1
00:00:00,000 --> 00:00:03,390
0, 0, 0.

2
00:00:03,390 --> 00:00:06,219
So you guys can see

3
00:00:06,440 --> 00:00:08,999
that we see just what I

4
00:00:08,999 --> 00:00:11,205
think what we see
is fine here.

5
00:00:11,205 --> 00:00:12,990
Yeah, good.

6
00:00:12,990 --> 00:00:14,640
And I will go
ahead and say

7
00:00:14,640 --> 00:00:16,214
welcome everyone
to read it.

8
00:00:16,214 --> 00:00:17,085
I pretty much
know everybody

9
00:00:17,085 --> 00:00:18,555
in the room right now.

10
00:00:18,555 --> 00:00:22,109
To the final session of

11
00:00:22,109 --> 00:00:23,385
the winter term

12
00:00:23,385 --> 00:00:25,439
System Science
Seminar Series.

13
00:00:25,439 --> 00:00:27,105
And we're pleased to
have Shane Dixon,

14
00:00:27,105 --> 00:00:29,084
one of our PhD students,

15
00:00:29,084 --> 00:00:30,719
present something released

16
00:00:30,719 --> 00:00:32,039
related to his research.

17
00:00:32,039 --> 00:00:34,305
If not spot on
his research.

18
00:00:34,305 --> 00:00:36,270
Go ahead and give
it give yourself

19
00:00:36,270 --> 00:00:37,770
a better introduction
on your background,

20
00:00:37,770 --> 00:00:39,535
please, if you
know my chain.

21
00:00:39,535 --> 00:00:40,445
Okay.

22
00:00:40,445 --> 00:00:42,350
Yeah, so thank you
for having me.

23
00:00:42,350 --> 00:00:44,840
I'm really excited to
be here and do this.

24
00:00:44,840 --> 00:00:46,430
That's something
I've been kind of

25
00:00:46,430 --> 00:00:47,960
thinking about
and preparing for

26
00:00:47,960 --> 00:00:49,699
since I sat down at

27
00:00:49,699 --> 00:00:53,495
my own first
seminar in 2017,

28
00:00:53,495 --> 00:00:56,585
just a fresh faced
PhD students.

29
00:00:56,585 --> 00:00:59,900
So here I am on the
other side and I,

30
00:00:59,900 --> 00:01:01,189
even though I'm not in

31
00:01:01,189 --> 00:01:02,675
harder house right now,

32
00:01:02,675 --> 00:01:04,580
I'm in harder
house and spirit,

33
00:01:04,580 --> 00:01:07,339
so excited to
get back there.

34
00:01:07,339 --> 00:01:11,629
So my background is

35
00:01:11,629 --> 00:01:16,955
that I attended undergrad
at Michigan State,

36
00:01:16,955 --> 00:01:19,279
interested in
public policy

37
00:01:19,279 --> 00:01:21,604
and political economy.

38
00:01:21,604 --> 00:01:26,390
I had some sort of

39
00:01:26,390 --> 00:01:29,044
confusion or
discomfort with

40
00:01:29,044 --> 00:01:31,430
the traditional
development story

41
00:01:31,430 --> 00:01:32,480
told to me there.

42
00:01:32,480 --> 00:01:34,190
And that set me on

43
00:01:34,190 --> 00:01:37,039
a long and winding path
that brought me into

44
00:01:37,039 --> 00:01:40,040
this study of
complex systems

45
00:01:40,040 --> 00:01:42,634
and systems science
here at Portland State.

46
00:01:42,634 --> 00:01:45,739
And now I'm getting into

47
00:01:45,739 --> 00:01:49,444
this open-ended
evolution question

48
00:01:49,444 --> 00:01:52,639
in the field of
artificial light.

49
00:01:52,639 --> 00:01:54,875
So that's what I'm going
to talk about today.

50
00:01:54,875 --> 00:01:57,484
Not my own work just yet,

51
00:01:57,484 --> 00:01:59,750
which is still a bit
of a rough form,

52
00:01:59,750 --> 00:02:03,570
but we're moving,
moving ahead.

53
00:02:03,580 --> 00:02:05,959
But just some of
the background.

54
00:02:05,959 --> 00:02:08,824
I'm open and
evolution, what it is,

55
00:02:08,824 --> 00:02:10,610
why it's important, how

56
00:02:10,610 --> 00:02:12,950
it fits into
artificial life.

57
00:02:12,950 --> 00:02:17,449
Okay, so to
preface this talk,

58
00:02:17,449 --> 00:02:19,039
I just want to give
everybody a chance

59
00:02:19,039 --> 00:02:22,460
to sort of meditate
on this con,

60
00:02:22,460 --> 00:02:24,440
concept of creativity and

61
00:02:24,440 --> 00:02:26,689
diversity in
the real-world.

62
00:02:26,689 --> 00:02:30,199
So we know that there are

63
00:02:30,199 --> 00:02:32,419
a number of these
complex systems

64
00:02:32,419 --> 00:02:34,219
that we look at here.

65
00:02:34,219 --> 00:02:36,470
It's System
Science, talking

66
00:02:36,470 --> 00:02:38,030
about the biosphere,

67
00:02:38,030 --> 00:02:39,650
the world economy,

68
00:02:39,650 --> 00:02:43,084
technological and social
political systems.

69
00:02:43,084 --> 00:02:44,569
And in all of these,

70
00:02:44,569 --> 00:02:47,105
there is just this
immense variety

71
00:02:47,105 --> 00:02:49,115
of different forms.

72
00:02:49,115 --> 00:02:52,069
And so then we've known

73
00:02:52,069 --> 00:02:56,600
that Darwinian
evolution is

74
00:02:56,600 --> 00:02:58,444
the source, or at least,

75
00:02:58,444 --> 00:03:00,485
you know, governing this,

76
00:03:00,485 --> 00:03:02,914
this production
of diversity

77
00:03:02,914 --> 00:03:05,615
through this or a
Darwinian mechanisms,

78
00:03:05,615 --> 00:03:09,514
variation, heredity
and natural selection.

79
00:03:09,514 --> 00:03:11,840
But what we think of

80
00:03:11,840 --> 00:03:15,109
as a simple process

81
00:03:15,109 --> 00:03:16,909
that we understand
very well,

82
00:03:16,909 --> 00:03:19,520
we're, we're not
able to recreate

83
00:03:19,520 --> 00:03:23,000
that same creative
potential

84
00:03:23,000 --> 00:03:25,040
when we implement
this process, right?

85
00:03:25,040 --> 00:03:27,664
So something has
to be missing.

86
00:03:27,664 --> 00:03:30,665
And so the question
is, what is a,

87
00:03:30,665 --> 00:03:33,349
what is the
missing thing that

88
00:03:33,349 --> 00:03:36,980
our simple evolutionary
algorithms aren't

89
00:03:36,980 --> 00:03:38,960
able to capture that

90
00:03:38,960 --> 00:03:43,804
the real-world is,
is demonstrating.

91
00:03:43,804 --> 00:03:48,124
Okay, so, whoops, the
outline of this talk,

92
00:03:48,124 --> 00:03:49,459
I'm going to briefly say,

93
00:03:49,459 --> 00:03:52,324
what is artificial
life as a discipline?

94
00:03:52,324 --> 00:03:54,260
I'm going to talk
a little bit about

95
00:03:54,260 --> 00:03:57,920
artificial evolution
and then get

96
00:03:57,920 --> 00:04:00,500
into actually saying
what is open-end ended

97
00:04:00,500 --> 00:04:01,820
evolution we're
going to talk

98
00:04:01,820 --> 00:04:04,610
about to define
and measure it.

99
00:04:04,610 --> 00:04:07,670
Some hypothesize
requirements.

100
00:04:07,670 --> 00:04:09,440
We're going to walk

101
00:04:09,440 --> 00:04:12,035
through some of
the key models.

102
00:04:12,035 --> 00:04:14,135
And then we're going
to wrap up by just

103
00:04:14,135 --> 00:04:17,285
talking about what's next.

104
00:04:17,285 --> 00:04:20,539
Okay? Just before we go

105
00:04:20,539 --> 00:04:22,819
any further questions or

106
00:04:22,819 --> 00:04:23,930
any kind of
questions going

107
00:04:23,930 --> 00:04:25,519
along like clarifying
questions?

108
00:04:25,519 --> 00:04:28,070
Or would you prefer
for us to kind of hold

109
00:04:28,070 --> 00:04:29,900
back and wait until

110
00:04:29,900 --> 00:04:31,520
you get to a
certain point.

111
00:04:31,520 --> 00:04:33,979
People can ask clarifying
questions as we go.

112
00:04:33,979 --> 00:04:37,399
I like, I like AS
interactive kinda deal.

113
00:04:37,399 --> 00:04:39,319
But I'll try to
move through this.

114
00:04:39,319 --> 00:04:40,670
I'm not a 100
percent sure how

115
00:04:40,670 --> 00:04:42,230
long this has
all been takes.

116
00:04:42,230 --> 00:04:43,129
So we may have to speed

117
00:04:43,129 --> 00:04:44,209
through the models if it,

118
00:04:44,209 --> 00:04:47,300
if it gets IT that way,

119
00:04:47,300 --> 00:04:48,424
but we'll see as we go.

120
00:04:48,424 --> 00:04:51,094
So okay, artificial life

121
00:04:51,094 --> 00:04:54,274
term itself is
coined by a guy,

122
00:04:54,274 --> 00:04:58,970
Chris Langton in 1986.

123
00:04:58,970 --> 00:05:01,879
But in 1988, he introduces

124
00:05:01,879 --> 00:05:05,345
the first our volume,

125
00:05:05,345 --> 00:05:07,955
artificial life
proceedings of a workshop

126
00:05:07,955 --> 00:05:10,849
on the synthesis and

127
00:05:10,849 --> 00:05:12,934
simulation of
living systems.

128
00:05:12,934 --> 00:05:15,409
He says, artificial
life is a study of

129
00:05:15,409 --> 00:05:17,179
man-made systems
that exhibit

130
00:05:17,179 --> 00:05:18,500
behaviors characteristic

131
00:05:18,500 --> 00:05:19,955
of natural
lighting systems.

132
00:05:19,955 --> 00:05:21,380
It compliments traditional

133
00:05:21,380 --> 00:05:23,179
biological sciences
concerned with

134
00:05:23,179 --> 00:05:24,979
the analysis of
living organisms by

135
00:05:24,979 --> 00:05:26,990
attempting to
synthesized life.

136
00:05:26,990 --> 00:05:28,744
Behaviors within computers

137
00:05:28,744 --> 00:05:30,335
and other
artificial media.

138
00:05:30,335 --> 00:05:31,759
By extending the empirical

139
00:05:31,759 --> 00:05:32,959
foundation upon which

140
00:05:32,959 --> 00:05:36,065
biology is base beyond
the carbon chain,

141
00:05:36,065 --> 00:05:37,519
life that has
evolved on Earth,

142
00:05:37,519 --> 00:05:39,439
artificial life
can contribute to

143
00:05:39,439 --> 00:05:40,760
theoretical biology

144
00:05:40,760 --> 00:05:42,755
by locating life
as we know it,

145
00:05:42,755 --> 00:05:44,089
within the larger picture

146
00:05:44,089 --> 00:05:45,830
of life as it could be.

147
00:05:45,830 --> 00:05:49,640
So this is an
extraordinarily stuff

148
00:05:49,640 --> 00:05:51,604
free knew of life.

149
00:05:51,604 --> 00:05:55,565
It's a, it's a good
system science fit

150
00:05:55,565 --> 00:05:59,990
for us this idea that
it is the organization

151
00:05:59,990 --> 00:06:03,890
or the pattern that
matters and not the

152
00:06:03,890 --> 00:06:08,600
media in which that
pattern is implemented.

153
00:06:08,600 --> 00:06:12,305
So hit brief
history and a life.

154
00:06:12,305 --> 00:06:14,525
Like I said, the
term was coined in a

155
00:06:14,525 --> 00:06:16,609
six by Chris
Langton but is

156
00:06:16,609 --> 00:06:18,799
informed by earlier work,

157
00:06:18,799 --> 00:06:21,394
notably Norbert Wiener,

158
00:06:21,394 --> 00:06:23,659
cybernetics or control and

159
00:06:23,659 --> 00:06:25,444
communication in
the animal machine,

160
00:06:25,444 --> 00:06:27,275
which connects

161
00:06:27,275 --> 00:06:29,600
artificial and
natural mechanisms

162
00:06:29,600 --> 00:06:32,090
by analyzing
both of those in

163
00:06:32,090 --> 00:06:34,730
terms of cybernetics
and information theory.

164
00:06:34,730 --> 00:06:36,904
And then by Newman in

165
00:06:36,904 --> 00:06:38,810
1966 with

166
00:06:38,810 --> 00:06:41,780
the theory of
self-reproducing automata.

167
00:06:41,780 --> 00:06:46,040
Which attempts to
formalize my sorry,

168
00:06:46,040 --> 00:06:47,809
a computable
formalization of

169
00:06:47,809 --> 00:06:50,315
the basic processes
of living systems.

170
00:06:50,315 --> 00:06:52,340
And so that's
where you get

171
00:06:52,340 --> 00:06:54,590
cellular automata,
which is,

172
00:06:54,590 --> 00:06:56,599
you know, some
of the earliest

173
00:06:56,599 --> 00:06:59,765
artificial life
that we see.

