What AI is -- and isn't | Sebastian Thrun and Chris Anderson

譯者: Lilian Chiu 審譯者: Helen Chang

Chris Anderson: Help us understand what machine learning is,

克里斯安德森：幫我們 了解一下機器學習是什麼，

because that seems to be the key driver

因為機器學習似乎是

of so much of the excitement and also of the concern

推動人工智慧

一些令人興奮及重要議題的關鍵動因，

around artificial intelligence.

How does machine learning work?

機器學習如何運作？

Sebastian Thrun: So, artificial intelligence and machine learning

賽巴斯汀索朗：人工智慧和機器學習

大約有六十年歷史，

is about 60 years old

and has not had a great day in its past until recently.

一直到近期才有輝煌的日子可言。

And the reason is that today,

原因是因為現今

we have reached a scale of computing and datasets

我們的計算能力和 資料集規模已經達到

讓機器變聰明所必要的條件。

that was necessary to make machines smart.

So here's how it works.

它的運作方式是這樣的。

If you program a computer today, say, your phone,

如果現在你要為一台電腦 寫程式，比如你的手機，

then you hire software engineers

你會僱用軟體工程師，

that write a very, very long kitchen recipe,

他們會寫一份 非常非常長的廚房食譜，

like, "If the water is too hot, turn down the temperature.

比如「如果水太熱，就把溫度調低。

If it's too cold, turn up the temperature."

如果水太冷，把溫度調高。」

The recipes are not just 10 lines long.

食譜長度並不是只有十行。

They are millions of lines long.

它們長達數百萬行。

A modern cell phone has 12 million lines of code.

一台現代手機有 1200 萬行的程式碼。

A browser has five million lines of code.

一個瀏覽器有五百萬行的程式碼。

And each bug in this recipe can cause your computer to crash.

食譜中的每一個錯誤， 都會造成你的電腦當機。

That's why a software engineer makes so much money.

那就是為什麼軟體工程師 能賺那麼多錢。

The new thing now is that computers can find their own rules.

現在的新發展是，電腦能 找到它們自己的規則。

So instead of an expert deciphering, step by step,

所以不再需要找一個專家， 來針對每個情況的規則

a rule for every contingency,

一步一步地做理解辨識，

what you do now is you give the computer examples

現在你的做法是，給電腦一些範例，

and have it infer its own rules.

讓它推導出它自己的規則。

A really good example is AlphaGo, which recently was won by Google.

最近 Google 的阿爾法圍棋贏得比賽， 就是一個很好的例子。

Normally, in game playing, you would really write down all the rules,

通常，在玩遊戲時， 你得要寫下所有的規則，

but in AlphaGo's case,

但在阿爾法圍棋的這個例子，

the system looked over a million games

系統是去看了一百萬場比賽，

and was able to infer its own rules

能推導出它自己的規則，

and then beat the world's residing Go champion.

然後打敗世界現在的棋王。

That is exciting, because it relieves the software engineer

這讓人很興奮， 因為軟體工程師能鬆口氣了，

of the need of being super smart,

他們不需要超聰明，

and pushes the burden towards the data.

這個重任已轉到資料上。

As I said, the inflection point where this has become really possible --

如我所言，這件事的反轉點在於──

very embarrassing, my thesis was about machine learning.

很慚愧，我的論文主題是機器學習，

It was completely insignificant, don't read it,

它完全不重要，請別去讀它，

because it was 20 years ago

因為那是二十年前寫的，

and back then, the computers were as big as a cockroach brain.

那時，電腦和蟑螂大腦一樣大。

Now they are powerful enough to really emulate

現在，電腦強大到能夠真正地模擬

kind of specialized human thinking.

人類的特定思想。

And then the computers take advantage of the fact

接著，電腦也因為

that they can look at much more data than people can.

可以比人類看更多的資料 進而取得優勢，

So I'd say AlphaGo looked at more than a million games.

阿爾法圍棋已經研究過 一百多萬場比賽。

No human expert can ever study a million games.

沒有任何人類專家能夠 研究到一百萬場比賽。

Google has looked at over a hundred billion web pages.

Google 已看過了一千億個網頁。

No person can ever study a hundred billion web pages.

從來沒有人有能力研究 一千億個網頁。

So as a result, the computer can find rules

因此，電腦能找出一些

that even people can't find.

人類找不出來的規則。

CA: So instead of looking ahead to, "If he does that, I will do that,"

克：換句話說，不太像是： 「如果他那樣下，我就這樣下。」

it's more saying, "Here is what looks like a winning pattern,

比較像是在說： 「下這裡像是獲勝的模式，

here is what looks like a winning pattern."

下那裡像是獲勝的模式。」

ST: Yeah. I mean, think about how you raise children.

賽：是的，想想看 你如何養育你的孩子。

你並不會花前十八年的時間， 對每種狀況給孩子一條規則，

You don't spend the first 18 years giving kids a rule for every contingency

and set them free and they have this big program.

然後放他們自由， 他們就會做出這個大程式。

他們會摔跤，會爬起來， 他們會被賞巴掌或打屁股，

They stumble, fall, get up, they get slapped or spanked,

and they have a positive experience, a good grade in school,

他們會有正向的經驗， 在學校有好成績，

and they figure it out on their own.

他們會靠自己去了解這些。

That's happening with computers now,

現在電腦也是這樣，

which makes computer programming so much easier all of a sudden.

