1
00:00:15,356 --> 00:00:22,996
Speaker 1: Pushkin. Every day around the world, more than a thousand

2
00:00:23,036 --> 00:00:26,356
Speaker 1: weather balloons are launched into the sky. The balloons float

3
00:00:26,436 --> 00:00:30,636
Speaker 1: high up into the atmosphere, sending back information you know, temperature,

4
00:00:30,996 --> 00:00:34,716
Speaker 1: wind speed, air pressure, et cetera, and then a couple

5
00:00:34,796 --> 00:00:39,156
Speaker 1: hours later, the balloons pop. This is basically the way

6
00:00:39,156 --> 00:00:42,316
Speaker 1: weather balloons have worked for decades, and the information the

7
00:00:42,316 --> 00:00:45,876
Speaker 1: balloons send back is really useful for weather forecasts. But

8
00:00:45,996 --> 00:00:49,196
Speaker 1: the information the balloon send back is also pretty limited

9
00:00:49,556 --> 00:00:51,516
Speaker 1: because the balloons only stay in the sky for a

10
00:00:51,516 --> 00:00:54,916
Speaker 1: couple hours, they don't fly very far. If we could

11
00:00:54,956 --> 00:00:57,916
Speaker 1: figure out how to make balloons stay up for longer,

12
00:00:58,356 --> 00:01:01,516
Speaker 1: they could blow in the wind and travel thousands of miles.

13
00:01:01,796 --> 00:01:05,356
Speaker 1: They could travel across continents and across oceans and send

14
00:01:05,396 --> 00:01:07,756
Speaker 1: us back a lot more data and give us a

15
00:01:07,836 --> 00:01:10,796
Speaker 1: much clearer picture of what weather is coming our way.

16
00:01:16,516 --> 00:01:18,756
Speaker 1: I'm Jacob Goldstein and this is What's Your Problem, the

17
00:01:18,796 --> 00:01:20,716
Speaker 1: show where I talk to people who are trying to

18
00:01:20,756 --> 00:01:25,396
Speaker 1: make technological progress. My guest today is Kai Marshland. He's

19
00:01:25,476 --> 00:01:28,836
Speaker 1: the co founder and chief product officer of Windborne Systems.

20
00:01:29,396 --> 00:01:33,116
Speaker 1: Kai's problem is this, can you build weather balloons that

21
00:01:33,196 --> 00:01:36,076
Speaker 1: stay in the air for weeks or months instead of hours,

22
00:01:36,556 --> 00:01:39,116
Speaker 1: And can you pair the data from those balloons with

23
00:01:39,196 --> 00:01:42,196
Speaker 1: AI to make weather forecasts that are way better than

24
00:01:42,236 --> 00:01:45,876
Speaker 1: anything we have today. Kai got interested in balloons back

25
00:01:45,876 --> 00:01:49,156
Speaker 1: in twenty fifteen. He was a freshman at Stanford and

26
00:01:49,196 --> 00:01:52,876
Speaker 1: he joined the Student Space Initiative, which sounds kind of

27
00:01:52,916 --> 00:01:55,796
Speaker 1: fancy but in fact was basically a bunch of college

28
00:01:55,876 --> 00:01:58,316
Speaker 1: kids trying to build their own weather balloons.

29
00:02:02,156 --> 00:02:06,516
Speaker 2: The first half dozen flights of our balloons were complete

30
00:02:06,556 --> 00:02:08,036
Speaker 2: and total failures.

31
00:02:08,516 --> 00:02:09,756
Speaker 1: When do things start working?

32
00:02:10,636 --> 00:02:14,876
Speaker 2: That would be back in June of twenty sixteen. This

33
00:02:15,076 --> 00:02:18,716
Speaker 2: is the end of my freshman year, and I think

34
00:02:18,756 --> 00:02:22,476
Speaker 2: our longest flight that year had been fourteen minutes, so

35
00:02:23,076 --> 00:02:26,876
Speaker 2: shorter than a normal weather balloon.

36
00:02:27,236 --> 00:02:28,756
Speaker 1: Buy a lot, Buy a lot.

37
00:02:29,156 --> 00:02:35,596
Speaker 2: And also the person who had started this project, who

38
00:02:35,916 --> 00:02:39,876
Speaker 2: then became my co founder, had just graduated and was

39
00:02:39,916 --> 00:02:42,756
Speaker 2: about to leave what we thought never to be seen again.

40
00:02:44,436 --> 00:02:48,076
Speaker 2: And school has ended. We can't even get into the

41
00:02:48,116 --> 00:02:51,916
Speaker 2: building because our key cards have stopped working, so we

42
00:02:52,036 --> 00:02:54,836
Speaker 2: have to tailgate someone in to get our balloon launch

43
00:02:54,836 --> 00:02:58,876
Speaker 2: equipment out. Go out to the middle of nowhere in

44
00:02:58,956 --> 00:03:05,236
Speaker 2: the central Valley of California, and if this launch didn't succeed,

45
00:03:05,676 --> 00:03:06,996
Speaker 2: windborne would not exist.

46
00:03:07,796 --> 00:03:11,516
Speaker 1: Okay, so I can guess what happened next based on

47
00:03:11,556 --> 00:03:12,956
Speaker 1: that fact. But tell me what happened.

48
00:03:13,036 --> 00:03:17,516
Speaker 2: Yeah. So it got to be about four am, and

49
00:03:19,516 --> 00:03:22,676
Speaker 2: we're out in the middle of this park, really cold,

50
00:03:22,916 --> 00:03:27,556
Speaker 2: and the sprinklers turn on. They start slowly swiveling around

51
00:03:27,916 --> 00:03:33,076
Speaker 2: towards our delicate electronics, and so I have to hurl

52
00:03:33,156 --> 00:03:37,636
Speaker 2: myself onto the sprinklers, getting soaked, just to save our electronics.

53
00:03:37,796 --> 00:03:39,876
Speaker 1: It's like the nerd version of throwing yourself on a

54
00:03:39,876 --> 00:03:40,916
Speaker 1: grenade exactly.

55
00:03:40,996 --> 00:03:43,436
Speaker 2: I really felt like I was in an action movie sequence.

56
00:03:43,596 --> 00:03:50,116
Speaker 2: Ah good, But that sacrifice really paid off because with

57
00:03:50,276 --> 00:03:53,916
Speaker 2: that flight, we managed to set the world record for

58
00:03:54,436 --> 00:03:55,636
Speaker 2: weather balloon endurance.

59
00:03:56,316 --> 00:03:57,396
Speaker 1: How long did it stay up?

60
00:03:57,476 --> 00:04:00,396
Speaker 2: It stayed up for seventy six hours?

61
00:04:00,436 --> 00:04:02,676
Speaker 1: Wow, and your previous best.

62
00:04:02,476 --> 00:04:04,276
Speaker 2: Was fourteen minutes exactly.

63
00:04:04,436 --> 00:04:06,716
Speaker 1: So you're in college, how do you build weather balloons

64
00:04:06,756 --> 00:04:10,956
Speaker 1: that are better than whatever the federal government? Noah's launching

65
00:04:10,956 --> 00:04:12,756
Speaker 1: our you know, state of the art weather balloon.

66
00:04:12,756 --> 00:04:17,756
Speaker 2: People. So the first big thing is that some new

67
00:04:17,796 --> 00:04:23,236
Speaker 2: consumer electronics have come out, in particular, things like micro

68
00:04:23,316 --> 00:04:30,876
Speaker 2: controllers and lightweight, low power satellite communications, and we realize

69
00:04:30,956 --> 00:04:35,156
Speaker 2: that we can use these innovations to make a balloon

70
00:04:35,476 --> 00:04:37,716
Speaker 2: smarter and to control its altitude.

71
00:04:38,236 --> 00:04:39,396
Speaker 1: What's a micro controller.