174
00:06:59,765 --> 00:07:02,659
Okay, so again,
960 LinkedIn

175
00:07:02,659 --> 00:07:05,884
coins the term. In 1987.

176
00:07:05,884 --> 00:07:08,344
You get the proceedings of

177
00:07:08,344 --> 00:07:11,869
an inter-district, an
inter-disciplinary,

178
00:07:11,869 --> 00:07:13,489
a workshop on
the synthesis

179
00:07:13,489 --> 00:07:15,574
and simulation of
living systems,

180
00:07:15,574 --> 00:07:17,749
which was actually
surprising to me,

181
00:07:17,749 --> 00:07:19,340
published by Santa Fe,

182
00:07:19,340 --> 00:07:22,235
but it was held at Los
Alamos National Labs.

183
00:07:22,235 --> 00:07:23,329
Didn't know that.

184
00:07:23,329 --> 00:07:25,904
Anyway, that is
artificial life.

185
00:07:25,904 --> 00:07:30,444
One. In 990, you get
artificial light to

186
00:07:30,444 --> 00:07:32,709
the second interdisciplinary
workshop at

187
00:07:32,709 --> 00:07:35,259
Santa Fe in 990 one,

188
00:07:35,259 --> 00:07:37,569
you get the first
European Conference

189
00:07:37,569 --> 00:07:39,564
on artificial life.

190
00:07:39,564 --> 00:07:43,645
9293 are back to

191
00:07:43,645 --> 00:07:46,044
Santa Fe workshops

192
00:07:46,044 --> 00:07:48,325
and another European
conference.

193
00:07:48,325 --> 00:07:49,569
And then in 990

194
00:07:49,569 --> 00:07:54,115
for the interdisciplinary
workshops at

195
00:07:54,115 --> 00:07:57,459
Santa Fe become the
International Conference

196
00:07:57,459 --> 00:07:58,540
on the simulation and

197
00:07:58,540 --> 00:08:00,639
synthesis of
living systems.

198
00:08:00,639 --> 00:08:02,185
I've highlighted it's

199
00:08:02,185 --> 00:08:03,789
synthesised in
simulation and

200
00:08:03,789 --> 00:08:05,469
one angle and simulation

201
00:08:05,469 --> 00:08:07,254
and synthesis because I,

202
00:08:07,254 --> 00:08:11,275
I thought I was mistakes
in my own typing.

203
00:08:11,275 --> 00:08:13,450
But if you go and look
at the conferences,

204
00:08:13,450 --> 00:08:14,919
they actually
flip-flop these

205
00:08:14,919 --> 00:08:17,275
words back and forth
the number of times.

206
00:08:17,275 --> 00:08:20,169
It's just a little I

207
00:08:20,169 --> 00:08:21,460
don't know. It
was funny to me.

208
00:08:21,460 --> 00:08:23,090
I like it.

209
00:08:23,220 --> 00:08:25,719
So we'll say after 90 for

210
00:08:25,719 --> 00:08:26,994
these conferences will

211
00:08:26,994 --> 00:08:29,095
alternate pretty
much annually.

212
00:08:29,095 --> 00:08:30,984
One-year is Europe.

213
00:08:30,984 --> 00:08:33,130
The European
conference, one year is

214
00:08:33,130 --> 00:08:36,010
the International
Conference until

215
00:08:36,010 --> 00:08:39,010
2018 when they
seem to merge

216
00:08:39,010 --> 00:08:42,984
into the conference
on artificial life.

217
00:08:42,984 --> 00:08:45,760
So I bring this up because

218
00:08:45,760 --> 00:08:48,114
its early history is

219
00:08:48,114 --> 00:08:49,869
I don't want to say murky,

220
00:08:49,869 --> 00:08:52,489
because once you
know the majority of

221
00:08:52,489 --> 00:08:53,630
the important
participants are

222
00:08:53,630 --> 00:08:55,669
still alive and
still working.

223
00:08:55,669 --> 00:08:57,815
But, you know,

224
00:08:57,815 --> 00:08:59,795
it took me a long
time and actually

225
00:08:59,795 --> 00:09:01,039
quite a bit of work to

226
00:09:01,039 --> 00:09:05,030
realize that
artificial life.

227
00:09:05,030 --> 00:09:07,369
One was the proceedings

228
00:09:07,369 --> 00:09:08,960
of this workshop and not

229
00:09:08,960 --> 00:09:12,890
volume one of the
journal artificial life.

230
00:09:12,890 --> 00:09:16,025
So you have the,

231
00:09:16,025 --> 00:09:18,200
the proceedings
which are called

232
00:09:18,200 --> 00:09:19,670
artificial life or a

233
00:09:19,670 --> 00:09:21,469
life and they're numbered.

234
00:09:21,469 --> 00:09:24,109
You also have the
Journal Artificial Life,

235
00:09:24,109 --> 00:09:26,000
which has various volumes.

236
00:09:26,000 --> 00:09:29,449
There's just a lot of
sort of confusion about

237
00:09:29,449 --> 00:09:35,979
what papers Our in what
public publications.

238
00:09:35,979 --> 00:09:38,590
And at least for me, I've

239
00:09:38,590 --> 00:09:39,939
had a fair amount of

240
00:09:39,939 --> 00:09:40,390
difficulty

241
00:09:40,390 --> 00:09:42,145
actually findings
to produce things.

242
00:09:42,145 --> 00:09:43,719
The first four,

243
00:09:43,719 --> 00:09:46,870
at least artificial
life proceedings

244
00:09:46,870 --> 00:09:49,419
are really only available.

245
00:09:49,419 --> 00:09:51,355
It seems in paper copies.

246
00:09:51,355 --> 00:09:54,444
It's very difficult to
find digital versions.

247
00:09:54,444 --> 00:09:56,139
Lot of times I know

248
00:09:56,139 --> 00:09:57,895
Dr. Dweck has them
all in the library,

249
00:09:57,895 --> 00:09:59,335
but I don't have them.

250
00:09:59,335 --> 00:10:02,514
And so it can borrowed.

251
00:10:02,514 --> 00:10:05,815
Well, I will I will
totally do that.

252
00:10:05,815 --> 00:10:08,289
Is just something
I I noticed it

253
00:10:08,289 --> 00:10:10,390
in that, you know,

254
00:10:10,390 --> 00:10:12,834
you'll you'll
reread a reference

255
00:10:12,834 --> 00:10:15,115
and then you go try to
find that reference.

256
00:10:15,115 --> 00:10:19,959
And it's just so these

257
00:10:19,959 --> 00:10:21,970
are my own struggles as I

258
00:10:21,970 --> 00:10:24,880
try to build up the
background and my work.

259
00:10:24,880 --> 00:10:28,119
Okay, So categorizing
artificial life,

260
00:10:28,119 --> 00:10:31,089
too easy ways to
categorize it.

261
00:10:31,089 --> 00:10:33,069
One, we can categorize

262
00:10:33,069 --> 00:10:35,560
it by philosophical
commitment.

263
00:10:35,560 --> 00:10:41,679
So you have strong
a life advocates

264
00:10:41,679 --> 00:10:44,380
who are saying
that a successful,

265
00:10:44,380 --> 00:10:47,065
a live system is
actually alive.

266
00:10:47,065 --> 00:10:47,680
And so this is

267
00:10:47,680 --> 00:10:50,559
a sort of ontological
question.

268
00:10:50,559 --> 00:10:52,389
If you were able to

269
00:10:52,389 --> 00:10:55,419
generate the right system,

270
00:10:55,419 --> 00:10:56,815
you could think of it as,

271
00:10:56,815 --> 00:10:59,480
as really being alive.

272
00:10:59,650 --> 00:11:02,510
By contrast, you have

273
00:11:02,510 --> 00:11:03,619
the weak position which

274
00:11:03,619 --> 00:11:04,940
says that as successful,

275
00:11:04,940 --> 00:11:06,349
a light system is,

276
00:11:06,349 --> 00:11:09,079
isn't, you know,
ontologically alive?

277
00:11:09,079 --> 00:11:11,614
But it is alive enough.

278
00:11:11,614 --> 00:11:13,460
It's lifelike enough
so that we can

279
00:11:13,460 --> 00:11:14,299
learn something about

280
00:11:14,299 --> 00:11:16,100
actually living systems.

281
00:11:16,100 --> 00:11:18,410
And my, I had

282
00:11:18,410 --> 00:11:21,080
originally wanted to
say something that,

283
00:11:21,080 --> 00:11:22,940
you know, that
weak position is

284
00:11:22,940 --> 00:11:24,634
basically uncontroversial.

285
00:11:24,634 --> 00:11:26,659
And it is uncontroversial

286
00:11:26,659 --> 00:11:28,010
within artificial life and

287
00:11:28,010 --> 00:11:29,210
probably within

288
00:11:29,210 --> 00:11:32,209
the larger community
of system science.

289
00:11:32,209 --> 00:11:33,619
And those of us who

290
00:11:33,619 --> 00:11:39,650
are fully bought into
that stuff free style

291
00:11:39,650 --> 00:11:42,229
of science and philosophy.

292
00:11:42,229 --> 00:11:44,810
Um, but it occurs
to me that's not,

293
00:11:44,810 --> 00:11:47,180
not a universal
position, right?

294
00:11:47,180 --> 00:11:48,619
So that's, that
one is still are

295
00:11:48,619 --> 00:11:50,390
also being sort of argued.

296
00:11:50,390 --> 00:11:52,640
That epistemological
argument is very

297
00:11:52,640 --> 00:11:57,394
real outside of a life
and system science.

298
00:11:57,394 --> 00:11:58,864
Okay?

299
00:11:58,864 --> 00:12:00,604
The other way to

300
00:12:00,604 --> 00:12:03,290
categorize artificial
life is by domain.

301
00:12:03,290 --> 00:12:05,750
We can talk about
what a life,

302
00:12:05,750 --> 00:12:09,170
which is synthetic
chemistry and biology,

303
00:12:09,170 --> 00:12:11,569
laboratory creating
proteins and

304
00:12:11,569 --> 00:12:13,549
cells and things
like that.

305
00:12:13,549 --> 00:12:16,475
We can talk about
hard a life,

306
00:12:16,475 --> 00:12:18,229
which is robotics,

307
00:12:18,229 --> 00:12:20,660
specifically
evolutionary robotics

308
00:12:20,660 --> 00:12:22,430
and swarm robotics.

309
00:12:22,430 --> 00:12:26,870
Life inspired
robot designs.

310
00:12:26,870 --> 00:12:30,950
And then finally,
and for me,

311
00:12:30,950 --> 00:12:32,720
the most important, we can

312
00:12:32,720 --> 00:12:34,324
talk about soft a life,

313
00:12:34,324 --> 00:12:35,599
which are computational

314
00:12:35,599 --> 00:12:37,100
models and simulations.

315
00:12:37,100 --> 00:12:38,494
And that is where our

316
00:12:38,494 --> 00:12:40,339
open ended
evolution question

317
00:12:40,339 --> 00:12:43,114
is going to be
basically located.

318
00:12:43,114 --> 00:12:43,939
Okay?

319
00:12:43,939 --> 00:12:48,840
So just some artificial
evolution background.

320
00:12:49,810 --> 00:12:53,209
In order to represent
evolution in

321
00:12:53,209 --> 00:12:54,349
a computer is a

322
00:12:54,349 --> 00:12:57,365
pretty simple,
basic algorithm.

323
00:12:57,365 --> 00:12:59,390
You're going to start

324
00:12:59,390 --> 00:13:01,715
with some randomized
population

325
00:13:01,715 --> 00:13:04,819
of initial
candidates solutions

326
00:13:04,819 --> 00:13:06,980
to a given problem, right?

327
00:13:06,980 --> 00:13:08,299
You're going to evaluate

328
00:13:08,299 --> 00:13:09,440
those candidates according

329
00:13:09,440 --> 00:13:11,660
to some explicit
fitness function,

330
00:13:11,660 --> 00:13:13,924
also called an
objective function.

331
00:13:13,924 --> 00:13:16,400
You are going to
preferentially

332
00:13:16,400 --> 00:13:18,649
select the higher
performing candidates

333
00:13:18,649 --> 00:13:20,014
for reproduction.

334
00:13:20,014 --> 00:13:21,934
And then you're
going to introduce

335
00:13:21,934 --> 00:13:24,799
some variation in
that reproduction

336
00:13:24,799 --> 00:13:28,430
through either
genetic crossover

337
00:13:28,430 --> 00:13:30,935
as in sexual
selection or sorry,

338
00:13:30,935 --> 00:13:33,005
sexual reproduction,
some kind of

339
00:13:33,005 --> 00:13:36,860
point mutation or other
genetic operators

340
00:13:36,860 --> 00:13:40,354
to create variation
in the genomes

341
00:13:40,354 --> 00:13:43,805
that represent these,

342
00:13:43,805 --> 00:13:45,560
these candidate solutions.

343
00:13:45,560 --> 00:13:47,779
And the purpose of
this approach is

344
00:13:47,779 --> 00:13:50,299
to incrementally
advance through

345
00:13:50,299 --> 00:13:52,879
some large search
space towards

346
00:13:52,879 --> 00:13:55,070
a global optimum outcome

347
00:13:55,070 --> 00:13:57,305
as defined by your
objective function.

