TED2015
Fei-Fei Li: How we're teaching computers to understand pictures
Fei-Fei Li: Hvordan vi lærer computere at forstå billeder
Filmed:
Readability: 4.5
2,702,344 views
Når et meget ungt barn kigger på billeder, kan hun identificere simple elementer: "Kat", "bog", "stol". Nu er computere ved at være så smarte at de kan gøre det også. Hvad er så det næste? I en spændende talk, computersyns ekspert Fei-Fei Li beskriver status - herunder en database med 15 millioner fotos som hendes team har samlet for at "lære" computere at forstå billeder - og stadigvæk mangler der meget.
Fei-Fei Li - Computer scientist
As Director of Stanford’s Artificial Intelligence Lab and Vision Lab, Fei-Fei Li is working to solve AI’s trickiest problems -- including image recognition, learning and language processing. Full bio
As Director of Stanford’s Artificial Intelligence Lab and Vision Lab, Fei-Fei Li is working to solve AI’s trickiest problems -- including image recognition, learning and language processing. Full bio
Double-click the English transcript below to play the video.
00:14
Let me show you something.
0
2366
3738
Lad mig vise jer noget.
00:18
(Video) Girl: Okay, that's a cat
sitting in a bed.
sitting in a bed.
1
6104
4156
(Video) Pige: Ok, det er en kat der
sidder i en seng.
sidder i en seng.
00:22
The boy is petting the elephant.
2
10260
4040
Drengen klapper elefanten.
00:26
Those are people
that are going on an airplane.
that are going on an airplane.
3
14300
4354
Det er folk, der går ind i en flymaskine.
00:30
That's a big airplane.
4
18654
2810
Det er en stor flymaskine.
00:33
Fei-Fei Li: This is
a three-year-old child
a three-year-old child
5
21464
2206
Fei-Fei Li: Det her er et treårigt barn
00:35
describing what she sees
in a series of photos.
in a series of photos.
6
23670
3679
der beskriver hvad hun ser i en
række billeder.
række billeder.
00:39
She might still have a lot
to learn about this world,
to learn about this world,
7
27349
2845
Hun har endnu masser at lære om verden
00:42
but she's already an expert
at one very important task:
at one very important task:
8
30194
4549
men hun er allerede ekspert i en
meget vigtig ting:
meget vigtig ting:
00:46
to make sense of what she sees.
9
34743
2846
at få mening ud af det hun ser.
00:50
Our society is more
technologically advanced than ever.
technologically advanced than ever.
10
38229
4226
Vort samfund er mere teknologisk
avanceret end nogensinde.
avanceret end nogensinde.
00:54
We send people to the moon,
we make phones that talk to us
we make phones that talk to us
11
42455
3629
Vi sender folk til månen, vi laver
telefoner der taler til os
telefoner der taler til os
00:58
or customize radio stations
that can play only music we like.
that can play only music we like.
12
46084
4946
eller tilpasser radio stationer der kan
spille musik som kun vi kan lide.
spille musik som kun vi kan lide.
01:03
Yet, our most advanced
machines and computers
machines and computers
13
51030
4055
Alligevel har de mest avancerede
maskiner og computere
maskiner og computere
01:07
still struggle at this task.
14
55085
2903
stadig problemer med dette.
01:09
So I'm here today
to give you a progress report
to give you a progress report
15
57988
3459
Jeg er her idag for at give jer
en statusrapport
en statusrapport
01:13
on the latest advances
in our research in computer vision,
in our research in computer vision,
16
61447
4047
på de nyeste fremskridt i
forskningen af computer vision
forskningen af computer vision
01:17
one of the most frontier
and potentially revolutionary
and potentially revolutionary
17
65494
4161
en af de mest banebrydende og
muligvis revolutionære
muligvis revolutionære
01:21
technologies in computer science.
18
69655
3206
teknologier i computer videnskab.
01:24
Yes, we have prototyped cars
that can drive by themselves,
that can drive by themselves,
19
72861
4551
Ja, vi har prototyper af biler,
der er selvkørende
der er selvkørende
01:29
but without smart vision,
they cannot really tell the difference
they cannot really tell the difference
20
77412
3853
men uden intelligent vision kan de
ikke se forskel
ikke se forskel
01:33
between a crumpled paper bag
on the road, which can be run over,
on the road, which can be run over,
21
81265
3970
mellem en krøllet pose på vejen,
der kan køres over
der kan køres over
01:37
and a rock that size,
which should be avoided.
which should be avoided.
22
85235
3340
og en sten af samme størrelse,
der skal undgås.
der skal undgås.
01:41
We have made fabulous megapixel cameras,
23
89415
3390
Vi har lavet fabelagtige megapixel
kameraer
kameraer
01:44
but we have not delivered
sight to the blind.
sight to the blind.
24
92805
3135
men vi har ikke givet blinde syn.
01:48
Drones can fly over massive land,
25
96420
3305
Droner kan flyve over store landområder
01:51
but don't have enough vision technology
26
99725
2134
men vi har ikke nok vision teknologi
01:53
to help us to track
the changes of the rainforests.
the changes of the rainforests.
27
101859
3461
til at følge forandringerne i regnskoven.
01:57
Security cameras are everywhere,
28
105320
2950
Sikkerhedskameraer er allevegne
02:00
but they do not alert us when a child
is drowning in a swimming pool.
is drowning in a swimming pool.
