为什么大家对AI的看法都错了（包括你）| Benedict Evans

它在告诉人们什么是被需要的。

It's telling people what's wanted.

它不仅仅是一个价值的信号。

It's not it's it's not just a signal of worth.

它是一个需求的信号。

It's a signal of demand.

我几年前读过一本很有趣的书，叫《红富余》，讲的是上世纪六七十年代苏联的中央计划。

There's a a fascinating book I read a while ago called Red Plenty, which is about Soviet central panning in the sixties, seventies, eighties.

它探讨的是，当中央计划无法应对六七十年代复杂精密经济的规模时会发生什么，而二十、三十年代只需生产粮食、拖拉机、机车和钢铁这类相对简单的东西时，情况则不同。

And it's about sort of what happens when you have a central planning that just cannot cope with the level of complexity of a sophisticated economy in the sixties and seventies as opposed to let's make grain and tractors and locomotives in the twenties and thirties and steel, which already kind of works.

但一旦你拥有了一个复杂的工业经济，中央计划就无法应对这种复杂性。

But once you actually have a sophisticated industrial economy, central planning can't handle the complexity.

于是你们试图在没有价格机制的情况下，建立激励机制和结构来弥补。

And so you try and create incentives and structures around that while not having pricing.

但这根本行不通。

And that just doesn't work.

我想这是一个普遍的观点，即市场经济是一种系统。

I suppose this is sort of a generalized point, which is like, a market economy is a system.

如果你在这里拉动一个杠杆，那里就会有所反应。

And if you pull a lever here, something will move there.

你不能只是在这里拉动一个杠杆，然后说：‘我不希望那样发生，因为这是民主。’

And you can't just pull a lever here and say, well, I don't want that to move because it's a democracy.

它还是会照常发生。

It will move anyway.

你必须理解这个系统是如何运作的，明白你希望从中获得什么后果，以及你在这个系统中的参数是什么。

You have to understand how the system works and understand what consequences you want from that and what your parameter parameters are within this.

我欣赏你的一点是，你以善于发现模式而闻名。

One of the things that I admire about you is that you're sort of known for spotting patterns.

我有一个关于如何学习模式识别的理论，很想听听你的反驳。

I have a theory on how to learn pattern matching, and I'd love to hear your your pushback on this.

我关于学习机制的理论，我称之为学习循环。

My theory on how we learned is I call it the learning loop.

我们有一个体验。

We have an experience.

我们反思这个体验，并对其进行压缩，这种压缩就成了我们的收获。

We reflect on that experience and we create a compression, and that compression becomes our takeaway.

所以我们可以看一部电影、读一本书，然后从中提炼出它的核心要点。

So we can watch a movie, read a book, and you come away with a compression of it.

但你可以从这个压缩后的要点反推回原始体验。

But you can work backwards from that compression to the experience.

但我们平时消费的大多是别人的压缩内容。

But what we consume most of the time is other people's compressions.

比如，当人们阅读你的通讯时，他们消费的是你所做工作的压缩版，而不是原始的、未经加工的内容。

So, like, when people read your newsletter, they're consuming a compression of the work that you've done, but not the actual raw work.

因此，如果你没有在那个领域真正做过相关工作，这种状态就只是一种知识的幻觉。

So in a way, it's an illusion of knowledge if if you haven't done the work in that area.

这很有趣。

It's funny.

我正在起草一些想法，关于大语言模型对网络搜索、出版、发现和电子商务的影响，这些都是一些宽泛、模糊的方面。

I'm I'm I have a a draft thinking about, like, what LLMs do to web search and publishing and discovery and ecommerce and, like, big foot of hand wavy, fuzzy, all of that stuff.

我一直在思考这个问题，二十年前有一位法国学者写过一本书，叫《如何谈论你没读过的书》，听起来有点讽刺。

And I was sort of thinking about this, and there's a book written by a French academic sort of twenty years ago or something called How to Talk About Books You Haven't Read, which sounds very kind of snide.

但他的观点是，比如你17岁时读过的一本书，当时根本没读懂。

But kind of his point is that, like, there's the book you read when you were 17 and you really didn't get it.

嗯。

Mhmm.

如果你现在再读，就能明白了。

And if you read it now, you'd get it.

还有一种书，是别人都读过的，所以你得说你读过，以示融入。

And there's the book that, like, he's got this kind of list of, like, there's the books that everybody else has read, so it's to say you've read them.

还有一种书，是你已经读过这位作者的其他三本书了，所以其实没必要再读这一本。

There's the books that, like, you've read three other books by that writer, so you don't really need to read this one too.

你懂的。

You get it.

比如，你真的需要再读一本马尔科姆·格拉德威尔的书吗？

Like, do you need, you know, do you need to read another Malcolm Gladwell book?

比如，如果你已经体验过马尔科姆·格拉德威尔的风格了。

Like, if you where you've kind of got the Malcolm Gladwell experience.

所以，有一种普遍的感觉，就是你所看到、部分看到、半记得的内容的模式和积累。

So there's this sort of generalized sense of pattern and accumulation of what you've seen, what you've half seen, what you half remember.

我认为，你的观众或听众可能会注意到，我有两三种不同的风格。

There's also, I think, you know, what your your your viewers, listeners might notice is I I kind of have two modes, two or three modes.

一种是较为发散、略显啰嗦、自由联想的风格，我会不时跑题，但最终 hopefully 会回到主题。

I have a mode that's sort of discursive and a slightly rambling and free associating, and I'll kind of spiral off in different directions and hopefully come back to the point.

还有一种风格，我想把事情搞清楚，拆解开来，看看这里究竟发生了哪两三个、四件事。

And then there's a mode where I want to try and pin the thing down and break it apart and say, are the two, three, four things that are happening here?

这正是你在幻灯片中看到的。

Which is what you see in the slides.

没有，比如，那到底是什么？

There's no, like, what is it?

就是这个，然后是这个，接着是那个。

It's this and then this and then that.

这样能概括出来吗？

And does that capture that?

这是一种试图理解这是什么的方式。

That's a way of trying to understand what this is.

我试图弄清楚我对这件事的想法、我的理解方式，以及如何通过明确界定来拆解它。

I try and work out what I think about this, how I understand it, how you can break it apart by kind of pinning it down.

关于数据，关于幻灯片和分析，我总是在问：谁在乎？

The thing about data is, and the thing about the slides and analysis is is, like, I'm always asking who cares?

而且

And

我总是在问：是的，但这里真正重要的是什么？

I'm always asking, yes, but what actually matters here?

你为什么要给我看这张幻灯片？

Why are you showing me this slide?

我为什么要给你看这个图表？

Why am I showing you this chart?

所以你得问，到底哪些才是真正的问题？

And so you kind of have to ask, like, well, what are the actual questions?

我们在人工智能领域没有问，但应该问的问题有哪些？

What are the questions we're not asking on AI that we should be asking?

我的意思是，我们确实问了，好吧。

I mean, we're asking okay.