348
00:13:57,305 --> 00:14:01,145
So, so far so good.

349
00:14:01,145 --> 00:14:03,649
It's a pretty neat way

350
00:14:03,649 --> 00:14:05,750
to implement what
we understand to

351
00:14:05,750 --> 00:14:08,629
be the basic features

352
00:14:08,629 --> 00:14:10,670
of Darwinian
evolution, right?

353
00:14:10,670 --> 00:14:12,650
And it's really effective

354
00:14:12,650 --> 00:14:15,005
at solving a lot of
engineering problems.

355
00:14:15,005 --> 00:14:17,149
You know,
assuming that you

356
00:14:17,149 --> 00:14:19,114
can specify the
problem well,

357
00:14:19,114 --> 00:14:20,989
and assuming that
you can come up

358
00:14:20,989 --> 00:14:23,689
with an encoding
that allows

359
00:14:23,689 --> 00:14:25,519
you to represent

360
00:14:25,519 --> 00:14:29,165
candidate solutions
to that problem.

361
00:14:29,165 --> 00:14:36,469
As some kind of as an
encoding, as a genome.

362
00:14:36,469 --> 00:14:38,659
Than artificial
evolution can

363
00:14:38,659 --> 00:14:39,979
help you move through

364
00:14:39,979 --> 00:14:42,589
those search
spaces in a way

365
00:14:42,589 --> 00:14:45,260
that exhaustive trial and

366
00:14:45,260 --> 00:14:47,690
error just cannot do.

367
00:14:47,690 --> 00:14:49,775
But some problems with

368
00:14:49,775 --> 00:14:51,440
artificial evolution
is that it's

369
00:14:51,440 --> 00:14:52,610
not always
straightforward to

370
00:14:52,610 --> 00:14:53,870
specify an objective

371
00:14:53,870 --> 00:14:55,160
function for your problem.

372
00:14:55,160 --> 00:14:57,620
Even if the problem
is well understood.

373
00:14:57,620 --> 00:15:00,559
And sense, what
artificial evolution

374
00:15:00,559 --> 00:15:02,299
is going to do is give

375
00:15:02,299 --> 00:15:04,969
you an optimal solution to

376
00:15:04,969 --> 00:15:06,649
whatever you
would specify it

377
00:15:06,649 --> 00:15:08,479
as your objective
function.

378
00:15:08,479 --> 00:15:11,224
It really, really matters.

379
00:15:11,224 --> 00:15:13,565
What that objective
function is,

380
00:15:13,565 --> 00:15:16,475
is you're going
to get the best.

381
00:15:16,475 --> 00:15:18,560
Well, you're hopefully
going to get

382
00:15:18,560 --> 00:15:21,289
the best available
answer to that question.

383
00:15:21,289 --> 00:15:22,279
So how you pose

384
00:15:22,279 --> 00:15:24,335
the question is
obviously important.

385
00:15:24,335 --> 00:15:28,249
For sufficiently
difficult tasks.

386
00:15:28,249 --> 00:15:32,945
You know, you might
not be able to get

387
00:15:32,945 --> 00:15:35,269
any traction in

388
00:15:35,269 --> 00:15:38,269
your initial population
of candidates.

389
00:15:38,269 --> 00:15:39,409
So all, if all of

390
00:15:39,409 --> 00:15:40,490
those initial candidates

391
00:15:40,490 --> 00:15:42,485
demonstrate 0 fitness,

392
00:15:42,485 --> 00:15:45,469
then there's no way to
preferentially select.

393
00:15:45,469 --> 00:15:48,455
This is called the
bootstrapping problem.

394
00:15:48,455 --> 00:15:51,080
You also have issues where

395
00:15:51,080 --> 00:15:53,719
candidates can become
stuck on local optima,

396
00:15:53,719 --> 00:15:57,470
where the objective
fitness functions are,

397
00:15:57,470 --> 00:15:59,090
it prevents them from,

398
00:15:59,090 --> 00:16:00,995
you know, climbing down

399
00:16:00,995 --> 00:16:02,269
locally in order to get

400
00:16:02,269 --> 00:16:04,159
backup globally, my date,

401
00:16:04,159 --> 00:16:05,479
they get stuck
on the top of

402
00:16:05,479 --> 00:16:07,220
the hill and the
objective function

403
00:16:07,220 --> 00:16:08,930
prevents them from
getting back down

404
00:16:08,930 --> 00:16:11,670
that hill to
find out more.

405
00:16:12,430 --> 00:16:15,575
A better overall
solution to the problem.

406
00:16:15,575 --> 00:16:16,610
So we're going to say all

407
00:16:16,610 --> 00:16:19,319
of these potential
problems

408
00:16:20,080 --> 00:16:23,539
have discussion,

409
00:16:23,539 --> 00:16:25,009
detailed
discussion and the

410
00:16:25,009 --> 00:16:27,514
literature on
Evolutionary computing,

411
00:16:27,514 --> 00:16:30,199
evolution, genetic
algorithms,

412
00:16:30,199 --> 00:16:31,819
evolutionary solutions.

413
00:16:31,819 --> 00:16:35,960
There's, there's
it's a Y literature.

414
00:16:35,960 --> 00:16:38,540
It's mostly an engineering
focus literature,

415
00:16:38,540 --> 00:16:41,150
but all of these
problems come up.

416
00:16:41,150 --> 00:16:42,574
They have solutions.

417
00:16:42,574 --> 00:16:44,165
It's, it's discussed.

418
00:16:44,165 --> 00:16:46,639
But an important one

419
00:16:46,639 --> 00:16:48,815
that remains for
the discussion of

420
00:16:48,815 --> 00:16:51,605
open ended
evolution is that

421
00:16:51,605 --> 00:16:55,140
nature is not an
optimization problem.

422
00:16:55,140 --> 00:16:57,669
The real-world, right?

423
00:16:57,669 --> 00:17:00,999
Doesn't actually
have there.

424
00:17:00,999 --> 00:17:03,190
There's nothing in, in

425
00:17:03,190 --> 00:17:05,529
general that all
living things

426
00:17:05,529 --> 00:17:07,210
are trying to
optimize for, right?

427
00:17:07,210 --> 00:17:09,085
So at their best,

428
00:17:09,085 --> 00:17:10,480
we're going to
say evolutionary

429
00:17:10,480 --> 00:17:11,980
algorithms will

430
00:17:11,980 --> 00:17:13,989
converge on a single
optimal solution

431
00:17:13,989 --> 00:17:15,340
to well-defined problem.

432
00:17:15,340 --> 00:17:17,260
But again, the biosphere,

433
00:17:17,260 --> 00:17:19,540
what is it trying
to optimize?

434
00:17:19,540 --> 00:17:23,199
What would be
an example of

435
00:17:23,199 --> 00:17:25,239
the single best technology

436
00:17:25,239 --> 00:17:28,374
or culture or social form?

437
00:17:28,374 --> 00:17:32,260
You know, my, my
father's and economists,

438
00:17:32,260 --> 00:17:34,330
I had this
conversation with m.

439
00:17:34,330 --> 00:17:37,184
What is the economy

440
00:17:37,184 --> 00:17:39,724
trying to optimize
as a whole?

441
00:17:39,724 --> 00:17:42,590
And with a total
straight face

442
00:17:42,590 --> 00:17:43,939
and an absolute
lack of irony.

443
00:17:43,939 --> 00:17:47,045
He said, it's maximizing
the general welfare.

444
00:17:47,045 --> 00:17:49,084
And I laughed, but

445
00:17:49,084 --> 00:17:51,259
he did not, he
didn't get that job.

446
00:17:51,259 --> 00:17:53,480
I guess I would say
what an economy

447
00:17:53,480 --> 00:17:56,570
composed of a set of
identical optimal firms

448
00:17:56,570 --> 00:17:58,909
exchanging goods and
services with a set of

449
00:17:58,909 --> 00:18:00,950
identical optimal
households really

450
00:18:00,950 --> 00:18:03,720
maximize the
general welfare.

451
00:18:04,390 --> 00:18:07,069
Useful, All
models are wrong.

452
00:18:07,069 --> 00:18:08,569
Some models are
useful, okay?

453
00:18:08,569 --> 00:18:10,264
Open-ended evolution.

454
00:18:10,264 --> 00:18:11,929
So what we're
going to say now

455
00:18:11,929 --> 00:18:13,399
is that evolution without

456
00:18:13,399 --> 00:18:16,475
an end or without an
objective is open-ended.

457
00:18:16,475 --> 00:18:18,514
And as a concept,

458
00:18:18,514 --> 00:18:20,839
it is an effort to capture

459
00:18:20,839 --> 00:18:23,119
that what it is about

460
00:18:23,119 --> 00:18:24,200
the real world and

461
00:18:24,200 --> 00:18:25,774
real evolving
systems that we have

462
00:18:25,774 --> 00:18:27,050
failed to capture with

463
00:18:27,050 --> 00:18:28,954
our evolutionary
algorithms,

464
00:18:28,954 --> 00:18:31,969
optimizing solutions

465
00:18:31,969 --> 00:18:35,930
to well-defined
objective functions.

466
00:18:35,930 --> 00:18:38,150
And so as a
research community

467
00:18:38,150 --> 00:18:39,935
and as a body
of literature,

468
00:18:39,935 --> 00:18:42,109
open-ended evolution
exists almost

469
00:18:42,109 --> 00:18:43,714
entirely within a life.

470
00:18:43,714 --> 00:18:45,770
You don't see it
talked about very

471
00:18:45,770 --> 00:18:48,799
much in, in biology.

472
00:18:48,799 --> 00:18:51,290
You don't see it
talked about very much

473
00:18:51,290 --> 00:18:55,294
in the social sciences,
anything like that.

474
00:18:55,294 --> 00:18:56,690
Too surprising
because it is

475
00:18:56,690 --> 00:18:59,435
such an important
seeming concept,

476
00:18:59,435 --> 00:19:01,609
arguably a
fundamental concept

477
00:19:01,609 --> 00:19:04,309
about the natural world,

478
00:19:04,309 --> 00:19:06,319
but it is only

479
00:19:06,319 --> 00:19:09,210
really well-defined
in this narrow,

480
00:19:09,730 --> 00:19:12,770
narrow body of literature.

481
00:19:12,770 --> 00:19:15,020
Surprisingly to me.

482
00:19:15,020 --> 00:19:16,490
But I guess, you know,

483
00:19:16,490 --> 00:19:18,140
everyone thinks that
what they do is

484
00:19:18,140 --> 00:19:19,820
the most important
thing in the world.

485
00:19:19,820 --> 00:19:23,104
So I guess that's what's
happening to me now.

486
00:19:23,104 --> 00:19:24,274
Okay.

487
00:19:24,274 --> 00:19:26,569
Brief History of
open-ended evolution.

488
00:19:26,569 --> 00:19:28,759
We're going to
say it has been

489
00:19:28,759 --> 00:19:31,009
an important goal sort of

490
00:19:31,009 --> 00:19:32,750
organizing concept since

491
00:19:32,750 --> 00:19:34,910
the beginning of
artificial life research

492
00:19:34,910 --> 00:19:36,754
in the late eighties,

493
00:19:36,754 --> 00:19:40,740
though again, going to
build on older work.

494
00:19:41,290 --> 00:19:44,479
But we'll say early 90s

495
00:19:44,479 --> 00:19:46,939
sees some foundational
measures and models.

496
00:19:46,939 --> 00:19:50,869
You have PEDOT and
Packard writing

497
00:19:50,869 --> 00:19:52,925
about evolutionary
activity statistics

498
00:19:52,925 --> 00:19:55,220
and the second artificial,

499
00:19:55,220 --> 00:19:57,949
artificial life to
the second workshop,

500
00:19:57,949 --> 00:20:01,910
re, Thomas re
describes his model,

501
00:20:01,910 --> 00:20:04,175
tiara and same volume.

502
00:20:04,175 --> 00:20:08,779
Africa and Dani put out
a vitae which is tear

503
00:20:08,779 --> 00:20:10,625
like we'll talk
about these models

504
00:20:10,625 --> 00:20:13,055
in a second and 990 three.

505
00:20:13,055 --> 00:20:17,880
And then Yeager has
poly world and 94.

506
00:20:18,070 --> 00:20:21,219
And this is just,
this is some of

507
00:20:21,219 --> 00:20:25,255
the sort of still
playing around.

508
00:20:25,255 --> 00:20:26,950
And it's just
trying to get

509
00:20:26,950 --> 00:20:29,920
even a loose sense
of what this is.

510
00:20:29,920 --> 00:20:31,570
There's still a lot
of disagreement

511
00:20:31,570 --> 00:20:33,970
about exactly
what constitutes

512
00:20:33,970 --> 00:20:36,610
open-end evolution
at this point.

513
00:20:36,610 --> 00:20:40,269
It's, this is an
exploratory phase right?

514
00:20:40,269 --> 00:20:42,220
Around the turn
of the century,

515
00:20:42,220 --> 00:20:45,235
we start getting some
clarification concepts

516
00:20:45,235 --> 00:20:47,740
and some other
important models.

