这是判断文字是否由AI生成的方法

我的意思是，暂且扮演一下反对者的角色，意图在识别AI生成的低质内容时重要吗？比如，你确实可以遇到一些恶意行为者，他们可能试图影响人们对某个话题的看法，于是他们在TwitterX上创建了大量机器人，用AI大量发布支持其观点的低质内容。

I mean, just to play devil's advocate for a second, does intent matter when it comes to identifying AI slop in the sense that, okay, I get you can have a bad actor who's maybe trying to influence how people feel about a particular topic, and maybe they've created a bunch of bots on TwitterX, and they're using AI to just flood the zone with a bunch of AI slop supporting their particular viewpoints.

另一方面，如果你是一名记者，你的工作就是撰写关于新闻话题的清晰易懂的稿件——我要澄清一下，我完全不支持这种做法。

On the other hand, if you're a journalist and your business is to write, you know, like basic understandable copy about a news topic, Just to be clear, I'm not advocating this at all.

但这种意图和通过纯粹的数量来影响舆论是完全不同的。

But that intent is very different to I'm gonna try to influence something by just sheer volume.

是的。

Yeah.

当然，这两种情况的严重程度相差很大。

Definitely, these are one is a lot more severe than the other.

但与此同时，如果你是一名记者，却用AI来逃避工作、不亲自撰写内容，我认为这也是个问题。

But I think at the same time, if you're a journalist and you're using AI to basically shirk your work and not do your work, I think that's also a problem.

这对媒体机构来说也是一种声誉风险，因为读者能察觉出来，而且他们会公开指出。

And I think it's a reputational risk to the outlet because people can tell and people are gonna call you out.

无论AI生成的内容来自哪里，都有很多人不想阅读这些低质内容。

There's a lot of people who don't wanna read AI slop regardless of where it's from.

是的。

Yeah.

这确实是真的。

This is definitely true.

你最终会用完可以用来训练的人类素材吗？

Are you ever gonna run out of human material to change on?

如果你找到一段发布于2023年之前、尤其是2019年或更早的互联网文本，你可以非常确定这是人类生成的。

You could be pretty confident that if you find some piece of text that was published on the internet prior to 2023, but certainly prior to 2019 or something like that, you could be extremely sure that this was human generated.

你是否担心未来会更难确定训练数据的来源？

Do you worry that in the future it's going to be harder to even establish the provenance of your training data?

是的，这确实是我们关注的问题。

Yeah, it's definitely a concern for us.

跟我们说说你们是怎么考虑这个问题的。

Talk to us about how you're thinking about that.

我们拥有一个近乎无限的2023年之前的预训练数据储备。

So we have a near infinite data reservoir of pre-twenty twenty three data.

我们有足够的数据来长期训练。

There's just more than enough for us to train on for a long, long time.

但问题的一部分是我们也希望用现代文本进行训练。

But part of the problem is we also want to train on modern text.

现在大家都在讨论，如果有人在写关于大语言模型或人工智能的内容，我们不希望错误地将这些内容标记为AI生成，因为我们的训练数据对这个话题没有认知。

There's all this talk about if somebody's writing about LLMs or about AI, we don't want to incorrectly flag that as AI because our training data has no sense of this topic.

我认为我们正在探索不同的方法来解决这个问题，但大多数方法都是在确定谁是可信赖的主体，谁在发布人类撰写的内容，我们可以在一定程度上利用我们的模型来识别这些主体。

I think we're looking at different ways to do this, but most of them are just figuring out who is a trusted actor, who do we know is putting out human written content, and we could use our model for that to some degree.

所以我们有已知的可信主体，我们知道他们发布的是人类撰写的内容，我们也可以使用他们的数据。

So we have known actors, we know they're putting out human written content, and then we could use their data as well.

一个稍微随机的问题，使用你们的模型，你们能估算出目前互联网上有多少比例是AI生成的垃圾内容吗？

Slightly random question, using your model, are you able to quantify what percentage of the internet at the moment is AI slop?

大约占40%。

It's about 40%.

哇。

Wow.

基于什么？

Based on what?

你是怎么得出这个数字的？

How'd you get that number?

互联网上很多都是SEO优化的文章。

A lot of the internet is just SEO written articles.

这些文章基本上是为了让网站在搜索中更频繁地出现，通过针对特定关键词来实现的。

It's articles written search basically so that your website comes up more often in search because it's targeting certain keywords.

很多这个行业已经转向使用AI，因为这样就不必支付写手费用，只需花极少的成本就能大量生成文章。

A lot of that industry has switched over to using AI because then instead of having to pay writers, you could churn out articles for pennies on the dollar.

但我认为这导致了互联网上大量内容都是AI撰写的。

But I think that kind of results in a lot of the Internet being AI written.

这还稍微有点取决于平台。

It's a little bit it's also kind of platform dependent.

从网页角度来看，大约占40%。

It's about 40% from an Internet page perspective.

大约一年半前，我们研究了Medium，发现超过50%的新发表Medium文章都是由AI生成的，这个比例高得惊人。

About a year and a half ago, we looked at Medium and found that over 50% of newly written Medium articles were AI generated, which was a crazy high number.

那Reddit呢？

What about Reddit?

Reddit一年前大概是7%，我认为现在略高于10%。

Reddit, it was 7% a year ago, I believe, a little over 10% today.

等等。

Wait.

实际上，这让我想起来一件事。

Actually, this reminds me.

我经常上Reddit，现在非常享受这个平台，但我确实担心其中有多少内容是AI生成的。

So I'm on Reddit a lot, and I really enjoy it nowadays as a platform, but I do worry about how much of it is being generated by AI.

我不太明白的是，到底有什么经济动机让人去大量发布AI生成的帖子并获得点赞？

And the thing I don't necessarily understand is what are the economic incentives to actually write a bunch of AI generated posts on Reddit and get upvoted?

这种系统或动机为什么会存在？

Why does that system or motivation even exist?

有一些初创公司。

There are startups.

我不会点名具体公司，因为我不想为它们做宣传，但它们会向企业承诺：我们会为你带来Reddit上的自然曝光。

I'm not going to name names because I don't want to promote them but they will sell a promise to companies that we're going to get you organic mentions on Reddit.

我们会运行看起来像真人一样的AI机器人，它们会自然地推荐你的产品，或在评论或帖子中提及你的产品。

We're going to run our AI bots that seem organic, and they're just going to naturally recommend your product or just mention your product in the comments or in a post.

所以我见过这方面的证据。

And so I've seen evidence of this.

我们可以找到这些机器人，它们本质上就像机器人农场，主要进行看似自然的互动，发一些简短回复，有时还会进行品牌提及。

We can find these they're basically like bot farms that are mostly engaging seemingly organically, just doing a short reply, and then sometimes they're doing this brand mention.

