1
00:00:00,040 --> 00:00:03,120
With Jenny I, it's become easier
and easier to create videos, 

2
00:00:03,520 --> 00:00:06,800
very cool, but also fake videos 
and deepfakes. 

3
00:00:07,160 --> 00:00:10,440
Today I'm joined by deepfake 
expert, CEO and Co founder of 

4
00:00:10,440 --> 00:00:14,240
Duckduckgoos, AI Pariah Lotfi. 
Believing in what you're 

5
00:00:14,240 --> 00:00:16,200
perceiving. 
That's the company motto. 

6
00:00:16,480 --> 00:00:20,120
And she shares many stories of 
what can happen to businesses, 

7
00:00:20,120 --> 00:00:23,440
but also to you and me and 
things that have happened to 

8
00:00:23,440 --> 00:00:25,560
businesses and to people like 
you and me. 

9
00:00:26,000 --> 00:00:27,600
I love this conversation so 
enjoy. 

10
00:00:31,040 --> 00:00:34,080
I was on the way to the office 
and I was on Reddit because 

11
00:00:34,080 --> 00:00:35,680
that's what I do till I relax 
and stuff. 

12
00:00:36,240 --> 00:00:39,240
And I saw this video and it was 
like the seagull on a car 

13
00:00:39,240 --> 00:00:42,440
dashboard and the guy throws a 
fry, but obviously the fry is 

14
00:00:42,440 --> 00:00:45,160
like behind the window screen. 
And then all of a sudden the 

15
00:00:45,160 --> 00:00:47,400
seagull bursts through and grabs
the fry bursts. 

16
00:00:47,400 --> 00:00:48,960
And I'm like, what the hell did 
I just watch? 

17
00:00:49,400 --> 00:00:51,600
And go in the comments section 
If you're like, well, this is AI

18
00:00:51,600 --> 00:00:55,280
generated obviously. 
And then other people said the 

19
00:00:55,280 --> 00:00:58,320
beginning of the video is not 
even is that's real and all of a

20
00:00:58,320 --> 00:01:00,920
sudden switches to AI. 
That's also what caught me 

21
00:01:00,920 --> 00:01:02,320
because I was like, the hell is 
this video? 

22
00:01:02,320 --> 00:01:03,400
And then it burst through the 
window. 

23
00:01:03,400 --> 00:01:07,320
I'm like, Oh my God, yeah, it's 
getting very subtle right now. 

24
00:01:07,520 --> 00:01:09,560
That's true to a point where 
it's ridiculous. 

25
00:01:09,560 --> 00:01:12,480
Yeah, yeah, yeah. 
They are mixed in reality with 

26
00:01:12,720 --> 00:01:16,720
fakeness using AI, Yeah. 
And that's very powerful 

27
00:01:16,720 --> 00:01:19,520
actually, because indeed, the 
first thing that you see is like

28
00:01:19,520 --> 00:01:21,200
something real, something 
familiar to you. 

29
00:01:21,640 --> 00:01:24,440
So you would believe it. 
And then all of a sudden changes

30
00:01:24,600 --> 00:01:27,280
and the chances that you would 
for example also believe that 

31
00:01:27,280 --> 00:01:31,040
part right now is higher. 
And that's also why we say the 

32
00:01:31,200 --> 00:01:35,640
the basically the difference 
between what is real and what is

33
00:01:35,640 --> 00:01:39,840
fake is getting everyday like 
more and more fake basically. 

34
00:01:40,080 --> 00:01:42,960
Can you also do like if you have
a video, can you do part of a 

35
00:01:42,960 --> 00:01:45,560
video frame and then fake that 
part as well? 

36
00:01:45,600 --> 00:01:48,200
Yeah, what happens right now 
also with these very accessible 

37
00:01:48,200 --> 00:01:51,080
large language models like 
ChatGPT and the others is that 

38
00:01:51,080 --> 00:01:53,600
you can upload one single frame 
picture, OK. 

39
00:01:53,800 --> 00:01:57,520
And it's going to make it fully 
like a moving basic D video for 

40
00:01:57,520 --> 00:01:59,640
you based on what you have been 
uploading. 

41
00:01:59,640 --> 00:02:02,400
I've seen a couple of examples 
of you what what you mentioned 

42
00:02:02,400 --> 00:02:06,240
basically just yet, but that was
with more like in a setup of 

43
00:02:06,240 --> 00:02:09,680
like politicians, they were all 
sitting Trump, Biden, etcetera. 

44
00:02:10,120 --> 00:02:13,120
The the first part was real and 
then in the second part, all of 

45
00:02:13,120 --> 00:02:16,280
a sudden they started fighting. 
Like, you know, they started 

46
00:02:16,280 --> 00:02:17,840
basically fighting. 
And I was like, wow, what 

47
00:02:17,840 --> 00:02:19,880
didn't, what did I just watch 
what had happened here? 

48
00:02:20,240 --> 00:02:22,840
And that's basically also how 
they play around with such 

49
00:02:22,840 --> 00:02:26,960
technology to not only make 
funny and nice things out of it,

50
00:02:26,960 --> 00:02:31,040
but also really fraudulent and 
with bad purposes, with bad 

51
00:02:31,040 --> 00:02:33,520
intentions basically. 
And that's basically also where 

52
00:02:33,800 --> 00:02:36,080
this dangerous part comes in. 
I'm not sure if you have seen 

53
00:02:36,080 --> 00:02:38,480
it, but there was an interview 
of Sam Altman, I think yesterday

54
00:02:38,480 --> 00:02:42,840
or the day before, where he also
mentioned it's unbelievable that

55
00:02:42,960 --> 00:02:47,720
these days we still rely on 
facial biometrics, voice 

56
00:02:47,720 --> 00:02:52,360
biometrics to, for example, do 
very high security jobs such as 

57
00:02:52,360 --> 00:02:55,200
logging into your bank account, 
transferring money. 

58
00:02:55,680 --> 00:02:59,280
And also for normal 
communication ways that we have 

59
00:02:59,280 --> 00:03:01,720
like we FaceTime the whole time 
basically, right? 

60
00:03:01,720 --> 00:03:06,000
We use Zoom, we use Teams. 
And also in those scenarios, 

61
00:03:06,000 --> 00:03:09,440
it's very likely that someone 
using only open source 

62
00:03:09,440 --> 00:03:11,880
technology, you know, can 
impersonate someone else. 

63
00:03:12,160 --> 00:03:15,520
And this can happen in a 
business setup or context, but 

64
00:03:15,520 --> 00:03:19,080
it can also happen to a mother 
who receives a call from their 

65
00:03:19,080 --> 00:03:22,240
child saying, mom, I got 
kidnapped. 

66
00:03:22,240 --> 00:03:25,880
You should pay this amount to 
this person or this bank 

67
00:03:25,880 --> 00:03:27,880
account, and then they will let 
me go. 

68
00:03:27,880 --> 00:03:32,000
And you can imagine with that, 
you know, emotional state of 

69
00:03:32,000 --> 00:03:34,200
mind what you would do. 
Of course you would. 

70
00:03:34,200 --> 00:03:36,440
Maybe, I think a high chance 
that you would do what they 

71
00:03:36,440 --> 00:03:38,960
asked for. 
And this is becoming our 

72
00:03:38,960 --> 00:03:41,000
reality. 
This is already the case. 

73
00:03:41,000 --> 00:03:43,560
Like the you can already do 
stuff like that, yeah. 

74
00:03:43,680 --> 00:03:46,200
Yeah, yeah, it's unbelievable. 
But you can already do stuff 

75
00:03:46,200 --> 00:03:49,760
like that also on a big scale. 
So the difference when it comes 

76
00:03:49,760 --> 00:03:53,720
to so generative AI based image 
and video manipulation and 

77
00:03:53,720 --> 00:03:57,760
generation compared to 
traditional ways of doing it 

78
00:03:57,760 --> 00:03:59,320
because it's it's been there 
always, right. 

79
00:03:59,320 --> 00:04:04,120
We have some pictures of Stalin 
basically with his friends and 

80
00:04:04,120 --> 00:04:06,560
then all of a sudden the friend 
was gone from the picture 

81
00:04:06,560 --> 00:04:08,400
because they were not friends 
anymore. 

82
00:04:08,440 --> 00:04:14,880
Yeah, removed from the history. 
So those were the, for example, 

83
00:04:14,880 --> 00:04:17,079
the first types of image 
manipulation. 

84
00:04:17,079 --> 00:04:20,920
It's always been there. 
The difference right now is that

85
00:04:20,920 --> 00:04:24,840
it is so much more scalable and 
so much easier for everyone to 

86
00:04:24,840 --> 00:04:29,560
be able to make it. 
We have seen cases where mostly 

87
00:04:29,560 --> 00:04:33,160
also in the US, but also in the 
Netherlands, you would receive a

88
00:04:33,160 --> 00:04:36,360
call or a voice note on WhatsApp
from someone, a loved one, and 

89
00:04:36,360 --> 00:04:38,920
then they claim things like I 
just mentioned. 

90
00:04:39,600 --> 00:04:44,160
I received also a call lately 
from it was, it was clearly an 

91
00:04:44,160 --> 00:04:46,080
AI voice. 
I received this call. 

92
00:04:46,080 --> 00:04:47,640
It was, it looked like a normal 
number. 

93
00:04:47,640 --> 00:04:49,200
So I was like, like, let me just
answer. 

94
00:04:49,880 --> 00:04:53,840
And then it said in Dutch, Hey, 
I think it also knew my name. 

95
00:04:53,840 --> 00:04:57,960
That was really surprising. 
Hey Pariya, just wanted to know 

96
00:04:57,960 --> 00:05:02,000
if you're interested in buying 
houses and real estate in 

97
00:05:02,000 --> 00:05:03,720
Almira. 
I was like, wow. 

98
00:05:04,760 --> 00:05:09,440
And the goal of the of the 
basically voice behind the phone

99
00:05:09,440 --> 00:05:12,760
was to make an appointment with 
me, which I don't didn't 

100
00:05:12,760 --> 00:05:15,240
understand because it was like 
an AI voice clearly. 

101
00:05:15,240 --> 00:05:16,360
And then making appointment with
me. 

102
00:05:16,360 --> 00:05:19,040
I'm like, why is that? 
Yeah, what's the purpose of it? 

103
00:05:19,040 --> 00:05:20,840
Or maybe they just wanted to 
hear my voice. 

104
00:05:20,840 --> 00:05:22,240
That's also possible. 
Oh really? 

105
00:05:22,240 --> 00:05:25,040
Because that's that's also 
really dangerous because right 

106
00:05:25,120 --> 00:05:29,960
nowadays they only need like 
maybe 3-4 seconds of someone's 

107
00:05:29,960 --> 00:05:33,400
voice to be able to generate a 
perfectly realistic sounding 

108
00:05:33,400 --> 00:05:34,840
clone of it. 
So this podcast. 

109
00:05:34,840 --> 00:05:36,960
It's a horrible for me, it's a 
horrible for you. 

110
00:05:37,880 --> 00:05:38,960
That's true, yeah. 
Yeah, I think. 

111
00:05:39,800 --> 00:05:45,720
About Nah, but like for me, I 
don't feel like those types of 

112
00:05:45,720 --> 00:05:48,480
technologies are very 
accessible, at least for me. 

113
00:05:48,960 --> 00:05:51,480
I tried to play around with like
video editing software. 

114
00:05:51,800 --> 00:05:53,720
I I don't edit the podcast 
anymore. 

115
00:05:53,720 --> 00:05:56,640
I used to so I was still playing
around with it and I saw that 

116
00:05:56,640 --> 00:06:00,040
the software that I use had this
AI functionality, seamless 

117
00:06:00,040 --> 00:06:03,360
transition and we use Gen. 
AI to do so It's like, OK, I 

118
00:06:03,360 --> 00:06:06,880
have my podcast, I'll just cut 
and exactly the same frame. 

119
00:06:06,880 --> 00:06:10,680
I'll put a little bit after and 
then I'll just have Jenny I in 

120
00:06:10,680 --> 00:06:11,600
the middle. 
That's what I thought. 

121
00:06:12,000 --> 00:06:13,480
And it's just me reacting like 
this. 

122
00:06:13,760 --> 00:06:17,880
And also I go like I basically 
blurb out in a weird way and I 

123
00:06:18,000 --> 00:06:19,680
was like, what is this? 
And then it just snaps back to 

124
00:06:19,680 --> 00:06:20,560
the frame. 
That's real life. 

125
00:06:20,560 --> 00:06:23,280
So it didn't look great. 
I, I could clearly see if 

126
00:06:23,280 --> 00:06:26,280
something's wrong, but then I do
see examples online that it's 

127
00:06:26,280 --> 00:06:30,640
like very, very specific and I 
have to double check and I have 

128
00:06:30,640 --> 00:06:33,480
to because I mainly see these on
like Twitter and Reddit and more

129
00:06:33,480 --> 00:06:35,720
community based things. 
I have to go to the comments 

130
00:06:35,720 --> 00:06:38,160
section and other people have to
educate me what I'm seeing. 

131
00:06:38,160 --> 00:06:39,960
Yeah, that's scary. 
That's scary. 

132
00:06:39,960 --> 00:06:41,640
Yeah, exactly. 
Yeah, and you're working on 

133
00:06:41,640 --> 00:06:46,240
software to signal that to at 
least see what is real and what 

134
00:06:46,240 --> 00:06:48,680
is fake? 
Yeah, that's exactly what we do.

135
00:06:48,680 --> 00:06:54,040
So we basically use AI to catch 
AI in that sense at the the 

136
00:06:54,040 --> 00:06:56,640
Goose, we develop explainable 
but also generalizable AI 

137
00:06:56,640 --> 00:06:59,800
technologies that would help 
human beings, companies, 

138
00:06:59,800 --> 00:07:03,400
citizens, everyone basically to 
be able to distinguish what is 

139
00:07:03,440 --> 00:07:06,120
fake from what is real when it 
when it comes to the digital 

140
00:07:06,120 --> 00:07:07,680
world. 
How do you do that? 

141
00:07:07,680 --> 00:07:11,760
Like can I say me with the AI to
actually create this? 

142
00:07:11,760 --> 00:07:13,920
Can I also have the same AI say 
as this real or fake? 

143
00:07:13,920 --> 00:07:15,840
Because I've tried that, it's 
not as good. 

144
00:07:16,200 --> 00:07:18,160
You have trained. 
Your own systems. 

145
00:07:18,160 --> 00:07:21,600
No, I haven't trained my like I,
I, for example, I received on 

146
00:07:21,600 --> 00:07:24,040
Twitter sponsorships 
opportunities. 

147
00:07:24,920 --> 00:07:26,440
I wrote a LinkedIn post about 
that. 

148
00:07:26,560 --> 00:07:28,880
It got pretty close actually, 
because they can now generate a 

149
00:07:28,880 --> 00:07:31,760
website and there it's like, oh,
you have to do this digital 

150
00:07:31,760 --> 00:07:33,800
signature. 
And it downloaded something on 

151
00:07:33,800 --> 00:07:35,200
my computer. 
And I was like, oh, that's, 

152
00:07:35,200 --> 00:07:37,560
that's a red flag. 
And then I thought, OK, 

153
00:07:38,000 --> 00:07:40,120
everything in this conversation,
there were many signals of red 

154
00:07:40,120 --> 00:07:41,800
flags. 
But because the companies were 

155
00:07:41,800 --> 00:07:44,760
super cool and the people had 
had Twitter accounts which were 

156
00:07:44,760 --> 00:07:48,000
verified for years with post 
history and I could see they 

157
00:07:48,000 --> 00:07:49,760
have LinkedIn accounts and 
they're actually people that 

158
00:07:49,760 --> 00:07:51,600
work there that I thought it was
real. 

