1
00:00:00,000 --> 00:00:02,790
This is Paola from the AIKR

2
00:00:02,790 --> 00:00:05,860
Community Group, Artificial

3
00:00:05,860 --> 00:00:08,450
Intelligence Knowledge

4
00:00:08,450 --> 00:00:14,860
presentation at WC. And it's

5
00:00:14,860 --> 00:00:18,200
June. Finally, I get around to

6
00:00:18,200 --> 00:00:20,980
show something. I've got

7
00:00:20,980 --> 00:00:24,000
something to show finally.

8
00:00:24,000 --> 00:00:26,570
So, about the slides very

9
00:00:26,570 --> 00:00:28,950
quickly. To provide an

10
00:00:28,950 --> 00:00:31,600
introduction because a few

11
00:00:31,600 --> 00:00:34,460
people are new to this work and

12
00:00:34,460 --> 00:00:37,230
they may still haven't quite

13
00:00:37,230 --> 00:00:40,340
grasped what is being done and

14
00:00:40,340 --> 00:00:41,000
why.

15
00:00:41,000 --> 00:00:43,400
And myself, I also, as this

16
00:00:43,400 --> 00:00:46,300
goalposts were shifting and the

17
00:00:46,300 --> 00:00:48,550
ideas behind this work were

18
00:00:48,550 --> 00:00:50,920
kind of shifting. I haven't

19
00:00:50,920 --> 00:00:53,350
been sure what I was doing for

20
00:00:53,350 --> 00:00:56,170
quite a while. So, it's just to

21
00:00:56,170 --> 00:00:59,000
say this is what we're doing.

22
00:00:59,000 --> 00:01:01,230
And also, there are a few

23
00:01:01,230 --> 00:01:03,450
people who would like to

24
00:01:03,450 --> 00:01:06,440
contribute. I thought we should

25
00:01:06,440 --> 00:01:09,440
make it easier for them. Also,

26
00:01:09,440 --> 00:01:12,140
there have been a couple of

27
00:01:12,140 --> 00:01:14,480
questions that I hope this

28
00:01:14,480 --> 00:01:17,000
slides will have to answer.

29
00:01:17,000 --> 00:01:20,440
Requesting expression of

30
00:01:20,440 --> 00:01:23,680
interest. This is very

31
00:01:23,680 --> 00:01:27,560
important. So, people are

32
00:01:27,560 --> 00:01:32,060
saying, do we have funding? Why

33
00:01:32,060 --> 00:01:35,000
do we have funding? And because

34
00:01:35,000 --> 00:01:35,000
I wasn't able to be very clear.

35
00:01:35,000 --> 00:01:35,000
Now, what can start? It's

36
00:01:35,000 --> 00:01:35,000
opening the proposal.

37
00:01:35,000 --> 00:01:38,180
And I'm hoping there will be

38
00:01:38,180 --> 00:01:41,770
people will be people will come

39
00:01:41,770 --> 00:01:44,830
on board who are capable of

40
00:01:44,830 --> 00:01:48,580
articulating budgeting and all

41
00:01:48,580 --> 00:01:52,310
of the things and handle the

42
00:01:52,310 --> 00:01:55,590
bureaucracy involved. Maybe

43
00:01:55,590 --> 00:01:58,880
institutional partners or, you

44
00:01:58,880 --> 00:02:02,000
know, backers of all sorts.

45
00:02:02,000 --> 00:02:04,550
And also, it's going to be a

46
00:02:04,550 --> 00:02:07,210
way of opening up feedback so

47
00:02:07,210 --> 00:02:09,770
people can see what has been

48
00:02:09,770 --> 00:02:12,250
done with a few slides and

49
00:02:12,250 --> 00:02:14,240
finally be able to give

50
00:02:14,240 --> 00:02:17,110
feedback on what has been done

51
00:02:17,110 --> 00:02:19,000
and contribute.

52
00:02:19,000 --> 00:02:21,480
So, people who have been in the

53
00:02:21,480 --> 00:02:23,950
CG, the community group know

54
00:02:23,950 --> 00:02:26,440
most of what has been going on.

55
00:02:26,440 --> 00:02:29,050
You just search W3C AI KRSG,

56
00:02:29,050 --> 00:02:31,190
you get to the landing page,

57
00:02:31,190 --> 00:02:34,040
you get to the wiki. Mostly, it's

58
00:02:34,040 --> 00:02:36,530
been discussions. But there has

59
00:02:36,530 --> 00:02:37,990
been quite a lot of discussions.

