0:00:00.498,0:00:03.097
Five years ago, I was a Ph.D. student

0:00:03.097,0:00:04.741
living two lives.

0:00:04.741,0:00:07.057
In one, I used NASA supercomputers

0:00:07.057,0:00:09.785
to design next-generation spacecraft,

0:00:09.785,0:00:12.364
and in the other I was a data scientist

0:00:12.364,0:00:14.783
looking for potential smugglers

0:00:14.783,0:00:18.022
of sensitive nuclear technologies.

0:00:18.022,0:00:20.881
As a data scientist, I did a lot of analyses,

0:00:20.881,0:00:22.348
mostly of facilities,

0:00:22.348,0:00:24.619
industrial facilities around the world.

0:00:24.619,0:00:27.307
And I was always looking for a better canvas

0:00:27.307,0:00:29.373
to tie these all together.

0:00:29.373,0:00:31.239
And one day, I was thinking about how

0:00:31.239,0:00:33.363
all data has a location,

0:00:33.363,0:00:35.096
and I realized that the answer

0:00:35.096,0:00:36.984
had been staring me in the face.

0:00:36.984,0:00:39.622
Although I was a satellite engineer,

0:00:39.622,0:00:42.793
I hadn't thought about using satellite imagery

0:00:42.793,0:00:44.359
in my work.

0:00:44.359,0:00:46.306
Now, like most of us, I'd been online,

0:00:46.306,0:00:48.088
I'd see my house, so I thought,

0:00:48.088,0:00:50.386
I'll hop in there and I'll start looking up

0:00:50.386,0:00:51.803
some of these facilities.

0:00:51.803,0:00:53.772
And what I found really surprised me.

0:00:53.772,0:00:55.446
The pictures that I was finding

0:00:55.446,0:00:57.347
were years out of date,

0:00:57.347,0:00:58.415
and because of that,

0:00:58.415,0:01:00.086
it had relatively little relevance

0:01:00.086,0:01:02.546
to the work that I was doing today.

0:01:02.546,0:01:04.007
But I was intrigued.

0:01:04.007,0:01:07.243
I mean, satellite imagery is pretty amazing stuff.

0:01:07.243,0:01:09.546
There are millions and millions of sensors

0:01:09.546,0:01:10.990
surrounding us today,

0:01:10.990,0:01:13.950
but there's still so much we[br]don't know on a daily basis.

0:01:13.950,0:01:17.703
How much oil is stored in all of China?

0:01:17.703,0:01:20.636
How much corn is being produced?

0:01:20.636,0:01:24.707
How many ships are in all of our world's ports?

0:01:24.707,0:01:27.266
Now, in theory, all of these questions

0:01:27.266,0:01:29.510
could be answered by imagery,

0:01:29.510,0:01:31.237
but not if it's old.

0:01:31.237,0:01:33.695
And if this data was so valuable,

0:01:33.695,0:01:35.627
then how come I couldn't get my hands

0:01:35.627,0:01:38.177
on more recent pictures?

0:01:38.177,0:01:41.312
So the story begins over 50 years ago

0:01:41.312,0:01:43.183
with the launch of the first generation

0:01:43.183,0:01:46.783
of U.S. government photo reconnaissance satellites.

0:01:46.783,0:01:48.479
And today, there's a handful

0:01:48.479,0:01:50.674
of the great, great grandchildren

0:01:50.674,0:01:52.339
of these early Cold War machines

0:01:52.339,0:01:54.348
which are now operated by private companies

0:01:54.348,0:01:57.057
and from which the vast majority of satellite imagery

0:01:57.057,0:01:59.882
that you and I see on a daily basis comes.

0:01:59.882,0:02:02.796
During this period, launching things into space,

0:02:02.796,0:02:05.473
just the rocket to get the satellite up there,

0:02:05.473,0:02:10.074
has cost hundreds of millions of dollars each,

0:02:10.074,0:02:11.826
and that's created tremendous pressure

0:02:11.826,0:02:14.115
to launch things infrequently

0:02:14.115,0:02:15.459
and to make sure that when you do,

0:02:15.459,0:02:19.098
you cram as much functionality in there as possible.