29
108270
5067
men de advarer os ikke, når et barn
drukner i en svømmepøl.
drukner i en svømmepøl.
02:06
Photos and videos are becoming
an integral part of global life.
an integral part of global life.
30
114167
5595
Foto og video er blevet en integreret
del af det globale liv.
del af det globale liv.
02:11
They're being generated at a pace
that's far beyond what any human,
that's far beyond what any human,
31
119762
4087
De bliver fremstillet hurtigere end
noget menneske
noget menneske
02:15
or teams of humans, could hope to view,
32
123849
2783
eller hold af folk kan nå at se
02:18
and you and I are contributing
to that at this TED.
to that at this TED.
33
126632
3921
og du og jeg bidrager til det
ved denne TED.
ved denne TED.
02:22
Yet our most advanced software
is still struggling at understanding
is still struggling at understanding
34
130553
5232
Vor mest avancerede software
kæmper stadig med at forstå
kæmper stadig med at forstå
02:27
and managing this enormous content.
35
135785
3876
og administrere dette enorme indhold.
02:31
So in other words,
collectively as a society,
collectively as a society,
36
139661
5272
Med andre ord er vi kollektivt som samfund
02:36
we're very much blind,
37
144933
1746
stadig meget blinde,
02:38
because our smartest
machines are still blind.
machines are still blind.
38
146679
3387
for vor klogeste maskiner er
stadig blinde.
stadig blinde.
02:43
"Why is this so hard?" you may ask.
39
151526
2926
"Hvorfor er det så svært?" spørger
du måske.
du måske.
02:46
Cameras can take pictures like this one
40
154452
2693
Kameraer tager billeder som dette
02:49
by converting lights into
a two-dimensional array of numbers
a two-dimensional array of numbers
41
157145
3994
ved at konvertere lys til en
to-dimensionel række af tal
to-dimensionel række af tal
02:53
known as pixels,
42
161139
1650
kendt som pixler
02:54
but these are just lifeless numbers.
43
162789
2251
men de er kun livløse tal.
02:57
They do not carry meaning in themselves.
44
165040
3111
De har ingen mening i sig selv.
03:00
Just like to hear is not
the same as to listen,
the same as to listen,
45
168151
4343
Ligesom at høre ikke er det samme
som at lytte
som at lytte
03:04
to take pictures is not
the same as to see,
the same as to see,
46
172494
4040
er det at tage billeder ikke det
samme som at se
samme som at se
03:08
and by seeing,
we really mean understanding.
we really mean understanding.
47
176534
3829
og ved at se mener vi at forstå.
03:13
In fact, it took Mother Nature
540 million years of hard work
540 million years of hard work
48
181293
6177
Faktisk tog det Moder Natur 540 mio. års
hårdt arbejde
hårdt arbejde
03:19
to do this task,
49
187470
1973
at løse denne opgave
03:21
and much of that effort
50
189443
1881
og megen af den indsats
03:23
went into developing the visual
processing apparatus of our brains,
processing apparatus of our brains,
51
191324
5271
gik til at udvikle det visuelle
apparat i vor hjerner
apparat i vor hjerner
03:28
not the eyes themselves.
52
196595
2647
og ikke selve øjnene.
03:31
So vision begins with the eyes,
53
199242
2747
Vision starter med øjnene
03:33
but it truly takes place in the brain.
54
201989
3518
men foregår virkelig i hjernen.
03:38
So for 15 years now, starting
from my Ph.D. at Caltech
from my Ph.D. at Caltech
55
206287
5060
I 15 år nu, fra min Ph.D. i Caltech
03:43
and then leading Stanford's Vision Lab,
56
211347
2926
og bagefter da jeg førte
Stanford´s Vision Lab,
Stanford´s Vision Lab,
03:46
I've been working with my mentors,
collaborators and students
collaborators and students
57
214273
4396
har jeg arbejdet med mine mentorer,
kolleger og elever
kolleger og elever
03:50
to teach computers to see.
58
218669
2889
med at lære computere at se.
03:54
Our research field is called
computer vision and machine learning.
computer vision and machine learning.
59
222658
3294
Vort forskningsfelt kaldes
computer vision og maskinlærdom.
computer vision og maskinlærdom.
03:57
It's part of the general field
of artificial intelligence.
of artificial intelligence.
60
225952
3878
Det er en del af det generelle felt
af kunstig intelligens.
af kunstig intelligens.
04:03
So ultimately, we want to teach
the machines to see just like we do:
the machines to see just like we do:
61
231000
5493
Vi vil lære maskiner at se ligesom os:
04:08
naming objects, identifying people,
inferring 3D geometry of things,
inferring 3D geometry of things,
62
236493
5387
identificere objekter, mennesker og
udlede 3D geometrien af ting
udlede 3D geometrien af ting
04:13
understanding relations, emotions,
actions and intentions.
actions and intentions.
63
241880
5688
at forstå relationer, følelser
handlinger og intentioner.
handlinger og intentioner.
04:19
You and I weave together entire stories
of people, places and things
of people, places and things
64
247568
6153
Du og jeg udleder hele historier
af folk, steder og ting
af folk, steder og ting
04:25
the moment we lay our gaze on them.
