本集简介
双语字幕
仅展示文本字幕，不包含中文音频；想边听边看，请使用 Bayt 播客 App。
本周是特别节目。
Special episode this week.
我们取消了本周的常规节目，只有三个人值得我们这么做：特朗普总统、耶稣和詹森。
We've preempted the weekly show, and there's only three people we preempt the show for, president Trump, Jesus, and Jensen.
你可以选择我们按什么顺序来讨论他们。
And I'll let you pick which order we do that.
但你取得了如此了不起的成就，还有这么棒的活动。
But what an amazing run you've had and and a great event.
每个行业都在这里。
Every industry is here.
每一家科技公司都在这里。
Every tech company is here.
每一家人工智能公司都在这里。
Every AI company is here.
太不可思议了。
Incredible.
不可思议。
Incredible.
非同寻常。
Extraordinary.
过去一年最了不起的公告之一就是Groq。
And one of the great announcements of the past year has been Groq.
你收购Groq的时候，有没有预料到查马斯会变得这么让人头疼？
When you made the purchase of Groq, did you realize how insufferable Chamath would become?
我有点预感，那个，那个，那个
I had a I had an inkling that that that
我们是他的朋友。
We're his friends.
我们每周都得应付他。
We have to deal with them every week.
我知道。
I know it.
你得在六周的封闭期内和他们打交道。
You had to deal with them for the six week close.
我知道。
I know it.
其实只有两周。
It's like two weeks.
两周？现在我想起来了。
Two It's all coming back to me now.
这让我感到相当不舒服。
It's it's making me rather uncomfortable.
问题是，我们的许多策略早在GTC上就提前几年公开了。
The the thing is many of our strategies are are presented in in broad daylight at GTC years in advance of when we do it.
两年半前，我介绍了AI工厂的操作系统，叫做Dynamo。
Two and a half years ago, I introduced the operating system of the AI factory, and it's called Dynamo.
Dynamo，你知道的，是西门子制造的一种设备，本质上是把水转化为电能的机器。
Dynamo, as you know, is a piece of instrument, a machine that was created by Siemens to turn essentially water into electricity.
Dynamo推动了上一次工业革命的工厂。
And Dynamo powered the factory of the last industrial revolution.
所以我认为，它是下一个工业革命——即那个工厂的操作系统的完美名称。
So I thought it was the perfect name for the operating system of the next industrial revolution, the factory of that.
在Dynamo内部，核心技术是分布式推理。
And so inside Dynamo, the fundamental technology is disaggregated inference.
杰森，我知道你非常懂技术。
Jason, I I know you're you're you're super technical.
当然。
Absolutely.
我知道。
I know it.
我来让你来解释这个。
I'll let you take this one.
请为观众详细说明一下。
Go ahead and divide it for the audience.
我不想打断你。
I don't wanna step on you.
是的。
Yeah.
谢谢。
Thank you.
我知道你刚才想插一句话。
I I I knew you wanted to jump in there for a second.
这就是分散推理，意味着推理的处理流程极其复杂。
It's it's disaggregated inference, which means the the pipeline, the processing pipeline of inference is extremely complicated.
事实上，它是当今最复杂的计算问题。
In fact, it is the most complicated computing problem today.
规模惊人，涉及各种各样的数学运算。
Incredible scale, lots of mathematics of different shapes and sizes.
我们提出了一个想法，就是将处理过程进行分解，让其中一部分在某些GPU上运行，其余部分在其他GPU上运行。
And we came to came up with the idea that you would change, you would you would disaggregate parts of the processing such that some of it can run on some GPUs, rest of it can run on different GPUs.
这让我们意识到，甚至分布式计算也可能有意义，我们可以拥有不同异构的计算方式。
And that led to us realizing that maybe even disaggregated computing could make sense, that we could have different heterogeneous nature of computing.
同样的理念促使我们提出了melonize的想法。
That same sensibility led us to melonize.
是的。
Yep.
你知道，如今英伟达的计算分布在GPU、CPU、交换机、扩展交换机、网络处理器上。
You know, today, NVIDIA's computing is spread across GPUs, CPUs, switches, scale up switches, scale out switches, networking processors.
现在我们要把Groq加入其中，并将合适的负载分配到合适的芯片上。
And now we're gonna add Groq to that, and we're gonna put the right workload on the right chips.
你知道，我们已经从一家GPU公司演变为一家AI工厂公司。
You know, we just really evolved from a GPU company to an AI factory company.
我的最大收获可能就是这一点。
I mean, I think that was probably the biggest takeaway that I had.
你正在看到这种根本性的解耦，我们已经从单一的GPU，发展到如今这种复杂多样的选择组合，未来还会更多。
You're seeing this fundamental disaggregation where we've gone from a GPU, and now you have this complexion of all these different options that will eventually exist.
你们在台上提到的一件事是，我希望那些从事高价值推理的人来听听这个，你们说，你们数据中心25%的空间应该分配给这个Groq LPU。
The thing that you guys said on stage or you said on stage was, I I would like the high value inference people to take a listen to this, and 25% of your data center space, you said, should be allocated to this Groq LPU
我们应该把Groq加入到数据中心约25%的Vera Rubins中。
GPU We should add Groq to about 25% of the Vera Rubins in the g in the data center.
所以你能告诉我们，行业如何看待这种现在基本形成的下一代解离式预填充、解码、解离的模式吗？
So can you tell us about how the industry looks at this idea of now basically creating this next generation form of disaggregated prefilled, decode, disag Mhmm.
人们会如何回应这种模式？
And how people do think will react to it?
是的。
Yeah.
让我们退一步来看。
And take a step back.
在我们引入这一技术时，我们从大语言模型处理转向了智能体处理。
And at the time that we added this, we went from large language model processing to agentic processing.
现在，当你运行一个智能体时，你是在访问工作内存。
Now, when you're running an agent, you're accessing working memory.
你在访问长期记忆。
You're accessing long term memory.
你在使用工具。
You're using tools.
你在对存储造成极大的压力。
You're really beating up on storage really hard.
代理之间正在相互协作。
You have agents working with other agents.
有些代理是大型模型。
Some of the agents are very large models.
有些是较小的模型。
Some of them are smaller models.
有些是扩散模型。
Some of them are diffusion models.
有些是自回归模型。
Some of them are autoregressive models.
因此，这个数据中心里有各种不同类型的模型。
And so there are all kinds of different types of models inside this data center.
我们创建了维拉·鲁宾，以运行这种极其多样的工作负载。
We created Vera Rubin to be able to run this extraordinarily diverse workload.
我的感觉是，我们把原本是一家单机架的公司扩展了。
My sense is and and so we added what used to be a one rack company.
我们现在又增加了四个机架。
We now add four more racks.
对。
Right.
所以，NVIDIA的总可用市场，如果说之前是多少的话，现在可能增长了33%到50%。
So NVIDIA's TAM, if you will, increased from what it whatever it was to probably something, call it, you know, 33%, 50% higher.
这33%或50%的增长中，很大一部分将是存储处理器。
Now part of that 33% or 50%, a lot of it's gonna be storage processors.
它被称为Bluefield。
It's called Bluefield.
其中一部分，我希望大部分会是Groq处理器，还有一部分会是CPU。
Some of it will be a lot of it, I'm hoping, will be Groq processors, and some of it will be CPUs.
而且其中很大一部分将是网络处理器。
And they're all and a lot of it's gonna be networking processors.
所有这些都将运行被称为智能体的AI革命计算机。
And so all of this is gonna be running basically the computer of the AI revolution called agents.
对。
Right.
现代工业的操作系统是什么？
The operating system of of What about modern modern industry.
嵌入式应用呢？
What about embedded applications?
你知道，我女儿家里的泰迪熊想和她说话。
So, you know, my daughter's teddy bear at home wants to talk to her.
里面装的是什么？
What goes in there?
它是定制的ASIC，还是会发展出一套更广泛的TAM，针对边缘不同使用场景开发不同的工具？
Is it a custom ASIC, or does there end up becoming much more kind of a broader set of TAM with developing tools that are maybe different for different use cases at the edge
在嵌入式应用领域呢？
and in an embedded application set?
当我们退一步来看，最大的规模上，我们认为存在三个计算机的问题。
We think that there's three computers problem at the at the largest at the largest scale when you stick take a step back.
其中一个计算机主要用于训练AI模型，开发和创建AI。
There's one computer that's really about training the AI model, developing, creating the AI.
另一个计算机则用于评估它。
Another computer for evaluating it.
根据你所面临问题的类型，比如你环顾四周，到处都是机器人、汽车之类的东西。
Depending on the type of problem you're having, like for example, you look around, there's all kinds of robots and cars and things like that.
你需要在一个模拟物理世界的虚拟健身房中评估这些机器人。
You have to evaluate these robots inside a virtual gym that represents the physical world.
因此，它必须是遵循物理定律的软件。
So it has to be software that obeys the laws of physics.
这是第二台计算机。
And that's a second computer.
我们称之为元宇宙。
We call that omniverse.
第三台计算机是边缘端的机器人计算机。
The third computer is the computer at the edge, the robotics computer.
这台机器人计算机中，其中一台可能是自动驾驶汽车。
That robotics computer, one of them could be self driving car.
另一台是机器人。
Another one's a robot.
还有一台可能是泰迪熊。
Another one could be a teddy bear.
一台用于泰迪熊的小型计算机。
Little tiny one for a teddy bear.
其中最重要的一台是我们正在研发的，它将电信基站转变为人工智能基础设施的一部分。
One of the most important ones is one that we're working on that basically turns the telecommunications base stations into part of the AI infrastructure.
所以现在这是一个价值2万亿美元的产业。
So now all of the it's a $2,000,000,000,000 industry.