517
00:20:47,740 --> 00:20:51,639
Again, Bento has
a classification

518
00:20:51,639 --> 00:20:53,815
of long-term
evolutionary dynamics.

519
00:20:53,815 --> 00:20:57,129
And 98 Geb comes

520
00:20:57,129 --> 00:21:01,030
out in 2001 from,
from Shannon.

521
00:21:01,030 --> 00:21:06,244
And that is a widely
believed, will say,

522
00:21:06,244 --> 00:21:09,620
to be the first digital
evolution system

523
00:21:09,620 --> 00:21:12,930
that is capable
of passing the,

524
00:21:13,360 --> 00:21:16,459
you know, the, the test of

525
00:21:16,459 --> 00:21:18,739
open-ended dynamics
that better

526
00:21:18,739 --> 00:21:20,930
put down and 998.

527
00:21:20,930 --> 00:21:22,730
So you're starting to get,

528
00:21:22,730 --> 00:21:24,934
at this point,
early 2000s,

529
00:21:24,934 --> 00:21:28,054
some efforts to say,

530
00:21:28,054 --> 00:21:29,810
you know, we need to all

531
00:21:29,810 --> 00:21:32,255
collectively figure
out what we all need.

532
00:21:32,255 --> 00:21:34,235
We need to be able to

533
00:21:34,235 --> 00:21:35,809
produce candidate
measures of this.

534
00:21:35,809 --> 00:21:37,130
We need to have
some models that

535
00:21:37,130 --> 00:21:38,735
respond to those
candidate measures.

536
00:21:38,735 --> 00:21:40,129
And so it's starting to

537
00:21:40,129 --> 00:21:42,199
organize itself
as a discipline.

538
00:21:42,199 --> 00:21:44,239
And then.

539
00:21:44,239 --> 00:21:45,949
For whatever
reason, there is

540
00:21:45,949 --> 00:21:47,254
a pretty large gap

541
00:21:47,254 --> 00:21:48,679
there between that turn of

542
00:21:48,679 --> 00:21:50,779
the century and then
when you start seeing

543
00:21:50,779 --> 00:21:54,004
a lot more active work
in open-end evolution,

544
00:21:54,004 --> 00:21:57,709
2015 is the first
workshop on

545
00:21:57,709 --> 00:21:59,269
open-ended evolution

546
00:21:59,269 --> 00:22:01,729
in the European Conference

547
00:22:01,729 --> 00:22:04,280
on artificial
life at York.

548
00:22:04,280 --> 00:22:07,985
And then 2016 you
got the second one.

549
00:22:07,985 --> 00:22:11,674
Again, there's that
synthesis and simulation.

550
00:22:11,674 --> 00:22:14,584
O e three is in Tokyo.

551
00:22:14,584 --> 00:22:16,159
We'll talk about the
importance of New York

552
00:22:16,159 --> 00:22:18,080
and Tokyo here
in a minute.

553
00:22:18,080 --> 00:22:19,879
And then 2019 is

554
00:22:19,879 --> 00:22:22,144
important because we
get a special edition,

555
00:22:22,144 --> 00:22:24,769
a special issue of
artificial life,

556
00:22:24,769 --> 00:22:27,740
volume 25, which

557
00:22:27,740 --> 00:22:30,019
features open-ended
abolition.

558
00:22:30,019 --> 00:22:32,990
And then finally,
OE for in

559
00:22:32,990 --> 00:22:37,384
2021 conference
held virtually

560
00:22:37,384 --> 00:22:40,655
and Central European
summertime.

561
00:22:40,655 --> 00:22:43,265
It was supposed to be
and the Czech Republic.

562
00:22:43,265 --> 00:22:48,379
But COVID forced
it to go remote.

563
00:22:48,379 --> 00:22:51,800
And so it was held from

564
00:22:51,800 --> 00:22:56,284
three to AM. Local time.

565
00:22:56,284 --> 00:22:58,009
Unfortunately I missed it,

566
00:22:58,009 --> 00:23:01,130
but maybe next time.

567
00:23:01,130 --> 00:23:03,350
Okay, so some

568
00:23:03,350 --> 00:23:06,184
definitions of
open-ended evolution.

569
00:23:06,184 --> 00:23:09,770
As I said, 2015,

570
00:23:09,770 --> 00:23:13,654
we have this first
workshop discussion,

571
00:23:13,654 --> 00:23:14,479
open discussion.

572
00:23:14,479 --> 00:23:16,250
At that workshop,
participants

573
00:23:16,250 --> 00:23:18,049
constructed a sort

574
00:23:18,049 --> 00:23:20,990
of outline of what
they think of

575
00:23:20,990 --> 00:23:23,149
as the key
behavioural hallmarks

576
00:23:23,149 --> 00:23:24,289
of open-ended evolution.

577
00:23:24,289 --> 00:23:27,140
And so what we have
are two things.

578
00:23:27,140 --> 00:23:30,274
One, the ongoing
generation

579
00:23:30,274 --> 00:23:32,884
of adaptive novelty,

580
00:23:32,884 --> 00:23:35,789
which is subdivided into

581
00:23:36,070 --> 00:23:39,740
the ongoing generation
of new adaptations.

582
00:23:39,740 --> 00:23:41,209
The ongoing generation of

583
00:23:41,209 --> 00:23:43,100
new kinds of entities.

584
00:23:43,100 --> 00:23:45,995
The emergence of a
dynamical hierarchy

585
00:23:45,995 --> 00:23:47,735
and major transitions,

586
00:23:47,735 --> 00:23:51,274
and the evolution
of evolvability.

587
00:23:51,274 --> 00:23:54,680
So new adaptations,
new kinds of entities,

588
00:23:54,680 --> 00:23:56,914
I think are fairly
straightforward.

589
00:23:56,914 --> 00:24:00,979
The idea of a
dynamical hierarchy is

590
00:24:00,979 --> 00:24:05,435
this idea of
when a certain,

591
00:24:05,435 --> 00:24:08,299
when a set of elements at

592
00:24:08,299 --> 00:24:11,809
one level are
organized in, in a,

593
00:24:11,809 --> 00:24:13,295
in a regular pattern

594
00:24:13,295 --> 00:24:15,634
of relationships
such that,

595
00:24:15,634 --> 00:24:18,785
that pattern
really deserves

596
00:24:18,785 --> 00:24:21,004
status as an entity.

597
00:24:21,004 --> 00:24:24,439
That's well, actually
that's bullet.

598
00:24:24,439 --> 00:24:25,940
That's a major
transition which I

599
00:24:25,940 --> 00:24:28,100
guess is a category of,

600
00:24:28,100 --> 00:24:30,199
of a dynamical hierarchy.

601
00:24:30,199 --> 00:24:31,999
But the idea is
that you've got

602
00:24:31,999 --> 00:24:34,700
this nesting of
organization.

603
00:24:34,700 --> 00:24:37,949
Where organization
at one level,

604
00:24:38,050 --> 00:24:41,525
the elements of a system
at a certain level,

605
00:24:41,525 --> 00:24:44,030
our than systems
at a lower level,

606
00:24:44,030 --> 00:24:45,335
which are composed of

607
00:24:45,335 --> 00:24:47,135
elements that is
still lower level,

608
00:24:47,135 --> 00:24:49,009
which are also
composed of elements.

609
00:24:49,009 --> 00:24:51,349
It's sort of this

610
00:24:51,349 --> 00:24:52,519
is a system
science seminar,

611
00:24:52,519 --> 00:24:53,765
so I feel like I should

612
00:24:53,765 --> 00:24:55,609
probably elaborate
too much time.

613
00:24:55,609 --> 00:24:59,164
A concept that you
all know and love.

614
00:24:59,164 --> 00:25:01,699
And okay, so then

615
00:25:01,699 --> 00:25:04,250
the ongoing growth
of complexity means

616
00:25:04,250 --> 00:25:07,040
both the individual
entities

617
00:25:07,040 --> 00:25:09,620
themselves are
becoming more complex.

618
00:25:09,620 --> 00:25:11,600
But that also the systems

619
00:25:11,600 --> 00:25:13,489
in which those entities
participate are

620
00:25:13,489 --> 00:25:17,180
becoming more
complex through

621
00:25:17,180 --> 00:25:21,455
an ongoing generation
of interactions.

622
00:25:21,455 --> 00:25:22,610
And so the, the

623
00:25:22,610 --> 00:25:24,110
complexity of
the interactions

624
00:25:24,110 --> 00:25:27,335
between those
entities increases.

625
00:25:27,335 --> 00:25:28,129
Okay?

626
00:25:28,129 --> 00:25:30,200
So there are some
problems with

627
00:25:30,200 --> 00:25:33,080
this in that there's
quite a bit of overlap,

628
00:25:33,080 --> 00:25:41,195
especially within York
category 111 b, c 1D.

629
00:25:41,195 --> 00:25:46,145
There's the same phenomena

630
00:25:46,145 --> 00:25:48,814
could easily be
said to fall into,

631
00:25:48,814 --> 00:25:50,015
you know, it's a
multiple of these

632
00:25:50,015 --> 00:25:51,379
categories at
the same time.

633
00:25:51,379 --> 00:25:56,435
So at the OEE three
workshop in Tokyo,

634
00:25:56,435 --> 00:25:57,979
there was an effort to

635
00:25:57,979 --> 00:26:00,349
refine those
are categories.

636
00:26:00,349 --> 00:26:03,365
And so we get these
revised categories,

637
00:26:03,365 --> 00:26:06,230
which consist of the
ongoing generation

638
00:26:06,230 --> 00:26:08,014
of the following
four kinds.

639
00:26:08,014 --> 00:26:09,259
Interesting new kinds of

640
00:26:09,259 --> 00:26:10,984
entities and interactions.

641
00:26:10,984 --> 00:26:13,250
Evolution of evolvability,

642
00:26:13,250 --> 00:26:16,909
major transitions and
semantic evolution,

643
00:26:16,909 --> 00:26:18,950
which is a new thing that

644
00:26:18,950 --> 00:26:20,029
comes from some of

645
00:26:20,029 --> 00:26:21,530
the new work that
was being done in,

646
00:26:21,530 --> 00:26:26,690
in Tokyo about how

647
00:26:26,690 --> 00:26:30,484
agents assign meanings and

648
00:26:30,484 --> 00:26:33,335
those meanings shift
sorted in time.

649
00:26:33,335 --> 00:26:34,640
How they, how they tagged

650
00:26:34,640 --> 00:26:36,140
phenomenon in the
world around them.

651
00:26:36,140 --> 00:26:38,060
It's interesting
stuff, but it's not

652
00:26:38,060 --> 00:26:40,129
we're going to talk
about here today.

653
00:26:40,129 --> 00:26:40,940
I'm going to say these

654
00:26:40,940 --> 00:26:42,499
are obviously called the

655
00:26:42,499 --> 00:26:44,089
Tokyo categories
because they

656
00:26:44,089 --> 00:26:45,935
were established in Tokyo.

657
00:26:45,935 --> 00:26:49,459
So interesting,
when we say

658
00:26:49,459 --> 00:26:50,884
interesting new kinds of

659
00:26:50,884 --> 00:26:52,969
entities and interactions,

660
00:26:52,969 --> 00:26:55,369
that word is doing a
lot of heavy work.

661
00:26:55,369 --> 00:27:00,139
And it has been
pointed out repeatedly

662
00:27:00,139 --> 00:27:02,345
throughout the
literature that's

663
00:27:02,345 --> 00:27:05,179
interesting is a very
subjective term.

664
00:27:05,179 --> 00:27:06,514
It's very vague.

665
00:27:06,514 --> 00:27:08,060
And so a lot of work

666
00:27:08,060 --> 00:27:09,499
is then being
done to try to

667
00:27:09,499 --> 00:27:12,140
formalize that
and establish

668
00:27:12,140 --> 00:27:14,464
what do we mean
by Interesting.

669
00:27:14,464 --> 00:27:16,955
So you have, Hence,

670
00:27:16,955 --> 00:27:19,130
who is describing
a system that's

671
00:27:19,130 --> 00:27:21,575
designed to produce
trivial novelty?

672
00:27:21,575 --> 00:27:25,370
And he does that
basically to

673
00:27:25,370 --> 00:27:26,900
point out that you don't

674
00:27:26,900 --> 00:27:29,165
novelty on its own
isn't valuable.

675
00:27:29,165 --> 00:27:30,499
And we really do need to

676
00:27:30,499 --> 00:27:33,155
understand what
interesting needs.

677
00:27:33,155 --> 00:27:35,150
And so the, the,

678
00:27:35,150 --> 00:27:38,600
the basics of that
model are you

679
00:27:38,600 --> 00:27:42,094
have a regular grid of

680
00:27:42,094 --> 00:27:45,335
sites and you initialize

681
00:27:45,335 --> 00:27:50,569
a path that traverses
those sites randomly.

682
00:27:50,569 --> 00:27:56,090
And then basically the
the path reproduces,

683
00:27:56,090 --> 00:27:58,730
you keep creating
new pads and you

684
00:27:58,730 --> 00:28:02,074
evaluate pads for
fitness where

685
00:28:02,074 --> 00:28:04,760
they basically
are more fit when

686
00:28:04,760 --> 00:28:07,445
they traverse more
sites that had

687
00:28:07,445 --> 00:28:09,274
not been traversed by any

688
00:28:09,274 --> 00:28:13,100
other any other
paths, right?

