We're building a dystopia just to make people click on ads | Zeynep Tufekci

譯者: Lilian Chiu 審譯者: Helen Chang

So when people voice fears of artificial intelligence,

當人們表達出對人工智慧的恐懼，

very often, they invoke images of humanoid robots run amok.

他們腦中的景象通常是 形象似人的機器人失控殺人。

You know? Terminator?

知道嗎？魔鬼終結者？

You know, that might be something to consider,

雖然考量那種情況的確沒錯，

but that's a distant threat.

但那是遙遠以後的威脅。

Or, we fret about digital surveillance

或者，我們擔心被數位監視，

with metaphors from the past.

有著來自過去的隱喻。

"1984," George Orwell's "1984,"

喬治歐威爾的《1984》

it's hitting the bestseller lists again.

再度登上了暢銷書的排行榜。

It's a great book,

雖然它是本很棒的書，

but it's not the correct dystopia for the 21st century.

但它並未正確地反映出 21 世紀的反烏托邦。

What we need to fear most

我們最需要恐懼的

is not what artificial intelligence will do to us on its own,

並不是人工智慧本身會對我們怎樣，

but how the people in power will use artificial intelligence

而是掌權者會如何運用人工智慧

to control us and to manipulate us

來控制和操縱我們，

in novel, sometimes hidden,

用新穎的、有時隱蔽的、

subtle and unexpected ways.

精細的、出乎意料的方式。

Much of the technology

那些會在不遠的將來

that threatens our freedom and our dignity in the near-term future

威脅我們自由和尊嚴的科技，

is being developed by companies

多半出自下面這類公司，

in the business of capturing and selling our data and our attention

他們攫取我們的注意力和資料，

to advertisers and others:

販售給廣告商和其他對象：

Facebook, Google, Amazon,

臉書、Google、亞馬遜、

Alibaba, Tencent.

阿里巴巴、騰訊。

Now, artificial intelligence has started bolstering their business as well.

人工智慧開始鞏固這些公司的事業。

And it may seem like artificial intelligence

看似人工智慧將是

is just the next thing after online ads.

線上廣告後的下一個產物。

It's not.

並非如此。

It's a jump in category.

它是個大躍進的類別，

It's a whole different world,

一個完全不同的世界，

and it has great potential.

它具有龐大的潛力，

It could accelerate our understanding of many areas of study and research.

能夠加速我們對於 許多研究領域的了解。

But to paraphrase a famous Hollywood philosopher,

但，轉述一位知名 好萊塢哲學家的說法：

"With prodigious potential comes prodigious risk."

「驚人的潛力會帶來驚人的風險。」

Now let's look at a basic fact of our digital lives, online ads.

先談談一個數位生活的 基本面向：線上廣告。

Right? We kind of dismiss them.

我們算是有點輕視了線上的廣告。

They seem crude, ineffective.

它們看似粗糙、無效。

We've all had the experience of being followed on the web

我們都曾經因為在網路上 搜尋或閱讀過某些內容，

by an ad based on something we searched or read.

而老是被一個廣告給跟隨著。

You know, you look up a pair of boots

上網搜尋一雙靴子，

and for a week, those boots are following you around everywhere you go.

之後的一週，你到哪兒 都會看見那雙靴子。

Even after you succumb and buy them, they're still following you around.

即使你屈服，買下了它， 它還是到處跟著你。

We're kind of inured to that kind of basic, cheap manipulation.

我們算是習慣了 那種基本、廉價的操縱，

We roll our eyes and we think, "You know what? These things don't work."

翻個白眼，心想： 「知道嗎？這些沒有用。」

Except, online,

只除了在線上，

the digital technologies are not just ads.

數位科技並不只是廣告。

Now, to understand that, let's think of a physical world example.

為瞭解這一點，我們先用 實體世界當作例子。

You know how, at the checkout counters at supermarkets, near the cashier,

你們有沒有看過，在超市結帳台 靠近收銀機的地方，

there's candy and gum at the eye level of kids?

會有放在孩子視線高度的 糖果和口香糖？

That's designed to make them whine at their parents

那是設計來讓孩子哀求

just as the parents are about to sort of check out.

正在結帳的父母用的。

Now, that's a persuasion architecture.

那是一種說服架構，

It's not nice, but it kind of works.

