TED2018
Supasorn Suwajanakorn: Fake videos of real people -- and how to spot them
Supasorn Suvajanakorn (Supasorn Suwajanakorn): Lažni snimci pravih ljudi - i kako ih uočiti
Filmed:
Readability: 4.6
1,453,308 views
Da li mislite da ste dobri u uočavanju lažnih video snimaka u kojima poznati ljudi izgovaraju stvari koje nikad nisu izgovorili u stvarnom životu? Pogledajte kako se oni prave u ovom zapanjujućem govoru i tehničkoj demonstraciji. Računarski stručnjak, Supasorn Suvajanakorn, pokazuje kako je, kao diplomac, koristio veštačku inteligenciju i 3D modelovanje da napravi lažne fotorealistične snimke ljudi sinhronizovane sa zvukom. Saznajte više o etičkim implikacijama i kreativnim mogućnostima ove tehnologije - i preduzetim koracima u borbi protiv njene zloupotrebe.
Supasorn Suwajanakorn - Computer scientist
Supasorn Suwajanakorn works on ways to reconstruct, preserve and reanimate anyone -- just from their existing photos and videos. Full bio
Supasorn Suwajanakorn works on ways to reconstruct, preserve and reanimate anyone -- just from their existing photos and videos. Full bio
Double-click the English transcript below to play the video.
00:12
Look at these images.
0
876
1151
Pogledajte ove slike.
00:14
Now, tell me which Obama here is real.
1
2051
2635
A sad, recite mi koji Obama je pravi.
(Video) Barak Obama: Da pomognemo
porodicama s refinansiranjem domova,
porodicama s refinansiranjem domova,
00:16
(Video) Barack Obama: To help families
refinance their homes,
refinance their homes,
2
4710
2861
00:19
to invest in things
like high-tech manufacturing,
like high-tech manufacturing,
3
7595
2647
da uložimo u stvari kao što su
visoko-tehnološka proizvodnja,
visoko-tehnološka proizvodnja,
00:22
clean energy
4
10266
1159
čista energija i infrastruktura
koja stvara dobre nove poslove.
koja stvara dobre nove poslove.
00:23
and the infrastructure
that creates good new jobs.
that creates good new jobs.
5
11449
2779
00:26
Supasorn Suwajanakorn: Anyone?
6
14647
1484
Supasorn Suvanjakorn: Bilo ko?
00:28
The answer is none of them.
7
16155
1874
Odgovor je: nijedan od njih.
00:30
(Laughter)
8
18053
1114
(Smeh)
00:31
None of these is actually real.
9
19191
1786
Nijedan od ovih zapravo nije pravi.
Dozvolite da objasnim
kako smo dospeli ovde.
kako smo dospeli ovde.
00:33
So let me tell you how we got here.
10
21001
1840
00:35
My inspiration for this work
11
23940
1578
Moja inspiracija
za ovaj posao bio je projekat
za ovaj posao bio je projekat
00:37
was a project meant to preserve our last
chance for learning about the Holocaust
chance for learning about the Holocaust
12
25542
5411
namenjen očuvanju naše poslednje šanse
za saznanje o holokaustu od preživelih.
za saznanje o holokaustu od preživelih.
00:42
from the survivors.
13
30977
1768
00:44
It's called New Dimensions in Testimony,
14
32769
2627
Zove se „Nove dimenzije u svedočenju“
00:47
and it allows you to have
interactive conversations
interactive conversations
15
35420
3126
i dozvoljava vam da vodite
interaktivne razgovore
interaktivne razgovore
00:50
with a hologram
of a real Holocaust survivor.
of a real Holocaust survivor.
16
38570
2556
sa hologramom pravog
preživelog iz holokausta.
preživelog iz holokausta.
00:53
(Video) Man: How did you
survive the Holocaust?
survive the Holocaust?
17
41793
1966
(Video) Čovek: Kako ste
preživeli holokaust?
preživeli holokaust?
00:55
(Video) Hologram: How did I survive?
18
43783
1668
(Video) Hologram: Kako sam preživeo?
00:57
I survived,
19
45912
1807
Preživeo sam,
01:00
I believe,
20
48419
1527
verujem,
01:01
because providence watched over me.