65
253721
2164
i det sekund vi kigger på dem.
04:28
The first step towards this goal
is to teach a computer to see objects,
is to teach a computer to see objects,
66
256955
5583
Første skridt mod dette mål er at
lære computere at se objekter,
lære computere at se objekter,
04:34
the building block of the visual world.
67
262538
3368
byggestenene i den visuelle verden.
04:37
In its simplest terms,
imagine this teaching process
imagine this teaching process
68
265906
4434
Enklest sagt, kan du forestille dig
denne lærdomsproces
denne lærdomsproces
04:42
as showing the computers
some training images
some training images
69
270340
2995
som at vise computere nogle
træningsbilleder
træningsbilleder
04:45
of a particular object, let's say cats,
70
273335
3321
af et bestemt objekt, f.eks. katte
04:48
and designing a model that learns
from these training images.
from these training images.
71
276656
4737
og designe en model der lærer
af disse træningsbilleder.
af disse træningsbilleder.
04:53
How hard can this be?
72
281393
2044
Hvor svært kan det være?
04:55
After all, a cat is just
a collection of shapes and colors,
a collection of shapes and colors,
73
283437
4052
En kat er jo en samling af
former og farver,
former og farver,
04:59
and this is what we did
in the early days of object modeling.
in the early days of object modeling.
74
287489
4086
og dét er hvad vi gjorde i starten
af objekt modelleringen.
af objekt modelleringen.
05:03
We'd tell the computer algorithm
in a mathematical language
in a mathematical language
75
291575
3622
Vi fortalte computer algoritmen i et
matematisk sprog
matematisk sprog
05:07
that a cat has a round face,
a chubby body,
a chubby body,
76
295197
3343
at en kat har et rundt ansigt,
en fyldig krop,
en fyldig krop,
05:10
two pointy ears, and a long tail,
77
298540
2299
to spidse ører og en lang hale
05:12
and that looked all fine.
78
300839
1410
og det virkede fint.
05:14
But what about this cat?
79
302859
2113
Men hvad med denne kat?
05:16
(Laughter)
80
304972
1091
(Latter)
05:18
It's all curled up.
81
306063
1626
Den er helt sammenfoldet.
05:19
Now you have to add another shape
and viewpoint to the object model.
and viewpoint to the object model.
82
307689
4719
Nu bliver du nødt til at tilføje en ny
form og udsigtspunkt til objektet.
form og udsigtspunkt til objektet.
05:24
But what if cats are hidden?
83
312408
1715
Men hvad hvis katten er gemt?
05:27
What about these silly cats?
84
315143
2219
Hvad med disse fjollede katte?
05:31
Now you get my point.
85
319112
2417
Nu ser I problemet.
05:33
Even something as simple
as a household pet
as a household pet
86
321529
3367
Selv noget så simpelt som et husdyr
05:36
can present an infinite number
of variations to the object model,
of variations to the object model,
87
324896
4504
kan lave uendelige variationer
til objekt modelleringen,
til objekt modelleringen,
05:41
and that's just one object.
88
329400
2233
og det er kun ét objekt.
05:44
So about eight years ago,
89
332573
2492
For omkring otte år siden
05:47
a very simple and profound observation
changed my thinking.
changed my thinking.
90
335065
5030
forandrede en meget simpel og dyb
observering min tænkemåde.
observering min tænkemåde.
05:53
No one tells a child how to see,
91
341425
2685
Ingen fortæller et barn hvordan det ser,
05:56
especially in the early years.
92
344110
2261
særlig i de tidligste år.
05:58
They learn this through
real-world experiences and examples.
real-world experiences and examples.
93
346371
5000
De lærer dette gennem virkelige
erfaringer og eksempler.
erfaringer og eksempler.
06:03
If you consider a child's eyes
94
351371
2740
Hvis man ser et barns øjne
06:06
as a pair of biological cameras,
95
354111
2554
som et par biologiske kameraer
06:08
they take one picture
about every 200 milliseconds,
about every 200 milliseconds,
96
356665
4180
tager de et billede ca. hver
200 millisekunder,
200 millisekunder,
06:12
the average time an eye movement is made.
97
360845
3134
gennemsnitstiden for hver bevægelse
i øjnene.
i øjnene.
06:15
So by age three, a child would have seen
hundreds of millions of pictures
hundreds of millions of pictures
98
363979
5550
Ved tre-årstiden vil et barn have set
hundrede mio. billeder
hundrede mio. billeder
06:21
of the real world.
99
369529
1834
af den virkelige verden.
06:23
That's a lot of training examples.
100
371363
2280
Det er mange træningseksempler.
06:26
So instead of focusing solely
on better and better algorithms,
on better and better algorithms,
101
374383
5989
Istedet for kun at fokusere på
bedre og bedre algoritmer
bedre og bedre algoritmer
06:32
my insight was to give the algorithms
the kind of training data
the kind of training data
102
380372
5272
var min indsigt at give algoritmerne
den slags træningsdata
den slags træningsdata
06:37
that a child was given through experiences
103
385644
3319
som et barn får gennem erfaringer
06:40
in both quantity and quality.
104
388963
3878
både i kvalitet og kvantitet.