72
00:04:39,596 --> 00:04:45,636
Speaker 2: Micro controller is a super small computer that you can

73
00:04:45,676 --> 00:04:48,116
Speaker 2: put in any device to give it a brain.

74
00:04:49,516 --> 00:04:50,156
Speaker 1: And they're cheap.

75
00:04:50,956 --> 00:04:54,156
Speaker 2: They're dirt cheap. Anytime you hear the word Internet of things,

76
00:04:54,236 --> 00:04:55,556
Speaker 2: that means micro controlers.

77
00:04:56,596 --> 00:04:58,996
Speaker 1: And nobody else was doing this at the time, Like

78
00:04:59,036 --> 00:05:02,756
Speaker 1: all the big, you know, well funded weather agencies weren't

79
00:05:02,916 --> 00:05:05,836
Speaker 1: sticking my micro controllers on their on their weather balloons.

80
00:05:06,276 --> 00:05:12,476
Speaker 2: Well, with these big agencies, they are really incentivized to

81
00:05:12,676 --> 00:05:16,556
Speaker 2: keep things stable because weather is this really important thing

82
00:05:16,676 --> 00:05:20,356
Speaker 2: that billions of people depend on, and so they're not

83
00:05:20,516 --> 00:05:24,356
Speaker 2: going to go and say, hey, what crazy experiments can

84
00:05:24,396 --> 00:05:24,796
Speaker 2: we run.

85
00:05:24,956 --> 00:05:27,556
Speaker 1: Their incentive is just like, don't screw it up exactly

86
00:05:27,556 --> 00:05:29,716
Speaker 1: the way we did it yesterday. Don't screw it up

87
00:05:29,716 --> 00:05:32,996
Speaker 1: and you won't get fired exactly. And so specifically, what

88
00:05:33,236 --> 00:05:35,756
Speaker 1: was it that you were doing that no one had

89
00:05:35,796 --> 00:05:36,396
Speaker 1: done before.

90
00:05:37,196 --> 00:05:41,316
Speaker 2: Yeah, So it's a very simple concept that is really

91
00:05:41,356 --> 00:05:45,636
Speaker 2: hard to make work and practice. The way we control

92
00:05:45,636 --> 00:05:49,476
Speaker 2: our altitude is when it gets too high up, we

93
00:05:49,756 --> 00:05:52,916
Speaker 2: vent some of our gas so it has less lift.

94
00:05:53,156 --> 00:05:56,356
Speaker 2: When we fall down too far, we drop some ballast

95
00:05:56,636 --> 00:06:00,356
Speaker 2: so it stops falling. But the hard part about this

96
00:06:00,676 --> 00:06:04,396
Speaker 2: is it has to function at negative seventy degrees celsius.

97
00:06:04,476 --> 00:06:07,716
Speaker 2: That's the temperature of dry ice. You have to have

98
00:06:08,116 --> 00:06:11,596
Speaker 2: a lot of on board software. That's where the micro

99
00:06:11,636 --> 00:06:15,876
Speaker 2: controllers come in to do things like understand is this

100
00:06:16,116 --> 00:06:19,076
Speaker 2: just turbulence that you're hitting or is it actually a

101
00:06:19,116 --> 00:06:21,636
Speaker 2: real change in lift that we need to drop ballast

102
00:06:21,716 --> 00:06:25,156
Speaker 2: or vent gas to account for. And in fact, we

103
00:06:25,236 --> 00:06:28,716
Speaker 2: stayed up twenty four to seven. Every single time it

104
00:06:28,836 --> 00:06:32,196
Speaker 2: talked to us. Every five minutes, we had it play

105
00:06:32,236 --> 00:06:35,716
Speaker 2: an airhorn sound because we were, of course falling asleep

106
00:06:35,756 --> 00:06:38,036
Speaker 2: in the middle of them to see what it was doing.

107
00:06:39,276 --> 00:06:41,756
Speaker 1: So, okay, so you set the record, you stay up

108
00:06:41,836 --> 00:06:45,196
Speaker 1: for what three days watching it? How far did it go?

109
00:06:46,276 --> 00:06:50,156
Speaker 2: It flew all the way across the country and landed

110
00:06:50,796 --> 00:06:55,916
Speaker 2: out in the Atlantic Ocean. But the real reason why

111
00:06:55,956 --> 00:06:59,076
Speaker 2: we decided to form a company around this was because

112
00:06:59,676 --> 00:07:03,636
Speaker 2: we realized the impact this could have. We got into

113
00:07:03,636 --> 00:07:06,956
Speaker 2: this from the engineering side. The lands are fun, but

114
00:07:07,956 --> 00:07:10,836
Speaker 2: we realized, wait a secd eighty five percent of the

115
00:07:10,876 --> 00:07:15,556
Speaker 2: Earth's atmosphere is invisible to humanity. Weather is this crazy,

116
00:07:15,676 --> 00:07:20,636
Speaker 2: unpredictable thing, and we can solve that. And weather is

117
00:07:20,716 --> 00:07:23,156
Speaker 2: so much more than just do you bring an umbrella

118
00:07:23,276 --> 00:07:29,516
Speaker 2: to work tomorrow is the single most immediately destructive aspect

119
00:07:29,876 --> 00:07:34,316
Speaker 2: of climate change, and improving the weather forecast it of

120
00:07:34,316 --> 00:07:37,996
Speaker 2: course helps with adapting to climate change, things like predicting

121
00:07:38,276 --> 00:07:41,756
Speaker 2: where a hurricane will make landfall, but it can also

122
00:07:41,876 --> 00:07:45,196
Speaker 2: help with preventing climate change in the first place. If

123
00:07:45,236 --> 00:07:49,756
Speaker 2: you have a better weather forecast, you can better route

124
00:07:49,876 --> 00:07:52,876
Speaker 2: ships and planes to save a huge amount of fuel.

125
00:07:53,636 --> 00:07:56,636
Speaker 2: You can accelerate the transition to renewables because you know

126
00:07:56,676 --> 00:07:59,996
Speaker 2: when the wind will be blowing or the sun will

127
00:08:00,036 --> 00:08:03,316
Speaker 2: be shining. And we looked at this and said, no

128
00:08:03,356 --> 00:08:06,716
Speaker 2: one else has our balloon technology. If we don't do this,

129
00:08:06,916 --> 00:08:10,596
Speaker 2: no one will. And if we want away from this opportunity,

130
00:08:10,636 --> 00:08:11,956
Speaker 2: we can't live with ourselves.

131
00:08:13,476 --> 00:08:15,596
Speaker 1: I mean, I buy that the stakes are high. I

132
00:08:15,676 --> 00:08:19,396
Speaker 1: buy that it could be very helpful, but like, why

133
00:08:19,436 --> 00:08:20,876
Speaker 1: wouldn't anybody else do it?

134
00:08:22,036 --> 00:08:26,876
Speaker 2: Yeah, a lot of people are trying to improve the

135
00:08:26,916 --> 00:08:32,756
Speaker 2: weather forecast. But what's unique about Windborne is the combination

136
00:08:33,156 --> 00:08:38,116
Speaker 2: of our balloon technology with our AI weather forecasts, because

137
00:08:38,236 --> 00:08:41,156
Speaker 2: really those are the two levers to pull to increase

138
00:08:41,196 --> 00:08:42,876
Speaker 2: the accuracy of a weather forecast.

139
00:08:44,156 --> 00:08:48,516
Speaker 1: You mentioned that eighty five percent of the world lacks

140
00:08:48,516 --> 00:08:52,276
Speaker 1: good weather data, which was surprising to me the first

141
00:08:52,356 --> 00:08:54,796
Speaker 1: time I heard it. But I would think, like, get

142
00:08:54,956 --> 00:08:57,876
Speaker 1: satellites do it. That's kind of I guess my naive

143
00:08:58,076 --> 00:08:59,956
Speaker 1: ish thought. It's like, I know there's a lot of

144
00:08:59,996 --> 00:09:01,916
Speaker 1: satellites looking down at the Earth all the time.

