1
00:00:02,750 --> 00:00:09,960
as you just heard I'm Robin victory and

2
00:00:05,839 --> 00:00:11,850
our Czech penetration tester work for

3
00:00:09,960 --> 00:00:16,440
company called commerson and based in

4
00:00:11,850 --> 00:00:20,400
Edinburgh and we do normal penetration

5
00:00:16,440 --> 00:00:24,650
testing stuff so YouTube check worth IT

6
00:00:20,400 --> 00:00:27,270
health checks code and build reviews

7
00:00:24,650 --> 00:00:30,779
internal and external penetration tests

8
00:00:27,270 --> 00:00:33,210
and mobile web application testing and

9
00:00:30,780 --> 00:00:36,420
it's on the internal penetration test

10
00:00:33,210 --> 00:00:40,170
that we almost always come up with some

11
00:00:36,420 --> 00:00:43,110
kind of password cracking as usually

12
00:00:40,170 --> 00:00:46,140
because we've got access to some

13
00:00:43,110 --> 00:00:49,890
low-hanging fruit on the network we've

14
00:00:46,140 --> 00:00:53,220
managed to Drabek is a TDS dead based on

15
00:00:49,890 --> 00:00:55,440
the domain controller and we're either

16
00:00:53,220 --> 00:00:59,610
turn to audited passwords for the client

17
00:00:55,440 --> 00:01:01,170
or we're trying to track passwords the

18
00:00:59,610 --> 00:01:05,250
main passwords to use on long domain

19
00:01:01,170 --> 00:01:08,130
connected devices so whichever we're

20
00:01:05,250 --> 00:01:12,030
doing we usually do the same fingers few

21
00:01:08,130 --> 00:01:13,350
much everyone else does we got a

22
00:01:12,030 --> 00:01:16,079
standard dictionary of what key lock

23
00:01:13,350 --> 00:01:20,939
types is a favorite one but in hash cap

24
00:01:16,079 --> 00:01:25,708
if varying rules depending on how much

25
00:01:20,939 --> 00:01:28,020
time we've got so yeah my favorite is

26
00:01:25,709 --> 00:01:33,539
one Baltimore more others use divert all

27
00:01:28,020 --> 00:01:36,509
those up over 54 and speedfest and that

28
00:01:33,539 --> 00:01:39,929
gets us a rather large amount of

29
00:01:36,509 --> 00:01:44,749
passwords you see people aren't very

30
00:01:39,929 --> 00:01:48,509
good with passwords but in order to

31
00:01:44,749 --> 00:01:53,759
improve things if and more what we need

32
00:01:48,509 --> 00:01:57,840
to do is start customizing the generic

33
00:01:53,759 --> 00:02:00,060
word lists to our customers and I start

34
00:01:57,840 --> 00:02:05,579
off with some of my own wordless which I

35
00:02:00,060 --> 00:02:10,649
carry straight from Wikipedia or other

36
00:02:05,579 --> 00:02:12,420
likes I got countries cities schools

37
00:02:10,649 --> 00:02:14,350
cottage place names with Spanish

38
00:02:12,420 --> 00:02:16,660
surnames

39
00:02:14,350 --> 00:02:18,220
I have registered Scotland towards my

40
00:02:16,660 --> 00:02:21,190
sister's all the baby names for each

41
00:02:18,220 --> 00:02:23,260
year and I stick all those at my

42
00:02:21,190 --> 00:02:26,970
password list cause I'm out by

43
00:02:23,260 --> 00:02:30,760
dictionary and that gets us a lot more

44
00:02:26,970 --> 00:02:32,350
but because this is Scotland I put in

45
00:02:30,760 --> 00:02:36,549
all the Scottish football clubs Allah

46
00:02:32,350 --> 00:02:40,120
well the football grounds every player I

47
00:02:36,550 --> 00:02:44,440
could get off Wikipedia from to pick

48
00:02:40,120 --> 00:02:50,680
four teams and again I guess it's

49
00:02:44,440 --> 00:02:54,130
another different passport but after

50
00:02:50,680 --> 00:03:00,870
that you have to start tearing things

51
00:02:54,130 --> 00:03:03,609
for a specific client looking at the

52
00:03:00,870 --> 00:03:06,430
normal thing you do is you spew stop

53
00:03:03,610 --> 00:03:08,620
calling the database using cool it gives

54
00:03:06,430 --> 00:03:11,860
you a list of words you see related to

55
00:03:08,620 --> 00:03:15,220