不太好，但算是有些效用，

That's why you see it in every supermarket.

因此在每個超級市場都看得到。

Now, in the physical world,

在實體世界中，

such persuasion architectures are kind of limited,

這種說服架構有點受限，

because you can only put so many things by the cashier. Right?

因為在收銀台那裡 只擺得下那麼點東西，對吧？

And the candy and gum, it's the same for everyone,

並且每個人看到的 是同樣的糖果和口香糖，

even though it mostly works

這招只對身旁

only for people who have whiny little humans beside them.

有小孩子喋喋不休吵著的大人有用。

In the physical world, we live with those limitations.

我們生活的實體世界裡有那些限制。

In the digital world, though,

但在數位世界裡，

persuasion architectures can be built at the scale of billions

說服架構的規模可達數十億的等級，

and they can target, infer, understand

它們會瞄準、臆測、了解，

and be deployed at individuals

針對個人來部署，

one by one

各個擊破，

by figuring out your weaknesses,

弄清楚個別的弱點，

and they can be sent to everyone's phone private screen,

且能傳送到每個人 私人手機的螢幕上，

so it's not visible to us.

別人是看不見的。

And that's different.

那就很不一樣。

And that's just one of the basic things that artificial intelligence can do.

那只是人工智慧 能做到的基本功能之一。

Now, let's take an example.

讓我舉個例子。

Let's say you want to sell plane tickets to Vegas. Right?

比如說，你要賣飛往賭城的機票。

So in the old world, you could think of some demographics to target

在舊式的世界裡，你可以想出 某些特徵的人來當目標，

based on experience and what you can guess.

根據你的經驗和猜測。

You might try to advertise to, oh,

你也可以試著打廣告，

men between the ages of 25 and 35,

像針對 25~35 歲的男性，

or people who have a high limit on their credit card,

或高信用卡額度的人，

or retired couples. Right?

或退休的夫妻，對吧？

That's what you would do in the past.

那是過去的做法。

With big data and machine learning,

有了大量資料和機器學習，

that's not how it works anymore.

方式就不一樣了。

So to imagine that,

試想，

think of all the data that Facebook has on you:

想想臉書掌握什麼關於你的資料：

every status update you ever typed,

所有你輸入的動態更新、

every Messenger conversation,

所有的訊息對話、

every place you logged in from,

所有你登入時的所在地、

all your photographs that you uploaded there.

所有你上傳的照片。

If you start typing something and change your mind and delete it,

如果你開始輸入些內容， 但隨後改變主意而將之刪除，

Facebook keeps those and analyzes them, too.

臉書會保留那些內容和分析它們。

Increasingly, it tries to match you with your offline data.

它越來越會試著將你 和你的離線資料做匹配，

It also purchases a lot of data from data brokers.

也會向資料仲介商購買許多資料。

It could be everything from your financial records

從你的財務記錄

to a good chunk of your browsing history.

到你過去瀏覽過的一大堆記錄。

Right? In the US, such data is routinely collected,

在美國，這些資料被常規地收集、

collated and sold.

校對和售出。

In Europe, they have tougher rules.

歐洲的規定比較嚴。

So what happens then is,

接下來發生的狀況是

by churning through all that data, these machine-learning algorithms --

透過攪拌所有這些資料， 這些機器學習演算法──

that's why they're called learning algorithms --

這就是為什麼它們 被稱為學習演算法──

they learn to understand the characteristics of people

它們學會了解過去購買機票

who purchased tickets to Vegas before.

飛往賭城的人有何特徵。

When they learn this from existing data,

當它們從既有的資料中 學到這些之後，

they also learn how to apply this to new people.

也學習如何將所學 套用到新的人身上。

So if they're presented with a new person,

如果交給它們一個新的人，

they can classify whether that person is likely to buy a ticket to Vegas or not.

它們能辨識那人可能 或不太可能買機票。

Fine. You're thinking, an offer to buy tickets to Vegas.

好。你心想，不就是提供 購買飛往賭城機票的訊息罷了，

I can ignore that.

可以忽略它。

But the problem isn't that.

但問題不在那裡。

The problem is,

問題是，

we no longer really understand how these complex algorithms work.

我們已經不能真正了解 這些複雜的演算法如何運作。

We don't understand how they're doing this categorization.