所有这些迟早都会演变为人工智能基础设施的延伸。
All of that in time will be transformed into an extension of the AI infrastructure.
因此，无线电、工厂、仓库，等等，都会成为边缘设备。
And so radios radios will become edge devices, factories, warehouses, you name it.
所以这三类基本的计算机，其实并不都需要。
And so so there are three these three basic computers, all of them, you know, aren't gonna be necessary.
詹森，去年我认为你比世界上其他人更早地指出，推理能力将提升一千倍。
Jensen, last last year, I think you were ahead of the the rest of the world in in in saying inference isn't gonna a thousand x.
就在去年。
Just last year.
是的。
Yes.
布拉德，你伤到我的感情了。
Brad, you're hurting my feelings.
会达到一百万倍吗？
Is it is it gonna 1,000,000 x?
会达到十亿倍。
It's gonna 1,000,000,000 x.
是的。
Yeah.
对吧？
Right?
我认为当时人们觉得这有点夸张，因为世界仍然专注于预缩放和训练。
And I think people at the time thought it was pretty hyperbolic because the world was still focused on prescaling, on training.
看看现在。
Here we are.
如今，推理已经爆炸式增长。
Now inference has exploded.
我们现在受限于推理能力。
We're inference constrained.
你宣布了一个我认为处于前沿的推理工厂，它的吞吐量将比下一代工厂高出10倍。
You announced an inference factory that I think is leading edge, that's gonna be 10 x better in terms of throughput to the next factory.
但如果你听一听外界的议论，他们会说你的推理工厂成本将达到500亿美元，而其他替代方案，比如AMD和其他定制ASIC，成本只有250亿到300亿美元，你会失去市场份额。
But yet if you if I listen to what the chatter is out there, it's that your inference factory is gonna cost 40 or 50,000,000,000, and the alternatives, the custom ASICs, AMD, others, are gonna cost 25 to 30,000,000,000, and you're gonna lose share.
所以你能不能跟我们说说？
So why don't you talk to us?
你看到了什么？
What are you seeing?
你怎么看待市场份额？
How do you think about share?
让所有这些公司为高出竞争对手两倍价格的产品买单，这合理吗？
And does it make sense for all these folks to pay something that's a two x premium to what others are marketing?
最重要的观点是，你不应该把工厂的价格和令牌的成本等同起来。
The big takeaway, the big idea is that you should not equate the price of the factory and the price of the tokens, the cost of the tokens.
非常有可能，这500亿美元的工厂实际上能为你生成成本最低的令牌，事实上，我可以证明这一点。
It is very likely that the $50,000,000,000 factory, and in fact, I can prove it that the $50,000,000,000 factory will generate for you the lowest cost tokens.
原因在于我们以非凡的效率生产这些令牌。
And the reason for that is because we produce these tokens at extraordinary efficiency.
十倍的差距，你知道的，就在500亿之间。
10 times, you know, the difference between 50,000,000,000.
事实上，200亿只是土地和外壳的成本。
Now it turns out 20,000,000,000 is just land power and shell.
对吧？
Right?
对。
Right.
除此之外，你本来就还需要存储、网络、CPU、服务器和冷却系统。
And then on top of that, you have storage anyways, networking anyways, you got CPUs anyways, you got servers anyways, you got cooling anyways.
GPU价格是一倍还是半倍的差别，并不是500亿和300亿之间的差距。
The difference between that GPU being one x price or half x price is not between 50,000,000,000 and 30,000,000,000.
对吧？
Right.
选一个你最喜欢的数字，比如说在500亿和400亿之间。
Pick your favorite number, but let's say between 50,000,000,000 and 40,000,000,000.
好的。
Okay.
当这个500亿美元的数据中心实际吞吐量高出10倍时，这个差距所占的百分比并不大。
That is not a large percentage when the $50,000,000,000 data center is actually 10 times the throughput.
对。
Right.
詹森，我
Jensen, I
这就是为什么我说，即使对于大多数芯片来说，如果你跟不上技术的发展和我们前进的步伐，即使芯片是免费的，也不够便宜。
wanna That's the reason why I said that even for most chips, if you can't keep up with the state of the technology and the pace that we're running, even when the chips are free, it's not cheap enough.
是的。
Yeah.
是的。
Yeah.
我可以问一个关于整体战略的问题吗？
Can I can I just ask a general strategy question?
当然可以。
Yep.
你的公司是全球最有价值的公司。
I mean, you're running the most valuable company in the world.
明年这家公司收入将超过3500亿美元，自由现金流达2000亿美元。
This thing is gonna do 350 plus billion of revenue next year, 200,000,000,000 of free cash flow.
它的增长速度如此惊人。
It's compounding at these crazy rates.
你如何决定该做什么？
How do you decide what to do?
你究竟是如何获取信息的？
Like, how do you actually get the information?
现在大家都知道，那些人们本该发给你的邮件。
I mean, it's famous now, these sort of emails that are people are meant to send you.
但你究竟如何做出判断，以直觉去塑造市场，决定在哪里加倍投入，哪里撤退，哪里进入全新领域？
But how do you really decide to get an intuition of how to shape the market, where to really double down, where to maybe pull back, where to actually go into a greenfield?
这些信息是如何传到你这里的？
How how does that information get to you?
你如何做出这些决策？
How do you decide these things?
最终来说，这是CEO的职责。
In a final analysis, that's the job of the CEO.
是的。
Yeah.
我们的工作是定义战略、定义愿景、制定策略。
And our job is to define the strategy, define the vision, define the strategy.
当然，我们会受到公司内杰出的计算机科学家、技术专家以及优秀人才的启发，但我们必须塑造未来。
We're informed, of course, by amazing computer scientists, amazing technologists, great people all over the company, but we have to shape that future.
其中一部分取决于，这件事是否难到难以置信？
Well, part of it has to do with, is this something that's insanely hard to do?
如果这件事做起来不难，我们就应该退避。
If it's not hard to do, we should back away from it.
原因在于，如果这件事很容易做，显而易见。
And the reason for that, if if it's easy to do, obviously.
会有大量竞争对手。
Lots of competitors.
很多竞争对手。
A lot of competitors.
是的。
Yeah.
这是否是以前从未有人做到过、且极其困难的事情？
Is this something that has never been done before that's insanely hard to do?
并且能充分利用我们公司独有的超强能力。
And that somehow taps into the special superpowers of our company.
因此，我必须找到这些因素的交汇点，以达到这个标准。
And so I have to find this confluence of things to that meets the standard.
最后，我们也知道，这将伴随着大量的痛苦和煎熬。
And in the end, we also know that a lot of pain and suffering is gonna go into it.
是的。
Yeah.
没有任何伟大的发明是仅仅因为容易做到、第一次尝试就成功的。
There are no great things that are invented because it was just easy to do and just like first try, here we are.
所以，如果这件事超级困难、前所未有，那么你几乎肯定会经历大量的痛苦和煎熬。
And so if it's super hard to do, nobody's ever done it before, it's very likely that you're gonna have a lot of pain and suffering.
你能并
Can you And
所以你最好享受这个过程。
so you better enjoy it.
你能看看你之前宣布的那三四个长期项目吗？比如太空数据中心，或者你在ADAS和汽车领域正在做的事，又或者是生物技术方面的努力？
So can you can you just look at maybe three or four of the more long tail things you announced and just talk about the long term viability of whether it's the data centers in space or whether it's what you're trying to do with ADAS and autos or, you know, what you're trying to do on the biology side?
给我们介绍一下，你是如何看待这些长期业务的增长曲线如何向上拐点的？
Just give us a sense of, like, how you see some of these curves inflecting upwards in some of these longer tail businesses.
太好了。
Excellent.
物理AI，一个大类别。
Physical AI, large category.
我们相信，正如我刚才提到的，我们有三个计算系统，以及运行在其上的所有软件平台。
We believe, and I just mentioned, we have three computing systems, all the software platforms on top of it.
物理AI作为一个大类别，是科技行业首次有机会进入一个价值5万亿美元、但迄今为止几乎没有任何技术介入的产业。
Physical AI as a large category, it's technology industry's first opportunity to address a $50,000,000,000,000 industry that has largely been, you know, void of technology until now.
因此，我们需要发明所有实现这一目标所需的技术。
And so we need to invent all of the technology necessary to do that.
我觉得这将是一段十年的旅程。
I felt that that was a ten year journey.
我们十年前就开始了。
We started ten years ago.
我们现在正看到它开始加速。
We're seeing it inflecting now.
这对我们来说是一个数十亿美元的业务。
It is a multi billion dollar business for us.
现在每年接近100亿美元。
It's close to $10,000,000,000 a year now.
因此这是一个庞大的业务，并且正在呈指数级增长。
And so it's a big business and it's growing exponentially.
所以这是第一点。
And so that's number one.
我认为在数字生物学领域，我们正处在数字生物学的ChatGPT时刻。
I think in the case of digital biology, I think we are literally near the chat GPT moment of digital biology.
我们即将掌握如何表示基因、蛋白质和细胞。
We're about to understand how to represent genes, proteins, cells.
我们已经知道如何理解化学物质。
We already know how to understand chemicals.
因此，我们能够表示和理解生物学基本构建模块的动态，这距离实现还有两到五年。
And so the ability for us to represent and understand the dynamics of the building blocks of biology, that's a couple, two, three, five years from now.
五年内，我完全相信医疗行业或数字生物学将发生转折。
In five years' time, I completely believe that the health care industry or digital biology is gonna inflect.
所以这些是其中一些非常棒的领域，而且你都能看到它们就在我们身边。
And so these are a couple of the really great ones, and you could see they're all around us.
农业。
Agriculture.
