克劳德代码如何编写代码

团队里有人正在用它缴税。

Someone on the team is using it to pay their taxes right now.

所以这也不是财务建议。

So also not financial advice.

但它实际上对各种各样的事情都很有用。

But it's actually quite useful for all this different kinds of stuff.

有一件事很难向别人解释：人们常问我，你用它来做什么？

This is one of the things that's also kinda hard to explain to people is people ask, what do I use it for?

我的回答是：差不多所有事情。

And my answer is, well, kinda everything.

它能帮你处理所有那些繁琐、不想亲手做的琐事，让你专注于真正想做的事情。

It's like all all the toil, like all the stuff you didn't wanna do by hand, it can just do so you can do the stuff you actually want to do.

是的。

Yeah.

所以我认为，好吧。

So I I think okay.

我们来谈谈报税吧，我知道这只是一个你随口举的例子，但我觉得它很好地体现了我经常思考的AI应用场景——比如整理我的截图，风险相对较低。

Let's talk about doing taxes, which I know is is just an example off the top of your head, but I think is is a useful sort of middle ground of the kinds of stuff that I I think about a lot with AI, where organize my screenshots is relatively low risk.

对吧？

Right?

比如，它可能会删除一些我不希望删除的东西，但总体来说，它只是把东西归位，删掉我电脑上不需要的文件。

Like, the the idea of it might delete a thing that I didn't want it to delete, but in general, it's just gonna put things in places and delete stuff off of my computer that I don't want.

我觉得人们可以很快适应在电脑上做这类事情。

And I think you can get people comfortable with doing things like that on their computers fairly quickly.

让AI帮我报税则自然会有更大的后果。

Have co work do my taxes just has naturally more consequences.

对吧？

Right?

我知道你经常被问到一个问题，也是我觉得很多这类工具的难点所在：让AI写代码是一回事，我之后可以去检查，即使我不懂。

And and I think part of I I know a question you get asked a lot, and also a a thing that I think is tricky with a lot of these tools is it's one thing to have it write code that I can then go check, even if I don't.

对吧？

Right?

责任又回到了我身上，我需要检查并确保自己理解它的运作方式，而且代码对我来说作为开发者是可读的。

The responsibility is back on me to check it and make sure that I understand where it is and and code is legible to me as a developer.

但如果我只是个普通人，让同事帮我报税，那么在现阶段，我对同事、Claude Code 或任何工具完全代我执行这项任务，能有多少合理的信任呢？

But if I'm just a person and I'm like, coworker, go do my taxes for me, how much faith is it reasonable or fair or rational to have in coworker or Claude Code or any tool to go just execute that entirely on my behalf at this point?

这些工具并不完美，现在也还处于早期阶段，但它们在人们通常认为不会擅长的领域却表现得出人意料地好。

The tools are not perfect, and it's still early, but they are surprisingly good at things that people often expect they would not be good at.

而且，每一代新模型都在不断进步。

And again, it just improves with every model.

对于报税这类事情，我肯定会仔细核对两遍。

For something like taxes, I would definitely double check it.

让工具帮我报税。

Have do the tax.

实际上，我会做的做法是：让工具报税，但然后三重核对它的结果。

And and actually, the thing that I that I would do is say do the taxes, but then triple check your results.

然后让同事帮你完成这项工作。

And just have Corework do that work for you.

而当你去检查时，它很可能已经相当不错了。

And then and then by the time you check it, there's a very high chance it's just gonna be pretty good.

是的。

Yeah.

实际上，关于你可以让它自我测试，这一点我觉得也非常强大。

Actually, the to your point about you you can have it test itself, that actually I think that there's there's something very powerful about that too.

但我提到税务的原因是，税务软件上一次重大创新是它能为你扫描W-2表格。

But part of the reason I bring up taxes is because the the last innovation in tax software was that it will scan your w twos for you.

我记得这在我生活中曾是一件大事，我不再需要手动输入，只需上传W-2的PDF，它就能自动提取所有信息。

I don't I this is this I remember this being a very big deal in my life where I didn't have to type out the I could just upload the PDF of my w two, and it would just pull in all the information.

我曾经有一刻觉得，好吧。

And I remember for a minute, it was like, okay.

你还是得检查一下，因为扫描系统并不完美。

You have to check that because the the scanners the scanning system isn't perfect.

软件不会完全准确地处理。

The software won't get it exactly right.

但现在，我不记得上一次核对数字是什么时候了。

But now, like, I don't remember the last time I double checked the numbers.

你只需要上传W-2表格就行了。

I just you just you just upload the w two.

它会自动显示在字段里，然后你就继续你的生活。

It shows up in the field, you move on with your life.

我在想，我们似乎正朝着这个方向一路狂奔，

And I I wonder it it feels like we are just barreling towards that with

是的。

Yeah.

随着所有这些工具的出现，就像你使用Claude Code的经历一样，

With all of these tools too, that it's like, there's gonna be a beat of I mean, I guess it's like your experience with Claude Code.

会有一个阶段你需要检查它的结果，然后进入一个阶段，就是我随便抽查一下，

There's gonna be a beat of I need to check its work, and then a beat of, well, I'll spot check it.

最后我们就不再担心了。

And then we get to, I'm just not worried about it anymore.

就这样了。

And that's it.

这才是正确的最终状态。

That's the right end state.

就是感觉我们还没有

It just it doesn't feel like we're

还没到那一步。

quite there yet.

对。

Right.

对。

Right.

我的意思是，有两件事同时发生。

I mean, it's like two things that happen at the same time.

就像模型变好了，产品也变好了，同时作为用户，我们也对这件事越来越习惯了。

It's like the model gets better and the product gets better, and then also it's like, as users, we get more comfortable with this thing.

而且，这两件事基本上是同时发生的。

And, like, both things kinda happen at the same time.

在发布CoWork之前，我在刚开始测试时就用它来管理我们团队的所有项目。

Before before we released CoWork, I was using it to do all of our project management for the team when I was, like, first testing it out.

而且我到现在每周还在用它做这件事。

And I still actually use it for this every week.

所以我们有一个表格，列出了团队正在做的所有事情，我们要求团队成员每周更新他们的进度。

So we have a spreadsheet of kind of all the things the team is working on, and we ask the team to just fill out their status every week.

比如，它是否在正轨上？

So to say, is it on track?

它是否偏离了轨道？

Is it off track?

所以我会让CoWork在Slack上提醒那些还没填写的人。

And so I just have CoWork ping people on Slack if they haven't filled it out.

所以我只需要说：嘿，CoWork。

And so all I do is I'm like, hey, co work.

打开电子表格。

Open the spreadsheet.

然后对那些还没填写的人，在Slack上给他们发消息。

And then for anyone that hasn't filled it out, message them on Slack.

它会完美地完成这一切。

It'll just do it perfectly.

但实际上有一个人的名字，不知什么原因，它在Slack上识别不出来，所以我得手动处理。

There's actually one person's name that it, for some reason, can't figure out on on Slack, so I I have to do that.

但除此之外，它都能自动完成。

But otherwise, it just does it.

我其实还挺惊讶的，因为我根本没想到它能做到这一点。

And I was actually, like, kinda taken aback because I didn't even realize that it would be able to do this.

