0:00:06.361,0:00:08.084
Marita Cheng: When I was growing up,

0:00:08.085,0:00:11.595
I had a family friend[br]who became blind in his 20s.

0:00:12.438,0:00:15.602
When we went out as a family,[br]he would say to me,

0:00:15.603,0:00:19.881
"Rita hold my hand, hold my arm,[br]and tell me what you see."

0:00:19.882,0:00:23.682
So I'd say, "There's [br]some flowers here to the left.

0:00:23.683,0:00:25.681
There's a gate here to the right.

0:00:25.682,0:00:28.212
There's a mountain over in the distance."

0:00:28.213,0:00:31.493
And he would say,[br]"What color are those flowers?

0:00:32.403,0:00:34.753
Can I use my hand,[br]reach out, and touch them?

0:00:34.754,0:00:36.322
Could you lead my hand to that?"

0:00:36.323,0:00:39.622
I'd say, "Oh, they're pink, they're blue."

0:00:39.623,0:00:42.833
And he'd say, "Tell me more,[br]tell me more about what you can see.

0:00:42.834,0:00:44.062
Share it with me."

0:00:44.063,0:00:48.592
About eight months ago,[br]Alberto and I decided to create an app

0:00:48.593,0:00:52.663
to enable blind people[br]to recognize their surroundings.

0:00:52.667,0:00:55.959
We used something called[br]convolutional neural networks,

0:00:55.972,0:01:00.283
which is a computer system[br]that's been trained on millions of images.

0:01:00.284,0:01:04.242
It learns the features of a dog.[br]It learns what a flower looks like.

0:01:04.244,0:01:07.964
It learns a fork, a knife,[br]everyday objects.

0:01:09.667,0:01:14.418
Using this system,[br]we created something called "Aipoly"

0:01:14.454,0:01:17.734
that recognizes[br]over 1,000 everyday objects.

0:01:17.735,0:01:21.598
So, a blind person just needs[br]to walk around with their phone,

0:01:21.599,0:01:26.924
and put it over various objects,[br]and it will say the name of the object.

0:01:27.444,0:01:32.773
Using voice over, the phone can relay[br]the word on the screen

0:01:32.774,0:01:34.113
to that blind person,

0:01:34.114,0:01:36.316
so they know exactly[br]what's in front of them.

0:01:36.317,0:01:38.193
(Applause)

0:01:46.103,0:01:49.103
Since we released[br]our application in January,

0:01:49.104,0:01:52.634
we've had over a 100,000 downloads[br]around the world.

0:01:53.536,0:01:58.444
The app has been so popular[br]we've translated it into seven languages.

0:01:59.167,0:02:02.376
Alberto Rizzoli: After experiencing[br]the technology the first time,

0:02:02.377,0:02:04.764
our users kept asking us for more.

0:02:05.555,0:02:09.384
We asked them

0:02:09.386,0:02:14.225
to think of our technology[br]as a superpower for a moment,

0:02:14.226,0:02:19.425
something they could effortlessly[br]evoke at any time,

0:02:19.426,0:02:22.346
and gain understanding[br]of what was in front of them.

0:02:23.080,0:02:25.705
And surprisingly,

0:02:25.706,0:02:29.545
nobody really wants[br]X-ray vision or telescopic goggles,

0:02:29.546,0:02:33.166
but what everyone wants[br]is more information.

0:02:33.167,0:02:34.495
It's not surprising

0:02:34.496,0:02:40.134
because 60% of the information[br]that we perceive comes through sight.

0:02:40.135,0:02:41.846
It is the main tool that we use

0:02:41.847,0:02:45.994
to understand our surroundings[br]and often, to make decisions.

0:02:45.995,0:02:50.356
If you're blind you must rely[br]on other senses like touch or hearing,

0:02:50.357,0:02:53.465
and you miss out[br]on the lightning-fast identification

0:02:53.466,0:02:58.717
that our brain and eyes do every second[br]of every day, if you're a sighted person.

0:02:59.077,0:03:02.154
We went

0:03:02.155,0:03:05.024
to the Santa Clara Valley Blind Center,

0:03:05.025,0:03:07.606
and we tried to build this superpower.

0:03:07.607,0:03:10.577
We tried to see what kind[br]of information people wanted,

0:03:10.578,0:03:13.565
and it's simple things[br]like whether a dish is clean or not,

0:03:13.566,0:03:17.655
whether you can cross the street,[br]what product am I looking at?

0:03:17.656,0:03:19.396
Things that can lead to a decision,

0:03:19.397,0:03:23.797
from a simple gaze, to an understanding[br]of the situation in front of you.

0:03:24.557,0:03:28.887
We asked what form factor[br]people preferred, and we built it.

0:03:29.707,0:03:31.485
So we put together

0:03:31.486,0:03:37.147
some jawbone conductive headphones,[br]a pair of sunglasses, and a tiny camera,

0:03:38.177,0:03:42.046
and we asked our friends[br]to think of a common situation

0:03:42.047,0:03:45.056
in which they had to make[br]many small decisions,

0:03:45.057,0:03:47.857
and we told them[br]we will be giving them the prototype,

0:03:47.858,0:03:50.166
and taking them[br]in the middle of that situation.

0:03:50.167,0:03:52.003
Let's see how it went. (Video starts)

0:03:52.004,0:03:53.351
[We asked blind individuals

0:03:53.352,0:03:56.089
[what are the hardest things to do[br]when visually impaired]

0:03:56.090,0:03:58.656
I mean it takes me forever[br]to go grocery-shopping.

