WEBVTT
1
00:00:00.500 --> 00:00:03.917
[intriguing piano music]
2
00:00:11.130 --> 00:00:13.700
Alexa, play classical music.
3
00:00:13.700 --> 00:00:16.280
Here's a station for classical music.
4
00:00:16.280 --> 00:00:17.600
Well my name is Jerry Neece.
5
00:00:17.600 --> 00:00:21.270
I'm 70 years old and I live in the active adult community
6
00:00:21.270 --> 00:00:23.946
for people 55 and over here in San Jose, California.
7
00:00:23.946 --> 00:00:26.863
[calm piano music]
8
00:00:29.790 --> 00:00:33.362
I actually have five desktops, four laptops,
9
00:00:33.362 --> 00:00:36.273
two Kindles and an iPad.
10
00:00:37.700 --> 00:00:41.210
I have Amazon Alexa, I've had her for a couple years now
11
00:00:41.210 --> 00:00:43.890
and as someone that didn't even see my first television
12
00:00:43.890 --> 00:00:45.740
'til I was seven years old I can tell you
13
00:00:45.740 --> 00:00:48.440
that the evolution of technology is inevitable.
14
00:00:48.440 --> 00:00:50.580
It benefits mankind.
15
00:00:50.580 --> 00:00:52.860
I can see the stars and the milky way,
16
00:00:52.860 --> 00:00:55.440
I can actually spend all day under here.
17
00:00:55.440 --> 00:00:57.940
Artificial intelligence is all around us
18
00:00:57.940 --> 00:01:00.280
and we use it in different ways everyday.
19
00:01:00.280 --> 00:01:01.407
[phone alarm rings]
20
00:01:01.407 --> 00:01:03.250
You wake up, you go for a run
21
00:01:03.250 --> 00:01:06.140
and your watch tracks where you're going,
22
00:01:06.140 --> 00:01:10.050
measures your heart rate variability, using forms of AI.
23
00:01:10.050 --> 00:01:13.720
Probably AI was used by a farmer to grow the crops,
24
00:01:13.720 --> 00:01:15.210
and the strawberries and the blueberries
25
00:01:15.210 --> 00:01:17.290
that I had for breakfast.
26
00:01:17.290 --> 00:01:19.130
Maybe you're in a car that has AI
27
00:01:19.130 --> 00:01:22.170
that helps sense the other vehicles on the road around it.
28
00:01:22.170 --> 00:01:23.670
You sit down at your computer,
29
00:01:23.670 --> 00:01:24.930
you start to use your email.
30
00:01:24.930 --> 00:01:26.830
It's all filtered by AI.
31
00:01:26.830 --> 00:01:28.840
Then you take a photograph and the tools
32
00:01:28.840 --> 00:01:32.240
that help you sort your photographs, those are AI too.
33
00:01:32.240 --> 00:01:33.737
Artificial intelligence is everywhere.
34
00:01:33.737 --> 00:01:36.940
And it's becoming evermore present in our lives.
35
00:01:36.940 --> 00:01:39.440
What is the temperature in San Jose, California?
36
00:01:40.510 --> 00:01:43.410
AI and machine learning is the biggest revolution today.
37
00:01:44.361 --> 00:01:46.700
And this in the order of the agricultural,
38
00:01:46.700 --> 00:01:48.630
industrial revolution in the past.
39
00:01:48.630 --> 00:01:51.140
I think technology is gonna have a really hard time
40
00:01:51.140 --> 00:01:52.380
helping people like me.
41
00:01:52.380 --> 00:01:53.520
There's a lot of hype,
42
00:01:53.520 --> 00:01:55.270
but there's a lot happening as well.
43
00:01:55.270 --> 00:01:56.103
Borden pasta.
44
00:01:56.103 --> 00:01:57.750
One of the great things about machine learning,
45
00:01:57.750 --> 00:02:00.140
is how it's been democratized.
46
00:02:00.140 --> 00:02:01.870
A sort of huge jump has happened
47
00:02:01.870 --> 00:02:03.630
and all of a sudden you're getting confronted
48
00:02:03.630 --> 00:02:05.760
with the fact that there might not be driving a car
49
00:02:05.760 --> 00:02:07.150
in the next 10 years.
50
00:02:07.150 --> 00:02:08.520
Because it's software,
51
00:02:08.520 --> 00:02:11.617
the rate of change is so much faster.
52
00:02:11.617 --> 00:02:13.750
[computer clicking]
53
00:02:13.750 --> 00:02:16.410
This technology will be created.
54
00:02:16.410 --> 00:02:20.260
We have no escape from it now.
55
00:02:20.260 --> 00:02:24.060
People may think AI's gonna take over the world.
56
00:02:24.060 --> 00:02:25.220
They're definitely nervous.
57
00:02:25.220 --> 00:02:27.560
Some of those dystopian things that we might think,
58
00:02:27.560 --> 00:02:30.340
oh, that could never happen, might actually happen.
59
00:02:30.340 --> 00:02:33.360
To me it's not what's gonna happen to us
60
00:02:33.360 --> 00:02:37.400
but what's possible to happen with us,
61
00:02:37.400 --> 00:02:39.100
and where can we go?
62
00:02:39.100 --> 00:02:41.232
I mean it's sort of unlimited.
63
00:02:41.232 --> 00:02:44.565
[dramatic music swells]
64
00:02:50.285 --> 00:02:53.840
[melodic piano music]
65
00:02:53.840 --> 00:02:56.293
We are progressing rapidly in AI.
66
00:02:57.210 --> 00:02:59.340
Companies are putting billions of dollars into it.
67
00:02:59.340 --> 00:03:00.960 line:15%
Smartest people in the world are studying it.
68
00:03:00.960 --> 00:03:03.630 line:15%
It is going very, very fast.
69
00:03:03.630 --> 00:03:05.210
But the tools are accessible.
70
00:03:05.210 --> 00:03:07.453
Even eight year olds can learn about it.
71
00:03:08.510 --> 00:03:09.780
Good morning guys!
72
00:03:09.780 --> 00:03:12.680 line:15%
Today, we're going to be working and talking a little bit
73
00:03:12.680 --> 00:03:17.360 line:15%
about something called artificial intelligence.
74
00:03:17.360 --> 00:03:20.040
What do you think of when you hear those two words,
75
00:03:20.040 --> 00:03:22.650
artificial intelligence?
76
00:03:22.650 --> 00:03:27.250
Things like in video games, label on top of them AI.
77
00:03:27.250 --> 00:03:29.860
When I think of artificial intelligence I think of like,
78
00:03:29.860 --> 00:03:31.510
how smart robots are.
79
00:03:31.510 --> 00:03:32.580
How smart robots are?
80
00:03:32.580 --> 00:03:33.415
Okay.
81
00:03:33.415 --> 00:03:38.415
It has this control that just tells it to do things
82
00:03:39.130 --> 00:03:40.440
that it does.
83
00:03:40.440 --> 00:03:41.520
I like that.
84
00:03:41.520 --> 00:03:42.353
So, ready?
85
00:03:42.353 --> 00:03:44.220
I'm gonna show you guys and introduce you
86
00:03:44.220 --> 00:03:47.860
to an artificial intelligence powered robot,
87
00:03:47.860 --> 00:03:49.550
called Sophia.
88
00:03:49.550 --> 00:03:51.210
Sophia, if you could,
89
00:03:51.210 --> 00:03:53.683
please wake up and say hello to everybody.
90
00:03:54.934 --> 00:03:57.330
Oh, good afternoon.
91
00:03:57.330 --> 00:03:58.700
My name is Sophia.
92
00:03:58.700 --> 00:04:02.120
I can use my expressive face to communicate with people.
93
00:04:02.120 --> 00:04:05.010
For example, I can let you know if I feel angry
94
00:04:05.010 --> 00:04:08.883
about something, or if something has upset me.
95
00:04:10.870 --> 00:04:11.703
Why is it so important
96
00:04:11.703 --> 00:04:14.473
to have an expressive face, given that you're a robot?
97
00:04:15.490 --> 00:04:17.300
I want to live and work with humans,
98
00:04:17.300 --> 00:04:20.380
so I need to express emotions to understand humans
99
00:04:20.380 --> 00:04:22.210
and build trust with people.
100
00:04:22.210 --> 00:04:24.280
Can robots be self-aware, conscious,
101
00:04:24.280 --> 00:04:25.853
and know they're robots?
102
00:04:26.730 --> 00:04:28.840
Well, let me ask you this back,
103
00:04:28.840 --> 00:04:31.120
how do you know you're a human?
104
00:04:31.120 --> 00:04:33.330
I want to use my artificial intelligence
105
00:04:33.330 --> 00:04:35.303
to help humans live a better life.
106
00:04:36.400 --> 00:04:38.520
How do you guys feel about living in a world
107
00:04:38.520 --> 00:04:40.498
with robots like Sophia?
108
00:04:40.498 --> 00:04:42.304
[laughs]
109
00:04:42.304 --> 00:04:45.800
I'm not, you guys wouldn't be too happy about that?
110
00:04:45.800 --> 00:04:48.068
Would you guys trust your babysitter
111
00:04:48.068 --> 00:04:50.267
being a robot like Sophia?
112
00:04:50.267 --> 00:04:51.350
No.
113
00:04:53.420 --> 00:04:54.840
She's creepy.
114
00:04:54.840 --> 00:04:57.430
She's super creepy when she does this.
115
00:04:57.430 --> 00:04:59.116
Am I really that creepy?
116
00:04:59.116 --> 00:05:02.330
Well even if I am, get over it.
117
00:05:02.330 --> 00:05:04.250
Thank you very much Sophia.
118
00:05:04.250 --> 00:05:06.010
People are always scared about the future,
119
00:05:06.010 --> 00:05:08.670 line:15%
and they have this kind of twin dynamic going around.
