AI Innovations in Disaster Management: A Deep Dive with Meta's Laura McGorman

1
00:00:00,040 --> 00:00:02,360
Well, Laura, thank you so much
for joining us.

2
00:00:02,360 --> 00:00:05,320
This will be excited to talk to
a Facebook expert.

3
00:00:05,320 --> 00:00:09,120
Welcome to the show.
Thank you so much for having me,

4
00:00:09,120 --> 00:00:11,120
Jonathan.
Well, this will be fun.

5
00:00:11,120 --> 00:00:15,520
So first tell folks what it what
what it is that you do at

6
00:00:15,520 --> 00:00:17,160
Facebook.
And then we'll kind of hop into

7
00:00:17,160 --> 00:00:21,160
some really cool tools that
Facebook has and how maybe those

8
00:00:21,160 --> 00:00:25,200
tools could help folks as they
navigate, you know, crazy things

9
00:00:25,200 --> 00:00:29,760
like weather disasters.
Yeah, well, I love talking about

10
00:00:29,760 --> 00:00:31,000
my job.
I think I have one of the

11
00:00:31,000 --> 00:00:33,880
coolest jobs in the world.
So I lead A-Team called AI for

12
00:00:33,880 --> 00:00:36,960
Good at Meta.
And my job is to take all the

13
00:00:36,960 --> 00:00:39,320
tools that Meta creates from a
technological perspective.

14
00:00:39,320 --> 00:00:42,560
That could be things like the
Facebook app that everyone knows

15
00:00:42,560 --> 00:00:46,200
and loves, or tools like our
open source AI models that I'll

16
00:00:46,200 --> 00:00:49,160
talk about today and try and
figure out how they can be used

17
00:00:49,160 --> 00:00:51,000
for social impact around the
world.

18
00:00:51,000 --> 00:00:53,480
And for this can be for
applications like public health

19
00:00:53,880 --> 00:00:56,480
that's been used for
applications like climate change

20
00:00:56,480 --> 00:01:04,680
on today the topic of commerce.
That's awesome, man.

21
00:01:04,680 --> 00:01:07,280
I bet that is a fun job.
I bet you have a lot of fun, and

22
00:01:07,280 --> 00:01:09,960
I'm sure you do tons of these
types of interviews.

23
00:01:10,240 --> 00:01:14,760
But today we get to kind of talk
about how Facebook and AI can

24
00:01:14,760 --> 00:01:17,880
help prep somebody for disaster,
disaster weather.

25
00:01:17,880 --> 00:01:20,400
Disasters happen all over the
US.

26
00:01:20,680 --> 00:01:25,400
We just had one here in Alaska.
Tell me about these tools that

27
00:01:25,400 --> 00:01:29,800
Facebook has and how the average
person could maybe utilize them

28
00:01:29,800 --> 00:01:32,720
to help them prepare.
Sure.

29
00:01:32,720 --> 00:01:36,160
So before we dive into the topic
of AI, which I think is really

30
00:01:36,160 --> 00:01:38,800
the future of how disaster
response is going to work in the

31
00:01:38,800 --> 00:01:41,120
US and abroad.
We have a lot of tools that are

32
00:01:41,120 --> 00:01:43,960
available on the Facebook app
right now that have actually

33
00:01:43,960 --> 00:01:45,720
been in market for a number of
years.

34
00:01:45,720 --> 00:01:48,040
I'll chat about 3:00 that I
think are the most important for

35
00:01:48,040 --> 00:01:50,800
people to know about 1 is called
safety check.

36
00:01:50,960 --> 00:01:53,760
And this is a tool that if
you're on Facebook and you're in

37
00:01:53,760 --> 00:01:56,760
a disaster affected area, you'll
essentially get a notification

38
00:01:56,760 --> 00:01:59,520
to your Facebook account that
allows you to mark yourself

39
00:01:59,520 --> 00:02:01,280
safe.
We think this is an important

40
00:02:01,280 --> 00:02:04,120
tool because, you know, during
disasters, it's really hard to

41
00:02:04,120 --> 00:02:06,080
stay in touch with your broader
community.