我們不了解它們如何分類。

It's giant matrices, thousands of rows and columns,

它是個巨大的矩陣， 有數以千計的直行和橫列，

maybe millions of rows and columns,

也許有上百萬的直行和橫列，

and not the programmers

程式設計者也無法了解，

and not anybody who looks at it,

任何人看到它都無法了解，

even if you have all the data,

即使握有所有的資料，

understands anymore how exactly it's operating

對於它到底如何運作的了解程度，

any more than you'd know what I was thinking right now

絕對不會高於你對我現在 腦中想什麼的了解程度，

if you were shown a cross section of my brain.

如果你單憑看我大腦的切面圖。

It's like we're not programming anymore,

感覺好像我們不是在寫程式了，

we're growing intelligence that we don't truly understand.

而是在栽培一種我們不是 真正了解的智慧。

And these things only work if there's an enormous amount of data,

只在資料量非常巨大的情況下 這些才行得通，

so they also encourage deep surveillance on all of us

所以他們也助長了 對我們所有人的密切監視，

so that the machine learning algorithms can work.

這樣機器學習才能行得通。

That's why Facebook wants to collect all the data it can about you.

那就是為什麼臉書要盡可能 收集關於你的資料。

The algorithms work better.

這樣演算法效果才會比較好。

So let's push that Vegas example a bit.

讓我們再談談賭城的例子。

What if the system that we do not understand

如果這個我們不了解的系統

was picking up that it's easier to sell Vegas tickets

發現比較容易把機票銷售給

to people who are bipolar and about to enter the manic phase.

即將進入躁症階段的躁鬱症患者。

Such people tend to become overspenders, compulsive gamblers.

這類人傾向於變成 花錢超支的人、強迫性賭徒。

They could do this, and you'd have no clue that's what they were picking up on.

他們能這麼做，而你完全不知道 那是他們選目標的根據。

I gave this example to a bunch of computer scientists once

有次，我把這個例子 給了一群電腦科學家，

and afterwards, one of them came up to me.

之後，其中一人來找我。

He was troubled and he said, "That's why I couldn't publish it."

他感到困擾，說：「那就是 為什麼我們無法發表它。」

I was like, "Couldn't publish what?"

我說：「不能發表什麼？」

He had tried to see whether you can indeed figure out the onset of mania

他曾嘗試能否在出現臨床症狀前 就預知躁鬱症快發作了，

靠的是分析社交媒體的貼文。

from social media posts before clinical symptoms,

and it had worked,

他辦到了，

and it had worked very well,

結果非常成功，

and he had no idea how it worked or what it was picking up on.

而他完全不知道是怎麼成功的， 也不知道預測的根據是什麼。

Now, the problem isn't solved if he doesn't publish it,

如果他不發表結果， 問題就沒有解決，

because there are already companies

因為已經有公司

that are developing this kind of technology,

在發展這種技術，

and a lot of the stuff is just off the shelf.

很多東西都已經是現成的了。

This is not very difficult anymore.

這已經不是很困難的事了。

Do you ever go on YouTube meaning to watch one video

你可曾經上 YouTube 原本只是要看一支影片，

and an hour later you've watched 27?

一個小時之後你卻已看了 27 支？

You know how YouTube has this column on the right

你可知道 YouTube 在網頁的右欄

that says, "Up next"

擺著「即將播放」的影片，

and it autoplays something?

而且會自動接著播放那些影片？

It's an algorithm

那是種演算法，

picking what it thinks that you might be interested in

選出它認為你可能會感興趣，

and maybe not find on your own.

但不見得會自己去找到的影片。

It's not a human editor.

並不是人類編輯者，

It's what algorithms do.

而是演算法做的。

It picks up on what you have watched and what people like you have watched,

它去了解你看過什麼影片， 像你這類的人看過什麼影片，

and infers that that must be what you're interested in,

然後推論出那就是你會感興趣、

想看更多的影片，

what you want more of,

然後呈現更多給你看。

and just shows you more.

It sounds like a benign and useful feature,

聽起來是個良性又有用的特色，

except when it isn't.

除了它不是這樣的時候。

So in 2016, I attended rallies of then-candidate Donald Trump

在 2016 年，我去了一場 擁護當時還是候選人川普的集會，

to study as a scholar the movement supporting him.