145
00:09:02,636 --> 00:09:04,876
Speaker 2: I'm actually a big fan of the space industry. I

146
00:09:04,956 --> 00:09:07,596
Speaker 2: used to work at SpaceX. But the problem is the

147
00:09:07,676 --> 00:09:12,036
Speaker 2: laws of physics fundamental limit what satellites can measure. Take

148
00:09:12,076 --> 00:09:16,796
Speaker 2: for example, pressure. It turns out pressure is extremely important

149
00:09:17,476 --> 00:09:20,836
Speaker 2: for predicting the weather because it determines where the winds

150
00:09:20,836 --> 00:09:24,076
Speaker 2: are blowing, how the weather systems are moving. But a

151
00:09:24,156 --> 00:09:29,316
Speaker 2: satellite fundamentally cannot measure pressure from space because it's not

152
00:09:29,476 --> 00:09:31,756
Speaker 2: in the atmosphere. It can't see what's going on.

153
00:09:32,316 --> 00:09:36,596
Speaker 1: So you start the company you're good at making weather

154
00:09:36,636 --> 00:09:38,716
Speaker 1: balloons that go up and stay up for a long time,

155
00:09:38,756 --> 00:09:41,676
Speaker 1: and that you can steer. I also read that you

156
00:09:41,836 --> 00:09:45,596
Speaker 1: launched thinking that you could collect data for a tenth

157
00:09:45,796 --> 00:09:50,676
Speaker 1: the cost of existing alternatives, which sounds compelling, but as

158
00:09:50,716 --> 00:09:53,916
Speaker 1: I understand was not compelling enough. Is that right?

159
00:09:54,596 --> 00:09:57,556
Speaker 2: It is. We're now at one hundred and fifty times

160
00:09:57,556 --> 00:09:59,636
Speaker 2: more data per dollar than alternatives.

161
00:09:59,996 --> 00:10:02,276
Speaker 1: Okay, was there some reason one tenth the cost was

162
00:10:02,356 --> 00:10:03,276
Speaker 1: not cheap enough.

163
00:10:04,196 --> 00:10:07,596
Speaker 2: Yeah. When we started the company, we thought it would

164
00:10:07,636 --> 00:10:11,356
Speaker 2: be cheap enough. But it turns out when you really

165
00:10:11,436 --> 00:10:14,876
Speaker 2: look at the scale of weather data collection, you need

166
00:10:14,956 --> 00:10:20,996
Speaker 2: to really understand the atmosphere. You want ten thousand balloons

167
00:10:21,156 --> 00:10:25,636
Speaker 2: aloft concurrently, that's a balloon every sixty miles in the atmosphere.

168
00:10:27,356 --> 00:10:32,276
Speaker 2: And in order to get to that level of scale,

169
00:10:32,556 --> 00:10:36,156
Speaker 2: it's the difference between your company spending one hundred million

170
00:10:36,236 --> 00:10:39,356
Speaker 2: dollars a year and spending a billion dollars a year.

171
00:10:39,796 --> 00:10:41,596
Speaker 2: One was a lot more in reach.

172
00:10:41,916 --> 00:10:43,796
Speaker 1: How do you drive down the cost to one one

173
00:10:43,916 --> 00:10:45,636
Speaker 1: hundred and fiftieth the cost? I mean, I'm sure it's

174
00:10:45,876 --> 00:10:50,516
Speaker 1: many many, many incremental you know, efficiency gains. But like,

175
00:10:50,556 --> 00:10:51,796
Speaker 1: what's an example of one.

176
00:10:51,996 --> 00:10:56,436
Speaker 2: Yeah, Well, one of the big pieces is improving the

177
00:10:56,636 --> 00:11:00,316
Speaker 2: software that flies on it so that the balloon can

178
00:11:00,356 --> 00:11:03,636
Speaker 2: fly for longer. I talked about in the student group

179
00:11:03,716 --> 00:11:07,316
Speaker 2: that first flight lasting seventy six hours. Now our longest

180
00:11:07,316 --> 00:11:09,756
Speaker 2: flight can fly for over four days.

181
00:11:10,476 --> 00:11:10,956
Speaker 1: Wow.

182
00:11:11,076 --> 00:11:15,036
Speaker 2: And the longer you fly, well, the hardware costs stays

183
00:11:15,076 --> 00:11:18,436
Speaker 2: the same, so that means you're collecting a lot more data.

184
00:11:18,916 --> 00:11:21,436
Speaker 1: Yeah. Yeah. So basically, make the balloon stamp for longer

185
00:11:21,596 --> 00:11:24,916
Speaker 1: is the fundamental way that you make it cheaper exactly,

186
00:11:25,476 --> 00:11:27,436
Speaker 1: And how do you make it fly longer?

187
00:11:27,956 --> 00:11:31,716
Speaker 2: Yeah, Well, one of the big things is improving the

188
00:11:31,756 --> 00:11:35,116
Speaker 2: software to better decide when do you vent gas, when

189
00:11:35,156 --> 00:11:38,396
Speaker 2: do you drop ballast or really one of the other

190
00:11:38,436 --> 00:11:43,116
Speaker 2: big things is just making everything smaller and lighter, because

191
00:11:43,596 --> 00:11:46,596
Speaker 2: the smaller it is, the lighter it is, the less

192
00:11:46,716 --> 00:11:50,236
Speaker 2: chance of a leak, the less ballast you have to use.

193
00:11:50,676 --> 00:11:55,916
Speaker 2: And things have just shrunk down so far. It really

194
00:11:56,316 --> 00:12:00,556
Speaker 2: surprises me sometimes when I'm like, wait, that tiny thing

195
00:12:00,636 --> 00:12:03,436
Speaker 2: the size of a dime replaces the thing that was

196
00:12:03,436 --> 00:12:05,716
Speaker 2: the size of a dinner plate before.

197
00:12:06,676 --> 00:12:10,076
Speaker 1: Yeah, that's amazing. So okay, so let's talk about where

198
00:12:10,116 --> 00:12:12,436
Speaker 1: you are today. Let's talk about sort of how it works.

199
00:12:13,716 --> 00:12:14,756
Speaker 1: How big are the balloons?

200
00:12:15,476 --> 00:12:20,116
Speaker 2: So the balloon is two pieces. There's the envelope that's

201
00:12:20,236 --> 00:12:24,076
Speaker 2: the bag that holds all parts the balloon part of

202
00:12:24,116 --> 00:12:30,356
Speaker 2: the balloon, and then there's the main unit, which is

203
00:12:30,396 --> 00:12:32,996
Speaker 2: the electronics and the ballast.

204
00:12:33,076 --> 00:12:36,836
Speaker 1: The stuff exactly, the balloon and the stuff the balloon part.

205
00:12:36,716 --> 00:12:39,116
Speaker 2: Of the balloon. The balloon part of the balloon is

206
00:12:39,436 --> 00:12:43,356
Speaker 2: five and a half meters tall. So that's two and

207
00:12:43,436 --> 00:12:45,796
Speaker 2: a half three times the height of person, depending on

208
00:12:45,836 --> 00:12:46,636
Speaker 2: how tall you are.

209
00:12:47,156 --> 00:12:51,996
Speaker 1: Yeah, so tall. Yeah, it's a big balloon. And how

210
00:12:52,236 --> 00:12:54,876
Speaker 1: how big is the stuff? How big is are the

211
00:12:54,916 --> 00:12:56,156
Speaker 1: sensors in the part that's not the.

212
00:12:56,316 --> 00:13:01,836
Speaker 2: Yeah, that stuff is. I guess this is an audio thing,

213
00:13:01,916 --> 00:13:04,516
Speaker 2: so I would say about this big.

214
00:13:06,116 --> 00:13:07,476
Speaker 1: The size of a basketball.

