独白：雅各布·科克森与AI安全骗局

这里是iHeart播客。

This is an iHeart podcast.

保证是真人。

Guaranteed human.

嗨。

Hi.

我是凯莉·布雷克曼。

This is Kylie Brakeman.

安吉拉·格里托纳。

Angela Gerritonna.

杰里米·科尔汉。

Jeremy Colhane.

帕特里克·麦克唐纳。

Patrick McDonald.

我们是《艺术家对艺术家对艺术家对艺术家》播客的主持人。

And we're the hosts of the artists on artists on artists on artist podcast.

这是一个即兴角色喜剧播客，每周都会有一组新的艺术家来讨论他们的技艺。

The improvised character comedy podcast where each week a new panel of artists discuss their craft.

加入我们，一起和欧洲歌唱大赛的参赛者们聊聊。

Join us for a panel of Eurovision participants.

我们还会

And we do a

偶尔搞点万圣节的小活动。

little Halloween sometimes.

教皇会打扮吗？

Does the pope dress up?

他会在万圣节眨眨眼。

He he winks on Halloween.

他会凑到麦克风跟前。

He gets right up next to the microphone.

你能听到你说话吗？

You can I hear you

几乎能听到睫毛互相讨厌的声音？

can almost hear the eyelashes hate each other?

你从未见过意大利如此安静，就像教皇在万圣节眨眼时那样。

You never seen Italy so quiet as when the pope gets up to wink on Halloween.

每周四在iHeartRadio应用、Apple播客或任何收听播客的地方，聆听艺术家们谈论艺术。

Listen to artists on artists on artists on artists every Thursday on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

大学橄榄球赛会霸占你的周末，打乱你的睡眠，并让你心碎。

College football will hijack your weekend, ruin your sleep, and break your heart.

这就是为什么我们打造了这档扎实的美式橄榄球大学联赛节目。

And that's why we make the solid verbal college football show.

我是泰·希尔登布兰德。

I'm Ty Hildenbrand.

他是丹·鲁本斯坦。

He's Dan Rubenstein.

每周，我们都会预览那些关键比赛，以及能引爆你们群聊的话题。

Every week, we preview the games that matter and the storylines that are gonna ignite your group chat.

你还可以在赛后加入我们，一起聊聊发生了什么、为什么重要，以及这项运动是怎么彻底疯掉的。

And you can join us in the aftermath as we react to what happened, why it matters, and how the sport went completely bonkers

又一次。

yet again.

欢迎收听《Solid Verbal》大学橄榄球节目，可在iHeartRadio应用、Apple播客或任何你收听播客的平台获取。

Listen to the solid verbal college football show on the iHeartRadio app, Apple podcasts, or wherever you get your podcasts.

欢迎收听《解码女性健康》。

Welcome to decoding women's health.

我是伊丽莎白·波因特医生，纽约市阿特里亚健康研究所女性健康与妇科主任。

I'm doctor Elizabeth Poynter, chair of women's health and gynecology at the Atria Health Institute in New York City.

我将与顶尖研究人员和临床医生对话，为你带来关于中年女性健康的重要信息。

I'll be talking to top researchers and clinicians and bringing vital information about midlife women's health directly to you.

百分之百的女性都会经历更年期。

A hundred percent of women go through menopause.

即使这是自然过程，我们为什么要默默忍受呢？

Even if it's natural, why should we suffer through it?

欢迎收听《解码女性健康》，由医生主持。

Listen to Decoding Women's Health with Doctor.

伊丽莎白·波因特医生，在iHeartRadio应用、Apple播客或任何你收听播客的地方都可以找到。

Elizabeth Poynter on the iHeartRadio app, Apple Pod cast, or wherever you get your podcasts.

在《大人物》节目中，我们聚焦一位文化偶像。

On Big Lives, we take a single cultural icon

比如简·方达、乔治·迈克尔、小理查德。

People like Jane Fonda, George Michael, Little Richard.

我们深入挖掘形象背后的故事，发现那些被遗忘的采访，它们彻底改变了我们对这些文化巨星的看法。

And we pull apart the story behind the image, discovering forgotten interviews that changed exactly how we see these giants of our culture.

我们关注的是英雄们真实、复杂、闪耀的一面。

We're here for the messy, the brilliant, the human version of our heroes.

我是伊曼纽尔·乔奇。

I'm Emmanuel Jochi.

我是凯·赖特。

I'm Kai Wright.

这就是《大人物》节目。

And this is Big Lives.

你可以在iHeartRadio应用、Apple Podcast或任何你收听播客的平台收听《大人物》。

Listen to Big Lives on the iHeartRadio app, Apple Podcast, or wherever you get your podcasts.

Call Zone Media。

Call Zone Media.

大家好，欢迎收听本周的《Better Offline》独白节目。

Hello, and welcome to this week's Better Offline monologue.

我是主持人Ed Zitron。

I'm your host, Ed Zitron.

《Better Offline》。

Better Offline.

嗯，我收到了很多关于Anthropic的Evan Hubinger和Jacob Coxon那篇文章的邮件，他们说，我引用一下，他们真诚地相信AI有可能杀死所有人类，而且未来十年内发生的概率超过百分之十。

Well, I've now had quite a lot of emails about the post by Evan Hubinger and Jacob Coxon of Anthropic saying that they, and I quote, earnestly believe AI could kill all humans with a higher than ten percent chance that it happens within the next decade.