06:44
Once we know this,
105
392841
1858
Når vi ved dette,
06:46
we knew we needed to collect a data set
106
394699
2971
kan vi begynde at samle et data sæt
06:49
that has far more images
than we have ever had before,
than we have ever had before,
107
397670
4459
som har langt flere billeder
end vi før har haft,
end vi før har haft,
06:54
perhaps thousands of times more,
108
402129
2577
måske tusindvis af gange mere,
06:56
and together with Professor
Kai Li at Princeton University,
Kai Li at Princeton University,
109
404706
4111
og sammen med professor Kai Li
i Princeton University
i Princeton University
07:00
we launched the ImageNet project in 2007.
110
408817
4752
startede vi ImageNet projektet i 2007.
07:05
Luckily, we didn't have to mount
a camera on our head
a camera on our head
111
413569
3838
Heldigvis behøvede vi ikke at sætte et
kamera på hovedet
kamera på hovedet
07:09
and wait for many years.
112
417407
1764
og vente i mange år.
07:11
We went to the Internet,
113
419171
1463
Vi tog på Internettet,
07:12
the biggest treasure trove of pictures
that humans have ever created.
that humans have ever created.
114
420634
4436
den største samling af billeder som
mennesket nogensinde har lavet.
mennesket nogensinde har lavet.
07:17
We downloaded nearly a billion images
115
425070
3041
Vi downloadede næsten en milliard billeder
07:20
and used crowdsourcing technology
like the Amazon Mechanical Turk platform
like the Amazon Mechanical Turk platform
116
428111
5880
og brugte crowdsource teknologier såsom
Amazon Mechanical Turk platformen
Amazon Mechanical Turk platformen
07:25
to help us to label these images.
117
433991
2339
til at hjælpe med at sortere billederne.
07:28
At its peak, ImageNet was one of
the biggest employers
the biggest employers
118
436330
4900
På sit højeste var ImageNet en af de
største arbejdsgivere
største arbejdsgivere
07:33
of the Amazon Mechanical Turk workers:
119
441230
2996
af alle Amazon Mechanical Turk ansatte:
07:36
together, almost 50,000 workers
120
444226
3854
tilsammen hjalp næsten 50,000 ansatte
07:40
from 167 countries around the world
121
448080
4040
fra 167 lande rundt omkring i verden
07:44
helped us to clean, sort and label
122
452120
3947
os med at rense, sortere og rubricere
07:48
nearly a billion candidate images.
123
456067
3575
næsten en milliard kandidat-billeder.
07:52
That was how much effort it took
124
460612
2653
Så megen indsats tog det
07:55
to capture even a fraction
of the imagery
of the imagery
125
463265
3900
at fange selv en brøkdel
af alle de billeder
af alle de billeder
07:59
a child's mind takes in
in the early developmental years.
in the early developmental years.
126
467165
4171
som et barns hjerne bearbejder
i de første leveår.
i de første leveår.
08:04
In hindsight, this idea of using big data
127
472148
3902
Set i bakspejlet er denne idé med at bruge
massive data
massive data
08:08
to train computer algorithms
may seem obvious now,
may seem obvious now,
128
476050
4550
til at træne computer algoritmer
måske indlysende,
måske indlysende,
08:12
but back in 2007, it was not so obvious.
129
480600
4110
men i 2007 var det ikke så enkelt.
08:16
We were fairly alone on this journey
for quite a while.
for quite a while.
130
484710
3878
Vi var ret alene på denne rejse
i et godt stykke tid.
i et godt stykke tid.
08:20
Some very friendly colleagues advised me
to do something more useful for my tenure,
to do something more useful for my tenure,
131
488588
5003
Nogle venlige kolleger gav mig det råd at
bruge arbejdstiden mere effektivt
bruge arbejdstiden mere effektivt
08:25
and we were constantly struggling
for research funding.
for research funding.
132
493591
4342
og vi arbejdede hårdt for at skaffe
forskningsmidler.
forskningsmidler.
08:29
Once, I even joked to my graduate students
133
497933
2485
Jeg jokede også med mine
kandidatstuderende
kandidatstuderende
08:32
that I would just reopen
my dry cleaner's shop to fund ImageNet.
my dry cleaner's shop to fund ImageNet.
134
500418
4063
at jeg ville genåbne mit renseri
for at skaffe penge til ImageNet.
for at skaffe penge til ImageNet.
08:36
After all, that's how I funded
my college years.
my college years.
135
504481
4761
Det er jo hvordan jeg finansierede
mine universitetsår.
mine universitetsår.
08:41
So we carried on.
136
509242
1856
Så vi fortsatte.
08:43
In 2009, the ImageNet project delivered
137
511098
3715
I 2009 leverede ImageNet projektet
08:46
a database of 15 million images
138
514813
4042
en database på 15 millioner billeder
08:50
across 22,000 classes
of objects and things
of objects and things
139
518855
4805
bestående af 22,000 klasser
af objekter og ting
af objekter og ting
08:55
organized by everyday English words.
140
523660
3320
organiseret af engelske dagligdags-ord.
08:58
In both quantity and quality,
141
526980
2926
I både kvalitet og kvantitet
09:01
this was an unprecedented scale.
142
529906
2972
var dette aldrig set før.