因此，这些帖子非常有价值。

And so that's why these posts are very valuable.

这真的很有趣。

That's really interesting.

我还得想象，这些内容之所以有价值，是因为所有模型都用Reddit的数据进行训练。

I have to also imagine it's valuable because all of the models train on Reddit.

对吧？

Right?

如果你希望你的产品名称出现在模型输出中，比如‘最好的鼻毛修剪器’之类的？

And if you want your product's name to appear in model outputs, it's like, what is the best nose hair trimmer or whatever?

Reddit上有很多机器人讨论过这款鼻毛修剪器，因此它更有可能出现在ChatGPT的回复中。

And there's a bunch of bots that on Reddit talked about this nose hair trimmer, and then that's probably more likely to show up in a ChattyPety request.

对吧？

Right?

是的。

Yeah.

对。

Yeah.

这已经被奇怪地操纵了。

It's been weirdly gamed.

你知道，以前你只是在谷歌上搜索‘最好的鼻毛修剪器’，但现在却有成千上万篇文章

You know, you used to just Google best nose hair trimmer, and now there's, like, a thousand articles

现在这些结果之所以排在前面，是因为人们都在那里寻找。

that results, like, show up first nowadays because that's where people are looking.

是的。

Yeah.

然后人们开始搜索‘最佳鼻毛修剪器 Reddit’，以获取关于它的Reddit评论。

And then people started searching best nose trimmer Reddit to to get their Reddit comments on it.

现在人们意识到，这就是人们在搜索的内容，因此你需要在Reddit上投放广告。

And now it's people have realized that that's what people are searching for, so you need to populate Reddit with your advertisements.

我是在《男性健康》杂志上看到的。

I'm I'm on the men's health.

你在找鼻毛修剪器吗？

Are you looking for nose hair trimmers?

松下耳鼻毛修剪器是《男性健康》杂志的首选。

The Panasonic ear nose hair trimmer is the number one choice on men's health.

优点是握感舒适。

Pro is easy to hold.

总之，不是这样的。

Anyway, it's not.

所有这些联盟链接

All these affiliate links

彻底毁了互联网。

just destroyed the internet.

我知道。

I know.

这真的很遗憾，但也没办法。

It's really too bad, but whatever.

再多跟我们讲讲整个流程吧。

Talk to us more about the whole pipeline.

我对这个想法非常感兴趣。

So I'm very fascinated by this idea.

就像是，你看到了一家丹尼餐厅的评价。

It's like, okay, you see this review for Denny's.

让AI模型尽可能地复制它。

You have the AI model try to replicate it as best as it could.

会有一些细微的差异。

There'll be these subtle differences.

跟我们讲讲整个流程。

Talk to us about the whole pipeline.

你们还用了哪些其他测试来验证真实性？因为我觉得你们想做的是找到最相似的数据集，其差异几乎难以察觉，从而进行极限压力测试。所以，跟我们详细说说这个完整流程吧。

What are the other tests that you're using to get the true because what I imagine you're trying to do is get the most similar data sets with an almost imperceptible difference to really stress test So the whole talk to us really about this whole pipeline.

是的。

Yeah.

所以我们真正想做的是，我们

So what we're really trying to do here is we're

作为一名模型开发者，我正在尝试，不。

As a model maker myself, I'm trying No.

不。

No.

抱歉。

Sorry.

继续说。

Keep going.

对。

Yeah.

作为AI专家。

As an AI expert.

对。

Yeah.

对。

Yeah.

作为AI专家。

As an AI expert.

我需要听

I need to hear

一些该领域的技巧。

some tips of the field.

是的。

Yeah.

所以我们真正寻找的是那些尽可能接近人类与AI边界的事例，这样我们的模型能学得更好。

So what we're really looking for is examples that are as close to the boundary between human and AI as possible so that our model learns better.

非常明显是AI生成的内容，我们的模型就学得不多。

Something that's very obviously AI is our model's not learning as much.

对于明显是人类生成的内容也是如此。

Same thing for something that's obviously human.

第一步是创建一个包含人类示例的合成镜像的数据集。

Step one is creating this dataset with synthetic mirrors of human examples.

然后我们训练一个模型。

And then we train a model.

第二步是所谓的主动学习。

And then step two is something called active learning.

然后我们用这个模型去扫描一个更大规模的数据集，查找错误、假阳性、假阴性，并将这些样本重新纳入训练集，从而训练出一个更好的模型，因为它已经见过这些错误，而这些错误我们认为更接近人类与AI之间的边界。

So we then take this model and use it to scan a much larger corpus of data and look for errors false positives, false negatives and then we pull those back into our training set are able to train a much better model because it's seen these errors, and these errors, we believe, are just much closer to the boundary between human and AI.

抱歉，为了明确一下，第一轮是已知的人类写作和已知的AI写作。

Sorry, just to be clear, the first pass is like, okay, you have known human writing and known AI writing.

你训练一个模型。

You train a model.

然后第二轮是未知的人类写作和已知的AI写作。

And then the next pass is once again unknown human and known AI writing.

你已经知道每个样本的答案，因此可以列出哪些地方判断错了，然后把这些结果反馈回第一轮模型。

You already know the answer of each of these, and therefore you could come up with a list of which you got wrong, and then that gets fed back into the first version.

没错。

Exactly.

一旦我们重新训练，模型就会变得好得多。

Once we retrain, then the model gets much, much better.

然后我们可以反复进行这个过程，让模型自我提升，每次训练都能变得更好。

And then we could do this as many times as we want to kinda just have a self improving model that gets better with every training run.

我还可以再深入讲讲我们如何处理人工智能编辑的问题，因为我认为

I can also tell you go a little bit more into how we deal with AI edits because I think

这是一个

that's an

日益重要的问题。

increasingly important problem.

我认为未来大多数写作都会有人工智能提供辅助。

I think most writing will be AI assisted in the future.

我觉得这个功能已经在谷歌文档里了，而且我用的谷歌输入法里也有。

I think it's already in Google Docs and it's in my Google keyboard.

语法检查工具Grammarly其实早在做类似的事情了。

Grammarly arguably has been doing this for a while.

没错。

Exactly.

Grammarly在后端就用到了大语言模型，我们总不能直接就断定‘现在所有的写作内容都是AI生成的’吧。

Grammarly uses LLMs on the back end, and we don't want to just say, All writing is AI now.

我们希望能够区分AI辅助和AI生成的内容。

We want to be able to differentiate between AI assisted and AI generated.

因此，我们使用了不同的提示语。

So what we do is we also have different prompts.

在人工审核Denny的评论时，我们不会说‘生成一篇这样的评论’，而是说‘帮助改进这篇’。

So for our human review of Denny's, rather than saying, Generate a review like this, we could say, Help improve this.