159
00:07:51,920 --> 00:07:53,240
I want to tell myself it was 
real. 

160
00:07:53,440 --> 00:07:56,360
But in the end, when it came to 
a digital signature, I'm not 

161
00:07:56,360 --> 00:07:59,120
going to download a program from
some. 

162
00:07:59,640 --> 00:08:02,520
If I look at the domain, a weird
sub domain that is like a tool 

163
00:08:02,520 --> 00:08:04,600
that I've used in the past, but 
it's not really the real thing. 

164
00:08:04,880 --> 00:08:06,240
Like those are all signals to 
me. 

165
00:08:06,680 --> 00:08:09,560
And then I asked because one of 
them said, OK, you have to sign 

166
00:08:09,560 --> 00:08:12,160
this NDA. 
It wasn't a here's the contract.

167
00:08:12,160 --> 00:08:14,800
Let's sign it was your first 
half to sign this NDA? 

168
00:08:15,080 --> 00:08:18,480
I got the NDA and the first 
thing I did was I added it to my

169
00:08:18,480 --> 00:08:20,440
ChatGPT at the time. 
And I was like, is this real 

170
00:08:20,440 --> 00:08:22,080
fake? 
And they were like, well, 

171
00:08:22,080 --> 00:08:24,960
there's no, not many reasons to 
think this is like fake. 

172
00:08:24,960 --> 00:08:26,840
So it could be real. 
I'm like one of the reasons and 

173
00:08:26,840 --> 00:08:29,920
they laid out the things that 
are excluded or like omitted 

174
00:08:29,920 --> 00:08:33,880
that for me made total sense. 
How long is the NDA valid for 

175
00:08:33,880 --> 00:08:36,080
which specific things? 
Does it make sense, yes or no? 

176
00:08:36,080 --> 00:08:37,960
Some legal things that were 
missing. 

177
00:08:37,960 --> 00:08:41,480
So I was like, OK, like all 
those signals and even my AI, 

178
00:08:41,520 --> 00:08:44,200
which was likely the same AI 
that they used to create an NDA 

179
00:08:44,200 --> 00:08:47,360
in the 1st place, yeah, Said OK,
it could be real, yeah. 

180
00:08:47,560 --> 00:08:50,720
Yeah, yeah. 
OK, that's that's neat. 

181
00:08:50,720 --> 00:08:52,400
And I won't understand what 
you're saying. 

182
00:08:52,680 --> 00:08:55,440
I think it sounds almost also 
like organized crime people 

183
00:08:55,440 --> 00:08:59,560
setting up websites and, you 
know, trying to get others to 

184
00:08:59,560 --> 00:09:02,000
communicate with or do something
else with them. 

185
00:09:02,000 --> 00:09:05,440
When it comes to large language 
models like Chi GPT being able 

186
00:09:05,440 --> 00:09:10,360
to distinguish what is fake, we 
have seen indeed also research 

187
00:09:10,360 --> 00:09:13,560
being done regarding that. 
The only difference is that 

188
00:09:13,600 --> 00:09:17,440
these models, they are general 
Transformers, let's say and they

189
00:09:17,440 --> 00:09:20,520
are really good in pattern 
recognition when it comes to 

190
00:09:20,520 --> 00:09:24,040
text. 
But when it comes to videos, you

191
00:09:24,040 --> 00:09:27,320
know, pictures and even like 
NDA's that you mentioned, 

192
00:09:27,680 --> 00:09:31,000
specifically asking them is this
fake or real might not be that 

193
00:09:31,000 --> 00:09:34,000
effective because they are not 
really meant for that purpose. 

194
00:09:34,400 --> 00:09:36,640
And that's the difference 
between, for example, your 

195
00:09:36,640 --> 00:09:39,880
question that you asked to chat 
versus our systems. 

196
00:09:39,880 --> 00:09:42,320
We have been out there for five 
years almost right now. 

197
00:09:42,320 --> 00:09:45,760
And that was also really preach 
IGPD era before it got really 

198
00:09:45,760 --> 00:09:47,520
popular and everyone started 
using it. 

199
00:09:48,400 --> 00:09:51,160
And what we have done throughout
these years is basically 

200
00:09:51,160 --> 00:09:54,160
training our very specific 
systems for very specific 

201
00:09:54,160 --> 00:09:57,800
purposes of catching different 
types of deepfakes with the 

202
00:09:57,800 --> 00:10:02,240
biggest basically also 
uniqueness to be able to explain

203
00:10:02,240 --> 00:10:06,880
as to why a video, a picture or 
audio is seen as a deepfake. 

204
00:10:07,040 --> 00:10:09,880
You can explain why. 
We can explain what kind of 

205
00:10:09,880 --> 00:10:12,960
reasoning the neural network is 
following to come up with that 

206
00:10:12,960 --> 00:10:14,240
conclusion. 
Interesting. 

207
00:10:14,240 --> 00:10:17,920
And AI being explainable is not 
really common. 

208
00:10:18,040 --> 00:10:20,800
Of course, it's one of the 
biggest challenges in this world

209
00:10:20,800 --> 00:10:24,400
right now. 
Since AI is by design, like kind

210
00:10:24,400 --> 00:10:26,360
of a black box, like magic. 
Exactly. 

211
00:10:26,560 --> 00:10:28,560
We don't know what's happening. 
Also at the the Coos, we're 

212
00:10:28,560 --> 00:10:31,600
trying to kind of understand 
what are the reasonings behind 

213
00:10:31,600 --> 00:10:35,760
this trace of thought. 
And we visually also show that 

214
00:10:35,760 --> 00:10:37,880
to the end user. 
So when you have, for example, a

215
00:10:37,880 --> 00:10:41,600
face swab or this bird that you 
mentioned going through a window

216
00:10:41,600 --> 00:10:47,480
getting this fry, we could for 
example make our users see which

217
00:10:47,480 --> 00:10:51,080
parts of this video have been 
manipulated specifically and 

218
00:10:51,080 --> 00:10:54,520
which type of defect generation 
is behind this manipulation or 

219
00:10:54,520 --> 00:10:56,520
creation being generation 
basically. 

220
00:10:56,880 --> 00:11:00,520
So in that sense, we are really 
specialized in basically 

221
00:11:00,520 --> 00:11:03,120
detection of deepfakes. 
And with deepfakes, we basically

222
00:11:03,120 --> 00:11:06,320
mean every type of audio visual 
media that has either been 

223
00:11:06,320 --> 00:11:10,160
manipulated or completely 
generated using AI technologies 

224
00:11:11,960 --> 00:11:15,680
in comparison to ChatGPT, which 
is more, I'm so sorry. 

225
00:11:15,880 --> 00:11:20,640
No problem. 
Okay, yeah, Which is more like a

226
00:11:20,640 --> 00:11:24,280
general type of AI for general 
usage scenarios. 

227
00:11:24,600 --> 00:11:27,760
And when you said it could kind 
of explain which type of 

228
00:11:27,760 --> 00:11:30,200
deepfakes have been applied to 
what you're seeing. 

229
00:11:30,200 --> 00:11:33,600
Are you then saying audio video 
as type? 

230
00:11:33,600 --> 00:11:35,600
Or even within video there's 
subtypes. 

231
00:11:35,800 --> 00:11:38,840
Within video. 
Even so, because we use this 

232
00:11:39,320 --> 00:11:42,200
comparison like defect 
generation techniques just pop 

233
00:11:42,200 --> 00:11:45,040
up like mushrooms from the 
ground like everywhere on a 

234
00:11:45,040 --> 00:11:47,080
daily basis. 
You have 10s of new defect 

235
00:11:47,080 --> 00:11:49,720
generation techniques only from 
the scientific world basically 

236
00:11:49,720 --> 00:11:52,680
coming. 
And then let alone what happens 

237
00:11:52,680 --> 00:11:56,800
if people who are also maybe 
part of organized crime groups 

238
00:11:57,080 --> 00:12:00,160
have bad intentions to use the 
effects to maybe open bank 

239
00:12:00,160 --> 00:12:02,840
accounts or scam companies, scam
individuals. 

240
00:12:03,360 --> 00:12:05,840
So in that sense, it's very 
important for us to know how a 

241
00:12:05,840 --> 00:12:09,520
certain deepfake has been made 
to also work on our threat 

242
00:12:09,520 --> 00:12:11,720
intelligence part. 
So when it comes to, for 

243
00:12:11,720 --> 00:12:15,840
example, a single frame simple 
image that can be a deepfake, 

244
00:12:15,840 --> 00:12:17,520
but it's a different type of 
deepfake. 

245
00:12:17,520 --> 00:12:19,040
Then you have a face swap 
deepfake. 

246
00:12:19,040 --> 00:12:21,600
So you can right now take a 
picture of me and put your face 

247
00:12:21,600 --> 00:12:24,000
on my face. 
Sometimes I kind of know blended

248
00:12:24,360 --> 00:12:29,080
sorry and the other type would 
be for example generating a 

249
00:12:29,080 --> 00:12:33,240
completely non existing human in
that sense not a picture 

250
00:12:33,240 --> 00:12:35,560
anymore, but something that has 
been created by a neural 

251
00:12:35,560 --> 00:12:38,240
network. 
So we can for example tell that 

252
00:12:38,240 --> 00:12:41,360
this image has been fully 
synthesized or this image is 

253
00:12:41,360 --> 00:12:44,200
partly real, partly face swap or
partly lip sync. 

254
00:12:44,200 --> 00:12:47,840
When it comes to videos to make 
someone say something that they 

255
00:12:47,840 --> 00:12:50,480
never have said, that's 
basically usually a lip sync 

256
00:12:50,480 --> 00:12:53,880
deep fake video where you don't 
have mostly this part of the 

257
00:12:53,880 --> 00:12:56,840
face manipulated to make you 
could synchronize to the certain

258
00:12:56,960 --> 00:12:59,280
audio that you want to put on 
the video itself. 

259
00:12:59,280 --> 00:13:03,600
Fascinating, how much education 
do you need to do to your 

260
00:13:03,600 --> 00:13:06,800
customers or organizations in 
general with regards to this? 

261
00:13:06,800 --> 00:13:10,640
Because I can imagine a big 
difference between kind of pre 

262
00:13:10,640 --> 00:13:13,120
Gen. 
AI era and now. 

263
00:13:13,200 --> 00:13:17,280
Because I saw this video just 
online so I am aware of the 

264
00:13:17,280 --> 00:13:20,960
possibilities or at least I have
a lot of assumptions in what is 

265
00:13:20,960 --> 00:13:23,280
possible and what is not. 
But also, I'm in the tech scene.

266
00:13:23,800 --> 00:13:26,760
I don't know many organizations 
that are not, but let's say more

267
00:13:26,760 --> 00:13:29,720
traditional banks, they do have 
IT capabilities and they have a 

268
00:13:29,720 --> 00:13:32,400
very low risk appetite. 
So maybe they are more aware. 

269
00:13:32,760 --> 00:13:34,680
But companies that don't have 
that, they can be like, well, 

270
00:13:34,680 --> 00:13:36,960
why would it hit us? 
Like what would be interesting 

271
00:13:36,960 --> 00:13:38,640
for us in this aspect? 
Yeah, yeah. 

272
00:13:38,920 --> 00:13:41,840
So once we started five years 
ago, we basically had to explain

273
00:13:41,840 --> 00:13:44,720
to everyone, like 99% of our 
interactions. 

274
00:13:44,720 --> 00:13:47,800
We had to explain or start by 
explaining what a deepfake 

275
00:13:47,800 --> 00:13:51,520
actually is. 
Right now there is much more 

276
00:13:51,520 --> 00:13:54,160
awareness regarding what 
generative AI is, what deepfakes

277
00:13:54,160 --> 00:13:57,280
is thanks to HIGPTH programs and
and tools. 

278
00:13:57,280 --> 00:14:02,920
Basically, we still need to 
touch upon every relevant use 

279
00:14:02,920 --> 00:14:06,280
case for the person or the 
company that we are talking to 

280
00:14:06,600 --> 00:14:10,280
that has approached us to make 
them understand to what extent 

281
00:14:10,280 --> 00:14:13,960
they can already be exposed to 
these deepfake based or Gen. 

282
00:14:13,960 --> 00:14:16,880
AI based fraud types. 
And it starts with like 

283
00:14:16,880 --> 00:14:20,160
individuals getting, you know, a
deep fake call from their loved 

284
00:14:20,160 --> 00:14:25,440
ones, but also a teenager being 
put in an 18 plus non consensual

285
00:14:25,440 --> 00:14:29,160
video up to market 
manipulations, fake fake news 

286
00:14:29,160 --> 00:14:32,200
that we all know about. 
And even like stealing money 

287
00:14:32,200 --> 00:14:35,040
from banks up to 1,000,000. 
Aside from the ones that have 

288
00:14:35,040 --> 00:14:38,080
made it to the headlines, we 
have seen several cases like 

289
00:14:38,080 --> 00:14:42,000
even Dutch small medium 
enterprise companies, they're 

290
00:14:42,000 --> 00:14:44,960
also getting targeted by a voice
that calls them that sounds like

291
00:14:44,960 --> 00:14:47,440
their colleague asking for 
either information or money. 

292
00:14:48,520 --> 00:14:52,200
And this basically showcases 
that the scale that is happening

293
00:14:52,200 --> 00:14:53,760
right now already is a real 
threat. 

294
00:14:53,760 --> 00:14:57,040
It's not theoretical anymore. 
We see it also happening a lot 

295
00:14:57,040 --> 00:15:00,280
in flows that you have lots of 
volume of volumes of data. 

296
00:15:01,360 --> 00:15:05,560
As I said, just with with the 
Sam Altman example, he just 

297
00:15:05,560 --> 00:15:08,560
said, you know, why are you 
still using face for her for 

298
00:15:08,560 --> 00:15:10,480
authentication and 
identification? 

299
00:15:11,600 --> 00:15:14,440
It's still a very secure way, it
seems like, but it is not 

300
00:15:14,440 --> 00:15:17,440
because deep fix Gen. 
AI are basically able and 

301
00:15:17,440 --> 00:15:20,880
capable of mimicking all these, 
you know, human like 

302
00:15:20,880 --> 00:15:23,640
characteristics that we have in 
a video or in a picture. 

303
00:15:24,120 --> 00:15:26,640
And what happens usually, of 
course, you use your, you know, 

304
00:15:26,640 --> 00:15:30,120
face, your voice to be able to 
open a bank account along with 

305
00:15:30,120 --> 00:15:32,880
your passport picture. 
And when it comes to, for 

306
00:15:32,880 --> 00:15:36,320
example, also banks and neo 
banks, they have on a daily 

307
00:15:36,320 --> 00:15:39,560
basis thousands of new clients 
that want to, you know, open a 

308
00:15:39,560 --> 00:15:42,680
new bank account and maybe like 
10s of thousands that want to 

309
00:15:42,680 --> 00:15:45,680
log into their account as well. 
And they use facial biometrics. 

310
00:15:46,160 --> 00:15:49,720
So in that sense, we also see a 
lot of deepfakes coming in to 

311
00:15:49,720 --> 00:15:53,720
basically be able to make an 
account based on a non traceable

312
00:15:53,720 --> 00:15:57,120
identity using like a picture of
a non existing person along with

313
00:15:57,120 --> 00:15:59,560
a fake passport. 
And then they use that to be 

314
00:15:59,560 --> 00:16:02,280
able to open a bank account. 
And it doesn't stop there. 

315
00:16:02,280 --> 00:16:05,560
Once they find a method, these 
people with bad intentions to 

316
00:16:05,560 --> 00:16:08,640
basically open that account, 
they go ahead and open maybe 10s

317
00:16:08,640 --> 00:16:12,160
of or hundreds of them and they 
eventually use it for really bad

318
00:16:12,160 --> 00:16:13,840
purposes such as money 
laundering. 

