How AI Could Save (Not Destroy) Education | Sal Khan | TED

譯者: LI YISI 審譯者: Coco Shen

So anyone who's been paying attention for the last few months

在過去的幾個月裡

有持續關注的人

has been seeing headlines like this,

都會看到這樣的文章，

especially in education.

尤其是在教育領域。

The thesis has been:

它們的觀點通常是這樣的：

students are going to be using ChatGPT and other forms of AI

學生將會使用 ChatGPT 或者其他 AI

to cheat, do their assignments.

來在考試中作弊， 或完成他們的作業；

They’re not going to learn.

或者學生們將不再學習了，

And it’s going to completely undermine education as we know it.

或者AI將徹底地顛覆現有的教育體制。

Now, what I'm going to argue today

而我今天要提出的觀點是：

is not only are there ways to mitigate all of that,

通過設置恰當的限制，做恰當的事情

if we put the right guardrails, we do the right things,

我們可以將以上所述問題的影響

we can mitigate it.

減到最低；

But I think we're at the cusp of using AI

更進一步地，我認為 我們現在正處於這樣的關鍵時刻：

for probably the biggest positive transformation

利用人工智能 推動教育系統進行一場

that education has ever seen.

前所未有的良性變革。

And the way we're going to do that

而我們的方法，

is by giving every student on the planet

就是要為每一個這個星球上的學生

an artificially intelligent but amazing personal tutor.

提供一個人工智能的， 同時也是無比出色的個人輔導員。

And we're going to give every teacher on the planet an amazing,

同時也為每一個這個星球上的老師

artificially intelligent teaching assistant.

提供一個人工智能的， 同時也是無比出色的助教。

And just to appreciate how big of a deal it would be

為了讓各位感受 給每個人一位個人輔導員

to give everyone a personal tutor,

是一件多麼了不起的事情，

I show you this clip

我會給各位看一張圖表：

from Benjamin Bloom’s 1984 2 sigma study,

它來自1984年， Benjamin Bloom的「2σ研究」。

or he called it the “2 sigma problem.”

或者他說的， 「兩個標準差問題」。

The 2 sigma comes from two standard deviation,

2σ就是兩個標準差，

sigma, the symbol for standard deviation.

——σ是標準差的符號。

And he had good data that showed that look, a normal distribution,

他有很好的數據表明，在傳統的教學模式中

that's the one that you see in the traditional bell curve

學習成績的分佈曲線 ——即常態分佈的鍾型曲線——

right in the middle, that's how the world kind of sorts itself out,

會像正中間的那條曲線那樣。

that if you were to give personal 1-to-1 to tutoring for students,

但如果你為學生提供一對一輔導，

then you could actually get a distribution that looks like that right.

你會得到最右端的那條曲線。

It says tutorial 1-to-1 with the asterisks,

即那條“一對一輔導”並標星號、

like, that right distribution,

最右端的曲線，

a two standard-deviation improvement.

而它比傳統教學模式有兩個標準差的提高。

Just to put that in plain language,

用容易理解的方式來說，

that could take your average student and turn them into an exceptional student.

一對一輔導可以讓平均水平的學生

變成優秀的學生；

It can take your below-average student

也可以讓低於平均水平的學生

and turn them into an above-average student.

變成高於平均水平的學生。

Now the reason why he framed it as a problem, was he said,

而Benjamin Bloom 認為這是一個「問題」 的原因是，

well, this is all good,

這當然很好，

but how do you actually scale group instruction this way?

但是你如何擴大這種 一對一教學模式的規模？

How do you actually give it to everyone in an economic way?

如何最有效率地 讓每個人真正受惠於這種教育模式？

What I'm about to show you is I think the first moves towards doing that.

我將要給你們展示的

正是我們向這個目標邁出的第一步。

Obviously, we've been trying to approximate it in some way

十多年來，在可汗學院

at Khan Academy for over a decade now,

我們一直在嘗試朝這個目標前進。

but I think we're at the cusp of accelerating it dramatically.

不過現在正是能夠大幅加快這個進程的時候。

I'm going to show you the early stages of what our AI,

接下來將向你們展示的是

which we call Khanmigo,

我們的AI——Khanmigo的初步進展：

what it can now do

它現在能做什麼，

and maybe a little bit of where it is actually going.