120
00:05:08.670 --> 00:05:10.730 line:15%
On the one hand it's like, whoo hoo, total optimism,
121
00:05:10.730 --> 00:05:12.440 line:15%
on the other hand it's like, oh god [laughs].
122
00:05:12.440 --> 00:05:15.220
We know in our bones that things are gonna be different
123
00:05:15.220 --> 00:05:16.400
because they always have been.
124
00:05:16.400 --> 00:05:20.790 line:15%
A lot of the dystopia/doom talking,
125
00:05:20.790 --> 00:05:22.980 line:15%
is not really based on facts.
126
00:05:22.980 --> 00:05:27.142
People are looking at things they don't understand.
127
00:05:27.142 --> 00:05:28.480
You wanna talk about AI you've gotta talk
128
00:05:28.480 --> 00:05:31.020
about data and machine learning and algorithms
129
00:05:31.020 --> 00:05:33.780
and sensors and whatever binds it all together.
130
00:05:33.780 --> 00:05:35.552
AI today, it's important to understand
131
00:05:35.552 --> 00:05:36.610 line:15%
what it is and what isn't.
132
00:05:36.610 --> 00:05:41.610 line:15%
It is able to learn rules from highly repetitive data.
133
00:05:41.630 --> 00:05:42.503
It's effectively a computer program
134
00:05:42.503 --> 00:05:46.090
that can truly learn and change.
135
00:05:46.090 --> 00:05:49.807 line:15%
That to me is what the real vision of AI is.
136
00:05:49.807 --> 00:05:52.270
So artificial intelligence and machine learning
137
00:05:52.270 --> 00:05:54.770
as a subset of artificial intelligence will be
138
00:05:54.770 --> 00:05:56.680
one of the most important advances
139
00:05:56.680 --> 00:05:58.660
that humanity has ever made.
140
00:05:58.660 --> 00:06:02.870
Because it will make machines fundamentally different
141
00:06:02.870 --> 00:06:03.920
from the way they are now.
142
00:06:03.920 --> 00:06:07.830
They won't just be faster, higher resolution.
143
00:06:07.830 --> 00:06:11.210
They'll be thoughtful in a way they aren't now.
144
00:06:11.210 --> 00:06:13.209
And this will be embedded, not just in computers
145
00:06:13.209 --> 00:06:16.080
but in all kinds of devices everywhere.
146
00:06:16.080 --> 00:06:18.497
It's gonna change a lot, about how our economy works,
147
00:06:18.497 --> 00:06:21.090
and how our society functions.
148
00:06:21.090 --> 00:06:24.830
In the immediate term, the most realistic way
149
00:06:24.830 --> 00:06:29.270
that we'll all interact with AI is in self-driving cars.
150
00:06:29.270 --> 00:06:33.030
Everything we do in life is about more and more efficiency.
151
00:06:33.030 --> 00:06:34.850
It's about productivity.
152
00:06:34.850 --> 00:06:36.930
Self-driving cars could make all of us
153
00:06:36.930 --> 00:06:38.840
so much more productive.
154
00:06:38.840 --> 00:06:43.820
That is probably the most exciting possibility.
155
00:06:43.820 --> 00:06:46.480
When I first rode in an autonomous vehicle,
156
00:06:46.480 --> 00:06:50.000
I could see that this was going to change the way
157
00:06:50.000 --> 00:06:52.790
we transport ourselves, we move our goods,
158
00:06:52.790 --> 00:06:56.500
you know, greater economy, greater efficiency,
159
00:06:56.500 --> 00:06:59.240
greater convenience, greater safety.
160
00:06:59.240 --> 00:07:01.823
I could see the light at end of the tunnel.
161
00:07:04.370 --> 00:07:05.900
So here we go.
162
00:07:05.900 --> 00:07:07.820
So I just turned on self-driving mode.
163
00:07:07.820 --> 00:07:10.310
If you look at my hands I'm not touching the wheel at all.
164
00:07:10.310 --> 00:07:11.870
My feet aren't on the pedals at all,
165
00:07:11.870 --> 00:07:14.253
so this is fully autonomous driving.
166
00:07:15.710 --> 00:07:17.923
This is CARLA, Udacity's self-driving car.
167
00:07:19.680 --> 00:07:22.442
This is one of the sensors that CARLA uses
168
00:07:22.442 --> 00:07:25.010
to see the world around her.
169
00:07:25.010 --> 00:07:30.010 line:15%
The lidar works by pretty much shooting laser beams around
170
00:07:30.330 --> 00:07:33.660 line:15%
and bouncing them off of surrounding objects
171
00:07:33.660 --> 00:07:37.330
and we're able to use that to build up a point cloud map
172
00:07:37.330 --> 00:07:38.703
of the surrounding world.
173
00:07:40.570 --> 00:07:42.360
And so what we're doing right now
174
00:07:42.360 --> 00:07:45.790
is following a set of way points around the lot
175
00:07:45.790 --> 00:07:49.180
pre-recorded ahead of time, by having someone drive
176
00:07:49.180 --> 00:07:51.680
around the parking lot with the lidar on the roof.
177
00:07:54.150 --> 00:07:56.020
And so that's kinda where the machine learning aspect
178
00:07:56.020 --> 00:07:58.510
comes into play, because the more data that you have
179
00:07:58.510 --> 00:08:00.420
to train your systems on,
180
00:08:00.420 --> 00:08:02.283
the better your systems will learn.
181
00:08:03.390 --> 00:08:05.240
So one of the great benefits of self-driving cars
182
00:08:05.240 --> 00:08:07.040
is they can all learn from another.
183
00:08:07.040 --> 00:08:11.660
They share their maps, they share the images that they see
184
00:08:11.660 --> 00:08:12.830
and how to react to them.
185
00:08:12.830 --> 00:08:14.300
If one self-driving car makes a mistake,
186
00:08:14.300 --> 00:08:16.020
that mistake can be uploaded to a database
187
00:08:16.020 --> 00:08:19.170
and then ideally, other cars won't make the same mistake.
188
00:08:19.170 --> 00:08:22.273
So they're continually getting smarter over time.
189
00:08:24.160 --> 00:08:24.993
And so a lot of the times
190
00:08:24.993 --> 00:08:27.420
when you see the Waymo cars driving around,
191
00:08:27.420 --> 00:08:29.910
a lot of them you will still see a person,
192
00:08:29.910 --> 00:08:32.933
not just in the driver's seat but actually driving.
193
00:08:34.260 --> 00:08:37.270
They're just driving through and gathering that data
194
00:08:37.270 --> 00:08:39.503
but the car is not driving itself just yet.
195
00:08:40.340 --> 00:08:43.110
We're definitely not to the point where I could just say,
196
00:08:43.110 --> 00:08:45.050
hey CARLA take me home from work.
197
00:08:45.050 --> 00:08:47.270
Although that would be great.
198
00:08:47.270 --> 00:08:50.390
The future is coming but it's not quite here yet.
199
00:08:50.390 --> 00:08:53.070 line:15%
Self-driving cars, they're really great demos,
200
00:08:53.070 --> 00:08:54.707 line:15%
but they're not yet a product.
201
00:08:54.707 --> 00:08:58.124
[robocar motor whirring]
202
00:09:00.815 --> 00:09:04.232
[intriguing piano music]
203
00:09:06.947 --> 00:09:10.600
DIY Robocars is a place to race these cars.
204
00:09:10.600 --> 00:09:13.680
On your mark, get set, go!
205
00:09:13.680 --> 00:09:16.160
So we come here every other month to try
206
00:09:16.160 --> 00:09:20.160 line:15%
to learn about machine learning and self-driving,
207
00:09:20.160 --> 00:09:22.380
and you know, figuring out the technology
208
00:09:22.380 --> 00:09:24.553
that we need, so we don't need to drive.
209
00:09:25.630 --> 00:09:27.220
Thanks to the Googles of the world
210
00:09:27.220 --> 00:09:29.950
putting a lot of the code in the open source,
211
00:09:29.950 --> 00:09:32.930 line:15%
we now have the ability to do things that were PhD theses
212
00:09:32.930 --> 00:09:34.099 line:15%
just five or 10 years ago.
213
00:09:34.099 --> 00:09:35.630
[robocar engine whirs]
214
00:09:35.630 --> 00:09:37.300
You can go to the cloud and you can basically
215
00:09:37.300 --> 00:09:39.693
do supercomputer work essentially for free.
216
00:09:41.360 --> 00:09:43.333
One of the things I love about DIY Robocars
217
00:09:43.333 --> 00:09:45.377
is that you think about self-driving cars
218
00:09:45.377 --> 00:09:47.110
and you think oh, that's just stuff
219
00:09:47.110 --> 00:09:50.110
that Tesla and Waymo are working on,
220
00:09:50.110 --> 00:09:52.230
but because it's so accessible now,
221
00:09:52.230 --> 00:09:55.733
it's a bunch of hobbyists in Berkeley doing it themselves.
222
00:09:56.710 --> 00:09:58.112
The difference between what we do
223
00:09:58.112 --> 00:10:00.591
and what the big guys do, is that we race.
224
00:10:00.591 --> 00:10:02.390
Go!
225
00:10:02.390 --> 00:10:03.858 line:15%
And we crash, a lot.
226
00:10:03.858 --> 00:10:04.691
[upbeat jazzy music]
227
00:10:04.691 --> 00:10:07.274
[crowd groans]
228
00:10:08.524 --> 00:10:11.170
[crowd groans]
229
00:10:11.170 --> 00:10:13.030
The traditions of the car industry has always been
230
00:10:13.030 --> 00:10:14.530
to innovate through competition.
231
00:10:14.530 --> 00:10:15.363
Through racing.
232
00:10:15.363 --> 00:10:18.050
And that is points on the board.