0:02:19.098,0:02:20.709
All of this has only made satellites

0:02:20.709,0:02:23.126
bigger and bigger and bigger

0:02:23.126,0:02:24.724
and more expensive,

0:02:24.724,0:02:29.761
now nearly a billion, with a b, dollars per copy.

0:02:29.761,0:02:31.460
Because they are so expensive,

0:02:31.460,0:02:32.927
there aren't very many of them.

0:02:32.927,0:02:34.272
Because there aren't very many of them,

0:02:34.272,0:02:36.728
the pictures that we see on a daily basis

0:02:36.728,0:02:38.232
tend to be old.

0:02:38.232,0:02:41.514
I think a lot of people actually[br]understand this anecdotally,

0:02:41.514,0:02:44.053
but in order to visualize just how sparsely

0:02:44.053,0:02:45.937
our planet is collected,

0:02:45.937,0:02:48.003
some friends and I put together a dataset

0:02:48.003,0:02:50.590
of the 30 million pictures that have been gathered

0:02:50.590,0:02:54.074
by these satellites between 2000 and 2010.

0:02:54.074,0:02:56.829
As you can see in blue, huge areas of our world

0:02:56.829,0:02:59.619
are barely seen, less than once a year,

0:02:59.619,0:03:01.891
and even the areas that are seen most frequently,

0:03:01.891,0:03:05.639
those in red, are seen at best once a quarter.

0:03:05.639,0:03:08.555
Now as aerospace engineering grad students,

0:03:08.555,0:03:11.983
this chart cried out to us as a challenge.

0:03:11.983,0:03:15.438
Why do these things have to be so expensive?

0:03:15.438,0:03:18.173
Does a single satellite really have to cost

0:03:18.173,0:03:22.936
the equivalent of three 747 jumbo jets?

0:03:22.936,0:03:25.099
Wasn't there a way to build a smaller,

0:03:25.099,0:03:28.271
simpler, new satellite design that could enable

0:03:28.271,0:03:30.289
more timely imaging?

0:03:30.289,0:03:33.740
I realize that it does sound a little bit crazy

0:03:33.740,0:03:35.251
that we were going to go out and just

0:03:35.251,0:03:37.098
begin designing satellites,

0:03:37.098,0:03:39.061
but fortunately we had help.

0:03:39.061,0:03:41.996
In the late 1990s, a couple of professors

0:03:41.996,0:03:45.477
proposed a concept for radically reducing the price

0:03:45.477,0:03:47.249
of putting things in space.

0:03:47.249,0:03:49.360
This was hitchhiking small satellites

0:03:49.360,0:03:52.376
alongside much larger satellites.

0:03:52.376,0:03:55.285
This dropped the cost of putting objects up there

0:03:55.285,0:03:57.283
by over a factor of 100,

0:03:57.283,0:04:00.369
and suddenly we could afford to experiment,

0:04:00.369,0:04:01.837
to take a little bit of risk,

0:04:01.837,0:04:04.103
and to realize a lot of innovation.

0:04:04.103,0:04:07.192
And a new generation of engineers and scientists,

0:04:07.192,0:04:08.791
mostly out of universities,

0:04:08.791,0:04:11.134
began launching these very small,

0:04:11.134,0:04:13.725
breadbox-sized satellites called CubeSats.

0:04:13.725,0:04:16.128
And these were built with electronics obtained

0:04:16.128,0:04:20.026
from RadioShack instead of Lockheed Martin.

0:04:20.026,0:04:22.866
Now it was using the lessons[br]learned from these early missions

0:04:22.866,0:04:25.561
that my friends and I began a series of sketches

0:04:25.561,0:04:27.250
of our own satellite design.

0:04:27.250,0:04:30.034
And I can't remember a specific day

0:04:30.034,0:04:31.894
where we made a conscious decision

0:04:31.894,0:04:34.529
that we were actually going to[br]go out and build these things,

0:04:34.529,0:04:36.959
but once we got that idea in our minds

0:04:36.959,0:04:39.120
of the world as a dataset,

0:04:39.120,0:04:41.610
of being able to capture millions of data points

0:04:41.610,0:04:44.663
on a daily basis describing the global economy,

0:04:44.663,0:04:47.183
of being able to unearth billions of connections

0:04:47.183,0:04:50.348
between them that had never before been found,

0:04:50.348,0:04:51.996
it just seemed boring

0:04:51.996,0:04:54.903
to go work on anything else.