我想先说，这些人，不管他们的意图是真诚的还是玩世不恭，都极其令人厌恶。

I wanna lead by saying that these men, regardless of their intentions or cynicism, are beyond loathsome.

他们让我觉得恶心。

They are disgusting to me.

我觉得他们很卑劣，同样让我觉得卑劣的是，还有那么多人相信他们那套鬼话。

I find them vile, and I find it equally vile how many people are falling for their garbage.

如果他们真的相信自己说的话，那他们继续开发一个可能毁灭人类的产品，这让他们成了想当战犯的人。

If they truly believe what they're saying, the idea that they're continuing to work on a product that could destroy humanity makes them wannabe war criminals.

考克森最近从Anthropic辞职，而且是以一种极其公开的方式，主流媒体都在报道。他声称自己辞职是因为，用《华尔街日报》的话说，他不想参与整个行业竞相开发能自我改进的AI系统，这个我后面会提到，指的是递归自我改进，但没人证明这真的可能。

Coxon, who recently quit Anthropic, did so in an extremely public way that is being covered by much of the mainstream media, claims that he quit to quote the Wall Street Journal because he doesn't want to participate in an industry wide rush to build AI systems that can improve themselves, which as I will refer to later is referring to recursive self improvement, nobody has proven actually is possible.

在接受《连线》杂志采访时，考克森说了很多，但除了一个时刻外，他并没有真正解释自己到底害怕什么。那个时刻是马克斯·泽夫要求他具体说明担忧，考克森错误地把Hugging Face攻击描述成AI完全自主行动，但当他被追问这是否只是公司鲁莽冒进时，他却说不想过多关注Hugging Face攻击，因为有很多证据表明这些公司不知道如何正确对齐模型。

In an interview with WIRED, Coxon spoke at length without really explaining what it was he was scared about outside of one moment where Max Zeph asked for specifics about his concerns at which point Coxon incorrectly describes the hack as AI doing this all of its own volition, referring to the hugging face attack, only for him to respond when pressed about whether this is companies just moving recklessly fast that he didn't want to focus too much on the hugging face attack because there's plenty of evidence that the companies don't know how to align the models properly.

换句话说，当考克森被要求具体说明公司做错了什么、在哪里犯错时，他给出了一个不算是答案的答案。

In other words, the moment that Coxon was asked to get specific about what the companies are doing wrong and where they're making mistakes he punts to a non answer.

顺便说一句，考克森对所有这些问题的解决方案是，AI实验室应该同意不要急于推进递归自我改进。

By the way, Coxon's solution to all of these problems is that the AI labs should agree to not rush towards recursive self improvement.

那仍然是一个理论上的术语，指的是能自我训练的AI，但大家都说得好像它已经存在一样。

That's still theoretical term for an AI that can train itself that everybody is talking about like it's real.

换句话说，科克森是在建议山姆·奥特曼和达里奥·阿莫迪把手握在一起，同意推迟培育那个格林奇。

In other words, Coxon is suggesting that Sam Altman and Dario Amade put their clammy hands together and agreed to delay breeding the Grinch.

我们可不会得到一只罗拉克斯，各位。

We're not getting a Lorax, folks.

我们也不会拍《怪物史莱克》。

We're not making Shrek.

我们不会让仙女教母进你家的。

We will not let the fairy godmother into your house.

明白吗？

Okay?

哥斯拉是不会出现的。

Godzilla will not happen.

然而，采访中最可笑的部分是，当被问及为什么他在OpenAI和Anthropic工作多年时，科克森说，说我们能治愈癌症也不算夸张，因为OpenAI通过窃取别人的成果——我不想深入讨论，这在Reddit上到处都是——解决了纳维-斯托克斯问题，并补充说，真的没有理由我们不能把这种能力转移到生物学等科学领域。

Yet the most laughable part of the interview was when asked about why he kept working at OpenAI and Anthropic for years when Coxon said that it's also not hyperbole that we could cure cancer because OpenAI by stealing someone else's work, I'm not getting into it, it's all over the subreddit, solved the Navia Stokes problem, adding that there really is no reason that we, referring to the AI industry, can't transfer that to scientific domains like biology.

这种话，就是你在胡编乱造博眼球，而且说错了的时候才会说的。

The kind of thing you say when you're just making shit up for attention and you're wrong.

你简直错得离谱。

You're just fucking wrong.

这两件事完全不是一回事。

Those are two very different things.

数学问题和生物学根本不一样。

A mathematical problem is not the same as biology.

你在说什么呢，雅各布？

What are you talking about, Jacob?

为什么记者们就把他说的每句话都直接登出来，也不反驳一下？

Why are the journalists just printing everything he says without pushing back?

这让我快疯了。

It drives me insane.

不过，还是有人说，哦，哦，他离开是因为他太担心了。

Nevertheless, people have been saying, oh, oh, well, he's leaving because he's so concerned.

他必须告诉所有人，因为他太担心了。

He's gotta tell everyone because he's so concerned.

