Name: Part 1：Eric Schmidt 與 李飛飛：超級人工智慧之後的人生 | EP #206 (Part 1: Eric Schmidt and Fei-Fei Li: Human Life After Artificial Superintelligence | EP #206)
Uploaded: 2026-01-22T12:56:11.000Z
Duration: 26 min 33 s

What does superintelligence mean and when is it likely to be here?

什麼是超級智能？它大概什麼時候會出現？

Superintelligence is defined as AI's ability to know from chemistry to biology to sports.

超級智能指的是 AI 能夠從化學、生物學到運動等各方面都瞭若指掌。

It's already super to human in many ways.

在很多方面，它已經超越人類了。

But the collaboration between humans and AI will be the most productive and fruitful way of doing things.

但人與 AI 的協作將是最高產、最有成果的做事方式。

All of the evidence and Fei Fei said this very well is going to be human and computer interaction.

所有證據，以及 Fei Fei 說得很好的一點，都指向人機互動。

I think that to get to real super intelligence, we probably need another algorithmic breakthrough.

我認為，要達到真正的超級智能，我們可能還需要一次演算法突破。

It's so important as we talk about AGI, asi that the most important thing that we keep in mind is.

當我們談論 AGI、ASI 時，最重要的一點是我們要記住。

Ladies and gentlemen, everybody, welcome to Eric Schmidt week.

各位女士、各位先生，大家晚安，歡迎來到 Eric Schmidt 週。

Two episodes this week, both with Eric talking about some of the most important developments.

這週有兩集節目，兩集都是 Eric 談論一些最重要的發展。

The first is with me and Eric and Fei Fei Lee, who's building world models.

第一集是我、Eric 和正在建構世界模型的 Fei Fei Lee。

This is a conversation we had live at FII in Saudi Arabia talking about digital superintelligence.

這是我們在沙烏地阿拉伯 FII 的現場對話，談論數位超級智能。

This is rare because Fei Fei doesn't travel.

這很難得，因為 Fei Fei 不常出門。

It was her first trip to Saudi, and it was an extraordinary conversation.

這是她第一次去沙烏地，而且是一場非凡的對話。

The second episode is between my moonshot mate, Dave Blunden and Eric Schmidt, talking about us versus China and how to avoid critical breakdowns and what might be catastrophes in the future due to accelerating AI.

第二集是我和我的登月計畫夥伴 Dave Blunden 與 Eric Schmidt 的對談，內容是關於我們與中國的對抗，以及如何避免因 AI 加速發展而可能發生的重大崩潰和災難。

All right, get ready for two incredible episodes.

好了，準備好迎接兩集精彩的節目吧。

The first one, myself, Eric and Fei Fei Li on digital superintelligence.

第一集，我和 Eric、Fei Fei Li 一起談論數位超級智能。

We're going to be discussing superintelligence.

What does that mean and what happens when it arrives?

那是什麼意思，又會在什麼時候出現呢？

We've been talking about AI, AGI, now perhaps digital superintelligence or asi.

我們一直在談論 AI、AGI，現在也許是數位超級智能或 ASI。

I want to start with the obvious question, and

it's one that I don't think anybody has it perfect answer for, but

而且這個問題，我覺得沒有人有完美的答案，但是

what does superintelligence mean and when is it likely to be here?

超級智能到底是什麼意思，又可能什麼時候會出現？

So a simple thank you, Peter, and thanks for everybody for being here.

所以，Peter，謝謝你，也謝謝大家今天來。

And obviously thanks to Fei Fei, our very close colleague.

當然也要謝謝 Fei Fei，我們非常親近的同事。

The general accepted definitions of general intelligence is its human level of intelligence.

通用智能的普遍接受定義是達到人類水準的智能。

AGI and human intelligence you can understand because we're all human.

AGI 和人類智能，因為我們都是人類，所以你應該能理解。

You have ideas, you have friends, you have.

Super intelligence is defined as the intelligence equal to the sum of everyone, right?

超級智能的定義是等於所有人總和的智能，對吧？

And there is a belief in our industry that we will get to super intelligence.

在我們的產業裡，大家相信我們會達到超級智能。

There's a group of people who I call the San Francisco Consensus consulting, because it's all.

有一群人，我稱他們為舊金山共識顧問團，因為他們都是。

Maybe it's the weather or the drugs or something, but

也許是天氣、藥物還是什麼的，但

they all basically think that it's within three to four years.

他們基本上都認為會在三到四年內達成。

I personally think it'll be longer than that.

But fundamentally, their argument is that there are compounding effects that we're seeing now which will race us to this much faster than people think.

但根本上，他們的論點是，我們現在看到的複利效應會比大家想的更快地將我們推向這個目標。

And Fei, Fei, I don't think anybody's expected the performance that AI has given us so far.

而 Fei，Fei，我不認為有人預期到 AI 到目前為止帶給我們的表現。

The scaling laws have given us capabilities that are extraordinary.

規模法則賦予了我們非凡的能力。

You know, you're the CEO of a new company, the founder of World Labs, you've been at Stanford working on this.

你知道，你是新公司的 CEO，World Labs 的創辦人，你在史丹佛大學也一直致力於這方面。

How do you think about superintelligence?

Do you discuss superintelligence at all in your work?

你的工作中有討論到超級智能嗎？

對啊，Peter，這是個很棒的問題。

And you know, when Alan Turing dared humanity with the question of can we create thinking machines?

你知道，艾倫·圖靈（Alan Turing）當年大膽地向人類提出了一個問題：我們能不能創造出會思考的機器？

He was thinking about the fundamental question of intelligence.

他當時思考的是智能的根本問題。

So the birth of AI is about intelligence, is about the, the.