60
00:02:37,990 --> 00:02:41,520
And there's been quite a lot of

61
00:02:41,520 --> 00:02:44,660
updates. So, the main task was

62
00:02:44,660 --> 00:02:47,080
try to understand what is going

63
00:02:47,080 --> 00:02:48,000
on in AI.

64
00:02:48,000 --> 00:02:50,690
Because this has been more easy.

65
00:02:50,690 --> 00:02:52,810
So, around the time when we

66
00:02:52,810 --> 00:02:54,660
started is the time when

67
00:02:54,660 --> 00:02:56,710
symbolic AI and knowledge

68
00:02:56,710 --> 00:02:59,280
presentation were being nailed

69
00:02:59,280 --> 00:03:01,280
in the coffin by the Turing

70
00:03:01,280 --> 00:03:04,190
Prize scholars. And at that

71
00:03:04,190 --> 00:03:06,820
time, myself, a few people

72
00:03:06,820 --> 00:03:09,230
started kind of, you know,

73
00:03:09,230 --> 00:03:11,000
getting nervous.

74
00:03:11,000 --> 00:03:13,300
And they were saying, "where is

75
00:03:13,300 --> 00:03:15,400
this going?" And soon after

76
00:03:15,400 --> 00:03:17,000
that, AI, how do you say,

77
00:03:17,000 --> 00:03:19,470
explainability risks and it all

78
00:03:19,470 --> 00:03:21,080
became mayhem of where, you

79
00:03:21,080 --> 00:03:22,760
know, what is it going to

80
00:03:22,760 --> 00:03:24,800
happen and how we're going to

81
00:03:24,800 --> 00:03:26,520
fix it. And a lot of people

82
00:03:26,520 --> 00:03:29,000
have jumped on the AI bandwagon.

83
00:03:29,000 --> 00:03:31,210
And WAGOR fits. Still not

84
00:03:31,210 --> 00:03:32,920
understanding AI, which is

85
00:03:32,920 --> 00:03:35,260
important. And I don't make any

86
00:03:35,260 --> 00:03:37,340
claims that I understand AI,

87
00:03:37,340 --> 00:03:39,170
but at least from this time

88
00:03:39,170 --> 00:03:41,350
onwards, it is documented on a

89
00:03:41,350 --> 00:03:43,260
mailing list that I've been

90
00:03:43,260 --> 00:03:45,440
paying attention, at least to

91
00:03:45,440 --> 00:03:48,000
whatever I could understand.

92
00:03:48,000 --> 00:03:49,550
Out of the need to understand

93
00:03:49,550 --> 00:03:50,970
machine learning, KR was

94
00:03:50,970 --> 00:03:52,910
involved. So, when we couldn't

95
00:03:52,910 --> 00:03:54,460
figure out what was happening

96
00:03:54,460 --> 00:03:56,390
in machine learning, the black

97
00:03:56,390 --> 00:03:57,940
box, we needed to call upon

98
00:03:57,940 --> 00:03:59,310
knowledge representation to

99
00:03:59,310 --> 00:04:03,190
figure. And at the time, people

100
00:04:03,190 --> 00:04:06,770
still weren't sure outside the

101
00:04:06,770 --> 00:04:10,000
niche of the KR people.

102
00:04:10,000 --> 00:04:13,900
We were able to communicate. So,

103
00:04:13,900 --> 00:04:17,340
we were able to communicate

104
00:04:17,340 --> 00:04:20,140
what was. And out of the need

105
00:04:20,140 --> 00:04:23,910
to understand machine learning,

106
00:04:23,910 --> 00:04:26,730
KR was needed. But I don't

107
00:04:26,730 --> 00:04:30,270
think anybody knew exactly how

108
00:04:30,270 --> 00:04:32,880
it was going to be used to

109
00:04:32,880 --> 00:04:37,000
achieve, to control the AI risk.

110
00:04:37,000 --> 00:04:38,260
We still don't know, but we're

111
00:04:38,260 --> 00:04:39,000
working on it.

112
00:04:39,000 --> 00:04:40,440
This is that progress.

113
00:04:41,180 --> 00:04:42,020
Took us a few years.

114
00:04:42,100 --> 00:04:42,820
So what is KR?

115
00:04:43,280 --> 00:04:44,120
We know that.

116
00:04:44,840 --> 00:04:46,280
You can read the rehab risk.

117
00:04:46,660 --> 00:04:48,420
Excellent education materials.