让它更正式一些。

Make it more formal.

修正语法错误。

Clean up the grammar.

因此，我们有一份很长的AI编辑提示列表，然后可以查看原始人类文本与...

And so we have a long list of AI editing prompts, and then we're able to look at the cosine difference, the distance between the original human text and

在高维空间中AI生成文本的距离。

The distance AI in hyper multidimensional space.

没错。

Exactly.

那么，AI改变了这段文本多少呢？

So how much did AI change this text?

而且

And

然后我们就能训练我们的模型，根据这个距离打点，判断这是中等程度的AI辅助、轻度AI辅助还是重度AI辅助。

then we're able to train our model to say, we're just going to put a point on this distance and say, This is moderate AI assistance, this is light AI assistance, this is heavy AI assistance.

很有趣。

Interesting.

我打算做一件我觉得自己以前从未做过的事——问一位创始人关于他们公司的使命。

I'm going to do something I don't think I've ever done before, which is ask a founder about their corporate mission.

但你创办了这家公司，当你思考自己在这里想要实现的目标时，这只是单纯的AI检测吗？比如，可能只有像教师这样的少数群体觉得这很有价值？

But you've set up this company, and when you think about what you're trying to do here, is it just basic AI detection in the sense that there might be, you know, a few groups of people like teachers that find this very valuable?

还是说，你的使命更宏大，你实际上是在试图改善互联网以及人们在上面看到的内容？

Or is the mission something broader where you're actually trying to improve the Internet and what people see on it?

我相信，能够检测AI生成内容的技术具有巨大的价值，这种价值不仅对教师有用，对每个行业的每个人都有用，包括律师、出版商，以及任何在互联网上消费内容的普通用户。

I believe the technology of being able to detect AI generated content is immensely valuable, and it's valuable not just for teachers, but for basically everybody in every profession, lawyers, publishers, just an individual who consumes content on the Internet.

我认为这对所有这些人来说都很有价值。

I think it's valuable for all these people.

但最终，是的，我们的高层次目标是帮助缓解日益增长的AI内容带来的某些负面影响。

But ultimately, yeah, our high level goal is to help mitigate some of these negative effects of growing AI content.

但举例来说，仅以产品评论为例，你们的愿景是否是像Yelp这样的平台希望使用这项技术来确保其系统不被操纵？

But for instance, just using the product review example, is the vision that, like, a Yelp, for instance, would wanna use this technology to make sure that its system isn't being gamed?

还是说，你们的愿景是，如果我是一个特别认真的消费者，有很多空闲时间，想出去吃饭，我可以把各个餐厅的评论都输入Pangram，从而真正判断出哪些是真实的口碑，哪些是炒作？

Or is the vision, like, if I am a particularly diligent consumer who has a lot of time on my hands and I'm looking to go out to a restaurant, I can run all these individual restaurant reviews through Pangram and then, like, actually figure out if it's real hype or not?

所以我认为目前主要是前者。

So I think right now it's a lot of the former.

我们与平台合作。

We work with platforms.

我们最大的客户之一是Quora，他们将大量内容通过Pangram进行处理。

One of our biggest customers is Quora, and they run a bunch of content through Pangram.

但我们还有许多其他平台使用Pangram来帮助审核并识别AI不良行为者，将他们从平台上移除。

But we have a lot of different platforms that use Pangram to help moderate and find AI bad actors and get them off their platform.

但我认为，个人消费者这一场景也在快速增长，我们非常希望在此领域推进。

But I also think, yeah, the individual consumer case has been growing a lot, and we're really interested in pushing here.

这是汤姆·基恩邀请您收听彭博市场观察播客。

This is Tom Keene inviting you to join us for the Bloomberg Surveillance Podcast.

我们每天致力于让您在商业上更明智。

It's about making you smarter every business day.

我是保罗·斯威尼。

I'm Paul Sweeney.

我们为您提供全面的覆盖，

We bring you complete coverage of

美国市场开盘。

The US market open.

我们覆盖股票、债券、商品，甚至加密货币，所有您需要的资讯，助您脱颖而出。

We cover stocks, bonds, commodities, even crypto, all the information you need to excel.

我是亚历克西斯·克里斯托弗。

And I'm Alexis Christophris.

Bloomberg Surveillance 还为您带来新闻背后的深度分析。

Bloomberg Surveillance also brings you the analysis behind the headlines.

我们通过与经济学、金融、投资和国际关系领域最杰出的人物对话来实现这一点。

We do that through conversations with the smartest names in economics, finance, investment, and international relations.

我们每天工作日都会实时进行这些讨论，为您呈现每日播客中最优质的分析。

We do all this live each and every weekday that bring you the best analysis in our daily podcast.

请在 Apple、Spotify、YouTube 或您收听播客的任何平台搜索 Bloomberg Surveillance。

Search for Bloomberg Surveillance on Apple, Spotify, YouTube, or anywhere else you listen.

在东海岸，午餐时间收听。

On the East Coast, listen at lunch.

而在

And on

西海岸，一醒来就收听。

the West Coast, listen as soon as you wake up.

这就是由汤姆·基恩、保罗·斯威尼和我，亚历克西斯·克里斯托弗里斯带来的 Bloomberg Surveillance 播客。

That's the Bloomberg Surveillance Podcast with Tom Keene, Paul Sweeney, and me, Alexis Christofferis.

今天就订阅吧，无论你在哪个平台收听播客。

Subscribe today wherever you get your podcasts.

Bloomberg Surveillance，每个交易日都值得收听。

Bloomberg Surveillance, essential listening each and every business day.

pangram.com 的免费版本，比如你每天只能获得几次测试之类的。

The free version of pangram.com, like you get a handful of tests a day or something like that.

如果有人拥有无限次数的 Pangram 响应，甚至能以无限规模访问 Pangram API，他们理论上能否训练出一个提示词，然后输入到 AI 中，生成类似人类的写作？

If someone had an unlimited number of Pangram responses and maybe had an access to the Pangram API at infinite scale, could they theoretically learn a prompt that they would then be able to put into an AI to generate human style writing?

我确实有个朋友做过这件事。

I actually had a friend do that.

他把他的 Claude 代码设成了循环运行。

He put his Claude code on a loop.

我给了他一些 API 积分，然后他的 Claude 代码就整晚自动运行，不断尝试生成提示词，试图让输出结果看起来像是人类写的，或者能被 Pangram 识别为人类创作的内容。

I I gave him some API credits, and then his Claude code just basically worked overnight writing a prompt, trying to get it to output something that's human written, or that came back from Pangram as human written.

它确实成功了，但生成的文本相当混乱。

It got there, but the text was pretty incoherent.

所以，是的，它生成的基本上都是长篇的胡言乱语。

So so, yeah, it was producing more or less long gibberish.

