API Documentation

       
                                                                                                        
                                                                                                                                                                                                                            WEBVTT

00:00:00.359 --> 00:00:03.879
this is Young Simon cole performing in

00:00:03.879 --> 00:00:05.799
America's Got Talent in front of Simon

00:00:05.799 --> 00:00:08.500
cole because this is deep fake and this

00:00:08.500 --> 00:00:09.000
is a deep fake of Ukraine's president

00:00:09.000 --> 00:00:12.940
falada mazelinski asking his soldiers to

00:00:12.940 --> 00:00:13.200
surrender

00:00:13.200 --> 00:00:15.420
one is entertaining while the other is

00:00:15.420 --> 00:00:18.180
alarming about the dangers of deep fake

00:00:18.180 --> 00:00:22.539
these are the two faces of deep fake

00:00:22.539 --> 00:00:24.000
due to the term fake the phrase deep

00:00:24.000 --> 00:00:27.760
fake evokes a negative feeling

00:00:27.760 --> 00:00:29.619
companies like metaphysic are doing

00:00:29.619 --> 00:00:30.180
everything they can to dispel the

00:00:30.180 --> 00:00:32.040
misconceptions about deep fakes in the

00:00:32.040 --> 00:00:33.180
public eye

00:00:33.180 --> 00:00:35.219
this is what they did on America's Got

00:00:35.219 --> 00:00:36.420
turns

00:00:36.420 --> 00:00:38.100
there are many practical applications

00:00:38.100 --> 00:00:40.320
for deep fakes across a wide range of

00:00:40.320 --> 00:00:43.140
sectors for instance news outlets May

00:00:43.140 --> 00:00:46.079
publish news 24 hours a day seven days a

00:00:46.079 --> 00:00:48.820
week without the actual reporter present

00:00:48.820 --> 00:00:50.739
by using a deep fake of their own

00:00:50.739 --> 00:00:51.879
reporter

00:00:51.879 --> 00:00:53.039
they may reduce studio and labor costs

00:00:53.039 --> 00:00:55.260
saving the business a great deal of time

00:00:55.260 --> 00:00:57.579
and money

00:00:57.579 --> 00:00:60.699
the news anchor Kim Juha from the Korean

00:00:60.699 --> 00:01:02.399
television network mbn was deep faked

00:01:02.399 --> 00:01:04.780
for viewers

00:01:04.780 --> 00:01:06.119
the news reader would be a fake as the

00:01:06.119 --> 00:01:08.920
channel had previously informed viewers

00:01:08.920 --> 00:01:11.619
and Kim Joo has still employed

00:01:11.619 --> 00:01:14.560
however mbn stated that it intended to

00:01:14.560 --> 00:01:15.000
keep using the deet fake for some

00:01:15.000 --> 00:01:17.619
breaking news reports

00:01:17.619 --> 00:01:19.960
and deep frame a the South Korean

00:01:19.960 --> 00:01:21.420
company that created the Deep fake has

00:01:21.420 --> 00:01:23.860
stated that it is looking for media

00:01:23.860 --> 00:01:26.560
buyers in China and the US which has

00:01:26.560 --> 00:01:27.360
some people worried that news readers

00:01:27.360 --> 00:01:29.979
may become obsolete

00:01:29.979 --> 00:01:31.380
a tick tock account solely dedicated to

00:01:31.380 --> 00:01:34.540
Tom Cruise deep fakes has been created

00:01:34.540 --> 00:01:36.640
as a result of how far deep fakes have

00:01:36.640 --> 00:01:38.619
advanced in recent years

00:01:38.619 --> 00:01:39.479
videos of Crews doing everything

00:01:39.479 --> 00:01:41.280
including washing his hands and playing

00:01:41.280 --> 00:01:44.740
golf as well as doing magic tricks are

00:01:44.740 --> 00:01:45.360
available online

00:01:45.360 --> 00:01:48.700