我以學者身份去研究支持他的運動。

I study social movements, so I was studying it, too.

我研究社會運動，所以也去研究它。

And then I wanted to write something about one of his rallies,

接著，我想要針對 他的某次集會寫點什麼，

so I watched it a few times on YouTube.

所以就在 YouTube 上 看了幾遍。

YouTube started recommending to me

YouTube 開始推薦給我

and autoplaying to me white supremacist videos

並為我自動播放， 白人至上主義的影片，

in increasing order of extremism.

一支比一支更極端主義。

If I watched one,

如果我看了一支，

it served up one even more extreme

它就會送上另一支更極端的，

and autoplayed that one, too.

並且自動播放它。

If you watch Hillary Clinton or Bernie Sanders content,

如果你看的影片內容是 希拉蕊柯林頓或伯尼桑德斯，

YouTube recommends and autoplays conspiracy left,

YouTube 會推薦並自動播放 陰謀論左派的影片，

and it goes downhill from there.

之後就每況愈下。

Well, you might be thinking, this is politics, but it's not.

你可能會想，這是政治。

但並不是，重點不是政治，

This isn't about politics.

This is just the algorithm figuring out human behavior.

這只是猜測人類行為的演算法。

I once watched a video about vegetarianism on YouTube

我曾經上 YouTube 看一支關於吃素的影片，

and YouTube recommended and autoplayed a video about being vegan.

而 YouTube 推薦並自動播放了 一支關於嚴格素食主義者的影片。

It's like you're never hardcore enough for YouTube.

似乎對 YouTube 而言 你的口味永遠都還不夠重。

(Laughter)

（笑聲）

So what's going on?

發生了什麼事？

Now, YouTube's algorithm is proprietary,

YouTube 的演算法是專有的，

but here's what I think is going on.

但我認為發生的事是這樣的：

The algorithm has figured out

演算法發現到，

that if you can entice people

如果誘使人們思索

into thinking that you can show them something more hardcore,

你還能提供他們更重口味的東西，

they're more likely to stay on the site

他們就更可能會留在網站上，

watching video after video going down that rabbit hole

看一支又一支的影片， 一路掉進兔子洞，

while Google serves them ads.

同時 Google 還給他們看廣告。

Now, with nobody minding the ethics of the store,

沒人在意商家倫理的情況下，

these sites can profile people

這些網站能夠描繪人的特性，

who are Jew haters,

哪些人痛恨猶太人，

who think that Jews are parasites

認為猶太人是寄生蟲，

and who have such explicit anti-Semitic content,

以及哪些人明確地反猶太人，

and let you target them with ads.

讓你針對他們提供廣告。

They can also mobilize algorithms

它們也能動員演算法，

to find for you look-alike audiences,

為你找出相近的觀眾群，

people who do not have such explicit anti-Semitic content on their profile

那些側看不怎麼明顯反猶太人，

but who the algorithm detects may be susceptible to such messages,

但是被演算法偵測出來 很容易受到這類訊息影響的人，

and lets you target them with ads, too.

讓你針對他們提供廣告。

Now, this may sound like an implausible example,

這可能聽起來像是個 難以置信的例子，

but this is real.

但它是真實的。

ProPublica investigated this

ProPublica 調查了這件事，

and found that you can indeed do this on Facebook,

且發現你的確可以 在臉書上做到這件事，

and Facebook helpfully offered up suggestions

且臉書很有效地提供建議，

on how to broaden that audience.

告訴你如何拓展觀眾群。

BuzzFeed tried it for Google, and very quickly they found,

BuzzFeed 用 Google 做了實驗，他們很快發現，

yep, you can do it on Google, too.

是的，你也可以在 Google 上這樣做。

And it wasn't even expensive.

而且甚至不貴。

The ProPublica reporter spent about 30 dollars

ProPublica 的記者 花了大約 30 美元

to target this category.

來針對這個類別。

So last year, Donald Trump's social media manager disclosed

去年川普的社交媒體經理透露，

that they were using Facebook dark posts to demobilize people,

他們利用臉書的隱藏廣告貼文 來「反動員」選民，

not to persuade them,

不是勸說或動員他們，

but to convince them not to vote at all.