它在语法上是错误的。

It was grammatically incorrect.

很多词根本毫无意义。

A lot of the words just didn't really make sense.

因为这正是我最初的想法。

Because this was my first thought.

当我看到这个时，我想，这会是个有趣的实验：看看你是否能收集所有输出结果，找出差异，不断迭代你给AI的提示，最终生成一个让Pangram认为是人类撰写的文本。

When I saw it, I was like, that would be a fun experiment to see if you could take all the outputs, find the difference, and just keep iterating on the prompt you would have to tell AI in order to eventually get an output that looked to Pangram like it was human generated.

是的。

Yeah.

我认为如果你还用一个LLM来评估连贯性，并同时用Pangram和连贯性评估器来给文本打分，是有可能实现的。

I think there's a way to do it if you also had an LLM judge on coherency and use Pangram and the coherency judge both to score your text.

我觉得这绝对可行，我也很期待有人去尝试，因为如果能做到这一点，我们的模型会变得更好、更稳健。

I think it's definitely possible, and I'm excited for someone to try to do it because we could make our model a lot better and more robust if this existed.

乔，我想知道你现在个人的令牌预算大概是多少，以至于你都会考虑这些事情？

Joe, I wanna know what your personal, like, token budget is nowadays that you're even, like, contemplating some of this You

你知道我感觉怎么样吗？

know what I feel like?

我用的是Clod Max套餐。

I have the Clod Max plan.

你知道的？

You know?

而且我上班的时候不会用。

And I don't work, like, when I'm at work.

我不会在任何我的氛围编码项目上工作。

I don't work on any of my vibe coding projects.

你知道吗，小时候，我记得如果我没吃完饭，总会有人说，世界上还有挨饿的孩子。

And you know, like, when we were kids, I don't know if you remember, like, if you didn't eat all your food, like, someone would say, oh, there's, starving kids in the world usually.

我就想，哦，还有挨饿的氛围编码者需要火鸡呢。

I'm like, oh, like, starving vibe coders that need the turkeys.

就像是，你没有意识到，我有四个小时的令牌额度，但我几乎从来不会用满。

It's like, oh, you didn't like, I have this four hour token window, and I'm almost never maxing it out.

我只是在想，世界上另一边有些孩子，多么希望拥有你的令牌，而你却没能充分利用这期间的全部额度。

And I'm just thinking it's like, there are kids on the other side of the world that wish they had your tokens, and you're you're not using all of your tokens for the window.

你怎么敢这样？

How dare you?

当我没有用满我的Clod Max令牌额度时，我会感到有点内疚。

I feel a little guilty when I don't max out my clawed max token program.

我也有Clod Max，而且确实，大多数日子我根本没怎么写代码。

I also have clawed max, and yeah, most days, I'm not doing much coding at all.

我没有用满额度，但有些日子我会用得远远超过。

I'm not maxing it out, Then some days I'm going way

那你对此感到内疚吗？

Do over feel guilty about that though?

哦，好吧。

Oh, okay.

是的。

Yeah.

所以我想问你，写作确实挺有意思，但这种技术在图像和视频生成方面前景如何呢？你肯定经常被问到这个问题。

So can I ask you, writing is kinda interesting, but what are the prospects of this being able to work on, say, and you must get this a lot, image and video generation?

这在理论上是否类似？

Is it at all theoretically similar?

有没有理由认为它能够被复制，还是说这是一个完全不同的问题？

Is there a reason to think that it will be replicable, or is this just a different beast of a problem?

我认为这种方法肯定是可行的。

I think the approach is definitely doable.

我认为一些经济因素会发生变化，尤其是当我们关注当前视频生成的成本时。

I think some of the economics change, especially if we look at video and the cost of generating video today.

好的。

Okay.

我们无法像生成文本那样大规模地生成视频。

We can't generate video at the same scale that we can generate text.

所以我们可能需要一种不同的方法。

And so we might need a kind of different approach.

但我相信，如果我们能为图像加上音频解决这个问题，那可能就足以一并解决视频问题了。

But I also believe that if we're able to solve this for image plus maybe audio, that could be enough to just solve it for video as well.

零样本。

Zero shot.

你能否想象，比如推出某种视频认证计划？

Could you ever envision, I don't know, launching some sort of certification program for video?

因为这看起来像是我爸爸这一代人，花很多时间在Facebook上。

Because this seems to be my dad's a boomer, spends a lot of time on Facebook.

这似乎是社会所需要的。

Like, this seems to be what society needs.

对吧？

Right?

比如，一段视频附带一个小标识，说明这不是AI生成的，并且有人真正盖章认证过。

Like, a video that comes with a little thing that says this is not AI generated, and someone has actually rubber stamped that.

有一个名为C2PA的组织，我认为他们在内容溯源方面做得相当不错。

So there's an organization called C2PA, and I think they're doing pretty good work on content provenance.

他们基本上在与手机制造商和硬件制造商合作，嵌入硬件签名，以证明图像和视频确实是由该硬件拍摄的。

Basically, they are working with phone makers and hardware makers to basically embed hardware signatures to prove that image and video were truly taken from the hardware.

basically 就是水印。

Like watermarks, basically.

是的。

Yeah.

没错。

Exactly.

所以，与其标记AI生成的内容，不如……

So so rather than marking the AI outputs Yeah.

我们反而是在真实拍摄的、真实存在的内容中嵌入一种真实性证明。

We're instead embedding, like, a proof of authenticity in the the, like, thing that's real and was captured in real life.

这很有趣。

That's interesting.

好的。

Alright.

所以从宏观角度看，互联网将走向何方？

So big picture, where's the internet going?

你提到互联网上已经有40%的内容是广告生成的，但也许这也不是世界末日。

You mentioned 40% of the internet is already ad generated, but maybe that's not the end of the world.

如果只是些我从不看的SEO页面，管他呢。

If it's just a bunch of SEO pages that I never read, I don't know, whatever.

给我们谈谈宏观层面的想法吧，无论Pangram和其他AI检测模型的普及程度如何，互联网的发展趋势会怎样？

Give us some thoughts high level about what the trajectory of the internet, regardless of the uptake of Pangram and other AI detection models.

说实话，我对互联网的现状有点担忧。

I'm a little bit worried about the state of the internet, I'm going be honest.

完全同意。

Totally.

我认为，如今互联网的运作在很大程度上依赖于信任和共识，而我们目前并没有准备好应对这种前所未有的、规模巨大的机器人潮。

I think right now, so much of it is built around trust and norms in a way that we're not really well equipped to suddenly deal with an onslaught of bots at a completely different scale than we've dealt with before.

这可能有好的一面和坏的一面。

There's maybe a good case and a bad case.

我认为坏的情况是互联网走向了‘死亡互联网’理论所描述的境地。

I would say the bad case is the internet goes the way of dead Internet theory.