the same firm that produced a convincing

00:01:48.700 --> 00:01:50.860
imitation of Simon Cowell metaphysic

00:01:50.860 --> 00:01:51.119
make this

00:01:51.119 --> 00:01:53.860
there are many potential uses of deep

00:01:53.860 --> 00:01:55.780
fakes in the film industry

00:01:55.780 --> 00:01:56.460
you may bring a deceased actor or

00:01:56.460 --> 00:01:58.780
actress back

00:01:58.780 --> 00:01:59.399
it is debatable for an ethical

00:01:59.399 --> 00:02:01.540
standpoint

00:02:01.540 --> 00:02:03.939
however if we then consider ethics it's

00:02:03.939 --> 00:02:05.860
quite doable and simple

00:02:05.860 --> 00:02:07.380
also it is probably far less expensive

00:02:07.380 --> 00:02:09.479
than alternative choices

00:02:09.479 --> 00:02:12.819
deep fakes enabled translated films that

00:02:12.819 --> 00:02:14.980
feature the original actors since they

00:02:14.980 --> 00:02:17.160
can modify footage and duplicate voices

00:02:17.160 --> 00:02:19.140
the voices resemble the original cast

00:02:19.140 --> 00:02:20.220
members

00:02:20.220 --> 00:02:22.440
and critically the lip motions even

00:02:22.440 --> 00:02:25.660
match the words at it with the help of

00:02:25.660 --> 00:02:27.060
positive deep fakes we can thus better

00:02:27.060 --> 00:02:30.180
exchange ideas movies and other creative

00:02:30.180 --> 00:02:33.940
works on a global scale even those with

00:02:33.940 --> 00:02:35.680
tighter finances

00:02:35.680 --> 00:02:37.599
this should increase the variety of

00:02:37.599 --> 00:02:39.000
entertainment and content we consume

00:02:39.000 --> 00:02:42.519
in other words deep fakes can overcome

00:02:42.519 --> 00:02:43.379
language barriers and increase the

00:02:43.379 --> 00:02:46.660
accessibility of content

00:02:46.660 --> 00:02:49.660
deep fakes aren't just found in videos

00:02:49.660 --> 00:02:51.940
a rapidly expanding field with a vast

00:02:51.940 --> 00:02:54.180
array of uses is deep fake audio

00:02:54.180 --> 00:02:57.580
deep learning algorithms can now create

00:02:57.580 --> 00:02:59.099
realistic audio deep fakes with just a

00:02:59.099 --> 00:03:01.780
few hours of audio of the person whose

00:03:01.780 --> 00:03:03.819
voice has been replicated

00:03:03.819 --> 00:03:05.459
once a model of a voice is created that

00:03:05.459 --> 00:03:08.800
person can be made to say anything

00:03:08.800 --> 00:03:10.080
thanks to deep fake audio programmers

00:03:10.080 --> 00:03:12.060
may now allow gamer characters to say

00:03:12.060 --> 00:03:14.400
anything in game rather than depending

00:03:14.400 --> 00:03:16.260
on a small number of lines that were

00:03:16.260 --> 00:03:19.140
produced before the game was released

00:03:19.140 --> 00:03:21.239
some people use this technology to

00:03:21.239 --> 00:03:24.519
commit financial fraud data breaches

00:03:24.519 --> 00:03:26.159
phishing scams automated misinformation

00:03:26.159 --> 00:03:29.680
attacks and the spread of misleading

00:03:29.680 --> 00:03:30.180
information from sources that are

00:03:30.180 --> 00:03:32.920
normally trustworthy

00:03:32.920 --> 00:03:34.959
critics also point out the risks of

00:03:34.959 --> 00:03:36.819
utilizing deed fakes to further a

00:03:36.819 --> 00:03:38.099
political agenda such as by creating

00:03:38.099 --> 00:03:40.900
credible fake news

00:03:40.900 --> 00:03:42.060
politicians have also employed deep fake

00:03:42.060 --> 00:03:43.260
videos

00:03:43.260 --> 00:03:46.319
in 2018 a Belgian political party made

00:03:46.319 --> 00:03:48.879