118
00:04:48,580 --> 00:04:49,800
I don't have to tell you.

119
00:04:50,580 --> 00:04:51,390
You say, okay, I'm going to

120
00:04:51,390 --> 00:04:52,080
school.

121
00:04:52,200 --> 00:04:54,120
I've been studying KR for years.

122
00:04:54,240 --> 00:04:55,220
I've been a researcher.

123
00:04:55,420 --> 00:04:56,920
I'm a professor in KR.

124
00:04:57,200 --> 00:04:58,690
There are a few, not many, a

125
00:04:58,690 --> 00:04:59,640
few people.

126
00:05:00,560 --> 00:05:03,690
But as you can see from the

127
00:05:03,690 --> 00:05:07,580
educational resources,

128
00:05:08,100 --> 00:05:09,820
which we are creating on the wiki,

129
00:05:09,820 --> 00:05:12,370
there are a lot of courses in

130
00:05:12,370 --> 00:05:13,980
knowledge of recitation.

131
00:05:14,180 --> 00:05:15,160
Some of them are excellent and

132
00:05:15,160 --> 00:05:16,160
they're accessible.

133
00:05:16,820 --> 00:05:17,880
And there are books.

134
00:05:19,180 --> 00:05:20,100
So one can say, okay, I'm going

135
00:05:20,100 --> 00:05:21,180
to study KR.

136
00:05:21,180 --> 00:05:24,170
But that doesn't mean that

137
00:05:24,170 --> 00:05:26,980
people who know KR, the

138
00:05:26,980 --> 00:05:30,820
professor or even the experts,

139
00:05:32,140 --> 00:05:33,820
know how to use it to address

140
00:05:33,820 --> 00:05:36,880
the current AI concerns.

141
00:05:37,040 --> 00:05:37,940
Or can't, you know.

142
00:05:38,040 --> 00:05:40,740
So nobody can do anything in

143
00:05:40,740 --> 00:05:43,740
absolute terms because AI is

144
00:05:43,740 --> 00:05:47,980
evolving, how do you say, organically.

145
00:05:48,200 --> 00:05:49,480
Nobody has control over, you

146
00:05:49,480 --> 00:05:51,110
know, these algorithms are out

147
00:05:51,110 --> 00:05:52,830
there, the computational power

148
00:05:52,830 --> 00:05:53,500
is there.

149
00:05:53,920 --> 00:05:55,270
And a lot of wonderful things

150
00:05:55,270 --> 00:05:56,720
are happening, a lot of useful

151
00:05:56,720 --> 00:05:57,640
applications.

152
00:05:59,800 --> 00:06:01,660
So we know that we need KR for

153
00:06:01,660 --> 00:06:06,130
accessibility and safety in

154
00:06:06,130 --> 00:06:07,640
general.

155
00:06:07,640 --> 00:06:09,640
So we're bridging the machine

156
00:06:09,640 --> 00:06:10,520
learning.

157
00:06:10,520 --> 00:06:12,440
So now that you see, KR

158
00:06:12,440 --> 00:06:15,020
typically has been used to make

159
00:06:15,020 --> 00:06:18,160
the machine to program machines.

160
00:06:18,160 --> 00:06:19,660
So we wanted an intelligent

161
00:06:19,660 --> 00:06:20,400
function.

162
00:06:20,400 --> 00:06:23,000
We need KR to tell the computer

163
00:06:23,000 --> 00:06:24,860
to execute it.

164
00:06:26,320 --> 00:06:27,940
Now the opposite is happening.

165
00:06:27,940 --> 00:06:30,200
We tell the computer is

166
00:06:30,200 --> 00:06:32,280
executing an intelligent

167
00:06:32,280 --> 00:06:33,480
function.

168
00:06:33,480 --> 00:06:35,650
And we need KR to figure out

169
00:06:35,650 --> 00:06:37,260
how it's doing it.

170
00:06:37,260 --> 00:06:39,330
So I think this is very

171
00:06:39,330 --> 00:06:41,820
exciting to see for me.

172
00:06:41,820 --> 00:06:44,000
And this is what makes it worth

173
00:06:44,000 --> 00:06:45,200
it, really.

174
00:06:45,200 --> 00:06:48,440
So in order to communicate what

175
00:06:48,440 --> 00:06:51,580
KR is, is such a huge domain,

176
00:06:51,580 --> 00:06:54,640
we need this vertical because

177
00:06:54,640 --> 00:06:57,200
there is so much of it.

