How algorithms shape our world | Kevin Slavin

譯者: Resa CC 審譯者: Kuo-Yuan Cheng

This is a photograph

這是張相片

by the artist Michael Najjar,

由藝術家Michael Najjar 拍攝的

and it's real,

這張相片是真的

in the sense that he went there to Argentina

也就是說,他親自到阿根廷,那座山的所在處

to take the photo.

拍攝這張照片。

But it's also a fiction. There's a lot of work that went into it after that.

但也可以說,這是張虛構的相片。這張相片的完作花了很多功夫。

And what he's done

他對相片動了些手腳:

is he's actually reshaped, digitally,

數位化重整

all of the contours of the mountains

整片山脈的形體輪廓,

to follow the vicissitudes of the Dow Jones index.

使其隨道瓊指數曲線變化。

So what you see,

你所看到的

that precipice, that high precipice with the valley,

那個峭壁, 那個有處凹陷的高聳峭壁

is the 2008 financial crisis.

代表2008年的金融危機。

The photo was made

拍攝這張相片時

when we were deep in the valley over there.

我們的金融情勢正處於低谷,

I don't know where we are now.

不曉得我們現在處於何種形勢。

This is the Hang Seng index

這是恆生指數,

for Hong Kong.

香港股市價格的重要指標。

And similar topography.

(兩張相片)地形相似,

I wonder why.

我想知道為什麼

And this is art. This is metaphor.

這是藝術;這是種象徵。

But I think the point is

但我認為重點是

that this is metaphor with teeth,

這個象徵有“牙齒”。

and it's with those teeth that I want to propose today

就是因為這些“牙齒”,我今天提議

that we rethink a little bit

我們稍微重新思考

about the role of contemporary math --

當代數學的角色;

not just financial math, but math in general.

不只金融數學,還有普通數學。

That its transition

「它」的演變:

from being something that we extract and derive from the world

從我們鑽研這個世界,抽絲撥繭而取得的發現

to something that actually starts to shape it --

到實際開始形成「它」的重要發現,

the world around us and the world inside us.

這含括我們的外在世界和我們內在的世界。

And it's specifically algorithms,

說明確些,它是演算法,

which are basically the math

基本上,是種數學─

that computers use to decide stuff.

─電腦用來測定東西的數學。

They acquire the sensibility of truth

演算法掌握高度精確的計量,

because they repeat over and over again,

因為它們一而再,再而三的重覆著;

and they ossify and calcify,

然後漸漸成型，發展出基本架構

and they become real.

然後它們變得實際且可靠。

And I was thinking about this, of all places,

我當時正在思考這點, 真是太湊巧了!

on a transatlantic flight a couple of years ago,

就在幾年前,橫越大西洋的班機上,

because I happened to be seated

因為我的座位碰巧

next to a Hungarian physicist about my age

在一位年紀與我相仿的匈牙利物理學家隔壁

and we were talking

我們談論關於

about what life was like during the Cold War

匈牙利冷戰期間

for physicists in Hungary.

物理學家的生活情況。

And I said, "So what were you doing?"

我說:「你那時在做什麼?」

And he said, "Well we were mostly breaking stealth."

他說:「嗯,我們大多在打擊祕密行動。」

And I said, "That's a good job. That's interesting.

「那是個好工作,有趣吧,

How does that work?"

那是怎麼運作的?」

And to understand that,

要了解那之前

you have to understand a little bit about how stealth works.

你必須稍稍了解祕密行動的運作。

And so -- this is an over-simplification --

這是個超簡單化的例子,

but basically, it's not like

基本上,它不像是

you can just pass a radar signal

你可以藉由156噸在天空飛的鋼鐵

right through 156 tons of steel in the sky.

傳送雷達信號。

It's not just going to disappear.

飛機不會消失不見。

But if you can take this big, massive thing,

但若你能將這個龐大、具規模的東西

and you could turn it into

變成

a million little things --

百萬個小玩意

something like a flock of birds --

─像鳥群一樣的東西─

well then the radar that's looking for that

那麼雷達偵測到那一群群的小東西

has to be able to see

必定會看到

every flock of birds in the sky.

在空中有“一群群的鳥”

And if you're a radar, that's a really bad job.