突然間讓電腦寫程式就變簡單了。

Now we don't have to think anymore. We just give them lots of data.

我們不用再花腦筋思考了。 只要給它們大量資料即可。

CA: And so, this has been key to the spectacular improvement

克：所以這是自動駕駛車的能力

in power of self-driving cars.

能夠有重大改善的關鍵。

I think you gave me an example.

我想你給了我一個例子。

Can you explain what's happening here?

你能否解釋一下這裡發生了什麼事？

ST: This is a drive of a self-driving car

賽：這是自動駕駛車的行車，

that we happened to have at Udacity

我們優達學城（Udacity）碰巧有，

and recently made into a spin-off called Voyage.

最近變成稱為 Voyage 的副產品。

We have used this thing called deep learning

我們用所謂的「深度學習」

to train a car to drive itself,

來訓練汽車自動駕駛，

and this is driving from Mountain View, California,

這趟行程從加州的山景城出發

to San Francisco

前往舊金山，

on El Camino Real on a rainy day,

在雨天行駛 El Camino Real 路名，

with bicyclists and pedestrians and 133 traffic lights.

路上有腳踏車騎士及行人， 途中經過 133 個交通燈號。

And the novel thing here is,

新奇的是，

許多個月前，我成立了 Google 自動駕駛汽車團隊，

many, many moons ago, I started the Google self-driving car team.

And back in the day, I hired the world's best software engineers

那時，我僱用了世界上 最好的軟體工程師，

to find the world's best rules.

來找出世界上最好的規則。

This is just trained.

這只是訓練出來的。

We drive this road 20 times,

這條路我們開了二十次，

we put all this data into the computer brain,

我們把所有資料放到電腦的大腦中，

and after a few hours of processing,

經過幾小時的處理之後，

it comes up with behavior that often surpasses human agility.

它所找出的行為， 通常都能勝過人類的機敏。

So it's become really easy to program it.

所以變得很容易為它寫程式。

This is 100 percent autonomous, about 33 miles, an hour and a half.

這是 100% 自主的， 大約 33 英哩，一小時半。

CA: So, explain it -- on the big part of this program on the left,

克：解釋一下這程式左半邊的大部分，

you're seeing basically what the computer sees as trucks and cars

我們可以看到電腦 所看到的卡車與汽車，

and those dots overtaking it and so forth.

還有那些超過它的點。

ST: On the right side, you see the camera image, which is the main input here,

賽：右側的是攝影機的影像， 也就是主要的輸入，

and it's used to find lanes, other cars, traffic lights.

用來找車道、其它車輛、交通號誌。

The vehicle has a radar to do distance estimation.

這車用個雷達來估算距離。

This is very commonly used in these kind of systems.

這是這類系統常用的方式。

On the left side you see a laser diagram,

左邊的是雷射圖，

where you see obstacles like trees and so on depicted by the laser.

可以看到雷射槍描繪出來的障礙， 如樹木等等。

But almost all the interesting work is centering on the camera image now.

但幾乎所有有趣的部份 都以攝影機影像為中心。

We're really shifting over from precision sensors like radars and lasers

我們其實在從精準的感測器， 像是雷達和雷射，

轉換到極便宜的一般感測器。

into very cheap, commoditized sensors.

A camera costs less than eight dollars.

一台攝影機的成本不到 $8。

CA: And that green dot on the left thing, what is that?

克：左邊的綠點是什麼？

Is that anything meaningful?

是有意義的嗎？

ST: This is a look-ahead point for your adaptive cruise control,

賽：這是「向前看」的點， 供自動調整航程控制用，

so it helps us understand how to regulate velocity

它會根據前車的距離

based on how far the cars in front of you are.

幫助我們調整速度。

CA: And so, you've also got an example, I think,

克：我想，你應該也可以舉個例說明

of how the actual learning part takes place.

學習的部份實際上如何進行。

Maybe we can see that. Talk about this.

也許我們可以 邊看那畫面，邊談這個。

ST: This is an example where we posed a challenge to Udacity students

賽：這是我們挑戰 Udacity 學生的一個例子，

to take what we call a self-driving car Nanodegree.

是取得「自駕車奈米學位」的挑戰。

We gave them this dataset

我們給他們這個資料集，

and said "Hey, can you guys figure out how to steer this car?"

說：「你們能不能想出 要如何駕駛這台車？」

And if you look at the images,

如果從影像來看，

it's, even for humans, quite impossible to get the steering right.

即使是人類操縱也很難駕駛好。

And we ran a competition and said, "It's a deep learning competition,

我們進行了一項競賽，並說： 「這是場深度學習競賽，

AI competition,"

人工智慧競賽。」

and we gave the students 48 hours.

我們給學生 48 小時。

So if you are a software house like Google or Facebook,

如果你是間軟體公司， 如 Google 或臉書，

something like this costs you at least six months of work.

像這樣的東西會花你 至少六個月的功夫。

So we figured 48 hours is great.

所以我們認為 48 小時是很棒的。

And within 48 hours, we got about 100 submissions from students,

在 48 小時內，我們得到了 約一百件學生提交的結果，

and the top four got it perfectly right.

前四名完全無誤。

It drives better than I could drive on this imagery,

它駕駛得比我能在 這影像上駕駛得還要好，

using deep learning.