that particular customer you might want

56
00:03:11,860 --> 00:03:17,320
to call all of our websites which are in

57
00:03:15,220 --> 00:03:21,760
the same industry so competitors like

58
00:03:17,320 --> 00:03:24,190
regulator websites or industry or other

59
00:03:21,760 --> 00:03:28,899
industry websites to find other terms if

60
00:03:24,190 --> 00:03:31,570
you used youth in that company and then

61
00:03:28,900 --> 00:03:35,680
you start using what I call common sense

62
00:03:31,570 --> 00:03:39,010
you think about what their company does

63
00:03:35,680 --> 00:03:41,200
and you try to think of what terms they

64
00:03:39,010 --> 00:03:43,060
might be using his password so their

65
00:03:41,200 --> 00:03:48,040
transport company and they've got a very

66
00:03:43,060 --> 00:03:52,000
drivers brands of lorries breath stops

67
00:03:48,040 --> 00:03:55,390
main roads their ferry company you're

68
00:03:52,000 --> 00:04:02,350
looking for courtship names multiple

69
00:03:55,390 --> 00:04:05,579
terms if you look at where the company

70
00:04:02,350 --> 00:04:09,370
is located is there you're looking for

71
00:04:05,580 --> 00:04:14,680
any local landmarks the history of the

72
00:04:09,370 --> 00:04:17,530
area and you can go on but more time you

73
00:04:14,680 --> 00:04:18,880
spending on this look the last time you

74
00:04:17,529 --> 00:04:22,260
actually spending on doing the actual

75
00:04:18,880 --> 00:04:24,400
penetration test which you came to do so

76
00:04:22,260 --> 00:04:27,230
what would be good is if it could

77
00:04:24,400 --> 00:04:32,330
automate a lot of this of the work

78
00:04:27,230 --> 00:04:36,440
and that's whereas not looking at much

79
00:04:32,330 --> 00:04:40,070
with language processing because natural

80
00:04:36,440 --> 00:04:41,180
language processing uses what one of the

81
00:04:40,070 --> 00:04:43,370
first me to do in natural language

82
00:04:41,180 --> 00:04:46,700
processing is you tokenize all the words

83
00:04:43,370 --> 00:04:49,880
and open a text corpus and you turn them

84
00:04:46,700 --> 00:04:52,159
into a form which you communicate are

85
00:04:49,880 --> 00:04:56,750
not done through something called word

86
00:04:52,160 --> 00:04:59,840
embedding which translates every word

87
00:04:56,750 --> 00:05:01,700
that it finds in text corpus into a big

88
00:04:59,840 --> 00:05:05,979
vector describing its relationship to

89
00:05:01,700 --> 00:05:05,979
all the other works that are in course

90
00:05:07,690 --> 00:05:15,950
DS so weight usually works if you have a

91
00:05:13,790 --> 00:05:21,320
sliding window to use across the text

92
00:05:15,950 --> 00:05:25,849
corpus so say have 8 words within that

93
00:05:21,320 --> 00:05:29,150
window and you count all words which Co

94
00:05:25,850 --> 00:05:34,640
occur within that window and then you

95
00:05:29,150 --> 00:05:37,849
built up a matrix of those occurrences

96
00:05:34,640 --> 00:05:41,810
and then you end up with a massive

97
00:05:37,850 --> 00:05:44,560
massive model that describes all the

98
00:05:41,810 --> 00:05:51,220
words you then compress that down

99
00:05:44,560 --> 00:05:53,450
getting rid of all of the relationships

100
00:05:51,220 --> 00:05:56,150
the words which aren't related to many

101
00:05:53,450 --> 00:06:00,430
other words and you end up with is

102
00:05:56,150 --> 00:06:03,460
smaller but still quite hefty model but

103
00:06:00,430 --> 00:06:08,330
what happens then what you find is that

104
00:06:03,460 --> 00:06:13,039
words which occur in the similar context

105
00:06:08,330 --> 00:06:15,380
in your test in your training focus also

106
00:06:13,040 --> 00:06:17,890
have vectors which are clustered

107
00:06:15,380 --> 00:06:26,300
together in vectors in the vector space

108
00:06:17,890 --> 00:06:29,300
so here because the kitchen license will

109