42
00:02:06,080 --> 00:02:08,520
Maybe you only stay in touch
with a couple close family

43
00:02:08,520 --> 00:02:10,520
members.
But this allows you to do a mark

44
00:02:10,520 --> 00:02:12,800
yourself safe and let your whole
social network know that you're

45
00:02:12,800 --> 00:02:15,600
out of harm's way.
The second floor is called local

46
00:02:15,600 --> 00:02:17,160
alerts.
This is really more about you

47
00:02:17,160 --> 00:02:20,640
getting information from
emergency response agencies in

48
00:02:20,640 --> 00:02:23,400
your area.
And this is a tool you can opt

49
00:02:23,400 --> 00:02:26,440
into getting notifications
essentially from either your

50
00:02:26,440 --> 00:02:29,160
local Police Department or your
local Emergency Management

51
00:02:29,160 --> 00:02:31,640
department.
If they have a Facebook page and

52
00:02:31,640 --> 00:02:34,600
they use local alerts, you'll be
able to get notifications

53
00:02:34,600 --> 00:02:37,320
directly from them as the
situation unfolds on the ground.

54
00:02:37,320 --> 00:02:40,000
And we believe this is a really
important tool that empowers

55
00:02:40,200 --> 00:02:42,560
local governments and 1st
responders to directly

56
00:02:42,560 --> 00:02:44,280
communicate with affected
communities.

57
00:02:45,000 --> 00:02:47,400
And the last but not least is
that many people don't know, but

58
00:02:47,400 --> 00:02:50,840
Facebook also has a really
robust set of fundraising tools

59
00:02:51,240 --> 00:02:53,800
that allow for individuals who
are out of a disaster affected

60
00:02:53,800 --> 00:02:57,240
area to raise funds for those
who are affected.

61
00:02:57,240 --> 00:03:00,120
Maybe they need supplies, maybe
they need funding, maybe they

62
00:03:00,120 --> 00:03:03,280
need housing assistance.
Fundraising is a really powerful

63
00:03:03,280 --> 00:03:05,800
tool in a way that you can
support communities affected by

64
00:03:05,800 --> 00:03:08,160
disaster.
And you can create a nonprofit

65
00:03:08,560 --> 00:03:12,560
fundraiser for everything from a
local animal shelter to a Red

66
00:03:12,560 --> 00:03:14,800
Cross in your area.
And we believe this, this is a

67
00:03:14,800 --> 00:03:16,960
really powerful tool for people
to know about as well.

68
00:03:17,680 --> 00:03:20,120
That's cool.
So the, the check, you know, the

69
00:03:20,120 --> 00:03:23,720
check mark safe is not just a
meme, it's an actual thing.

70
00:03:23,960 --> 00:03:26,920
We've all seen the memes of
like, you know, I've been, I'm

71
00:03:26,920 --> 00:03:29,720
safe from whatever, you know,
funny thing was happening that

72
00:03:29,720 --> 00:03:31,040
week.
But this is an actual thing

73
00:03:31,040 --> 00:03:34,360
where people could go in and
mark themselves safe.

74
00:03:34,360 --> 00:03:37,640
Maybe they're in a hurricane or
tornado or a flood or something

75
00:03:37,640 --> 00:03:41,720
like that, and they could inform
loved ones that I'm alive.

76
00:03:41,720 --> 00:03:42,920
And that's an easy way to do
that.

77
00:03:42,920 --> 00:03:44,840
Is that kind of the way it
works?

78
00:03:45,240 --> 00:03:47,680
That's exactly how it works.
And, and I I find it to be a

79
00:03:47,680 --> 00:03:50,360
really helpful tool.
Sometimes things are happening

80
00:03:50,360 --> 00:03:52,280
in other countries.
I'll have a friend from Graduate

81
00:03:52,280 --> 00:03:54,800
School that's in a country where
there was a horrible earthquake

82
00:03:54,800 --> 00:03:56,720
or typhoon.
It's always a relief when I open

83
00:03:56,720 --> 00:03:59,480
my Facebook account to see
somebody who I know is in a

84
00:03:59,480 --> 00:04:01,560
disaster affected area.
You know, it doesn't require me

85
00:04:01,560 --> 00:04:03,360
reaching out to them directly,
but if you see that they've

86
00:04:03,360 --> 00:04:05,680
marked themselves themselves
safe, you know they're OK.

