本·汤普森：Anthropic、五角大楼与私人权力的局限

你可能对政治不感兴趣，但政治对你却感兴趣。

You might not be interested in politics, but politics has an interest in you.

什么是政治？

What is politics?

以其他方式进行的战争。

War by other means.

你可能对那并不感兴趣。

You might not be interested in that.

它终将对你产生兴趣。

It is going to have an interest in you.

如果我们像达里奥·阿马德一再做的那样，将其比作核武器，你就必须思考在一个私人公司研发出核武器的世界里会发生什么。

If we're going to analogize it to nuclear weapons, as Dario Amade has done repeatedly, you have to think through what would happen in a world where a private company developed nuclear weapons.

它的关注已经酝酿多年。

Its attention has been brewing for years.

是的。

Yeah.

也就是说，你们是一家美国公司，是否受美国法律甚至超越法律的道德义务驱使，去支持美国军方？

Which is, are you an American company subject to American law and even beyond law, just morally compelled to support The US Military or not?

一家私营公司开发出了足够强大的技术，以至于政府威胁要摧毁它，因为它拒绝合作。

A private company built something powerful enough that the government threatened to destroy it for not cooperating.

这并非假设。

That's not hypothetical.

就在上周，战争部将Anthropic列为供应链风险，因为该公司拒绝移除防止大规模国内监控和自主武器的防护措施。

It happened last week when the Department of War designated Anthropic a supply chain risk after the company refused to remove safeguards against mass domestic surveillance and autonomous weapons.

本·汤普森的回应并不是为任何一方辩护。

Ben Thompson's response wasn't to defend either side.

而是指出了几乎没人提到的一点。

It was to point out what almost no one was saying.

如果人工智能真如其开发者所声称的那样强大，那么手握枪支的人一定会想要发言。

If AI is as powerful as its builders claim, the people with guns are going to want to say.

无论是美国政府强制要求访问权限，还是中国因美国过于强大而采取行动，这些都不再是理论问题。

Whether that means the US government compelling access or China deciding to act because America is getting too powerful, These are no longer theoretical questions.

在这段之前在TBPN播出的对话中，约翰·库根和乔丹·海斯与Stratechery的创始人本·汤普森讨论了他关于Anthropic与对齐的论文。

In this conversation previously aired on TBPN, John Coogan and Jordi Hays speak with Ben Thompson, founder of Stratechery, about his essay Anthropic and Alignment.

我们有来自Stratechery的本·汤普森正在Restream等候室里。

We have Ben Thompson in the Restream waiting room from Stratechery.

欢迎来到

Welcome to the

节目，本。

show, Ben.

你好吗？我很好。

How are I'm you good.

希望这次我打开的是正确的麦克风。

Hopefully, I have the right microphone turned on this time.

是的，声音效果非常好。

You do, and it sounds fantastic.

非常感谢你抽空这么快加入。

Thank you so much for joining on short notice.

谢谢你写了《Anthropic与对齐》这篇文章。

Thank you for writing Anthropic and Alignment.

这是一篇非常出色的文章，我认为它涵盖了我所有的疑问。

It is a fantastic piece that I think covers all of my questions.

但我想先问问，你是怎么度过那个周末的？

But I wanna start with, like, just how did you process the the the weekend?

你是怎么走到这一步的？

How did you get to this particular place?

那么，你关于Anthropic与对齐的核心观点是什么？

And then, like, what is your key thesis with Anthropic and Alignment?

我的意思是，这篇文章恰好在周末结束时发布，我不知道这是好是坏，但我有充足的时间去思考它。

I mean, this is one of those ones I don't know if it's good or bad that it came out sort of at the end of the week, so I had a lot of time to think about it.

是的。

Yeah.

最终，我认为这是好事，因为我不确定是否有人像我这样明确地提出过这个观点。

Ultimately, I think it was good because I'm not sure anyone very as explicitly made the point I did.

是的。

Yeah.

也许这并不好，因为我觉得里面有很多限制条件。

And maybe it was bad because, I feel like there's a lot of, like, caveats.

也许事后看来，我应该在文章中加入一些内容，来回应人们感到不满的诸多观点。

Maybe in retrospect, I should have put in the article that would have addressed a lot of the points that people are upset about.

是的。

Yeah.

总的来说，放大视角来看。

Basically zooming out.

嗯。

Mhmm.

这并不是一篇规范性的文章，我没有说正在发生的事情是好是坏。

This was not a normative article where I'm saying what's happening is good or bad.

嗯。

Mhmm.

这正是我真希望当时能在文章里加上的一点保留意见。

And that's really the one caveat I really wish I would have put on there.

