Seeing Data From Space

there is no pool. This is fine. I don't need to look at Google Maps to know that my neighbors are living their best lives now, and I perhaps am not. But there are other instances, other use cases, if you will, when you might want satellite images that are updated more

frequently than every year or so. For example, if you were a hedge fund who wanted to know how many cars were in the parking lot of every home depot in America last weekend, or an NGO monitoring deforestation in the Amazon, or a government that wanted to see if foreign troops were massing near your border. If you were any of those things, you might want satellite images that are updated much more frequently than once a year. You might even want a new picture from space every single day.

I'm Jacob Goldstein and this is What's Your Problem, the show where I talked to entrepreneurs and engineers about how they're going to change the world once they solve a few problems. My guest today is Will Marshall, the co founder and CEO of Planet. Planet is a private company and they own a fleet of over two hundred satellites that takes a picture of the entire Earth every day. Will's current problem how to turn a daily picture of

the Earth into useful information. Will started his career as an engineer at NASA, and while he was there in a colleague came up with the idea for Planet. But to get started, they faced a simple, massive problem. Satellites were way, way, way too expensive, about a billion dollars per satellite, and they knew they'd never be able to get a whole fleet of satellites into space at that price. So they asked themselves a question, what's something that is

a lot like a satellite, but much cheaper. We had looked at a phone and said, well, that's got most of the ingredients of a satellite. If you look at a smartphone, it has GPS telling you where you are, It has cameras, It has hard drives, it has accelerometers, It has rate gyros that used for gaming, so people know when they're tipping the thing. Actually, that's kind of handy for spacecraft. The only thing it doesn't have that spacecraft has is continuous power supply. It has a battery, right,

but it doesn't have the ability to power itself. So and it doesn't have a bright enough radio if you like. It's it's radio is the most short range, whereas we need a more substantive radio. But it has a lot of ingredients. But also it doesn't cost a billion dollars on the plus, well that's the other thing. Yeah, so because maybe because five hundred dollars not five hundred million dollars. And so we thought it would be cunning to throw some of these phones in space and see what happened

to them. And we did. This is when we're as you like, put it in a box with the solar panel and a stronger radio, yeah exactly, binger, and then put it into space. They were tumbling around and they took a few pictures of the Earth with the cameras on the phone, just with their regular smartphone here. Yeah, and then we had amateur radio observers pick up the signals from these, they're like texting the selfie of the Earth batter Yes, And then we recompiled all the bits

of the pictures into a picture of the Earth. And then then we were like, oh, okay and dipso did it work? Yes, it worked. The phones worked just fine in space. We flew three of them, three Google Nexus ones, because they were easier to hack at than the iPhone. And that was when we decided, okay, we could do this very differently, but it's probably makes more sense as a private company because NASA set about doing exploration and science,

but this was more applied. This is more about commercial benefits or humanitarian benefits or it's very applied in nature. And so we thought we could do it a little bit faster outside of NASA. So we left NASA to do this project of setting out this goal of launching about one hundred satellites to image the whole Earth every

single day. So you leave NASA at this moment when a satellite costs a billion dollars, so you need something between a five hundred dollars smartphone and a billion dollars satellite, right, So what is the thing you've got to build, like more or less, how much does it have to cost?

How cheap do you have to be able to make a satellite when you set out for it to work, we thought that we could do it of order a couple of hundred thousand dollars per satellite, and so to get from a billion to three hundred thousand all the way down, Like, can you just tell me maybe one problem you solved, like one thing you had to figure out? Obviously, I'm sure you had to figure out a thousand things, a billion things to save a billion dollars, but like,

what's one little problem you solved on their way? One is radios. So the smallest radio that would work in the frequency band that is designated for satellites to do Earth imaging costs more than a million dollars I think a few million dollars and was bigger than our entire satellite is meant to be. And we were like, well, damn it, that's not going to work because most of the satellite needs to be our telescope. We need to scram it into a little motherboard and have a little

antenna on the side. And by the way, that that system would only produce about a few megabits a second, and we needed gigabits a second, so it was not only too expensive and too massive, it didn't even do the job. And so we built our own radios from scratch, designed these antenna that was based on pieces of wire, but their machine learning designs so that they even though it's just like a paper clip, almost we now get it's the perfect paper clip. Yes, it's a perfect dedesigned

paper clip. And we can now get one point eight gigabits a second over a thousand kilometer arrange, which is unheard of. And how much does it cost? How much does it cost of order a thousand dollars thousand dollars instead of a million, Yeah, exactly. You know, we didn't