87
00:04:05,680 --> 00:04:09,120
And that can be really a relief
to their broader social network.

88
00:04:09,480 --> 00:04:13,040
And this is probably something
that if you can navigate

89
00:04:13,040 --> 00:04:16,680
Facebook, you could easily be
able to do this part of it.

90
00:04:16,680 --> 00:04:18,920
It's not like you don't need
like a degree in.

91
00:04:19,360 --> 00:04:22,880
No.
Coding or developing Or if

92
00:04:22,880 --> 00:04:25,440
you're able to upload a Facebook
photo, you could probably do

93
00:04:25,440 --> 00:04:27,560
this part.
Yeah, you'll get a notification

94
00:04:27,560 --> 00:04:29,080
essentially at the top of your
news feed.

95
00:04:29,080 --> 00:04:31,120
It's essentially one Click to
mark yourself safe.

96
00:04:31,120 --> 00:04:33,280
And so we've tried to make it as
easy as possible for people to

97
00:04:33,280 --> 00:04:35,600
do that.
And do people, I mean, I'm sure

98
00:04:35,600 --> 00:04:38,560
you have data on this, but my
guess is people are using this

99
00:04:39,080 --> 00:04:41,440
pretty actively all over the
globe when it comes to

100
00:04:41,440 --> 00:04:43,920
disasters.
Indeed, billions of people over

101
00:04:43,920 --> 00:04:45,200
years have marked themselves
safe.

102
00:04:45,200 --> 00:04:48,240
And we, we believe it's a really
important tool and we have kept

103
00:04:48,240 --> 00:04:51,480
it going at launched over a
decade ago and we wow.

104
00:04:53,280 --> 00:04:57,200
OK, so let's talk about some of
these new really cool AI tools

105
00:04:57,200 --> 00:05:01,240
and what they are and maybe what
it means to being prepared for

106
00:05:01,240 --> 00:05:03,280
disasters.
Sure.

107
00:05:03,280 --> 00:05:06,680
So many people maybe don't
realize, but in addition to all

108
00:05:06,680 --> 00:05:10,440
of the sort of social networking
tools that Meta offers, we also

109
00:05:10,440 --> 00:05:14,760
are an AI company and we have
open source AI tools, and I'll

110
00:05:14,760 --> 00:05:17,960
talk about a few of them today.
One is called Llama, which is a

111
00:05:18,040 --> 00:05:21,640
large language model, similar to
other models that power tools

112
00:05:21,640 --> 00:05:24,720
like ChatGPT.
We also have what are called

113
00:05:24,720 --> 00:05:27,920
computer vision models.
So these are AI models that work

114
00:05:27,920 --> 00:05:32,040
on imagery, namely things like
photographs, images, video.

115
00:05:32,920 --> 00:05:35,240
And we've made a lot of these
models open source, which means

116
00:05:35,240 --> 00:05:38,080
anyone can download them and use
them free of charge.

117
00:05:38,680 --> 00:05:41,400
And so in the context of
disaster response, we have been

118
00:05:41,400 --> 00:05:44,200
working with partners around the
United States and around the

119
00:05:44,200 --> 00:05:46,720
globe to figure out how
artificial intelligence, whether

120
00:05:46,720 --> 00:05:49,600
or not it's large language
models like Llama or computer

121
00:05:49,600 --> 00:05:52,200
vision models like Segment
anything, can essentially

122
00:05:52,200 --> 00:05:56,080
improve emergency response
times, improve repairings, great

123
00:05:56,080 --> 00:05:59,000
tools for first responders.
And we're already seeing that

124
00:05:59,000 --> 00:06:01,840
happen in collaboration with
nonprofits and research

125
00:06:01,840 --> 00:06:05,200
institutions across.
That's awesome.

126
00:06:05,200 --> 00:06:09,040
So if there's an agency, let's
say in Alaska, where I live that

127
00:06:09,040 --> 00:06:11,520
wanted to utilize these tools,
how would they go about

128
00:06:11,520 --> 00:06:15,240
utilizing these tools?
Is it just simple sign up or do

129
00:06:15,240 --> 00:06:17,000
they need to work with somebody
on your team?