用的就是深度學習。

And again, it's the same methodology.

同樣的方法，

It's this magical thing.

很神奇，

When you give enough data to a computer now,

當你提供電腦足夠的資料，

and give enough time to comprehend the data,

並給它足夠時間來理解那些資料，

it finds its own rules.

它就會自己找到規則。

CA: And so that has led to the development of powerful applications

克：所以那就導致了 強大應用程式的發展，

in all sorts of areas.

在各領域都有。

You were talking to me the other day about cancer.

之前你有和我談過癌症的事。

Can I show this video?

我能播那段影片嗎？

ST: Yeah, absolutely, please. CA: This is cool.

賽：當然，請放。 克：這很酷。

ST: This is kind of an insight into what's happening

賽：這有點像是對完全不同的領域

in a completely different domain.

洞察所發生的事。

This is augmenting, or competing --

在旁觀者眼裡，

這是擴增，或者可說是

it's in the eye of the beholder --

with people who are being paid 400,000 dollars a year,

與年賺 $40 萬美元的人競爭：

dermatologists,

皮膚科醫生，

highly trained specialists.

他們是受過高度訓練的專家，

It takes more than a decade of training to be a good dermatologist.

要受十年以上的訓練才可能 成為好的皮膚科醫生。

What you see here is the machine learning version of it.

這裡所看到的是它的機器學習版本。

It's called a neural network.

稱為「神經網路」，

"Neural networks" is the technical term for these machine learning algorithms.

神經網路是機器學習 演算法的專有名詞，

They've been around since the 1980s.

大約出現於 1980 年代。

This one was invented in 1988 by a Facebook Fellow called Yann LeCun,

這個是在 1988 年由臉書的 研究專員揚勒丘恩所發明，

and it propagates data stages

它傳播數據的階段

through what you could think of as the human brain.

透過一種你可視為是人腦的方式。

It's not quite the same thing, but it emulates the same thing.

它不是人腦，但它模仿人腦。

It goes stage after stage.

一個階段接著一個階段，

In the very first stage, it takes the visual input and extracts edges

在第一個階段取得視覺輸入，

粹取出邊緣、細竿，和點。

and rods and dots.

And the next one becomes more complicated edges

下個階段就變成更複雜的邊緣

and shapes like little half-moons.

以及形狀，像是半月。

And eventually, it's able to build really complicated concepts.

最終，它能建立出非常複雜的概念。

Andrew Ng has been able to show

吳恩達就展示過，

that it's able to find cat faces and dog faces

它能夠在非常大量的影像中找出

貓和狗的臉。

in vast amounts of images.

What my student team at Stanford has shown is that

我在史丹佛的學生團隊也展示過，

if you train it on 129,000 images of skin conditions,

如果你用十二萬九千張 皮膚症狀的影像來訓練它，

including melanoma and carcinomas,

包括黑色素瘤和癌，

you can do as good a job

你就能和最好的人類皮膚科醫生

as the best human dermatologists.

做得一樣好。

And to convince ourselves that this is the case,

為了說服我們自己確實是如此，

we captured an independent dataset that we presented to our network

我們取得了一個獨立的資料集， 拿給我們的網路看，

and to 25 board-certified Stanford-level dermatologists,

也拿給 25 位認證過的 史丹佛水準的皮膚科醫生看，

and compared those.

來做比較。

And in most cases,

在大部份狀況，

they were either on par or above the performance classification accuracy

在分類正確性上， 網路的表現都和人類皮膚科醫生

of human dermatologists.

並駕齊驅或更好。

CA: You were telling me an anecdote.

克：你跟我說過一則軼事。

I think about this image right here.

上面的這張影像。

What happened here?

這裡發生了什麼事？

ST: This was last Thursday. That's a moving piece.

賽：時間是上星期四， 是個進行中的故事。

What we've shown before and we published in "Nature" earlier this year

我們之前展示過，且今年稍早 也刊在「Nature」期刊中，

was this idea that we show dermatologists images

想法是，我們讓皮膚科醫生看影像，

and our computer program images,

也讓我們的電腦程式看，

and count how often they're right.

計算它們多常判斷正確。

But all these images are past images.

但所有影像都是過去的影像。

They've all been biopsied to make sure we had the correct classification.

都已經過切片檢查，確保分類正確。

This one wasn't.

這一張卻不是。

This one was actually done at Stanford by one of our collaborators.

這張其實是在史丹佛 由我們的合作者之一做的。

The story goes that our collaborator,

故事是，我們的合作者

who is a world-famous dermatologist, one of the three best, apparently,

是世界知名的皮膚科醫生， 很顯然是最好的三位之一，

looked at this mole and said, "This is not skin cancer."

他看著這個痣，說： 「這不是皮膚癌。」

And then he had a second moment, where he said,

他想了一下，接著又說：

"Well, let me just check with the app."

「讓我用應用程式確認一下。」

So he took out his iPhone and ran our piece of software,

他拿出他的 iPhone， 執行我們的軟體，

our "pocket dermatologist," so to speak,

可說是我們的「口袋皮膚科醫生」，

and the iPhone said: cancer.

而 iPhone 說：癌症。

It said melanoma.

它說是黑色素瘤。

And then he was confused.