215
00:13:07,556 --> 00:13:12,156
Speaker 2: Yeah, it's a roughly the size of a basketball. It

216
00:13:12,236 --> 00:13:16,156
Speaker 2: weighs just over four pounds, so about the weight of

217
00:13:16,196 --> 00:13:22,396
Speaker 2: a large duck. Okay, and is kind of long and skinny.

218
00:13:22,436 --> 00:13:25,516
Speaker 2: It kind of looks like a fish. I think of

219
00:13:25,556 --> 00:13:28,716
Speaker 2: it as a trout attached to a giant bag.

220
00:13:30,156 --> 00:13:33,636
Speaker 1: How many of your balloons are in the sky right now? About?

221
00:13:34,396 --> 00:13:37,076
Speaker 2: Yeah, there are a few dozen aloft right now.

222
00:13:37,556 --> 00:13:37,836
Speaker 1: Okay.

223
00:13:37,876 --> 00:13:40,796
Speaker 2: We launch around one hundred a month and are quickly

224
00:13:40,876 --> 00:13:41,636
Speaker 2: ramping that up.

225
00:13:41,676 --> 00:13:43,076
Speaker 1: So where do you launch your balloons from?

226
00:13:43,796 --> 00:13:47,476
Speaker 2: We launch them every day from three continents, South Korea,

227
00:13:47,676 --> 00:13:50,076
Speaker 2: paloelto New York and Cabo Verde.

228
00:13:50,396 --> 00:13:53,156
Speaker 1: Cabo Verde is just off off the west coast of Africa.

229
00:13:53,236 --> 00:13:53,676
Speaker 1: Is that right?

230
00:13:53,796 --> 00:13:55,916
Speaker 2: You got it. One of the reasons why we have

231
00:13:56,036 --> 00:13:59,356
Speaker 2: launch site in Cabo Verde right now. That's right where

232
00:13:59,516 --> 00:14:02,316
Speaker 2: a lot of hurricanes are forming, and so the fact

233
00:14:02,356 --> 00:14:06,116
Speaker 2: that we're collecting data around there is going to be

234
00:14:06,316 --> 00:14:10,716
Speaker 2: really impactful for better predicting the path of these hurricanes. Huh.

235
00:14:10,756 --> 00:14:13,876
Speaker 1: So they start there and then they essentially travel across

236
00:14:13,916 --> 00:14:16,396
Speaker 1: the Atlantic and into the Americas, and if you can

237
00:14:16,516 --> 00:14:19,876
Speaker 1: understand what's going on there, ideally you can sort of

238
00:14:19,996 --> 00:14:22,716
Speaker 1: understand the hurricane where it's going to go in the

239
00:14:22,796 --> 00:14:26,276
Speaker 1: dream scenario, even just as it's becoming a tropical storm.

240
00:14:26,356 --> 00:14:30,476
Speaker 2: Exactly. Yeah, Noah did an analysis of the Impactor data

241
00:14:30,516 --> 00:14:35,756
Speaker 2: had on the twenty twenty two hurricane season, and it

242
00:14:36,436 --> 00:14:40,836
Speaker 2: made the forecasts for Hurricane Fiona about twenty percent better.

243
00:14:43,876 --> 00:14:46,316
Speaker 1: Now it's time for a few ads. After the ads,

244
00:14:46,556 --> 00:14:50,716
Speaker 1: Kai and I will talk about AI. Of course, we'll

245
00:14:50,716 --> 00:14:54,156
Speaker 1: also talk about Winborne's business, what they're selling, and we'll

246
00:14:54,156 --> 00:14:56,876
Speaker 1: talk about the company's quest to build balloons that can

247
00:14:56,916 --> 00:14:59,476
Speaker 1: stay in the air for months at a time and

248
00:14:59,476 --> 00:15:16,916
Speaker 1: make multiple trips around the world. What is your business

249
00:15:16,956 --> 00:15:18,636
Speaker 1: right now, like, what are you selling and who you're

250
00:15:18,636 --> 00:15:19,036
Speaker 1: selling it.

251
00:15:18,996 --> 00:15:22,236
Speaker 2: To right now. Our business has two pieces to it.

252
00:15:22,596 --> 00:15:26,276
Speaker 2: We are selling the observations of the atmosphere we collect

253
00:15:26,516 --> 00:15:30,356
Speaker 2: to the government right now, so that those governments can

254
00:15:31,436 --> 00:15:36,236
Speaker 2: use this data to improve your weather forecast. And just

255
00:15:36,276 --> 00:15:40,116
Speaker 2: a couple of weeks ago, we announced our AI based

256
00:15:40,156 --> 00:15:43,836
Speaker 2: weather model, which is the world's most accurate global weather

257
00:15:43,916 --> 00:15:44,996
Speaker 2: model bar none.

258
00:15:45,036 --> 00:15:46,836
Speaker 1: I feel like you need a little TM when you

259
00:15:46,876 --> 00:15:50,116
Speaker 1: say the world's most accurate global weather model bar none.

260
00:15:50,276 --> 00:15:51,916
Speaker 2: We really do, we really do.

261
00:15:52,076 --> 00:15:54,596
Speaker 1: So let's talk about that. Is that claim validated? I mean,

262
00:15:54,836 --> 00:15:57,196
Speaker 1: I know there was news about that just this year,

263
00:15:57,236 --> 00:16:01,476
Speaker 1: and congratulations, but also, how do I know that's true? Respectfully?

264
00:16:01,636 --> 00:16:06,116
Speaker 2: Yeah, I love that question because anytime a weather company

265
00:16:06,156 --> 00:16:10,636
Speaker 2: makes a claim, you should look at it closely. So,

266
00:16:10,996 --> 00:16:15,116
Speaker 2: first off, we've published our results on our website and

267
00:16:15,196 --> 00:16:18,876
Speaker 2: so you can dig into some of the raw numbers there. Second,

268
00:16:18,996 --> 00:16:23,076
Speaker 2: we're in the process of submitting to weather bench which

269
00:16:23,156 --> 00:16:28,836
Speaker 2: is these benchmarks run by Google, which were the previous

270
00:16:29,116 --> 00:16:33,916
Speaker 2: holders of this record. So we're actively talking with them

271
00:16:33,996 --> 00:16:38,956
Speaker 2: right now about submitting our results getting them validated into

272
00:16:38,996 --> 00:16:41,956
Speaker 2: live up there. It's just a process that takes a

273
00:16:41,996 --> 00:16:42,996
Speaker 2: little while, so.

274
00:16:43,036 --> 00:16:45,676
Speaker 1: Let's talk about the model you built. So I mean

275
00:16:46,676 --> 00:16:49,236
Speaker 1: generally when I talk to people who are working on AI,

276
00:16:49,356 --> 00:16:51,516
Speaker 1: they talk a lot about data, right, And it seems

277
00:16:51,596 --> 00:16:57,156
Speaker 1: like to some significant degree the models are kind of commoditized,

278
00:16:57,196 --> 00:17:00,316
Speaker 1: not quite, but like models are pretty similar, it seems

279
00:17:00,356 --> 00:17:02,676
Speaker 1: in many settings at least one to the other, and

280
00:17:02,756 --> 00:17:06,476
Speaker 1: the differentiator often ends up being the data. Is that

281
00:17:06,516 --> 00:17:08,076
Speaker 1: the case in this instance? I mean, do you think

282
00:17:08,116 --> 00:17:10,596
Speaker 1: you're the best. If you're the best because you have

283
00:17:10,836 --> 00:17:13,596
Speaker 1: all this data from all your balloons.

284
00:17:14,316 --> 00:17:16,556
Speaker 2: I think that you're spot on. Data is the real

285
00:17:16,596 --> 00:17:20,436
Speaker 2: mote here, because what really sets us apart is the

286
00:17:20,476 --> 00:17:24,356
Speaker 2: ability to have all of this data that no one

287
00:17:24,356 --> 00:17:28,716
Speaker 2: else has and to use that both for training the

288
00:17:28,756 --> 00:17:32,236
Speaker 2: model but then also for running the model in real time.

