埃兹拉·克莱因秀：人工智能代理将多快席卷经济？

由Anthropic撰写，由Claude完成。

Written at Anthropic to be written by Claude.

这件事发生了吗？

Has that happened?

Anthropic是否按计划推进了？

Is Anthropic on on track for that?

我的意思是，现在有多少代码是由系统自己编写的？

I mean, how much coding is now being done by the system itself?

我认为绝大多数代码都是由系统完成的。

I would say comfortably the majority of code is being done by the system.

我们的某些系统，比如Claude Code，几乎完全由Claude编写。

Some of our systems like Claude Code are almost entirely written by Claude.

我的意思是，Boris负责Claude Code，他说他已经不写代码了。

I mean, Boris, leads Claude Code, says, I don't code anymore.

我只是和Claude Code来回互动，来构建Claude Code。

I just go back and forth with Claude Code to build Claude Code.

如果进展真的非常迅速，我们今年年底可能达到99%。

We could be 99% by the end of the year if things speed up really aggressively.

如果我们真的能很好地让这些系统在需要的地方编写代码的话。

If we are actually good at getting these systems to be able to write code everywhere they need to.

因为通常的障碍是组织性的繁琐事务，而不是系统本身的限制。

Because often the impediment is organizational schlep rather than any limiter in the system.

但据我所知，如今Anthropic拥有软件工程技能的人比两年前更多了。

But it is also true, as I understand it, that there are more people with software engineering skills working at Anthropic today than there were two years ago.

是的。

Yeah.

这绝对是真的。

That's absolutely true.

但人员结构正在发生变化。

But the distribution is changing.

我们发现，拥有非常精准直觉和品味的资深人员价值在上升，而初级人员的价值则有些不确定了。

Something that we found is that we are the value of more senior people with really, really well calibrated intuitions and taste is going up, and the value of more junior people is, like, a bit more dubious.

仍然有一些岗位需要引入年轻人，但我们正面临的问题是，哇。

There are still certain roles where you want to bring in, like, younger people, but an issue that we're staring at is, wow.

Claude代码或我们的编码系统能完成的那些基础任务，我们需要的是有丰富经验的人。

The really basic tasks Claude code or our coding systems can do, what we need is someone with tons of experience.

在这方面，我看到了未来经济的一些问题。

In this, I see some issues for the future economy.

对吧？

Right?

让我们先把这个，入门级岗位的问题放一放。

Let me put a pin in that, the the entry level job question.

我们很快就会再回到这个问题。

We're gonna come back to that quite shortly.

但这些程序员现在都在做什么？

But what are all these coders now doing?

如果Claude代码有望编写99%的代码，但你又没有解雇那些会写代码的人，那他们今天的工作内容和一年前相比有什么不同？

If Cloud Code is on track to be writing 99% of code, but you've not fired the people who know how to write code, what are they doing today compared to what they were doing a year ago?

其中一部分工作是构建工具来监控这些代理，无论是在Anthropic内部还是外部。

Some of it is just building tools to monitor these agents, both inside Anthropic and outside Anthropic.

既然我们现在有了这些高效运行的系统，就开始想要了解代码库中哪些部分变化最快，哪些部分变化最慢。

You know, now that we have all of these productive systems working for us, you start to want to understand where the code base is changing the fastest, where it's changing the least.

你希望弄清楚哪里存在瓶颈。

You want to understand where the blockages are.

有一段时间，一个主要的障碍是代码合并，因为合并代码需要人类和其他系统来检查其正确性。

You know, one one blocker for a while was being able to merge in code because merging code requires humans and other systems to check it for correctness.

但现在如果你生成的代码量大幅增加，我们就必须大幅改进这个系统。

But now if you're producing way more code, we had to go and massively improve that system.

我对这个问题有一个普遍的经济理论，叫做O型环自动化，其基本观点是自动化受限于链条中最慢的环节。

There's a general economic theory I like for this called O ring automation, which basically says automation is bounded by the slowest link in the chain.

此外，当你自动化公司的一部分时，人类会涌向最未被自动化的部分，提升其质量，并推动它达到可以被自动化的程度，然后你再进入下一个循环。

And also, as you automate parts of a company, humans flood towards what is least automated and both improve the quality of that thing and get it to a point where it eventually can be automated, then you move to the next loop.

因此，我认为我们不断发现一些奇怪地缓慢的环节，但我们可以改进它们，为机器的跟进铺平道路，然后找到下一个目标。

And so I think we're just continually finding areas where things are oddly slow, but we can improve to sort of make way for the machines to come behind us, and then you find the next thing.

所以Claude Code是一个相对较新的产品。

So Claude Code is a fairly new product.

Claude能够进行高级编码的时间跨度只有几个月吗？

The amount of time at which Claude has been capable of doing high level coding is can be measured in months?

一年？

A year?

可能有一年了。

Maybe a year.

是的。

Yeah.

Claude本身是一个非常有价值的产品。

Claude itself is a very valuable product.

你把一种非常新的技术，相对宽松地应用在一个非常有价值的产品上。

So you've you've set a very new technology, somewhat loose on a very valuable product.

你可能正在生成更多的代码。

You're probably producing more code.

很多人告诉我云代码的一个优点是它能正常工作。

One thing many people say about Cloud Code to me is that it works.

它并不优雅，但确实能用。

It's not elegant, but it works.

是的。

Yep.

但你现在对代码库的理解，很可能不如以前那么好了，因为你的工程师不再亲手编写代码了。

But presumably now you now understand the code base less well than you did before because your engineers are not writing it by hand.

你是否担心自己正在制造大量的技术债务、网络安全风险，以及与软件底层语言运行机制的直觉越来越脱节？

Are you worried that you're creating huge amounts of technical debt, cybersecurity risk, just an increasing distance from an intuition for what is happening inside the fundamental language of the software?

是的。

Yes.

这正是整个社会都将面临的问题。

And this is the issue that all of society is going to contend with.

世界上很大一部分领域，现在都将由AI系统来完成许多低层次的决策和具体工作，而我们需要理解这一切。

Just large chunks of of the world are going to now have many of the kind of low level decisions and bits of work being done by AI systems, and we're going to need to make sense of it.

要理解这一切，需要开发许多你可能称之为监管技术的东西，就像大坝有装置来调节不同时间点的水流一样，我们将最终建立起对所有系统完整性的概念，明确AI可以在哪里快速流动、在哪里应该放慢、在哪里必须有人类监督。

And making sense of it is going to require building many technologies that you might think of as kind of oversight technologies or, you know, in the same way that a dam has things that regulate, like how much water can go through it at different levels of, different points in time, we're going to end up developing some notion of integrity of all of our systems and where where AI can kind of flow quickly, where it should be slow, where you definitely need human oversight.

在未来几年，这不仅是AI公司，更是所有机构的任务：弄清楚在我们把大量繁琐工作交给代表我们行事的机器之后，这种治理体制应该是什么样子。

And that's gonna be the task of not just for AI companies, but institutions in general in the coming years is figuring out what does this this governance regime look like now that we've given a load of, basically schlep work over to machines that work on our behalf.

那你如何

And how are you

做到这一点呢？

doing it?

你说这是每个人的问题，但你们在应对这个问题上走在了前面，而如果你们搞砸了，后果对你们来说会非常严重。

You said it's everybody's problem, but you're ahead on facing this problem, and the consequences of getting it wrong for you are pretty high.

对吧？

Right?

如果因为你们把编码工作交给了Claude，结果Claude出问题了，那会让Anthropic看起来相当糟糕。

If Claude blows up because you handed over your coding to Claude code, that's gonna make Anthropic look fairly bad.

如果Claude真的把整个文件系统都删了，那对Anthropic来说可就是个糟糕的日子了。

It would be a bad day for Anthropic if if Claude, like, RMRF'd for entire file system.

我不知道那是什么意思，但很好。

I have no idea what that means, but great.

如果Claude删除了代码，那会很糟糕。

If Claude deleted the code, it would be bad.

是的。

Yeah.

看起来很糟糕。

Seems bad.

所以，既然你比我们其他人更早面对这个问题，别把责任推给社会。

So as you're facing this before the rest of us are, like, don't pass the the buck over to society here.

你有什么

What have

你正在做什么？

you what are you doing?

公司内部以及我所管理的团队正在做的最重要事情，就是构建监控系统，来监控所有这些工作现在发生的地方。

The biggest thing that that is happening across the company and on teams that I manage basically building monitoring systems to monitor this all of the different places that the work is now happening.

我们最近发表了一项研究，探讨人们如何使用智能代理，以及人们如何让代理随着时间推移逐渐提交越来越多的代码。

So we recently published research on studying how people use agents and how people let agents kind of push increasingly large amounts of code over time.

