1
00:00:02,043 --> 00:00:03,930
This is Henry,

2
00:00:03,930 --> 00:00:05,328
a cute boy,

3
00:00:05,328 --> 00:00:07,834
and when Henry was three,

4
00:00:07,834 --> 00:00:13,387
his mom found him having
some febrile seizures.

5
00:00:13,575 --> 00:00:13,958
Febrile seizures are seizures that occur

6
00:00:13,958 --> 00:00:18,253
when you also have a fever,

7
00:00:18,253 --> 00:00:20,980
and the doctor said,

8
00:00:20,980 --> 00:00:23,746
"Don't worry too much.
Kids usually outgrow these."

9
00:00:23,746 --> 00:00:26,144
When he was four,

10
00:00:26,144 --> 00:00:27,838
he had a convulsive seizure,

11
00:00:27,838 --> 00:00:30,909
the kind that you lose consciousness

12
00:00:30,909 --> 00:00:31,740
and shake,

13
00:00:31,740 --> 00:00:33,957
generalized tonic-clonic seizure,

14
00:00:33,957 --> 00:00:38,456
and while the diagnosis of epilepsy

15
00:00:38,456 --> 00:00:41,672
was in the mail,

16
00:00:41,672 --> 00:00:45,141
Henry's mom went to get him
out of bed one morning,

17
00:00:45,525 --> 00:00:48,939
and as she went in his room,

18
00:00:48,939 --> 00:00:52,039
she found his cold, lifeless body.

19
00:00:52,039 --> 00:00:58,302
Henry died of SUDEP,

20
00:00:58,302 --> 00:01:01,867
Sudden Unexpected Death in Epilepsy.

21
00:01:01,867 --> 00:01:05,446
I'm curious how many of you
have heard of SUDEP?

22
00:01:05,796 --> 00:01:10,233
This is a very well-educated audience,
and I see only a few hands.

23
00:01:10,233 --> 00:01:14,113
SUDEP is when an otherwise
healthy person with epilepsy

24
00:01:14,113 --> 00:01:18,113
dies and they can't attribute it
to anything they can find in an autopsy.

25
00:01:20,322 --> 00:01:24,434
There is a SUDEP
every seven to nine minutes.

26
00:01:24,434 --> 00:01:27,313
That's on average two per TED Talk.

27
00:01:31,651 --> 00:01:33,315
Now, a normal brain

28
00:01:33,315 --> 00:01:35,392
has electrical activity.

29
00:01:35,392 --> 00:01:38,170
You can see some of the electrical waves

30
00:01:38,170 --> 00:01:40,792
coming out of this picture
of a brain here.

31
00:01:40,792 --> 00:01:44,408
And these should look
like typical electrical activity

32
00:01:44,408 --> 00:01:46,465
that an EEG could read on the surface.

33
00:01:46,465 --> 00:01:48,255
When you have a seizure,

34
00:01:48,255 --> 00:01:50,951
it's a bit of unusual electrical activity,

35
00:01:50,951 --> 00:01:52,772
and it can be focal.

36
00:01:52,772 --> 00:01:54,863
It can take place in just
a small part of your brain.

37
00:01:54,863 --> 00:01:56,463
When that happens,

38
00:01:56,463 --> 00:01:58,385
you might have a strange sensation.

39
00:01:58,385 --> 00:02:01,160
One could be happening,
several could be happening

40
00:02:01,160 --> 00:02:02,511
here in the audience right now

41
00:02:02,511 --> 00:02:04,323
and the person next to you
might not even know.

42
00:02:04,323 --> 00:02:08,082
However, if you have a seizure
where that little brush fire

43
00:02:08,082 --> 00:02:10,359
spreads like a forest fire over the brain,

44
00:02:10,359 --> 00:02:12,059
then it generalizes,

45
00:02:12,059 --> 00:02:14,317
and that generalized seizure

46
00:02:14,317 --> 00:02:16,581
takes your consciousness away

47
00:02:16,581 --> 00:02:18,968
and causes you to convulse.