319
00:16:13,840 --> 00:16:16,960
And then from there on also 
supporting organized crime even 

320
00:16:16,960 --> 00:16:19,280
more. 
And that's one of the flows that

321
00:16:19,280 --> 00:16:23,040
we see lots of defects basically
up to 10% on a daily basis in 

322
00:16:23,040 --> 00:16:27,240
peak times coming up. 
And that's only one use case, 

323
00:16:27,240 --> 00:16:29,200
let's say with remote 
identification and 

324
00:16:29,200 --> 00:16:32,040
authentication. 
We also had cases in which 

325
00:16:32,040 --> 00:16:35,400
people like individual people, 
maybe also more vulnerable 

326
00:16:36,120 --> 00:16:39,560
groups of society calling us, 
saying I just heard you talking 

327
00:16:39,560 --> 00:16:43,440
on the radio. 
I have a girlfriend online, 

328
00:16:43,880 --> 00:16:45,760
never seen her. 
Oh no. 

329
00:16:45,760 --> 00:16:47,640
But I send money to her because 
she. 

330
00:16:47,840 --> 00:16:48,880
She needs it. 
She needs it. 

331
00:16:48,880 --> 00:16:52,360
Yeah, of course. 
But yeah, it got me thinking, 

332
00:16:52,400 --> 00:16:55,320
can you please check it for me? 
And of course, we went along, we

333
00:16:55,320 --> 00:16:58,000
checked and we said please don't
send money again to this person 

334
00:16:58,000 --> 00:17:02,040
because you know, it's she's, 
she's a deep fake, which is. 

335
00:17:02,600 --> 00:17:05,160
This use case you understand 
because they reached out to you.

336
00:17:05,160 --> 00:17:08,040
But when it comes to the other 
use cases, the more let's say 

337
00:17:08,040 --> 00:17:11,400
Baker skill fraud, how do you 
know these things already exist?

338
00:17:11,400 --> 00:17:12,720
Like how do you get your 
information? 

339
00:17:12,720 --> 00:17:15,240
Because if I'm a bank and this 
happens, I wouldn't want to 

340
00:17:15,240 --> 00:17:17,720
publish this because I'll become
a target. 

341
00:17:17,720 --> 00:17:21,160
And it's kind of a shame thing. 
And you know, these cases exist 

342
00:17:21,160 --> 00:17:22,200
in the first place. 
Yeah. 

343
00:17:22,400 --> 00:17:24,440
So that's basically by applying 
technology. 

344
00:17:24,440 --> 00:17:27,160
And also it's all starts first 
with a simple question. 

345
00:17:27,160 --> 00:17:29,960
Am I exposed already? 
And that's a very powerful 

346
00:17:29,960 --> 00:17:32,160
question, I would say, because 
if you'd go ahead and say, you 

347
00:17:32,160 --> 00:17:35,400
know, doesn't happen to me, it's
still a technology that needs to

348
00:17:35,520 --> 00:17:37,640
mature 1st and then maybe it'll 
happen to me. 

349
00:17:37,920 --> 00:17:39,920
And once it happens, then I will
take action. 

350
00:17:40,200 --> 00:17:42,240
We can also do that. 
But it's like I would say it's 

351
00:17:42,240 --> 00:17:45,800
not really a smart way of 
dealing with new types of fraud 

352
00:17:45,800 --> 00:17:48,760
specifically in this age of AI 
and everything automated. 

353
00:17:48,880 --> 00:17:50,880
Yeah, you want to prevent. 
Exactly. 

354
00:17:50,880 --> 00:17:53,320
Yeah, you want to prevent 
monitor mostly as well, because 

355
00:17:53,320 --> 00:17:57,880
cybersecurity on it's own is 
really, it's, it's like it's a 

356
00:17:57,880 --> 00:17:59,800
cat and mouse game. 
You have these fraud, these 

357
00:17:59,800 --> 00:18:02,680
fraudsters, they're also usually
very clever people, also 

358
00:18:02,680 --> 00:18:05,040
collaborative. 
And what's surprising to me 

359
00:18:05,040 --> 00:18:08,400
because one of the most 
prominent reasons that we see so

360
00:18:08,400 --> 00:18:12,440
many defects in banking flows, 
for example, is that on the dark

361
00:18:12,440 --> 00:18:15,520
net, but also on some other 
places and online sources, you 

362
00:18:15,520 --> 00:18:18,920
have groups of people basically 
brainstorming together what 

363
00:18:18,920 --> 00:18:22,480
types of defects, what types of 
material and input material is 

364
00:18:22,480 --> 00:18:26,640
useful to be able to spoof a 
banks identity verification 

365
00:18:26,640 --> 00:18:28,640
system. 
And they basically they. 

366
00:18:29,000 --> 00:18:33,280
Discuss this, you can see this. 
On the darkness, if you know 

367
00:18:33,280 --> 00:18:34,800
which channels. 
You know where to look. 

368
00:18:35,520 --> 00:18:39,200
And once they also find this 
way, they don't stop by only 

369
00:18:39,200 --> 00:18:41,560
using it themselves, but they go
ahead and monetize their 

370
00:18:41,560 --> 00:18:43,440
knowledge. 
Oh shows a business. 

371
00:18:43,440 --> 00:18:46,360
Yeah, it's a real. 
And when I heard this, I was 

372
00:18:46,360 --> 00:18:47,320
like, it was a couple of years 
ago. 

373
00:18:47,320 --> 00:18:50,400
I was like, yeah, I wouldn't buy
deepfakes from these criminal 

374
00:18:50,400 --> 00:18:52,600
people in the darknet, right, 
Because you can give them your 

375
00:18:52,600 --> 00:18:54,920
money, even though it's. 
A scamming other people. 

376
00:18:54,920 --> 00:18:56,600
Exactly. 
They can scam it and you know, 

377
00:18:56,600 --> 00:18:59,080
they can get away with it. 
And I cannot go to the police, 

378
00:19:01,000 --> 00:19:03,400
no customer service, but 
apparently they do have a 

379
00:19:03,400 --> 00:19:06,800
customer service. 
This doesn't make any sense. 

380
00:19:07,160 --> 00:19:09,160
Yeah, it's, it's real. 
It's like for real. 

381
00:19:09,160 --> 00:19:12,800
I I talked to an expert who is 
following also along what we do 

382
00:19:13,480 --> 00:19:16,160
and, and he was like, yeah, if 
the deepfakes don't work for 

383
00:19:16,160 --> 00:19:18,600
your purpose, so you cannot open
this bank account, you can go 

384
00:19:18,680 --> 00:19:21,320
back to them and they will give 
you a new one or they will give 

385
00:19:21,320 --> 00:19:23,280
your money back. 
Man, that's good customer 

386
00:19:23,280 --> 00:19:25,760
service. 
They they know their business. 

387
00:19:27,680 --> 00:19:31,200
So, yeah, but that's basically 
knowing what's happening on the 

388
00:19:31,200 --> 00:19:34,440
other side, being the criminal 
people side. 

389
00:19:35,000 --> 00:19:38,880
And then, yeah, basically trying
to answer this simple question 

390
00:19:38,880 --> 00:19:41,000
of to what extent are we already
exposed? 

391
00:19:41,000 --> 00:19:42,840
What should we protect ourselves
from? 

392
00:19:43,400 --> 00:19:45,920
And once you do that, there is 
already technology that you can 

393
00:19:45,920 --> 00:19:48,400
put in place, like the one that 
we develop, but also other 

394
00:19:48,400 --> 00:19:52,000
things, of course, other methods
that you can apply and to 

395
00:19:52,000 --> 00:19:55,880
basically train your, you know, 
stuff, train your own mindset 

396
00:19:55,880 --> 00:19:58,240
regarding this new reality that 
we have right now, the fake 

397
00:19:58,240 --> 00:20:01,720
reality. 
And yeah, and, and that sense 

398
00:20:01,720 --> 00:20:03,920
when you start monitoring, then 
you see things happening and 

399
00:20:03,920 --> 00:20:07,240
then you can also, once you do 
that, you can preventively stop 

400
00:20:07,240 --> 00:20:09,840
things from happening. 
Instead of letting these people 

401
00:20:09,840 --> 00:20:12,240
make this bank account 1st and 
then letting them put their 

402
00:20:12,240 --> 00:20:16,680
money on their black money 
usually and wait for anti money 

403
00:20:16,680 --> 00:20:18,320
laundering systems to get 
triggered. 

404
00:20:18,320 --> 00:20:19,680
And then you're going to off 
board them. 

405
00:20:19,680 --> 00:20:21,720
You also do a lot of 
investigations before. 

406
00:20:22,120 --> 00:20:25,160
And once this start happening on
on scale, you also have big 

407
00:20:25,160 --> 00:20:28,800
problems with, of course, 
legislation as a bank, taking 

408
00:20:28,800 --> 00:20:33,120
that as an example. 
Yeah, and and really high fines 

409
00:20:33,120 --> 00:20:34,960
basically. 
So in that sense, your 

410
00:20:34,960 --> 00:20:37,160
reputation is also in danger, 
your own. 

411
00:20:37,800 --> 00:20:40,360
Yeah, systems and the flows are 
in danger. 

412
00:20:41,040 --> 00:20:43,920
It's a very interesting and also
very king, king. 

413
00:20:43,920 --> 00:20:46,880
It's very dark because we heard 
that it's happening in South 

414
00:20:46,880 --> 00:20:48,960
America a lot as well in Europe 
too. 

415
00:20:48,960 --> 00:20:51,680
But in South America, they are 
really creative when it comes to

416
00:20:51,680 --> 00:20:56,160
using deep fakes and the cartels
and mafia groups they're 

417
00:20:56,160 --> 00:20:59,440
basically use it for. 
So once they have this, this 

418
00:20:59,440 --> 00:21:02,320
golden way of opening accounts 
based on non traceable DPEC 

419
00:21:02,320 --> 00:21:04,440
identities, they make hundreds 
of them. 

420
00:21:04,440 --> 00:21:07,640
And then every account they will
put a small amount of their 

421
00:21:07,640 --> 00:21:10,960
black money, which is basically 
millions or billions even to 

422
00:21:10,960 --> 00:21:13,760
make sure that the AML systems 
are not getting triggered. 

423
00:21:13,760 --> 00:21:15,680
And it looks like a normal bank 
account. 

424
00:21:15,680 --> 00:21:18,400
It looks like normal bit of a 
couple of thousands of money 

425
00:21:18,400 --> 00:21:21,280
being put on it and then they 
use it for human trafficking 

426
00:21:21,280 --> 00:21:24,920
purposes, for supporting 
terrorism purposes, which is of 

427
00:21:24,920 --> 00:21:26,960
course, really, yeah, a really 
bad thing, I would say. 

428
00:21:27,320 --> 00:21:30,080
Yeah. 
It's like, I agree. 

429
00:21:30,080 --> 00:21:33,240
It has become way more 
accessible than it used to be. 

430
00:21:33,280 --> 00:21:35,000
Yeah. 
And I think if I were to find 

431
00:21:35,000 --> 00:21:36,880
and like create a video that 
looks real, I could probably 

432
00:21:36,880 --> 00:21:38,360
achieve it like as a means to an
end. 

433
00:21:38,360 --> 00:21:41,760
I could do that. 
And then it's the looking from 

434
00:21:41,760 --> 00:21:43,280
the other side. 
How do you prevent that? 

435
00:21:43,280 --> 00:21:45,520
Because accessibility is huge 
now. 

436
00:21:45,520 --> 00:21:48,720
Everyone can do things and 
people with questionable morals 

437
00:21:48,720 --> 00:21:51,040
will do questionable things. 
And then you're there. 

438
00:21:51,040 --> 00:21:52,800
You already mentioned there's 
stuff being discussed in the 

439
00:21:52,800 --> 00:21:56,480
dark web, so you also talk to 
experts and you can also do kind

440
00:21:56,480 --> 00:21:59,960
of your own investigation. 
Is that usually how you think of

441
00:21:59,960 --> 00:22:02,080
the perspective that these 
people have? 

442
00:22:02,360 --> 00:22:04,960
People with questionable morals.
The fraudsters. 

443
00:22:04,960 --> 00:22:07,680
The people that are scam 
scamming other people. 

444
00:22:08,680 --> 00:22:10,200
Yeah, I'm not sure what they're 
thinking. 

445
00:22:10,200 --> 00:22:14,680
It's basically, yeah, it's 
another way, of course utilizing

446
00:22:14,680 --> 00:22:17,960
technology to for your own 
purposes, which might seem good 

447
00:22:17,960 --> 00:22:20,920
to you, but of course the person
that got scammed, they are not 

448
00:22:20,920 --> 00:22:24,120
happy for sure. 
So in that sense they are. 

449
00:22:24,400 --> 00:22:27,440
Technology is open source and 
it's basically a neutral thing. 

450
00:22:27,440 --> 00:22:31,720
So you have a lot of deep fake 
tools out there, applications 

451
00:22:31,720 --> 00:22:33,600
that you can just download. 
You can use it for funny 

452
00:22:33,600 --> 00:22:37,480
purposes, but you can also try 
to use the same tool or try to 

453
00:22:37,480 --> 00:22:41,800
use the basically the tool maybe
in a modified way for your own 

454
00:22:41,800 --> 00:22:44,240
bad purposes. 
So about that, we cannot do 

455
00:22:44,240 --> 00:22:47,960
anything I would say drastically
effective. 

456
00:22:48,440 --> 00:22:50,560
We cannot really prevent. 
That was actually one of the 

457
00:22:50,560 --> 00:22:53,160
aspects that a couple of years 
ago the Dutch government 

458
00:22:53,160 --> 00:22:59,280
contacted us about saying we are
investigating ways to be able to

459
00:22:59,760 --> 00:23:02,400
prevent defect technology 
online. 

460
00:23:03,080 --> 00:23:07,600
And my Co founder and I were 
like, yeah, even though you've 

461
00:23:07,680 --> 00:23:09,680
if you would do that, how are 
you going to enforce that? 

462
00:23:09,680 --> 00:23:12,920
You know, how are you going to 
ban all those applications, all 

463
00:23:12,920 --> 00:23:15,800
those websites, all those open 
source code that is being put on

464
00:23:15,800 --> 00:23:19,320
GitHub? 
So it's basically not doable. 

465
00:23:20,160 --> 00:23:23,240
And when it comes to specific 
dangerous use cases, of course 

466
00:23:23,240 --> 00:23:24,560
there are there are things that 
you can do. 

467
00:23:24,560 --> 00:23:26,880
And also, Patrick, one, one of 
the things that we see is that 

468
00:23:27,200 --> 00:23:30,160
it's still a new phenomenon and 
people are trying to kind of 

469
00:23:30,440 --> 00:23:34,320
figure out how to deal with it. 
But also sometimes, actually, 

470
00:23:34,320 --> 00:23:37,680
usually we see companies that 
want to take measures. 

471
00:23:37,680 --> 00:23:40,320
They're like, yeah, we actually 
want to do it, but there's no 

472
00:23:40,320 --> 00:23:44,640
legislation yet that kind of, 
you know, motivates us more to 

473
00:23:44,640 --> 00:23:48,280
be able to go ahead with, for 
example, putting technology in 

474
00:23:48,280 --> 00:23:50,240
place to be able to detect 
deepfakes. 