00:06:26,300 --> 00:06:35,420
trusted nicely over now so what we can

110
00:06:29,300 --> 00:06:38,000
do if is you try to locate kitchen words

111
00:06:35,420 --> 00:06:38,680
is find all the words of effectiveness

112
00:06:38,000 --> 00:06:41,440
of cost

113
00:06:38,680 --> 00:06:45,460
causing me to take bathroom and it'll

114
00:06:41,440 --> 00:06:48,430
give you Elizabeth you also find that

115
00:06:45,460 --> 00:06:49,900
words have relationships get

116
00:06:48,430 --> 00:06:53,229
communication but that can be

117
00:06:49,900 --> 00:06:58,448
represented as possible vectors so if

118
00:06:53,229 --> 00:07:02,469
you you're fine disturbed the effector

119
00:06:58,449 --> 00:07:04,870
for our capital city and defector for

120
00:07:02,470 --> 00:07:09,780
the country which refers to always have

121
00:07:04,870 --> 00:07:13,030
a similar distance to so here we have

122
00:07:09,780 --> 00:07:14,770
the vectors between various countries

123
00:07:13,030 --> 00:07:19,119
and the capital cities and you see it

124
00:07:14,770 --> 00:07:21,299
all pretty similar and that also applies

125
00:07:19,120 --> 00:07:23,789
to other friends so there codes with the

126
00:07:21,300 --> 00:07:26,350
relationship between cream cream cream

127
00:07:23,789 --> 00:07:33,729
and women are also pretty single

128
00:07:26,350 --> 00:07:37,060
intensive vector representations so in

129
00:07:33,729 --> 00:07:38,860
terms of generating your orders you

130
00:07:37,060 --> 00:07:42,720
don't really want to talk yourself but

131
00:07:38,860 --> 00:07:49,630
luckily there's some big retains courses

132
00:07:42,720 --> 00:07:51,520
that you can get and they have trained

133
00:07:49,630 --> 00:07:55,659
these courses those obtain these models

134
00:07:51,520 --> 00:07:59,710
on publicly available text courses

135
00:07:55,659 --> 00:08:01,510
sure as long as that so pick one normal

136
00:07:59,710 --> 00:08:04,000
year so the Wikipedia top we find most

137
00:08:01,510 --> 00:08:08,860
listened to their 210 models use that

138
00:08:04,000 --> 00:08:10,800
and some former and this tears newswire

139
00:08:08,860 --> 00:08:16,419
the detectives a couple of constant use

140
00:08:10,800 --> 00:08:18,870
use wire tater as well and this is where

141
00:08:16,419 --> 00:08:23,830
you can get those future courses

142
00:08:18,870 --> 00:08:28,780
what's up details in order to use them

143
00:08:23,830 --> 00:08:33,900
we use a library called Jensen which was

144
00:08:28,780 --> 00:08:36,939
a - library that's how you pretty simple

145
00:08:33,900 --> 00:08:39,699
panted how did that she learned the

146
00:08:36,940 --> 00:08:45,970
models in the problem and this is how

147
00:08:39,700 --> 00:08:47,620
you actually an important load model and

148
00:08:45,970 --> 00:08:51,529
then finds that the word Summers read

149
00:08:47,620 --> 00:08:56,329
work so guess what trace

150
00:08:51,529 --> 00:08:59,480
expansion motion so that should return

151
00:08:56,329 --> 00:09:03,319
to us to find most similar words to

152
00:08:59,480 --> 00:09:04,660
editor and this is what it actually

153
00:09:03,319 --> 00:09:08,779
comes out with

154
00:09:04,660 --> 00:09:15,680
so on the left day you find words on the

155
00:09:08,779 --> 00:09:18,249
right you have ranges between 0 & 1

156
00:09:15,680 --> 00:09:23,209
that represents the similarity to the

157
00:09:18,249 --> 00:09:27,290
original way and here it's managed to

158
00:09:23,209 --> 00:09:29,089
find it all of the cities were decided

159
00:09:27,290 --> 00:09:32,599
by debt versus City and it's given us

160
00:09:29,089 --> 00:09:34,399
lots of Scottish cities through the

161
00:09:32,600 --> 00:09:39,170
summer

162
00:09:34,399 --> 00:09:47,420
the downside is that 3 minutes and 21

163
00:09:39,170 --> 00:09:49,490