48
00:02:18,968 --> 00:02:22,947
There are more SUDEPs
in the United States every year

49
00:02:22,947 --> 00:02:25,622
than Sudden Infant Death Syndrome.

50
00:02:25,622 --> 00:02:29,464
Now, how many of you have heard
of Sudden Infant Death Syndrome.

51
00:02:29,464 --> 00:02:31,544
Right? Pretty much every hand goes up.

52
00:02:31,544 --> 00:02:33,566
So what's going on here?

53
00:02:33,566 --> 00:02:38,405
Why is this so much more common
and yet people haven't heard of it?

54
00:02:38,609 --> 00:02:40,934
And what can you do to prevent it?

55
00:02:40,934 --> 00:02:44,500
Well, there are two things
scientifically shown

56
00:02:44,500 --> 00:02:47,355
that prevent or reduce the risk of SUDEP.

57
00:02:47,355 --> 00:02:49,863
The first is follow
your doctor's instructions,

58
00:02:49,863 --> 00:02:51,306
take your medications.

59
00:02:51,306 --> 00:02:53,619
Two thirds of people who have epilepsy

60
00:02:53,619 --> 00:02:55,738
get it under control
with their medications.

61
00:02:56,088 --> 00:03:00,283
the second thing that reduces
the risk of SUDEP is companionship.

62
00:03:00,283 --> 00:03:05,001
It's having somebody there
at the time that you have a seizure.

63
00:03:05,001 --> 00:03:09,270
Now, SUDEP, even though
most of you have never heard of it,

64
00:03:09,270 --> 00:03:13,197
is actually the number two cause
of years of potential life lost

65
00:03:13,197 --> 00:03:17,159
of all neurological disorders.

66
00:03:17,159 --> 00:03:25,128
The vertical axis is the number of deaths
times the remaining lifespan,

67
00:03:25,351 --> 00:03:28,814
so higher is much worse impact.

68
00:03:28,814 --> 00:03:31,760
SUDEP, however, unlike these others

69
00:03:31,760 --> 00:03:37,476
is something that people right here
could do something to push that down.

70
00:03:37,964 --> 00:03:45,185
Now, what is Ros Picard, an AI researcher,
doing here telling you about SUDEP? Right?

71
00:03:45,185 --> 00:03:47,587
I'm not a neurologist.

72
00:03:47,587 --> 00:03:50,349
When I was working at the Media Lab

73
00:03:50,349 --> 00:03:51,927
on measurement of emotion,

74
00:03:51,927 --> 00:03:55,072
trying to make our machines
more intelligent about our emotions,

75
00:03:55,072 --> 00:03:58,011
we started doing a lot of work
measuring stress.

76
00:03:58,011 --> 00:04:00,704
We built lots of sensors

77
00:04:00,704 --> 00:04:03,817
that measured it
in lots of different ways.

78
00:04:03,817 --> 00:04:06,161
But one of them in particular

79
00:04:06,161 --> 00:04:10,549
grew out of some of this very old work
with measuring sweaty palms

80
00:04:10,549 --> 00:04:12,121
with an electrical signal.

81
00:04:12,121 --> 00:04:14,655
This is a signal of skin conductance
that's known to go up

82
00:04:14,655 --> 00:04:17,229
when you get nervous,

83
00:04:17,229 --> 00:04:19,380
but it turns out it also goes up with
a lot of other interesting conditions.

84
00:04:19,380 --> 00:04:22,153
But measuring it with wires on your hand
is really inconvenient.

85
00:04:22,153 --> 00:04:26,284
So we invented a bunch of other ways
of doing this at the MIT Media Lab,

86
00:04:26,284 --> 00:04:28,588
And with these wearables,

87
00:04:28,588 --> 00:04:33,417
we started to collect the first ever
clinical quality data 24/7.

88
00:04:33,417 --> 00:04:38,705
Here's a picture of what that looked like
the first time that an MIT student

89
00:04:38,705 --> 00:04:43,025
collected skin conductance
on the wrist 24/7.