我的意思是，我现在竟然被像尼尔·帕特尔这样的人指责，仿佛我完全支持法西斯主义似的。

I mean, I'm being out there accused by, like, a Neil Patel, like, the full throated fascist endorsement of fascism or something like that.

别紧张。

And it's like, relax.

好吧。

Okay.

我能不能为过去这些年的工作得到一些认可？

Can I get get some some some credit for the last x number of years?

基本上，我长期以来一直有一个深层的担忧，现在我甚至都不太敢再使用‘有效利他主义’这个术语了，因为自从上周的事件后，它已经被政治化了。

Basically, the and there there is a deep rooted concern that I've had for a long time about and I'm now hesitant to even use sort of EA as a term because it's kind of now politicized, thanks thanks to the events of the last week.

但无法正视一个充满枪支的世界，说到底就是这么回事。

But a failure to grapple with a world of guns is basically the long and short of it.

事实上，我认为阿里扎是唯一一个对此坦诚相待的人，他曾在《时代》杂志上写过关于未来可能轰炸数据中心的文章。

And I actually think Alizar has been the one guy who's been honest about this where he wrote that Time article about potentially bombing data center someday.

是的

Yeah.

这其实是一个值得提出的观点，即所有这些技术目前都存在于数字领域，涉及机器人学和其他潜在应用，并且显然已被用于军事行动。

And that's actually a point worth bringing up, which is all this stuff is right now in the digital realm with robotics and potential other applications, and it's obviously being used for military operations.

它正在跨越到物理世界。

It's crossing over into the physical realm.

但如果人工智能真如人们所说的那样强大，那么现实世界必然会对此做出反应。

But if AI is as powerful as people say it's going to be, then there are going to be real world reactions to that.

如果我们像达里奥·阿马德反复做的那样，将其类比为核武器，你就必须思考：在一个私人公司研发出核武器的世界里，会发生什么。

And if we're going to analogize it to nuclear weapons, as Dario Amade has done repeatedly, you have to think through what's what would happen in a world where a private company developed nuclear weapons.

嗯

Mhmm.

政府会如何应对？

What would the government's response be?

这并不是说在这种情况下政府的反应是好是坏。

And that's not to say that the government response in that case is good or bad.

是的。

Yeah.

还是说它遵循某种宪法原则或其他类似的东西？

Or does it follow sort of constitutional principles or whatever it might be?

显然，我希望它们能这样。

Obviously, I want them to.

关于监控这一点，多年来我一直对计算机在我们监控法律中的应用感到担忧。

On the surveillance point, I've been concerned about the application of computers to our surveillance laws for years.

比如，我们社会中的许多事情都假定执行某些行为存在一定的阻力，而计算机已经消除了这种阻力，人工智能则会将这种效应放大数倍。

Like, so many things in our society assumed a certain level of friction in doing things that computers already obviated, and AI is gonna just do that on steroids.

嗯。

Mhmm.

我认为我们需要新的法律。

I do think we need new laws.

我觉得所有这些说法都是正确的。

I think all this stuff is is correct.

我认为，将人工智能应用于这些商业购买的数据集是一个巨大的问题，我不希望这种情况发生。

And I think the idea that AI being applied to these commercially purchased datasets, for example, is a huge problem that I don't want to happen.

我担心的是，如果这项技术像预期的那样强大，单方面实施限制——即使这些限制本身是好的——也不仅仅关乎谁来统治我们，正如帕尔默·拉基所深刻指出的民主问题。

The concern I have is that if this technology is as powerful as it is on pace to be, unilaterally imposing restrictions, even if those restrictions are good, isn't just an issue as far as who rules us, the democracy issue that sort of Palmer Lucky, I think, very eloquently raised.

这种做法会为那些主张如此行事的人带来非常糟糕的结果。

It's inviting very bad outcomes for those asserting that in general.

嗯。

Mhmm.

我觉得人们对这个问题缺乏足够的认识。

And I feel there's been a lack of awareness of this.

这就是为什么我提到了台湾与中国的问题。

That's why I brought up the the Taiwan China thing.

这一直是我对Anthropic公司的不满之处。

This has been a frustration I've had with Anthropic generally.

他们谈论说，Alimate在反对向中国出售芯片方面非常直言不讳，尽管只是针对某个狭窄的方面，但理由非常充分。

They talk about, you know, Alimate has been very outspoken in terms of opposing selling chips to China for in a narrow rep, you know, aspect, very, very good reasons.

我一直以来的质疑是，如果我们获得了超级强大的人工智能，而中国没有，会发生什么？

My pushback has always been what happens if we get super powerful AI and China doesn't?