若你是一個雷達,事情可就糟了。

And he said, "Yeah." He said, "But that's if you're a radar.

他說:「對,但那是,如果你是個雷達。

So we didn't use a radar;

所以我們不用雷達,

we built a black box that was looking for electrical signals,

我們建造一個黑箱子,用它搜尋電波,

electronic communication.

電子通訊。

And whenever we saw a flock of birds that had electronic communication,

任何時候,我們發現帶有電子通訊的鳥群,

we thought, 'Probably has something to do with the Americans.'"

我們會認為這很可能跟美國人有關。」

And I said, "Yeah.

我接著說:「是啊,

That's good.

真行,

So you've effectively negated

你們成功地消磨了

60 years of aeronautic research.

60年的航空學研究心血。

What's your act two?

你接著要做什麼?

What do you do when you grow up?"

當你長大成人以後,你從事什麼工作?」

And he said,

他回答:

"Well, financial services."

「嗯,金融服務業。」

And I said, "Oh."

我驚呼:「喔!」

Because those had been in the news lately.

那一陣子相關報導一直在新聞出現。

And I said, "How does that work?"

我問:「進行的如何?」

And he said, "Well there's 2,000 physicists on Wall Street now,

他說:「現有2,000名物理學家在華爾街(美國金融中心),

and I'm one of them."

我是他們其中一人。」

And I said, "What's the black box for Wall Street?"

我接著問:「華爾街用的『黑箱』是什麼?」

And he said, "It's funny you ask that,

他回說:「你這樣問很好笑,

because it's actually called black box trading.

事實上,人們會稱它為『黑箱交易』

And it's also sometimes called algo trading,

有時也稱為

algorithmic trading."

「演算法交易」(algorithmic trading 或algo trading）

And algorithmic trading evolved in part

「演算法交易」的演進發展,有部分是因為

because institutional traders have the same problems

某些機構交易人遇到相同的問題;

that the United States Air Force had,

而那問題美國空軍也同樣遭遇到,

which is that they're moving these positions --

他們都在「移動這些位置」──

whether it's Proctor & Gamble or Accenture, whatever --

不論是寶僑(Proctor&Gamble)或埃森哲(Accenture:管理顧問、技術服務公司)

they're moving a million shares of something

他們都在移動一百萬股的東西,

through the market.

透過市場交易而進行。

And if they do that all at once,

如果他們一次就挪動全部,

it's like playing poker and going all in right away.

就像玩撲克牌,把剩下的所有籌碼一次全部壓上,

You just tip your hand.

你只會過早洩露底餡;

And so they have to find a way --

所以他們必須找到方法

and they use algorithms to do this --

─他們用演算法,有系統的操作─

to break up that big thing

將龐然大數化整為零

into a million little transactions.

成為百萬個小交易。

And the magic and the horror of that

恐怖的是這個魔術正是

is that the same math

「相同的數學」

that you use to break up the big thing

─用來瓦解龐然巨物

into a million little things

變成百萬個小東西─

can be used to find a million little things

可以用來計算出百萬個零星單位

and sew them back together

又將他們統整在一起

and figure out what's actually happening in the market.

並推算出實際在市場上發生的事情。

So if you need to have some image

如果你立即需要

of what's happening in the stock market right now,

一些股市交易的樣貌,

what you can picture is a bunch of algorithms

你可以構想到的是,成串的運算法

that are basically programmed to hide,

基本上被設計為隱藏不顯示

and a bunch of algorithms that are programmed to go find them and act.

和成串的運算法被設計為可搜尋並執行。

And all of that's great, and it's fine.

整個設計的真是太棒了,又精確。

And that's 70 percent

那是百分之七十的

of the United States stock market,

美國股票市場,

70 percent of the operating system

這個百分之七十的營運系統

formerly known as your pension,

之前堪稱為某些人的“退休金”

your mortgage.

某人的“抵押借款”。

And what could go wrong?

會有什麼錯呢?

What could go wrong

事情出了差池:

is that a year ago,

一年前

nine percent of the entire market just disappears in five minutes,

整體股市的百分之九突然消失了五分鐘,

and they called it the Flash Crash of 2:45.

