Dan Berkenstock: The world is one big dataset. Now, how to photograph it ...

譯者: Jen-Huei Wu 審譯者: Ying Wang

Five years ago, I was a Ph.D. student

我五年前還是個博士生

living two lives.

過著兩種生活

In one, I used NASA supercomputers

一個是用美國太空總署的超級電腦

to design next-generation spacecraft,

設計下一代的太空船

and in the other I was a data scientist

另一個身分則是數據科學家

looking for potential smugglers

找出誰有可能走私

of sensitive nuclear technologies.

性能高的核能科技

As a data scientist, I did a lot of analyses,

那作為一個數據科學家

mostly of facilities,

我做了不少的數據分析在設備方面

industrial facilities around the world.

像世界各地的工業設備

And I was always looking for a better canvas

我總是在尋找更好的畫面

to tie these all together.

把落在各地的工業設備拍在一起

And one day, I was thinking about how

有天突然靈機一動

all data has a location,

想到全部的數據不都有存放的地方

and I realized that the answer

瞬間發現

had been staring me in the face.

答案早就在眼前了

Although I was a satellite engineer,

雖然我是衛星工程師

I hadn't thought about using satellite imagery

卻從未想在工作中

in my work.

使用衛星成像

Now, like most of us, I'd been online,

你們大都知道我常在線上

I'd see my house, so I thought,

我看得到我家，所以我就想

I'll hop in there and I'll start looking up

我可以跳入這個領域，然後開始尋找

some of these facilities.

這類的設備

And what I found really surprised me.

我對搜尋的結果感到訝異

The pictures that I was finding

我所找到的圖片

were years out of date,

已經遠遠過時了

and because of that,

也因為這樣

it had relatively little relevance

過時的圖對我

to the work that I was doing today.

現今的工作毫無關聯

But I was intrigued.

卻引起我的好奇心

I mean, satellite imagery is pretty amazing stuff.

用衛星拍出圖像是很酷的事

There are millions and millions of sensors

今天有好幾百萬的感應器

surrounding us today,

充斥在你我的周遭

but there's still so much we don't know on a daily basis.

我們卻對每日的變動一無所知

How much oil is stored in all of China?

像中國蘊藏多少油?

How much corn is being produced?

有多少穀物生產?

How many ships are in all of our world's ports?

今日全球有多少船駐入港口?

Now, in theory, all of these questions

理論上來講，這些問題

could be answered by imagery,

可以看圖像回答

but not if it's old.

但若用過時的可就難了

And if this data was so valuable,

倘若這數據很珍貴

then how come I couldn't get my hands

我怎會沒法子

on more recent pictures?

取得更多得近照呢?

So the story begins over 50 years ago

所以要從 50 年前講起

with the launch of the first generation

當美國政府發射出第一代

of U.S. government photo reconnaissance satellites.

衛星拍攝照片以利勘察

And today, there's a handful

今日

of the great, great grandchildren

到處可見這些

of these early Cold War machines

早期冷戰留下來的機器

which are now operated by private companies

現在都由私營企業經營

and from which the vast majority of satellite imagery

也就是你我看到每日

that you and I see on a daily basis comes.

由衛星拍出的圖像

During this period, launching things into space,

在這期間，射東西到外太空

just the rocket to get the satellite up there,

像用火箭射出衛星到外太空

has cost hundreds of millions of dollars each,

每射出一次就要花好幾億

and that's created tremendous pressure

壓力大到喘不過氣來

to launch things infrequently

所以久久才會射到外太空

and to make sure that when you do,

而且還要確保射時

you cram as much functionality in there as possible.

要盡一切塞進一堆功能

All of this has only made satellites

這只讓衛星

bigger and bigger and bigger

變得越來越來越大

and more expensive,

越來越貴

now nearly a billion, with a b, dollars per copy.

現在都要十億美元了，每份都是億開頭的

Because they are so expensive,

正因為它們超級貴

there aren't very many of them.

所以數量不多

Because there aren't very many of them,

也因為數量不多

the pictures that we see on a daily basis

平常看到的照片

tend to be old.

都早就過時以久了

I think a lot of people actually understand this anecdotally,

我想大家都聽得懂我再講什麼

but in order to visualize just how sparsely

但為了能用想像得出

our planet is collected,

地球的照片真是少得可憐

some friends and I put together a dataset

我與朋友就開始集中數據

of the 30 million pictures that have been gathered

集到三千萬張的照片

by these satellites between 2000 and 2010.

都是在 2000-2010 年的衛星拍的

As you can see in blue, huge areas of our world

那一大片藍色代表 地球罕見的地方

are barely seen, less than once a year,

甚至一年也沒見過一次

and even the areas that are seen most frequently,

就算是最常看見的地方

those in red, are seen at best once a quarter.

