伊利亚·苏茨克沃对德瓦克什·帕特尔反应的看法，英伟达回应谷歌AI进展，特朗普公布Genesis | Diet TBPN

时间线上周末和昨天一片混乱。

Timeline was in turmoil over the weekend and yesterday.

我们稍微谈到了Nucleus，时间线上的尘埃落定。

We covered a little bit about the Nucleus, dust up on the timeline.

科技和AI领域最大的新闻是Ilya Sutskever和Dwarkash Patel的播客发布了。

The biggest news in tech in AI is that Ilya Sutskever, Dwarkash Patel podcast has dropped.

开场片段非常经典。

The opening clip is is iconic.

这很有趣。

It's very funny.

有点像是麦克风没关的尴尬瞬间。

It's a it's a bit of a hot mic moment.

听听这个。

Listen to that.

这一切都是真实的。

All of this is real.

是吗？

Yeah?

什么意思？

Meaning what?

你不这么认为吗？

Don't you think so?

什么意思？

Meaning what?

就像所有这些人工智能的东西，所有这些。

Like all this AI stuff and all this Yeah.

它已经发生了。

Bay That it's happened.

是不是很科幻？

Like, isn't it straight out of science fiction?

是啊。

Yeah.

另一件疯狂的事情是，这种缓慢起飞的感觉竟然显得如此正常。

Another thing that's crazy is like how normal the slow takeoff feels.

我们投资GDP的1%的想法，就像AI一样。

The idea that we'd be investing 1% of GDP It's like AI.

甚至感觉从未觉得这是个大事。

Hasn't even feel like never felt like a deal.

你知道吗？

You know?

现在感觉就像

Where right now it just feels like

而且事实证明我们适应新事物的速度相当快。

And we get used to things pretty fast turns out.

是啊。

Yeah.

但它也有点抽象，这到底意味着什么？

But also it's kinda like it's abstract, what does it mean?

你在新闻里看到的那些意味着什么，是的。

What it means that you see it in the news Yeah.

某某公司宣布了某某金额。

That such and such company announced such and such dollar amount.

没错。

Right.

这就是你能看到的全部了。

That's that's all you see.

对。

Right.

到目前为止，还没有真正以其他方式感受到它。

It's not really felt in any other way so far.

是啊。

Yeah.

我们真的应该从这里开始吗？

Should we actually begin here?

觉得这是个有趣的讨论。

Think this is an interesting discussion.

当然。

Sure.

这是最棒的播客实习生之一

It's one of the greatest podcast interns

有史以来。

the all time.

从某个角度来看。

Point of view.

太棒了。

So good.

太棒了。

So good.

这将成为新常态。

That's gonna be a new meta.

是的。

Yes.

是的。

Yes.

你没法伪装那种状态。

You you can't you can't fake that.

这太不可思议了。

It's amazing.

而且这很有趣，因为这本质上就是被现场麦克风捕捉到的真实反应。

Also, it's just funny because, it's effectively getting caught on a hot mic.

但我当时是在开玩笑。

But I was joking.

我当时就想，在所有你可能在坐下前被麦克风录到的话里——'哦好吧我们其实已经在录制了'——这句话完全印证了我们对伊利亚·塞茨克里特的所有认知。

I was like, of all the things that you could say on the hot mic before you sit down, oh, Okay, we're actually recording, his is just completely reaffirming everything we know about Ilya Setskrit.

这简直一模一样，好吧。

It's just completely the same like, Okay.

他是个真正的信徒。

He is a true believer.

他并不是坐下来然后说，诺德现金，我们得坐我的私人飞机去。

It's not like he was sitting down and being like, like, Nordcash, we got to go on my private plane.

我刚卖掉了大量二手货。

I just sold so much secondary.

这玩意儿发生的事情太疯狂了。

It's crazy what's going on with this stuff.

就像，如果人们真以为AI这玩意儿能成，我就赚个盆满钵满。

Like, if people really think this AI thing's gonna pan out, I'm making billions of dollars, and I'm I'm cashing out.

我我我不信这一套。

I'm I I don't believe any of this stuff is real.