有些人说，我们所有的缓存都会集中在模型里。

Well, there are some people who are saying all of our cache is gonna be in the in the in the in the models.

这种观点存在一种有趣的分歧：一部分人只关注模型变得越来越好，而另一些人则认为，所有价值都将集中在应用层，那么哪些公司值得关注？我们应该投资Cursor，投资他们的一切，但为什么还没有出现面向消费者的突破性产品？

This is kind of funny split between people who are just talking about the models getting better and everybody else who's saying, well, all the value is going to be in the application layer, and, you know, what are the companies, and let's fund Cursor, and let's fund all of their stuff, and why isn't there a consumer breakout yet?

还有人说，你说什么？还没有出现面向消费者的突破？

And other people are saying, what do you mean there isn't a consumer breakout?

人人都在用ChatGPT，对此我的回应是：其实并没有，这就是我的观点。

Everyone's using ChatGPT, to which the answer is, well, not exactly, which is my data point.

有些人正在使用ChatGPT。

Some people are using ChatGPT.

大多数人看了却看不懂。

Most people look at it and don't get it.

但还是很迷人。

Still, just fascinating.

这里有一个核心问题：价值在哪里捕获？

There's this kind of core where's the value capture question.

然后还有一些问题，我们两年前本该问的，但现在依然没有答案。

Then there's, like, a bunch of questions that we could have asked two years ago where we don't have an answer.

错误率是否能被控制或管理？

Will the error rate ever be controllable or manageable?

你能否让模型知道它什么时候错了？

Will you ever get to a mold that knows when it's wrong?

这对我来说似乎是个矛盾，因为统计系统本质上就是如此，但也许你可以列出十几个人们在2023年初本该问的问题。

Which to me seems just like, given the statistics statistical system seems like a contradiction in terms, but maybe it was a bunch of code you could make a list of, like, a dozen questions we could have asked in early twenty three.

我们实际上对这些问题都没有答案。

We don't really have answers to any of those.

我的意思是，曾经有些人问过，这些东西是商品吗？

I mean, there were some people who were asking, are these things commodities?

中国会赶上来吗？

Will China catch up?

即使当时我们回答了，那显然是肯定的，事实也的确如此，DeepSeek 就证明了这一点。

Which we answer even then was obviously yes, of course, which is what happened, which Deepsea kind of demonstrated.

但从那以后，我们并没有提出太多新问题。

But we don't have that many new questions since then.

我现在困惑的是，首先，正如我所说，存在一个关于什么是 LLM 的 L 和 SCA 的整体关联。

The thing that I puzzle about right now is, first of all, there's this whole nexus, as I said, of, like, what is l l what is SCA for LLMs?

你知道，我们有无限的产品、无限的零售、无限的媒体。

You know, we have infinite product, infinite retail, infinite media.

你该如何选择买什么？

How will you choose what to buy?

如果我问一个大语言模型：我该买哪张床垫，会发生什么？

What happens if I go to an LLM and say, what mattress should I buy?

我该买哪种人寿保险？

What life insurance what life insurance should I get?

这又是怎么运作的？

How does that work?

这带来了数十个我们目前尚不清楚的问题。

That that poses dozens of questions we don't know yet.

然后是关于大语言模型作为产品的差异性问题。

Then there's the question around, like, the differentiation in the LLMs as product.

在我看来，现在你可以对Grok、Claude、Gemini、Mistral、DeepSeek进行双盲测试，用同样的提示词输入，我打赌大多数人根本分辨不出哪个是哪个。

Like, seems to me right now, you could do, a double blind test of the same prompt given to Grok, Claude, Gemini, Mistral, DeepSeek, do a double blind test, I bet most people wouldn't be able to tell which is which.

关于是否存在产品差异性的问题。

That question of, like, is there product differentiation?

大语言模型作为消费品，能否实现产品差异化？

Can there be product differentiation around the LLM as consumer product?

因为目前这些模型都像是大宗商品，但ChatGPT的使用量远远、远远、远远、远远、远远超过其他。

Because right now, the models are commodities, but ChatGPT is way, way, way, way, way more usage.

所以ChatGPT在应用商店排名中位居榜首。

So ChatGPT is at the top of the app store rank.

Gemini的排名在50到100之间波动。

Gemini bubbles between, like, 50 and a 100.

其他所有模型都没进入前100名。

None of the others were in the top 100.

在谷歌趋势、使用数据和收入方面也都是如此。

Same in Google Trends, same in the usage numbers, same in the revenue.

企业API层面是有收入的。

There's revenue for a corporate API.

企业市场是另一回事。

Corporate is a whole other story.

但作为消费级产品，ChatGPT现在已经成了品牌代名词。

But as a consumer thing, it's like ChatGPT is now like the brand.

这是默认选择。

It's the default.

它就像谷歌一样。

It's the Google.

你使用它是因为你听说过它，而其他任何产品都没能超越它。

You use it because you've heard of it, and none of the others have broken it.

我们现在就是这种情况吗？

Is that where we are now?

但如果你看看这些产品，其实这些模型都差不多。

Is that but but then if you look at the products, the products not any of the models are all the same.

底层模型都是一样的。

The underlying model is the same.

产品也都差不多。

The product's all the same.

对吧？

Right?

很难分辨出区别，除了它们有不同的配色方案和不同的图标。

It's really hard to tell the difference except, like, they've got different color schemes and different icons.

不同的品牌标识。

Different branding.

不同的品牌标识。

Different branding.

但产品都是一样的。

But the I product's all the same.

这让我想起了浏览网页时的情况，因为所有浏览器都差不多。

And this reminded me of looking at browsers, in that browsers are all the same.

底层的渲染引擎可能不同，就像大语言模型可能不同一样。

The rendering engine underneath might be different, just as the LLM might be different.

但你有一个应用，一个输入框，一个输出框，而输出框在思考Renamend能给你什么。

But you've got an app, an input box, an output box, and the output box wonders what the Renamend can give you.

过去二十五年里，浏览器唯一的创新基本上就是标签页和将搜索功能合并到地址栏中。

And the only innovation in browsers in the last twenty five years is basically tabs and merging search in the app address part.

是的。

Yeah.

而且这就像浏览器项目一样。

And it's like there's a lot of the browser project.

现在有人在尝试做这件事，但没成功。

There's people trying to do it now, but it hasn't worked.

它没有获得关注。

It hasn't got traction.

LLM会不会也是这样，关键在于分发和品牌，而不是产品或模型本身？

And is that sort of how LLNs will work in that it's about the distribution and the brand is not actually about the product or the model?

还是说它更像社交领域，比如拍照是种商品，但Instagram和Flickr之间有巨大差异？

Or is it maybe more like social in that, yeah, photo showing is a commodity, but you there's a big difference between Instagram and Flickr.

是的。

Yeah.

还有其他所有尝试做拍照分享的人。

And all the other people that try to do photo showing.

所以你必须

And so you have to

几乎可以说这是一种赢家通吃的模式。

be almost be an argument that it's sort of winner take all.