289
00:17:32,476 --> 00:17:35,956
Speaker 2: Data isn't just about training. It's about essentially giving your

290
00:17:35,956 --> 00:17:40,436
Speaker 2: model the prompt of what's going to happen next, and

291
00:17:40,676 --> 00:17:43,076
Speaker 2: you need new data every single day for that.

292
00:17:44,396 --> 00:17:48,676
Speaker 1: Just one thing to clarify. So, if you're selling the

293
00:17:48,796 --> 00:17:53,556
Speaker 1: data to governments, are they not sharing it widely? Like

294
00:17:53,596 --> 00:17:56,076
Speaker 1: they're paying for the data but they're not. But not

295
00:17:56,156 --> 00:17:58,156
Speaker 1: everybody can get at the data? Is that why it's

296
00:17:58,156 --> 00:17:58,756
Speaker 1: a mode for you?

297
00:17:59,396 --> 00:18:03,236
Speaker 2: Yeah, so what they're using the data we sell them

298
00:18:03,436 --> 00:18:09,236
Speaker 2: for is putting it into conventional physics based weather model

299
00:18:09,596 --> 00:18:14,796
Speaker 2: that they then release. So they can't redistribute the data itself,

300
00:18:15,756 --> 00:18:20,116
Speaker 2: but the general public can benefit from its use in

301
00:18:20,156 --> 00:18:21,116
Speaker 2: these weather models.

302
00:18:21,436 --> 00:18:24,356
Speaker 1: And so just to distinguish and this is a distinction

303
00:18:24,516 --> 00:18:27,636
Speaker 1: that's not particular to your company, but you mentioned the

304
00:18:27,676 --> 00:18:30,676
Speaker 1: sort of traditional physics based weather models. I mean, I

305
00:18:30,676 --> 00:18:33,516
Speaker 1: think it's worth spending just a moment here to distinguish, right,

306
00:18:33,596 --> 00:18:39,476
Speaker 1: like between the classic weather model, you know, the sort

307
00:18:39,516 --> 00:18:42,436
Speaker 1: of pre AI weather model, and the AI weather model. Like,

308
00:18:43,316 --> 00:18:44,996
Speaker 1: what's the basic difference between those two?

309
00:18:45,436 --> 00:18:49,516
Speaker 2: Yeah, So a conventional physics based weather model, it takes

310
00:18:49,516 --> 00:18:52,956
Speaker 2: the initial state of the atmosphere and then runs a

311
00:18:53,156 --> 00:18:59,116
Speaker 2: bunch of fluid dynamics on it to simulate what's going

312
00:18:59,156 --> 00:19:01,196
Speaker 2: to happen to all of these different fluids.

313
00:19:01,676 --> 00:19:04,036
Speaker 1: There's a sort of kind of classic sort of feels

314
00:19:04,036 --> 00:19:08,036
Speaker 1: like kind of nineteenth century deterministic. Give me the initial

315
00:19:08,796 --> 00:19:10,876
Speaker 1: condition and I'll tell you what's going to happen in

316
00:19:10,876 --> 00:19:13,516
Speaker 1: the future. It's that right, it's sort of rules based.

317
00:19:13,836 --> 00:19:21,076
Speaker 2: Exactly exactly, and that is nice because you can see

318
00:19:21,116 --> 00:19:25,236
Speaker 2: the exact physics that is being used, but you need

319
00:19:26,316 --> 00:19:29,996
Speaker 2: compute clusters. They cost hundreds of millions of dollars in

320
00:19:30,076 --> 00:19:33,076
Speaker 2: order to run this because the atmosphere is so big.

321
00:19:34,396 --> 00:19:39,076
Speaker 2: And by contrast, a AI based weather model can run

322
00:19:39,236 --> 00:19:41,396
Speaker 2: on just a gaming laptop.

323
00:19:42,316 --> 00:19:42,516
Speaker 1: Huh.

324
00:19:43,356 --> 00:19:48,956
Speaker 2: And what it does, by contrast, is effectively picking out statistics.

325
00:19:49,116 --> 00:19:51,196
Speaker 1: I mean, it's just it's machine learning. I presume it's

326
00:19:51,236 --> 00:19:52,596
Speaker 1: pattern matching exactly.

327
00:19:53,316 --> 00:19:57,276
Speaker 2: It's our model is a transformer based architecture, the same

328
00:19:57,316 --> 00:19:58,676
Speaker 2: thing that powers chat GPT.

329
00:19:59,556 --> 00:20:02,516
Speaker 1: Our A models in general better at this point than

330
00:20:02,556 --> 00:20:04,316
Speaker 1: the physics based traditional models.

331
00:20:04,916 --> 00:20:07,276
Speaker 2: It depends on the use case. But if you're talking

332
00:20:07,276 --> 00:20:12,876
Speaker 2: about hurricanes, yeah, we ran a case study on Hurricane

333
00:20:12,916 --> 00:20:19,476
Speaker 2: Ian and our model would have predicted landfall by two

334
00:20:19,596 --> 00:20:21,556
Speaker 2: hundred kilometers closer.

335
00:20:21,876 --> 00:20:24,956
Speaker 1: I mean, that's another one where like I'm intrigued by

336
00:20:24,956 --> 00:20:27,676
Speaker 1: what you said, but I'm very eager for I'm not

337
00:20:27,796 --> 00:20:31,356
Speaker 1: eager for this hurricane season to come, but like it

338
00:20:31,396 --> 00:20:34,676
Speaker 1: will resonate more with me when you do that prospectively, right,

339
00:20:34,756 --> 00:20:39,156
Speaker 1: of course, I mean, will you this year be predicting

340
00:20:39,196 --> 00:20:40,676
Speaker 1: where hurricanes are going to make landfall?

341
00:20:41,276 --> 00:20:45,236
Speaker 2: We will be predicting that for public safety reasons. We

342
00:20:45,436 --> 00:20:47,956
Speaker 2: aren't going to be you won't tell me.

343
00:20:48,116 --> 00:20:49,756
Speaker 1: Yeah you will, but you won't tell me we.

344
00:20:49,796 --> 00:20:55,236
Speaker 2: Will know a hurricane. Okay, but fair, you of course

345
00:20:55,276 --> 00:20:58,956
Speaker 2: don't want to say, hey, don't listen to Noah about

346
00:20:58,956 --> 00:21:00,396
Speaker 2: these evacuation orders.

347
00:21:01,076 --> 00:21:05,236
Speaker 1: No, no, fair that that is very responsible. So basically,

348
00:21:05,396 --> 00:21:09,916
Speaker 1: you're selling data. You are likely soon to be selling forecasts.

349
00:21:10,116 --> 00:21:14,956
Speaker 1: That's sort of the business. On the technical side, what

350
00:21:15,636 --> 00:21:17,596
Speaker 1: is the frontier? What are you trying to figure out

351
00:21:17,596 --> 00:21:18,876
Speaker 1: that you haven't figured out yet?

352
00:21:19,636 --> 00:21:23,316
Speaker 2: Yeah, well, one of the big areas for innovation is

353
00:21:23,996 --> 00:21:28,596
Speaker 2: on the AI modeling side. We were kind of surprised

354
00:21:28,716 --> 00:21:31,716
Speaker 2: that our models did as well as they did because

355
00:21:32,356 --> 00:21:35,836
Speaker 2: there are a lot of quite frankly obvious things that

356
00:21:36,116 --> 00:21:40,596
Speaker 2: we haven't done, things like just increasing the amount of

357
00:21:40,676 --> 00:21:44,676
Speaker 2: compute we're using to train these models. We use something

358
00:21:44,756 --> 00:21:49,396
Speaker 2: like a fifteenth as much as Google did. So what

359
00:21:49,516 --> 00:21:51,676
Speaker 2: happens when you train it for longer?