130
00:06:17,000 --> 00:06:19,400
What are some of those steps
that people could take?

131
00:06:20,200 --> 00:06:22,120
Sure.
Starting with some of the more

132
00:06:22,120 --> 00:06:25,640
sort of on platform consumer
ready tools that we talked about

133
00:06:25,640 --> 00:06:28,480
at the very beginning, if you
are a local response agency in

134
00:06:28,480 --> 00:06:30,880
Alaska and you're not yet
enrolled in local alerts,

135
00:06:31,000 --> 00:06:33,120
there's a very straightforward
process to do that.

136
00:06:33,120 --> 00:06:37,440
If you have a Facebook page, you
can essentially see if you're on

137
00:06:37,440 --> 00:06:39,920
the Facebook page.
You can also Google the workflow

138
00:06:39,920 --> 00:06:42,640
very easily.
But for things like our AI

139
00:06:42,640 --> 00:06:45,320
tools, there's a number of
different ways you can have it.

140
00:06:45,320 --> 00:06:47,920
So if you have an engineering
team or if you have a team

141
00:06:48,360 --> 00:06:51,200
that's familiar with artificial
intelligence, these are things

142
00:06:51,200 --> 00:06:53,920
you can download and use right
out-of-the-box.

143
00:06:54,200 --> 00:06:58,040
We have a website, ai.meta.com,
and it has all of our open

144
00:06:58,040 --> 00:07:00,280
source models there.
Tools like Llama actually have

145
00:07:00,280 --> 00:07:02,840
their own website, llama.com,
again, which is our large

146
00:07:02,840 --> 00:07:04,560
language model.
But then there's another way you

147
00:07:04,560 --> 00:07:06,480
could go about it.
And that this is the way that we

148
00:07:06,480 --> 00:07:09,520
think most productive is we
think that in partnership with

149
00:07:09,520 --> 00:07:12,640
your local research institution,
you may already learn a lot.

150
00:07:12,640 --> 00:07:15,480
So when Meta releases these open
source models, you'll find that

151
00:07:15,480 --> 00:07:18,480
the research community uses them
sometimes the day that they're

152
00:07:18,480 --> 00:07:21,080
open sourced.
And so sometimes if you have a

153
00:07:21,080 --> 00:07:23,920
local research institution or
university that's doing research

154
00:07:23,920 --> 00:07:27,440
and disaster response, they have
made already gotten their hands

155
00:07:27,440 --> 00:07:30,920
dirty with some of Meta's open
source models and collaboration

156
00:07:30,920 --> 00:07:33,480
in that regard as well.
That's cool.

157
00:07:33,480 --> 00:07:36,880
So what about from like a
practical standpoint, how do

158
00:07:36,880 --> 00:07:40,440
these tools, the AI 1
specifically, how do they help

159
00:07:40,440 --> 00:07:44,160
during, you know, a hurricane or
an earthquake, let's say there

160
00:07:44,160 --> 00:07:48,240
is a institution that's using
them in a proactive way to, you

161
00:07:48,240 --> 00:07:51,360
know, help with disaster
preparedness.

162
00:07:51,520 --> 00:07:54,120
What does that look like?
Like, how are the tools actually

163
00:07:54,560 --> 00:07:58,640
practically helping that, you
know, average John and Joe out

164
00:07:58,640 --> 00:07:59,840
in the world?
Yeah.

165
00:07:59,960 --> 00:08:02,520
So when I think we think about
artificial intelligence, the

166
00:08:02,520 --> 00:08:06,440
major thing people should be
thinking is about making what

167
00:08:06,440 --> 00:08:11,080
were previously manual workflows
much more automated and faster.

168
00:08:11,600 --> 00:08:14,400
And so to give an example from
the computer vision space, this

169
00:08:14,400 --> 00:08:16,640
is a space that I spent a lot of
time working in.