对吧？

Right?

以Cload为例，它很难竞争。

It's very hard for like, use Cload as an example.

如果Cload没有足够的用户基础来持续投入，它就很难竞争。

It'd be very hard for Cload to compete if they don't have enough usage to continuously make the investments.

嗯，这稍微有点不同。

Well, that's a slight this is a slightly different thing.

在那里，赢家并没有形成一种自我强化的循环：因为更多人使用，所以产品变得更好，进而吸引更多人使用。

So the winner there doesn't appear to be a sort of self reinforcing cycle in which more people use it because more people use it.

产品因为更多人使用而变得更好，从而吸引更多人使用，这就像操作系统那样：因为有更多应用，所以有更多用户，进而又有更多应用。

The product gets better because more people use it, so more people use it, which is what you have with operating systems because you have more apps, therefore more users, therefore more apps.

所以谷歌搜索的情况是，谷歌拥有大量用户使用它的反馈数据，这些数据让搜索引擎变得更好。

So what you have with Google Search, that Google has all the feedback from how people use it that makes the search engine better.

社交媒体存在网络效应：你在那里，是因为你在那儿，因为你的朋友在那里，因为你在那里。

You have a network effect in social media that you're there because you're there, because your friends are there, because you're there.

目前，大语言模型中还没有明显的类似机制。

There's no apparent equivalent in LLMs right now.

没有理由认为大语言模型会因为更多人使用而变得更好。

There's no reason why the LLMs get better because more people use them.

但这可能会出现。

Now that may come.

OpenAI 等公司已经在做记忆功能，能记住你之前问过的问题，但这更像是切换成本，而不是网络效应。

You have that the OpenAI and people have been doing memory, where it remembers what else you've asked, but that seems more like a switching cost than a network effect.

而且，你也可以直接问它关于你的信息，然后把同样的内容告诉 Claude，或者反过来。

And also, it might be easier for you to just ask it what it knows about you and then tell Claude or vice versa.

所以目前还不清楚，但我们正处在类似浏览器时代的阶段，正在思考：有没有办法在这里建立用户粘性？

So it's not clear if that but it we are at that sort of stage where you're looking at the browser and saying, there a way that you can create stickiness here?

或者你能在浏览器上创造网络效应吗？

Or that you can create network effect on the browser?

还是说浏览器本身只是一种商品？

Or is it just that the browser itself is a commodity?

现在，在这种模式下，资本并不具有赢家通吃的效果。

Now, capital is not a winner takes all effect in that convention.

所以，无论如何，都存在另一种类型的赢家通吃效应。

So it always and anyway, there's a different kind of winner takes all effect.

我的意思是，我通常不会把资本看作一种网络效应。

I mean, I wouldn't conventionally think of capital as a network effect.

或者这并不是产品本身固有的东西。

Or is it it's not a product that's in it's it's not something that's inherent in the product.

它是其他某种东西。

It's something else.

也许确实如此，ChatGPT 和 OpenAI 拥有更多的资金，因此能改进他们的模型。

It may be that, yes, chat g p that OpenAI has more money so they can make their model better.

在我们转向新话题之前，我想先深入探讨六个方向。

There's just, like, six rabbit holes I wanna go down before we move on to something new.

如果OpenAI能像你提到的那样，因为人们使用而让AI变得更好，那么成为OpenAI是不是就拥有巨大优势？

If OpenAI can I I like your point about sort of at the point where AI gets better because people are using it, then there's a huge advantage to being OpenAI?

我们目前还看不到这会是什么样子。

We don't have any we don't have visibility on what that would be yet.

但到了那时，领先者可能会

At that point, though, whoever's in the lead would sort of

如果真是这样，就可能形成一种失控的局面。

If it did, then you could get kind of a runaway.

但我们应该回头想想MySpace。

But we should kind of go back and think about Myspace.

嗯。

Mhmm.

因为在这些事物的早期阶段，你会看到类似MySpace的情况，就像早期个人电脑行业一样，当时有十几家公司在竞争。

Because you have, like and this it would be Myspace, you know, in the early phases of these things, and you see the same thing with the early PC industry, you've got a dozen of them.

通常会有一个早期的领先者后来逐渐衰落。

And there's often an early leader that falls away later.

因此，MySpace就是那个后来衰落的早期领先者。

And so Myspace was the early leader that fell away later.

然后你会进入一个后期阶段，S型曲线趋于平缓，网络效应和产品品质都已稳固。

Then you get a late stage where the s curve is kind of flattened out, where all the network effects have kind of solidified and the product quality has solidified.

在早期，人们其实很容易在MySpace、Facebook、Bebo、Friends Reunited、Orkut以及其他所有平台之间来回切换。

It was very easy actually to get people to switch back and forth between Myspace and Facebook and and Bebo and and and Friends reunited and whatever the Orcut and all the other things in the early days.

对。

Right.

然后你会逐渐看到这种分化出现。

Then you kind of get this this separation out.

但接着，Instagram出现了。

But then, of course, then you get then Instagram comes along.

然后TikTok也出现了。

And then TikTok comes along.

对。

Right.

所以，一旦出现一个不同的产品主张，并且很容易就把用户拉走，你知道，谷歌失去了YouTube，他们不得不买下YouTube。

And so as soon as you have something that's a different proposition that turned out to be extremely easy to pull that away, you know, read Google Google lost YouTube, they had to buy YouTube.

Facebook失去了Instagram和WhatsApp，不得不把两者都买下。

Facebook lost to to Instagram and WhatsApp, and they had to buy them both.

所以这些winner-takes-all效应相当脆弱且狭窄，或者至少看起来是这样。

So those are quite fragile and quite narrow winner takes all effects, or at least they they they appear to be.

我们不知道会是什么样子，或者各种模式会如何呈现。

We don't know what that would be or what the modalities would look like.

模式，抱歉，这个词太空洞了。

Modalities, sorry, that's a great meaningless word.

就像说‘社会性的’一样。

It's like saying societal.

不知道会是什么样子，因此我们也不知道它会有多稳固，或者如何运作，因为我们还没有它。

Don't know what that would look like, and therefore, we can't we don't know how rigid it would be or how it would work because we don't have it yet.

你不能，正如你所知，他们并不是一直在用数据重新训练模型。

You can't as as I'm sure you know, you're not they're not retraining the models all the time with the data.

所以你不会出现那种持续流动、查询越多结果就越好的失控效应。

So you don't have that kind of runaway effect as, like, continuous flow and more queries produces, you know, better results.

因此，目前要做到这一点还挺棘手的。

So it's kind of tricky to do that yet.

我想回到你刚才说的一点。

I wanna come back to something you said.

你说有些人看ChatGPT却看不懂。

You said some people look at ChatTPT and don't get it.

是的。

Yeah.

我觉得这非常重要。

I think this is really important.

有很多调查数据是关于有多少人在使用这些东西的。

There's a whole bunch of survey data on how people how many people are using this stuff.

