Latest / What's Your Problem? / Building Boundary-Breaking Balloons

Transcript

Building Boundary-Breaking Balloons · What's Your Problem?

0:15
Speaker 1: Pushkin. Every day around the world, more than a thousand
0:23
Speaker 1: weather balloons are launched into the sky. The balloons float
0:26
Speaker 1: high up into the atmosphere, sending back information you know, temperature,
0:30
Speaker 1: wind speed, air pressure, et cetera, and then a couple
0:34
Speaker 1: hours later, the balloons pop. This is basically the way
0:39
Speaker 1: weather balloons have worked for decades, and the information the
0:42
Speaker 1: balloons send back is really useful for weather forecasts. But
0:45
Speaker 1: the information the balloon send back is also pretty limited
0:49
Speaker 1: because the balloons only stay in the sky for a
0:51
Speaker 1: couple hours, they don't fly very far. If we could
0:54
Speaker 1: figure out how to make balloons stay up for longer,
0:58
Speaker 1: they could blow in the wind and travel thousands of miles.
1:01
Speaker 1: They could travel across continents and across oceans and send
1:05
Speaker 1: us back a lot more data and give us a
1:07
Speaker 1: much clearer picture of what weather is coming our way.
1:16
Speaker 1: I'm Jacob Goldstein and this is What's Your Problem, the
1:18
Speaker 1: show where I talk to people who are trying to
1:20
Speaker 1: make technological progress. My guest today is Kai Marshland. He's
1:25
Speaker 1: the co founder and chief product officer of Windborne Systems.
1:29
Speaker 1: Kai's problem is this, can you build weather balloons that
1:33
Speaker 1: stay in the air for weeks or months instead of hours,
1:36
Speaker 1: And can you pair the data from those balloons with
1:39
Speaker 1: AI to make weather forecasts that are way better than
1:42
Speaker 1: anything we have today. Kai got interested in balloons back
1:45
Speaker 1: in twenty fifteen. He was a freshman at Stanford and
1:49
Speaker 1: he joined the Student Space Initiative, which sounds kind of
1:52
Speaker 1: fancy but in fact was basically a bunch of college
1:55
Speaker 1: kids trying to build their own weather balloons.
2:02
Speaker 2: The first half dozen flights of our balloons were complete
2:06
Speaker 2: and total failures.
2:08
Speaker 1: When do things start working?
2:10
Speaker 2: That would be back in June of twenty sixteen. This
2:15
Speaker 2: is the end of my freshman year, and I think
2:18
Speaker 2: our longest flight that year had been fourteen minutes, so
2:23
Speaker 2: shorter than a normal weather balloon.
2:27
Speaker 1: Buy a lot, Buy a lot.
2:29
Speaker 2: And also the person who had started this project, who
2:35
Speaker 2: then became my co founder, had just graduated and was
2:39
Speaker 2: about to leave what we thought never to be seen again.
2:44
Speaker 2: And school has ended. We can't even get into the
2:48
Speaker 2: building because our key cards have stopped working, so we
2:52
Speaker 2: have to tailgate someone in to get our balloon launch
2:54
Speaker 2: equipment out. Go out to the middle of nowhere in
2:58
Speaker 2: the central Valley of California, and if this launch didn't succeed,
3:05
Speaker 2: windborne would not exist.
3:07
Speaker 1: Okay, so I can guess what happened next based on
3:11
Speaker 1: that fact. But tell me what happened.
3:13
Speaker 2: Yeah. So it got to be about four am, and
3:19
Speaker 2: we're out in the middle of this park, really cold,
3:22
Speaker 2: and the sprinklers turn on. They start slowly swiveling around
3:27
Speaker 2: towards our delicate electronics, and so I have to hurl
3:33
Speaker 2: myself onto the sprinklers, getting soaked, just to save our electronics.
3:37
Speaker 1: It's like the nerd version of throwing yourself on a
3:39
Speaker 1: grenade exactly.
3:40
Speaker 2: I really felt like I was in an action movie sequence.
3:43
Speaker 2: Ah good, But that sacrifice really paid off because with
3:50
Speaker 2: that flight, we managed to set the world record for
3:54
Speaker 2: weather balloon endurance.
3:56
Speaker 1: How long did it stay up?
3:57
Speaker 2: It stayed up for seventy six hours?
4:00
Speaker 1: Wow, and your previous best.
4:02
Speaker 2: Was fourteen minutes exactly.
4:04
Speaker 1: So you're in college, how do you build weather balloons
4:06
Speaker 1: that are better than whatever the federal government? Noah's launching
4:10
Speaker 1: our you know, state of the art weather balloon.
4:12
Speaker 2: People. So the first big thing is that some new
4:17
Speaker 2: consumer electronics have come out, in particular, things like micro
4:23
Speaker 2: controllers and lightweight, low power satellite communications, and we realize
4:30
Speaker 2: that we can use these innovations to make a balloon
4:35
Speaker 2: smarter and to control its altitude.