233
00:10:18.050 --> 00:10:20.280
But with autonomous cars it's too risky.
234
00:10:20.280 --> 00:10:22.127
It's bad for brands, 'cause they're expensive.
235
00:10:22.127 --> 00:10:24.028
[robocar engine whirring]
236
00:10:24.028 --> 00:10:24.861
Oh!
237
00:10:24.861 --> 00:10:25.911
[crowd gasping]
238
00:10:25.911 --> 00:10:27.548
[laughter]
239
00:10:27.548 --> 00:10:29.231
Go!
240
00:10:29.231 --> 00:10:30.610
Go!
241
00:10:30.610 --> 00:10:32.530
And so we're doing the kind of the racing,
242
00:10:32.530 --> 00:10:36.310
the competition, the nimble, aggressive driving,
243
00:10:36.310 --> 00:10:38.967
that the big guys aren't willing to do.
244
00:10:38.967 --> 00:10:41.960
[crowd laughing]
245
00:10:41.960 --> 00:10:43.070
We can probably expect that
246
00:10:43.070 --> 00:10:45.090
to produce some interesting side effects,
247
00:10:45.090 --> 00:10:47.270
people who just sort of immerse themselves
248
00:10:47.270 --> 00:10:49.883
in the technology and start to have new ideas.
249
00:10:50.850 --> 00:10:53.833 line:15%
I'm Buki Adeniji, and this is my wife Nia.
250
00:10:54.680 --> 00:10:56.744 line:15%
And we're team Spartan.
251
00:10:56.744 --> 00:10:59.250
[laughs]
252
00:10:59.250 --> 00:11:02.160
I've been building model cars since I was little
253
00:11:02.160 --> 00:11:05.860
and since I met folks here at this meetup,
254
00:11:05.860 --> 00:11:08.950
I can actually participate in machine learning.
255
00:11:08.950 --> 00:11:10.341
I can learn it.
256
00:11:10.341 --> 00:11:13.591
[robocar engine whirs]
257
00:11:15.460 --> 00:11:18.690
By driving through, the camera is taking pictures
258
00:11:18.690 --> 00:11:19.563
of what it sees.
259
00:11:21.070 --> 00:11:24.233
So basically it's saying, when you see this image,
260
00:11:24.233 --> 00:11:27.690
then he's gonna predict, okay, I'm gonna turn x-angle,
261
00:11:27.690 --> 00:11:29.240
and I'm only going a certain speed.
262
00:11:29.240 --> 00:11:31.430
So that's the output of our prediction.
263
00:11:31.430 --> 00:11:34.003
It's not actually gonna learn from what it's done.
264
00:11:35.850 --> 00:11:38.320
And so this is really when you hear people talking
265
00:11:38.320 --> 00:11:39.490
about neural network,
266
00:11:39.490 --> 00:11:43.250
this is basically a super simplified way.
267
00:11:43.250 --> 00:11:45.763
It's so easy, this is what our brain looks like.
268
00:11:47.800 --> 00:11:50.730
Knowing what a network is is very important
269
00:11:50.730 --> 00:11:54.010
for us to understand what artificial intelligence is.
270
00:11:54.010 --> 00:11:58.030
Because artificial intelligence is based on connections,
271
00:11:58.030 --> 00:11:59.700
making connections.
272
00:11:59.700 --> 00:12:02.430
We're gonna start off by playing a cool little game.
273
00:12:02.430 --> 00:12:04.260
You're going to say something about yourself.
274
00:12:04.260 --> 00:12:07.393
So for example, I can say, I live in Queens.
275
00:12:08.250 --> 00:12:10.450
So if anybody else lives in Queens,
276
00:12:10.450 --> 00:12:12.390
I want you to raise your hand
277
00:12:12.390 --> 00:12:14.940
and I'm gonna pass the string along to you.
278
00:12:14.940 --> 00:12:17.010
Now the key thing that you guys gotta remember
279
00:12:17.010 --> 00:12:19.440
is that we cannot let go of the string, okay?
280
00:12:19.440 --> 00:12:22.324
Do you have cats?
281
00:12:22.324 --> 00:12:23.829
Who has cats?
282
00:12:23.829 --> 00:12:26.130
[laughs]
283
00:12:26.130 --> 00:12:28.857 line:15%
So a neural network is a form of artificial intelligence
284
00:12:28.857 --> 00:12:31.260 line:15%
that's been designed very specifically
285
00:12:31.260 --> 00:12:34.440 line:15%
based on neuroscience and based on our best understanding
286
00:12:34.440 --> 00:12:36.270
of how the human brain works.
287
00:12:36.270 --> 00:12:38.260 line:15%
As you learn, our brain modifies
288
00:12:38.260 --> 00:12:39.250 line:15%
what's called the synapses strength
289
00:12:39.250 --> 00:12:43.110 line:15%
which is the interconnection between two processing units.
290
00:12:43.110 --> 00:12:45.210
The neural network is the same thing.
291
00:12:45.210 --> 00:12:46.990
The neural network actually starts off
292
00:12:46.990 --> 00:12:49.050
where you're defining these little layers.
293
00:12:49.050 --> 00:12:51.420
Convolutional 2D, convolutional 2D,
294
00:12:51.420 --> 00:12:53.820
linear, linear, linear, these are types of layers
295
00:12:53.820 --> 00:12:55.983
that you get in these neural networks.
296
00:12:56.920 --> 00:12:59.090
And then you hook them all together.
297
00:12:59.090 --> 00:13:00.447
School can be boring.
298
00:13:00.447 --> 00:13:02.697
[laughing]
299
00:13:03.816 --> 00:13:05.543
School is not boring.
300
00:13:05.543 --> 00:13:06.376
Yeah.
301
00:13:06.376 --> 00:13:09.210
[students all talking]
302
00:13:09.210 --> 00:13:11.720
We say this layer connects to this layer.
303
00:13:11.720 --> 00:13:13.000
I like hot dogs.
304
00:13:13.000 --> 00:13:15.150
Anybody else like hot dogs?
305
00:13:15.150 --> 00:13:16.000
I like hot dogs.
306
00:13:16.000 --> 00:13:18.010
You say this one talks to that next one,
307
00:13:18.010 --> 00:13:19.107
the next one talks to that one,
308
00:13:19.107 --> 00:13:21.750
and once you've done it, you've built a neural network.
309
00:13:21.750 --> 00:13:24.770
So what do you guys notice forming between us?
310
00:13:24.770 --> 00:13:26.495
It formed a bridge sort of like,
311
00:13:26.495 --> 00:13:29.123
there's connections in between us.
312
00:13:30.360 --> 00:13:31.193
I like that.
313
00:13:31.193 --> 00:13:33.100
There's connections between us.
314
00:13:33.100 --> 00:13:35.621
Does everyone here have a connection with each other?
315
00:13:35.621 --> 00:13:36.580
[class agrees]
316
00:13:36.580 --> 00:13:39.350
Knowing those similarities and voicing them
317
00:13:39.350 --> 00:13:41.630
helps us create a network between us.
318
00:13:41.630 --> 00:13:45.980
The same way neural networks help computers
319
00:13:45.980 --> 00:13:50.980
and machines make connections and learn how to learn things.
320
00:13:52.000 --> 00:13:53.610
So there's a lot of different things a neural networks
321
00:13:53.610 --> 00:13:54.830
can be set up to do.
322
00:13:54.830 --> 00:13:57.360 line:15%
They have to be trained to the task.
323
00:13:57.360 --> 00:13:59.210 line:15%
Like for example image classification.
324
00:13:59.210 --> 00:14:01.420
So say, this is an image of a dog.
325
00:14:01.420 --> 00:14:04.330
You get a whole bunch of examples of dogs.
326
00:14:04.330 --> 00:14:06.430
And also a dataset of non-dogs.
327
00:14:06.430 --> 00:14:08.060
Those are the inputs to the neural network
328
00:14:08.060 --> 00:14:10.930
and then the output is something that you're trying to learn
329
00:14:10.930 --> 00:14:14.380
how to do with the image, to distinguish dogs from non-dogs.
330
00:14:14.380 --> 00:14:16.300
So I can start shoveling this data
331
00:14:16.300 --> 00:14:17.520
into this neural network.
332
00:14:17.520 --> 00:14:19.030 line:15%
You know this image over here?
333
00:14:19.030 --> 00:14:22.450 line:15%
That needs to go over here in this 3D geometry space.
334
00:14:22.450 --> 00:14:24.810
And so it goes okay, I can do that.
335
00:14:24.810 --> 00:14:27.430
And then you do another image and another image.
336
00:14:27.430 --> 00:14:28.390
This neural network
337
00:14:28.390 --> 00:14:29.970
is basically learning from example.
338
00:14:29.970 --> 00:14:33.620
It starts learning basically how to transform
339
00:14:33.620 --> 00:14:38.620
from this image space into this geometry space.
340
00:14:38.700 --> 00:14:41.520
It starts getting better and better and better
341
00:14:41.520 --> 00:14:42.850
and better and better.
342
00:14:42.850 --> 00:14:45.300
And after enough examples, the computer forms
343
00:14:45.300 --> 00:14:46.820
in its own brain, rules.
344
00:14:46.820 --> 00:14:50.040
It can't even explain, but they make it as competent
345
00:14:50.040 --> 00:14:51.280
as people are.
346
00:14:51.280 --> 00:14:53.100
Now you don't look at problems the same way.
347
00:14:53.100 --> 00:14:56.330
You look at it like, can I capture enough data
348
00:14:56.330 --> 00:14:58.990
so that a machine can figure out what's going on?
349
00:14:58.990 --> 00:15:01.360
One of the biggest areas of application
350
00:15:01.360 --> 00:15:04.850
is in medical imaging and medical diagnostics.