178
00:06:57,200 --> 00:06:58,980
And we need to also explain the

179
00:06:58,980 --> 00:07:00,830
role of KR in explainability,

180
00:07:00,830 --> 00:07:02,880
which becomes a bit circular.

181
00:07:02,880 --> 00:07:06,630
We have to open the black box.

182
00:07:06,630 --> 00:07:10,100
And we have to use KR to create

183
00:07:10,100 --> 00:07:15,170
reliable, safe, ethical, unbiased,

184
00:07:15,170 --> 00:07:17,400
and all of that.

185
00:07:17,400 --> 00:07:19,730
So now this is the other very

186
00:07:19,730 --> 00:07:22,650
exciting bit, using KR in the

187
00:07:22,650 --> 00:07:25,200
creation of agents. Non-trivial

188
00:07:25,200 --> 00:07:27,190
because the KR is a huge

189
00:07:27,190 --> 00:07:29,650
subject. So you can point

190
00:07:29,650 --> 00:07:31,700
people to KR, educational

191
00:07:31,700 --> 00:07:34,370
resources, and there are books

192
00:07:34,370 --> 00:07:37,140
and exercises and lectures they

193
00:07:37,140 --> 00:07:38,500
can follow.

194
00:07:38,500 --> 00:07:41,750
But will they, is anyone that,

195
00:07:41,750 --> 00:07:45,310
I don't know anybody who can

196
00:07:45,310 --> 00:07:47,540
guarantee a safe, ethical AI

197
00:07:47,540 --> 00:07:50,320
system, whether they have

198
00:07:50,320 --> 00:07:52,500
studied KR or not.

199
00:07:52,500 --> 00:07:55,530
So, you know, but still we need

200
00:07:55,530 --> 00:07:58,150
to be able to, to summarize

201
00:07:58,150 --> 00:08:00,990
what KR is or which concepts

202
00:08:00,990 --> 00:08:03,480
and tools and methods in KR can

203
00:08:03,480 --> 00:08:06,220
be leveraged to guarantee AI

204
00:08:06,220 --> 00:08:07,500
safety.

205
00:08:07,500 --> 00:08:09,340
Because even experts in KR may

206
00:08:09,340 --> 00:08:11,160
not be familiar with the entire

207
00:08:11,160 --> 00:08:12,940
domain or with the state of the

208
00:08:12,940 --> 00:08:13,500
art.

209
00:08:13,500 --> 00:08:15,390
So typical expert people I've

210
00:08:15,390 --> 00:08:17,530
spoken with, they, they were

211
00:08:17,530 --> 00:08:18,980
experts 50 years ago when

212
00:08:18,980 --> 00:08:20,500
things have changed.

213
00:08:20,500 --> 00:08:23,150
AI has changed, computational,

214
00:08:23,150 --> 00:08:25,500
computer science has changed.

215
00:08:25,500 --> 00:08:27,470
Because using KR to make the

216
00:08:27,470 --> 00:08:29,370
black box transparent may

217
00:08:29,370 --> 00:08:31,610
require advancing the state of

218
00:08:31,610 --> 00:08:33,500
the art in many fields.

219
00:08:33,500 --> 00:08:35,630
KR but also logic. And this is

220
00:08:35,630 --> 00:08:37,890
another exciting area. So, at

221
00:08:37,890 --> 00:08:39,690
least we, so we have to start

222
00:08:39,690 --> 00:08:41,820
by mapping the knowledge domain

223
00:08:41,820 --> 00:08:42,500
for KR.

224
00:08:42,500 --> 00:08:45,240
So ultimately we're doing an

225
00:08:45,240 --> 00:08:47,520
ontology for knowledge

226
00:08:47,520 --> 00:08:49,500
representation.

227
00:08:49,500 --> 00:08:51,970
The scenario, important thing

228
00:08:51,970 --> 00:08:53,910
to keep in mind is that AI,

229
00:08:53,910 --> 00:08:56,300
whatever is going to be, which

230
00:08:56,300 --> 00:08:58,230
there is a lot here, which

231
00:08:58,230 --> 00:09:00,390
there is so much here I cannot

232
00:09:00,390 --> 00:09:01,500
describe.

233
00:09:01,500 --> 00:09:04,990
So, AI is actually driving and

234
00:09:04,990 --> 00:09:07,150
influencing human evolution,

235
00:09:07,150 --> 00:09:10,620
cognition, information, belief,

236
00:09:10,620 --> 00:09:13,210
decision, the politics, the

237
00:09:13,210 --> 00:09:14,500
economics.