而是說服他們根本不去投票。

And to do that, they targeted specifically,

為做到這一點，他們準確設定目標，

for example, African-American men in key cities like Philadelphia,

比如像費城這樣 關鍵城市的非裔美國男性，

and I'm going to read exactly what he said.

讓我把他的話一字不漏讀出來。

I'm quoting.

以下為引述。

They were using "nonpublic posts

他們使用「非公開貼文，

whose viewership the campaign controls

那些貼文的觀看權限 由競選團隊來控制，

so that only the people we want to see it see it.

所以只有我們挑的讀者才看得到。

我們為此建立了模型，

We modeled this.

It will dramatically affect her ability to turn these people out."

會嚴重影響到她（指希拉蕊） 動員那些人去投票的能力。」

What's in those dark posts?

那些隱藏廣告貼文中有什麼內容？

We have no idea.

我們不知道。

Facebook won't tell us.

臉書不告訴我們。

So Facebook also algorithmically arranges the posts

所以臉書也用演算法的方式 來安排你的朋友

that your friends put on Facebook, or the pages you follow.

在臉書的貼文或是你追蹤的頁面。

它並不會照時間順序 來呈現所有內容。

It doesn't show you everything chronologically.

It puts the order in the way that the algorithm thinks will entice you

呈現順序是演算法認為

能引誘你在網站上逗留久一點的順序。

to stay on the site longer.

Now, so this has a lot of consequences.

所以，這麼做有許多後果。

You may be thinking somebody is snubbing you on Facebook.

你可能會認為有人在臉書上冷落你。

The algorithm may never be showing your post to them.

也許是演算法根本沒把 你的貼文呈現給他們看。

The algorithm is prioritizing some of them and burying the others.

演算法優先呈現其中某些， 而埋藏掉其他的。

Experiments show

實驗顯示，

that what the algorithm picks to show you can affect your emotions.

演算法選擇呈現給你的內容， 會影響你的情緒。

But that's not all.

但不止這樣，

It also affects political behavior.

它也會影響政治行為。

So in 2010, in the midterm elections,

在 2010 年的期中選舉時，

Facebook did an experiment on 61 million people in the US

臉書做了一個實驗， 對象是美國 6100 萬人，

that was disclosed after the fact.

該實驗後來被揭露出來。

So some people were shown, "Today is election day,"

有些人看到的是「今天是選舉日」，

the simpler one,

簡單的版本，

and some people were shown the one with that tiny tweak

有些人看到的是有 小小調整過的版本，

with those little thumbnails

用小型照片縮圖來顯示出

of your friends who clicked on "I voted."

你的朋友中按了 「我已投票」的那些人。

This simple tweak.

這是個小小的調整。

OK? So the pictures were the only change,

唯一的差別就是照片，

and that post shown just once

這篇貼文只被顯示出來一次，

turned out an additional 340,000 voters

結果多出了 34 萬的投票者

in that election,

在那次選舉投了票，

according to this research

根據這研究指出，

as confirmed by the voter rolls.

這結果已經由選舉人名冊確認過了。

A fluke? No.

是僥倖嗎？不是。

Because in 2012, they repeated the same experiment.

因為在 2012 年， 他們重覆了同樣的實驗。

And that time,

那一次，

that civic message shown just once

只顯示一次的公民訊息

turned out an additional 270,000 voters.

造成投票者多出了 27 萬人。

For reference, the 2016 US presidential election

供參考用：2016 年 總統大選的結果，

was decided by about 100,000 votes.

大約十萬選票的差距決定了江山。

Now, Facebook can also very easily infer what your politics are,

臉書也能輕易推論出你的政治傾向，

即使你未曾在臉書上透露過。

even if you've never disclosed them on the site.

Right? These algorithms can do that quite easily.

對吧？那些演算法很輕易就做得到。

What if a platform with that kind of power

一旦具有那種力量的平台決定要使

decides to turn out supporters of one candidate over the other?

一位候選人的支持者出來投票， 另一位的則不，會如何呢？

How would we even know about it?

我們如何得知發生了這種事？

Now, we started from someplace seemingly innocuous --

我們討論的起始點看似無害──

online adds following us around --

線上廣告跟著我們到處出現──

and we've landed someplace else.