你对某个智能代理越熟悉，就越倾向于将任务委托给它。

So the more familiar you get with an agent, the more you tend to delegate to it.

这让我们意识到需要建立各种评估体系来应对这些模式。

That queues us to all kinds of patterns that we need to build systems of evaluation for.

基本上就是说，好吧。

Basically saying, oh, okay.

在这个人与AI系统互动的这个阶段，他们很可能已经将大量工作委托给了系统。

At this person's point of working with the AI system, it's likely that they're massively delegating it.

因此，在这些时刻，我们需要加强所有关于正确性检查的措施。

So anything that we're doing to check correctness needs to be kind of turned up in these moments.

但你所描述的这个世界，是一个由AI代理编写代码、AI代理监督代码、AI代理监督对监督的再监督的系统吗？

But is this world you're talking about, a system where you have AI agents coding, AI agents overseeing the code, AI agents overseeing the meta overseeing of it?

对吧？

Right?

我们的讨论是不是只是在说模型层层递进？

Like, are we just talking about models all the way down?

最终，是的。

Eventually, yes.

我认为我们现在花所有时间在做的事情，就是让这一点对我们变得可见。

And I think that the thing that we are now spending all of our time on is making that visible to us.

一两年前，我们构建了一个系统，能够以保护隐私的方式，查看人们与我们的AI系统进行的对话。

A year or two ago, we built a system that let us, in a privacy preserving way, look at the conversations that people were having with our AI system.

然后我们获得了这张地图——一张关于人们与Claude交谈的所有主题的庞大地图。

And then we gained this map, this giant map of all of the topics that people were talking to Claude about.

第一次，我们可以从整体上看到世界与我们的系统进行的对话。

And for the first time, we could see in aggregate the conversation the world was having with our system.

我们需要构建许多类似的新系统，以提供不同的观察视角。

We're gonna need to build many new systems like that which allow for different ways of seeing.

我刚才提到的这个系统，使我们能够构建一个名为‘Anthropic经济指数’的东西，因为现在我们可以定期发布关于人们与Claude讨论的不同主题及其与各类工作关系的数据，这首次为Anthropic之外的经济学家提供了了解这些系统及其对经济影响的切入点。

And that system that I just named allowed us to then build this thing called the Anthropic Economic Index because now we can release regular data about the different topics people are talking about with Claude and how that relates to different types of jobs, which for the first time gives economists outside Anthropic some hook into these systems and what they're doing to the economy.

公司的重心将越来越多地转向构建监控和监管运行于公司内部的AI系统的机制。

The work of the company is increasingly going to shift to building a monitoring and oversight system of the AI systems running the company.

最终，我们所采用的任何治理框架都可能要求一定程度的透明度和对这些知识系统的访问权限。

And ultimately, any kind of governance framework we end up with will probably demand some level of transparency and some level of access into these systems of knowledge.

因为如果我们认真对待这些AI公司（包括Anthropic）的目标，那就是打造有史以来最强大的技术，并最终将其部署到各个角落。

Because if we take as literal the goals of these AI companies, including Anthropic, it's to build the most capable technology ever, which eventually gets deployed everywhere.

这在我看来，意味着AI最终将与整个世界融为一体，届时你不会希望只有AI公司才了解世界上正在发生的一切。

Well, that sounds a lot to me like an eventually AI becomes indistinguishable from the world writ large, at which point you don't want only AI companies to have a sense of what's going on with the entire world.

所以，政府、学术界和第三方机构都将参与其中。

So it's going to be governments, academia, third parties.

大量公司外部的利益相关者将希望了解正在发生的事情，并在社会层面展开对话，讨论哪些是合适的、哪些让我们感到不安、哪些我们需要更多信息。

A huge set of stakeholders outside the companies are going to want to see what's going on and then have a conversation as society about what's appropriate and what what do we feel discomfort about, what do we need more information about.

等等。

Wait.

我想回到刚才那个话题。

I wanna go back on that.

你是说Anthropic和See My Chats？

You're saying Anthropic and See My Chats?

我们不会查看任何人的聊天内容。

We cannot see no human looks at your chats.

聊天记录会临时存储，用于信任与安全目的，并运行分类器进行分析。

Chats are temporarily stored for trust and safety purposes, running running classifiers over them.

我们可以让Claude阅读并总结这些内容，然后删除。

And we can have Claude read it, summarize it, and toss it out.

所以我们从不查看这些内容，Claude也不会记住它们。

So we never see it, and Claude has no memory of it.

它所做的只是尝试生成一个高度概括的摘要。

All it does is try to write a very high level summary.

比如，如果你在讨论园艺，Claude会将其总结为‘此人正在谈论园艺’，并归入一个名为‘园艺’的类别。

So say you were having a conversation about gardening, Claude would summarize that as this person's talking about gardening, and it leads to a cluster we can see that just says gardening.

但随着时间推移，这感觉可能会陷入许多社交媒体已陷入的令人不快的境地——从人们与系统之间的高度私密互动中收集大量元数据。

This feels though, like, over time, it could get into the quite unpleasant territory a lot of social media has gotten to, where the amount of metadata being gathered from a quite personal interaction people are having with a a system could be a lot.

是的。

Yes.

我的意思是，这里有几点。

I mean, a a couple of things here.

一年前，我们开始思考我们在消费者方面的立场，并采取了不投放广告的立场，因为我们认为人们对于这类事情显然存在焦虑。

A year ago, we started thinking about our our position on on consumer, and we adopted this position of not running ads because we think that's an area that people obviously have have anxieties about with regard to this kind of thing.

除此之外，我们努力向用户展示他们的数据，并在网站上提供一个按钮，让用户下载他们与Claude共享的所有数据，以便他们至少能够查看。

In addition to that, we try and show people their data, and we have a button on the site that lets you download all the data that you shared with Claude so that you can at least see it.

总体而言，我们致力于在数据处理方式上对用户保持极度透明。

Generally, we're trying to be extremely transparent with people about how we handle their data.

最终，在我看来，人们会希望拥有大量可操作的控制选项，我认为我们和其他公司未来会逐步实现这些功能。

And ultimately, the way I see it is people are gonna want a load of controls that they can use, which I think we and others will build out over time.

你有多大的信心，认为当这些模型变得更加复杂时，我们仍能进行这种监控和评估？如果我们真的进入一种情况，即Claude的代码能够自主地以远超软件工程师阅读代码库速度的方式自我改进？

How confident are you that we can do this kind of monitoring and evaluation as these models become more complicated, as if we do enter a situation where Claude code is autonomously improving Claude at a rate faster than software engineers could possibly keep up with reading that code base.

我们已经简要讨论过，你如何看待模型表现出一定程度的欺骗性，以及它们追求自身目标的倾向。

We already talked briefly about how you see the models exhibit some levels of deception, some levels of pursuing their own goals.

我的意思是，Anthropic公司由Chris Ola等人在可解释性方面做了令人惊叹的工作，但这些工作还很初级。

I mean, there's been amazing interpretability work at Anthropic under Chris Ola and others, but it's rudimentary.

因此，你正在使用自己并不完全理解的AI系统来监控另一些你也不完全理解的AI系统，如果事情按照你预期的方向发展，这些系统正在以加速的速度相互增强。

So you're using AI systems you don't totally understand to monitor AI systems you don't totally understand, And the systems are making each other stronger at an accelerating rate if things go the way you think they're gonna go.

你有多大的信心认为我们能够理解这一切？

How confident are you that we're gonna understand that?

这正是多年来人们一直警告的情况——将某些具有轻微不可理解性和不可预测性的系统委托出去。

This is one of the situations which people warned about for years, some form of delegation to systems that have slightly inscrutable and unpredictable aspects.

而这种情况正在发生。

And so this is happening.

我们对此极其重视。

We take this really, really seriously.

我认为，完全有可能构建一个系统，能够完成这里所需工作的绝大部分。

I think it's absolutely possible that you can build a system that does the the vast majority of what needs to be done here.

这个问题具有分形的特性。

This has the property of being a fractal problem.

你知道，如果我想衡量埃兹拉，我可以构建几乎无限多的测量方式来描述你。

You know, if I wanted to measure Ezra, I could build an almost infinite number of measurements to characterize you.

但问题是，我需要在多高的精确度上来衡量你？

But the question is, at what level of fidelity do I need to be measuring you?

我认为我们会达到足以应对安全问题和社会问题的精确度，但这需要公司投入巨大的资源。

I think we'll get to the level of fidelity to deal with the safety issues and societal issues, but it's going to take a huge amount of investment by the companies.

