1
00:00:02,280 --> 00:00:08,130
[Music]

2
00:00:08,130 --> 00:00:10,140
[Applause]

3
00:00:10,140 --> 00:00:16,619
[Music]

4
00:00:24,240 --> 00:00:26,400
hello

5
00:00:26,400 --> 00:00:28,560
good morning uh good afternoon depending

6
00:00:28,560 --> 00:00:30,000
on where you are i'm patrick tucker

7
00:00:30,000 --> 00:00:32,880
technology editor at defense one joining

8
00:00:32,880 --> 00:00:33,920
me today

9
00:00:33,920 --> 00:00:37,040
is richard marco ceo of eset

10
00:00:37,040 --> 00:00:39,600
uh fantastic company we're going to talk

11
00:00:39,600 --> 00:00:40,399
today about

12
00:00:40,399 --> 00:00:42,399
applications of artificial intelligence

13
00:00:42,399 --> 00:00:44,800
and cyber security both the good and the

14
00:00:44,800 --> 00:00:47,039
bad and uh this is a vitally important

15
00:00:47,039 --> 00:00:47,520
topic

16
00:00:47,520 --> 00:00:49,840
because if you are like me or you've

17
00:00:49,840 --> 00:00:51,760
been awake for the last few years you

18
00:00:51,760 --> 00:00:52,640
know that uh

19
00:00:52,640 --> 00:00:54,399
we're all beset all the time by very

20
00:00:54,399 --> 00:00:56,160
aggressive marketing about

21
00:00:56,160 --> 00:00:58,160
what artificial intelligence can do for

22
00:00:58,160 --> 00:01:00,000
businesses what it can perhaps do for

23
00:01:00,000 --> 00:01:00,960
individuals

24
00:01:00,960 --> 00:01:03,359
uh we don't hear a lot about how it

25
00:01:03,359 --> 00:01:04,479
might cause havoc

26
00:01:04,479 --> 00:01:05,920
unless you're watching science fiction

27
00:01:05,920 --> 00:01:07,680
movies so there's this strange

28
00:01:07,680 --> 00:01:09,200
divergence right in terms of the

29
00:01:09,200 --> 00:01:10,880
marketing that we consume and the the

30
00:01:10,880 --> 00:01:12,000
content that we consume

31
00:01:12,000 --> 00:01:13,840
where on the one hand we're being told

32
00:01:13,840 --> 00:01:15,280
all of the great things that ai

33
00:01:15,280 --> 00:01:17,280
do for us we're always set all the time

34
00:01:17,280 --> 00:01:18,640
scared through uh

35
00:01:18,640 --> 00:01:21,680
hyperbole and and and fiction

36
00:01:21,680 --> 00:01:25,040
uh about what it might do uh when it's

37
00:01:25,040 --> 00:01:28,320
unleashed but uh richard here you

38
00:01:28,320 --> 00:01:30,479
actually operate at the center of where

39
00:01:30,479 --> 00:01:32,400
those two worlds collide creating cyber

40
00:01:32,400 --> 00:01:33,840
security solutions through artificial

41
00:01:33,840 --> 00:01:35,680
intelligence and also watching

42
00:01:35,680 --> 00:01:38,400
uh how potential adversaries might use

43
00:01:38,400 --> 00:01:40,000
the most important emerging technology

44
00:01:40,000 --> 00:01:42,159
of the 21st century to cause problems

45
00:01:42,159 --> 00:01:43,439
and wreak havoc

46
00:01:43,439 --> 00:01:46,399
so first i wonder if you can tell us

47
00:01:46,399 --> 00:01:47,920
anything about

48
00:01:47,920 --> 00:01:50,640
whether you've seen actual examples

49
00:01:50,640 --> 00:01:51,920
today

50
00:01:51,920 --> 00:01:55,119
of artificial intelligence whether

51
00:01:55,119 --> 00:01:56,079
machine learning

52
00:01:56,079 --> 00:01:58,000
or something more complex like neural

53
00:01:58,000 --> 00:02:00,399
networks used as part

54
00:02:00,399 --> 00:02:02,640
of a cyber security attack is this

55
00:02:02,640 --> 00:02:04,479
something that exists right now

56
00:02:04,479 --> 00:02:09,598
perhaps right so um

57
00:02:09,598 --> 00:02:11,599
it's it's definitely a field that is

58
00:02:11,599 --> 00:02:13,360
attracting a lot of attention and

59
00:02:13,360 --> 00:02:14,319
everybody

60
00:02:14,319 --> 00:02:17,280
is like feeling you know you you said it

61
00:02:17,280 --> 00:02:18,239
very correctly

62
00:02:18,239 --> 00:02:20,640
there is quite a bit of hype behind it

63
00:02:20,640 --> 00:02:23,120
some artificial intelligence so

64
00:02:23,120 --> 00:02:26,560
uh in many cases people just imagine

65
00:02:26,560 --> 00:02:27,920
that there is some

66
00:02:27,920 --> 00:02:30,160
independent intelligence working in the

67
00:02:30,160 --> 00:02:30,959
background

68
00:02:30,959 --> 00:02:33,680
and in this case producing some

69
00:02:33,680 --> 00:02:34,720
malicious

70
00:02:34,720 --> 00:02:37,840
attack software and so on

71
00:02:37,840 --> 00:02:41,519
what in general you can't

72
00:02:41,519 --> 00:02:44,959
easily tell whether whether machine

73
00:02:44,959 --> 00:02:47,120
learning or artificial intelligence was

74
00:02:47,120 --> 00:02:47,840
used in

75
00:02:47,840 --> 00:02:50,640
a particular attack unless you you have

76
00:02:50,640 --> 00:02:51,120
a

77
00:02:51,120 --> 00:02:53,920
like a clear evidence so you might see

78
00:02:53,920 --> 00:02:55,040
signs of it

79
00:02:55,040 --> 00:02:58,159
but unless you confiscate servers and

80
00:02:58,159 --> 00:02:59,760
you see that they were using the

81
00:02:59,760 --> 00:03:02,159
attackers for using the machine learning

82
00:03:02,159 --> 00:03:04,480
or artificial intelligence in general

83
00:03:04,480 --> 00:03:06,319
you you can't have a proof

84
00:03:06,319 --> 00:03:09,120
but among the things that we consider

85
00:03:09,120 --> 00:03:11,680
are probably influenced by or

86
00:03:11,680 --> 00:03:14,239
might be using artificial intelligence

87
00:03:14,239 --> 00:03:15,440
are things like

88
00:03:15,440 --> 00:03:18,879
for example uh the spam can

89
00:03:18,879 --> 00:03:21,040
uh campaigns you know you need to

90
00:03:21,040 --> 00:03:22,879
produce a text that is

91
00:03:22,879 --> 00:03:24,959