21
49970
3023
zato što je proviđenje bdelo nada mnom.
01:05
SS: Turns out these answers
were prerecorded in a studio.
were prerecorded in a studio.
22
53573
3454
SS: Ispostavlja se da su ovi odgovori
unapred snimljeni u studiju.
unapred snimljeni u studiju.
01:09
Yet the effect is astounding.
23
57051
2452
Pa ipak, efekat je zapanjujuć.
01:11
You feel so connected to his story
and to him as a person.
and to him as a person.
24
59527
3619
Osećate se privrženo
njegovoj priči i njemu kao osobi.
njegovoj priči i njemu kao osobi.
01:16
I think there's something special
about human interaction
about human interaction
25
64011
3301
Mislim da postoji nešto posebno
u vezi sa ljudskom interakcijom
u vezi sa ljudskom interakcijom
01:19
that makes it much more profound
26
67336
2757
što je čini mnogo dubljom
01:22
and personal
27
70117
2198
i ličnijom
01:24
than what books or lectures
or movies could ever teach us.
or movies could ever teach us.
28
72339
3485
nego što bi knjige, predavanja ili filmovi
ikada mogli da nas nauče.
ikada mogli da nas nauče.
01:28
So I saw this and began to wonder,
29
76267
2425
Dakle, video sam ovo i počeo da se pitam:
01:30
can we create a model
like this for anyone?
like this for anyone?
30
78716
2810
da li možemo da napravimo
ovakav model za bilo koga?
ovakav model za bilo koga?
01:33
A model that looks, talks
and acts just like them?
and acts just like them?
31
81550
2975
Model koji izgleda, priča
i ponaša se baš kao oni?
i ponaša se baš kao oni?
01:37
So I set out to see if this could be done
32
85573
2007
Pokrenuo sam se da vidim
da li je izvodljivo
da li je izvodljivo
01:39
and eventually came up with a new solution
33
87604
2310
i naposletku došao do novog rešenja
01:41
that can build a model of a person
using nothing but these:
using nothing but these:
34
89938
3220
koje može da napravi model osobe
koristeći ništa drugo osim ovoga:
koristeći ništa drugo osim ovoga:
01:45
existing photos and videos of a person.
35
93747
2214
postojećih fotografija i snimaka osobe.
01:48
If you can leverage
this kind of passive information,
this kind of passive information,
36
96701
2617
Ako možete da iskoristite
ovu vrstu pasivnih informacija,
ovu vrstu pasivnih informacija,
01:51
just photos and video that are out there,
37
99342
2007
samo dostupne fotografije i snimke,
to je ključno za primenu na svakoga.
01:53
that's the key to scaling to anyone.
38
101373
2056
01:56
By the way, here's Richard Feynman,
39
104119
1777
Uzgred, evo Ričarda Fajnmana,
01:57
who in addition to being
a Nobel Prize winner in physics
a Nobel Prize winner in physics
40
105920
3413
ko je uz to što je osvojio
Nobelovu nagradu za fiziku,
Nobelovu nagradu za fiziku,
02:01
was also known as a legendary teacher.
41
109357
2453
takođe bio poznat
kao legendaran predavač.
kao legendaran predavač.
02:05
Wouldn't it be great
if we could bring him back
if we could bring him back
42
113080
2198
Zar ne bi bilo sjajno
kada bismo mogli da ga vratimo
kada bismo mogli da ga vratimo
02:07
to give his lectures
and inspire millions of kids,
and inspire millions of kids,
43
115302
3265
da drži svoja predavanja
i inspiriše milione klinaca,
i inspiriše milione klinaca,
02:10
perhaps not just in English
but in any language?
but in any language?
44
118591
2992
možda čak ne samo na engleskom,
već na bilo kom jeziku?
već na bilo kom jeziku?
02:14
Or if you could ask our grandparents
for advice and hear those comforting words
for advice and hear those comforting words
45
122441
4602
Ili ako biste mogli da pitate naše bake
i deke za savet i čujete utešne reči
i deke za savet i čujete utešne reči
02:19
even if they're no longer with us?