但後來卻談到別的現象。

As a public and as citizens,

身為大眾、身為公民，

we no longer know if we're seeing the same information

我們不再知道 我們看到的資訊是否相同，

or what anybody else is seeing,

或是其他人看到了什麼，

and without a common basis of information,

沒有共同的資訊基礎，

little by little,

漸漸地，

public debate is becoming impossible,

就會變成不可能公開辯論了，

and we're just at the beginning stages of this.

我們目前只是在 這個過程的初始階段。

These algorithms can quite easily infer

這些演算法很容易推論出

things like your people's ethnicity,

比如你的種族、

religious and political views, personality traits,

宗教和政治觀點、個人特質、

intelligence, happiness, use of addictive substances,

智力、快樂程度、 是否使用上癮式物質、

parental separation, age and genders,

父母離異、年齡和性別，

just from Facebook likes.

只從臉書按的讚就能知道。

These algorithms can identify protesters

這些演算法能夠辨識抗議者，

even if their faces are partially concealed.

即使遮蔽他們部份的臉也能辨識。

These algorithms may be able to detect people's sexual orientation

這些演算法或許能偵測人的性向，

just from their dating profile pictures.

只要有他們的約會側寫照片即可。

Now, these are probabilistic guesses,

這些是用機率算出的猜測，

so they're not going to be 100 percent right,

所以不見得 100% 正確，

but I don't see the powerful resisting the temptation to use these technologies

但我並沒有看到因為這些技術有 假陽性結果（實際沒有被預測為有）

just because there are some false positives,

大家就抗拒使用它們，

which will of course create a whole other layer of problems.

因而這些假陽性結果 又造成全然另一層的問題。

Imagine what a state can do

想像一下國家會怎麼用

with the immense amount of data it has on its citizens.

所擁有的大量國民資料。

China is already using face detection technology

中國已經在使用面部辨識技術

to identify and arrest people.

來識別和逮捕人。

And here's the tragedy:

不幸的是，

we're building this infrastructure of surveillance authoritarianism

起初我們建立這個 專制監視的基礎結構，

merely to get people to click on ads.

僅僅為了要讓人們點閱廣告。

And this won't be Orwell's authoritarianism.

這不會是歐威爾的專制主義。

This isn't "1984."

這不是《1984》。

Now, if authoritarianism is using overt fear to terrorize us,

如果專制主義 公然利用恐懼來恐嚇我們，

we'll all be scared, but we'll know it,

我們會害怕，但我們心知肚明，

we'll hate it and we'll resist it.

我們會厭惡它，也會抗拒它。

But if the people in power are using these algorithms

但如果掌權者用這些演算法

to quietly watch us,

悄悄地監看我們、

to judge us and to nudge us,

評斷我們、輕輕推使我們，

to predict and identify the troublemakers and the rebels,

用這些演算法來預測和辨識出 問題製造者和叛亂份子，

to deploy persuasion architectures at scale

部署大規模的說服結構，

and to manipulate individuals one by one

並個別操弄每一個人，

using their personal, individual weaknesses and vulnerabilities,

利用他們個人、個別的缺點和弱點，

and if they're doing it at scale

如果規模夠大，

through our private screens

透過我們私人的螢幕，

so that we don't even know

那麼我們甚至不會知道

what our fellow citizens and neighbors are seeing,

其他公民及鄰居看到了什麼內容，

that authoritarianism will envelop us like a spider's web

那種專制主義會像蜘蛛網 一樣把我們緊緊地包裹起來，

and we may not even know we're in it.

而我們甚至不會知道 自己被包在裡面。

So Facebook's market capitalization

所以，臉書的市場資本化

is approaching half a trillion dollars.

已經接近五千億美元。

It's because it works great as a persuasion architecture.

因為它是個很成功的說服架構。

But the structure of that architecture

但用的架構一樣，

is the same whether you're selling shoes

不論你銷售的是鞋子

or whether you're selling politics.

或是政治。

The algorithms do not know the difference.

演算法不知道差別。

The same algorithms set loose upon us

那個被鬆綁了的、

為使我們更容易 被廣告左右的演算法，

to make us more pliable for ads

are also organizing our political, personal and social information flows,

同時也正組織著我們的 政治、個人和社會的資訊流，

and that's what's got to change.

這點必須要被改變才行。

Now, don't get me wrong,

別誤會我，

we use digital platforms because they provide us with great value.