90
00:04:43,025 --> 00:04:45,696
Let's zoom in a little bit here.

91
00:04:45,696 --> 00:04:48,998
What you see is 24 hours
from left to right,

92
00:04:48,998 --> 00:04:50,593
and here is two days of data,

93
00:04:50,593 --> 00:04:53,515
and first what's surprised us

94
00:04:53,515 --> 00:04:56,885
was the sleep was
the biggest peak of the day.

95
00:04:56,885 --> 00:04:58,659
Now, that sound broken. Right?

96
00:04:58,659 --> 00:05:01,182
You're calm when you're asleep,
so what's going on here?

97
00:05:01,182 --> 00:05:05,616
Well, it turns out that
our physiology during sleep

98
00:05:05,616 --> 00:05:08,342
is very different than
our physiology during wake,

99
00:05:08,342 --> 00:05:11,855
and while there's still a bit of a mystery
why these peaks are usually

100
00:05:11,855 --> 00:05:13,981
the biggest of the day during sleep,

101
00:05:13,981 --> 00:05:17,255
we now believe they're related
to memory consolidation

102
00:05:17,255 --> 00:05:19,882
and memory formation during sleep.

103
00:05:20,135 --> 00:05:23,386
We also saw things that were
exactly what we expected.

104
00:05:23,386 --> 00:05:26,426
When an MIT student
is working hard in the lab

105
00:05:26,426 --> 00:05:28,117
or on homeworks,

106
00:05:28,117 --> 00:05:31,925
there is not only emotional stress,
but there's cognitive load,

107
00:05:32,088 --> 00:05:36,973
and it turns out that cognitive load,
cognitive effort, mental engagement,

108
00:05:36,973 --> 00:05:39,120
excitement about learning something,

109
00:05:39,120 --> 00:05:41,595
those things also make the signal go up.

110
00:05:42,818 --> 00:05:47,031
Unfortunately, to the embarrassment
of the MIT professors,

111
00:05:47,031 --> 00:05:52,070
the low point every day
is classroom activity.

112
00:05:53,050 --> 00:05:55,191
Now, I am just showing you
one person's data here,

113
00:05:55,191 --> 00:05:58,755
but this unfortunately is true in general.

114
00:06:00,712 --> 00:06:05,117
This sweatband has inside it
a home-built skin conductance sensor,

115
00:06:05,117 --> 00:06:09,137
and one day, one of our undergrads
knocked on my door

116
00:06:09,137 --> 00:06:13,093
right at the end of the December semester,

117
00:06:13,093 --> 00:06:15,621
and he said, "Professor Picard,

118
00:06:15,621 --> 00:06:17,850
can I please borrow
one of your wrist band sensors?

119
00:06:17,850 --> 00:06:20,151
My little brother has autism,

120
00:06:20,151 --> 00:06:21,386
he can't talk,

121
00:06:21,386 --> 00:06:24,497
and I want to see
what's stressing him out."

122
00:06:24,770 --> 00:06:27,493
And I said, "Sure, in fact,
don't just take one, take two,"

123
00:06:27,493 --> 00:06:30,280
because they broke easily back then.

124
00:06:30,280 --> 00:06:32,534
So he took them home,
he put them on his little brother.

125
00:06:32,534 --> 00:06:36,209
Now, I was back in MIT
looking at the data on my laptop,

126
00:06:36,209 --> 00:06:38,775
and the first day, I thought,
hmm, that's odd,

127
00:06:38,775 --> 00:06:41,852
he put them on both wrists
instead of waiting for one to break.

128
00:06:42,014 --> 00:06:44,453
OK, fine, don't follow my instructions.

129
00:06:44,453 --> 00:06:46,195
I'm glad he didn't.

130
00:06:46,195 --> 00:06:50,157
Second day, chill,
looked like classroom activity.

131
00:06:50,741 --> 00:06:53,644
A few more days ahead.