农业。
Agriculture.
现在正在发生转折。
Inflecting now.
毫无疑问。
No question.
是的。
Yeah.
詹森，我想从数据中心聊到桌面端。
Jensen, I wanna take you from the data center to the desktop.
这家公司很大程度上是建立在爱好者、电子游戏玩家以及早期的那些显卡基础上的。
The company was built in large part on hobbyists, video gamers, and and all those graphic cards in the beginning.
你曾在这里，面对大约一万人提到，亲手攻克了OpenClaw，编写了代码，而如今智能体已经成为了巨大的革命。
And you mentioned in front of, I think, 10,000 people here, just clawed OpenClaw, clawed code, and what a revolution agents have become.
特别是那些爱好者，正是我们看到的大量能量和创新突破所聚焦的桌面端。
And specifically, the hobbyists who are really where a lot of energy we see, you know, a lot of the innovation breaks want desktops.
你在这里发布了一款产品。
You announced one here.
我相信是戴尔6800。
I believe it's the Dell 6,800.
这是一台非常强大的工作站，用于运行本地模型，拥有750吉字节的内存。
This is a very powerful workstation to run local models, 750 gigs of RAM.
显然，Mac Studio在全球范围内都售罄了。
Obviously, the the Mac studio sold out everywhere.
在我的公司，我们正在全面转向OpenClaw。
In my company, we're moving to OpenClaw everything.
弗里德伯格刚被爪化了。
Friedberg just got claw pilled.
你也被爪化了。
You got claw pilled.
我明白你对这些着迷。
I understand that you're obsessed with these.
从街头兴起的创建开源代理并在桌面使用开源的运动，对你来说意味着什么？
What is this from the streets movement of creating open source agents and using open source on the desktop mean to you?
太棒了。
So great.
这会走向哪里？
Where is that going?
是的。
Yeah.
太棒了。
So great.
首先，让我们退一步来看。
First of all, let's take a step back.
在过去两年里，我们见证了三个关键的转折点。
In the last two years, we saw basically three inflection points.
第一个是生成式AI。
The first one was generative.
ChatGPT 让人工智能走进了普通大众的视野，引起了我们的关注。
ChatGPT brought AI to the common everybody, to our awareness.
但事实上，这项技术在 GPT 出现前几个月就已经摆在明面上了。
But the fact of the matter is the technology sat in plain sight months before GPT.
直到 ChatGPT 为其提供了用户界面，让它变得易于使用，生成式 AI 才真正开始爆发。
It wasn't until ChatGPT put a user interface around it, made it easy for us to use that generative AI took off.
如今，正如你所知，生成式 AI 不仅生成用于外部使用的令牌，也生成用于内部使用的令牌。
Now generative AI, as you know, generates tokens for internal consumption as well as external consumption.
是的。
Yeah.
内部消耗是思考，这催生了推理能力。
Internal consumption is thinking, which led to reasoning.
O1和O3延续了ChatGPT基于事实信息的浪潮，使AI不仅能回答问题，还能以更切实有用的方式回答问题。
O one and o three continued that wave of chat GBT grounded information, made AI not only answer questions, but answer questions in a more grounded way useful.
我们开始看到OpenAI的收入和经济模式开始发生转折。
We started seeing the revenues and the the economic model of OpenAI start to inflect.
第三个转折点则仅在行业内被我们察觉。
Then the third one was only inside the industry that we saw.
Code Interpreter（代码解释器）。
Claw code.
第一个非常有用的自主代理系统。
The first agentic system that was very useful.
真正革命性的东西。
Really revolutionary stuff.
但Code Interpreter仅对企业提供。
But on but Claw code was only available for enterprises.
大多数外部人士直到OpenClaw出现前，从未听说过Claw code。
Most people outside never saw anything about Claw code until OpenClaw.
OpenClaw让大众真正意识到AI代理能做什么。
OpenClaw basically put into the popular consciousness what an AI agent can do.
是的。
Mhmm.
这就是为什么从文化角度看，OpenClaw如此重要。
That's the reason why OpenClaw is so important from a cultural perspective.
现在，它如此重要的第二个原因是，OpenClaw是开源的，并且它构建了一种计算模型，从根本上重新定义了计算方式。
Now the second second reason why it's so important is that OpenClaw is open, but it formulates, it structures a type of computing model that is basically reinventing computing altogether.
它拥有一个记忆系统。
It has a memory system.
Scratch是一个短期记忆的文件系统。
Scratch is a short term memory file system.
它具备扩展能力。
It has Scales.
它有尺度。
It has it has scales.
是的。
Yeah.
你说的是技能还是尺度？
Did you say skills or scales?
技能。
Skills.
哦，技能。
Oh, skills.
如果你有尺度，是的。
If you have scales Yeah.
理论上。
Theoretically.
是的。
Yeah.
是的。
Yeah.
技能。
Skills.
所以第一个
So the first
首先，它，你知道的，它有资源。
thing first thing, it it, you know, it has resources.
它管理资源。
It it manages resources.
它进行调度。
It's it does scheduling.
是的。
Yep.
对吧？
Right?
它还能处理定时任务。
And it cron jobs.
它可以启动代理程序。
It could it could spawn off agents.
它可以分解任务，并像调度一样解决问题。
It could, you know, it could decompose a task and and cause and solve problems as does scheduling.
它拥有I/O子系统。
It has IO subsystems.
它可以进行输入。
It could, you know, input.
它具有输出功能。
It has output.
它可以连接到WhatsApp。
It connect to WhatsApp.
此外，它还提供一个API，用于运行多种称为技能的应用程序。
And also, it has a API that allows it to run multiple types of applications called skills.
是的
Yeah.
这四个要素从根本上定义了计算机。
These four elements fundamentally define a computer.
是的
Yeah.
那么，我们得到了什么？
And therefore, what do we have?
我们第一次拥有了个人人工智能计算机。
We have a personal artificial intelligence computer for the very first time.
开源的。
Open source.
它是开源的。
It's open source.
它几乎可以在任何地方运行。
It runs literally everywhere.
因此，这现在就是——这基本上是现代计算的蓝图和操作系统。
And so this is now the this is the op this is basically the blueprint, the operating system of modern computing.
是的。
Yeah.
它将真正运行在每一个地方。
And it's gonna run literally everywhere.
当然，我们必须帮助它实现的一件事是：当你拥有自主软件时，必须确保这些自主软件能够访问敏感信息。
Now, of course, one of the things that we had to help it do is whenever you have agentic software, you have to make sure that an agentic software has access to sensitive information.
它能执行代码。
It execute code.
它能与外部通信。
It could communicate externally.
我们必须确保所有这些都受到管控，都具备安全性，并且我们制定的策略要让这些代理同时只拥有这三者中的两项，而不是全部三项。
We have to make sure that all of it has to be governed, all of it has to be secure, and that we have policies that gives these agents two of the three things but not all three things at the same time.
对。
Right.
因此，治理部分的工作，我们贡献给了彼得。
And so the governance part of it, we contributed to Peter.
彼得·斯坦伯格在这里，我们有一大批优秀的工程师与他合作，帮助确保并维护这一系统，以保护我们的隐私和安全。
Peter Steinberger was here, and and so we've got a mound of great engineers working with him to help secure and keep that thing so that it could protect our privacy, protect our security.
詹森，这种范式转变是否使得全国各地通过的那些监管人工智能的立法以及许多拟议的立法变得无效？
Jensen, that paradigm shift makes legislation that has passed around the country to regulate AI and a lot of the proposed legislation effectively moot, doesn't it?
你能简单评论一下，这种范式转变如何迅速使许多人工智能监管模型过时吗？这在当前政治中正成为一个非常热门的话题。
Can you just comment for a second on how quickly the paradigm shift kind of obviates a lot of the models for regulatory oversight of AI, which is becoming a very hot topic in politics right now?
这就是我们与政策制定者打交道时的关键所在——我们必须始终走在他们前面。
Well, this is this is the part that that we just with policymakers, we need to we need to always get in front of them.
布拉德，你在这方面做得非常好。
And, Brad, you do a great job doing this.
我们要走在他们前面，向他们说明技术的现状——它是什么，又不是什么。
We get in front of them and inform them about the state of the technology, what it is, what it is not.
嗯。
Mhmm.
它不是一个生物体。
It is not a biological being.
它不是外星生命。
It is not alien.
它没有意识。
It is not conscious.
它是一种计算机软件。
It is computer software.
是的。
Yeah.
没错。
Exactly.
而且它也不是我们常说的那种‘我们完全不了解’的东西。
And and it is not something that, we say things like we don't understand it at all.
对。
Right.
这不是真的。
It is not true.
我们并不是完全理解它。
We don't understand it all.
我们对这项技术了解很多。
We understand a lot of things about this technology.
所以我认为，我们必须确保继续向政策制定者传达信息，不要让末日论和极端主义影响政策制定者对这项技术的看法和理解。
And and so so I think one, we have to make sure that we continue to inform the policy makers and not affect not allow doomerism and extremism to affect how policy makers think and understand about this technology.
然而，我们仍必须认识到，这项技术发展得非常迅速，不要让政策走在技术前面太快。
However however, we still have to recognize this technology is moving really fast and don't get policy ahead of the technology too quickly.
我们国家面临的风险，也是我们在人工智能方面最大的国家安全担忧，就是其他国家在采用这项技术，而我们却因为对它感到愤怒、恐惧或某种偏执，导致我们的产业和社会无法利用人工智能。
And the risk that we we run as a nation, our greatest source of national security concern with respect to AI is that other countries adopt this technology while we are so angry at it or afraid of it or somehow paranoid of it that our industries, our society don't take advantage of AI.