seconds to him also in terms of using

164
00:09:47,420 --> 00:09:51,559
that to find of our candidates for

165
00:09:49,490 --> 00:09:55,189
password cracking that's not very

166
00:09:51,559 --> 00:09:57,410
accessible luckily and most of that is

167
00:09:55,189 --> 00:09:59,839
upfront cost and when you're looking at

168
00:09:57,410 --> 00:10:02,240
models and the initial search so here

169
00:09:59,839 --> 00:10:05,930
she's looking for additional words on

170
00:10:02,240 --> 00:10:12,379
top of that it takes roughly 3 seconds

171
00:10:05,930 --> 00:10:15,309
of code word so really we don't really

172
00:10:12,379 --> 00:10:17,809
be doing this through every word and

173
00:10:15,309 --> 00:10:19,790
don't be this button is for every word

174
00:10:17,809 --> 00:10:25,160
but you saw it once and you can once at

175
00:10:19,790 --> 00:10:29,120
it it's a deep so we can also improve

176
00:10:25,160 --> 00:10:33,949
the speed by indexing did there's a the

177
00:10:29,120 --> 00:10:36,670
indexes tend to reduce the accuracy

178
00:10:33,949 --> 00:10:38,449
which we have a somewhat more word about

179
00:10:36,670 --> 00:10:40,670
increases startup cost

180
00:10:38,449 --> 00:10:44,420
diversity in terms of the team and then

181
00:10:40,670 --> 00:10:46,910
is just a lot of costs modification so

182
00:10:44,420 --> 00:10:52,959
this is the same thing one as with

183
00:10:46,910 --> 00:10:52,959
indexing this uses the annoy indexer and

184
00:10:53,290 --> 00:10:59,029
qual from adding the index row to finish

185
00:10:55,699 --> 00:11:01,189
the same code but you won't find the

186
00:10:59,029 --> 00:11:03,600
results are different from that

187
00:11:01,189 --> 00:11:07,740
so here it's found Edinboro zone perfect

188
00:11:03,600 --> 00:11:11,010
which is most twice but Sofia where

189
00:11:07,740 --> 00:11:13,260
results are slightly different and not

190
00:11:11,010 --> 00:11:17,819
it's not going back with selfish cities

191
00:11:13,260 --> 00:11:22,560
of basking but they're all related in

192
00:11:17,820 --> 00:11:23,340
some way over to my his matches for

193
00:11:22,560 --> 00:11:26,430
other words

194
00:11:23,340 --> 00:11:29,990
so whatever language you sell whiskey

195
00:11:26,430 --> 00:11:38,160
says compared with other whiskeys

196
00:11:29,990 --> 00:11:46,110
fish and topic um it's all combat both

197
00:11:38,160 --> 00:11:51,380
Faris that it was so the next from this

198
00:11:46,110 --> 00:11:54,320
is that a wrapper around half cap and

199
00:11:51,380 --> 00:12:01,140
and we do an initial run of hash cap

200
00:11:54,320 --> 00:12:04,740
with the standard dictionary and then it

201
00:12:01,140 --> 00:12:10,110
finds a bunch of words with passwords

202
00:12:04,740 --> 00:12:11,970
you're gonna kill them up using basic

203
00:12:10,110 --> 00:12:13,500
regular expression to remove all the

204
00:12:11,970 --> 00:12:14,700
bits and pieces people and you had to

205
00:12:13,500 --> 00:12:19,020
teach enter that are supposed to make

206
00:12:14,700 --> 00:12:19,590
them London and who puts the words in

207
00:12:19,020 --> 00:12:24,960
slaw

208
00:12:19,590 --> 00:12:29,460
Jensen script it will follow similar

209
00:12:24,960 --> 00:12:34,050
words and then we put those back in

210
00:12:29,460 --> 00:12:37,500
dictionary as we move on have cut with

211
00:12:34,050 --> 00:12:40,280
that you repeat until we had we had

212
00:12:37,500 --> 00:12:43,800
borders not producing any new passwords

213
00:12:40,280 --> 00:12:50,640
and we seed return words if we want to

214
00:12:43,800 --> 00:12:51,599
or we do that and that is what I

215
00:12:50,640 --> 00:12:53,819
accomplished so far

216
00:12:51,600 --> 00:13:09,650
also I'm working on round so any

217
00:12:53,820 --> 00:13:16,809
questions yes policies houses

218
00:13:09,650 --> 00:13:16,809
[Music]