他们会怎么做？

What are they going to do?

当然。

Sure.

最理想的做法就是直接摧毁台积电，因为即便带来巨大代价，这也会突然变得有利可图。

The optimal thing would be to just bomb TSMC out of existence because suddenly that becomes optimal even with all the cost that that that does.

然后呢？

And then what?

那你们打算怎么办？

Then what are gonna do?

是的。

Yeah.

就像我们正在进入这样一个领域，我不喜欢涉足政治话题。

Like, we're entering this like, I don't like getting into political Same.

帖子。

Posts.

这

It's

一点也不有趣。

not fun at all.

我根本没觉得这有什么乐趣。

I've I've I've not having fun with this.

这并不令人愉快。

It's not enjoyable.

我可以向你保证这一点。

I could promise you this.

是的。

Yeah.

有些人会说，那你本来可以把帖子设为私密的。

And some people are like, well, you should've just made the post private.

我心想，不行。

I'm like, no.

我真的很希望Anthropic公司以及与此相关的人能读到这段话，因为长期以来人们一直在推测当人工智能变得更强大时会发生什么，而现在这些情况真的开始发生了。

I actually I really want Anthropic and people associated with this to read this because people have theorized for a while about what's going to happen as AI becomes more powerful, and now it's starting to happen for real.

我想，上周末的一部分原因是我感到有责任说出这些话，并且为自己鼓足了勇气。

And I've I guess over the weekend, part of it was just I felt compelled to say this and girding myself to do so.

即使如此，我也没能等太久，这事儿我早就觉得没什么意思，但事情就是这样。

And even then, I still wasn't I haven't I haven't waited in this for a while, it's it's no fun, but it is what it is.

你能再详细解释一下你发的那条推文吗？你当时在Dario的文章里搜索了台湾，发现根本没有提到。

Can you unpack a little bit more of that that tweet that you posted where you did the find on the Dario article for Taiwan and saw that wasn't mentioned.

我的意思是，我之前就已经对这个问题抱怨过不少了。

Is I mean, I've got I've sort of griped about this in general.

我觉得。

I I think that

所以你是觉得他应该更明确地谈论台湾问题吗？

So so so do you just think he should be he should be talking about the Taiwan issue more deliberately?

他应该传达这个观点吗？

He should be messaging that?

为什么他不提台湾这件事很重要？为什么这很关键？

Like, why is it important that, why is it why is it significant that he doesn't mention Taiwan?

嗯，我认为不向中国出售芯片的立场是完全合理的。

Well oh, I think the position about not selling chips to China is a totally legitimate one.

我理解这个论点。

I understand the argument.

如果需要，我也可以提出这个论点。

I could make that argument if I needed to.

是的。

Like Yeah.

我曾经主张过相反的观点。

I have advocated the opposite.

首先，我们不仅应该向中国出售芯片，而且是出售落后一两代的技术，这一直是我们对芯片的常规做法，嗯。

That number one, not only should we should we be selling chips to China and a generation or two behind, which has always been sort of our standard practice with chips Mhmm.

我们也应该允许中国公司使用台积电代工。

We should also be allowing Chinese companies to fab with TSMC.

这项限制是后来才实施的。

That is a restriction that has come down.

现在这些华为芯片不知怎么地是由台积电制造的。

Now these Huawei chips are somehow manufactured by TSMC.

我们别太较真了，但我们应该明确允许这样做。

Let's let's not look too closely at it, but we should explicitly be allowing it.

好的。

Okay.

这样做的原因是，我认为让中国依赖台湾是更安全的平衡状态。

And the reason for that is I think it is a safer equilibrium to have China dependent on Taiwan Mhmm.

而不是试图切断他们与台湾的联系，是的。

Than to try to cut them off from Taiwan Yeah.

而我们自己却依赖台湾。

While we are dependent on Taiwan.

嗯。

Mhmm.

台湾距离中国海岸仅70英里。

Taiwan is 70 miles off the coast of China.

我们依赖台湾，而中国却不依赖台湾，这种全球格局并不理想。

It's not an ideal position in the world for us to have a dependency on it and China to not have a dependency on it.

是的。

Yeah.

所以，这就是问题所在。

So this and this is the problem.

所有这些未来的发展都伴随着巨大的权衡。

All this stuff has everything going forward has massive trade offs.

是的。

Yeah.

允许中国与台积电合作制造芯片，或允许他们购买英伟达芯片，其后果是使中国获得这些极其强大的AI能力，而这正是整个辩论的核心。

The implication of letting China fab with TSMC or the implication of letting them buy NVIDIA chips is that they gain these incredibly powerful AI capabilities that is driving this entire debate.