所有开放和可访问的空间都被机器人充斥，人们唯一能进行真实交流的地方，只剩下像Discord服务器这样高度封闭的‘围墙花园’——在那里，每个人的身份都是已知的，你知道这里没有机器人。

Just like every space that's open and accessible is just flooded by bots, and then the only place people are able to communicate authentically is in, like, very walled garden, like, closed servers, like like, Discord servers, for example, where, you know, everybody's identity is known and you know you don't have bots in here.

所以这可能是那种糟糕的 scenario。

So that's maybe the, like, bad scenario.

我能跟你说一个我有过的大胆想法吗？

Can I tell you an insane thought that I've had?

说吧。

Go on.

就只是

Just

我们要把这给踢出去。

we're gonna kick out of this.

就只是只是

Just just

那一个。

the one.

你听说过吗，那种针对不良行为者的概念，我忘了他们怎么叫了。

Heard of, like, I forget what they call, like, this idea of, like, for the bad actors.

它叫什么‘天堂模式’或者‘天堂封禁’。

It's called, like, heaven mode or heaven banning.

你听说过这个吗？

Have you heard of that?

是的。

Yeah.

所以有一个想法，对付网络上的不良行为者的一种方式是，嗯。

So there's this thought that one way you could deal with bad actors Mhmm.

在互联网上，突然间他们被置于一个类似推特的版本中

On the Internet is suddenly they're on a version of, say, Twitter

好的。

Okay.

在那里只有机器人，而且每个人都总是同意他们的一切，这会让他们发疯之类的。

In which they're only bots, and everyone always agrees with them on everything, and it drives them crazy and stuff like that.

而他们永远不会意识到，因为他们会想，哦，这真棒。

And they would never know it because they're like, oh, that's cool.

每个人都在，然后慢慢地，是的。

Everyone's and then it's sort like, slowly, like, yeah.

这是一种惩罚人的方式，把他们放到一个永远不会发生争执的互联网环境中。

They just this is like a way you could punish people by putting them on an Internet where they will never get any fighting.

你被天堂封禁了，被关进了牢笼。

You get heaven banned and put into basically jail.

你正在跟一帮傻子说话。

You're talking to a bunch of fuck.

没错。

That's right.

那将会是被天堂封禁的监狱。

That would be jail where you're heaven banned.

但我想说的是，我自己搭建了这个小型AI模型，然后我给朋友们看了。

But I thought and again, this is you know, like, built this little AI model myself, and I, like, showed it to my friends.

他们说：‘哇，真酷啊，乔。’

They're like, oh, it's really cool, Joe.

我真的很佩服。

I'm really impressed.

我真的很佩服你居然能做成这个。

Like, I'm really impressed by, like, that you're able to do this.

我当时就想：人们是在对我诚实吗？

And I was like, are people being honest with me?

我是不是已经被天堂封禁了？

Have I been heaven banned?

因为我觉得，如果这东西很烂，你大可以坦率地告诉我。

Because I just, like like, you can be honest with me if it sucks.

而且我确实有点这种恐惧。

And I'm and I sort of have this fear.

这是

This is

史上最谦虚的炫耀了。

the biggest humble brag ever.

我是认真的。

I'm very serious.

我做了这件事，大家都觉得它很棒。

I did this thing, and everyone thought it was great.

我只是想说，我觉得人们可能是出于觉得‘真酷’才对我客客气气的，我担心的就是这个。

I'm just saying, like, people are, like I think people I'm worried that, like, people are being nice to me because, like, oh, cool.

是的。

Yeah.

这真的令人印象深刻。

That's really impressive.

你真的做了这件事。

You, like, did that.

我有一种深深的焦虑，担心人们并没有对我坦诚相告。

And I have this, like, deep anxiety that, like, people aren't giving it to me straight about it.

我知道这听起来像是谦虚地炫耀，但真的不是。

I know that sounds like a humble brag, but it's really not.

这就是为什么你永远不能太成功。

That's why you can never get too successful.

坎耶·韦斯特身边全是只会说好话的人。

Kanye West is surrounded by a bunch of yes men.

他从来得不到这样的反馈：哦，这是他第一次尝试用氛围编码做点什么。

He never gets any Oh, this is his first try at doing something with vibe coding.

我非常焦虑。

I'm deeply anxious.

不，你直接告诉我它很烂就行了。

No, you can just tell me if it sucks.

这没问题。

That's fine.

这正是我的担忧。

That's my worry.

我不担心这个。

I I don't worry about this.

如果我发推说我在吃牛排，会有一百个人来批评我。

If I tweet that I'm eating a steak, I will get, like, a 100 people criticizing me

为了一些事情。

for something.

肉类争议。

The meat kerabala.

是的。

Yeah.

所以还有另一件事，那就是你永远不该发推的两件事：肉类烹饪和享受生活。

So that's the other thing, which is that the two things you are never allowed to tweet about, meat preparation and enjoying life.

因为如果你一旦享受生活，或者一旦烹饪肉类，人们就会在网上对你大加抨击。

Because if you ever enjoy life and if you ever enjoy and if you ever prepare meat, people will flip out at you on the Internet.

这两件事是你在网络上绝对不能做的。

Those are the two things that you're not allowed to do online.

非常对。

Very true.

还有一个相关的问题，但让我们回到方法论上。

And sort of related question, but just going back to the methodology.

如果你专注于这种路径依赖的想法，我想象它就像一棵巨大的决策树。

If if you're focused on this sort of, like, path dependent idea, I'm kind of envisioning it as like a giant decision tree.

对吧？

Right?

随着模型变得越来越好，而且我们知道它们已经在输出中注入了一定程度的随机性，这有可能吗？

Is there a possibility that as the models get better and better and we know that they're already injecting, like, some degree of randomness into their output.

虽然我知道肯定会有人较真，给我发消息说：‘你知道计算机根本无法实现真正的随机性。’

Although I know there's gonna be a pedant out there who messages me and says, Well, you know computers can't do true randomness.

暂且放下这一点，要知道它们正在调整。

Setting that aside, know that they're adjusting.

它们正以惊人的速度变得更加复杂。

They're becoming more sophisticated at an incredible rate.

我们知道它们正试图调整并注入一些随机性，以避免这种类型的检测。

We know that they're trying to adjust and inject some randomness in order to avoid exactly this kind of detection.

你担心它们自身的适应能力吗？

Do you worry about their own adaptation at all?

我注意到，随着模型能力的提升，我认为它们的输出分布变得越来越复杂。

I have noticed that the models, as they get more capable, I believe their output distribution gets more complex.

用一个简单的模型来学习更困难，这就是为什么我们一直在扩大模型规模，以捕捉能够表征大语言模型输出的更高复杂度函数。

It's harder to learn with a simple model, which is why we've been increasing our model size to capture a higher complexity function that can capture the LLM output.