4:38
Speaker 1: What's a micro controller.
4:39
Speaker 2: Micro controller is a super small computer that you can
4:45
Speaker 2: put in any device to give it a brain.
4:49
Speaker 1: And they're cheap.
4:50
Speaker 2: They're dirt cheap. Anytime you hear the word Internet of things,
4:54
Speaker 2: that means micro controlers.
4:56
Speaker 1: And nobody else was doing this at the time, Like
4:59
Speaker 1: all the big, you know, well funded weather agencies weren't
5:02
Speaker 1: sticking my micro controllers on their on their weather balloons.
5:06
Speaker 2: Well, with these big agencies, they are really incentivized to
5:12
Speaker 2: keep things stable because weather is this really important thing
5:16
Speaker 2: that billions of people depend on, and so they're not
5:20
Speaker 2: going to go and say, hey, what crazy experiments can
5:24
Speaker 2: we run.
5:24
Speaker 1: Their incentive is just like, don't screw it up exactly
5:27
Speaker 1: the way we did it yesterday. Don't screw it up
5:29
Speaker 1: and you won't get fired exactly. And so specifically, what
5:33
Speaker 1: was it that you were doing that no one had
5:35
Speaker 1: done before.
5:37
Speaker 2: Yeah, So it's a very simple concept that is really
5:41
Speaker 2: hard to make work and practice. The way we control
5:45
Speaker 2: our altitude is when it gets too high up, we
5:49
Speaker 2: vent some of our gas so it has less lift.
5:53
Speaker 2: When we fall down too far, we drop some ballast
5:56
Speaker 2: so it stops falling. But the hard part about this
6:00
Speaker 2: is it has to function at negative seventy degrees celsius.
6:04
Speaker 2: That's the temperature of dry ice. You have to have
6:08
Speaker 2: a lot of on board software. That's where the micro
6:11
Speaker 2: controllers come in to do things like understand is this
6:16
Speaker 2: just turbulence that you're hitting or is it actually a
6:19
Speaker 2: real change in lift that we need to drop ballast
6:21
Speaker 2: or vent gas to account for. And in fact, we
6:25
Speaker 2: stayed up twenty four to seven. Every single time it
6:28
Speaker 2: talked to us. Every five minutes, we had it play
6:32
Speaker 2: an airhorn sound because we were, of course falling asleep
6:35
Speaker 2: in the middle of them to see what it was doing.
6:39
Speaker 1: So, okay, so you set the record, you stay up
6:41
Speaker 1: for what three days watching it? How far did it go?
6:46
Speaker 2: It flew all the way across the country and landed
6:50
Speaker 2: out in the Atlantic Ocean. But the real reason why
6:55
Speaker 2: we decided to form a company around this was because
6:59
Speaker 2: we realized the impact this could have. We got into
7:03
Speaker 2: this from the engineering side. The lands are fun, but
7:07
Speaker 2: we realized, wait a secd eighty five percent of the
7:10
Speaker 2: Earth's atmosphere is invisible to humanity. Weather is this crazy,
7:15
Speaker 2: unpredictable thing, and we can solve that. And weather is
7:20
Speaker 2: so much more than just do you bring an umbrella
7:23
Speaker 2: to work tomorrow is the single most immediately destructive aspect
7:29
Speaker 2: of climate change, and improving the weather forecast it of
7:34
Speaker 2: course helps with adapting to climate change, things like predicting
7:38
Speaker 2: where a hurricane will make landfall, but it can also
7:41
Speaker 2: help with preventing climate change in the first place. If
7:45
Speaker 2: you have a better weather forecast, you can better route
7:49
Speaker 2: ships and planes to save a huge amount of fuel.
7:53
Speaker 2: You can accelerate the transition to renewables because you know
7:56
Speaker 2: when the wind will be blowing or the sun will
8:00
Speaker 2: be shining. And we looked at this and said, no
8:03
Speaker 2: one else has our balloon technology. If we don't do this,
8:06
Speaker 2: no one will. And if we want away from this opportunity,
8:10
Speaker 2: we can't live with ourselves.
8:13
Speaker 1: I mean, I buy that the stakes are high. I
8:15
Speaker 1: buy that it could be very helpful, but like, why
8:19
Speaker 1: wouldn't anybody else do it?
8:22
Speaker 2: Yeah, a lot of people are trying to improve the
8:26
Speaker 2: weather forecast. But what's unique about Windborne is the combination
8:33
Speaker 2: of our balloon technology with our AI weather forecasts, because
8:38
Speaker 2: really those are the two levers to pull to increase
8:41
Speaker 2: the accuracy of a weather forecast.
8:44
Speaker 1: You mentioned that eighty five percent of the world lacks
8:48
Speaker 1: good weather data, which was surprising to me the first
8:52
Speaker 1: time I heard it. But I would think, like, get
8:54
Speaker 1: satellites do it. That's kind of I guess my naive
8:58
Speaker 1: ish thought. It's like, I know there's a lot of
8:59
Speaker 1: satellites looking down at the Earth all the time.