689
00:28:13,100 --> 00:28:15,244
So you're explicitly
rewarding novelty.

690
00:28:15,244 --> 00:28:16,790
And what you get are

691
00:28:16,790 --> 00:28:19,339
a lot of different
pads that are

692
00:28:19,339 --> 00:28:21,049
wandering through

693
00:28:21,049 --> 00:28:24,230
the ever expanding
grid space.

694
00:28:24,230 --> 00:28:26,240
And so you say
yes, technically,

695
00:28:26,240 --> 00:28:28,264
each one of
these is novel,

696
00:28:28,264 --> 00:28:32,480
but they're all just
sequences of up,

697
00:28:32,480 --> 00:28:34,879
down, left, right,
through a regular grid.

698
00:28:34,879 --> 00:28:36,530
And so no one
cares, right?

699
00:28:36,530 --> 00:28:38,029
They're just not, it's,

700
00:28:38,029 --> 00:28:39,440
it's, it's novel,

701
00:28:39,440 --> 00:28:40,730
but it's not at
all interesting.

702
00:28:40,730 --> 00:28:42,110
It's, it's almost by

703
00:28:42,110 --> 00:28:44,270
definition, uninteresting.

704
00:28:44,270 --> 00:28:47,210
So then not necessarily

705
00:28:47,210 --> 00:28:49,505
as a response to
hints because they,

706
00:28:49,505 --> 00:28:51,410
they do come out earlier,

707
00:28:51,410 --> 00:28:52,489
but I'm just
going to set it

708
00:28:52,489 --> 00:28:53,840
up as a contrast.

709
00:28:53,840 --> 00:28:56,150
You have work from bonds,

710
00:28:56,150 --> 00:28:59,855
f et al, Taylor
from Stephanie.

711
00:28:59,855 --> 00:29:02,060
And they are, they
have a set of

712
00:29:02,060 --> 00:29:04,849
related works that create

713
00:29:04,849 --> 00:29:08,450
these sort of
overlapping schemes

714
00:29:08,450 --> 00:29:10,969
for, for categorizing
novelty.

715
00:29:10,969 --> 00:29:15,860
And so they use this idea

716
00:29:15,860 --> 00:29:18,230
that gets laid
out by bonds

717
00:29:18,230 --> 00:29:21,034
up about models
and Meta models.

718
00:29:21,034 --> 00:29:23,510
And this idea of

719
00:29:23,510 --> 00:29:25,640
the state space
within a model.

720
00:29:25,640 --> 00:29:27,425
So you have

721
00:29:27,425 --> 00:29:29,815
the first and most
basic novelty is

722
00:29:29,815 --> 00:29:32,224
novelty that
remains within

723
00:29:32,224 --> 00:29:35,479
the state space
for a given model.

724
00:29:35,479 --> 00:29:38,030
So it's, it's expected
novelty, right?

725
00:29:38,030 --> 00:29:41,554
It's the kind of stuff
that is, you know,

726
00:29:41,554 --> 00:29:44,884
sort of just

727
00:29:44,884 --> 00:29:47,494
basic agent-based
modeling stuff.

728
00:29:47,494 --> 00:29:49,325
You've given it a
rank, you've given

729
00:29:49,325 --> 00:29:50,209
the agents or range of

730
00:29:50,209 --> 00:29:51,259
things that they can do.

731
00:29:51,259 --> 00:29:52,969
And sure enough,
they're going to

732
00:29:52,969 --> 00:29:55,724
do all of them if you
give them enough time.

733
00:29:55,724 --> 00:29:57,940
So you may see one that's

734
00:29:57,940 --> 00:30:01,239
never done that before,

735
00:30:01,239 --> 00:30:03,010
but it was always
able to do it

736
00:30:03,010 --> 00:30:04,210
and you always expected

737
00:30:04,210 --> 00:30:04,990
it to be able to do it.

738
00:30:04,990 --> 00:30:07,825
So it's, it's
just variation,

739
00:30:07,825 --> 00:30:09,340
It's just
exploratory novelty.

740
00:30:09,340 --> 00:30:10,389
It's fully expected.

741
00:30:10,389 --> 00:30:13,045
Innovation or
expansive novelty.

742
00:30:13,045 --> 00:30:14,949
Is that novelty which

743
00:30:14,949 --> 00:30:18,070
extends the state space
of a mile, right.

744
00:30:18,070 --> 00:30:23,229
So you you have a model?

745
00:30:23,229 --> 00:30:23,965
Well, yeah.

746
00:30:23,965 --> 00:30:26,169
I mean, I it it extends
the state space.

747
00:30:26,169 --> 00:30:29,079
It's, it's
making new kinds

748
00:30:29,079 --> 00:30:31,420
of things for your
agents to be able to do.

749
00:30:31,420 --> 00:30:33,849
And then the idea
of emergence or

750
00:30:33,849 --> 00:30:37,104
transformational
novelty is novelty,

751
00:30:37,104 --> 00:30:40,310
which also changing
the model state-space,

752
00:30:40,310 --> 00:30:41,450
but it's changing it in

753
00:30:41,450 --> 00:30:43,999
such a way that
your Meta model,

754
00:30:43,999 --> 00:30:47,239
for the model
requires a change in

755
00:30:47,239 --> 00:30:49,204
order to conceptualize

756
00:30:49,204 --> 00:30:50,795
or capture that change.

757
00:30:50,795 --> 00:30:53,554
And so basically,

758
00:30:53,554 --> 00:30:55,129
they will describe
this as, you know,

759
00:30:55,129 --> 00:30:58,549
variation is a new way

760
00:30:58,549 --> 00:31:00,634
to solve a
familiar problem.

761
00:31:00,634 --> 00:31:04,204
Innovation is solving
a new problem.

762
00:31:04,204 --> 00:31:06,049
And then emergence or

763
00:31:06,049 --> 00:31:07,309
a transformational
novelty is

764
00:31:07,309 --> 00:31:08,869
something like
we've talked about.

765
00:31:08,869 --> 00:31:13,130
A major transition
where the solution to

766
00:31:13,130 --> 00:31:15,050
a problem
fundamentally changes

767
00:31:15,050 --> 00:31:18,020
the way that other agents,

768
00:31:18,020 --> 00:31:20,089
or even approaching the
solutions of problems

769
00:31:20,089 --> 00:31:23,089
or what we think of
as the problems.

770
00:31:23,089 --> 00:31:25,130
And I'd say it's

771
00:31:25,130 --> 00:31:26,570
a little bit fuzzy
this difference

772
00:31:26,570 --> 00:31:29,885
between our innovation
and emergence.

773
00:31:29,885 --> 00:31:31,459
But I think some of

774
00:31:31,459 --> 00:31:32,930
that is just that
emergence itself

775
00:31:32,930 --> 00:31:37,009
is a difficult
concept and it's,

776
00:31:37,009 --> 00:31:38,479
there aren't very
many examples

777
00:31:38,479 --> 00:31:39,815
of this Point app.

778
00:31:39,815 --> 00:31:41,360
So it's, it's hard to get

779
00:31:41,360 --> 00:31:44,480
a very clear and
concrete differentiation

780
00:31:44,480 --> 00:31:45,245
there.

781
00:31:45,245 --> 00:31:47,270
Okay, I gotta move fast.

782
00:31:47,270 --> 00:31:50,404
So measures of
open-ended evolution.

783
00:31:50,404 --> 00:31:52,490
First of all,
we talked about

784
00:31:52,490 --> 00:31:56,135
meadows, evolutionary
activity statistics.

785
00:31:56,135 --> 00:31:58,145
He set out trying
to classify

786
00:31:58,145 --> 00:32:01,129
long-term evolutionary
dynamics by

787
00:32:01,129 --> 00:32:03,230
using these
statistics which

788
00:32:03,230 --> 00:32:04,849
measured what
he called the

789
00:32:04,849 --> 00:32:07,354
evolutionary activity
of components.

790
00:32:07,354 --> 00:32:10,820
So this is sensitive

791
00:32:10,820 --> 00:32:13,025
to what you think
of as a component.

792
00:32:13,025 --> 00:32:17,855
But the basic idea is
that your components,

793
00:32:17,855 --> 00:32:20,870
which are often species or

794
00:32:20,870 --> 00:32:23,419
Tax or something
like that.

795
00:32:23,419 --> 00:32:26,249
Categories of entities.

796
00:32:26,530 --> 00:32:30,410
They accumulate
activity for

797
00:32:30,410 --> 00:32:31,459
each time step that

798
00:32:31,459 --> 00:32:33,200
they're active
in the model.

799
00:32:33,200 --> 00:32:35,330
And when they're
not active,

800
00:32:35,330 --> 00:32:37,519
I would say they're not
going to reactivate.

801
00:32:37,519 --> 00:32:39,380
So then what
we're gonna do

802
00:32:39,380 --> 00:32:41,074
is we're going
to calculate

803
00:32:41,074 --> 00:32:43,174
diversity as the number

804
00:32:43,174 --> 00:32:45,739
of components that
are active at time

805
00:32:45,739 --> 00:32:48,274
t. We're going
to calculate

806
00:32:48,274 --> 00:32:51,274
mean cumulative
evolutionary activity

807
00:32:51,274 --> 00:32:53,449
as the sum of

808
00:32:53,449 --> 00:32:55,040
the activity of

809
00:32:55,040 --> 00:32:57,544
all the components
at a given time.

810
00:32:57,544 --> 00:33:00,230
And then divided by
the diversity, right?

811
00:33:00,230 --> 00:33:01,489
So this is a startup per

812
00:33:01,489 --> 00:33:04,654
component measure of
evolutionary activity.

813
00:33:04,654 --> 00:33:05,990
All right, So again,
it's like if you have

814
00:33:05,990 --> 00:33:07,969
a component that's
been in the model for

815
00:33:07,969 --> 00:33:10,939
10 timesteps and a
component that's been in

816
00:33:10,939 --> 00:33:14,630
the model for a
100 timesteps,

817
00:33:14,630 --> 00:33:18,154
then you're summing
that to a 110,

818
00:33:18,154 --> 00:33:19,609
dividing it by two.

819
00:33:19,609 --> 00:33:20,990
And your mean cumulative

820
00:33:20,990 --> 00:33:22,204
evolution activity is,

821
00:33:22,204 --> 00:33:27,830
is 50 buys steps, right?

822
00:33:27,830 --> 00:33:32,074
Or whatever the unit
for, for activity.

823
00:33:32,074 --> 00:33:36,619
And then new
evolutionary activity is

824
00:33:36,619 --> 00:33:38,044
designed to be a sort of

825
00:33:38,044 --> 00:33:40,520
subset of the
cumulative activity.

826
00:33:40,520 --> 00:33:41,734
What you want to know is

827
00:33:41,734 --> 00:33:46,369
our new adaptations
flowing

828
00:33:46,369 --> 00:33:48,275
into the model, right?

829
00:33:48,275 --> 00:33:50,600
Are our new positive

830
00:33:50,600 --> 00:33:54,724
adaptive components
coming into being.

831
00:33:54,724 --> 00:33:58,879
And so this is tricky
because we don't

832
00:33:58,879 --> 00:34:03,545
want to count a new
component right away.

833
00:34:03,545 --> 00:34:06,109
Because it's possible
that the component is

834
00:34:06,109 --> 00:34:09,050
either neutral or
actually maladaptive.

835
00:34:09,050 --> 00:34:11,450
So what you need
to do is you

836
00:34:11,450 --> 00:34:14,479
need to find a
range of activity,

837
00:34:14,479 --> 00:34:17,554
where below that range,

838
00:34:17,554 --> 00:34:19,519
we can exclude

839
00:34:19,519 --> 00:34:23,315
neutral or maladaptive
components.

840
00:34:23,315 --> 00:34:24,800
And above that range,

841
00:34:24,800 --> 00:34:28,670
we can exclude
this sort of

842
00:34:28,670 --> 00:34:33,019
ongoing active activity of

843
00:34:33,019 --> 00:34:34,309
those components
that are highly

844
00:34:34,309 --> 00:34:35,959
adaptive so that
we can focus

845
00:34:35,959 --> 00:34:40,489
on newly adaptive
components,

846
00:34:40,489 --> 00:34:40,775
right,

847
00:34:40,775 --> 00:34:42,979
to get this new

848
00:34:42,979 --> 00:34:44,704
evolutionary
activity statistic.

849
00:34:44,704 --> 00:34:48,140
And so establishing this
requires this use of

850
00:34:48,140 --> 00:34:50,029
a neutral shadow model

851
00:34:50,029 --> 00:34:52,969
that operates without
selection pressure.

852
00:34:52,969 --> 00:34:56,464
The idea is to
capture the point

853
00:34:56,464 --> 00:35:02,610
where the
distribution of the

854
00:35:03,970 --> 00:35:09,719
basically where the
distribution of activity

855
00:35:09,719 --> 00:35:13,165
for the set of components

856
00:35:13,165 --> 00:35:15,714
cross between your
neutral model

857
00:35:15,714 --> 00:35:17,919
and your real model.

858
00:35:17,919 --> 00:35:20,665
So that you get the
sense of how many,

859
00:35:20,665 --> 00:35:24,054
how much activity can
be expected by chance.

