0:00:00.556,0:00:04.573
Our emotions influence[br]every aspect of our lives,

0:00:04.573,0:00:08.149
from our health and how we learn,[br]to how we do business and make decisions,

0:00:08.149,0:00:09.922
big ones and small.

0:00:10.672,0:00:14.162
Our emotions also influence[br]how we connect with one another.

0:00:15.132,0:00:19.108
We've evolved to live[br]in a world like this,

0:00:19.108,0:00:23.427
but instead, we're living[br]more and more of our lives like this --

0:00:23.427,0:00:26.561
this is the text message[br]from my daughter last night --

0:00:26.561,0:00:29.301
in a world that's devoid of emotion.

0:00:29.301,0:00:31.252
So I'm on a mission to change that.

0:00:31.252,0:00:35.343
I want to bring emotions[br]back into our digital experiences.

0:00:36.223,0:00:39.300
I started on this path 15 years ago.

0:00:39.300,0:00:41.366
I was a computer scientist in Egypt,

0:00:41.366,0:00:45.871
and I had just gotten accepted to[br]a Ph.D. program at Cambridge University.

0:00:45.871,0:00:47.984
So I did something quite unusual

0:00:47.984,0:00:52.209
for a young newlywed Muslim Egyptian wife:

0:00:53.599,0:00:56.598
With the support of my husband,[br]who had to stay in Egypt,

0:00:56.598,0:00:59.616
I packed my bags and I moved to England.

0:00:59.616,0:01:02.844
At Cambridge, thousands of miles[br]away from home,

0:01:02.844,0:01:06.257
I realized I was spending[br]more hours with my laptop

0:01:06.257,0:01:08.486
than I did with any other human.

0:01:08.486,0:01:13.339
Yet despite this intimacy, my laptop[br]had absolutely no idea how I was feeling.

0:01:13.339,0:01:16.550
It had no idea if I was happy,

0:01:16.550,0:01:19.538
having a bad day, or stressed, confused,

0:01:19.538,0:01:22.460
and so that got frustrating.

0:01:23.600,0:01:28.831
Even worse, as I communicated[br]online with my family back home,

0:01:29.421,0:01:32.703
I felt that all my emotions[br]disappeared in cyberspace.

0:01:32.703,0:01:37.858
I was homesick, I was lonely,[br]and on some days I was actually crying,

0:01:37.858,0:01:42.786
but all I had to communicate[br]these emotions was this.

0:01:42.786,0:01:44.806
(Laughter)

0:01:44.806,0:01:49.780
Today's technology[br]has lots of I.Q., but no E.Q.;

0:01:49.780,0:01:52.956
lots of cognitive intelligence,[br]but no emotional intelligence.

0:01:52.956,0:01:55.153
So that got me thinking,

0:01:55.153,0:01:58.777
what if our technology[br]could sense our emotions?

0:01:58.777,0:02:02.853
What if our devices could sense[br]how we felt and reacted accordingly,

0:02:02.853,0:02:05.866
just the way an emotionally[br]intelligent friend would?

0:02:06.666,0:02:10.230
Those questions led me and my team

0:02:10.230,0:02:14.607
to create technologies that can read[br]and respond to our emotions,

0:02:14.607,0:02:17.697
and our starting point was the human face.

0:02:18.577,0:02:21.750
So our human face happens to be[br]one of the most powerful channels

0:02:21.750,0:02:25.766
that we all use to communicate[br]social and emotional states,

0:02:25.766,0:02:28.776
everything from enjoyment, surprise,

0:02:28.776,0:02:32.979
empathy and curiosity.

0:02:32.979,0:02:37.907
In emotion science, we call each[br]facial muscle movement an action unit.

0:02:37.907,0:02:40.832
So for example, action unit 12,

0:02:40.832,0:02:42.870
it's not a Hollywood blockbuster,

0:02:42.870,0:02:46.312
it is actually a lip corner pull,[br]which is the main component of a smile.

0:02:46.312,0:02:49.300
Try it everybody. Let's get[br]some smiles going on.

0:02:49.300,0:02:51.954
Another example is action unit 4.[br]It's the brow furrow.

0:02:51.954,0:02:54.192
It's when you draw your eyebrows together

0:02:54.192,0:02:56.459
and you create all[br]these textures and wrinkles.