238
00:09:14,500 --> 00:09:17,910
So AI is not just by itself. It's

239
00:09:17,910 --> 00:09:20,750
something which is very

240
00:09:20,750 --> 00:09:22,570
fundamental in human

241
00:09:22,570 --> 00:09:26,370
contemporary evolution and in

242
00:09:26,370 --> 00:09:28,500
social technologies.

243
00:09:28,500 --> 00:09:31,750
So, KR can help to mitigate

244
00:09:31,750 --> 00:09:35,510
existential risks. To be honest

245
00:09:35,510 --> 00:09:37,680
here, I have a whole talk on

246
00:09:37,680 --> 00:09:40,500
this. Systemic aberration.

247
00:09:40,500 --> 00:09:43,400
So, what is happening now is

248
00:09:43,400 --> 00:09:46,480
that things are being done

249
00:09:46,480 --> 00:09:49,110
which are not being described

250
00:09:49,110 --> 00:09:50,500
properly.

251
00:09:50,500 --> 00:09:52,490
And things that are said that

252
00:09:52,490 --> 00:09:54,820
are not being thought. There is

253
00:09:54,820 --> 00:09:57,500
a lot of confusion by mislabeling.

254
00:09:57,500 --> 00:10:01,040
And we can only use knowledge

255
00:10:01,040 --> 00:10:03,930
representation to help us disambiguate

256
00:10:03,930 --> 00:10:06,110
and get a grip with what is

257
00:10:06,110 --> 00:10:07,500
going on.

258
00:10:07,500 --> 00:10:09,410
So, if you want, I can share

259
00:10:09,410 --> 00:10:11,210
the link if you cannot click it.

260
00:10:11,210 --> 00:10:13,070
If you're interested, that's

261
00:10:13,070 --> 00:10:14,500
about an hour tall.

262
00:10:14,500 --> 00:10:16,980
If the standards being

263
00:10:16,980 --> 00:10:20,470
developed to ensure AI safety,

264
00:10:20,470 --> 00:10:23,830
and I have sifted through a few

265
00:10:23,830 --> 00:10:26,830
AI standards on the hub, are

266
00:10:26,830 --> 00:10:29,580
missing out on this essential

267
00:10:29,580 --> 00:10:31,900
AI safety concept, then the

268
00:10:31,900 --> 00:10:34,780
risk of critical failures in AI

269
00:10:34,780 --> 00:10:36,500
safety increase.

270
00:10:36,500 --> 00:10:39,370
So, I hope this is fairly

271
00:10:39,370 --> 00:10:41,800
logical and clear, despite a

272
00:10:41,800 --> 00:10:44,880
lot of effort. So, there is a

273
00:10:44,880 --> 00:10:47,410
lot of money and noise in AI

274
00:10:47,410 --> 00:10:48,500
safety.

275
00:10:48,500 --> 00:10:51,940
Nonetheless, the safety

276
00:10:51,940 --> 00:10:54,960
standards are missing out on

277
00:10:54,960 --> 00:10:58,930
essential safety concepts. And

278
00:10:58,930 --> 00:11:02,710
therefore, I think we need to

279
00:11:02,710 --> 00:11:05,840
use leverage KR to get out of

280
00:11:05,840 --> 00:11:07,500
this loop.

281
00:11:07,500 --> 00:11:10,620
So, I'm mapping the KR domain,

282
00:11:10,620 --> 00:11:13,640
simply because when people ask

283
00:11:13,640 --> 00:11:16,400
me what is KR, I want to point

284
00:11:16,400 --> 00:11:19,650
to something that is cognitively

285
00:11:19,650 --> 00:11:22,510
can be processed by a human and

286
00:11:22,510 --> 00:11:25,500
by the machine obviously.

287
00:11:25,500 --> 00:11:28,580
But, at the moment, KR is all

288
00:11:28,580 --> 00:11:31,630
over the place. So, I have

289
00:11:31,630 --> 00:11:33,500
divided into sub domain.

290
00:11:33,500 --> 00:11:36,060
So, I'm saying the knowledge

291
00:11:36,060 --> 00:11:38,690
presentation domain is made up

292
00:11:38,690 --> 00:11:41,830
of upper foundational ontologies,

293
00:11:41,830 --> 00:11:43,980
knowledge presentation

294
00:11:43,980 --> 00:11:46,420
languages and formalisms,

295
00:11:46,420 --> 00:11:49,130
domain ontologies. Oops,

296
00:11:49,130 --> 00:11:51,500
spelling mistake here.