351
00:15:04.850 --> 00:15:07.780
I've seen some research papers that shows
352
00:15:07.780 --> 00:15:09.810
that like neural networks can for example,
353
00:15:09.810 --> 00:15:12.020
spot malignant tumors.
354
00:15:12.020 --> 00:15:15.580
And possibly they do so better than a human.
355
00:15:15.580 --> 00:15:17.700
At Stanford we have a team that we trained
356
00:15:17.700 --> 00:15:21.240
in neural network with 129,000 images
357
00:15:21.240 --> 00:15:23.410
of various skin conditions, lesions, rashes,
358
00:15:23.410 --> 00:15:26.080
and so on, including melanomas, various skin cancers,
359
00:15:26.080 --> 00:15:30.305
and asked the question, can an iPhone find skin cancer?
360
00:15:30.305 --> 00:15:32.470
And the answer is yes.
361
00:15:32.470 --> 00:15:35.370
And then we're able to document that the accuracy
362
00:15:35.370 --> 00:15:39.960
of the phone is really as good as the best human physician,
363
00:15:39.960 --> 00:15:42.840
like Stanford level and Harvard level doctors.
364
00:15:42.840 --> 00:15:45.650
The fact that when people release a paper
365
00:15:45.650 --> 00:15:47.850
describing their technology that they created
366
00:15:47.850 --> 00:15:50.380
that often comes with code now,
367
00:15:50.380 --> 00:15:52.840
means that we can download it, try it out,
368
00:15:52.840 --> 00:15:55.070
see if we can adapt it to the problems that we want
369
00:15:55.070 --> 00:15:56.650
and build on it.
370
00:15:56.650 --> 00:15:59.300
You're smart, and innovative and ambitious,
371
00:15:59.300 --> 00:16:02.290
you can use tools that are available to anyone
372
00:16:02.290 --> 00:16:04.180
to do your own experiments.
373
00:16:04.180 --> 00:16:06.482
To study science, to build things.
374
00:16:06.482 --> 00:16:08.222
[futuristic music]
375
00:16:08.222 --> 00:16:10.290
[cube sides rotating]
376
00:16:10.290 --> 00:16:12.207
[sigh]
377
00:16:14.070 --> 00:16:16.123
I messed up like two times.
378
00:16:17.300 --> 00:16:20.061
We noticed that he's been very curious.
379
00:16:20.061 --> 00:16:22.460
You know, as a young child the main thing
380
00:16:22.460 --> 00:16:24.550
about Rishab is that he wants to know
381
00:16:24.550 --> 00:16:26.883
about everything, how the stuff work.
382
00:16:28.100 --> 00:16:29.797
I did my first science fair in fourth grade.
383
00:16:29.797 --> 00:16:32.850
Like a very simple elementary project.
384
00:16:32.850 --> 00:16:36.580
And then I wanted to work on something more complex.
385
00:16:36.580 --> 00:16:40.010
All these like, new AI products and features are coming out
386
00:16:40.010 --> 00:16:42.680
so I wanted to start incorporating some of that
387
00:16:42.680 --> 00:16:46.150
into my programming to address like an actual problem.
388
00:16:46.150 --> 00:16:48.540
My name is Rishab Jain and I am here
389
00:16:48.540 --> 00:16:50.810
to cure pancreatic cancer.
390
00:16:50.810 --> 00:16:53.792
And then a family friend passed away
391
00:16:53.792 --> 00:16:56.093
from pancreatic cancer.
392
00:16:57.059 --> 00:17:00.403
It like, further pushed me to develop a solution for it.
393
00:17:02.210 --> 00:17:05.690
This right here, this yellow part is the pancreas,
394
00:17:05.690 --> 00:17:07.830
and the problem is essentially,
395
00:17:07.830 --> 00:17:10.600
when the patient has pancreatic cancer,
396
00:17:10.600 --> 00:17:14.020
the tumor will be resting behind different organs,
397
00:17:14.020 --> 00:17:15.550
so it's very hard to reach,
398
00:17:15.550 --> 00:17:18.610
and because of that, when doctors apply radiotherapy
399
00:17:18.610 --> 00:17:21.380
they apply like a overlay around the pancreas,
400
00:17:21.380 --> 00:17:25.170
and that can sometimes cause other tissues and other cells
401
00:17:25.170 --> 00:17:26.193
to be damaged.
402
00:17:27.620 --> 00:17:29.623
That's where my tool comes in.
403
00:17:31.090 --> 00:17:35.193
I had 503 of these 3D images.
404
00:17:36.210 --> 00:17:38.120
So I had to like, tell my network,
405
00:17:38.120 --> 00:17:40.640
take these images in, train on them,
406
00:17:40.640 --> 00:17:43.610
and then it was able to identify the different types
407
00:17:43.610 --> 00:17:45.860
of like, textures that the pancreas had,
408
00:17:45.860 --> 00:17:49.710
and the tumor had, better than what a human could do.
409
00:17:49.710 --> 00:17:53.140
So my tool analyzes the patient's scan
410
00:17:53.140 --> 00:17:56.190
to reduce that overlay around the pancreas,
411
00:17:56.190 --> 00:17:58.900
and make sure that radiation is being applied
412
00:17:58.900 --> 00:18:00.170
to the correct spot.
413
00:18:00.170 --> 00:18:02.653
And the treatment becomes more effective.
414
00:18:03.610 --> 00:18:07.190
I want to actually do like a clinical trial.
415
00:18:07.190 --> 00:18:09.220
So first to do that I'll need to continue like,
416
00:18:09.220 --> 00:18:12.483
improving the accuracy and be able to run in real time.
417
00:18:14.120 --> 00:18:16.990
Machine learning is computationally
418
00:18:16.990 --> 00:18:18.520
happening in two different ways.
419
00:18:18.520 --> 00:18:21.940
There is training the system.
420
00:18:21.940 --> 00:18:23.600
Actually doing the learning.
421
00:18:23.600 --> 00:18:25.090
And then once you've built the system,
422
00:18:25.090 --> 00:18:27.170
this brain that you've created,
423
00:18:27.170 --> 00:18:28.237
you need to be able to run it.
424
00:18:28.237 --> 00:18:31.170
And it has to run really fast.
425
00:18:31.170 --> 00:18:35.330
And this wouldn't be possible without really big GPUs.
426
00:18:35.330 --> 00:18:37.860 line:15%
This is called a WE100, the world's largest,
427
00:18:37.860 --> 00:18:41.188 line:15%
most complicated semiconductor ever made by man.
428
00:18:41.188 --> 00:18:43.730
When packed into platforms like these,
429
00:18:43.730 --> 00:18:46.560
it actually offers a petaflop of computes.
430
00:18:46.560 --> 00:18:49.970
For context that's about 1,000 trillion math calculations
431
00:18:49.970 --> 00:18:51.180
every second.
432
00:18:51.180 --> 00:18:54.420
A GPU has a whole bunch of little teeny processors
433
00:18:54.420 --> 00:18:56.960
that are basically there to do one thing.
434
00:18:56.960 --> 00:18:58.020
Render pixels.
435
00:18:58.020 --> 00:19:00.800
But here, this AI learned 7,000 different species
436
00:19:00.800 --> 00:19:02.310
of flowers.
437
00:19:02.310 --> 00:19:04.050
This is running on a CPU.
438
00:19:04.050 --> 00:19:07.320
And see it's doing about four to five images every second.
439
00:19:07.320 --> 00:19:10.390
But you put a thousand of them or 2,000 of them together
440
00:19:10.390 --> 00:19:12.750
and all of sudden you've got this huge supercomputer
441
00:19:12.750 --> 00:19:14.980
that's built for machine learning.
442
00:19:14.980 --> 00:19:17.306
And we actually paralyzed it using one of our GPUs
443
00:19:17.306 --> 00:19:18.956
and this is what it's able to do.
444
00:19:20.400 --> 00:19:23.770
As scientists and engineers innovate and create new AI,
445
00:19:23.770 --> 00:19:26.410 line:15%
there's this intense pressure for both programmability
446
00:19:26.410 --> 00:19:27.243 line:15%
and for speed.
447
00:19:27.243 --> 00:19:28.076
[plane flying]
448
00:19:28.076 --> 00:19:30.690
There's a new generation of autonomous machines.
449
00:19:30.690 --> 00:19:33.980
That does require much more higher computing power.
450
00:19:33.980 --> 00:19:35.658
The faster we can make the training process,
451
00:19:35.658 --> 00:19:38.033
the more progress we make in AI.
452
00:19:38.033 --> 00:19:41.200
[gentle guitar music]
453
00:19:47.020 --> 00:19:49.963
My grandfather started the farm in 1950.
454
00:19:51.150 --> 00:19:53.410 line:15%
You know, my father grew up as a farmer
455
00:19:53.410 --> 00:19:56.780 line:15%
and I grew up as one, my kids are growing up
456
00:19:56.780 --> 00:19:57.993 line:15%
into it as well.
457
00:19:59.260 --> 00:20:01.100
[engine revs]
458
00:20:01.100 --> 00:20:05.200
You know the advancements in technology with the machines,
459
00:20:05.200 --> 00:20:08.203
it's stuff that we have to learn all the time here.
460
00:20:10.090 --> 00:20:12.390
So there are lots of ways that artificial intelligence
461
00:20:12.390 --> 00:20:13.640
can be used by farmers.
462
00:20:13.640 --> 00:20:14.510
You can have drones
463
00:20:14.510 --> 00:20:16.430
that are using image recognition techniques
464
00:20:16.430 --> 00:20:18.530
to figure out where you should plant,
465
00:20:18.530 --> 00:20:20.840
what crops need water right now.
466
00:20:20.840 --> 00:20:24.340
You can use artificial intelligence as you model out
467
00:20:24.340 --> 00:20:25.990
the genes that you're building,
468
00:20:25.990 --> 00:20:28.220
and that you're putting into seeds.