你们有来自ChatOpenAI的数字，他们预测说，我们有这么多周活跃用户。

You've got the numbers from so so so so chat OpenAI predicts, say, well, we've got this many weekly active users.

有趣的是，社交媒体兴起时，人们会谈论注册用户。

Funny thing about social is people when social happened, people would talk about registered users.

你知道，在互联网早期，人们会谈论访问量。

You know, in the early days of the Internet, people would talk about hits.

是的。

Yeah.

然后我们意识到，如果你的网页菜单栏有七个项目，那就是七个GIF，也就是七次访问。

And then we realized that if your web page has seven items in the menu bar, that's seven GIFs, so that's seven hits.

所以访问量毫无意义，后来我们改用页面浏览量，再后来是注册用户，然后是月活跃用户。

So hits was meaningless, and you had to switch to page impressions, and then it's registered users, then it's monthly active users.

在社交媒体上，人们说，等等。

And on social, people said, hang on.

如果你一个月只用一次Instagram，那你根本没在用。

If you're using Instagram once a month, you're not using it.

相反，活跃用户根本毫无意义。

Instead, the active users are nothing.

每周活跃用户，我们也不喜欢。

And weekly active users, we don't like either.

现在OpenAI在使用每周活跃用户这个指标。

Now OpenAI is doing weekly active users.

山姆·沃尔顿曾是社交媒体初创公司的创始人。

And Sam Walton was a social media startup founder.

他懂这个。

He knows this.

这纯粹是个荒谬的数字。

It's a bullshit number.

你看看调查数据，我上次做演示时做的幻灯片里，有五个来自美国的调查，时间是去年底到今年初，结果都差不多。

You look at survey data, and I had to did the slide in the last presentation I did of, like, five different surveys from The US from late last year, earlier this year, and it's all roughly the same.

大约有10%的人，根据不同调查上下浮动三到四个百分点，表示他们每天都在使用这个平台。

It's like something around 10%, give or take three or 4% of people, depending on the survey, are using this say they're using this every day.

另有大约15%到20%的人表示他们每周使用一次。

Another sort of 15 to 20% of people say they're using it every week.

然后你就有了大约10%的人每天使用它。

And then there's so you've got, like, say, 10% of people using it every day.

大约15%到20%的人每周使用一次。

Say, 15 or 20% of people using it every week.

另外20%到30%的人表示他们每月或每两个月使用一次。

Another 20 or 30% of people who say I use it every month or two.

还有另外20%到30%的人说，是的。

And another 20 or 30 people of percent of people who said, yeah.

我看过一眼。

I had a look.

我没搞明白。

I didn't get it.

然后你看到另一项调查，说70%的人在使用人工智能。

And then you have this survey where people say 70% of people are using AI.

我心想，等等。

I'm like, wait.

你什么意思？

What what do you mean?

还有另一个完全不同的问题，就是人们会说，你用过 Snapchat 的面部滤镜吗？

There's a whole other rabbit hole, which is, you know, people say, well, did you use Snapchat's face filters?

我会说你在使用人工智能。

I'll say you're using AI.

我们说的 AI 到底是指什么？

What do we mean by AI?

所以让我们具体一点。

So which is let's be specific.

我们来谈谈，你有没有使用面向消费者的大型语言模型聊天机器人？

Let's talk about your are you using a consumer facing LLM chatbot?

比如，你有没有使用 ChatGPT 或 Claude 来提问？

Like, you're going to chat GPT or Claudine asking questions.

这就是我们需要关注的数字。

Like, that's the number we want to look at.

大多数人不会把收件箱当作一个系统，但我却会。

Most people don't think about their inbox as a system, but I do.

邮件曾经是我经营业务的得力助手。

Email used to be the thing that helped me run my business.

最近，它却成了阻碍我工作的因素。

Lately, it felt like the thing getting in the way of it.

我花了太多时间筛选低优先级的消息，生怕错过那唯一或两封真正重要的邮件，这让我精力耗尽。

I'd spend too much time weeding through low priority messages, trying not to miss the one or two that actually mattered, and it was draining my focus.

然后我开始使用Notion Mail，一切发生了改变。

Then I started using Notion Mail, and everything changed.

Notion Mail 是一个像你一样思考的收件箱。

Notion Mail is the inbox that thinks like you.

它自动化、个性化且灵活，终于能按照你原本的方式工作。

It's automated, personalized, and flexible to finally work the way that you were.

借助能学习你关注内容的AI，它可以整理你的收件箱、标记邮件、草拟回复，甚至安排会议。

With AI that learns what matters to you, it can organize your inbox, label messages, draft replies, and even schedule meetings.

无需手动分类。

No manual sorting required.

我创建了自定义视图，根据紧急程度和主题对收件箱进行分类，让我能专注工作而不受干扰。

I've created custom views that split my inbox by urgency and topic so I can focus without distraction.

我还使用片段快速发送最常见的邮件、跟进信息、介绍信和日程安排，无需重复编写内容。

And I use snippets to fire off my most common emails, follow ups, intros, and scheduling without rewriting anything.

最棒的是，它能与我的Notion工作区无缝集成，并由Notion提供支持——这是超过一半财富500强公司信赖的工具。

The best part is it works seamlessly with my Notion workspace and is powered by Notion, the tool trusted by over half of Fortune 500 companies.

Notion以其强大的连接性、直观的功能和提升生产力的能力而闻名。

Notion is known for powerful connectivity, intuitive functionality, and the ability to supercharge productivity.

立即前往 notion.com/knowledgeproject 免费获取 Notion Mail，体验这个能像你一样思考的收件箱。

Get Notion Mail for free right now at notion.com/knowledgeproject and try the inbox that thinks like you.

全部是小写字母：notion.com/knowledgeproject，立即免费获取 Notion Mail。

That's all lowercase letters, notion.com/knowledgeproject to get Notion mail for free right now.

当您使用我们的链接时，也在支持我们的节目。

When you use our link, you're supporting our show too.

notion.com/knowledgeproject。

Notion.com/knowledgeproject.

对我来说，你可以从多个角度来分析这个问题。

And to me, there's a there's a bunch of you could matrix this.

所以有些方面还处于早期阶段。

So some of this is it's early.

这里有一个相反的观点，如果人们做这个图表并说：天哪。

There's a counterpoint here, which if people do the chart and they say, my god.

进展太快了。

It's so fast.

它比智能手机还要快。

It's like, it's faster than smartphones.

是的。

Yes.

因为你不需要花一千美元买一部智能手机。

Because you didn't need to buy a thousand dollar smartphone.

对。

Right.

它比个人电脑还快。

It's faster than PCs.

对。

Yes.

你知道八十年代一台个人电脑的价格，按通货膨胀调整后是多少吗？

Because you know what PCs cost in the eighties adjusted for inflation?

大概五美元。

It's like $5.

是的。

Yeah.

对很多人来说，它是免费的。

It's free for a lot of them.

它是免费的。

It's free.