9:02
Speaker 2: I'm actually a big fan of the space industry. I
9:04
Speaker 2: used to work at SpaceX. But the problem is the
9:07
Speaker 2: laws of physics fundamental limit what satellites can measure. Take
9:12
Speaker 2: for example, pressure. It turns out pressure is extremely important
9:17
Speaker 2: for predicting the weather because it determines where the winds
9:20
Speaker 2: are blowing, how the weather systems are moving. But a
9:24
Speaker 2: satellite fundamentally cannot measure pressure from space because it's not
9:29
Speaker 2: in the atmosphere. It can't see what's going on.
9:32
Speaker 1: So you start the company you're good at making weather
9:36
Speaker 1: balloons that go up and stay up for a long time,
9:38
Speaker 1: and that you can steer. I also read that you
9:41
Speaker 1: launched thinking that you could collect data for a tenth
9:45
Speaker 1: the cost of existing alternatives, which sounds compelling, but as
9:50
Speaker 1: I understand was not compelling enough. Is that right?
9:54
Speaker 2: It is. We're now at one hundred and fifty times
9:57
Speaker 2: more data per dollar than alternatives.
9:59
Speaker 1: Okay, was there some reason one tenth the cost was
10:02
Speaker 1: not cheap enough.
10:04
Speaker 2: Yeah. When we started the company, we thought it would
10:07
Speaker 2: be cheap enough. But it turns out when you really
10:11
Speaker 2: look at the scale of weather data collection, you need
10:14
Speaker 2: to really understand the atmosphere. You want ten thousand balloons
10:21
Speaker 2: aloft concurrently, that's a balloon every sixty miles in the atmosphere.
10:27
Speaker 2: And in order to get to that level of scale,
10:32
Speaker 2: it's the difference between your company spending one hundred million
10:36
Speaker 2: dollars a year and spending a billion dollars a year.
10:39
Speaker 2: One was a lot more in reach.
10:41
Speaker 1: How do you drive down the cost to one one
10:43
Speaker 1: hundred and fiftieth the cost? I mean, I'm sure it's
10:45
Speaker 1: many many, many incremental you know, efficiency gains. But like,
10:50
Speaker 1: what's an example of one.
10:51
Speaker 2: Yeah, Well, one of the big pieces is improving the
10:56
Speaker 2: software that flies on it so that the balloon can
11:00
Speaker 2: fly for longer. I talked about in the student group
11:03
Speaker 2: that first flight lasting seventy six hours. Now our longest
11:07
Speaker 2: flight can fly for over four days.
11:10
Speaker 1: Wow.
11:11
Speaker 2: And the longer you fly, well, the hardware costs stays
11:15
Speaker 2: the same, so that means you're collecting a lot more data.
11:18
Speaker 1: Yeah. Yeah. So basically, make the balloon stamp for longer
11:21
Speaker 1: is the fundamental way that you make it cheaper exactly,
11:25
Speaker 1: And how do you make it fly longer?
11:27
Speaker 2: Yeah, Well, one of the big things is improving the
11:31
Speaker 2: software to better decide when do you vent gas, when
11:35
Speaker 2: do you drop ballast or really one of the other
11:38
Speaker 2: big things is just making everything smaller and lighter, because
11:43
Speaker 2: the smaller it is, the lighter it is, the less
11:46
Speaker 2: chance of a leak, the less ballast you have to use.
11:50
Speaker 2: And things have just shrunk down so far. It really
11:56
Speaker 2: surprises me sometimes when I'm like, wait, that tiny thing
12:00
Speaker 2: the size of a dime replaces the thing that was
12:03
Speaker 2: the size of a dinner plate before.
12:06
Speaker 1: Yeah, that's amazing. So okay, so let's talk about where
12:10
Speaker 1: you are today. Let's talk about sort of how it works.
12:13
Speaker 1: How big are the balloons?
12:15
Speaker 2: So the balloon is two pieces. There's the envelope that's
12:20
Speaker 2: the bag that holds all parts the balloon part of
12:24
Speaker 2: the balloon, and then there's the main unit, which is
12:30
Speaker 2: the electronics and the ballast.
12:33
Speaker 1: The stuff exactly, the balloon and the stuff the balloon part.
12:36
Speaker 2: Of the balloon. The balloon part of the balloon is
12:39
Speaker 2: five and a half meters tall. So that's two and
12:43
Speaker 2: a half three times the height of person, depending on
12:45
Speaker 2: how tall you are.
12:47
Speaker 1: Yeah, so tall. Yeah, it's a big balloon. And how
12:52
Speaker 1: how big is the stuff? How big is are the
12:54
Speaker 1: sensors in the part that's not the.
12:56
Speaker 2: Yeah, that stuff is. I guess this is an audio thing,
13:01
Speaker 2: so I would say about this big.