所以我认为我们可能需要继续提升我们的模型。

So I think we may have to continue to make our models better.

我们必须努力跟上它们的步伐。

We're gonna have to work to keep up with it.

我们不能躺在功劳簿上睡大觉。

We can't just rest on our laurels.

什么是突发性和困惑度？

What are burstiness and perplexity?

是的。

Yeah.

这是一个被一些AI检测工具使用但Pangram不使用的指标。

So this is a metric that's used by some AI detectors but not Pangram.

好的。

Okay.

我可以解释一下它是如何工作的。

And so I can explain a bit about how it works.

所以困惑度基本上是一种衡量标准

So perplexity is basically a measure

这并不是perplexity.ai那个网站。

of And this is not perplexity.ai, the website.

这是一个技术术语。

This is a technical term.

好的。

Okay.

很好。

Good.

这是一个指标。

This is a metric.

这是衡量一段文本对语言模型而言有多令人困惑的指标。

This is a measure of how confusing a piece of text is to a language model.

基本上，对于每个词元，我们都可以计算出某种困惑度，也就是这个词元有多预期。

Basically, for example, with every token, we can calculate some perplexity, which is basically how expected this is.

例如，如果句子是‘我回家去看我的宠物’，下一个词元是‘龙猫’，那么这个词元的困惑度会比‘我的宠物狗’高得多。

For example, if it's I went home to my pet and then the next token is chinchilla, that would be a much higher perplexity token than my pet dog.

因此，低困惑度的文本，或者说大语言模型的输出，通常具有低困惑度。

So low perplexity texts or really LLM outputs, tend to be low perplexity.

它们不会生成让自己感到意外的输出。

They're not going to produce outputs that are surprising to themselves.

这是一种相当不错的方法，可以实现约90%到95%准确率的AI检测器，但它存在一些问题。

This is a decent way to get an AI detector that's around 90% to 95% accurate, but it has some problems.

最主要的问题是，你无法在此基础上进一步改进。

The main one is that you can't improve upon it.

基本上，它会产生误报。

Basically, it has false positives.

非英语母语者写的文章通常 perplexity 很低，因为当你在

Text written by non native English speakers often is low perplexity just because when you're

学习时，他们不会冒太多风险。

learning They don't take as many risks.

没错。

Exactly.

我很感兴趣。

I'm interested.

是的

Yeah.

这就是为什么许多早期的AI检测器对非母语英语使用者产生了大量误报。

So that's why a lot of the early AI detectors had a bunch of false positives with ESL speakers.

因为他们的文本具有低困惑度。

It's because their text was low perplexity.

所以我认为这是一个非常棒的指标，但它并不是Pangram的路径。

So I think this is a very cool metric, but it is not the path for Pangram.

相反，我们采用了深度学习的方法，以便能够做得更好

Instead, we went the deep learning approach so we can do better than

那什么是突发性？

And what's burstiness?

这仅仅是另一面吗？

Is that just the opposite side of the coin?

是的

Yeah.

突发性实际上我也不知道该怎么定义它。

Burstiness is basically actually, I don't know if I can define it.

好的。

Okay.

行。

Fine.

是的。

Yeah.

没问题。

All good.

突发性听起来就像是那种男性社群术语，对吧？

Burstiness just sounds like one of those of, I guess, manosphere terms, doesn't it?

哦，对。

Oh, yeah.

他一直在用高突发性进行smaxing

He's been look smaxing with high burstiness

或者类似什么‘突发性 Mug’的东西。

or something like Burstiness mug.

是的。

Yeah.

那太好了。

That's great.

是的。

Yeah.

我认为这可能只是衡量句子长度以及文本起伏的一种方式。

I think it might just be a measure of sentence length and how the ups and downs of the text.

但如果我们假设世界普遍关注AI垃圾，并希望对此采取行动，那么在互联网经济、监管或技术（比如你正在开发的东西）方面，最重大的单一改变会是什么，才能真正帮助减少垃圾内容？

But if we assume that the world is collectively concerned about AI slop and wants to do something about it, what would be, like, the single biggest change to the system, either in terms of, like, the economics of the Internet or regulation or technology, like what you're developing, that would actually help reduce slop?

是的。

Yeah.

我认为最重要的是规范。

I think the biggest one is norms.

已经有一些很棒的博客文章讨论了，未经告知就将AI生成的内容发给别人是一种不礼貌的行为。

There have been a couple great blog posts written about how it is rude to send other people undisclosed AI outputs.

我认为我完全同意这一点。

And I think I completely agree here.

我觉得如果有人在互联网上提问，另一个人却去用ChatGPT搜索答案再粘贴回去，这有点不礼貌。

I think if somebody asks a question on the internet and then somebody else goes and puts into ChatGPT and then paste the answer, It's kind of rude.

我当时是想问问我的朋友或关注者们的看法，而不是去问ChatGPT。

I was going here to ask the opinions of my friends or my followers, not ChatGPT.

我自己本来就可以做这件事。

I could have done that myself.

因此，我认为在这样一项新技术出现时，建立这种规范非常重要，我们需要快速行动，但我相信这对社会会有很大帮助。

And so I think building this norm is something that it's very new technology, so we need to do it quickly, but I think this would help a lot for society.

那么，这实际上引出了我一个疑问：我觉得主要的互联网平台实际上正在朝完全相反的方向发展。

Well then, actually, this gets to a question that I have then, which is I feel as though the major Internet platforms are actually moving the exact opposite direction.

我的意思是，我感到非常震惊。

I mean, I'm stunned.

也许我曾经不小心点到了什么，但我经常收到邮件，打开Gmail准备回复时，却看到那里有一段幽灵文字，问：‘你只是想让Gemini来回复这封邮件吗？’

Maybe I accidentally clicked on something at some point, but the frequency with which I get an email and then I open it up to respond in Gmail, and there's that ghost text there that, I do you just want Gemini to respond to this?

我从未这么做过。

I've never done that.

我也认为，那样做会极其无礼。

I also consider I think that would be extremely rude.

我从未用AI生成的回复来回复过任何邮件，但这些系统却在暗示你这么做。

I've never responded to any email with AI response, but they're basically telling you to do that.

它们的做法恰恰相反。

They're doing the exact opposite.

它们正在破坏这些准则。

They're blowing up these norms.

所以我想知道，从你的角度来看，你提到你和Quora合作，但根据你的观察，主要的互联网平台是否认为这是一个值得解决的问题？

And so I'm curious from your perspective, you mentioned you work with Quora, But from your impression, do the major Internet platforms think this is a problem worth solving?

还是说，它们的顾虑是：‘你知道吗？’

Or from their concern is like, you know what?