that looks like it's coming from a human

92
00:03:24,959 --> 00:03:26,959
being that is easily readable

93
00:03:26,959 --> 00:03:29,360
he's able to persuade people to do

94
00:03:29,360 --> 00:03:30,239
something

95
00:03:30,239 --> 00:03:32,239
it's written in different languages and

96
00:03:32,239 --> 00:03:33,519
this is where

97
00:03:33,519 --> 00:03:35,920
artificial intelligence as an example

98
00:03:35,920 --> 00:03:36,959
can play a very

99
00:03:36,959 --> 00:03:40,319
uh very important role in the past like

100
00:03:40,319 --> 00:03:41,200
let's say

101
00:03:41,200 --> 00:03:43,920
five to ten years ago we saw many

102
00:03:43,920 --> 00:03:46,319
examples where you saw this kind of

103
00:03:46,319 --> 00:03:48,879
phishing email coming to you and he knew

104
00:03:48,879 --> 00:03:50,319
right from the beginning that

105
00:03:50,319 --> 00:03:52,720
it's a mess it was just words you know

106
00:03:52,720 --> 00:03:54,080
in wrong orders and

107
00:03:54,080 --> 00:03:56,879
so on right now it's a lot better and

108
00:03:56,879 --> 00:03:58,640
it's only improving

109
00:03:58,640 --> 00:04:00,640
huh okay so this is to your mind

110
00:04:00,640 --> 00:04:02,080
something that we're actually already

111
00:04:02,080 --> 00:04:04,000
experiencing we're already experiencing

112
00:04:04,000 --> 00:04:06,159
applications of artificial intelligence

113
00:04:06,159 --> 00:04:07,360
in cyber attacks

114
00:04:07,360 --> 00:04:09,680
um tell us a little bit about the state

115
00:04:09,680 --> 00:04:10,720
of the art of

116
00:04:10,720 --> 00:04:13,040
artificial intelligence in defending

117
00:04:13,040 --> 00:04:14,799
from not only those attacks but really

118
00:04:14,799 --> 00:04:15,680
all of them

119
00:04:15,680 --> 00:04:19,040
uh i know from my coverage of this area

120
00:04:19,040 --> 00:04:20,720
that one of the things ai does

121
00:04:20,720 --> 00:04:23,759
is rapidly accelerate the pace of

122
00:04:23,759 --> 00:04:25,360
operations

123
00:04:25,360 --> 00:04:29,040
and defending against potential cyber

124
00:04:29,040 --> 00:04:29,840
attackers

125
00:04:29,840 --> 00:04:31,600
is something that you have to do very

126
00:04:31,600 --> 00:04:33,440
promptly and in many ways almost

127
00:04:33,440 --> 00:04:34,400
predictively

128
00:04:34,400 --> 00:04:36,639
so what is the current state of the art

129
00:04:36,639 --> 00:04:38,720
of using ai and cyber defense what does

130
00:04:38,720 --> 00:04:41,199
that look like right now

131
00:04:41,199 --> 00:04:45,360
so the machine learning of

132
00:04:45,360 --> 00:04:49,759
like the current version deep learning

133
00:04:49,759 --> 00:04:53,280
is especially powerful when you when you

134
00:04:53,280 --> 00:04:55,520
are able to process you want to process

135
00:04:55,520 --> 00:04:56,400
some

136
00:04:56,400 --> 00:04:59,280
big data you have a lot of like data

137
00:04:59,280 --> 00:05:01,039
that has some

138
00:05:01,039 --> 00:05:03,039
it contains some patterns that can be

139
00:05:03,039 --> 00:05:04,800
classified so let's say

140
00:05:04,800 --> 00:05:07,440
malicious non-malicious so if you have

141
00:05:07,440 --> 00:05:07,840
that

142
00:05:07,840 --> 00:05:10,000
then you can use these machine learning

143
00:05:10,000 --> 00:05:11,120
algorithms

144
00:05:11,120 --> 00:05:14,639
to train your module as it is called

145
00:05:14,639 --> 00:05:17,039
to be able to distinguish it but you

146
00:05:17,039 --> 00:05:18,160
need to be careful

147
00:05:18,160 --> 00:05:20,720
on what you put as an input into it if

148
00:05:20,720 --> 00:05:21,280
you

149
00:05:21,280 --> 00:05:24,080
if you are too general you expect too

150
00:05:24,080 --> 00:05:25,199
much from it

151
00:05:25,199 --> 00:05:27,120
then the results might be just

152
00:05:27,120 --> 00:05:29,440
disappointing so in our case and i

153
00:05:29,440 --> 00:05:31,919
i believe we are at the edge of using

154
00:05:31,919 --> 00:05:33,600
machine learning in

155
00:05:33,600 --> 00:05:37,600
in malicious software kind of protection

156
00:05:37,600 --> 00:05:40,960
is that we use several kind of

157
00:05:40,960 --> 00:05:43,520
algorithms they're not machine learning

158
00:05:43,520 --> 00:05:46,720
in pre-processing inputs

159
00:05:46,720 --> 00:05:50,160
into the then models

160
00:05:50,160 --> 00:05:53,199
that use deep learning to to do the

161
00:05:53,199 --> 00:05:55,680
final decisions and usually we use

162
00:05:55,680 --> 00:05:58,000
several of such models for different

163
00:05:58,000 --> 00:05:59,039
scenarios

164
00:05:59,039 --> 00:06:01,600
so do you envision uh based on what

165
00:06:01,600 --> 00:06:02,800
you've seen

166
00:06:02,800 --> 00:06:05,120
that artificial intelligence is going to

167
00:06:05,120 --> 00:06:07,039
accelerate both the pace of cyber

168
00:06:07,039 --> 00:06:07,600
attacks

169
00:06:07,600 --> 00:06:10,160
and the pace of necessary defense so it

170
00:06:10,160 --> 00:06:11,120
becomes almost

171
00:06:11,120 --> 00:06:13,199
like an arms race if you will is that

172
00:06:13,199 --> 00:06:14,880
the future that we're headed towards

173
00:06:14,880 --> 00:06:17,919
uh i would say so because uh the example

174
00:06:17,919 --> 00:06:18,319
i

175
00:06:18,319 --> 00:06:21,120
i gave about the fishing emails it's

176
00:06:21,120 --> 00:06:22,000
just

177
00:06:22,000 --> 00:06:24,800
the the part you know it's generally

178
00:06:24,800 --> 00:06:26,560
accepted that the human beings are

179
00:06:26,560 --> 00:06:27,440
usually

180
00:06:27,440 --> 00:06:31,039
the weakest point in in the defense so