它是一个网站。

It's a website.

你直接去那里就行了。

You just go there.

当然，它的普及速度更快。

Of course, it's got faster adoption.

而且现在在线的人也多得多。

And there's way more people online as well.

所以，即使绝对数字也比二十年前、十五年前的Facebook增长得更快，因为现在在线的人多了很多。

So even the absolute numbers are faster than they were for Facebook twenty years ago, fifteen years ago, because there's way more people online now.

是的。

Yes.

所以，这再次说明了我那个不公平但相关的比较。

So so that's that's again an example of my unfair but relevant comparison.

你可以说是站在巨人的肩膀上。

You're sort of standing on the shoulders of giants.

所以当然，你能更快地接触到更多人，但你是不是还得不断问：那又怎样？

So of course, you can get to way more people quicker, but do you have to keep asking, well, what?

是的。

Yes.

但为什么有这么多人都看了却看不懂呢？

But why do so many more people look at this and not get it?

或者更糟的是，那种看不懂的情况，其实还能看得出来。

Or even worse, like, the not getting it, can kinda see.

因为人们看很多东西都看不懂。

Because people look at everything and don't get it.

为什么有些人看了就懂了，还每周都回来，但只每周回来一次？

Why is it that somebody looks at this and gets it and goes back every week, but only every week?

对。

Right.

为什么他们每周只能想到一次与这个相关的事情呢？

Why is it they can only think of something to do with this once a week?

我担心这样的人。

I worry about those people.

我的意思是，我只是在想，如果这些数据准确的话，我花最多时间相处的那10%、15%的人，其实都属于那10%。

I mean, I'm just thinking if these numbers are accurate, the 10%, the 15, you know, 90% of the people that I spend the most time with are within that 10%.

嗯，我不是这样。

Well, I'm not.

有意思。

Interesting.

让我多说说，其实，我得先说明一下，我的孩子们已经不再使用谷歌了。

Tell me more about well, here, actually, I'll I'll preface this conversation with my kids don't use Google anymore.

他们只用谷歌查找电话号码、本地商家或地点距离。

They have they use it to find phone numbers or local businesses or places distance.

其他所有事情，他们基本上都已经转用ChatGPT了。

Everything else, they they basically have defaulted to ChatGPT now.

再说一遍，所有的AI对话似乎都是类比，所以我完全同意。

Again, I'm going to all AI conversations seem to be analogies, so I'll A 100%.

这就像核武器一样。

I'll I'll it's like it's like nuclear weapons.

不，不是这样的。

It's like, no, it's not.

我认为这里一个有趣的比较是，虽然不完美，但值得一看：早期的电子表格软件与纸质表格的对比。

The comparison I think is interesting here, it was not perfect, but it's interesting, is to look at early spreadsheets for software spreadsheets, spreadsheets for paper.

丹·布雷克林创造了PhysiCalc，但我记不起另一个人的名字了。

Dan Brecklin creates phys and I can't remember the other guy's name.

他在七十年代末创造了PhysiCalc。

He created PhysiCalc in the late seventies.

我想，当时让一台Apple II运行它，配上屏幕等设备，成本大约是15美元（经通货膨胀调整后）。

And I think to get an Apple II to run it with a screen and everything costs, like, $15 adjusted for inflation.

你把这个展示给会计师看，他们会发现，你在这里修改利率，其他所有数字都会随之变化。

And you show this to an accountant, and then it's like, you can change the interest rate here and all the other numbers change.

我们现在看到这种情况，就会说：没错。

And we see that now and we're like, yes.

1978年，那可是整整一周的工作量。

1978, that was a week of work.

我的意思是，几乎是字面意义上的。

I mean, almost literally

那简直太惊人了。

That was like amazing.

是的。

Yeah.

他半小时就能完成一周的工作。

He would do a week of work in half an hour.

是的。

Yeah.

甚至更短。

Or less.

他有很多关于会计师的故事，你知道，他们会被分配一个月的项目，却能在一周内完成，然后去打三周的高尔夫球，部分原因是他们可能不想告诉客户‘我其实一周就能做完’，因为客户会以为他们没好好干活。

And he has all these stories about accountants who would, you know, they would be given a one month project and they'd get it done in a week, and then they'd like go and play golf for three weeks, because they partly because they could probably they didn't actually want to tell the client I needed it in a week because the client would think they hadn't done it properly.

所以我看到ChatGPT时就想，好吧，我不写代码。

So I look at ChatGPT and I think, right, I don't write code.

我对那些愿意替我写代码的人毫无用处。

I have zero use for some people that will write code for me.

我其实不做头脑风暴。

I don't really do brainstorming.

好吧。

Okay.

我不做内容总结。

I don't do a summarization of things.

我不做那些简单直接、显而易见的事情。

I don't do the things where it's sort of out of the box easy and obvious.

然后还有一种心理负担，就是我得试着想想，有哪些事情是我正在做的，而它能替我完成的。

And then there's a sort of mental load of, okay, I've got to kind of try and think of what things am I doing that it could do for me.

而且大多数人并不这样思考。

And that most people don't think like that.

所以就像我刚才说的，存在一种矩阵。

So there's a sort of said a moment ago, there's like a matrix.

这种矩阵涉及谁拥有它擅长的那种任务，这很明显。

There's a matrix of, like, who has the kinds of tasks that it's good at, obviously.

谁擅长拥有它擅长的那种任务，这并不明显。

Who is good at has the kind of tasks that it's good at, not obviously.

谁擅长思考他们正在做的事情的新工具？

Who is good at thinking about new tools for the things that they're doing?

谁不擅长？

Who isn't?

如果你不本能地使用AI，我简直震惊了。

If you I'm kinda blown away you don't reflexively use AI.

如果你

If you

你们在使用Salesforce，上面有一个按钮写着‘帮我草拟一封回复客户的邮件’，对吧。

were using Salesforce and you had a button that said draft me an email to reply to this client Yeah.

然后这种功能就会被广泛采用。

Then that gets massive adoption.

嗯，那

Well, that

那是另一回事。

that's different thing.

是的。

Yeah.

是那种聊天机器人产品吗？你面对一个空白屏幕，得盯着它，挠头思考：我到底能用它做什么？

Is it is it the chatbot as product where you get this blank screen and you kind of have to look at it and you scratch your head and you just have to think, well, what is it that I would do with this?

然后你还得围绕它养成新的习惯。

And then you have to form new habits around it.

还是说，它被封装在一个产品和用户界面中，有人已经告诉你：这个功能用在这里会很有用，不是吗？

Or is it that it's wrapped in product and UI where somebody else has said, it would be really useful for this, wouldn't it?

然后你看着它，说：哦，对了。

And then you look at it and go, oh, yeah.

这个我能做。

That would I I could do that.

你觉得是定性分析好还是定量分析好？

Do you think it's better with qualitative or quantitative

分析？

analysis?