360
00:21:51,956 --> 00:21:54,076
Speaker 1: I mean, I'll say, on the AI side, it's not

361
00:21:54,396 --> 00:21:57,556
Speaker 1: obvious that it's optimal for you to be doing the

362
00:21:57,596 --> 00:22:00,316
Speaker 1: AI as well. Right, there's a universe where you are

363
00:22:00,756 --> 00:22:04,236
Speaker 1: optimizing the balloons and the data, and then someone else

364
00:22:05,276 --> 00:22:08,156
Speaker 1: like Google, who has all the computers and all the

365
00:22:08,196 --> 00:22:12,276
Speaker 1: AI knows what to do with doing the AI side.

366
00:22:12,036 --> 00:22:17,116
Speaker 2: Of it, right, Yes, and no, I think that where

367
00:22:17,156 --> 00:22:21,396
Speaker 2: that falls apart is coupling our data with the models

368
00:22:21,516 --> 00:22:24,676
Speaker 2: much more closely. In that there's so much to be

369
00:22:24,836 --> 00:22:28,156
Speaker 2: done in terms of figuring out how to better take

370
00:22:28,196 --> 00:22:33,516
Speaker 2: advantage of our data into particular and also things like saying,

371
00:22:34,556 --> 00:22:37,316
Speaker 2: based on this weather model, where should we be flying

372
00:22:37,316 --> 00:22:41,236
Speaker 2: our balloons to improve the forecast? And we're already using

373
00:22:41,436 --> 00:22:45,356
Speaker 2: our own AI weather forecasts to do that flight plan

374
00:22:45,476 --> 00:22:48,876
Speaker 2: optimization to figure out where our balloons are going to fly.

375
00:22:49,276 --> 00:22:52,796
Speaker 2: So it's really this beneficial effect where the better our

376
00:22:52,876 --> 00:22:56,116
Speaker 2: weather forecast, the better we can fly our balloons. And

377
00:22:56,836 --> 00:23:00,076
Speaker 2: we can then target our balloons to fly to the

378
00:23:00,076 --> 00:23:02,716
Speaker 2: places that will most improve the weather forecast.

379
00:23:03,396 --> 00:23:04,516
Speaker 1: That's a good feedback loop.

380
00:23:04,676 --> 00:23:06,716
Speaker 2: There's also a lot to do on the balloon side

381
00:23:06,716 --> 00:23:10,796
Speaker 2: of things. I wish I could tell you about our

382
00:23:11,476 --> 00:23:16,436
Speaker 2: project that will increase flight time by another factor.

383
00:23:16,076 --> 00:23:18,876
Speaker 1: Of ten, another factor of ten.

384
00:23:19,076 --> 00:23:21,676
Speaker 2: Yeah, so what what are you at now?

385
00:23:21,716 --> 00:23:23,716
Speaker 1: What's the like median flight time?

386
00:23:23,996 --> 00:23:29,076
Speaker 2: Medium flight time is seven to ten days depending on

387
00:23:29,356 --> 00:23:30,236
Speaker 2: how we're targeting it.

388
00:23:30,396 --> 00:23:33,996
Speaker 1: And so you think you're going to get to ninety days. Yep,

389
00:23:34,276 --> 00:23:38,956
Speaker 1: we around the world in eighty days. Yeah, you say

390
00:23:38,956 --> 00:23:40,356
Speaker 1: you wish you could tell me like you feel like

391
00:23:40,476 --> 00:23:43,076
Speaker 1: you're about to do it. Yeah, tell me, tell me

392
00:23:43,116 --> 00:23:44,716
Speaker 1: without telling me what you can't tell me.

393
00:23:44,996 --> 00:23:49,756
Speaker 2: Okay, I'll do my best. So we have had some

394
00:23:50,356 --> 00:23:56,836
Speaker 2: very successful flight tests where we have demonstrated the ability

395
00:23:57,076 --> 00:24:03,116
Speaker 2: to fly without using any finite resources.

396
00:24:03,356 --> 00:24:06,636
Speaker 1: So what what is the key finite resource? The the

397
00:24:06,676 --> 00:24:08,876
Speaker 1: lift gas, the what is it helium or hydrogen?

398
00:24:08,916 --> 00:24:13,236
Speaker 2: What you used to keep Our balloons are compatible with

399
00:24:13,276 --> 00:24:16,556
Speaker 2: both helium and hydrogen, So which we use is dependent

400
00:24:16,636 --> 00:24:17,276
Speaker 2: on the country.

401
00:24:17,676 --> 00:24:20,316
Speaker 1: When you say flying without any finite resource, that's what

402
00:24:20,356 --> 00:24:21,556
Speaker 1: I think of, yep.

403
00:24:22,196 --> 00:24:25,036
Speaker 2: Okay, yeah, that and ballast.

404
00:24:25,716 --> 00:24:30,836
Speaker 1: Oh right, So so you want to you can? I should?

405
00:24:30,876 --> 00:24:31,316
Speaker 1: I guess.

406
00:24:31,596 --> 00:24:35,476
Speaker 2: I mean, there's heat there, leave it there and just say.

407
00:24:37,156 --> 00:24:39,076
Speaker 1: If you get that, could you fly forever?

408
00:24:40,036 --> 00:24:45,636
Speaker 2: That the dream someday, someday we're we're definitely ninety.

409
00:24:45,396 --> 00:24:47,996
Speaker 1: Days for a weather balloon is sort of forever, right.

410
00:24:47,996 --> 00:24:51,436
Speaker 2: It really is. And so, yeah, we've had some very

411
00:24:51,476 --> 00:24:52,876
Speaker 2: successful flight tests.

412
00:24:53,196 --> 00:24:55,596
Speaker 1: What's the longest flight you've had to this point?

413
00:24:55,796 --> 00:24:57,236
Speaker 2: Forty days?

414
00:24:57,276 --> 00:24:58,316
Speaker 1: Forty days? Okay?

415
00:24:58,436 --> 00:24:58,716
Speaker 2: Long?

416
00:24:58,836 --> 00:24:59,156
Speaker 1: Yeah?

417
00:24:59,236 --> 00:25:02,356
Speaker 2: Long? And that was without using this technology, was that

418
00:25:02,436 --> 00:25:07,316
Speaker 2: like a fluke. It was definitely on the side of

419
00:25:07,356 --> 00:25:10,956
Speaker 2: the bell curve. But we've had multiple month long flights.

420
00:25:11,756 --> 00:25:16,236
Speaker 1: What why do your balloons usually what do you say fall?

421
00:25:16,316 --> 00:25:19,076
Speaker 1: Crash sounds a little extreme crash fall.

422
00:25:19,956 --> 00:25:23,996
Speaker 2: Yeah, it's that they run out of ballast and lifting gas,

423
00:25:24,236 --> 00:25:29,156
Speaker 2: and so these finite resources have been used up. You

424
00:25:29,196 --> 00:25:32,156
Speaker 2: don't have a more ballast to drop, you don't have

425
00:25:32,196 --> 00:25:35,796
Speaker 2: a more gas event, and so you can no longer

426
00:25:35,876 --> 00:25:36,916
Speaker 2: control your altitude.

427
00:25:37,316 --> 00:25:40,836
Speaker 1: What happens when it falls crashes comes down.

428
00:25:41,636 --> 00:25:44,756
Speaker 2: So by the time it falls, it has used up

429
00:25:44,996 --> 00:25:49,956
Speaker 2: all of the ballast. So instead of weighing four pounds,

430
00:25:50,036 --> 00:25:55,476
Speaker 2: it weighs about two hundred grams.

431
00:25:56,036 --> 00:25:58,516
Speaker 1: Two hundred grams is what half a pound?

432
00:25:58,676 --> 00:26:02,996
Speaker 2: Ish? Yes, I'd looked it up. Zero point four to

433
00:26:03,036 --> 00:26:03,756
Speaker 2: four pounds.

