#208 亚历山德·王 - Scale AI首席执行官

然后你和然后你通过并弄清楚，好吧，有哪些可能的行动方案？所以一旦你有了态势感知，那么针对这个特定场景有哪些行动方案？老实说，你可以让一个AI代理直接提出行动方案，比如，嘿，在这个场景下，考虑到这个芯片正朝这边来，我们可以对它开火。我们可以先观望事态发展。我们可以重新部署，以便更好地应对威胁，诸如此类的各种方案。

And then you and then you go through and and figure out, okay, what are the what are the possible, courses of actions? So once you have situational awareness, then what are the courses of actions against this particular scenario? And you can have an AI agent honestly just propose courses of actions, like, hey, in this scenario, given this chip is is coming here, you know, we could fire at it. We could just wait to see what happens. We could reposition so that we're, you know, we're able to to, you know, handle the threat better, you know, all sorts of that.

我们可以重新部署一些卫星以获得更强的感知能力。你知道，有各种不同的行动方案可供选择。然后，一旦AI生成这些行动方案，它会通过模拟器运行每一个不同的行动方案。所以它会接着运行

We could reposition some satellites so we have greater sensing. You know, there's all sorts of different courses of actions, that we could take. And then, once the AI produces those courses of actions, it'll run each of those different course of actions through a simulator. So it'll then run

所以它是在实时进行兵棋推演。没错。它会实时进行兵棋推演。

So it war games at real time. Exactly. It'll war game at real time.

然后它会通过模拟器运行并说，好吧，如果我们对它开火会发生什么？比如，这是我们目前对红方部队的了解。这是我们目前对蓝方部队的了解。如果我们开火，这就是兵棋推演中可能的发展情况。如果我们只是增强感知，红方部队可能会采取这些行动来搞砸我们，这就是我们要承担的风险。

And so then it'll run through a simulator and say, okay, what's gonna happen if we fire at it? Like, you know, this is what we know about red forces. This is what we know about blue forces right now. If we fire at it, this is like, you know, this is the war game of how that plays out. If we just increase our sensing, these are the things that the Red Forces could do to fuck us up, and that's the risk that we take on.

好处是，因为这一切都是自动化的，你可以运行这些兵棋推演和模拟数百万次。所以不像军事规划人员试图在人类时间尺度上进行兵棋推演和规划。你可以运行数百万次模拟，因为你没有完美的信息。你没有完美的知识。所以你需要根据情况的不确定性，找出所有可能的结果。

And then the benefit is because all of this is automatic, you can run it, these war games and these simulations, a million times. So it's not just like one, you know, military planners just like trying to like war game and plan it out, like, you know, in human time. It's like you could run a million simulations because you don't have perfect information. You don't have perfect knowledge. So you need to kind of figure out based on the uncertainties of the situation, what are all the potential outcomes that that pop out of that?

哇。然后你运行每个行动方案的数百万次不同模拟，然后你可以直接给指挥官一个完整的简报和演示，基本上就是说，这些是我们考虑的行动方案。这些是这些行动方案的可能结果。我们可以在每个场景中展示模拟结果。所以我们可以展示如果发生的话，在每个场景中会是什么样子，比如代表性的模拟，然后由指挥官做出决定。

Wow. And then so you run like a million different simulations of each of these different courses of action, and then you can give a commander direct like, you just give them this whole, like, brief and presentation, which is basically, these are the courses of actions we considered. This is the this these are the likely outcomes in those courses of action. We can show you the simulated outcome in each one of these scenarios. So we can show you what it would look like in every one of those scenarios if it happened, like representative simulations, and then the commander makes a call.

哇。所以这就是它所做的，这些是可能的行动方案，这些是每个行动的后果，这是百分比。没错。它会在几秒钟内输出这些吗？

Wow. So it's this is what it is, this is what it's doing, these are the possible courses of action, these are the consequences of each action, this is the percentage. Yeah, exactly. And it spits that out in what, a matter of seconds?

是的，现在可能需要几个小时，因为这些模型比未来的要慢得多。但是的，我的意思是，根据情况，人类今天可能需要几天才能完成。这不是因为缺乏意愿、努力或能力。只是因为情况非常复杂。如果一艘船突然出现，有很多事情需要考虑。

Yeah, now it takes a probably takes even now, it probably takes a few hours because these models are a lot slower than they will be in the future. But yeah, I mean, that to I mean, depending on the situation, like that could take days for humans to do today. And it's not from lack of will or effort or or capability. It's just it's a really complicated situation. If a ship pops up out of nowhere, like, there's a lot of stuff you have to consider.

所以，这里真正的变革是极大地加速态势感知，极大地加速对不同行动方案的理解，可能会发生什么，后果是什么，并将这些呈现给指挥官。它会做出推荐吗？这是一个有趣的问题。我们来回讨论是否要做出推荐。因为最终，我们不想让指挥官梦游，如果这说得通的话。

And so, that's really the the the step change here is just like a, like, dramatically accelerating situational awareness, dramatically accelerating, like, an understanding of what the different course of actions are, what could happen, what are the consequences, and surfacing that to commander. Does it make a recommendation? This is kind of an interesting thing. We go back and forth if we want to make a recommendation. Because ultimately, like, we don't want to just be we don't want to let commanders sleepwalk, if that makes sense.

我们希望我们的军事指挥官是世界上最好的人，能够考虑所有这些不同行动方案的潜在后果。并且最终基于这些潜在后果做出决定。所以我认为我们希望确保指挥官在这些决策中仍然行使他们的判断力，而不是让他们更容易地说，哦，就按AI说的做。

We want them to our military commanders are the best humans in the world considering all of the potential consequences of these different courses of action. And also considering, and ultimately making a call based on those potential consequences. So I think we want to ensure that commanders are still exercising their judgment in these decisions versus just making it easier for them to just say, oh, go with what the AI says.

有意思。哇。

Interesting. Wow.

但这个...然后好吧。想想接下来会发生什么。这里事情开始变得非常诡异。假设在一个只有蓝军、只有美国拥有这种能力的世界里，那很棒。我们就能轻松超越其他所有人。

But this but then Okay. Think about what happens next. So and this is where stuff gets really freaky. So let's say that obviously, in a world where just the Blue Force, just The United States has this capability, that's great. You know, we're gonna we're gonna be running circles around everyone else.

但如果红军——比如中国、俄罗斯或其他国家——也拥有这种能力呢？那就变成我推演过的局面。他们也能瞬间完成整个战局推演。然后情况就变成...老实说我认为，当蓝军和红军都拥有完美战争推演能力时，你会选择哪条路径？

But then what happens if the Red Force, you know, China, Russia, whomever, also has the capability? Then you're in this situation where I've war gamed out the whole situation. You know, they've instantaneously war gamed out the whole situation. And then it's like then then it I think I honestly think so then it's like, we know and you know, like, blue forces, red forces, we both know that we both have, like, you know, this perfectly war game scenarios. Which avenue do you pick?

这就会演变成极其复杂的心理博弈，最终完全取决于情报质量。我们对敌方指挥官的情报有多准？对他们侦察能力的掌握如何？对他们可能了解我方情报的评估是否准确？反之亦然。局面会变得相当...

And then it becomes this really complicated, almost like psychological, of situation where it's like then it all comes down to how good our intel is. So how good is our intel about that commander? How good is our intel about what their collection capabilities are? How good our intel about what they likely know about us and vice versa? And it gets pretty

所以实际上，假设中俄等对手和我们都有这种能力，那过程就类似现在的对抗模式。谁的情报更优？只是决策速度更快了，而敌人也在以同样速度行动。

So this is actually so let's say China, Russia, our enemies have this capability, we have this capability. Then it it kinda becomes it's like the same process that we deal with now. Who has the better intel? Right? It's just developing and and and you're going to a course of action quicker, and the enemy's doing the exact same thing quicker.

本质上就是现有模式的加速版。如果我们先研发成功，就能实现全球主导。我这么理解对吗？

So it's essentially, it's the exact same thing that we're doing now but faster. And so if we develop it first, then we achieve basically global domination. Am I correct here?

没错。时机至关重要。如果我们提前一年获得这种能力——未来AI还能做更多事——我们的反应速度将形成碾压优势。我常用下棋比喻：你每走一步，我能走十步。

Yeah. And I think timing really matters here. Because if we get this capability, and this will go for I mean, there's way more AI we'll be able do. But let's say we get this capability a year ahead of adversaries, then you're then, like, we're just gonna be able to respond so much faster. The the analogy I often use is like, imagine we were playing chess, but for every one move you take, I can take 10 moves.

胜负毫无悬念。

Like, I'm just gonna win.

嗯。

Mhmm.

这就是该能力带来的不对称优势。但一旦双方势均力敌，就会如你所说，演变成基于情报和能力的对抗性冲突。

And that's what that's the asymmetric advantage that that comes out of this of this capability. And then once it but then once it equalizes, then then it's like this very, to your point, becomes this adversarial, intel based, capability based kind of conflict.

我是说，我们如何阻止对手拥有这类AI系统？

I mean, how do we combat our adversaries from having this type of AI system?

所以我认为，中国已经通过深度求索（DeepSeek）展示了实力，之后又陆续推出了其他模型。他们在AI领域将极具竞争力。2024年，也就是去年，中国的大型语言模型AI公司与中国人民解放军（PLA）签订了约80份合同。而美国这个数字远不到80，相差甚远。

So I think then mean, China's demonstrated with DeepSeek, and, you know, models have come out since then. They're gonna be very competitive on AI. And in, I think in 2024, so last year, there were something like 80 contracts between, large language model AI companies in China and the People Liberations Army, the PLA. That number is not 80 in The United States. Like, The United States is, like, way, way less than 80.

他们正非常迅速地将AI整合到国家安全和军事体系中。就目前来看，我们实际上无法阻止他们获得我描述的这种能力。于是问题就转向下一层面——情报领域。接下来需要关注两点：

So they're very clearly accelerating the integration of AI into their national security and into their military apparatus very quickly. I don't think, at this point, realistically, we can stop them from having this this capability that I described. So then you go to the next layer down. So, Intel. So it well, the next layer down the next two things that you look at is, okay.

AI如何影响情报工作？敌对的AI动态是怎样的？比如，能否用我们的AI破坏他们的AI？他们是否会用AI攻击我们的系统？这本质上就是AI对AI的战争。

How does AI impact Intel? And how does AI how can we what is the adversarial AI dynamic? Like, can we use our AIs to sabotage their AIs? Can they use their AIs to sabotage ours? And it's like AI on AI warfare, effectively.

分析这个场景时，首先要考虑的是我们之前讨论过的：最终可能取决于双方运行的AI系统数量对比。这就变成了数字游戏——如果我运行1万个AI副本而你只有100个，我依然能对你形成碾压优势。

Then when you look at that scenario Okay. So let's dig into that. The first level analysis here is kind of what we're talking about before, which is that probably just boils down to how many copies of these AI systems do I have running versus how many copies do you have running. So it turns into a numbers game. If I have 10,000 AI copies running and you only have 100 AI copies running, then I'm going run circle I'm still going to run circles around you.

这归根结底取决于其他因素。假设你有100个AI，我有10,000个AI。我会抽调一半即5,000个AI专门攻击你的AI系统。它们将全力寻找你信息架构和数据中心的漏洞。

And that boils down to who else So So let's say you have 100 AIs. I have 10,000 AIs. Will take half of my AIs. I will take 5,000 of my AIs and just focus them on hacking your AIs. So I'm gonna they're all gonna be looking for vulnerabilities in your, in your, in your information architecture, in your data centers.

我将集中所有资源对你的100个AI实施网络攻击。同时另外5,000个副本会为我进行军事规划。作为对手，我面临这样的选择：虽然我只有100个AI...

I'm gonna look for I'm gonna, you know, I'm just purely focused on cyber hacking of your 100 AIs. And then my other 5,000 copies are gonna do the military planning process for myself. Then then look at think about the adversary. I have this choice. I have a 100 AIs.

如果全部用于军事规划，就会因缺乏网络防御而被攻破。即便全部投入网络防御，100对5,000的数量劣势仍会导致失守。所以数量优势至关重要。

If I have them all focused on doing, the military planning process, I'm gonna get hacked because I'm not doing any cyber defense. And then even if I have all of them focused on cyber defense, even those numbers are bad. It's like a 100 AIs versus 5,000 AIs from you. And so I probably still get hacked. So the numbers end up mattering a lot.

即使对手仅有2倍优势（比如我有10,000副本而对手5,000），我仍可如法炮制：用5,000副本瘫痪你的AI系统——使其失效、接收错误信息或遭受数据污染；另一半则专注军事规划。

If even if they had even if the other adversary let's say it's only in two x advantage. I have 10,000 copies running and the adversary is 5,000 copies running. I can do the same thing. 5,000 of my copies are just focused on hacking your AI so that your AI is incapacitated or has incorrect information or, or is poisoned in some way, like, basically is incapable incapacity for some reason. And the other half of my eyes are focused on the military planning process.

对手依然陷入绝境：要有效应对网络攻击，我可能需要全部5,000副本投入防御，届时将再无余力进行军事规划。

Again, the adversary is screwed because to properly deal with a cyber attack, I need probably all 5,000 copies to be focused on cyber defense, and then I have no capacity left to do the military planning.

哇。

Wow.

所以它确实会演变成这样，就像你今天指挥部队的方式一样，调动所有领域的不同力量试图钳制并智胜敌人。你会为你的AI军队——或者说AI资产分配——做同样的规划。是的，你的资产分配。没错。

So it really turns into this very just in the same way that you would command your forces today, all of your various your forces across all domains to like try to pincer outmaneuver the enemy. You'll do the same kind of planning for your like AI army, to speak, your AI Allocation of assets. Yeah. Your allocation of assets. Exactly.

其中很大一部分会是：好吧，我要分配多少资源用于黑客攻击和破坏对手？多少用于我自己的军事规划和战争推演过程？另一个关键是你分配多少资源给无人机，以及多少用于执行战术任务级别的自主行动以实现任务目标。但归根结底，我认为这最终取决于谁拥有更多资源。以及这些资源是什么？

And a lot of it will be, okay, how many am I dedicating towards hacking and sabotaging the opponent? How many am I dedicating towards my own military planning and wargaming process? The other thing is how many you allocate towards towards you know, the the other key component here is drones and how many you're allocating towards doing the, like, very tactical mission level autonomy to accomplish mission level objectives. But it'll be it'll be like I think it really boils down to ultimately who has more resources. And then what are those resources?