所以你可以先试着实验一下，反复确认，直到你感到放心，但我觉得我们很快就能做到。

So I would just, like, experiment with this, double check until you're comfortable, but I think we'll be there pretty soon.

在这种情况下，它是以你的身份还是以机器人的身份在Slack上发消息？

In a case like that, does it message people on Slack as you or as, like, a a bot?

我让它以CoWork的身份发送消息。

I asked it to sign as messages as CoWork.

那没问题。

So that's fine.

好的。

Okay.

是的。

Yeah.

而且现在，CoWork已经变成了一件无须强制的事情。

And CoWork, actually, now it's a forceless thing.

在Claude Code中，我们有一个叫做Claude dot m d的概念。

In Claude Code, we have this idea called Claude dot m d.

它就像一个特殊文件，但本质上，它是你希望Claude每次都能考虑的所有指令。

It's just like a special file, but essentially, it's like all the instructions you want Claude to take into account every time.

所以CoWork现在也支持这个功能。

So CoWork also supports this now.

所以你只需说，比如当你在Slack上给他人发消息时，自动以CoWork、机器人之类的名义签名。

So you can just say, like, whenever you message people on Slack, sign yourself as you know, sign sign a message as like CoWork or bot or something.

它就会自动这么做。

It'll just do that.

是的。

Yeah.

这很聪明。

That's smart.

对。

Yeah.

我觉得这种透明度有点意思。

I think that that kind of there's a little bit of transparency there that I think is interesting.

比如我记得你曾在一次采访中提到，CoWork在为你做事时，偶尔会替你发推文，这总让我觉得有点奇怪。

Like, I remember you said in one interview I was I was watching that coworker would occasionally, in the course of doing stuff for you, go and tweet on your behalf, and that that always felt kinda strange.

是的。

Yeah.

是的。

Yeah.

对。

Yeah.

这挺有趣的。

It's it's funny.

实际上，Claude Code 现在也经常这么做。

Actually, Claude Code does this too pretty consistently now.

当我正在调试某些东西时，Claude 有时会说：嘿。

When when I'm, like, debugging something, sometimes Claude will be like, hey.

这段代码有点奇怪。

This code is kinda weird.

让我看看历史记录。

Let me, like, look at the history.

所以为了查看 Git 中代码的历史记录，它偶尔会发现某人做出的非常奇怪的更改，然后直接在 Slack 上给那位工程师发消息以获取上下文。

So to look at the history of the code in Git, once in a while, it sees a really weird change by someone, and it'll message that engineer on Slack just to get context.

它会等待回复。

It'll wait on the response.

而且我也见过它提出异议。

And then I've also seen it push back.

所以，工程师会说，是的。

So, like, the engineer is like, yeah.

我做这个改动是有原因的。

I did this change for this reason.

然后Claude Code会说，我觉得这个理由并不充分。

And then and then Claude Code is like, well, I don't think that's a very good reason.

而且我认为你实际上引入了一个bug，让我来帮你修复一下。

And I think you actually introduced a bug, so let me, like, go ahead and fix that.

你对这部分功能的其他界面有什么想法？

How are you thinking about the rest of the UI around this stuff?

我觉得Claude、Code和Cowork的大部分功能仍然非常依赖聊天形式。

I think, like so much of Claude, Code and Cowork are are very chat based still.

从长远来看，你觉得这个界面是合适的吗？还是说那里还需要做更多工作？

Does that feel like the right UI to you going forward, or is there more work to do there?

我们一直在不断尝试新想法。

We are constantly experimenting with new ideas.

我认为未来的界面还没有被发现。

I think the UI of the future has not been discovered yet.

所以我们目前有很多实验正在进行中。

So we have a lot of experiments in flight.

我预计它会有所变化。

I would expect it to change.

我们会测试很多东西。

There's gonna be a lot of things that we test.

最重要的是了解人们真正想要什么。

The single most important thing is just seeing what people want.

所以我整天都在推特和Threads上浏览，因此错过了团队的很多内容。

And so I'm on Twitter and threads all day, and so lose a lot of the team.

我们就是喜欢和人交谈。

We just love talking to people.

我们很喜欢获得反馈，因为我们有很多想法，但只有通过观察人们怎么说、人们喜欢什么，才能找出哪些是正确的想法。

We love getting the feedback because we have a lot of ideas, but the only way to figure out what the right ideas are are to see what people say and to see what people enjoy.

我同意你的观点，未来的用户界面还没有被发现。

I agree with you that the UI of the future has not been discovered yet.

在2026年初，你对它可能是什么样子有什么假设吗？

Do you have a hypothesis at this moment in early twenty twenty six about what it might be?

我还没有。

I don't yet.

说实话，我觉得我们还没找到。

I don't think we found it, to be honest.

我觉得现在有很多想法，比如主动性，像Claude在它知道你需要帮助时主动介入，但很难把握好这个界限，因为你不想最后搞出一个像Flippy那样的东西。

I think there's a lot of ideas around, like, proactivity and Claude kinda jumping in when it knows that you're gonna need help, but it's kinda hard to get this boundary right because you don't wanna end up with something like Flippy.

这其实也反映了你提到的某种进展，比如从拉小提琴到指挥交响乐团的转变。

And it it speaks to, I think, the the progression you're talking about a little bit, a, from, you know, playing the violin to conducting the orchestra.

当你这么思考时，你能使用的一套工具就完全不同了。

It's just a different set of tools that are available to you when when that's what you're thinking about.

但你也提到了，从基本上写代码到使用工具，再到使用计算机的转变。

But, also, you you mentioned going from basically code to tool use to computer use.

你能给我详细讲讲这个过程是怎样的吗？

Can you just walk me through what that progression looks like as we go through?

因为我觉得我们已经听了很多关于智能体的内容，以至于‘智能体’这个词实际上已经没什么意义了。

Because I think we we've heard a lot about agents to the point where I think the word agent essentially means nothing.

智能体就像是你电脑上发生的魔法，嗯，好吧。

Agent is just like magic that happens on your computer, and it's like, sure.

随便吧。

Whatever.

但我认为你对这个问题的思考方式更实际，更关注于如何赋予这个系统更多能力。

But I I think you're you're thinking about this in a in a much more sort of practical how do we give this thing more powers kind of way.

是的。

Mhmm.

为什么是从代码、工具使用到计算机使用？

Why does it go code, tool use, computer use?

是的。

Yeah.

天啊。

Oh my god.

别让我开始说这个。

Don't don't get me started.

比如‘okay’这个词。

Like, the word okay.

我真要开始了。

I will get started.

我觉得‘agent’这个词，每个人都用错了。

The word agent, I feel like I feel like everyone just misuses it.

当你谈论人工智能研究或工程时，这个词其实有非常特定的含义。

Like, it has a really specific meaning when you talk about AI research, when you talk about engineering.

所以，智能体就是一个你可以与之对话的大型语言模型，但它能使用工具。

So an agent is an LLM that you talk to, but the LLM can use tools.

这正是让它成为智能体的关键。

This is the thing that makes it an agent.

它就像能够使用工具一样。

It's like it it can use tools.

如果没有工具使用能力，智能体只能写代码。

And so if you think about without tool use, the agent can write code.