我們使用數位平台，是因為 它們能提供我們極大的價值。

I use Facebook to keep in touch with friends and family around the world.

我用臉書來和世界各地的 朋友家人保持聯絡。

I've written about how crucial social media is for social movements.

我寫過關於社交媒體對於 社會運動有多重要的文章。

I have studied how these technologies can be used

我研究過這些技術能如何

to circumvent censorship around the world.

被用來規避世界各地的審查制度。

But it's not that the people who run, you know, Facebook or Google

但，不是臉書 或 Google 的營運者

are maliciously and deliberately trying

在惡意、刻意地嘗試

to make the country or the world more polarized

讓國家或世界變得更兩極化、

and encourage extremism.

或鼓勵極端主義。

I read the many well-intentioned statements

我讀過許多出發點很好的聲明，

that these people put out.

都是這些人發出來的。

But it's not the intent or the statements people in technology make that matter,

但重點並不是科技人的意圖或聲明，

it's the structures and business models they're building.

而他們建造的結構與商業模型

And that's the core of the problem.

才是問題的核心。

Either Facebook is a giant con of half a trillion dollars

要不就臉書是個大騙子， 詐騙了半兆美元，

and ads don't work on the site,

該網站上的廣告沒有用，

it doesn't work as a persuasion architecture,

它不以說服架構的形式運作；

or its power of influence is of great concern.

要不就它的影響力很讓人擔心。

It's either one or the other.

只會是兩者其一。

It's similar for Google, too.

Google 也類似。

So what can we do?

所以，我們能做什麼？

This needs to change.

這必須要改變。

Now, I can't offer a simple recipe,

我無法提供簡單的解決之道，

because we need to restructure

因為我們得要重建

the whole way our digital technology operates.

整個數位技術的運作方式；

Everything from the way technology is developed

每一樣──從發展技術的方式

to the way the incentives, economic and otherwise,

到獎勵的方式，不論是實質 或其他形式的獎勵──

are built into the system.

都要被建置到系統中。

We have to face and try to deal with

我們得要面對並試圖處理

the lack of transparency created by the proprietary algorithms,

專有演算法所造成的透明度缺乏，

the structural challenge of machine learning's opacity,

難懂的機器學習的結構性挑戰，

all this indiscriminate data that's being collected about us.

所有被不分皂白地收集走、 與我們相關的資料。

We have a big task in front of us.

我們面對巨大的任務。

We have to mobilize our technology,

我們得要動員我們的科技、

our creativity

我們的創意、

and yes, our politics

以及我們的政治。

so that we can build artificial intelligence

以讓我們建立的人工智慧

that supports us in our human goals

能夠支持我們人類的目標，

but that is also constrained by our human values.

那些同時也被人類價值 所限制住的目標。

And I understand this won't be easy.

我知道這不容易。

We might not even easily agree on what those terms mean.

我們甚至無法輕易取得 那些用語意義的共識。

But if we take seriously

但如果我們認真看待

how these systems that we depend on for so much operate,

我們如此依賴的這些系統如何運作，

I don't see how we can postpone this conversation anymore.

我看不出我們怎能再延遲對話。

These structures

這些結構

are organizing how we function

正在組織我們運作的方式，

and they're controlling

並且控制了

what we can and we cannot do.

我們能做什麼、不能做什麼。

And many of these ad-financed platforms,

許多這類由廣告贊助的平台，

they boast that they're free.

它們誇說它們是免費的。

In this context, it means that we are the product that's being sold.

在這個情境下，意思就是說 「我們」就是被銷售的產品。

We need a digital economy

我們需要一個數位經濟結構，

where our data and our attention

在這個結構中，我們的 資料和注意力是非賣品，

is not for sale to the highest-bidding authoritarian or demagogue.

不能售與出價最高的 專制主義者或煽動者。

(Applause)

（掌聲）

So to go back to that Hollywood paraphrase,

回到前面說的好萊塢改述，

we do want the prodigious potential

我們的確希望人工智慧與數位科技的

of artificial intelligence and digital technology to blossom,

巨大潛能能夠綻放，

but for that, we must face this prodigious menace,

但為此，我們必須要 面對這個巨大的威脅，

open-eyed and now.

睜開眼睛，現在就做。

Thank you.

謝謝。

(Applause)

（掌聲）