13:06
Speaker 1: The size of a basketball.
13:07
Speaker 2: Yeah, it's a roughly the size of a basketball. It
13:12
Speaker 2: weighs just over four pounds, so about the weight of
13:16
Speaker 2: a large duck. Okay, and is kind of long and skinny.
13:22
Speaker 2: It kind of looks like a fish. I think of
13:25
Speaker 2: it as a trout attached to a giant bag.
13:30
Speaker 1: How many of your balloons are in the sky right now? About?
13:34
Speaker 2: Yeah, there are a few dozen aloft right now.
13:37
Speaker 1: Okay.
13:37
Speaker 2: We launch around one hundred a month and are quickly
13:40
Speaker 2: ramping that up.
13:41
Speaker 1: So where do you launch your balloons from?
13:43
Speaker 2: We launch them every day from three continents, South Korea,
13:47
Speaker 2: paloelto New York and Cabo Verde.
13:50
Speaker 1: Cabo Verde is just off off the west coast of Africa.
13:53
Speaker 1: Is that right?
13:53
Speaker 2: You got it. One of the reasons why we have
13:56
Speaker 2: launch site in Cabo Verde right now. That's right where
13:59
Speaker 2: a lot of hurricanes are forming, and so the fact
14:02
Speaker 2: that we're collecting data around there is going to be
14:06
Speaker 2: really impactful for better predicting the path of these hurricanes. Huh.
14:10
Speaker 1: So they start there and then they essentially travel across
14:13
Speaker 1: the Atlantic and into the Americas, and if you can
14:16
Speaker 1: understand what's going on there, ideally you can sort of
14:19
Speaker 1: understand the hurricane where it's going to go in the
14:22
Speaker 1: dream scenario, even just as it's becoming a tropical storm.
14:26
Speaker 2: Exactly. Yeah, Noah did an analysis of the Impactor data
14:30
Speaker 2: had on the twenty twenty two hurricane season, and it
14:36
Speaker 2: made the forecasts for Hurricane Fiona about twenty percent better.
14:43
Speaker 1: Now it's time for a few ads. After the ads,
14:46
Speaker 1: Kai and I will talk about AI. Of course, we'll
14:50
Speaker 1: also talk about Winborne's business, what they're selling, and we'll
14:54
Speaker 1: talk about the company's quest to build balloons that can
14:56
Speaker 1: stay in the air for months at a time and
14:59
Speaker 1: make multiple trips around the world. What is your business
15:16
Speaker 1: right now, like, what are you selling and who you're
15:18
Speaker 1: selling it.
15:18
Speaker 2: To right now. Our business has two pieces to it.
15:22
Speaker 2: We are selling the observations of the atmosphere we collect
15:26
Speaker 2: to the government right now, so that those governments can
15:31
Speaker 2: use this data to improve your weather forecast. And just
15:36
Speaker 2: a couple of weeks ago, we announced our AI based
15:40
Speaker 2: weather model, which is the world's most accurate global weather
15:43
Speaker 2: model bar none.
15:45
Speaker 1: I feel like you need a little TM when you
15:46
Speaker 1: say the world's most accurate global weather model bar none.
15:50
Speaker 2: We really do, we really do.
15:52
Speaker 1: So let's talk about that. Is that claim validated? I mean,
15:54
Speaker 1: I know there was news about that just this year,
15:57
Speaker 1: and congratulations, but also, how do I know that's true? Respectfully?
16:01
Speaker 2: Yeah, I love that question because anytime a weather company
16:06
Speaker 2: makes a claim, you should look at it closely. So,
16:10
Speaker 2: first off, we've published our results on our website and
16:15
Speaker 2: so you can dig into some of the raw numbers there. Second,
16:18
Speaker 2: we're in the process of submitting to weather bench which
16:23
Speaker 2: is these benchmarks run by Google, which were the previous
16:29
Speaker 2: holders of this record. So we're actively talking with them
16:33
Speaker 2: right now about submitting our results getting them validated into
16:38
Speaker 2: live up there. It's just a process that takes a
16:41
Speaker 2: little while, so.
16:43
Speaker 1: Let's talk about the model you built. So I mean
16:46
Speaker 1: generally when I talk to people who are working on AI,
16:49
Speaker 1: they talk a lot about data, right, And it seems
16:51
Speaker 1: like to some significant degree the models are kind of commoditized,
16:57
Speaker 1: not quite, but like models are pretty similar, it seems
17:00
Speaker 1: in many settings at least one to the other, and
17:02
Speaker 1: the differentiator often ends up being the data. Is that
17:06
Speaker 1: the case in this instance? I mean, do you think
17:08
Speaker 1: you're the best. If you're the best because you have
17:10
Speaker 1: all this data from all your balloons.
17:14
Speaker 2: I think that you're spot on. Data is the real
17:16
Speaker 2: mote here, because what really sets us apart is the
17:20
Speaker 2: ability to have all of this data that no one
17:24
Speaker 2: else has and to use that both for training the
17:28
Speaker 2: model but then also for running the model in real time.