比如说，你给它一个提示，它就能写出一些HTML代码之类的。

So let's say you give it a prompt, and it can kinda write some HTML or something.

然后作为用户，你需要把这些代码复制粘贴到IDE之类的工具里。

And then as a user, you take this and you kinda copy and paste it into, like, an IDE or something like this.

这仅仅是一种编码能力。

So this is just like the coding capability.

随着模型越来越智能，它在处理大型代码库方面也会变得越来越好。

And as the model gets smarter, it gets better and better at working with big code bases.

但这里仍然存在一个问题：在某个时刻，你无法一次性提供它所需的所有上下文，但你知道，模型实际上确实知道它需要的上下文，因为它能够进行搜索，并查看整个代码库。

But there's still kind of this problem that you hit where at some point, you just can't give it all the context it needs, but, you know, the model actually does know the context that it needs because it's able to search around, and it's able to look throughout the entire code base.

它能够查看 Slack。

It's able to look at Slack.

它能够查看代码的历史记录。

It's able to look at, like, the history of the code.

它能够做所有这些事情，但信息量实在太大了。

It's able to do all of this, but it's just too much information.

就像，你不可能把所有信息都提前提供给它。

Like, you wouldn't be able to give it all the information upfront.

所以答案就是工具。

And so the answer is tools.

你给模型提供工具，它就可以使用工具来查看代码。

You give the model tools, and it can use a tool to look at the code.

它能够加载更多的文件。

It can it can pull in more files.

它可以查看历史记录。

It can look at history.

它可以做所有这些事情。

It can do all this stuff.

这就是为什么工具使用很重要。

And so this is why tool use is important.

这和人是一样的。

It's the same as a person.

如果你没有工具，光靠双手其实做不了多少事。

If you don't have tools, you actually can't do a lot just with your hands.

对吧？

Right?

你需要键盘。

You need keyboards.

你需要铲子。

You need shuffles.

如果你在厨房做饭，你需要一个打蛋器。

You need if you're cooking in the kitchen, you need a whisk.

这些东西都很重要，没有它们你几乎什么都做不了。

These things are just very there's not a lot you can do without it.

所以对模型来说，情况也差不多。

So it's kind of the same thing for a model.

当你想到计算机使用时，有很多事情仅靠工具很难直接交互。

And then when you think about computer use, there's just like a lot of things that are kind of hard to interact with just with tools.

所以，想想看，在计算机上用工具到底能做什么？比如MCP、API或者命令行界面，但并不是所有东西都有这些接口。

So if you think about, like, what what can you actually do with a tool on a computer, it's something like MCP or it's an API or it's a command line interface, but not everything has that.

所以，比如说，假设你有，我不知道。

So, you know, like, if you have, like, I don't know.

比如这个，这个叫采蛤蜊的东西。

Like, this this, like, clamming thing.

我之前在申请一个采蛤蜊的许可证。

I was getting this, like, clamming license.

而且，你知道，这没有API，但有一个网站。

And, you know, there's no API for that, but there's a website.

对。

Right.

要使用这个网站，你需要让模型能够使用浏览器。

And to use the website, you want the model to be able to use a browser.

你希望它能够使用电脑。

You want it to be able to use a computer.

所以这是一种自然的演进。

And so this is kind of this natural evolution.

你从编程开始，然后转向工具，这就是与世界互动的方式。

So you start with coding, then you move on to tools, and this is the way to interact with the world.

因此，你不必事无巨细地为模型提供上下文。

And so you don't have to spoon feed the model context.

它可以直接使用工具来获取上下文。

It can just use the tools to pull in context.

然后，计算机是最后一步，因为这样一来，模型就能使用一切了。

And then computers are kind of the last the last thing, because then the model can just use everything.

好的。

Okay.

你觉得，随着人工智能持续发展，如果它真的像很多人预测的那样全面接管所有软件和计算，那么‘使用计算机’这一部分最终会不会变得多余？

Do you think as AI continues to grow, and if it if it sort of takes over all of software and computing the way that that a lot of people think it's going to, that the the computer use part eventually becomes sort of obviated.

比如，如果有足够多的工具、足够的MCP访问权限，以及你所说的那些东西，‘使用计算机’会不会只是个优雅的临时方案，用来绕过未来可能存在的更高级解决方案，而那时我们就不需要它了？

Like, if there were enough tools and enough MCP access and enough of the the stuff that you're talking about, does compute is computer use just sort of an elegant hack that gets around the stuff that maybe will exist later and we won't need it?

在早期使用模型进行编程时，人们曾讨论过设计专门的编程语言，以便让模型更好地编程。

Early on, in the early days of using the model for coding, people were talking about designing special programming languages to make it so the model can code better.

对。

Right.

但我一直觉得这有点可笑，因为模型自己就能搞明白。

And I always thought this was kind of silly because the model can just figure it out.

你知道的。

You know?

这不像我们人类，有人喜欢Python，

It's not like us where there's a programmer that likes Python.

有人喜欢JavaScript，坚决不碰Python。

There's another one that likes JavaScript and won't touch Python.

模型可不是这样的。

The model's not like that.

它可以写出任何语言。

It can just write whatever language.

它根本不介意。

It doesn't care.

所以我觉得这里也是同样的道理。

So I think it's kind of the same thing here.

我认为随着时间推移，模型根本不在乎。

I think over time, the model doesn't care.

你给它什么工具，它都能弄明白，并用这些工具为你做事。

Whatever tools you give it, it will be able to figure it out, and it can use those tools to do, you know, things for you.

谈谈人们在向Claude Code或CodeWork这样的系统授予访问其数据、电脑、文件、照片等权限时，应该如何看待自己的风险承受能力。

Talk to me about how how people should think about kind of their own risk profile in giving access to their data and their computer and their files and their photos and whatever to a system like Claude Code or or CodeWork.

我觉得你有很强的动机告诉我，这完全没问题。

I think you you have, you know, lots of incentive to tell me that it's it's totally fine.

你可以拥有我电脑上的所有这些东西。

You can have all this stuff on my computers.

我们正在建立安全防护措施。

We're we're putting the safeguards in.

但人们应该如何理解授予Claude Code访问我电脑上某个文件夹的含义，哪怕只是这样的操作？

But how should people think about what it means to give Claude Code access to a folder on my computer, even something like that.

完全正确。

Totally.

我会从几个层面来思考这个问题。

So I would think about it on a few levels.

最基本的一点是，Anthropic存在的意义是什么？

So the most basic level is, why does Anthropic exists?

我们存在的目的是创造安全的通用人工智能。

We exist to make safe AGI.

我们最初有一群创始人，他们离开了另一家人工智能实验室，来到这里并创立了这家公司。

We you know, initially, we have a bunch of founders that, you know, like, left a different AI lab and came and started and brought it have heard of.

是的。

Yeah.

我对此很熟悉。

I'm I'm familiar.

但这就是我们存在的原因。

But this is the reason we exist.

有很多人对安全性提出质疑。

And there's a lot of quarrelers to safety.

如果你想实现真正的安全性，安全措施至关重要。

Security is actually very important if you wanna get safety right.

如果你想实现真正的安全性，隐私保护同样至关重要。

Privacy is very important if you wanna get safety right.