0:02:56.459,0:03:00.754
We don't like them, but it's[br]a strong indicator of a negative emotion.

0:03:00.754,0:03:02.960
So we have about 45 of these action units,

0:03:02.960,0:03:06.350
and they combine to express[br]hundreds of emotions.

0:03:06.350,0:03:10.251
Teaching a computer to read[br]these facial emotions is hard,

0:03:10.251,0:03:13.223
because these action units,[br]they can be fast, they're subtle,

0:03:13.223,0:03:15.777
and they combine in many different ways.

0:03:15.777,0:03:19.515
So take, for example,[br]the smile and the smirk.

0:03:19.515,0:03:23.268
They look somewhat similar,[br]but they mean very different things.

0:03:23.268,0:03:24.986
(Laughter)

0:03:24.986,0:03:27.990
So the smile is positive,

0:03:27.990,0:03:29.260
a smirk is often negative.

0:03:29.260,0:03:33.136
Sometimes a smirk[br]can make you become famous.

0:03:33.136,0:03:35.960
But seriously, it's important[br]for a computer to be able

0:03:35.960,0:03:38.815
to tell the difference[br]between the two expressions.

0:03:38.815,0:03:40.627
So how do we do that?

0:03:40.627,0:03:42.414
We give our algorithms

0:03:42.414,0:03:46.524
tens of thousands of examples[br]of people we know to be smiling,

0:03:46.524,0:03:49.589
from different ethnicities, ages, genders,

0:03:49.589,0:03:52.400
and we do the same for smirks.

0:03:52.400,0:03:53.954
And then, using deep learning,

0:03:53.954,0:03:56.810
the algorithm looks for all these[br]textures and wrinkles

0:03:56.810,0:03:59.390
and shape changes on our face,

0:03:59.390,0:04:02.592
and basically learns that all smiles[br]have common characteristics,

0:04:02.592,0:04:05.773
all smirks have subtly[br]different characteristics.

0:04:05.773,0:04:08.141
And the next time it sees a new face,

0:04:08.141,0:04:10.440
it essentially learns that

0:04:10.440,0:04:13.473
this face has the same[br]characteristics of a smile,

0:04:13.473,0:04:17.751
and it says, "Aha, I recognize this.[br]This is a smile expression."

0:04:18.381,0:04:21.181
So the best way to demonstrate[br]how this technology works

0:04:21.181,0:04:23.317
is to try a live demo,

0:04:23.317,0:04:27.230
so I need a volunteer,[br]preferably somebody with a face.

0:04:27.230,0:04:29.564
(Laughter)

0:04:29.564,0:04:32.335
Cloe's going to be our volunteer today.

0:04:33.325,0:04:37.783
So over the past five years, we've moved[br]from being a research project at MIT

0:04:37.783,0:04:38.939
to a company,

0:04:38.939,0:04:42.131
where my team has worked really hard[br]to make this technology work,

0:04:42.131,0:04:44.540
as we like to say, in the wild.

0:04:44.540,0:04:47.210
And we've also shrunk it so that[br]the core emotion engine

0:04:47.210,0:04:50.530
works on any mobile device[br]with a camera, like this iPad.

0:04:50.530,0:04:53.316
So let's give this a try.

0:04:54.756,0:04:58.680
As you can see, the algorithm[br]has essentially found Cloe's face,

0:04:58.680,0:05:00.372
so it's this white bounding box,

0:05:00.372,0:05:02.943
and it's tracking the main[br]feature points on her face,

0:05:02.943,0:05:05.799
so her eyebrows, her eyes,[br]her mouth and her nose.

0:05:05.799,0:05:08.786
The question is,[br]can it recognize her expression?

0:05:08.786,0:05:10.457
So we're going to test the machine.

0:05:10.457,0:05:14.643
So first of all, give me your poker face.[br]Yep, awesome. (Laughter)

0:05:14.643,0:05:17.456
And then as she smiles,[br]this is a genuine smile, it's great.

0:05:17.456,0:05:19.756
So you can see the green bar[br]go up as she smiles.

0:05:19.756,0:05:20.978
Now that was a big smile.

0:05:20.978,0:05:24.021
Can you try a subtle smile[br]to see if the computer can recognize?