我认为目前是这样，我会说个绝对点的结论：我认为现在它对定量分析毫无价值。

I think it is presently, and I'm gonna get a binary statement, I think today it has zero value for quantitative analysis.

哦，有意思。

Oh, interesting.

因为，让我先澄清一下。

Because if well, let me let me qualify that.

数字需要精确还是大致准确？

Do the numbers need to be right or roughly right?

嗯。

Mhmm.

因为所有这些事情的作用都是给你一个大致正确的东西。

Because what all of these things do is they give you something that's roughly right.

而大致正确是一个范围，但

And roughly is a spec the spectrum, but

你总是不希望圆周率是3.1。

it's always You don't want pi to be 3.1.

这取决于你测量的圆有多大。

Depends how big the circle you're measuring is.

对。

Right.

你知道，这就是关于圆周率的那句话，我们可以计算它。

You know, this is the line about pi that, you know, we can calculate it.

然而，即使我们有了这么多位数，也足以计算出宇宙的直径之类的，但人们仍然在不断添加更多数字。

The light then however many digits we have is like enough to calculate, you know, the diameter of the gal of the universe or something, but people still adding more numbers.

所以，你知道，这里面有一点芝诺悖论的意思，就是你无限接近。

So, you know, there's a little bit of Zeno's paradox in here, you know, like you get infinitely close.

但在某个点上，这就不重要了。

At a certain point, it doesn't matter.

从高层次来看，这正是关于AGI的一些争论：如果一个东西无限接近于推理，却从未真正进行推理，这有关系吗？

And this is, at a high level, this is some of the AGI argument, that if the thing gets infinitely close to reasoning without ever actually reasoning, does it matter?

对。

Right.

比如，在某个点上，如果一个东西总是对的，如果它每十亿年才错一次，那它不是始终正确，这有关系吗？

Like, at a certain point, the thing if the thing is always right, if the if the thing is only wrong once in a billion years, does it matter that it's not that it's not always right?

今天的问题是，它并不是每十亿年才错一次。

The problem today is it's not wrong once in a billion years.

它每页就会错上十几回。

It's wrong a dozen times a page.

是啊。

Yeah.

你不想把这玩意儿吐出来给别人。

You don't wanna spit that out and give it to somebody.

我不知道。

And I don't know.

嗯。

Yeah.

嗯。

Yeah.

我有个很早的例子，当时我打算在2023年初的一个活动上发言，主办方让我提供一份详细的个人简介，但我没有，现在也没有。

So I had a very early example of this, and I was going to speak at an event at beginning the of twenty twenty three, and the conference people had asked me for a long biography of myself, and I didn't have I don't have still don't have one.

于是他们自己做了一份，还用ChatGPT生成的，没告诉我，只是发给我让我核对，我一看就问：这他妈是什么垃圾？

And so they'd made one, and they'd used it with ChatTVT and not told me, and they just sent it to me to check, and I looked at them and they said, what the fuck is this bullshit?

那是2020年的事。

That's 2020.

那简直是上个时代了，而且

That's like But but generations ago, and

这和我要说的重点无关。

that's not relevant to the point I'm gonna make.

我要说的是，第一，那确实是一种合适的传记。

The point I'm gonna make is, a, it was always the right kind of biography.

它有合适的学历、合适的大学、合适的经验、合适的工作。

It was the right kind of degree, the right kind of university, the right kind of experience, the right kind of jobs.

但其实并不是真正合适的内容。

It just wasn't actually the right things.

但第二，我可以拿这个来修正它。

But b, I could take that and fix it.

嗯。

Mhmm.

所以对他们来说，这毫无用处。

So for them, it was useless.

对。

Right.

对我来说，这非常有用。

For me, it was completely very useful.

我只需要花三十秒就解决了，而不是花一小时抓耳挠腮，这就是为什么我说对错取决于具体情况，这简直是个非常法国哲学家式的问题。

I was like, spent thirty seconds fixing it instead of spending an hour scratching my head, which is why I say right or wrong depends, which is a very kind of French philosopher kind of question.

答案是，提示本身是否有错误？

Is the answer is the prompt does it does the prompt have errors?

这在很大程度上取决于你想要它的原因。

It kind of depends why you wanted it.

好吧。

Okay.

我并没有需要大致正确内容的使用场景。

Well, that I don't have use cases where I want something that's roughly right.

我没有这样的使用场景：想要十个点子、想让它帮我头脑风暴、想让它帮我起草邮件、想让它写代码，或者想让它生成一些图片。

I don't have use cases where I want a list of 10 ideas or I want it to brainstorm or I want it to draw off me an email or I want it to write code or I don't want it to generate some images.

你知道吗，有个朋友在咨询公司工作，他们需要概念的草图，现在他们可以直接用Pidjourney来制作这些。

You know, a friend who works at a consultancy and they want pencil sketches of concepts, and now they can just use Pidjourney to make those.

太好了。

That's great.

那个素描，比如后面那个人，有三条腿吗？

Does that sketch, like, that person at the back have three legs?

现在没有了。

Not anymore.

没有。

No.

即使有，也没关系。

And if they did, it wouldn't matter.

你可以用Photoshop修掉。

You could Photoshop that out.

我不做那种事。

I don't do that.

那就是我不生成图像。

That's I don't create images.

所以我无法很好地将这些内容适合早期使用的部分，与我实际做的事情对应起来。

So I don't have a good mapping of the stuff this is good for early against the stuff that I do.

而它可能有用的地方，恰恰是目前还不是很擅长的领域。

And the stuff that it that it maybe would be useful for is the stuff where it's actually not yet very good.

至于那些你可以通过说‘我会去修正它’来弥补的问题，我并不做那些事情。

And the things where you would mitigate that by saying, well, I would fix it, I don't do those things.

好的。

Okay.

这很好，因为我想回到你之前说过的一件事。

So this is a good thing because I wanted to come back to something you said.

你说你是通过写作来思考的。

You said you think by writing.

但在一个你使用AI生成内容并加以编辑的世界里，这和写作是不同的。

And in a world where you're taking something generated by AI and editing it, that's different than writing.

跟我谈谈‘通过写作来思考’这件事。

Talk to me about thinking by writing.

我为什么改变了我的使用场景，这更多是一种思维模型而非实际应用：我写一些东西，然后思考它是否真的有价值——过去我总会问，就像你提到的模式识别，我会看某样东西并问自己：我在这里是否增添了价值？

I why she my my my change beauty use case, which is more mental model than a practical thing is I write something and I think it's such a I I would what I was always would ask in the past is kind of your point about pattern recognition is, I look at something and say, am I adding value here?

我是否说了些有用的东西？

Am I saying something useful?

我是否说了些不同的东西？

Am I saying something different?

我是否提出了关键问题？

Am I asking the key question?

我是否在深入挖掘？

Am I pushing further?

我是否在提出下一个问题，而不是仅仅回答显而易见的问题？

Am I pushing the I asking the next question rather than just answering the obvious questions?

现在我只需要问：这是ChatGPT会说的话吗？

Now I can just say, is this what ChatGPT would have said?