181
00:06:31,039 --> 00:06:32,960
if you are able to trick the right

182
00:06:32,960 --> 00:06:35,840
people then you are able to penetrate

183
00:06:35,840 --> 00:06:38,240
the overall defense quite well usually

184
00:06:38,240 --> 00:06:39,440
so

185
00:06:39,440 --> 00:06:42,160
uh this artificial intelligence is

186
00:06:42,160 --> 00:06:44,080
exactly going in the direction to be

187
00:06:44,080 --> 00:06:44,639
able to

188
00:06:44,639 --> 00:06:47,039
interact with human beings in a way that

189
00:06:47,039 --> 00:06:48,000
they don't

190
00:06:48,000 --> 00:06:50,639
realize this is uh artificial

191
00:06:50,639 --> 00:06:52,560
intelligence in the background or

192
00:06:52,560 --> 00:06:54,160
they don't realize they are being

193
00:06:54,160 --> 00:06:55,680
manipulated

194
00:06:55,680 --> 00:06:58,880
into doing some steps so yeah it's it is

195
00:06:58,880 --> 00:07:01,199
going to accelerate this and make it

196
00:07:01,199 --> 00:07:01,599
more

197
00:07:01,599 --> 00:07:04,400
tricky and can you describe for us very

198
00:07:04,400 --> 00:07:05,120
quickly

199
00:07:05,120 --> 00:07:08,560
in this accelerated future what is the

200
00:07:08,560 --> 00:07:09,919
role for

201
00:07:09,919 --> 00:07:12,160
classically trained cybersecurity

202
00:07:12,160 --> 00:07:14,000
professionals it's it's something of an

203
00:07:14,000 --> 00:07:14,639
art

204
00:07:14,639 --> 00:07:17,360
as well as the science you have to learn

205
00:07:17,360 --> 00:07:18,240
a lot about

206
00:07:18,240 --> 00:07:20,400
the psychology that attackers might use

207
00:07:20,400 --> 00:07:22,960
in developing phishing emails like this

208
00:07:22,960 --> 00:07:25,520
and you you have to understand different

209
00:07:25,520 --> 00:07:27,120
methodologies that different attackers

210
00:07:27,120 --> 00:07:29,120
have used it requires a lot of

211
00:07:29,120 --> 00:07:31,039
experience to become really good at this

212
00:07:31,039 --> 00:07:32,639
and the best practitioners in cyber

213
00:07:32,639 --> 00:07:33,599
security

214
00:07:33,599 --> 00:07:36,560
uh have very deep uh experience with

215
00:07:36,560 --> 00:07:39,199
specific attackers so

216
00:07:39,199 --> 00:07:41,680
what is the role of professionals like

217
00:07:41,680 --> 00:07:42,720
that

218
00:07:42,720 --> 00:07:44,720
in a future where this tool is very

219
00:07:44,720 --> 00:07:47,680
widely available

220
00:07:48,000 --> 00:07:51,680
so uh we we need to realize that uh

221
00:07:51,680 --> 00:07:55,199
even these so-called deep

222
00:07:55,199 --> 00:07:58,000
like networks so they contain a number

223
00:07:58,000 --> 00:07:58,879
of layers

224
00:07:58,879 --> 00:08:01,759
number of neurons they are far far from

225
00:08:01,759 --> 00:08:02,240
what

226
00:08:02,240 --> 00:08:06,319
the human brain is like the capacity of

227
00:08:06,319 --> 00:08:10,160
human brain so usually the

228
00:08:10,160 --> 00:08:12,319
the artificial intelligence is still

229
00:08:12,319 --> 00:08:13,680
very much dependent on

230
00:08:13,680 --> 00:08:16,960
human intelligence to kind of direct it

231
00:08:16,960 --> 00:08:19,039
to train it in the right way

232
00:08:19,039 --> 00:08:22,000
what is artificial intelligence better

233
00:08:22,000 --> 00:08:22,800
than human

234
00:08:22,800 --> 00:08:26,080
is uh the ability to use

235
00:08:26,080 --> 00:08:29,280
you know these i t tools like a lot of

236
00:08:29,280 --> 00:08:31,039
memory a lot of informations

237
00:08:31,039 --> 00:08:33,839
and process it quickly so human being is

238
00:08:33,839 --> 00:08:35,599
just not able to do this

239
00:08:35,599 --> 00:08:39,760
in the same way so this is where i see

240
00:08:39,760 --> 00:08:40,399
uh

241
00:08:40,399 --> 00:08:43,599
this connection you have the you you

242
00:08:43,599 --> 00:08:45,680
create your models your

243
00:08:45,680 --> 00:08:48,480
neural networks they are able to process

244
00:08:48,480 --> 00:08:50,160
a lot of information

245
00:08:50,160 --> 00:08:52,800
based on what what you train them for

246
00:08:52,800 --> 00:08:53,680
you know like

247
00:08:53,680 --> 00:08:57,040
identifying animal anomalies in a

248
00:08:57,040 --> 00:08:59,680
network of you know traffic or whatever

249
00:08:59,680 --> 00:09:00,480
it is

250
00:09:00,480 --> 00:09:02,800
because you can't do it like human being

251
00:09:02,800 --> 00:09:03,839
can't do that

252
00:09:03,839 --> 00:09:07,519
but yeah messy you know a neural network

253
00:09:07,519 --> 00:09:08,000
can

254
00:09:08,000 --> 00:09:10,880
yeah uh and this is i think also an

255
00:09:10,880 --> 00:09:12,240
important point because we talk a little

256
00:09:12,240 --> 00:09:13,680
bit about machine learning and you also

257
00:09:13,680 --> 00:09:15,040
talk about neural networks and there's

258
00:09:15,040 --> 00:09:16,000
there's something of a difference

259
00:09:16,000 --> 00:09:17,680
between the machine learning is uh the

260
00:09:17,680 --> 00:09:19,200
application of somewhat conventional

261
00:09:19,200 --> 00:09:20,560
statistical methods

262
00:09:20,560 --> 00:09:24,080
uh to a problem using big data in very

263
00:09:24,080 --> 00:09:25,120
large compute

264
00:09:25,120 --> 00:09:28,959
and neural network methodology relies on

265
00:09:28,959 --> 00:09:31,200
somewhat more abstract and complex

266
00:09:31,200 --> 00:09:32,240
algorithms

267
00:09:32,240 --> 00:09:35,680
and the process by which a neural

268
00:09:35,680 --> 00:09:36,240
network

269
00:09:36,240 --> 00:09:38,640
reaches its conclusion or reaches its

270