紅色代表至少一季見過一次

Now as aerospace engineering grad students,

身為航空航天工程學的研究生

this chart cried out to us as a challenge.

這圖表簡直觸動我每一根神經

Why do these things have to be so expensive?

有必要那麼貴嗎?

Does a single satellite really have to cost

一顆衛星真得有必要貴到

the equivalent of three 747 jumbo jets?

約 3 台波音 747 飛機合起來的價格嗎?

Wasn't there a way to build a smaller,

難道就沒有方法建造

simpler, new satellite design that could enable

更小、更簡單、更新穎的設計

more timely imaging?

更有效率的拍出圖像嗎?

I realize that it does sound a little bit crazy

我知道這聽來像瘋了

that we were going to go out and just

但我們決定要到外面亂闖

begin designing satellites,

只為了重新設計衛星

but fortunately we had help.

好在我們有貴人

In the late 1990s, a couple of professors

在 1990 年末

proposed a concept for radically reducing the price

有幾位教授打膽提出

of putting things in space.

讓衛星射到外大空更便宜的方案

This was hitchhiking small satellites

這些小衛星

alongside much larger satellites.

都是隨著較大的射出

This dropped the cost of putting objects up there

這大幅降低成本

by over a factor of 100,

超過 100 倍

and suddenly we could afford to experiment,

突然間，研究變成負擔得起了

to take a little bit of risk,

冒一點風險

and to realize a lot of innovation.

就可成就更多的創新

And a new generation of engineers and scientists,

新一代的工程師和科學家

mostly out of universities,

大多大學畢業

began launching these very small,

就開始射出小的

breadbox-sized satellites called CubeSats.

像土司箱的人工衛星叫作CubeSats

And these were built with electronics obtained

這些電子器件都取自無線電屋公司

from RadioShack instead of Lockheed Martin.

不是洛克西德公司的

Now it was using the lessons learned from these early missions

接下來就是跟前人取經

that my friends and I began a series of sketches

朋友跟我開始畫出

of our own satellite design.

一系列有關衛星的草圖

And I can't remember a specific day

記不得是哪一天

where we made a conscious decision

不知是哪根蔥促使我們

that we were actually going to go out and build these things,

決心將這些草圖變成實體

but once we got that idea in our minds

一旦那慾望跑進我們的腦海裡

of the world as a dataset,

要將世界變成數據

of being able to capture millions of data points

要能夠捕捉好幾百萬的數據點

on a daily basis describing the global economy,

日日都能收到 相關全球經濟的資料

of being able to unearth billions of connections

能夠接收到幾十億地球外的資訊

between them that had never before been found,

而且是前所未聞的

it just seemed boring

這麼一想

to go work on anything else.

要去做別的事情就沒勁了

And so we moved into a cramped,

所以我們就搬到帕羅奧圖城市

windowless office in Palo Alto,

又窄小又沒窗戶的辦公室

and began working to take our design

開始將草圖上的設計

from the drawing board into the lab.

搬進實驗室做出實體

The first major question we had to tackle

我們遇到第一大的問題

was just how big to build this thing.

就是要建造的有多大

In space, size drives cost,

在外太空，尺寸越大價格越貴

and we had worked with these very small,

而我們在學校裡

breadbox-sized satellites in school,

有做過像土司箱大小的衛星

but as we began to better understand the laws of physics,

隨著對物理定律的了解加深

we found that the quality of pictures

發現由衛星拍出

those satellites could take was very limited,

圖像的品質很有限

because the laws of physics dictate

因為物理定律名確指出

that the best picture you can take through a telescope

要能拍出質感好的圖像

is a function of the diameter of that telescope,

跟望眼鏡的直徑有關

and these satellites had a very small,

但這些衛星都非常小

very constrained volume.

容量也有限

And we found that the best picture we would

我們發現拍出最棒的

have been able to get looked something like this.

圖像卻長這樣子

Although this was the low-cost option,

雖然成本低

quite frankly it was just too blurry

但真得太模糊了

to see the things that make satellite imagery valuable.

這樣子衛星的優點都沒發揮出來

So about three or four weeks later,

所以約 3-4 個星期後

we met a group of engineers randomly

偶遇一個團隊的工程師

who had worked on the first

衛星首次傳送圖像

private imaging satellite ever developed,

就是他們打造的

and they told us that back in the 1970s,

他們說 1970 年時

the U.S. government had found a powerful

美國政府有找到強而有力的方式

optimal tradeoff --

拍出清晰圖像

that in taking pictures at right about one meter resolution,

拍照時用 100 公分的解析度

being able to see objects one meter in size,

就能以同樣的尺寸拍到物體

they had found that they could not just get very high-quality images,

他們發現怎樣就是 拍不出解析度高的圖像

but get a lot of them at an affordable price.