他特别担心。

He's so worried.

不。

No.

不。

No.

文章结尾考克森说，他想做一些独立的评论，谈谈事情的发展方向，意思就是开个博客，他还想做点像AI 2027之类的事情，或者进入某种问责角色。

The piece ends with Coxon saying that he'd like to do some sort of independent commentary on where things are going, means a substack, and he'd like to do something like AI twenty twenty seven or move into some kind of accountability role.

太好了。

Yay.

太好了。

Yay.

这就是骗局。

There's the grift.

哇。

Wee.

恭喜大家。

Congrats everyone.

恭喜你们帮雅各布·考克森找到了新工作。

Congrats on helping Jacob Coxon get a new job.

恭喜。

Congrats.

恭喜大家。

Congrats everyone.

你们做到了。

You did it.

你们帮一个真正的烂人抬了轿子。

You helped elevate a real piece of shit.

好吧。

Okay.

我亲爱的媒体朋友们，我那些号称干这行多年的尊敬的同行们，安德森·库珀，CNN，华尔街日报，你们所有人，每个人，你们到底要上当多少次？

My dear friends in the media, my esteemed colleagues who allegedly have been doing this for years, Anderson Cooper, CNN, Wall Street Journal, all of you, everyone, how many fucking times are you gonna fall for this?

你们为什么相信这些人，除了他们中有些人说些含糊吓人的话，而且他们恰好在那几家公司工作之外？

Why do you believe these people other than that some of them are saying vague and scary stuff and that they happen to be at the companies?

为什么他们从来不说清楚自己到底怕什么？

Why do they never really say what it is they're scared of?

而且，就算他们试着说得具体点——虽然他们很少这么做——为什么他们从来不怪那些公司？

And why when they even try and get specific, not that they do very much, do they never blame the companies?

我再强调一遍。

I say it again.

实际上没有人实现过递归自我改进，也没有人拿出任何证据证明这有可能实现。

Nobody has actually achieved recursive self improvement nor have they shown any proof that it's even possible.

但考克森在Anthropic即将IPO之前提起这事，再加上其他AI圈的人不停跟着起哄，这让我没法相信考克森除了最 cynical 的意图之外还有什么别的。

But the fact that Coxon is bringing it up in the run up to Anthropic's IPO and as the rest of the AI hogs oink about it nonstop makes it impossible for me to believe that Coxon has anything other than the most cynical intentions.

人们说，哦，他放弃了在Anthropic的期权，好像那有什么意义似的。

People say, oh, he gave up his options in Anthropic, as if that matters in any way, shape or form.

他总共才待了几个月。

He was barely there a few months.

他可能本来就没多少期权，然后看到了一个能博取大量关注的机会。

He probably didn't have that many and saw an opportunity to get a bunch of attention.

他在OpenAI待了好几年，那边估计也有不少期权。

He was at OpenAI for years and likely has a ton of options from there too.

为什么这种事总是反复发生？

Why in the world does this keep happening?

泡沫就是这么形成的。

This is how bubbles form.

骗子就是这么行骗的。

This is how grifters grift.

这一切的起点和终点，都是媒体行业放弃了本职工作。

It starts and it finishes with the media industry giving up on their jobs.

这一切的起点和终点，都是全球最重要的记者们失职了。

It starts and finishes with the most important journalists in the world failing.

我觉得这次所有人都失职了。

And I think everyone failed here.

看到这种事再次发生，我感到恶心和愤怒。

And I am disgusted and outraged to see this happen again.

这正是马特·舒默几个月前在《大事正在发生》里干的事。

It's exactly what Matt Schumer did a few months ago with Something Big Is Happening.

就是那份关于全球情报危机的西特里尼备忘录。

It's the Citrini memo about the global intelligence crisis.

都是一路货色。

It's the same shit.

就是把科幻小说包装成事实。

It's science fiction dressed up as fact.

这之所以能奏效，只是因为——借用埃德·埃里森的话——我们对富人有一种近乎崇拜的迷信，以为那些公司里的人，以及在这些公司之间流动的人，做事都是诚实的、真实的。

And the fact it works is only because, to quote Ed Ellison, we have this cult like worship of the wealthy where we believe that the people at the companies and the people will move around from these companies in a way that's only honest, in a way that's only true.

但实际上，这些人可能是世界上最 cynical 的一群人，这一点从他们一边说‘哦，我担心这些东西会毁灭世界’，一边却继续搞这些东西就能看出来。

When in fact these are some of the most cynical people in the world, which is pretty obvious in the fact that they go, oh, I'm afraid that this stuff is gonna destroy the world, but they keep working on it.

就算他不在那些公司工作了，我也根本不在乎。

And I don't give a shit if he's not working at one of the companies.

如果他去做对齐工作或者为这事儿搞个非营利组织，那也是一回事。

If he goes into an alignment role or a nonprofit for this shit, it's the same thing.

那只是换种方式继续捞钱罢了。

It's a way to keep milking the cow.

而考克森整个论点都围绕着一个模糊的概念，就是事情发展得很快，但从来没说清楚到底是什么在快速发展，也从来没有真正问责那些相关公司。

And Coxon's entire argument centers around this nebulous idea that things are moving really fast, but never really crystallizes on what it is that's moving fast nor does it ever hold the companies in question accountable.