所有这些工作我们都需要去做，我们非常幸运的是，我们关心安全，而我们最重要的目标客户——企业与公司——也同样关心安全。

And all of this stuff we sort of have to do, we're very lucky that we care about safety, and so does our most important target customer, which is enterprise and companies.

你知道，有很多消费者在使用Anthropic的产品。

You know, there's a lot of, like, consumers that use Anthropic products.

这太棒了，这也是我们非常乐于看到的。

This is awesome, and this is something we love to see.

我们会为你打造产品。

We will build for you.

但实际上，我们真正关注的主要市场是企业客户。

But, actually, like, the main market we care about is enterprise as a company.

当然。

Sure.

我们非常幸运，并且有意识地选择了这个市场，因为我们知道企业非常重视安全、保密和隐私。

And we're very lucky, and we picked this market on purpose because we know enterprises care a ton about safety and security and privacy.

因此，我们为他们量身打造了产品。

And so we built for them.

所以如果你看一下这个产品，对我来说其实有点烦人，因为如果有人提交了一个四重代码错误报告之类的问题，我根本看不到你的数据。

And so if you look at the product, it's actually kinda annoying for me because if someone has a quad code bug report or something, I literally cannot see your data.

所以我需要你提供能复现问题的材料，但我真的无法访问这些数据来查看这个问题。

So I need you to give me reproduction stuff so I can reproduce it, but I literally can't access the data see this issue.

因此，我们设置了大量类似的控制机制。

So there's a lot of controls like that in place.

此外，由于我们非常重视安全性，我们在使模型本身更加对齐和可解释方面投入了大量工作。

Also, because we care about safety a lot, there's a lot of work that goes into just making the model inherently more aligned and interpretable.

这一点同样非常重要，并且与此密切相关。

And this is also just it's very important and also very related to this.

而且，是的，最后一点是我们为产品内置了大量功能。

And, yeah, I mean, the final thing is there's just a lot of stuff that we build into the product.

Cowork 只能访问你授权给它的文件夹。

Cowork can only see the folders that you give it access to.

它无法查看你电脑上的任何其他内容。

It cannot see anything else on your computer.

我们在Cowork中内置了一个完整的虚拟机，以确保它与你的数据之间存在坚固的安全边界，使其无法访问你未授权的任何内容。

We put an entire virtual machine in Cowork to make sure it's a really hard security boundary to you know, so it can't access stuff that you don't give it access to.

最需要担心的是类似提示注入之类的攻击，这类攻击可能会窃取你的数据。

The biggest thing to worry about is attacks like prompt injection, anything like this that would kind of exfiltrate your data.

我们为此设置了多重防护措施，而Opus 4.6是我们迄今为止针对提示注入问题对齐程度最高的模型。

We have a lot of protections in place for this, and Opus 4.6 is just the most aligned model that we've ever built for prompt injection in particular.

此外，我们还部署了许多运行时分类器和安全机制来应对这一问题。

And there's also a lot of, like, runtime classifiers and kind of safeguard that we put in place for this.

但我觉得最重要的是，当你让同事或Claude Code与互联网交互时，要谨慎选择它访问的网站，它会主动向你请求权限。

But this is the biggest thing that I would think about is as you have coworkers, as you have Claude Code interact with the Internet, just be thoughtful about what websites it is using, and it will ask you for permission.

但这一点仍需留意，因为这个问题尚未完全解决。

But it's a thing to keep an eye on because this is not a solved problem yet.

它已经相当不错了，但还没有彻底解决。

It's quite good, but it's not yet solved.

这是个

That's a

很好。

good one.

好的。

Alright.

给我一个正常的职场人类活动吧，很多人都应该做，你要么做过、要么正在开发、要么听别人说过，但可能不是每个人都意识到他们应该去做。然后我就放你走。

Give me one, like, normal human co work activity that lots of people should do that you you've either done or building or have heard from people that not everybody might expect that they should go do, and then I'm gonna let you go.

哦，正常的普通人。

Oh, a normal human.

好吧。

Okay.

一个是回邮件。

One is just, like, responding to email.

打开我的Gmail，看看最上面三封需要回复的邮件，起草回复内容。

Just, like, open my Gmail, look at the top three things I should respond to, draft responses.

它在这方面做得相当不错。

So it can do that quite well.

我做的第二件事是取消订阅。

A second one that I do is just like canceling subscriptions.

所以我实际上用它来取消了一个我不再看的电视服务。

So I actually use it to cancel like I canceled like a TV thing that I wasn't watching.

这真的是最让人烦不胜烦的事情。

That's the the most unbelievably annoying thing to do.

我要让Claude Code帮我取消所有我不再想看的电子邮件通讯。

I'm gonna make Claude Code unsubscribe to all of my email newsletters that I don't want anymore.

这对我来说会很管用。

This is this is gonna work for me.

是的。

Yeah.

对。

Yeah.

我超爱这种双轨并行的方式。

I I love this, like, dual track.

你可以用它来写邮件，也可以用来取消订阅邮件。

Like, you can use it for your write the emails and also unsubscribe for email.

是的。

Yeah.

没错。

Exactly.

我再也不想看我的邮件了。

I just never wanna look at my email ever again.

如果克劳德能实现这一点，我们就真的做到了一些事情。

If if Claude can make that happen, we will have accomplished something.

通用人工智能。

AGI.

好吧。

Alright.

博里斯，谢谢你。

Boris, thank you

太多了。

so much.

非常感谢你做这件事。

I really appreciate you doing this.

是的。

Yeah.

是的。

Yeah.

谢谢，大卫。

Thanks, David.

我们马上回来。

We'll be right back.

本节目由Built Rewards赞助。

Support for the show comes from Built Rewards.

不幸的是，每个月按时交房租不会有人拍拍你的背表示赞赏，但你可以通过Built为房租获得奖励，这就像有人拍拍你的背，只不过是对你的钱包的鼓励。

Unfortunately, no one will pat you on the back for paying your rent on time every month, but you can earn rewards on your rent with Built, which is like a pat on the back, but for your wallet.

Built 是为租客设计的忠诚度计划，每月为您奖励积分和社区内的独家福利。

Built is the loyalty program for renters and rewards you monthly with points and exclusive benefits in your neighborhood.

使用 Built，每次支付房租时，您都能获得积分，可用于兑换航班、酒店、Lyft 打车、亚马逊购物等众多权益。

With Built, every time you pay your rent, you earn points that you can put towards flights, hotels, Lyft rides, amazon.com purchases, and so much more.

现在，这不再仅限于租客。

And now it's not just for renters.

Built 会员也可以在偿还房贷时累积积分。

Built members can earn points on mortgage payments too.

成为 Built 会员还能解锁超过 45,000 家餐厅、健身工作室、药房及其他社区合作伙伴的独家优惠。

Being a Built member also unlocks exclusive benefits with more than 45,000 restaurants, fitness studios, pharmacies, and other neighborhood partners.

立即前往 joinbuilt.com/verge 加入为租客打造的忠诚计划。

Join the loyalty program for renters at joinbuilt.com/verge.

网址是 joinbilt.com/verge。

That's joinbilt.com/verge.

请务必使用我们的链接，以便他们知道是我们推荐的。

Make sure to use our URL so they know we sent you.

本节目由Shopify提供支持。

Support for the show comes from Shopify.