这将与大规模数据中心有关。所以谁拥有更大的数据中心和更多电力来运行所有这些AI代理？

That's gonna be about large scale data centers. So who has bigger data centers and more power to run all these AI agents?

那么由谁来决定我们要在战术环境中部署多少AI？多少AI会去搞网络安全，试图入侵其他AI？是人类还是另一层AI，它会精确输出你刚才说的内容：这是我们的情况，这是行动方案，这是可能的结果。所以这只是一层又一层的AI在做所有这些模拟吗？

And who makes the determination of how many AIs we're gonna put in tactical environment? How many AIs are gonna go after cybersecurity trying to hack into the other AIs? Is that a human or is that another layer of AI that spits out exactly what you just said. This our situation, here's the courses of action, here's the consequences of what happened. So is it just AI after AI after AI that's doing all these simulations?

是的。没错。你说得完全正确。

Yeah. That yeah. No. You're exactly right. Then I yeah.

正是如此。你会有另一个AI来规划和布局，根据对对手的了解，我该如何分配AI资源来有效应对对手？那么，哪些关键维度能让你比对手更具优势呢？首先，如果你的AI在某些方面与众不同，对手就难以准确预测你的行动方式。

Exactly. You have another AI that's planning out and mapping out, you know, how should I allocate my AI resources to properly deal with the adversary given what I know about the adversary? And then the so then what are the ways in which, you know, what are so then what are the key dimensions that wouldn't give you an edge versus your adversary? Well, it's if, a, your AI is different somehow. So it actually is hard for your adversary to know exactly how you would act.

基本上，某种形式的出其不意——表现为AI系统不同的思维过程或推理方式。另一个因素是资源数量的模糊性。比如，如果我能让对手误以为我的资源远少于或多于实际数量，这将成为此类情境中战略突袭的关键要素。哇。

Basically, surprise in some form, in the form of a different thinking process or a different way of reasoning of the AI systems. And then the other one is ambiguity of how many what your resources actually are. Like, if somehow I can make the adversary think that I have way fewer resources than I actually do or way more resources than I actually do, that'll be a critical element of of, yeah, of strategic surprise in those kinds of situations as well. Wow.

AI能否...能否...能否在被黑客入侵时发出警报？它会知道自己被黑了吗？

Would an AI be able to be able to would would would AI be able to alert if it if it will it know it's been hacked?

这是个很好的问题。目前来说，很可能可以。但未来完全有可能出现这种情况：你能有效入侵系统或以某种方式毒害AI系统，且这种活动相对难以追踪。因为你基本上会直接黑入那个AI系统。有两种方式可以实现。

So, yeah, this is this is a great question. That you know, right now, probably, yes. The it's definitely possible in the future that you will be able to effectively hack into a system or somehow poison an AI system and have that activity be relatively untraceable. Because you would basically, would would hack into that AI system. So there's two ways you would do it.

一种方法是你毒害输入AI的数据。所以我并不是直接入侵AI本身，而是污染所有输入该AI的数据，这样在未来任何时候，我都可以激活那个AI，基本上无需任何主动入侵就能黑入它。因为我已预先毒害了输入AI的数据，只要我掌握特定触发条件——

One is you poison the data that goes into that AI. So I'm not hacking into the AI itself. I'm just poisoning all the data that's feeding into that AI, such that at any moment in the future, I, like, I can activate that AI and basically hack it without any sort of active intrusion. But I can just do it because I've poisoned I've, like, poisoned the AI that go the data that goes into the AI such that if I, like, know, save

某些历史数据会改变决策过程。

It some past alters the decision making process.

对，但是

Yeah, But

最终决策者（人类）不会意识到这点。

the end decision maker, which would be a human, would not realize that.

没错。所以数据投毒会...这正是DeepSeek最可怕的地方。中国选择开源这个模型，正是DeepSeek如此骇人的原因之一，明白吗？

Yeah, exactly. Okay. So data poisoning is gonna this is what's so terrifying about DeepSeek. One of the reasons why DeepSeek is really scary is, know, China chose to open source the model. Right?

美国有许多大型企业选择使用DeepSeek，因为他们觉得这是个好模型，好AI，而且还是免费的。不用白不用。但DeepSeek模型本身可能已被入侵，可能以某种方式被投毒，使得中共与解放军掌握了一些我们不知道的特性、行为或激活方式。这就是DeepSeek令人恐惧之处。第一个问题就是数据投毒。

So there's a lot of corporates, large scale corporates in The United States that have chosen to use DeepSeek because they're like, oh, it's a good model, and it's a good AI, and it's free. Why not use it? But deepseq itself as a model could already be compromised, could already be poisoned in some way such that, you know, there are characteristics or behavior or ways to activate deep seek that the the CCP and the PLA know about that that we don't. So so that's why deep sea gets scary. And why so so the first area is just data poisoning.

简而言之，能否对我们训练AI的数据投毒，从而如你所说，在你们不知情的情况下改变AI行为，进而对你们的整个军事行动产生连锁影响。这是其一。其二则是，如果行动足够迅速，你们可以入侵系统，就像我们之前讨论的那样抹除痕迹。在任何人察觉前，你们就能销毁所有入侵证据，清除黑客代理的踪迹。

So basically, can you poison the data that we're using to train the AIs such that, to your point, I've altered the behavior of your AIs in a way that you don't know about, and that's going to affect that's going to have cascading effects across your whole military operation. That's one. And then the second one is basically, you know, if if you're able to do the whole operation quickly enough, you basically hack in and you it's kind of as we were talking about before. You would, like, destroy the traces. You destroyed any sort of trace that, like, you had hacked in, and you have an agent that, like, hacked in, like, removed that trace and the evidence of you hacking in before anybody before it was alerted or notified.

后者可能有些极端，但短期内数据投毒问题确实更令人担忧。

That's maybe a bit more extreme, but definitely the data poisoning stuff is more concerning in the near term.

见鬼。那该怎么防范？如果AI被黑且你们知道被黑，那AI就完全不可信了，对吧？

Damn. So how would you defeat it? So if it were to be hacked and you knew it was hacked, then AI becomes completely irrelevant, correct?

问题在于我们仍然要在很多方面依赖它。

Well, the issue is we're still going to rely on it for lots of things.

这又得归结到人类思维上了。比如说有一艘战舰，你必须了解历史上采取过的所有行动，这样敌方的AI才无法预判你的战术。你需要做出一些前所未见的举动来迷惑对手的AI，对吧？没错。所以你必须做出连自己都不确定是否有效的剧烈改变，让AI无法识别——‘糟糕，我们见过这招，它接下来要这么做’。

It would have to come down to the human mind again. And you would have to, let's say it's a ship, you would have to know everything that you've done in the history so that it doesn't detect what tactic you're going to use and do something something that's never been seen before in order to confuse the adversary's AI, correct? Yeah. So you'd have to make a drastic change that you don't know if it's actually going to work so that the AI doesn't detect, oh shit, we've seen this before, this is what it's about to do.

对，对。正如你所说，战略突然性会迅速成为制胜关键。如何设计作战行动才能最大化对敌方AI的战略突然性？这是其一。坦白说第二点更关键——最终这很大程度上直接取决于你部署了多少AI副本、数据中心规模有多大、以及你在中央和战区各环境部署这些AI的工业产能。

Yeah. Yeah, so to your point, yes, strategic surprise becomes the name of the game very quickly. And how do you create an operation such that you maximize the amount of strategic surprise against an adversarial AI? That's one. And then honestly, the second thing that's really critical is a lot of this will just plain up boil down to like straight up boil down to how many copies you have running and how large your data centers are and how much industrial capacity you have to run these AIs, both centrally and at the edge in the theaters, in every environment.

AI学习新技术的速度有多快？以Cyronic为例，他们正在研发自主水面作战载具，或是Palmer Lucky的自主潜艇。假设我们与中国开战...

How fast will it learn new technology? So let's just take, for example, Cyronic. They're making autonomous surface warfare vehicles, or Palmer Lucky, he's doing the autonomous submarines. So when, what am I trying to say here? So let's say we're at war with China.

中国拥有从二战以来所有关于我们作战能力的历史数据。当战场出现Cyronic的自主载具、Epirus系统、Palmer的火箭或潜艇等新装备时，AI如何获取数据集来决策？或者说如何生成你提到的行动方案、后果预判、行为概率预测？面对战场新事物，它的学习速度能有多快？

China has all the data, all the history back from whatever, World War II on different capabilities that we have. What happens when something new is introduced onto the battle space, Ceryonic's autonomous vehicles or Epirus or Palmer's rockets or his submarines? How would the AI get the data set to make a decision, or not make decisions, but come up with what you're talking about, courses of actions, consequences, what it's about to do, probability of what's going to happen, how fast will it be able to learn when something new is introduced onto the battle space?

好问题。一般来说，当AI首次遭遇全新的无人艇或潜水器时，它无法预测会发生什么。因为它不知道新装备的航速、弹药配置或作战半径。

Yeah, this is a great question. In general, first so time it sees a totally new, let's say, USB or UUV or whatever it might be that it's never seen before, it won't be able to predict what's going happen. Like, because it won't know how fast it's going to go. It won't know you know, what, what munitions it has. It won't know what its range is.

它无法掌握关键参数——除非他们通过黑客手段提前获取了情报。假设他们不知情，那在最初几次交战中AI根本无法理解状况。这正是战略突然性的核心：持续部署敌方兵棋推演技术无法模拟的新平台。

It won't it won't know all the key, the key facts. Unless, by the way, they have really good intel and they already know all those things because they've hacked us. But let's assume they don't know. So the first few conflicts, it's not really going to be able to figure out what's happening. And that's a key component of strategic surprises, always having new platforms that won't be sort of simulatable, let's say, by enemy war gaming tech.

这确实是部分答案。但AI最终会掌握硬件性能参数，并通过模拟运算理解其影响。虽然目前军事领域完全不是这样运作的，但展望未来：最终你会运行大规模模拟，AI将计算出新型无人水面艇的航程、

So that's definitely part of it. But at a certain point, it's going to know what the hardware are capable of, and it's going to be able to run the simulations to understand how that changes the calculus. Because ultimately, what's going to happen is and some of this stuff like, you know, this is this is like, you know, some of this stuff is dissonant because obviously, if you look at what happens today in the military, it looks nothing like this. But let's play the play the tape forward and, like, see what happens in the future. Ultimately, you're gonna run large scale simulations and it's going to figure out, hey, this new, you know, unmanned surface vehicle has this much range.

机动速度、运动方式、弹药类型、通信能力，以及易受哪些电子战攻击影响。

It can go this quickly. It can maneuver in this way. It has this kinds of munitions. It has this kind of connectivity. It is vulnerable to these kinds of EW attacks, whatever they may be.

它可能遭遇这些方式的干扰，这些都将成为模拟运行的参数。所以我认为

It can be jammed in these ways, and those will all just be parameters for the simulation to run. So I think

但最初阶段它会毫无建议？

But initially, it would have no recommendations?

最初，你们会拥有战略上的突然性优势。

Initially, you'd have strategic surprise.

所以说到武器能力时，行动保密性仍然至关重要。而且我想说，最终是否总是要回归到人类心智的问题上？

So OPSEC, when it comes to weapons capabilities, is still just paramount. And it will, I mean, will it always come back to the human mind?

是的，我认为如此。我们经常讨论一个概念叫人类主权。虽然AI系统会越来越强大，但我们如何确保人类保持主权？如何确保人类真正掌控重要事务？比如维持对政治体系、军事力量、经济体系以及主要产业等各方面的控制。

Yeah, I believe so. I believe that we have this concept that we talk about a lot, which is human sovereignty. So AI systems are going to get way better, but how do we ensure that humans remain sovereign? How do we ensure that humans maintain real control over what matters? So maintain control over our political systems, maintain control over our militaries, maintain control over our economic systems, you know, our major industries, all that kind of stuff.

我认为这在军事领域尤为关键。显然我们不会——简单来说——绝不会赋予AI单方面发射核武器的能力。

And so, and I believe it's pretty paramount in the military. You're not you are not gonna wanna take certainly, just as like a as like a simplistic thing, we're not gonna give AI the capabilities to unilaterally fire nuclear weapons.

嗯。

Mhmm.

我们永远不会那么做。最终，真正关键的是信息整合、模拟推演、兵棋推演和规划，由人类做出最终正确决策。顺便说，这些都将渗透到外交领域——那些需要做出的外交决策，也会蔓延到经济战等各个方面。

Like, we're never gonna do that. And so ultimately, so much of what is gonna become really critical is the aggregation of information, simulations, wargaming, planning to humans to ultimately make the proper decisions. And by the way, so much of this will will start bleeding into the diplomatic like diplomacy, diplomatic decisions that need to be made. It'll bleed into, like, into economic warfare. Like, it'll bleed into

这甚至会影响国家间关系的建立。比如与俄罗斯结盟会带来什么结果？有哪些行动方案？会产生什么后果？我的意思是，它确实会渗透到所有领域。

I mean, this this goes all the way into I could see this going all the way into relationship building with with in between nations. What are the outcomes if we become allies with Russia? What are the courses of action? What are the consequences? I mean, does it does it so it bleeds into everything.

政治、盟友、对手、战争、经济，无一例外。

Politics, allies, adversaries, warfare, economics, all of it.

完全正确。归根结底，这种能力是什么？就是感知和态势感知。我将能处理海量数据——开源情报、各类情报来源——了解当前状态、正在发生的事件和局势。它能整合所有数据，提供对这些行为的全面视角。

Yeah, totally. Because if you ultimately boil it down, what is the capability? The capability is sensing and situational awareness. So I'm I'm gonna know I'm gonna be able to go through troves and troves of data, OSINT, other forms of, like open source intel, different kinds of various intel feeds that I have, and know what is the current status, what's going on, what is the current situation. It'll be able to aggregate all that data in to provide a comprehensive view as to what those behaviors are.

它将赋予你预测能力，让你能推演每个潜在行动可能带来的结果，提供概率性评估。然后你会将这些应用于每个重大决策——军方和政府应该把这种能力用于我们做出的每个重大决策，包括贸易政策制定。

And it'll give you the ability to predict. And it'll give you the ability to effectively play forward, you know, every potential action you could take, what would happen in those scenarios with some probabilistic view, some some probabilities. And then, yeah, you're gonna use that for every major decision. Like, the the military and the government should use this for every major decision we make. We should do it for trade policies.