所以用能負擔的價格多拍幾張

From our own computer simulations,

我們用電腦模擬

we quickly found that one meter really was

迅速發現 100 公分

the minimum viable product

小物體是可行的

to be able to see the drivers of our global economy,

能夠見到全球經濟的驅動

for the first time, being able to count

第一次我們能夠算出

the ships and cars and shipping containers and trucks

有多少船、車輛、運輸、貨櫃、卡車

that move around our world on a daily basis,

每天在全球運作

while conveniently still not being able to see individuals.

雖然還不能夠很簡單的看清楚人

We had found our compromise.

我們也找到妥協的方法

We would have to build something larger

我們需要建造比原形

than the original breadbox,

吐司箱大小更大的東西

now more like a mini-fridge,

看起來像個迷你冰箱

but we still wouldn't have to build a pickup truck.

還不到建造小貨車尺寸的地步

So now we had our constraint.

現在出現另一個限制

The laws of physics dictated

物理定律指出

the absolute minimum-sized telescope that we could build.

我們能建造出最小的望眼鏡尺寸

What came next was making the rest of the satellite

接下來的任務就是打造出

as small and as simple as possible,

越小越簡單的衛星越好

basically a flying telescope with four walls

基本上就是一個 會飛的四面望眼鏡

and a set of electronics smaller than a phone book

有著一組電子器材 比一本電話簿還小

that used less power than a 100 watt lightbulb.

用量低於100瓦燈泡

The big challenge became actually taking

挑戰接著變成

the pictures through that telescope.

能否順利拍下圖像?

Traditional imaging satellites use a line scanner,

傳統衛星拍的圖像都用線掃描器

similar to a Xerox machine,

就像一台影印機一樣

and as they traverse the Earth, they take pictures,

然後它們就會橫越地球拍照

scanning row by row by row

每行每行的掃描

to build the complete image.

以便建立完整的圖像

Now people use these because they get a lot of light,

現在人們都能用因為他們亮多了

which means less of the noise you see

這意味著畫質清晰多了

in a low-cost cell phone image.

不像用廉價的手機拍照一樣的粗糙

The problem with them is they require

然而，又有問題出現了

very sophisticated pointing.

它們需要很精緻的數據點

You have to stay focused on a 50-centimeter target

每秒能夠橫越超過七公里

from over 600 miles away

就能擁有很棒的解析度

while moving at more than seven kilometers a second,

但離物體遠超過 600 英呎的地方

which requires an awesome degree of complexity.

卻只能聚焦物體的 50 公分

So instead, we turned to a new generation of video sensors,

所以我們轉向新一代的錄影感應器

originally created for use in night vision goggles.

原本是要用於夜視眼鏡

Instead of taking a single, high quality image,

與其一次拍下高值圖像

we could take a videostream

我們就用視訊流拍下

of individually noisier frames,

一張張粗糙的圖像

but then we could recombine

然後幫它們

all of those frames together

結成一體

into very high-quality images

成為高質圖像

using sophisticated pixel processing techniques

使用的是精密的機器處理畫素

here on the ground,

就在這地上

at a cost of one one hundredth a traditional system.

是原來成本的百分之一

And we applied this technique

我們也使用這科技

to many of the other systems on the satellite as well,

在別的衛星系統上

and day by day, our design evolved

我們的設計每篇都會做更換

from CAD to prototypes

從電腦輔助設計軟體、原型

to production units.

到生產線

A few short weeks ago,

幾星期前

we packed up SkySat 1,

我們製的衛星 SkySat 一號準備好了

put our signatures on it,

簽上我們的名字後

and waved goodbye for the last time on Earth.

就等著與地球拜拜

Today, it's sitting in its final launch configuration

今日它仍在它當初射出的位置

ready to blast off in a few short weeks.

準備在幾星期後射出

And soon, we'll turn our attention to launching

很快的，我們的注意力轉到發射

a constellation of 24 or more of these satellites

一群 24 台或更多的衛星

and beginning to build the scalable analytics

開始提供數據分析

that will allow us to unearth the insights

讓我們看到地球外的視野

in the petabytes of data we will collect.

用拍位元組收集而成

So why do all of this? Why build these satellites?

用意是什麼? 為何要製造這些衛星?

Well, it turns out imaging satellites

因為用衛星傳送圖像

have a unique ability to provide global transparency,

才有辦法提供國際的透明度

and providing that transparency on a timely basis

為此透明度提供有效的時間單位

is simply an idea whose time has come.

就是靈感的來源

We see ourselves as pioneers of a new frontier,

我們是自己為新領域的先驅

and beyond economic data,

經濟數據也時時刻刻

unlocking the human story, moment by moment.

為人類提供圖像

For a data scientist

作為數據科學家

that just happened to go to space camp as a kid,

感覺就像是個小孩 去了一趟外太空野營

it just doesn't get much better than that.

沒有什麼比這感到更棒了

Thank you.

謝謝

(Applause)

(掌聲)