我们必须说出一些让我们自己感到不舒服的话，包括在那些我们对自己系统的能力或局限性尚不明确的领域。

And we're going to have to say things that are uncomfortable for us to say, including in areas where we may be deficient in what we can or can't know about our systems.

Anthropic 在这方面一直有长期的讨论和警示历史，并且一直在努力解决这些问题。

And Anthropic has a long history of talking about and warning about some of these issues while working on it.

我们的总体原则是，通过公开讨论这些问题，也让自己承担责任。

Our general principle is we talk about things to also make ourselves culpable.

这是一个我们必须说得更多的领域。

This is an area where we're going to have to say more.

理论上，我知道这种事情在任何家庭中都可能发生。

In theory, I knew that this kind of thing can happen in any family.

任何人的表亲都可能在密谋谋杀。

Anyone's first cousin could be plotting murder.

这里是UCE四七三五，今天的话题是：正直的公民总是被揭露出是隐藏的罪犯。

This is UCE four seven three five, and today is Upstanding citizens are always turning out to be secret criminals.

与艾伦·盖森的公民对话。

Civic wording with Alan Gessen.

我甚至不会称我的表亲艾伦为杰出公民。

And I wouldn't even call my cousin Alan an outstanding citizen.

你知道，我的客户都是黑帮级别的家伙。

You know, my clients are cartel level guys.

他们都是狠角色。

They're all badasses.

他们他们他们，但知道是一回事，真正永久地解决却是另一回事。

They're they they But it's one thing to know There's a more permanent way to do it.

是吗？

Is it?

是的

Yeah.

越来越不同了。

More and more different.

永久的。

Permanent.

还有另一件事需要明白。

And another thing to understand.

艾伦杀了我。

Alan murder me.

结果比我想象的要糟糕得多。

It ended up being so much worse than I thought I knew.

价格非常合理。

The price is eminently reasonable.

好的。

Okay.

但它到底值什么？艾伦当时在想什么？

But what it's worth What the hell was Alan thinking?

比如说吧

Like, let's just say that

我有点生气了。

I'm a little bit pissed stuff.

你知道吗，我不是那个意思。

You know what I'm No.

我明白了。

I get it.

是的。

Yeah.

由Serial Productions和《纽约时报》出品，我是艾姆·格森，这是《傻瓜》。

From Serial Productions and The New York Times, I'm Em Gessen, and this is The Idiot.

在您收听播客的任何平台收听。

Listen wherever you get your podcasts.

我已经读过足够多关于人工智能超级智能和爆发的恐慌性观点，知道在几乎每一个这样的故事中，关键的转折点都是人工智能系统开始递归自我改进。

I have read enough of the frightened ideas about AI superintelligence and takeoff to know that in almost every single one of them, the key move in the story is that the AI systems become recursively self improving.

嗯哼。

Mhmm.

它们在编写自己的代码。

They're writing their own code.

它们在部署自己的代码。

They're deploying their own code.

它变得越来越快。

It's getting faster.

它们在编写

They're writing

得更快。

it faster.

它们在更快地部署它。

They're deploying it faster.

现在你将进入越来越快的迭代循环。

Now you're going to faster and faster iteration cycles.

你对此感到担忧吗？

Are you worried about it?

你对此感到兴奋吗？

Are you excited about it?

我休完陪产假回来后，为西耶拉做的两个主要项目是：公开发布关于人工智能与经济的更准确信息，以及内部建立更完善的系统，以了解我们在多大程度上自动化了人工智能开发的各个部分。

I came back from paternity leave, and my two big projects for Sierra are better information about AI and the economy that we will release publicly and generating much better information and systems of knowing information internally about the extent to which we are automating aspects of AI development.

我认为目前这还只是以非常边缘的方式发生。

I think right now, it's happening in a very peripheral way.

研究人员的效率正在被提升。

Researchers are being sped up.

人工智能系统正在运行不同的实验。

Different experiments are being run by the AI system.

如果你能完全闭合这个循环，那将极其重要。

It would be extremely important to know if you're fully closing that loop.

我认为我们实际上还有一些技术工作要做，以构建能够监控我们内部开发环境的方法，从而观察趋势的变化。

And I think that we actually have some technical work to do to build ways of instrumenting our internal development environment so that we can see trends over time.

我担心吗？

Am I worried?

我读过你读过的所有内容，当事情开始出错时，这正是故事的关键转折点。

I have read the same things that you have read, and this is the pivotal point in the story when things begin to go awry, if things do.

当我们对此有更完善的数据时，我们会指出这一趋势。

We will call out this trend as we have better data on it.

我认为这是一个需要极其谨慎对待的领域，因为很容易出现将太多事情交由系统处理的情况，一旦系统出错，错误会迅速累积并失控。

And I think that this is an area to tread with, like, extraordinary caution because it's very easy to see how you delegate so many things to the system that if the system goes wrong, the wrongness compounds very quickly and gets away from you.

但让我一直感到担忧的是，每个人都知道。

But the thing that always strikes me and has always struck me as being dangerous about this is everybody knows.

如果我问任何一家公司的员工是否希望在这里保持谨慎，他们会告诉我他们确实希望如此。

And if I ask a member of any of the companies whether or not they wanna be cautious here, they will tell me they they do.

但另一方面，这几乎是他们彼此之间唯一的竞争优势。

On the other hand, it is their almost only advantage over each other.

你们刚刚撤销了 OpenAI 使用 Cloud Code 的权限，因为据我理解，你们认为它确实加快了你们的进度，而你们不希望它也加速他们的进度。

And you all just revoked OpenAI's ability to use Cloud Code because, as best I can tell, you think it is genuinely speeding you up, and you don't want it to speed them up.

在这些力量的权重和威力之间，你们所有人都清楚自己正在玩弄什么，同时又有着极其强烈的要抢先一步的动机。

There is something here between the weight of the forces, the power of the forces that I think you all know you're playing with, and the very, very, very strong incentives to be first.

我真的可以想象自己身处 Anthropic，心想：最好是我们，而不是 OpenAI？

And I I can I can really imagine being inside Anthropic and thinking, well, better us than OpenAI?

最好是我们，而不是 Alphabet、谷歌。

Better us than Alphabet, Google.

最好是我们，而不是中国。

Better us than China.

这成为了一个非常强有力的理由，让我们不能放慢脚步。

And that being a very strong reason to not slow down.

我甚至不确定这个问题你们是否能回答，但你们该如何平衡这一点呢？

I don't even know that this is a question I believe you can answer, but how do you balance that?

嗯，也许我这里有个答案。

Well, maybe I have something of an answer here.

今天，我们的系统以及其他公司的系统都会接受第三方测试，包括政府机构对国家安全特性、生物武器、网络攻击等方面进行评估。

Today, our systems and the other systems from other companies are tested by third parties, including parts of government for national security properties, biological weapons, cyber offense, other things.

这是一个明显的问题领域，世界需要知道这种情况是否正在发生。

It's clearly a problem area where the world needs to know if this is happening.

我认为，如果你随机调查街头任何一个人，向他们解释什么是递归自我改进后，问他们是否认为AI公司应该被允许这样做，

And you almost certainly, I think if you polled any person on the street and said, do you think AI companies should be allowed to do, like, recursive self improvement after explaining what that was?

他们几乎肯定会说：不。

Without checking with anyone, they would say, no.

这听起来相当危险。

That sounds sounds pretty risky.

我希望存在某种形式的监管。

Like, I would like there to be some form of regulation.

但它们可能根本不会出现，或者即使出现也不会那么有力。

But they're probably either won't be or it won't be that strong.

我的意思是，当我与你们这些AI公司高层交谈时，这有时让我感到沮丧：你们竟然天真地寄希望于一种‘机械降神’式的监管，而实际上你们很清楚监管环境是什么样子。

I mean, this actually sometimes frustrates me when I talk to all of you at the top of the AI companies, which is the emergence of, like, a very naive deus ex machina of regulation, where you all know what the regulatory landscape looks like.

目前，主要的争论在于我们是否会完全 preempt 各州对人工智能的监管。

Right now, the big debate is whether or not we're gonna completely preempt any state I AI regulation.

你知道事情进展有多慢。

And you know how slowly things move.

国会到目前为止在这方面根本没有通过任何重大法案，是的。

There has been nothing major passed by congress on this at all Yep.

我会这么说。

I would say.

建立一个所有不同实验室都认可的独立测试与评估系统会很难。

And setting up some kind of independent testing and evaluation system that all the different labs buy into, it would be hard.

这会很复杂。

It would be complicated.