public a video showing Donald Trump

00:03:48.879 --> 00:03:50.099
pressing Belgium to leave the Paris

00:03:50.099 --> 00:03:52.260
climate Accord in a speech

00:03:52.260 --> 00:03:56.739
Trump never gave that speech however it

00:03:56.739 --> 00:03:57.239
was a deep fake

00:03:57.239 --> 00:03:60.519
it can be used to scam people by using

00:03:60.519 --> 00:04:02.560
the identities of popular people

00:04:02.560 --> 00:04:04.540
and there is enough data on social media

00:04:04.540 --> 00:04:06.640
to identify a certain influencer or

00:04:06.640 --> 00:04:07.959
celebrity

00:04:07.959 --> 00:04:09.700
there are thousands of photos and videos

00:04:09.700 --> 00:04:11.159
online which can be used as training

00:04:11.159 --> 00:04:13.200
data to create frauds in the name of

00:04:13.200 --> 00:04:15.239
authentic people this is some of the

00:04:15.239 --> 00:04:17.400
reasons why the Ali Community is afraid

00:04:17.400 --> 00:04:19.199
of making deep faking available to

00:04:19.199 --> 00:04:21.760
everyone regardless of their background

00:04:21.760 --> 00:04:24.060
just like AI image generators

00:04:24.060 --> 00:04:27.300
just like deep fakes AI image generators

00:04:27.300 --> 00:04:29.220
can be used for illegal activities as

00:04:29.220 --> 00:04:31.199
platforms like Dali become hyper

00:04:31.199 --> 00:04:34.840
realistic and look genuine

00:04:34.840 --> 00:04:37.540
on GitHub a lot of deep fake software

00:04:37.540 --> 00:04:38.340
may be discovered

00:04:38.340 --> 00:04:41.560
deed fake production is legal because

00:04:41.560 --> 00:04:42.240
some of these apps are only used for

00:04:42.240 --> 00:04:45.520
pure entertainment but other apps are

00:04:45.520 --> 00:04:47.100
much more likely to be used maliciously

00:04:47.100 --> 00:04:49.199
many experts predict that as technology

00:04:49.199 --> 00:04:52.540
advances deep fakes will become much

00:04:52.540 --> 00:04:54.580
more sophisticated and pose more

00:04:54.580 --> 00:04:56.800
substantial hazards to the public in the

00:04:56.800 --> 00:04:58.780
form of electoral meddling political

00:04:58.780 --> 00:05:01.380
unrest and increased criminal activity

00:05:01.380 --> 00:05:03.300
while deep fakes will only get more

00:05:03.300 --> 00:05:05.860
realistic with time as techniques

00:05:05.860 --> 00:05:07.840
improve we're not wholly defenseless

00:05:07.840 --> 00:05:09.820
when it comes to countering them

00:05:09.820 --> 00:05:11.740
deep fake detection techniques are being

00:05:11.740 --> 00:05:14.500
developed by several businesses some of

00:05:14.500 --> 00:05:16.500
which are startups for instance sensity

00:05:16.500 --> 00:05:18.479
has created a platform for deep fake

00:05:18.479 --> 00:05:21.520
identification that works like an

00:05:21.520 --> 00:05:23.800
antivirus and notifies users through

00:05:23.800 --> 00:05:25.720
email when they are watching something

00:05:25.720 --> 00:05:27.699
that contains the telltale signs of AI

00:05:27.699 --> 00:05:29.860
generated fake media

00:05:29.860 --> 00:05:31.960
the Deep learning techniques employed by

00:05:31.9...
                                                                                                                                                                                                                    
                                                                                                    

       
                                                                                                        