在孤立的情况下，这并不是一件好事。

That is, in a vacuum, not a good thing.

但没有任何事情是孤立的。

But nothing's in a vacuum.

是的。

Yeah.

每件事都是一种权衡。

Everything is a trade off.

在这一特定领域，我认为人们一再绝对化地看待芯片问题，而我却对没有看到任何关于这一点的公开评论感到沮丧，这并不公平。

And in that specific area, I think that just it's repeatedly, again and again, being absolutist about the chip issue when I am frustrated to not see any public comment about the that's not quite fair.

他曾发表过一些评论，比如，是的。

He has made comments about, oh, yeah.

如果台湾遭到轰炸，这会减缓长期的应对进程。

That would slow down sort of the adoption of anti the long run if Taiwan got got bombed.

我觉得，这在我心目中是对台湾可能被轰炸这一可能性的考虑不足。

I'm like, that's in my mind, that's an insufficient consideration of the possibility of Taiwan getting bombed.

现在，我在这一点上是有偏见的。

Now, again, I biased in that regard.

我在那里生活了将近二十年。

I lived there for for nearly two decades.

但我之所以在这个语境下提到这一点，是因为如果人工智能真的如我们所见，那么手握武器的人一定会想要发言权。

But it's just the the reason I brought it up in this context is if AI is what it is, the people with guns are going to want to have a say.

是的。

Yeah.

无论是在国内还是国际上，都可能表现为美国政府直接接管，试图打压你的公司，因为他们觉得你不配合；也可能表现为中国决定必须采取行动，因为美国正在变得过于强大。

Whether that be domestically, whether that be internationally, that might be in the context of the US government just taking it, trying to kill your company because they feel you're not cooperating, or it might be the context of China deciding it has to act because The US is becoming too powerful.

嗯。

Mhmm.

因为你知道吗？

Because the you know?

这并不是一个轻松的讨论。

And, it's not a fun debate.

确实，我认为核角度是个不错的切入点。

It it it does I do think the nuclear angle is a good one.

这让人联想到核扩散、相互确保摧毁等问题。

It has echoes of the proliferation, question of mutual assured destruction, all those sorts of things.

这将成为未来辩论的现实。

And that's just gonna be the reality of the debate going forward.

而且，这确实不太令人愉快，但我认为逃避这个问题也是不负责任的。

And again, it's not very fun, but I think it's also irresponsible to sort of run away from

上周，国防部和Anthropic之间的信息不对称起到了多大作用？

How much attention or or or what kind of factor do you think the information asymmetry between the Department of War and Anthropic played last week?

事后看来，国防部知道自己即将陷入一场重大、如今看来会旷日持久的冲突。

It felt like in hindsight, Department of War knows they're headed into a major, what is now looking like a drawn out conflict.

而Anthropic却坐在那里，想着：我们有个随意的截止日期。

Anthropic sitting there thinking, hey, we got this like arbitrary deadline.

为什么我们现在需要重新谈判？

Why do we need to renegotiate this now?

如果按照埃米尔·迈克尔的时间线来看，他们似乎直到最后一刻还在努力促成这笔交易。

And then if going off of Emil Michael's timeline, it sounds like they were still in the final hour trying to make a deal happen.

根据埃米尔的说法，达里奥当时正在开会，忙于其他事务，并没有认真对待这个截止日期，也许他认为这个期限有些随意。

And according to Emil, Dario was in a meeting and was busy and wasn't really respecting the deadline, which maybe he felt was kind of artificial.

但事后看来，这显得非常重要，因为国防部当时正将国家推向一场冲突，想要知道：嘿。

But in hindsight, now looks like it was significant because the Department of War was, you know, taking the country into a conflict and wanted to know, hey.

我们能否依靠我们的某一位AI合作伙伴？

Can we lean on one of our one of our AI partners?

我不知道。

I don't know.

我的意思是，我觉得设定这样一个截止日期似乎很随意，我也不太愿意随意猜测。

I mean, I I think that seems pretty arbitrary to have cut I mean, I'm hesitant to speculate.

我不清楚当时发生了什么。

I don't know what was going on.

我不了解其中的细节。

I don't know the angles.

我认为这就是为什么我没有深入探讨它。

I think and that's why I didn't sort of delve too deeply into it.

而且我认为像这种供应链风险这样的具体说法可能过于宽泛了。

And I also think some of the specifics like this supply chain risk probably over broad.

是的。

Yeah.

而且几乎可以肯定，推文中的表述如果真正去阅读并重新阅读相关法律条文，绝对是过于宽泛的。

And almost certainly the way it was stated in the tweet is definitely over broad if you actually go and read reread the statute.