434
00:26:04,076 --> 00:26:06,796
Speaker 1: Okay. By the way, when you're getting rid of ballast,

435
00:26:06,876 --> 00:26:08,676
Speaker 1: is it just like sand or something? You just like

436
00:26:08,716 --> 00:26:14,836
Speaker 1: sprinkle sand out exactly? Okay, So it's half a pound,

437
00:26:14,916 --> 00:26:17,556
Speaker 1: so it wouldn't hurt if it hit me on the head?

438
00:26:17,636 --> 00:26:19,316
Speaker 1: Is that is that part of the reason half a

439
00:26:19,316 --> 00:26:21,156
Speaker 1: pound is important? Would it if it hit me on that?

440
00:26:21,396 --> 00:26:25,196
Speaker 2: Is it? Anybody on that it's never hit anybody. We

441
00:26:25,276 --> 00:26:29,516
Speaker 2: do landing simulations and control where it lands to direct

442
00:26:29,556 --> 00:26:36,236
Speaker 2: it towards unpopulated areas. But remember it also has this

443
00:26:36,836 --> 00:26:40,356
Speaker 2: envelope attached to it, which acts like a parachute, so

444
00:26:40,556 --> 00:26:45,516
Speaker 2: it actually falls very slowly. It's very light and not

445
00:26:45,636 --> 00:26:46,356
Speaker 2: an issue.

446
00:26:46,516 --> 00:26:49,156
Speaker 1: And then what happens. Then you have this big deflated

447
00:26:49,156 --> 00:26:51,756
Speaker 1: balloon sitting on the ground or floating in the ocean,

448
00:26:51,836 --> 00:26:54,196
Speaker 1: and is it just what happens to you?

449
00:26:54,316 --> 00:26:58,316
Speaker 2: Yeah? So I love that question because it's a chance

450
00:26:58,396 --> 00:27:02,236
Speaker 2: to talk about how our balloons can reduce the amount

451
00:27:02,316 --> 00:27:05,076
Speaker 2: of waste going out into the world compared to the

452
00:27:05,116 --> 00:27:07,996
Speaker 2: half a million that are launched every year, of which

453
00:27:08,036 --> 00:27:09,476
Speaker 2: only a fifth are covered.

454
00:27:10,156 --> 00:27:12,756
Speaker 1: Just get answer directly. I appreciate that you want to

455
00:27:12,796 --> 00:27:15,956
Speaker 1: contextualize it, but like, first, let's talk about what happens

456
00:27:15,956 --> 00:27:18,476
Speaker 1: with your balloons and then feel free to provide the bigger.

457
00:27:18,156 --> 00:27:22,996
Speaker 2: Cart sounds good, sounds good? Yeah, yeah, So our balloons

458
00:27:23,076 --> 00:27:26,996
Speaker 2: are out in the world, and we recover as many

459
00:27:27,036 --> 00:27:30,516
Speaker 2: of them as we can. In the long term, we're

460
00:27:30,636 --> 00:27:34,596
Speaker 2: aiming to recover essentially all of them because we'll direct

461
00:27:34,636 --> 00:27:36,276
Speaker 2: them to specific landing sites.

462
00:27:36,716 --> 00:27:39,996
Speaker 1: I mean, the ocean has got to be tough to write.

463
00:27:40,036 --> 00:27:41,636
Speaker 1: I presume when they fall in the ocean, for the

464
00:27:41,676 --> 00:27:42,836
Speaker 1: most part, you don't recover them.

465
00:27:42,876 --> 00:27:45,196
Speaker 2: Is that right, exactly? And that's one of the reasons

466
00:27:45,236 --> 00:27:49,636
Speaker 2: why improving endurance is so exciting, because if you circumnavigate

467
00:27:49,716 --> 00:27:52,836
Speaker 2: multiple times, well, on the last leg, you bring it

468
00:27:52,916 --> 00:27:55,996
Speaker 2: down in a field, have somebody drive around and collect

469
00:27:56,036 --> 00:27:59,476
Speaker 2: all of them. They are about half million weather balloons

470
00:27:59,556 --> 00:28:04,316
Speaker 2: launched each year, and only a fifth of those are recovered.

471
00:28:04,636 --> 00:28:09,196
Speaker 2: So when each balloon flies for only two hours, well,

472
00:28:09,356 --> 00:28:12,196
Speaker 2: you have to launch a lot more balloons to collect

473
00:28:12,276 --> 00:28:14,996
Speaker 2: the same amount of data. So we really see this

474
00:28:15,036 --> 00:28:18,196
Speaker 2: as an opportunity to reduce the amount of stuff that's

475
00:28:18,236 --> 00:28:19,556
Speaker 2: going out there in the first place.

476
00:28:20,916 --> 00:28:25,236
Speaker 1: Right, So there will actually be fewer balloons going into

477
00:28:25,276 --> 00:28:26,596
Speaker 1: the world if you succeed.

478
00:28:26,956 --> 00:28:27,516
Speaker 2: Exactly.

479
00:28:28,516 --> 00:28:30,596
Speaker 1: So, you're at this point where you've learned a lot.

480
00:28:30,676 --> 00:28:32,356
Speaker 1: You sort of feels like you're kind of on the

481
00:28:32,396 --> 00:28:35,356
Speaker 1: precipice of a lot more. And I want to talk

482
00:28:35,356 --> 00:28:37,996
Speaker 1: about sort of two futures. Right. One is a future

483
00:28:38,036 --> 00:28:42,076
Speaker 1: where where your company doesn't succeed for any number of

484
00:28:42,076 --> 00:28:45,756
Speaker 1: reasons with which you're truly more familiar than I. And

485
00:28:45,796 --> 00:28:47,916
Speaker 1: then the other is if you do succeed. Right, So,

486
00:28:48,596 --> 00:28:50,476
Speaker 1: in the sort of sad version where you don't succeed,

487
00:28:50,516 --> 00:28:52,316
Speaker 1: what are some reasons it might not work out?

488
00:28:52,956 --> 00:28:56,916
Speaker 2: Yeah, I think that probably the biggest reason is that

489
00:28:57,356 --> 00:29:02,676
Speaker 2: scaling up hardware is hard. We need to increase manufacturing

490
00:29:02,836 --> 00:29:06,196
Speaker 2: by a factor of one hundred effectively, and.

491
00:29:06,116 --> 00:29:08,956
Speaker 1: So it's scaling up that final assembly by a factor

492
00:29:08,956 --> 00:29:12,076
Speaker 1: of one hundred is complicated and hard and just kind

493
00:29:12,116 --> 00:29:17,956
Speaker 1: of classic hard building. A business gets more capital intensive exactly,

494
00:29:18,076 --> 00:29:21,796
Speaker 1: and the money before you can get the money exactly. Okay,

495
00:29:21,916 --> 00:29:26,516
Speaker 1: So that's an easy to imagine sad outcome. Let's talk

496
00:29:26,516 --> 00:29:30,356
Speaker 1: about the happy outcome. Like it works. You scale up

497
00:29:31,156 --> 00:29:35,396
Speaker 1: your balloons, stay up for months at a time. What's

498
00:29:35,436 --> 00:29:37,996
Speaker 1: the world look like? What are you doing in that scenario.

499
00:29:38,276 --> 00:29:41,516
Speaker 2: We want everybody to see twice as far into the

500
00:29:41,556 --> 00:29:44,516
Speaker 2: future when it comes to weather. So making the ten

501
00:29:44,596 --> 00:29:47,996
Speaker 2: day forecast is accurate is the current five day forecast.