take space grade components. We took components that we knew of work in a vacuum and tried them and had instead of trying to put all our energies into one spacecraft, we would build lots of them and if one or two spacecraft die, it's not the end of the world either. So by doing that it enabled us to leverage the latest technologies. Because if you don't care so much, if it's for sure going to work. You know, you just want it to be like eighty percent probable of working

or ninety percent probable of working. You can take bigger risk, which actually means you can more use more latest gadgets tree. Yeah, so that's a bigger idea. So by using cheaper components, it's okay if they fail, because you're launching one hundred satellites and if ninety five of them naked, that's just fun. That's just fine. And let me give you a concrete example of the last launch we did just a few

weeks ago on a SpaceX rocket. We launched forty four satellites and one of them hasn't worked, and we barely you know, blinked as an organization, you know it was. In fact, that's part of the place they're they're built to fail. What how is there some amount of time like you expect ten percent failure rate or five percent or do you have that sort of penciled down? You know, I've I've asked the team to shoot for more like ten percent or twenty percent failure rate, but we've never

achieved that. We've always been much lower. You need to take more risks, you need to make the right or whatever right, right, right right. That's really just that we're being too conservative. So you build the satellites. It works. You got your cheap satellites. Now you got to get them to space. Yes, correct. We call it piggybacking, piggyback payloads.

So obviously there's a big satellite paying for most of the cust of the bucket, and we say, well, there's a little bit of space there, can we shove a few satellites in. It's like we're hitchhikers. So it's like we're going to the Florida and we're sticking up our thumb and saying, hey, can we can we come up with you guys? Well, and your satellites are in fact small, right, just like yeah, very small. Tell me what one looks like. It is about the size of a loafer bread is

ten by ten by thirty centimeters. It weighs about five kilograms, So it's so tiny for a satellite, tiny, smaller than anybody would think of a satellite being absolutely, you can take them on our hand luggage. In fact, that's how we did get them to the launch site. Early on. You checked them on the you carried out, just carried them on a carry on luggage. Yeah, you put them in the overhead or just put them in the overhead.

I mean we wrap them up a bit and stuff, and we have to tell you really have carried them on the plane. Oh yeah, yeah to the launch site. Yeah. So I mean in a sense like the satellites are cool, but you're a data business, right, Your business is not really launching satellites. It's selling data and specific So you take a picture of the whole earth every day. Who do you sell it to? What do you do with it?

So the biggest one is agriculture, the one I just mentioned, and we help them improve efficiency by twenty or forty percent, decrease use of fertilizer, similar sort of amounts. So that's a big deal for big industry like that forestry. We work with mapping like Google, help improve the maps, civil government with disaster response, so floods, fires, earthquakes. We help countries monitor marine protected areas and coral reef zones and

make sure that there's not illegal fishing. Insurance, hedge funds. Spying comes to mind, like our intelligence agencies clients of yours, Yes they are, but we take a super important principle which is why I think it's different than the classical way the earth imaging can be used in this arena, which is non exclusivity. We can provide that data because we scan automatically. We are not mainly just taking task for a specific location for a specific customer, say the CIA.

You know, we are scanning the whole earth, and if they want to buy it, that's okay. But other people could also buy the same thing, and that's really important. Let me ask you about privacy though, is there a creepy version of your company? Not that you're that version, but like, do you know what I'm saying? Basically, the answer is, we don't really get into personal privacy. And here's why. With five hundred kilometers up. That's like me in San Francisco pointing my camera at Los Angeles and

expecting to see personal details. It turns out you can't. You're too far away that pixel size. Maybe that will go in half in time, you know, but you're still not going to be identify a person. You know, even if it goes in ten x, you still won't be able to identify a person. So we're a very long way from being able to do anything like that. After the break Will's next big problem? How do you turn a picture of the earth into useful, actionable information. That's

the end of the ads. Now we're going back to the show. So I want to talk about some of the problems you haven't figured out how to solve yet, and I think in particular with the data business. Right, I mean, it seems like that's really the sort of frontier for you. Right. You have these pictures every day of the earth, and it's sort of how do you make them useful? Right? So what is something? Yeah, what's

a problem you haven't solved in that part of your business? Well? Many, because the use cases are many, And you're right, a picture, it can be useful, but most people, most of the time want more than a picture. That agriculture of farmer actually doesn't want the picture. They might look at it, they know what their field look, but they more want where do I should I add fertilizers? Where should I order?

The civil government doesn't want to have pictures. They want to know where other roads that are still working after the flood or the bridges down after the you know, a hurricane or whatever said that they can send in the relief supplies. So they want more detailed information that it's in principle in those images, but you have to extract out. So the big project across all of these things that we're doing is something that I call queriable Earth.