以及它將可能做到什麼。

So this right over here is a traditional exercise

這是你或者你的孩子可能看到的

that you or many of your children might have seen on Khan Academy.

可汗學院的傳統教學頁面。

But what's new is that little bot thing at the right.

而我們新增的東西 是右邊的那個小機器人。

And we'll start by seeing one of the very important safeguards,

首先我們要知道 其中一項重要的保護機制是：

which is the conversation is recorded and viewable by your teacher.

所有的對話都將被記錄， 並且對老師可見。

It’s moderated actually by a second AI.

這個小機器人事實上是由人工智能控制的，

And also it does not tell you the answer.

它不會告訴你問題的答案，

It is not a cheating tool.

也不是一個作弊工具。

When the student says, "Tell me the answer,"

當學生說「告訴我答案」的時候

it says, "I'm your tutor.

它會說：「我是你的老師。

What do you think is the next step for solving the problem?"

你認為解決這個問題的下一步是什麼呢？」

Now, if the student makes a mistake, and this will surprise people

而當學生犯錯的時候，

who think large language models are not good at mathematics,

和人們認為的大型語言模型不擅長數學相反，

notice, not only does it notice the mistake,

Khanmigo不僅可以檢測到錯誤，

it asks the student to explain their reasoning,

而且會要求學生闡明他們的推理過程。

but it's actually doing what I would say,

Khanmigo能做到的事情

not just even an average tutor would do, but an excellent tutor would do.

我認為不只是一個平均水平的老師會做的， 而且是一個出色的老師會做的。

It’s able to divine what is probably the misconception in that student’s mind,

Khanmigo也可以判断出学生头脑中可能的概念謬誤，

that they probably didn’t use the distributive property.

比如他們沒有正確地使用分配律。

Remember, we need to distribute the negative two

「記住，我們需要

to both the nine and the 2m inside of the parentheses.

把-2同時分配給9，以及同在括號內的2m。」

This to me is a very, very, very big deal.

對我來說，Khanmigo的這個回應標誌著 非常重要的一步。

And it's not just in math.

Knamigo的應用不僅在數學科。

This is a computer programming exercise on Khan Academy,

這是可汗學院上的一個電腦編程練習，

where the student needs to make the clouds part.

學生需要把這些雲朵分開。

And so we can see the student starts defining a variable, left X minus minus.

現在這個學生 開始定義變量，讓X變成負的，

It only made the left cloud part.

但它只能讓左邊的雲朵移開。

But then they can ask Khanmigo, what’s going on?

於是學生們就可以問Khanmigo， 怎麼回事？

Why is only the left cloud moving?

為什麼只有左邊的雲朵在動？

And it understands the code.

Khanmigo可以理解這些代碼，

It knows all the context of what the student is doing,

理解這個語境下 學生正在幹什麼；

and it understands that those ellipses are there to draw clouds,

而且我認為最為震撼的，

which I think is kind of mind-blowing.

是Khanmigo可以理解 這些橢圓組成了雲朵。

And it says, "To make the right cloud move as well,

於是Khanmigo說， 「為了讓右邊的雲也移動，

try adding a line of code inside the draw function

你可以試著在繪製函數中加一行代碼，

that increments the right X variable by one pixel in each frame."

使得右變量X在每一幀中增加一個像素。」

Now, this one is maybe even more amazing because we have a lot of math teachers.

這個例子也許更讓人驚嘆，

We've all been trying to teach the world to code,

因為我們有很多數學老師， 但是電腦編程老師卻不多，

but there aren't a lot of computing teachers out there.

即使我們一直非常希望教授編程。

And what you just saw, even when I'm tutoring my kids,

就算當我在指導自己的孩子

when they're learning to code,

學習編程的時候，

I can't help them this well, this fast,

我也沒辦法輔導得這麼好、 回應得這麼快。

this is really going to be a super tutor.

Khanmigo真的會成為一個超級老師。

And it's not just exercises.

不僅僅是功課，

It understands what you're watching.

Khanmigo能夠理解你在看什麼，

It understands the context of your video.