创业之初，可能是一段孤独的旅程，尤其是在刚开始的时候。

Starting a new business, it could be a lonely endeavor, especially in the beginning.

如果你刚刚起步，确保手头拥有正确的工具就显得尤为重要。

And if you're just starting out, it's more important than ever to make sure you have the right tools at hand.

如果你的业务涉及电子商务，那么下一步可以尝试使用Shopify。

If your business includes ecommerce, a great next step is to try Shopify.

Shopify是全球数百万企业依赖的电商平台，用于在线销售他们的产品。

Shopify is the commerce platform that millions of businesses around the world rely on to sell their products online.

你可以通过数百个即用型模板，轻松启动自己的设计工作室。

You can get started with your own design studio with hundreds of ready to use templates.

Shopify帮助你打造一个与品牌风格相匹配的精美在线商店。

Shopify helps you build a beautiful online store that matches your brand's style.

如果你在想，如果人们还没听说过我的品牌怎么办？

If you're asking yourself, what if people haven't heard about my brand?

Shopify 通过简单易操作的电子邮件和社交媒体活动帮助你找到客户。

Shopify helps you find your customers with easy to run email and social media campaigns.

如果你遇到困难，Shopify 随时提供帮助，其屡获殊荣的 24/7 客户支持团队会为你提供建议。

And if you get stuck, Shopify is always around to share advice with their award winning 20 four seven customer support.

是时候用 Shopify 将这些‘如果’变为现实了。

It's time to turn those what ifs into with Shopify today.

你可以以每月 1 美元的价格注册试用，今天就开始在 shopify.com/vergecast 销售产品。

You could sign up for your $1 per month trial and start selling today at shopify.com/vergecast.

前往 shopify.com/vergecast。

Go to shopify.com/vergecast.

就是 shopify.com/vergecast。

That's shopify.com/vergecast.

嘿。

Hey.

这里是卡拉·斯威舍。

Kara Swisher here.

我想告诉你，Vox Media 将重返德克萨斯州奥斯汀的西南偏南大会，进行你最爱播客的现场录制。

I wanna let you know that Vox Media is returning to South by Southwest in Austin Austin for live tapings of your favorite podcasts.

请在3月13日至15日加入我们，现场录制《今日解析》《Tuffy Talks》《Prof G 市场》以及当然还有你最爱的两档播客《Pivot》和《On with Kara Swisher》。

Join us from March 13 through the fifteenth for live tapings of Today Explained, Tuffy Talks, Prof G Markets, and, of course, your two favorite podcasts, Pivot and On with Kara Swisher.

舞台还将呈现来自蕾妮·布朗、亚当·格兰特、马尔凯斯·布朗利、基思·李、维维安·图和罗宾·阿尔赞的专场活动。

The stage will also feature sessions from Renee Brown and Adam Grant, Marques Brownlee, Keith Lee, Vivian Tu, and Robin Arzan.

这一切都是由Odoo赞助的Vox Media播客舞台在西南偏南大会上的精彩内容。

It's all part of the Vox Media podcast stage at South by Southwest presented by Odoo.

访问 voxmedia.com/sxsw 预注册，获取你的创新通行证专属折扣。

Visit voxmedia.com/sxsw to preregister and get your special discount on your innovation badge.

要注册的话，请访问 voxmedia.com/sxsw。

That's voxmedia.com/sxsw to register.

真的，你该去注册了。

Really, you should register.

活动票会售罄，我们期待在那里见到你。

We sell out, and we hope to see you there.

好的。

Alright.

我们回来了。

We're back.

这里是《The Verge》的高级AI记者海登·菲尔德。

Hayden Field, Verge senior AI reporter is here.

嗨，海登。

Hi, Hayden.

嗨。

Hi.

你最近来过，我们聊过Moltbook、OpenClaw，还有你能用AI工具和AI代理在电脑上做的各种疯狂事情。

So you were here recently, and we were talking about Moltbook and OpenClaw and and all of the insane things that you can do on your computer with AI tools and AI agents.

我们还稍微讨论了隐私问题，以及如何思考是否应该使用这些工具、安装它们，还有应该向它们提供哪些数据。

And we talked a little bit about privacy and kind of how to think about whether or not you should engage with these tools and install them and what kind of data you should give to them.

我意识到，我一直在对AI工具产生不同程度的存在主义危机。

And I've realized I've been having kind of varying levels of existential crises about AI tools.

首先是，我确实有过一次使用 OpenClaw 的真实经历：我下载了 OpenClaw 的安装程序到我的电脑上。

Starting with, like, I had a real experience with OpenClaw where I I downloaded the installer for OpenClaw onto my computer.

我有一台 Mac mini，还有一台 MacBook Air。

I have a I have a Mac mini, and I have a MacBook Air.

我当时在 MacBook Air 上，下载了 OpenClaw，心想：好吧，我要用用看，体验一下，彻底试试它。

I And was on the MacBook Air and I downloaded OpenClaw and I like, I'm gonna use this, get into it, see what it's like, try the whole thing out.

但在安装进行到一半时，我突然觉得：这太蠢了。

And I got literally halfway through the install process and was like, this is so stupid.

这台电脑里存着我生活中所有在意的信息，所有关于我认识的人的资料，包括作为一名记者所掌握的重要机密信息，却要把这些交给一个完全不可知的 AI 代理，这简直疯狂。

Like, this computer is full of all the information I care about in the world and all of this stuff that I know about everyone that I know, including, like, important confidential information as a journalist, giving this unknowable AI agent access to this is insane.

这是第一个层面。

So that's one level.

但另一方面，我主要用 Claude 来做 AI 相关的事情。

But then even, like, I use I mostly use Claude for AI stuff.

而 Claude 非常希望你连接你的 Gmail 和 Google 日历。

And one thing Claude really wants you to do is connect your Gmail and connect your Google Calendar.

我曾经有过这样的时刻，觉得这样做是不是不负责任？

And I've had moments of being like, is this an irresponsible thing to do?

我这样给Claude访问我的邮箱，是不是太傻了？

Like, am I being stupid giving Claude access to my email?

所以，我想尽可能地好好思考一下，该如何看待你的数据与人工智能的关系。

So what I wanna do as best we can is just try to think through sort of how to think about your data and AI framework.

这听起来合理吗？

Does that seem reasonable?

完美。

Perfect.

我一直在问同样的问题。

I've been asking the same questions.

好的。

Okay.

那我们先从宏观的角度开始吧。

So let's just start kind of big picture.

我想深入一些细节，我 literally 想让你告诉我，我该不该给 Claude 访问我的 Gmail 的权限，我们会说到这一点。

I wanna get into some sort of nitty gritty I literally want you to tell me if I should give Claude access to my Gmail, we'll get to that.

你一直在这方面做报道，采访专家，并且自己也在认真思考这个问题。

You've been reporting on this a lot and talking to experts and and trying to think through this for yourself.

你有没有一些宏观的建议，能告诉人们应该如何思考这类问题？

Do you have kind of big picture guidance on just how people should be thinking about this stuff?

是的。

Yes.

这再好不过了，因为宏观层面最容易把握，毕竟每个人的决策都不同，取决于他们的风险承受能力以及生活方式的其他方面。

And this is perfect because the big picture is the easiest to get at, you know, because it's it's really a different it's a different decision for each person depending on their risk tolerance and, like, the other ways they live their life.