不。

No.

他不是那样4. 他并不是那样被热麦抓到的。

He wasn't caught on a hot mic like that.

他他他的麦克风漏音时刻简直，哇。

He he his hot mic moment is like, wow.

完全就像科幻小说一样。

It's exactly like science fiction.

一切它

Everything It's

都是真的。

all real.

都是真的。

It's all real.

是啊。

Yeah.

简直太经典了。

Which is just iconic.

泰勒，你速通过程中还有其他收获吗？

Tyler, did you have any other takeaways from your speed run?

你正在以五倍速收听。

You're listening to it at five f.

对吧？

Right?

是的。

Yeah.

他是否戳破了扩张泡沫？

Does he pop the scaling bubble?

他是否对‘结束了吗’持悲观看法？

Does he give a bearish take about Is it over?

有什么观点吗？

Any point?

我不会说他是反扩张派，但他确实提出了一个有趣的观点，大意是说AI公司面临的问题是创意太少

So I wouldn't say he's, like, anti scaling, but he does kind of give this interesting take, which he he basically says that, like, AI companies like, there's too few ideas Mhmm.

相对于现有企业数量和我们所处的规模阶段

For the amount of companies and and for for the the scale that we're at.

嗯。

Mhmm.

你可以将AI的进步视为处于这些不同的时代。

You you can think of AI progress as being in these kind of distinct ages.

对吧？

Right?

所以他说，2012年到2020年是研究的时代，嗯。

So he says, 2012 to 2020 was like the age of research Mhmm.

那时人们尝试了各种不同的想法，而且规模非常小。

Where you're trying all these like different ideas and and the scale of things is very small.

对吧？

Right?

比如，训练最初的AlexNet只用到了两块GPU。

Like, to train the original AlexNet was like two GPUs.

最初的Transformer模型大概用了八块，也许是六十四块，但你知道，GPU的数量非常少。

To do the original transformer was like eight, maybe 64, but like, you know, very small amount of GPUs.

当我们基本确认transformer架构可行后，就进入了规模扩展的时代。

Once we kinda figured out that that transformers work, we entered this age of scaling.

嗯。

Mhmm.

这个阶段基本上是从2020年到2025年。

And that's basically from from 2020 to 2025.

现在我们正处于这样一个节点：是的，你可以继续扩大规模，模型性能会继续提升。

And now we're basically at this point where like, yes, you can keep scaling and models will get better.

但即便规模再扩大100倍——嗯。

But even if you scale a 100 x Mhmm.

我们真的能因此获得超级智能吗？

Like, are we really gonna get super intelligence?

模型在基准测试上的表现确实会更好，没错。

It'll get better on the benchmarks Yep.

而且它们会变得更加实用。

And they'll become more useful.

但他并不认为仅靠单纯的规模扩展就能让我们实现目标。

But it's not like this he doesn't think that just raw scaling alone is basically what's gonna bring us there.

我的意思是，很多人都认同这个观点。

I mean, this has been echoed by a lot of people.

我们仍然需要几种不同的范式

We still need a couple different kind of paradigms

对。

Yeah.

要实现这一点。

For this to work.

OPUS 4.5表现更好的原因不仅仅是因为他们扩大了预训练规模。

The reason that that OPUS 4.5 was better is not just because they scaled pre training.

是的。

Yeah.

这是整体规模的扩大。

It's scaling generally.

规模扩展已经从预训练阶段发展到了现在的强化学习阶段。

The scaling has gone from pre training and now it's RL.

是的。

Yeah.

所以我们基本上需要找到另一种范式。

And so we basically we need to find another paradigm.

而实现这一目标的方法就是进行研究。

And the way you do that is just doing like research.

他谈到SSI本质上就是这样，像是

And and so he talks about SSI as basically being this, like

回归研究。

Return to research.

进行研究。

To research.

没错。

Yeah.

这是些小型训练项目。

It it's small kind of training runs.

尽管，你知道，他们只筹集了30亿，与其他相比规模很小

Even though, you know, they only raise 3,000,000,000, which is like small compared to other

当然。

Sure.

与其他研究相比，是的。

To to other research Yeah.