00:09:38,640 --> 00:09:40,080
output

271
00:09:40,080 --> 00:09:42,399
is often very mysterious to the operator

272
00:09:42,399 --> 00:09:44,320
it's difficult to understand

273
00:09:44,320 --> 00:09:46,880
and i know from writing about this space

274
00:09:46,880 --> 00:09:48,399
that policy makers

275
00:09:48,399 --> 00:09:50,560
and people that have to be accountable

276
00:09:50,560 --> 00:09:51,519
to

277
00:09:51,519 --> 00:09:55,519
elected electors and at voters

278
00:09:55,519 --> 00:09:58,320
they are sometimes very hesitant to

279
00:09:58,320 --> 00:09:59,040
employ

280
00:09:59,040 --> 00:10:00,959
neural networks precisely because they

281
00:10:00,959 --> 00:10:02,480
have difficulty

282
00:10:02,480 --> 00:10:05,760
describing the process by which that

283
00:10:05,760 --> 00:10:09,040
uh neural network reached the conclusion

284
00:10:09,040 --> 00:10:09,760
that it did

285
00:10:09,760 --> 00:10:11,360
how do you tell a voter that a neural

286
00:10:11,360 --> 00:10:13,200
network decided that uh

287
00:10:13,200 --> 00:10:15,200
they were suspicious and deserved extra

288
00:10:15,200 --> 00:10:16,959
security check like it's it's a very

289
00:10:16,959 --> 00:10:18,800
inconvenient conversation

290
00:10:18,800 --> 00:10:22,000
so where are we right now in terms of

291
00:10:22,000 --> 00:10:23,519
taking neural networks

292
00:10:23,519 --> 00:10:26,640
and making them more explainable to

293
00:10:26,640 --> 00:10:29,519
uh the public and to operators or is

294
00:10:29,519 --> 00:10:31,279
that not something that's actually

295
00:10:31,279 --> 00:10:32,320
possible

296
00:10:32,320 --> 00:10:36,320
uh given the nature of that tool

297
00:10:36,320 --> 00:10:38,959
yes that's that's a very very good kind

298
00:10:38,959 --> 00:10:40,560
of indication that

299
00:10:40,560 --> 00:10:43,519
it's in my opinion the nature of of this

300
00:10:43,519 --> 00:10:46,079
technology the neural networks that it's

301
00:10:46,079 --> 00:10:49,360
it's kind of universal tool to

302
00:10:49,360 --> 00:10:52,800
uh to work with data and and uh

303
00:10:52,800 --> 00:10:54,959
to be able to make some decisions and

304
00:10:54,959 --> 00:10:56,560
you have some universal

305
00:10:56,560 --> 00:10:59,519
universal algorithm how you train this

306
00:10:59,519 --> 00:11:00,720
these networks

307
00:11:00,720 --> 00:11:03,600
but then it's usually very difficult to

308
00:11:03,600 --> 00:11:06,399
know exactly why they do the decisions

309
00:11:06,399 --> 00:11:08,720
that they do and what is being more and

310
00:11:08,720 --> 00:11:10,560
more uncovered

311
00:11:10,560 --> 00:11:13,760
and shown is that while there are those

312
00:11:13,760 --> 00:11:15,360
fancy examples of

313
00:11:15,360 --> 00:11:18,720
neural networks doing almost magic then

314
00:11:18,720 --> 00:11:21,600
uh it's quite often very easy to fool

315
00:11:21,600 --> 00:11:24,640
them to make completely wrong decisions

316
00:11:24,640 --> 00:11:26,480
for things that for human beings are

317
00:11:26,480 --> 00:11:28,000
obvious you know because we

318
00:11:28,000 --> 00:11:31,000
just don't see all the all the

319
00:11:31,000 --> 00:11:33,680
vulnerabilities of the design

320
00:11:33,680 --> 00:11:36,160
so and this speaks to a growing area of

321
00:11:36,160 --> 00:11:37,519
research and study which is

322
00:11:37,519 --> 00:11:40,240
safety in artificial intelligence uh

323
00:11:40,240 --> 00:11:40,800
it's

324
00:11:40,800 --> 00:11:41,920
something that's becoming more and more

325
00:11:41,920 --> 00:11:43,360
important especially as governments and

326
00:11:43,360 --> 00:11:45,120
militaries begin to adopt artificial

327
00:11:45,120 --> 00:11:46,480
intelligence solutions on top of

328
00:11:46,480 --> 00:11:48,399
businesses it's one thing if you're

329
00:11:48,399 --> 00:11:50,639
if your algorithm outputs you a wrong

330
00:11:50,639 --> 00:11:52,720
suggestion for showing an ad to somebody

331
00:11:52,720 --> 00:11:55,040
it's very different if your output

332
00:11:55,040 --> 00:11:56,720
sends you wrong targeting information

333
00:11:56,720 --> 00:11:59,120
for a military operation that's very

334
00:11:59,120 --> 00:12:02,160
bad generally speaking so

335
00:12:02,160 --> 00:12:04,800
when you look at the resources that are

336
00:12:04,800 --> 00:12:06,480
required to really

337
00:12:06,480 --> 00:12:09,519
run a good artificial intelligence

338
00:12:09,519 --> 00:12:10,000
program

339
00:12:10,000 --> 00:12:12,079
and there are are three you need a lot

340
00:12:12,079 --> 00:12:13,040
of data

341
00:12:13,040 --> 00:12:15,519
uh you need a lot of compute power which

342
00:12:15,519 --> 00:12:16,959
is today available through

343
00:12:16,959 --> 00:12:19,920
uh enterprise compute solutions like

344
00:12:19,920 --> 00:12:21,760
azure from microsoft aws

345
00:12:21,760 --> 00:12:23,680
all sorts i don't have any connection to

346
00:12:23,680 --> 00:12:26,320
them and you need these statistical

347
00:12:26,320 --> 00:12:28,959
algorithms when you look at those

348
00:12:28,959 --> 00:12:30,160
resources

349
00:12:30,160 --> 00:12:33,680
do you see a future where attackers

350
00:12:33,680 --> 00:12:36,480
even very low level criminal groups will

351
00:12:36,480 --> 00:12:37,519
potentially have

352
00:12:37,519 --> 00:12:40,639
access to them in the same way as

353
00:12:40,639 --> 00:12:43,040
large institutions as governments and as

354
00:12:43,040 --> 00:12:44,720
entities that are

355
00:12:44,720 --> 00:12:46,959
bound to defend are those resources

356
00:12:46,959 --> 00:12:49,440