170
00:08:16,640 --> 00:08:20,600
Again, this is AI that's focused
on analyzing imagery and video,

171
00:08:21,600 --> 00:08:24,880
more big issues across coastal
communities in the United

172
00:08:24,880 --> 00:08:28,320
States, Alaska included, in
terms of actually predicting

173
00:08:28,560 --> 00:08:30,840
floods.
And so normally what you're

174
00:08:30,840 --> 00:08:33,280
going to have to do in order to
predict if an area is going to

175
00:08:33,280 --> 00:08:37,320
flood is maybe you'll have some
sort of automated monitoring.

176
00:08:37,320 --> 00:08:40,039
This could be some sort of
camera sensor that's placed at

177
00:08:40,039 --> 00:08:42,400
the coastline.
But in traditional workflows,

178
00:08:42,400 --> 00:08:45,120
what you'll have to do is hire
somebody back at some

179
00:08:45,120 --> 00:08:49,000
headquarters to look at these
photos coming in from the field

180
00:08:49,000 --> 00:08:52,120
and do a bunch of manual
measurements and figure out, OK,

181
00:08:52,120 --> 00:08:56,080
the the sort of coast starts
here and, and the water seems

182
00:08:56,200 --> 00:08:58,520
moving at this rate.
And we believe that the, you

183
00:08:58,520 --> 00:09:01,520
know, the likelihood of flooding
in the next week is X.

184
00:09:01,520 --> 00:09:04,160
And it's all hyper manual and
requires a lot of people to do a

185
00:09:04,160 --> 00:09:06,520
lot of work.
What you can do with the advent

186
00:09:06,520 --> 00:09:09,320
of artificial intelligence and a
model like segment anything, a

187
00:09:09,320 --> 00:09:12,480
segment anything basically can
cut out any object from an

188
00:09:12,480 --> 00:09:14,240
image.
So in an automated fashion, I

189
00:09:14,240 --> 00:09:16,040
can say this is where the
coastline is.

190
00:09:16,040 --> 00:09:19,160
This is how quickly and the
calculations required become

191
00:09:19,160 --> 00:09:23,000
much faster to then the teams
actually sending notifications

192
00:09:23,000 --> 00:09:26,520
for things like evacuation
orders or other types of

193
00:09:26,520 --> 00:09:29,680
preparedness efforts that can
happen essentially without that

194
00:09:29,680 --> 00:09:31,600
manual workflow needing to take
place.

195
00:09:31,600 --> 00:09:34,080
And so we think there's already
again been a bunch of research

196
00:09:34,080 --> 00:09:36,760
collaborations that will have
used segment anything for

197
00:09:36,760 --> 00:09:39,160
coastal flood monitoring in an
automated fashion.

198
00:09:39,440 --> 00:09:44,120
And we think what the next phase
of was going to look like is

199
00:09:44,120 --> 00:09:47,160
essentially having those
research collaborations be

200
00:09:47,160 --> 00:09:50,840
brought into into Emergency
Management workflows and and

201
00:09:50,840 --> 00:09:54,400
being scaled up in a way that
makes these these evacuations

202
00:09:54,400 --> 00:09:57,720
and fairness efforts working.
That's great.

203
00:09:57,720 --> 00:10:01,200
So you know, lots of cities,
probably almost every county or

204
00:10:01,200 --> 00:10:04,520
borough across the US, including
here in Alaska, they have

205
00:10:04,520 --> 00:10:06,240
departments of Emergency
Management.

206
00:10:06,360 --> 00:10:11,960
And has Facebook found success
in working with these types of

207
00:10:11,960 --> 00:10:16,680
local Emergency Management
entities and is there training

208
00:10:16,680 --> 00:10:18,840
that comes along with it from
Facebook side?

209
00:10:19,480 --> 00:10:21,160
Sure.
So we've done this in a lot of

210
00:10:21,160 --> 00:10:23,080
states.
Generally speaking, again, I

211
00:10:23,080 --> 00:10:27,520
think it's a partnership model
that spans things like Meta

212
00:10:27,840 --> 00:10:30,360
research institutions are
nonprofits and the public

213
00:10:30,360 --> 00:10:32,520
sector.
So we tend to work actually as

214
00:10:32,520 --> 00:10:35,880
one big consortium of actors.
And so Meta will be the

215
00:10:35,880 --> 00:10:38,680
organization that supplies the
open source AI model.