17:32
Speaker 2: Data isn't just about training. It's about essentially giving your
17:35
Speaker 2: model the prompt of what's going to happen next, and
17:40
Speaker 2: you need new data every single day for that.
17:44
Speaker 1: Just one thing to clarify. So, if you're selling the
17:48
Speaker 1: data to governments, are they not sharing it widely? Like
17:53
Speaker 1: they're paying for the data but they're not. But not
17:56
Speaker 1: everybody can get at the data? Is that why it's
17:58
Speaker 1: a mode for you?
17:59
Speaker 2: Yeah, so what they're using the data we sell them
18:03
Speaker 2: for is putting it into conventional physics based weather model
18:09
Speaker 2: that they then release. So they can't redistribute the data itself,
18:15
Speaker 2: but the general public can benefit from its use in
18:20
Speaker 2: these weather models.
18:21
Speaker 1: And so just to distinguish and this is a distinction
18:24
Speaker 1: that's not particular to your company, but you mentioned the
18:27
Speaker 1: sort of traditional physics based weather models. I mean, I
18:30
Speaker 1: think it's worth spending just a moment here to distinguish, right,
18:33
Speaker 1: like between the classic weather model, you know, the sort
18:39
Speaker 1: of pre AI weather model, and the AI weather model. Like,
18:43
Speaker 1: what's the basic difference between those two?
18:45
Speaker 2: Yeah, So a conventional physics based weather model, it takes
18:49
Speaker 2: the initial state of the atmosphere and then runs a
18:53
Speaker 2: bunch of fluid dynamics on it to simulate what's going
18:59
Speaker 2: to happen to all of these different fluids.
19:01
Speaker 1: There's a sort of kind of classic sort of feels
19:04
Speaker 1: like kind of nineteenth century deterministic. Give me the initial
19:08
Speaker 1: condition and I'll tell you what's going to happen in
19:10
Speaker 1: the future. It's that right, it's sort of rules based.
19:13
Speaker 2: Exactly exactly, and that is nice because you can see
19:21
Speaker 2: the exact physics that is being used, but you need
19:26
Speaker 2: compute clusters. They cost hundreds of millions of dollars in
19:30
Speaker 2: order to run this because the atmosphere is so big.
19:34
Speaker 2: And by contrast, a AI based weather model can run
19:39
Speaker 2: on just a gaming laptop.
19:42
Speaker 1: Huh.
19:43
Speaker 2: And what it does, by contrast, is effectively picking out statistics.
19:49
Speaker 1: I mean, it's just it's machine learning. I presume it's
19:51
Speaker 1: pattern matching exactly.
19:53
Speaker 2: It's our model is a transformer based architecture, the same
19:57
Speaker 2: thing that powers chat GPT.
19:59
Speaker 1: Our A models in general better at this point than
20:02
Speaker 1: the physics based traditional models.
20:04
Speaker 2: It depends on the use case. But if you're talking
20:07
Speaker 2: about hurricanes, yeah, we ran a case study on Hurricane
20:12
Speaker 2: Ian and our model would have predicted landfall by two
20:19
Speaker 2: hundred kilometers closer.
20:21
Speaker 1: I mean, that's another one where like I'm intrigued by
20:24
Speaker 1: what you said, but I'm very eager for I'm not
20:27
Speaker 1: eager for this hurricane season to come, but like it
20:31
Speaker 1: will resonate more with me when you do that prospectively, right,
20:34
Speaker 1: of course, I mean, will you this year be predicting
20:39
Speaker 1: where hurricanes are going to make landfall?
20:41
Speaker 2: We will be predicting that for public safety reasons. We
20:45
Speaker 2: aren't going to be you won't tell me.
20:48
Speaker 1: Yeah you will, but you won't tell me we.
20:49
Speaker 2: Will know a hurricane. Okay, but fair, you of course
20:55
Speaker 2: don't want to say, hey, don't listen to Noah about
20:58
Speaker 2: these evacuation orders.
21:01
Speaker 1: No, no, fair that that is very responsible. So basically,
21:05
Speaker 1: you're selling data. You are likely soon to be selling forecasts.
21:10
Speaker 1: That's sort of the business. On the technical side, what
21:15
Speaker 1: is the frontier? What are you trying to figure out
21:17
Speaker 1: that you haven't figured out yet?
21:19
Speaker 2: Yeah, well, one of the big areas for innovation is
21:23
Speaker 2: on the AI modeling side. We were kind of surprised
21:28
Speaker 2: that our models did as well as they did because
21:32
Speaker 2: there are a lot of quite frankly obvious things that
21:36
Speaker 2: we haven't done, things like just increasing the amount of
21:40
Speaker 2: compute we're using to train these models. We use something
21:44
Speaker 2: like a fifteenth as much as Google did. So what
21:49
Speaker 2: happens when you train it for longer?