所以说实话，宏观层面最容易理清，每个人都可以为自己制定规则。

So honestly, the big picture is the easiest to kinda square and then everyone can kinda make the rules for themselves.

但这一周我采访了一堆专家，就是为了确认我的直觉是否跟真正的隐私专家和科技领袖们的看法一致。

But I did a bunch of expert interviews this week just to make sure my instincts are kind of on track with what actual privacy experts and, you know, tech leaders are thinking.

看起来他们的观点基本一致：很难给出一套能长期有效的建议来指导人们处理这类问题。

And it seemed like they were in that basically, it's hard to give people good advice on this stuff that stays current over time.

这正是我所交谈的一些隐私专家所说的。

That's what some of the privacy experts I was talking to said.

你知道，事情可能每六个月就变一次，每个月、每年都可能变。

You know, it's like every six months things could change, every month, every year.

所以，就目前而言，很多人基本上忽略了他们通常评估风险承受能力的方式，只是因为害怕错过（FOMO）或觉得这些AI工具能极大简化生活，就盲目采用了那些正在走红或被广泛讨论的工具。

So, you know, that as of this moment in time, a lot of people are essentially kind of ignoring the way that they usually, you know, evaluate their risk tolerance and just kind of adopting AI tools that are going viral or being talked about a lot just because of FOMO or like, you know, the promise of making your life a lot easier.

作为人类，我们都希望让生活更轻松。

We all as humans want to make our lives easier.

我认识的一位专家说，这就像塞壬的歌声，或者说是‘青少年模式’。

You know, one expert I talked to said it was like the siren song or like teenager mode.

就是说，你只想获得短期收益，让生活更轻松。

It's like, you know, you're just- you want to have the short term gain and make your life easier.

有时候，你真的不想去考虑长期的事情。

You don't really wanna think about the long term stuff sometimes.

这没关系。

That's fine.

但是

But

用‘青少年模式’来形容真是太贴切了。

Teenager mode is such a good way to think of it.

是的。

Yeah.

这太精彩了。

That was incredible.

完全是YOLO模式，哈哈，什么都不重要。

Full like YOLO mode, like LOL, nothing matters.

但这有点像我的大脑还没完全发育成熟。

But it's it's a little bit like my brain is just not yet fully developed.

而且他正是这么告诉我的。

And exactly, told me that.

就像系安全带 versus 仅仅觉得‘没关系’。

It was like, you know, doing your seat belt versus just being like, oh, it's fine.

我会直接从高速公路上开走。

I'll just drive off the highway.

所以，没错，就是这样。

So, yeah, exactly.

这就像是，你基本上需要像对待任何其他大量索取你数据的服务一样来对待AI工具，甚至要更加谨慎，因为这些公司更新，经受考验的时间更短，而且它们更有动力快速推进，同时它们所面临的监管框架也相对较少。

It's like, you know, basically you need to treat AI tools the exact same as you would treat any other service that was requesting a lot of data from you and maybe even with a sharper eye because these companies are newer, they're less time tested, and they're also more incentivized to move quickly and they have a little bit less, you know, regulatory frameworks on them.

所以，很多时候它们是自愿遵守某些规则的，这当然很好。

So, you know, a lot of times they're voluntarily complying with certain rules, you know, and that's all well and good.

但是一些专家告诉我的是，是的，它们随时都可以改变这些规则。

But, you know, something that a bunch of experts told me is, yeah, they can change that at any time.

它们可以在暗地里悄悄调整这种自愿性框架，随时改变自己的使命。

They can kind of on the DL shift that voluntary framework, shift that little our mission anytime.

如果它们这么做了，也不会有任何严厉的惩罚降临到它们头上。

And there's no hammer coming down on them if they do.

它们只是自愿遵守而已。

They are just it's voluntary.

所以他们是在做好事。

So they're doing it as a favor.

这意味着他们随时可以改变，调整如何处理你的数据、与谁共享你的数据，以及是否用你的数据来训练他们自己的系统。

And that means that they can shift it any given time and change how they treat your data, who they share your data with, how they use your data to train their own systems or not.

另一点是，这些公司最终可能会被收购。

And the other thing is these companies may get bought eventually.

我采访的一位专家说，如果你不希望五年后你的雇主知道关于你的某些信息，那么当OpenAI被收购、把你的信息卖给最高出价者时，你也应该警惕。

One expert I spoke with was like, if you wouldn't feel comfortable with your employer knowing certain things about you five years from now when OpenAI gets sold and they're selling off the info to the highest bidder.

这确实是个极端的例子，但他还是说，别分享这些信息。

Again, that was like an extreme scenario, but he was like, yeah, don't share it.

所以在分享健康信息时，这也是需要记住的一点。

So, that's something to keep in mind too with sharing health stuff.

你希望保险公司得知关于你的某些信息，然后调整你的保费吗？

Like, do you want insurance companies to find out certain things about you and change your premium?

这同样是个极端的例子。

Again, that's an extreme scenario.

希望这种情况永远不会发生，但你永远无法确定，因为这一切都太新了。

Hopefully it would never happen, but you never really know because all this stuff is so new.

所以，我不会说，你绝对不能做X、Y或Z。

So it's like, I'm never going to be like, don't do X, Y, or Z for you.

我可以这么说，因为我了解你。

I can do that because I know you.

但你知道，其他听众可能会说：不，我就是想把我的健康数据发给物理治疗师。

But like, you know, other listeners are going be like, no, I want to get my health data to chat to PT.

它对我帮助太大了。

It helped me so much.

比如，你知道，医疗系统正在让我失望。

Like the, you know, medical system is failing me.

这没问题。

That's fine.

是的。

Yeah.

医疗系统太糟糕了。

The medical system sucks.

所以，如果你确实想从你的健康记录中找出规律，并且你能接受这种程度的风险，那没问题。

So if you do want to find patterns in your health records and you feel comfortable with that level of risk tolerance, okay.

但如果你这么做，一定要确保你是在像ChatuchPeautyHealth这样的专业平台内操作，而不是用普通的聊天机器人，不过这依然有相当大的风险。

Just if you do that, of course, make sure you're doing it like within like ChatuchPeautyHealth and not like just regular chatbot, but it's still pretty risky.

是的。

Yeah.

我想，如果继续用青少年的类比，我觉得这有点像很多父母给他们的青少年孩子提出的建议，比如那些在发送敏感照片的孩子。

I think I I to to continue using the teenage analogy, I feel like it it's a little bit like the advice you hear a lot of parents give to, like, their their teenage children who are sending, let's say, sensitive pictures.

这种观点是，你应该假设任何你发送给他人或数字创建的内容最终都会公之于众。

This this idea of, like, you should assume that anything you send to someone or create digitally will eventually be public.

你的基本准则应该是：不要分享任何你不愿意让所有人看到的东西。

And that that is your framework is is you shouldn't share anything that you wouldn't share with everybody.

我认为每个人对这条界限的界定都截然不同。

And I think every everybody draws that line really differently.

对吧？

Right?

而且画这条线没有绝对的对错，但这种方式确实相当极端。

And there's kind of no wrong or right place to draw that line, but that is it's a pretty extreme way to draw that line.