216
00:10:38,680 --> 00:10:42,320
Again, we we make our many of
our models available in a

217
00:10:42,320 --> 00:10:44,520
fashion that's completely free
and available to anybody who

218
00:10:44,520 --> 00:10:47,280
wants to use them.
And then what we often do will

219
00:10:47,280 --> 00:10:51,720
partner with a local university.
So to give an example of Texas,

220
00:10:51,720 --> 00:10:54,160
we've been working with Texas
A&M for a number of years.

221
00:10:54,480 --> 00:10:57,280
They actually house a research
institution called the Institute

222
00:10:57,280 --> 00:11:00,000
for Disaster Resilient Texas
that was formed after Hurricane

223
00:11:00,000 --> 00:11:02,440
Harvey.
They're the ones taking Meadows

224
00:11:02,440 --> 00:11:06,520
Llama model and building custom
tooling for the Texas set of use

225
00:11:06,520 --> 00:11:08,840
cases and the Emergency
Management set of use cases.

226
00:11:09,080 --> 00:11:11,400
And then they're the ones
sharing that with the Emergency

227
00:11:11,400 --> 00:11:13,160
Management department in Harris
County.

228
00:11:13,800 --> 00:11:16,080
So I've gone to to Harris County
and trained people.

229
00:11:16,640 --> 00:11:19,440
So I've gone to California and
trained people on our tools.

230
00:11:19,440 --> 00:11:22,480
We just actually did a disaster
response work work back in

231
00:11:22,480 --> 00:11:25,440
Philadelphia that was really
more a preparedness effort in

232
00:11:25,440 --> 00:11:28,000
anticipation of the 2026 World
Cup game.

233
00:11:28,000 --> 00:11:31,000
So certainly work hand in hand
with local government, but we

234
00:11:31,000 --> 00:11:33,880
also try to bring our local
partners who can really provide

235
00:11:33,880 --> 00:11:38,360
that sustainability plan to be
able to use our tools locally

236
00:11:38,360 --> 00:11:41,960
for their own individual use
over the long term.

237
00:11:43,040 --> 00:11:45,800
Nice.
And, and, you know, with most of

238
00:11:45,800 --> 00:11:50,040
this being open source, there is
not a huge cost there to the to

239
00:11:50,040 --> 00:11:52,720
the to the borough, to the
Emergency Management side.

240
00:11:52,960 --> 00:11:56,640
But, you know, often times that
is a stumbling block for

241
00:11:57,040 --> 00:11:59,120
government agencies or
nonprofits.

242
00:11:59,520 --> 00:12:02,560
Even if something's open source,
it's tough to get a position

243
00:12:02,560 --> 00:12:06,240
funded or whatever it is.
Does Meta or Facebook do any

244
00:12:06,240 --> 00:12:10,240
sort of grants to help kind of
spur this work forward in a

245
00:12:10,240 --> 00:12:13,360
county or a city or borough?
Sure.

246
00:12:13,360 --> 00:12:16,520
We have a program called the
Llama Impact Grants program that

247
00:12:16,520 --> 00:12:19,480
looks at sort of seed funding
for exactly these kind of use

248
00:12:19,560 --> 00:12:22,440
phases.
We funded quite a number of

249
00:12:22,440 --> 00:12:25,560
social impact ventures around
the world essentially trying to

250
00:12:25,560 --> 00:12:28,080
ensure their AI models are used
in this regard.

251
00:12:28,360 --> 00:12:30,920
So we do have those for grant
programs in place.

252
00:12:30,960 --> 00:12:34,400
But again, I also think that
what's neat about partnering not

253
00:12:34,400 --> 00:12:37,240
just directly with the public
sector is it's really neat to

254
00:12:37,240 --> 00:12:40,680
see the way research
institutions can sometimes also

255
00:12:40,680 --> 00:12:42,640
tap into things like private
philanthropy.

256
00:12:42,640 --> 00:12:45,680
If they studies that as open
source models in an Emergency

257
00:12:45,680 --> 00:12:48,480
Management capacity or in
another social impact capacity,

258
00:12:48,480 --> 00:12:51,800
there's often additional funding
opportunities for those projects

259
00:12:51,800 --> 00:12:53,280
as well.
Nice.