人們稱之為『瞬間當機2:45』

All of a sudden, nine percent just goes away,

突然, 百分之九就這樣不見了,

and nobody to this day

直到今天,仍沒有人

can even agree on what happened

對發生的事取得一致的意見,

because nobody ordered it, nobody asked for it.

因為沒人“下令”當機;沒人自找麻煩。

Nobody had any control over what was actually happening.

大家對實際正在發生的事情束手無策

All they had

他們只有

was just a monitor in front of them

盯著面前的電腦螢幕,

that had the numbers on it

電腦螢幕上的數字,

and just a red button

和一顆紅色按紐

that said, "Stop."

上面寫著: 『停止』

And that's the thing,

事情就是這樣,

is that we're writing things,

我們正在編寫的「東西」,

we're writing these things that we can no longer read.

我們正在編寫這些連自己都看不懂的東西。

And we've rendered something

我們已經對「某種東西」投降了,

illegible,

某種「難以辨識」的東西。

and we've lost the sense

而且我們失去了

of what's actually happening

對實際正發生之事的判別力

in this world that we've made.

就在我們自己創造的這個世界中,

And we're starting to make our way.

況且我們正開始邁向成功。

There's a company in Boston called Nanex,

在波士頓有間公司叫Nanex(該公司開發市場數據供給系統),

and they use math and magic

他們用數學和魔法

and I don't know what,

和我不知道的什麼來的

and they reach into all the market data

他們深入研究市場數據資料

and they find, actually sometimes, some of these algorithms.

他們確實發現值得重視的東西:某些演算法

And when they find them they pull them out

當他們發現這些演算程序,便把它們擷取出來

and they pin them to the wall like butterflies.

並將它們像蝴蝶一樣釘在牆上。

And they do what we've always done

他們做大家總是會做的事情,

when confronted with huge amounts of data that we don't understand --

當面臨龐大又不懂的數據資料時,

which is that they give them a name

為其命名

and a story.

和揑造故事。

So this is one that they found,

這是他們的發現:

they called the Knife,

他們稱為『刀』

the Carnival,

『嘉年華會』(Carnival)

the Boston Shuffler,

『波士頓通勤者』(Boston Shuffler )

Twilight.

『暮光』

And the gag is

好玩的是

that, of course, these aren't just running through the market.

當然,這些不光是存在於金融市場;

You can find these kinds of things wherever you look,

你能在任何你看得到的地方,發現這些東西,

once you learn how to look for them.

一旦你明白如何找尋到它們(演算法)。

You can find it here: this book about flies

從這兒你可以發現:這是本關於蒼蠅的書,

that you may have been looking at on Amazon.

你可能已在亞馬遜看到這本書;

You may have noticed it

你可能已經注意到

when its price started at 1.7 million dollars.

它的價格從一百七十萬元起價時,

It's out of print -- still ...

這本書是絶版的......仍然絶版中。

(Laughter)

(笑笑)

If you had bought it at 1.7, it would have been a bargain.

如果能以一百七十萬的價格買下它是很划算的

A few hours later, it had gone up

稍後幾小時,它飆漲至

to 23.6 million dollars,

兩千三百六十萬元,

plus shipping and handling.

包含運費和手續費。

And the question is:

問題是:

Nobody was buying or selling anything; what was happening?

這並無產生任何買賣行為；發生了什麼事?

And you see this behavior on Amazon

你在亞馬遜見到這樣的行為,

as surely as you see it on Wall Street.

確實跟你在華爾街看到的一般。

And when you see this kind of behavior,

當你見到這種行為:

what you see is the evidence

你所看到的顯然正是

of algorithms in conflict,

矛盾的演算程序,

algorithms locked in loops with each other,

演算程序被彼此套住,卡在電腦程式回路中;

without any human oversight,

沒有任何“人類監管”

without any adult supervision

沒有任何“成人監護”

to say, "Actually, 1.7 million is plenty."

來告訴你,“其實,一百七十萬已經夠多了!”

(Laughter)

(笑笑)

And as with Amazon, so it is with Netflix.

如同亞馬遜,Netflix(美國公司,經營線上串流影片)也一樣。

And so Netflix has gone through

多年來, Netflix採用過

several different algorithms over the years.