机构。

Institutions.

事实上他们基本上把所有资源都押在这类...我不确定是否算大胆尝试，但就是这些进行实验的小型训练项目。

The fact that they're basically putting it all on these kind of I mean, I don't know if they're moonshots, but they're these small training runs where they're doing experiments.

没错。

Yep.

然后他们最终会扩大规模。

And then they're gonna scale it up eventually.

是的。

Yep.

但他们不仅仅是通过简单扩大规模和做和别人一样的事情来试图赢得AI竞赛。

But they're not just basically trying to win the AI race by just scaling up and do the same thing as someone else.

对。

Yeah.

没错。

Yeah.

他们试图找到一种新方法来真正改变规模曲线，发现新的规模定律，或者找到可以对抗的新技术。

They're trying to find a new a way to actually bend the scaling curve, find a new scaling law, or find a new technology that they can scale against.

我在想Ilya去年在NeurIPS上的演讲。

I was thinking about Ilya's talk at NeurIPS last year.

他展示了哺乳动物体重与脑容量关系的图表，那是一条相当线性的曲线。

He pulls up this chart of the relationship between the mammal's mass and the brain volume, and it's a pretty linear graph.

就像大象比老鼠大得多，所以它的大脑与身体体积的比例也相应更大。

And so like the elephant is a lot bigger than the mouse, and so it has a proportionally larger brain to its body volume.

这是一条完美的线性曲线。

And it's this perfect linear curve.

我应该试着弄清楚是否能把它发短信发出来。

I should just try and figure it out if I can maybe text it in.

基本上，哺乳动物呈现出这种非常清晰的线性趋势。

So basically, the the mammals have this, like, very clear linear trend.

但非人类灵长类动物在图表上的位置略高一些，表现稍好一点。

But then the nonhuman primates are a little bit higher up on the chart, and they're just doing a little bit better.

而人科动物，也就是真正的人类，则呈现出明显不同的曲线。

But then hominids, the actual humans, have a different there's a very distinctly different curve.

这让我想到，或许我们在思考这个问题时应该看到的就是这个现象。

It was making me think like maybe that's what we're supposed to see when we think about, yeah, this, this.

当我们谈论对数图上的直线时，当我们说当前架构正在发生规模扩展时，我们参照的是哪条线？

When we say like straight lines on log graphs, when we say we are seeing scaling happen with the current architectures, which line are we scaling against?

我们实际上是在沿着人类曲线进行扩展吗？

Are we are we actually scaling on the on the human curve?

还是我们在等待当前缩放定律出现分歧？

Or are we waiting for divergence from that current scaling law?

缩放已经耗尽了所有讨论空间。

Scaling has taken all the air out of the room.

嗯哼。

Mhmm.

对吧？

Right?

基本上，我们有足够的算力去尝试这些不同的想法。

Where, like, basically, we have more than enough compute to try these, like, different ideas Yeah.

但它们全都直接投入到训练下一个大模型上了

But they're just all going straight into to training the next big model

是啊。

Yeah.

采用下一个范式。

Using the next paradigm.

也许它略有不同。

And maybe it's slightly different.

对吧？

Right?

你们有不同方式的强化学习或其他方法，但本质上仍是同一回事。

You have a different way of doing RL or whatever, but it is still fundamentally the same thing.

对吧？

Right?

他提到持续学习可能是更好的途径。

And he talks about maybe continual learning is really the the better approach.

对吧？

Right?

我们处于这种预训练模式的时代太久了，以至于认为AI就像训练完就发布，然后就完成了。

We've been in this era of like having a pre training thing for so long that we think of like AI is like you train this thing and then you release it and it's like done.

而现在的强化学习有点不同，因为

And RL is like a little bit different now because

是的。

Yeah.

有一种后训练的概念，你可以某种程度上整合不同的

There's this idea of post training and you can kind of integrate different

还有事物。

also things.

有趣的是，预训练时你使用了整个互联网，所以无需做任何决定。

Thought the interesting thing was with with pre training, you use the whole Internet, so you don't have to decide anything.