因此，我主要担心的是人工智能在美国的融合问题。
And so I'm just mostly worried about the fusion of AI here in The United States.
如果你坐在Anthropic公司董事会的座位上，面对整个与国防部之间的流言蜚语，你能详细说说吗？
Can you just double click if you were in the seat in the boardroom of Anthropic over that whole scuttlebutt with the Department of War?
这在一定程度上延续了人们不知该如何看待的想法。
It sort of builds on this idea of people didn't know what to think.
它进一步加剧了人们对AI软件层面的怨恨、恐惧或普遍不信任。
It's sort of added to this layer of either resentment or fear or just general mistrust that people have sometimes at the software levels of AI.
你认为你会给达里奥和他们的团队什么建议，以不同的方式处理，从而改变这种结果和这种认知？
What would you would do you think you would have told Dario and that team to do maybe differently try to change some of this outcome and some of this perception?
关于Anthropic，我首先要说的是，这项技术非常出色。
The first thing that I I would I would say about Anthropic is first of all, the technology is incredible.
非常出色。
Incredible.
我们是Anthropic技术的大量使用者。
We are a large consumer of Anthropic technology.
是的。
Yeah.
非常钦佩他们对安全的专注，非常钦佩他们对安全性的重视。
Really admire their focus on security, really admires their focus on safety.
他们做事的文化，以及他们在技术上的卓越，真的非常出色。
The the the culture by which they went about it, the the technology excellence by which they went about it, really fantastic.
我认为，他们希望提醒人们关注这项技术的能力，这一点也非常棒。
I I would say that that the the desire to warn people about the capability, the technology is is also, really terrific.
我们只需要确保理解这个世界是一个连续谱，提醒是好的，制造恐慌就不太好了。
We just have to make sure that we understand that the world has a spectrum and that that warning is good, scaring is less good.
对。
Right.
因为这项技术对我们来说太重要了。
And because this technology is too important to us.
对。
Right.
我认为预测未来是可以的，但我们需要更加谨慎。
And and I think that it is fine to predict the future, but we need to be a little bit more circumspect.
我们需要保持一点谦逊，因为我们实际上无法完全预测未来。
We need to have a little bit more humility that in fact we can't completely predict the future.
而且，说出那些毫无证据支持的极端、灾难性言论，其危害可能比人们想象的更大。
And the ability and to say things that that are quite extreme, quite catastrophic, that there's no evidence of it happening, could be more damaging than people think.
当然，我们是技术领域的领导者。
And and of course, we are technology leaders.
曾经有一段时间，没人听我们说话。
There were there was a time when nobody listened to us.
是的。
Yeah.
但如今，由于技术在社会结构中如此重要，是一个如此关键的产业，对国家安全如此重要，我们的言辞确实有分量。
But now because technology is so important in the social fabric, such an important industry, so important to national security, our words do matter.
我认为我们必须更加谨慎。
And I think we have to be much more circumspect.
我们必须更加持中。
We have to be more moderate.
我们必须更加平衡。
We have to be more balanced.
展开剩余字幕（还有 480 条）
我们必须更加深思熟虑。
We have to be more for more thoughtful.
嗯，我觉得你应该是合适的人选。
Well, I, you know, I would nominate you.
我认为整个行业必须团结起来。
I think the industry's gotta get together.
人工智能在美国的受欢迎程度为17%。
17% popularity of AI in The United States.
我的意思是，我们看看核能发生了什么。
I mean, we see what happened to nuclear.
对吧？
Right?
我们基本上让整个核工业停摆了，现在中国正在建设100座裂变反应堆，而美国一座都没有。
We basically shut down the entire nuclear industry, and now we have a 100 fission reactors being built in China and zero in The United States.
我们听说了对数据中心的暂停令，所以我认为我们必须在这方面更加积极主动。
We hear about moratoriums on data centers, so I think we have to be a lot more proactive about that.
但我想回到你公司内部正在发生的智能代理爆发现象，以及由此带来的效率和生产力提升。
But but I want to go back to this agentic explosion that you're seeing inside your company, the efficiencies, the productivity gains inside your company.
关于我们是否看到了投资回报，目前有很多争论。
There's a lot of debate whether or not we're seeing ROI.
对吧？
Right?
你和我今年初进入这个领域时，最大的问题是：收入会不会出现？
And you and I entering into this year, the big question was, are the revenues going to show up?
收入会像智能一样规模化增长吗？
Are the revenues gonna scale like intelligence?
然后我们迎来了一个类似奥本海默时刻——Anthropic在二月份实现了560亿美元的月收入。
And then we had this kind of Oppenheimer moment, a $56,000,000,000 month by Anthropic in February.
你展望未来时，是否认为你宣布的未来几年内黑威尔和维拉·鲁宾将带来万亿美元的收入前景是可信的？
Do you think, as you look ahead, you announced a trillion dollar you know, visibility into a trillion dollars of just Blackwell and Vera Rubin over the course of the next couple years.
当你看到Anthropic和OpenAI正在发生这一切时，你认为我们现在是否已经处于这样的曲线之上：收入将像智能一样实现规模化增长？
When you see this happening at Anthropic and OpenAI, do you think we're on that curve now where we're going to see revenues scale in the way that intelligence is scaling?
当你环顾四周时，我会从几个不同的角度来回答这个问题。
When you look around when you I'll answer this a couple of different ways.
当你观察在场的观众时，你会发现Anthropic和OpenAI都有代表在这里。
When you look around this audience, you will see that Anthropic and OpenAI is represented here.
但事实上，这里99%以上的内容都是AI，而不是Anthropic和OpenAI的。
But in fact, every but 99 of everything that is here is all AI and is not Anthropic and OpenAI.
对。
Right.
对。
Right.
原因在于AI非常多元化。
And the reason for that is because AI is very diverse.
我认为，作为一类，第二受欢迎的模型是开源模型。
I would say that the second most popular model as a category is open models.
第一是是的。
Number one is yeah.
开源。
Open open source.
开放的方式。
Open ways.
开源。
Open source.
OpenAI 是第一。
OpenAI is number one.
开源是第二。
Open source is number two.
Anthropic 位居非常遥远的第三，这说明了在场所有人工智能公司的规模差异。
Very distant third is Anthropic, and that tells you something about the scale of all of the AI companies that are here.
所以，认识到这一点很重要。
And so so it's important to recognize recognize that.
让我回来再说几件事。
Let me let me come back and say couple things.
第一，当我们从生成式转向推理时，所需的计算量大约增加了100倍。
One, when we went from generative to reasoning, the amount of computation we needed was about a 100 times.
对。
Right.
当我们从推理转向自主代理时，计算量可能又增加了100倍。
When we went from reasoning to agentic, the computation is probably another 100 times.
现在我们看到，仅仅两年内，计算量就提升了整整一万倍。
Now we're looking at in just two years, computation went up by a fact 10,000 x.
与此同时，人们为信息付费，但大多数人是为工作付费。
Meanwhile, people pay for information, but people mostly pay for work.
对。
Right.
和聊天机器人交谈并获得答案非常好。
Talking to a chatbot and getting an answer is super great.
对。
Right.
帮我做一些研究，简直难以置信。
Helping me do some research, unbelievable.
但要是让我完成工作，我愿意为此付费。
But getting work done, I'll pay for.
确实如此。
Indeed.
这就是我们目前所处的状况。
And so that's where we are.
是的。
Yeah.
代理系统能够完成工作。
Agentic systems get work done.
它们正在帮助我们的软件工程师完成工作。
They're helping our software engineers get work done.
然后你就把这个拿去。
And and so then you take that.
你拥有了十万倍的计算能力。
You got 10,000 x more compute.
此时此刻，你的消耗量可能增加了上百倍。
You get probably, at this point, a 100 x more consumption now.
是的。
Yes.
对。
Yeah.
我们甚至还没有开始扩展。
And we haven't even started scaling yet.
我们绝对已经达到了百万倍的规模。
We are absolutely at a million x.
这确实是。
Which is Yeah.
我觉得现在可以聊聊公司里的人数了，大概有两万到三万人左右？
I think a great place to talk about the number of people have, twenty, thirty thousand at the company, something like that?
我们有43,000名员工。
We have 43,000 employees.
我想说，其中约有38,000人是工程师。
You know, I would say 38,000 are engineers.
我们在这个播客里已经多次讨论过，天哪。
The conversation we've had on the pod a number of times is, oh my god.
看看我们公司里的令牌使用量。
Look at the token usage in our companies.
它正在急剧增长。
It is growing massively.
有些人会问，嘿。
And some people are asking, hey.
我加入一家公司时，能获得多少令牌？
When I join a company, how many tokens do I get?
因为我想成为一名高效的员工。
Because I wanna be an effective employee.
我相信在你两个半小时的主旨演讲中——演讲非常精彩——你提出过，你们在花费
And you postulated, I believe, during your two and a half hour keynote, pretty long keynote, well done, that you were spending
如果做得好，应该会更短一些。
If it was well done, it would be shorter.
是的。
Yeah.
我只是想说明一下。
I just wanna let know.
该说声‘是的’了。
Time to do a Yeah.
他们没时间花一个小时去写它
They didn't have time to write it for an hour
和四十五分钟。
and forty five.
你们都知道，你们都知道，根本没有练习。
You guys so you guys know so you guys know, there is no practice.
是的。
Yeah.
所以这既扣人心弦又令人震撼。
And so it's a gripping and ripping.
撕裂。
Rip.
是的。
Yeah.
是的。
Yeah.
我喜欢这个。
I love it.
所以我想让你知道，我是在演讲的同时写演讲稿的。
So so I just wanna let you know I was writing the speech while I was giving the speech.
好吗？
Okay?
所以永远不知道。
So never know.
是的。
Yeah.