                                                                                                                                                                                                                            [{"text": "this is Young Simon cole performing in", "start": 0.359, "duration": 4.44}, {"text": "America's Got Talent in front of Simon", "start": 2.879, "duration": 4.621}, {"text": "cole because this is deep fake and this", "start": 4.799, "duration": 4.201}, {"text": "is a deep fake of Ukraine's president", "start": 7.5, "duration": 4.44}, {"text": "falada mazelinski asking his soldiers to", "start": 9.0, "duration": 4.2}, {"text": "surrender", "start": 11.94, "duration": 3.48}, {"text": "one is entertaining while the other is", "start": 13.2, "duration": 4.98}, {"text": "alarming about the dangers of deep fake", "start": 15.42, "duration": 6.119}, {"text": "these are the two faces of deep fake", "start": 18.18, "duration": 5.82}, {"text": "due to the term fake the phrase deep", "start": 21.539, "duration": 5.221}, {"text": "fake evokes a negative feeling", "start": 24.0, "duration": 4.619}, {"text": "companies like metaphysic are doing", "start": 26.76, "duration": 3.42}, {"text": "everything they can to dispel the", "start": 28.619, "duration": 3.421}, {"text": "misconceptions about deep fakes in the", "start": 30.18, "duration": 3.0}, {"text": "public eye", "start": 32.04, "duration": 3.179}, {"text": "this is what they did on America's Got", "start": 33.18, "duration": 3.24}, {"text": "turns", "start": 35.219, "duration": 2.881}, {"text": "there are many practical applications", "start": 36.42, "duration": 3.9}, {"text": "for deep fakes across a wide range of", "start": 38.1, "duration": 5.04}, {"text": "sectors for instance news outlets May", "start": 40.32, "duration": 5.759}, {"text": "publish news 24 hours a day seven days a", "start": 43.14, "duration": 4.68}, {"text": "week without the actual reporter present", "start": 46.079, "duration": 3.66}, {"text": "by using a deep fake of their own", "start": 47.82, "duration": 3.059}, {"text": "reporter", "start": 49.739, "duration": 3.3}, {"text": "they may reduce studio and labor costs", "start": 50.879, "duration": 4.381}, {"text": "saving the business a great deal of time", "start": 53.039, "duration": 3.54}, {"text": "and money", "start": 55.26, "duration": 4.439}, {"text": "the news anchor Kim Juha from the Korean", "start": 56.579, "duration": 5.82}, {"text": "television network mbn was deep faked", "start": 59.699, "duration": 4.081}, {"text": "for viewers", "start": 62.399, "duration": 3.72}, {"text": "the news reader would be a fake as the", "start": 63.78, "duration": 4.14}, {"text": "channel had previously informed viewers", "start": 66.119, "duration": 4.5}, {"text": "and Kim Joo has still employed", "start": 67.92, "duration": 5.64}, {"text": "however mbn stated that it intended to", "start": 70.619, "duration": 4.381}, {"text": "keep using the deet fake for some", "start": 73.56, "duration": 3.059}, {"text": "breaking news reports", "start": 75.0, "duration": 3.96}, {"text": "and deep frame a the South Korean", "start": 76.619, "duration": 4.801}, {"text": "company that created the Deep fake has", "start": 78.96, "duration": 3.9}, {"text": "stated that it is looking for media", "start": 81.42, "duration": 4.14}, {"text": "buyers in China and the US which has", "start": 82.86, "duration": 4.5}, {"text": "some people worried that news readers", "start": 85.56, "duration": 3.419}, {"text": "may become obsolete", "start": 87.36, "duration": 4.02}, {"text": "a tick tock account solely dedicated to", "start": 88.979, "duration": 4.561}, {"text": "Tom Cruise deep fakes has been created", "start": 91.38, "duration": 4.26}, {"text": "as a result of how far deep fakes have", "start": 93.54, "duration": 4.079}, {"text": "advanced in recent years", "start": 95.64, "duration": 3.839}, {"text": "videos of Crews doing everything", "start": 97.619, "duration": 3.661}, {"text": "including washing his hands and playing", "start": 99.479, "duration": 4.261}, {"text": "golf as well as doing magic tricks are", "start": 101.28, "duration": 4.08}, {"text": "available online", "start": 103.74, "duration": 3.96}, {"text": "the same firm that produced a convincing", "start": 105.36, "duration": 4.5}, {"text": "imitation of Simon Cowell metaphysic", "start": 107.7, "duration": 3.419}, {"text": "make this", "start": 109.86, "duration": 3.0}, {"text": "there are many potential uses of deep", "start": 111.119, "duration": 3.661}, {"text": "fakes in the film industry", "start": 112.86, "duration": 3.6}, {"text": "you may bring a deceased actor or", "start": 114.78, "duration": 3.0}, {"text": "actress back", "start": 116.46, "duration": 2.939}, {"text": "it is debatable for an ethical", "start": 117.78, "duration": 2.76}, {"text": "standpoint", "start": 119.399, "duration": 3.54}, {"text": "however if we then consider ethics it's", "start": 120.54, "duration": 4.32}, {"text": "quite doable and simple", "start": 122.939, "duration": 4.441}, {"text": "also it is