living two lives.

我五年前還是個博士生

In one, I used NASA supercomputers

過著兩種生活

to design next-generation spacecraft,

一個是用美國太空總署的超級電腦

and in the other I was a data scientist

設計下一代的太空船

looking for potential smugglers

另一個身分則是數據科學家

of sensitive nuclear technologies.

找出誰有可能走私

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

性能高的核能科技

mostly of facilities,

那作為一個數據科學家

industrial facilities around the world.

我做了不少的數據分析在設備方面

And I was always looking for a better canvas

像世界各地的工業設備

to tie these all together.

我總是在尋找更好的畫面

And one day, I was thinking about how

把落在各地的工業設備拍在一起

all data has a location,

有天突然靈機一動

and I realized that the answer

想到全部的數據不都有存放的地方

had been staring me in the face.

瞬間發現

Although I was a satellite engineer,

答案早就在眼前了

I hadn't thought about using satellite imagery

雖然我是衛星工程師

in my work.

卻從未想在工作中

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

使用衛星成像

I'd see my house, so I thought,

你們大都知道我常在線上

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

我看得到我家，所以我就想

some of these facilities.

我可以跳入這個領域，然後開始尋找

And what I found really surprised me.

這類的設備

The pictures that I was finding

我對搜尋的結果感到訝異

were years out of date,

我所找到的圖片

and because of that,

已經遠遠過時了

it had relatively little relevance

也因為這樣

to the work that I was doing today.

過時的圖對我

But I was intrigued.

現今的工作毫無關聯

I mean, satellite imagery is pretty amazing stuff.

卻引起我的好奇心

There are millions and millions of sensors

用衛星拍出圖像是很酷的事

surrounding us today,

今天有好幾百萬的感應器

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

充斥在你我的周遭

How much oil is stored in all of China?

我們卻對每日的變動一無所知

How much corn is being produced?

像中國蘊藏多少油?

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

有多少穀物生產?

Now, in theory, all of these questions

今日全球有多少船駐入港口?

could be answered by imagery,

理論上來講，這些問題

but not if it's old.

可以看圖像回答

And if this data was so valuable,

但若用過時的可就難了

then how come I couldn't get my hands

倘若這數據很珍貴

on more recent pictures?

我怎會沒法子

So the story begins over 50 years ago

取得更多得近照呢?

with the launch of the first generation

所以要從 50 年前講起

of U.S. government photo reconnaissance satellites.

當美國政府發射出第一代

And today, there's a handful

衛星拍攝照片以利勘察

of the great, great grandchildren

今日

of these early Cold War machines

到處可見這些

which are now operated by private companies

早期冷戰留下來的機器

and from which the vast majority of satellite imagery

現在都由私營企業經營

that you and I see on a daily basis comes.

也就是你我看到每日

During this period, launching things into space,

由衛星拍出的圖像

just the rocket to get the satellite up there,

在這期間，射東西到外太空

has cost hundreds of millions of dollars each,

像用火箭射出衛星到外太空

and that's created tremendous pressure

每射出一次就要花好幾億

to launch things infrequently

壓力大到喘不過氣來

and to make sure that when you do,

所以久久才會射到外太空

you cram as much functionality in there as possible.

而且還要確保射時

All of this has only made satellites

要盡一切塞進一堆功能

bigger and bigger and bigger

這只讓衛星

and more expensive,

變得越來越來越大

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

越來越貴

Because they are so expensive,

現在都要十億美元了，每份都是億開頭的

there aren't very many of them.

正因為它們超級貴

Because there aren't very many of them,

所以數量不多

the pictures that we see on a daily basis

也因為數量不多

tend to be old.

平常看到的照片

I think a lot of people actually understand this anecdotally,

都早就過時以久了

but in order to visualize just how sparsely

我想大家都聽得懂我再講什麼

our planet is collected,

但為了能用想像得出

some friends and I put together a dataset

地球的照片真是少得可憐

of the 30 million pictures that have been gathered

我與朋友就開始集中數據

by these satellites between 2000 and 2010.

集到三千萬張的照片

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

都是在 2000-2010 年的衛星拍的

are barely seen, less than once a year,

那一大片藍色代表 地球罕見的地方

and even the areas that are seen most frequently,

甚至一年也沒見過一次

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

就算是最常看見的地方

Now as aerospace engineering grad students,

紅色代表至少一季見過一次

this chart cried out to us as a challenge.

身為航空航天工程學的研究生

Why do these things have to be so expensive?

這圖表簡直觸動我每一根神經

Does a single satellite really have to cost

有必要那麼貴嗎?

the equivalent of three 747 jumbo jets?

