually computers.

这让我想到，这种现象在每个行业似乎都存在。

It's sort of and and it occurred to me this sort of applies in every industry.

比如两年前，我们听到了WPP的创始人马丁·索罗尔的演讲，WPP是全球最大的广告网络之一。

So I, you know, was an event, like, two years ago, we're listening to Martin Soul, who founded WPP, which is one of the big global ad networks.

他在谈论人工智能时说：不。

And he was talking about AI, and he says he said, no.

不。

No.

人工智能的主要影响并不是生成更多的广告素材。

The big impact of AI is not making more advertising assets.

人工智能的主要影响是，实际上每家大型广告公司都有整栋楼的人在做极其枯燥无聊的工作，比如处理电子表格、发送传真、把广告上传到Facebook。

The big impact of AI is is that actually every big ad agency has got buildings full of people doing really, really boring crap with spreadsheets and faxes loading the ads into Facebook.

嗯哼。

Mhmm.

几乎是字面意思。

Almost literally.

人工智能对广告业的一个巨大影响在于运营环节，而不是你所不了解的那些内容——你可能会以为它只是能生成更多图片。

One of the huge impacts of AI on AI on advertising is on the operations, not on the stuff that you you know nothing about advertising, you would just think, well, it'll make more pictures.

是的，没错。

Like, well, yes.

但不是这样的。

But no.

关键在于后台那些忙着处理文件的人。

The big thing is all the guys in the back office shuffling paper around.

我认为这一点适用于每个行业：当你从外面看时，你会觉得，当然，AI会做这件事。

And I think that the kind of a point a point applies to every industry that, you you're sitting on the outside, you think, oh, of course, AI will do that.

但如果你身处其中，就会发现，事情在这里，而AI的作用在那里。

And if you're on the inside, you're like, like, over here, and it's over here.

那并不是最难的部分。

And that's not the hard part.

是的。

Yeah.

作为你咨询和公开演讲的一部分，你接触过很多大公司。

And you talk to a lot of big corporations as as part of your consulting and your public speaking.

你对整体情绪有什么看法？

What's your sense of the overall sentiments?

人们现在比以前更困惑了吗？

Are people more bewildered than they used to be?

大家有没有开始逐渐领会到这一点？

Is some of this starting to sink in?

你观察到他们都采取了哪些行动？

What do you see them do?

首先我肯定不会用“困惑”这个词来形容。

So I certainly wouldn't say bewildered.

我的意思是，至少他们不会比科技行业的任何人更困惑。

I mean, at least not any more bewildered than anyone in tech.

而且我觉得，你懂的，一旦...对的。

And if I think, you know, once yeah.

处在赛道中间线的位置，如果你能掌控节奏，那你的前进速度就足够快了。

In the center line, you know, if you're in control, you're going fast enough.

就好比说，要是你觉得自己把这一切都搞懂了，那只能说明你根本没走心。

Like, if you understand any of this, you're not paying attention.

我前几天刚好看到一位很有名的物理学教授说，他要开一门为期八周的量子力学课程。

I was like, was came across the other day that the physic famous physics professor says, you know, I'm gonna teach you an eight week class on quantum.

今天，我不懂量子力学。

Today, I don't understand quantum.

上完这门课后，你也一样不会懂量子力学。

At the end of this class, you won't understand quantum either.

是的。

Yeah.

这种说法当然适用于广告技术，但你知道，这就像是施莱彻-科尔斯塔德问题。

This sort of certainly applies to ad tech, but, you know, it's like the Schleicher Kolstad question.

你知道，好好想想这个问题。

You know, the the the think about that.

哦，施莱彻-科尔斯塔德问题是一个著名的复杂十九世纪政治问题。

Oh, the Schleicher Kolstad question is this famously complicated nineteenth century politics question.

我认为帕尔默斯顿说过，真正理解这个问题的只有三个人：一位德国教授，他已经疯了；一位俾斯麦身边的人，已经去世了；还有我自己，但我已经忘了。

And I think Palmerston said that only three people have ever understood that my the German professor who's gone mad, somebody at Bismarck who is dead, and myself who has forgotten.

总之，你可以把那段删掉。

Anyway, the you can cut that out.

关键是，没人真正知道答案，比如AI和AGI会怎样发展，明年会怎样，但每个人都已经部署了一堆东西。

The point is that, like, no one really knows the answers to, like, does it go to h AI and AGI and what will it do next year and but everyone's got a bunch of stuff deployed.

每个人都部署了Copilot，然后说：哦，好吧。

Everyone deployed Copilot and went, oh, okay.

这并不太成功。

That wasn't very successful.

每个人都在像1997年给所有人发放互联网一样，说：喏，给你了，去提高效率吧。

Everyone is kind of like giving everybody the Internet in 1997 and saying, there you are, be more productive.

这其实没起什么作用。

It doesn't didn't really work.

现在每个人都有一堆试点项目。

Everyone now has a bunch of pilots.

其中一些试点项目已经进入了生产环境。

Some of those pilots have made their way into production.

还有一些没有。

Some of them haven't.