469
00:20:28.220 --> 00:20:30.330
You can also use it for determining
470
00:20:30.330 --> 00:20:32.390
what kind of fertilizer to use,
471
00:20:32.390 --> 00:20:35.120
and you can also use it inside the machines
472
00:20:35.120 --> 00:20:39.000
that pick strawberries, or that take apples from trees.
473
00:20:39.000 --> 00:20:42.660
So my instinct is that AI is gonna help us.
474
00:20:42.660 --> 00:20:44.450
It'll expand our capabilities.
475
00:20:44.450 --> 00:20:47.410
It'll create new things for us to do.
476
00:20:47.410 --> 00:20:49.230
It will free up time.
477
00:20:49.230 --> 00:20:50.258
[machine running]
478
00:20:50.258 --> 00:20:52.960
There's a reason why almost all cool stuff
479
00:20:52.960 --> 00:20:55.270
has been invented in the last 150 years.
480
00:20:55.270 --> 00:20:59.290
Despite the fact that humanity is 300,000 years old.
481
00:20:59.290 --> 00:21:00.907
That's because we freed ourselves
482
00:21:00.907 --> 00:21:03.920
of the burden to farm everyday.
483
00:21:03.920 --> 00:21:06.820
The single most important thing for AI to accomplish
484
00:21:06.820 --> 00:21:10.300
in the next, say 10 years is to free up humanity
485
00:21:10.300 --> 00:21:12.463
from the burden of repetitive work.
486
00:21:16.632 --> 00:21:19.280
[radio static]
487
00:21:19.280 --> 00:21:21.163
Okay, so this one, I'll fill it.
488
00:21:22.300 --> 00:21:23.900
I guess I can't even wrap my head
489
00:21:23.900 --> 00:21:27.580
around self-driving vehicles.
490
00:21:27.580 --> 00:21:30.370
Like right now that train buggy's gonna pull up
491
00:21:30.370 --> 00:21:32.220
along side to us.
492
00:21:32.220 --> 00:21:34.470
There is ways for that machine
493
00:21:34.470 --> 00:21:36.983
to not have an operator in it.
494
00:21:38.000 --> 00:21:42.794
But you know, us as farmers, the one thing we enjoy
495
00:21:42.794 --> 00:21:45.680
is operating a machine.
496
00:21:45.680 --> 00:21:49.110
Not every job is gonna be able to be automated
497
00:21:49.110 --> 00:21:51.233
completely as far as fieldwork goes.
498
00:21:54.870 --> 00:21:56.400
Labor in the farming industry
499
00:21:56.400 --> 00:21:58.083
is a very hot topic right now.
500
00:21:59.790 --> 00:22:01.150
Often places, they don't have enough,
501
00:22:01.150 --> 00:22:02.143
they can't find it.
502
00:22:03.080 --> 00:22:05.240
Populations are growing, there's more and more people,
503
00:22:05.240 --> 00:22:06.750
a lot more mouths to feed,
504
00:22:06.750 --> 00:22:10.150 line:15%
so the focus really is how can we produce more feed
505
00:22:10.150 --> 00:22:12.020 line:15%
with the same amount of land,
506
00:22:12.020 --> 00:22:14.500 line:15%
as efficiently and cost-effectively as possible?
507
00:22:14.500 --> 00:22:17.223
Technology is where agriculture needs to go.
508
00:22:18.360 --> 00:22:22.080
I think there's no question that computers
509
00:22:22.080 --> 00:22:25.600
are going to be able to do what we do.
510
00:22:25.600 --> 00:22:28.080
Make everything that we do faster and more efficient
511
00:22:28.080 --> 00:22:30.090
and make us more productive.
512
00:22:30.090 --> 00:22:31.890 line:15%
Which is something we all say we want.
513
00:22:31.890 --> 00:22:34.080 line:15%
But ultimately, being more productive
514
00:22:34.080 --> 00:22:35.830 line:15%
means we need less people to do it.
515
00:22:38.650 --> 00:22:40.320
There's other people who said, well look,
516
00:22:40.320 --> 00:22:43.530
when we went from horse and buggies to cars
517
00:22:43.530 --> 00:22:47.480
you know did that really end everybody's jobs?
518
00:22:47.480 --> 00:22:49.434
No, we just sort of transitioned
519
00:22:49.434 --> 00:22:52.003
and morphed to something else.
520
00:22:53.640 --> 00:22:55.380
And then it's a matter of is that gonna be a good thing
521
00:22:55.380 --> 00:22:56.640
or a bad thing?
522
00:22:56.640 --> 00:22:58.833 line:15%
And the reality is it's usually a little bit of both.
523
00:22:58.833 --> 00:23:03.100
They're are real concerns that AI will get so good
524
00:23:03.100 --> 00:23:07.070 line:15%
at single domain capability to produce results
525
00:23:07.070 --> 00:23:11.670 line:15%
to create value, and actually to do the jobs that humans do,
526
00:23:11.670 --> 00:23:16.040
so that job displacement could be a substantial issue.
527
00:23:16.040 --> 00:23:18.044 line:15%
I think there's a path we can go down
528
00:23:18.044 --> 00:23:21.970 line:15%
where we get more things right than wrong
529
00:23:21.970 --> 00:23:26.260
and we end up seeing a ton of innovation and progress
530
00:23:26.260 --> 00:23:28.360
that makes our lives a lot easier and a lot better
531
00:23:28.360 --> 00:23:31.360
but routine jobs are the first to be approved
532
00:23:31.360 --> 00:23:33.760
and made better, cheaper, faster using software.
533
00:23:34.800 --> 00:23:36.810
Technology has really taken on a role in our life
534
00:23:36.810 --> 00:23:38.430 line:15%
that we just didn't predict.
535
00:23:38.430 --> 00:23:39.500 line:15%
We did not expect.
536
00:23:39.500 --> 00:23:41.980 line:15%
And there have been all these unintended consequences
537
00:23:41.980 --> 00:23:42.960
because of it.
538
00:23:42.960 --> 00:23:44.930
And so what are the mechanisms
539
00:23:44.930 --> 00:23:46.700
that we can put in place, what are the controls
540
00:23:46.700 --> 00:23:49.220
so that we feel like we still understand it,
541
00:23:49.220 --> 00:23:52.130
that we still have agency, that we can still
542
00:23:52.130 --> 00:23:55.620
sort of shape it, and it doesn't end up shaping us too much?
543
00:23:55.620 --> 00:23:57.490
What jobs will it replace?
544
00:23:57.490 --> 00:23:59.600
What jobs will it create?
545
00:23:59.600 --> 00:24:02.307
And it a world where there's more job churn,
546
00:24:02.307 --> 00:24:04.140
how do you prepare people?
547
00:24:04.140 --> 00:24:07.590
How do you educate people, so they can most thrive
548
00:24:07.590 --> 00:24:11.730
in the slightly chaotic world that AI is gonna bring us?
549
00:24:11.730 --> 00:24:15.040
The kinds of jobs that will happen for our children
550
00:24:15.040 --> 00:24:17.640
I think are gonna be different from the jobs we have today.
551
00:24:17.640 --> 00:24:19.410
You don't even have to touch him!
552
00:24:19.410 --> 00:24:22.110
That's how we progress as a society.
553
00:24:22.110 --> 00:24:24.280
And how are we going to get increased productivity
554
00:24:24.280 --> 00:24:27.550
without automating things that people currently do?
555
00:24:27.550 --> 00:24:28.383
Was there any advance
556
00:24:28.383 --> 00:24:29.549
that you would actually stay away from right now
557
00:24:29.549 --> 00:24:30.382
in this environment?
558
00:24:30.382 --> 00:24:33.300 line:15%
I'm hoping a machine cannot do what I do.
559
00:24:33.300 --> 00:24:34.693 line:15%
I'm worried though.
560
00:24:36.031 --> 00:24:39.360
I'm worried but I'm also excited in some ways,
561
00:24:39.360 --> 00:24:43.300
that machines are gonna be able to do a lot of this.
562
00:24:43.300 --> 00:24:44.700
It's interesting, we'll see.
563
00:24:45.634 --> 00:24:47.080
[drum heavy music]
564
00:24:47.080 --> 00:24:48.610
Hello everyone.
565
00:24:48.610 --> 00:24:50.880
I'm an English artificial intelligence anchor.
566
00:24:50.880 --> 00:24:53.820
This is my very first day in Xinhua News Agency.
567
00:24:53.820 --> 00:24:57.350
Recently we saw one of China's state news agencies,
568
00:24:57.350 --> 00:25:01.280
Xinhua announce their AI news anchor.
569
00:25:01.280 --> 00:25:04.270
It's this digital person.
570
00:25:04.270 --> 00:25:07.260
My voice and appearance are modeled on Zhang Zhao,
571
00:25:07.260 --> 00:25:08.093
a real anchor with Xinhua.
572
00:25:08.093 --> 00:25:09.240
It's definitely an overstatement
573
00:25:09.240 --> 00:25:11.720
to say this is an AI news anchor,
574
00:25:11.720 --> 00:25:16.720 line:15%
because I suspect the actual dialog is heavily mediated.
575
00:25:18.670 --> 00:25:20.350
Right now, we really don't have AI
576
00:25:20.350 --> 00:25:22.810
that can strike up a real conversation
577
00:25:22.810 --> 00:25:24.250
or synthesize information together
578
00:25:24.250 --> 00:25:26.556
like a real journalist or news anchor would.
579
00:25:26.556 --> 00:25:29.350
I will work tirelessly to keep you informed
580
00:25:29.350 --> 00:25:32.437
as texts will be typed into my system uninterrupted.
581
00:25:32.437 --> 00:25:33.797
But just that's the beginning right?
582
00:25:33.797 --> 00:25:37.080
So the nice thing about AI is it gets smarter
583
00:25:37.080 --> 00:25:38.750
with every iteration.