becoming more available

357
00:12:49,440 --> 00:12:52,000
well i i would say that you know there's

358
00:12:52,000 --> 00:12:53,519
simply a lot of

359
00:12:53,519 --> 00:12:56,480
money in in cyber crime and you have all

360
00:12:56,480 --> 00:12:57,360
the different

361
00:12:57,360 --> 00:13:00,720
like groups you know some some are

362
00:13:00,720 --> 00:13:04,399
really just pure uh

363
00:13:04,399 --> 00:13:06,399
criminals looking for some financial

364
00:13:06,399 --> 00:13:08,680
gains but we know that there is

365
00:13:08,680 --> 00:13:11,600
state-sponsored you know cyber uh thing

366
00:13:11,600 --> 00:13:12,800
going on so

367
00:13:12,800 --> 00:13:14,639
i i think there is already enough money

368
00:13:14,639 --> 00:13:16,320
to use whatever

369
00:13:16,320 --> 00:13:18,320
tools that are available if we go down

370
00:13:18,320 --> 00:13:21,120
to low-level criminals as you indicated

371
00:13:21,120 --> 00:13:24,720
i i would say even right now the

372
00:13:24,720 --> 00:13:28,959
the ability to design your own uh

373
00:13:28,959 --> 00:13:31,120
you know neural networks and to use it

374
00:13:31,120 --> 00:13:33,120
to process your own set of data is

375
00:13:33,120 --> 00:13:35,279
available basically to anybody already

376
00:13:35,279 --> 00:13:36,240
right now

377
00:13:36,240 --> 00:13:38,880
and i think this availability will be

378
00:13:38,880 --> 00:13:40,639
even bigger and so it will be more and

379
00:13:40,639 --> 00:13:41,760
more easy to

380
00:13:41,760 --> 00:13:44,160
design more and more complex networks

381
00:13:44,160 --> 00:13:45,199
and use all the

382
00:13:45,199 --> 00:13:47,519
results of the research that is going on

383
00:13:47,519 --> 00:13:48,720
all the time so yes

384
00:13:48,720 --> 00:13:51,040
from this perspective definitely is but

385
00:13:51,040 --> 00:13:52,399
to be able to

386
00:13:52,399 --> 00:13:55,760
correctly use or efficiently use this

387
00:13:55,760 --> 00:13:58,800
technology for your for your means

388
00:13:58,800 --> 00:14:00,639
is is a different thing it's not so

389
00:14:00,639 --> 00:14:02,720
obvious it's not so

390
00:14:02,720 --> 00:14:05,680
it as i say it's it's not intelligence

391
00:14:05,680 --> 00:14:07,760
by itself it's not working on its own

392
00:14:07,760 --> 00:14:08,399
you know

393
00:14:08,399 --> 00:14:11,839
you need to employ it yeah and

394
00:14:11,839 --> 00:14:14,560
to your mind legislative efforts

395
00:14:14,560 --> 00:14:15,680
regulations

396
00:14:15,680 --> 00:14:18,959
like gdpr like different regulation

397
00:14:18,959 --> 00:14:19,920
regimes

398
00:14:19,920 --> 00:14:22,399
that apply specifically to data and

399
00:14:22,399 --> 00:14:23,199
often

400
00:14:23,199 --> 00:14:25,440
personal or private data do you see

401
00:14:25,440 --> 00:14:28,000
those as having a positive or negative

402
00:14:28,000 --> 00:14:28,959
effect

403
00:14:28,959 --> 00:14:31,600
on this threat that you mentioned it

404
00:14:31,600 --> 00:14:33,199
obviously makes it a little bit more

405
00:14:33,199 --> 00:14:34,560
complicated to run

406
00:14:34,560 --> 00:14:36,800
a legitimate business because data is

407
00:14:36,800 --> 00:14:39,120
harder to come by because of gdpr

408
00:14:39,120 --> 00:14:41,279
but does it also decrease the threat

409
00:14:41,279 --> 00:14:44,320
what is your opinion

410
00:14:44,320 --> 00:14:47,279
so i i noticed there there is a draft of

411
00:14:47,279 --> 00:14:48,720
what is called i think

412
00:14:48,720 --> 00:14:51,600
artificial intelligence act that is

413
00:14:51,600 --> 00:14:52,079
being

414
00:14:52,079 --> 00:14:55,360
uh created by european union

415
00:14:55,360 --> 00:14:57,839
european commissioner and it really

416
00:14:57,839 --> 00:14:59,040
tries to address the

417
00:14:59,040 --> 00:15:02,000
some of the known concerns like for

418
00:15:02,000 --> 00:15:02,800
example

419
00:15:02,800 --> 00:15:06,000
uh using uh artificial intelligence to

420
00:15:06,000 --> 00:15:08,880
kind of score how people behave if you

421
00:15:08,880 --> 00:15:10,399
cross the street

422
00:15:10,399 --> 00:15:11,920
with the red light and these kind of

423
00:15:11,920 --> 00:15:14,959
things and and it explicitly says like

424
00:15:14,959 --> 00:15:18,079
this is prohibited this is not allowed

425
00:15:18,079 --> 00:15:19,440
way of using this

426
00:15:19,440 --> 00:15:21,760
and it goes to many different areas and

427
00:15:21,760 --> 00:15:23,519
actually sounds quite

428
00:15:23,519 --> 00:15:26,639
reasonable to me but when we talk about

429
00:15:26,639 --> 00:15:28,399
cyber crime of course they are not

430
00:15:28,399 --> 00:15:31,839
limited by by legislation so it's more

431
00:15:31,839 --> 00:15:33,120
about

432
00:15:33,120 --> 00:15:35,120
uh probably the leadership the

433
00:15:35,120 --> 00:15:37,519
government's not misusing

434
00:15:37,519 --> 00:15:39,680
machine learning in areas where it

435
00:15:39,680 --> 00:15:43,120
already can do some harm

436
00:15:43,360 --> 00:15:44,560
there's a couple of questions that we've

437
00:15:44,560 --> 00:15:46,639
got and i'd like to ask the first one

438
00:15:46,639 --> 00:15:48,160
this is an area that i'm very interested

439
00:15:48,160 --> 00:15:50,639
in i think that uh anybody that

440
00:15:50,639 --> 00:15:52,000
lives in eastern europe and is very

441
00:15:52,000 --> 00:15:54,639
concerned about

442
00:15:54,639 --> 00:15:57,120
russian misinformation efforts should

443
00:15:57,120 --> 00:15:57,759
also be

444
00:15:57,759 --> 00:16:00,240
interested in how does disinformation