21:51
Speaker 1: I mean, I'll say, on the AI side, it's not
21:54
Speaker 1: obvious that it's optimal for you to be doing the
21:57
Speaker 1: AI as well. Right, there's a universe where you are
22:00
Speaker 1: optimizing the balloons and the data, and then someone else
22:05
Speaker 1: like Google, who has all the computers and all the
22:08
Speaker 1: AI knows what to do with doing the AI side.
22:12
Speaker 2: Of it, right, Yes, and no, I think that where
22:17
Speaker 2: that falls apart is coupling our data with the models
22:21
Speaker 2: much more closely. In that there's so much to be
22:24
Speaker 2: done in terms of figuring out how to better take
22:28
Speaker 2: advantage of our data into particular and also things like saying,
22:34
Speaker 2: based on this weather model, where should we be flying
22:37
Speaker 2: our balloons to improve the forecast? And we're already using
22:41
Speaker 2: our own AI weather forecasts to do that flight plan
22:45
Speaker 2: optimization to figure out where our balloons are going to fly.
22:49
Speaker 2: So it's really this beneficial effect where the better our
22:52
Speaker 2: weather forecast, the better we can fly our balloons. And
22:56
Speaker 2: we can then target our balloons to fly to the
23:00
Speaker 2: places that will most improve the weather forecast.
23:03
Speaker 1: That's a good feedback loop.
23:04
Speaker 2: There's also a lot to do on the balloon side
23:06
Speaker 2: of things. I wish I could tell you about our
23:11
Speaker 2: project that will increase flight time by another factor.
23:16
Speaker 1: Of ten, another factor of ten.
23:19
Speaker 2: Yeah, so what what are you at now?
23:21
Speaker 1: What's the like median flight time?
23:23
Speaker 2: Medium flight time is seven to ten days depending on
23:29
Speaker 2: how we're targeting it.
23:30
Speaker 1: And so you think you're going to get to ninety days. Yep,
23:34
Speaker 1: we around the world in eighty days. Yeah, you say
23:38
Speaker 1: you wish you could tell me like you feel like
23:40
Speaker 1: you're about to do it. Yeah, tell me, tell me
23:43
Speaker 1: without telling me what you can't tell me.
23:44
Speaker 2: Okay, I'll do my best. So we have had some
23:50
Speaker 2: very successful flight tests where we have demonstrated the ability
23:57
Speaker 2: to fly without using any finite resources.
24:03
Speaker 1: So what what is the key finite resource? The the
24:06
Speaker 1: lift gas, the what is it helium or hydrogen?
24:08
Speaker 2: What you used to keep Our balloons are compatible with
24:13
Speaker 2: both helium and hydrogen, So which we use is dependent
24:16
Speaker 2: on the country.
24:17
Speaker 1: When you say flying without any finite resource, that's what
24:20
Speaker 1: I think of, yep.
24:22
Speaker 2: Okay, yeah, that and ballast.
24:25
Speaker 1: Oh right, So so you want to you can? I should?
24:30
Speaker 1: I guess.
24:31
Speaker 2: I mean, there's heat there, leave it there and just say.
24:37
Speaker 1: If you get that, could you fly forever?
24:40
Speaker 2: That the dream someday, someday we're we're definitely ninety.
24:45
Speaker 1: Days for a weather balloon is sort of forever, right.
24:47
Speaker 2: It really is. And so, yeah, we've had some very
24:51
Speaker 2: successful flight tests.
24:53
Speaker 1: What's the longest flight you've had to this point?
24:55
Speaker 2: Forty days?
24:57
Speaker 1: Forty days? Okay?
24:58
Speaker 2: Long?
24:58
Speaker 1: Yeah?
24:59
Speaker 2: Long? And that was without using this technology, was that
25:02
Speaker 2: like a fluke. It was definitely on the side of
25:07
Speaker 2: the bell curve. But we've had multiple month long flights.
25:11
Speaker 1: What why do your balloons usually what do you say fall?
25:16
Speaker 1: Crash sounds a little extreme crash fall.
25:19
Speaker 2: Yeah, it's that they run out of ballast and lifting gas,
25:24
Speaker 2: and so these finite resources have been used up. You
25:29
Speaker 2: don't have a more ballast to drop, you don't have
25:32
Speaker 2: a more gas event, and so you can no longer
25:35
Speaker 2: control your altitude.
25:37
Speaker 1: What happens when it falls crashes comes down.
25:41
Speaker 2: So by the time it falls, it has used up
25:44
Speaker 2: all of the ballast. So instead of weighing four pounds,
25:50
Speaker 2: it weighs about two hundred grams.
25:56
Speaker 1: Two hundred grams is what half a pound?
25:58
Speaker 2: Ish? Yes, I'd looked it up. Zero point four to
26:03
Speaker 2: four pounds.