但这是否意味着如果我们回退
But does that mean if we do back
对不起。
I apologize.
稍微算一下数学。
Back a little math.
对。
Yeah.
每位工程师75,000个token左右？
75,000 in tokens for each engineer or something like that?
所以你们现在在NVIDIA上为工程团队花20亿美元在token上吗？
So are you spending in NVIDIA a billion, $2,000,000,000 on tokens for your engineering team right now?
我们正在努力。
We're trying to.
让我给你做一个思想实验。
Let me give you the thought experiment.
假设你有一位软件工程师或AI研究员，你每年支付他们50万美元。
Let's say you have a software engineer or AI researcher and you pay them $500,000 a year.
我们经常这么做。
We do that all the time.
是的。
Yeah.
明白吗？
Okay?
这种情况一直在发生。
This is happening all of the time.
到了年底，我会问这位50万美元的工程师：你花在令牌上的钱有多少？
That 500,000 engineer at the end of the year, I'm gonna ask him how many how much did you spend in tokens?
那个人说花了5000美元，我会气得发疯。
And that person said $5,000, I will go ape something else.
是的。
Yes.
没错。
Right.
如果那个50万美元的工程师一年内没有消耗至少25万美元的token，我会非常担忧。
If that if that $500,000 engineer did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed.
明白吗？
Okay?
这和我们的一位芯片设计师说‘猜猜怎么着？’没什么不同。
And this is no different than one of our chip designers who says, guess what?
我打算只用纸和笔。
I'm just gonna use paper and pencil.
我觉得我不需要任何CAD工具。
I don't think I'm gonna need any CAD tools.
对。
Right.
对。
Right.
这是一种真正的范式转变，开始思考这些顶尖员工。
This is a real paradigm shift to start thinking about these all star employees.
这几乎让我想起我们在NBA学到的东西，当勒布朗·詹姆斯每年花一百万美元来维护自己的健康和身体时。
It almost reminds me of of what we learned in the NBA when LeBron James started spending a million dollars a year just on his health of his body, like, in maintaining it.
没错。
That's right.
他41岁了，还在打球。
Here he is at age 41 still playing.
确实如此，嘿。
It really is, hey.
如果这些是杰出的知识工作者，我们为什么不能赋予他们超人的能力呢？
If these are incredible knowledge workers, why wouldn't we give them superhuman abilities?
完全正确。
That's exactly right.
这该去哪儿？
Where does that go?
如果我们再往前推两三年，这位全明星员工在英伟达的效率和他们能完成的工作会是什么样子？
If we if we extrapolate out two or three years from now, what is the efficiency of that all star at NVIDIA and what they're able to accomplish?
那会是什么样子？
What does even look like?
首先，那些‘天啊，这太难了’的想法已经消失了。
Well, first of all, things that that that, wow, this is too hard.
这种想法已经不存在了。
That thought is gone.
这会花很长时间。
This is gonna take a long time.
这种想法已经不存在了。
That thought is gone.
我们需要很多人。
We're gonna need a lot of people.
那个想法已经消失了。
That thought is gone.
这和上一次工业革命没什么不同。
This is no different than in this in the last industrial revolution.
有人会说，哇，那栋建筑看起来真重。
Somebody goes, boy, that building really looks heavy.
没人会这么说。
Nobody says that.
对。
Right.
没人会说哇。
Nobody wow.
那座山看起来太大了。
That mountain looks too big.
没人会这么说。
Nobody says that.
对。
Right.
所有太大、太重、耗时太长的想法，这些观念都已经消失了。
Everything that's too big, too heavy, takes too long, those thought those ideas are all gone.
你削弱了创造力。
You reduce the creativity.
没错。
That's right.
你能想出什么？
What can you come up with?
正是。
Exactly.
对。
Right.
这意味着现在的问题是，你该如何与这些代理协作？
Which means now the question is, how do you how do you work with these agents?
这只是一种新的计算机编程方式。
Well, it's just a new way of doing computer programming.
在过去，我们编写代码。
In the few in the past, we code.
在未来，我们将撰写想法、架构和规范。
In the future, we're gonna we're gonna write ideas, architectures, specifications.
我们将组织团队。
We're gonna organize teams.
我们将帮助他们定义如何评估好坏的标准。
We're gonna we're gonna help them define how to evaluate the definition of good versus bad.
对。
Right.
当某个结果非常出色时，它会是什么样子？
What's the what does it look like when something is a great outcome?
如何与你一起迭代？
How to iterate with you?
如何进行头脑风暴？
How to brainstorm?
这正是你所追求的。
That's really what you're looking for.
而且我认为每位工程师都将拥有100个代理。
And I'm I think that every engineer is gonna 100, a 100 agents.
回到当前行业面临的PR问题。
Back to the PR problem the industry has right now.
你有像All Hollow的戴维·弗里德伯格这样的高管，他们正通过使用你的技术和人工智能，减少生产的卡路里数量并制造高质量的卡路里。
You have executives, like David Friedberg with All Hollow, who's looking at literally taking through the use of technology, your technology and AI, the number of calories produced and making high quality calories.
你认为能将成本降低的因素是什么，弗里德伯格？
What is the factor you think you can bring the cost down, Friedberg?
这个愿景对你的工作有何影响？
What impact does this vision have for what you're doing?
零样本基因组建模，而且它有效。
Zero shot genomic modeling, and it works.
是的。
Yeah.
然后你就迎来了那个时刻，你会说，天啊，真的，这还是在人们一夜之间就替换了整个企业软件栈之后。
And then you have that moment, and you're like, holy Honestly, like and and and that's after people are replacing entire enterprise software stacks in a night.
我花了九十分钟就做了一件事。
I did something in ninety minutes.
我跟那些人说，我们替换了整个软件栈，还有大量的工作负载。
I was telling the guys about replaced the whole software stack and, like, a whole bunch of workload.
在云端花了九十分钟，运行了这个智能体系统，构建了整个东西，部署了它，我们做到了。
Ninety minutes on Cloud, ran this agentic system, built the whole thing, deployed it, and we got we were
周日晚上。
Sunday night.
周日晚上十点。
On a Sunday night, 10PM.
我11点半就完成了。
I was done at 11:30.
我去睡觉了。
I went to bed.
作为CEO，你替换了什么？
As the CEO, you replaced?
是的。
Yeah.
我管理团队的每个人都必须在周末做类似的练习。
Everyone on my management team had to do a similar exercise over the weekend.
周一我们看到的情况让我觉得，这事儿结束了。
What we saw on Monday, I was like, it's over.
但技术方面、科学方面，我们用AutoResearch在三十分钟内做了一些事，我很赞同你对AutoResearch的看法，它让我们意识到在效率上我们还有多远的路要走。
But the technical stuff, the science stuff, we did something in thirty minutes using autoresearch, and I love your view on autoresearch and what that tells us about how far we still have to go in terms of efficiency.
但通过AutoResearch和一部分数据，我们内部发布了一项成果，大家都惊呼：天哪。
But using autoresearch and a chunk of data, something was published internally that we said, oh my god.
这通常需要七年时间才能完成一篇博士论文。
And that would normally be a PhD thesis that would take seven years.
这将成为我们领域有史以来最受赞誉的博士论文之一，并发表在《科学》期刊上。
That would be one of the most celebrated PhD theses we've ever seen in this field, and it would be in the journal Science.
而这一切是在一台台式电脑上，仅用三十分钟，通过autoresearch和我们刚刚导入的所有数据完成的。
And it was done in thirty minutes on a desktop computer running on autoresearch with all the data we just ingested.
我们在周五得到了结果，当时就想着，嘿，我们来试试吧。
We got it on Friday, we're like, hey, let's try it.
启动后，我们去了GitHub，下载了autoresearch并运行了它。
Try booted up, going to GitHub, downloaded autoresearch, and ran it.
你看到每个人的脸上都露出震惊的表情，而这项技术为我们解锁的潜力，原本需要七年才能实现，现在却在三十分钟内完成了。
And you see everyone's face just go like and then the potential of what this is unlocking for us is like the kind of thing that would take seven years, and it happened in thirty minutes.
我们在基因组学领域正亲身体验这一切，简直难以置信。
And we're experiencing it in genomics, and we're like, this is unbelievable.
所以我认为，这种加速正在以几年前你无法想象的方式，为每个人拓宽了可能性的边界。
So I I think, like, the acceleration is widening the aperture for everyone in a way that, like, you didn't imagine a few years ago.
但回到autoresearch这个点上，你能评论一下吗？这个工具在周末用600行代码就发布了，还能在本地运行，利用各种不同数据集实现如此强大的功能，这告诉我们关于算法和硬件优化的早期阶段，究竟意味着什么？
But just going back to the autoresearch point, can you just comment on what you think about the fact that this thing got published with 600 lines of code in a weekend and the capacity that it has to run locally and achieve what it can achieve with all of these diverse datasets and what that tells us about the early stages we are in terms of optimization on algorithms and hardware.
OpenClaw如此了不起的根本原因，第一点在于它与大型语言模型突破的完美契合。
The fundamental reason why OpenClaw is so incredible, number one, is its its confluence, its timing with the breakthroughs in large language model.
是的。
Yeah.
它的时机简直完美。
Its timing was perfect.
时机无可挑剔。
It was impeccable.
对。
Yeah.
在很多方面，如果不是因为Claude、GPT和ChatGPT已经达到了非常出色的水平，彼得可能根本不会提出这个想法。
Now in a lot of ways, Peter wouldn't have come up with it probably if not for the fact that Claude and GPT and Chad GPT have reached a level that is really very good.
没错。
Right.
它还带来了一种新能力，使这些模型能够使用工具。
It is also a new capability that allows these models to tool use.