502
00:29:48,156 --> 00:29:50,596
Speaker 2: Making the twenty day forecast as accurate as the ten

503
00:29:50,676 --> 00:29:54,156
Speaker 2: day forecast, so we want people to see twice as

504
00:29:54,196 --> 00:29:57,516
Speaker 2: far into the future. We want to pinpoint where hurricanes

505
00:29:57,556 --> 00:30:01,956
Speaker 2: are going to make landfall a week in advance, and

506
00:30:02,436 --> 00:30:07,276
Speaker 2: we want day to day weather that businesses rely on

507
00:30:07,756 --> 00:30:15,356
Speaker 2: to be really accurate, things like never having to cancel

508
00:30:15,396 --> 00:30:18,796
Speaker 2: a flight last minute because a bomb cyclone popped up,

509
00:30:20,556 --> 00:30:24,156
Speaker 2: saving fuel with all of your shipping because you know

510
00:30:24,436 --> 00:30:28,996
Speaker 2: where there will be headwinds. We want to accelerate the

511
00:30:29,036 --> 00:30:33,276
Speaker 2: transition to renewables, and we want weather to go from

512
00:30:33,316 --> 00:30:39,036
Speaker 2: this crazy, unpredictable source of uncertainty to something that humans

513
00:30:39,116 --> 00:30:39,836
Speaker 2: just know about.

514
00:30:42,916 --> 00:30:56,076
Speaker 1: We'll be back in a minute with the lightning round. Okay,

515
00:30:56,116 --> 00:31:01,636
Speaker 1: let's do the lightning round. Would you rather have it

516
00:31:01,676 --> 00:31:03,036
Speaker 1: be too hot or too cold?

517
00:31:03,276 --> 00:31:04,516
Speaker 2: Too cold any day?

518
00:31:06,556 --> 00:31:10,836
Speaker 1: Okay? The next one is multiple choice. What is your

519
00:31:10,836 --> 00:31:15,756
Speaker 1: favorite movie prominently featuring balloons? Wizard of Oz around the

520
00:31:15,756 --> 00:31:19,556
Speaker 1: World in eighty days, up the Red Balloon, or none

521
00:31:19,596 --> 00:31:20,036
Speaker 1: of the above.

522
00:31:20,196 --> 00:31:22,476
Speaker 2: It's got to be up brilliant.

523
00:31:21,996 --> 00:31:27,196
Speaker 1: Movie, Blimps, overrated or underrated?

524
00:31:27,476 --> 00:31:31,116
Speaker 2: Ooh, so here, I've got to make a distinction between

525
00:31:31,236 --> 00:31:34,316
Speaker 2: blimps and Zeppelins because I actually used to work at

526
00:31:34,476 --> 00:31:35,916
Speaker 2: a zeppelin company.

527
00:31:37,676 --> 00:31:40,996
Speaker 1: Is that the Larry Page Zeppelin company? It sure is

528
00:31:41,596 --> 00:31:45,996
Speaker 1: perhaps the only zeppelin company. Yeah, and so is the

529
00:31:46,076 --> 00:31:48,996
Speaker 1: distinction that a zeppelin has a rigid frame?

530
00:31:49,436 --> 00:31:50,156
Speaker 2: Got it in? Won?

531
00:31:51,036 --> 00:31:53,596
Speaker 1: Okay? So how about this. Let me reformulate it to

532
00:31:54,396 --> 00:31:59,076
Speaker 1: see if I get it this time. Airships overrated or underrated?

533
00:31:59,276 --> 00:32:06,556
Speaker 2: Underrated? I thought that zeppelins are so cool. I don't

534
00:32:06,636 --> 00:32:10,716
Speaker 2: have a great mission driven answer here other than you know,

535
00:32:11,436 --> 00:32:14,396
Speaker 2: I think it's amazing. I read too many books growing

536
00:32:14,476 --> 00:32:18,236
Speaker 2: up which prominently featured airships.

537
00:32:18,596 --> 00:32:22,556
Speaker 1: Isn't the isn't the Larry Page airship company dreams some

538
00:32:22,716 --> 00:32:25,436
Speaker 1: kind of cargo, like the idea that, for like really

539
00:32:25,476 --> 00:32:28,076
Speaker 1: heavy things, giant airships would be an efficient way to

540
00:32:28,476 --> 00:32:29,116
Speaker 1: move cargo.

541
00:32:29,636 --> 00:32:32,756
Speaker 2: Full disclosure, It's been a long time since I worked there,

542
00:32:32,796 --> 00:32:37,356
Speaker 2: so I can't speak to that company in particular. But yeah,

543
00:32:38,036 --> 00:32:43,116
Speaker 2: faster deliveries than a ship, more efficient deliveries than a plane.

544
00:32:43,556 --> 00:32:49,956
Speaker 1: Aha, It's like a niche an Until what's something besides

545
00:32:49,996 --> 00:32:52,516
Speaker 1: the weather that AI will be really good at predicting?

546
00:32:53,236 --> 00:33:02,076
Speaker 2: Oh? Great question, let's see. Uh. I don't have a

547
00:33:02,076 --> 00:33:05,316
Speaker 2: good answer off the top of my head. I think

548
00:33:05,396 --> 00:33:10,956
Speaker 2: that it will be really interesting for economics in terms

549
00:33:11,036 --> 00:33:18,276
Speaker 2: of taking short term data, various real time indicators, and

550
00:33:18,476 --> 00:33:22,436
Speaker 2: making predictions about what the final readings are going to be.

551
00:33:24,196 --> 00:33:27,036
Speaker 1: I'm sure a lot of very smart people are working

552
00:33:27,076 --> 00:33:29,076
Speaker 1: on that and they may in fact already have good

553
00:33:29,116 --> 00:33:31,116
Speaker 1: models and they're not telling us.

554
00:33:31,276 --> 00:33:32,276
Speaker 2: I think that's likely.

555
00:33:32,356 --> 00:33:33,676
Speaker 1: Is there anything else you want to say?

556
00:33:35,076 --> 00:33:37,516
Speaker 2: I think that covers it. Yeah, great to be on

557
00:33:37,556 --> 00:33:37,876
Speaker 2: the show.

558
00:33:39,236 --> 00:33:40,996
Speaker 1: Thanks for your time. It was lovely to talk with you.

559
00:33:41,116 --> 00:33:41,596
Speaker 1: Good luck.

560
00:33:41,716 --> 00:33:46,156
Speaker 2: Thanks and if you ever are in Paloelto, you're welcome

561
00:33:46,156 --> 00:33:47,316
Speaker 2: to come see a balloon launch.

562
00:33:48,436 --> 00:33:52,996
Speaker 1: Great. So do you still you launch from where you launch?

563
00:33:53,036 --> 00:33:55,876
Speaker 1: From that Air Force base or like or just from.

564
00:33:56,196 --> 00:33:56,996
Speaker 2: From our part part?

565
00:33:57,036 --> 00:33:59,676
Speaker 1: Are they easy? Yeah? From the parking lot, that's cool.

566
00:33:59,836 --> 00:34:00,156
Speaker 2: Yeah.

567
00:34:00,196 --> 00:34:02,236
Speaker 1: How much space do you need to launch a weather balloon?

568
00:34:03,756 --> 00:34:08,076
Speaker 2: Not that much? About fifty by fifty feet. It's really

569
00:34:08,156 --> 00:34:10,596
Speaker 2: just a matter of making sure there's nothing that the

570
00:34:10,596 --> 00:34:12,996
Speaker 2: balloon will blow into right after releasing it.

571
00:34:13,556 --> 00:34:17,196
Speaker 1: Uh huh. Seems like it'll be fun to see. Yeah,

572
00:34:17,236 --> 00:34:22,996
Speaker 1: I like a balloon. Kai Marshland is the co founder

573
00:34:23,036 --> 00:34:27,716
Speaker 1: and chief product officer of Windborne Systems. Today's show was

574
00:34:27,756 --> 00:34:31,436
Speaker 1: produced by Gabriel Hunter Chang and edited by Lydia Jean Kott,

575
00:34:31,596 --> 00:34:34,796
Speaker 1: who was engineered by Sarah Buguer. You can email us

576
00:34:34,836 --> 00:34:38,716
Speaker 1: at problem at Pushkin dot FM. I'm Jacob Goldstein and

577
00:34:38,756 --> 00:34:41,116
Speaker 1: we'll be back next week with another episode of What's

578
00:34:41,116 --> 00:34:50,596
Speaker 1: Your Problem,