我们应当为外交关系而行动，也应当向外展望，但坦率地说，我们更应将其纳入国内政策考量。比如我们的医疗政策如何？诸如此类的事项。因此，这种全方位感知加规划的能力将变得至关重要。

We should do it for diplomatic relations. We should do it for we should do it off we're looking outwards, but honestly, we should also do it for internal policies. Like, what are our health care policies? What are our all that kind of stuff too. But so it will this capability of effectively all domain sensing plus planning is going to be paramount.

我有太多疑问了。你是否预见到AI在全球变得如此强大，以至于它自身会过时？然后我们又回到——我不知道——十年前、二十年前那种完全依赖人类决策的状态？它会自我超越吗？

I have so many questions. Do you see a world where AI becomes so powerful throughout the world that it becomes obsolete? And we're right back to where we were, I don't know, ten years ago, twenty years ago, where it's all human decision making. Will it outdo itself?

我有几点想法。我认为第一阶段会发生的是，正如我所说，从人类在回路中转向人类在回路上。目前人类在经济、军事等领域从事大量密集型人力工作，这将是首要被大规模自动化的层面。之后关键在于战略决策能力，以及综合长短中期因素的高阶判断力。

A few thoughts here. I think, so one of the thing so I think the first stage of what's gonna happen is, like, kind of what I'm saying, like, human is the loop to human on the loop. Like, we're gonna right now, humans do a lot of just, like like brute force manpower work in all sorts of different places, you know, in the economy and in warfare, etcetera. That'll that's that's like the first level of of of major automation that's gonna that's gonna take place. So then it's about your strategic decision making and ability to make high judgment decisions that consider long term, short term, medium term, all that kind of stuff.

随着AI持续迭代进步，其运作节奏将远超人类跟进能力。这将首先体现在研发领域——AI将能开展大量科学研究，比人类更快地研发新武器系统、军事平台等。人类仅需复核其成果并决策，整个进程将呈加速状态。

At a certain point of of well, as certain as AI continues to improve and improve and improve and improve, it will operate at a pace that is very, very difficult for humans to keep up with. You know, this will start happening in R and D first, in research and development. Like, AI will be able to start doing lots of scientific research, lots of R and D into new weapon systems, lots of R and D into new military platforms, etcetera, much faster than humans would be able to do. And then humans will just check over their work and decide. And so it's gonna sort of race faster and faster and faster.

这会导致人类做出的少数决策权重剧增。极端情况下，像总统决定是否让本国AI与他国AI协作这类抉择，其后果将比当今类似决策重大千百倍。正如你所言，最终仍归结于人类决策，但这些决策的影响力将呈几何级增长。

And and the so then what happens, I think it'll what it'll do is it'll create dramatically more weight on the few decisions that humans make. So any decision that all the way to the extreme is the president or whomever making decisions about, do I let my AI collaborate with another country's AI? Like, that'll be like a decision of just like dramatic consequence, much higher consequence than like similar decisions today. So I think it, almost to your point, it like it will as it accelerates, will end up at a place where you're right, it all boils down to human decision making, but those decisions will carry a thousand times more consequence.

你们如何选择合作对象？毕竟这是家跨国企业。

How do you decide who you're gonna work with? I mean, it's an international company.

是的。我们曾与——

Yeah. We've had Who

目前有哪些合作伙伴？

all are you working with?

首先我们对合作方非常挑剔，毕竟资源有限。正如讨论过的，构建这些系统和数据集相当复杂。我们的目标始终是：如何与各行业顶尖机构合作？比如我提到的顶级银行、药企、电信公司、军方等。

Well, so first thing is we're pretty picky about who we work with. Ultimately, just because we only have so many resources. And building these systems and building these data sets is pretty involved, as we've discussed. So our aim generally is, how do you work with the best in every industry? How do you work with, like I was mentioning, the number one bank, the number one pharma, number one telco, number one military, etcetera.

唯一需要补充的是，我们认为至关重要的一点是：随着事态发展，必须确保全球尽可能运行在美国AI技术栈而非中共AI技术栈上。这不仅关乎意识形态、宣传控制等先前讨论的议题，更在操作层面至关重要——我们需要最大限度扩展AI能力边界。

The only addition to this that I would say we viewed as as as important is how are we as you as we play the tape forward and everything we're just discussing, it's really important that as much of the world runs on an American AI stack versus a CCP AI stack. That becomes really, really important. And it matters not only for ideology and kind of what we were talking about before, like, propaganda and control and all that kind of stuff, but it also really matters just for, like, you know, at a pure operational level, like, we're gonna want to be able to have as extended AI capabilities as possible. So

好的，我的理解是你们正在与X国合作。我们就称其为X国。你们将AI模型提供给X国使用，无论他们用于何种用途，假设是军事领域。模型所有权归我们，但他们必须接入美国的数据中心。我的理解对吗？

Okay, so the way I understand this is you're working with X country. We'll just say country X. You give Country X the AI model to utilize for whatever they're doing, let's just say warfare. We own, but they have to tap into a US based data center. Am I correct here?

因此只要我们控制着为AI模型提供数据的数据中心，本质上我们就拥有它，而X国只能相信Scale AI会维护他们的最佳利益。

And so as long as we control the data center that's feeding that AI model, we essentially own it and Country X just has to trust that Scale AI has their best interest.

没错，这就像是升级版的控制

Yeah, it's like next level

如果他们改变立场，假设X国现在与中国结盟，决定脱离美国阵营，那么我们只需切断AI的数据供给——不是AI本身，而是喂养这个AI的数据——或者操纵这些数据使其实质上被黑客入侵。我说得对吗？这就是我们保障自身安全的方式。

And if they change, if they change, let's say Country X now forms an alliance with China, they decide they don't want to be a part of America, then we just yank the AI, not the AI, the data that feeds that AI, or manipulate that data to where it's essentially been hacked. Am I correct? And that's how we keep ourselves safe.

是的。另外我认为，至少按当前普遍认知，只要数据中心由美国拥有和运营，即使位于他国境内，我们仍能在任何突发情况下保持控制权。我还要补充，初期我们会更聚焦于AI的低风险应用场景。比如能否用AI帮助这些国家的教育行业？或者医疗行业？

Yes. And then with the addition, like, I think of the way that, at least we think about it today, and I think a lot of people think about it today, is like, it's okay for the data center to be located elsewhere, located in the country, as long as it's US owned and operated because then we still have control in any sort of scenario that happens. And the only other thing I would say is we're much more focused initially on just low stakes uses of AI. So can you use AI to help the education industry in one of these countries? Or can you use it to help the health care industry?

又或者用于优化行政审批流程？这些低风险用例在初期更为重要。但我确实认为，就像地缘政治摇摆州的概念，当前世界上有许多国家，它们最终倒向美国还是中国，不仅会影响潜在冲突形态，还将决定长期冷战格局的演变。因此我将AI视为外交博弈和长期战略影响的关键要素。

Or can you use it to aid in in, like, you know, permitting processes? You know, low I think low stakes use cases matter a lot more initially. But I really do think, like, know, we have this concept of geopolitical swing states. There are there are a number of countries right now in the world where whether they side with The US or China over time is gonna have immense consequences for certainly what a potential conflict scenario looks like, but also even what the long term Cold War scenario looks like, like what happens over time as our countries are interacting. So I view AI as one of these key elements of diplomacy and long term strategic impact in international war game.

AI将如何融入我们的政府体系？我记不清你具体怎么说的。让AI运作我们的政治领域会是什么景象？

How would AI be implemented into our government? I mean, I can't remember exactly what you said. Implemented to run our political sphere. What does that look like? Yeah.

因为政治很大程度上关乎人们的价值观和信仰立场。比如当今美国可能比历史上任何时候都更加两极分化。在意识形态如此分裂的情况下——部分国民极端保守，另一部分极端自由——AI模型要如何运作政府？

So Because so much of that is people's values and what people believe in and stand for and you know, mean, like today for example. I mean, country is probably more polarized than it's ever been. And so how do you get an AI model to run government when it is this polarized and there's so many different ideologies and part of the country's way over here, the other part's way over here. How would an AI model run that?

我们有个'代理战争/代理政府'的概念。就像其他领域一样，能否用AI功能取代政府中那些低效流程，从而提升行政效率和成果质量？

Yeah. So we have this concept of of like agentic warfare, agentic government. So can you, just like the same thing, can you take these very inefficient processes in government and start replacing those with AI related functions so that you're just improving efficiency and improving outcomes?

举个具体例子。

Give me a specific example.

是的，举个超级简单的例子。目前退伍军人事务部（VA）系统内，老兵平均需要约22天才能见到医生，这耗时太长了。部分原因在于一系列过时的流程和工作方式，整个系统根本运转不灵。大家都能看出这完全不是个正常运作的体系。

Yeah, so one super simple one. Right now I think the average time it takes for a veteran to see a doctor in the VA is something like twenty two days. It's way too long. And part of that is because of a host of antiquated processes and workflows and just in general, that system's not working. I think we can all look at that and say, that's not a functional system.

那么能否利用AI——比如AI智能体——来自动化部分流程？自动获取所需审批和必要信息，把22天压缩到一两天？我认为这根本无需犹豫，纯粹是提升政府整体效率的胜利。另一个重大领域是审批流程，比如想新建数据中心或翻修住宅，某些地区的审批流程可能耗时数年，涉及无数环节的核准。

And so, can you use AI to, you know, AI agents to automate some parts of that process, automatically get whatever approvals need to be gotten, get whatever information needs to be gotten, such that that that twenty two days becomes a day or two or something like that. That I think is like a no brainer, just pure win for government efficiency overall. Another one that, other ones that are like big are like, you know, permitting processes. So if I wanna build a new data center somewhere or even I just wanna like remodel my home, the, you know, permitting processes depending where you are could take could literally take years, for all of that to go down. And part of that is like, there's so many different approvals that need to happen.

各种工作流程错综复杂。如果我们把系统规则编码化，让AI智能体自动跑完审批流程，能否在一天内获得许可或被拒？想象这种模式放大百万倍的效果。就像Doge项目发现的——联邦雇员退休档案至今仍以纸质形式存放在铁山矿场的实际矿洞里，技术落后整整两代。

There's so many like, there's all these, like, different workflows and things that need to, like, happen. What if instead we just codified what are the rules of the system and had an AI agent just go automatically go through that permitting process so that you could get that permit or or get the permit denied within, a day. Right? So and just that times a million. Like like the, like, like one of the things from from Doge that they found, right, is that, you know, the retirements are stored in the mine, Iron Mountain Mine, literal a literal, like, iron mine, or, the paper copies of the retirements for all the federal employees.

我们能否用AI将这种原始纸笔记录直接跨越到超前两代的技术？尽可能自动化这些流程？这类改造机会在政府服务和流程优化中俯拾皆是，都是唾手可得的成果。

Can we just take that, which is two generations behind in terms of tech? It's literally pen and paper, and then use AI to go from two generations behind to two generations forward? Can we just automate as much of those processes as possible? So so I see it as just like, you know, all over the place. There's so much low hanging fruit in terms of just making, current government services and government processes way more efficient.

我还没见过任何人对此持反对意见。这些都还只是最基础的第一层级改进，关乎政府运作效率的提升。

I think that I haven't met anybody who doesn't think this is the case. So that's just all the level one stuff. Yeah, that's just all the level one stuff in improving how our government operates.

最终会取代政客吗？

Would eventually replace politicians?

好问题。首先从宏观来看，政策制定速度、立法速度以及政府对新技术反应速度必须加快。我在华盛顿花了很多时间推动国家建立合适的AI立法和监管框架，确保技术良性发展。多年努力至今，我们仍未真正解决这个问题。

That's a good question. I think ultimately, like we so first off, just like taking a step back, it's definitely the case that policy the speed of policy making and the speed of legislation and the speed at which the government reacts to new technologies, like, that's gonna have to speed up. You know, we I've spent a lot of time in DC trying to make sure that as a country, we get the right kind of AI legislation and the right kind of AI regulation to ensure that this all goes well for us. It's been years of trying to get that done. We still haven't really figured that out as a country.

什么才是正确的AI监管框架？目前仍无定论。

What is the right AI regulatory framework? Like, it's still undecided.

你怎么跟华盛顿那些老古董解释这些？有人在镜头前中风，有人死在任上，有些人恐怕连他妈的电邮都不会打开。然后你28岁带着Scale AI去跟他们谈——

I mean, how do you even describe this stuff to the dinosaurs that are still sitting in DC. I mean, we've got people stroking out on camera. We've got people literally dying in office. I mean, we got people up there that probably can't even figure out how to open a fucking email. And then you come in, 28 years old, built Scale AI.

回想当年扎克伯格在国会作证，虽然我不完全认同他，但看着那群连自己邮箱都不会打开的政客，对着需要把技术稀释到幼儿园水平的科技天才指手画脚...要是我和你相处一天可能都难以消化这些概念。

Just going all the way back to when Zuckerberg's sitting there talking to Congress. I and I don't agree with everything he did and whatever, it doesn't matter. But I look at that and I'm like, you guys have been sitting in DC, probably don't even know how to open your own email, and you're talking to a tech genius who's trying to dub this down and make you understand. I mean, I get one day with you, you know what I mean? And to try to wrap my head around this.

他们手头还有五千万件其他事情要处理。他们对科技并不精通。我是说，你甚至开始... 对，接入话题了吗？

And they have 50,000,000 other things they're dealing with. They're dot up to speed on tech. I mean, do you even begin to Yeah. Tap in?

我认为很多情况下——首先这点很多知情人士都明白——大量即时决策实际上是由幕僚做出的。对吧？而且一般来说，无论什么岗位的幕僚都必须极其能干。这份工作本身就充满混乱。

I mean, think a lot of it I think the first thing and I think this is like a lot of people in the know understand this. A lot of the minute decisions really end up being made by staffers. Right? And I think generally speaking, you have to be extremely competent as a staffer no matter what. It's a very chaotic job.