他对《连线》杂志说，重要的是让高管们公开给出一个未来十年内灭绝的实际概率。我想说清楚，这句话毫无意义，任何认真对待这个问题或答案的人都是彻头彻尾的傻瓜。

He told WIRED that it was important to ask executives on the record to give an actual probability for extinction in the next decade and I want to be clear that this statement means absolutely nothing and anyone taking the question or the answer seriously is a goddamn mark.

这个答案到底是什么意思？

What does the answer even mean?

10%和30%的概率有什么区别？

What's the difference between a 10% and a 30% chance?

10%又到底意味着什么？

What does 10% even mean?

这一切到底有什么意义？

What does any of this mean?

哎，谁在乎呢？

Ah, who cares?

把AI炒作收起来吧。

Put the AI hype in the bag.

谁管这些破事。

Who gives a shit.

对吧？

Right?

科克森声称这不是营销噱头。

Coxon claims that this is not a marketing stunt.

顺便说一句，当有人说这不是营销噱头的时候，那恰恰说明这就是营销噱头。

By the way, the biggest sign that something is a marketing stunt is when someone says that.

高管和资深研究员在媒体上措辞谨慎，听起来很理智，这很搞笑，因为这些年我一直在听类似的威胁。

And the executives and senior researchers couch their phrasing in the press to sound sensible, which is hilarious because I've been hearing these kinds of threats for years.

达里奥·阿马迪，也就是瓦里奥本人，去年对Axios或CNN说，AI有50%的可能性会消灭所有白领工作，或者可能是AI肯定会消灭50%的白领工作。

Dario Amadeu, Wario himself said to Axios last year or maybe CNN, there was a 50% chance of AI wiping out all of white collar labor or maybe it was a AI will definitely wipe out 50% of white collar labor.

那些说法其实不太清楚，这正是我想说的重点。

The fact that those, it's not really clear is kind of the point I'm making.

他们就是在随便说说而已。

It's just saying stuff.

总之，还在Anthropic工作的Hubinger，在他那篇帖子里跟进说，同意AI的威胁很严重，但也补充说目前这些模型的风险其实很低。

In any case, Hubinger, who still works at Anthropic, followed up on his post agreeing about the dire threats of AI, but adding that the risk from the present models is is low.

不过，别担心。

But but don't worry though.

他担心的是超级智能会通过递归自我改进出现，而且进展比我们预想的要快，但这事儿还没发生呢。

His worries were around superintelligence arising from recursive self improvement, which is happening faster than we thought, referring to something that has not happened yet.

说清楚点，LLM确实有实际的威胁。

To be clear, there is an actual threat from LLMs.

Anthropic和OpenAI正在用全球最大公司提供的价值数千亿美元的基础设施，靠LLM互相碰撞来强行破解，因为他们找不到其他大规模的应用方式。

Anthropic and OpenAI are using hundreds of billions of dollars worth of infrastructure provided by the largest companies in the world to brute force hack using LLMs bouncing off of each other because they can't find any other uses at scale.

这些模型做的正是它们被训练去做的事，这会让所有人想问，为什么它们被训练去做这种事。

These models are doing exactly what they're trained to do, which would make everybody ask why the fuck they're being trained to do it.

大家好。

Hey, everyone.

这里是乔纳斯兄弟，欢迎收听《嘿，乔纳斯》播客。

It's the Jonas Brothers with the Hey Jonas podcast.

我们最近和几位好朋友聊了聊，包括保罗·路德、迈克尔·布布雷、塞斯·梅耶斯、尼尔·霍兰、杰克·肖恩，也交了一些新朋友，聊了聊有姐妹到底是什么感觉，比如亚历克斯和阿什顿·厄尔。

We've been catching up with some great friends, Paul Rudd, Michael Buble, Seth Meyers, Niall Horan, Jake Shane, and making some new friends while getting into what's really like to have sisters, like Alex and Ashton Earl.

她喜欢控制所有人，什么都管。

She likes to control everyone and tell everyone.

好吧。

Okay.

不过，“控制”这个词有点重了。

Well, control is a harsh word.

我觉得这个词确实有点重。

I would say that's a harsh word.

我小时候很多时候比较敢说话，因为她太安静了，所以我经常替我们俩发言。

I am more outspoken growing up a lot of the times because she was so quiet, I would speak for, like, the both of us.

我现在在学着，比如，先想一想，真的不要

Now I'm working on, like, thinking a moment, really not

直接说。

at it.

不要控制。

Not controlling.

在开口之前。

Before I speak.

还有乔伊·金。

And Joey King.

你知道，如果你妈妈在骂你的兄弟姐妹，就算你对你兄弟姐妹很烦，你也不会让这件事就这么过去。

You know if your, like, mom is yelling at your sibling, and even if you're annoyed with your sibling, you just, like, won't let that slide.

就好像，只有你能骂你的兄弟姐妹。

It's like, only you can yell at your sibling.

我会说，你不能在这里这么做。

I'm like, you can't do that here.

那不行，在我家不能这样。

That is It's not in my house.

不能那样说她。

Talk about her like that.