如果答案是‘这正是ChatGPT会说的’，那我就不会发布它。

And if the answer is this is what ChatGPT would have said, then I don't publish it.

不是因为人们可以从ChatGPT那里得到它，而是因为任何人都会这么说。

Not because people can get it from ChatGPT, but because anyone would have said that.

这种分析非常到位，因为它提高了什么是洞见的标准。

That's a perfect analysis in the sense that it raises the baseline of what qualifies as insight.

区别在于洞见的上升斜率。

The difference is the slope of the insights.

所以，如果ChatGPT会这么说，你就不要说，当然也可以对此提出质疑。

And so you wouldn't say it if ChatGPT is gonna say, and push back on this, by all means.

但以ChatGPT级别的洞见为例，无论是Claude、Grok还是其他任何模型，其进步速度都比大多数人快得多。

But the the ChatGPT level of insight, to use an example, it could be Claude, it could be Grok, it could be any of them, is increasing at a faster pace than most people.

最终，这些斜率会相交，可能已经交汇了，比如达到了实习生水平。

And eventually those slopes intersect and it's probably intercepted or intersected already with, you know, maybe up to intern level.

明年，它可能达到硕士水平，甚至在某些领域（如数学）早已远远超越。

And next year, it might be master's level or it might even far surpass that and it has in some domains in terms of math.

再过一年，它可能会继续提升，也许五年后它才会超越本尼迪克特，四年后超越其他人，而可能早在很久以前，它就已经超越了我。

The year after, it might and so maybe it's like five years before it it passes Benedict, and maybe it's four years before it passes somebody else, and maybe it like passed me way long time ago.

所以我认为我们可以在这一点上采取两到三个方向。

So I think there's a there's there's two or three directions we we could take that.

其中一个方向是，这是一个有趣且具有哲学意味的问题，即原创性——也就是说，AlphaGo 能够走出原创的棋步，因为它可以尝试所有可能的走法，并走出前人从未走过的步法，而它并不知道人类之前做过什么。

One of them is that's an interesting theoretical philosophical question, is originality, which is to say AlphaGo could do original moves because it could do all the moves and do do moves that no one had done before, not knowing what people had done before.

但它有一个外部的评分系统，而那个步法本身并不一定好。

But it had an external scoring system and neither that move was good.

嗯。

Mhmm.

因为他们没有反馈。

They because they didn't have feedback.

是的。

Yeah.

他们有一个反馈循环，因为每一步都有一个评分。

They had a feedback loop because every move has a score.

他们可以对每一步进行评分评估。

They can evaluate the score every move.

经典的猴子打字机寓言，或者你所说的波吉亚影响者图书馆，是没有反馈循环的。

The classic parable of the monkeys and typewriters, or you know the Borgias Influencer Library, is there's no feedback loop.

是的。

Yeah.

对。

Yes.

波吉亚影响者图书馆包含随机生成的新杰作。

The Borgias Influencer Library contains new masterpieces generated at random.

而猴子打字机虽然也会生成新的杰作，但因为没有反馈循环，所以无从判断。

While the monkeys with typewriters would generate new masterpieces, but there's no feedback loops, so there's no way of knowing.

你现在在音乐领域就能看到这种情况。

You'll see this in music now.

你可以生成新的音乐，也能生成一些你原本想不到的东西。

You can generate new music and you could generate new stuff that you wouldn't know.

对于大语言模型来说，变异是坏事。

For an LLM, variance is bad.

原创性得分较低。

Originality is is is a lower score.

那么，原创但优秀的东西，其反馈机制是什么？

So what's the feedback loop for original but good?

现在，这可能和问‘这真的是推理，还是只是重复了99次，你知道的，九九九后面跟着很多零？’一样，是个虚假的问题。

Now it might be that that's the same sort of false question as saying, is it really reasoning, or is it just right 99 times, you know, nine nine nine followed by many zeros?

它真的理解了吗，还是只是在不理解的情况下总是对的？

Does it actually understand, or is it just always right without understanding?

它真的知道这是原创且不同的吗，还是不知道？

Does it actually know that's original and different, or does it not?

这是一个令人困惑的问题，我认为我们还不知道答案，也许这个问题本身就有误，但它确实是个谜：这些系统究竟能在多大程度上创造出既不同于训练数据又优秀的东西？

And that's a that's a kind of a puzzling I don't think we know the answer for that, and it may be the wrong wrong question, but it's kind of a a puzzle as to how would these how far can these things make things that are both different from the training data and good?

知道这是不同的，与知道这是好的，它们真的不同吗，还是只是在更长的频率上匹配模式？

And is knowing that this is different than good, is it really different, or is it just matching the pattern on a longer frequency?

你

Do you

你明白我的意思吗？

see what I mean?

如果给足数据，你实际上能预测到多少内容，以至于它并没有真正超出模式？

And how much could you actually have predicted that given enough data that it's not actually outside the pattern?

如果你放大来看，它只是看起来像是那样。

It just kinda looks like it is if you're zoomed in more.

而如果你拉远来看，那其实就是在匹配模式。

And if you zoom out more, then that is matching the pattern.

你怎么知道人们会喜欢朋克音乐呢？举个例子，你是怎么知道人们会喜欢这种音乐的？

How would you know that people it's like, you know, thinking about music, like, how would you know that people would like punk?

你可以很容易想象出一个生成式AI系统，它能生成更多听起来像Yes乐队的东西，或者更多听起来像平克·弗洛伊德的东西。

You could get you can very easily imagine generative AI system that can make you more stuff that sounds like yes, or more stuff that sounds like Pink Floyd.

它可能听起来不像真正的平克·弗洛伊德的好作品，但你可以想象它会生成更多听起来像感恩而死乐队的东西。

It might not sound like good Pink Floyd for Floyd, but you could imagine it would make more stuff that sounds like like the Grateful Dead.

你知道感恩而死乐队的粉丝在毒品用完后会说什么吗？

You know what Grateful Dead fans say when they run out of drugs?

这音乐太糟糕了。

This this music's terrible.

你可以想象，如果我这么说有点刻薄，但你能想象到，现在人们已经对七十年代的前卫摇滚感到厌倦，他们非常渴望一些新的东西，而这种新东西就是朋克，而且它会成功。

You can imagine this if the you can imagine I'm being unkind, but like, can imagine the challenges knowing that now people are really fed up of seventies prog rock, and they would really like something else, and that something else would be punk, and that would work.

就像知道四十年代的人们对战争感到厌倦，渴望奢华，而克里斯汀·迪奥的‘新风貌’会成功并表达这种需求一样。

Is knowing that people in the forties were really fed up of the war and would want luxury and that Christian Dior's new look would work and would express that.

一个大语言模型能做到这一点吗？

Could an LLM do that thing?

你需要多少差异性？

How much variance do you need?

我不知道。

I don't know.

提出这个问题是个很有趣的思维实验。

It's an interesting, like, thought experiment to ask that question.