219
00:13:43,290 --> 00:13:52,748
writers fluticasone tactics

220
00:13:47,790 --> 00:13:56,139
yes so in terms of but it definitely

221
00:13:52,749 --> 00:14:00,399
catch but what i'd motion sensor system

222
00:13:56,139 --> 00:14:05,589
has digital parts of the words you can

223
00:14:00,399 --> 00:14:13,269
use the same techniques with pangrams

224
00:14:05,589 --> 00:14:14,920
so you can use words release both so you

225
00:14:13,269 --> 00:14:17,350
can use if you think technique wave

226
00:14:14,920 --> 00:14:24,009
engrams so Robert complete words to you

227
00:14:17,350 --> 00:14:26,920
you're using sections of works and that

228
00:14:24,009 --> 00:14:32,470
has more flexibility as mode it doesn't

229
00:14:26,920 --> 00:14:36,118
require them to have your test was the

230
00:14:32,470 --> 00:14:50,889
simple words already in model it can

231
00:14:36,119 --> 00:14:54,279
find words which so I know that this

232
00:14:50,889 --> 00:14:57,220
miss muffet put if you took up apple

233
00:14:54,279 --> 00:14:59,379
pear orange you in order to it for this

234
00:14:57,220 --> 00:15:02,649
who would have any kind of success you

235
00:14:59,379 --> 00:15:07,749
can just split that up into half a pair

236
00:15:02,649 --> 00:15:11,829
of orange again there as there's

237
00:15:07,749 --> 00:15:15,240
techniques for doing that which I'm

238
00:15:11,829 --> 00:15:17,589
actually playing on one putting in this

239
00:15:15,240 --> 00:15:19,419
but that would be an additional step if

240
00:15:17,589 --> 00:15:20,980
you have to spit spit there to pass it

241
00:15:19,419 --> 00:15:23,410
up in this kind of person up the top

242
00:15:20,980 --> 00:15:25,209
pair orange then you've got a pair

243
00:15:23,410 --> 00:15:27,550
so each individual word in the database

244
00:15:25,209 --> 00:15:29,888
and then the rules that have to put them

245
00:15:27,550 --> 00:15:32,399
back in together again tonight creation

246
00:15:29,889 --> 00:15:32,399
of passwords

247
00:15:34,740 --> 00:15:42,100
that's of answer you question sloth fr

248
00:15:39,430 --> 00:15:43,089
yeah then not one is over this won't

249
00:15:42,100 --> 00:15:45,129
miss won't really help with the

250
00:15:43,089 --> 00:15:48,009
situation you're describing but an

251
00:15:45,129 --> 00:15:51,519
additional step just flipped it to take

252
00:15:48,009 --> 00:15:55,300
it yeah combined password spit it off

253
00:15:51,519 --> 00:16:10,379
before it goes of day it will cope with

254
00:15:55,300 --> 00:16:14,319
that I'm just using Jenson users Milton

255
00:16:10,379 --> 00:16:16,360
I'm not but this been so long since I've

256
00:16:14,319 --> 00:16:18,128
done in your house line was so I'm just

257
00:16:16,360 --> 00:16:31,720
using what is with the lifestyle that

258
00:16:18,129 --> 00:16:33,459
might inspire you use so yeah I did a

259
00:16:31,720 --> 00:16:35,410
long time ago but I might tend to use

260
00:16:33,459 --> 00:16:37,329
natural ash battles on these days and

261
00:16:35,410 --> 00:16:53,290
that's partly from familiarity more than

262
00:16:37,329 --> 00:16:55,920
anything else yes bass mom there's money

263
00:16:53,290 --> 00:16:58,779
there orange tada mm-hm

264
00:16:55,920 --> 00:17:03,539
you've got to know originally that easy

265
00:16:58,779 --> 00:17:07,419
don't mess with those three elements but

266