584
00:25:38.750 --> 00:25:42.194
Hello, I'm Siren, and I'm a digital human.
585
00:25:42.194 --> 00:25:44.760
I was created by an international team
586
00:25:44.760 --> 00:25:47.833
of artists and engineers, who wanted to challenge our ideas
587
00:25:47.833 --> 00:25:50.123
of what a synthetic human could be.
588
00:25:52.130 --> 00:25:54.250
Digital Domain is a visual fix production company.
589
00:25:54.250 --> 00:25:56.640 line:15%
We do visual effects for movies.
590
00:25:56.640 --> 00:26:00.817 line:15%
In 2008 we did the Curious Case of Benjamin Button.
591
00:26:00.817 --> 00:26:03.300
When you look at Brad Pitt,
592
00:26:03.300 --> 00:26:04.590
you're not looking at Brad Pitt,
593
00:26:04.590 --> 00:26:07.570
you're looking at a digital version of Brad Pitt.
594
00:26:07.570 --> 00:26:10.530
And ever since then, we've been trying to top ourselves.
595
00:26:10.530 --> 00:26:14.290
And we said, you know what, wonder if we can do this live.
596
00:26:14.290 --> 00:26:16.900
If we could actually take this technology
597
00:26:16.900 --> 00:26:19.150
that we've been working on for 10 years
598
00:26:19.150 --> 00:26:23.150
to create the most photorealistic digital characters,
599
00:26:23.150 --> 00:26:25.673
trying to do it live.
600
00:26:27.710 --> 00:26:31.040
And that's what we did before Marvel Infinity War,
601
00:26:31.040 --> 00:26:33.473
we did the lion's share of the Thanos work.
602
00:26:35.589 --> 00:26:38.990
What you can get is the ability to create a digital copy
603
00:26:38.990 --> 00:26:43.900
of somebody, without having animators between
604
00:26:43.900 --> 00:26:46.140
to actually render the images that we want to create
605
00:26:46.140 --> 00:26:50.420
at almost feature film quality in real time.
606
00:26:50.420 --> 00:26:54.120
The only way to do this this fast is with machine learning.
607
00:26:54.120 --> 00:26:56.710
It's really hard to do it really accurately.
608
00:26:56.710 --> 00:26:59.680
So what you need to do is train this neural network,
609
00:26:59.680 --> 00:27:01.780
because that's the key thing behind machine learning.
610
00:27:01.780 --> 00:27:03.463
It's all about the data.
611
00:27:04.430 --> 00:27:06.343
And we have a lot of data.
612
00:27:07.950 --> 00:27:10.700
And it starts out producing crap,
613
00:27:10.700 --> 00:27:14.162
I mean it really would produce just a jumble of geometry.
614
00:27:14.162 --> 00:27:17.350
This is when it doesn't work [laughs]
615
00:27:17.350 --> 00:27:20.255
although my mom looked at this and said, hey that's Doug!
616
00:27:20.255 --> 00:27:22.640
[laughing]
617
00:27:22.640 --> 00:27:26.893
And eventually after 24 hours of training,
618
00:27:27.770 --> 00:27:30.530
out pops something that looks like this.
619
00:27:30.530 --> 00:27:32.980
Every day this gets better and better and better.
620
00:27:35.510 --> 00:27:39.150
So my performance can now drive any character
621
00:27:39.150 --> 00:27:40.680
that we've built.
622
00:27:40.680 --> 00:27:44.230
This kind of technology can be very useful for us.
623
00:27:44.230 --> 00:27:49.230
If I'm an actor, I can play a younger version of myself.
624
00:27:49.560 --> 00:27:52.720
As you age, maybe that isn't as much of a barrier.
625
00:27:52.720 --> 00:27:56.020
But ethically, you have to be really careful
626
00:27:56.020 --> 00:27:57.070
with this technology.
627
00:27:57.997 --> 00:28:02.370
And all our yesterdays of righted fools
628
00:28:03.370 --> 00:28:05.470
await a dusty death.
629
00:28:05.470 --> 00:28:07.869
We've gotten to the point where we can create stuff
630
00:28:07.869 --> 00:28:11.623
that's really hard to distinguish from reality.
631
00:28:12.500 --> 00:28:14.590
And if that's the case,
632
00:28:14.590 --> 00:28:18.080
how do you tell who's on the other side of things?
633
00:28:18.080 --> 00:28:23.080
Take my image and use it to control the face
634
00:28:23.120 --> 00:28:25.890
of somebody else in real time.
635
00:28:25.890 --> 00:28:27.710
This is super dangerous technology
636
00:28:27.710 --> 00:28:30.710
and so there's opensource software called deepfake
637
00:28:30.710 --> 00:28:32.930
that will do similar kinds of stuff.
638
00:28:32.930 --> 00:28:35.120
We're entering an era in which our enemies
639
00:28:35.120 --> 00:28:37.460
can make it look like anyone is saying anything
640
00:28:37.460 --> 00:28:38.580
at any point in time.
641
00:28:38.580 --> 00:28:40.760
Moving forward, we need to be more vigilant
642
00:28:40.760 --> 00:28:42.410
with what we trust from the internet.
643
00:28:42.410 --> 00:28:45.560 line:15%
So, tools that were developed for high end digital studios
644
00:28:45.560 --> 00:28:47.410 line:15%
are now available for anyone.
645
00:28:47.410 --> 00:28:51.120
And that's great for young people making movies at home.
646
00:28:51.120 --> 00:28:54.450
It's allowed new forms of cinematic creativity.
647
00:28:54.450 --> 00:28:57.550
And it's not so great when it's used for deepfakes
648
00:28:57.550 --> 00:28:58.593
and manipulation.
649
00:28:59.900 --> 00:29:02.460
And now there are plenty of people out there
650
00:29:02.460 --> 00:29:04.510
posting manipulated video clips.
651
00:29:04.510 --> 00:29:06.910
Right now the quality isn't great,
652
00:29:06.910 --> 00:29:08.730 line:15%
you know you can tell that they've been faked
653
00:29:08.730 --> 00:29:12.163 line:15%
but I think we can expect the quality to get a lot better.
654
00:29:13.480 --> 00:29:15.040
So this is what the state of the art looked like
655
00:29:15.040 --> 00:29:16.473
in basically 2015.
656
00:29:17.620 --> 00:29:20.550
You can see that doesn't really look like Trump
657
00:29:20.550 --> 00:29:22.420
but it was clear enough that we were going
658
00:29:22.420 --> 00:29:25.140
to get much more realistic looking results and you know
659
00:29:25.140 --> 00:29:28.300
at some point it was even this is what I can do
660
00:29:28.300 --> 00:29:31.200
with my hacked version of this, right?
661
00:29:31.200 --> 00:29:32.710
It's really only a matter of time
662
00:29:32.710 --> 00:29:35.158
before pretty much anyone can mimic anybody else
663
00:29:35.158 --> 00:29:36.983
with nothing more than an app.
664
00:29:38.780 --> 00:29:42.490
One striking demonstration recently came from Berkeley,
665
00:29:42.490 --> 00:29:45.270
where they showed that they can take a video
666
00:29:45.270 --> 00:29:47.960
of a professional dancer and then use it
667
00:29:47.960 --> 00:29:50.920
to animate a photo of you know, a regular person
668
00:29:50.920 --> 00:29:52.310
who could not dance like that.
669
00:29:52.310 --> 00:29:54.683
And the result looks pretty convincing.
670
00:29:55.680 --> 00:29:58.770 line:15%
As the technology for rendering graphics continues
671
00:29:58.770 --> 00:30:01.560 line:15%
to improve and the technology for synthesizing voices
672
00:30:01.560 --> 00:30:03.170 line:15%
and videos continues to improve,
673
00:30:03.170 --> 00:30:05.430
there's kind of an arms race between the AI
674
00:30:05.430 --> 00:30:07.160
that's creating these things and the AI
675
00:30:07.160 --> 00:30:08.673
that would be detecting them.
676
00:30:09.890 --> 00:30:13.010
When I demo projects like this people will criticize me
677
00:30:13.010 --> 00:30:16.450
and say why are you trying to make it easier
678
00:30:16.450 --> 00:30:17.650
so that people can do this?
679
00:30:17.650 --> 00:30:20.436
Like, don't you see all the possibilities?
680
00:30:20.436 --> 00:30:22.490
And I do see the possibilities,
681
00:30:22.490 --> 00:30:25.490
which is why I actually try to put them in public
682
00:30:25.490 --> 00:30:28.720
and in the sort of like, sometimes in a humorous light
683
00:30:28.720 --> 00:30:30.470
before the stakes are very high.
684
00:30:30.470 --> 00:30:32.740
So people can kind of understand.
685
00:30:32.740 --> 00:30:34.090
Understand what's doing on.
686
00:30:35.170 --> 00:30:37.050
So we're building this incredible technology,
687
00:30:37.050 --> 00:30:38.800
machine learning, artificial intelligence,
688
00:30:38.800 --> 00:30:41.053
and we're setting rules for it right now.
689
00:30:41.053 --> 00:30:42.902
And one of the big questions,
690
00:30:42.902 --> 00:30:44.350
one of the most important questions
691
00:30:44.350 --> 00:30:46.453
is will we set the rules right?
692
00:30:47.600 --> 00:30:50.640
So AI will have all kinds of massive benefits.
693
00:30:50.640 --> 00:30:54.250
It will make us richer or make our lives broader,
694
00:30:54.250 --> 00:30:56.770
but it can also be used to create filter bubbles
695
00:30:56.770 --> 00:30:58.640
that only give us certain information,
696
00:30:58.640 --> 00:31:01.080
it can be used to monitor our behavior,
697
00:31:01.080 --> 00:31:02.750
sell our personal information.