0:04:54.903,0:04:57.876
And so we moved into a cramped,

0:04:57.876,0:05:00.855
windowless office in Palo Alto,

0:05:00.855,0:05:02.822
and began working to take our design

0:05:02.822,0:05:05.760
from the drawing board into the lab.

0:05:05.760,0:05:08.418
The first major question we had to tackle

0:05:08.418,0:05:10.752
was just how big to build this thing.

0:05:10.752,0:05:13.935
In space, size drives cost,

0:05:13.935,0:05:15.948
and we had worked with these very small,

0:05:15.948,0:05:18.042
breadbox-sized satellites in school,

0:05:18.042,0:05:20.593
but as we began to better[br]understand the laws of physics,

0:05:20.593,0:05:22.715
we found that the quality of pictures

0:05:22.715,0:05:25.799
those satellites could take was very limited,

0:05:25.799,0:05:27.503
because the laws of physics dictate

0:05:27.503,0:05:30.609
that the best picture you[br]can take through a telescope

0:05:30.609,0:05:32.848
is a function of the diameter of that telescope,

0:05:32.848,0:05:34.510
and these satellites had a very small,

0:05:34.510,0:05:36.216
very constrained volume.

0:05:36.216,0:05:38.011
And we found that the best picture we would

0:05:38.011,0:05:40.496
have been able to get looked something like this.

0:05:40.496,0:05:42.741
Although this was the low-cost option,

0:05:42.741,0:05:44.414
quite frankly it was just too blurry

0:05:44.414,0:05:47.983
to see the things that make[br]satellite imagery valuable.

0:05:47.983,0:05:50.129
So about three or four weeks later,

0:05:50.129,0:05:52.946
we met a group of engineers randomly

0:05:52.946,0:05:54.587
who had worked on the first

0:05:54.587,0:05:57.028
private imaging satellite ever developed,

0:05:57.028,0:05:59.146
and they told us that back in the 1970s,

0:05:59.146,0:06:01.328
the U.S. government had found a powerful

0:06:01.328,0:06:02.971
optimal tradeoff --

0:06:02.971,0:06:06.059
that in taking pictures at right[br]about one meter resolution,

0:06:06.059,0:06:08.943
being able to see objects one meter in size,

0:06:08.943,0:06:11.634
they had found that they could not[br]just get very high-quality images,

0:06:11.634,0:06:14.527
but get a lot of them at an affordable price.

0:06:14.527,0:06:16.503
From our own computer simulations,

0:06:16.503,0:06:18.508
we quickly found that one meter really was

0:06:18.508,0:06:20.272
the minimum viable product

0:06:20.272,0:06:23.351
to be able to see the drivers of our global economy,

0:06:23.351,0:06:24.912
for the first time, being able to count

0:06:24.912,0:06:27.633
the ships and cars and shipping[br]containers and trucks

0:06:27.633,0:06:30.457
that move around our world on a daily basis,

0:06:30.457,0:06:34.261
while conveniently still not[br]being able to see individuals.

0:06:34.261,0:06:35.882
We had found our compromise.

0:06:35.882,0:06:37.486
We would have to build something larger

0:06:37.486,0:06:39.049
than the original breadbox,

0:06:39.049,0:06:40.700
now more like a mini-fridge,

0:06:40.700,0:06:43.461
but we still wouldn't have to build a pickup truck.

0:06:43.461,0:06:46.085
So now we had our constraint.

0:06:46.085,0:06:47.842
The laws of physics dictated

0:06:47.842,0:06:51.166
the absolute minimum-sized[br]telescope that we could build.

0:06:51.166,0:06:54.220
What came next was making the rest of the satellite

0:06:54.220,0:06:55.961
as small and as simple as possible,

0:06:55.961,0:06:58.613
basically a flying telescope with four walls

0:06:58.613,0:07:01.554
and a set of electronics smaller than a phone book

0:07:01.554,0:07:04.992
that used less power than a 100 watt lightbulb.