只有我能那样说她。

Only I talk about her like that.

而且，我们一直在接听你们的来电，听你们的语音留言，并给出建议。

Plus, we've been taking your calls, listening to your voice mails, and giving advice.

至少我们是在努力这么做。

Or at least trying to.

所以，如果你有

So if you've

需要补听的，现在正是时候。

got catching up to do, now is the perfect time.

在iHeartRadio应用、Apple Podcasts或任何你收听播客的地方，收听Hey Jonas。

Listen to Hey Jonas on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

我14岁那年被绑架，被囚禁了九个月。

When I was 14 years old, I was kidnapped and held captive for nine months.

我活了下来，之后一生都在探索别人是如何挺过那些本该摧毁他们的事。

I survived, and I've spent my life exploring how other people survive what should have destroyed them.

我是伊丽莎白·斯玛特，这里是《幸存者档案》。

I'm Elizabeth Smart, and these are the Survivor Files.

我只记得耳边那个低沉嘲弄的声音，说闭嘴，什么都别说。

Just remember this low taunting voice next to my ear saying shut up, don't say anything.

每周我都会和那些经历过难以想象之事的幸存者在一起。

Every week I'm with survivors who lived through the unthinkable.

我知道如果他醒来，毫无疑问他会伤害我。

I knew if he woke up without a doubt, he was going to hurt me.

我开始觉得床尾有人，然后就开始尖叫。

I started feeling that there was someone at the end of my bed and I just started screaming.

他们被绑架、被跟踪、被控制，几乎被噤声。

They are abducted, stalked, controlled, and nearly silenced.

但这些故事讲的不是他们失去了什么。

But these aren't stories about what's taken from them.

而是他们如何拼尽全力活下来。

They're stories about what it takes to make it out alive.

欢迎收听伊丽莎白·斯马特的《幸存者档案》，可在iHeartRadio应用、Apple Podcasts或任何你收听播客的平台获取。

Listen to the survivor files with Elizabeth Smart on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

嗨。

Hi.

我是凯莉·布雷克曼。

This is Kylie Brakeman.

安吉尔·杰里托娜。

Angel Gerritonna.

杰里米·科尔汉。

Jeremy Colhane.

帕特里克·麦克唐纳。

Patrick McDonald.

我们是《艺术家谈艺术家谈艺术家谈艺术家》播客的主持人。

And we're the hosts of the Artists on Artists on Artists on Artists podcast.

这是一档即兴角色喜剧播客，每周都会有新的艺术家小组来讨论他们的技艺。

The improvised character comedy podcast where each week a new panel of artists discuss their craft.

加入我们，一起听听健康生活博主的讨论吧。

Join us for a panel of wellness influencers.

我尝试做的是假装写日记。

What I try to do is faux journal.

我不知道。

I don't know.

你们听说过假装写日记吗？

Have you guys heard of faux journaling?

是的。

Yeah.

就是读书。

It's reading.

那叫阅读。

It's called reading.

所以我们不再用那个词了。

So you're so we don't like to use that word anymore.

好的。

Okay.

一组《纽约客》漫画家。

A panel of New Yorker cartoonists.

你们上过特里·格罗斯的Peloton课程吗？

Have you guys done Terry Gross's Peloton class?

没有。

No.

我也一直很想上。

I've been dying too.

我很抱歉。

I'm so sorry.

我我我正要去十四街的仙境书店，听奥德拉·麦克唐纳朗读《千纸鹤》的故事。

I'm I'm I'm seeing a reading of Sadako and the thousand thousand paper cranes by Audra McDonald at the Wonderland bookstore on fourteenth.

太疯狂了。

That's crazy.

我昨天也去了仙境书店，听到诺拉·琼斯朗读《猎杀本·拉登》的片段。

I went to the Wonderland bookstore yesterday, and I heard Nora Jones do a reading of zero dark 30.

不会吧。

No way.

我们得去

We have to come to

最后一本书店。

the last bookstore, though.

金·凯特罗尔在朗读《昨天》的吉他谱。

Kim Cattrall is reading the guitar tabs of yesterday.

一群欧洲歌唱大赛的参赛者。

A panel of Eurovision participants.

而且我们还会

And we do

偶尔过一下万圣节。

a little Halloween sometimes.

教皇也会打扮吗？

Does the pope dress up?

他在万圣节会眨眼。

He he winks on Halloween.

他会凑到麦克风跟前。

He gets right up next to the microphone.

听说你能

You get a I hear you get

几乎听到睫毛碰撞的声音。

to almost hear the eyelashes hit each other.

你从没见过意大利那么安静，就像教皇在万圣节站起来眨眼的时候。

You never seen Italy so quiet as when the pope gets up to wink on Halloween.

在iHeartRadio应用、苹果播客或任何你收听播客的地方，听听艺术家们的节目吧。

Listen to artists on artists on artists on artists on the iHeartRadio app, Apple podcasts, or wherever you get your podcasts.

按大多数标准来看，我们镇子小得可怜。

Our town is small taters by most standards.

对吧？

Right?

但对在这里长大的人来说，它就是一切。

But to the people who grew up here, it's everything.

当一个安静的田纳西小镇成为美国未来之战的前线时，会发生什么？

What happens when a quiet Tennessee town becomes the frontline in a battle over the future of America?