132
00:06:53,644 --> 00:06:58,182
The next day, one missed signal was flat

133
00:06:58,182 --> 00:07:02,140
and the other had
the biggest peak I've ever seen,

134
00:07:02,736 --> 00:07:05,078
and I thought, what's going on?

135
00:07:05,078 --> 00:07:08,562
We've stressed people out at MIT
every way imaginable.

136
00:07:08,562 --> 00:07:12,254
I've never seen a peak this big.

137
00:07:12,254 --> 00:07:14,732
And it was only on one side.

138
00:07:14,732 --> 00:07:17,483
How can you be stressed on one side
of your body and not the other?

139
00:07:17,483 --> 00:07:20,950
So I thought one or both
sensors must be broken.

140
00:07:21,473 --> 00:07:23,840
Now, I'm an electro engineer by training,

141
00:07:23,840 --> 00:07:26,184
so I started a whole bunch of stuff
to try to debug this,

142
00:07:26,184 --> 00:07:28,110
and long story short,
I could not reproduce this.

143
00:07:28,110 --> 00:07:32,259
So I resorted to old-fashioned debugging.

144
00:07:32,259 --> 00:07:35,522
I called the student at home on vacation.

145
00:07:35,748 --> 00:07:39,932
"Hi, how's your little brother?
How's your Christmas?

146
00:07:39,932 --> 00:07:43,150
Hey, do you have any idea
what happened to him?"

147
00:07:43,150 --> 00:07:45,316
And I gave this particular date and time,

148
00:07:45,316 --> 00:07:46,698
and the data.

149
00:07:46,698 --> 00:07:50,019
And he said, "I don't know,
I'll check the diary."

150
00:07:50,019 --> 00:07:54,022
Diary? MIT student keeps a diary?

151
00:07:54,022 --> 00:07:55,865
So I waited and he came back.

152
00:07:55,865 --> 00:07:57,428
He had the exact date and time,

153
00:07:57,428 --> 00:08:01,281
and he says, "That was right before
he had a grand mal seizure."

154
00:08:02,188 --> 00:08:06,312
Now, at the time, I didn't know
anything about epilepsy,

155
00:08:06,312 --> 00:08:08,851
and did a bunch of research,

156
00:08:08,851 --> 00:08:11,767
realized that another student's dad
is chief of neurosurgery

157
00:08:11,767 --> 00:08:13,781
at Children's Hospital Boston,

158
00:08:13,781 --> 00:08:16,139
screwed up my courage
and called Dr. Joe Madsen.

159
00:08:16,139 --> 00:08:18,876
"Hi, Dr. Madsen,
my name's Rosalind Picard.

160
00:08:18,876 --> 00:08:23,075
Is it possible somebody could have

161
00:08:23,075 --> 00:08:27,065
a huge sympathetic
nervous system surge" --

162
00:08:27,065 --> 00:08:29,406
that's what drives the skin conductance --

163
00:08:29,406 --> 00:08:32,158
"20 minutes before a seizure?"

164
00:08:32,158 --> 00:08:34,865
And he says, "Probably not."

165
00:08:34,865 --> 00:08:37,073
He says, "It's interesting.

166
00:08:37,073 --> 00:08:40,060
We've had people whose hair
stands on end on one arm

167
00:08:40,060 --> 00:08:42,491
20 minutes before a seizure."

168
00:08:42,491 --> 00:08:44,439
And I'm like, "On one arm?"

169
00:08:44,439 --> 00:08:46,793
I didn't want to tell him that initially

170
00:08:46,793 --> 00:08:49,375
because I thought this was too ridiculous.

171
00:08:49,375 --> 00:08:51,284
He explained how
this could happen in the brain,

172
00:08:51,284 --> 00:08:53,720
and he got interested.
I showed him the data.