你只是将这个算法应用于所有数据和计算资源，没有任何决策过程。

You're just applying this algorithm to just all the data, all the compute, and there's no decisions.

但到了强化学习阶段，你必须决定：我们要输入这些数学方程，可能不输入其他内容，因为我们实际上是在创造数据，而不仅仅是

But then with RL, you have to decide, okay, we're putting in these math equations and we're maybe not putting in something else because we're actually creating the data and we're and it's not just this

这或许就是为什么我们看到这类表现得非常出色的模型。

This is like this is maybe why we see these kind of like models that are super well.

没错。

Yep.

它们在评估中表现非常出色，但实际应用却不尽如人意

They do super well in evals, but not so much

是啊

Yeah.

有些过拟合和

Some of overfitting and

原因在于我们选择的数据并非正确数据，研究人员可能被奖励机制误导了 对

And the reason is because the data that we choose is not the correct data because researchers are basically being reward hacked maybe Yeah.

只专注于解决基准测试问题

Into like just solving for benchmarks.

听你这么说很有意思，结论是我们需要另一个突破

It's interesting to hear this the conclusion is we need another breakthrough.

但同时共识认为我们肯定会在未来十年内实现那个突破

And then simultaneously consensus be but we're definitely going to get that breakthrough in the next decade.

我觉得这和Mike Newp一直说的很相似

I feel like it echoes a lot of even what Mike Newp has been saying.

我们需要新的想法。

We need new ideas.

是啊，他是

Yeah, He's

几个月来一直在说这个。

been saying this for months.

但预测突破性进展的到来速度要困难得多，相比之下，你可以实际规划出：资本的积累、建设数据中心所需的时间、制造一批GPU并上架运行训练所需的时间。

But it's way harder to predict the rate at which breakthroughs will arrive as opposed to you can actually chart out, Okay, the formation of capital, the time it takes build a data center, how long it takes to to, you know, manufacture a bunch of GPUs, rack them, run the training around.

这比人类想出新的算法要可预测得多。

Like, that's much more predictable than, like, human came up with new algorithm.

这有点随机。

That's sort of random.

他提出这一点作为你看到公司这样做的原因，因为如果你在融资，说'我们要买这个数据中心'来证明融资合理性要容易得多

And he he brings this up as as the reason why you see companies doing this because it's just if you're raising money, it's so much easier to to justify the raise by saying, we're gonna buy this data center

完全同意。

Totally.

然后去做

And do

这次训练运行

this training run.

完全正确

Totally.

成本会完全一样

It's gonna cost exactly as much.

哦，这非常符合正确方式

Oh, It's very under right way.

是的

Yeah.

然后模型就能达到这个水平，我们就可以这样变现

Then the model will be this good, then we can use it to monetize this way.

完全正确

Totally.

如果你只是说，哦，是的。

Where if you're just saying, like, oh, yeah.

我们就是要付钱给一群，比如说，非常聪明的研究人员。是的。

We're gonna just pay a bunch of, like, really smart researchers Yeah.

让他们做一堆研究，然后他们会想出些东西来。

To do a bunch of research, and then they'll they'll figure something out.

没错。

Yep.

这就像，你其实不能... 是的。

That it like, you can't really Yeah.

在某些方面，感觉SSI是为某种小型AI寒冬准备的。嗯。

In some ways, it it feels like SSI is set up for, like, somewhat of a mini AI winter Mhmm.

或者说，至少是在驾驭炒作周期的下行阶段。

Or, like, at least riding the hype cycle down.

是的。

Yeah.

因为这听起来不像他坐在那里说，我们筹集了30亿美元，并将在未来12个月内花光。

Because it doesn't sound like he's sitting there being, we raised 3,000,000,000 and we're spending it in the next twelve months.

就像是

It's like

2.9亿是那个数字。

2.9 was that.

不。

No.

不不不

Not not not

不。

No.

不。

No.

不。

No.

不。

No.

这才是重点。

That's the point.

不是这样的。

It's not.

这就像股权。

It's like equity.

它只是闲置在那里，他显然可以随时撤回。

It's just sitting It's he can you you clearly pull back out.