好幾個不同的演算程序。

They started with Cinematch, and they've tried a bunch of others --

他們從Cinematch(推薦系統軟體)開始,也試了一連串其他的軟體。

there's Dinosaur Planet; there's Gravity.

有Dinosaur Planet團隊、Gravity團隊各別研發的推薦系統。

They're using Pragmatic Chaos now.

他們現在使用 Pragmatic Chaos研發的系統。

Pragmatic Chaos is, like all of Netflix algorithms,

像所有Netflix的運算系統,

trying to do the same thing.

Pragmatic Chaos研發的推薦系統,試圖做相同的事。

It's trying to get a grasp on you,

它試著去掌控你們,

on the firmware inside the human skull,

控制人類頭顱內的思考邏輯,

so that it can recommend what movie

以便它能推薦你

you might want to watch next --

下次你也許想看的電影─

which is a very, very difficult problem.

─這是非常高難度的難題。

But the difficulty of the problem

但問題和事實的艱難度

and the fact that we don't really quite have it down,

─我們不是真的掌握問題的事實─

it doesn't take away

並沒減損

from the effects Pragmatic Chaos has.

Pragmatic Chaos的影嚮。

Pragmatic Chaos, like all Netflix algorithms,

Pragmatic Chaos,如同所有Netflix運算系統,

determines, in the end,

至終裁定

60 percent

百分之六十的

of what movies end up being rented.

哪些電影最後會被租借。

So one piece of code

所以一片程式編碼

with one idea about you

─紀錄著你們看片的喜好─

is responsible for 60 percent of those movies.

得為百分之六十的電影負責。

But what if you could rate those movies

但倘若你能評估這些電影,

before they get made?

在電影製作前作預測呢?

Wouldn't that be handy?

那不就簡便多了?

Well, a few data scientists from the U.K. are in Hollywood,

嗯,在好萊塢,一些來自英國的數據科學家

and they have "story algorithms" --

擁有故事情節演算程式系統──

a company called Epagogix.

一間公司叫Epagogix(英國一家預測劇本未來票房好壞的公司)

And you can run your script through there,

你可以拿劇本請這間公司幫你預測;

and they can tell you, quantifiably,

他們會提供你數據:

that that's a 30 million dollar movie

那是一部可賣三千萬的電影

or a 200 million dollar movie.

或是一部兩億的賣座電影。

And the thing is, is that this isn't Google.

事情是......這不是Google;

This isn't information.

這不是情報資料;

These aren't financial stats; this is culture.

這些不是金融統計;這是文化。

And what you see here,

你們在這裡見到的,

or what you don't really see normally,

或者說,實際上,你通常不會察覺的

is that these are the physics of culture.

是物理文化

And if these algorithms,

而且若這些演算系統

like the algorithms on Wall Street,

像華爾街的演算系統

just crashed one day and went awry,

某天突然當機,出岔子了

how would we know?

我們如何會知道.....

What would it look like?

那會如何?

And they're in your house. They're in your house.

再者,它們就在你的房子內,它們就在你的房子內

These are two algorithms competing for your living room.

兩個演算系統在競爭你的客廳。

These are two different cleaning robots

兩個不同的清潔機器人

that have very different ideas about what clean means.

對乾淨的定義有不同的概念。

And you can see it

而且你能從中看到演算程序,

if you slow it down and attach lights to them,

如果讓它慢下來,為它們裝上LCD燈的話,你們就能見識到。

and they're sort of like secret architects in your bedroom.

而且他們有點像在你卧房內的袐密建築師。

And the idea that architecture itself

況且建築學本身的概念

is somehow subject to algorithmic optimization

從某種角度而言,是基於演算法的最佳化

is not far-fetched.

一點也不牽強喔,

It's super-real and it's happening around you.

超真實而且就在存在你週遭。

You feel it most

你感受最深的時刻是,

when you're in a sealed metal box,

當你在一個密閉的金屬箱子內

a new-style elevator;

─一臺新型的電梯─

they're called destination-control elevators.

他們被稱為「終點控制電梯」。

These are the ones where you have to press what floor you're going to go to

這些是電梯,你可以按鈕到你要去的樓層

before you get in the elevator.