09:04
As an example, in the case of cats,
143
532878
3461
F.eks. i eksemplet med katte
09:08
we have more than 62,000 cats
144
536339
2809
har vi nu mere end 62,000 katte
09:11
of all kinds of looks and poses
145
539148
4110
af alle slags udseender og positioner
09:15
and across all species
of domestic and wild cats.
of domestic and wild cats.
146
543258
5223
og af alle racer af både hus-
og vilde katte.
og vilde katte.
09:20
We were thrilled
to have put together ImageNet,
to have put together ImageNet,
147
548481
3344
Vi var utrolig glade for at have
sammensat ImageNet,
sammensat ImageNet,
09:23
and we wanted the whole research world
to benefit from it,
to benefit from it,
148
551825
3738
og vi ville, at hele forskningsverdenen
ville få gavn af dette,
ville få gavn af dette,
09:27
so in the TED fashion,
we opened up the entire data set
we opened up the entire data set
149
555563
4041
så på ren TED-manér åbnede vi
for hele data-sættet
for hele data-sættet
09:31
to the worldwide
research community for free.
research community for free.
150
559604
3592
gratis til hele forsknings-fællesskabet.
09:36
(Applause)
151
564636
4000
(Bifald)
09:41
Now that we have the data
to nourish our computer brain,
to nourish our computer brain,
152
569416
4538
Nu, da vi har data til at nære
vor computer-hjerne
vor computer-hjerne
09:45
we're ready to come back
to the algorithms themselves.
to the algorithms themselves.
153
573954
3737
er vi klar til at tage fat på
selve algoritmerne.
selve algoritmerne.
09:49
As it turned out, the wealth
of information provided by ImageNet
of information provided by ImageNet
154
577691
5178
Det viste sig at overfloden af information
som ImageNet gav
som ImageNet gav
09:54
was a perfect match to a particular class
of machine learning algorithms
of machine learning algorithms
155
582869
4806
passede perfekt til en særlig klasse af
maskinlærings-algoritmer
maskinlærings-algoritmer
09:59
called convolutional neural network,
156
587675
2415
kaldet sammenfoldede neurale netværk,
10:02
pioneered by Kunihiko Fukushima,
Geoff Hinton, and Yann LeCun
Geoff Hinton, and Yann LeCun
157
590090
5248
udviklet af Kunihiko Fukushima,
Geoff Hinton og Yann LeCun
Geoff Hinton og Yann LeCun
10:07
back in the 1970s and '80s.
158
595338
3645
i 1970´erne og 80´erne.
10:10
Just like the brain consists
of billions of highly connected neurons,
of billions of highly connected neurons,
159
598983
5619
Ligesom hjernen består af milliarder af
sammenkædede neuroner,
sammenkædede neuroner,
10:16
a basic operating unit in a neural network
160
604602
3854
er et grundlæggende armatur i et
neuralt netværk
neuralt netværk
10:20
is a neuron-like node.
161
608456
2415
som en neuron-lignende node.
10:22
It takes input from other nodes
162
610871
2554
Det tager input fra andre noder
10:25
and sends output to others.
163
613425
2718
og sender output til andre.
10:28
Moreover, these hundreds of thousands
or even millions of nodes
or even millions of nodes
164
616143
4713
Disse hundrede tusinder eller endda
millioner af noder
millioner af noder
10:32
are organized in hierarchical layers,
165
620856
3227
er organiseret i hierarkiske lag,
10:36
also similar to the brain.
166
624083
2554
ligesom i hjernen.
10:38
In a typical neural network we use
to train our object recognition model,
to train our object recognition model,
167
626637
4783
I et typisk neuralt netværk vi bruger
til at træne vor objekt-genkendelsesmodel,
til at træne vor objekt-genkendelsesmodel,
10:43
it has 24 million nodes,
168
631420
3181
er der 24 millioner knuder,
10:46
140 million parameters,
169
634601
3297
140 millioner parametre,
10:49
and 15 billion connections.
170
637898
2763
og 15 milliarder tilslutninger.
10:52
That's an enormous model.
171
640661
2415
Det er en enorm model.
10:55
Powered by the massive data from ImageNet
172
643076
3901
Drevet af de massive data fra ImageNet,
10:58
and the modern CPUs and GPUs
to train such a humongous model,
to train such a humongous model,
173
646977
5433
og de moderne CPU´er og GPU´er brugt til
at træne en så stor model,
at træne en så stor model,
11:04
the convolutional neural network
174
652410
2369
begyndte det sammenfoldede neurale netværk
11:06
blossomed in a way that no one expected.
175
654779
3436
at blomstre på en måde ingen forventede.
11:10
It became the winning architecture
176
658215
2508
Det blev den førende arkitektur
11:12
to generate exciting new results
in object recognition.
in object recognition.
177
660723
5340
til at genere spændende nye resultater
i objekt-genkendelse.
i objekt-genkendelse.
11:18
This is a computer telling us
178
666063
2810
Dét her er en computer, der fortæller os
11:20
this picture contains a cat
179
668873
2300
at dette billede indeholder en kat
11:23
and where the cat is.
180
671173
1903
og hvor katten er.