我打算给每位研究员、所有这些不同团队一些尝试的机会。

I'm gonna give each researcher, all these different teams, like, shots on goal.

不。

No.

我喜欢这个主意。

I love it.

他会持续尝试这些机会，直到显然能够随时再筹集100亿美元，尤其是在他获得关键突破性见解时，他们可以率先将其规模化。

Gonna keep taking those shots until, obviously, he'd be able to raise, like, another $10,000,000,000 whenever he wants, especially if he has like a key breakthrough insight and they can be first to scale that.

我们对谷歌的成功感到欣喜。

We're delighted by Google's success.

他们在人工智能领域取得了重大进展，而我们持续为谷歌提供支持。

They've made great advances in AI and we continue to supply to Google.

英伟达领先行业一代。

NVIDIA is a generation ahead of the industry.

这是唯一能在所有计算场景下运行各种AI模型的平台。

It's the only platform that runs every AI model and does it everywhere computing is done.

与专为特定AI框架或功能设计的ASIC相比，英伟达提供了更高的性能、多功能性和可替代性。

NVIDIA offers greater performance Versatility Versatility fungibility fungibility than ASICs are designed for specific AI frameworks or functions.

发布这样的内容真是疯狂。

That is a crazy thing to post.

疯狂，太疯狂了，居然发布这种东西。

Crazy, crazy, crazy thing to post.

有时候你会看到这样的内容说，

Sometimes you get stuff says,

我不知道各位，但世界上最大的公司发推文来捍卫他们的主要产品，这并不能让人安心。

I don't know, boys, but having the largest company in the world sending tweets to defend their main product is not very reassuring.

我觉得如果用另一种方式表达会好得多。

I feel like this would be so much better delivered.

其实我对这里的文字内容本身并没有太大意见。

I actually don't have that much of a problem with the actual text here.

这段话应该由Jensen在对话场合中用更微妙的方式表达出来。

This should be delivered by Jensen with some nuance in a conversational setting.

但当这段话在早上9点整准时发布时，感觉就完全不同了，明显是预先安排好的，明显是在文档里打出来的。

It just hits a lot different when this is in at exactly 9AM, like, clearly scheduled, clearly typed out in a document.

这感觉像是一份新闻稿，而当它本应是对某个问题的回答时，就显得很奇怪。

It it feels like a press release, which is just an odd odd thing when in when it should be there should be an answer to a question.

聊天室里的Bobby Cosmic说，哦，主流媒体现在才开始报道Gemini三的故事。

Someone, Bobby Cosmic in the chat was saying like, oh, the mainstream media is just now picking up on the Gemini three story.

《华尔街日报》和其他地方的文章都在说，哦，也许谷歌要卷土重来了。

And there's articles in The Wall Street Journal and other places saying like, oh, maybe Google's back.

比如，你知道的，买入谷歌股票。

Like, you know, buy Google.

看起来非常令人兴奋。

Like, it's very exciting.

所以英伟达觉得有必要对此作出回应。

And so NVIDIA feels feels the need to respond to that.

但当它真正是一个回应，而不仅仅是一份新闻稿时，感觉就大不相同了。

But it's a lot different when it's actually a response instead of just like, we're putting out a press release.

谁知道为什么呢？

Like, who knows why?

是啊。

Like Yeah.

而不是像黄仁勋那样说，既然你问了，对脱口秀主持人、新闻主播、播客主持人，或者他在交谈的对象——多拉·卡什，或者可能是在对我们说。

As opposed to, like, Jensen saying, like, well, since you asked, talk show host or news anchor or whoever he's podcast host, whoever he's talking to, Dora Cash, whoever he's talking to, maybe us.

我们很欢迎他来。

We'd love to have him.

我可以问他这个问题。

I can ask him that question.

他可以在这里为自己辩护。

He can defend this here.

时机似乎很关键，因为他们目前正面临巨大压力。

Timing is is seems important because they are coming under a huge amount of pressure right now.

今早《巴伦周刊》上有篇泰·金的文章。

There was an article in Yep.

是的。

Barons this morning by Tae Kim.

没错。

Yep.