860
00:35:24,054 --> 00:35:25,420
And so you say below

861
00:35:25,420 --> 00:35:26,739
that, we want to say, Oh,

862
00:35:26,739 --> 00:35:28,944
it could be that component

863
00:35:28,944 --> 00:35:31,210
just exists and it
might mean something.

864
00:35:31,210 --> 00:35:32,725
It might not
mean something.

865
00:35:32,725 --> 00:35:34,735
But above that
you can say,

866
00:35:34,735 --> 00:35:37,090
for a component to it
that existed this long,

867
00:35:37,090 --> 00:35:38,440
it must be providing

868
00:35:38,440 --> 00:35:40,779
some adaptive value or it

869
00:35:40,779 --> 00:35:43,689
must have some
adaptive value.

870
00:35:43,689 --> 00:35:44,829
And that's the idea.

871
00:35:44,829 --> 00:35:45,415
Okay?

872
00:35:45,415 --> 00:35:47,584
So using these statistics,

873
00:35:47,584 --> 00:35:49,109
we can classify

874
00:35:49,109 --> 00:35:51,570
long-term
evolutionary dynamics

875
00:35:51,570 --> 00:35:54,314
into three basic
categories.

876
00:35:54,314 --> 00:35:55,859
In class one, you

877
00:35:55,859 --> 00:35:57,359
have no evolutionary
activity,

878
00:35:57,359 --> 00:35:59,009
which is bounded
diversity,

879
00:35:59,009 --> 00:36:02,655
no new activity, no mean
cumulative activity.

880
00:36:02,655 --> 00:36:04,710
In class 2, we're going to

881
00:36:04,710 --> 00:36:06,810
say we have bounded
evolutionary activity,

882
00:36:06,810 --> 00:36:10,409
which means we do have
positive new activity.

883
00:36:10,409 --> 00:36:12,060
We do have positive mean,

884
00:36:12,060 --> 00:36:13,274
cumulative activity,

885
00:36:13,274 --> 00:36:14,879
but diversity is bounded

886
00:36:14,879 --> 00:36:16,544
and beyond a
certain point,

887
00:36:16,544 --> 00:36:21,480
you are not increasing
the population, right?

888
00:36:21,480 --> 00:36:24,120
So this might
be a population

889
00:36:24,120 --> 00:36:25,844
that's cyclin, right?

890
00:36:25,844 --> 00:36:27,539
Where you're, you know,

891
00:36:27,539 --> 00:36:32,030
you are getting new
adaptive components,

892
00:36:32,030 --> 00:36:33,500
but those new
adaptive components

893
00:36:33,500 --> 00:36:35,270
are replacing
previous components,

894
00:36:35,270 --> 00:36:37,205
not expanding on the space

895
00:36:37,205 --> 00:36:40,385
of components rates are
not adding diversity.

896
00:36:40,385 --> 00:36:42,275
And then Class 3,

897
00:36:42,275 --> 00:36:44,000
unbounded evolutionary
activity,

898
00:36:44,000 --> 00:36:45,680
which is the target
we're aiming for,

899
00:36:45,680 --> 00:36:47,960
has unbounded
diversity with

900
00:36:47,960 --> 00:36:49,159
positive new activity and

901
00:36:49,159 --> 00:36:51,034
puzzling cumulative
activity.

902
00:36:51,034 --> 00:36:51,934
Okay?

903
00:36:51,934 --> 00:36:56,225
So our guy, Alice Cannon,

904
00:36:56,225 --> 00:36:59,990
who created the
gap model and

905
00:36:59,990 --> 00:37:02,674
ran the

906
00:37:02,674 --> 00:37:05,105
evolutionary activity
statistics on it,

907
00:37:05,105 --> 00:37:08,690
trying to classify the
evolutionary dynamics.

908
00:37:08,690 --> 00:37:12,699
He found.
Excitingly for him.

909
00:37:12,699 --> 00:37:14,700
I passed the test.

910
00:37:14,700 --> 00:37:16,935
But in passing it,

911
00:37:16,935 --> 00:37:18,120
he realized that he

912
00:37:18,120 --> 00:37:20,609
had some problems
with the way

913
00:37:20,609 --> 00:37:23,459
that better had formulated

914
00:37:23,459 --> 00:37:26,084
those evolutionary
activity statistics.

915
00:37:26,084 --> 00:37:29,219
So he argues that
the components in

916
00:37:29,219 --> 00:37:31,649
his real model are
likely to be more

917
00:37:31,649 --> 00:37:33,179
tightly clustered
than those

918
00:37:33,179 --> 00:37:34,665
in a neutral shadow model.

919
00:37:34,665 --> 00:37:37,919
And so the mutation of

920
00:37:37,919 --> 00:37:40,319
a real component would
be more likely to

921
00:37:40,319 --> 00:37:43,005
produce another highly
active component,

922
00:37:43,005 --> 00:37:44,939
then the mutation of
a shadow component.

923
00:37:44,939 --> 00:37:47,985
And so in order to
correct for this problem,

924
00:37:47,985 --> 00:37:50,849
he has this method of
periodically resetting

925
00:37:50,849 --> 00:37:52,379
the shadow to match

926
00:37:52,379 --> 00:37:54,060
the components and
the activity history

927
00:37:54,060 --> 00:37:54,989
of the real run.

928
00:37:54,989 --> 00:37:59,744
Then in order to,

929
00:37:59,744 --> 00:38:02,460
you know, he wants to have

930
00:38:02,460 --> 00:38:05,159
a new normalized
activity measure

931
00:38:05,159 --> 00:38:07,049
to rebuild these
statistics.

932
00:38:07,049 --> 00:38:09,540
And so now
instead of saying

933
00:38:09,540 --> 00:38:12,090
the activity of a
component increments

934
00:38:12,090 --> 00:38:14,085
at each timestep
where it's present.

935
00:38:14,085 --> 00:38:15,809
He's going to increment

936
00:38:15,809 --> 00:38:18,930
the activity of
a component only

937
00:38:18,930 --> 00:38:20,790
if it's present in

938
00:38:20,790 --> 00:38:23,744
the real model and not
in the shadow model.

939
00:38:23,744 --> 00:38:24,960
And in fact, he will

940
00:38:24,960 --> 00:38:27,569
decrement the activity if

941
00:38:27,569 --> 00:38:29,429
he finds that the
component is present

942
00:38:29,429 --> 00:38:31,740
in the shadow model and
nine, the real model.

943
00:38:31,740 --> 00:38:33,150
And so then using that

944
00:38:33,150 --> 00:38:35,580
new normalized
activity statistic,

945
00:38:35,580 --> 00:38:37,834
he calculates
for that, sorry,

946
00:38:37,834 --> 00:38:41,210
the normal, normalized
activity measure.

947
00:38:41,210 --> 00:38:42,410
He's going to calculate.

948
00:38:42,410 --> 00:38:47,464
New diversity means
cumulative activity,

949
00:38:47,464 --> 00:38:53,405
and new adapted exit
statistical model.

950
00:38:53,405 --> 00:38:57,005
Lipid. Heard no model.
The base model.

951
00:38:57,005 --> 00:38:59,885
That's comparing the
real model or cute.

952
00:38:59,885 --> 00:39:02,839
Yeah, the shadow
honesty is tricky

953
00:39:02,839 --> 00:39:04,385
and I will say that

954
00:39:04,385 --> 00:39:06,304
from what I understand
about the shadow,

955
00:39:06,304 --> 00:39:08,990
I don't actually
totally understand

956
00:39:08,990 --> 00:39:11,600
Shannon's complaint at and

957
00:39:11,600 --> 00:39:12,980
so what I gather is he

958
00:39:12,980 --> 00:39:15,080
talked to better
extensively about it.

959
00:39:15,080 --> 00:39:17,779
So better it seems
to understand it.

960
00:39:17,779 --> 00:39:20,430
That makes me feel
like I needed

961
00:39:20,430 --> 00:39:21,599
to do more to try

962
00:39:21,599 --> 00:39:23,399
to understand what
the issue is.

963
00:39:23,399 --> 00:39:26,384
But the shadow, the
basic idea of the shadow

964
00:39:26,384 --> 00:39:29,520
is that it is

965
00:39:29,520 --> 00:39:31,844
a copy of the real run

966
00:39:31,844 --> 00:39:34,319
where it's got
the same sequence

967
00:39:34,319 --> 00:39:35,714
of births and deaths.

968
00:39:35,714 --> 00:39:38,235
So every time a new agent

969
00:39:38,235 --> 00:39:39,599
or a new component

970
00:39:39,599 --> 00:39:41,310
comes into being
and the real run,

971
00:39:41,310 --> 00:39:42,449
a new component comes

972
00:39:42,449 --> 00:39:44,205
into being and
the shadow run.

973
00:39:44,205 --> 00:39:49,139
Every time component
dies and the real run,

974
00:39:49,139 --> 00:39:52,589
a component dies
and the shadow run.

975
00:39:52,589 --> 00:39:54,090
But in the real run,

976
00:39:54,090 --> 00:39:56,279
births and deaths
are governed by

977
00:39:56,279 --> 00:39:59,850
the actual
performance under

978
00:39:59,850 --> 00:40:01,965
selection pressure
of these components.

979
00:40:01,965 --> 00:40:03,285
Whereas in the shadow run,

980
00:40:03,285 --> 00:40:05,429
it's totally
randomized, right?

981
00:40:05,429 --> 00:40:07,334
So the idea would
be that it's,

982
00:40:07,334 --> 00:40:10,680
it's just a sort
of diffusion.

983
00:40:10,680 --> 00:40:14,399
And so again, from
my understanding,

984
00:40:14,399 --> 00:40:15,809
the purpose of the shadow

985
00:40:15,809 --> 00:40:17,609
is to just get
a sense of how

986
00:40:17,609 --> 00:40:22,439
long should a component
exist by chance.

987
00:40:22,439 --> 00:40:25,229
So that you can get this,

988
00:40:25,229 --> 00:40:28,949
this range of
activity that

989
00:40:28,949 --> 00:40:32,849
tells you when a
new component,

990
00:40:32,849 --> 00:40:34,259
a new adaptive
component is

991
00:40:34,259 --> 00:40:36,330
new and when it's
adapting and

992
00:40:36,330 --> 00:40:37,589
when you can say
for sure this is

993
00:40:37,589 --> 00:40:40,560
an adaptive component
versus when it just,

994
00:40:40,560 --> 00:40:43,064
you know, happens
to be a component.

995
00:40:43,064 --> 00:40:44,940
So I guess I'm not I'm

996
00:40:44,940 --> 00:40:47,999
not totally sure I get
what the problem is.

997
00:40:47,999 --> 00:40:50,909
Because even if
it is drifting,

998
00:40:50,909 --> 00:40:54,089
I don't see why that
should affect the time.

999
00:40:54,089 --> 00:40:56,669
There isn't any selection
in the shadow run,

1000
00:40:56,669 --> 00:40:58,529
so it shouldn't matter how

1001
00:40:58,529 --> 00:41:01,289
far it supposedly drips
from the real run,

1002
00:41:01,289 --> 00:41:02,879
but I'd have to
look into it.

1003
00:41:02,879 --> 00:41:06,405
And if anybody has
any thoughts, chain,

1004
00:41:06,405 --> 00:41:10,920
isn't it the case
that the shot all

1005
00:41:10,920 --> 00:41:13,619
had the same components

1006
00:41:13,619 --> 00:41:15,555
as the real model, right?

1007
00:41:15,555 --> 00:41:17,279
So it does I mean,

1008
00:41:17,279 --> 00:41:18,899
it, it begins that way.

1009
00:41:18,899 --> 00:41:20,189
Yeah, so that's, that's

1010
00:41:20,189 --> 00:41:21,674
the problem with Manson's.

1011
00:41:21,674 --> 00:41:25,425
The real model changes
its components.

1012
00:41:25,425 --> 00:41:28,230
Then the shadow
model is still using

1013
00:41:28,230 --> 00:41:30,870
components that
previously existed

1014
00:41:30,870 --> 00:41:32,100
in the real model.

1015
00:41:32,100 --> 00:41:35,279
And so it's no longer
an F her while it

1016
00:41:35,279 --> 00:41:38,430
no longer shadows
the real model.

1017
00:41:38,430 --> 00:41:41,774
It shadows what the
real model used to be.

1018
00:41:41,774 --> 00:41:43,230
Well, got it.

1019
00:41:43,230 --> 00:41:45,240
Yeah, when there
is a birth,

1020
00:41:45,240 --> 00:41:46,859
there is still a mutation

1021
00:41:46,859 --> 00:41:48,000
that's happening, right?

1022
00:41:48,000 --> 00:41:49,380
So there's no
selective pressure,

1023
00:41:49,380 --> 00:41:50,640
but there's
still mutation.

1024
00:41:50,640 --> 00:41:52,290
So what's happening
is that the shadow

1025
00:41:52,290 --> 00:41:53,969
is diffusing out through

1026
00:41:53,969 --> 00:41:55,574
the space of

1027
00:41:55,574 --> 00:41:58,979
the phenotype space
of the components.

1028
00:41:58,979 --> 00:42:03,254
I like anyway, like I
said, I still right.

1029
00:42:03,254 --> 00:42:05,730
It's where the conversation
I would love to

1030
00:42:05,730 --> 00:42:07,829
talk to anybody
about that.