在你“進電梯前”按鈕。

And it uses what's called a bin-packing algorithm.

它使用所謂的「裝著演算法的盒子」。

So none of this mishegas

也就是說,這一點也不異常或瘋狂,

of letting everybody go into whatever car they want.

讓每個人選擇進入任何一台電梯。

Everybody who wants to go to the 10th floor goes into car two,

要到十樓的人進入二號電梯;

and everybody who wants to go to the third floor goes into car five.

要到三樓的人進入五號電梯。

And the problem with that

問題是

is that people freak out.

人們嚇壞了

People panic.

人們驚慌失措。

And you see why. You see why.

你看看為什麼......你看看為什麼......

It's because the elevator

原因是:

is missing some important instrumentation, like the buttons.

電梯缺少了某些種要的儀表,譬如說「按鈕」

(Laughter)

(笑笑)

Like the things that people use.

人們會使用那個東西。

All it has

電梯內只顯示

is just the number that moves up or down

上樓或下樓的數字

and that red button that says, "Stop."

還有紅色的按鈕,寫著:『停止』

And this is what we're designing for.

而這是我們正在設計的,

We're designing

我們正在設計

for this machine dialect.

這種「機器方言」。

And how far can you take that? How far can you take it?

你可以作到什麼樣程度?你可以利用它到何種境界?

You can take it really, really far.

你可以“搭乘它(演算法)”至無遠弗界。

So let me take it back to Wall Street.

讓我們退回到華爾街,

Because the algorithms of Wall Street

因為華爾街的演算系統

are dependent on one quality above all else,

仰賴某種性質更勝於一切

which is speed.

即「速度」。

And they operate on milliseconds and microseconds.

他們以毫秒和微秒運作

And just to give you a sense of what microseconds are,

讓你了解什麼是微秒,

it takes you 500,000 microseconds

你需要花五十萬微秒

just to click a mouse.

去點擊滑鼠;

But if you're a Wall Street algorithm

若你是華爾街的演算法

and you're five microseconds behind,

而你落後了五微秒,

you're a loser.

你就是失敗者。

So if you were an algorithm,

所以,倘若你是一個演算法,

you'd look for an architect like the one that I met in Frankfurt

你會找一個建築師,像我在法蘭克福市遇到的那位,

who was hollowing out a skyscraper --

掏空摩天大樓,

throwing out all the furniture, all the infrastructure for human use,

扔掉所有傢俱、所有供人類使用的基礎建設,

and just running steel on the floors

只有鋼鐵舖地

to get ready for the stacks of servers to go in --

準備好讓大批的伺服器入駐。

all so an algorithm

整個如此的演算程序

could get close to the Internet.

能使網路通路密切而有效率。

And you think of the Internet as this kind of distributed system.

再者,你們認為網路是種分散式系統。

And of course, it is, but it's distributed from places.

當然,它是;可是,是從各個定點分散

In New York, this is where it's distributed from:

在紐約,這裡是分佈的中心據點:

the Carrier Hotel

電信機房(Carrier Hotel)

located on Hudson Street.

座落在哈德森街(Hudson Street)

And this is really where the wires come right up into the city.

這裡的確是電纜貫穿整座城市的源頭。

And the reality is that the further away you are from that,

事實是,離那裡越遠

you're a few microseconds behind every time.

每一次就落後數微秒。

These guys down on Wall Street,

在華爾街這一帶的“這些傢伙”

Marco Polo and Cherokee Nation,

Marco Polo和Cherokee Nation

they're eight microseconds

他們落後八微秒,

behind all these guys

落後所有“這些傢伙”

going into the empty buildings being hollowed out

這些傢伙進入被掏空的建築物

up around the Carrier Hotel.

而這些建築座落接近電信機房的周邊。

And that's going to keep happening.

而且那將會持續發生

We're going to keep hollowing them out,

─這些建築物將會持續被掏空─

because you, inch for inch

因為每一英寸

and pound for pound and dollar for dollar,

每一磅和每一(美)元

none of you could squeeze revenue out of that space

你們沒人能從那個空間距離強擠出收益

like the Boston Shuffler could.

