His company provides accurately labeled data to help companies train their AI tools.

他的公司提供精確標註的數據，幫助公司訓練人工智能工具。

And back in 2022, he became the youngest self-made billionaire in the world.

早在 2022 年，他就成為世界上最年輕的白手起家的億萬富翁。

Pretty amazing.

太神奇了

Thanks for having me on.

謝謝你邀請我上節目。

I want to go straight to what we were talking about off camera, which is the idea of where the U.S. is on AI versus China, because you have some very surprising statistics that I think will probably, frankly, freak out some of the viewers.

我想直接談談我們在鏡頭外討論的話題，也就是美國與中國在人工智能方面的差距，因為你有一些非常令人吃驚的數據，坦白說，我認為這些數據可能會嚇到一些觀眾。

So, yeah, first of all, the AI race and the AI war between U.S. and China, I think, is one of the most important issues of today.

所以，是的，首先，我認為中美之間的人工智能競賽和人工智能戰爭是當今最重要的問題之一。

We took out a full page ad on The Washington Post on Tuesday saying that, you know, America must win the AI war.

我們週二在《華盛頓郵報》上刊登了整版廣告，說美國必須贏得人工智能戰爭。

And so this sort of relative race in AI between the U.S. and China is critical.

是以，中美兩國在人工智能領域的相對競爭至關重要。

Today, we released Humanities Last Exam, which is a new evaluation or benchmark of AI models that we produced by getting, you know, math, physics, biology, chemistry professors to provide the hardest questions they could possibly imagine that are relevant to their recent research to really put the test of the models, to give you a sense no model is getting above 10 percent on this test.

今天，我們發佈了 "人文科學最後一考"，這是我們對人工智能模型進行的一項新評估或基準測試，我們請數學、物理、生物、化學教授提供他們所能想到的、與他們近期研究相關的最難的問題，以真正對模型進行測試。

That being said, you know, what we found is that DeepSeek, which is the leading Chinese AI lab, their model is actually the top performing or roughly on par with the best American models, which are a one from.

儘管如此，我們發現，作為中國領先的人工智能實驗室，DeepSeek 的模型實際上是表現最好的，或者說與美國最好的模型大致相當，後者是來自美國的一個模型。

OK, so I think we have been all under the impression that the U.S. was way ahead of China as it relates to AI in large part because we have access to, you know, NVIDIA GPUs and chips and other things that that supposedly the Chinese do not have.

好吧，我想我們一直都有一個印象，那就是美國在人工智能方面遙遙領先於中國，這在很大程度上是因為我們有機會獲得英偉達™（NVIDIA）圖形處理器和芯片以及其他據說中國沒有的東西。

I keep hearing from people all week from people, Chinese AI executives, that they say, well, we're so close.

我一週都在聽人們說，中國的人工智能高管們說，我們已經很接近了。

And by the way, we're doing it with one hand tied behind our back.

順便說一句，我們是在一隻手被綁在背後的情況下完成的。

Our algos are better.

我們的算法更好

We're actually going to figure out how to do this, do it better than the U.S. and in even a more energy efficient way, because we don't need these super powerful chips.

實際上，我們會想出辦法，比美國做得更好，甚至更節能，因為我們不需要這些超級強大的芯片。

Are they happen to be right?

他們是對的嗎？

There's two things happening.

有兩件事正在發生。

First, it is true.

首先，這是事實。

It has been true for a long time that the United States has been ahead.

長期以來，美國一直處於領先地位。

And that's been true for, you know, maybe the past decade.

你知道，也許在過去的十年裡一直如此。

That being said, you know, the very recent event on Christmas Day, you know, about a month ago, DeepSeek released a model, which, by the way, I think is symbolic that the Chinese lab releases, you know, an earth shattering model on Christmas Day when, you know, the rest of us are sort of celebrating the holiday.

話雖如此，你知道，最近在聖誕節那天，也就是大約一個月前，DeepSeek發佈了一個模型，順便說一下，我認為這具有象徵意義，因為中國實驗室在聖誕節那天發佈了一個震撼人心的模型，而我們其他人都在慶祝這個節日。

And and they released it to too much fanfare.

而且，他們還大張旗鼓地發行了它。

And then they followed up with their reasoning model, DeepSeek R1, which is the one that we evaluated as top of the leaderboard.

隨後，他們又推出了他們的推理模型 DeepSeek R1，這也是我們評估為排名第一的模型。

You know, the the the reality is yes and no.

要知道，現實情況是有也有。

So, you know, the Chinese labs, they have more H100s than than people think.

是以，中國實驗室擁有的 H100 比人們想象的要多。

You know, the.