11:25
Of course there are more things than cats,
181
673076
2112
Der er selvfølgelig andre ting end katte
11:27
so here's a computer algorithm telling us
182
675188
2438
så her er en computer algoritme
der fortæller
der fortæller
11:29
the picture contains
a boy and a teddy bear;
a boy and a teddy bear;
183
677626
3274
at dette billede indeholder en
dreng og en bamse;
dreng og en bamse;
11:32
a dog, a person, and a small kite
in the background;
in the background;
184
680900
4366
en hund, en person og en
lille drage i baggrunden;
lille drage i baggrunden;
11:37
or a picture of very busy things
185
685266
3135
eller et billede af meget travle ting
11:40
like a man, a skateboard,
railings, a lampost, and so on.
railings, a lampost, and so on.
186
688401
4644
som en mand, et skateboard, rækværk,
en lygtepæl, og så videre.
en lygtepæl, og så videre.
11:45
Sometimes, when the computer
is not so confident about what it sees,
is not so confident about what it sees,
187
693045
5293
Sommetider, når computeren ikke er så
sikker på hvad den ser,
sikker på hvad den ser,
11:51
we have taught it to be smart enough
188
699498
2276
har den lært at være klog nok
11:53
to give us a safe answer
instead of committing too much,
instead of committing too much,
189
701774
3878
til at give os et sikkert svar i stedet
for at være for skråsikker sig,
for at være for skråsikker sig,
11:57
just like we would do,
190
705652
2811
ligesom vi ville,
12:00
but other times our computer algorithm
is remarkable at telling us
is remarkable at telling us
191
708463
4666
men som regel er algoritmen
utrolig god til at fortælle os
utrolig god til at fortælle os
12:05
what exactly the objects are,
192
713129
2253
præcis hvad objekterne er,
12:07
like the make, model, year of the cars.
193
715382
3436
som fremstillingen, modellen og
årgangen af bilerne.
årgangen af bilerne.
12:10
We applied this algorithm to millions
of Google Street View images
of Google Street View images
194
718818
5386
Vi anvendte algoritmerne i mio.
af Google Street View billeder
af Google Street View billeder
12:16
across hundreds of American cities,
195
724204
3135
i hundreder af amerikanske byer,
12:19
and we have learned something
really interesting:
really interesting:
196
727339
2926
og vi har lært noget meget
interessant:
interessant:
12:22
first, it confirmed our common wisdom
197
730265
3320
vi fik bekræftet den gængse viden
12:25
that car prices correlate very well
198
733585
3290
at bilpriser korrelerer meget godt
12:28
with household incomes.
199
736875
2345
med husholdningsindtægter.
12:31
But surprisingly, car prices
also correlate well
also correlate well
200
739220
4527
Men overraskende nok korrelerer
bilpriser godt
bilpriser godt
12:35
with crime rates in cities,
201
743747
2300
med kriminalitet i byer,
12:39
or voting patterns by zip codes.
202
747007
3963
eller stemmemønstre ved postnumre.
12:44
So wait a minute. Is that it?
203
752060
2206
Vent lidt. Er dét alt?
12:46
Has the computer already matched
or even surpassed human capabilities?
or even surpassed human capabilities?
204
754266
5153
Har computeren allerede nået
eller endda overhalet menneskelige evner?
eller endda overhalet menneskelige evner?
12:51
Not so fast.
205
759419
2138
Øjeblik.
12:53
So far, we have just taught
the computer to see objects.
the computer to see objects.
206
761557
4923
Indtil videre har vi kun lært
computeren at se objekter.
computeren at se objekter.
12:58
This is like a small child
learning to utter a few nouns.
learning to utter a few nouns.
207
766480
4644
Det er ligesom et barn, der
lærer at udtale nogle få navneord.
lærer at udtale nogle få navneord.
13:03
It's an incredible accomplishment,
208
771124
2670
Det er en utrolig stor udrettelse,
13:05
but it's only the first step.
209
773794
2460
men det er kun første skridt.
13:08
Soon, another developmental
milestone will be hit,
milestone will be hit,
210
776254
3762
Snart vil en ny udviklingsmæssig
milepæl blive nået,
milepæl blive nået,
13:12
and children begin
to communicate in sentences.
to communicate in sentences.
211
780016
3461
og barnet begynder at
kommunikere i sætninger.
kommunikere i sætninger.
13:15
So instead of saying
this is a cat in the picture,
this is a cat in the picture,
212
783477
4224
Istedet for at sige at dette
er en kat på billedet,
er en kat på billedet,
13:19
you already heard the little girl
telling us this is a cat lying on a bed.
telling us this is a cat lying on a bed.
213
787701
5202
har vi allerede hørt pigen sige, at dette
er en kat der ligger på en seng.
er en kat der ligger på en seng.
13:24
So to teach a computer
to see a picture and generate sentences,
to see a picture and generate sentences,
214
792903
5595
For at lære en computer at se et billede
og lave sætninger,
og lave sætninger,
13:30
the marriage between big data
and machine learning algorithm
and machine learning algorithm
215
798498
3948
må sammenføringen af massive data og
maskinlærings-algoritmer
maskinlærings-algoritmer
13:34
has to take another step.
216
802446
2275
tages til et nyt niveau.
13:36
Now, the computer has to learn
from both pictures
from both pictures
217
804721
4156
Nu må computeren lære fra både billeder
13:40
as well as natural language sentences
218
808877
2856
og naturlige sprog-sætninger
13:43
generated by humans.
219
811733
3322
fra mennesker.