我认为我当时的目标是——再说一遍，这时候我真希望当时多加些限定说明，比如：注意，我实际上并不是在讨论所有那些事情。

I think the goal that I was and, again, this is where I wish I had sort of put more caveats to say, look, I'm not actually talking about all that stuff.

我真的不关心。

I I don't really care.

我在意，但那不是这篇文章的重点。

I do care, but that's not the point of this article.

这篇文章的重点是，现在大家都在谈论对齐问题。

The point of this article is there's all this talk about alignment.

这就是为什么我把这一点放在标题里。

That's why I put that in in the headline.

一方面，对齐是指让人工智能与人类整体保持一致。

And on one hand, alignment is aligning AI with humanity generally.

但在可预见的未来，即使你可以在哲学上争论国家在互联网时代、更不用说人工智能时代及其未来形态中的长期存续性，这一话题无疑比以往任何时候都更加紧迫。

But for the foreseeable future, and you could have a philosophical argument about the long term viability of nation states in the age of the Internet, much less the age of AI and whatever that might be, that certainly is, you know, a more pressing conversation than probably ever before.

Anthropic 公司存在于美国的背景下。

Anthropic exists in the context of The United States.

嗯。

Mhmm.

这就是为什么我引用了那句话。

And that's why I put that quote.

你或许对政治不感兴趣，但政治却对你感兴趣。

You may not be interested in politics, but politics has an interest in you.

什么是政治？

What is politics?

以其他方式进行战争。

War by other means.

你可能对这个不感兴趣。

You might not be interested in that.

但它会对你产生兴趣。

It is going to have an interest in you.

正如我所说，长期以来一直有一种挫败感，没有充分正视这一事实，反而在宿舍里空谈AGI。

And my there's a like I said, a certain long standing frustration of not fully grappling with that fact, having dorm room theoretical arguments about AGI.

你回想一下圣诞节那篇帖子，说一百年后AGI会导致没人有工作，或变得无价值、无意义等等，其中隐含了对一百五十年后财产权仍如今天这般存在的假设。

You go back to that post over Christmas about, like, AGI in, like, a hundred years and no one having any jobs or being worthless or pointless or whatever, which included some implicit assumptions around property rights existing in a hundred and fifty years as they exist today.

醒醒吧，如果真发生了那种情况，今天的财产权将不复存在。

News flash, if that happens, property rights as they exist today are going away.

所有这些权利，这都是一个哲学问题。

All all these rights and this is a philosophic argument.

因此我才会从国际法的概念开始谈起。

That's why I'd start with the international law concept.

所有这些权利和法律，都取决于被治理者是否同意遵守它们。

All these rights, all these laws are subject to the agreement of those governed by them to follow them.

嗯。

Mhmm.

最终的决定权属于那些成功施加暴力的人。

And the final say is those who successfully inflict violence.

嗯。

Mhmm.

而且，这并不是一个令人愉快的思考话题。

And again, this isn't fun to think about.

它并不让人舒服。

It's not pleasant.

你希望假设我们生活在一个所有人都遵守法律的世界里。

You would like to assume we operate in a world of laws that everyone follows them and goes by them.

但鉴于人工智能的影响和力量如此巨大，这些我们认为早已被解决数百甚至数千年的问题，将再次被提出来。

But to the extent AI is as impactful and powerful as it is, the more these questions, fundamental questions that we thought have been settled for hundreds of years, if not thousands of years, are going to be raised.

这只是我预计会发生的若干集中的第一集。

And this is just the first of several episodes where I think that's going to happen.

我成长于后冷战时期，不需要躲进掩体。

I grew up in sort of, like, post cold war, no ducking cover.

我对核末日没什么恐惧，但达里奥·阿马迪是这本书的粉丝——关于原子弹的制造。

Didn't have a lot of fear of nuclear Armageddon, but Dario Amade is, you know, a fan of this book, the the the making of the nuclear bomb.

他似乎预测，如果人工智能变得极其强大，美国可能会采取类似当年管控核武器的方式。

And and it seemed like he sort of predicted that if if AI becomes super powerful, The US might take a similar approach that they did with nuclear regulation of nuclear weapons.

当我思考这一点时，我对核武器的管控方式感到相当乐观。

And I'm and when as I was thinking about that, I feel sort of good about the way nuclear weapons are are regulated.

我觉得我们得到了一个好结局，七十年来都没有核武器被使用过。

Like, I I feel like it it it we got the good ending, and we haven't had nuclear weapons drop in seventy years.