0:05:24.021,0:05:26.352
It does recognize subtle smiles as well.

0:05:26.352,0:05:28.477
We've worked really hard[br]to make that happen.

0:05:28.477,0:05:31.439
And then eyebrow raised,[br]indicator of surprise.

0:05:31.439,0:05:35.688
Brow furrow, which is[br]an indicator of confusion.

0:05:35.688,0:05:39.695
Frown. Yes, perfect.

0:05:39.695,0:05:43.188
So these are all the different[br]action units. There's many more of them.

0:05:43.188,0:05:45.220
This is just a slimmed-down demo.

0:05:45.220,0:05:48.368
But we call each reading[br]an emotion data point,

0:05:48.368,0:05:51.337
and then they can fire together[br]to portray different emotions.

0:05:51.337,0:05:55.990
So on the right side of the demo --[br]look like you're happy.

0:05:55.990,0:05:57.444
So that's joy. Joy fires up.

0:05:57.444,0:05:59.371
And then give me a disgust face.

0:05:59.371,0:06:03.643
Try to remember what it was like[br]when Zayn left One Direction.

0:06:03.643,0:06:05.153
(Laughter)

0:06:05.153,0:06:09.495
Yeah, wrinkle your nose. Awesome.

0:06:09.495,0:06:13.226
And the valence is actually quite[br]negative, so you must have been a big fan.

0:06:13.226,0:06:15.926
So valence is how positive[br]or negative an experience is,

0:06:15.926,0:06:18.712
and engagement is how[br]expressive she is as well.

0:06:18.712,0:06:22.126
So imagine if Cloe had access[br]to this real-time emotion stream,

0:06:22.126,0:06:24.935
and she could share it[br]with anybody she wanted to.

0:06:24.935,0:06:27.858
Thank you.

0:06:27.858,0:06:32.479
(Applause)

0:06:33.749,0:06:39.019
So, so far, we have amassed[br]12 billion of these emotion data points.

0:06:39.019,0:06:41.630
It's the largest emotion[br]database in the world.

0:06:41.630,0:06:44.593
We've collected it[br]from 2.9 million face videos,

0:06:44.593,0:06:47.193
people who have agreed[br]to share their emotions with us,

0:06:47.193,0:06:50.398
and from 75 countries around the world.

0:06:50.398,0:06:52.113
It's growing every day.

0:06:52.603,0:06:54.670
It blows my mind away

0:06:54.670,0:06:57.865
that we can now quantify something[br]as personal as our emotions,

0:06:57.865,0:07:00.100
and we can do it at this scale.

0:07:00.100,0:07:02.277
So what have we learned to date?

0:07:03.057,0:07:05.388
Gender.

0:07:05.388,0:07:09.034
Our data confirms something[br]that you might suspect.

0:07:09.034,0:07:10.891
Women are more expressive than men.

0:07:10.891,0:07:13.574
Not only do they smile more,[br]their smiles last longer,

0:07:13.574,0:07:16.478
and we can now really quantify[br]what it is that men and women

0:07:16.478,0:07:18.614
respond to differently.

0:07:18.614,0:07:20.904
Let's do culture: So in the United States,

0:07:20.904,0:07:24.108
women are 40 percent[br]more expressive than men,

0:07:24.108,0:07:27.753
but curiously, we don't see any difference[br]in the U.K. between men and women.

0:07:27.753,0:07:30.259
(Laughter)

0:07:31.296,0:07:35.323
Age: People who are 50 years and older

0:07:35.323,0:07:38.759
are 25 percent more emotive[br]than younger people.

0:07:39.899,0:07:43.751
Women in their 20s smile a lot more[br]than men the same age,

0:07:43.751,0:07:47.590
perhaps a necessity for dating.

0:07:47.590,0:07:50.207
But perhaps what surprised us[br]the most about this data

0:07:50.207,0:07:53.410
is that we happen[br]to be expressive all the time,

0:07:53.410,0:07:56.243
even when we are sitting[br]in front of our devices alone,

0:07:56.243,0:07:59.517
and it's not just when we're watching[br]cat videos on Facebook.

0:08:00.217,0:08:03.227
We are expressive when we're emailing,[br]texting, shopping online,

0:08:03.227,0:08:05.527
or even doing our taxes.