0:03:58.657,0:04:02.316
Even with someone helping me[br]that has known me for years.

0:04:02.317,0:04:05.827
I'll say, "What's in that cabinet?"

0:04:07.097,0:04:10.397
Or develop a system,[br]right to left, top to bottom.

0:04:10.398,0:04:13.058
[So we took them grocery shopping[br]with our technology]

0:04:13.358,0:04:14.728
Computer: Oranges.

0:04:16.048,0:04:18.618
Man: This is great. I'm really liking it.

0:04:19.567,0:04:23.237
Apples, grapes, carrots.

0:04:23.238,0:04:25.626
I'm looking, I'm looking.

0:04:25.627,0:04:27.576
Computer: Lilies.[br]Man: Lilies.

0:04:27.577,0:04:31.317
Computer: Bouquet.[br]Man: A bouquet, ahh!

0:04:31.318,0:04:33.211
Computer: Roses, flowers.

0:04:33.212,0:04:35.177
Man: Can I take these home?

0:04:35.178,0:04:36.778
This is great.

0:04:36.792,0:04:38.710
Computer: Roses.[br]Man: Roses.

0:04:38.711,0:04:41.367
Computer: Bouquet, tulips.[br]Man: Tulips.

0:04:41.368,0:04:44.019
Computer: Pineapple.[br]Woman: It's a pineapple.

0:04:45.089,0:04:46.830
Computer: Mango.[br]Woman: Mango.

0:04:46.831,0:04:48.389
Computer: M&Ms.[br]Woman: M&Ms.

0:04:48.375,0:04:50.750
Computer: Tic Tac.[br]Woman: It just said, "Tic Tac."

0:04:50.759,0:04:52.377
Computer: Tic Tac.[br]Woman: Tic Tac.

0:04:52.378,0:04:54.309
Computer: Paper note, calendar.

0:04:54.310,0:04:55.719
Woman: Calendar, you got it.

0:04:56.329,0:04:59.318
Wow, I didn't know what that was at all.

0:04:59.319,0:05:01.288
Computer: Pretzels.[br]Woman: Pretzels.

0:05:01.289,0:05:03.609
Computer: Pretzels.[br]Woman: It said, "Pretzels."

0:05:03.610,0:05:05.409
Computer: Lipton tea.[br]Woman 2: Lipton?

0:05:05.410,0:05:06.589
Tea?

0:05:06.590,0:05:08.548
Computer: Lipton teabags.

0:05:08.549,0:05:12.209
Woman 2: It's like I'm seeing it,[br]but I'm not, it's seeing it for me.

0:05:12.210,0:05:14.971
Computer: Coffee mate.[br]Woman 2: Mate; coffee mate.

0:05:16.281,0:05:18.739
It didn't say "coffee,"[br]but it kept saying "mate."

0:05:18.740,0:05:21.369
Computer: Mate, mate.

0:05:22.190,0:05:24.731
Man 2: I put on the glasses,

0:05:24.732,0:05:27.869
and right away, it told me[br]there was an apple, there were oranges,

0:05:27.870,0:05:32.561
and there was this, and there was that,[br]and it's like, "This is great!"

0:05:33.420,0:05:34.799
Instant love.

0:05:34.800,0:05:35.966
(Video ends)

0:05:35.967,0:05:37.301
(Applause)

0:05:45.251,0:05:48.759
AR: That little pair of glasses[br]connected to their phones

0:05:48.760,0:05:52.589
could identify four[br]to 5,000 objects in real time.

0:05:52.590,0:05:56.080
That's about the capacity[br]of a five-year-old child.

0:05:57.440,0:06:03.519
A simple accessory[br]can now expand a person's perception

0:06:03.520,0:06:05.900
to thousands of new possibilities.

0:06:06.630,0:06:11.456
This is the power of marrying[br]artificial and human intelligence,

0:06:12.316,0:06:14.666
and the potential[br]is still vastly untapped.

0:06:14.667,0:06:18.209
This isn't going to be a revolution[br]just because GPUs are getting faster,

0:06:18.210,0:06:20.550
or the research is getting more open,

0:06:20.551,0:06:24.250
but because the barriers of entry[br]to impacting millions of lives

0:06:24.251,0:06:27.472
for artificial intelligence[br]are getting lower and lower.

0:06:28.142,0:06:31.110
The Paralympic games[br]are starting in a few weeks,

0:06:31.111,0:06:36.731
an event where sheer force of will,[br]training, and technology

0:06:36.732,0:06:40.122
turn people with a disability[br]into super humans.

0:06:40.532,0:06:46.961
and so, too, will all ability to think,[br]perceive, make decisions, and learn

0:06:46.962,0:06:48.951
increase exponentially.

0:06:48.952,0:06:52.472
You will be building[br]the tools to make this happen.

0:06:54.032,0:06:57.291
So tomorrow, with your morning coffee,

0:06:57.292,0:07:00.652
take 40 minutes[br]and try out a tutorial on deep learning.

0:07:01.212,0:07:03.232
Build yourself a small superpower.

0:07:03.233,0:07:05.902
All it takes is your laptop,[br]and a bunch of data,

0:07:05.903,0:07:07.763
like your holiday pictures.

0:07:08.063,0:07:12.593
Superpower engineer -[br]that's a great dream job.

0:07:14.073,0:07:18.283
The good news is that the world[br]needs many, many more of them

0:07:18.632,0:07:22.303
so we can't wait to see[br]what you will be building next.

0:07:22.603,0:07:23.721
Thank you.

0:07:23.722,0:07:25.114
(Applause)