这个标题可不是英伟达公关团队希望看到的。

The headline is not what NVIDIA's comms teams would have liked it to be.

英伟达在内部备忘录中回应会计质疑，坚称自己不是安然公司

NVIDIA says it's not Enron in private memo refuting accounting questions.

这种说法太疯狂了

That's a crazy thing to say.

让我来让我来

Let me get let me get

切入报道内容

into the to the coverage.

Tay指出：一系列引人注目的股票抛售和会计违规指控，使英伟达成为关于人工智能及其相关股票价值辩论的中心

So Tay says, a series of prominent stock sales and allegations of accounting irregularities have put Nvidia in the middle of a debate about the value of artificial intelligence and its related stocks.

现在英伟达正在反击

Now Nvidia pushing back.

英伟达投资者关系团队上周末向华尔街分析师发送了一份七页的机密备忘录，这家芯片制造商直接回应了持怀疑态度投资者提出的十余项指控

In a private seven page memo sent by NVIDIA's investor relations team to Wall Street analysts over the weekend, the chipmaker directly addressed a dozen claims made by skeptical investors.

这份采用英伟达标志性绿色字体的备忘录，开篇便回应了迈克尔·伯里上周在社交媒体上批评该公司股票薪酬、股权稀释和股票回购的帖子

NVIDIA's memo, which includes fonts in the company's trademark green color, begins by addressing a social media post from Michael Burry last week, which criticized the company for stock based comp, dilution, and stock buybacks.

巴里在2008年金融危机前做空次级抵押贷款的事迹被电影《大空头》描绘过。

Burry's bet against subprime mortgages before the two thousand and eight financial crisis was depicted in the movie The Big Short, of course.

英伟达自2018年以来回购了910亿股，而非1120亿股。

Nvidia repurchased 91,000,000,000 shares since 2018, not 112,000,000,000.

先生

Mr.

巴里似乎错误地包含了RSUs。

Burry appears to have incorrectly included RSUs.

RSU税收和员工股权授予不应与回购计划的绩效混为一谈.

RSU taxes, employee equity grants should not be conflated with the performance of the repurchase program.

英伟达在备忘录中表示，员工从股价上涨.

Nvidia said in the memo, employees benefiting from a rising share price does not indicate the original equity grants were excessive at the time of issuance.

That makes sense.

That makes sense.

《巴伦周刊》审阅了这份备忘录，该备忘录最初出现在 speaker.

Barron's reviewed the memo, which initially appeared in social media posts over the weekend and confirmed its authenticity.

巴里向《巴伦周刊》表示，他不同意英伟达的回应，并坚持自己的分析。

Burry told Barron's he disagrees with Nvidia's response and stands by his analysis.

他表示将更详细地讨论公司股票薪酬的话题。

He said he would discuss the topic of the company's stock based comp in more details.

当然，巴里现在转战Substack平台了。

Burry is, of course, now over on Substack.

他每年收费380美元。

He's charging $380 a year.

如果你是个永远看空的人，这简直就像提前到来的圣诞节。

And if you are a permabear, I can't this is like Christmas coming early.

英伟达未回应《巴伦周刊》的置评请求。

Nvidia didn't respond to Barron's for a request for comment.

但他们也回应了关于当前情况与历史会计欺诈案（如安然、世通和朗讯）类似的指控，这些案件都涉及供应商融资和特殊目的实体。

But they also responded to claims that the current situation is analogous to historical accounting frauds, Enron, WorldCom, and Lucent, that featured vendor financing and SPVs.

与安然不同，英伟达并未使用特殊目的实体来隐藏债务或虚增收入。

Unlike Enron, Nvidia does not use special purpose entities to hide debt or inflate revenue.

英伟达还回应了关于其客户——大型科技公司——未能正确核算英伟达硬件经济价值的指控。

NVIDIA also addressed allegation that its customers, large technology companies, aren't properly accounting for the economic value of NVIDIA hardware.

其中一些公司采用六年折旧周期来处理GPU，我们此前讨论过这一做法。

Some of the companies use we've talked about this use a six year depreciation schedule for GPUs.