46
127067
1770
čak i ako više nisu s nama?
02:21
Or maybe using this tool,
book authors, alive or not,
book authors, alive or not,
47
129683
3396
Ili bi možda korišćenjem ovog alata
pisci knjiga, živi ili ne,
pisci knjiga, živi ili ne,
02:25
could read aloud all of their books
for anyone interested.
for anyone interested.
48
133103
2937
naglas mogli da čitaju
sve svoje knjige za sve zainteresovane.
sve svoje knjige za sve zainteresovane.
02:29
The creative possibilities
here are endless,
here are endless,
49
137199
2437
Kreativne mogućnosti za ovo su beskrajne,
02:31
and to me, that's very exciting.
50
139660
1713
i za mene, to je vrlo uzbudljivo.
02:34
And here's how it's working so far.
51
142595
2002
A evo kako to radi za sad.
02:36
First, we introduce a new technique
52
144621
1667
Prvo, uvodimo novu tehniku
02:38
that can reconstruct a high-detailed
3D face model from any image
3D face model from any image
53
146312
4572
koja može da rekonstruiše
veoma detaljan 3D model lica
veoma detaljan 3D model lica
sa bilo koje slike,
02:42
without ever 3D-scanning the person.
54
150908
2119
bez ikakvog 3D skeniranja osobe.
02:45
And here's the same output model
from different views.
from different views.
55
153890
2642
Evo istog izlaznog modela
iz različitih uglova.
iz različitih uglova.
02:49
This also works on videos,
56
157969
1502
Ovo funkcioniše i na snimcima,
02:51
by running the same algorithm
on each video frame
on each video frame
57
159495
2852
korišćenjem istog algoritma
na svakoj slici snimka
na svakoj slici snimka
02:54
and generating a moving 3D model.
58
162371
2222
i generisanjem pokretnog 3D modela.
02:57
And here's the same
output model from different angles.
output model from different angles.
59
165538
2772
Evo istog izlaznog modela
iz različitih uglova.
iz različitih uglova.
03:01
It turns out this problem
is very challenging,
is very challenging,
60
169933
2534
Ispostavlja se da je ovo
veoma problematično,
veoma problematično,
03:04
but the key trick
is that we are going to analyze
is that we are going to analyze
61
172491
2525
ali ključni trik leži u tome
da ćemo unapred analizirati
veliku zbirku fotografija osobe.
veliku zbirku fotografija osobe.
03:07
a large photo collection
of the person beforehand.
of the person beforehand.
62
175040
2966
03:10
For George W. Bush,
we can just search on Google,
we can just search on Google,
63
178650
2539
Za Džordža V. Buša možemo prosto
da potražimo na Guglu
da potražimo na Guglu
03:14
and from that, we are able
to build an average model,
to build an average model,
64
182309
2499
i iz toga, u stanju smo
da napravimo prosečan model,
da napravimo prosečan model,
03:16
an iterative, refined model
to recover the expression
to recover the expression
65
184832
3111
iterativan, prerađen model
da povratimo izraz u finim detaljima,
kao što su brazde i bore.
kao što su brazde i bore.
03:19
in fine details,
like creases and wrinkles.
like creases and wrinkles.
66
187967
2336
03:23
What's fascinating about this
67
191326
1403
Ono što je fascinantno kod ovoga je
03:24
is that the photo collection
can come from your typical photos.
can come from your typical photos.
68
192753
3423
da zbirka fotografija može nastati
od vaših tipičnih fotografija.
od vaših tipičnih fotografija.
03:28
It doesn't really matter
what expression you're making
what expression you're making
69
196200
2603
Nije stvarno bitno kakav izraz imate
03:30
or where you took those photos.
70
198827
1885
ili gde ste snimili fotografije.
03:32
What matters is
that there are a lot of them.
that there are a lot of them.
71
200736
2400
Ono što je bitno je da ih ima mnogo.
03:35
And we are still missing color here,
72
203160
1736
A ovde nam još uvek fali boja,
pa dalje razvijamo novu tehniku mešanja
03:36
so next, we develop
a new blending technique
a new blending technique
73
204920
2348
03:39
that improves upon
a single averaging method
a single averaging method
74
207292
2836
koja koristi samo jednu
metodu uprosečavanja
metodu uprosečavanja
03:42
and produces sharp
facial textures and colors.
facial textures and colors.