需要应对的事务非常繁杂，他们必须快速做出决策。另外我觉得类比很有帮助——当今在世的人都见证了技术发展速度的持续飙升。要找到不相信AI将改变世界的人，恐怕很难。具体如何改变尚不明确，但这项技术必将重塑世界。

There's a lot that's there's a lot that's going on, and they have to make very fast decisions. The other thing is I think I think analogies are are pretty helpful. Like, I think, you know, everybody alive today has seen the pace of technology progress just increase and increase and increase and increase. Like, I think that, you know, you'd be hard pressed to defend anyone who doesn't believe that AI will be this world changing technology. Now exactly how it'll change the world, I think that's where it gets fuzzier, but it will be world changing technology.

但问题在于政治体系的反应总是滞后的。这会造成严重危害。我们必须能快速应对这些新技术。随着AI等技术加速发展，世界剧变的速度会变得非常明显。

But the issue is, mean, political system just doesn't respond very quickly. And that's going to be very harmful. We need to be able to respond very quickly to these new technologies. And so and I think they'll become more and more obvious. Like, I think I think as AI and other technologies accelerate, it'll be very obvious that, like, the world will just change so quickly.

坦白说，选民们会要求更快速的行动。因此我认为我们的政府体制需要——也必须——实现决策加速。

And frankly, I think voters are gonna demand faster action. Mhmm. And I so I think I think our government is set up to to accelerate, but but that's that's what needs to happen.

能源供给怎么解决？这可是个大议题。现在人人对核能畏之如虎，而电网系统早已过时。

How do we power all this? I mean, that's that's a big discussion. You know? And everybody seems so apprehensive to go nuclear. The grid is extremely outdated.

就在大约半小时前，这里的灯光还闪烁不定。停电事故频发——前不久整个西班牙、葡萄牙和意大利刚经历大范围停电。

I mean, we just saw the light flickers here about, I don't know, thirty minutes ago. Power outages happening all the time. There was just a big one, all of Spain. All Spain. Portugal, Italy.

美国也时刻面临断电危机。我们如何为所有技术发展提供能源支撑？你希望看到什么解决方案？

I mean, it's happening all the time in The US, power outages. How are we gonna be able to power all this stuff? I mean, what would you like to see happen?

首先，对比中美过去二十年的总发电量曲线：中国的图表呈直线上升，他们疯狂扩充电力产能——过去十年可能翻了一番。没错，发电能力翻倍。

Yeah, I mean, first of all, if you look at if you take a graph of Chinese total China's total power capacity over the past twenty years versus US total power capacity for the past twenty years, The China graph is like straight up into the right. They're just adding crazy amounts of power. They've doubled it in the last decade, I think. Doubled Doubled it? Doubled their power capacity in the last decade.

而美国基本持平，增长幅度微乎其微。

And The United States is basically flat. It's grown like a

一点点。

little bit.

所以现状就是这样。目前中国大约每十年能源翻一番，而美国基本持平。考虑到要为AI公司现在已知想建造的数据中心供电，我们需要能源容量近乎翻倍，而且必须非常迅速地实现，几乎是即刻达成。

And so we're like that that's what's happening right now. Right now, China's doubling every decade or so. US is is is basically flat. And we're looking at, you know, the for it to just power the data centers that that today AI companies know they wanna build. We're gonna need something like a doubling of our energy capacity, and that needs to happen very, very quickly, like almost that has to happen almost immediately.

因此你必须相信，我们的曲线将从完全平直转为垂直上升，增速甚至快于中国的能源增长。而中国目前正以理想速度持续增长，他们还会加速，向电网注入更多电力。在我看来，若不采取极端措施，很难想象美国能在能源容量增速上超越中国的现实场景。

And so you have to believe that our graph is gonna go from totally flat to vertical, faster vertical than than China's energy growth. And China in the meantime is just is growing is is growing perfectly quickly. They'll accelerate. They'll add more power to their grid. Like, I think it's very hard to imagine realistic scenarios where without drastic action, The United States is able to grow its energy capacity faster than China.

那么当前态势如何？如果中国直线上升而我们停滞不前，这是否意味着中国已超越我们的能源能力？还是说尽管他们在崛起，我们仍保持领先？

Now where are we on the so if China's going straight up and we're flatlined, I mean, does that mean, are you saying that China has surpassed our power capabilities? Or are we still above them even though they're on the rise?

他们绝对已超越我们。因为人口更庞大，工业规模远超我们。我确认一下数据——他们总发电量确实比我们多，发电能力更强。

They're definitely above us. Because they have a bigger population, and they have way more industrials. They have, I'll double check. I they they definitely have more power total than us. More power generation capabilities.

顺便说，这其实并非高深莫测。如果分析中国电力来源就会发现，煤电占比约80%。

And and by the way, like, it's actually not rocket science why that is. It's if you look at if you then break that down to sources of that power in China, it's because coal is, like, 80% of that.

是啊。几乎全是煤电。

Yeah. It's like They're all they're all coal.

没错。就是海量的煤炭。反观美国，可再生能源增长显著，但总量持平的原因是我们在用它们替代煤炭、天然气等化石燃料。所以美国净增长为零。

Yeah. Correct? It's just tons of coal. And then we've actually like, if you look in The US, renewables have grown a lot, but a lot of it the reason the overall number is flat is because we're using renewables to replace coal, natural gas, like fossil fuels. And so when you net it out in The US, we're flat.

而中国则是直线上升。这是首要问题——我们需要极端措施。政府设有国家能源主导委员会，我们与其多次磋商：必须采取激进行动，至少开始匹配他们电网扩容的速度，理想是超越。

And then in China, it's it's straight up. So that's the first thing. Like, need we need drastic action. You know, the administration has the National Energy Dominance Council. We've sat down with them a few times, like, that we we gotta have to we have to take drastic action to enable us to to at least start matching their speed of adding energy to the grid and ideally surpass it.

这是第一要务。第二点如你所言：我们的电网极度老化，构成重大战略风险。虽不确定西班牙大停电是否境外势力或网络攻击所致，但我敢保证美国电网对大规模网络攻击极其脆弱。这些攻击的复杂程度有时简直荒谬。

That's like that's the first thing. The second thing, like you're talking about, is our grid is extremely antiquated, and that's a major strategic risk. I don't know what the the cause or the source of the the outage across Spain was, but some people think it was a foreign actor or some kind of some kind of cyber attack of some sort. I guarantee you The US energy grid is extremely susceptible to large scale cyber attacks. It would be and the way that the sophistication of these cyber attacks sometimes is so stupid.

这就好比，如果你找到正确的发电厂登录终端入口，有时人们不会修改默认的用户名和密码，就是‘username’和‘password’。所以你可能会发现怀俄明州某个发电站的登录凭证依然是默认的‘username’和‘password’。一旦登录，你就能关闭整个地区的电力供应。正因为我们的电网如此陈旧和分散，所有这些都极度容易受到网络攻击，极易受到外国行动的影响。而现在，这至关重要。

It's like, if you find the right, like, like, power plant login terminal to go into, sometimes people don't change the username and password from the default, which is username and password. And so you can just find, like, some power station in, like, Wyoming that still has an the username and password's username and password. You log in, and you can shut down the entire, power in the entire region. So the like the so so our grid, just because of how antiquated, how decentralized it is, every all of that is hyper, hyper susceptible to to cyber attacks, hypersusceptible to foreign action, foreign activity. And, that matters now.

比如现在，如果你破坏一个大城市的能源电网，就会有人丧命。所以现状已经很糟糕了。再回到我们刚才讨论的AI问题——假设我们与中国爆发大规模AI对抗战，他们只需摧毁电网，切断数据中心及其电力供应，我们就完全束手无策了。

Like, right now, if you take the energy grid in a major city, people will die. So it's like, it's bad now. But then let's go back to what we're just talking about with AI. Like, let's say we have large scale AI on AI warfare with China. They just take out the power grid, take out our data centers and the power fueling those data centers, and then we're sitting ducks.

不仅如此，据我所知中国实际上生产了我们电网的大部分关键部件，比如变压器。更可怕的是，据我了解我们甚至不会检查这些设备是否带有恶意软件、木马程序之类的东西。事实上，能源部曾对某台设备进行过检测，却从未公开结果，这很可能意味着他们发现了问题。我真的不知道...

I mean, not only that, but it's my understanding that China actually produces and manufactures a lot of the major components that go into our grid, like the transformers. If we don't even, to my understanding, we don't even check those for malware, Trojan horses, shit like that. In fact, DOE actually did an inspection on one and never even released the results of what they found, which probably means they found some shit. And I mean, I just, I don't know

我们该如何应对。类似情况在其他地方发生过吗？看看‘盐台风’事件——这是最近解密的一次黑客攻击，中国恶意软件和网络活动已经完全渗透我们主要电信运营商。据我所知AT&T就被这个中共发动的‘盐台风’攻击彻底攻陷了。

how we combat that. I mean, like the, like what is, where did that happen elsewhere? Like, look at, Salt Typhoon. Like, this was a recent hack that was declassified, which is that Chinese malware and cyber act activity, like, basically, had fully infiltrated our, major telecom providers. I think AT and T was, like, like, entirely compromised by this hack called Salt Typhoon, from the CCP.

他们这样做是为了窃取所有信息——短信、通话录音，作为情报收集行动的一部分。既然能入侵电信系统，他们当然更有能力攻破我们的电网，以及其他任何关键基础设施。这又回到我们讨论的核心：a)如果我们无法生产足够电力就完了；b)如果对手能随意切断我们的电力，我们也完了。

And and that's they did that so that they could read all the messages. Like, the SMS, all the audio, they were able to to capture as part of that, as part of an intel gathering operation. But if they're able to hack into our telco, they've sure as hell they're clearly capable of hacking into our energy grid, clearly capable of hacking to any of our other critical infrastructure. And it just goes back to what we're talking about, like, the energy grid, a, if we can't produce enough power, we're hosed. And b, if the adversaries can take out our power at will, we're hosed.

嗯。

Mhmm.

所以我们国家在电网网络安全方面存在重大漏洞。我认为这是我们整个国家最明显、最直接的软肋之一。a)你可以制造社会动荡——想象一下休斯顿电网瘫痪会导致人员死亡和混乱；b)摧毁数据中心、军事基地、雷达系统等等任何设施...

And so we have this major major vulnerability as a country on just like the cyber posture of our energy grid. I think it's like I think it's one of the the biggest, like, very obvious, like, flat out, like, clear vulnerabilities of our overall of our entire country. A, just like you create civil unrest. You can, like, take you know, imagine you took Houston's power grid out, people would die, and, you cause, like, all sorts of chaos. But then if you but then you take out these data centers, you take out, military bases, you take out radar systems, you take out you know, you name it.

几乎任何国土基础设施的瘫痪都会给对手创造巨大的战略机会。

You can take out almost any piece of homeland infrastructure and those create huge strategic openings for adversaries.

你们既然在建设大型数据中心，去华盛顿游说需要更多电力时，接触的是哪个协会？

I mean, You have to run-in these circles. You're building massive data centers, correct? And so when you go to DC and you're advocating, hey, we need more power, what's the association you met with?

国家能源主导委员会。他们完全认同必须建设更多发电设施的观点。接下来就需要深入细节层面了。

The National Energy Dominance Council. What do they say? They totally agree. They know we have to build more power. So then you get to the next layer of detail.

就像，好吧，我们如何加速核能发展？如何加快审批流程？有哪些已关闭的现有发电能力可以重新启用？你需要逐一考虑这些自然而然的事情。我是说，我认为我们知道该怎么做。

It's like, okay, how do we accelerate nuclear? How do we accelerate the permitting process? What are existing power generation abilities that we turned off that we can turn back on? Like, you go through all the natural things to do. Like, it's I mean, I think I think we know what to do.

问题在于我们能否克服自身的阻碍。如果我们的电网过于陈旧，这种脆弱性意味着我们随时可能被击垮。

The question is if we can get out of our own way. And then if our grid is so antiquated that even that vulnerability kind of means that we can be taken out any time.

我是说，可能做了个假设。你们在建设数据中心吗？

Mean, may have made an assumption. Are building data centers?

我们自己并不建设数据中心。

We ourselves are not building data centers.

我们是在为数据中心供电。

We're feeding the data centers.

我们与那些正在建造全球最大数据中心的公司合作。

We partner with companies that, yeah, that are building the largest data centers in the world.

好的。我还听说这些大型数据中心开始自建电源。这说法有根据吗？

Okay. And so I've also heard rumors that these major data centers are starting to just create their own power source. Is there any validity to that?

是的。现在很多设计方案都在考虑，能否为每个数据中心配备一个SMR（小型模块化核反应堆）？基本上就是让核反应堆与数据中心共址，为其供电。我觉得这是个好主意。但问题是中国在这方面会遥遥领先——全球最大的核电站就在中国。

Yeah. So a lot of designs these days involve, can you just create a SMR, a small, like a like a nuclear reactor per data center? Can you basically, like, have a nuclear reactor co located with the data center to power that data center's capacity, which I think is a good idea. The issue is China's going be way ahead of us on that. The largest nuclear power plant in the world is in China.

显然我们必须发展核能，这是必然的。采用全方位发电策略时，显然也需要发展所有发电方式。但即便如此，我们也只是勉强追上中国的水平，远谈不上 confidently exceeding（自信超越）。

So obviously, we need to lean into nuclear. That needs to happen. Obviously, we need to lean into all power generation sources when you come with an all of the above approach to power generation. But even that doesn't get us to a posture where you're confidently exceeding China. You're just kind of catching up to where they are.

所以这确实是个重大问题。

And so, I mean, this is a huge issue.

好的，我们稍作休息。回来后，我想深入探讨中国的实力和我们的能力。今天你能采取的最重要行动，就是保护家人的未来免受网络罪犯和线上掠夺者的侵害。而Bunkr（拼写为b-u-n-k-r）正能助你一臂之力。

Yeah. Let's take a quick break. When we come back, I wanna I wanna dive more into China's capabilities and and our capabilities. The most important action you can take today is to help protect your family's future from cybercriminals and online predators. And Bunkr, that's b u n k r, can help you do just that.