297
00:11:51,500 --> 00:11:54,890
So, I can never remember what

298
00:11:54,890 --> 00:11:59,510
ODDS stands for. Domain? Ontology

299
00:11:59,510 --> 00:12:03,010
domain? Basically, these are

300
00:12:03,010 --> 00:12:07,020
domain ontologies. And there is

301
00:12:07,020 --> 00:12:10,410
a whole field and I'm looking

302
00:12:10,410 --> 00:12:12,960
at it. When I can remember what

303
00:12:12,960 --> 00:12:15,650
it is. AI safety standards. So,

304
00:12:15,650 --> 00:12:17,210
here is where I found the

305
00:12:17,210 --> 00:12:19,670
critical issues. Some critical

306
00:12:19,670 --> 00:12:20,500
issues.

307
00:12:20,500 --> 00:12:23,960
which are my worry, and I think

308
00:12:23,960 --> 00:12:27,500
they should be your worry too.

309
00:12:27,500 --> 00:12:32,560
So we have a lot happening in

310
00:12:32,560 --> 00:12:33,800
the knowledge presentation

311
00:12:33,800 --> 00:12:35,550
learning and ontology-driven

312
00:12:35,550 --> 00:12:37,500
agents, very exciting stuff.

313
00:12:37,500 --> 00:12:40,930
We cannot conceptualize all

314
00:12:40,930 --> 00:12:44,590
this massive field without a

315
00:12:44,590 --> 00:12:47,960
bit of sweat. So this is what

316
00:12:47,960 --> 00:12:50,600
we're doing now, and I

317
00:12:50,600 --> 00:12:53,770
appreciate very much your

318
00:12:53,770 --> 00:12:57,790
attention and the concentration

319
00:12:57,790 --> 00:13:01,500
you're applying to follow.

320
00:13:01,500 --> 00:13:03,470
Now, this is the point where

321
00:13:03,470 --> 00:13:05,740
there was a bit of a question.

322
00:13:05,740 --> 00:13:08,200
Peter Rivett said, "Oh, you're

323
00:13:08,200 --> 00:13:09,930
saying that you're going to do

324
00:13:09,930 --> 00:13:11,820
something that can make the

325
00:13:11,820 --> 00:13:13,500
system reliable."

326
00:13:13,500 --> 00:13:16,560
He wasn't quite sure this is

327
00:13:16,560 --> 00:13:20,210
something to be taken seriously

328
00:13:20,210 --> 00:13:24,390
or feasible. So I thought, okay,

329
00:13:24,390 --> 00:13:27,530
we cannot have any AI ethical

330
00:13:27,530 --> 00:13:30,500
safety standards unless...

331
00:13:30,500 --> 00:13:33,120
Unless we understand the notion

332
00:13:33,120 --> 00:13:35,930
of reliability engineering. So

333
00:13:35,930 --> 00:13:38,820
I'm intersecting these two here

334
00:13:38,820 --> 00:13:40,570
and I'm happy to look at it

335
00:13:40,570 --> 00:13:41,500
later.

336
00:13:41,500 --> 00:13:44,560
So today we are starting from

337
00:13:44,560 --> 00:13:50,160
this. We are modeling. So these

338
00:13:50,160 --> 00:13:52,720
are subdomains. Each of these

339
00:13:52,720 --> 00:13:54,570
is the subdomain for the

340
00:13:54,570 --> 00:13:56,940
knowledge representation in

341
00:13:56,940 --> 00:13:59,500
that domain, which is huge.

342
00:13:59,500 --> 00:14:02,250
And today we are just focusing

343
00:14:02,250 --> 00:14:05,180
on this state. Because I think

344
00:14:05,180 --> 00:14:07,640
we should take this step by

345
00:14:07,640 --> 00:14:10,290
step. So first question, are

346
00:14:10,290 --> 00:14:13,250
the subdomains adequate? And

347
00:14:13,250 --> 00:14:14,980
this is a nice little

348
00:14:14,980 --> 00:14:18,260
evaluation task for anyone who

349
00:14:18,260 --> 00:14:20,500
has a spare neuron.