445
00:16:00,240 --> 00:16:02,240
come into the future of ai

446
00:16:02,240 --> 00:16:04,079
is it a threat to its future development

447
00:16:04,079 --> 00:16:05,279
so a little bit on

448
00:16:05,279 --> 00:16:08,399
disinformation and uh

449
00:16:08,399 --> 00:16:10,639
influence campaigns and the role of

450
00:16:10,639 --> 00:16:12,480
artificial intelligence in those in the

451
00:16:12,480 --> 00:16:14,000
future

452
00:16:14,000 --> 00:16:16,560
right i i guess this is unfortunately

453
00:16:16,560 --> 00:16:18,160
the space where

454
00:16:18,160 --> 00:16:21,040
where artificial intelligence even even

455
00:16:21,040 --> 00:16:21,680
with

456
00:16:21,680 --> 00:16:25,120
the current state is able to

457
00:16:25,120 --> 00:16:28,399
to help like to do to do harm so

458
00:16:28,399 --> 00:16:32,240
uh you

459
00:16:32,320 --> 00:16:34,240
what body what the artificial

460
00:16:34,240 --> 00:16:36,320
intelligence is able to do is to

461
00:16:36,320 --> 00:16:39,199
use the you know the existing text the

462
00:16:39,199 --> 00:16:39,759
existing

463
00:16:39,759 --> 00:16:41,600
information available on the internet

464
00:16:41,600 --> 00:16:44,399
and then to combine to produce new text

465
00:16:44,399 --> 00:16:46,480
and the text might even look like

466
00:16:46,480 --> 00:16:48,320
reasonable but they might be

467
00:16:48,320 --> 00:16:50,399
false they might be not based on any

468
00:16:50,399 --> 00:16:52,000
real um

469
00:16:52,000 --> 00:16:54,800
evidence or whatever then we have this

470
00:16:54,800 --> 00:16:55,519
uh

471
00:16:55,519 --> 00:16:57,759
relatively well known like deep deep

472
00:16:57,759 --> 00:16:59,680
fake you know this fake videos

473
00:16:59,680 --> 00:17:02,000
right that are already quite good and

474
00:17:02,000 --> 00:17:04,880
will only be better in the future so

475
00:17:04,880 --> 00:17:08,000
so like it really can be helped to

476
00:17:08,000 --> 00:17:10,559
to create chaos to to make it very

477
00:17:10,559 --> 00:17:13,280
difficult to distinguish reality from

478
00:17:13,280 --> 00:17:16,880
from like just whatever

479
00:17:16,880 --> 00:17:20,880
do you see a role of ai in helping

480
00:17:20,880 --> 00:17:23,919
attackers determine which individuals

481
00:17:23,919 --> 00:17:26,160
might be most susceptible to

482
00:17:26,160 --> 00:17:27,919
disinformation

483
00:17:27,919 --> 00:17:30,960
as well as what messages they might be

484
00:17:30,960 --> 00:17:34,080
vulnerable to is that also a role for ai

485
00:17:34,080 --> 00:17:37,280
uh sure because i i think we already see

486
00:17:37,280 --> 00:17:39,120
it with uh

487
00:17:39,120 --> 00:17:42,559
being used in in uh commercial practice

488
00:17:42,559 --> 00:17:43,600
you know like

489
00:17:43,600 --> 00:17:46,080
youtube and so on you you get you know

490
00:17:46,080 --> 00:17:48,240
the recommendations for your next videos

491
00:17:48,240 --> 00:17:48,960
and uh

492
00:17:48,960 --> 00:17:51,200
probably all of us whoever used it we

493
00:17:51,200 --> 00:17:53,760
see that it actually works well so

494
00:17:53,760 --> 00:17:56,640
the these algorithms are able to kind of

495
00:17:56,640 --> 00:17:58,160
identify the bubbles

496
00:17:58,160 --> 00:18:01,360
like these these pieces that make

497
00:18:01,360 --> 00:18:03,120
people believe that this is something

498
00:18:03,120 --> 00:18:05,440
they want to read or they are interested

499
00:18:05,440 --> 00:18:07,360
in that and so then you can couple it

500
00:18:07,360 --> 00:18:08,880
with whatever

501
00:18:08,880 --> 00:18:11,600
misinformation you want and probably the

502
00:18:11,600 --> 00:18:13,200
uh the impact will be

503
00:18:13,200 --> 00:18:15,520
significant yeah so i think this is a

504
00:18:15,520 --> 00:18:16,640
really important point because if you

505
00:18:16,640 --> 00:18:17,760
watched

506
00:18:17,760 --> 00:18:19,280
russian disinformation campaigns aimed

507
00:18:19,280 --> 00:18:21,280
at the u.s electorate in 2016 for

508
00:18:21,280 --> 00:18:22,400
example

509
00:18:22,400 --> 00:18:26,000
what you saw was an expert use

510
00:18:26,000 --> 00:18:29,120
of what is given away information to

511
00:18:29,120 --> 00:18:30,799
social media outlets

512
00:18:30,799 --> 00:18:32,320
like facebook information that we all

513
00:18:32,320 --> 00:18:33,840
give away basically all the time whether

514
00:18:33,840 --> 00:18:35,440
we kind of realize it or not even if

515
00:18:35,440 --> 00:18:36,799
we're not on those networks

516
00:18:36,799 --> 00:18:40,320
uh currently that used in order to

517
00:18:40,320 --> 00:18:41,760
deliver very specific people very

518
00:18:41,760 --> 00:18:42,720
deliverance and

519
00:18:42,720 --> 00:18:44,240
information messages and the

520
00:18:44,240 --> 00:18:45,840
accelerating effect of ai just means a

521
00:18:45,840 --> 00:18:47,760
lot more of that so that is the

522
00:18:47,760 --> 00:18:49,120
i think highlights the importance of

523
00:18:49,120 --> 00:18:51,600
potential regulation and on regulation

524
00:18:51,600 --> 00:18:53,440
very good question from tyson barker

525
00:18:53,440 --> 00:18:55,600
what do you think of the explainability

526
00:18:55,600 --> 00:18:56,559
requirement

527
00:18:56,559 --> 00:19:00,160
in the draft ai regulation

528
00:19:00,160 --> 00:19:03,440
um so

529
00:19:03,440 --> 00:19:06,480
um just please try to explain i

530
00:19:06,480 --> 00:19:10,160
uh yeah so the draft ai regulation

531
00:19:10,160 --> 00:19:13,120
uh right explain ability okay i see

532
00:19:13,120 --> 00:19:13,600
right

533