伯里表示他认为芯片的实际使用寿命短于六年，这意味着英伟达客户通过长期分摊高额折旧成本来虚增利润。

Burry said he believes the useful lives of the chips are shorter than six years, meaning NVIDIA's customers are inflating profits by spreading out deep depreciation costs over a long period.

将TPU等同于对英伟达不利的观点堪称最愚蠢的论断——可能比DeepSeek更离谱，因为它完全误解了过去六周的真实情况。

The TPUs equal bad for NVIDIA Take is up there with the dumbest, maybe worse than DeepSeek, as it completely misses what actually happened in the last six weeks.

我会记住动物园里谁是谁，这是我的观点。

And I will remember who is who in the zoo, my view.

第一，AI需求正呈爆发式增长。

One, demand for AI is bananas.

无人能够满足这种需求。

No one can meet demand.

所有人都在增加投入。

Everyone is spending more.

谷歌昨天刚表示，他们必须每六个月将容量翻倍才能跟上需求。

Google said just yesterday they have to double capacity every six months to keep up.

两条规模法则依然有效。

Two scaling laws are intact.

他指的是Gemini三号。

He's referencing Gemini three.

飞轮即将加速运转。

The flywheel is about to speed up.

不知为何，中期曲线派认为这是零和竞争。

Somehow the mid curve crew thinks this is zero sum competition.

这些都没有表明这一点。

None of this suggests that.

如果你认为现在的竞争已经很激烈，那就等着看大型连贯黑井集群能带来什么吧。

If you think the race is hot now, wait until you see what comes out of large coherent black well clusters.

所谓'神机'的所有魔力基本上仍基于Hopper架构。

All the magic from the quote god machines is pretty much still hopper based.

最后，快速提一下，GPU的实际表现往往低于包装盒上标榜的成本与性能参数。

Lastly, a quick GPU t p less than the cost and performance specs on the box aren't what you get in real life.

而且谷歌也将获得丰厚的利润，甚至翻倍。

And Google is going to get a fat margin too, doubled up.

关键在于系统层面的有效令牌与功耗、成本及总体拥有成本的比值。

What matters is system level effective tokens to watt to dollars and TCO.

英伟达GPU之所以拥有更高的FMU（功能模块利用率），是因为它们已深度集成于工作流程中，且生态系统极为庞大。

NVIDIA GPUs have higher FMU because they are they're already embedded in workflows slash the ecosystem is massive.

顺便说，这是个很好的测试。

By the way, this is a good test.

如果你对这个话题有看法，却需要查FMU的含义，或许该筛选更优质的信息源。

If you have an opinion on this topic, but you have to look up FMU, then perhaps curate better source.

什么？

What?

MFU。

M f u.

M f u。

M f u.

你刚才说的是FMU。

You said FMU.

随着Rubin的推出，上述有效token与功耗的差距可能还会进一步扩大。

The above effective token watt gap also likely widens with Rubin.

再加上黄仁勋确实能在供应紧张的市场中实现量产，以及未来的灵活性、多云适配能力、可编程性以适应范式转变，他将持续多年保持GPU供不应求的销售态势。

Add in that Jensen can actually deliver volume in a tight market, plus future flexibility, multi cloud capable, programmable for paradigm shifts, and he'll sell every GPU he makes for years.

谷歌也会如此，因为大家都需要第二供应商，而TPU是一款出色的芯片。

Google will too since everyone wants a second supplier and TPU is a fantastic chip.

但这远非二选一的问题。

But this is as far from either or as it gets.

这种混乱局面唯一的好处是，可能会让谷歌短暂登顶全球市值最高公司的宝座。

The one benefit of this confusion is that it is likely to give Google a brief stint as the world heavyweight champion, the most valuable company.

我估计那些半吊子分析师不出两周就会给他们扣上冠军腰带了。

I would guess the midwits put the strap on them in less than two weeks.

给他们上套？

Put the strap on them?

这是什么意思？

What does that mean?

就像...就像一窝蜂涌入？

Just like like like pile in?