350
00:14:20,500 --> 00:14:24,240
So, okay. So, upper ontology,

351
00:14:24,240 --> 00:14:27,620
knowledge representation

352
00:14:27,620 --> 00:14:32,010
languages, domain ontology,

353
00:14:32,010 --> 00:14:34,360
knowledge representation

354
00:14:34,360 --> 00:14:37,650
learning, and all of these. So

355
00:14:37,650 --> 00:14:39,960
is there anything in this subdomains

356
00:14:39,960 --> 00:14:41,940
that shouldn't be there, that

357
00:14:41,940 --> 00:14:43,760
doesn't belong? Or is there

358
00:14:43,760 --> 00:14:46,140
anything fundamentally missing?

359
00:14:46,140 --> 00:14:47,500
Please let me know.

360
00:14:47,500 --> 00:14:50,100
And so what are we doing? What

361
00:14:50,100 --> 00:14:52,550
I have done so far is that I've

362
00:14:52,550 --> 00:14:54,900
taken all the upper top level

363
00:14:54,900 --> 00:14:57,500
ontologies that I could put my

364
00:14:57,500 --> 00:14:58,500
hands on.

365
00:14:58,500 --> 00:15:01,810
I have identified the sources

366
00:15:01,810 --> 00:15:04,430
of knowledge, which were

367
00:15:04,430 --> 00:15:07,070
textbooks and papers and

368
00:15:07,070 --> 00:15:10,200
repositories and abstracted

369
00:15:10,200 --> 00:15:13,740
terms. In particular, I have

370
00:15:13,740 --> 00:15:17,580
extracted the classes, I think.

371
00:15:17,580 --> 00:15:19,690
And so now it's a matter of

372
00:15:19,690 --> 00:15:22,720
evaluating. So then we're going

373
00:15:22,720 --> 00:15:25,120
to develop a definition for

374
00:15:25,120 --> 00:15:27,500
each class or category.