Queriable like searchable, like a search query, searchable earth shot. And the reason I think about it the way about this. Google indexed what's on the Internet and made it searchable. Right. It made a database of where all the pages are, what contents are where, so that when you type in whatever, you get an answer that's relevant. We have the ability with the machine learning. The same tools that pick out cats and dogs from pictures that you post online and

can pull that out is the same too. We can use to put out this is a tree, this is a road, this is a house, this is a building, this is a ship. There's a plane out of our images. So we call it object identification using computer vision or machine learning or artificial intelligence is the broadest category. Then you can imagine that the interface to our data becomes less and less the imagery of the earth and more and more a little search box and you can put in, hey,

how many roads are there in India? Give me a plot of that, versus time, tell me where the deforestation is between this month and last month in the Congo basin or the Amazon. The kind of data that you should be able to pull out and I call it yeah, queriabile Earth. So that's the dream. That's the problem. It's a big problem. I mean, it's fundamentally an AI machine learning problem. Like, are you guys? Do you guys have

to solve that AI machine learning problem? Is it sort of the frontier of machine learning and really turning images into the data right or at least into language? I mean, what has to happen? How do you solve that problem? Well, luckily computer vision has come a long way and we don't have to solve it all ourselves. We are doing some of that machine learning ourselves, but we also work with partners, especially because a lot of it's bespoke to

different regions, to different problems, statements. So we have done some core things that we think a lot of people are interested in, roads, building ships, planes, there's a number more who are we going after? And just to be clear, can anybody who wants to just go to your website and search whatever how many roads are there in Michigan? No, No,

you can't. Actually you can sign on a trial account and you can get some of these capabilities, but generally this is set up for at the present more sophisticated users. I don't mean to I don't know how sophisticate you are personally sitting here in my closet recording a podcast.

I'm happy to be an unsophisticated US. But you know, Google uses our APIs, and those big agg companies use our APIs, and they have teams of people and but but that's exactly the problem we're trying to crack, is to help everyone to be able to have access to this data, not just the big organizations that have teams

of people with PhDs in satellite imagery processing. You know, we want to enable it to be that little insurance company, that little that little ngo, that university team that's doing research on penguins or whatever, you know, Like, we want them all to be able to leverage the data. So how do we make it easier to use How do we help them to build the models that they need and build So we're trying to build infrastructure that helps them refine this data to make it useful. I mean,

you need like automation in the middle. Right, there's like you got all the pictures, and then you got people who want information out of the pictures, and right now it's hard and expensive to get the information out of it, and so you've got to automate a lot of that process. Bingo. Okay, let's get back to the show. We're gonna close with the lightning round. Okay, uh, just have a very few quick lightning round for the end. What are the chances

that are Kardashian will go to space? Before you? What is a Kardashian? Let me ask it another way, what are the chances you'll go to space? I think that there's a good chance that anyone that wants to go to space will go if they really you know, in our lifetime, do you want to go to space? Do you want to? I would quite like to, but it's not my main goal in life. I would love to

see the Earth. I'd especially enough to go around the moon and look back at the Earth like four times wider than the Moon in the sky and be able to see the beautiful cradle of civilization. That's a good one. Um Star Wars or Star Trek Star check for sure. What's one piece of advice you'd give somebody trying to solve a hard problem. Um, stick with it, you know. I think the biggest thing that differentiates those succeed those that don't less the idea than the timing, both of

which are important, but more of the staying power. It's it's perseverance and I think, you know you've got to do keep on going in the face of a lot of people saying it's not possible. Good. Do you have any favorite images that Planet has ever captured. I think the very first image we ever captured is an Oregon area of partially forested, partially agricultural lands, pasture lands, and we could immediately see from that first picture that we

could identify individual trees in the field. And then we were like, oh my god, we can we can do this mission. You think you'll ever leave Planet? Leave planet? Did you say? Yeah, like the moment when it will be time for you to do something else. Last question, it's it's definitely my life's work, and I'd be surprised if we're not working here in some capacity for a while. I mean, look, I enjoy it, and I think I'm

more likely to leave the planet than Planet. Will Marshall is co founder and CEO of Planet, big news in the What's Your Problem universe. This week we got our own email address, It's problem at pushkin dot Fm. I would love to hear what you think of the show. If there's topics you want to cover, guests, you want to hear, things you like us to do differently, Please please do let us know again It's problem at pushkin dot Fm. Today's show was edited by Robert Smith, produced

by Edith Russlo, and engineered by Amanda Kwong. I'm Jacob Goldstein and we'll be back next week with another episode of What's Your Problem