75
210152
2818
i proizvodi jasne teksture lica i boje.
03:45
And this can be done for any expression.
76
213779
2771
A ovo može da se uradi za bilo koji izraz.
03:49
Now we have a control
of a model of a person,
of a model of a person,
77
217485
2499
Sada imamo kontrolu nad modelom osobe,
03:52
and the way it's controlled now
is by a sequence of static photos.
is by a sequence of static photos.
78
220008
3795
a način na koji se sada kontroliše
je preko niza nepokretnih fotografija.
je preko niza nepokretnih fotografija.
03:55
Notice how the wrinkles come and go,
depending on the expression.
depending on the expression.
79
223827
3126
Obratite pažnju na to
kako se bore pojavljuju i nestaju
kako se bore pojavljuju i nestaju
u zavisnosti od izraza.
04:00
We can also use a video
to drive the model.
to drive the model.
80
228109
2746
Možemo da koristimo i snimak
da upravljamo modelom.
da upravljamo modelom.
04:02
(Video) Daniel Craig: Right, but somehow,
81
230879
2593
(Video) Danijel Krejg: Da,
ali nekako smo uspeli
ali nekako smo uspeli
04:05
we've managed to attract
some more amazing people.
some more amazing people.
82
233496
3771
da privučemo neke još neverovatnije ljude.
04:10
SS: And here's another fun demo.
83
238021
1642
SS: A evo još jednog zabavnog demo snimka.
04:11
So what you see here
are controllable models
are controllable models
84
239687
2246
Ovde vidite upravljive modele ljudi,
04:13
of people I built
from their internet photos.
from their internet photos.
85
241957
2444
koje sam napravio
iz njihovih internet fotografija.
iz njihovih internet fotografija.
04:16
Now, if you transfer
the motion from the input video,
the motion from the input video,
86
244425
2904
Sada, ako se prebaci
pokret sa ulaznog snimka
pokret sa ulaznog snimka
04:19
we can actually drive the entire party.
87
247353
2152
zapravo možemo
da upravljamo celom družinom.
da upravljamo celom družinom.
Džordž V. Buš:
„Ovaj nacrt zakona će teško proći
„Ovaj nacrt zakona će teško proći
04:21
George W. Bush:
It's a difficult bill to pass,
It's a difficult bill to pass,
88
249529
2172
04:23
because there's a lot of moving parts,
89
251725
2303
zato što ima mnogo pokretnih delova,
04:26
and the legislative processes can be ugly.
90
254052
5231
a proces zakonodavstva ume da bude gadan.“
04:31
(Applause)
91
259307
1630
(Aplauz)
04:32
SS: So coming back a little bit,
92
260961
1837
SS: Dakle, da se vratim malo unazad,
04:34
our ultimate goal, rather,
is to capture their mannerisms
is to capture their mannerisms
93
262822
3191
naš konačni cilj je radije
da uhvatimo njihovu mimiku
da uhvatimo njihovu mimiku
04:38
or the unique way each
of these people talks and smiles.
of these people talks and smiles.
94
266037
3045
ili jedinstven način na koji svako
od ovih ljudi priča ili se smeši.
od ovih ljudi priča ili se smeši.
04:41
So to do that, can we
actually teach the computer
actually teach the computer
95
269106
2313
Da bismo to uradili,
možemo li zaista naučiti računar
možemo li zaista naučiti računar
04:43
to imitate the way someone talks
96
271443
2222
da podražava način na koji neko govori
04:45
by only showing it
video footage of the person?
video footage of the person?
97
273689
2420
pokazujući mu samo video snimke osobe?
Tako sam pustio računaru da gleda
04:48
And what I did exactly was,
I let a computer watch
I let a computer watch
98
276898
2577
04:51
14 hours of pure Barack Obama
giving addresses.
giving addresses.
99
279499
3277
14 sati isključivo Baraka Obame
i njegovih obraćanja.
i njegovih obraćanja.
04:55
And here's what we can produce
given only his audio.
given only his audio.