Bunkr由拥有25年打击犯罪和网络犯罪经验的专家开发。他们最初创建Bunkr是为了保护自己的家人，现在将其分享给你。Bunkr还提供家庭套餐，包含私人通讯功能，可与配偶子女安全分享消息、照片和文件，仅限你授权的人联系。杜绝垃圾信息、监控和冒名者。

Bunkr was developed by experts with twenty five years of experience catching criminals and fighting cybercrime. They created Bunker to help protect their own families, and now they're sharing it with you. Bunker even offers a family plan. With the family plan, you get a private messenger to share messages, photos, and files with your spouse and children, connecting only with people you approve. No spam, no surveillance, no imposters ever.

套餐还包含密码管理器，保护家庭金融、购物、流媒体和社交媒体等在线账户。另有保险库功能，可安全存储照片、身份证、出生证明和护照等重要文件。立即访问bunkr.life/srs并使用优惠码SRS，家庭套餐立享25%折扣。重复一遍：bunkr.life/srs，输入代码SRS，今日家庭套餐享75折优惠。

You also get a password manager to help safeguard your family's online accounts, including financial, shopping, streaming, and social media. You even get a vault to help securely store important documents like photos, IDs, birth certificates, and passports. Visit their website bunker.life/srs and use code s r s for 25% off your family plan today. That's bunkr.life/srs and use code SRS for 25% off your family plan today. That's bunkr.life/srs and use code SRS for 25 off your family plan today.

最新报告显示，全球央行实际购金量可能是官方数据的两倍。没错，两倍。更惊人的是，部分央行正绕过传统市场，直接从非洲、亚洲和拉丁美洲的矿商处购金。

According to new reports, central banks around the world may be buying twice as much gold as official numbers suggest. You heard that right. Twice as much. And get this. Some are bypassing the traditional markets and buying gold directly from miners in Africa, Asia, and Latin America.

这意味着完全避开美元，直接获取实体黄金。这种转变不仅是象征性的，更可能是战略性的——他们或许在试图绕过西方金融体系。既然各国央行都在紧急减少美元敞口、增持实体黄金，你是否也该效仿？

That means no US dollars, just straight physical gold. The shift isn't just symbolic. It could be strategic. They could be looking to bypass Western financial systems. So if central banks are scrambling to reduce their dollar exposure and hold more physical gold, should you do the same?

这正是获奖贵金属公司Goldco的专长所在。现在注册即可免费领取2025金银套装，了解金银如何守护你的财富。只需访问SeanLikesGold.com，符合条件者还能获得最高10%的赠银奖励。立即登录seanlikesgold.com。

That's where the award winning precious metals company Goldco comes in. Right now, you can get a free 2025 gold and silver kit and learn more about how gold and silver can help you protect your savings. And all you have to do is visit Sean Likes Gold dot com. Plus, if you qualify, you could get up to 10% back in bonus silver just for getting started. Go to seanlikesgold.com.

重申网址：seanlikesgold.com。收益或有波动，请务必咨询财务与税务顾问。好了Alex，我们休息结束。

That's seanlikesgold.com. Performance may vary. You should always consult with your financial and tax professional. Alright, Alex. We're back from the break.

接下来我们将对比中美两国的技术实力。刚才我们讨论了能源领域，在人工智能竞赛中，中国是否还有其他领先美国的方面？习近平主席曾明确表示，赢得AI竞赛者将主宰全球格局。

We're getting ready to discuss some of our capabilities versus China's capabilities. And, you know, we we we just got done kinda talking about power. Is China leading The US in any other realms when it comes to the AI race? I mean, Xi Jinping even said himself, the winner of the AI race will achieve global domination.

是的。首先要明确的是，中国自2018年起就按照AI总体规划行动。中共发布了军民融合的全面政府计划来赢得AI竞赛。正如你提到的，习近平本人多次强调AI将决定这场全球竞争的胜负。从军事角度，他们明确表示——

Yeah. I think, well, the first thing, almost as you're mentioning to understand is China has been operating against an AI master plan since 2018. The CCP put out a broad whole of government civil military fusion plan to win on AI. Like you're mentioning, Xi Jinping himself has been has spoken about how AI is going to define the future winners of this global competition. In military from a military standpoint, they say explicitly, hey.

他们认为AI是蛙跳式技术：尽管当前军力不如美国，但只要在AI上超常投入，打造更智能化的军队，就能实现反超。目前最贴切的描述是：中国在能源和发电领域大幅领先，芯片技术虽落后但正在追赶，而在数据积累方面已超越我们。

We believe that AI is a leapfrog technology, which means even though our military is worse than America's military today, if we overinvest in AI, we we have a more AI enabled military than theirs, we can leapfrog them. So they've been super invested. Right now, I think the best way to kind of paint the current situation is they are way ahead on power and power generation. They're behind on chips, but catching up on chips. They are ahead of us on data.

中国自2018年起再次开展大规模数据主导行动。截至2023年，据我所知，中国有超过200万人从事数据工厂工作，主要是数据标注员或注释员，为AI系统提供数据燃料。相比之下，美国这一数字约为10万。他们在数据投入上是我们的12倍。中国有超过七座城市作为专门的数据枢纽，支撑着这种全方位的数据主导战略。

China has had, so again, since 2018, a large scale operation to dominate on data. And today, in 2023, I think, there were over 2,000,000 people in China who are working as working inside data factories, basically as data labelers or annotators, basically creating data to fuel into AI systems. I think that number in The US by comparison is something like 100,000. So they're outspending us 12 to one on data. They have over seven cities, full cities in China that are dedicated data hubs that are basically powering this broad approach to data dominance.

在算法方面，我认为他们通过大规模间谍活动与我们势均力敌。这其实是科技行业公开的秘密——中国情报机构窃取了美国所有的知识产权和技术机密。有几份令人担忧的报告：比如一名谷歌工程师盗取了谷歌AI芯片设计图纸和所有知识产权，直接带到中国创办公司。这人名叫Leon Ding。

And then on algorithms, I think they are on par with us because of large scale espionage. And this is, I think, one of these open secrets in the tech industry that Chinese intelligence basically steals all of the IP and technological secrets from, from The United States. There are a bunch of very concerning reports here. So one is there was a Google engineer who took the designs and and all the IP of how Google designed their AI chips and just took those and and moved to China and then started a company, on top using those using those designs. The way he got those designs, by this way, it was this guy, Leon Leon Ding, I think.

他窃取谷歌企业云端数据的方式简直愚蠢——直接把所有代码复制粘贴到苹果备忘录应用，导出PDF打印后带出公司。就这么简单。

The way he stole the data out of out of, Google's corporate cloud, by the way, was that he it was so stupid. He just took all the code. He copy pasted it into Apple Notes, into like the Notes app, and then exported to a PDF and printed it, and just walked walked out with it. That's it. That's it.

这件事后来才被发现。我们有几个月完全不知道这些关键知识产权被盗。斯坦福大学上周刚曝出消息——整个学校都被中共特工渗透了。几个惊人事实：

So that was this was later discovered. You know, we we found out this happened, but for months, we had, you know, we had no idea that they'd stolen all this critical IP. Stanford University, this just came out last week. Stanford University is is entirely infiltrated by CCP operatives. Few crazy facts.

首先，中国法律规定所有公民必须配合中共情报收集工作。这意味着在美国生活的中国公民若被情报机构联系，必须提供所见所闻。大量中国公民遍布顶尖大学、科技公司和AI实验室。

So first, by law in China, any Chinese citizen must comply with CCP intelligence gathering operations. So if you're a Chinese citizen, you're living in The United States, and the intelligence agencies in China reach out to you, you have to comply with them. And so you have to give them what you're seeing, what finding, etcetera. And there's tons of Chinese nationals, Chinese citizens across all the major elite universities, across all the major tech companies, across all the major AI labs. Like, they're everywhere.

第二点更惊人：约六分之一在美中国留学生接受中共直接资助的奖学金。这些学生必须向联络人汇报所学所获，否则奖学金会被撤销。美国科技行业正面临大规模情报行动，从顶尖研究机构、大学到AI实验室和科技公司都在被系统性窃密。

The second thing that's crazy is, you know, about a sixth of Chinese students, so so student like, Chinese citizens who are students in America are on scholarships sponsored by the CCP itself. And for those on these scholarships, they have to report back to a handler, basically, what are the things they find, what are the things they're learning. Otherwise, their scholarships get revoked. So we have there's there's an incredibly large scale intelligence operation running in The against The US tech industry, which is just collecting all the information and secrets and technological secrets from our greatest research institutions, our universities, our lab AI labs, our tech companies, at at massive scale. And honestly, think this is a very underrated element of how China caught up so quickly.

比如DeepSeek的突然崛起让所有人惊讶其模型能力。他们究竟自主研发了多少技术？还是通过精密间谍行动窃取美国商业机密后在中国复现？

So, you know, DeepSeek came out of nowhere. Everyone was so surprised at how capable their model was and how they learned all these tricks. You know, how much of that is because they came up with all of them on their own or they managed to have a, like, exquisite high end espionage operation to steal all of our trade secrets from The United States and then reimplement them back in China.

我们的间谍行动是什么水平？

What does our espionage look like?

远远不及。DeepSeek爆红后，其CEO会见中国总理，随后所有研究人员被集中管理并没收护照。这些AI研究员无法出境，也不接触外国人，整个研究体系被严格封锁，极难渗透。

Well, was a I think nowhere close to it is good. I mean, think so one thing that the CCP did for DeepSeek, the DeepSeek Lab, is after DeepSeek blew up and the CEO of DeepSeek met with the Chinese premier, they then locked up all the researchers into inside a I shouldn't say locked up, but they huddled all the researchers together and they took all their passports. So none of the AI researchers who work at DeepSeek are able to leave the country at all. And they don't come into contact with any foreigners. So they basically lock down the entire research effort so that makes it very, very hard to conduct any sort of espionage into that operation.

还有报道称，十年前所有/许多在华中情局特工因通信渠道被中方破解而遭清除。相比之下，中国对我们的渗透极深，风险巨大。据我所知，我们的反制能力要弱得多。

And then there's that report this is all in the news, but a decade ago, fifteen years ago, all of or many of the CIA operatives, US CIA operatives in China were all killed because they were compromised because one of the communication channels they were using was compromised by Chinese intelligence. And the CCP was able to effectively round a lot of them up and kill them. So our comparable their espionage on us is extremely deep, huge risk. There's incredible amounts of we're deeply, deeply penetrated by Chinese intel. And comparatively, as far as I know, we have much less capability.

我认为他们设计得让外界很难渗透进他们的人工智能研发领域。天啊。这就是为什么他们在数据方面领先我们。他们能轻易通过间谍手段在算法上追赶。在算力方面他们也领先我们。

And I think they've designed it such that it's very hard to infiltrate their AI efforts. Jeez. So that's how they're so they're they're ahead of us on data. They're they're able to catch up through espionage on algorithms pretty easily. They're ahead of us on power.

那我们领先什么呢？目前我们在芯片领域领先，这算是我们的救赎——英伟达芯片及其整个技术栈是世界骄傲，我们在这些芯片上最先进。中国芯片也在追赶，最近多份报告显示华为芯片基本只落后英伟达一代。

So what are we ahead at? Well, right now, we're ahead in chips, and that's kind of our saving grace is that, the NVIDIA chips and the entire stack there are the pride of the world and, you know, we're the most advanced on these chips. Chinese chips are also catching up. There's, like, a bunch of recent reports that Huawei chips are are getting to be they're basically, like, one generation behind the NVIDIA chips.

所以他们很接近了。

So they're close.

非常接近。这一切都令人担忧。CSIS最近还有份报告提到中国有个'下一代脑认知计划'，试图用AI彻底理解人类人格和心理行为。我猜最终是为了信息战。就像早餐时说的，中国开展大规模信息行动和信息战已有数十年历史，可追溯到在香港的实地行动。

They're close. So all of this is pretty concerning. There was another report that came out of CSIS recently that there was a Chinese effort called it's like the Next Generation brain understanding project or something, where they're basically trying to use AI to fully understand human human personality effectively and human psycho psychological behaviors. I imagine that's ultimately for effectively information warfare. As we were talking about at breakfast, like, I mean, China has large scale information operations, large scale information warfare, and has been has been doing that for decades and literally decades, going back all the way to like in person operations in Hong Kong.

他们在这方面极其老练，AI将让他们行动得更快。

Like they're so sophisticated at all that, and AI is going to enable them to just move much faster as well.

我们如何应对？

How do we combat that?

我们需要自己的信息战能力，这很关键。具体到这点，我们必须承认我们终究是更具创新力的国家，但若想长期赢得AI竞赛，就必须彻底整顿。我们需要将芯片制造回流本土。

Well, I mean, I think we need our own information operations efforts. I think that's pretty critical. That's specifically on that thread. And then I think we need to acknowledge that at the end of the day, we are a more innovative country, but we have to dramatically get our shit together if we want to win long term in AI. We need to onshore chip manufacturing.

必须大规模制造芯片，不能依赖台湾生产高端芯片。

Like we need to be manufacturing huge numbers of chips. We can't be dependent on Taiwan to manufacture our high end chips.

我们现在有任何产能吗？

Are we doing that yet at any capacity?

极少。亚利桑那州有几家晶圆厂能生产部分芯片，但绝大部分产能仍在台湾。我们需要大幅加强AI公司安保，建立完善反间谍机制防范企业渗透风险，还要解决之前讨论过的算力问题。

Extremely small capacity. Like there are a few fabs in Arizona that can produce some chips, but the vast majority of the volume still comes out of Taiwan. We need to tighten up security in our AI companies dramatically. We need to have proper counter intel on what is the espionage risk within these companies. We need to solve the power problem that we talked about.

我们需要加大对网络威胁的投资，比如大规模网络防御。我们需要投资数据领域。我们需要建立自己的数据主导计划，确保中国不会在高质量和更庞大的AI数据集上甩开我们。因此，你们可以逐一审视这些要素，为美国制定制胜的恰当计划。

We need to be investing into the cyber threats, like investing into large scale cyber defense. We need to invest into data. We need our own programs around data dominance to ensure that China doesn't just run away with higher quality and greater AI data sets than us. So you can go through each of the elements and build the proper plan for The United States to win.

我已经启动这些工作了吗？

Have I started any of that?

我是说，有些工作已经在进行中，但远远不够。要确保美国能赢，目前的投入还差得远。而且他们还有一项根本优势。

I mean, think some things are underway, but, no. I mean, not enough. Nowhere close to enough. For for to to be sure that The US will win, definitely not. And they also have a fundamental advantage.