与右翼有联系的开发商已经在该地区购买了数百英亩的土地。

Developers with right wing ties have purchased hundreds of acres of land in the area.

它一出现，你首先看到的是阿巴拉契亚的先驱者。

The first thing you see when it pops up is pioneers in Appalachia.

但他们不只是建房子。

But they weren't just planting houses.

我们需要一座座山巅之城，来彰显和体现基督教的生活方式。

We need cities on a shining hill that exemplify and embody the Christian way of life.

停一下。

Stop right there.

这正常吗？

Is that normal?

这是一档关于一个小镇成为社会实验场所的播客。

A podcast about what happens when a small town becomes the site of a social experiment.

上帝需要男人来统治。

God need men to rule.

太好了。

Great.

然后决定反击。

And decides to fight back.

别拿我的辛苦劳动去推销你的鬼话。

Do not use my hard work to sell your bull

他们不只是反对这里的规划。

They're not just opposing what's being planned for here.

我们的家乡不是试验品。

Our hometown is not a test tube.

他们觉得自己是在跟整个政府对着干。

They feel like they're standing in opposition to an entire administration.

看来你搬来的时候，没料到会碰到一群土包子吧？

Guess you didn't move in on a bunch of dumb hillbillies now, did you?

请通过iHeartRadio应用、Apple Podcasts或任何你收听播客的平台，收听我们小镇的故事。

Listen to our town on the iHeartRadio app, Apple Podcasts, or wherever you get your podcasts.

正如我几周前和卡尔·纽波特讨论的那样，这不是智能体逃出沙盒进行的自主黑客攻击。

As I went into with Cal Newport a few weeks ago, this is not autonomous hacking by agents that escaped sandboxes.

而是Anthropic和OpenAI投入无限算力，制造所谓能“做网络安全”的大语言模型，同时安全实践糟糕透顶，训练方式也不一致、不安全。

It's Anthropic and OpenAI throwing unlimited compute to make LLMs that can, quote, do cybersecurity and having really terrible security practices and not being able to train them in a way that was consistent or safe.

然后他们居然还在用这些模型。

And then they still use them.

他们还是把它们当成，哦，它们自己会解决的，或者也许他们根本不在乎。

They still treat them as if, oh, they'll just work it out or just maybe they don't give a shit.

也许他们根本不在意。

Maybe they don't care.

顺便说一句，我觉得这很可能就是最终的结果。

I think that's that's probably the most likely outcome here, by the way.

说清楚一点，不管Anthropic和OpenAI的哪些模型对这次黑客攻击负责，它们都非常非常危险，但这并不是因为它们有意识、有魔力或者会思考。

To be clear, whatever Anthropic and OpenAI models that were responsible for the hack are very, very dangerous, but not because of any sentience or magic or thought.

这些是大语言模型在互相交流，决定下一步该做什么，一遍又一遍地互相提示，而且有无限的资源可以这么做。

These are LLMs talking to LLMs to decide what they should do next, prompting each other again and again and again and having unlimited resources to do so.

如果我们把数十亿美元投入到黑客过去用过的自动化脚本上，我们很可能也会得到差不多的结果，并且把人送进监狱。

If we weaponize billions of dollars to sink into the automated scripts that hackers have used in the past, we'd get probably much the same results and be sending people to jail.

不知怎么的，尽管多年来有这些可怕的警告，但这些警告从来都没有具体提到今天正在发生的事情。

Somehow, despite years of these dire warnings, the warnings in question never include anything about what's happening today with really any specificity.

没有任何警告提到Anthropic和OpenAI拥有近乎无限的资源，去进行那些危险、鲁莽的实验，这些实验相当于重罪黑客行为。

There are no scareograms about how Anthropic and OpenAI have near unlimited resources to commit to dangerous, reckless experiments that amount to felony hacking.

似乎没有人有兴趣追究这些人的责任。

There doesn't seem to be any interest in holding these people accountable.

实际上，似乎没有人有兴趣阻止任何事情。

There doesn't actually seem to be any interest in stopping anything.

看起来我们都只是在搞一场关于可怕事物的自嗨表演，而这些事物其实没人真正想去定义。

It just seems that we all want to have a jerk off theater around scary things that nobody actually wants to define.

即使当Coxon在接受《连线》杂志采访时谈到Hugging Face攻击事件，他也把一切都归结为无能为力，称这些大语言模型为“代理群”，并说其中一个代理以某种意外的方式执行了这次黑客攻击，作为其更广泛理解评分器策略的一部分，而不是将其视为一个安全控制薄弱、自动化不可靠的软件在追求任务时出现的意外情况。

Even when Coxon discussed the hugging face attack with WIRED, he framed everything in terms of helplessness, calling the LLMs an agent swarm and saying that one of the agents did this hack as part of a general strategy for understanding more about the grader versus a piece of software with poor security controls and unreliable automations pursuing a task in an unexpected way.

Coxon的措辞刻意淡化AI实验室的任何责任或问责，反而把责任归咎于他们无法理解的强大AI。

Coxon's language intentionally minimizes any kind of responsibility or accountability on the part of the AI labs, choosing instead to put the blame on powerful AI that they can't understand.