260
00:12:53,520 --> 00:12:54,960
So, you know, this is your
world.

261
00:12:54,960 --> 00:12:58,320
You get to meet folks all over
the US prior over the globe.

262
00:12:58,440 --> 00:13:03,000
Do you have a moment in your job
in this capacity where you've

263
00:13:03,000 --> 00:13:05,840
kind of looked back and said,
man, we're making a we're

264
00:13:05,840 --> 00:13:07,320
actually making a difference
here.

265
00:13:07,320 --> 00:13:08,840
Tell me about what that looked
like for you.

266
00:13:10,440 --> 00:13:13,600
I think the peak of that was
actually probably during the

267
00:13:13,600 --> 00:13:17,920
COVID pandemic, so a different
set of use cases, not in in sort

268
00:13:17,920 --> 00:13:21,160
of disaster response and extreme
weather, but in this case public

269
00:13:21,160 --> 00:13:24,720
health emergencies.
So as you might imagine, before

270
00:13:24,720 --> 00:13:27,640
the onset of things like the
COVID vaccine, we were being

271
00:13:27,640 --> 00:13:30,280
told that there were few things
that we could do to keep

272
00:13:30,280 --> 00:13:33,160
ourselves safe.
One was to stay home, and there

273
00:13:33,160 --> 00:13:36,640
was to wear masks.
And I don't know if you're

274
00:13:36,640 --> 00:13:39,440
familiar with most doctors
offices or health surveillance

275
00:13:39,440 --> 00:13:42,760
systems, but they tend to not
have data on what percentage of

276
00:13:42,760 --> 00:13:44,920
the population is wearing masks
or staying home.

277
00:13:44,920 --> 00:13:47,400
These are not things we've
historically had to measure as

278
00:13:47,400 --> 00:13:50,760
part of a global pandemic.
And so the early version of my

279
00:13:50,760 --> 00:13:53,080
program actually focused much
more on open data.

280
00:13:53,880 --> 00:13:58,600
We wondered if I would say every
country I could name at the time

281
00:13:58,600 --> 00:14:02,720
to share open source mobility
data so that people could ask

282
00:14:02,720 --> 00:14:05,600
they'll track whether or not
populations were adhering to

283
00:14:05,600 --> 00:14:09,040
stay at home owners.
Facebook also did something

284
00:14:09,040 --> 00:14:12,040
which was incredibly impressive
is that we also launched a

285
00:14:12,040 --> 00:14:16,360
global survey in 200 countries
and 55 languages every single

286
00:14:16,360 --> 00:14:20,040
day for almost 2 years to try to
better understand at scale

287
00:14:20,040 --> 00:14:22,520
whether or not people were
wearing masks and other sort of

288
00:14:22,520 --> 00:14:25,240
preventive measures relevant to
the COVID pandemic.

289
00:14:25,680 --> 00:14:30,240
And those tools actually became
the go to data sources for

290
00:14:30,880 --> 00:14:33,960
international forecasting
institutions around the world

291
00:14:33,960 --> 00:14:35,400
and research institutions around
the world.

292
00:14:35,400 --> 00:14:38,280
There really wasn't any other
data from traditional sources

293
00:14:38,280 --> 00:14:41,360
that they could rely on.
So I think what global

294
00:14:41,360 --> 00:14:43,880
emergencies tell us is that, you
know, when we break down these

295
00:14:43,880 --> 00:14:46,640
institutional barrier, we really
just try to solve problems

296
00:14:46,640 --> 00:14:48,800
together.
We can get in quarter of work

297
00:14:48,800 --> 00:14:51,480
done and it's a period of my
career I'm very proud of and

298
00:14:51,760 --> 00:14:53,600
look back on.
It was obviously horrible that

299
00:14:53,600 --> 00:14:57,080
we were facing COVID at the time
in terms of the level of

300
00:14:57,080 --> 00:15:00,680
collaboration and just
cooperation that took place

301
00:15:00,680 --> 00:15:03,240
across public, private and and
nonprofit sectors.

302
00:15:03,240 --> 00:15:04,680
It was pretty exceptional in
that regard.