475
00:23:50,920 --> 00:23:55,240
On the other side, legislation 
is really slow when it comes to 

476
00:23:55,240 --> 00:23:59,320
coming with new aspects for, you
know, protection regarding new 

477
00:23:59,320 --> 00:24:02,360
technologies like deepfakes. 
And then in the meantime, we 

478
00:24:02,360 --> 00:24:05,920
have normal citizens. 
So not the private company world

479
00:24:05,920 --> 00:24:09,720
anymore, but normal citizens 
also suffering once a teenage 

480
00:24:09,720 --> 00:24:13,120
girl, you know, gets deep faked 
in front of her whole school and

481
00:24:13,120 --> 00:24:15,720
then they eventually end up 
depressed and Oh yeah, 

482
00:24:15,720 --> 00:24:19,720
humiliated basically. 
And the question that since is 

483
00:24:20,040 --> 00:24:22,160
where do you start investigating
this problem? 

484
00:24:22,160 --> 00:24:25,360
And I would say us having be 
having been in this space for 

485
00:24:25,360 --> 00:24:28,160
five years, it's actually a 
problem that you would need to 

486
00:24:28,160 --> 00:24:32,800
collaborate a lot on together to
first of all identify the scope 

487
00:24:32,840 --> 00:24:36,280
in each vertical in each place 
of the society. 

488
00:24:36,640 --> 00:24:39,880
And then see how together with 
of course, education and 

489
00:24:39,880 --> 00:24:43,160
technology used in the right 
place, we can come up with a new

490
00:24:43,160 --> 00:24:45,600
framework to deal with this new 
fake reality. 

491
00:24:45,600 --> 00:24:49,680
As I just said, I was also 
seeing some deep fakes lately 

492
00:24:49,680 --> 00:24:52,280
with all these wars going on, 
you know, around the world. 

493
00:24:53,360 --> 00:24:56,120
One of the things of course, 
that we need to also ensure is 

494
00:24:56,120 --> 00:24:58,640
the integrity of news and media,
right? 

495
00:24:58,640 --> 00:25:01,920
So we see videos, you know, 
coming on coming from Iran, 

496
00:25:02,200 --> 00:25:04,920
where I'm also originally from, 
by the way, with this war that 

497
00:25:04,920 --> 00:25:08,800
was there, the 12 day, the 12 
day war or, or Gaza or African 

498
00:25:08,800 --> 00:25:12,000
countries that are also in war. 
And then you see, like I saw 

499
00:25:12,000 --> 00:25:14,840
literally a couple of deep fakes
being published from these 

500
00:25:14,840 --> 00:25:19,640
bombings in Tehran, in Iran, 
only because it was the deep 

501
00:25:19,640 --> 00:25:23,080
fake basically was supporting 
the framing of this news 

502
00:25:23,080 --> 00:25:25,840
channel. 
And it got me really thinking 

503
00:25:25,840 --> 00:25:29,280
like if these news channels that
all have their own framings, I 

504
00:25:29,280 --> 00:25:32,520
was following different channels
when this 12th day war was 

505
00:25:32,520 --> 00:25:35,840
ongoing between Iran and Israel 
because I wanted to get a lot of

506
00:25:35,840 --> 00:25:37,320
information. 
I still have a lot of family 

507
00:25:37,320 --> 00:25:39,560
living in Iran. 
They're all good, by the way, 

508
00:25:39,560 --> 00:25:43,200
for now. 
I saw clearly the difference 

509
00:25:43,200 --> 00:25:47,360
between a lot of media channels 
that I was following and then 

510
00:25:47,360 --> 00:25:50,400
them using a couple of them, two
of them using deepfakes on their

511
00:25:50,400 --> 00:25:53,600
social media and under other 
channels to basically support 

512
00:25:53,600 --> 00:25:56,480
their own point without 
questioning even, is it real or 

513
00:25:56,480 --> 00:25:57,920
is it not? 
I'm not sure if they made deep 

514
00:25:57,920 --> 00:26:00,760
fix themselves, maybe not, but 
they published it and they do 

515
00:26:00,760 --> 00:26:03,400
have a big audience. 
And that got me really thinking 

516
00:26:03,400 --> 00:26:05,760
that. 
Yeah, it's basically weaponized.

517
00:26:05,760 --> 00:26:08,240
You know, it's, it's they can 
weaponize it to basically also 

518
00:26:08,280 --> 00:26:12,000
achieve bigger purposes to. 
Yeah, mislead people. 

519
00:26:12,000 --> 00:26:13,560
Yeah. 
Spread less information. 

520
00:26:13,680 --> 00:26:15,560
Go back to like proper 
propaganda. 

521
00:26:15,640 --> 00:26:16,680
Exactly. 
Yeah. 

522
00:26:16,680 --> 00:26:19,160
And this, this got me thinking 
like if you don't do anything 

523
00:26:19,160 --> 00:26:22,680
properly about it, soon our 
reality will become everything 

524
00:26:22,680 --> 00:26:27,840
that we see digitally is fake 
unless the otherwise has been 

525
00:26:27,840 --> 00:26:30,680
proven. 
The the, the contradictory has 

526
00:26:30,680 --> 00:26:32,680
been proven. 
And then that will be it's not 

527
00:26:32,680 --> 00:26:35,480
fake, it's real. 
But again, that's also a 

528
00:26:35,480 --> 00:26:37,600
complicated question. 
So how would you prove if 

529
00:26:37,600 --> 00:26:39,960
something is real or not? 
Yeah, this is getting really 

530
00:26:39,960 --> 00:26:43,040
scary. 
Like pre Internet era, you had 

531
00:26:43,040 --> 00:26:45,800
the newspaper and there wasn't 
as much global knowledge. 

532
00:26:45,800 --> 00:26:48,240
Whatever was in your newspaper 
was the truth. 

533
00:26:48,440 --> 00:26:51,600
And then it's very easy to have 
one source of information and 

534
00:26:51,600 --> 00:26:53,240
then have propaganda there as 
well. 

535
00:26:53,680 --> 00:26:56,600
Then it's decentralized because 
with the Internet, I can go on 

536
00:26:56,600 --> 00:26:58,600
Twitter and I see people posting
things. 

537
00:26:58,960 --> 00:27:00,600
I'm like, well, that's a 
different lens, right? 

538
00:27:00,600 --> 00:27:03,000
And those people are more, maybe
more accurate and maybe more 

539
00:27:03,000 --> 00:27:05,080
there. 
And now with general 

540
00:27:05,080 --> 00:27:08,600
availability of Jeep deep fakes 
and it being more accessible, 

541
00:27:09,160 --> 00:27:10,480
there's going to be a lot of 
noise out there. 

542
00:27:10,920 --> 00:27:13,360
Like worse than the truth. 
I think it's going to be 

543
00:27:13,440 --> 00:27:15,480
incredibly scary. 
If you're working on 

544
00:27:15,640 --> 00:27:19,400
cybersecurity, It's not a domain
I'm I'm very much involved in 

545
00:27:19,400 --> 00:27:21,880
nor aware of. 
But for any technology that you 

546
00:27:21,880 --> 00:27:25,080
use that is kind of 
preventative, these fraudsters 

547
00:27:25,080 --> 00:27:27,480
are like your audience. 
And I feel like you need to know

548
00:27:27,480 --> 00:27:30,200
your audience to be able to 
reenact a scenario. 

549
00:27:30,200 --> 00:27:32,880
If you're developing a piece of 
software, you can't really user 

550
00:27:32,880 --> 00:27:34,800
test with fraudsters. 
It's like, yeah, you have to 

551
00:27:34,800 --> 00:27:37,200
know your audience. 
How do you develop something 

552
00:27:37,200 --> 00:27:38,960
that is preventative in nature 
like this? 

553
00:27:39,000 --> 00:27:40,880
Yeah, that's that's a great 
question. 

554
00:27:41,720 --> 00:27:45,440
At Doctor Goose, we do have a 
lot of red teaming activity. 

555
00:27:45,440 --> 00:27:49,520
So it's basically us trying to 
think as these criminals and. 

556
00:27:49,520 --> 00:27:51,000
Is that what red teaming is 
thinking? 

557
00:27:51,000 --> 00:27:51,640
Yeah, OK. 
Yeah. 

558
00:27:51,760 --> 00:27:53,760
And then you kind of try to 
attack, you know, the defence 

559
00:27:53,760 --> 00:27:56,560
systems based on that. 
And then we have also our blue 

560
00:27:56,560 --> 00:27:58,920
team that tries to, you know, 
it's it's basically this kind of

561
00:27:58,920 --> 00:28:00,000
mouse game. 
Gotcha. 

562
00:28:00,160 --> 00:28:02,680
Yeah. 
So that helps a lot because we 

563
00:28:02,680 --> 00:28:05,520
really need to understand the 
mindset of the people who have 

564
00:28:05,520 --> 00:28:08,520
intentions to attack, you know, 
some system or some person, some

565
00:28:08,520 --> 00:28:11,200
company. 
Otherwise you just blindly 

566
00:28:11,200 --> 00:28:12,960
develop technology. 
And then once you put it in 

567
00:28:12,960 --> 00:28:15,080
production, you'll see, oh, it's
not working. 

568
00:28:15,720 --> 00:28:17,960
And of course, a lot of 
monitoring, as I said at the 

569
00:28:17,960 --> 00:28:20,560
beginning, that would help too. 
Because this threat intelligence

570
00:28:20,560 --> 00:28:23,880
that you're gaining right now 
comes partly from indeed these 

571
00:28:23,880 --> 00:28:26,120
conversations that we see 
happening in the dark net and 

572
00:28:26,120 --> 00:28:27,960
these great customer services 
that they have. 

573
00:28:28,320 --> 00:28:30,360
But also in production 
environments, once we are 

574
00:28:30,360 --> 00:28:33,960
monitoring data and indeed 
trying to identify defect types 

575
00:28:33,960 --> 00:28:36,760
that are being used to make 
these defects that we are seeing

576
00:28:37,200 --> 00:28:42,760
in production, that helps a lot 
in basically trying to adapt our

577
00:28:42,960 --> 00:28:46,080
strategy as well when it comes 
to designing our defenses. 

578
00:28:46,320 --> 00:28:48,840
Yeah. 
Which topics do you focus on? 

579
00:28:48,840 --> 00:28:50,240
Because you mentioned a lot of 
things, right? 

580
00:28:50,240 --> 00:28:52,520
It can go from large scale fraud
in banking. 

581
00:28:52,880 --> 00:28:55,920
And I've also seen many posts 
online because getting a job in 

582
00:28:55,920 --> 00:28:58,000
tech is now harder. 
People are being laid off, 

583
00:28:58,560 --> 00:29:01,800
people that are graduates are 
finding it harder to get their 

584
00:29:01,800 --> 00:29:04,600
first job. 
Might even automate something 

585
00:29:04,600 --> 00:29:08,320
with AI to apply to jobs and 
like fake a voice thing and get 

586
00:29:08,320 --> 00:29:11,000
through a first screening call. 
I feel like all of it's fraud. 

587
00:29:11,720 --> 00:29:12,640
That's. 
Happening already. 

588
00:29:12,640 --> 00:29:13,960
That's happening already as well
there. 

589
00:29:13,960 --> 00:29:17,760
Was last year, I think, or the 
year before even this American 

590
00:29:17,760 --> 00:29:20,880
cybersecurity company, you know,
before they published A blog 

591
00:29:20,960 --> 00:29:24,920
written by the CEO saying that 
they mistakenly on boarded a 

592
00:29:24,920 --> 00:29:28,520
North Korean spy as a new remote
employee. 

593
00:29:28,600 --> 00:29:29,560
Yeah, big whoops. 
Yeah. 

594
00:29:29,800 --> 00:29:32,560
Yeah, exactly. 
And it's really nice that 

595
00:29:32,560 --> 00:29:36,880
they're being open about it. 
And they also, right before this

596
00:29:36,880 --> 00:29:39,680
person was about to do these 
fraudulent activities, they 

597
00:29:39,680 --> 00:29:41,080
caught the person. 
So that was good. 

598
00:29:41,960 --> 00:29:45,400
But in the whole process, the 
person used a deep fake face, 

599
00:29:45,400 --> 00:29:48,400
apparently to be in those Zoom 
calls or Teams calls. 

600
00:29:48,480 --> 00:29:50,800
And I don't know which platform 
because it's possible with all 

601
00:29:50,800 --> 00:29:52,600
of those platforms. 
Yeah, it doesn't matter. 

602
00:29:52,600 --> 00:29:54,160
I even have a deep fake mask. 
And. 

603
00:29:54,400 --> 00:29:56,840
Yeah, I don't have it with me 
right now, but we wait. 

604
00:29:56,920 --> 00:29:59,920
What does that mean? 
It's basically if I would put 

605
00:29:59,920 --> 00:30:02,520
you in a Teams call right now, 
you can get my face and then. 

606
00:30:02,560 --> 00:30:05,160
Oh, really? 
Can we do that after I I would 

607
00:30:05,160 --> 00:30:07,240
love to see that. 
Yeah, 'cause I'm gonna call, I I

608
00:30:07,320 --> 00:30:11,800
need to call a colleague. 
We can even like if if we take 

609
00:30:11,800 --> 00:30:14,320
your face right now from this 
video recording on YouTube later

610
00:30:14,320 --> 00:30:17,560
and we can also train a model 
that is able to mimic your face 

611
00:30:17,560 --> 00:30:20,840
in a Zoom or Teams call, OK. 
So I might do, I might do a 

612
00:30:20,840 --> 00:30:24,240
separate clip, yeah. 
So look at the camera as well 

613
00:30:25,720 --> 00:30:26,840
for 10. 
Seconds setting up, yeah. 

614
00:30:28,880 --> 00:30:29,920
Wow. 
So yeah, it's happening already 

615
00:30:29,920 --> 00:30:32,440
in terms of remote job 
applications. 

616
00:30:32,440 --> 00:30:35,160
There is even like a whole North
Korean, I think group of people 

617
00:30:35,160 --> 00:30:37,120
doing this, attacking American 
companies. 

618
00:30:37,520 --> 00:30:40,440
And in the case of no, before, 
they only found about it once 

619
00:30:40,440 --> 00:30:43,600
they sent them, send this person
the hardware of their company so

620
00:30:43,600 --> 00:30:46,240
he could start working. 
And then they saw that he is 

621
00:30:46,240 --> 00:30:48,800
trying to inject malware 
basically into the system and 

622
00:30:48,800 --> 00:30:50,560
that that's the point where they
blocked it. 

623
00:30:51,640 --> 00:30:53,000
So yeah, it happens already a 
lot. 

624
00:30:53,000 --> 00:30:56,840
And if you have like a company 
that is sensitive to lose 

625
00:30:56,840 --> 00:31:01,080
information like ASMR or other 
companies, let's say, and there 

626
00:31:01,080 --> 00:31:03,880
is like a shortage of stuff as 
well, so you would tend to maybe

627
00:31:03,880 --> 00:31:07,880
remotely hire people, then yeah,
there is a good chance that you 

628
00:31:07,880 --> 00:31:09,400
have a deep. 
And also to answer your 

629
00:31:09,400 --> 00:31:13,640
question, where do we focus on 
right now? 

630
00:31:13,640 --> 00:31:16,400
It's for us, it's mostly the 
fintech and the financial use 

631
00:31:16,400 --> 00:31:19,520
cases along with video 
conferencing as well. 

632
00:31:19,520 --> 00:31:24,120
So and, and that's mostly also 
for companies that are required 

633
00:31:24,120 --> 00:31:27,360
and also desirable to have high 
security measures basically in 

634
00:31:27,360 --> 00:31:29,880
place. 
On the other side, if there are 

635
00:31:29,880 --> 00:31:32,360
requests such as this person 
that I mentioned calling us 

636
00:31:32,360 --> 00:31:35,200
saying I have a girlfriend, can 
you tell me if she's real or 

637
00:31:35,200 --> 00:31:36,560
not? 
We also do that. 