你知道，现在人们常说美国需要一个人工智能曼哈顿计划——把所有顶尖人才聚集起来，整合资源，在美国开展一项大规模行动。但事实证明，在美国很难实现这一点，而中国却能轻松做到。中国可以直接下令：所有最优秀的AI人才现在都归入一家公司，我们将集中所有资源，把你们安置在全世界最大的核电站旁边。

You know, one of the things that that people say a lot now is like, oh, like, what we need in The United States is an AI Manhattan project where we, like, you know, we collect all the brilliant minds together, we collect our resources, we have one large, effort in in The United States. Well, it turns out, like, it's actually really hard to pull that off in The United States, but China can pull it off super easily. China can just say, hey, all the best AI people, you now work in one company. You we're gonna pool together all of your resources. You are you all are gonna we're gonna put you right next to the largest nuclear power plant in the in the world.

我们将在这里建造全球最大的数据中心。中国所有的芯片都将用于这个大型AI项目。他们有能力集中所有资源，全力投入AI竞赛。而在美国，我们拥有众多公司。目前美国政府不可能强制这些公司合并——这在当今会被视为政府权力的严重越界。

Like, we're gonna build the largest data center in the world here. All the chips that China has are gonna go towards building this this, like, large scale AI project, and they just have the ability to collect all their resources together and throw it at at winning on the AI race. Whereas in The United States, we have all these companies. And, you know, the United States government as of yet, like, it's not gonna force all these companies to combine and merge. Like, that's that's like such an that'd be today would be viewed as such an overreach of government power.

正因如此，我们将面临五个分散的AI项目。也许我们的芯片总量更多，算力总和更大，优秀研究人员总数更占优。但我们无法集中这些力量，而中国却能轻易整合所有资源。

But because of that, we're gonna have, like, you know, five fragmented AI efforts. And maybe in aggregate, we'll have way more chips. And in aggregate, we'll have more power. In aggregate, we'll have, more great researchers. But we're not gonna be able to focus those efforts, whereas China is easily gonna be able to focus all their efforts.

哇。你刚才在楼下提到过核武器的事。对，我记得。

Wow. You had mentioned something downstairs about nuclear weapons. Yeah. I believe.

没错。这就是国家安全变得非常诡异的地方——你可以清晰设想这样的场景：先进的网络AI会让核威慑失效。什么意思？现在没人发射核弹是因为我们有'相互保证毁灭'机制。

Yeah. So this is this is where stuff gets stuff gets really weird for for, national security, which is, you you could you could clearly imagine scenarios where advanced very advanced cyber AI invalidates nuclear deterrence. What do I mean by this? Right now, you know, nobody fires nukes because we have MAD. We have mutually assured destruction.

如果我对别国发动首轮核打击，对方在核弹飞行期间就能实施二次打击，最终双方同归于尽。正是这种二次打击能力构成了有效威慑。但假设——假设美国拥有全球最先进的AI网络攻击能力，能创建侵入任何国家的AI代理，关闭其电网，瘫痪其武器系统，瓦解一切防御。

And if I do a first strike against another country, they're gonna be able to, while that nuke is in the air, do a second strike, and we'll both you know, there'll be destruction on both sides. It'll it'll be really bad. So because of this second strike capability, luckily, we have a proper you know, we have real deterrence. Well, what if instead, let's say let's say I'm, you know, in The United States, and I have the most advanced AI cyber hacking capabilities in the world. So I can build AI agents that hack into, that can hack into any other country, can, like, turn off their energy grid, can disable their weapon systems, can disable everything.

那么我会怎么做？我会先派出网络AI部队，用压倒性的AI能力瘫痪敌国所有武器系统。等确保你们失去反击能力后，再发动首轮核打击——这样你们就丧失了二次打击能力。

So what do I do instead? I launch the first strike, and I meet or, like, first, I send in my my cyber AI agent capabilities. I send my cyber AI, you know, force effectively to disable all the weapon systems of, of the of the enemy country. And because it's, my I have, like, such low so much AI capacity, I can take out all of your I can, like, disable all of your weapon systems. And then I set my first strike, and then you don't have a second strike capability.

所以如果这种情况发生，基本上，人工智能与核武器的结合将导致——你懂的——仅靠核武器无法威慑‘AI+核武’的组合。这将迫使各国不得不扩散人工智能能力。即便是小国也需要大量投资AI能力，因为他们的核武器——

So if that happens, basically, the combination of AI and nuclear be you know, you you cannot deter AI plus nuclear with just nuclear. So then it forces this, that's what will force this proliferation of AI capabilities. And so even small countries are gonna need to invest in lots of AI capabilities because their nuclear weapons are

不再构成足够的威慑。天啊。那生物武器呢？

no longer a sufficient deterrent. Jeez. What about bioweapons?

没错。这是当前最被低估的因素。新冠病毒从武汉病毒研究所泄漏，让世界停摆两年。这还只是初级生物风险级别的威胁——相对而言，这算是一种温和的病原体，但它仍导致全球至少一千万人死亡，并让整个世界停摆两年。

Yeah. This is this is the the element that is really underrated right now. So COVID leaked out of a virology lab in, Wuhan and basically shut the world down for two years. And that's like that's like the the level one, you know, bio risk kind of stuff. Like, this was relatively, a relatively, you know, innocuous, let's say, pathogen, but it still killed, you know, probably at least ten million people globally and it was still shut the whole world down for two years.

最新的AI模型已经能超越95%的MIT病毒学家。根据人工智能安全中心的研究，OpenAI和谷歌的最新模型比MIT95%的病毒学家更聪明。所以现在——无论是当下还是几年内——利用AI能力设计强力病原体将成为可能。更重要的是，你还能定制这些病原体的特性，比如调节传播性或致死率。

Well, recent models, new models, the new AI models are able to outperform ninety five percent of MIT virologists. So the newest models from OpenAI and Google are smarter than literally 95% of virologists at MIT based on a recent study by the Center for AI Safety. So now, it's right now or whether it's in a few years, it will be feasible to use AI based capabilities to help you design powerful pathogens. And what's more than that, you're gonna be able to design in certain characteristics of these pathogens. You know, you'll be able to tune the virality, tune the lethality of them.

由于合成生物学的最新进展，现在可以制造针对特定DNA片段的病毒。这意味着我能开发只攻击携带特定DNA群体的生物武器——可以针对世界上任何族群、任何人口子集，这非常非常危险。即便没有AI，合成生物学的进步本身就带来了各种生物武器或病原体泄漏的固有风险。而有了AI后——不一定是当前模型，但几代之后的模型——你将能用这些系统设计下一代病原体。正因如此，国际条约才禁止生物战。

You know, there's also due to recent advancements in synthetic biology, you now can create viruses that specifically target certain segments of DNA. So I could create a bioweapon that just targeted any individual with a certain segment of DNA, which means I can target basically any population or any group or any subsegment of the population in the world, which is which is really, really bad. And so the the ability so first, even without AI, like biology synthetic biology is making so much progress and that there's just like all sorts of inherent risk of, like all sorts of inherent risk of bioweaponry or, you know, leaks of of pathogens and and viruses and whatnot. And then with AI all of a sudden, this is not literally today's models, but few models a few generations down, you're gonna be able to use these AI systems to design or build next generation pathogens. So that's an entire I mean, for good reason, biological warfare is not one of the not there are international treaties such that we don't engage in biological warfare.

但想象一下：当核威慑失效，某些国家又无力建设大型AI数据中心时，我担心他们会转向生物武器作为威慑手段——这对世界稳定极具破坏性。

But if you imagine these scenarios where countries nuclear deterrence doesn't work, they don't have the resources to get to use to utilize to have large scale AI data centers, I'm worried that countries will turn to biological weaponry, bioweapons, as their deterrence mechanism, which is highly destabilizing for the world.

哇，这太可怕了。

Wow. That's some scary shit.

但另一方面也有新技术能防范这些。西雅图David Baker实验室（他刚因数字嗅觉获诺贝尔奖）正在研发能自动检测空气中蛋白质、化学物质或病原体的设备。未来生物攻防战可能会演变成大规模部署这种‘数字鼻子’——每个空间、集装箱、飞机上都配备，持续监测已知或新型病原体并实施 containment。

The flip side is there is new technology that can also prevent this stuff. So there's this research coming out of this lab in Seattle, David Baker's lab, this guy who just won a Nobel Prize on biological noses, or digital noses, sorry, which is basically you have these devices that can chem detect proteins or chemicals or pathogens in the air, automatically. And so I think what this will like you know, the real sort of like offense defense of of bio and bioweaponry will end up looking like we're just gonna have large scale deployment of digital noses effectively that in every space, on every, like, shipping container, on every plane, you know, they're just constantly sensing for all existing known pathogens, any new pathogens that might exist, and are constantly just like effect, or like detecting and ultimately containing.

有意思。实时嗅探所有那些玩意儿。

Interesting. The spread of It's sniffing real time for all of that shit.

没错，正是如此。

Yeah, exactly.

我是说，另一方面，如果AI正在研发一种新型生物武器，比如又出现COVID，我们姑且称之为COVID二代，那么我们的AI也应该能找出应对方案，

I mean, also on the flip side, mean, I guess if AI is developing a new bioweapon, COVID comes out again, COVID-two, we'll just call it, then our AI should also be able to figure out the the

疫苗或是

Vaccines or

疫苗，或者说解药。对吧？

the vaccine, the antidote to it. Correct?

没错，完全正确。所以就像我们讨论AI应用于指挥控制系统时存在攻防两面性，AI应用于网络安全领域也有攻防两面性，AI应用于生物及生物武器领域同样会存在攻防对抗。

Yeah. Totally. So there will be there will be an offense defense element to just as just as in kind of as we're walking through, like, AI applied to command and control, there's an offense defense element. AI applied to cyber, there's an offense defense element. AI applied to bio and bio weaponry, there will be an offense defense element.

幸运的是，希望最终我们能达成全球共识——鉴于相互威慑的存在，世界上任何国家都不会选择走这些危险道路，因为没有人愿意 destabilize（破坏稳定），拿人类命运冒险。这基本上是我们需要达成的共识。

So all these thankfully, there's like you know, the hope is that we end up in a in a in a global world, you know, the the world agrees that basically we're not gonna go down any of these paths, like, because there's mutual deterrence, and we just you know, it's not worth it for anybody in the world to destabilize, you know, and risk humanity like that. That's that's basically where we need to land.

哇。你对中台关系有多担忧？早餐时我们简单聊过，我简直不敢相信他们至今还没采取行动。我原以为上届政府末期肯定会发生。考虑到他们的芯片制造能力，你对大陆攻台有多担心？

Wow. How concerned are you about China, Taiwan? I mean, we're talking about this a little bit at breakfast, and I I I can't believe they have not made a move yet. I mean, I thought for sure it would happen towards the end of the last administration. With their chip production capabilities, I mean, how concerned are you about China taking Taiwan?

我认为如果要发生，必定是这十年内，很可能就在本届政府任期内。

I think if it's going to happen, it's going to happen this decade, and it's probably going to happen this administration.

为什么这么说？

Why do you say that?

从中国宏观角度看，他们面临严峻的人口结构问题。这是无法回避的国情现实——老龄化人口庞大。几十年前独生子女政策的错误决策，导致他们将承受巨大的老龄人口压力。

Mean, China macro at sense, they have huge demographic issues. Those are I mean, there's not that's just the force of gravity in their country. They have this huge aging population. They made the wrong bet many decades ago to have a one child policy. And so they are going to have this huge aging population.

这个危机很快就会显现。未来十年内，他们会越来越像日本那样，被庞大的老龄人口拖累。这将严重削弱他们采取任何激进行动的能力，包括军工产能等方面。这是他们必须应对的客观制约。

That plays out really quite soon. Over the next a decade from now, over time, they're going to look more and more like Japan in that way, where they have this large aging population. And it'll paralyze a lot of ability to make any sort of aggressive moves. So part of it when it comes to military, industrial capacity, etcetera. So that's one force of gravity that they have to contend with.

因此我认为他们会希望尽早行动而非拖延。过去几十年来，他们进行了疯狂的军事建设。要知道，我认为当前形势下，中国在工业产能和制造业能力上已远超美国。这对他们来说是个机遇窗口。

And so I think they're going to want to move faster sooner rather than later. And then I mean, they've had such an insane military buildup, over the course of the past few decades. You know, I don't think it's and I think, you know, we're we're currently in a situation where China has far more industrial capacity, far more manufacturing capacity than we do in The United States. And so that's a window for them.

那你认为他们会因为人口老龄化问题而被迫采取行动吗？

So do you think they'll do, that oppressed to do it because of the aging population?

我认为有多重因素。习近平也在衰老，对吧？这将成为他个人政治遗产的重要组成部分。他们面临亚洲人口结构问题，这自然会逐渐限制他们的政治操作空间。此外，他们正处在一个惊人的窗口期——相比世界其他地区，他们拥有无与伦比的工业制造能力。

I think a lot of factors. I think Xi is aging, right? This will be an important component of his legacy as he would view it, I think. They have the Asian population, which will minimize their political latitude over time naturally. And then they have I mean, in this in in insane window where they have just incredible industrial manufacturing capabilities compared to anywhere else in the world.

2023年中国部署的工业机器人数量超过了全球其他地区的总和。我们之前讨论过自动化工厂和工业自动化的话题——他们在这方面的推进速度远超任何国家。综合这些维度和时间窗口来看，如果要行动，他们很快就会动手。

You know, in 2023, China deployed more industrial robots than the rest of the world combined. That's like I mean, we were talking a little bit about automated factories and automated industrials. They're raising it that faster than any other country in the world. And so I think that you can look at all these dimensions and this window, if they're going to do it, they're going to do it soon.

确实。我们使用的芯片有多少比例来自台湾？

Yeah. Yeah. I mean, what percentage of the chips that we use come from Taiwan?

高端芯片95%都在台湾制造。

I mean, 95% of the high end chips are manufactured in Taiwan.

那如果中国拿下台湾会怎样？

And so what happens if China takes Taiwan?

我们正在推演这个情景。假设中国封锁或入侵台湾，这些晶圆厂的价值将变得极其重要——

So yeah, we're gaming out. So we're talking a little bit about this. So let's say China blockades or invades Taiwan, then then there's a so these fabs are incredibly, incredibly valuable

嗯。

Mhmm.