13:47
Just like the brain integrates
vision and language,
vision and language,
220
815055
3853
Ligesom hjernen integrerer vision og sprog
13:50
we developed a model
that connects parts of visual things
that connects parts of visual things
221
818908
5201
har vi udviklet en model, der
sammenfører dele af visuelle ting,
sammenfører dele af visuelle ting,
13:56
like visual snippets
222
824109
1904
eller visuelle klip,
13:58
with words and phrases in sentences.
223
826013
4203
med ord og fraser i sætninger.
14:02
About four months ago,
224
830216
2763
For fire måneder siden
14:04
we finally tied all this together
225
832979
2647
fik vi endelig sammensat alt dette
14:07
and produced one of the first
computer vision models
computer vision models
226
835626
3784
og producerede en af de første
computervision modeller
computervision modeller
14:11
that is capable of generating
a human-like sentence
a human-like sentence
227
839410
3994
der er i stand til at lave
en menneskelignende sætning
en menneskelignende sætning
14:15
when it sees a picture for the first time.
228
843404
3506
når den ser et billede for første gang.
14:18
Now, I'm ready to show you
what the computer says
what the computer says
229
846910
4644
Nu er jeg klar til at vise jer
hvad computeren siger,
hvad computeren siger,
14:23
when it sees the picture
230
851554
1975
når den ser billedet af
14:25
that the little girl saw
at the beginning of this talk.
at the beginning of this talk.
231
853529
3830
hvad den lille pige så
i starten af dette foredrag.
i starten af dette foredrag.
14:31
(Video) Computer: A man is standing
next to an elephant.
next to an elephant.
232
859519
3344
(Video) Computer: En mand står
ved siden af en elefant.
ved siden af en elefant.
14:36
A large airplane sitting on top
of an airport runway.
of an airport runway.
233
864393
3634
Et stort fly står på en landingsbane
i en lufthavn.
i en lufthavn.
14:41
FFL: Of course, we're still working hard
to improve our algorithms,
to improve our algorithms,
234
869057
4212
FFL: Selvfølgelig arbejder vi stadig hårdt
med at forbedre vor algoritmer,
med at forbedre vor algoritmer,
14:45
and it still has a lot to learn.
235
873269
2596
og den har stadig meget at lære.
14:47
(Applause)
236
875865
2291
(Bifald)
14:51
And the computer still makes mistakes.
237
879556
3321
Og computeren laver stadig fejl.
14:54
(Video) Computer: A cat lying
on a bed in a blanket.
on a bed in a blanket.
238
882877
3391
(Video) Computer: En kat ligger i et tæppe
på en seng.
på en seng.
14:58
FFL: So of course, when it sees
too many cats,
too many cats,
239
886268
2553
FFL: Så selvfølgelig, når den ser
for mange katte,
for mange katte,
15:00
it thinks everything
might look like a cat.
might look like a cat.
240
888821
2926
tror den at alt ligner en kat.
15:05
(Video) Computer: A young boy
is holding a baseball bat.
is holding a baseball bat.
241
893317
2864
(Video) Comp.: En lille dreng
holder et baseball bat.
holder et baseball bat.
15:08
(Laughter)
242
896181
1765
(Latter)
15:09
FFL: Or, if it hasn't seen a toothbrush,
it confuses it with a baseball bat.
it confuses it with a baseball bat.
243
897946
4583
FFL: Eller hvis den ikke har set en tand-
børste før, tror den at det er et bat.
børste før, tror den at det er et bat.
15:15
(Video) Computer: A man riding a horse
down a street next to a building.
down a street next to a building.
244
903309
3434
(Video) Comp.: En mand rider en hest
på en vej ved siden af en bygning.
på en vej ved siden af en bygning.
15:18
(Laughter)
245
906743
2023
(Latter)
15:20
FFL: We haven't taught Art 101
to the computers.
to the computers.
246
908766
3552
FFL: Vi har ikke lært computere kunst.
15:25
(Video) Computer: A zebra standing
in a field of grass.
in a field of grass.
247
913768
2884
(Video) Comp.: En zebra står på en eng.
15:28
FFL: And it hasn't learned to appreciate
the stunning beauty of nature
the stunning beauty of nature
248
916652
3367
FFL: Og den har ikke lært at værdsætte
naturens utrolige skønhed
naturens utrolige skønhed
15:32
like you and I do.
249
920019
2438
som du og jeg.
15:34
So it has been a long journey.
250
922457
2832
Så det har været en lang rejse.
15:37
To get from age zero to three was hard.
251
925289
4226
Det var svært at nå treårsalderen.
15:41
The real challenge is to go
from three to 13 and far beyond.
from three to 13 and far beyond.
252
929515
5596
Udfordringen ligger i at nå 13-årsstadiet
og endnu længere frem.
og endnu længere frem.
15:47
Let me remind you with this picture
of the boy and the cake again.
of the boy and the cake again.
253
935111
4365
Lad mig minde jer om billedet af drengen
og kagen igen.
og kagen igen.
15:51
So far, we have taught
the computer to see objects
the computer to see objects
254
939476
4064
Indtil videre har vi lært computere
at se objekter
at se objekter
15:55
or even tell us a simple story
when seeing a picture.
when seeing a picture.
255
943540
4458
eller endda fortælle en simpel historie
når den ser et billede.
når den ser et billede.