他很困惑。

And he decided, "OK, maybe I trust the iPhone a little bit more than myself,"

他決定：「好，也許我應該相信 iPhone 比相信我自己多一點點。」

and he sent it out to the lab to get it biopsied.

他把它送去實驗室做切片檢查。

And it came up as an aggressive melanoma.

結果是惡性黑色素瘤。

So I think this might be the first time that we actually found,

我想，這可能是我們第一次

真正在深度學習的實做中遇到，

in the practice of using deep learning,

an actual person whose melanoma would have gone unclassified,

如果沒有深度學習的話，

這個人的黑色素瘤就不會被發現。

had it not been for deep learning.

CA: I mean, that's incredible.

克：那很了不起。

(Applause)

（掌聲）

It feels like there'd be an instant demand for an app like this right now,

感覺現在對於像這樣的應用程式， 有很迫切的需求，

that you might freak out a lot of people.

你可能會嚇壞很多人。

Are you thinking of doing this, making an app that allows self-checking?

你有想過要這麼做嗎？ 做個自我檢測的應用程式？

賽：我的收件匣被關於癌症 應用程式的信件給淹滿了，

ST: So my in-box is flooded about cancer apps,

with heartbreaking stories of people.

信上都是人們的心碎故事。

I mean, some people have had 10, 15, 20 melanomas removed,

有些人已經移除了 10、15、20 個黑色素瘤，

and are scared that one might be overlooked, like this one,

很害怕會漏掉任何一個，就像這個，

and also, about, I don't know,

還有些內容是，我不知道，

flying cars and speaker inquiries these days, I guess.

飛天車、演說邀請，我猜是吧。

My take is, we need more testing.

我的反應是，我們需要更多測試。

I want to be very careful.

我想要非常小心。

It's very easy to give a flashy result and impress a TED audience.

很容易就可以丟出亮眼的結果 來讓 TED 觀眾印象深刻。

It's much harder to put something out that's ethical.

要端出合乎道德的東西就難很多。

And if people were to use the app

如果人們要用這個應用程式，

and choose not to consult the assistance of a doctor

且選擇不去諮詢醫生的協助，

because we get it wrong,

而我們弄錯的話，

I would feel really bad about it.

我就會感覺非常糟。

So we're currently doing clinical tests,

所以我們目前在做臨床實驗，

and if these clinical tests commence and our data holds up,

如果這些實驗開始之後， 我們的資料站得住腳，

we might be able at some point to take this kind of technology

在某個時點，我們或許可以把這技術

and take it out of the Stanford clinic

拿到史丹佛臨床課之外，

and bring it to the entire world,

把它帶給全世界，

places where Stanford doctors never, ever set foot.

帶到史丹佛的醫生 從來沒有去過的地方。

CA: And do I hear this right,

克：我有沒有聽正確，

that it seemed like what you were saying,

聽起來像是你在說

because you are working with this army of Udacity students,

因為你在和這支 Udacity 學生大軍合作，

that in a way, you're applying a different form of machine learning

以某種方式，你們在應用 一種不同形式的機器學習，

than might take place in a company,

可能會發生在公司中的形式，

which is you're combining machine learning with a form of crowd wisdom.

也就是你們將機器學習 與一種群眾智慧結合。

Are you saying that sometimes you think that could actually outperform

你是不是在說， 有時你認為那能夠勝過

公司所能做到的，甚至大型公司？

what a company can do, even a vast company?

賽：我相信現在有一些 讓我很興奮的例子，

ST: I believe there's now instances that blow my mind,

and I'm still trying to understand.

我還在試著了解。

What Chris is referring to is these competitions that we run.

克里斯指的是

我們的競賽才進行了大約 四十八小時就打開來用；

We turn them around in 48 hours,

and we've been able to build a self-driving car

而我們建的自駕車

that can drive from Mountain View to San Francisco on surface streets.

能從山景城開上馬路去到舊金山；

It's not quite on par with Google after seven years of Google work,

它尚未趕上 Google 投入七年心血的成果，

but it's getting there.

但是就快追上了。

And it took us only two engineers and three months to do this.

我們的研發只花了兩個工程師 用了三個月就做到這樣，

And the reason is, we have an army of students

原因是，我們有一支學生大軍，

who participate in competitions.

參與競賽的那些學生。

We're not the only ones who use crowdsourcing.

我們並非唯一使用「群眾外包」的人，

Uber and Didi use crowdsource for driving.

Uber 和 Didi 用群眾外包做駕駛，

Airbnb uses crowdsourcing for hotels.

Airbnb 用群眾外包做飯店。

There's now many examples where people do bug-finding crowdsourcing

現在有許多例子是 群眾外包除錯工作

or protein folding, of all things, in crowdsourcing.

或群眾外包蛋白質摺疊等。

But we've been able to build this car in three months,

但我們能在三個月內建造這台車，

so I am actually rethinking

所以我其實在重新思考，

how we organize corporations.

我們要如何組織企業。

We have a staff of 9,000 people who are never hired,

我們有從未被僱用的九千名員工，

that I never fire.

我也從未開除他們，

They show up to work and I don't even know.

我不知道他們什麼時候工作。

Then they submit to me maybe 9,000 answers.

後來他們提交大約九千份答案給我。

I'm not obliged to use any of those.

我沒有義務要用任何一個答案。

I end up -- I pay only the winners,

最後我只付錢給贏家，

so I'm actually very cheapskate here, which is maybe not the best thing to do.