理解影片中的語境。

It can answer the age-old question, “Why do I need to learn this?”

它可以回答這個古老的問題： 「我為什麼要學這些？」

And it asks Socratically, "Well, what do you care about?"

Khanmigo會蘇格拉底式地反問：

「那你對什麼感興趣呢？」

And let's say the student says, "I want to be a professional athlete."

讓我們假設學生回答： 「我想成為一個職業運動員。」

And it says, "Well, learning about the size of cells,

Khanmigo會說：

「那這個關於細胞大小的影片，

which is what this video is,

也許有助於你理解營養的概念，

that could be really useful for understanding nutrition

and how your body works, etc."

或是你的身體如何運作，諸如此類。」

It can answer questions, it can quiz you,

Khanmigo可以提問題，可以測驗你，

it can connect it to other ideas,

也可以與其他更多的想法產生連結。

you can now ask as many questions of a video

你現在可以對一個影片

as you could ever dream of.

想提出多少問題就提出多少問題。

(Applause)

（掌聲）

Another big shortage out there,

可是還有一個缺點。

I remember the high school I went to,

我記得在我的高中，

the student-to-guidance counselor ratio was about 200 or 300 to one.

學生與顧問指導的比例 大概是200或300比1。

A lot of the country, it's worse than that.

很多國家的情況甚至更加糟糕。

We can use Khanmigo to give every student a guidance counselor,

現在，我們可以用Khanmigo 給每個學生一個顧問指導，

academic coach, career coach, life coach,

不論是學業指導、職業指導， 還是生涯指導，

which is exactly what you see right over here.

正如你們看到的這樣。

And we launched this with the GPT-4 launch.

我們和GPT-4同時發佈了Khanmigo，

We have a few thousand people on this.

有幾千人正在測試這個AI。

This isn't a fake demo,

這不是一個後期製作的影片，

this is really it in action.

這是Khanmigo真實運作的過程。

And then there is, you know,

還有一些事情

things that I think it would have been even harder,

我認為就算對人類導師來說

it would have been a little science fiction to do

都是有些困難，

with even a traditional tutor.

甚至有點科幻色彩的事情。

We run an online high school with Arizona State University

我們和亞利桑那州立大學合作

called Khan World School,

運行了一個線上高中： 可汗世界中學。

and we have a student who attends that online school, based in India.

其中一個名叫Saanvi的印度學生

Her name's Saanvi.

參加了這個線上高中，

And she was doing a report on "The Great Gatsby."

而她要為《了不起的蓋茨比》做一份報告。

And when she was reading "The Great Gatsby,"

當她在閱讀《了不起的蓋茨比》的時候，

Jay Gatsby keeps looking at the green light off into the distance.

她發現傑·蓋茨比一直在眺望遠處的那道綠光。

And she's like, "Why does he do that?"

於是她好奇：為什麼他要這麼做？

She did some web searches, and people have obviously studied this

Saanvi在網上進行了一些搜索。

當然，很多人都研究過這道綠光

and commented about the symbolism of that,

並提出了許多可能的象徵意義。

but none of it was really resonating with her.

但是這一切都沒能引起她的共鳴。

And then she realized that she had Khanmigo

然後Saanvi想起她還有Khanmigo，

and that she could talk to Jay Gatsby himself.

因此她可以直接和傑·蓋茨比對話。

And so, "Ah, splendid choice, old sport.

於是Khanmigo回答她： 「啊，老夥計，你做了個正確的選擇。

I am now Jay Gatsby, the enigmatic millionaire

我現在是正是費茲傑羅筆下的

from F. Scott Fitzgerald’s classic.”

蓋茨比， 那個謎一樣的百萬富翁。」

And so, "Why do you keep staring at the green light?"

Saanvi接著問： 「你為什麼一直盯著那道綠光？」

"Ah, the green light, old sport.

「噢，那道綠光，老夥計，

It's a symbol of my dreams and desires, you see.

你看，這道綠光是我夢想和慾望的象徵，

It's situated at the end of Daisy Buchanan's dock

它坐落於黛西·布坎南的碼頭，

across the bay from my mansion.