15:59
(Video) Computer: A person sitting
at a table with a cake.
at a table with a cake.
256
947998
3576
(Video) Computer.: En person sidder ved
et bord med en kage.
et bord med en kage.
16:03
FFL: But there's so much more
to this picture
to this picture
257
951574
2630
FFL: Men der er så meget mere ved
dette billede
dette billede
16:06
than just a person and a cake.
258
954204
2270
end kun en person og en kage.
16:08
What the computer doesn't see
is that this is a special Italian cake
is that this is a special Italian cake
259
956474
4467
Hvad computeren ikke ser, er at dette
er en speciel italiensk kage
er en speciel italiensk kage
16:12
that's only served during Easter time.
260
960941
3217
som kun bliver serveret ved påske.
16:16
The boy is wearing his favorite t-shirt
261
964158
3205
Drengen er i sin yndlings t-shirt
16:19
given to him as a gift by his father
after a trip to Sydney,
after a trip to Sydney,
262
967363
3970
som hans far gav ham efter en
tur til Sydney,
tur til Sydney,
16:23
and you and I can all tell how happy he is
263
971333
3808
og vi kan se hvor glad han er
16:27
and what's exactly on his mind
at that moment.
at that moment.
264
975141
3203
og præcis hvad han tænker i dette øjeblik.
16:31
This is my son Leo.
265
979214
3125
Dette er min søn Leo.
16:34
On my quest for visual intelligence,
266
982339
2624
På min mission for visuel intelligens
16:36
I think of Leo constantly
267
984963
2391
tænker jeg konstant på Leo
16:39
and the future world he will live in.
268
987354
2903
og fremtidens verden han vil leve i.
16:42
When machines can see,
269
990257
2021
Når maskiner kan se
16:44
doctors and nurses will have
extra pairs of tireless eyes
extra pairs of tireless eyes
270
992278
4712
vil læger og sygeplejersker have
ekstra utrættelige øjne
ekstra utrættelige øjne
16:48
to help them to diagnose
and take care of patients.
and take care of patients.
271
996990
4092
til at hjælpe med at diagnosticere
og pleje patienter.
og pleje patienter.
16:53
Cars will run smarter
and safer on the road.
and safer on the road.
272
1001082
4383
Biler vil være klogere og sikrere
på vejene.
på vejene.
16:57
Robots, not just humans,
273
1005465
2694
Robotter, ikke kun mennesker,
17:00
will help us to brave the disaster zones
to save the trapped and wounded.
to save the trapped and wounded.
274
1008159
4849
vil hjælpe med at redde sårede
mennesker på ulykkessteder.
mennesker på ulykkessteder.
17:05
We will discover new species,
better materials,
better materials,
275
1013798
3796
Vi vil opdage nye arter,
bedre materialer,
bedre materialer,
17:09
and explore unseen frontiers
with the help of the machines.
with the help of the machines.
276
1017594
4509
og udforske nye territorier
ved hjælp af maskiner.
ved hjælp af maskiner.
17:15
Little by little, we're giving sight
to the machines.
to the machines.
277
1023113
4167
Vi lærer maskiner at se med tiden.
17:19
First, we teach them to see.
278
1027280
2798
Først lærer vi dem at se.
17:22
Then, they help us to see better.
279
1030078
2763
Så lærer de os at se bedre.
17:24
For the first time, human eyes
won't be the only ones
won't be the only ones
280
1032841
4165
For første gang vil menneskelige øjne
ikke være de eneste
ikke være de eneste
17:29
pondering and exploring our world.
281
1037006
2934
som iagttager og udforsker vor verden.
17:31
We will not only use the machines
for their intelligence,
for their intelligence,
282
1039940
3460
Vi vil ikke blot bruge maskiner for
deres intelligens,
deres intelligens,
17:35
we will also collaborate with them
in ways that we cannot even imagine.
in ways that we cannot even imagine.
283
1043400
6179
vi vil også samarbejde med dem på
måder vi end ikke kan forestille os.
måder vi end ikke kan forestille os.
17:41
This is my quest:
284
1049579
2161
Dette er min mission:
17:43
to give computers visual intelligence
285
1051740
2712
at give computere visuel intelligens
17:46
and to create a better future
for Leo and for the world.
for Leo and for the world.
286
1054452
5131
og at skabe en bedre fremtid
for Leo og for verden.
for Leo og for verden.
17:51
Thank you.
287
1059583
1811
Tak skal I have.
17:53
(Applause)
288
1061394
3785
(Bifald)
ABOUT THE SPEAKER
Fei-Fei Li - Computer scientistAs Director of Stanford’s Artificial Intelligence Lab and Vision Lab, Fei-Fei Li is working to solve AI’s trickiest problems -- including image recognition, learning and language processing.
Why you should listen
Using algorithms built on machine learning methods such as neural network models, the Stanford Artificial Intelligence Lab led by Fei-Fei Li has created software capable of recognizing scenes in still photographs -- and accurately describe them using natural language.
Li’s work with neural networks and computer vision (with Stanford’s Vision Lab) marks a significant step forward for AI research, and could lead to applications ranging from more intuitive image searches to robots able to make autonomous decisions in unfamiliar situations.
Fei-Fei was honored as one of Foreign Policy's 2015 Global Thinkers.
Fei-Fei Li | Speaker | TED.com