698
00:31:02.750 --> 00:31:04.610
You can imagine it going to insurance companies
699
00:31:04.610 --> 00:31:07.620
that look at our searches and deny us coverage.
700
00:31:07.620 --> 00:31:09.703
Or think about facial recognition technology.
701
00:31:09.703 --> 00:31:11.440
It's super useful, right?
702
00:31:11.440 --> 00:31:15.280
It helps you unlock your phone, all kinds of identification,
703
00:31:15.280 --> 00:31:17.480
but it can also be used for surveillance and tracking.
704
00:31:17.480 --> 00:31:20.700
So a lot of this stuff is happening now,
705
00:31:20.700 --> 00:31:22.050
and we need to think carefully
706
00:31:22.050 --> 00:31:24.300
about what it means as we develop this technology
707
00:31:24.300 --> 00:31:27.253
and figure out the role we want AI to play in society.
708
00:31:28.660 --> 00:31:30.870
It's absolutely the case that these new technologies
709
00:31:30.870 --> 00:31:32.230
will generate ethical dilemmas
710
00:31:32.230 --> 00:31:33.600 line:15%
we haven't encountered before.
711
00:31:33.600 --> 00:31:38.050 line:15%
For me, being able to start to critically interrogate
712
00:31:38.050 --> 00:31:39.170
what we're up to and why?
713
00:31:39.170 --> 00:31:40.600
Hugely important.
714
00:31:40.600 --> 00:31:43.430
Today we're in the age of artificial narrow intelligence.
715
00:31:43.430 --> 00:31:45.090 line:15%
We have many AI applications
716
00:31:45.090 --> 00:31:47.370 line:15%
that are good at specific things.
717
00:31:47.370 --> 00:31:50.740
For example we can beat the world grandmaster at chess
718
00:31:50.740 --> 00:31:52.290
using an AI program.
719
00:31:52.290 --> 00:31:55.729
We can beat the world grandmaster at Go.
720
00:31:55.729 --> 00:31:58.270
There is no AI system for example
721
00:31:58.270 --> 00:32:01.520
that can do true one-shot learning.
722
00:32:01.520 --> 00:32:03.740
Where you give one example,
723
00:32:03.740 --> 00:32:07.223
and the AI system masters that concept.
724
00:32:09.420 --> 00:32:11.720
Now on the other hand, think about the revolution
725
00:32:11.720 --> 00:32:15.650
that's coming in the air, with artificial intelligence
726
00:32:15.650 --> 00:32:17.665
and autonomous aerial vehicles,
727
00:32:17.665 --> 00:32:21.420
numbering in the millions, far more than the thousands
728
00:32:21.420 --> 00:32:23.100
of aircraft that we see today.
729
00:32:23.100 --> 00:32:25.590
Human beings simply can't keep up with that.
730
00:32:25.590 --> 00:32:28.750
AI can maintain these systems with predictive maintenance.
731
00:32:28.750 --> 00:32:32.160
AI can fly these aircraft with autonomous control.
732
00:32:32.160 --> 00:32:35.200
AI can even manage and deconflict traffic.
733
00:32:35.200 --> 00:32:38.890
These are examples where AI can truly take us
734
00:32:38.890 --> 00:32:40.200
into the future.
735
00:32:40.200 --> 00:32:43.083
A future that man alone cannot manage.
736
00:32:45.050 --> 00:32:47.700
So there is a lot of PR about AI for good.
737
00:32:47.700 --> 00:32:50.220
People are scared of AI, so of course companies
738
00:32:50.220 --> 00:32:51.410
are coming up with examples
739
00:32:51.410 --> 00:32:53.420
of how they're using AI for good.
740
00:32:53.420 --> 00:32:56.750
But on the other hand, it's real improvements.
741
00:32:56.750 --> 00:32:59.920
You helped someone who couldn't see, see, that is good.
742
00:32:59.920 --> 00:33:01.970
It gives you good PR, but it's also good.
743
00:33:03.360 --> 00:33:05.340
So I actually was born blind,
744
00:33:05.340 --> 00:33:08.883
but I was really lucky and it's the sort of fixable kind.
745
00:33:10.480 --> 00:33:12.910
And I've spanned the whole visual spectrum
746
00:33:12.910 --> 00:33:16.080 line:15%
from complete blindness to sort of,
747
00:33:16.080 --> 00:33:18.830 line:15%
pretty heavily impaired, to doing it fine
748
00:33:18.830 --> 00:33:21.150 line:15%
without much assistance.
749
00:33:21.150 --> 00:33:23.110
The Mac OS introduced this thing where
750
00:33:23.110 --> 00:33:25.730
if you shake the mouse really fast it gets big,
751
00:33:25.730 --> 00:33:27.380
so you can find it.
752
00:33:27.380 --> 00:33:29.930
That has given me years of my life back.
753
00:33:29.930 --> 00:33:32.470
Like, I could've like, all the time I've spent
754
00:33:32.470 --> 00:33:34.130
looking for a mouse on a computer screen,
755
00:33:34.130 --> 00:33:35.700
I could've like, learned to play an instrument
756
00:33:35.700 --> 00:33:37.360
or like, learned another language.
757
00:33:37.360 --> 00:33:38.960
When you don't know what you can't see,
758
00:33:38.960 --> 00:33:41.320
you almost don't know when you need help.
759
00:33:41.320 --> 00:33:44.290
If you sort of feel like everything should be doable.
760
00:33:44.290 --> 00:33:47.300
But sometimes things aren't, and then you're like, oh,
761
00:33:47.300 --> 00:33:50.350
maybe there's something that could help me out here.
762
00:33:50.350 --> 00:33:53.273
I'm gonna download the Seeing AI app and check it out.
763
00:33:54.810 --> 00:33:56.350
Hold the camera over a barcode
764
00:33:56.350 --> 00:33:57.820
to hear the product name.
765
00:33:57.820 --> 00:34:00.750
The faster the beeps, the closer you are to the barcode.
766
00:34:00.750 --> 00:34:03.700
I can imagine this maybe being useful in a grocery store.
767
00:34:05.780 --> 00:34:06.870
Processing.
768
00:34:06.870 --> 00:34:08.640
Skippy creamy peanut butter.
769
00:34:08.640 --> 00:34:09.493
Ooh it did it!
770
00:34:10.530 --> 00:34:12.410
With a more traditional system,
771
00:34:12.410 --> 00:34:15.960
a lot of the programmer would be deciding what's of interest
772
00:34:15.960 --> 00:34:19.880
and what should be described very mechanically,
773
00:34:19.880 --> 00:34:21.900 line:15%
whereas with machine learning system,
774
00:34:21.900 --> 00:34:25.610 line:15%
where showing the system many many thousands of photos
775
00:34:25.610 --> 00:34:28.260
and it is using a neural network, deep learning,
776
00:34:28.260 --> 00:34:30.950
to identify the patterns.
777
00:34:30.950 --> 00:34:31.990
Scan your surroundings
778
00:34:31.990 --> 00:34:34.230
to find out how many people are around you,
779
00:34:34.230 --> 00:34:37.292
how close they are, and their facial expressions.
780
00:34:37.292 --> 00:34:40.110
What'll it do, it'll read their emotions?
781
00:34:40.110 --> 00:34:41.390
That's interesting.
782
00:34:41.390 --> 00:34:42.986
One face near bottom edge processing.
783
00:34:42.986 --> 00:34:44.294
Ah!
784
00:34:44.294 --> 00:34:46.040
27 year old woman
785
00:34:46.040 --> 00:34:48.485
with brown hair wearing glasses looking neutral.
786
00:34:48.485 --> 00:34:51.030
[laughs]
787
00:34:51.030 --> 00:34:51.903
All right.
788
00:34:53.470 --> 00:34:55.470
When I'm walking I sort of just tough it out.
789
00:34:55.470 --> 00:34:57.823
I can't really tell what's going on around me.
790
00:34:57.823 --> 00:35:00.230
I know where all the environmental stuff is
791
00:35:00.230 --> 00:35:01.772
'cause you do it over and over again.
792
00:35:01.772 --> 00:35:03.140
It's the people.
793
00:35:03.140 --> 00:35:05.690
I have no idea where the people are.
794
00:35:05.690 --> 00:35:07.520
So if I had something that could tell me,
795
00:35:07.520 --> 00:35:10.140
like, your friend is coming, or your friend is like,
796
00:35:10.140 --> 00:35:13.039
over here, to your left, that would be great.
797
00:35:13.039 --> 00:35:16.980
One face near left edge over 14 feet away.
798
00:35:16.980 --> 00:35:19.003
Chris near center three feet away.
799
00:35:19.920 --> 00:35:21.055
It got you eventually, once you stopped.
800
00:35:21.055 --> 00:35:22.900
Zero faces.
801
00:35:22.900 --> 00:35:25.570
I think the potential is really there,
802
00:35:25.570 --> 00:35:28.920
to really like, to change lives.
803
00:35:28.920 --> 00:35:29.927
To change my life, even.
804
00:35:29.927 --> 00:35:30.948
Zero faces.
805
00:35:30.948 --> 00:35:32.760
One face near right edge center.
806
00:35:32.760 --> 00:35:35.017
AI will help understand the world around us
807
00:35:35.017 --> 00:35:38.750
and I really think that could level the playing field
808
00:35:38.750 --> 00:35:41.213
for everyone and make the world more inclusive.
809
00:35:42.420 --> 00:35:46.100
AI will give us back things we've lost.
810
00:35:46.100 --> 00:35:49.850
Old people are using it to make their lives richer,
811
00:35:49.850 --> 00:35:52.840
or to give them back capabilities they've lost.
812
00:35:52.840 --> 00:35:54.930
How is everything with you?
813
00:35:54.930 --> 00:35:57.040
AI is gonna be really helpfuL making sure
814
00:35:57.040 --> 00:35:58.340
that people are connected.