在这方面，事情进展得相当不错，尽管存在这种极其危险的技术，但它尚未被部署。

And it seems like things are going well there as well as they can considering that there's this amazing there you know, tremendous, like, dangerous technology that exists, but it hasn't been deployed.

它实际上并没有轰炸过任何人。

It hasn't actually, you know, bombed anyone.

但你认为他是如何消化那本书的内容的？

But how do you think he's processing that book?

你认为我们应该如何看待政府采用与核武器相同的监管策略这一想法？

How do you think you're how do you think we should be processing that idea of of the government running the same playbook that they did with nuclear weapons?

这相当有趣。

It's pretty interesting.

我的意思是，一方面，仅从物理角度来看，处理权重和软件确实如此。

I mean, on one hand, just from sort of a physical perspective, dealing with weights and software Mhmm.

这与处理可裂变材料的情况大不相同。

Is very different than dealing with fishable material.

或者我猜那些超级炸弹，它们实际上就像是聚变装置。

Or I guess the the the super bombs are like they're actually like fusion devices.

对吧？

Right?

是的。

Yeah.

这是可以追踪的。

That is trackable.

是的，确实如此。

It is Mhmm.

是可以拦截的。

Interceptable.

你知道，以伊朗为例，

You know when Iran, to take a pertinent example

是的。

Yeah.

正在试图建立浓缩设施。

Is trying to build enrichment facilities.

嗯。

Mhmm.

所有这些都使问题更容易解决。

All of which makes the problem easier to solve.

是的。

Yeah.

所以这是第一个区别。

So that's difference number one.

第二个区别，我真希望我能把这一点包含进去。

Difference number two, and I really wish I would I I had this included.

我删掉了它，以便让这篇文章更加紧凑。

I cut it so that the sort of the the article will be tighter.

但在技术史上有一个非常有趣的点，那就是英特尔的早期阶段。

But there is a very interesting point in technological history, which was the early days of Intel.

嗯。

Mhmm.

罗伯特·诺伊斯做出决定，我们会向政府销售，但不会为政府设计芯片。

And Bob Noyce made the decision that we will sell to the government, but we're not going to design chips for the government.

这里的区别在于，我们有保证的订单，这非常好。

And the distinction there was you had guaranteed orders, which was great.

政府会拿走你的知识产权，而在他看来，更重要的是市场需求有限。

The government would take your IP, and there was and in his mind, the more important thing is there was limited volume.

嗯。

Mhmm.

他正确地预见到了，设计芯片图案将是一个前期资本投入巨大的过程。

And the way that he foresaw correctly that this was going to be a very up front capital intensive process of designing shapes.

你必须设计这些图案。

You have to design them.

你必须拥有设备，而如今这些设备的投入都是以十亿美元计的。

You have to have the equipment, all of which is in the billions of dollars today.

当时，投入是几千万到几亿美元，你需要找到尽可能大的市场，也就是消费和商业市场。

Back then, it in the tens of millions and hundreds of millions is you need to find the largest possible market, which was the consumer slash business market.

你为这个市场设计产品，这将极大加速你的技术改进和能力提升，最终你将制造出比政府所能要求或自行制造的设备更先进的产品。

You design for that, that will accelerate your improvement and your capabilities so much that you will end up having better devices than the government could have ever requested or made for itself.

是的。

Yeah.

在人工智能领域， stakes 被放大到了极致。

That is at stake on steroids with AI.

是的。

Yeah.

有人问我，为什么政府不直接找人开发自己的模型呢？

People like, I was talking to someone, like, why doesn't the government just has just get someone to make their own model?

你想想政府合同，我们谈论的只是个位数的十亿美元级别。

It's like, because it's you you talk about government contracts, we're, like, single digit billions.

我们说的是在资本支出上，这些模型的成本。

We're talking about for the the amount that's going into CapEx, the cost of these models.

我们说的是模型本身需要数亿甚至接近每年一万亿美元的投入。

We're talking hundreds of mill you know, hundreds of millions of dollars for the models and hundreds of billions of dollars approaching a trillion dollars a year Yeah.

在资本支出方面。

In CapEx.

只有面向所有人销售，这种模式才是可持续且可行的。

That is only sustainable and viable if you're selling to everyone.

但这引入了全新的动态，就像政府曾研发核武器一样。

And but that introduces these entire new dynamics where the government built nuclear.

它始于那里，并且基于许多假设，因为这是一个政府项目。

It started there, and it started with a lot of assumptions because it was a government program.

由于所有前期成本的投入，我们出于经济原因，不得不依赖私营公司，而政府只是众多客户之一。

We are necessarily for economic reasons because of all the upfront costs entailed, starting with private companies of which the government is one of many customers.