所以在這裡我算是個小氣鬼， 這不見得是最好的做法。

But they consider it part of their education, too, which is nice.

但他們認為這是他們 教育的一部份，這樣想很好。

But these students have been able to produce amazing deep learning results.

但這些學生能夠產出非常 了不起的深度學習結果。

So yeah, the synthesis of great people and great machine learning is amazing.

所以，厲害的人結合 偉大的機器學習是很驚人的。

克：加里卡斯帕洛夫 在（TED 2017）第一天說，

CA: I mean, Gary Kasparov said on the first day [of TED2017]

that the winners of chess, surprisingly, turned out to be two amateur chess players

很意外的，棋賽的贏家 是兩位業餘的棋手，

with three mediocre-ish, mediocre-to-good, computer programs,

用三個平庸、中上的電腦程式

that could outperform one grand master with one great chess player,

就勝過了一個大師 和一個很棒的棋手，

like it was all part of the process.

就像這過程的一部份，

And it almost seems like you're talking about a much richer version

幾乎和你談的想法同樣，

of that same idea.

而是更豐富的版本。

ST: Yeah, I mean, as you followed the fantastic panels yesterday morning,

賽：是的，昨天早上的小組討論很棒，

two sessions about AI,

兩場關於人工智慧的討論，

robotic overlords and the human response,

機器超載和人類回應，

many, many great things were said.

說到很多很棒的內容。

But one of the concerns is that we sometimes confuse

但是讓人擔心的事情之一 是有時我們混淆了

what's actually been done with AI with this kind of overlord threat,

人工智慧實際做的事 和機器超載的威脅，

where your AI develops consciousness, right?

也就是人工智慧發展出意識，對吧？

The last thing I want is for my AI to have consciousness.

我最不想要人工智慧有意識。

I don't want to come into my kitchen

我可不想進到廚房，

and have the refrigerator fall in love with the dishwasher

發現冰箱愛上了洗碗機，

and tell me, because I wasn't nice enough,

然後告訴我，因為我不夠好，

my food is now warm.

我的食物現在溫的。

I wouldn't buy these products, and I don't want them.

我不會買這些產品， 我也不想要它們。

But the truth is, for me,

但，事實是，對我來說，

AI has always been an augmentation of people.

人工智慧一直都是人的擴增。

It's been an augmentation of us,

它一直是我們的擴增，

to make us stronger.

讓我們更強大。

And I think Kasparov was exactly correct.

我認為卡斯帕洛夫完全正確。

It's been the combination of human smarts and machine smarts

一直都是人類的聰明 結合機器的聰明，

that make us stronger.

才讓我們更強。

The theme of machines making us stronger is as old as machines are.

機器讓我們更強的主題， 就像機器本身一樣老。

The agricultural revolution took place because it made steam engines

發生農業革命是因為 做出了蒸汽引擎以及耕作設備，

and farming equipment that couldn't farm by itself,

它們不會自己耕作或取代我們，

that never replaced us; it made us stronger.

而是會讓我們更強。

And I believe this new wave of AI will make us much, much stronger

而我相信，這波新的人工智慧風潮

會讓我們人類更強大許多。

as a human race.

CA: We'll come on to that a bit more,

克：我們等等會再談那個話題，

but just to continue with the scary part of this for some people,

但先繼續聊這個 對一些人來說很駭人的部份，

like, what feels like it gets scary for people is when you have

對人們來說，會覺得害怕的是

a computer that can, one, rewrite its own code,

你讓電腦能重寫它自己的程式，

so, it can create multiple copies of itself,

它就能複製多個自己，

try a bunch of different code versions,

嘗試各種不同版本的程式，

possibly even at random,

甚至可能是隨機嘗試，

and then check them out and see if a goal is achieved and improved.

然後再確認看看 目標是否有達成或改善。

So, say the goal is to do better on an intelligence test.

所以，比如，目標是要在一項 智力測驗中得到更好的成績。

You know, a computer that's moderately good at that,

一台電腦只要還算擅長，

you could try a million versions of that.

就能嘗試一百萬個版本，

You might find one that was better,

可能會找到一版比較理想，

and then, you know, repeat.

重覆做下去。

And so the concern is that you get some sort of runaway effect

擔心的是，你會有某種失控效應，

where everything is fine on Thursday evening,

在星期四晚上一切都很好，

and you come back into the lab on Friday morning,

你星期五早上回到實驗室，

and because of the speed of computers and so forth,

因為電腦的速度等等，

一切就天翻地覆，突然間──

things have gone crazy, and suddenly --

ST: I would say this is a possibility,

賽：我會說，這有可能，

but it's a very remote possibility.

卻是非常遙遠的可能。

So let me just translate what I heard you say.

所以讓我翻譯一下我剛聽你說的。

In the AlphaGo case, we had exactly this thing:

阿爾法圍棋的例子就有這樣的狀況：

the computer would play the game against itself

電腦會自己對抗自己來下棋，

and then learn new rules.

接著學習新規則。

And what machine learning is is a rewriting of the rules.

機器學習就是重寫規則。

It's the rewriting of code.

就是重寫程式。

But I think there was absolutely no concern

但我認為完全不用擔心

that AlphaGo would take over the world.

阿爾法圍棋會佔領世界。

It can't even play chess.