1031
00:42:07,829 --> 00:42:10,395
We're going to talk
about it at length.

1032
00:42:10,395 --> 00:42:11,819
Yeah, it's, it's,
it's funds.

1033
00:42:11,819 --> 00:42:14,684
Okay? Um, and so on.

1034
00:42:14,684 --> 00:42:15,959
Another set of measures

1035
00:42:15,959 --> 00:42:19,109
for, for open-ended
evolution,

1036
00:42:19,109 --> 00:42:22,994
we have the modes
toolbox from Jolson al,

1037
00:42:22,994 --> 00:42:26,369
working out at
Michigan State with,

1038
00:42:26,369 --> 00:42:30,285
with Africa and
the venous system.

1039
00:42:30,285 --> 00:42:33,405
So they're outlining,
basically,

1040
00:42:33,405 --> 00:42:36,494
they want to take the
opposite take and say,

1041
00:42:36,494 --> 00:42:38,070
when can we be
sure a system

1042
00:42:38,070 --> 00:42:41,700
is not evolving
open-endedly?

1043
00:42:41,700 --> 00:42:44,489
And so they're going
to say, you know,

1044
00:42:44,489 --> 00:42:46,665
well, you know,

1045
00:42:46,665 --> 00:42:48,855
such a system must
have changed.

1046
00:42:48,855 --> 00:42:51,990
Potential, must have
novelty, potential,

1047
00:42:51,990 --> 00:42:54,884
complexity, potential
ecological potential,

1048
00:42:54,884 --> 00:42:57,074
transition potential.

1049
00:42:57,074 --> 00:42:59,730
They organize them
into a hierarchy

1050
00:42:59,730 --> 00:43:00,929
and say that, you know,

1051
00:43:00,929 --> 00:43:03,030
novelty potential
obviously requires

1052
00:43:03,030 --> 00:43:04,290
change potential
because you

1053
00:43:04,290 --> 00:43:05,460
can't have something new,

1054
00:43:05,460 --> 00:43:07,650
you can't add
something different.

1055
00:43:07,650 --> 00:43:09,509
They think complexity

1056
00:43:09,509 --> 00:43:10,800
and ecological potential

1057
00:43:10,800 --> 00:43:14,234
also require
novelty potential.

1058
00:43:14,234 --> 00:43:15,885
For similar reasons.

1059
00:43:15,885 --> 00:43:17,924
Transition potential

1060
00:43:17,924 --> 00:43:20,729
obviously requires
novelty potential.

1061
00:43:20,729 --> 00:43:22,260
They say they're
not sure about

1062
00:43:22,260 --> 00:43:23,910
the relationship
between complexity

1063
00:43:23,910 --> 00:43:25,319
of ecological potential

1064
00:43:25,319 --> 00:43:28,110
without doing
empirical work.

1065
00:43:28,110 --> 00:43:31,439
And similarly at
these dotted lines

1066
00:43:31,439 --> 00:43:33,029
here are saying
they're not sure about

1067
00:43:33,029 --> 00:43:34,994
the relationship
between complexity

1068
00:43:34,994 --> 00:43:36,209
and ecological potential

1069
00:43:36,209 --> 00:43:37,365
and transition potential.

1070
00:43:37,365 --> 00:43:38,910
They say, oh,
they, they might

1071
00:43:38,910 --> 00:43:42,270
be important for
transitions,

1072
00:43:42,270 --> 00:43:44,624
but they might not be.
We'll have to see.

1073
00:43:44,624 --> 00:43:46,530
And so the measures that

1074
00:43:46,530 --> 00:43:48,075
they actually give, they,

1075
00:43:48,075 --> 00:43:50,624
they only have metrics for

1076
00:43:50,624 --> 00:43:55,185
the first four of
their properties.

1077
00:43:55,185 --> 00:43:57,630
And these are actually
fairly simple metrics,

1078
00:43:57,630 --> 00:43:59,669
but it's designed
that way so

1079
00:43:59,669 --> 00:44:02,774
that they can be
applicable widely.

1080
00:44:02,774 --> 00:44:04,919
They say they're
going to assume that

1081
00:44:04,919 --> 00:44:06,750
the components have
been filtered in

1082
00:44:06,750 --> 00:44:10,530
the same way that
an activity x or

1083
00:44:10,530 --> 00:44:13,079
filtering out neutral

1084
00:44:13,079 --> 00:44:15,195
or maladaptive components.

1085
00:44:15,195 --> 00:44:16,425
They want to say
we want to make

1086
00:44:16,425 --> 00:44:17,804
sure that we
have filtered to

1087
00:44:17,804 --> 00:44:22,935
only look at the
adaptive components.

1088
00:44:22,935 --> 00:44:25,275
And then they say
once we've done that,

1089
00:44:25,275 --> 00:44:27,959
we're going to
look at change.

1090
00:44:27,959 --> 00:44:31,020
Our measure of change
is just the number of

1091
00:44:31,020 --> 00:44:33,179
components in a current
time point which

1092
00:44:33,179 --> 00:44:35,879
were not present in
a prior time point.

1093
00:44:35,879 --> 00:44:37,590
They're going
to say novelty

1094
00:44:37,590 --> 00:44:39,144
is the number of
components in

1095
00:44:39,144 --> 00:44:40,559
a current time
point which were

1096
00:44:40,559 --> 00:44:42,299
not present in any
previous time.

1097
00:44:42,299 --> 00:44:43,859
Funny, they're going to

1098
00:44:43,859 --> 00:44:46,139
say that complexity is

1099
00:44:46,139 --> 00:44:48,749
the highest
observed count of

1100
00:44:48,749 --> 00:44:50,564
meaningful genome sites

1101
00:44:50,564 --> 00:44:52,485
across all components.

1102
00:44:52,485 --> 00:44:54,299
And I actually
have a pretty

1103
00:44:54,299 --> 00:44:55,890
interesting
discussion about what

1104
00:44:55,890 --> 00:44:59,865
constitutes meaningful
genome sites.

1105
00:44:59,865 --> 00:45:03,629
But once they've decided

1106
00:45:03,629 --> 00:45:05,430
what constitutes a
meaningful geno site,

1107
00:45:05,430 --> 00:45:07,560
then complexity is
just measured by

1108
00:45:07,560 --> 00:45:09,390
the individual
component that

1109
00:45:09,390 --> 00:45:11,235
has the most of them.

1110
00:45:11,235 --> 00:45:13,170
So kind of, again,

1111
00:45:13,170 --> 00:45:17,639
a simple but ready
measure of complexity.

1112
00:45:17,639 --> 00:45:21,104
And then for their
ecological measure,

1113
00:45:21,104 --> 00:45:23,879
they actually use the
Shannon entropy of

1114
00:45:23,879 --> 00:45:26,040
the frequency
distribution of

1115
00:45:26,040 --> 00:45:28,229
components at
a time point.

1116
00:45:28,229 --> 00:45:31,575
And this is where I
will make a comment

1117
00:45:31,575 --> 00:45:34,439
that Shannon entropy is

1118
00:45:34,439 --> 00:45:36,360
good measure of diversity,

1119
00:45:36,360 --> 00:45:37,380
but I'm not sure it's

1120
00:45:37,380 --> 00:45:40,319
a good measure of
ecological diversity.

1121
00:45:40,319 --> 00:45:42,405
I don't actually know
that I agree with that.

1122
00:45:42,405 --> 00:45:45,015
Bet. You know,

1123
00:45:45,015 --> 00:45:47,024
maximum diversity
in it, in an,

1124
00:45:47,024 --> 00:45:48,449
in an ecology
means that you

1125
00:45:48,449 --> 00:45:49,950
have a variety
of types that

1126
00:45:49,950 --> 00:45:51,420
are each equally

1127
00:45:51,420 --> 00:45:53,445
represented in
the population.

1128
00:45:53,445 --> 00:45:55,110
I think, you know,

1129
00:45:55,110 --> 00:45:56,490
real ecologies

1130
00:45:56,490 --> 00:46:00,255
feature significant
differences

1131
00:46:00,255 --> 00:46:02,399
in the frequencies
that you'd

1132
00:46:02,399 --> 00:46:04,050
expect different types of

1133
00:46:04,050 --> 00:46:05,220
things to show up, right?

1134
00:46:05,220 --> 00:46:06,600
You should expect a

1135
00:46:06,600 --> 00:46:09,659
lot more primary producers

1136
00:46:09,659 --> 00:46:12,224
than consumers,

1137
00:46:12,224 --> 00:46:14,940
and a lot more
primary consumers.

1138
00:46:14,940 --> 00:46:16,349
Secondary consumers.

1139
00:46:16,349 --> 00:46:18,284
All right? This sort of

1140
00:46:18,284 --> 00:46:20,475
way you build
a food chain.

1141
00:46:20,475 --> 00:46:22,679
But that's neither here

1142
00:46:22,679 --> 00:46:24,944
nor there. I will
write them about it.

1143
00:46:24,944 --> 00:46:26,129
Okay?

1144
00:46:26,129 --> 00:46:27,869
Then we're going
to have a set

1145
00:46:27,869 --> 00:46:29,280
of several sets of

1146
00:46:29,280 --> 00:46:31,080
hypothesise requirements

1147
00:46:31,080 --> 00:46:33,240
for open-ended evolution.

1148
00:46:33,240 --> 00:46:36,510
Back to bonds off at
how they're going to

1149
00:46:36,510 --> 00:46:38,039
use their concept of

1150
00:46:38,039 --> 00:46:39,674
models, that meta models.

1151
00:46:39,674 --> 00:46:41,909
And they've given
us this scheme

1152
00:46:41,909 --> 00:46:43,799
for categorizing novelty.

1153
00:46:43,799 --> 00:46:46,274
Again, variation

1154
00:46:46,274 --> 00:46:48,734
is within a model
state-space.

1155
00:46:48,734 --> 00:46:50,699
Innovation extends
the model space,

1156
00:46:50,699 --> 00:46:53,369
state-space and
emergence requires

1157
00:46:53,369 --> 00:46:54,899
a change to the
Meta model.

1158
00:46:54,899 --> 00:46:58,304
So what they say is that
an open-ended event,

1159
00:46:58,304 --> 00:47:01,169
one that results
in innovation or

1160
00:47:01,169 --> 00:47:02,879
emergence in

1161
00:47:02,879 --> 00:47:05,145
their novelty
categorization scheme.

1162
00:47:05,145 --> 00:47:06,239
And so an open-ended

1163
00:47:06,239 --> 00:47:07,890
system must be capable of

1164
00:47:07,890 --> 00:47:09,509
the continual
production of

1165
00:47:09,509 --> 00:47:13,455
open-ended events.
Pretty straightforward.

1166
00:47:13,455 --> 00:47:16,590
Sorrows and
Stanley layout,

1167
00:47:16,590 --> 00:47:18,510
a separate set
of requirements.

1168
00:47:18,510 --> 00:47:20,534
They talk about
a non-trivial,

1169
00:47:20,534 --> 00:47:23,504
minimal criterion
for reproduction.

1170
00:47:23,504 --> 00:47:29,519
And this is part
of Stanley coming

1171
00:47:29,519 --> 00:47:32,924
from more of a computer
science background

1172
00:47:32,924 --> 00:47:34,649
interested in this idea

1173
00:47:34,649 --> 00:47:36,554
of novelty search, right?

1174
00:47:36,554 --> 00:47:39,270
Again, as a solution
to that sort

1175
00:47:39,270 --> 00:47:44,405
of problem in
artificial evolution,

1176
00:47:44,405 --> 00:47:46,969
where you get
solutions that

1177
00:47:46,969 --> 00:47:50,420
converge or you know,
prematurely converge.

1178
00:47:50,420 --> 00:47:53,420
So you get this idea
of novelty search,

1179
00:47:53,420 --> 00:47:56,555
which tries to directly
reward novelty.

1180
00:47:56,555 --> 00:47:58,594
And then this idea of

1181
00:47:58,594 --> 00:48:03,650
minimal criteria
in evolution is,

1182
00:48:03,650 --> 00:48:05,269
is an extension
of that work.

1183
00:48:05,269 --> 00:48:09,364
But it's basically
you just,

1184
00:48:09,364 --> 00:48:11,254
you want to give
them something,

1185
00:48:11,254 --> 00:48:13,010
some simple barrier
that they have

1186
00:48:13,010 --> 00:48:15,244
to overcome in
order to reproduce.

1187
00:48:15,244 --> 00:48:17,029
But only that, right?

1188
00:48:17,029 --> 00:48:18,589
So you're not,
you're not saying

1189
00:48:18,589 --> 00:48:20,719
that they're more
or less fit.

1190
00:48:20,719 --> 00:48:24,809
You're not selecting
the top half

1191
00:48:24,809 --> 00:48:26,969
or the top 20 percent
of a population,

1192
00:48:26,969 --> 00:48:30,600
you're giving them a
bar to jump over rather

1193
00:48:30,600 --> 00:48:33,600
than selecting the winners

1194
00:48:33,600 --> 00:48:34,725
of a competition
basically.

1195
00:48:34,725 --> 00:48:36,120
So everybody who can cross

1196
00:48:36,120 --> 00:48:38,325
the bar gets to live.