内容越多越好。

The more content, the better.

对他们来说，这里的激励是混合的。

There's mixed incentives there for them.

这很有趣，因为谷歌似乎在两头下注。

It's funny because Google seems to be playing both sides.

一方面，他们做过一个广告，引发了很多争议，广告里说孩子们现在可以通过AI发送他们对英雄的敬意信件，而不是自己亲手写信。

On one hand, they had that advertisement which people blew up about where it's like, Oh, children can now send their heroes notes on how much they respect them by using AI instead of writing the note themselves.

我觉得这不对。

I'm like, This is wrong.

这在社会层面上是糟糕的。

This is societally bad.

但与此同时，他们正在努力解决互联网搜索结果中的AI垃圾内容，确保用户获得的是真实内容，而不是AI生成的垃圾内容。

But at the same time, they're working very hard to deal with the AI slop on the internet in search results to make sure people get served real content and not AI slop content.

所以我认为，显然有很多激励因素在影响产品人员，他们被推动去推广AI，因为这是公司的使命。

So I think I mean, I think obviously there's a lot of incentives that play around product people who are incentivized to push AI because that is the corporate mandate.

但我觉得，总的来说，即使在我这一圈AI研究人员中，普遍共识也是AI是个强大的工具，但垃圾内容很糟糕。

But, yeah, I think overall, even in my sphere of a bunch of people who are AI researchers, generally consensus is that AI is a powerful tool, but slop is bad.

这让我想起一件事。

This reminds me.

我父母以前总让我亲手制作圣诞贺卡，给所有亲戚朋友。

My parents used to make me do these handmade greeting cards for Christmas, for all relatives and stuff.

这本意是让我展示自己用心沟通的诚意。

And it was supposed to be a demonstration of my commitment to communicating.

我觉得这挺好的。

Think that's great.

但它让我留下了心理阴影。

It traumatized me forever.

因此我特别讨厌贺卡，就是因为以前花几个小时手工制作这些东西。

And I hate greeting cards as a result of them, of doing this, just spending hours manufacturing these things.

但更搞笑的是，后来我们有了电子贺卡，我父母立刻就改用电子贺卡了。

But then secondly, the funniest thing was once we got ecards, my parents immediately switched to using ecards.

而且现在这事儿也特别搞笑。

Just and now this is also the funniest thing.

我爸爸用电子贺卡，他发现电子贺卡系统能告诉他你有没有打开过。

My dad uses ecard he figured out that the ecard system can tell him whether or not you opened it.

所以他现在就用它来日常沟通了。

So he just uses it as, like, day to day communication now.

这太搞笑了。

That's so funny.

我刚给你女儿发了一封邮件，我是通过电子贺卡发的。

I just sent an email to your daughter, and I do it via e card.

你看，我都注意到你还没打开我发的国际热狗日电子贺卡。

It's like, noticed you haven't opened up my e card for International Hot Dog Day.

麻烦告诉我你最近怎么样。

Please, let me know what's going on.

我小时候字写得特别差，我妈就让我亲手写所有感谢信，感谢别人在我成年礼上送的礼物。

I did terrible handwriting as a kid, and my mother made me write all of these handwritten notes to thank people for the gifts I got for my bar mitzvah.

是的

Yeah.

我讨厌那样。

I hated it.

但你知道吗？

But you know what?

但你知道吗？我与那些人建立了深厚的联系，这些联系一直延续至今。虽然那痛苦的一周里我手写到手腕酸痛，但我觉得这一切都值得。

I have deep connections with all of those people that have lasted And over the that miserable one week where I just rode and I got, you know, hand cramped, I think it paid off.

好吧。

Alright.

想象一下，差不多十六年里，一直不停地做这种事，永无止境。

Well, imagine doing that for, like, sixteen years, basically, in a never ending stream.

马克斯·斯佩罗，非常感谢你做客《Odd Lots》。

Max Spero, thank you so much for coming on Odd Lots.

这真是太有趣了。

That was a lot of fun.

我对这场对话非常着迷。

I'm fascinated by this conversation.

非常感谢你邀请我。

Thanks so much for having me.

是的。

Yeah.

讨论这个话题真的很令人兴奋，我认为垃圾信息正在成为一个日益严重的问题。

Really exciting to talk about this, and I think Slop is a growing problem.

所以希望我们能做得很好。

So hopefully Awesome.

我们能够应对这个问题。

We're able to deal with it.

互联网上有40%的内容。

40% of the Internet.

我不确定我是对这个数字感到惊讶，还是不惊讶。

I can't tell if I'm surprised by that or not.

那么明年这个时候会变成什么样呢？

And what's it gonna be next year at this time?

天哪。

Oh, man.

我不知道。

I don't know.

可能会这样，很难说。

It'll be, like Hard to say.

占多数。

Over a majority

肯定大部分都是。

of them for sure.

是的。

Yeah.

几乎可以肯定。

Almost certainly.

太疯狂了。

Crazy.

好的。

Alright.

谢谢你参加《Odd》节目。

Thanks for coming on Odd

很多。

Lots.

谢谢。

Thanks.

特蕾西，我喜欢这场对话。

Tracy, I love that conversation.

我只是觉得这就像一个非常有趣的谜题。

I just think it's like a really fun puzzle.

对吧？

Right?

不。

No.

完全正确。

Totally.

这看起来像是一个很有趣的谜题要解决。

It's very like, it seems like a fun question to solve.

我对这个想法着迷：无论是人类还是人工智能，我们所知道的和我们能够表达出来的之间，必然存在一道鸿沟。

And I'm fascinated by this idea of how, like, with both humans and AI, there is gonna be this gap inevitable between what we know and what we can articulate.

因为你

Because you

我和你，撇开人工智能和文本不谈，我们都知道一些事情。

and I, setting aside AI versus text, there are things that we both know.

例如，这个有新闻价值，那个没有。

For example, this is newsworthy, this isn't.

这是一个不错的播客节目。

This is a good episode of a podcast.

这个不是。

This isn't.

这个听起来像是个可信的嘉宾，而这个不是。

This is a credible sounding guest, and this isn't.

这种认知和能够解释清楚其中原因之间的差距，就像是你只是隐约知道而已。

The gap between that and then being able to explain why, it's like, well, you just sort of know it.

对吧？

Right?

你就是有种这样的感觉。

You just sort of have this feeling.

这是一种直觉。

There's an intuition.

是的。

Yeah.

这种直觉是通过大量例子积累起来的，某种程度上，这和AI的训练方式是一样的。

And that intuition is built up from numerous examples, which is the same way in a sense that, like, the AI is trained.

就像这些你只能从模式中感知到的东西，你能看到它们，却无法完全清晰地表达出究竟发生了什么。

It's like these things that you only know from patterns, and you can see them without fully being able to, like, articulate exactly what's going on.