它不會下西洋棋。

CA: No, no, no, but now, these are all very single-domain things.

克：不，不，現在這些 都還是非常單一領域的東西。

But it's possible to imagine.

但有可能去想像，

I mean, we just saw a computer that seemed nearly capable

我是指，我們剛看到幾乎有能力

of passing a university entrance test,

通過大學入學測驗的電腦，

that can kind of -- it can't read and understand in the sense that we can,

就像它無法用 我們的方式去閱讀及了解，

but it can certainly absorb all the text

但它絕對可以吸收所有的文字，

and maybe see increased patterns of meaning.

也許能看到越來越多有意義的模式。

Isn't there a chance that, as this broadens out,

有沒有可能，當拓展更廣時，

there could be a different kind of runaway effect?

會是不同種類的失控效應？

ST: That's where I draw the line, honestly.

賽：老實說，我會把底線設在那裡。

And the chance exists -- I don't want to downplay it --

存在這可能性，我不想低估它，

but I think it's remote, and it's not the thing that's on my mind these days,

但我認為它很遙遠， 現在我腦中不會去想這個，

因為我認為大革命是另一回事。

because I think the big revolution is something else.

Everything successful in AI to the present date

目前為止，人工智慧所有的成功，

has been extremely specialized,

都是極度專門化的，

and it's been thriving on a single idea,

一直以來，它能興盛全靠一個辦法：

which is massive amounts of data.

大量的資料。

The reason AlphaGo works so well is because of massive numbers of Go plays,

阿爾法圍棋能如此成功 是因為下過大量的圍棋棋譜，

and AlphaGo can't drive a car or fly a plane.

阿爾法圍棋無法開車或開飛機。

The Google self-driving car or the Udacity self-driving car

Google 的自動駕駛汽車或 Udacity 的自動駕駛汽車

thrives on massive amounts of data, and it can't do anything else.

能成功是因為有大量的資料，

它們無法做其他事， 甚至無法開摩托車。

It can't even control a motorcycle.

It's a very specific, domain-specific function,

它是非常明確、專門領域的功能，

and the same is true for our cancer app.

我們的癌症應用程式也是如此。

There has been almost no progress on this thing called "general AI,"

所謂的「一般性人工智慧」幾無進展，

where you go to an AI and say, "Hey, invent for me special relativity

就是你可以叫它：

「嘿，為我發明 狹義相對論或弦理論」的那種

or string theory."

It's totally in the infancy.

完全還在嬰兒期。

The reason I want to emphasize this,

我想要強調這點的理由

I see the concerns, and I want to acknowledge them.

是我知道人們擔心，我聽見了。

But if I were to think about one thing,

但如果要我思考一件事，我會自問：

I would ask myself the question, "What if we can take anything repetitive

「如果我們能夠把任何重覆事物的

and make ourselves 100 times as efficient?"

效率提高一百倍，會如何？」

It so turns out, 300 years ago, we all worked in agriculture

事實證明，三百年前我們都從事農業，

and did farming and did repetitive things.

耕種，做重覆性的事。

Today, 75 percent of us work in offices

現今，我們有 75% 的人 在辦公室工作，

and do repetitive things.

做重覆性的事。

We've become spreadsheet monkeys.

我們已變成了試算表猴子。

And not just low-end labor.

不只是低階勞工，

We've become dermatologists doing repetitive things,

我們的皮膚科醫生 已經開始做重覆性工作，

lawyers doing repetitive things.

律師也做重覆性工作。

I think we are at the brink of being able to take an AI,

我認為我們正處於 能夠採用 AI 的邊緣，

look over our shoulders,

保持警覺，

and they make us maybe 10 or 50 times as effective in these repetitive things.

可以提高我們執行 重複性工作的效率十或五十倍。

That's what is on my mind.

我在想的是這個。

CA: That sounds super exciting.

克：那聽起來非常讓人興奮。

The process of getting there seems a little terrifying to some people,

對於一些人來說，要達成 那樣的過程似乎有點嚇人，

because once a computer can do this repetitive thing

因為一旦電腦能做重覆性的事，

much better than the dermatologist

且做得比皮膚科醫生好，

or than the driver, especially, is the thing that's talked about

尤其做得比司機好，

這是現在熱門的話題，

so much now,

suddenly millions of jobs go,

突然間，幾百萬個工作就沒了，

and, you know, the country's in revolution

你知道的，這個國家正在革命之中，

before we ever get to the more glorious aspects of what's possible.

我們都還來不及去做到 可能達成的輝煌面。

ST: Yeah, and that's an issue, and it's a big issue,

賽：是啊，那是個課題，大課題，

and it was pointed out yesterday morning by several guest speakers.

昨天早上有幾位嘉賓指出這一點。

Now, prior to me showing up onstage,

在我上台之前，

I confessed I'm a positive, optimistic person,

我坦白說我是個正面、樂觀的人，

so let me give you an optimistic pitch,

讓我為各位定個樂觀的調，

which is, think of yourself back 300 years ago.

就是，試想你回到三百年前，

Europe just survived 140 years of continuous war,

歐洲剛結束了持續 140 年的戰爭，

none of you could read or write,

你們都不會讀或寫，

there were no jobs that you hold today,

沒有你們現在做的工作，

like investment banker or software engineer or TV anchor.

比如投資銀行家、 軟體工程師、電視台主播，

We would all be in the fields and farming.