我们长期以来开发的工具，比如网页浏览器、Excel表格，以及在芯片设计领域的Synopsys、Cadence、Omniverse、Blender和Autodesk等。
The tools that we've created over time, web browsers and Excel spreadsheets and, you know, in the case of chip design, Synopsys and Cadence and, Omniverse and Blender and Autodesk.
所有这些工具都将继续被使用。
And all of these tools are gonna continue to be used.
有些人认为，企业IT软件行业将被摧毁。
There's some some people say that that the enterprise IT software industry is gonna get destroyed.
让我给你一个不同的观点。
There's it's there's a let me give you the alternative view.
企业软件行业受限于用户数量和席位。
The enterprise software industry is limited by butts and seats.
它即将迎来上百倍的智能代理在这些工具上操作。
It's about to get a 100 times more agents banging on those tools.
这些代理将直接操作SQL。
They're gonna be agents banging on SQL.
它们将会是代理在向向量数据库施压，代理在向Blender施压，代理在向Photoshop施压，原因在于这些工具首先做得非常好。
They're gonna be agents banging on vector databases, agents banging on Blender, agents banging on Photoshop, and the reason for that is because those tools are, first of all, do a very good job.
其次，这些工具是我们与外界沟通的渠道。
Second, those tools are the conduit between us.
最终，当工作完成时，它必须以我能掌控的方式呈现给我。
In the final analysis, when the work is done, it has to be represented back to me in a way that I can control.
对。
Right.
对。
Right.
而我知道如何掌控这些工具。
And I know how to control those tools.
所以我需要所有东西都回到Synopsys中。
And so I need everything to be put back into synopsis.
我希望所有东西都回到Cadence，因为这才是我掌控它的方式。
I want everything to be put put back into cadence because that's how I control it.
这就是我建立真实数据的方式。
That's how I've ground truth.
我问你一个问题，关于开源。
Let me ask you a question about open source.
我们有这些闭源模型。
So we have these closed source models.
它们非常出色。
They're excellent.
是的。
Yeah.
我们还有这些开放权重的模型。
We have these open weight models.
许多中国模型非常棒，简直不可思议。
Many of the Chinese models are incredible, absolutely incredible.
两天前，你可能没看到这个，因为你正在台上忙碌，但有一个训练运行在名为BitTensor的加密项目中发生了。
Two days ago, you may not have seen this because you were busy on stage, but there was a training run that happened in this crypto project called BitTensor.
子网三，他们成功训练了一个40亿参数的LAMA模型，完全分布式，由许多人贡献了多余的计算资源。
Subnet three, they managed to train a 4,000,000,000 parameter LAMA model, totally distributed, with a bunch of people contributing excess compute.
但他们能够以有状态的方式完成训练，我认为这是一项相当惊人的技术成就。
But they were able to do it statefully and manage a training run, which I thought was like a pretty crazy technical accomplishment
是的。
Yeah.
因为这是由一些随机的人完成的，而且是的。
Because it's like random people, and Yeah.
每个人都能获得一把小椅子。
Each person gets a little chair.
我们现代版的Folding@home。
Our our modern version of folding at home.
没错。
Exactly.
是的。
Yeah.
那么，你对开源的最终形态有什么看法？
So what what do you think about the end state of open source?
你是否也认为架构的去中心化和计算资源的去中心化，能够支持开源权重和完全开源的方式，以确保AI对每个人广泛可用？
Do you see this decentralization of architecture as well and decentralization of compute to support open weights and a totally open source approach to making sure AI is broadly available to everyone?
我认为我们从根本上需要既拥有作为专有产品的模型，也拥有作为开源的模型。
I believe we fundamentally need models as a first class product, proprietary product, as well as models as open source.
这两者不是二选一，而是两者都需要。
These two things are not a or b, it's a and b.
这一点毫无疑问。
There's no question about it.
原因在于，模型是一种技术，而不是一种产品。
And the reason for that is because models is a technology, not a product.
模型是一种技术，而不是一种服务。
Model is a technology, not a service.
对于绝大多数用户来说，在横向层面上的通用智能，我真心希望不必自己去微调。
For the vast majority of consumers, the horizontal layer, the general intelligence, I would really, really love not to go fine tune my own.
对。
Right.
对。
Right.
我真的很喜欢使用ChatGPT。
I would really love to keep using ChatGPT.
我喜欢使用云服务。
I love to use Cloud.
我喜欢使用Gemini。
I love to use Gemini.
我喜欢使用x。
I love to use x.
它们都有各自的特点，你知道的，这完全取决于我的心情和我想解决的问题。
And they all have their own personalities, as you know, which just kinda depends on my mood and depends on what problem I'm trying to solve.
你知道，我可能会用x来做，也可能用ChatGPT来做。
You know, I might, you know, do it on x or I might do it on on ChatGPT.
因此，这个行业这一部分正在蓬勃发展。
And so that that segment of the of the industry is thriving.
这将会很棒。
It's gonna be great.
然而，所有这些行业，它们的专业知识和专长必须以一种它们能够掌控的方式被整合和捕捉，而这只能通过开源模型实现。
However, there all these industries, their domain expertise, their specialization has to be channeled, has to be captured in a way that they can control, and that it can only come from open models.
我们正在为开源模型行业做出巨大贡献。
The open model industry, we're contributing tremendously to.
它已经接近前沿。
It is near the frontier.
坦率地说，即使它达到了前沿，我认为作为服务的产品、世界级的产品即模型，仍将继续繁荣。
And quite frankly, even if it reaches the frontier, I think that products as a service, world class products as as a models as a product is gonna continue to thrive.
我们现在投资的每一个初创公司都是先采用开源，然后再转向专有模型。
Every startup we're investing in now is open source first and then going to the proprietary models.
是的。
Yeah.
而美妙之处在于，你有了一个优秀的路由器，从第一天起每天连接它，就能随时访问世界上最好的模型，这为你赢得了时间来降低成本、进行微调和专业化。
And the beautiful thing is because you have a great router, you connect it to by on on first day, every single day, you're gonna have access to the world's best model, and and then it gives you time to cost reduce and fine tune and specialize.
因此，你每次都能获得世界级的能力。
And so you're gonna have world class capabilities out to shoot every single time.
詹森，我可以问个问题吗？
Let Jensen, can I ask a question?
没有人比你更希望美国赢得全球人工智能竞赛。
Nobody wants The US to win the global AI race more than you.
对吧？
Right?
但一年前，拜登时代的扩散规则实际上是遏制人工智能在全球范围内的传播。
But a year ago, the Biden era diffusion rule really was an anti American diffusion of AI around the world.
如今，新政府上任已一年。
So here we are a year into the new administration.
给我们打个分吧。
Give us a grade.
在全球技术扩散方面，我们在哪里？我们向世界传播美国人工智能技术的速度如何？
Where is where are we in terms of global diffusion and the rate at which we're spreading US AI technology around the world?
我们是A级吗？
Are we an a?
我们是B级吗？
Are we a b?
我们是C级吗？
Are we a c?
哪些方面奏效了？
What what's working?
哪些方面没奏效？
What's not working?
首先，特朗普总统希望美国产业占据领先地位。
Well, first of all, president Trump wants American industry to lead.
他希望美国科技产业占据领先地位。
He wants American technology industry to lead.
他希望美国科技行业获胜。
He wants American technology industry to win.
他希望我们把美国技术推广到世界各地。
He wants us to spread American technology around the world.
他希望美国成为世界上最富有的国家。
He wants United States to be the wealthiest country in the world.
他希望这一切都能实现。
He wants all of that.
就在我们说话的此刻，英伟达在世界第二大市场放弃了95%的市场份额，而我们目前为0%。
At the current moment as we speak, NVIDIA gave up a 95% market share in the second largest market in the world, and we're at 0%.
特朗普总统。
President Trump.
没错。
That's right.
特朗普总统希望我们重新进入市场，首先要做的是为我们将要销售的公司获得许可。
President Trump wants us to get back in there, and and, the first thing is, to get license licensed for the companies that we're gonna be able to sell to.
我们有许多公司已经申请了许可证。
We've got many companies who have requested for licenses.
我们已经为它们提交了许可证申请，并获得了哈特·卢德尼克部长的批准。
We've applied licenses for them, and we've got approved licenses from secretary Hut Ludnick.
现在，我们已经通知了中国公司，其中许多已经向我们下了采购订单。
Now, we've we've informed the Chinese companies, and many of them have given us purchase orders.
因此，我们正着手重新启动供应链，准备发货。
And so we're gonna we're gonna we're in the process of cranking up our supply chain again to go ship.
我认为，从最高层面来看，布拉德，我们应该承认这一点。
I think at the highest level, Brad, I think one of the things that we should acknowledge is this.
当我们无法获取微型电机和稀土矿物时，我们的国家安全就会受到削弱。
Our national security is diminished when we don't have access to miniature motors, rare earth minerals.
当我们无法掌控电信网络时，我们的国家安全也会受到削弱。
It's diminished when we don't control our telecommunications networks.
当我们无法为本国提供可持续能源时，我们的国家安全同样会受到削弱。
It's diminished when we can't provide for sustainable energy for our country.
这是根本性的削弱。
It is fundamentally diminished.
每一个这样的产业都是我不希望人工智能行业所走向的例子
Every single one of these industries is an example of what I don't want the AI industry to
那样。
be.
对。
Right.
当我们展望未来，思考我们想要什么时。
When we look forward in time and we say, what do we want?
我们希望美国科技产业、美国人工智能产业引领世界时，会是什么样子？我们都可以承认，人工智能模型不可能是全球统一的。
What is the what does it look like when American technology industry, American AI industry leads the world, we can all acknowledge that there is no way that AI models is one universally.