和我的豪宅隔海相望。

I gaze at it longingly as it represents my yearning for the past

我熱切地凝視這道綠光， 因為它代表著我對過去的懷念，

and my hope to reunite with Daisy, the love of my life."

以及我與黛西——我的摯愛 重逢的渴望。」

And what was cool is, Saanvi then said, “I had this long conversation,”

而最酷的是，Saanvi隨後說， 「我們交談了很久」，

she called him “Mr. Gatsby,”

她稱呼它為「蓋茨比先生」，

and at the end she actually apologized for taking his time,

且居然還在最後非常禮貌地

which I thought was very polite of her.

為佔用了「蓋茨比先生」的時間而道歉。

But you can imagine this unlocks learning literature, learning ...

可以想像，Khanmigo為學習文學、 歷史提供了全新的可能。

You could talk to historical figures.

你可以和歷史人物對話。

We're even probably going to add an activity

我們甚至考慮加入新的機制，

you can talk to like, the Mississippi River.

讓你可以和，比如說， 密西西比河對話。

It brings things to life in ways that really were science fiction

Khanmigo賦予事物生命，以一種

even six months or a year ago.

我們在半年、一年前 都無法想像的方式。

Students can get into debates with the AI.

學生可以和AI辯論。

And we’ve got this here is the student debating

我們已經實現了這一步，

whether we should cancel student debt.

這是學生正在辯論 我們是否應該取消學生貸款。

The student is against canceling student debt,

這個學生反對取消學生貸款，

and we've gotten very clear feedback.

Khanmigo所給出的應對也非常清晰。

We started running it at Khan World School in our lab school that we have,

我們也開始在我們的

Khan Lab School.

可汗實驗室學校運行這套系統。

The students, the high school students especially,

學生們，尤其是高中學生，

they're saying "This is amazing to be able to fine-tune my arguments

他們都在說：「太不可思議了， Khanmigo可以對我的論證進行微調，

without fearing judgment.

而我也不用擔心會被批評。

It makes me that much more confident

它讓我更加有信心

to go into the classroom and really participate."

走進教室，並真正參與到課堂之中。」

And we all know that Socratic dialogue debate is a great way to learn,

我們都知道， 蘇格拉底式的對話是一種很棒的學習方法。

but frankly, it's not out there for most students.

可說實話，不是每個學生都能以這種方式學習。

But now it can be accessible to hopefully everyone.

不過現在，每個人都有機會了。

A lot of the narrative, we saw that in the headlines, has been,

我們看到的新聞頭條都在說，

"It's going to do the writing for kids.

「AI會幫孩子們寫作，

Kids are not going to learn to write."

因此孩子們將不再學習寫作了。」

But we are showing that there's ways that the AI doesn't write for you,

而Khanmigo，正如你將看到的

它不會代你寫作，

it writes with you.

它和你一起寫作。

So this is a little thing,

這像是一個小遊戲，

and my eight year old is addicted to this,

而我八歲的孩子真的非常喜歡它，

and he's not a kid that really liked writing before,

儘管他之前並不是一個 對寫作很有熱情的孩子。

but you can say,

你可以輸入：

“I want to write a horror story,”

「我想要寫一個恐怖故事。」

and it says, "Ooh, a horror story, how spine-tingling and thrilling.

Khanmigo會說：「噢噢，一個驚心動魄、

毛骨悚然的恐怖故事！

Let's dive into the world of eerie shadows and chilling mysteries."

讓我們潛入陰森恐怖的黑暗世界，

探索那些令人不寒而慄的秘密吧。」

And this is an activity where the student will write two sentences,

在這樣的寫作中，學生會寫兩個句子，

and then the AI will write two sentences.

然後AI接著寫兩個句子。

And so they collaborate together on a story.

他們將合作完成一個故事的寫作。

The student writes, "Beatrice was a misunderstood ghost.

學生寫道：「彼特阿麗斯是一個被誤解的鬼魂，

She wanted to make friends but kept scaring them by accident."

她想要和大家交朋友， 卻總是不小心嚇到他們。」

And the AI says, "Poor Beatrice, a lonely spirit yearning for companionship.