一顆衛星真得有必要貴到

Wasn't there a way to build a smaller,

約 3 台波音 747 飛機合起來的價格嗎?

simpler, new satellite design that could enable

難道就沒有方法建造

more timely imaging?

更小、更簡單、更新穎的設計

I realize that it does sound a little bit crazy

更有效率的拍出圖像嗎?

that we were going to go out and just

我知道這聽來像瘋了

begin designing satellites,

但我們決定要到外面亂闖

but fortunately we had help.

只為了重新設計衛星

In the late 1990s, a couple of professors

好在我們有貴人

proposed a concept for radically reducing the price

在 1990 年末

of putting things in space.

有幾位教授打膽提出

This was hitchhiking small satellites

讓衛星射到外大空更便宜的方案

alongside much larger satellites.

這些小衛星

This dropped the cost of putting objects up there

都是隨著較大的射出

by over a factor of 100,

這大幅降低成本

and suddenly we could afford to experiment,

超過 100 倍

to take a little bit of risk,

突然間，研究變成負擔得起了

and to realize a lot of innovation.

冒一點風險

And a new generation of engineers and scientists,

就可成就更多的創新

mostly out of universities,

新一代的工程師和科學家

began launching these very small,

大多大學畢業

breadbox-sized satellites called CubeSats.

就開始射出小的

And these were built with electronics obtained

像土司箱的人工衛星叫作CubeSats

from RadioShack instead of Lockheed Martin.

這些電子器件都取自無線電屋公司

Now it was using the lessons learned from these early missions

不是洛克西德公司的

that my friends and I began a series of sketches

接下來就是跟前人取經

of our own satellite design.

朋友跟我開始畫出

And I can't remember a specific day

一系列有關衛星的草圖

where we made a conscious decision

記不得是哪一天

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

不知是哪根蔥促使我們

but once we got that idea in our minds

決心將這些草圖變成實體

of the world as a dataset,

一旦那慾望跑進我們的腦海裡

of being able to capture millions of data points

要將世界變成數據

on a daily basis describing the global economy,

要能夠捕捉好幾百萬的數據點

of being able to unearth billions of connections

日日都能收到 相關全球經濟的資料

between them that had never before been found,

能夠接收到幾十億地球外的資訊

it just seemed boring

而且是前所未聞的

to go work on anything else.

這麼一想

And so we moved into a cramped,

要去做別的事情就沒勁了

windowless office in Palo Alto,

所以我們就搬到帕羅奧圖城市

and began working to take our design

又窄小又沒窗戶的辦公室

from the drawing board into the lab.

開始將草圖上的設計

The first major question we had to tackle

搬進實驗室做出實體

was just how big to build this thing.

我們遇到第一大的問題

In space, size drives cost,

就是要建造的有多大

and we had worked with these very small,

在外太空，尺寸越大價格越貴

breadbox-sized satellites in school,

而我們在學校裡

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

有做過像土司箱大小的衛星

we found that the quality of pictures

隨著對物理定律的了解加深

those satellites could take was very limited,

發現由衛星拍出

because the laws of physics dictate

圖像的品質很有限

that the best picture you can take through a telescope

因為物理定律名確指出

is a function of the diameter of that telescope,

要能拍出質感好的圖像

and these satellites had a very small,

跟望眼鏡的直徑有關

very constrained volume.

但這些衛星都非常小

And we found that the best picture we would

容量也有限

have been able to get looked something like this.

我們發現拍出最棒的

Although this was the low-cost option,

圖像卻長這樣子

quite frankly it was just too blurry

雖然成本低

to see the things that make satellite imagery valuable.

但真得太模糊了

So about three or four weeks later,

這樣子衛星的優點都沒發揮出來

we met a group of engineers randomly

所以約 3-4 個星期後

who had worked on the first

偶遇一個團隊的工程師

private imaging satellite ever developed,

衛星首次傳送圖像

and they told us that back in the 1970s,

就是他們打造的

the U.S. government had found a powerful

他們說 1970 年時

optimal tradeoff --

美國政府有找到強而有力的方式

that in taking pictures at right about one meter resolution,

拍出清晰圖像

being able to see objects one meter in size,

拍照時用 100 公分的解析度

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

就能以同樣的尺寸拍到物體

but get a lot of them at an affordable price.