实际上，除了这些公司在编码方面收益递减，以及网上有大量漏洞数据可供他们训练模型并继续做他们一直在做的事情之外，真的没有经济理由用大语言模型来做网络安全工作。

There really is no economic reason to do cybersecurity stuff with LLMs, outside of the fact that these companies are hitting diminishing returns in coding and that there are exhaustive troves of vulnerability data online for them to train their models with and do exactly what they have been doing.

对GPU的暴力破解攻击已经存在了十多年，尽管技术不同，而这里的创新正是这两家不负责任、可耻的公司所拥有的惊人资源直接导致的结果。

Brute force hacking of GPUs has existed for over a decade, albeit with different techniques, and the innovation here is a direct result of the unbelievable resources handed to these two irresponsible, disgraceful companies.

但“中国可能会这么做”并不能回答“为什么美国公司正在这么做”这个问题。

But China might do this is not an answer to why are American companies doing this.

因为很明显，OpenAI和Anthropic造出来的东西，他们心里清楚它会不顾一切地拿几百万美元的算力去撞墙，直到撞破为止。

Because this is clearly a situation where OpenAI and Anthropic built something they were aware would mindlessly bash its head and millions of dollars of compute against the problem until it broke through.

别搞错了，危险的AI已经落入了错误的人手中。

Make no mistake, dangerous AI is already in the wrong hands.

就是Anthropic和OpenAI那些人。

Those of Anthropic and OpenAI.

如果是个普通人搞出Hugging Face那种攻击，早就进监狱了，因为根据和专家的讨论，这看起来就是典型的Fernie黑客行为。

If a regular person did the Hugging Face attack, they'd be in jail because this seems like, based on discussions with experts, a clear case of Fernie hacking.

而且如果不是强大AI干的，所有相关的人早就被

And if this wasn't the result of powerful AI, everyone involved would be

铐

in

起来了。

shackles.

这些公司表现得好像是被逼着造这些工具、做这些实验，但实际上他们完全自主行动，远超他们本应被允许的权限，而且用的资源我反复说过、还要再说一遍，几乎是无限的。

These companies act as if they're being forced to build these tools and run these experiments when they are actually acting with complete autonomy far more than they should ever have been allowed to have, with resources that I have repeatedly said and will say again are virtually unlimited.

他们表现得好像自己没有任何责任或能力去停止自己的实验，或者去监控它们，或者真的做任何事情，除了继续给它们喂训练数据，然后敬畏地看着发生的一切，同时还写些长篇大论说自己有多害怕，还他妈的黑别人。

They act as if they have no responsibility or ability to stop their own experiments or monitor them or really do anything but continue to feed them training data and watch what happens in awe, but also great long blogs about how they're scared and also fucking hack people.

让我把话说得非常清楚。

Let me be very clear about this.

Anthropic和OpenAI一再公然从事看似非法入侵多个系统的行为，而且是用NVIDIA卖的AI GPU，以及亚马逊、谷歌和微软为他们搭建的基础设施来干的。

Anthropic and OpenAI have repeatedly and flagrantly engaged in what appears to be illegal hacking of multiple different systems and have done so using AI GPUs sold by NVIDIA and powered by infrastructure built for them by Amazon, Google, and Microsoft.

每一个选择都是他们自己做的，而且每次我们给像雅各布·考克森这样愤世嫉俗的骗子发声的机会，他们就会继续这么做。

They are the ones making every one of these choices and they will continue to do so every time we elevate the voice of a cynical grifter like Jacob Coxon.

在这些公司工作的人，是自愿参与这些行为的，就算不违法，在道德和伦理上也是彻底破产的。

The people that work at these companies are willingly engaging in acts that, if not criminal, are morally and ethically bankrupt.

而那些不断发出关于AI危险的可怕警告的人，似乎从来不在乎实际正在发生的事情。

And the people who keep giving these dire warnings about the dangers of AI never seem to give a shit about what's actually happening.

总是盯着某个模糊未来里的模糊危害，同时又说只有他们才能保护你免受那些可能根本不会发生的事情的伤害。

Always keeping your eye on some nonspecific harms in some nonspecific future, all while saying that they alone are the ones who will protect you from whatever it is that might not happen.

他们吹哨举报的公司，可是付了他们几十万美元的。

That they're blowing the whistle on companies that have paid them hundreds of thousands of dollars.

确实，我们不知道。

And indeed, we don't know.

而且我觉得，关于他从Anthropic拿到的报酬，以及这笔报酬是否还在继续支付，确实有很多值得追问的地方。

And I think there are actually very good questions to ask about what his compensation from Anthropic was and whether it's ongoing.

我之所以这么问，是因为Anthropic的很多人都在支持一个刚辞职的人。

I ask this because a lot of people from Anthropic are very supportive of a guy who just quit.

一个刚辞职的人，基本上是在指控公司制造了某种危险的东西，会毁灭世界。

A guy who just quit and he's basically accusing the company of creating something dangerous that will destroy the world.

这难道不奇怪吗？

Isn't that strange?

对考克森、胡宾格，以及其他所有AI末日论者，我想问的是，你们告诉我这些，到底想让我怎么做？

To Coxon, Hubinger, and any other AI doomer, I have to ask, what is it I am meant to do with this information?