但鉴于你提到的这一切，我们现在根本不清楚，也没有任何监管，甚至还没有得到行业普遍认可。

But given all of the stuff you're saying we just don't know now and isn't regulated and isn't even sort of industry accepted yet.

我认为，我们实际上向谷歌、脸书等公司提供了大量信息。

I think that there are a lot of ways that, you know, we give a lot of information to Google and we give a lot of information to Facebook and whatever.

但至少现在，这些数据的处理方式已经有了公认的规范，如果这些规范改变，将会带来严重的问题。

But there are at least now sort of accepted norms in how that data is treated, and there would be real problematic ramifications if that changed.

但在人工智能领域，目前似乎没有任何类似的规范存在。

It doesn't seem like any of that exists in AI right now.

而且这也很困难，因为即使这些公司确实有这些保护措施，它们通常还是会保留一部分你的数据，哪怕只是匿名化或去除了个人信息。

And it's also hard because even if they do have those protections in place, you know, like most of these companies do retain some semblance of your data, even if it's like, you know, anonymized or the personal stuff is stripped out.

它们通常还是会以某种方式使用这些数据。

They like use it in some degree usually.

在过去十年里，去匿名化数据其实很容易。

And so and we've seen like in the past like ten years, it's pretty easy to de anonymize data.

而且，系统如何判断什么是敏感信息、什么不是，这门科学本身也不完善。

And it's also imperfect science on like what a system knows is sensitive versus not.

Darktrace的玛格丽特·坎宁安告诉我，聊天机器人很难区分电话号码和社保号码，或者街道地址和账户号码。

Margaret Cunningham from Darktrace was telling me it's hard for chatbots to tell the difference between a phone number and a social security number or a street address and an account number.

所以问题在于，即使他们尽了最大努力，这些安全防护措施也并不完美。

So it's tough because one, even if they're trying their best, the like guardrails here are not perfect.

即使这些防护措施效果很好，你仍然可以对数据进行去匿名化处理。

And even if they do work great, you can de anonymize data.

我不知道Chatuche Beauty Health是否真的能做到这一点，但总的来说，大家普遍认为，用于保护个人数据的匿名化系统非常非常不完善。

You know, I don't know for sure if Chatuche Beauty Health would be able to do that, but I'm just saying in general, it's a pretty understood rule that anonymization systems for protecting personal data are very, very imperfect.

因此，在做决定之前，真的需要清楚了解其中的风险。

So it's like, just really need to know the risks here before you make a decision.

如果你确实想提供你的数据，那也没问题。

And if you do want to give your data over, that's fine.

你只是需要以一种知情的方式去做，而不是像摘掉安全带那样无所谓地说：‘这可能会在十年后找上门，但我根本不在乎。’

It's just you need to do it in an informed way without just like taking off your seatbelt and being like, whatever, know, this may come back to buy me in ten years, but I don't care.

如果你就想这样，那也没问题。

If you wanna be like that, sure.

但请带着清醒的认知去做。

But do it with an informed take.

这正是我所要求的。

That's all I'm asking.

是的。

Yeah.

我属于那种在大学期间，把别人在每个派对上拍的每一张照片都发到Facebook上的世代。

I am very much of the the generation that shared every photo that anyone took at every party in college on Facebook.

几年后，我们才意识到，必须回头彻底清理所有在Facebook上分享过的照片。

And then, boy, did we all learn several years later to go back and pretty ruthlessly comb through all of the pictures that we had shared on Facebook.

我觉得这一代人也会经历完全相同的事情。

And it feels like this is the generation that's gonna go through that exact same thing.

我记得我以前常和朋友爬到我高中的屋顶上。

I remember I used to climb on my high school's rooftop with my friends.

那真的特别有趣。

Like, it was, like, so fun.

我们会爬过藤架，待在上面闲逛，但我从来不敢把照片发到Facebook上，因为我妈是那所学校的老师。

We would, like, go up the trellis and just hang out up there, and I could never share the photos on Facebook because my mom was a teacher at the school.

那栋楼现在已经被拆了，所以我现在可以讲这个故事了。

The the The building has now been torn down, so I can share that story.

但说实话，幸好我当时决定没把这些照片分享出去。

But it's like, yeah, I mean, that was like thank God I decided not to share those.

但我们当时对一切都毫不在意，因为我们成长于互联网时代，渴望别人在我们的墙面上留言，喜欢我们的Facebook相册。

But yeah, we were just being super willy nilly about everything because we grew up in the age of the Internet and we wanted to have people writing on our walls and like, you know, liking our Facebook albums.

所以，是啊，这确实挺难的。

So yeah, it's tough.

我觉得人们应该在这里也运用同样的思考方式，即使感觉像是私密的一对一对话，但我们都见过ChatGPT的记录被公开，因为链接变得可搜索，公司高管甚至把公司的财务数据输入了系统。

Like, I think people should, you know, kind of apply the same thought process here, Even though it feels private because it seems like a one on one conversation, you know, we've seen like ChatGovity records go public because like the link became searchable, you know, and like company execs were like giving financial data from their company into the system.

然后任何公众人物都可以搜索到。

And then like any public person could search it.

这一点后来已经修复了，但类似的事情仍可能发生。

That has since been fixed, but things like that can happen.

你永远无法预知它们会在何时、何地或为何发生，因为这项技术还相对新颖。

And you never know how, when, or why they're gonna happen because this technology is relatively new.

是的。

Yeah.

我经常看到很多人担心和害怕的一件事是，这些公司会用我的数据来训练他们的模型。

One thing I see a lot of people wondering about and and being fearful of is this idea that these companies are gonna use my data to train their models.

这既是我们在交互中的现实情况，也像你所说的，我上传到ChatGPT的重要财务数据会被用来训练下一个版本的ChatGPT，这在某种程度上构成了隐私风险或泄露。

And that's both sort of the the facts of our interaction, but also, like like you said, the the the important financial data that I upload into ChatGPT is going to be used to train the next version of ChatGPT and that that is a a privacy risk or breach in some way.

你怎么看这个问题？

What do you make of that?

人们应该如何思考自己的哪些数据被用于训练AI模型，以及这在隐私方面意味着什么？

How should people think about what of their data is being used to train AI models and what that means privacy wise?

这很难，因为这些公司会在措辞上非常谨慎，你知道的？

It's hard because these companies will be very careful with their wording, you know?

所以你永远无法完全了解他们是否、如何或为何使用你的数据来训练他们的模型。

So you never really know the full extent on how they're how or why or if they're using your data to train their models.

比如，他们明确表示，ChatGPT Health会严格保密你的健康数据，不会用于训练他们的AI模型。

Like, they will say, for example, chat GPD Health, they say explicitly, your health data will be kept confidential and it won't be used to train their AI models.

但这是否意味着，你所说的话的某种匿名化、简化版本仍会以某种方式被用于模型训练呢？

But does that mean some, anonymized stripped down version of what you say won't be in some way used to train the models.

我不知道。

Don't know.

你知道，他们就是不这么做，这才是关键。

You know, they they don't that's the thing.

他们根本不会深入解释具体如何操作，而且他们也没有义务这么做，因为这一切都是自愿的。

They don't really go into the the how here, and they don't have to because it's all voluntary.