看来他预测人们会过度解读英伟达的看跌观点，同时高估谷歌的机会。

It seems like he's predicting that that people will overplay the Nvidia bear take and Overplay the Google opportunity.

这个机会将导致谷歌成为全球最有价值的公司。

Opportunity, that will result in Google becoming the most valuable company in the world.

他在多个场合使用了‘给他们上套’这个短语

And he uses the phrase put the strap on them in multiple

没错。

Yep.

在不到两周的时间内。

In less than two weeks.

根据《华尔街日报》今日报道，人工智能相关投资占GDP增长的一半。

According to today's Wall Street Journal, AI related investment accounts for half of GDP growth.

逆转趋势可能导致经济衰退。

A reversal would risk recession.

我们承受不起倒退的代价。

We can't afford to go backwards.

这篇文章题为《美国经济如何对AI支出上瘾》。

The article is how The US economy Became Hooked on AI Spending.

唐纳德·J·特朗普总统

President Donald J.

特朗普公布'创世使命'计划以加速AI推动科学发现。

Trump unveils the Genesis Mission to Accelerate AI for Scientific Discovery.

今日特朗普签署行政命令启动'创世使命'，这项国家新计划旨在利用人工智能改变科研方式并加速科学发现进程。

Today, Trump signed an executive order launching the Genesis Mission, a new national effort to use artificial intelligence to transform how scientific research is conducted and accelerate the speed of scientific discovery.

'创世使命'要求能源部长统筹国家实验室资源，将美国最杰出的头脑、最强大的计算机和庞大的科学数据整合为协同研究系统。

The Genesis Mission charges the Secretary of Energy with leveraging our national laboratories to unite America's brightest minds, most powerful computers, and vast scientific data into one cooperative system for research.

该行政令指示能源部创建一个闭环AI实验平台，整合我国世界一流的超级计算机和独特数据集，以生成科学基础模型并驱动机器人实验室。

The order directs the Department of Energy to create a closed loop AI experimentation platform that integrates our nation's world class supercomputers and unique datasets to generate scientific foundation models and power robotic laboratories.

该行政令要求总统科技助理协调这项国家倡议，整合来自联邦政府各部门的数据和基础设施。

The order instructs the assistant to the president for science and technology to coordinate the national initiative and integrate an integration of data and infrastructure from across the federal government.

这里还有一条关于巩固美国AI霸主地位的备注。

There's one more note here on strengthening America's AI dominance.

特朗普继续将美国在AI领域的全球主导地位作为优先事项，以开启人类繁荣、经济竞争力和国家安全的新黄金时代。

Trump continues to prioritize America's global dominance in AI to usher in a new golden age of human flourishing, economic competitiveness and national security.

嗯。

Yeah.

我我非常想了解这里的公私合作机制实际是如何运作的。

I'm I'm very interested to hear how, how the public private partnership actually works here.

曾几何时，基本上所有酷炫技术都出自DARPA，出自美国政府。

There was a time when every basically, cool technology was coming out of DARPA, coming out of the US government.

美国政府实现了登月。

The US government landed on the moon.

自那以后，你知道，我认为科技界的许多人对美国政府监管技术发展失去了信心。

And since then, you know, I I think a lot of people in technology have lost faith in the US government overseeing the development of technology.

甚至学术界，我是说，人们认为，比如，通用人工智能会从一家私营企业诞生。

Even academia, I mean, the the people people think, like, you know, AGI will emerge from a private c corp.

人们相信那里才会做出最出色的工作。

That's where people believe that the best work will be done.

给伊利亚·苏茨克沃这样的顶尖科学家30亿美元，让他放手去干。

Give Ilya Sutskever, give the best scientist $3,000,000,000, let him go cook.

这就是当前的主流观点。

Like, that's the thesis currently.

这在某种程度上感觉像是对那种观点的某种否定。

This feels like somewhat of a rejection of that in some ways.

显然，在很多领域，政府拥有AI资源和科技资源是非常合理的。

There's obviously lots of different places where having AI resources, having science and technology resources within the government make a ton of sense.

但观察这两者之间的接口在哪里会很有趣。

But it'll be interesting to see, like, where are the interfacing points between the two between the two categories.