0:08:05.527,0:08:07.919
Where is this data used today?

0:08:07.919,0:08:10.682
In understanding how we engage with media,

0:08:10.682,0:08:13.166
so understanding virality[br]and voting behavior;

0:08:13.166,0:08:15.906
and also empowering[br]or emotion-enabling technology,

0:08:15.906,0:08:20.527
and I want to share some examples[br]that are especially close to my heart.

0:08:21.197,0:08:24.265
Emotion-enabled wearable glasses[br]can help individuals

0:08:24.265,0:08:27.493
who are visually impaired[br]read the faces of others,

0:08:27.493,0:08:31.680
and it can help individuals[br]on the autism spectrum interpret emotion,

0:08:31.680,0:08:34.458
something that they really struggle with.

0:08:35.918,0:08:38.777
In education, imagine[br]if your learning apps

0:08:38.777,0:08:41.587
sense that you're confused and slow down,

0:08:41.587,0:08:43.444
or that you're bored, so it's sped up,

0:08:43.444,0:08:46.413
just like a great teacher[br]would in a classroom.

0:08:47.043,0:08:49.644
What if your wristwatch tracked your mood,

0:08:49.644,0:08:52.337
or your car sensed that you're tired,

0:08:52.337,0:08:54.885
or perhaps your fridge[br]knows that you're stressed,

0:08:54.885,0:09:00.951
so it auto-locks to prevent you[br]from binge eating. (Laughter)

0:09:00.951,0:09:03.668
I would like that, yeah.

0:09:03.668,0:09:05.595
What if, when I was in Cambridge,

0:09:05.595,0:09:07.908
I had access to my real-time[br]emotion stream,

0:09:07.908,0:09:11.437
and I could share that with my family[br]back home in a very natural way,

0:09:11.437,0:09:15.408
just like I would've if we were all[br]in the same room together?

0:09:15.408,0:09:18.550
I think five years down the line,

0:09:18.550,0:09:20.887
all our devices are going[br]to have an emotion chip,

0:09:20.887,0:09:24.951
and we won't remember what it was like[br]when we couldn't just frown at our device

0:09:24.951,0:09:29.200
and our device would say, "Hmm,[br]you didn't like that, did you?"

0:09:29.200,0:09:32.961
Our biggest challenge is that there are[br]so many applications of this technology,

0:09:32.961,0:09:35.864
my team and I realize that we can't[br]build them all ourselves,

0:09:35.864,0:09:39.360
so we've made this technology available[br]so that other developers

0:09:39.360,0:09:41.474
can get building and get creative.

0:09:41.474,0:09:45.560
We recognize that[br]there are potential risks

0:09:45.560,0:09:47.627
and potential for abuse,

0:09:47.627,0:09:50.576
but personally, having spent[br]many years doing this,

0:09:50.576,0:09:53.548
I believe that the benefits to humanity

0:09:53.548,0:09:55.823
from having emotionally[br]intelligent technology

0:09:55.823,0:09:59.399
far outweigh the potential for misuse.

0:09:59.399,0:10:01.930
And I invite you all to be[br]part of the conversation.

0:10:01.930,0:10:04.484
The more people who know[br]about this technology,

0:10:04.484,0:10:07.661
the more we can all have a voice[br]in how it's being used.

0:10:09.081,0:10:13.655
So as more and more[br]of our lives become digital,

0:10:13.655,0:10:17.153
we are fighting a losing battle[br]trying to curb our usage of devices

0:10:17.153,0:10:19.382
in order to reclaim our emotions.

0:10:20.622,0:10:24.536
So what I'm trying to do instead[br]is to bring emotions into our technology

0:10:24.536,0:10:26.765
and make our technologies more responsive.

0:10:26.765,0:10:29.435
So I want those devices[br]that have separated us

0:10:29.435,0:10:31.897
to bring us back together.

0:10:31.897,0:10:36.485
And by humanizing technology,[br]we have this golden opportunity

0:10:36.485,0:10:39.782
to reimagine how we[br]connect with machines,

0:10:39.782,0:10:44.263
and therefore, how we, as human beings,

0:10:44.263,0:10:46.167
connect with one another.

0:10:46.167,0:10:48.327
Thank you.

0:10:48.327,0:10:51.640
(Applause)