那么，我对这一点的另一个疑问是，从长远来看，这真的重要吗？毕竟，互联网的很大一部分已经建立在机器人和虚假注意力经济之上。

Well, the other question I would have on that is, is it even gonna matter in the long run if you think about, like, much of the Internet is already built on bots and the sort of, like, false attention economy?

如果我们的整个世界观都被人工智能驱动的垃圾内容所塑造，

Like, if our entire, like, worldview becomes shaped by AI driven drivel

是的。

Yeah.

如果互联网的经济模式仍然依赖于单个机器人账号之类的东西，那也无所谓了。

Doesn't matter if, like, the economics of the Internet are still attached to individual bot accounts and things like that.

我不确定我有没有把这解释清楚。

I don't know if I'm if I'm explaining this.

但是

But

不。

No.

不。

No.

我觉得这很有道理。

I think it makes a lot of sense.

而且我认为，确实很重要，我们必须彻底改变我们的思维方式，就像马克斯开头说的那样，我也一直在思考这个问题：过去，如果你看到一篇文字，标点符号非常规范，拼写也很完美，读起来逻辑清晰，你就会想，嗯，这肯定是聪明人写的。

And I do think, like, it is important like, we're gonna have to change the entire way we think, and Max said at the beginning, which is and I've thought about this, which is that it used to be that if you came across a piece of writing and the punctuation was excellent Mhmm.

标点和拼写都很出色，读起来也很有条理，你就会觉得，好吧，

And the spelling was excellent and it was, like, cogent sounding, you're like, okay.

这篇文章肯定是聪明人写的。

I this has been written by a smart person.

我会认真对待它的内容。

I will take the content seriously.

对吧？

Right?

但现在，技巧和输出之间完全脱节了，因为你确实可以做到这一点。

Now there is this complete severance of sort of, like, craft and output because you could and you do this.

让克劳德写一篇论证，支持最荒谬的主张——比如，为我论证里根在20世纪80年代初推行减税的原因，与70年代的UFO目击报告有关。

Ask Claude to write an argument in favor of the most absurd proposition Ask Claude to write an argument for me that the reason why Reagan wanted to do tax cuts in the early 1980s related to these reports of UFO sightings in the 1970s.

而它写出的内容不仅语法正确，还会竭力构建出这个论点最完美的版本。

And it will write something that not only is it grammatically correct, it'll actually strain to come up with the best version of this argument before.

而且，如果你在读完之前还觉得，也许这个人真的认真对待了这个论点，但现在这个论点完全是凭空捏造的，我们就必须彻底改变对这类事情的判断标准。

And again, if prior to that having read it, like, oh, maybe the person like, this person took this argument seriously, but now this argument is just created ex nihilo, we're gonna have to really, like, change our heuristics about this stuff.

我们创造出了无限多语法极好却纯粹是怪人的声音。

We've created an unlimited stream of basically cranks with really good grammar.

是的。

Yeah.

没错。

That's right.

没错。

That's right.

因为过去我们能认出怪人，是因为他们语法很差，或者会给我们发邮件，一半单词是黄色的，另一半还带下划线。

Because it used to be we knew the crank because they had bad grammar, or they would email us, and, like, half the words would be in yellow and the other half would be underlined.

绿色墨水是经典的例子。

Green ink was the classic example.

我们用来判断一个人是怪人的那些工具，

The tools that we use to just, like, oh, this person's crank.

比如，他们一半的单词都是大写，诸如此类。

They, like, you know, half their words are in all caps and stuff like that.

这些方法现在不管用了。

Those don't work anymore.

好的。

Alright.

说到这个，我们就这样结束吧？

On that note, shall we leave it there?

我们就到这里吧。

Let's leave it there.

以上就是本期《奇怪想法播客》的全部内容。

This has been another episode of the Odd Thoughts Podcast.

我是特蕾西·阿拉韦。

I'm Tracy Allaway.

你可以关注我，账号是Tracy Allaway。

You can follow me at Tracy Allaway.

我是乔·维森塔尔。

And I'm Joe Wiesenthal.

你可以关注我，账号是The stalwart。

You can follow me at The stalwart.

关注我们的嘉宾马克斯·斯佩罗。

Follow our guest, Max Spero.

他的账号是max_spero_。

He's at max underscore Spero underscore.

关注我们的制作人：卡门·罗德里格斯，账号是CarmenArmen；乔·贝内特，账号是Dashbot；以及凯尔·布鲁克斯，账号是Kale Brooks。

Follow our producers, Carmen Rodriguez at Carmen Armen, dash Joe Bennett at Dashbot, and Kale Brooks at Kale Brooks.

如需获取更多Odd Lots内容，请访问bloomberg.com/oddlots，那里有我们所有节目的每日简报，你还可以在我们的Discord频道discord.gg/oddlots中24小时不间断地讨论这些话题。

And for more Odd Lots content, go to bloomberg.com/oddlots where have a daily newsletter on all of our episodes, and you could chat about all of these topics twenty four seven in our Discord, discord.gg/oddlots.

如果你喜欢《Odd Lots》这个节目，喜欢我们讨论互联网有40%是乱七八糟的内容，那就请在你最喜欢的播客平台上给我们一个好评。

And if you enjoy Odd Lots, if you like it when we talk about how the Internet is 40% slopped, then please leave us a positive review on your favorite podcast platform.

记住，如果你是彭博的订阅用户，你可以免费收听我们所有的节目，没有任何广告。

Remember, And if you are a Bloomberg subscriber, you can listen to all of our episodes absolutely ad free.

你只需要在Apple Podcasts上找到彭博频道，然后按照那里的说明操作即可。

All you need to do is find the Bloomberg channel on Apple Podcasts and follow the instructions there.

感谢收听。

Thanks for listening.

我是弗朗辛·拉克鲁瓦，一位获奖记者，我推出了一档新播客《弗朗辛·拉克鲁瓦的领袖》，由彭博播客出品。

I'm Francine Lacroix, an award winning journalist, and I've got a new podcast, Leaders with Francine Lacroix from Bloomberg Podcasts.

我采访过从国家元首到时尚偶像等各种人物，探讨当下新闻，但我一直很好奇，这些人在作为领导者时究竟是怎样的。

I've interviewed everyone from heads of state to fashion icons about the news of the moment, but I've always been curious who are these people as leaders.

我认为并没有一种

I don't think there's one

正确的领导方式。

right way to be a leader.

做出决定。

Make decisions.

一个糟糕的决定总比不做决定要好。

A poor decision is always better than no decision.

每隔一周的星期一收听新一期节目。

Listen to new episodes every other Monday.

在您收听播客的平台关注《Francine Lacroix 的领导者》。

Follow leaders with Francine Lacroix wherever you get your

播客。

podcasts.