100
283443
3516
I evo šta možemo da izvedemo
samo iz njegovog zvučnog snimka.
samo iz njegovog zvučnog snimka.
04:58
(Video) BO: The results are clear.
101
286983
1777
(Video) BO: Rezultati su jasni.
05:00
America's businesses have created
14.5 million new jobs
14.5 million new jobs
102
288784
4349
Američki biznisi su stvorili
14,5 miliona novih poslova
14,5 miliona novih poslova
05:05
over 75 straight months.
103
293157
2774
tokom 75 meseci neprekidno.
05:07
SS: So what's being synthesized here
is only the mouth region,
is only the mouth region,
104
295955
2905
SS: Dakle, ono što se ovde spaja
je samo oblast usta,
je samo oblast usta,
05:10
and here's how we do it.
105
298884
1540
i evo kako to radimo.
05:12
Our pipeline uses a neural network
106
300764
1826
Naš sistem koristi neuronsku mrežu
05:14
to convert and input audio
into these mouth points.
into these mouth points.
107
302614
2936
da pretvori i dovede audio snimak
u ove tačke usta.
u ove tačke usta.
05:18
(Video) BO: We get it through our job
or through Medicare or Medicaid.
or through Medicare or Medicaid.
108
306547
4225
(Video) BO: Dobijamo to preko našeg posla
ili preko zdravstvene zaštite.
ili preko zdravstvene zaštite.
05:22
SS: Then we synthesize the texture,
enhance details and teeth,
enhance details and teeth,
109
310796
3420
SS: Tada stvaramo teksturu,
poboljšavamo detalje i zube
poboljšavamo detalje i zube
05:26
and blend it into the head
and background from a source video.
and background from a source video.
110
314240
3074
i stapamo u glavu i pozadinu
iz izvornog snimka.
iz izvornog snimka.
(Video) BO: Žene mogu
besplatno da se pregledaju
besplatno da se pregledaju
05:29
(Video) BO: Women can get free checkups,
111
317338
1905
05:31
and you can't get charged more
just for being a woman.
just for being a woman.
112
319267
2968
i ne može vam se naplatiti više
samo zato što ste žena.
samo zato što ste žena.
05:34
Young people can stay
on a parent's plan until they turn 26.
on a parent's plan until they turn 26.
113
322973
3306
Mladi mogu da ostanu
u roditeljskom programu do 26. godine.
u roditeljskom programu do 26. godine.
05:39
SS: I think these results
seem very realistic and intriguing,
seem very realistic and intriguing,
114
327267
2952
SS: Mislim da ovi rezultati
izgledaju vrlo realistično i interesantno,
izgledaju vrlo realistično i interesantno,
05:42
but at the same time
frightening, even to me.
frightening, even to me.
115
330243
3173
ali istovremeno zastrašujuće,
čak i za mene.
čak i za mene.
05:45
Our goal was to build an accurate model
of a person, not to misrepresent them.
of a person, not to misrepresent them.
116
333440
4015
Naš cilj je bio da napravimo
precizan model osobe,
precizan model osobe,
ne da je pogrešno predstavimo.
05:49
But one thing that concerns me
is its potential for misuse.
is its potential for misuse.
117
337956
3111
Ali, ono što me brine
je potencijal za zloupotrebu.
je potencijal za zloupotrebu.
05:53
People have been thinking
about this problem for a long time,
about this problem for a long time,
118
341958
2971
Ljudi su dugo razmišljali o ovom problemu,
još otkad je „Fotošop“ izašao na tržište.
05:56
since the days when Photoshop
first hit the market.
first hit the market.
119
344953
2381
05:59
As a researcher, I'm also working
on countermeasure technology,
on countermeasure technology,
120
347862
3801
Kao istraživač, takođe radim
na tehnologiji za protivmeru
na tehnologiji za protivmeru
06:03
and I'm part of an ongoing
effort at AI Foundation,
effort at AI Foundation,
121
351687
2942
i učestvujem u tekućem poduhvatu
u fondaciji za veštačku inteligenciju
u fondaciji za veštačku inteligenciju
06:06
which uses a combination
of machine learning and human moderators
of machine learning and human moderators
122
354653
3397
gde se koristi kombinacija
mašinskog učenja i ljudskih moderatora
mašinskog učenja i ljudskih moderatora
06:10
to detect fake images and videos,
123
358074
2144
da se otkriju lažne slike i video-snimci,
06:12
fighting against my own work.