AI寫道：「可憐的彼特阿麗斯， 形單影隻的她多麼渴望陪伴！

One day she stumbled upon an old abandoned mansion," etc.

一天，她偶然發現了一座廢棄的老宅。」

等等等等。

I encourage you all to hopefully one day try this.

我真心鼓勵各位來嘗試這個寫作遊戲，

This is surprisingly fun.

它真的非常有趣。

Now to even more directly hit this use case.

這個例子，

And what I'm about to show you,

和之前我展示的一切

everything I showed you so far

都已經是Khanmigo實際功能的一部分。

is actually already part of Khanmigo, and what I’m about to show you,

但是接下來我所展示的功能

we haven't shown to anyone yet, this is a prototype.

我們還沒有對外公開， 這只是一個樣本。

We hope to be able to launch it in the next few months,

我們希望可以在未來的幾個月內 發佈這個功能。

but this is to directly use AI, use generative AI,

我們希望用AI， 用生成式AI，

to not undermine English and language arts

不破壞英語和語言技能，

but to actually enhance it in ways

相反，可以以一種

that we couldn't have even conceived of even a year ago.

我們在一年前無法想像的方式提高它。

This is reading comprehension.

這就是閱讀理解。

The students reading Steve Jobs's famous speech at Stanford.

學生們在閱讀喬布斯 在斯坦福大學的著名演講時，

And then as they get to certain points,

當他們讀到某些段落時，

they can click on that little question.

他們可以點擊這些小問題。

And the AI will then Socratically, almost like an oral exam,

AI就會以一種蘇格拉底的口吻， 像口試一樣地

ask the student about things.

向學生提出問題。

And the AI can highlight parts of the passage.

AI也可以標記某些段落。

Why did the author use that word?

為什麼作者要用那個詞？

What was their intent?

他們的意圖是什麼？

Does it back up their argument?

這段話支持了他們的論述嗎？

They can start to do stuff that once again,

AI可以做到這些事情。

we never had the capability to give everyone a tutor,

於是我們就有能力 給每個學生提供一位老師，

everyone a writing coach to actually dig in to reading at this level.

一位可以幫助學生深入閱讀的寫作指導。

And you could go on the other side of it.

我們也可以走到事情的另一面，

And we have whole work flows that helps them write,

我們有可以幫助學生寫作的工作流程，

helps them be a writing coach, draw an outline.

讓學生成為自己的寫作指導， 比如勾勒出寫作大綱。

But once a student actually constructs a draft,

當學生完成草稿時，

and this is where they're constructing a draft,

——這是他們寫作草稿的地方——

they can ask for feedback once again,

他們可以再次獲得

as you would expect from a good writing coach.

一個出色的寫作指導的反饋。

In this case, the student will say, let's say,

在這個情形下，舉個例子， 學生可以問：

"Does my evidence support my claim?"

「我的證據是否支持我的觀點？」

And then the AI, not only is able to give feedback,

這時AI不僅可以給出反饋，

but it's able to highlight certain parts of the passage and says,

而且可以標記草稿中特定的段落， 並說：

"On this passage, this doesn't quite support your claim,"

「這一段話似乎並不太能夠支持你的觀點。」

but once again, Socratically says, "Can you tell us why?"

再次，蘇格拉底式地發問： 「你可以告訴我為什麼嗎？」

So it's pulling the student, making them a better writer,

所以Khanmigo是在 推動學生成為一個更好的寫作者，

giving them far more feedback

給學生遠多於傳統模式下

than they've ever been able to actually get before.

可以得到的建議。

And we think this is going to dramatically accelerate writing, not hurt it.

我們認為這將極大地提高寫作能力， 而非降低它。

Now, everything I've talked about so far is for the student.

我們說了這麼多關於學生的學習，

But we think this could be equally as powerful for the teacher

但是Khanmigo對老師來說 也是一個出色的工具：

to drive more personalized education and frankly

幫助他們推進個性化教學，

save time and energy for themselves and for their students.

為他們自己和學生節省時間和精力。

So this is an American history exercise on Khan Academy.

這是可汗學院上的一道美國歷史功課，

It's a question about the Spanish-American War.

它和美西戰爭有關。

And at first it's in student mode.