00:17:03,539 --> 00:17:10,720
yeah it based on that principle was in

267
00:17:07,419 --> 00:17:14,169
using upper and lower case as soon as

268
00:17:10,720 --> 00:17:16,630
you start with djt in unusual soon as

269
00:17:14,169 --> 00:17:19,480
she said in that into that complexity

270
00:17:16,630 --> 00:17:23,760
when yes and again it though she would

271
00:17:19,480 --> 00:17:28,270
process taking consideration that time

272
00:17:23,760 --> 00:17:30,460
so you need to know you need to find a

273
00:17:28,270 --> 00:17:34,270
password in there that has it say apple

274
00:17:30,460 --> 00:17:36,570
pear orange you don't need to split that

275
00:17:34,270 --> 00:17:40,270
up enter

276
00:17:36,570 --> 00:17:45,909
look for similar was so I come over file

277
00:17:40,270 --> 00:17:48,879
and honor of a fruity a tablet for your

278
00:17:45,910 --> 00:17:53,530
rules fact you have cat will then we

279
00:17:48,880 --> 00:17:55,870
combine words surrounding words together

280
00:17:53,530 --> 00:18:02,139
to make of our combinations so we might

281
00:17:55,870 --> 00:18:05,739
come up with banana orange pear that

282
00:18:02,140 --> 00:18:08,380
gets very complex very fast so you have

283
00:18:05,740 --> 00:18:11,680
to eat you know means you have to limit

284
00:18:08,380 --> 00:18:15,340
the amount that key value to hold Weis

285
00:18:11,680 --> 00:18:21,480
you cannot be there forever but and

286
00:18:15,340 --> 00:18:21,480
usually we on a limited time for this so

287
00:18:23,010 --> 00:18:27,760
yes I lost crackin question boutique

288
00:18:25,720 --> 00:18:29,200
yeah just thinking or perish

289
00:18:27,760 --> 00:18:31,120
complexities firstly have you come to

290
00:18:29,200 --> 00:18:32,260
know how many wizards or not raise it is

291
00:18:31,120 --> 00:18:34,840
in it but I've just a little Friday

292
00:18:32,260 --> 00:18:37,810
which is turning but most things about

293
00:18:34,840 --> 00:18:40,270
you give me Scylla call bought that

294
00:18:37,810 --> 00:18:45,879
Plymouth in 21 days brother

295
00:18:40,270 --> 00:18:49,470
patience to the Sun she stopped brief to

296
00:18:45,880 --> 00:18:54,670
use those three words will surely throw

297
00:18:49,470 --> 00:19:01,000
eulogy yeah so yeah you could probably

298
00:18:54,670 --> 00:19:05,800
spit and spit but they you're unlikely

299
00:19:01,000 --> 00:19:08,080
to find in seller come off turf or ready

300
00:19:05,800 --> 00:19:10,780
to pass was virtually the first place an

301
00:19:08,080 --> 00:19:12,699
isthmus will only find pass wasn't a

302
00:19:10,780 --> 00:19:20,350
related in some form to ones which are

303
00:19:12,700 --> 00:19:23,620
already you discovered if you did find

304
00:19:20,350 --> 00:19:28,060
food that's listening scientist it's not

305
00:19:23,620 --> 00:19:31,209
a call but let's say you got onto a

306
00:19:28,060 --> 00:19:33,899
machine and you you you learning Nene

307
00:19:31,210 --> 00:19:38,740
can ski water and you found

308
00:19:33,900 --> 00:19:41,860
silat coral tariff as a cash password on

309
00:19:38,740 --> 00:19:43,420
a machine and you use that as was one of

310
00:19:41,860 --> 00:19:44,998
the initialization a supposed to be in

311
00:19:43,420 --> 00:19:49,600
Europe

312
00:19:44,999 --> 00:19:51,039
but you because you can you sped it up

313
00:19:49,600 --> 00:19:54,009
into a setup :