你知道

And these are the highest powered Nvidia chips that they were not supposed to have.

而這些都是他們本不應該擁有的最高性能的 Nvidia 芯片。

Yes.

是的。

My understanding is that is that DeepSeek has about 50,000 H100s, which they can't talk about, obviously, because it is against the export controls that the United States has put in place.

我的理解是，"深海探索 "擁有大約 5 萬架 H100 型飛機，他們顯然不能談論這些，因為這違反了美國實施的出口管制。

And I think it is true that, you know, I think they have more chips than other people expect, but also going to go forward basis, they are going to be limited by the chip controls and the export controls that we have in place.

我認為，他們的芯片數量確實比其他人預期的要多，但在未來的發展中，他們也會受到芯片管制和我們現有的出口管制的限制。

How do you I mean, you work with all you work with everybody.

你是怎麼做到的，我是說，你和所有人一起工作，你和所有人一起工作。

So I don't know if it's fair or unfair, but how do you stack rank these large language models and who ultimately is going to be a winner?

是以，我不知道這是否公平或不公正，但你如何對這些大型語言模型進行排名，最終誰會成為贏家？

Or are they all so close and it gets commoditized?

還是說，它們都很接近，因而被商品化了？

The interesting thing that we see right now.

我們現在看到的有趣現象是

So we actually specialize in this.

是以，我們實際上是這方面的專家。

We've produced our SEAL evaluations, our safety evaluations and alignment labs evaluations, which which measure across many different dimensions and we measure across math capabilities, coding capabilities, multilingual capabilities and reasoning capabilities and many different dimensions, including tool use and agent capabilities.

我們已經進行了 SEAL 評估、安全評估和對齊實驗室評估，這些評估從多個不同的維度進行衡量，我們衡量的維度包括數學能力、編碼能力、多語言能力和推理能力，以及工具使用和代理能力等多個不同的維度。

And what we see is different models are better at different things.

而我們看到的是，不同的機型在不同的方面更勝一籌。

So it's hard to put a clear stack ranking among all the models.

是以，很難在所有機型中給出一個明確的排名。

You know, for example, the open AI models are extremely good at reasoning, but the anthropic models make you really good at code.

比如說，開放人工智能模型非常擅長推理，但人類學模型卻讓你非常擅長編碼。

And sort of there's a diversity of of capabilities of the models.

這些模型的功能多種多樣。

That being said, I think what we're seeing in general is the space is becoming more competitive, not less competitive.

儘管如此，我認為我們看到的總體情況是，這個領域的競爭越來越激烈，而不是越來越少。

I keep hearing from business leaders here that they're all playing around with, you know, open AI or they're playing around with Claude, which is the anthropic model or they're playing around with Gemini, et cetera.

我一直聽到這裡的商業領袖們說，他們都在玩開放式人工智能，或者在玩克勞德，也就是人類學模型，或者在玩雙子座，等等。

And then they're going and using Lama.

然後他們又去利用喇嘛。

They're going to find some open source version to try to get close to what they could approximate these other guys doing because of just the different price points of these things.

由於這些東西的價位不同，他們會尋找一些開放源碼版本，試圖接近其他公司的做法。

Do you think that's the future of this?

你認為這就是未來嗎？

I feel like in a Linux world, there's there's definitely a dimension.

我覺得在 Linux 世界裡，肯定有一個維度。

You know, it comes down to ultimately the level of capabilities and intelligence that are required for your use case.

要知道，這歸根結底取決於您的使用案例所需的功能和智能水準。

I think ultimately what we're going to see is, you know, what we do with all the leading labs, including OpenAI and Google DeepMind and Meta and many others, is continuing to push the frontier and push the boundaries.

我認為，最終我們會看到的是，你知道，我們與所有領先實驗室（包括 OpenAI、谷歌 DeepMind 和 Meta 以及許多其他實驗室）一起所做的事情，就是繼續推動前沿技術的發展，不斷突破界限。

And so how do we leverage data, given that, you know, as a as an industry, we sort of run out of publicly available data?

那麼，我們該如何利用數據，因為，你知道，作為一個行業，我們的公開數據已經用完了？

How do we generate new data to keep pushing the frontiers?

我們如何生成新的數據，以不斷推動前沿技術的發展？

And our belief is that, you know, advanced capabilities are going to enable incredible use cases where where you're going to be willing to pay for those for those increased capabilities.

我們的信念是，你知道，先進的功能將帶來令人難以置信的用例，在這些用例中，你將願意為這些增強的功能付費。

But for the more simplistic use cases, those will probably go more towards open source or or more basic models.