这就引出了一个假设：这是一个拥有私有财产权的私营公司，所有这些我都希望是真实的。

And that introduces the assumption that, well, it's a private company with private property rights and all those sorts of things, all of which I want to be true.

再说一遍，我完全不喜欢这种发展方式。

Again, I don't like how this is going down at all.

是的。

Yeah.

这里的关键是，之所以没有朝这个方向发展，是有充分理由的。

The point here is to say there's a good reason why it's not going down that way.

而且我们必须意识到，即使这是一个正在构建通用模型的私营公司，出于非常好的原因，它也希望施加限制。

And there needs to be cognizance that even though this is a private company that is building the model general purpose and for very good reasons wants to put restrictions.

再说一遍，我认为这种相同的变体论点是一个非常有力的论点，我同意。

Again, I think the the the same variance one is a very powerful argument that I agree with.

问题是，你必须意识到，政府确实是一个小客户。

The problem is that you just need to be aware of, yes, the government is a small customer.

政府也是那个拥有武力的实体，虽然我不想这么说。

The government is also the entity, again, not to be but with guns.

我的意思是，你知道，我为什么要缴税？

Like, they they you know, like, why do I pay taxes?

因为法律规定你要缴税。

Because the law says you pay taxes.

是的。

Yep.

不。

No.

归根结底，我缴税是因为，如果你非要简化来说，如果我不缴，就会有人带着枪来我家，把我扔进监狱。

At the end of the day, I pay taxes because, you know, if you really wanna distill down, if I don't, someone with guns will come to my house and throw me in jail.

对吧？

Right?

我们根本不会去想这一点。

Like like, we don't think about that.

但归根结底，这些假设、法律和权利源自哪里？

But at the end of the day, where do these assumptions and laws and rights flow from?

只要这种情况依然存在，它就必须成为这些公司决策时的考量因素。

And as long as that is still the case, that it needs to be a decision making factor for these companies.

嗯。

Mhmm.

你觉得这对Anthropic来说会如何发展？

What how do you think this plays out for Anthropic?

这只是一个很小的合同，但在时代精神中却如此重要。

It's such a small contract, but it's so important in the zeitgeist.

有很多人因为这件事在为Anthropic声援。

There's a lot of people that are rallying around Anthropic because of this.

有很多人因为这件事正在远离Anthropic。

There's a lot of people that are pulling away from Anthropic because of this.

感觉可以打造一种不与政府合作的商业模式，却依然提供编码模型、知识检索系统以及一系列非常有价值的产品和技术，最终也能做得很好。

It feels like there is a business to be built that doesn't work with the government but delivers coding models and knowledge retrieval systems and a whole bunch of really valuable products and technology, and it winds up being fine.

但与此同时，你也不希望这种与政府之间紧张对立的复杂关系持续太久。

But at the same time, you don't want this, like, hairy relationship with the government adversarial to go on for a long time.

我希望他们能向政府销售产品，让国会通过一项法律来解决这些数字监控问题。

I would like them to sell to the government, now like Congress, to pass a law addressing these digital surveillance issues.

是的。

Yeah.

很多人觉得这不现实，我对此持开放态度。

And a lot of people are like, that's unrealistic, which I'm amenable to.

但归根结底，如果你不把‘合法还是非法’作为指导标准，唯一的替代方案就是必须有人来做决定。

But at the end of the day, if you don't have it's legal or not legal as your guiding standard, the only alternative is someone has to decide.

是的。

Yeah.

而如果这不足以成为正当理由，那就意味着是由私人高管在做决定。

And the implication of that not being a sufficient justification is that means a private executive is deciding.

是的。

Yeah.

如果人工智能真的如我们所见，我认为这将是——我用这个词——不可容忍的。

And if AI is what it is, I think that's going to be I use this word intolerable.

我不是说对我而言不可容忍。

I didn't mean intolerable to me.

我是说，对于那些握有权力的人来说，让私人高管来做这些决定，或者不做决定，是不可容忍的。

I meant intolerable to those with power to have a private executive making those decisions or not.

如果你想想，如果权力——如果我们采用一种非常粗暴的分析，认为权力来源于法律，或者法律来源于权力。

And if you think about if power if we're gonna have this very sort of brute analysis that power flows from or laws flow from power.

人工智能是一种权力来源。

AI is a source of power.

是的。

Yeah.

所以这不仅仅是这样，我认为这正是供应链问题的动机所在，虽然我并不支持这一点。

So it's not just that and I think this is where the supply chain again, which I'm not endorsing, but I think that's where the motivation is coming from.

目标并不是简单地决定我们不再使用Anthropic。

The goal isn't to find we just won't use Anthropic.