00:19:13,600 --> 00:19:16,720
um so

534
00:19:16,720 --> 00:19:20,320
you know you you like know you compared

535
00:19:20,320 --> 00:19:22,400
artificial intelligence to the

536
00:19:22,400 --> 00:19:23,600
statistical

537
00:19:23,600 --> 00:19:26,480
kind of models and methods and yes with

538
00:19:26,480 --> 00:19:27,280
statistical

539
00:19:27,280 --> 00:19:29,760
models we usually you can usually tell

540
00:19:29,760 --> 00:19:30,559
like

541
00:19:30,559 --> 00:19:32,960
why the model is doing what what it is

542
00:19:32,960 --> 00:19:35,120
doing with artificial intelligence is

543
00:19:35,120 --> 00:19:36,000
not

544
00:19:36,000 --> 00:19:38,480
uh usually not possible at all so

545
00:19:38,480 --> 00:19:39,360
probably

546
00:19:39,360 --> 00:19:42,400
like trying to to be able to

547
00:19:42,400 --> 00:19:45,280
to tell like why it works the way it

548
00:19:45,280 --> 00:19:47,039
works is going to be difficult

549
00:19:47,039 --> 00:19:50,640
but you can do some countermeasures to

550
00:19:50,640 --> 00:19:53,120
ensure higher safety so

551
00:19:53,120 --> 00:19:55,679
a very simple one or very general is

552
00:19:55,679 --> 00:19:57,520
just human supervised on

553
00:19:57,520 --> 00:19:59,840
critical like decisions you know that

554
00:19:59,840 --> 00:20:00,960
it's because

555
00:20:00,960 --> 00:20:03,200
human being has this ability to

556
00:20:03,200 --> 00:20:05,120
distinguish obvious

557
00:20:05,120 --> 00:20:08,240
wrongness from uh

558
00:20:08,240 --> 00:20:11,919
from reasonable recommendation

559
00:20:11,919 --> 00:20:13,360
i think that this speaks a little bit to

560
00:20:13,360 --> 00:20:15,280
some of the efforts that we've seen

561
00:20:15,280 --> 00:20:17,919
uh by militaries particularly the us

562
00:20:17,919 --> 00:20:18,960
military

563
00:20:18,960 --> 00:20:21,600
to employ artificial intelligence in a

564
00:20:21,600 --> 00:20:23,520
way that keeps human beings accountable

565
00:20:23,520 --> 00:20:24,720
i don't know if you're

566
00:20:24,720 --> 00:20:27,120
familiar with it but uh the u.s military

567
00:20:27,120 --> 00:20:28,720
has adopted a long list

568
00:20:28,720 --> 00:20:30,880
of ethical principles that they'll use

569
00:20:30,880 --> 00:20:33,280
when applying ai to their operations

570
00:20:33,280 --> 00:20:35,280
uh hopefully uh it's very good that they

571
00:20:35,280 --> 00:20:36,880
adopted the the list

572
00:20:36,880 --> 00:20:38,720
whether or not it's employed as

573
00:20:38,720 --> 00:20:40,480
prescribed that becomes something that

574
00:20:40,480 --> 00:20:41,760
individual commanders will make

575
00:20:41,760 --> 00:20:42,960
decisions about

576
00:20:42,960 --> 00:20:45,840
uh but to your mind have you seen any of

577
00:20:45,840 --> 00:20:46,799
the lists

578
00:20:46,799 --> 00:20:48,400
coming out of google coming out of

579
00:20:48,400 --> 00:20:50,480
microsoft coming out of other

580
00:20:50,480 --> 00:20:53,360
big silicon valley players lists of

581
00:20:53,360 --> 00:20:54,960
ethical principles

582
00:20:54,960 --> 00:20:57,280
that apply to the use of artificial

583
00:20:57,280 --> 00:20:58,320
intelligence

584
00:20:58,320 --> 00:20:59,919
and and what do you think of them are

585
00:20:59,919 --> 00:21:01,840
they sufficient or do they not go far

586
00:21:01,840 --> 00:21:04,159
enough

587
00:21:04,159 --> 00:21:07,280
uh well i i didn't study this in detail

588
00:21:07,280 --> 00:21:11,039
i noticed this is a kind of regular kind

589
00:21:11,039 --> 00:21:12,559
of input so

590
00:21:12,559 --> 00:21:15,520
i you know you can you can stay on a

591
00:21:15,520 --> 00:21:17,120
quite general level and

592
00:21:17,120 --> 00:21:19,360
on that level it would just sound good

593
00:21:19,360 --> 00:21:21,440
you simply don't want to have

594
00:21:21,440 --> 00:21:24,159
artificial intelligence being used

595
00:21:24,159 --> 00:21:26,400
against humans

596
00:21:26,400 --> 00:21:29,679
misuse you know their weaknesses

597
00:21:29,679 --> 00:21:32,320
against them but then what it means in

598
00:21:32,320 --> 00:21:33,360
practice is

599
00:21:33,360 --> 00:21:35,520
more tricky because we you gave an

600
00:21:35,520 --> 00:21:37,200
example facebook

601
00:21:37,200 --> 00:21:40,000
but we know that for facebook is like

602
00:21:40,000 --> 00:21:42,000
from a commercial perspective is

603
00:21:42,000 --> 00:21:44,640
is kind of helping them if if you know

604
00:21:44,640 --> 00:21:45,440
when people

605
00:21:45,440 --> 00:21:48,960
are getting into you know fights and

606
00:21:48,960 --> 00:21:51,919
arguing about uh their point of views

607
00:21:51,919 --> 00:21:52,400
and

608
00:21:52,400 --> 00:21:55,520
bubbles clashing and so on and so

609
00:21:55,520 --> 00:21:59,840
you know like balancing these

610
00:21:59,840 --> 00:22:03,120
ethical aspects with uh with the pure

611
00:22:03,120 --> 00:22:03,679
business

612
00:22:03,679 --> 00:22:06,640
that's that's a tricky thing yeah uh

613
00:22:06,640 --> 00:22:07,440
very interesting

614
00:22:07,440 --> 00:22:10,240
research paper out quite uh recently

615
00:22:10,240 --> 00:22:10,720
that

616
00:22:10,720 --> 00:22:13,039
looks at viral tweets viral social

617
00:22:13,039 --> 00:22:14,880
messages and the ones that go most viral

618
00:22:14,880 --> 00:22:16,240
tend to be the ones that uh

619
00:22:16,240 --> 00:22:18,799
state an angry political opinion believe

620
00:22:18,799 --> 00:22:19,679
it or not

621
00:22:19,679 --> 00:22:21,679
so uh you can imagine that being

622
00:22:21,679 --> 00:22:22,799
weaponized

623