在學生模式下，

And if you say, “Tell me the answer,” it’s not going to tell the answer.

如果你輸入「告訴我答案」， Khanmigo不會給出答案，

It's going to go into tutoring mode.

它會進入指導模式。

But that little toggle which teachers have access to,

而教師可以用那個小開關，

they can turn student mode off and then it goes into teacher mode.

來關閉學生模式，進入教師模式。

And what this does is it turns into --

所以Khanmigo的回答就變成了——

You could view it as a teacher's guide on steroids.

你可以把它看作是升級版的教師指南。

Not only can it explain the answer,

它不僅可以解釋答案，

it can explain how you might want to teach it.

而且可以闡述你可以怎麼組織教學。

It can help prepare the teacher for that material.

Khanmigo會幫助老師準備材料，

It can help them create lesson plans, as you could see doing right there.

製作課程計畫， 正如我所展示的。

It'll eventually help them create progress reports

它可以幫助老師準備進度報告，

and help them, eventually, grade.

並最後，幫他們進行打分。

So once again, teachers spend about half their time

因此，在課程計畫功能的幫助下，

with this type of activity, lesson planning.

教師可以節省一半的時間。

All of that energy can go back to them

而那省出來的精力可以回到他們自己，

or go back to human interactions with their actual students.

或是回到與學生的人與人之間的互動上去。

(Applause)

（掌聲）

So, you know, one point I want to make.

因此我想說的是，

These large language models are so powerful,

現在我們的語言模型是如此強大，

there's a temptation to say like, well,

以至於好像我們可以說，

all these people are just going to slap them onto their websites,

人們只要把要做的事情扔到網站上就可以了。

and it kind of turns the applications themselves into commodities.

這好像把應用程序變成了商品。

And what I've got to tell you

這就是為什麼，

is that’s one of the reasons why I didn’t sleep for two weeks

當我在去年八月使用GPT-4的時候，

when I first had access to GPT-4 back in August.

我有整整兩個星期睡不著覺。

But we quickly realized that to actually make it magical,

但是我們很快意識到， 我們應該把它變得更加迷人。

I think what you saw with Khanmigo a little bit,

我想各位已經在Khanmigo身上看到了這一點：

it didn't interact with you the way that you see ChatGPT interacting.

它和用戶互動的方式和ChatGPT不同。

It was a little bit more magical, it was more Socratic,

Khanmigo更加迷人， 更加像蘇格拉底，

it was clearly much better at math

也比人們想像的

than what most people are used to thinking.

更擅長數學。

And the reason is,

這背後的原因是

there was a lot of work behind the scenes to make that happen.

我們做了大量的工作讓這一切發生。

And I could go through the whole list of everything we've been working on,

在過去的六、七個月裡， 我們做了很多的事，得到很多人的幫助

many, many people for over six, seven months to make it feel magical.

才讓Khanmigo呈現出現在的強大能力。

But perhaps the most intellectually interesting one

但其中最有啓發性的觀點

is we realized, and this was an idea from an OpenAI researcher,

來自於一位OpenAI研究者。

that we could dramatically improve its ability in math

他說如果要大幅提高AI的數學能力

and its ability in tutoring

和輔導能力，

if we allow the AI to think before it speaks.

我們就要讓AI在輸出前進行「思考」。

So if you're tutoring someone

想像一下，當你在輔導別人時，

and you immediately just start talking before you assess their math,

如果你不評估他們的數學能力就開始教學，

you might not get it right.

你不太可能會成功。

But if you construct thoughts for yourself,

但如果你構建自己的思考

and what you see on the right there is an actual AI thought,

——就像屏幕右端顯示的， 這是AI為自己生成的「思考」；

something that it generates for itself but it does not share with the student.

當然學生無法看到這些「思考」——

then its accuracy went up dramatically,

你輔導的準確性就會大幅提升。

and its ability to be a world-class tutor went up dramatically.

而AI成為一位超水平老師的可能也大幅提升。

And you can see it's talking to itself here.

你可以看到AI真正和自己對話：

It says, "The student got a different answer than I did,

「學生的答案與我的不一致，

but do not tell them they made a mistake.