375
00:15:27,500 --> 00:15:30,630
or streamline the duplicate

376
00:15:30,630 --> 00:15:33,550
clean up boil it down a little

377
00:15:33,550 --> 00:15:36,820
bit make a list and effect well

378
00:15:36,820 --> 00:15:37,120
so when I

379
00:15:37,120 --> 00:15:38,680
say list it looks like I'm

380
00:15:38,680 --> 00:15:40,440
doing nothing you know but I'm

381
00:15:40,440 --> 00:15:41,930
looking I'm making a bag of

382
00:15:41,930 --> 00:15:43,060
words which is

383
00:15:43,060 --> 00:15:45,270
a fundamental thing and I'll

384
00:15:45,270 --> 00:15:47,410
pay mm-hmm so I have you can

385
00:15:47,410 --> 00:15:49,410
click here if you cannot I'm

386
00:15:49,410 --> 00:15:50,380
gonna send a

387
00:15:50,380 --> 00:15:53,000
link separately a parentology

388
00:15:53,000 --> 00:15:55,580
book up typically I would put

389
00:15:55,580 --> 00:15:57,680
this in an open spreadsheet

390
00:15:57,680 --> 00:15:58,420
open for

391
00:15:58,420 --> 00:16:02,390
anyone to open but because

392
00:16:02,390 --> 00:16:07,160
there have been people who have

393
00:16:07,160 --> 00:16:11,560
kind of taken advantage of the

394
00:16:11,560 --> 00:16:13,440
openness we

395
00:16:13,440 --> 00:16:16,300
don't know who they are we have

396
00:16:16,300 --> 00:16:19,080
there is no way to track whose

397
00:16:19,080 --> 00:16:21,950
access to what resource maybe

398
00:16:21,950 --> 00:16:22,800
using it

399
00:16:22,800 --> 00:16:24,730
without credit to see doesn't

400
00:16:24,730 --> 00:16:26,800
matter because ultimately these

401
00:16:26,800 --> 00:16:28,660
are going to be published and

402
00:16:28,660 --> 00:16:31,550
licensed for everyone to use

403
00:16:31,550 --> 00:16:33,530
they're gonna be on an open

404
00:16:33,530 --> 00:16:36,840
access but in the meantime some

405
00:16:36,840 --> 00:16:37,560
people could

406
00:16:37,560 --> 00:16:40,150
just use this work to do their

407
00:16:40,150 --> 00:16:42,850
own thing and not acknowledge

408
00:16:42,850 --> 00:16:44,480
no so this this is

409
00:16:44,480 --> 00:16:45,900
unfortunately

410
00:16:45,900 --> 00:16:47,910
happened so here is the upper

411
00:16:47,910 --> 00:16:49,710
ontology vocab if you would

412
00:16:49,710 --> 00:16:51,600
like to look at it or provide

413
00:16:51,600 --> 00:16:52,800
feedback in

414
00:16:52,800 --> 00:16:55,580
any way please request and if

415
00:16:55,580 --> 00:16:57,810
you cannot see it please

416
00:16:57,810 --> 00:17:00,990
request access I will however

417
00:17:00,990 --> 00:17:02,100
already managed

418
00:17:02,100 --> 00:17:05,250
to I will send the vocab to the

419
00:17:05,250 --> 00:17:08,350
people who are already working

420
00:17:08,350 --> 00:17:11,010
on this one so two or three

421
00:17:11,010 --> 00:17:12,340
people have

422
00:17:12,340 --> 00:17:14,830
already provided feedback on a

423
00:17:14,830 --> 00:17:17,320
drums and these will receive it

424
00:17:17,320 --> 00:17:19,420
probably in an email with a

425
00:17:19,420 --> 00:17:20,460
request not

426
00:17:20,460 --> 00:17:22,400
to share it outside at the

427
00:17:22,400 --> 00:17:24,890
community group at this stage.

428
00:17:24,890 --> 00:17:27,160
Please check the current list

429
00:17:27,160 --> 00:17:27,900
of terms.

430
00:17:30,380 --> 00:17:30,560
Annotate it. So in the Excel

431
00:17:30,560 --> 00:17:33,240
sheet you may be able to create

432
00:17:33,240 --> 00:17:35,420
a column with your name

433
00:17:37,340 --> 00:17:38,210
and then in the column you

434
00:17:38,210 --> 00:17:41,680
could write something for each

435
00:17:41,680 --> 00:17:45,100
term, something like add,

436
00:17:46,860 --> 00:17:49,330
you know, this doesn't,

437
00:17:49,330 --> 00:17:52,980
whatever. So feel free to annotate

438
00:17:52,980 --> 00:17:55,500
each term or to create an

439
00:17:55,500 --> 00:17:58,140
annotation that

440
00:17:58,140 --> 00:17:59,940
applies to the entire

441
00:17:59,940 --> 00:18:02,740
vocabulary or even just email.

442
00:18:02,740 --> 00:18:05,440
So are there any terms in cost?

443
00:18:05,440 --> 00:18:07,500
About 200 terms now.

444
00:18:08,460 --> 00:18:09,890
Anything to be added? Anything

445
00:18:09,890 --> 00:18:12,340
that should be deleted? How can

446
00:18:12,340 --> 00:18:16,220
these upper ontology

447
00:18:17,020 --> 00:18:19,850
vocabulary, upper ontology be

448
00:18:19,850 --> 00:18:24,040
used? Is it useful? Complete?

449
00:18:24,040 --> 00:18:27,590
How can it be used? Make some

450
00:18:27,590 --> 00:18:28,940
use cases?

451
00:18:30,140 --> 00:18:31,680
Then eventually this is going

452
00:18:31,680 --> 00:18:34,440
to be published and it's going

453
00:18:34,440 --> 00:18:37,090
to be constituting the first

454
00:18:37,090 --> 00:18:37,980
step into

455
00:18:37,980 --> 00:18:40,840
the publication of a series of

456
00:18:40,840 --> 00:18:44,500
vocabularies which overall will

457
00:18:44,500 --> 00:18:46,440
represent the knowledge

458
00:18:46,440 --> 00:18:48,140
representation

459
00:18:49,260 --> 00:18:51,730
domain. And probably the basis

460
00:18:51,730 --> 00:18:54,040
for an ontology of knowledge

461
00:18:54,040 --> 00:18:56,650
representation. So if anyone

462
00:18:56,650 --> 00:18:57,580
interested

463
00:18:57,580 --> 00:18:59,830
in working on the list, please

464
00:18:59,830 --> 00:19:02,160
get in touch. Thank you so much

465
00:19:02,160 --> 00:19:04,590
for listening. And I'll see you

466
00:19:04,590 --> 00:19:05,820
in the group.

467
00:19:08,220 --> 00:19:08,500
Wow, this is being disclosed

468
00:19:08,500 --> 00:19:11,880
already. Look at the machine.

469
00:19:11,880 --> 00:19:14,320
This is the machine. It's doing

470
00:19:14,320 --> 00:19:14,620
it.

471
00:19:17,420 --> 00:19:19,900
The machine is opening. Bye.