638
00:31:36,560 --> 00:31:39,200
We have like a specific team 
taking a look at those requests 

639
00:31:39,200 --> 00:31:41,560
as well. 
The ultimate goal that Duck Duck

640
00:31:41,560 --> 00:31:44,680
Goose has is to become and in 
three to five years become a 

641
00:31:44,680 --> 00:31:48,160
platform that everyone, not only
banks, but everyone can use 

642
00:31:48,160 --> 00:31:51,400
basically to go there to see if 
the video, the picture, the 

643
00:31:51,400 --> 00:31:54,120
audio that they are seeing or 
hearing digitally is real or 

644
00:31:54,120 --> 00:31:56,520
not. 
And that also includes Citizens,

645
00:31:56,680 --> 00:31:59,440
yeah. 
Yeah, I feel like everyone now 

646
00:31:59,520 --> 00:32:02,920
in especially in tech, in kind 
of the onboarding process is 

647
00:32:02,920 --> 00:32:05,360
dealing with some form of AI 
indeed. 

648
00:32:05,360 --> 00:32:07,920
It could be someone that's not 
really the real person. 

649
00:32:07,920 --> 00:32:10,800
It could be someone whose voice 
is different, could be someone 

650
00:32:10,800 --> 00:32:13,720
that's not even that same person
and just another person doing 

651
00:32:13,720 --> 00:32:15,840
the onboarding thing, which is 
not even really AI. 

652
00:32:16,320 --> 00:32:19,080
It could be something that 
someone uses with regards to 

653
00:32:19,440 --> 00:32:21,840
kind of asking a question real 
quick to AI and then reading 

654
00:32:21,840 --> 00:32:23,960
from a screen or a script or 
something like that. 

655
00:32:24,320 --> 00:32:28,280
I feel like onboarding is right 
now already being kind of 

656
00:32:28,280 --> 00:32:31,360
compromised. 
Let's say XEBIA, we have a eight

657
00:32:31,360 --> 00:32:34,480
hour, 8 hour assessment. 
It used to be 8 hours. 

658
00:32:34,520 --> 00:32:37,640
Likely with AI, it's no longer. 
And we were like, well, should 

659
00:32:37,640 --> 00:32:40,760
we prevent people from using AI 
because we do use AI to be more 

660
00:32:40,760 --> 00:32:41,600
productive? 
Yeah. 

661
00:32:41,720 --> 00:32:45,960
So we were like, okay, we will 
allow people to use AI, but we 

662
00:32:45,960 --> 00:32:48,640
will kind of frame our questions
and then that interview round 

663
00:32:48,640 --> 00:32:50,680
with the assumption that people 
have used AI. 

664
00:32:51,160 --> 00:32:53,960
And if they use AI with a 
certain technology, we already 

665
00:32:53,960 --> 00:32:56,280
have pre prompted kind of our 
prompts. 

666
00:32:56,560 --> 00:32:59,360
So we can say tailor it to this 
technology and we can then 

667
00:32:59,360 --> 00:33:03,400
compare outputs because then we 
can see if our AI has this, this

668
00:33:03,400 --> 00:33:05,480
and this, it takes us X amount 
of hours. 

669
00:33:05,800 --> 00:33:07,520
This person has delivered this 
in eight hours. 

670
00:33:07,520 --> 00:33:10,040
What are the differences? 
And then we specifically can ask

671
00:33:10,040 --> 00:33:12,760
questions on those, but that's 
because we're aware of this. 

672
00:33:12,760 --> 00:33:14,240
We're already changing our 
onboarding. 

673
00:33:14,600 --> 00:33:18,000
I feel like if you're not aware 
then even kind of changing your 

674
00:33:18,000 --> 00:33:21,440
onboarding is not there. 
Yeah, yeah, that's that, that's 

675
00:33:21,440 --> 00:33:25,400
very true. 
I would say trying to yeah, ban 

676
00:33:25,520 --> 00:33:27,880
AI usage or tell people you're 
not allowed to use it. 

677
00:33:27,880 --> 00:33:32,160
Also, even like students at 
universities or or in I I have a

678
00:33:32,160 --> 00:33:35,640
cousin, she's 11 even she uses 
she lives in Paris, she uses 

679
00:33:35,640 --> 00:33:38,760
chachi PT for her homework. 
So I would say it's, it's like 

680
00:33:38,760 --> 00:33:42,480
maybe the time that these 
digital calculators, we could 

681
00:33:42,480 --> 00:33:44,240
also say to people like don't 
use it. 

682
00:33:44,240 --> 00:33:47,120
It's I would say it's a tool 
that we can use to become more 

683
00:33:47,120 --> 00:33:49,880
efficient. 
Right now, if you use chat or 

684
00:33:49,880 --> 00:33:53,360
cloud or perplexity, it cannot 
write a good quality report for 

685
00:33:53,360 --> 00:33:56,280
you on a specific topic, being 
cybersecurity when it comes to 

686
00:33:56,280 --> 00:33:56,880
Gen. 
AI. 

687
00:33:58,400 --> 00:34:02,160
But you can use it as something 
much faster than Google to come 

688
00:34:02,160 --> 00:34:04,360
up with relevant sources when it
comes to perplexity. 

689
00:34:04,360 --> 00:34:08,760
For example, to have your, you 
know, database of, you know, 

690
00:34:08,920 --> 00:34:10,520
papers knowledge that you want 
to gather. 

691
00:34:10,520 --> 00:34:13,320
And then based on that, you can 
maybe even ask ChatGPT to 

692
00:34:13,320 --> 00:34:16,199
summarize it for you because I, 
I think ChatGPT is better in 

693
00:34:16,199 --> 00:34:17,840
summarizing that than 
perplexity. 

694
00:34:18,840 --> 00:34:22,159
You can use those insights to 
make your own work faster, but 

695
00:34:22,159 --> 00:34:24,600
it wouldn't really. 
I went with you right now with 

696
00:34:24,600 --> 00:34:27,159
this stage of, you know, that 
the quality that they have trust

697
00:34:27,159 --> 00:34:31,440
in AI to write an essay for me 
or write a code for me. 

698
00:34:31,440 --> 00:34:36,360
Basically right now we do use a 
lot of AI for more like indeed 

699
00:34:36,360 --> 00:34:40,920
what you said, junior level of 
programming, cloud mostly it's 

700
00:34:40,920 --> 00:34:43,800
like I think €160.00 a month for
the professional version and you

701
00:34:43,800 --> 00:34:47,120
have like a kind of a junior, 
you know, AI assistant there. 

702
00:34:48,440 --> 00:34:51,000
The code is suitable for, for 
example, front end. 

703
00:34:51,159 --> 00:34:53,159
But if you want to use it for 
the back end, you shouldn't 

704
00:34:53,159 --> 00:34:56,679
actually unless you have a 
senior person, for example, like

705
00:34:56,679 --> 00:35:01,360
we have in the team that looks 
really at the code to see, you 

706
00:35:01,360 --> 00:35:04,320
know, if everything is good and 
OK and scalable for later. 

707
00:35:04,320 --> 00:35:07,240
If you want to change things, is
it in good quality and 

708
00:35:07,240 --> 00:35:09,920
changeable later and then deploy
that code. 

709
00:35:10,240 --> 00:35:13,080
So it's kind of basically there 
to make things faster, but 

710
00:35:13,120 --> 00:35:16,320
definitely not to replace, I 
would say for now. 

711
00:35:16,880 --> 00:35:19,400
On the other hand, I also heard,
for example, the BIG4 have 

712
00:35:19,400 --> 00:35:23,120
always been really targeting 
fresh graduates and even like 

713
00:35:23,120 --> 00:35:28,480
interns, graduate interns fresh 
out of the university right now.

714
00:35:28,480 --> 00:35:31,120
Apparently what I read is that 
in the UK they completely 

715
00:35:31,120 --> 00:35:34,760
stopped doing it because of AI. 
They're like junior work can be 

716
00:35:34,760 --> 00:35:37,840
done by AI. 
It sounds scary. 

717
00:35:38,640 --> 00:35:43,040
It's also maybe for us as human 
beings, like, oh, my job is not 

718
00:35:43,040 --> 00:35:44,680
necessarily, you know, needed 
anymore. 

719
00:35:45,440 --> 00:35:49,320
On the other hand, I think it's 
an opportunity for us to become 

720
00:35:49,320 --> 00:35:51,600
better in other things. 
I mean, the same thing also 

721
00:35:51,600 --> 00:35:54,400
happened in like 1800s, right? 
With the industry revolution. 

722
00:35:54,400 --> 00:35:57,840
You had machines, they were 
faster and it was like the 

723
00:35:57,840 --> 00:36:00,240
people are kind of only serving 
these machines. 

724
00:36:00,520 --> 00:36:03,280
But then we got the chance to 
become better in technology, 

725
00:36:03,400 --> 00:36:05,840
right? 
We had then we started making 

726
00:36:05,840 --> 00:36:08,560
these tools and, and right now 
look, we are where we are and 

727
00:36:08,560 --> 00:36:10,640
without those machines, it 
wouldn't maybe have been 

728
00:36:10,640 --> 00:36:14,120
possible because we would still 
be doing, you know, a lot of job

729
00:36:14,120 --> 00:36:17,360
with our hands and, and, and not
not even out of my. 

730
00:36:17,360 --> 00:36:20,640
So I compare it to the same 
basically what happened also 200

731
00:36:20,640 --> 00:36:24,680
years ago, but then the scale is
much faster and it's that's 

732
00:36:24,680 --> 00:36:27,360
maybe why people are scared. 
Yeah, I can see that. 

733
00:36:27,440 --> 00:36:30,160
Yeah. 
I mean, I think so with digital,

734
00:36:30,160 --> 00:36:32,720
we had digital natives. 
And I feel like with generative 

735
00:36:32,720 --> 00:36:36,240
AI, you can have people that are
very well equipped that are not 

736
00:36:36,560 --> 00:36:39,720
afraid to try out new tooling 
that can be very effective maybe

737
00:36:39,720 --> 00:36:42,920
from a learning and education 
standpoint and can get up to a 

738
00:36:42,920 --> 00:36:45,960
certain level that they need to 
be effective very, very quickly.

739
00:36:45,960 --> 00:36:48,560
Or they maybe don't have the 
same depth, but they can go 

740
00:36:48,560 --> 00:36:51,040
really wide and they know a lot 
about a lot of topics. 

741
00:36:51,600 --> 00:36:55,120
I feel like now more than ever, 
the way to distinguish yourself 

742
00:36:55,120 --> 00:36:57,680
if you're starting out, if 
you're early in career, is to 

743
00:36:57,680 --> 00:37:00,600
indeed find your skill set, find
what you're really good at, and 

744
00:37:00,600 --> 00:37:03,920
leverage whatever tooling is out
there to distinguish yourself 

745
00:37:03,920 --> 00:37:06,080
from others. 
It's just, yeah, there's a lot 

746
00:37:06,080 --> 00:37:09,000
of people and the technology is 
ever accessible. 

747
00:37:09,680 --> 00:37:13,080
But I do think we need to create
software and all companies will 

748
00:37:13,080 --> 00:37:16,320
still live in a digital era to 
create things. 

749
00:37:16,800 --> 00:37:18,720
And likely, this is kind of my 
hope. 

750
00:37:18,960 --> 00:37:20,840
A lot of companies are kind of 
creating the same thing, 

751
00:37:20,840 --> 00:37:23,000
slightly different and making 
money with it. 

752
00:37:23,360 --> 00:37:26,120
I hope it also enables people to
create new things and unique 

753
00:37:26,120 --> 00:37:27,880
things and innovate. 
Yeah, yeah. 

754
00:37:28,200 --> 00:37:30,680
Then I think companies will be 
unstoppable. 

755
00:37:30,680 --> 00:37:33,600
They're just developing things. 
You have anthropic open air, all

756
00:37:33,600 --> 00:37:36,920
of them. 
Google, Microsoft, then the next

757
00:37:36,920 --> 00:37:39,720
big thing that the Western world
is talking about is artificial 

758
00:37:39,720 --> 00:37:42,760
general intelligence. 
Of course, that's like 

759
00:37:42,840 --> 00:37:46,040
artificial intelligence that is 
not only statistical models, you

760
00:37:46,040 --> 00:37:49,000
know, but human. 
Like, I'm not sure what you 

761
00:37:49,000 --> 00:37:53,360
think about that. 
But yeah, I would say that that 

762
00:37:53,360 --> 00:37:55,400
could be that's something that 
personally scares me. 

763
00:37:55,400 --> 00:37:58,040
So if you have, I saw a video on
LinkedIn, I think it was 

764
00:37:58,040 --> 00:38:01,040
yesterday or day before there 
was this robot in a Chinese 

765
00:38:01,040 --> 00:38:03,240
factory who was able to replace 
his own battery. 

766
00:38:03,360 --> 00:38:04,160
Oh, OK. 
And it was. 

767
00:38:04,200 --> 00:38:05,440
It was. 
Really like a self surgery, 

768
00:38:05,440 --> 00:38:07,960
yeah. 
Yeah. 

769
00:38:07,960 --> 00:38:12,800
And like, it's maybe like a 
first mini step in basically 

770
00:38:13,880 --> 00:38:16,960
preventing yourself from 
shutting down because of battery

771
00:38:16,960 --> 00:38:18,720
shortage. 
So you replace your own battery.

772
00:38:18,720 --> 00:38:22,920
And then so AGI, if possible, 
ever combined with these 

773
00:38:22,920 --> 00:38:25,520
robotics. 
That scares me personally. 

774
00:38:25,640 --> 00:38:26,520
Yeah. 
What do you think? 

775
00:38:26,920 --> 00:38:32,040
I mean, so in a vacuum, I'm a 
very logical, objective person. 

776
00:38:32,960 --> 00:38:35,880
I don't think we know enough 
about our own intelligence and 

777
00:38:35,880 --> 00:38:38,760
like the human brain, the 
perception of the world, the 

778
00:38:38,760 --> 00:38:41,880
perception of interactions, what
thought actually is, where does 

779
00:38:41,880 --> 00:38:43,680
that live? 
I mean, we know it from a brain 

780
00:38:43,680 --> 00:38:45,520
perspective. 
What does it actually mean? 

781
00:38:45,520 --> 00:38:47,040
What happens when we die and 
stuff like that? 

782
00:38:47,520 --> 00:38:49,720
We are so unaware. 
Like even how the body works. 

783
00:38:49,720 --> 00:38:52,960
The fact that we have placebo is
incredibly interesting. 

784
00:38:53,440 --> 00:38:55,840
So then how can we, the fact 
that we don't know everything, 

785
00:38:55,960 --> 00:38:59,880
create something that will kind 
of supersede intelligence? 

786
00:39:00,000 --> 00:39:02,360
I don't know, like in my head 
that doesn't add up. 

787
00:39:02,760 --> 00:39:05,080
I know right now a lot of models
are just prediction models. 

788
00:39:05,640 --> 00:39:08,960
It's like a huge decision tree. 
I cannot even comprehend the 

789
00:39:08,960 --> 00:39:11,240
decision trees. 
And sure, it might be really 

790
00:39:11,240 --> 00:39:15,760
good at math and I I'm really 
bad at that level of math and 

791
00:39:15,760 --> 00:39:18,560
maybe that's sad, but that's 
also what computers are good at.

792
00:39:18,560 --> 00:39:20,280
Like it makes sense. 
It's very black and white to a 

793
00:39:20,280 --> 00:39:22,480
certain degree. 
And yet prediction models are 

794
00:39:22,480 --> 00:39:26,040
still prediction. 
That's why we have now non 

795
00:39:26,040 --> 00:39:29,800
deterministic code and non non 
deterministic features because 

796
00:39:29,800 --> 00:39:31,000
we don't know what's going to 
come out. 