0:07:04.992,0:07:06.892
The big challenge became actually taking

0:07:06.892,0:07:09.090
the pictures through that telescope.

0:07:09.090,0:07:12.134
Traditional imaging satellites use a line scanner,

0:07:12.134,0:07:13.895
similar to a Xerox machine,

0:07:13.895,0:07:16.295
and as they traverse the Earth, they take pictures,

0:07:16.295,0:07:18.491
scanning row by row by row

0:07:18.491,0:07:20.166
to build the complete image.

0:07:20.166,0:07:23.170
Now people use these because they get a lot of light,

0:07:23.170,0:07:25.001
which means less of the noise you see

0:07:25.016,0:07:27.752
in a low-cost cell phone image.

0:07:27.752,0:07:30.113
The problem with them is they require

0:07:30.113,0:07:32.353
very sophisticated pointing.

0:07:32.353,0:07:34.899
You have to stay focused on a 50-centimeter target

0:07:34.899,0:07:36.651
from over 600 miles away

0:07:36.651,0:07:38.883
while moving at more than[br]seven kilometers a second,

0:07:38.883,0:07:41.632
which requires an awesome degree of complexity.

0:07:41.632,0:07:45.069
So instead, we turned to a new[br]generation of video sensors,

0:07:45.069,0:07:48.160
originally created for use in night vision goggles.

0:07:48.160,0:07:51.059
Instead of taking a single, high quality image,

0:07:51.059,0:07:52.446
we could take a videostream

0:07:52.446,0:07:55.369
of individually noisier frames,

0:07:55.369,0:07:57.113
but then we could recombine

0:07:57.113,0:07:58.579
all of those frames together

0:07:58.579,0:08:00.575
into very high-quality images

0:08:00.575,0:08:03.168
using sophisticated pixel processing techniques

0:08:03.168,0:08:04.689
here on the ground,

0:08:04.689,0:08:07.826
at a cost of one one hundredth a traditional system.

0:08:07.826,0:08:08.979
And we applied this technique

0:08:08.979,0:08:11.953
to many of the other systems on the satellite as well,

0:08:11.953,0:08:14.793
and day by day, our design evolved

0:08:14.793,0:08:18.282
from CAD to prototypes

0:08:18.282,0:08:21.170
to production units.

0:08:21.170,0:08:22.827
A few short weeks ago,

0:08:22.827,0:08:24.678
we packed up SkySat 1,

0:08:24.678,0:08:26.451
put our signatures on it,

0:08:26.451,0:08:28.674
and waved goodbye for the last time on Earth.

0:08:28.674,0:08:32.155
Today, it's sitting in its final launch configuration

0:08:32.155,0:08:35.216
ready to blast off in a few short weeks.

0:08:35.216,0:08:37.558
And soon, we'll turn our attention to launching

0:08:37.558,0:08:40.589
a constellation of 24 or more of these satellites

0:08:40.589,0:08:43.139
and beginning to build the scalable analytics

0:08:43.139,0:08:45.478
that will allow us to unearth the insights

0:08:45.478,0:08:48.667
in the petabytes of data we will collect.

0:08:48.667,0:08:52.640
So why do all of this? Why build these satellites?

0:08:52.640,0:08:55.251
Well, it turns out imaging satellites

0:08:55.251,0:08:58.961
have a unique ability to provide global transparency,

0:08:58.961,0:09:01.587
and providing that transparency on a timely basis

0:09:01.587,0:09:04.998
is simply an idea whose time has come.

0:09:04.998,0:09:08.521
We see ourselves as pioneers of a new frontier,

0:09:08.521,0:09:10.285
and beyond economic data,

0:09:10.285,0:09:13.886
unlocking the human story, moment by moment.

0:09:13.886,0:09:15.449
For a data scientist

0:09:15.449,0:09:18.199
that just happened to go to space camp as a kid,

0:09:18.199,0:09:20.914
it just doesn't get much better than that.

0:09:20.914,0:09:23.129
Thank you.

0:09:23.129,0:09:27.263
(Applause)