173
00:08:53,720 --> 00:08:56,092
We made a whole bunch more devices,
got them safety certified,

174
00:08:56,092 --> 00:08:58,390
90 families were being enrolled in a study

175
00:08:58,390 --> 00:09:01,890
all with children who were going
to be monitored 24/7

176
00:09:01,890 --> 00:09:05,403
with gold standard EEG on their scalp

177
00:09:05,403 --> 00:09:06,885
for reading the brain activity,

178
00:09:06,885 --> 00:09:08,814
video to watch the behavior,

179
00:09:08,814 --> 00:09:12,656
electrocardiogram ECG,
and now EDA, electrodermal activity,

180
00:09:12,656 --> 00:09:15,015
to see if there was
something in this periphery

181
00:09:15,015 --> 00:09:17,711
that we could easily pick up
related to a seizure.

182
00:09:18,363 --> 00:09:24,968
We found in 100 percent
of the first batch of grand mal seizures

183
00:09:24,968 --> 00:09:28,625
these whopper of responses
in the skin conductance.

184
00:09:28,625 --> 00:09:30,652
The blue in the middle, the boy's sleep,

185
00:09:30,652 --> 00:09:32,648
is usually the biggest peak of the day.

186
00:09:32,648 --> 00:09:36,161
These three seizures you see here
are popping out of the forest

187
00:09:36,161 --> 00:09:38,783
like redwood trees.

188
00:09:39,238 --> 00:09:43,142
Furthermore, when you couple
these skin conductance at the top

189
00:09:43,142 --> 00:09:46,481
with the movement from the wrist

190
00:09:46,481 --> 00:09:51,448
and you get lots of data
and train machine learning and AI on it,

191
00:09:51,448 --> 00:09:54,486
you can build an automated AI
that detects these patterns

192
00:09:54,486 --> 00:10:00,418
much better than just
a shake detector can do.

193
00:10:00,646 --> 00:10:04,669
So we realized that we needed
to get this out,

194
00:10:04,669 --> 00:10:07,694
and with the PhD work
of ??

195
00:10:07,694 --> 00:10:10,085
and later great
improvements by Empatica,

196
00:10:10,085 --> 00:10:13,953
this has made progress and the seizure
detection is much more accurate.

197
00:10:13,953 --> 00:10:16,776
But we also learned some other things
about SUDEP during this.

198
00:10:16,776 --> 00:10:18,278
One thing we learned

199
00:10:18,278 --> 00:10:20,116
is that SUDEP is,

200
00:10:20,116 --> 00:10:23,254
while it's rare after
a generalized tonic-clonic seizure,

201
00:10:23,254 --> 00:10:26,515
that's when it's most likely
to happen after that type,

202
00:10:26,515 --> 00:10:29,554
and when it happens,
it doesn't happen during the seizure,

203
00:10:29,554 --> 00:10:32,712
and it doesn't usually happen
immediately afterwards,

204
00:10:32,712 --> 00:10:36,820
but immediately afterwards when the person
just seems very still and quiet,

205
00:10:36,820 --> 00:10:42,439
they may go into another phase
where the breathing stops,

206
00:10:42,439 --> 00:10:45,194
and then after the breathing stops,
later the heart stops.

207
00:10:45,392 --> 00:10:47,459
So there's some time
to get somebody there.

208
00:10:47,459 --> 00:10:53,445
We also learned that there is a region
deep in the brain called the amygdala,

209
00:10:53,445 --> 00:10:56,366
which we had been studying
in our emotion research a lot.

210
00:10:56,366 --> 00:10:57,782
We have two amygdalas,

211
00:10:57,782 --> 00:10:59,711
and if you stimulate the right one,

212
00:10:59,711 --> 00:11:02,366
you get a big right
skin conductance response.

213
00:11:02,565 --> 00:11:06,538
Now, you have to sign up right now
for a craniotomy to get this done,

214
00:11:06,728 --> 00:11:09,140
not exactly something
you're going to volunteer to do,

215
00:11:09,140 --> 00:11:11,601
but it causes a big right skin
conductance response.

216
00:11:11,601 --> 00:11:15,556
Stimulate the left one, big left
skin conductance response on the palm.