但對於更簡單的用例，可能會更傾向於開源或更基本的模式。

We've been talking all morning about Stargate and the debate happening on Twitter between Sam Altman and Elon Musk about whether they really have 100 billion dollars or 500 billion dollars.

我們一上午都在討論 "星際之門"，以及山姆-奧特曼和埃隆-馬斯克在推特上關於他們到底是擁有 1000 億美元還是 5000 億美元的爭論。

Satya Nadella was sitting in your chair just yesterday saying he's got 80 billion dollars, his money's real.

薩蒂亞-納德拉（Satya Nadella）昨天還坐在你的椅子上，說他有 800 億美元，他的錢是真的。

He took to Twitter.

他在 Twitter 上寫道。

What do you make of all this?

你怎麼看這一切？

You know, all these players, you know, so much is on Twitter anyway.

你知道，所有這些球員，你知道，反正很多東西都在推特上。

So so I'm not sure I have an X, we should say.

所以，我不確定我是否有 X。

Oh, yeah.

哦，是的

X.

X.

But I mean, I think one thing that is very real, regardless of of sort of Stargate specifically as a program, is that the United States is going to need a huge amount of computational capacity, a huge amount of infrastructure.

但我的意思是，無論 "星際之門 "是否是一個具體的項目，我認為有一件事是非常現實的，那就是美國將需要大量的計算能力、大量的基礎設施。

So this was actually in we wrote a letter to the Trump administration to on recommendations on how to ensure that the U.S. stays ahead.

是以，這實際上是我們給特朗普政府寫了一封信，就如何確保美國保持領先地位提出建議。

And one of them was really around infrastructure.

其中之一就是基礎設施。

We need to unleash U.S. energy to enable this AI boom.

我們需要釋放美國的能量，讓人工智能蓬勃發展。

And that's clearly what we're seeing right now, which is, you know, in addition to the Stargate program, many of the major AI companies and major clouds are going to be looking to produce to build a giant data center.

這顯然就是我們現在所看到的，你知道，除了星際之門計劃，許多大型人工智能公司和大型雲計算公司都將尋求建立一個巨型數據中心。

So the reason I asked about the different companies doing this, do you ultimately think we need five, six, seven companies all trying to build frontier models?

所以，我之所以問不同的公司都在做這件事，你最終認為我們需要五、六、七家公司都在嘗試建立前沿模型嗎？

Or I mean, there's been a talk forever that, you know, in a different if Lena Khan hadn't been running the FTC, would have Amazon wanted to buy Anthropic already, for example, or would have Microsoft bought open AI or would have some of these folks.

或者我是說，一直以來都有人在說，如果莉娜-可汗（Lena Khan）沒有掌管聯邦貿易委員會，亞馬遜會不會已經想收購人類學公司（Anthropic），比如，微軟會不會收購開放人工智能公司（open AI），或者這些人中會不會有一些人。

So there wouldn't be as many everybody competing against each other in the same way.

這樣就不會有那麼多的人以同樣的方式互相競爭了。

I don't know.

我不知道。

Maybe you think the competition is great.

也許你認為競爭很激烈。

I just don't know how long long term, how many models there ultimately will be like that.

我只是不知道，從長遠來看，最終會有多少型號是這樣的。

I mean, our view is actually that this is potentially going to be one of the greatest markets or the greatest industries ever.

我的意思是，我們的觀點是，這有可能成為有史以來最偉大的市場或行業之一。

You know, right now, let's say there's between 10 and 20 billion dollars of LLM based revenue.

你知道，現在，假設有 100 億到 200 億美元的法律碩士收入。

And if you believe that we're actually on a track towards superintelligence or AGI, then it stands to reason that that's going to go to a trillion dollars or more revenue.

如果你相信我們正在朝著超級智能或 AGI 的方向發展，那麼這將會帶來一萬億美元或更多的收入。

And so if you're looking at a market that's going to go from, let's say, 10 billion to one trillion over who knows how many years, I tend to believe a fewer number of years.

是以，如果你看到的是一個不知道要經過多少年才能從 100 億增長到 1 萬億的市場，我傾向於相信更短的年限。

I think we're sort of in the two to four range, two to four to get to AGI.

我認為我們的目標是兩到四倍，兩到四倍達到 AGI。

What's your version of AGI?

你的 AGI 版本是什麼？

I think obviously there's many definitions.

我想顯然有很多定義。

You know, the definition I believe in is is are powerful AI systems that are able to use a computer just like you or I could and could use all the tools that a computer could and could basically be a remote worker in the most capable way.

你知道，我所相信的定義是，強大的人工智能系統能夠像你我一樣使用計算機，並能使用計算機所能使用的所有工具，而且基本上能以最得心應手的方式成為遠程工作者。