797
00:39:31,560 --> 00:39:35,040
We can't predict what we have a 
high likelihood, but that's Jen 

798
00:39:35,040 --> 00:39:37,520
AI from me right now. 
So I don't see it evolving to 

799
00:39:37,520 --> 00:39:40,280
that unless proven otherwise. 
Yeah, exactly. 

800
00:39:40,560 --> 00:39:43,280
What about you? 
I I think I mostly agree with 

801
00:39:43,280 --> 00:39:44,880
you. 
I was thinking about this myself

802
00:39:44,880 --> 00:39:47,640
as well. 
The, the like, AG is everywhere.

803
00:39:47,640 --> 00:39:49,640
Everyone. 
I was like, OK, but have you 

804
00:39:49,640 --> 00:39:51,880
ever defined what intelligence 
actually means? 

805
00:39:51,880 --> 00:39:53,880
That you're creating the 
artificial version of it? 

806
00:39:54,400 --> 00:39:57,240
The human brain is so complex. 
I used to listen a lot to Andrew

807
00:39:57,240 --> 00:40:00,840
Huberman. 
And he also, he was, he's a 

808
00:40:00,840 --> 00:40:04,200
neuroscientist trying to make 
information about our brain 

809
00:40:04,200 --> 00:40:06,640
accessible to everyone by 
explaining it in his podcast in 

810
00:40:06,680 --> 00:40:10,320
a more simple way. 
But like, neuroscientists like 

811
00:40:10,320 --> 00:40:12,680
himself and also including 
Andrew Huberman are already 

812
00:40:12,680 --> 00:40:16,520
saying we, we know only so 
little about the human brain and

813
00:40:16,520 --> 00:40:19,240
indeed, per SE, but a lot of 
other things as well. 

814
00:40:19,240 --> 00:40:23,400
So yeah. 
I I don't know how and when we 

815
00:40:23,400 --> 00:40:25,480
will be able to develop 
something that is really human 

816
00:40:25,480 --> 00:40:29,240
like, but if it happens in 
theory, yeah, I'm not sure if 

817
00:40:29,240 --> 00:40:29,960
you'll see it. 
Maybe. 

818
00:40:30,840 --> 00:40:32,720
Maybe after our lifetimes. 
I don't know. 

819
00:40:32,720 --> 00:40:35,080
Yeah, it could be. 
This could also age very poorly.

820
00:40:36,160 --> 00:40:38,800
It's like in a few months it's 
like, oh, we're actually quite 

821
00:40:38,800 --> 00:40:40,120
close. 
Yeah, well, who knows? 

822
00:40:41,280 --> 00:40:43,400
That could be that's that's also
one of the aspects of this 

823
00:40:43,400 --> 00:40:45,000
modern world. 
It's really unpredictable. 

824
00:40:45,360 --> 00:40:47,280
So yeah. 
Yeah, with things that move 

825
00:40:47,600 --> 00:40:51,320
super quickly, Google can 
release something which all of a

826
00:40:51,320 --> 00:40:55,720
sudden creates a new version of 
generative AI in videos, looks 

827
00:40:55,720 --> 00:40:58,680
super realistic. 
Things are moving quite well and

828
00:40:58,680 --> 00:41:01,120
quite fast in a high quality 
manner. 

829
00:41:01,400 --> 00:41:04,240
How do you as a company kind of 
try and manoeuvre all this thing

830
00:41:04,240 --> 00:41:05,840
pops up? 
Do we actually do something for 

831
00:41:05,840 --> 00:41:07,680
this? 
Do we need to focus on what we 

832
00:41:07,680 --> 00:41:10,000
were doing? 
Do we really need to account for

833
00:41:10,000 --> 00:41:11,880
this now? 
How do you make those priority 

834
00:41:11,880 --> 00:41:13,600
calls? 
That's that's a great question. 

835
00:41:13,600 --> 00:41:18,800
So also things like AGI, agentic
AI, like everywhere at every 

836
00:41:18,800 --> 00:41:21,120
event that we go, it's either 
about Gen. 

837
00:41:21,120 --> 00:41:27,160
AI, defects, AGI or yeah, 
agentic AI sometimes also about 

838
00:41:27,160 --> 00:41:29,760
stable coin. 
But it's like, I'm like, is this

839
00:41:29,760 --> 00:41:31,760
like only hype? 
Why is everyone talking about 

840
00:41:31,760 --> 00:41:33,120
it? 
And I would say personally, I 

841
00:41:33,120 --> 00:41:35,520
think right now there are every 
setup that I see out there also 

842
00:41:35,520 --> 00:41:38,360
in the Netherlands is about 
something agentic basically. 

843
00:41:38,880 --> 00:41:41,400
And it could also be because of 
the high. 

844
00:41:41,400 --> 00:41:45,520
But I would say maybe in five 
years, 5, five years time, maybe

845
00:41:45,520 --> 00:41:48,960
like 60% of these HNDK 
applications won't be there 

846
00:41:48,960 --> 00:41:51,040
because they started right now. 
And then there on we'll see how 

847
00:41:51,040 --> 00:41:54,080
the usage is not really that 
optimized and and let's you 

848
00:41:54,080 --> 00:41:56,600
know, let's not do it anymore. 
On the other hand, I heard in 

849
00:41:56,600 --> 00:41:59,760
countries like China, they are 
much more instead of trying out 

850
00:41:59,760 --> 00:42:04,640
these high or or maybe 
glamorized terminology such as 

851
00:42:04,640 --> 00:42:10,680
agentic AI or AGI, they mostly 
are focusing on which parts of 

852
00:42:10,680 --> 00:42:14,320
our society and of our normal 
daily works can we improve using

853
00:42:14,480 --> 00:42:17,640
automation, using indeed 
prediction models instead of, 

854
00:42:17,800 --> 00:42:20,840
you know, going with the hype. 
That's also more the mindset 

855
00:42:20,840 --> 00:42:23,440
that I personally have together 
with our team as well. 

856
00:42:23,840 --> 00:42:27,240
So we are right now, as I said, 
DPLAY technology generate 

857
00:42:27,240 --> 00:42:28,920
technology is evolving really 
fast. 

858
00:42:29,440 --> 00:42:32,840
And the video that you mentioned
just yet with this bird example 

859
00:42:32,840 --> 00:42:36,120
is maybe that's something that 
we should be able to catch in 

860
00:42:36,120 --> 00:42:37,640
five years. 
But right now there is not 

861
00:42:37,640 --> 00:42:40,480
really any harmful activity 
behind it. 

862
00:42:41,680 --> 00:42:44,080
So in that sense, we also 
distinguish a lot of context 

863
00:42:44,080 --> 00:42:47,120
saying OK in in that sense, 
there's a new type of deep fake.

864
00:42:47,360 --> 00:42:48,960
Should we be able to catch it or
not? 

865
00:42:48,960 --> 00:42:50,320
Is it harmful? 
First of all? 

866
00:42:50,640 --> 00:42:53,080
Is it something that is bad for 
our clients and for our 

867
00:42:53,080 --> 00:42:56,200
basically people that are 
trusting Dr. Goose to catch 

868
00:42:56,200 --> 00:42:59,080
their fraudulent AI based 
activities? 

869
00:42:59,920 --> 00:43:02,840
And that's one of the questions 
that we have basically go back 

870
00:43:02,840 --> 00:43:07,640
to a lot when it comes to all 
the other yeah words. 

871
00:43:08,280 --> 00:43:13,360
Yeah, Thank you. 
No, yeah, it kind of indeed I 

872
00:43:13,360 --> 00:43:16,680
get this question a lot like do 
do you, are you scared for 

873
00:43:16,680 --> 00:43:18,480
indeed Google developing 
something? 

874
00:43:19,080 --> 00:43:21,240
I'm not sure. 
Yeah, if they do it, that's like

875
00:43:21,240 --> 00:43:26,720
kind of a maybe validation that 
even more validation that what 

876
00:43:26,720 --> 00:43:29,000
we are doing is really useful at
the the goose. 

877
00:43:29,600 --> 00:43:33,240
And I would also be like it. 
It's kind of a strategy. 

878
00:43:33,240 --> 00:43:35,400
Of course you can also, you 
know, merge with these companies

879
00:43:35,400 --> 00:43:38,040
and do things on a bigger scale.
Right now there are lots of 

880
00:43:38,040 --> 00:43:41,240
investment in the counterpart of
what we are doing being the 

881
00:43:41,240 --> 00:43:43,720
generation part. 
And the detection also of 

882
00:43:43,720 --> 00:43:46,480
course, needs to be there to 
make sure that these generation 

883
00:43:46,480 --> 00:43:49,760
techniques that are out there 
and everyone can use them are 

884
00:43:49,760 --> 00:43:52,720
not really used for, yeah, bad 
purposes. 

885
00:43:52,880 --> 00:43:55,760
Yeah. 
I mean, from a personal sense, 

886
00:43:55,760 --> 00:43:58,520
you work on this 4-5 years ago, 
right? 

887
00:43:59,160 --> 00:44:02,000
No one was talking about it. 
There were probably some videos 

888
00:44:02,000 --> 00:44:04,160
that all of a sudden proved that
oh, this is a thing. 

889
00:44:04,600 --> 00:44:06,560
But now with Jen AI, things have
accelerated. 

890
00:44:06,680 --> 00:44:10,000
As you mentioned, investments in
the fact that we are able to do 

891
00:44:10,000 --> 00:44:12,440
this and that it's accessible 
that I only have to swipe my 

892
00:44:12,440 --> 00:44:15,600
credit card and I could do 
something similar is insane. 

893
00:44:15,920 --> 00:44:19,600
Prevention needs to kind of also
have the same acceleration and 

894
00:44:19,600 --> 00:44:21,920
you were already there and now 
you're in the middle of it. 

895
00:44:22,120 --> 00:44:24,400
How does that feel like? 
Is the team super excited to 

896
00:44:24,400 --> 00:44:26,680
work on these things? 
You don't need to educate 

897
00:44:26,680 --> 00:44:29,080
customers as much. 
It's completely different. 

898
00:44:29,120 --> 00:44:33,480
Yeah, exactly. 
It's it's, yeah, it's, it's 

899
00:44:33,480 --> 00:44:37,120
really nice, I would say. 
So once we started that, we 

900
00:44:37,120 --> 00:44:41,160
started that the Cusoversa at 
the university and of course, it

901
00:44:41,160 --> 00:44:43,040
was like starting part time 
really only. 

902
00:44:44,160 --> 00:44:47,080
And it's for example, also 
caused me to be able to get my 

903
00:44:47,200 --> 00:44:50,640
you know, my my school paper 
later, for example. 

904
00:44:51,040 --> 00:44:55,320
But if I would go back, I would 
choose it like 100 times again 

905
00:44:55,320 --> 00:44:57,400
and again and again. 
It's it's such an incredible 

906
00:44:57,400 --> 00:45:02,360
journey being able to have a 
team of really ambitious people 

907
00:45:03,120 --> 00:45:05,560
together with your Co founders, 
with the other team members 

908
00:45:05,880 --> 00:45:08,800
working on something so 
innovative, so challenging as 

909
00:45:08,800 --> 00:45:10,640
well. 
Because deep detection, I would 

910
00:45:10,640 --> 00:45:13,240
say is maybe one of the most 
challenging problems that we 

911
00:45:13,240 --> 00:45:16,160
have right now in the world of 
AI, computer vision, etcetera, 

912
00:45:16,160 --> 00:45:20,520
cybersecurity and being able to 
work on that, being able to see 

913
00:45:20,520 --> 00:45:23,760
that your product is being used 
by banks basically and you give 

914
00:45:23,760 --> 00:45:27,360
them value. 
It's really, yeah, it's really a

915
00:45:27,360 --> 00:45:29,240
next level feeling for me, I 
would say. 

916
00:45:29,720 --> 00:45:32,760
So I'm really grateful that we 
decided to start Duck Duck Goose

917
00:45:32,760 --> 00:45:34,920
like five years ago from 
University Project. 

918
00:45:36,160 --> 00:45:39,360
And yeah, I, I sometimes also go
still to Tudelph University of 

919
00:45:39,360 --> 00:45:44,680
Technology to talk to students. 
And yeah, I always would say if 

920
00:45:44,680 --> 00:45:47,200
you have this ambition in you 
that you would like to create 

921
00:45:47,200 --> 00:45:49,200
something for yourself, just go 
for it. 

922
00:45:49,200 --> 00:45:50,880
And you can't even start during 
your student time. 

923
00:45:51,160 --> 00:45:53,800
I was lately at a talk where 
also parents were there. 

924
00:45:54,760 --> 00:45:56,920
So I was like, OK, shall I say 
this because parents usually 

925
00:45:56,920 --> 00:46:00,600
don't like their children. 
I was like, OK, I will say it. 

926
00:46:01,280 --> 00:46:02,800
Yeah. 
So there are, I think, 200 

927
00:46:02,800 --> 00:46:05,480
parents plus 200 children. 
To like start a company and 

928
00:46:05,480 --> 00:46:07,720
don't like don't worry about 
school too much. 

929
00:46:07,720 --> 00:46:11,280
You said that. 
Kind of a bit. 

930
00:46:11,400 --> 00:46:12,920
That's quite funny, is. 
It parents. 

931
00:46:12,920 --> 00:46:13,760
Sorry bud. 
Yeah. 

932
00:46:13,760 --> 00:46:16,200
Yeah, I'm gonna say this. 
Yeah, Yeah. 

933
00:46:16,280 --> 00:46:19,000
So I would say it's really 
incredible and there are of 

934
00:46:19,000 --> 00:46:21,520
course, ups and downs. 
It's difficult, it's 

935
00:46:21,520 --> 00:46:26,320
challenging, but I sometimes or 
usually enjoy solving those 

936
00:46:26,320 --> 00:46:28,560
challenges. 
Like there was a time two years 

937
00:46:28,560 --> 00:46:31,560
ago, three years ago that 
everything was quite stable. 

938
00:46:31,640 --> 00:46:33,560
You know, things went OK. 
It was good it. 

939
00:46:33,680 --> 00:46:34,320
Was before Gen. 
AI. 

940
00:46:34,320 --> 00:46:35,840
Yeah, exactly. 
Yeah, yeah. 

941
00:46:36,520 --> 00:46:39,240
Before things exploded. 
Before things started exploding 

942
00:46:39,240 --> 00:46:42,160
exactly and I was like you know,
I would like to have, you know, 

943
00:46:42,160 --> 00:46:43,720
a problem, I would like to solve
a problem. 

944
00:46:44,200 --> 00:46:48,880
So I was like OK, universe give 
me a problem and then I got 2 

945
00:46:48,880 --> 00:46:50,880
problems, two big problems to 
solve. 

946
00:46:50,880 --> 00:46:54,480
But even like solving those big 
problems was I really enjoy when

947
00:46:54,480 --> 00:46:57,680
after that I felt so empowered 
together with the team. 

948
00:46:57,680 --> 00:47:00,480
And you can see sometimes also, 
of course you get tired and we 

949
00:47:00,480 --> 00:47:03,400
are all humans. 
But then you have your team that

950
00:47:03,520 --> 00:47:06,120
I've also had days that I was 
like, oh, I don't know, it's I'm

951
00:47:06,120 --> 00:47:08,040
tired, just tired, you know, 
just getting tired. 

952
00:47:08,400 --> 00:47:11,120
But then I was like, yeah, I 
will just continue only for our 

953
00:47:11,120 --> 00:47:12,880
team. 
We have these amazing people in 

954
00:47:12,880 --> 00:47:14,360
the team, my amazing Co 
founders. 