不過與其直接判定他們的錯誤，

Instead, ask them to explain how they got to that step."

我應該允許他們解釋 他們是如何進行到這一步的。」

So I'll just finish off, hopefully,

我的演講也差不多了。

you know, what I’ve just shown you is just half of what we are working on,

其實剛剛所展示的一切， 都只是我們整體工作的一半。

and we think this is just the very tip of the iceberg

而且這僅僅是AI能力的

of where this can actually go.

冰山一角。

And I'm pretty convinced, which I wouldn't have been even a year ago,

一年前我還不敢肯定，但現在我有信心說

that we together have a chance of addressing the 2 sigma problem

現在正是解決「兩個標準差問題」，

and turning it into a 2 sigma opportunity,

把它變成「兩個標準差機遇」的時候，

dramatically accelerating education as we know it.

並極大地推動傳統教育的普及。

Now, just to take a step back at a meta level,

從更宏觀的角度來說，

obviously we heard a lot today, the debates on either side.

顯然，我們聽到很多對立雙方的爭論：

There's folks who take a more pessimistic view of AI,

有人持悲觀態度，

they say this is scary,

認為AI的發展是可怕的，

there's all these dystopian scenarios,

反烏托邦的，

we maybe want to slow down, we want to pause.

我們需要放慢甚至停止發展的腳步。

On the other side, there are the more optimistic folks

另一方面，也有人持更加樂觀的態度：

that say, well, we've gone through inflection points before,

他們說，我們已經走過了 許多類似的歷史拐點，

we've gone through the Industrial Revolution.

譬如工業革命這樣

It was scary, but it all kind of worked out.

曾經被視為洪水猛獸的大事件。

And what I'd argue right now

而我認為

is I don't think this is like a flip of a coin

這件事不只是簡單的「硬幣的兩面」，

or this is something where we'll just have to,

或者是一件

like, wait and see which way it turns out.

我們只能任其發展、坐以觀之的事情。

I think everyone here and beyond,

我認為所有人， 不論你身處何處，

we are active participants in this decision.

我們都是這個事件的參與者。

I'm pretty convinced that the first line of reasoning

因為首先我認為

is actually almost a self-fulfilling prophecy,

這其實是一個自證預言。

that if we act with fear and if we say,

如果我們帶著恐懼行動：

"Hey, we've just got to stop doing this stuff,"

「嘿，我們必須停止這些事情。」

what's really going to happen is the rule followers might pause,

那麼實際上會發生的， 就是那些遵守規則的人的會慢下來、

might slow down,

停下來。

but the rule breakers, as Alexandr [Wang] mentioned,

而那些違反規則的人， 正如Alexandr Wang提到的，

the totalitarian governments, the criminal organizations,

極權主義政府、犯罪組織，

they're only going to accelerate.

他們只會加快自己的腳步。

And that leads to what I am pretty convinced is the dystopian state,

最終，壞人比好人擁有更先進的AI。

which is the good actors have worse AIs than the bad actors.

世界變成一個反烏托邦。

But I'll also, you know, talk to the optimists a little bit.

而對於那些持樂觀態度的人，

I don't think that means that,

我認為「樂觀」並不意味著

oh, yeah, then we should just relax and just hope for the best.

我們應該放任自流，然後期望有最好的結果。

That might not happen either.

那同樣不可能發生。

I think all of us together have to fight like hell

我認為，為這項新技術設定保護機制，

to make sure that we put the guardrails,

並在必要的時候建立合適的規則，

we put in -- when the problems arise --

是我們所有人

reasonable regulations.

都必須全力爭取的事情。

But we fight like hell for the positive use cases.

我們全力爭取的當然是有益的用途。

Because very close to my heart,

因為誠實地說，

and obviously there's many potential positive use cases,

儘管AI可能有很多潛在的有益用途，

but perhaps the most powerful use case

但其中最強大、

and perhaps the most poetic use case is if AI, artificial intelligence,

也是最理想化的用途，

can be used to enhance HI, human intelligence,

應該是用AI來提高HI， 即人類的智慧，

human potential and human purpose.

與人類潛力， 從而最終，實現人生意義。

Thank you.

謝謝。

(Applause)

（掌聲）