他們發現怎樣就是 拍不出解析度高的圖像

From our own computer simulations,

所以用能負擔的價格多拍幾張

we quickly found that one meter really was

我們用電腦模擬

the minimum viable product

迅速發現 100 公分

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

小物體是可行的

for the first time, being able to count

能夠見到全球經濟的驅動

the ships and cars and shipping containers and trucks

第一次我們能夠算出

that move around our world on a daily basis,

有多少船、車輛、運輸、貨櫃、卡車

while conveniently still not being able to see individuals.

每天在全球運作

We had found our compromise.

雖然還不能夠很簡單的看清楚人

We would have to build something larger

我們也找到妥協的方法

than the original breadbox,

我們需要建造比原形

now more like a mini-fridge,

吐司箱大小更大的東西

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

看起來像個迷你冰箱

So now we had our constraint.

還不到建造小貨車尺寸的地步

The laws of physics dictated

現在出現另一個限制

the absolute minimum-sized telescope that we could build.

物理定律指出

What came next was making the rest of the satellite

我們能建造出最小的望眼鏡尺寸

as small and as simple as possible,

接下來的任務就是打造出

basically a flying telescope with four walls

越小越簡單的衛星越好

and a set of electronics smaller than a phone book

基本上就是一個 會飛的四面望眼鏡

that used less power than a 100 watt lightbulb.

有著一組電子器材 比一本電話簿還小

The big challenge became actually taking

用量低於100瓦燈泡

the pictures through that telescope.

挑戰接著變成

Traditional imaging satellites use a line scanner,

能否順利拍下圖像?

similar to a Xerox machine,

傳統衛星拍的圖像都用線掃描器

and as they traverse the Earth, they take pictures,

就像一台影印機一樣

scanning row by row by row

然後它們就會橫越地球拍照

to build the complete image.

每行每行的掃描

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

以便建立完整的圖像

which means less of the noise you see

現在人們都能用因為他們亮多了

in a low-cost cell phone image.

這意味著畫質清晰多了

The problem with them is they require

不像用廉價的手機拍照一樣的粗糙

very sophisticated pointing.

然而，又有問題出現了

You have to stay focused on a 50-centimeter target

它們需要很精緻的數據點

from over 600 miles away

每秒能夠橫越超過七公里

while moving at more than seven kilometers a second,

就能擁有很棒的解析度

which requires an awesome degree of complexity.

但離物體遠超過 600 英呎的地方

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

卻只能聚焦物體的 50 公分

originally created for use in night vision goggles.

所以我們轉向新一代的錄影感應器

Instead of taking a single, high quality image,

原本是要用於夜視眼鏡

we could take a videostream

與其一次拍下高值圖像

of individually noisier frames,

我們就用視訊流拍下

but then we could recombine

一張張粗糙的圖像

all of those frames together

然後幫它們

into very high-quality images

結成一體

using sophisticated pixel processing techniques

成為高質圖像

here on the ground,

使用的是精密的機器處理畫素

at a cost of one one hundredth a traditional system.

就在這地上

And we applied this technique

是原來成本的百分之一

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

我們也使用這科技

and day by day, our design evolved

在別的衛星系統上

from CAD to prototypes

我們的設計每篇都會做更換

to production units.

從電腦輔助設計軟體、原型

A few short weeks ago,

到生產線

we packed up SkySat 1,

幾星期前

put our signatures on it,

我們製的衛星 SkySat 一號準備好了

and waved goodbye for the last time on Earth.

簽上我們的名字後

Today, it's sitting in its final launch configuration

就等著與地球拜拜

ready to blast off in a few short weeks.

今日它仍在它當初射出的位置

And soon, we'll turn our attention to launching

準備在幾星期後射出

a constellation of 24 or more of these satellites

很快的，我們的注意力轉到發射

and beginning to build the scalable analytics

一群 24 台或更多的衛星

that will allow us to unearth the insights

開始提供數據分析

in the petabytes of data we will collect.

讓我們看到地球外的視野

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

用拍位元組收集而成

Well, it turns out imaging satellites

用意是什麼? 為何要製造這些衛星?

have a unique ability to provide global transparency,

因為用衛星傳送圖像

and providing that transparency on a timely basis

才有辦法提供國際的透明度

is simply an idea whose time has come.

為此透明度提供有效的時間單位

We see ourselves as pioneers of a new frontier,

就是靈感的來源

and beyond economic data,

我們是自己為新領域的先驅

unlocking the human story, moment by moment.

經濟數據也時時刻刻

For a data scientist

為人類提供圖像

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

作為數據科學家

it just doesn't get much better than that.

感覺就像是個小孩 去了一趟外太空野營

Thank you.

沒有什麼比這感到更棒了

(Applause)

謝謝