我們都在田野裡耕種。

Now here comes little Sebastian with a little steam engine in his pocket,

現在，來了一個小賽巴斯汀， 口袋中有個小蒸氣引擎，

saying, "Hey guys, look at this.

說：「嘿，各位，看看這個。

It's going to make you 100 times as strong, so you can do something else."

它會讓你強大一百倍， 這樣你們就可以做其它事了。」

And then back in the day, there was no real stage,

在那個年代，沒有真正的舞台，

but Chris and I hang out with the cows in the stable,

但克里斯和我在畜舍中 和乳牛在一起，

and he says, "I'm really concerned about it,

他說：「我真的很擔心這事，

because I milk my cow every day, and what if the machine does this for me?"

我每天給乳牛擠奶， 如果讓機器來為我擠，會如何？

The reason why I mention this is,

我提到這一點的原因是，

we're always good in acknowledging past progress and the benefit of it,

我們向來都很擅長認可 過去的進展和它帶來的益處，

like our iPhones or our planes or electricity or medical supply.

就像我們的 iPhone、 飛機、電力、醫材。

We all love to live to 80, which was impossible 300 years ago.

我們都喜歡活到八十歲， 這在三百年前是不可能的。

But we kind of don't apply the same rules to the future.

但我們似乎不太會用 同樣的規則看未來。

So if I look at my own job as a CEO,

如果我看我自己的工作，執行長，

I would say 90 percent of my work is repetitive,

我會說，我 90% 的工作 是重覆性的，

I don't enjoy it,

我並不享受做那些，

I spend about four hours per day on stupid, repetitive email.

我每天要花大約四小時在 愚蠢、重覆性的電子郵件上。

And I'm burning to have something that helps me get rid of this.

我極渴望有什麼能協助我擺脫這些。

Why?

為什麼？

Because I believe all of us are insanely creative;

因為我相信我們所有人 都極度有創意；

I think the TED community more than anybody else.

我認為，比起其他人， TED 社區更是如此。

But even blue-collar workers; I think you can go to your hotel maid

但即使藍領階級勞工；我認為 你可以去找你的飯店服務員，

and have a drink with him or her,

和他或她喝杯飲料，

and an hour later, you find a creative idea.

一小時後，你會找到一個創意想法。

What this will empower is to turn this creativity into action.

人工智慧能賦予人能力， 將創意轉化為行動。

Like, what if you could build Google in a day?

比如，如果你能在一天就 建造出 Google，會如何？

What if you could sit over beer and invent the next Snapchat,

如果你能坐著喝啤酒，就發明出 下一個 Snapchat，會如何？

whatever it is,

不論發明的是什麼，

and tomorrow morning it's up and running?

明早它就可以開始運作，會如何？

And that is not science fiction.

那不是科幻小說。

What's going to happen is,

將會發生的事是，

we are already in history.

我們已經在歷史中。

We've unleashed this amazing creativity

我們已經釋放出了這了不起的創意，

by de-slaving us from farming

讓我們脫離耕種的奴役，

and later, of course, from factory work

當然，之後又脫離了 工廠工作的奴役，

and have invented so many things.

且發明出了這麼多東西。

It's going to be even better, in my opinion.

依我所見，將來還會更好。

And there's going to be great side effects.

將來會有很大的副作用。

One of the side effects will be

其中一項副作用會是，

that things like food and medical supply and education and shelter

很多東西，比如食物、 醫材、教育、庇護所

and transportation

以及交通，

will all become much more affordable to all of us,

都會變為大家更負擔得起，

not just the rich people.

不只是有錢人的專利。

CA: Hmm.

克：嗯。

So when Martin Ford argued, you know, that this time it's different

所以，當馬丁福特主張， 這次會有所不同，

because the intelligence that we've used in the past

因為我們在過去用來

to find new ways to be

找出新方式的智慧，

will be matched at the same pace

將會以同樣的速度

by computers taking over those things,

被接手那些事的電腦給比過，

what I hear you saying is that, not completely,

我聽到你說並不完全如此，

because of human creativity.

因為人類有創意。

Do you think that that's fundamentally different from the kind of creativity

你認為那和電腦能做的那種創意

that computers can do?

在根本上是不同的嗎？

ST: So, that's my firm belief as an AI person --

賽：身為人工智慧人，我堅定相信

that I haven't seen any real progress on creativity

我尚未看到任何真正創意上的進展，

and out-of-the-box thinking.

也沒有創造性思維。

What I see right now -- and this is really important for people to realize,

我現在看到的── 人們很需要了解這一點，

because the word "artificial intelligence" is so threatening,

因為「人工智慧」這詞 深具威脅性的，

and then we have Steve Spielberg tossing a movie in,

史帝芬史匹柏拍了一部電影，

電影中電腦突然成了我們的主人，

where all of a sudden the computer is our overlord,

but it's really a technology.

但它其實只是一項技術，

It's a technology that helps us do repetitive things.

協助我們做重覆工作的技術，

And the progress has been entirely on the repetitive end.

而進展完全在重覆性方面：

It's been in legal document discovery.

在法律文件探索上有進展，

It's been contract drafting.

在合約起草上有進展，

It's been screening X-rays of your chest.

在判讀胸腔 X 光上有進展。

And these things are so specialized,

這些工作都很專門化，

I don't see the big threat of humanity.