我认为真正的目标是打击Anthropic。

I do think the goal is to hurt Anthropic.

是的。

Yeah.

如果你不顺从我们，你就不会被允许建立一个权力基础。

And you're if you're not going to be subservient to us, you're not gonna be allowed to build a power base.

就此为止。

Period.

嗯哼。

Mhmm.

再说一遍，我并不认同所有这些。

And again, I'm not endorsing all this.

是的。

Yeah.

这并不令人意外，事情正在发生。

It's just a matter of it's not a surprise this is happening.

是的。

Yeah.

这确实是一个真正的风险因素，必须在所有这些决策中加以考虑。

And it'd be this needs to be just a this is a real risk factor, a real that has to be considered in all these decisions.

从达里奥的角度来看，我在想有没有其他方式可以实现目标，同时减少敌意。

Putting on my Dario hat, I'm thinking about a different way to achieve the goals with maybe less acrimony.

我提出了一个想法，也许更好的解决方案是与政府合作，然后游说通过一项监控法案，真正去推动

And I threw out this idea that maybe the better solution is, like, work with the government, but then lobby for a surveillance act and actually try

我希望白宫能出来表示，是的。

to I wish the White House would come out and say, yeah.

存在一个数字监控问题。

There's a digital surveillance problem.

我们来稍微讨论一下。

Let's work on a bit.

我的意思是，我确实不，是的。

Like, I I I don't Yeah.

我另一个可能的遗憾是，把所有这些都归咎于Anthropic。

Probably another regret I have is sort of putting this all on Anthropic.

那正是我所担心的角度。

That that was sort of the angle I was concerned about.

这让我显得很容易被人批评，认为我只是在为白宫的做法辩护。

I and that left me, I think, fairly open to the critique that this is just, like, defending the White House's approach.

而那再次说明，我当时试图从更高的层面来看，说：看。

And that was again, that was I was trying to be a higher level that's saying, look.

这就是将会发生的事。

This is what's gonna happen.

但确实如此。

But yeah.

我只是在想，从寻找一个中间立场的角度来看。

The the I'm just thinking from the perspective of to find a middle ground here.

我只是在想，如果白宫是一个不可改变的存在，而你又参与了Anthropic，那么一个建议是：嘿。

I'm just thinking of, like, from the perspective of, like, the if the White House is, like, this immutable thing, I mean, but you are in you know, involved in Anthropic, like like, one advice would be, hey.

好的。

Okay.

与其直接与政府对抗，不如成立一个政治行动委员会，通过民主程序游说实现你想要的变革。

Instead of going in and having this confrontation with the government directly, go and start a political action committee that lobbies for change in the way that you want through the democratic process.

是的。

Yes.

这才是理想的过程。

That is the ideal process.

我理解为什么人们对此感到沮丧和怀疑。

I understand why people are frustrated and skeptical about this.

好的。

Okay.

我过去经常在反垄断和聚合平台的背景下进行这样的辩论。

I used to have this debate a lot in the context of antitrust and aggregators.

关于聚合平台和反垄断，我的一个观点是，反垄断法从根本上不适合应对聚合平台，因为反垄断法历来关注的是对供应的控制，而聚合平台的力量源于对需求的控制。

And one of my sort of theses about the aggregators and antitrust is that the the antitrust laws are fundamentally unsuited to dealing with aggregators because antitrust law has historically been about control of supply, and the power of aerators flows from control of demand.

因此，你最终会得到所有这些我称之为‘推绳子’的解决方案。

And so you end up with all these solutions that I call pushing on a string.

你只是试图让人们改变他们的行为。

You're just trying to get people to change how they behave.

是的。

Yeah.

但这并不太有效。

And that doesn't work very well.

像谷歌一直以来都是对的。

Like like Google has always been right.

竞争始终只需点击一下就能实现。

Competition has always been just a click away.

问题是人们根本不会点击。

The problem is people aren't clicking.

所以，那些专注于供应端的解决方案，在供应充足但没人选择的情况下是行不通的。

And like like so so the solutions focused on the supply angle doesn't work in a world where the supply is there, just no one's choosing it.

是的。

Yeah.

因此，我的建议是，你需要制定新法律，而不是试图将这些旧法律强行套用到它们并不适用的新情境中。

And therefore, my prescription is you actually need to pass new laws, not try to retrofit these old laws to this new use case where they don't work.

而反应总是：这不可能。

And the reaction is always, that's impossible.

我们不可能通过新法律。

We can't pass new laws.

好吧。

And okay.

但你要意识到你所说的话背后的含义。

But realize the implications of what of what you're saying.