哦，有意思。

Oh, interesting.

这里有一个完全不同的角度，那就是呼吁精品店、线下活动，以及独特、精选和个性化的体验。

There's a completely different place to take this, which is to say, this is an appeal for boutiques and in person events and the unique and the curated and the individual.

我总是会提到一家店。

There's a shop I always used always talk about.

我不确定它现在是否还开着。

I'm not sure if it's actually still there.

东京银座有一家店，只卖一本书，每个月都会更换书目。

There's a shop in Tokyo that just sells one book, in Ginza that just sells one book, and they change what it is once a month.

它可能十年前就关门了。

It may may have closed ten years ago.

我已经谈论它二十年了。

I've been talking about it for twenty years.

但重点是，你进店后不需要费心挑选该买哪本书。

But the point is you don't go into the shop and have to work out what book to buy.

嗯。

Mhmm.

但你得知道这家店存在。

But you have to know the shop exists.

嗯。

Mhmm.

或者你可以选择亚马逊，他们有五亿种商品，他们知道自己什么都有。

Or you can be Amazon, and you've got 500,000,000 SKUs, and they know they've got everything.

实际上，有些东西他们没有，因为他们想保持独特。

Like, actually, there's some stuff they don't have because they want to be individual.

比如，他们没有路易威登。

Like, they don't have LVMH.

但就论点而言，亚马逊什么都有，但你不能去亚马逊问：‘哪本书好？’

But for the sake of argument, Amazon has everything, but you can't go to Amazon and say, what's a good book?

或者，你知道的，‘哪盏灯好？’

Or, you know, what's a good lamp?

他们什么灯都有。

They have all the lamps.

你不能只是去那里问：我该买哪盏灯？

You can't just go to it and say, what lamp should I buy?

对。

Right.

整个零售、商品陈列和广告行业都关乎你处于这个光谱的哪个位置，以及你还能做些什么？

All of retailing and merchandising and advertising is about where are you on that spectrum and what else do you do?

你是把钱花在租金、广告还是物流上？

Do you do you spend the money on rent or advertising or shipping?

还是其他什么？

Or what?

这又是怎么运作的？

And how does that work?

而如今，两极分化越来越明显：如果我知道自己想要什么，我十二小时内就能拿到；但我要怎么知道自己想要什么呢？

And the more there's a sort of a polarization between, well, if I know I want the thing, I can get it within twelve hours, but how do I know I want the thing?

正如我前面提到的，大型语言模型或许恰恰能推荐给你那个独特而个性化的物品。

And as I alluded to earlier, an LLM is one might paradoxically, an an LLM could suggest you the unique individual thing.

LLM也会创造出那个独一无二的东西吗？

Would the LLM also create the unique individual thing?

这是一个很高的评价，但第二个问题是不同的。

That's a high award and a second that's that's a different question.

这是更后面的问题了。

It's a question further down the pipe.

但LLM越能做每个人都会做或说的事情，你就越需要往其他地方推进。

But the more that the LLM can do what everybody would probably do or say what everyone would probably say, then the more you push to other places.

这很有道理。

That makes a lot of sense.

我的意思是，无论来自LLM还是人类，洞察力的市场总是存在的。

I mean, there's always gonna be a market for insight, it comes from LLMs or people.

你必须提供洞察力。

You you have to be providing insight.

你知道，我们这些所谓的内容创作者，其世界涵盖了做着非常不同事情的广泛人群。

You know, our world as quote unquote content creators is a very wide spectrum of people who do very different stuff.

有人在做AI垃圾内容，也有人在做，你知道的，那个叫什么来着？

And there's, you know, there's people doing AI slop and there's people doing, you know, the the, the what is it called?

就是那种被动收入的东西。

The, you know, the passive income thing.

但有些人做着完全不同的内容，来自不同的地方，出于不同的原因。

But there's people who do very different kinds of content coming from different places for different reasons.

你知道，斯科特·加洛韦做的东西和我完全不同。

You know, Scott Galloway does very different kind of stuff to me.

玛丽·米克尔做的东西和我完全不同。

Mary Meeker does very different kind of stuff to me.

你做的东西也和我完全不同。

You do very different kind of stuff to me.

这其中一部分关乎你是谁、你的故事以及它的真实性。

It's just part of that is about who you are and your story and the authenticity of it.

而另一部分则在于，没人关心你是谁，但你说的东西很有趣。

And some of it is about no one cares who you are, but you're saying interesting stuff.

还有一些是推荐算法或其他原因。

And some of it's the recommendation algorithm and or something else.

左拉写过一本关于百货商店创建的书，名为《女士的幸福》，意思是女性的快乐。

There's a book by Zola about the creation of department stores called Bonnet de Dame, which means the happiness of women.

这本书基本上讲述了一位十九世纪的杰夫·贝佐斯，凭借意志力凭空创造出了百货商店。

And it's basically about a nineteenth century Jeff Bezos calling a an apartment store department store into existence out of thin air through force of will.

他发明了明码标价，从而可以实现折扣、亏损引流、邮购和广告。

And, like, he invents fixed prices so that you can have discounts and loss leaders and mail order and advertising.

他把滞销的昂贵商品放在商店顶部，而把食品和化妆品放在商店底部。

And, you know, he puts the the slow moving expensive stuff at the top of the store, and he puts food and makeup on the bottom of the store.

太阳底下没有新鲜事。

There's nothing new under the sun.

与此同时，街对面的店主们却在说：你看到那个疯子现在在干什么吗？

And meanwhile, the shopkeepers on the other side of the street are saying, like, have you seen what that maniac doing now?

他居然把帽子和手套放在同一个店里卖。

He's selling hats and gloves in the same shop.

他连一个痣都没有。

He's got no moles.

他接下来就要卖鱼了，当然，整个情节的核心就是亏本引流。

He'll be selling fish next, and of course, he's got the like, there's, like, the whole plot point is about loss leaders.

所以，你又得退后一步想想，人们以前就对工业化大规模生产的产品感到恐慌过。

And so there's, like, you you can again, you you had to step back and think, well, people have freaked out about industrial industrialized mass produced product before.

人们以前就对内容过多感到恐慌过。

People have freaked out about there being too much content.

伊拉斯谟是最后一个读过所有书的人。

There's a line that Erasmus was the last person to have read every book.

对吧？

Right?

现在互联网上充斥着太多AI生成的垃圾内容。

There's too much AI content slop on the Internet now.

是啊。

Like, yeah.

你认为1980年出版了多少本书？

How many books do you think were being published in 1980?

你觉得那时候每个人都在读所有的书吗？

Do you think everyone was reading all the books then?

是的。

Yeah.

情况一样。

Same thing.

只是规模不同而已，我想。

Just different scales, I guess.

你对今天的大学生有什么建议？

What advice would you give students today?

当我还是学生的时候，我们都被要求学日语。

Well, when I was a student, we were all supposed to be learning Japanese.

我想那只是那个时代的尾声。

I think that was just the tail end of that.