26:04
Speaker 1: Okay. By the way, when you're getting rid of ballast,
26:06
Speaker 1: is it just like sand or something? You just like
26:08
Speaker 1: sprinkle sand out exactly? Okay, So it's half a pound,
26:14
Speaker 1: so it wouldn't hurt if it hit me on the head?
26:17
Speaker 1: Is that is that part of the reason half a
26:19
Speaker 1: pound is important? Would it if it hit me on that?
26:21
Speaker 2: Is it? Anybody on that it's never hit anybody. We
26:25
Speaker 2: do landing simulations and control where it lands to direct
26:29
Speaker 2: it towards unpopulated areas. But remember it also has this
26:36
Speaker 2: envelope attached to it, which acts like a parachute, so
26:40
Speaker 2: it actually falls very slowly. It's very light and not
26:45
Speaker 2: an issue.
26:46
Speaker 1: And then what happens. Then you have this big deflated
26:49
Speaker 1: balloon sitting on the ground or floating in the ocean,
26:51
Speaker 1: and is it just what happens to you?
26:54
Speaker 2: Yeah? So I love that question because it's a chance
26:58
Speaker 2: to talk about how our balloons can reduce the amount
27:02
Speaker 2: of waste going out into the world compared to the
27:05
Speaker 2: half a million that are launched every year, of which
27:08
Speaker 2: only a fifth are covered.
27:10
Speaker 1: Just get answer directly. I appreciate that you want to
27:12
Speaker 1: contextualize it, but like, first, let's talk about what happens
27:15
Speaker 1: with your balloons and then feel free to provide the bigger.
27:18
Speaker 2: Cart sounds good, sounds good? Yeah, yeah, So our balloons
27:23
Speaker 2: are out in the world, and we recover as many
27:27
Speaker 2: of them as we can. In the long term, we're
27:30
Speaker 2: aiming to recover essentially all of them because we'll direct
27:34
Speaker 2: them to specific landing sites.
27:36
Speaker 1: I mean, the ocean has got to be tough to write.
27:40
Speaker 1: I presume when they fall in the ocean, for the
27:41
Speaker 1: most part, you don't recover them.
27:42
Speaker 2: Is that right, exactly? And that's one of the reasons
27:45
Speaker 2: why improving endurance is so exciting, because if you circumnavigate
27:49
Speaker 2: multiple times, well, on the last leg, you bring it
27:52
Speaker 2: down in a field, have somebody drive around and collect
27:56
Speaker 2: all of them. They are about half million weather balloons
27:59
Speaker 2: launched each year, and only a fifth of those are recovered.
28:04
Speaker 2: So when each balloon flies for only two hours, well,
28:09
Speaker 2: you have to launch a lot more balloons to collect
28:12
Speaker 2: the same amount of data. So we really see this
28:15
Speaker 2: as an opportunity to reduce the amount of stuff that's
28:18
Speaker 2: going out there in the first place.
28:20
Speaker 1: Right, So there will actually be fewer balloons going into
28:25
Speaker 1: the world if you succeed.
28:26
Speaker 2: Exactly.
28:28
Speaker 1: So, you're at this point where you've learned a lot.
28:30
Speaker 1: You sort of feels like you're kind of on the
28:32
Speaker 1: precipice of a lot more. And I want to talk
28:35
Speaker 1: about sort of two futures. Right. One is a future
28:38
Speaker 1: where where your company doesn't succeed for any number of
28:42
Speaker 1: reasons with which you're truly more familiar than I. And
28:45
Speaker 1: then the other is if you do succeed. Right, So,
28:48
Speaker 1: in the sort of sad version where you don't succeed,
28:50
Speaker 1: what are some reasons it might not work out?
28:52
Speaker 2: Yeah, I think that probably the biggest reason is that
28:57
Speaker 2: scaling up hardware is hard. We need to increase manufacturing
29:02
Speaker 2: by a factor of one hundred effectively, and.
29:06
Speaker 1: So it's scaling up that final assembly by a factor
29:08
Speaker 1: of one hundred is complicated and hard and just kind
29:12
Speaker 1: of classic hard building. A business gets more capital intensive exactly,
29:18
Speaker 1: and the money before you can get the money exactly. Okay,
29:21
Speaker 1: So that's an easy to imagine sad outcome. Let's talk
29:26
Speaker 1: about the happy outcome. Like it works. You scale up
29:31
Speaker 1: your balloons, stay up for months at a time. What's
29:35
Speaker 1: the world look like? What are you doing in that scenario.
29:38
Speaker 2: We want everybody to see twice as far into the
29:41
Speaker 2: future when it comes to weather. So making the ten
29:44
Speaker 2: day forecast is accurate is the current five day forecast.