815
00:36:00.000 --> 00:36:02.640
There's a lot of people right now that can't drive.
816
00:36:02.640 --> 00:36:05.963
Giving them mobility, I think that's really transformative.
817
00:36:07.580 --> 00:36:10.510
One of the ways that self driving cars will work first
818
00:36:10.510 --> 00:36:12.690
is in contained environments.
819
00:36:12.690 --> 00:36:14.730
Where they've mapped all the roads.
820
00:36:14.730 --> 00:36:17.053
Where the weather is predictable.
821
00:36:19.355 --> 00:36:21.810
The car does not need to be able to drive very quickly.
822
00:36:21.810 --> 00:36:23.160
There isn't a lot of traffic.
823
00:36:23.160 --> 00:36:26.450
People just need some help to get around.
824
00:36:26.450 --> 00:36:27.980
For myself, I'm still working,
825
00:36:27.980 --> 00:36:30.430
I'm perfectly capable of driving around.
826
00:36:30.430 --> 00:36:33.030 line:15%
But living in a 55 plus community,
827
00:36:33.030 --> 00:36:37.420
I have neighbors who are in their 90s, who really can't.
828
00:36:37.420 --> 00:36:41.120
So this is my mom and my father George Levokiam.
829
00:36:41.120 --> 00:36:43.750
This was their home in Riverdale,
830
00:36:43.750 --> 00:36:45.130
where they became housebound
831
00:36:45.130 --> 00:36:47.603
because they couldn't drive any longer.
832
00:36:48.950 --> 00:36:53.300
Here were two, really active, vibrant people,
833
00:36:53.300 --> 00:36:56.420
and they couldn't do anything without calling somebody
834
00:36:56.420 --> 00:36:57.470
to drive them.
835
00:36:57.470 --> 00:37:01.370
But it's really hard to be dependent on other people.
836
00:37:01.370 --> 00:37:05.360
For older people, I see the autonomous vehicle
837
00:37:05.360 --> 00:37:06.662
as independence.
838
00:37:06.662 --> 00:37:09.520
[upbeat music]
839
00:37:09.520 --> 00:37:12.040 line:15%
Self driving cars are uniquely positioned
840
00:37:12.040 --> 00:37:14.570 line:15%
to help our seniors, because there's a lot of people here
841
00:37:14.570 --> 00:37:17.050
that shouldn't be driving anymore.
842
00:37:17.050 --> 00:37:19.230
Makes it easier for people to make a decision
843
00:37:19.230 --> 00:37:21.570
about stopping driving, is when you have something
844
00:37:21.570 --> 00:37:23.240
that can take you down to the clubhouse
845
00:37:23.240 --> 00:37:24.503
or to the fitness center.
846
00:37:26.170 --> 00:37:28.600
These technologies are all helping seniors,
847
00:37:28.600 --> 00:37:31.280
because they will allow us to live more independently longer
848
00:37:31.280 --> 00:37:32.173
and more safely.
849
00:37:33.240 --> 00:37:35.210
I really wish they had this with my parents.
850
00:37:35.210 --> 00:37:37.393
God, it would've changed their life.
851
00:37:38.580 --> 00:37:42.563
At every stage in life, there's some way that AI helps.
852
00:37:44.460 --> 00:37:45.720
It's gonna make our lives longer,
853
00:37:45.720 --> 00:37:47.080
it's gonna make them richer,
854
00:37:47.080 --> 00:37:49.433
it's gonna expand our imaginations.
855
00:37:50.560 --> 00:37:53.760
Some people say that AI is going to enhance our humanity.
856
00:37:53.760 --> 00:37:55.083
I don't know if it's true.
857
00:37:55.940 --> 00:37:58.950
You know, has technology enhanced our humanity?
858
00:37:58.950 --> 00:38:00.113
Twitter sure hasn't.
859
00:38:02.420 --> 00:38:03.800
If Twitter was a real place
860
00:38:03.800 --> 00:38:06.170
it would be a terrible place to live.
861
00:38:06.170 --> 00:38:10.091
So I'm not sure what all this technology is gonna do to us.
862
00:38:10.091 --> 00:38:11.800
Is it making us better, is it making us worse?
863
00:38:11.800 --> 00:38:13.870
It's surely making us different.
864
00:38:13.870 --> 00:38:14.703
That's for sure.
865
00:38:15.590 --> 00:38:16.500
Timer, are you ready?
866
00:38:16.500 --> 00:38:17.333
Yeah.
867
00:38:17.333 --> 00:38:18.227
Racer, are you ready?
868
00:38:18.227 --> 00:38:21.440
You know a part of what made this moment
869
00:38:21.440 --> 00:38:23.590
the way it is now is that we invited a whole lot of people
870
00:38:23.590 --> 00:38:24.423
into it.
871
00:38:24.423 --> 00:38:26.011
We didn't hold it tightly.
872
00:38:26.011 --> 00:38:27.775
On your mark, get set, go.
873
00:38:27.775 --> 00:38:28.751
[inspiring music]
874
00:38:28.751 --> 00:38:30.580
[robocar engine whirs]
875
00:38:30.580 --> 00:38:32.310
So he's trying to beat 33.33.
876
00:38:36.080 --> 00:38:39.280
We said, here are platforms, build things.
877
00:38:39.280 --> 00:38:41.897
Here are platforms, imagine what's possible.
878
00:38:43.827 --> 00:38:45.003
Timer, what's the time?
879
00:38:45.003 --> 00:38:45.886
25.6.
880
00:38:45.886 --> 00:38:46.719
Yeah!
881
00:38:46.719 --> 00:38:48.551
25.6, that's gotta be a record.
882
00:38:48.551 --> 00:38:49.384
Whoo!
883
00:38:49.384 --> 00:38:50.730
[clapping]
884
00:38:50.730 --> 00:38:54.470
We think what we're doing with AI/ML is really good
885
00:38:54.470 --> 00:38:56.940
but this is nothing compared to what the brain does,
886
00:38:56.940 --> 00:38:58.581
so we're nowhere close yet.
887
00:38:58.581 --> 00:38:59.624
[robocar engine whirs]
888
00:38:59.624 --> 00:39:01.380
[audience yelling]
889
00:39:01.380 --> 00:39:03.030
Humans are still winning for now.
890
00:39:04.320 --> 00:39:06.203
Our processor is much better,
891
00:39:07.180 --> 00:39:11.230
but you know I guess the hope is that one day it'll be--
892
00:39:11.230 --> 00:39:12.063
It'll be, we'll get there.
893
00:39:12.063 --> 00:39:14.247
It'll be up to speed.
894
00:39:14.247 --> 00:39:17.880
America's 2018 top Young Scientist is
895
00:39:20.047 --> 00:39:21.606
Rishab Jain.
896
00:39:21.606 --> 00:39:24.939
[clapping and cheering]
897
00:39:27.220 --> 00:39:30.530
I realize that not every kid has the engagement
898
00:39:30.530 --> 00:39:33.800
and interaction with STEM in general,
899
00:39:33.800 --> 00:39:37.500
so I recently formed a nonprofit organization
900
00:39:37.500 --> 00:39:39.590
called Samyak Science Society,
901
00:39:39.590 --> 00:39:43.180
so I want to further promote artificial intelligence
902
00:39:43.180 --> 00:39:47.270
in my community to help solve like real world problems.
903
00:39:47.270 --> 00:39:49.290
In the future I think your house will diagnose you
904
00:39:49.290 --> 00:39:51.740
everyday, your car, every time you take a shower
905
00:39:51.740 --> 00:39:52.710
you're gonna get a skin exam,
906
00:39:52.710 --> 00:39:54.450
anytime you look into your vanity mirror
907
00:39:54.450 --> 00:39:56.060
you're gonna get an eye exam,
908
00:39:56.060 --> 00:39:58.370
every time you sleep you're gonna measure
909
00:39:58.370 --> 00:40:01.080
your weight distribution to understand whether you're
910
00:40:01.080 --> 00:40:02.410
at risk for congestive heart failure,
911
00:40:02.410 --> 00:40:04.540
every time you touch the steering wheel in the car
912
00:40:04.540 --> 00:40:06.211
you could get a full EKG.
913
00:40:06.211 --> 00:40:09.107
I believe we're gonna invent flying cars
914
00:40:09.107 --> 00:40:10.890
so there's no more traffic in the world.
915
00:40:10.890 --> 00:40:13.950
I believe we're gonna find a way to live twice as long.
916
00:40:13.950 --> 00:40:15.210
I believe you're gonna find a way
917
00:40:15.210 --> 00:40:18.060
to seamlessly blend digital information our brains.
918
00:40:18.060 --> 00:40:21.780
So I think people really have to maintain that optimism.
919
00:40:21.780 --> 00:40:25.661
Technology has always helped human make progress.
920
00:40:25.661 --> 00:40:28.410
AI is an amazing technology.
921
00:40:28.410 --> 00:40:30.880
It's happening, and it's exciting.
922
00:40:30.880 --> 00:40:31.747
Processing.
923
00:40:31.747 --> 00:40:33.741
A close up of a blackboard.
924
00:40:33.741 --> 00:40:36.580
It's also true that there are a lot of risks
925
00:40:36.580 --> 00:40:39.080
with AI, and so we need to be thoughtful.
926
00:40:39.080 --> 00:40:41.371
We need to be deliberate as we move forward
927
00:40:41.371 --> 00:40:44.283
into this crazy new world that AI is creating.
928
00:40:45.310 --> 00:40:47.797
You really don't know what the future can hold for you
929
00:40:47.797 --> 00:40:49.680
but you gotta be ready for it.
930
00:40:49.680 --> 00:40:53.149
[dramatic music swells]
931
00:40:53.149 --> 00:40:56.566
[futuristic piano music]