考虑到人们推进的速度以及系统已经表现出的异常行为。

And it is given how fast people are moving and how strange the behaviors the systems are already exhibiting are.

即使你能以极快的速度制定出正确的政策，测试是否能发现一个快速自我改进系统中所有你关心的问题，仍然是一个悬而未决的问题。

Even if you could get the policy right at a high speed, the question of whether or not the testing would be capable of finding everything you want on a rapidly self improving system is a very open question.

我于2021年与我的合著者杰斯·惠特尔斯顿在英国共同撰写了一篇题为《政府应如何及为何监控人工智能发展》的研究论文。

I wrote a research paper in 2021 called how and why governments should monitor AI development with, my co author Jess Whittleston in England.

我认为这里并不在强调因果关系，但在该论文发表后的两年内，美国和英国的AI安全研究所已经开始对实验室的成果进行测试，大致监控其中一些内容。

And I think I'm not attributing a causal factor here, but within two years of that paper, we had the AI safety institutes in The US and UK testing things from the labs, roughly monitoring some of these things.

所以我们确实能做成这件困难的事。

So we we can do this hard thing.

这种事情已经在某一领域发生了。

It has already happened in one domain.

我并不是在依赖某种像看不见的神秘力量之类的东西。

And I'm not relying on some, like, invisible big other force here.

我更想说的是，公司们已经开始在自己的系统中对这些问题进行测试和监控。

I'm more saying that companies are starting to test for this and monitor for this in their own systems.

仅仅拥有一个非监管性质的外部测试，来验证你是否真正进行了相关测试，就非常有帮助。

Just having a nonregulatory external test of whether you truly are testing for that is is extremely helpful.

你认为我们在测试方面做得足够好吗？

And do you think we're good enough at the testing?

我的意思是，我持怀疑态度的原因并不是我不认为我们能建立一个声称是测试的机制。

I mean, I think one reason I am skeptical is not that I don't think we can set up something that claims to be a test.

正如你所说，我们已经做到了。

As you say, we have done that already.

而是相比投入在加速这些系统上的资源，投入在测试上的资源实在太少；而且我已经看到Anthropic的报告称，Claude可能知道自己正在接受测试，并据此调整行为。

It is that the resources going into that compared to the resources going into speeding these systems and already, I am reading Anthropic reports that Claude maybe knows when it's being tested and alters its behavior accordingly.

所以，在一个越来越多代码由Claude编写、而越来越少被人类理解的世界里。

So a world where more of the code is being written by Claude and less of it is being understood.

我只是清楚资源流向了哪里。

I just know where the resources are going.

它们似乎并没有流向测试这一侧。

They don't seem to be going into the testing side.

我见过我们在两到两年半的时间内，从零建立起一个被普遍认为有效的生物武器测试体系。

I've seen us go from zero to having what I think people generally feel is an effective bioweapon testing regime in maybe two years, two and a half.

所以这是可以做到的。

So it can be done.

这确实很难，但我们已经有了一个实例。

It's really hard, but we have a proof point.

所以我认为我们能够达到这一目标，今年你们会听到我们更多地谈论我们如何开始构建针对这一领域的监控和测试机制。

So I think that we can get there, and you should expect us to kind of speak more about this this year about precisely how we're starting to try and build monitoring and testing things for this.

我认为，这是我们和其他AI公司需要更加公开地分享我们发现的领域。

And I think this is an area where we and the other AI companies will need to be, significantly more more public about what we're finding.

我们并不是不公开信息。

We're not we're not not being public now.

这些信息都包含在模型卡片等你可以查阅的材料中。

It's in the model cards and things that you can really read.

但显然，有人开始阅读这些内容并说：等等。

But clearly, are starting to read this and say, hang on.

这看起来相当令人担忧，他们正期待我们提供更多的数据。

This looks like quite concerning, and they are looking to us to produce more data.

现在我想回到初级职位的问题上。

I wanna go back now to the entry level jobs question.

你的首席执行官达里奥·阿马德表示，他认为在未来几年内，人工智能可能会取代一半的初级白领工作。

Your CEO, Dario Amade, has said that he thinks AI could displace half of all entry level white collar jobs in the next couple of years.

当我看到相关报道时，我总觉得人们忽略了‘初级’这个关键词。

I always think that that the people sort of miss the entry level language there when I see it reported on.

但首先，你同意这个观点吗？

But first, do you agree with that?

你是否担心在未来几年内，一半的初级白领工作会被取代？

Do you worry that half of all entry level white collar jobs can be replaced in the next couple of years?

我相信这项技术将渗透到广泛的知识经济领域，并影响大多数初级岗位。

I believe that this technology is gonna make its way into the broad knowledge economy, and it will touch the majority of entry level jobs.

这些岗位是否真的会发生变化，这是一个更加微妙的问题，从现有数据中并不容易得出明确结论。

Whether those jobs actually change is a much more, like, subtle question and it's not obvious from the data.

比如，如果我们看看当前发布的一些数据，或许能发现研究生招聘放缓的迹象，也可能看到生产率提升的端倪，但这还非常早期，很难做出明确判断。

Like, we maybe see the hints of a slowdown in graduate hiring, maybe, if you look at some of the data coming out right now, we maybe see the signatures of a productivity boom, but it's very, very early and it's hard to be definitive.

但我们确知，所有这些岗位都将发生变化。

But we do know that all of these jobs will change.

所有初级职位最终都会发生变化，因为人工智能使某些事情成为可能，并将改变公司的招聘计划。

All of the entry level jobs are eventually going to change because AI has made certain things possible, and it's going to change for hiring plans of companies.

因此，作为一群人才，你可能会看到初级职位的招聘机会减少。

So as a cohort, you might see fewer job openings for entry level jobs.

这将是所有人对这一现象的一种天真预期。

That would be one naive expectation out of all of this.

但让我们谈谈，这甚至可能根本不是一种天真的预期。

But let's talk about that maybe not even being a naive expectation.

你说在Anthropic，你已经看到了这种情况的发生

You say it's already happening at Anthropic that what you're seeing

我看到的是我们偏好的转变。

I'm seeing a shift our preference.

没错。

Exactly.

我猜测，这种情况在其他地方也会发生。

And I I my guess is that that would be happening elsewhere.

当然。

Of course.

就我们现在所处的位置而言，就我使用这些系统的方式来看，我认为Claude、ChatGPT、Gemini或其他任何系统都很少能超越某个领域中最优秀的人。

And where we are right now, I mean, even in the way I use some of these systems, it is rare, I think, that Claude or ChatGPT or Gemini or any of the other systems is better than the best person in a field.

嗯。

Mhmm.

它们还没有在这方面实现重大突破，而且它们还不能做很多事。

It has not heavily breached that, and there's all kinds of things they can't do.

但它们比普通大学毕业生更强吗？

But are they better than your median college graduate Mhmm.

在很多方面都是吗？

At a lot of things?

是的。

Yeah.

它们确实更强。

They are.

在一个对普通大学毕业生需求减少的世界里，我看到人们争论的一个问题是，这些系统目前是否能做得比一般或替代水平的工作更好。

And in a world where you need fewer of your median college graduates, One thing I've seen people arguing about is whether these systems at this point can do better than sort of average or replacement level work.

但我每次看到这种说法都会非常担忧，因为一旦我们接受它们可以胜任一般或替代水平的工作，那么根据定义，大多数工作和大多数从事这些工作的人都是普通的。

But I always really worry when I see that because once we have accepted they can do average or placement level work, well, by definition, most of the work done and most of the people doing it Is average.

是普通的。

Is average.

对吧？

Right?

最优秀的人只是少数例外。

The best people are the exceptions.

而且，人们变得更好的方式，是通过那些能让他们学习的工作。

And, also, the way people become better is that they have jobs where they learn.

嗯。

Mhmm.

我的职业生涯中，花了很多时间招聘年轻的记者。

I mean, I have spent a lot of time hiring young journalists over my career.

当你从大学招聘人员时，在某种程度上，你是在根据他们当时可能写出的文章和工作来雇佣他们。

And when you hire people out of college, to some degree, you're hiring them for their possible articles and work at that exact moment.

是的。

Mhmm.

但某种程度上，你也在对他们进行投资，相信他们的能力会随着时间推移不断进步，最终带来回报。

But to some degree, you're making a investment in them that you think will only pay off over time as they get better and better and better.

因此，这个世界中，人工智能对初级岗位产生了真实的影响，而这种情形对我来说并不遥远，它引发了关于人口技能提升、未来如何培养高级岗位人才，以及人们在成长过程中错过了哪些学习机会的深刻问题。

So this world where you have a a potential real impact on entry level jobs, and that that world does not feel far away to me, seems to me to have really profound questions it is raising about the upskilling of the population, how you end up with people for senior level jobs down the road, what people aren't learning along the way.