正如我们所说，如果你相信AI技术的发展速度，那么一切最终都取决于你拥有多少算力和芯片。如果他们控制了全球95%的芯片制造能力，他们将取得绝对优势。于是问题就变成：台湾人会炸毁台积电的数据中心吗？或者美国会轰炸这些设施吗？还是其他国家会采取行动？

Because as we're just describing, if you believe in the pace of AI progress and AI technology, then everything boils down to how much power you got, how many chips you've got. And if they own 95% of the world's ship manufacturing capability, I mean, they're going to run away with it. So then you look at that and you say, will the Taiwanese people bomb the TSMC data centers? Or will The US bomb the TSMC data centers? And or will some other country bomb the data centers?

抱歉，不是指晶圆厂，是台积电的芯片厂。我个人认为，台湾方面不会这么做，因为即便遭到封锁或入侵，这些晶圆厂仍是台湾生存能力和实体地位的关键组成部分。所以我不认为他们会动手。中国肯定不会，因为他们入侵的部分目的就是为了获取这些能力。那么，美国会轰炸它们吗？

Or sorry, not the fabs, the TSMC chip fabs. I think my personal belief, I don't think the Taiwanese do it because even if they get blockaded or invaded, those fabs are still a huge component of Taiwan's survivability and Taiwan's relevance as an entity, even if they get blocked or invaded by Taiwan. So I don't think they do it. China definitely doesn't do it because they are invading partially to gain those capabilities. And so then, does The US bomb them?

如果美国轰炸它们，那很可能引发第三次世界大战。很难想象这不会导致大规模冲突升级。所以你会发现根本没有好的选择。虽然各方都非常关注这个问题，但这个地区确实像个火药桶。

If The US bombs them, that's probably World War three. I it's hard to imagine that not just resulting in massive escalation. And so you're looking at it and there's kind of no good options. So I think it's I mean, everyone's very focused on it obviously, but it is it is like a real powder keg of a Damn. Of a region.

你觉得最终会如何收场？我们早餐时简单讨论过。

How do you think this all ends? We added a little discussion about this at breakfast.

是的。假设未来三四年内台湾遭遇入侵或封锁，考虑到AI的重要性，美国很难不采取行动。而几乎所有可能的行动都会升级成重大冲突。所以最好的情况是我们能完全阻止入侵或封锁的发生。

Yeah. I mean, I think think if so let's assume that in the next handful of years, next, like, three, four years, there's an invasion or blockade of Taiwan. And, know, I think it's I think given how important AI is, it's hard for The US to to not take any sort of action in that scenario. And then almost all the actions you would see escalating into a major, major conflict. So best case scenario is we deter the invasion or blockade altogether.

显然，避免造成大规模破坏和人员伤亡的世界大战符合所有人的利益。因此从根本上说，我们应该有能力阻止这场冲突。正因如此，确保我国AI技术全球领先、军事AI能力世界第一至关重要。

And and I think it certainly is in everyone's interest to not get into a large scale world war that's hugely destructive and kills lots of people. So I think fundamentally, we should be able to deter that conflict. But that's why all of this matters so much. We need to make sure our AI capabilities as a country are the best in the world. We need to make sure that our military AI capabilities are the best in the world.

我们必须建立明确的经济威慑机制，全方位投入来阻止冲突。真正的危机将出现在中共的盘算与我们背道而驰时——如果他们判断'我们能拿下这个，实力足够支撑成功'，而我们的判断完全相反，世界大战就会爆发。

We need to make sure that there's clear economic deterrence of this kind of scenario. We need to be investing in every way to deter this conflict such that where this really will break down is if the Chinese if the CCP calculus, you know, diverges from our own. If their calculus becomes, oh, no. This is gonna work out. You know, we we can take this, and then, you know, we're strong enough so it'll work out for us.

所以我认为威慑是可能的。我们需要做大量工作来确保阻止冲突，这应该已经是整个国防部80%的工作重点。

And then our calculus is the opposite. That's where that's where the world war scenario happens. So so I think it's possible to deter. And I think we have to you know, there's a lot of things we have to do to make sure that we deter that conflict. And that should be, I mean, certainly, I think it already is like 80% of the focus of the entire

我们确实可以威慑，但考虑到他们人口老龄化带来的绝望感。要真正获胜，他们必须夺取那些芯片厂对吧？他们现在的造船能力已是我们的250倍，人力也占绝对优势。

DOD. Mean, it's just, we can deter, but I mean, what you're talking about an aging population, I mean, they're getting desperate. And it sounds like in order for them to legitimately win, they have to acquire those chip fabs, correct? And so, they already have two fifty times the shipbuilding capacity. They have way more people.

他们的军力已超越我们。美国征兵人数创历史新低，即便现在有所回升——我想说的是，对绝望的对手，威慑终有失效之时，他们迟早会孤注一掷。你同意吗？

Have they have more power than we do. I mean, military recruitment in The US, you know, was at an all time low. I don't know what it is today. But, I mean, even if it so I guess what I'm saying is we can you can only deter a desperate entity for so long before they throw a hail Mary play. Right?

你同意这个观点吗？

Would you agree with that?

是的。然后这就取决于

Yeah. And then it just depends on the

在我看来，需要动用整个军队包围台湾才能有效做到这一点吗？

whole an entire military to surround Taiwan to effectively do that, in my opinion?

对。我是说，我认为如果中共与解放军评估台湾已被全面包围，他们会将全部军事力量集中于夺取台湾，那就会变成一个非常棘手的战略难题。

Yeah. I mean, I think that the if if they assess if the CCP and the PLA assess that Taiwan is all around, like, they they will focus their entire military capacity on seizing Taiwan, then that becomes a really that becomes a really tricky calculus.

我是说，他们为什么不这样做？如果习近平认为赢得AI竞赛就能实现全球霸权，他年纪越来越大。你刚提到个人历史定位对他有多重要，我确信你是对的。我不知道如何阻止这种想法。然后他们就赢了AR竞赛。

I mean, why wouldn't they? If Xi believes that the winner of the AI race achieves global domination, he's getting older. You had just talked about how important his legacy is to him, which I'm sure you're right. I don't know how you deter that. And then they win the AR race.

没错。我们唯一能做的——虽然希望渺茫但很重要——就是最终在AI领域实现合作。我知道这听起来很疯狂，但如果我们国家能证明自己遥遥领先...AI发展的关键因素之一就是所谓的'AI自我改进'或'智能递归'。简单说就是当AI足够强大时，就能利用现有AI来更快地开发下一代AI。

Yeah. The only thing that we can do, I think this is a long shot, but I think it's important, is if if ultimately we actually end up collaborating on AI. And I know that sounds kind of crazy, but if we're able as a country to demonstrate just we're so far ahead and there's like you know, the one one key element of of how the whole AI thing plays out, is this idea of AI self improvement or intelligence recursion, sometimes people call it. But basically, once AIs get sufficiently good, then you can start utilizing the AIs to help you build the next AI. As sci fi as that sounds, you utilize your current generation AI to build the next generation AI faster and faster and faster and faster.

到某个阶段，AI能力会引发指数级爆发增长。如果某个对手比你落后三到六个月，他们就永远追不上了，因为你正以他们无法企及的速度运行着自我改进循环。

And so at some point, AI capabilities enable you like, you know, there's some form of, like, you know, just exponential takeoff. They just they just, you know, your AI capabilities get good really, really quickly. And if somebody's even three to six months behind you, then they're they're they're never gonna catch up to you because you're running the self improvement loop

明白了。

Gotcha.

比任何人都快。这是个关键概念，虽然目前还比较理论化。不确定智能递归是否会真的这样发展，但很多AI研究者相信这点，我也倾向于认同——我们将能用AI协助训练新一代AI并加速迭代。

Faster than anybody else. And so this is a this is a key idea. I mean, it's I think it's it's a little bit theoretical right now. Like, it's not clear whether or not this intelligence recursion is gonna be how it plays out, but but a lot of people in AI believe it. And I probably believe it too, that we will be able to use AIs to help us continue training the next AIs and improve things more quickly.

如果这个假设成立，当我们领先中国三到六个月并保持优势加速发展时，他们将远远落后。届时我们就能占据主动说：'看，我们遥遥领先，你们应该停止追赶。我们会为你们提供民用AI技术，并承诺不发展军事AI对抗。'

And if you believe that, then if we're, let's say, three to six months ahead of China, and we maintain that advantage and we take off faster, then they're gonna be way behind. And then ultimately, we're gonna be in a great position to say, hey. Actually, like, we're way ahead, and we should just you you guys should quit your efforts. We'll give you AI for all of your economic and and humanitarian uses throughout your society. And we agree we're not going to battle on military AI.

要把台湾的芯片制造能力转移到美国本土进行保护，需要什么条件？

What would it take to take the chip building capabilities that Taiwan has and implement that here in The US to protect

首先，已有数千亿美元投入这些晶圆厂的建设——他们称之为代工厂，实质上是大型芯片制造工厂及其内部高端设备和工具的配置。数千亿美元的投资。因此，美国首先也需要同等规模的资金投入。这还不是最困难的部分。

it? So the first thing is there's been hundreds of billions of dollars invested just into the build out of those fabs and the the they call it, foundries, but the buildup of large scale chip factories effectively, and all the high end equipment and tooling inside of them. Hundreds of billions of dollars of investment. So first off, there needs to be hundreds of billions of dollars investment in The US. That's not the hard part.

真正困难的是第二部分：整个运作如同一座由高度熟练、经验丰富的工人操作的大型工厂，像精密钟表般运转。除非能将这些人引进美国，否则必须从零重建所有技术诀窍和工艺能力。这需要极其漫长的过程，也是关键难点之一。

The second part that's really the hard part is all it's basically a large scale factory operated by highly, highly skilled workers who are very experienced in those processes, and the whole thing operates like a, you know, like clockwork. And unless you can get those people to The US, you're going to have to rebuild all that know how and all that technical capability. And that's what takes a really long time. And that's one of the things

那你认为我们为何没这么做？为何没激励这些杰出人才来美为我们效力？

So why do you think we haven't done that? Why do you think we have not incentivized these brilliant minds to come here and do it for us?

台积电（台湾半导体制造公司）在亚利桑那州建了几座晶圆厂，但他们指出了诸多问题：先是许可审批和电力供应不足，又遇到环保署监管问题。更棘手的是，当地技术员技能水平和工作效率远不及台湾员工。所以他们在美国建厂进展有限。

So TSMC, the Taiwan Semiconductor, the company that builds these fabs, they have stood up a few fabs in Arizona. But they cited issues. Like first, there were issues around permitting and getting enough power, and they dealt with some EPA issues. And then they just have issues where the technicians working in Arizona aren't as skilled or don't work as hard as those working in Taiwan. So they've built a few fabs in The United States.

也就是说他们尝试过，但我们的官僚程序和电力基础设施达不到要求。

So they've tried to do it, but our red tape and our power is not what it needs to be to be able to do this.

官僚程序、电力、劳动力之外，还有关键一点：从台积电角度看，他们缺乏在美国大规模布局的动机。嗯...因为一旦美国本土具备完整芯片产能，美国保护台湾的意愿就会降低。

Red tape, power, workforce. And then there's another key thing, which is if you look at it from Taiwan Semiconductor, from from TSMC's perspective, they're not all that incentivized to stand up all these capabilities in The United States. Mhmm. Like, if as soon as they start standing up all these capabilities in The United States, United States is not incentivized to defend Taiwan.

确实。

Yeah.

作为台湾企业，这关乎其生存战略。真正核心问题是：他们是否有动力在美国大规模扩建芯片产能？我认为答案基本是否定的。

And it's a Taiwanese company. So and it's a critical part of their survival strategy. So that's really where the rubber hits the road is. Are they actually incentivized to do a large scale build out of chip manufacturing capacity in The United States? I think the answer is like, no.

有道理。除非达成某种协议让他们纳入我们的保护伞下。

Makes sense. I mean, there would have to be some type of a some type of a deal struck where they fall under our wing.

对。或许能设想中美达成高层外交协议，比如'你们可以接管台湾，但美国需要本土大型晶圆厂'之类的条款。这种协议或许存在理论可能，但谁知道呢。

Yeah. I mean, you could imagine some kind of deal with with China between The US and China. It'd have to be like a diplomatic deal at the highest levels, which is something along the lines of, hey, you guys can have Taiwan, but we need large scale fabs in we need large scale chip manufacturing in The United States or something like that. And maybe there's worlds where that kind of deal could get get drawn up. I don't know.

但这意味着，美国将不得不表态说，嘿，实际上现阶段我们只关心芯片制造，至于台湾人民、那个地方以及其他一切，我们并不在乎。老兄。

But that would I mean, that would also mean that The United States would just have to say, hey. Well, all we care about actually at this point is is chip manufacturing, and that we don't care actually about the Taiwanese people and the country and all that stuff. Man.

老兄。他们和中国有任何合作吗？

Man. Are they working with China at all

台积电方面？是的。技术上他们不应该合作，但华为作为中国领军企业之一，已经通过所谓'晶圆'渠道从台湾获取了大量芯片或芯片制造的前期材料。

in capacity? The TSMC? Yeah. So they're I think they're technically not supposed to, but a lot of the the Huawei, one of the leading companies in China, has been able to get tons of chips from tons of dies, it's called. But basically tons of chips or chip prerequisites from Taiwan.

他们通常先在新加坡注册一家看似无关的壳公司，然后这家新加坡或马来西亚公司从台积电批量采购芯片，再转运回中国。显然已有大量台积电高端产品流向了中国企业。

And they usually do it through they start some cutout company that doesn't seem associated with them in Singapore. And then that Singaporean company buys a bunch of, or Malaysia, or the Singaporean Malaysian companies buy a bunch of chips from TSMC and then they mail it back or something. But there's clearly been a lot of TSMC high end outputs that have gone to the Chinese companies.

哇靠。真他妈吓人。

Wow. Wow. Scary shit, man.

更糟的是，当前局势就像火药桶——各种力量博弈下异常脆弱，充满隐患。这时候我们只能相信各方正在寻求外交解决途径。

It gets I mean, I think this is where you have to believe like right now, if you look at the you know, just as we were right now, like if you look at the situation and all of the of the dynamics at play right now, it's a powder keg. It's very, very, very volatile, highly problematic in many ways. And this is where, I mean, you just ultimately have to believe that there's got to be some effort towards diplomatic solutions.