314
00:19:51,039 --> 00:19:56,649
Kenneth ante 12 fine it will almost

315
00:19:54,009 --> 00:20:01,629
certainly find words related to each of

316
00:19:56,649 --> 00:20:03,879
those words they probably won't come up

317
00:20:01,629 --> 00:20:06,399
with any any matches in any other

318
00:20:03,879 --> 00:20:09,039
passwords because it will find chair

319
00:20:06,399 --> 00:20:11,979
basis another similar to cellar you'll

320
00:20:09,039 --> 00:20:15,100
find Paul Smith as a single words at the

321
00:20:11,980 --> 00:20:17,409
tariff and it will finds anything aside

322
00:20:15,100 --> 00:20:22,059
and public asunder comes up with all the

323
00:20:17,409 --> 00:20:25,480
new but for we think that there's

324
00:20:22,059 --> 00:20:26,440
something similar it's cold you thought

325
00:20:25,480 --> 00:20:28,990
those words together

326
00:20:26,440 --> 00:20:33,299
ngo it's very unlikely to match anyone

327
00:20:28,990 --> 00:20:37,799
else's pass within your organization but

328
00:20:33,299 --> 00:20:41,649
it will find a lot of our face so hip it

329
00:20:37,799 --> 00:20:46,509
it's not there to help you to get every

330
00:20:41,649 --> 00:20:49,629
possible bit what this there to take a

331
00:20:46,509 --> 00:20:53,289
lot of the domain knowledge did you you

332
00:20:49,629 --> 00:20:54,879
need to when you get when you go into

333
00:20:53,289 --> 00:20:57,249
their fur and into the clients and you

334
00:20:54,879 --> 00:20:59,019
yeah you're home to spend time trying to

335
00:20:57,249 --> 00:21:02,249
think of what he chooses task force what

336
00:20:59,019 --> 00:21:04,840
what are they in their terms that

337
00:21:02,249 --> 00:21:06,399
important to their industry and you've

338
00:21:04,840 --> 00:21:08,320
got the limited time and you're trying

339
00:21:06,399 --> 00:21:15,219
to work on other things at the same time

340
00:21:08,320 --> 00:21:17,110
as this and this is just to take out the

341
00:21:15,220 --> 00:21:40,090
requirement for you to actually wasting

342
00:21:17,110 --> 00:21:42,699
time doing that while the nation's other

343
00:21:40,090 --> 00:21:52,769
cities but it will be very very strange

344
00:21:42,700 --> 00:21:57,909
to see how six reduce the incidence of

345
00:21:52,769 --> 00:21:59,430
reductive words so that's time to the

346
00:21:57,909 --> 00:22:05,200
rules you use

347
00:21:59,430 --> 00:22:07,720
for Africa so this just provides the

348
00:22:05,200 --> 00:22:10,810
Prime Minister of was for your

349
00:22:07,720 --> 00:22:12,610
dictionary the rules which he write half

350
00:22:10,810 --> 00:22:15,820
that world government how you combine

351
00:22:12,610 --> 00:22:21,729
nicely up to you I just used the use

352
00:22:15,820 --> 00:22:24,040
again she's a title or one rule but it

353
00:22:21,730 --> 00:22:32,320
is not very part of the natural language

354
00:22:24,040 --> 00:22:33,879
versus very yeah I can't really give any

355
00:22:32,320 --> 00:22:38,159
way to actually tell them that without

356
00:22:33,880 --> 00:22:38,160
outside the rule those rules

357
00:22:38,190 --> 00:22:41,309
[Music]