所以另一点是，这些规定随时可能改变，你知道的，他们随时可能改变主意。

So the other part of it is that can change, you know, they can change their mind at any time.

所以我会说，你知道，如果他们明确表示，对于任何特定服务，我们不会使用您的数据来训练我们的模型。

So I would say, you know, if they explicitly say for any given service, we don't use your data to train our models.

在大多数情况下，你可以相当确信他们确实没有这么做，至少对于这些服务是这样，但这种情况可能会改变，而对于那些没有明确说明的公司，情况就更不确定了。

You can be pretty sure that they don't for the most part, at least, for that, but that may change and certainly not for the ones that they don't explicitly say that.

这里需要记住的另一点是，如果一个产品是免费的，那么你就是产品本身。

The other thing to keep in mind here is that if it's a free product, you are the product.

所以，你知道，如果你为某个产品付费，他们使用你的数据进行训练的可能性就较小，或者至少程度较轻。

So like, you know, if you're paying for a product, there's less of a chance they're using your data to train or at least to a lesser extent.

但如果它是免费的，那一切就都难说了。

But if it's free, like all bets are kind of off.

我们从OpenClaw身上学到了这一点，也从之前无数类似产品中得到了同样的教训。

That's what we learned with OpenClaw and what we've learned with like a million products way before that.

但是的，如果你付费，你会稍微安全一点。

But yeah, would say you're a little bit safer if you're paying.

如果你是企业的用户，那就安全得多。

And if you're an enterprise user of something, you're way safer.

你知道，Chatuch Beauty Health 这家公司特别安全，因为他们在这方面说得非常明确，但这只限于现在。

You know, Chatuch Beauty Health specifically, you're pretty safe because they're pretty explicit about that stuff, but that's just for now.

谁知道呢？

Who knows?

你知道，Anthropic 也有类似的产品，是符合 HIPAA 标准的，但即便如此，你也不能完全放心。

You know, Anthropic has a similar product, like that's HIPAA compliant, but still like, you know, you're that's you're not bound.

所以我也说不准。

So I don't know.

这事儿确实很难办。

It's it's just a tough thing.

你在这里做事得稍微打点折扣，保持一点警惕。

You need to kinda operate with with a little bit of a grain of salt here.

是的。

Yeah.

这很好。

That's good.

我能读给你看一个我找到的、让我困惑了很久的吗？

Can I can I read you one that I found that just like flummoxed me forever?

可以。

Yes.

我觉得这正好能证明你所说的话。

And I think is a is a good proof of what you're talking about.

这是来自Anthropic对Claude的服务条款。

So this is from Anthropic's terms of service for Claude.

它说：我们不会用您的Gmail或日历集成数据来训练我们的模型，以确保您的私人信息保持私密。

It says, we do not train our models on your Gmail or calendar integration data, ensuring your private information remains private.

简单明了。

Simple, straightforward.

对吧？

Right?

再说一遍，很多这类情况都源于我使用Claude。

Again, this is so much of this comes from, I use Claude.

我喜欢在Claude中查找信息时，能够从我的Google Drive和Gmail中提取数据的这个想法。

I like the idea of being able to pull information from my Google Drive and my Gmail in Claude as I look for things.

我会这么做吗？

Do I do this?

所以我正在读这个，查这个。

So I'm reading this, looking this up.

这句话完全说得通。

That sentence makes perfect sense.

对吧？

Right?

我们不会训练我们的模型。

We do not train our models.

这是下一句话。

This is the next sentence.

请注意，如果您正在使用我们的消费者产品，例如免费版、专业版和Max版的Claude，在使用这些账户的Claude Code时——顺便说一下，我刚才读到的那对双括号——并且您已选择允许我们使用您的聊天和编码会话来训练模型，那么您从Gmail、日历或Claude回复中复制粘贴的任何内容，包括这些集成中的具体信息，都可能被用于改进我们的模型。

Note, if you are using our consumer products, e g Claude, free, pro, and max, when using Claude Code with those accounts, that was a double parenthesis I just read to you by the way, and you have chosen to allow us to use your chats and coding sessions for model training, then any content you copy paste from your Gmail or calendar or Claude responses, which includes specific information from these integrations may be used to improve our models.

什么？

What?

没错。

Exactly.

这正是我想说的。

This is what I'm saying.

就好像你真的无法知道。

It's like it's like you can't really know.

他们会给你设置各种双括号陷阱。

They will be pulling all sorts of double parentheses on you.

是的。

Yeah.

他们会使用双重否定。

They will be doing double negatives.

你根本无从知晓。

You just don't know.

所以那也是我的观点。

So that that's my thing too.

就像，是的。

It's like, yeah.

好的。

Okay.

比如，它可能不会直接使用你的邮件，但如果你把邮件里的内容复制粘贴进去，根据你刚才说的，它似乎会使用这些内容。

Like maybe it's not gonna directly use your emails, but if you're copy and pasting things from your email into it, okay, it'll it seems like it'll use that based on what you just said.

那如果它从你的邮件中提取内容，并说：‘这是你今天收到的所有邮件的摘要’呢？

And, what if it's returning stuff from your email and saying, Hey, here's a summary of all the emails you got today.

好的。

Okay.

看起来这些内容也会被用来训练模型，所以就是这样。

It seems like that's also going to be used to train so it's like, okay.

而且，正如我所说，他们并没有提到企业版。

And again, like I said, they didn't mention enterprise there.

所以企业产品相对来说比较安全，但消费类产品你就真说不准了。

So it's like enterprise is pretty safe, but the consumer products, you don't really know.

这真的很棘手，因为这里没有太多明确固定的规则。

It's just tough because there's not a lot of hard and fast rules here.

而且他们随时可以更改这些条款，你刚才读到的内容，可能一两周后就会有点不一样。

And the fact that they can change these things at any time, the thing you just read, maybe that'll look a little bit different in a week or two.

我目前设置了一个追踪器，用来监控这些公司更改其使命宣言的情况。

I have a current, like a tracker set up for when these companies change, like their mission statements.

你根本不知道我有多频繁地收到提醒，说‘哦，这个稍微改了’。

And like, you have no idea how often I see an alert that's like, oh, this changed slightly.

你知道吧？

You know?

我的意思是，通常都是一些琐碎的事，比如他们删掉了一两个括号，但有时候也不是。

I mean, usually it's something dumb, like they took out a couple parentheses, but sometimes it's not.

所以是的。

So Yeah.

是的。

Yeah.

我认为我们应该把这些文件视为动态的、草稿性质的、不断变化的文档。

I think it's like, it's a con we should treat these documents as like living documents that are drafts and constantly changing.

如果你不能接受几个月后这项政策看起来有所不同，那我通常会采取谨慎的态度。

And if you're not okay with, you know, that policy looking different in a couple months, you know, on the side of caution is what how I operate.

是的。

Yeah.

这很有道理。

That makes sense.

是的。

Yeah.

对我来说，整个实验一个非常有趣的结论是，这一切最终都指向了Gemini，因为Gemini提供了许多相似的功能。

I think one really interesting outcome of this whole experiment for me has been that it it all sort of leads to Gemini in a really funny way because Gemini offers a lot of the same things.

对吧？

Right?