955
00:47:14,840 --> 00:47:17,000
So we'll do it, we'll do it 
together. 

956
00:47:17,000 --> 00:47:20,360
So it's also this feeling of 
connectedness, you know, being 

957
00:47:20,360 --> 00:47:23,440
in this together, being united. 
I love that a lot. 

958
00:47:23,440 --> 00:47:26,280
I also love how full circle it 
went with you then educating 

959
00:47:26,280 --> 00:47:28,440
kind of your journey to other 
people. 

960
00:47:28,880 --> 00:47:33,560
You mentioned in there that this
part of cybersecurity, detecting

961
00:47:33,560 --> 00:47:36,440
fraud in this manner with 
regards to defects is also 

962
00:47:36,440 --> 00:47:39,080
becoming super challenging. 
And I only have assumptions 

963
00:47:39,080 --> 00:47:40,880
because I can see there's a lot 
of investment here. 

964
00:47:41,160 --> 00:47:45,000
From my human eye. 
I can barely, I, I can see some 

965
00:47:45,000 --> 00:47:46,560
distinguishments. 
But yeah, I go to a comment 

966
00:47:46,560 --> 00:47:48,640
section. 
I, I use the community to figure

967
00:47:48,640 --> 00:47:50,680
out what is real and what is not
real, which is not really great 

968
00:47:50,680 --> 00:47:54,080
signal, which means that I can 
see from my perspective that I 

969
00:47:54,080 --> 00:47:55,200
can be more and more 
challenging. 

970
00:47:55,440 --> 00:47:56,880
But what type of problem is 
that? 

971
00:47:56,880 --> 00:48:00,320
Is that the current models that 
do detection versus generation 

972
00:48:00,320 --> 00:48:02,800
are not as accurate? 
Is it a data problem? 

973
00:48:03,200 --> 00:48:06,080
I don't even know if you need 
let's say sample data or like 

974
00:48:06,080 --> 00:48:07,960
data to go off of. 
Is there not enough data? 

975
00:48:07,960 --> 00:48:10,160
What type of problem is it? 
It's a combination of all the 

976
00:48:10,160 --> 00:48:12,800
things you mentioned. 
It's, it's all about data as 

977
00:48:12,800 --> 00:48:16,160
well. 
I I don't believe more data will

978
00:48:16,160 --> 00:48:17,600
solve all the problems that you 
have. 

979
00:48:17,800 --> 00:48:19,720
Definitely not. 
It will also maybe even buy your

980
00:48:20,000 --> 00:48:21,640
bias, your systems in a certain 
way. 

981
00:48:22,680 --> 00:48:25,360
The the challenge comes with 
first of all, all these 

982
00:48:25,360 --> 00:48:28,560
different deepfake generation 
types out there, these methods 

983
00:48:28,560 --> 00:48:31,480
and techniques. 
The second one is also once a 

984
00:48:31,480 --> 00:48:33,880
fraudster has really its 
intention of doing something, 

985
00:48:33,880 --> 00:48:36,120
they won't stop with using only 
one method. 

986
00:48:37,600 --> 00:48:39,200
For example, the insurance use 
case. 

987
00:48:39,200 --> 00:48:41,760
There is also like it's really 
even in the Netherlands, it's 

988
00:48:41,760 --> 00:48:43,160
booming. 
It was on the news I think 

989
00:48:43,160 --> 00:48:47,960
recently that someone managed to
get 150,000 tons of EUR 

990
00:48:47,960 --> 00:48:53,440
basically because he was able to
show that he couldn't go on his 

991
00:48:53,440 --> 00:48:56,200
30K vacation. 
It was from 5 or 6 year 

992
00:48:56,960 --> 00:48:59,600
insurances. 
And then he had this Jenny, I 

993
00:48:59,600 --> 00:49:03,760
made a doctor proof that he's, 
he has been sick so he couldn't 

994
00:49:03,760 --> 00:49:06,840
go there for him to cancel. 
So insurance pay my money back. 

995
00:49:07,560 --> 00:49:10,200
And of course, you go ahead and 
generate this type of, you know,

996
00:49:10,200 --> 00:49:13,120
proof with maybe ChatGPT or some
other tools that you have to 

997
00:49:13,120 --> 00:49:15,120
help yourself based on open 
source things. 

998
00:49:15,920 --> 00:49:17,520
But then you won't stop there. 
You won't. 

999
00:49:17,680 --> 00:49:20,400
You will also maybe use like a 
signature to put it there. 

1000
00:49:20,400 --> 00:49:23,560
So you will go ahead and do 
other types of modifications as 

1001
00:49:23,560 --> 00:49:26,200
well. 
So that's one of the challenges,

1002
00:49:26,200 --> 00:49:27,880
for example. 
The other one is what happens 

1003
00:49:27,880 --> 00:49:30,280
if, for example, when it comes 
to the news and media use case, 

1004
00:49:30,560 --> 00:49:34,920
what happens if a video that has
been made gets forwarded on 

1005
00:49:34,920 --> 00:49:38,400
WhatsApp and other social media 
channels many times causing it 

1006
00:49:38,400 --> 00:49:41,600
to have a like a worse 
resolution. 

1007
00:49:42,240 --> 00:49:45,440
And in that sense, the image is 
then or the video is not of a 

1008
00:49:45,440 --> 00:49:48,200
good quality anymore. 
And when the quality drops, 

1009
00:49:48,200 --> 00:49:51,400
specifically if you have like 
dark videos with no light or 

1010
00:49:51,400 --> 00:49:54,680
small faces in it, it becomes 
challenging to detect them. 

1011
00:49:55,720 --> 00:49:58,800
So those are purposefully, I 
think, activities that people 

1012
00:49:58,800 --> 00:50:01,960
can do to make it more 
challenging to be able to take 

1013
00:50:01,960 --> 00:50:04,080
to detect, detect a deep fake in
that sense. 

1014
00:50:04,400 --> 00:50:06,600
And there are also, of course, 
adversarial attacks happening. 

1015
00:50:06,600 --> 00:50:09,840
So that means that basically if 
they know how to fool your 

1016
00:50:09,840 --> 00:50:12,920
system, so if they somehow have 
an idea of how deep fake 

1017
00:50:12,920 --> 00:50:16,160
detection works, they can go 
ahead and basically use your own

1018
00:50:16,560 --> 00:50:19,680
strategy against you to put it 
there to fool you to think 

1019
00:50:19,680 --> 00:50:22,160
something is a deep fake or not 
a deep fake basically. 

1020
00:50:22,600 --> 00:50:25,840
So it's, it's complicated. 
Data is there to help, but it's 

1021
00:50:25,840 --> 00:50:29,040
also not everything I would say.
So it's a combination of threat 

1022
00:50:29,040 --> 00:50:31,720
intelligence, seeing what is 
needed, really focusing. 

1023
00:50:31,720 --> 00:50:36,520
Also as a start up, once we 
start it, you need to focus. 

1024
00:50:36,520 --> 00:50:40,320
There needs to be a focus. 
Right now we're skating, we have

1025
00:50:40,320 --> 00:50:43,440
worldwide clients from Singapore
to Brazil to US in Europe, of 

1026
00:50:43,440 --> 00:50:48,480
course, as well. 
So being this company in deep 

1027
00:50:48,480 --> 00:50:52,440
tech detection, one of the 
industry leaders, it's also 

1028
00:50:52,440 --> 00:50:55,600
again requires, you know, 
focusing on a in a balanced way 

1029
00:50:55,600 --> 00:50:57,600
because we want to scale up, of 
course, also when it comes to 

1030
00:50:57,600 --> 00:51:00,640
developing, developing our own 
platform, being able to also 

1031
00:51:00,640 --> 00:51:03,200
detect those videos of those 
birds you mentioned. 

1032
00:51:04,000 --> 00:51:07,640
But then again, it needs focus 
along the along along the way as

1033
00:51:07,640 --> 00:51:09,000
well. 
So the milestones need to be 

1034
00:51:09,000 --> 00:51:10,840
focused. 
And that's how I would say maybe

1035
00:51:10,840 --> 00:51:14,240
the biggest challenge that that 
we have that goes also maybe for

1036
00:51:14,240 --> 00:51:16,480
every start about. 
Their Yeah, yeah, I can see 

1037
00:51:16,480 --> 00:51:18,280
that. 
I mean, my assumption is that 

1038
00:51:18,280 --> 00:51:21,800
your software is slightly 
different, right, because yours 

1039
00:51:21,800 --> 00:51:24,560
needs to have a certain level of
accuracy. 

1040
00:51:24,560 --> 00:51:26,920
If you label something as deep 
fake and it's not a deep fake, 

1041
00:51:26,920 --> 00:51:29,080
then it's even worse 
reputationally. 

1042
00:51:29,560 --> 00:51:31,760
So you can't put out a half 
baked product. 

1043
00:51:32,120 --> 00:51:35,200
And then you also have a very 
complex domain in that if you do

1044
00:51:35,200 --> 00:51:38,400
reconnaissance and you go on the
dark web, I don't know, I'm not 

1045
00:51:38,400 --> 00:51:40,680
I all my knowledge of like 
police and what is legal 

1046
00:51:40,680 --> 00:51:42,320
illegal, a lot of it comes from 
TV. 

1047
00:51:42,320 --> 00:51:44,240
So I don't know how much of that
is legal or illegal. 

1048
00:51:44,240 --> 00:51:46,360
You know, you may be on the 
borderline there. 

1049
00:51:46,640 --> 00:51:50,800
So it's also very hard to 
actually like if I were you, I 

1050
00:51:50,800 --> 00:51:54,040
cannot, my assumption is go and 
ask for a fake bank account. 

1051
00:51:54,040 --> 00:51:56,040
That doesn't make any sense. 
So I would never be able to 

1052
00:51:56,240 --> 00:51:58,920
experience the real good 
customer service that some 

1053
00:51:58,920 --> 00:52:01,880
fraudulent people offer. 
But then you also don't know 

1054
00:52:01,880 --> 00:52:03,800
what type of new attacks are out
there. 

1055
00:52:03,840 --> 00:52:05,840
But you do need to know that 
because you're creating 

1056
00:52:05,840 --> 00:52:08,360
preventative self. 
So it's like very challenging in

1057
00:52:08,360 --> 00:52:09,600
many aspects. 
That's true, yeah. 

1058
00:52:10,040 --> 00:52:12,960
But it also, I think, makes it a
very interesting playing field 

1059
00:52:12,960 --> 00:52:14,520
to play out, yeah. 
Yeah, exactly. 

1060
00:52:14,520 --> 00:52:18,920
It's really interesting. 
It's about again, monitoring, 

1061
00:52:18,920 --> 00:52:21,360
you know, threat intelligence, 
following different sources to 

1062
00:52:21,360 --> 00:52:24,880
know indeed what is the newest 
that they use and also making 

1063
00:52:24,880 --> 00:52:26,840
our systems explainable. 
So when it comes to really 

1064
00:52:26,840 --> 00:52:30,600
sensitive use cases, we do have,
for example, clients in 

1065
00:52:30,600 --> 00:52:33,080
Singapore as well and those are 
defence agencies. 

1066
00:52:33,920 --> 00:52:36,680
There are cases that are quite 
highly sensitive. 

1067
00:52:36,680 --> 00:52:40,080
So in that sense, we don't tell 
them, oh, just trust our 

1068
00:52:40,080 --> 00:52:42,280
technology. 
We are honest with them. 

1069
00:52:42,280 --> 00:52:45,640
We say this is the accuracy that
we can offer, 95%, let's say 

1070
00:52:46,120 --> 00:52:48,440
this is the recall and the false
rejection rate. 

1071
00:52:48,840 --> 00:52:50,320
Also important to know of 
course. 

1072
00:52:50,800 --> 00:52:54,600
And if the case is highly 
sensitive, we can help you with 

1073
00:52:56,040 --> 00:53:01,160
basically manual expert analysis
as well based on, yeah, our 

1074
00:53:01,160 --> 00:53:05,320
forensic basically expertise and
knowledge to be able to validate

1075
00:53:05,320 --> 00:53:09,000
the output of our technology. 
So that works for use cases that

1076
00:53:09,000 --> 00:53:12,000
are highly sensitive and volume 
wise not so big. 

1077
00:53:12,040 --> 00:53:15,520
So occasional use cases that can
happen when it comes to, for 

1078
00:53:15,520 --> 00:53:19,640
example, banks flows or social 
media content, that's of course 

1079
00:53:19,880 --> 00:53:24,560
about huge volumes, but the 
chance and the risk appetite 

1080
00:53:24,720 --> 00:53:26,960
there is also of course 
different than the one, for 

1081
00:53:26,960 --> 00:53:28,840
example, the defence agency in 
Singapore has. 

1082
00:53:29,000 --> 00:53:31,680
Yeah, I can see that. 
If there's a person that's 

1083
00:53:31,680 --> 00:53:34,760
responsible more from an 
organization standpoint, what is

1084
00:53:34,760 --> 00:53:37,600
some low hanging fruit that 
organizations can use or adopt? 

1085
00:53:38,400 --> 00:53:41,120
Have some preventative measures 
with regards to this. 

1086
00:53:41,160 --> 00:53:44,760
So when it comes to Gen. 
AI based fraud, our technology 

1087
00:53:44,760 --> 00:53:48,520
for example to give to give 
basically an impression of it is

1088
00:53:48,520 --> 00:53:51,840
really easy to be deployed 
really easily to be basically 

1089
00:53:51,840 --> 00:53:56,320
put there to monitor your 
existing data, production data. 

1090
00:53:56,760 --> 00:53:59,880
And that already is a great step
in knowing to what scope you're 

1091
00:53:59,880 --> 00:54:01,240
already dealing with the 
problem. 

1092
00:54:01,520 --> 00:54:02,960
It could be there at a big 
scale. 

1093
00:54:02,960 --> 00:54:05,920
It could also be not there yet 
or only a couple of them, but we

1094
00:54:05,920 --> 00:54:08,680
see still sold. 
First we had to educate around 

1095
00:54:08,960 --> 00:54:14,360
what is a deep fake and now we 
still do a lot of education 

1096
00:54:14,360 --> 00:54:17,800
regarding basically what is the 
scope of the problem that you're

1097
00:54:17,800 --> 00:54:19,600
dealing with right now because 
of Gen. 

1098
00:54:19,600 --> 00:54:22,760
AI based fraud. 
And that's the question that 

1099
00:54:22,760 --> 00:54:26,840
still a lot of companies haven't
found an answer to. 

1100
00:54:26,920 --> 00:54:28,400
Yeah. 
Yeah. 

1101
00:54:28,520 --> 00:54:31,440
What about the consumer side? 
So me, you already mentioned 

1102
00:54:31,440 --> 00:54:34,320
that some of my 2 factor 
authentication which I thought 

1103
00:54:34,320 --> 00:54:39,160
was really good, Face ID, voice 
ID in some cases are compromised

1104
00:54:39,160 --> 00:54:41,600
and I understand that. 
Now what can? 

1105
00:54:41,600 --> 00:54:43,360
What can consumers do? 
Or what do they need to pay 

1106
00:54:43,360 --> 00:54:45,360
attention to? 
Yeah, that's a that's a great 

1107
00:54:45,360 --> 00:54:47,000
question. 
So consumers are, I would say 

1108
00:54:47,000 --> 00:54:51,400
when it comes to your bank or 
your Apple ID, iPhone, iPhone, 

1109
00:54:51,720 --> 00:54:53,960
it's I would say it's the 
responsibility of the bank or 

1110
00:54:53,960 --> 00:54:55,600
the company behind it to make it
secure. 

1111
00:54:55,600 --> 00:55:00,760
Of course, consumers are 
attacked by different types of 

1112
00:55:00,760 --> 00:55:03,760
defects, like the ones on social
media, the ones that you can, 

1113
00:55:04,120 --> 00:55:06,200
yeah, you can expect a call from
a loved one. 