00:22:22,799 --> 00:22:25,360
uh and automated and as as you say

624
00:22:25,360 --> 00:22:26,000
accelerated

625
00:22:26,000 --> 00:22:27,679
and scaled up to the point where human

626
00:22:27,679 --> 00:22:29,120
beings can't uh

627
00:22:29,120 --> 00:22:31,600
properly control or defend the onslaught

628
00:22:31,600 --> 00:22:32,240
of that

629
00:22:32,240 --> 00:22:34,080
uh new messaging and then that makes

630
00:22:34,080 --> 00:22:35,440
virality in

631
00:22:35,440 --> 00:22:38,720
uh in internet context almost irrelevant

632
00:22:38,720 --> 00:22:41,520
uh just destroying the joy of social

633
00:22:41,520 --> 00:22:42,480
media for everyone

634
00:22:42,480 --> 00:22:45,200
if you had any uh very good question

635
00:22:45,200 --> 00:22:45,840
here

636
00:22:45,840 --> 00:22:49,200
from uh dr vera robiscrinka

637
00:22:49,200 --> 00:22:51,200
russia and china are investing in cyber

638
00:22:51,200 --> 00:22:52,400
capabilities

639
00:22:52,400 --> 00:22:55,440
how do you see a role of the eu in

640
00:22:55,440 --> 00:22:57,360
tackling potential future

641
00:22:57,360 --> 00:23:00,159
superiority in the cyber domain so

642
00:23:00,159 --> 00:23:01,919
russia and china are also investing a

643
00:23:01,919 --> 00:23:03,520
lot in artificial intelligence

644
00:23:03,520 --> 00:23:06,000
uh they both have cyber operations

645
00:23:06,000 --> 00:23:06,880
offensive

646
00:23:06,880 --> 00:23:08,880
uh and of course have constructed

647
00:23:08,880 --> 00:23:10,799
elaborate cyber defensive

648
00:23:10,799 --> 00:23:13,840
measures as well what is the role of the

649
00:23:13,840 --> 00:23:14,880
eu

650
00:23:14,880 --> 00:23:18,640
in helping democratic nation states

651
00:23:18,640 --> 00:23:22,080
better compete in that area

652
00:23:22,080 --> 00:23:25,840
right so i i guess the

653
00:23:25,840 --> 00:23:29,440
uh the focus uh already moved from

654
00:23:29,440 --> 00:23:31,679
from fighting and battles like a

655
00:23:31,679 --> 00:23:32,480
physical

656
00:23:32,480 --> 00:23:35,600
uh to more like virtual right

657
00:23:35,600 --> 00:23:38,480
and so it's it's logical that there is

658
00:23:38,480 --> 00:23:38,880
uh

659
00:23:38,880 --> 00:23:41,200
quite a bit of investment in this from

660
00:23:41,200 --> 00:23:42,960
all the super powers

661
00:23:42,960 --> 00:23:46,159
now the eu just needs to

662
00:23:46,159 --> 00:23:49,679
decide what uh what role

663
00:23:49,679 --> 00:23:52,320
they want to what we want to we want to

664
00:23:52,320 --> 00:23:52,960
have

665
00:23:52,960 --> 00:23:55,200
i would we know the statistics the

666
00:23:55,200 --> 00:23:57,279
statistics are saying the investment

667
00:23:57,279 --> 00:24:01,039
to cyber security in general in startups

668
00:24:01,039 --> 00:24:03,440
like 90 percent goes to u.s and let's

669
00:24:03,440 --> 00:24:04,720
say uh

670
00:24:04,720 --> 00:24:07,840
israel for example so it's just a tiny

671
00:24:07,840 --> 00:24:10,080
portion that is left for the rest

672
00:24:10,080 --> 00:24:14,240
so i i would suppose that

673
00:24:14,240 --> 00:24:17,520
we we already see that the cyber

674
00:24:17,520 --> 00:24:20,720
security is becoming more uh

675
00:24:20,720 --> 00:24:24,000
kind of china is closing what what kind

676
00:24:24,000 --> 00:24:24,320
of

677
00:24:24,320 --> 00:24:26,880
technologies are allowed to be used by

678
00:24:26,880 --> 00:24:28,159
chinese

679
00:24:28,159 --> 00:24:31,520
uh organizations or institutions we see

680
00:24:31,520 --> 00:24:33,200
this in other parts of the world russia

681
00:24:33,200 --> 00:24:34,400
for example

682
00:24:34,400 --> 00:24:38,240
so probably european union also needs to

683
00:24:38,240 --> 00:24:41,200
ensure that we that there is enough uh

684
00:24:41,200 --> 00:24:43,279
research and development going on within

685
00:24:43,279 --> 00:24:44,720
european union

686
00:24:44,720 --> 00:24:46,320
in this space whether it's cyber

687
00:24:46,320 --> 00:24:48,400
security in general or machine learning

688
00:24:48,400 --> 00:24:50,320
artificial intelligence

689
00:24:50,320 --> 00:24:53,840
um like a field of itself yeah so

690
00:24:53,840 --> 00:24:56,159
so i think like recognizing or

691
00:24:56,159 --> 00:24:57,919
identifying the

692
00:24:57,919 --> 00:25:00,080
what we already have and how to boost it

693
00:25:00,080 --> 00:25:02,400
further so that we are not dependent

694
00:25:02,400 --> 00:25:04,559
on the technologies being being imported

695
00:25:04,559 --> 00:25:05,520
from abroad

696
00:25:05,520 --> 00:25:08,720
okay uh so a uh a final note

697
00:25:08,720 --> 00:25:11,279
a a endorsement for the importance of

698
00:25:11,279 --> 00:25:12,400
alliances and the

699
00:25:12,400 --> 00:25:13,919
uh endorsement for the importance of

700
00:25:13,919 --> 00:25:15,520
individual individual and independent

701
00:25:15,520 --> 00:25:17,120
research dollars in this very important

702
00:25:17,120 --> 00:25:17,760
field

703
00:25:17,760 --> 00:25:19,279
uh that is all the time we have for this

704
00:25:19,279 --> 00:25:21,120
session i want to thank you and remind

705
00:25:21,120 --> 00:25:22,400
you all please stay here

706
00:25:22,400 --> 00:25:25,120
there's a session on 5g coming up uh

707
00:25:25,120 --> 00:25:26,080
right now

708
00:25:26,080 --> 00:25:27,840
but thank you very much please join me

709
00:25:27,840 --> 00:25:43,560
in thanking my guests

710
00:25:43,560 --> 00:25:45,920
[Music]

711
00:25:45,920 --> 00:25:47,950
[Applause]

712
00:25:47,950 --> 00:25:51,650
[Music]

713
00:26:14,880 --> 00:26:16,960
you