我们观察到，有一类年轻人已经连续几年沉浸并深入使用人工智能。

And one thing we see is that there is a certain type of young person that has just lived and breathed AI for several years now.

我们会雇佣他们。

We hire them.

他们非常出色，能够以完全新颖的方式思考如何让Claude为他们工作。

They're excellent and they think in entirely new ways about basically how to get Claude to work for them.

这就像那些从小在互联网环境中长大的孩子。

It's like kids who grow up on the Internet.

他们天生就精通这一点，而他们所加入的许多组织中的人却并不具备这种能力。

They they were naturally versed in it in a way that many people in the organizations they were coming into weren't.

因此，弄清楚如何教授这种基本的实验思维和对这些系统的探索精神，并加以鼓励，将变得至关重要。

So figuring out how to teach that basic experimental mindset and curiosity about these systems and to encourage it is going to be really important.

花大量时间摆弄这些东西的人会培养出非常宝贵的直觉，他们进入组织后能够变得极其高效。

People that spend a lot of time playing around with this stuff will develop very valuable intuitions, and they will come into organizations and be able to be extremely productive.

同时，我们还得弄清楚，我们希望培养哪些手工技能，或许可以发展一种类似行会的理念来维持人类的卓越，并思考组织该如何教授这些技能。

At the same time, we're gonna have to figure out what artisanal skills we want to almost develop maybe a guild style philosophy of maintaining human excellence in and how organizations choose how to teach those skills.

好的。

Okay.

那那些处于中间位置的人怎么办？

Then what about all those people in the middle of that?

在硅谷以外的现实经济中，事物的进展总是很缓慢。

Things move slowly in the real economy outside Silicon Valley.

我认为我们常常看到软件工程，就以为它能代表整个经济的运作方式，但事实往往并非如此。

I think that we often look at software engineering and think that this is a proxy for how the rest of the economy works, but it's often not.

这常常是一种不类比。

It's often a disanalogy.

组织会把人员调配到人工智能系统尚未发挥作用的地方。

Organizations will move people around to where the AI systems don't yet work.

我认为，就业结构不会立即发生巨大变化，但人们被要求从事的工作类型将会发生显著变化。

And I think that you won't see vast immediate changes in the makeup of employment, but you will see significant changes in the types of work people are being asked to do.

最擅长调动人员的组织将变得极其高效。

And the organizations which are best at sort of moving their people around are going to be extremely effective.

而不这么做的一些组织，最终可能不得不做出非常艰难的决定，比如裁员。

And ones that don't may end up having to make like really, really hard decisions involving involving laying off workers.

人工智能这一领域的不同之处在于，它可能比以往的技术发生得快得多。

The difference with this AI stuff is it maybe happens a lot faster than previous technologies.

我认为，包括Anthropic在内的许多人对此的焦虑是：这种速度会不会带来根本性的差异？

And I think many of the anxieties people might have about this, including at Anthropic, is is the speed of this going to make all of this difference?

它是否引入了我们此前从未遇到过的临界点？

Does it introduce sheer points that we haven't encountered before?

如果你要打个赌，三年后大学毕业生的失业率会和现在一样吗？

If you had to bet three years from now, is the unemployment rate for college graduates, is it the same as it is now?

会更高吗？

Is it higher?

会更低吗？

Is it lower?

我猜会略高一点。

I would guess it is higher but not by much.

我的意思是，今天有些专业领域已经因为人工智能的介入而彻底改变，甚至可能以不利于那些拥有特定专长的人的方式重塑了就业市场。

And what I mean by that is there'll be some disciplines today which actually AI has come in and and completely changed and completely changed the structure of that employment market, maybe in a way that's adverse to to people that have that specialism.

但大多数情况下，我认为三年后，人工智能将推动整个经济实现显著增长。

But mostly, I think three years from now, AI will have driven a pretty tremendous growth in the entire economy.

因此，你会看到许多新的工作类型因这一变化而出现，而这些是我们目前无法预测的。

And so you're going to see lots of new types of jobs that show up as a consequence of this that we can't yet predict.

而且我预计，毕业生们会大量涌入这些新岗位。

And you will see graduates kind of flood into that, I I expect.

你知道无法预测这些新工作，但如果你要猜猜它们可能是什么样子，会怎么想？

Do you have a I know you can't predict those new jobs, but if you had to guess what some of them might look like.

一方面，就是微型创业者这种现象。

I mean, one thing is just the phenomenon of the the kind of micro entrepreneur.

现在有很多方式可以在线创业，而AI系统能帮你完成大部分工作，让你无需雇佣大量人员来处理启动企业所需的繁重琐事。

I mean, there are lots and lots of, ways that you can start businesses online now which are just made massively easier by having the AI systems do it for you, and you don't need to hire a whole load of people to help you do the huge amounts of schlep work that involves getting a business off the ground.

如果你有一个清晰的想法和明确的商业愿景，那么现在正是创业的最佳时机，你只需极少的成本就能快速起步。

It's more a case of if you're a person with a clear idea and a clear vision of something to do a business in, it's now the best time ever to start a business and you can get up and running for pennies on the dollar.

我预计我们会看到大量具有这种特征的新事物涌现。

I expect we'll see tons and tons and tons of stuff that has that nature to it.

我还预计会看到一种‘AI对AI’经济的出现，即AI代理和AI企业之间相互进行交易，同时会出现一些人找到方法，通过一些奇特的新组织从中获利。

I also expect that we're gonna see the emergence of what you might think of as the AI to AI economy, where AI agents and AI businesses will be doing business with one another, and we'll have people that have figured out ways to basically profit off of that in the forms of strange new organizations.

比如，一家专门从事AI对AI法律合同的公司会是什么样子？

Like, what would it look like to have a firm which specializes in AI to AI legal contracts?

因为我打赌，今天就已经有人能想出创造性的方法来启动这样的业务。

Because I bet you there's a way that you can figure out creative ways to start that business today.

会有大量类似的东西出现。

There'll be a lot of stuff of that flavor.

所以，我既担心又认为最有可能的情况是，如果你告诉我，Anthropic将在一年后发布Claude Plus。

So the version of this that I both worry about and think to be the likeliest, if you told me what was gonna happen was that Anthropic was gonna release Claude plus in a year.

而Claude Plus会成为一个完全成型的同事，能够端到端地模仿各种职业的技能，甚至达到高管级别。

And Claude plus is somehow a fully formed coworker, and it can mimic end to end the skills of a lot of different professions up to the c suite level.

是的。

Mhmm.

而且这一切会突然发生，给企业带来巨大的即时压力，迫使它们裁员以保持竞争力。

And it's gonna happen all at once, and it's gonna create tremendous all at once pressure for businesses to downsize, to remain competitive with each other.

从政策层面来看，如果这种颠覆性变化像COVID那样一次性爆发，人人都待在家里，我反而没那么担心。

At a policy level, the fact that that would be so disruptive in that big bang, everybody stays home because of COVID style way, it worries me less.

因为当事情成为紧急状况时，我们会做出反应。

Because when things are emergencies, we respond.

我们确实会制定政策。

We actually do policy.

但如果你告诉我，未来会发生的是市场营销毕业生的失业率会上升175%到300%，但即便如此，失业率也不会高到哪里去。

But if you told me that what's gonna happen is that the unemployment rate for marketing graduates is going to go up by, you know, a 175%, 300% to still not be that high.

我的意思是，在大衰退期间，整体失业率最高也就徘徊在9%左右。

I mean, the overall employment rate during the Great Recession topped around, you know, in the nine ish percentile range.

所以，你完全可以经历巨大的冲击，而不会让一半的人失去工作。

So you can have a lot of disruption without having 50% of people thrown out of work.

对吧？

Right?

如果你有10%、15%的失业率，我的天，那已经是非常非常高了。

If you have 10%, 15%, I mean, that's very, very, very high.

但它还没高到那种地步。

But it's not so high.

而且如果这种变化只发生在少数几个行业，影响的只是毕业生，而不是整个行业的所有人，那也许只是你不够优秀。

And if it's only happening in a couple of industries at a time and it's grads, not everybody in the industry being thrown out of work, well, maybe it's just that you're not good enough.

是的。

Yep.

对吧？

Right?

你知道，那些顶尖的、非常优秀的毕业生仍然能找到工作。

You know, the the superstars, really good graduates are still getting jobs.

你应该更努力一些。

You should have worked hard.

你应该去一所更好的学校。

Should have gone to a better school.