1114
00:55:06,680 --> 00:55:08,920
And then again, one of the 
challenges that I mentioned 

1115
00:55:09,200 --> 00:55:12,160
along this conversation was we 
don't know who's responsible for

1116
00:55:12,160 --> 00:55:14,120
what. 
So I would say when it comes to 

1117
00:55:14,120 --> 00:55:18,400
banks, they are definitely 
responsible to not to let a deep

1118
00:55:18,400 --> 00:55:21,880
fake fool their systems. 
But when it comes to, for 

1119
00:55:21,880 --> 00:55:25,000
example, getting this call 
through WhatsApp, then I would 

1120
00:55:25,000 --> 00:55:28,160
say Mehta is also responsible. 
But then Mehta can say no, the 

1121
00:55:28,160 --> 00:55:31,280
person who received the call 
should be aware and not accept 

1122
00:55:31,280 --> 00:55:34,400
that. 
And then you go to the police 

1123
00:55:34,400 --> 00:55:36,800
saying I got scammed. 
The police says, yeah, how 

1124
00:55:36,800 --> 00:55:38,400
should, how are we supposed to 
find this person? 

1125
00:55:38,400 --> 00:55:41,800
They are professional criminals.
And so in that sense, there is 

1126
00:55:41,800 --> 00:55:46,800
still not really, I would say 
clear image, clear strategy for 

1127
00:55:46,800 --> 00:55:51,520
the for the society, where to 
start basically coming up with a

1128
00:55:51,520 --> 00:55:55,800
solution for this big problem. 
Yeah, also feels like kind of a 

1129
00:55:55,800 --> 00:55:59,040
gap in the market, right, 
Because right now maybe 

1130
00:55:59,040 --> 00:56:02,440
consumers are not really aware 
of fraudulent behavior or 

1131
00:56:02,440 --> 00:56:04,320
compromises or risks that 
they're there. 

1132
00:56:04,720 --> 00:56:08,280
They're also not like targets 
necessarily, maybe on a more 

1133
00:56:08,280 --> 00:56:10,560
smaller scale. 
But then when big problems 

1134
00:56:10,560 --> 00:56:13,240
happen, if there is a piece of 
software that all of a sudden 

1135
00:56:13,240 --> 00:56:15,360
says, OK, now that's our 
responsibility and we'll make 

1136
00:56:15,360 --> 00:56:18,520
sure that when you use that, 
that's not going to happen, then

1137
00:56:18,520 --> 00:56:20,240
that's all of a sudden like a 
USB. 

1138
00:56:20,240 --> 00:56:22,680
That would differentiate if 
we're talking about a chat or a 

1139
00:56:22,680 --> 00:56:26,200
voice or video calling service, 
something from WhatsApp where 

1140
00:56:26,200 --> 00:56:28,280
Meta says, OK, it's just your 
own responsibility. 

1141
00:56:28,400 --> 00:56:30,920
Yeah, that's very interesting. 
It's really interesting because 

1142
00:56:30,960 --> 00:56:33,640
I think it happened quite 
recently, a couple of months ago

1143
00:56:33,640 --> 00:56:38,640
that I think Facebook had 
something that you they they 

1144
00:56:38,640 --> 00:56:42,240
would require you to say if your
content that you would upload is

1145
00:56:42,240 --> 00:56:43,280
Gen. 
AI or not. 

1146
00:56:43,280 --> 00:56:44,720
Yeah. 
YouTube also does that. 

1147
00:56:44,760 --> 00:56:47,200
Yeah, but I think Facebook has 
removed it. 

1148
00:56:47,520 --> 00:56:50,360
I was like, wow, this is and it 
specifically happened after the 

1149
00:56:50,360 --> 00:56:54,800
new administration in the US. 
And then I also get questions 

1150
00:56:54,800 --> 00:56:59,880
like like, do you have any 
interest and you know, or any 

1151
00:56:59,880 --> 00:57:03,960
benefit by letting deep fix on 
your platform that can mislead 

1152
00:57:03,960 --> 00:57:06,720
people and send the funny ones? 
They are, they're nice. 

1153
00:57:06,720 --> 00:57:08,480
I also watched them. 
There's hilarious ones. 

1154
00:57:09,680 --> 00:57:13,040
Yeah, yeah, yeah. 
But the yeah, but, but the other

1155
00:57:13,040 --> 00:57:15,280
ones then like the really 
fraudulent ones. 

1156
00:57:16,080 --> 00:57:19,120
So in that sense, I also don't 
know what we did these these big

1157
00:57:19,120 --> 00:57:22,560
tech companies have, as you 
know, as their strategy when it 

1158
00:57:22,560 --> 00:57:25,160
comes to fake content. 
Yeah, I have no clue. 

1159
00:57:25,680 --> 00:57:27,160
I mean, fake content is super 
cool. 

1160
00:57:27,280 --> 00:57:29,520
I saw this. 
I think it must have been like 5

1161
00:57:29,520 --> 00:57:30,960
seconds. 
Someone made it and it wasn't 

1162
00:57:30,960 --> 00:57:34,280
even for that company. 
It was like an IKEA ad where 

1163
00:57:34,280 --> 00:57:37,160
there's a big box and all of a 
sudden there's an empty room and

1164
00:57:37,160 --> 00:57:39,560
it explodes and there's IKEA 
furniture everywhere. 

1165
00:57:40,160 --> 00:57:42,880
Super cool, super great 
marketing for IKEA. 

1166
00:57:42,880 --> 00:57:44,720
This person was not affiliated 
with IKEA. 

1167
00:57:44,720 --> 00:57:48,040
Made it with Google VO and like 
spent X amount of money and then

1168
00:57:48,360 --> 00:57:49,560
that's it. 
Video blew up. 

1169
00:57:49,640 --> 00:57:51,800
Yeah, super great. 
It's it's really nice it. 

1170
00:57:51,840 --> 00:57:55,320
Becomes really scary if like 
those super cool things, yeah, 

1171
00:57:55,320 --> 00:57:56,760
happen for fraudulent things, 
yeah. 

1172
00:57:56,920 --> 00:57:59,680
And sadly, that's reality. 
Yeah, I think I even McDonald's 

1173
00:57:59,680 --> 00:58:02,760
had themselves made a fully AI 
generated ad not from a 

1174
00:58:02,760 --> 00:58:03,720
marketing. 
Perspective. 

1175
00:58:03,760 --> 00:58:06,880
Yeah, that that's true, yeah. 
You also see like a new type of.

1176
00:58:06,880 --> 00:58:09,680
Of course, influencers are 
relatively a new type of 

1177
00:58:09,680 --> 00:58:13,120
industry and job, but there are 
now also AI, fully AI 

1178
00:58:13,120 --> 00:58:16,760
influencers. 
Then you would see like Swedish 

1179
00:58:16,760 --> 00:58:17,880
looking girl. 
Let's say. 

1180
00:58:17,880 --> 00:58:21,120
They also say like they have a 
personality, like I'm Diana 

1181
00:58:21,120 --> 00:58:24,360
from, I don't know, Stockholm. 
And then they go around the 

1182
00:58:24,360 --> 00:58:28,280
world. 
They have nice clothes and it's 

1183
00:58:28,280 --> 00:58:32,480
basically like a teenager in the
room behind this Diana. 

1184
00:58:32,480 --> 00:58:34,880
That's it. 
And it's it's becoming really 

1185
00:58:34,880 --> 00:58:37,320
prominent as though they have 
millions of followers on on 

1186
00:58:37,320 --> 00:58:41,000
TikTok, on Instagram. 
And yeah, it's nice to see. 

1187
00:58:41,000 --> 00:58:44,120
But again, like, personally, I'm
questioning what is the purpose 

1188
00:58:44,120 --> 00:58:46,680
behind, you know, following such
an account or even making this 

1189
00:58:46,680 --> 00:58:48,440
account. 
They could be making money. 

1190
00:58:48,440 --> 00:58:50,520
I understand. 
But like following an AI account

1191
00:58:50,800 --> 00:58:53,080
which has nice pictures could 
also be nice to see. 

1192
00:58:53,080 --> 00:58:55,120
I don't know but I was 
questioning that a bit. 

1193
00:58:55,440 --> 00:58:57,800
I understand that. 
Like for me I I never was 

1194
00:58:57,800 --> 00:58:59,400
interested in. 
Like Instagram and stuff like 

1195
00:58:59,400 --> 00:59:00,920
that. 
But the fact that you're 

1196
00:59:00,920 --> 00:59:03,880
following, let's say an AI 
person versus a person that you 

1197
00:59:03,880 --> 00:59:07,000
know exists, but you don't 
really know who that person is. 

1198
00:59:07,000 --> 00:59:09,680
You've just seen them online. 
Like there's a, it's quite close

1199
00:59:09,680 --> 00:59:11,320
to each other. 
I would say one just lives in 

1200
00:59:11,320 --> 00:59:13,880
the real world and one doesn't, 
but they do the same thing. 

1201
00:59:13,880 --> 00:59:17,000
Yeah, I think there was one that
that's a real really. 

1202
00:59:17,000 --> 00:59:21,760
Sad example, there was one 
teenager who thought that he 

1203
00:59:21,760 --> 00:59:27,760
fell in love with such an AI 
person and I think it was also 

1204
00:59:27,760 --> 00:59:32,560
like maybe a HHPT ish AI that 
would talk to talk to this 

1205
00:59:32,560 --> 00:59:35,280
person to this teenager. 
He was 14 years old, I think. 

1206
00:59:35,800 --> 00:59:40,560
And eventually he committed 
suicide because he wanted to go 

1207
00:59:40,560 --> 00:59:44,120
to his beloved girlfriend, AI 
girlfriend. 

1208
00:59:44,680 --> 00:59:47,160
And the mother was of, of 
course, devastated. 

1209
00:59:47,160 --> 00:59:50,440
And she was like, you know, Cha 
Chibi did this to my son because

1210
00:59:50,880 --> 00:59:52,760
the AI would say, come to me, 
you know, I'm here. 

1211
00:59:52,760 --> 00:59:56,560
And the the person just, yeah, 
yeah, there's no possibility. 

1212
00:59:57,240 --> 01:00:00,680
That's really sad. 
Yeah, yeah. 

1213
01:00:00,720 --> 01:00:04,680
It's like it's, I also feel like
with consumers and right now 

1214
01:00:04,680 --> 01:00:08,040
this is not a problem. 
But if I were to go and I'm I'm 

1215
01:00:08,040 --> 01:00:10,760
victim of fraud, if I go to the 
police, I don't know if they're 

1216
01:00:10,760 --> 01:00:13,480
well equipped. 
To help me, I don't know what 

1217
01:00:13,480 --> 01:00:15,880
cybersecurity capabilities are 
already. 

1218
01:00:15,880 --> 01:00:19,040
Now I feel like we have too many
people, too many issues for 

1219
01:00:19,040 --> 01:00:20,400
already the current police 
system. 

1220
01:00:20,720 --> 01:00:22,960
Add something like a digital 
layer on top of it. 

1221
01:00:23,360 --> 01:00:26,800
I would love to call like a 
cyber police hotline but it's 

1222
01:00:26,800 --> 01:00:28,800
not there. 
I think we need something. 

1223
01:00:28,800 --> 01:00:31,640
If this becomes a bigger problem
and like more towards the 

1224
01:00:31,640 --> 01:00:34,800
consumer rather than the people 
that create the apps for 

1225
01:00:34,800 --> 01:00:37,240
consumers, then we need 
something like that. 

1226
01:00:37,240 --> 01:00:38,680
With regards to the fence. 
Exactly. 

1227
01:00:38,680 --> 01:00:40,280
The police are specifically in 
that last. 

1228
01:00:40,280 --> 01:00:43,400
They are super. 
Busy when it comes to cyber 

1229
01:00:43,400 --> 01:00:46,120
criminality and you know, cyber 
attacks. 

1230
01:00:46,800 --> 01:00:48,720
I don't know what they're doing 
right now because they, they 

1231
01:00:48,720 --> 01:00:51,480
could do takedown requests. 
There are also specific hotlines

1232
01:00:51,480 --> 01:00:53,560
for that to do a takedown 
request when it comes to, for 

1233
01:00:53,560 --> 01:00:57,240
example, 18 plus videos of 
someone, not not consensually, 

1234
01:00:57,640 --> 01:00:59,920
but of what happens after, 
because the video can also be 

1235
01:00:59,920 --> 01:01:02,760
already circulating in social 
media channels and maybe even 

1236
01:01:02,760 --> 01:01:06,800
WhatsApp chats or telecom chats.
It's, it's indeed like a big 

1237
01:01:06,800 --> 01:01:11,040
skill of it's a quite fast speed
of sharing material, even if, if

1238
01:01:11,040 --> 01:01:13,360
it's a deep fake. 
And then at that point I'm not 

1239
01:01:13,360 --> 01:01:15,080
training it. 
What the police can do, I don't 

1240
01:01:15,080 --> 01:01:16,120
know. 
Have you? 

1241
01:01:16,160 --> 01:01:18,520
Because I'm not aware of any of 
these cases. 

1242
01:01:18,880 --> 01:01:21,240
But my assumption is usually 
it's someone close to a person 

1243
01:01:21,240 --> 01:01:22,560
right? 
If there's something that's in 

1244
01:01:22,560 --> 01:01:25,200
fake of a specific person, it 
might be someone in a close 

1245
01:01:25,200 --> 01:01:27,800
circle of friends or family that
just doesn't like this person or

1246
01:01:27,800 --> 01:01:28,880
something happened. 
Yeah. 

1247
01:01:28,880 --> 01:01:31,040
Is that usually the case? 
That's true if it's a non 

1248
01:01:31,040 --> 01:01:35,160
celebrity or non pep. 
Then politically exposed person,

1249
01:01:35,520 --> 01:01:38,920
then it's a need either a 
classmate that doesn't like this

1250
01:01:38,920 --> 01:01:43,320
person or they think that 
they're funny or also like 

1251
01:01:43,640 --> 01:01:45,880
people who have ended their 
relationships, their romantic 

1252
01:01:45,880 --> 01:01:48,560
relationships have ended and 
right now they're enemies. 

1253
01:01:49,040 --> 01:01:51,840
And then the the deep fake 
mostly also like the girl, but 

1254
01:01:51,840 --> 01:01:55,440
also the boy can get deep fake. 
And it's also used sometimes for

1255
01:01:55,440 --> 01:01:58,200
blackmailing purposes, which is 
also really bad. 

1256
01:01:58,200 --> 01:01:59,160
Of course. 
It's quite messed up. 

1257
01:01:59,160 --> 01:02:02,120
Yeah, yeah, I must say I have. 
Learned a lot. 

1258
01:02:02,440 --> 01:02:05,640
I I knew there was depth to 
this, but there's much more 

1259
01:02:05,640 --> 01:02:08,000
depth to this than I think. 
Thank you so much for coming on 

1260
01:02:08,000 --> 01:02:10,720
and sharing. 
But yeah, this was a a lovely 

1261
01:02:10,720 --> 01:02:12,160
conversation. 
Thank you for having me. 

1262
01:02:12,360 --> 01:02:14,200
Yeah, I'm going to round it off 
here. 

1263
01:02:14,320 --> 01:02:16,000
Normally cameras on this. 
Level we switched. 

1264
01:02:16,760 --> 01:02:19,080
Thank you so much for listening.
Leave a like if you like the 

1265
01:02:19,080 --> 01:02:21,080
episode, let us know in the 
comment section what you thought

1266
01:02:21,080 --> 01:02:22,080
and we'll see you on the next 
one.