217
00:11:15,782 --> 00:11:17,404
And furthermore,

218
00:11:17,404 --> 00:11:19,846
when somebody stimulates your amygdala

219
00:11:19,846 --> 00:11:22,542
while you're sitting there
and you might just be working,

220
00:11:22,542 --> 00:11:25,201
you don't show any signs of distress,

221
00:11:25,201 --> 00:11:28,404
but you stop breathing,

222
00:11:28,710 --> 00:11:32,839
and you don't start again
until somebody stimulates you.

223
00:11:32,839 --> 00:11:34,376
"Hey, Ros, are you there?"

224
00:11:34,376 --> 00:11:36,583
And you open your mouth to talk.

225
00:11:36,583 --> 00:11:39,326
As you take that breath to speak,

226
00:11:39,326 --> 00:11:41,191
you start breathing again.

227
00:11:43,002 --> 00:11:46,841
So we had started with work on stress,

228
00:11:46,841 --> 00:11:50,446
which had enabled us
to build lots of sensors

229
00:11:50,446 --> 00:11:51,893
that were gathering
high quality enough data

230
00:11:51,893 --> 00:11:53,347
that we could leave the lab
and start to get this in the wild,

231
00:11:53,347 --> 00:11:56,665
accidentally found a whopper
of a response with the seizure,

232
00:11:56,665 --> 00:11:59,508
neurological activation that can cause
a much bigger response

233
00:11:59,508 --> 00:12:00,894
than traditional stressors,

234
00:12:00,894 --> 00:12:02,959
lots of partnership with hospitals
and an epilepsy monitoring unit,

235
00:12:02,959 --> 00:12:05,968
especially Children's Hospital Boston

236
00:12:05,968 --> 00:12:07,983
and the Brigham,

237
00:12:07,983 --> 00:12:09,997
and machine learning and AI on top of this

238
00:12:09,997 --> 00:12:13,051
to take and collect lots more data

239
00:12:13,051 --> 00:12:15,907
in service of trying
to understand these events

240
00:12:15,907 --> 00:12:18,005
and if we could prevent SUDEP.

241
00:12:18,263 --> 00:12:21,253
This is now commercialized by Empatica,

242
00:12:21,253 --> 00:12:24,763
a startup that I had
the privilege to cofound,

243
00:12:24,763 --> 00:12:27,035
and the team there has done an amazing job
improving the technology

244
00:12:27,035 --> 00:12:30,776
to make a very beautiful sensor

245
00:12:30,776 --> 00:12:34,391
that not only tells time and does steps
and sleep and all that good stuff,

246
00:12:34,391 --> 00:12:37,857
but this is running real time
AI and machine learning

247
00:12:37,857 --> 00:12:40,335
to detect generalized
tonic-clonic seizures

248
00:12:40,335 --> 00:12:42,596
and send an alert for help

249
00:12:42,596 --> 00:12:45,812
if I were to have a seizure
and lose consciousness.

250
00:12:45,812 --> 00:12:48,652
This just got FDA-approved

251
00:12:48,652 --> 00:12:50,774
as the first smartwatch

252
00:12:50,774 --> 00:12:52,792
to get approved in neurology.

253
00:12:54,235 --> 00:13:00,552
(Applause)

254
00:13:02,914 --> 00:13:07,112
Now the next slide is what made
my skin conductance go up.

255
00:13:07,242 --> 00:13:08,971
One morning, I'm checking my email

256
00:13:08,971 --> 00:13:11,214
and I see a story from a mom

257
00:13:11,214 --> 00:13:13,012
who said she was in the shower

258
00:13:13,012 --> 00:13:15,547
and her phone was
on the counter by the shower,

259
00:13:15,547 --> 00:13:18,436
and it said her daughter
might need her help.

260
00:13:18,436 --> 00:13:21,239
So she interrupts her shower and goes
running to her daughter's bedroom,

261
00:13:21,239 --> 00:13:25,608
and she finds her daughter
face down in bed blue and not breathing.