我们都可以承认，那种结果是毫无意义的。
It is we can all acknowledge that that is an outcome that makes no sense.
然而，我们可以想象，美国的技术栈——从芯片到计算系统再到平台——被全世界广泛使用，各国可以在此基础上开发自己的人工智能，使用公共人工智能或私有人工智能，无论何种方式，都能在自己的社会中构建应用程序。
However, we can all imagine that the American tech stack from chips to computing systems to the platforms are used broadly by the world where they build their own AI, they use public AI, they use private AI, whatever, and they can build their applications in their society.
我希望美国的技术栈能占据全球90%的份额。
I would love that the American tech stack is 90% of the world.
是的。
Yes.
我也希望如此。
I would love that.
如果情况像太阳能、稀土、磁体、电机、电信那样，我认为这对国家安全来说是极糟糕的结果。
The alternative, if it looks like solar, rare earth, magnets, motors, telecommunications, I consider that a very bad outcome for national security.
同意。
Agreed.
对。
Yep.
对。
Yep.
你现在对全球各地的冲突局势监控得如何？
How much are you monitoring the situation with the conflicts around the world right now?
这让你多担心呢，詹森？
And how much does it worry you, Jensen?
所以中国和台湾，以及从中东出来的氦气供应，我知道这对半导体制造构成供应链风险。
So China and Taiwan and then helium availability coming out of The Middle East, I understand can be a supply chain risk to semiconductor manufacturing.
这些情况让你多担心？
How much do these situations worry you?
你们在这些事情上花了多少钱？
How much are you spending on them?
首先，我认为在中东，我们有六千个家庭。
Well, first of all, I think the in Middle East, I have we have 6,000 families there.
是的。
Yeah.
我们在英伟达有很多伊朗人，他们的家人仍然在伊朗。
We have a lot of Iranians, at NVIDIA, and their families are still in Iran.
所以我们那里有很多家庭。
And so so we have we have a lot of families there.
首先，他们非常焦虑。
The first thing is is they're quite anxious.
他们非常担忧，非常害怕。
They're quite concerned, quite scared.
我们一直在惦记着他们。
We're thinking about them all the time.
我们一直在监控并密切关注他们。
We're monitor and keeping an eye on them all the time.
他们得到了我们100%的支持。
They have a 100% of our support.
我被问过好几次，我们是否还在考虑留在以色列？
I've been asked several times, are we still considering, being in Israel?
我们100%留在以色列。
We are 100% in Israel.
我们100%支持那里的家庭。
We are a 100% behind the families there.
我们完全位于中东。
We are a 100% in The Middle East.
我也被问到，鉴于中东当前的局势，这是否是我们认为可以拓展人工智能的领域？
I was also asked, you know, given what's happening in The Middle East, is that an area where we believe that we can expand artificial intelligence to?
我相信我们发动战争是有原因的，我相信战争结束后，中东将比以往更加稳定。
I believe that there there's a reason we went to war, and I believe at the end of the war, Middle East will be more stable than before.
因此，如果我们以前考虑过这一点，那么战后我们更应该考虑。
And so if we were there if we're considering it before, we should absolutely be considering it after.
因此，我对此完全支持。
And so I'm a 100% in on that.
关于台湾，我们必须做三件事。
With respect to with with with with with respect to to Taiwan, we have to do three things.
第一，我们必须尽快重新实现美国的工业化。
One, we have to make sure that we re industrialize The United States as fast as we can.
是的。
Yeah.
无论是芯片制造工厂、计算机制造工厂，还是AI工厂，进展如何？
And whether it's the chip manufacturing plants, the the computer manufacturing plants, or the AI factories How are
这方面进展怎么样？
doing on that?
我们进展得非常顺利。
We're doing excellent.
通过获得战略支持，通过赢得台湾供应链的友谊与支持，我们得以以惊人的速度在亚利桑那州、德克萨斯州和加利福尼亚州建立工厂。
With by by gaining the strategic support, by gaining the friendship of the supply chain of Taiwan, by gaining their friendship, by gaining their support, we were able to build Arizona and Texas, California at incredible rates.
他们确实是我们的战略伙伴。
They're they are genuinely a strategic partner.
我们真的应该支持他们。
We we we really they deserve our support.
他们值得我们的友谊。
They deserve our friendship.
他们值得我们的慷慨，而且他们正在尽一切努力加速为我们的制造进程。
They deserve our, generosity, and they're doing everything they can to accelerate the manufacturing process for us.
所以我认为这是第一点。
And so so I think that's number one.
第二点，我们应该多元化制造供应链。
Number two, we ought to diversify the manufacturing supply chain.
无论是韩国、日本还是欧洲，我们都应该多元化供应链，增强其韧性。
And whether it's South Korea, whether it's it's Japan, it's Europe, we ought to we ought to diversify the supply chain, make it more resilient.
第三点，让我们展现出克制。
And number three, let's be let's let's demonstrate restraint.
在我们减少依赖、提升多样性和韧性的过程中，不要过度施压。
And while we're reducing, increasing our diversity and resilience, let's not press, push
没有必要。
Unnecessarily.
没必要。
Unnecessary.
我们需要有耐心。
We need to be patient.
是的。
It's Yeah.
深思熟虑。
Thoughtful.
氦气是个问题吗？
Is helium a problem?
很多报告都提到
A lot of reports
你知道，我觉得氦气可能是个问题，但供应链很可能已经有很多缓冲空间
You know, I I think helium could be a problem, but it's also the case that the supply chain probably has a lot of buffer in
了。
it.
是的。
Yeah.
这类事情通常都有很多缓冲，但你知道吗？
These kind of things tend to have a lot of buffer, but but you know?
是的。
Yeah.
你在自动驾驶领域取得了巨大进展。
You've made massive progress in self driving.
你发布了一个重大消息。
You made a big announcement.
你增加了许多合作伙伴，包括比亚迪。
You've added many more partners, including BYD.
刚刚还播放了一段视频，显示你驾驶着一辆奔驰汽车，并宣布了与优步的重大合作，你们将从多家制造商那里部署大量车辆上路。
There was just a video of you driving around in a Mercedes and a huge announcement, with Uber that you're gonna have a number of cars on the road from many different manufacturers.
我相信，你的设想是会出现一个类似Android的开源平台，你将在其中扮演重要角色，与数十家汽车厂商合作；而在另一端，可能也会出现类似iOS的封闭平台，由特斯拉或Waymo主导。
Your bet is, I believe, that there's going to be an Android type open source platform that you're gonna play a major part in with dozens of car providers, and then maybe on the other side, there could be an iOS with Tesla or Waymo.
你对此的战略思考是什么？这个棋盘将如何展开？
What's your strategy thinking there and how that chessboard emerges?
因为感觉你拥有非常深厚的积累，某种程度上你在竞争，而在其他方面你又在合作。
Because it feels like you have a a pretty deep stack, and in some ways, you're competing, and in other places, you're collaborative.
是的。
Yeah.
我们需要退一步思考。
It's taking a step back.
我们相信，未来所有移动的载具都将完全或部分实现自动驾驶，这是第一点。
We believe that everything that moves will be autonomous completely or partly someday, number one.
第二点，我们并不想自己制造自动驾驶汽车，而是希望让全球每一家汽车公司都能制造自动驾驶汽车。
Number two, we don't wanna build self driving cars, but we wanna enable every car company in the world to build self driving cars.
因此，我们构建了所有三类计算机：训练计算机、仿真计算机、评估计算机，以及车载计算机。
And so we built all three computers, the training computer, the simulation computer, the valuation evaluation computer, as well as the car computer.
我们开发了世界上最安全的驾驶操作系统。
We developed the world's safest driving operating system.
我们还创造了世界上第一辆具备推理能力的自动驾驶汽车，使其能够将复杂场景分解为它已知如何应对的简单场景，就像人类的推理系统一样。
We also created the world's first reasoning autonomous vehicle so that it could decompose complicated scenarios into simpler scenarios that it knows how to navigate through, just like us, reasoning systems.
因此，这个名为Alpomayo的推理系统使我们取得了惊人的成果。
And so that reasoning system called Alpomayo has enabled us to achieve incredible results.
我们进行这种垂直优化。
We open this we we vertical optimization.
我们进行横向创新，并让每个人自行决定。
We horizontally innovate, and we let everybody decide.
你想要从我们这里购买一台计算机吗？
Do you wanna buy one computer from us?
在埃隆和特斯拉的情况下，他们购买我们的训练计算机。
In the case of Elon and Tesla, they buy our training computers.
他们想要购买我们的训练计算机和模拟计算机吗？
Do they wanna buy our training computer and our simulation computers?
还是你希望我们与你合作，一起完成这三者，甚至将计算机集成到你的汽车中？
Or do you wanna let us, work with us to do all three and even put the car computer in your car?
所以，你知道，我们的态度是希望解决问题。
So we you know, our attitude is we wanna solve the problem.
我们不是解决方案的提供者，无论你如何与我们合作，我们都非常高兴。
We're not the solution provider, and we're delighted however you work with us.
让我接着这个话题说一下，因为我觉得这非常有趣。
Let me build on this question because I think it's like it's so fascinating.
你确实打造了一个平台。
You actually do create this platform.
百花齐放，但确实也有一些公司想回到底层，试图和你竞争一下。
A thousand flowers are blooming, but it's also true that some of those flowers wanna now go back down in the stack and try to compete with you a little bit.
谷歌有TPU。
Google has TPU.
亚马逊有Inferentia和Trainium。
Amazon has Inferentia and Trainium.
每个人都在开发自己的版本，觉得自己能超越NVIDIA，尽管他们自己也往往是NVIDIA的大客户。
You know, everybody's sort of spinning up their own version of, I think I can out NVIDIA NVIDIA, even though they also tend to be huge customers.