29:48
Speaker 2: Making the twenty day forecast as accurate as the ten
29:50
Speaker 2: day forecast, so we want people to see twice as
29:54
Speaker 2: far into the future. We want to pinpoint where hurricanes
29:57
Speaker 2: are going to make landfall a week in advance, and
30:02
Speaker 2: we want day to day weather that businesses rely on
30:07
Speaker 2: to be really accurate, things like never having to cancel
30:15
Speaker 2: a flight last minute because a bomb cyclone popped up,
30:20
Speaker 2: saving fuel with all of your shipping because you know
30:24
Speaker 2: where there will be headwinds. We want to accelerate the
30:29
Speaker 2: transition to renewables, and we want weather to go from
30:33
Speaker 2: this crazy, unpredictable source of uncertainty to something that humans
30:39
Speaker 2: just know about.
30:42
Speaker 1: We'll be back in a minute with the lightning round. Okay,
30:56
Speaker 1: let's do the lightning round. Would you rather have it
31:01
Speaker 1: be too hot or too cold?
31:03
Speaker 2: Too cold any day?
31:06
Speaker 1: Okay? The next one is multiple choice. What is your
31:10
Speaker 1: favorite movie prominently featuring balloons? Wizard of Oz around the
31:15
Speaker 1: World in eighty days, up the Red Balloon, or none
31:19
Speaker 1: of the above.
31:20
Speaker 2: It's got to be up brilliant.
31:21
Speaker 1: Movie, Blimps, overrated or underrated?
31:27
Speaker 2: Ooh, so here, I've got to make a distinction between
31:31
Speaker 2: blimps and Zeppelins because I actually used to work at
31:34
Speaker 2: a zeppelin company.
31:37
Speaker 1: Is that the Larry Page Zeppelin company? It sure is
31:41
Speaker 1: perhaps the only zeppelin company. Yeah, and so is the
31:46
Speaker 1: distinction that a zeppelin has a rigid frame?
31:49
Speaker 2: Got it in? Won?
31:51
Speaker 1: Okay? So how about this. Let me reformulate it to
31:54
Speaker 1: see if I get it this time. Airships overrated or underrated?
31:59
Speaker 2: Underrated? I thought that zeppelins are so cool. I don't
32:06
Speaker 2: have a great mission driven answer here other than you know,
32:11
Speaker 2: I think it's amazing. I read too many books growing
32:14
Speaker 2: up which prominently featured airships.
32:18
Speaker 1: Isn't the isn't the Larry Page airship company dreams some
32:22
Speaker 1: kind of cargo, like the idea that, for like really
32:25
Speaker 1: heavy things, giant airships would be an efficient way to
32:28
Speaker 1: move cargo.
32:29
Speaker 2: Full disclosure, It's been a long time since I worked there,
32:32
Speaker 2: so I can't speak to that company in particular. But yeah,
32:38
Speaker 2: faster deliveries than a ship, more efficient deliveries than a plane.
32:43
Speaker 1: Aha, It's like a niche an Until what's something besides
32:49
Speaker 1: the weather that AI will be really good at predicting?
32:53
Speaker 2: Oh? Great question, let's see. Uh. I don't have a
33:02
Speaker 2: good answer off the top of my head. I think
33:05
Speaker 2: that it will be really interesting for economics in terms
33:11
Speaker 2: of taking short term data, various real time indicators, and
33:18
Speaker 2: making predictions about what the final readings are going to be.
33:24
Speaker 1: I'm sure a lot of very smart people are working
33:27
Speaker 1: on that and they may in fact already have good
33:29
Speaker 1: models and they're not telling us.
33:31
Speaker 2: I think that's likely.
33:32
Speaker 1: Is there anything else you want to say?
33:35
Speaker 2: I think that covers it. Yeah, great to be on
33:37
Speaker 2: the show.
33:39
Speaker 1: Thanks for your time. It was lovely to talk with you.
33:41
Speaker 1: Good luck.
33:41
Speaker 2: Thanks and if you ever are in Paloelto, you're welcome
33:46
Speaker 2: to come see a balloon launch.
33:48
Speaker 1: Great. So do you still you launch from where you launch?
33:53
Speaker 1: From that Air Force base or like or just from.
33:56
Speaker 2: From our part part?
33:57
Speaker 1: Are they easy? Yeah? From the parking lot, that's cool.
33:59
Speaker 2: Yeah.
34:00
Speaker 1: How much space do you need to launch a weather balloon?
34:03
Speaker 2: Not that much? About fifty by fifty feet. It's really
34:08
Speaker 2: just a matter of making sure there's nothing that the
34:10
Speaker 2: balloon will blow into right after releasing it.
34:13
Speaker 1: Uh huh. Seems like it'll be fun to see. Yeah,
34:17
Speaker 1: I like a balloon. Kai Marshland is the co founder
34:23
Speaker 1: and chief product officer of Windborne Systems. Today's show was
34:27
Speaker 1: produced by Gabriel Hunter Chang and edited by Lydia Jean Kott,
34:31
Speaker 1: who was engineered by Sarah Buguer. You can email us
34:34
Speaker 1: at problem at Pushkin dot FM. I'm Jacob Goldstein and
34:38
Speaker 1: we'll be back next week with another episode of What's
34:41
Speaker 1: Your Problem,