我担心的是，我们对这种类型的失业反应不佳，就像来自中国带来的那种失业，这种失业似乎更有可能发生，因为它不均衡，而且发生的速度让我们还能把个人的命运归咎于他们自己。

And one of my worries is that we don't respond to that kind of job displacement well, right, which is the kind of job displacement we got from China, which is the kind of job displacement that seems likelier because it's uneven, and it's happening at a rate where we can still blame people for their own fortunes.

我很想知道你是怎么看待这个说法的。

I'm curious how you think about that story.

我认为默认的结果就是你所描述的那样。

I think the default outcome is something like what you described.

但实现这个结果其实是一种选择，我们可以做出不同的选择。

But getting there is actually a choice, and we can make different choices.

我们发布Anthropic经济指数的全部目的，是为了获得与职业、与经济中实际工作相关联的数据。我们这样做是有意为之，因为这能随着时间推移绘制出AI如何逐步渗透到不同工作岗位的图景，并帮助Anthropic之外的经济学家将这些信息整合起来。

The whole purpose of what we release in the form of the Anthropic Economic Index is the ability to have data that ties to occupations, that tie to real jobs in the economy, we do that very intentionally because it is building a map over time of how this AI is making its way into different jobs and will empower economists outside Anthropic to tie it together.

我相信，如果我们能对工作变动或转型的原因做出更扎实、更有证据支持的论断，我们就能在政策上做出不同的选择。

I believe that we can choose different things in policy if we can make much more well evidenced claims about what the cause of a job disruption or change is.

而我们面临的挑战是：我们能否充分描述这个新兴的AI经济，使其变得如此清晰明确？

And the challenge in front of us is can we characterize this emerging AI economy well enough that we can make this extremely stark?

然后，我认为我们才能真正就这个问题展开政策讨论。

And then I think that we can actually have a policy discussion about it.

那我们来谈谈政策讨论吧。

Well, let's talk about the policy discussion.

对。

Yep.

我之所以特别邀请你来，其中一个原因是你曾在OpenAI从事政策工作。

One reason I wanted to have you in particular on is you did policy at OpenAI.

对。

Yep.

你在Anthropic做政策工作。

You do policy at Anthropic.

所以你一直参与这些政策辩论。

So you've been around these policy debates for a long time.

你长期以来一直在你的通讯中追踪模型的能力。

You've been tracking model capabilities in your newsletter for a long time.

我的感觉是，我们已经就人工智能和就业问题展开了多年的讨论。

My perception is we are many, many years into the debate about AI and jobs.

嗯。

Mhmm.

早在ChatGPT出现之前很久，就已经有多年的讨论了。

Many, many years dating far before ChatGPT

嗯。

Mhmm.

关于人工智能和就业，我们该怎么做，早在阿斯彭和其他地方就举办过许多会议。

Of there being conferences at Aspen and everywhere else about, you know, what are we gonna do about AI and jobs?

但不知为何，我依然看不到任何切实可行的政策，能够应对我刚才描述的那种情况——如果突然间，大量行业的初级岗位都变得极其难找。

And somehow, I still see almost no policy that seems to me to be actionable if the situation I just described begins showing up, where all of a sudden entry level jobs are getting much harder to come by across a large range of industries all at once Mhmm.

以至于经济无法将这些市场营销专业的毕业生重新安置到数据中心建设、护理或其他行业。

Such that the economy cannot reshift all these marketing majors into data center construction or nurses or something.

嗯。

Mhmm.

所以，好吧，你比我更深入地参与了这场讨论。

So, okay, you've been deeper in this conversation than I've been.

当你说到我们可以就这个问题展开政策讨论时，其实我们已经在进行政策讨论了。

When you say we can have a policy conversation about that we've been having a policy conversation.

但我们有具体的政策吗？

Do we have policy?

我们对人工智能对经济和就业的影响感到普遍焦虑。

We have generalized anxiety about the effect of AI on the economy and on jobs.

但我们没有清晰的政策建议。

We don't have clear policy ideas.

这部分原因是，民选官员并不单单或主要被高层次的政策讨论所影响。

Part of that is that elected officials are not moved solely or mostly by the high level policy conversation.

他们更关注选民身上发生的事情。

They're moved by what happens to their constituents.

就在几个月前，我们才刚刚能够提供各州层面的经济指数数据，现在你才能开始进行政策讨论。

Only a few months ago were we able to produce state level views for our economic index, And now you can start having the policy conversation.

我们曾与民选官员讨论过这一点，现在我们可以这样说：哦，你是来自印第安纳州的。

And we've had this with elected officials where now we can say, oh, you're from you're from Indiana.

比如，这是你州内人工智能的主要应用领域，我们可以将其与主要就业来源结合起来。

Like, here's for, like, major uses of AI in your state, and we can join it with major sources of employment.

我们开始看到的是，这能激发他们的行动，因为这将政策与选民联系起来，而选民会把这一切与政客的作为挂钩：你做了什么？

And what we're starting to see is that activates them because it makes it tied to their constituents who are going to tie it to the the politician of what did you do.

对于这个问题，你需要采取一种多层次的应对方式，从延长失业救济（特别是针对那些我们知道将受到最严重影响的职业）到考虑学徒计划等措施。

Now, what you do about this is going to need to be an extremely kind of multilayered response ranging from extending unemployment for especially occupations that we know are going to be hardest hit to thinking about things like apprenticeship programs.

随着情况变得越来越严重，你可能还需要扩展到更广泛的社会项目，比如补贴那些你希望人们转向的经济领域中的工作岗位——而这只有在经历由显著经济增长带来的富足时才可能实现。

And then as the scenarios get more and more significant, you may extend to much larger social programs or things like subsidizing jobs in the parts of the economy where you want to move people to that you're only able to do if you experience the kind of abundance that comes from significant economic growth.

但经济增长可能通过为一些你能做的事情提供资金来帮助解决这些其他政策挑战。

But the economic growth may help solve some of these other policy challenges by funding some of the things you can do.

我总是觉得这个答案令人沮丧。

I always find this answer depressing.

我要说实话。

I'm gonna be honest.

失业是一种非常糟糕的处境。

Unemployment is a terrible thing to be on.

我们需要这个项目，但领取失业救济的人并不为此感到开心。

It's a program we need, but people on unemployment are not happy about it.

是的。

Mhmm.

这也不是任何人的长期解决方案。

And it's not a good long term solution for anybody.

学徒再培训项目，其成效并不理想。

Apprentice retraining programs, they don't have great track records.

我们在帮助那些因制造业岗位外流而失业的人再培训方面做得并不好。

We were not good at retraining people out of having their manufacturing jobs outsourced.

我不是说从理论上讲我们不可能在这方面变得更好，但我们必须迅速提升这方面的能力。

I'm not saying it is conceptually impossible that we could get better at it, but we would need to get better at it fast.

嗯。

Mhmm.

而我们并没有投入足够的实践、实验、制度建设或能力提升来实现这一点。

And we have not been putting in the reps or the experimentation or the institution or capacity building to do that.

至于大规模社会福利改革这个更广泛的问题，我觉得这似乎非常困难。

And the the broader question of big social insurance changes doesn't seem I mean, that seems tough to me.

我想就这一点再深入探讨一下。

I wanna push on this Please.

就在我们明知有一种干预措施，能比其他任何方法都更有效地帮助人们应对经济变化时。

Just a bit where we know that there is one intervention that helps people dealing with, like, a changing economy more than almost anything else.

那就是时间。

It is just time.

给人们时间去寻找本行业的工作，或寻找互补性的工作。

Giving the person time to find either a job in their industry or to find a job that's complementary.

如果人们没有时间，他们就会接受更低工资的工作。

If people don't have time, they take lower wage jobs.

他们会脱离原本所处的经济阶层，跌落下去。

They fall out of their whatever economic rung they're on, they fall down at.

能够给予人们时间去寻找工作的政策干预，我认为是一种非常有效的措施，而且在政策制定中，有很多可以调整的杠杆可以发挥作用。

Policy interventions that can just give people time to search is, I think, a a robustly useful intervention and one where for many, like, dials to turn in a policy making sense that you can use.

我认为这一点得到了大量经济学文献的有力支持。

And I think this is just well supported by lots of the economic literature.

所以我们已经有了这一点。

So we have that.

如果我们最终陷入像你提到的那些更极端的情境中，我认为这将促使我们展开关于如何应对这项技术的更大范围的国家对话，而这种对话已经开始出现了。

Now if we end up in a more extreme scenario like some of the ones that you're talking about, I think that will just bring us to the larger national conversation on what to do about this technology, which which is beginning to happen.