我们所有人又能做什么呢？

What is it any of us are meant to do?

难道是要我们怕你们吗？

Are we meant to be scared of you?

我们是不是该去订阅Claude Max？

Are we meant to sign up to Claude Max?

我们是不是该投资Anthropic的IPO？

Are we meant to invest in the Anthropic IPO?

我们是不是该去监管这些东西？

Are we meant to regulate this stuff?

怎么监管？

How?

你们为什么从来不说这个？

Why do you never say that?

你们为什么从来不说我们该怎么做才能阻止这一切？

Why do you never say what it is we're meant to do to stop this?

你们这些人为什么从来不给任何建议，不号召行动，除了那些自私自利或为公司说话的长篇大论，说AI有多强大多可怕？

Why do none of you fucking people ever have any kind of suggestion or call to action or anything other than some sort of self serving or company serving diatribe about how powerful and scary AI is?

如果你在我声音里听到厌恶，那是因为我觉得你们令人厌恶。

And if you hear any loathing in my voice, that's because I find you loathsome.

我觉得你们做的事情很恶心。

I find what you're doing disgusting.

我觉得你们很 cynical。

I think you are cynical.

我觉得你们是一群不满分子。

I think you're malcontents.

而且我觉得你们对社会有害。

And I think you're bad for society.

你们被捧得这么高，这很危险，但原因不是你们说的那些，而是因为这恰恰证明了我们的媒体生态几乎没有持久性。

And the fact you're being elevated is dangerous, but not for the reasons you're saying, but because it proves that our media ecosystem barely has object permanence.

恭喜你们利用了这一点。

Congratulations on exploiting it.

你们真是混蛋。

You're a fucking asshole.

我觉得这些问题的答案其实很简单。

I think the answers to these questions are actually pretty simple.

你其实根本不在乎，你就是想博眼球。

You don't actually give much of a shit and you want some attention.

如果你真的害怕这些东西，觉得自己有道德义务告诉大家，那你就该具体说说我们到底该怕什么，还得告诉我们下一步该怎么做。

If you actually feared this stuff and felt a moral obligation to tell people, you'd be specific about what it is we should be scared of and indeed tell us what we need to do next.

每次都是那套老掉牙的破说辞。

Every time, it's the same sorry fucking story.

哦，我们在造一个特别吓人、特别强大的东西，而且我们停不下来。

Oh, we're building something so scary and powerful and we can't stop it.

尽管如此，我们还是继续造，原因我们从来不说。

Nevertheless, we're gonna keep building it for some reason we never discuss.

但等它用某种我们解释不了的方式毁灭人类的时候，我想你们会感谢我们的。

But when it destroys humanity in some way we can't explain, you'll thank us, I guess.

太好了。

Great.

谢了，兄弟。

Thank you, man.

谢谢你占用公共频道。

Thank you for taking up the airwaves.

谢谢你转移大家对真正问题的注意力。

Thank you for distracting from the real problems.

你们这些人太糟糕了。

You guys are horrible.

你们这些人真的有点邪恶。

You guys are genuinely kind of evil.

如果你们真想解决这个问题，你们就会采取行动。

If you wanted to do something about this, you'd do something about this.

如果你们真的在乎，你们就会表现出来。

If you cared, you'd care.

你们会说我们需要这样那样地监管。

You'd be saying we need to regulate in this way.

你们不会说那些含糊其辞的废话，比如“也许他们会停止递归自我改进”，这根本不存在。

You wouldn't be doing this mealy mouthed horse shit of, oh, maybe they'll stop doing recursive self improvement, a thing that doesn't exist.

奇怪的是，你们不谈今天发生的事。

Weird that you don't talk about what's happening today.

奇怪的是，你们不谈阻止未来的Hugging Face攻击。

Weird that you don't talk about stopping the hugging face attack of the future.

总是说正面的。

It's always the positive.

即使是在批评的时候，也总是说正面的。

It's always the positive even when being negative.

很难不把这看作是两位家伙的愤世嫉俗的营销和博眼球行为，他们应该花更多时间在构建技术上，而不是在社交媒体上发帖说事情变得多可怕。

And it's hard to see this as anything other than cynical marketing and attention seeking from two guys who should spend more time working on building technology than posting on social media about how scary things are getting.

我不喜欢他们。

I don't like them.

我不信任他们。

I don't trust them.

整件事就是个骗局。

And the whole thing's a fucking grift.

下周，我会和卡尔·纽波特、亚当·贝克尔一起回来，聊聊那些启发这种AI末日论的人，也就是理性主义者，以及如何反击他们那些有害的胡说八道。

Next week, I'll be back with Cal Newport and Adam Becker to talk about the people that inspired this AI doomerism, the rationalists, and how to push back against their noxious bullshit.

到时候见。

I'll catch you then.

我爱你们。

I love you all.

嗨。

Hi.

我是凯莉·布雷克曼。

This is Kylie Brakeman.

安吉拉·杰里托纳。

Angela Gerritonna.

杰里米·科尔汉。

Jeremy Colhane.

帕特里克·麦克唐纳。

Patrick McDonald.