124
360242
1514
boreći se protiv sopstvenog posla.
06:14
And one of the tools we plan to release
is called Reality Defender,
is called Reality Defender,
125
362675
3190
Jedan od alata koji planiramo da izdamo
se zove „Branilac stvarnosti“ -
se zove „Branilac stvarnosti“ -
06:17
which is a web-browser plug-in
that can flag potentially fake content
that can flag potentially fake content
126
365889
4039
dodatak za internet pretraživače koji može
da označi potencijalno lažne sadržaje
da označi potencijalno lažne sadržaje
06:21
automatically, right in the browser.
127
369952
2533
automatski, direktno u samom pretraživaču.
06:24
(Applause)
128
372509
4228
(Aplauz)
06:28
Despite all this, though,
129
376761
1453
Uprkos svemu ovome,
06:30
fake videos could do a lot of damage,
130
378238
1840
lažni video-snimci mogu
da učine dosta štete
da učine dosta štete
06:32
even before anyone has a chance to verify,
131
380102
3294
čak i pre nego što iko
stigne da ih proveri,
stigne da ih proveri,
06:35
so it's very important
that we make everyone aware
that we make everyone aware
132
383420
2722
tako da je vrlo važno da svakoga osvestimo
06:38
of what's currently possible
133
386166
2007
o tome šta je trenutno moguće,
06:40
so we can have the right assumption
and be critical about what we see.
and be critical about what we see.
134
388197
3369
tako da možemo pravilno da pretpostavimo
i da posmatramo kritički.
i da posmatramo kritički.
06:44
There's still a long way to go before
we can fully model individual people
we can fully model individual people
135
392423
5007
I dalje smo daleko od toga
da možemo potpuno da modelujemo pojedince
da možemo potpuno da modelujemo pojedince
06:49
and before we can ensure
the safety of this technology.
the safety of this technology.
136
397454
2786
i pre nego što osiguramo
bezbednost ove tehnologije.
bezbednost ove tehnologije.
06:53
But I'm excited and hopeful,
137
401097
1587
Ali ja sam uzbuđen i pun nade,
06:54
because if we use it right and carefully,
138
402708
3539
jer ako ga koristimo
na pravi način i pažljivo,
na pravi način i pažljivo,
ovaj alat može da omogući da pozitivni
uticaj ma kog pojedinca na svet
uticaj ma kog pojedinca na svet
06:58
this tool can allow any individual's
positive impact on the world
positive impact on the world
139
406271
4309
07:02
to be massively scaled
140
410604
2190
bude mnogostruko uveličan
07:04
and really help shape our future
the way we want it to be.
the way we want it to be.
141
412818
2742
i zaista pomogne da oblikujemo
budućnost u onakvu kakvu želimo.
budućnost u onakvu kakvu želimo.
07:07
Thank you.
142
415584
1151
Hvala vam.
07:08
(Applause)
143
416759
5090
(Aplauz)
ABOUT THE SPEAKER
Supasorn Suwajanakorn - Computer scientistSupasorn Suwajanakorn works on ways to reconstruct, preserve and reanimate anyone -- just from their existing photos and videos.
Why you should listen
Can we create a digital avatar that looks, acts and talks just like our sweet grandma? This question has inspired Supasorn Suwajanakorn, a recent PhD graduate from the University of Washington, to spend years developing new tools to make it a reality. He has developed a set of algorithms that can build a moving 3D face model of anyone from just photos, which was awarded the Innovation of the Year in 2016. He then introduced the first system that can replicate a person's speech and produce a realistic CG-animation by only analyzing their existing video footage -- all without ever bringing in the person to a Hollywood capture studio.
Suwajanakorn is working in the field of machine learning and computer vision. His goal is to bring vision algorithms out of the lab and make them work in the wild.
Supasorn Suwajanakorn | Speaker | TED.com