如果你看看各州以及州一级层出不穷的立法，是的，并非所有政策都恰好是正确的应对方式，但它们确实表明了人们希望就这一问题展开更系统、更连贯的对话。

If you look at the states and the flurry of legislation at the at the state level, yes, not all of it is, like, the exactly the right policy response, but it is indicative of a a desire for there to be some larger coherent conversation about this.

我认为，时间是描述这个问题的一个很好的方式，因为我同意你的观点。

Well, I think time is a really good way of describing what the question is because I agree with you.

我的意思是，当我说失业保险不是一个理想的社会福利项目时，并不是说人们不需要它。

I mean, when I say unemployment insurance isn't a great program to be on, I don't mean people don't need to be on it.

是的。

Yep.

我的意思是，他们希望尽快离开这个项目。

I mean, they wanna get off of it.

当然。

Absolutely.

因为人们想要通过工作获得收入。

Because people for they want money from jobs.

他们想要尊严。

They want dignity.

他们想要与他人相处。

They want to be around other human beings.

通常情况下，当你帮助人们争取时间时，你是在帮他们度过一段有明确期限的中断期。

Usually, what you're doing when you are helping people buy time is you're helping them wait out a time delimited disruption.

嗯哼。

Mhmm.

并不总是如此。

Not always.

对吧？

Right?

中国冲击并不完全如此，但你预期它会过去，然后市场会恢复正常。

The China shock wasn't exactly like that, but that you expect to pass, and then the the market is sort of normal.

在这种情况下，你面对的是一种技术，如果希望发生的事情真的发生了，这种技术正在加速发展。

In this case, what you have is a technology that if what you wanna have happen happens, the technology is accelerating.

嗯哼。

Mhmm.

所以，你现在看到的是三种不同的速度在同时发生。

So what you have is, like, three different speeds happening here.

你有个人调整的速度。

You have the speed at which individual people can adjust.

我多快能学会新技能，适应新世界，学习人工智能，不管它是什么？

How fast can I learn new skills, figure out a new world, learn AI, whatever it might be?

你有AI系统的发展速度，几年前这些系统还无法胜任优秀大学毕业生的工作，还有政策的速度。

You have the speed at which the AI systems, which a couple of years ago were not capable of doing the work of a median college grad from a good school, and you have the speed of policy.

系统变得越来越好、能做更多事情的速度非常快。

And the speed at which the systems are getting better and able to do more things is quite fast.

我的意思是，你比我能更深刻地体会到这一点，但我发现连跟进都很难，因为你知道，三个月内，又会有新东西出现，彻底改变可能实现的事情。

I mean, that is you you experience this more than I do, but I find it hard to even cover this because, you know, within three months, something else will have come out that has significantly changed what is possible.

我最近有了一个孩子，休完陪产假回来后，发现我们构建的新系统。

I had a baby recently and came back from paternity leave to the new systems we built.

我感到非常惊讶。

Was deeply surprised.

个体人类的适应速度比这慢得多，而政策和政府机构的运作速度比个体人类还要慢得多。

Individual humans are moving more slowly than that, and policy and government institutions move a lot more slowly than individual human beings.

因此，通常来说，时间会像你说的那样 favor 工作者。

And so, typically, the the intervention is that time favors the worker, as you're saying.

而在这里，会帮助工作者。

And here, will help the worker.

但我认为令人担忧的问题是，时间是否实际上只是为破坏加剧争取了更多时间。

But I think the scary question is whether time just actually creates time for the disruption to get worse.

你知道的。

You know?

也许你原本想转去做数据中心建设，但事实上现在我们不需要那么多数据中心建设了。

Maybe you wanted to move over to data center construction, but actually now we don't need as much data center construct.

对吧？

Right?

你可以这样理解。

Like, you can think of it like that.

我的意思是，在你所描述的情况下，经济将会非常火热。

I mean, under the situation you're describing, the economy will be running extremely hot.

这些人工智能系统将产生巨大的经济活动。

Huge amounts of economic activity will be generated by these AI systems.

在大多数这种情况下，我认为GDP不会保持不变或萎缩。

And under most scenarios where this is happening, I don't think you're going to be seeing GDP stay the same or or shrink.

对吧？

Right?

它将显著增长。

It's going to be getting substantially larger.

我认为西方已经很久没有经历过重大的GDP增长了，所以我们渐渐忘记了这种增长在政策制定上能带来什么。

I think we just haven't experienced major GDP growth in the West in a long time, and we sort of forget what that affords you in a policy making sense.

我认为我们可以开展许多大型项目来创造新型工作，但这需要经济规模变得极其庞大，从而为这些项目腾出空间。

I think that there are huge projects that we could do that would allow you to create new types of jobs, but it requires the economic growth to be so kind of profoundly large that it creates space to do those projects.

正如你对丰裕运动的研究所深入了解的那样，这需要社会有意愿相信我们能够建造东西，并且愿意去建造。

And, you know, as you're deeply familiar with with with your work on on the abundance movement, it requires for, like, social will to believe that we can build stuff and to want to build stuff.

但我认为这两者都可能会随之而来。

I But think both of those things might come along.

我认为，由于这种巨大的经济增长，我们可能会进入一个非常令人兴奋的场景，能够自主决定如何分配社会中的巨大资源。

I think that we could end up being in a pretty exciting scenario where we get to choose how to allocate, like, great efforts in in society due to this large amount of economic growth that has happened.

这将迫使人们展开一场关于‘这不是暂时现象’的讨论，我想你正是在暗示这一点，而这也是最难向政策制定者传达的：这项技术并没有一个自然的终点。

That is going to require the conversation to be forced about this isn't temporary, which I think is what you're gesturing at, and is in a sense, the hardest thing to communicate to policymakers is there isn't a there isn't a natural stopping point for this technology.

它会持续变得更好，它带来的变化也将与社会其他部分不断叠加。

It's gonna keep getting better, and the changes it brings are going to keep compounding with the rest of society.

因此，这需要政治意愿发生转变，需要人们重新愿意去接纳那些我们许久未曾考虑过的事物。

So that will need to create a change in in political will and a willingness to entertain things which we haven't in some time.

所以现在我想换个角度，来反问你一个问题。

So now I wanna flip it, the question I'm asking.

你提到了富足。

You brought up abundance.

我在从事这项工作中学到的一点是，我绝对不认为社会中缺乏的是改善事物的好点子，我们的政策之所以不够好，并不是因为政策库枯竭了。

One of the things I have learned doing that work is that it is certainly not my view that what is scarce in society is ideas for better ways of doing things, that our policy isn't better than it is because our policy cupboard is dry.

是的。

Mhmm.

对吧？

Right?

这不对。

That's not true.

我们有很多好的政策。

We have lots of good policies.

我能举出一大堆例子。

I could name a bunch of them.

但以我们目前的政治体制，这些政策很难推行下去。

They're very hard to get through our political systems as they're currently constituted.

人工智能未来最无吸引力的版本，是一个你只是创造了一种方式，把年轻的白领工作者赶出工作岗位，用普通水平的人工智能取代他们的世界。

The least inspiring version of the AI future is a world where what you have done is create a way to throw young white collar workers out of work and replace them with average level AI intelligence.

更令人兴奋的版本，用达里奥的比喻来说，是数据中心里的天才。

The more exciting version, to use Dario's metaphor, is geniuses in a data center.

嗯。

Mhmm.

我认为这确实令人兴奋。

And I do think that's exciting.

当我听到你或他谈论，比如如果我们每年实现10个百分点的GDP增长，或者20个百分点的GDP增长时，我不禁想问。

And I wonder when I hear him or you talk about, well, what if we had 10 percentage point GDP growth year on year, 20 percentage point GDP growth year on year?

我好奇，我们面临的许多问题，是否真的只是受限于创意层面。

I wonder how many of our problems are really bounded at the ideas level.

对吧？

Right?

我们现在就可以去找诺贝尔奖得主，问他们：这个国家我们应该怎么做？

We could go to Nobel Prize winners right now and say, what should we do in this country?

他们中的很多人能提出一些我们目前尚未实施的好点子。

And a lot of them could cause some good ideas that we are not currently doing.

有时我会担心，或者根据我在其他问题上的经验感到疑惑：我们是否高估了自身面临的障碍——以为阻碍我们走向繁荣经济的，主要是缺乏足够的智慧和由此产生的创意，而实际上，我们执行能力的削弱更为严重。

I do worry sometimes or wonder, given my experience on other issues, whether we have overstated to ourselves how much of what stands between us and the expanding abundant economy we want is that we don't have enough intelligence and the ideas that that intelligence could create versus our actual ability to implement things is very weakened.