我看不出對人類有什麼大威脅。

In fact, we as people --

事實上，我們人類──

I mean, let's face it: we've become superhuman.

我們得承認，我們已經變成超人。

We've made us superhuman.

我們已經把自己變成超人。

We can swim across the Atlantic in 11 hours.

我們可以在 11 小時泳渡大西洋。

We can take a device out of our pocket

我們能從口袋中拿出一個裝置

and shout all the way to Australia,

然後對著遙遠的澳洲大吼，

and in real time, have that person shouting back to us.

而且對方還會即時吼回來。

That's physically not possible. We're breaking the rules of physics.

在物理上是不可能的， 我們打破了物理的規則。

When this is said and done, we're going to remember everything

說到底，

我們會記得曾經說過和看過的一切，

we've ever said and seen,

you'll remember every person,

你們將會記得每個人，

which is good for me in my early stages of Alzheimer's.

對在阿滋海默前期的我是件好事。

Sorry, what was I saying? I forgot.

抱歉，我剛在說什麼？我忘了。

CA: (Laughs)

克：（笑聲）

ST: We will probably have an IQ of 1,000 or more.

賽：我們將來可能會有 超過 1,000 的智商。

There will be no more spelling classes for our kids,

我們的孩子將不用再學習拼字，

because there's no spelling issue anymore.

因為將不再有拼字問題。

There's no math issue anymore.

將不再有數學問題。

And I think what really will happen is that we can be super creative.

我認為會發生的是， 我們會超級有創意。

And we are. We are creative.

我們是有創意的，

That's our secret weapon.

那是我們的秘密武器。

CA: So the jobs that are getting lost,

克：所以正在消失中的工作，

in a way, even though it's going to be painful,

在某個層面上，即使會很痛苦，

humans are capable of more than those jobs.

人類的能力是超過這些工作的。

This is the dream.

這是個夢想。

The dream is that humans can rise to just a new level of empowerment

夢想是人類可以崛起

爬升到賦能與探索的新層級。

and discovery.

That's the dream.

這是個夢想。

ST: And think about this:

賽：想想這一點：

if you look at the history of humanity,

如果你去看人類的歷史，

that might be whatever -- 60-100,000 years old, give or take --

也許 6~10 萬年前左右，

almost everything that you cherish in terms of invention,

幾乎你所珍惜的一切，

發明、科技、我們建造的東西，

of technology, of things we've built,

has been invented in the last 150 years.

都是在最近的 150 年間發明的。

If you toss in the book and the wheel, it's a little bit older.

如果你把書和輪子放進來， 那就古老一些。

Or the axe.

或是斧頭。

But your phone, your sneakers,

但你的手機、你的運動鞋、

these chairs, modern manufacturing, penicillin --

這些椅子、現代工業、盤尼西林──

the things we cherish.

我們珍視的東西。

Now, that to me means

對我來說，那意味著，

the next 150 years will find more things.

接下來的 150 年會發現更多東西。

In fact, the pace of invention has gone up, not gone down, in my opinion.

事實上，依我所見，發明的速度 已經變快了，不是變慢。

I believe only one percent of interesting things have been invented yet. Right?

我相信，我們才只發明了 1% 有趣的東西出來。對吧？

We haven't cured cancer.

我們還沒有治癒癌症。

We don't have flying cars -- yet. Hopefully, I'll change this.

我們沒有飛天車，還沒有。 希望我能改變這一點。

That used to be an example people laughed about. (Laughs)

以前那是個會讓人發笑的例子。

（笑聲）

It's funny, isn't it? Working secretly on flying cars.

很有趣，是吧？ 暗地裡致力發明飛天車。

We don't live twice as long yet. OK?

我們的壽命還沒到兩倍長。是嗎？

We don't have this magic implant in our brain

我們在大腦中還沒有神奇的植入物

that gives us the information we want.

能給予我們想要的資訊。

And you might be appalled by it,

你可能會覺得它很駭人，

but I promise you, once you have it, you'll love it.

但我保證，一旦你有了它 就會愛上它。

I hope you will.

我希望你會。

It's a bit scary, I know.

它有點可怕，我知道。

There are so many things we haven't invented yet

還有好多我認為我們應該 發明的東西還沒被發明出來。

that I think we'll invent.

我們沒有重力保護罩。

We have no gravity shields.

我們無法把自己從一地 用光束傳送到另一地。

We can't beam ourselves from one location to another.

那聽起來很荒謬，

That sounds ridiculous,

但大約 200 年前，

but about 200 years ago,

experts were of the opinion that flight wouldn't exist,

專家認為飛機不會存在，

even 120 years ago,

甚至 120 年前。

and if you moved faster than you could run,

還有認為如果你移動速度 比你的跑步速度快，

you would instantly die.

你就會馬上死掉。

So who says we are correct today that you can't beam a person

所以現在誰敢肯定說 我們不能把一個人用光束

from here to Mars?

從這裡傳送到火星？

CA: Sebastian, thank you so much

克：賽巴斯汀，非常謝謝你

for your incredibly inspiring vision and your brilliance.

分享啟發靈感的遠景和你的智慧。

Thank you, Sebastian Thrun.

謝謝你，賽巴斯汀索朗。

That was fantastic. (Applause)

賽：很棒的經驗。（掌聲）