是的。
Yeah.
你如何应对这种情况？
How do you navigate that?
是的。
And Yeah.
你认为随着时间推移，这些因素会如何发展，它们在这个愿景的整体格局中扮演什么角色？
What do you think happens over time, and where do those things play in the complexion of this kind of vision?
是的。
Yeah.
非常好。
Really great.
首先，我们是唯一一家人工智能公司。
You know, first of all, we're the only AI company.
我们是一家人工智能公司。
We're an AI company.
我们构建基础模型。
We build foundation models.
我们在许多不同领域都处于前沿。
We're at the frontier in many different domains.
我们构建了每一个层级、每一个栈。
We build every single every single layer, every single stack.
我们是全球唯一一家与所有人工智能公司合作的人工智能公司。
We're the only AI company in the world that works with every AI company in the world.
他们从不向我展示他们在构建什么，而我总是向他们展示我正在构建的全部内容。
They never show me what they're building and I always show them exactly what I'm building.
对。
Right.
是的。
Yeah.
因此，这种信心就来源于此。
And so so the confidence comes from this.
第一，我们非常乐意在最顶尖的技术上展开竞争。
One, we are delighted to compete on what is the best technology.
只要我们能持续快速前进，我相信对于他们来说，继续从英伟达采购仍然是最经济的选择，对此我充满信心。
And to the extent that to the extent that we can continue to run fast, I believe that buying from NVIDIA still is one of the most economic things they could do, and I just have incredible confidence there.
第一。
Number one.
第二，我们是唯一能够在所有云平台上部署的架构，这为我们带来了根本性的优势。
Number two, we're the only architecture that could be in every cloud, and that gives us some fundamental advantages.
我们是唯一一种可以从云平台迁移到本地、汽车或任何地区的架构。
We're the only architecture you could take from a cloud and put into on prem, in the car, in any region.
在太空中。
In space.
没错。
That's right.
在太空中。
In space.
因此，我们市场中有一大部分，约占我们业务的40%。
And so there's a whole whole part of our market, about 40% of our of our business.
大多数人并不了解这一点。
Most people don't realize this.
我们业务的40%，如果你没有CUDA堆栈，不能构建完整的AI工厂，客户就不知道如何与你合作。
40% of our business, unless you have the CUDA stack, unless you can build an entire AI factory, you have the customers don't know what to do with you.
他们并不是想制造芯片。
They're not trying to build chips.
他们也不是想购买芯片。
They're not trying to buy chips.
他们想构建AI基础设施，因此希望你带来完整的解决方案，而我们拥有全套技术。
They're trying to build AI infrastructure, and so they want you to come in with a full stack, and we've got the whole stack.
因此，令人惊讶的是，英伟达正在赢得市场份额。
And so surprisingly, NVIDIA's gaining market share.
如果你看看我们今天的位置，我们正在扩大市场份额。
If you look at where we are today, we're gaining share.
但你觉得这些公司尝试之后会意识到：天哪。
But you think what happens is these guys try and they realize, oh my god.
这太复杂了，然后他们会回来找我们。
It's too much, and then they come back.
这就是份额增长的原因吗？
Is that why the share grows?
我们份额增长有几个原因。
Well, we're gaining share for several reasons.
第一，我们的推进速度加快了，我们帮助人们意识到，关键不在于制造芯片。
One, our velocity has gone we help people realize it's not about building the chip.
而在于构建整个系统。
It's about building the system.
系统就是。
System is.
而这个系统非常难以构建。
And that system's really hard to build.
因此，他们与我们的业务往来正在增加。
And and so their their their business with us is increasing.
以AWS为例，我想他们刚刚宣布，我想是昨天，未来几年将采购一百万颗芯片。
In the case of AWS, I think they just announced, I think it was yesterday, that they're gonna buy a a million chips, in the next couple years.
我的意思是，AWS采购这么多芯片，这还不包括他们之前已经购买的所有芯片。
I mean, that's a lot of chips from from AWS, and that's on top of all the chips they've already bought.
因此，我们非常高兴能与他们合作。
And so we're delighted to do that.
但首先，过去几年我们市场份额的增长，是因为Anthropic现在转向了英伟达。
But number one, we're gaining share this last couple of years because we now have Anthropic coming to NVIDIA.
Meta也转向了英伟达，开源模型的增长势头惊人，而这一切都基于英伟达。
Meta SL is coming to NVIDIA, and the growth of open models is incredible, and that's all on NVIDIA.
因此，我们市场份额的增长得益于模型数量的增加。
And so we're growing in share because of the number of models.
我们市场份额的增长还因为这些公司都位于云之外，它们在企业、行业和边缘计算领域实现区域扩张，而如果只是单纯制造ASIC，要实现整个这一增长板块是非常困难的。
We're also growing in share because outs all of these companies are outside the cloud, and they're growing regionally in enterprise, in industries, at the edge, and that entire segment of growth is, you know, really hard to do if it's just building an ASIC.
布拉德。
Brad.
是的。
Yeah.
与此相关，我不打算深究具体数字，但分析师似乎并不相信。
Related to that, and not to get in the weeds on the numbers, but analysts don't seem to believe.
对吧？
Right?
所以如果你看一下共识预测，你说计算能力可能增长一千倍。
So if you look at the consensus forecast, you said compute could 1,000,000 x.
对吧？
Right?
但他们预测你们明年增长30%，后年增长20%，而到了2029年——本应是爆发性的一年——却只有7%。
And yet they have you growing next year at 30%, the year after that at 20%, and in 2029, which is supposed to be a monster year, at 7%.
对吧？
Right?
所以，如果你用你们的总潜在市场，再套用他们的增长率，就意味着你们的市场份额将急剧下滑。
So if you just if you take your TAM and you apply their growth numbers, it suggests that your share will plummet.
你在未来的订单簿中看到任何能证明这种预测成立的迹象吗？
Do you see anything in your future order book that would make that correct?
是的。
Yeah.
首先，他们根本不理解人工智能的规模和广度。
First of all, they just don't understand the scale and the breadth of AI.
对。
Yes.
是的。
Yeah.
是的。
Yeah.
我认为这是真的。
I think that's true.
大多数人认为人工智能只存在于五大超大规模云服务商中。
Most people think that AI is in the top five hyperscalers.
对。
Right.
没错。
That's right.
围绕着大数定律还存在一种固有观念，你知道，他们必须回到自己的投资银行风险委员会，展示某个模型。
There's also an orthodoxy around these law of large numbers where, you know, they have to go back to their investment banking risk committee and show some model.
他们心里根本无法相信，从5万亿会增长到15万亿。
They're not gonna believe in their minds that 5,000,000,000,000 goes to 15,000,000,000,000.
它
It's
已经无法控制了。
it's out of the bag.
可以达到7。
Can go 7.
到
To
或者他们应该拥有一家10万亿美元的公司。
Or they should have a $10,000,000,000,000 company.
这都只是些自保的表面功夫，我觉得
It's all just CYA stuff that I think
以前从未发生过，所以你不能说它一定会发生。
never happened before, so you can't say it will.
而且因为你必须重新定义你所做的事情。
And and because because you have to redefine what it is that you do.
最近有个人提出了一个观点：NVIDIA 怎么可能在服务器领域比英特尔还大？
There was somebody who made an observation recently that NVIDIA how can you be larger than Intel in servers?
原因在于，整个数据中心的 CPU 市场每年大约是 250 亿美元。
And the reason for that is because the CPU market of the entire data center was about $25,000,000,000 a year.
对。
Right.
对。
Right.
我们每年的营收也是 250 亿美元，就像你们知道的，在我们坐在这里的这段时间里。
We do $25,000,000,000 a year, as you guys know, in a very in the time that we were sitting here.
所以，很明显，这只是一个笑话。
And so, obviously, obviously, that was a joke.
不。
No.
但它确实
It's but
这全是播客里的内容。
it's All in podcast.
这真的没错。
It's really true.
确实如此。
It is.
别担心。
Don't worry.
节目里的所有内容都是粗糙的。
Everything on the show is rough.
天啊。
God.
别担心，
Don't worry,
老兄。
man.
这都是节目的一部分。
It's all in.
你不能，你真的不能。
You can't you can't.
我的意思是，
I mean,
医生，不管怎样。
doc, anyway.
这并不是指导。
It's not guidance.
但不管怎样，重点在于你能做到多大，是的。
But anyhow anyhow, it the the point is how big you can be Yeah.
这取决于你具体做什么。
Depends on what is it that you make.
对。
Right.
英伟达并不是在制造芯片。
NVIDIA is not making chips.
第一，制造芯片已经无法帮助你解决人工智能基础设施问题了。
Number one, making chips does not help you solve the AI infrastructure problem anymore.
这太复杂了。
It's too complicated.
第三，大多数人认为人工智能只局限于他们谈论、听到和看到的那些方面。
Number three, most people think that AI is narrowly in the things that they talk about and hear and see.
对。
Right.
AI非常强大，OpenAI更是令人惊叹。
It's AI is much OpenAI is incredible.
它们将会变得极其庞大。
They're gonna be enormous.
Anthropic也非常出色。
Anthropic is incredible.
它们将会变得极其庞大。
They're gonna be enormous.
但AI的规模将远不止于此。
But AI is going to be much, much bigger than that.
是的。
Yeah.
让我们重新谈谈那一部分。
Tell us readdress that segment.
谈谈太空中的数据中心吧。
Tell us about data centers in space for a second.
关于 Bayt 播客
Bayt 提供中文+原文双语音频和字幕，帮助你打破语言障碍，轻松听懂全球优质播客。