262
00:13:25,608 --> 00:13:29,046
She flips her over, human stimulation,

263
00:13:29,046 --> 00:13:32,746
and her daughter takes a breath,
and another breath,

264
00:13:32,936 --> 00:13:36,890
and her daughter turns pink and is fine.

265
00:13:37,791 --> 00:13:40,851
I think I turned white reading this email.

266
00:13:40,851 --> 00:13:43,385
My first response is,
"Oh no, it's not perfect.

267
00:13:43,385 --> 00:13:45,234
The Bluetooth could break.
The battery could die.

268
00:13:45,234 --> 00:13:47,590
All these things could go wrong.
Don't rely on this."

269
00:13:47,590 --> 00:13:49,435
And she said, "It's OK.

270
00:13:49,435 --> 00:13:51,667
I know no technology is perfect.

271
00:13:51,667 --> 00:13:54,601
None of us can always
be there all the time.

272
00:13:54,601 --> 00:13:59,450
But this, this device plus AI

273
00:13:59,450 --> 00:14:02,984
enabled me to get there in time
to save my daughter's life.

274
00:14:06,245 --> 00:14:08,103
Now, I've been mentioning children,

275
00:14:08,103 --> 00:14:13,834
but SUDEP peaks actually among people
in their 20s, 30s, and 40s,

276
00:14:13,834 --> 00:14:16,832
and the next line I'm going to put up

277
00:14:16,832 --> 00:14:18,044
is probably going to make
some people uncomfortable,

278
00:14:18,044 --> 00:14:22,908
but it's less uncomfortable
than we'll all be

279
00:14:22,908 --> 00:14:24,807
if this list is extended
to somebody you know.

280
00:14:24,807 --> 00:14:27,437
Could this happen to somebody you know?

281
00:14:27,437 --> 00:14:30,133
And the reason I bring up
this uncomfortable question

282
00:14:30,133 --> 00:14:34,926
is because one in 26 of you
will have epilepsy at some point,

283
00:14:34,926 --> 00:14:37,398
and from what I've been learning,

284
00:14:37,398 --> 00:14:40,547
people with epilepsy often don't tell
their friends and their neighbors

285
00:14:40,547 --> 00:14:41,698
that they have it.

286
00:14:41,698 --> 00:14:46,995
So if you're willing to let them
use an AI or whatever

287
00:14:46,995 --> 00:14:51,321
to summon you in a moment
of possible need,

288
00:14:51,321 --> 00:14:53,032
if you would let them know that,

289
00:14:53,032 --> 00:14:55,386
you could make a difference in their life.

290
00:14:56,000 --> 00:15:00,397
Why do all this hard work to build AIs?

291
00:15:00,397 --> 00:15:01,927
A couple of reasons here:

292
00:15:01,927 --> 00:15:03,856
one is Natasha, the girl who lived,

293
00:15:03,856 --> 00:15:07,575
and her family wanted me
to tell you her name.

294
00:15:07,575 --> 00:15:09,115
Another is her family

295
00:15:09,115 --> 00:15:11,849
and the wonderful people out there

296
00:15:11,849 --> 00:15:13,970
who want to be there to support people
who have conditions

297
00:15:13,970 --> 00:15:18,063
that they've felt uncomfortable
in the past mentioning to others.

298
00:15:18,253 --> 00:15:20,311
And the other reason is all of you,

299
00:15:20,311 --> 00:15:22,849
because we have the opportunity

300
00:15:22,849 --> 00:15:25,481
to shape the future of AI.

301
00:15:25,481 --> 00:15:28,273
We can actually change it,

302
00:15:28,273 --> 00:15:30,203
because we are the ones building it.

303
00:15:30,203 --> 00:15:32,352
So let's build AI

304
00:15:32,352 --> 00:15:35,991
that makes everybody's lives better.

305
00:15:35,991 --> 00:15:38,588
Thank you.

306
00:15:38,588 --> 00:15:41,691
(Applause)