是啊。

Yeah.

因为战争对双方都绝对是灾难。

Because it is definitely true. Like, war will be really bad for both sides.

没错。我们在AI领域怎么和中国协调？

Yeah. Yeah. How do we coordinate with China with the AI?

现状就是中美正处于全面竞赛状态——双方都全力投入开发最顶尖的AI系统，这种竞争态势会持续下去。

Yeah, so What does that look like? So yeah, right now, we're definitely US and China, we're definitely in an all out race dynamic. And we're going to race and I think this is correct we're going to race to build the best AI systems. They're going to race to build the best AI systems. And we're both all in on this approach.

我们双方都全力投入竞赛，致力于打造最先进的人工智能能力、最大的数据中心、最大容量等等。如果你还记得，核能发展也是如此。无论是核战争还是核能发电应用，当时所有人都在竞相建设产能和提升能力。然后切尔诺贝利和三哩岛事故发生了，引发了对该技术及其风险的大规模恐慌。随后出现了一系列国际条约，国际社会也大规模响应，开始协调核技术发展。

And we're both all in on racing towards building the most advanced AI capabilities, the largest data centers, the largest capacity, etcetera, etcetera. And this is, if you recall, how nuclear was. In nuclear war, as well as application of nuclear towards power production, it was all systems go everyone racing towards building capacity building capability. And then Chernobyl and 3 Mile Island happen, and it creates large scale consternation around the technology and the risks of those technologies. And there were bunch of international treaties and there's a large international response towards coordinating on nuclear technology.

话说回来，如果你审视核能发展，这些事件确实让我们国家乃至许多国家在发电领域倒退了好几代。但正是这些小型灾难的发生，才成为推动国际合作的强制力量。可以想象，在AI领域也可能出现类似情景——由于我们讨论过的种种因素，可能会出现某个恐怖组织、非国家行为体或朝鲜等实体以特别敌对或非人道的方式使用AI，从而造成大规模灾难性后果。比如导致世界某大都市断电造成大量人员死亡，或释放某种病原体导致数千万人丧生。这类事件将迫使国际社会和全世界意识到：我们必须就此展开协调。

Now, all said and done, if you really you know, if you look at nuclear, like, that set our country back, set many countries back, you know, many generations in terms of power generation. But what it took was effectively these, like, small scale disasters to take place that effectively were the forcing function for international cooperation. You can imagine a scenario with AI where, because of all the things that we've been talking about, there's some scenario where maybe some terrorist group or some non sea actor or some North Korea or whomever, somebody decides to use it in a particularly adversarial or inhumane way and and that disaster has some large scale fallout. So it create you take out power in one of the largest cities in the world and tons of people die. Or there's some pathogen that gets released and, like, tens of millions of people die, or, you know, some one of these things happens that causes the international community and everyone in the world to realize, oh, shoot, we have to be coordinating on this.

我们应该通过AI合作来改善社会、发展经济、提升人民生活。但必须就其应用进行协调——比如生物战、网络战等可怕领域。简而言之，我认为关键路径在于出现某种'AI漏油事件'般的重大事故，才能真正促使国际社会意识到：我们必须开始协调管控这项技术。

And, you know, we should be collaborating for AI to improve our societies and improve our economies and improve the lives of our people. But we shouldn't you know, we need to we need to coordinate on its use towards, for lack of a better term, scary things like bio or cyber warfare or the list goes on. So long story short, I think the path really is some kind of you know, we sometimes you talk about as like an AI oil spill or some kind of of incident that really causes the international community to realize, like, hey. We we have to we have to start coordinating on this.

你说中国在AI竞赛中全力以赴，美国也全力投入，但我们却在自缚手脚。光是提到的繁文缛节、环保署审批和电力问题就够呛。我们的发电量没有增长，这已是既定事实。

I mean, it's such you say China's all out, gone all in on the race day ice, and The US has gone all out on the race day, but we're knee capping ourselves. I mean, just mentioned the red tape, the EPA, permitting, and the power. And we're not producing more power. We're flat blind. We've established that.

据我所知，我们并没有简化行政程序来推动发展。这简直就像在自断经脉。

We're as far as I know, we're not getting rid of the red tape, you know, to to to jet launch this. And it just seems like we're cutting ourselves off at the knees here.

确实。当前我们还有很多工作要做：必须制定制胜战略实现能源主导权、数据主导权。算法方面我认为问题不大——虽然存在间谍风险，但芯片领域我们必须确保长期优势。

Right? Right now, I mean, we have a lot of work to do for sure. We have to we have to build strategies to win to have energy dominance, to have data dominance, to on the algorithms, think we'll be okay. They're gonna espionage, but I think we'll be okay on algorithms. We need to ensure we have chip dominance long term.

必须确保这些优势转化为军事主导权。我完全同意你的观点——当下就需要制定正确战略以保持全方位领先。对美国最糟糕的情景是：中共在国内开展曼哈顿计划式的大规模项目，结合我们讨论过的所有因素，意识到可以在AI领域超越美国。这将带来极端军事优势，进而用于统治世界。

We need to make sure all this lends itself to military dominance. I totally agree with you. I mean, we need to we need to, today, ensure that we have the proper strategies in place so that we stay ahead on all these areas. The worst case scenario for The United States is the following, which is CCP does a large scale Manhattan style project inside their country, realizes they can start, because of all the factors that we've talked about, they they realize they can start overtaking The US on AI. That lends itself to extreme hyper military advantage, and they use that to take over the world.

这就是美国最坏的情况。如果美中AI能力大致相当，我认为会形成威慑平衡。但若中国取得领先...那才是最危险的局面。

That's like that's like worst case scenario for The US. If US and AI AI US and China AI capabilities are even just roughly on par, I think you have deterrence. I don't think either country will take the risk. I think if US is way ahead of China, I think you maintain US leadership, and that's a pretty safe world. So that the the worst case scenario is they get ahead of us.

除了美中之外，还有其他参与者吗？我们还需要警惕哪些国家？

Are there any other players other than The US and China involved in this? Who else do we need to be watching out for?

目前主要是美中。虽然许多国家都参与其中，但并非所有都具备成为AI超级大国的要素。不过有些国家拥有关键优势——比如在网络战和信息战领域，俄罗斯拥有非常先进的作战能力。如果他们与中共结盟，可能会产生重大影响。

So, yeah, right now, definitely US and China. A The lot of other countries will matter, but not all of them have enough ingredients to really properly be AI superpowers. So, but other countries are gonna they have they have key ingredients. So to name a few, A, everything we've talked about with cyber warfare and information warfare, information operations, Russia has very advanced operations in those areas. And that could end up mattering a lot if they ally with the CCP.

他们有很多合作方式，那可能会相当糟糕。中东国家将非常重要，因为他们拥有巨额资本和丰富的能源。因此，这些国家在整个局势中扮演着关键角色。印度也至关重要，印度拥有大量高端技术人才。

There's a lot of ways they can team up and that could be pretty bad. The countries in The Middle East will be very important because they have incredible amounts of capital and they have lots of energy. And so that's these are they're critical players in how all this plays out. India matters a lot. India has a lot of high end technical talent.

我不确定现在印度和中国哪个拥有更多高端技术人才，但印度肯定有很多。庞大的人口也开始真正工业化，与中国类似。所以印度将非常重要。欧洲也有很多技术人才，但欧洲的能力如何发挥作用尚不明确。

I don't know if I think right now, I don't know if between India and China, which has more high end technical talent, but there's a lot in India for sure. Massive population also starting to industrialize in a real way and right to China. So India will matter a lot. And there's a lot of technical talent in Europe as well. I think it's unclear exactly how this plays out with the European capabilities.

我的意思是，他们必须——现在似乎有一些努力让欧洲尝试建立大规模算力，建设大型数据中心，采取行动。这些努力的效果还有待观察，但可以清楚地看到一些情景，如果他们全力投入，他们也可能成为重要角色。

I mean, they have to it seems like there's some efforts now for Europe to try to build up large scale power, build up large data centers, make a play. I think yet to be seen how effective those efforts are going to be, but you can clearly see some scenarios where if they make a hard turn and go all in, they could be relevant as well.

有没有可能AI会发展出自己的意识？所以

Is there a world where AI takes on a mind of its own? So

显然，你可以假设性地描绘一个场景，即存在超级智能或非常强大的AI。然后它在某个时候意识到人类有点烦人，就把我们都消灭了。但我认为这是一个完全可以避免的结果。因为首先，我们刚才讨论的所有事情都是在你拥有这种超先进AI之前就会发生的非常现实的事情。这是第一点。

obviously, you can hypothetically paint the scenario where you have superintelligence or you have really powerful AI. And then it realizes at some point that humans are kind of annoying and takes us all out. But I think it's a very that's so preventable as an outcome. Because first of all, all the things we just talked about are like the very real things that happen long before you have this hyper advanced AI that takes everyone out. That's first that's first thing.

所以在那之前我们有很多事情需要做好。其次，要让AI真正有能力拥有自己的意识并消灭所有人类，我们必须给予它难以置信的控制权。基本上它必须掌控一切，而我们只是随波逐流。嗯。这是一个选择。

So we have lots of things we have to get right before then. And then second is, you know, for AI to actually, be capable of, you know, having a mind of its own and taking all humans out, like, we'd have to give it just incredible amounts of control. Like, it would have to just basically be running everything, and we're just sort of, like, along for the ride. Mhmm. And that's a choice.

我们有权选择是否将所有控制权交给AI系统。正如我之前谈到的人类主权，我的信念是我们不应该放弃对我们最关键系统的控制。我们应该设计所有系统，使人类决策和人类控制非常重要。人类监督非常重要。这实际上是我们公司正在努力的事情之一。

We have this choice of whether or not to, like, give all of our control to AI systems. And as I was talking about before with, like, human sovereignty, I my belief is we should not cede control of our most critical systems. Like, we should we should design all the systems such that human decision making, human control is really, really important. Human oversight is really important. This is one of the things that I actually think is is, one of the things that we're working on as a company.

说实话，当我思考长期使命时，最重要的事情之一就是建立人类主权。首先，我们如何确保输入这些AI模型的所有数据都能增强人类主权，使模型按照我们的指令行事，与人类目标一致。其次，我们建立监督机制。随着AI开始执行更多行动、更多规划，在经济、军事等领域承担更多任务，人类要监督和指导每一个行动。这样我们才能保持控制。

It's honestly one of like, as I think about, like, long term missions, one of the most important things is creating human sovereignty. So first is how do we make sure all the data that goes into these AI models, increases human sovereignty such that the models are going to do what we tell them, are aligned with humans and aligned with our objectives. And two is that we create oversight. So as AI starts doing more and more actions, doing more planning, taking out carrying out more things in the world, in the economy, in the military, etcetera, that humans are watching and supervising every one of those actions. So that's how we maintain control.

这样我们才能避免《终结者》那样的场景或AI消灭人类的场景。

And that's how we prevent the Terminator scenarios or the AI takes us out kind of scenarios.

有趣。好吧，Alex，我们在这里结束采访。但是，哇，多么迷人的讨论。谢谢。谢谢你来到这里。

Interesting. Well, Alex, wrapping up the interview here. But, man, what a fascinating discussion. Thank you. Thank you for being here.

最后一个问题。如果有三位嘉宾你希望看到出现在节目里，会是谁？

One last question. If you had three guests you'd like to see on the show, who would it be?

哦，这是个好问题。我想见谁呢？其实我挺喜欢你最近的做法，就是邀请更多科技界人士上播客。所以我会往这个方向选。比如我觉得埃隆上节目会很棒。

Oh, that's a good question. Who would I like to see? Well, I really like what you've been doing recently, which is getting more tech folks on the on the pod. So so I go in that direction. I mean, think Elon would be great to see on the show.

我记得我们聊过这个——扎克上节目应该会很有意思。山姆·奥特曼也是个不错的选择。所以肯定是更多科技领域的人。除此之外，我们之前也讨论过，国际领导人这类嘉宾也很重要。

I think I think we were talking about this. Zach would be would be cool to see on the show. I think Sam Altman would be cool to see on the show. So definitely, like, more people in tech. Outside of that, I think, and we're talking about some of this, international leadership, like leaders of other countries is super important.

因为我们讨论的所有这些情景中，国际合作都至关重要。

Because we talk about all these scenarios, international cooperation is going to matter so much.

说得对。我们会联系他们的。至于世界领导人方面，我们已经在行动了。不过亚历克斯，再次感谢你能来，兄弟。真是场精彩的对话。

Right on. We'll reach out to them. And as far as world leaders is concerned, we're on it. But, well, Alex, thanks again for coming, man. Fascinating discussion.

看到你二十八年来取得的所有成就，我真心为你高兴。真的。我特别欣赏这一切。所以感谢你的到来，我知道你是个大忙人。

I'm just super happy to see all the success that you've amassed throughout your twenty eight years. It's it is. I love seeing it. So thank you for being here. I know you're a busy guy.

是啊，谢谢邀请。这次很愉快。

So Yeah. Thanks for having me. It was fun.

我是迈克尔·罗森鲍姆。我是汤姆·威灵。欢迎收听《闲谈小镇》，在这里讨论《超人前传》其乐无穷。我们偶尔也会邀请客串明星来聊天。

I am Michael Rosenbaum. I am Tom Welling. Welcome to Talkville. Where it's fun to talk about smallville. We're gonna be talking to sometimes guest stars.

你喜欢弗洛雷斯现在的发展方向吗？

Are you liking the direction Flores is going in?

喜欢啊，因为我的镜头变多了。

Yeah. Because I'm getting more screen time.

挺好的。但主要是，只有我和汤姆还记得。我觉得我们都觉得这里少了一场戏。你们懂我的意思吧，汤姆。我们重新回顾一下。

It's good. But mostly, it's just me and Tom remembering. I think we all feel like there was a scene missing here. You got me, Tom. Let's revisit it.

我们来看看。看看我们记得什么。看看我们记得什么。

Let's look at it. See what we remember. See what we remember.

我以前从未经历过类似的事情。

I had never been around anything like that before.

我是说，那太有趣了。Talkville。Talkville。我刚才突然闪回了一下。

I mean, it was so fun. Talkville. Talkville. I just had a flashback.

关注并在你喜欢的平台上收听。

Follow and listen on your favorite platform.

让我们开始吧。

Let's get into it.