所以 AI 的誕生，就是關於智能，關於智能的意義。

The profound general ability of what intelligence means.

So from that point of view, AI is already born as a field that tries to push the boundary of what intelligence mean.

所以從這個角度來看，AI 已經誕生，成為一個試圖拓展智能邊界的領域。

Now, in Fast forward to 75 years after Alan Turing, this phrase superintelligence, is pretty hot in Silicon Valley.

現在，在艾倫·圖靈之後的 75 年，超級智能這個詞在矽谷非常熱門。

And I do agree with Eric that the colloquial definition is what is the capability of AI and computers that's better than any human?

我也同意 Eric 的看法，大家口語上說的超級智能，就是 AI 和電腦的能力是否比任何人類都強？

But I do think we need to be a little careful.

但我認為我們需要稍微謹慎一點。

First of all, some part of today's AI is already better than any human.

首先，現今的 AI 有些部分已經比任何人類都強了。

For example, AI's ability of speaking many different languages, translating between, you know, dozens and dozens of language.

像是 AI 能夠說很多種不同的語言，並且可以在數十種語言之間進行翻譯。

Or AI's ability to calculate things really fast.

或是 AI 能夠非常快速地進行計算。

AI's ability to know from chemistry to biology to sports, know the vast amount of knowledge.

AI 能夠從化學到生物學再到體育，知道大量的知識。

So it's already super to human in many ways.

所以在很多方面，它已經超越人類了。

But it remains a question that, can AI ever be Newton?

但問題仍然是，AI 能夠成為牛頓嗎？

For example, we have all the celestial data of the movement of the stars that we observe today.

舉例來說，我們擁有今天觀測到的所有恆星運動的天體數據。

Give that data to any AI algorithm, it will not be able to deduce Newtonian law of motion.

把這些數據給任何一個 AI 演算法，它都無法推導出牛頓運動定律。

It's the combination of creativity, abstraction.

I do not see today's AI or tomorrow's AI being able to do that yet, Eric.

Eric，我其實看不到今天的 AI 或明天的 AI 能做到這一點。

Fei, Fei, of course, got it right, is to think about if you had all of the knowledge in a computer that existed in 1902, could you invent relativity?

Fei，Fei，當然，他理解對了，就是想想看，如果你在電腦裡擁有 1902 年存在的所有知識，你能不能發明相對論？

So for example, if you look at what is called test time compute where the systems are doing reasoning, they can't take the reasoning that they learned and feed it back into themselves very quickly.

所以舉例來說，如果你看看所謂的測試時間計算，系統正在進行推理，它們無法非常快速地將學到的推理反饋給自己。

Whereas if you're a mathematician, you prove something, you can base your next proof on that.

但如果你是個數學家，你證明瞭某件事，你就可以基於它來進行下一個證明。

It's hard for the systems today, although there are appro.

現今的系統很難做到，雖然有近似的方法。

Approximations, so we don't know where the boundaries are.

近似的方法，所以我們不知道界線在哪裡。

The example that I'd like to use is, let's imagine that we can get computers that can solve everything that we normally can do as humans, except for these amazing set of creativities.

我想舉的例子是，想像一下我們可以讓電腦解決我們平常身為人類能做的所有事情，除了這些驚人的創造力之外。

真正有創意的人是怎麼做到的？

The best examples are that they are experts in one area, they see another area and they have an intuition that the same mechanism will solve a problem of a completely different area There.

最好的例子是，他們在一個領域是專家，看到另一個領域，然後他們直覺認為相同的機制能解決那個完全不同領域的問題。

That's an example of something we have to learn how to do with AI.

這就是我們必須學會如何用 AI 來做的事情的例子。

An alternative would be to simply do it in brute force using reinforcement learning.

另一種方法是單純用強化學習來暴力破解。

The problem is that combinatorially the cost of that is insane.

問題是，從組合學的角度來看，成本高到離譜。

And we're already running out of electricity and so forth.

So I think that to get to real superintelligence, we probably need another algorithmic breakthrough.

所以我認為，要達到真正的超級智能，我們可能需要另一個演算法突破。

另一種處理、處理這個的方法。

The technical, the technical term is called non stationarity of objectives.

技術上，技術上的術語叫做目標的非平穩性。

What's happening is the systems are trained against objectives.

發生的是系統是針對目標進行訓練的。

But to do this kind of creativity that Fei Fei is talking about, you need to be able to change the objectives as you're doing them.

但要做到 Fei Fei 談到的那種創造力，你需要在進行的過程中能夠改變目標。

We've seen this past year, I think GPT5Pro reach an IQ of like 148, which is extraordinary.

我認為我們在過去一年看到 GPT5Pro 達到了 148 的智商，這是非凡的。

And of course there is no ceiling on, on this.

I mean it, it loses meaning at some point.

我的意思是，在某個點上它就失去意義了。

But the ability for every human on the planet to have an Einstein level, not in the creativity side, but

但地球上的每個人都能擁有愛因斯坦級別的能力，不是在創造力方面，而是

字幕列表影片播放

Part 1：Eric Schmidt 與李飛飛：超級人工智慧之後的人生 | EP #206 (Part 1: Eric Schmidt and Fei-Fei Li: Human Life After Artificial Superintelligence | EP #206)

Part 1：Eric Schmidt 與 李飛飛：超級人工智慧之後的人生 | EP #206 (Part 1: Eric Schmidt and Fei-Fei Li: Human Life After Artificial Superintelligence | EP #206)

Part 1：Eric Schmidt 與李飛飛：超級人工智慧之後的人生 | EP #206 (Part 1: Eric Schmidt and Fei-Fei Li: Human Life After Artificial Superintelligence | EP #206)