probably far less expensive", "start": 124.86, "duration": 4.619}, {"text": "than alternative choices", "start": 127.38, "duration": 4.439}, {"text": "deep fakes enabled translated films that", "start": 129.479, "duration": 4.501}, {"text": "feature the original actors since they", "start": 131.819, "duration": 5.341}, {"text": "can modify footage and duplicate voices", "start": 133.98, "duration": 5.16}, {"text": "the voices resemble the original cast", "start": 137.16, "duration": 3.06}, {"text": "members", "start": 139.14, "duration": 3.3}, {"text": "and critically the lip motions even", "start": 140.22, "duration": 4.44}, {"text": "match the words at it with the help of", "start": 142.44, "duration": 4.62}, {"text": "positive deep fakes we can thus better", "start": 144.66, "duration": 5.52}, {"text": "exchange ideas movies and other creative", "start": 147.06, "duration": 5.88}, {"text": "works on a global scale even those with", "start": 150.18, "duration": 4.5}, {"text": "tighter finances", "start": 152.94, "duration": 3.659}, {"text": "this should increase the variety of", "start": 154.68, "duration": 4.32}, {"text": "entertainment and content we consume", "start": 156.599, "duration": 4.92}, {"text": "in other words deep fakes can overcome", "start": 159.0, "duration": 4.379}, {"text": "language barriers and increase the", "start": 161.519, "duration": 4.141}, {"text": "accessibility of content", "start": 163.379, "duration": 5.281}, {"text": "deep fakes aren't just found in videos", "start": 165.66, "duration": 5.28}, {"text": "a rapidly expanding field with a vast", "start": 168.66, "duration": 5.52}, {"text": "array of uses is deep fake audio", "start": 170.94, "duration": 5.64}, {"text": "deep learning algorithms can now create", "start": 174.18, "duration": 4.919}, {"text": "realistic audio deep fakes with just a", "start": 176.58, "duration": 4.2}, {"text": "few hours of audio of the person whose", "start": 179.099, "duration": 3.72}, {"text": "voice has been replicated", "start": 180.78, "duration": 4.679}, {"text": "once a model of a voice is created that", "start": 182.819, "duration": 4.981}, {"text": "person can be made to say anything", "start": 185.459, "duration": 4.621}, {"text": "thanks to deep fake audio programmers", "start": 187.8, "duration": 4.26}, {"text": "may now allow gamer characters to say", "start": 190.08, "duration": 4.32}, {"text": "anything in game rather than depending", "start": 192.06, "duration": 4.2}, {"text": "on a small number of lines that were", "start": 194.4, "duration": 4.74}, {"text": "produced before the game was released", "start": 196.26, "duration": 4.979}, {"text": "some people use this technology to", "start": 199.14, "duration": 4.379}, {"text": "commit financial fraud data breaches", "start": 201.239, "duration": 4.92}, {"text": "phishing scams automated misinformation", "start": 203.519, "duration": 5.161}, {"text": "attacks and the spread of misleading", "start": 206.159, "duration": 4.021}, {"text": "information from sources that are", "start": 208.68, "duration": 3.24}, {"text": "normally trustworthy", "start": 210.18, "duration": 3.779}, {"text": "critics also point out the risks of", "start": 211.92, "duration": 3.899}, {"text": "utilizing deed fakes to further a", "start": 213.959, "duration": 4.14}, {"text": "political agenda such as by creating", "start": 215.819, "duration": 4.081}, {"text": "credible fake news", "start": 218.099, "duration": 3.961}, {"text": "politicians have also employed deep fake", "start": 219.9, "duration": 3.36}, {"text": "videos", "start": 222.06, "duration": 4.259}, {"text": "in 2018 a Belgian political party made", "start": 223.26, "duration": 4.619}, {"text": "public a video showing Donald Trump", "start": 226.319, "duration": 3.78}, {"text": "pressing Belgium to leave the Paris", "start": 227.879, "duration": 4.381}, {"text": "climate Accord in a speech", "start": 230.099, "duration": 5.64}, {"text": "Trump never gave that speech however it", "start": 232.26, "duration": 4.979}, {"text": "was a deep fake", "start": 235.739, "duration": 3.78}, {"text": "it can be used to scam people by using", "start": 237.239, "duration": 4.321}, {"text": "the identities of popular people", "start": 239.519, "duration": 4.021}, {"text": "and there is enough data on social media", "start": 241.56, "duration": 4.08}, {"text": "to identify a certain influencer or", "start": 243.54, "duration": 3.419}, {"text": "celebrity", "start": 245.64, "duration": 3.06}, {"text": "there are thousands of photos and videos", "start": 246.959, "duration": 4.2}, {"text": "online which can be used as training", "start": 248.7, "duration": 4.5}, {"text": "data to create frauds in the name of", "start": 251.159, "duration": 4.08}, {"text": "authentic people this is some of the", "start": 253.2, "duration": 4.2}, {"text": "reasons why the Ali Community is afraid", "start": 255.239, "duration": 3.96}, {"text": "of making deep faking available to", "start": 257.4, "duration": 3.36}, {"text": "everyone regardless of their background", "start": 259.199, "durat...