像『波士頓通勤者』那般。

But if you zoom out,

但如果縮小地圖

if you zoom out,

縮小地圖

you would see an 825-mile trench

你會看到825英里(1327.7公里)的溝渠

between New York City and Chicago

在紐約和芝加哥之間,

that's been built over the last few years

已建立有幾年了

by a company called Spread Networks.

由Spread Networks 經營。

This is a fiber optic cable

這一道光纖電纜

that was laid between those two cities

被設置在兩城市間

to just be able to traffic one signal

只為一個信號的傳遞

37 times faster than you can click a mouse --

能以37倍速快過點擊滑鼠─

just for these algorithms,

─只為了這些演算系統;

just for the Carnival and the Knife.

只為了『嘉年華會』和『刀』。

And when you think about this,

當你們想著這點時,

that we're running through the United States

我們正以炸藥與岩石鋸貫穿、

with dynamite and rock saws

損耗美國,

so that an algorithm can close the deal

以便一個演算法能快速達成交易

three microseconds faster,

─以減少三微秒的速度─

all for a communications framework

全都為了一個人類

that no human will ever know,

將永不會明瞭的通訊機制

that's a kind of manifest destiny;

那是一種顯而易見的定數

and we'll always look for a new frontier.

且將永遠不斷地尋找未開拓的新領域。

Unfortunately, we have our work cut out for us.

不幸的是,我們必須要完成這個任務。

This is just theoretical.

這只是一個理論。

This is some mathematicians at MIT.

這是某些在麻省理工學院(MIT)的數學家製作的

And the truth is I don't really understand

事實上,我不真的都了解

a lot of what they're talking about.

他們在談論些什麼

It involves light cones and quantum entanglement,

它涉及光圓錐體和量子糾結

and I don't really understand any of that.

我不真的了解那是什麼

But I can read this map,

但我會讀這面地圖。

and what this map says

這面地圖指示

is that, if you're trying to make money on the markets where the red dots are,

如果你試圖在有紅色點點的市場中賺錢

that's where people are, where the cities are,

也就是在人們聚集的地方及市鎮重心,

you're going to have to put the servers where the blue dots are

你就必須將伺服器設置在藍色點點的地方

to do that most effectively.

讓運作效率最大化。

And the thing that you might have noticed about those blue dots

你也許注意到那些藍色點點的分佈,

is that a lot of them are in the middle of the ocean.

很多藍色點點在海的中央;

So that's what we'll do: we'll build bubbles or something,

所以,我們要怎麼做:我們要建立透明圓外罩(bubbles意同泡泡)或什麼來的

or platforms.

或者很多平臺。

We'll actually part the water

我們將能確實分開海水

to pull money out of the air,

將錢從空氣中抽取出,

because it's a bright future

未來是光明閃亮的

if you're an algorithm.

如果你自己就是一個演算法的話。

(Laughter)

(笑笑)

And it's not the money that's so interesting actually.

然而,事實上,不是錢有趣

It's what the money motivates,

而是錢激發的東西引人入勝─

that we're actually terraforming

─我們能確實地地球化(terraforming)

the Earth itself

地球本身,

with this kind of algorithmic efficiency.

透過演算法具有的最佳效率(能)。

And in that light,

根據這點,

you go back

咱們回到前面,

and you look at Michael Najjar's photographs,

看著Michael Najjar的相片

and you realize that they're not metaphor, they're prophecy.

我們領悟到:他們不是象徵;他們是預言

They're prophecy

他們預言了

for the kind of seismic, terrestrial effects

數學之地震效應、陸地效應

of the math that we're making.

即將發生在我們創造出來的數學世界中。

And the landscape was always made

而且這風貌過去一直是由自然界和人之間

by this sort of weird, uneasy collaboration

不可思議的協作及不易妥協而創作出來的,

between nature and man.

是自然界和人之間的對話。

But now there's this third co-evolutionary force: algorithms --

但現在有第三股共同演化勢力:演算系統

the Boston Shuffler, the Carnival.

『波士頓通勤者』、『嘉年華會』

And we will have to understand those as nature,

我們必須明白這些皆為自然。

and in a way, they are.

在某種程度上,它們是!

Thank you.

謝謝大家

(Applause)

(掌聲熱烈)