303
00:15:05,520 --> 00:15:07,120
Nice.
Well, my last question to you is

304
00:15:07,120 --> 00:15:11,600
this, Laura is what's it, what's
the maybe end goal or big

305
00:15:11,600 --> 00:15:15,480
success look like specifically
with the AI disaster

306
00:15:15,480 --> 00:15:17,960
preparedness tools that that
meta has?

307
00:15:18,320 --> 00:15:21,520
You know, let's say it's five
years from now, What do you

308
00:15:21,520 --> 00:15:25,920
consider is going to be the big
win for you all looking back and

309
00:15:25,920 --> 00:15:28,080
say, OK, we're doing something
good here.

310
00:15:28,080 --> 00:15:31,040
We're headed in the right
direction And you know, your

311
00:15:31,040 --> 00:15:33,440
team can sit back and say that's
a big W.

312
00:15:33,480 --> 00:15:34,720
What?
What's that look like for you?

313
00:15:35,760 --> 00:15:37,440
Yeah.
I mean, I think we would hope to

314
00:15:37,440 --> 00:15:40,360
be able to get ahead of things
as much as possible.

315
00:15:40,360 --> 00:15:44,400
And this is really challenging.
So I think for rural

316
00:15:44,400 --> 00:15:47,440
communities, it's going to be
how we work together with both

317
00:15:47,480 --> 00:15:51,600
online and offline solutions.
So if you think about a workflow

318
00:15:51,600 --> 00:15:53,960
that involves artificial
intelligence, how can something

319
00:15:53,960 --> 00:15:57,960
like a flood siren that may have
maybe has to work in an offline

320
00:15:58,000 --> 00:16:01,080
fashion still sort of get the
right notification and the right

321
00:16:01,080 --> 00:16:05,000
time to go off so that
communities can remove

322
00:16:05,000 --> 00:16:07,680
themselves from harm's way?
I mean, I work on a global

323
00:16:07,680 --> 00:16:10,760
program too.
So I think a lot about, I would

324
00:16:10,800 --> 00:16:13,760
love to be able to see what we
can achieve in, you know, an

325
00:16:13,840 --> 00:16:16,840
urban setting in New York City
or in California.

326
00:16:17,040 --> 00:16:19,760
I would want us to be able to be
able to achieve that not only in

327
00:16:19,760 --> 00:16:22,960
rural settings in the United
States, but around the world.

328
00:16:22,960 --> 00:16:26,680
And I think that's going to be,
we don't have to partner with

329
00:16:26,680 --> 00:16:30,000
the research community on.
And then also to really think

330
00:16:30,000 --> 00:16:33,960
about how we get these solutions
out of the very technical

331
00:16:33,960 --> 00:16:36,280
communities and sometimes in
which they're created and into

332
00:16:36,280 --> 00:16:38,320
the hands of people on the
ground.

333
00:16:38,720 --> 00:16:41,600
So at the end of the day, it's
about, you know, saving more

334
00:16:41,600 --> 00:16:44,280
lives from things like Rivera
natural disasters.

335
00:16:44,280 --> 00:16:46,480
And I think we it's a tall order
to get there for sure.

336
00:16:47,200 --> 00:16:48,480
Well, this has been exciting
folks.

337
00:16:48,480 --> 00:16:51,360
I'll put the links to those new
AI tools in the podcast

338
00:16:51,360 --> 00:16:53,440
description.
Laura, I run on.

339
00:16:53,640 --> 00:16:55,600
I want to thank you.
I appreciate you joining us here

340
00:16:55,600 --> 00:16:56,960
on the show.
Do you have any last minute

341
00:16:56,960 --> 00:16:58,960
thoughts here before we head
off?

342
00:16:58,960 --> 00:17:02,080
The floor is yours.
Oh well, I just thank you for

343
00:17:02,080 --> 00:17:04,599
having me, Jonathan.
Happy Thanksgiving and wishing

344
00:17:04,599 --> 00:17:06,599
everybody a safe and happy
holiday season.

345
00:17:07,400 --> 00:17:08,400
Awesome.
Well, thanks for joining us.

346
00:17:08,400 --> 00:17:09,480
You're welcome back anytime.