而人工智能将加剧这一瓶颈，因为会有更多东西被推给系统去执行，包括愚蠢的点子、错误信息和垃圾内容。

And what AI is gonna create is a larger bottlenecks around that because there'll be more being pushed at the system to implement, including dumb ideas and disinformation and slot.

对吧？

Right?

就像，它在账目的另一侧也会有一些东西。

Like, it'll have things on the other side of the ledger too.

你怎么看待这些限制因素？

How do you think about these rate limiters?

这里有一个有趣的教训，来自人工智能公司或一般公司，尤其是科技公司：通常，新想法来自公司内部创建所谓的‘创业中的创业’，也就是把那些随着时间推移积累起来的后台官僚主义或繁琐工作剥离出来，然后对公司内部一个很小的团队说：你们什么都不用管。

There's kind of a funny lesson here from the AI companies or companies in general, especially tech companies, where often new ideas come out of companies by them creating what they always call the startups within a startup, which is basically taking whatever process has, like, built up over time leading to back end bureaucracy or schlep work and saying to a very small team inside the company, you don't have any of this.

去干点事吧。

Go and do some stuff.

这就是像Claude Code和其他一些东西是如何被创造出来的。

And and this is, you know, how things like Claude Code and other stuff get created.

目前开始浮现的一些想法是：在更广泛的经济体系中，创建一种无许可的创新结构会是什么样子？

Ideas that kind of are starting to float around are what would it look like to sort of create that permissionless innovation structure in in the larger economy?

这真的非常困难，因为它还有一个额外的特性：经济体系与民主制度是相互关联的。

And it's really, really hard because it has the additional property that, you know, economies are are linked to democracies.

民主制度需要权衡众多民众的偏好，而所有政治都是地方性的。

Democracies weigh the preferences of many, many people, and all politics is local.

因此，正如你在基础设施建设中所遇到的那样，如果你想建立一个无许可的创新体系，就会遇到产权和人们偏好这类问题，这时你就陷入了一个无法解决的困境。

So often, as you've encountered with infrastructure build outs, if you want to create a permissionless innovation system, you run into things like property rights and what people's preferences are, and now you're in an intractable place.

但我的感觉是，这正是我们不得不面对的主要问题。

But my sense is that's the main thing that we're going to have to confront.

而人工智能可能给我们带来的一个优势是，如果运用得当，它本质上是一种能吞噬官僚体系的机器；但如果用得不好，它也可能成为制造官僚体系的机器？

And the one advantage that AI might give us is it is kind of a native bureaucracy eating machine if done correctly, or a bureaucracy creating machine if done badly?

你有没有看到，有人创建了一个系统，你可以把附近新开发项目的文件输入进去？

Did you see that somebody created a a system that basically you feed it in the documents of a new development near you?

哦，它会自动生成环境评估报告？

Oh, and it writes environmental review things?

或者

Or

它会写出极其复杂的反对意见。是的。

It writes incredibly sophisticated challenges Yep.

在你可以提出质疑的每一个代码层面。

Across every level of the code that you could possibly challenge on.

所以，大多数人并没有足够的钱去聘请一家非常专业的律师事务所，来阻止隔壁建公寓楼。

So most people don't have the money when they wanna stop an apartment building from going up down the block to hire a very sophisticated law firm to figure out how to stop that apartment building.

但基本上，这在规模上实现了这一点。

But, basically, this created that at scale.

所以，正如你所说，它既能吞噬官僚体系，也能极大地强化官僚体系。

And so as you say, right, it could eat bureaucracy, could also supercharge bureaucracy.

是的。

Yep.

AI的每一件事都有其另一面。

It's for everything in AI has the other side of the coin.

我们有一些客户使用我们的AI系统，大幅缩短了提交新药候选材料所需的时间，时间被大幅削减。

We have customers that have used our AI systems to massively reduce the time it takes them to produce all of the materials they need when they're submitting new new drug candidates, and it's cut that time massively.

这正是你刚才所描述的镜像世界版本。

It's the Mirror World version of what you just described.

我对这个问题没有简单的答案。

I don't have an easy answer to this.

我认为，只有当这种情况明显演变为危机，并且能够在社会层面进行讨论时，它才会变得可操作。

I think that this is the kind of thing that becomes actionable when it is more obviously a crisis and actionable when it's something that you can discuss at a societal level.

我想，这场对话中一直萦绕的问题是，人工智能的变化几乎会在各个地方发生，而其风险则以一种分散且难以捉摸的方式出现，导致我们很难准确识别它并采取行动。

I guess the thing that was circling around in this conversation is that the changes of AI will kind of happen almost everywhere, And the risks of it, it happens in a diffuse unknowable way such that it is very hard to call it for what it is and take actions on it.

但机会在于，如果我们能够真正看清这一现象，并帮助世界看清正是这种变化在推动这一切，我相信这会凸显出问题，促使我们摆脱一些惯性，从而学会如何与这些系统共处并从中受益。

But the opportunity is that if we can actually see the thing and help the world see the thing that is causing this change, I do believe it will dramatize the issues to kind of shake us out of some of this stuff and help us figure out how to work with with these systems and benefit from them.

我在这所有事情中注意到的是，据我所知，公众对人工智能没有任何明确的议程。

What I notice in all this is that there is, as far as I can tell, zero agenda for public AI.

社会希望人工智能做什么？

What does society want from AI?

它希望这项技术具备哪些能力？

What does it want this technology to be able to do?

有哪些事情是你需要通过商业模式、奖励机制，或者某种政府补贴、政策来塑造市场或激励体系的？

What are things that maybe you would have to create a business model or a prize model or some kind of government payout or some kind of policy to shape a market or to shape a system of incentives?

所以我们现在有系统在解决的不只是私营市场知道如何付费的问题，还有那些没人负责、只能由公众和政府来解决的问题。

So we have systems that are solving not just problems that the private market knows how to pay for, but problems that it's nobody's job but the public Mhmm.

而政府则需要想办法解决这些问题。

And the government to to figure out how to solve.

我认为，考虑到过去几年人们对AI的大量讨论以及这些系统变得多么强大，我本该看到更多这方面的提案了。

I think I would have bet given how much discussion there's been of AI over the past couple of years and how strong some of these systems have gotten, that I would have seen more proposals for that by now.

我和一些人讨论过这个问题，也一直在思考，但我很好奇你是怎么看待这个问题的。

And I've talked to people about it and wondered about it, but I I guess I'm curious on how you think about this.

如果能有一个与所有私营AI发展激励并行的议程，会是什么样子？这个议程不是关于我们害怕AI会对公众造成什么影响。

What would it look like to have, at least parallel to all the private incentives for AI development, an actual agenda for not what we are scared AI will do to the public.

我们确实也需要这样一个议程。

We need an agenda for that too.

而是关于我们希望AI能做什么，从而让像你们这样的公司有动力往这个方向投资？

But what we want it to do such that companies like yours have reasons to invest in that direction?

我很喜欢这个问题。

I love this question.

我认为这里存在一个典型的鸡生蛋还是蛋生鸡的问题：如果你亲自使用这项技术，就会对它能做什么产生非常强烈的直觉。

I think there's a real chicken and egg problem here where if you work with the technology, you develop these very strong intuitions for just how much it can do.

而私营市场非常擅长推动这些直觉的形成。

And the private market is great at forcing those intuitions to get developed.

但我们尚未在公共领域大规模部署这项技术。

We haven't had massive large scale public side deployments of this technology.

因此，许多公共部门的人还没有这些直觉。

So many of the people in the public sector don't yet have those those intuitions.

一个积极的例子是能源部正在开展的‘创世计划’，他们的科学家正与所有实验室（包括Anthropic）合作，探索如何有意识地加速科学进展的某些环节。

One positive example is something the Department of Energy is doing called the Genesis Project, where their scientists are working with all of the labs, including Anthropic, to figure out how to actually go and intentionally speed up bits of science.

要达到这个阶段，我们和其他实验室花了多次黑客松和与能源部科学家的多次会议，直到他们不仅形成了直觉，还变得兴奋起来，并开始思考可以将这项技术用在哪些方向上。

Getting there took us and other labs doing multiple hack days and meetings with scientists at the Department of Energy to the point where they not only had intuitions, but they became excited, and they had ideas of what you could turn this toward.

要将这种模式推广到更广泛的公共生活领域——比如医疗或教育——这些领域触及大多数人，需要企业深入这些社区，与他们面对面交流，开展基层努力。

How we do that for the larger parts of the public life that touch most people, like health care or education, is going to be a combination of grassroots efforts from companies going into those communities and and meeting with them.

但到了某个阶段，我们必须将其转化为政策。

But at some point, we'll have to translate it to policy.