1197
00:48:38,325 --> 00:48:40,289
But the, but the
idea is that

1198
00:48:40,289 --> 00:48:42,600
the bar has to
be nontrivial.

1199
00:48:42,600 --> 00:48:44,835
Okay? So then
we're going to say

1200
00:48:44,835 --> 00:48:46,769
novel individuals,

1201
00:48:46,769 --> 00:48:49,485
whenever they emerge,
need to create

1202
00:48:49,485 --> 00:48:50,609
novel opportunities to

1203
00:48:50,609 --> 00:48:52,724
satisfy the
minimal criteria.

1204
00:48:52,724 --> 00:48:55,589
And then they say
that the agents

1205
00:48:55,589 --> 00:48:57,300
themselves must choose how

1206
00:48:57,300 --> 00:48:58,934
and where to interact
with the world.

1207
00:48:58,934 --> 00:49:00,660
And that the
potential size and

1208
00:49:00,660 --> 00:49:01,350
complexity at

1209
00:49:01,350 --> 00:49:03,970
phenotypes needs
to be unbounded.

1210
00:49:04,810 --> 00:49:07,909
Taylor has a related set.

1211
00:49:07,909 --> 00:49:11,030
Requirements. Basically
starts out by

1212
00:49:11,030 --> 00:49:12,740
arguing there should be

1213
00:49:12,740 --> 00:49:14,884
unlimited phenotype space,

1214
00:49:14,884 --> 00:49:16,519
mutational pathways

1215
00:49:16,519 --> 00:49:18,515
between those
potential phenotypes

1216
00:49:18,515 --> 00:49:21,214
and changing
adaptive landscapes.

1217
00:49:21,214 --> 00:49:23,555
And then he goes
on to note that

1218
00:49:23,555 --> 00:49:26,240
the complexity of the
physical environment

1219
00:49:26,240 --> 00:49:27,844
has a lot to do with that.

1220
00:49:27,844 --> 00:49:31,070
And whether the organism

1221
00:49:31,070 --> 00:49:32,674
is embedded in
the environment.

1222
00:49:32,674 --> 00:49:34,549
Which again is a
pretty rich discussion

1223
00:49:34,549 --> 00:49:36,874
that I seem to matter
of time to go into.

1224
00:49:36,874 --> 00:49:42,469
And he refines those
requirements later into

1225
00:49:42,469 --> 00:49:45,319
five new
requirements which

1226
00:49:45,319 --> 00:49:48,070
are robustly reproductive
individuals,

1227
00:49:48,070 --> 00:49:51,840
meaning, you know,
robust against

1228
00:49:51,840 --> 00:49:55,080
mutation and

1229
00:49:55,080 --> 00:49:56,969
disturbance by
other individuals.

1230
00:49:56,969 --> 00:49:59,909
A medium allowing
the possibility of

1231
00:49:59,909 --> 00:50:01,665
unlimited type

1232
00:50:01,665 --> 00:50:03,840
Tokyo Type 1
open-end abolition

1233
00:50:03,840 --> 00:50:05,415
at various levels
of complexity.

1234
00:50:05,415 --> 00:50:06,989
Individuals that
can produce

1235
00:50:06,989 --> 00:50:08,580
more complex offspring,

1236
00:50:08,580 --> 00:50:10,560
mutational pathways
between by

1237
00:50:10,560 --> 00:50:11,669
all individuals and

1238
00:50:11,669 --> 00:50:13,785
a driver for
continued evolution.

1239
00:50:13,785 --> 00:50:15,419
And then, oh good,

1240
00:50:15,419 --> 00:50:17,415
That's my last
one of those.

1241
00:50:17,415 --> 00:50:19,004
Okay.

1242
00:50:19,004 --> 00:50:22,589
So these requirements
are laid

1243
00:50:22,589 --> 00:50:26,970
out as, again,
hypotheticals.

1244
00:50:26,970 --> 00:50:29,190
This is what should
be necessary and

1245
00:50:29,190 --> 00:50:32,480
sufficient for a model

1246
00:50:32,480 --> 00:50:34,999
of open-ended evolution
to actually produce

1247
00:50:34,999 --> 00:50:37,504
open-ended
evolution and pass

1248
00:50:37,504 --> 00:50:40,219
that O's and, and
others tests.

1249
00:50:40,219 --> 00:50:45,365
Okay, so let's talk about
some famous models.

1250
00:50:45,365 --> 00:50:47,254
Um, and first, we're
going to talk about

1251
00:50:47,254 --> 00:50:50,375
Tara is probably
the most famous.

1252
00:50:50,375 --> 00:50:52,609
It's arguably
the original,

1253
00:50:52,609 --> 00:50:55,039
widely known model that

1254
00:50:55,039 --> 00:50:57,920
aimed for open-ended
evolutionary dynamics.

1255
00:50:57,920 --> 00:51:00,409
And it's important because

1256
00:51:00,409 --> 00:51:04,550
unlike evolutionary
algorithms of the period,

1257
00:51:04,550 --> 00:51:07,324
it doesn't have an
explicit fitness function.

1258
00:51:07,324 --> 00:51:11,240
Tara is a virtual
machine executing

1259
00:51:11,240 --> 00:51:12,890
a custom machine language

1260
00:51:12,890 --> 00:51:15,405
designed to be
robust to mutation.

1261
00:51:15,405 --> 00:51:17,219
It's organisms are

1262
00:51:17,219 --> 00:51:19,635
basically simple
self-replicating programs,

1263
00:51:19,635 --> 00:51:20,879
and they are effectively

1264
00:51:20,879 --> 00:51:22,170
competing for address

1265
00:51:22,170 --> 00:51:25,484
space in the memory of
that virtual machine.

1266
00:51:25,484 --> 00:51:28,169
But they have kind

1267
00:51:28,169 --> 00:51:29,850
of limited potential
to interact with each

1268
00:51:29,850 --> 00:51:31,590
other beyond being able

1269
00:51:31,590 --> 00:51:37,245
to point at each other's
code and borrowing.

1270
00:51:37,245 --> 00:51:38,910
This is how you
get the emergence

1271
00:51:38,910 --> 00:51:40,740
of parasitism,

1272
00:51:40,740 --> 00:51:43,410
where one program will use

1273
00:51:43,410 --> 00:51:45,329
the copy loop of
another program

1274
00:51:45,329 --> 00:51:47,324
instead of having
its own copy loop.

1275
00:51:47,324 --> 00:51:50,639
And so then it can fit
into a smaller space

1276
00:51:50,639 --> 00:51:52,079
and copy itself more

1277
00:51:52,079 --> 00:51:55,239
widely because of
it being smaller.

1278
00:51:55,239 --> 00:51:58,069
Some interesting
stuff appears here,

1279
00:51:58,069 --> 00:52:00,170
but it ultimately
does not produce

1280
00:52:00,170 --> 00:52:02,119
the unbounded
evolutionary dynamics

1281
00:52:02,119 --> 00:52:03,950
that he was looking for.

1282
00:52:03,950 --> 00:52:06,380
And so then you
get a vitae,

1283
00:52:06,380 --> 00:52:09,454
which is very
much like tiara,

1284
00:52:09,454 --> 00:52:12,035
featuring self-replicating
computer programs

1285
00:52:12,035 --> 00:52:13,580
and a virtual machine.

1286
00:52:13,580 --> 00:52:16,594
But each of these run
in their own machine.

1287
00:52:16,594 --> 00:52:18,290
They're not competing
directly for

1288
00:52:18,290 --> 00:52:20,480
memory or memory
or CPU cycles.

1289
00:52:20,480 --> 00:52:22,309
They can run at
different speeds.

1290
00:52:22,309 --> 00:52:24,994
Basically they
are performing

1291
00:52:24,994 --> 00:52:28,025
can't basic logic
computations on

1292
00:52:28,025 --> 00:52:30,769
numbers in the
environments where

1293
00:52:30,769 --> 00:52:34,025
they absorb resources
and earn merit,

1294
00:52:34,025 --> 00:52:38,049
which can give them
more CPU timers.

1295
00:52:38,049 --> 00:52:39,660
Their CPU cycles.

1296
00:52:39,660 --> 00:52:41,220
They don't
interact directly

1297
00:52:41,220 --> 00:52:42,689
with other organisms,

1298
00:52:42,689 --> 00:52:45,989
but by reproducing,

1299
00:52:45,989 --> 00:52:47,039
they can kill

1300
00:52:47,039 --> 00:52:49,694
their neighbors when
they reproduce.

1301
00:52:49,694 --> 00:52:53,235
Copy themselves into
the sites that they're,

1302
00:52:53,235 --> 00:52:55,440
that they're
neighbor occupies.

1303
00:52:55,440 --> 00:52:58,259
And there are both

1304
00:52:58,259 --> 00:53:00,509
two-dimensional sort
of grid topologies are

1305
00:53:00,509 --> 00:53:02,550
the Beta and also

1306
00:53:02,550 --> 00:53:04,440
everything is
a neighbor of

1307
00:53:04,440 --> 00:53:06,720
everything topologies
that are Beta.

1308
00:53:06,720 --> 00:53:11,399
I'm a vita is very
detailed and very complex.

1309
00:53:11,399 --> 00:53:13,409
It's in its fourth
major version.

1310
00:53:13,409 --> 00:53:18,090
Now, there is a
simplified a b to add,

1311
00:53:18,090 --> 00:53:19,470
which is for
instructional use

1312
00:53:19,470 --> 00:53:21,240
at the high school
or undergrad level.

1313
00:53:21,240 --> 00:53:22,995
It's still under
active development,

1314
00:53:22,995 --> 00:53:25,379
which makes it,
to my knowledge,

1315
00:53:25,379 --> 00:53:27,089
the, the longest running

1316
00:53:27,089 --> 00:53:30,030
single model of
open-ended evolution.

1317
00:53:30,030 --> 00:53:31,770
And that, you
know, a number

1318
00:53:31,770 --> 00:53:33,584
of people are
still using it.

1319
00:53:33,584 --> 00:53:35,850
They're still
working on it.

1320
00:53:35,850 --> 00:53:39,119
Africa at Michigan
State and

1321
00:53:39,119 --> 00:53:40,590
the digital evolution lab

1322
00:53:40,590 --> 00:53:43,800
is continuing to
work on this model,

1323
00:53:43,800 --> 00:53:46,515
continuing to
improve it. Okay.

1324
00:53:46,515 --> 00:53:48,855
Next we have echo.

1325
00:53:48,855 --> 00:53:50,579
Echo is very
near and dear to

1326
00:53:50,579 --> 00:53:52,139
my heart because
it's focused

1327
00:53:52,139 --> 00:53:56,100
on the interactions
between agents.

1328
00:53:56,100 --> 00:54:00,015
And it's a tornado
toroidal lattice sites.

1329
00:54:00,015 --> 00:54:02,025
It has resource fountains.

1330
00:54:02,025 --> 00:54:05,670
Echo agents are
chromosomes as substrings.

1331
00:54:05,670 --> 00:54:08,370
They're organized
into these tags.

1332
00:54:08,370 --> 00:54:10,950
And then it's this kind

1333
00:54:10,950 --> 00:54:12,539
of tag matching process

1334
00:54:12,539 --> 00:54:15,030
decides which agents can

1335
00:54:15,030 --> 00:54:16,439
interact with
which other agents

1336
00:54:16,439 --> 00:54:18,555
and what happens
when they do.

1337
00:54:18,555 --> 00:54:19,949
You know, sort of

1338
00:54:19,949 --> 00:54:21,330
limited set of
interactions,

1339
00:54:21,330 --> 00:54:23,475
trade meeting and combat,

1340
00:54:23,475 --> 00:54:25,860
and then accumulating

1341
00:54:25,860 --> 00:54:27,270
environmental resources.

1342
00:54:27,270 --> 00:54:29,070
But echo unfortunately has

1343
00:54:29,070 --> 00:54:31,330
also not been bound
to demonstrate.

1344
00:54:31,330 --> 00:54:34,400
Open-ended
evolution. I'm going

1345
00:54:34,400 --> 00:54:36,350
to move on and talk
about poly world,

1346
00:54:36,350 --> 00:54:38,359
which is, it exists

1347
00:54:38,359 --> 00:54:39,935
in a graphical
environment,

1348
00:54:39,935 --> 00:54:41,884
which was cool and new.

1349
00:54:41,884 --> 00:54:44,059
You've got trap
it, a trapezoid

1350
00:54:44,059 --> 00:54:45,199
digital organisms.

1351
00:54:45,199 --> 00:54:46,985
They're moving around
in a flat plane.

1352
00:54:46,985 --> 00:54:48,349
They're foraging for food,

1353
00:54:48,349 --> 00:54:50,104
they're metabolizing
energy.

1354
00:54:50,104 --> 00:54:51,739
They have vision because

1355
00:54:51,739 --> 00:54:53,419
they're in a graphical
environment.

1356
00:54:53,419 --> 00:54:55,594
And so they're
controlled by IES,

1357
00:54:55,594 --> 00:54:58,369
artificial neural networks
which select from

1358
00:54:58,369 --> 00:55:00,214
some predefined
set of behaviors

1359
00:55:00,214 --> 00:55:02,539
using that input from
that vision system.

1360
00:55:02,539 --> 00:55:05,059
But again, they have a
pretty small number of

1361
00:55:05,059 --> 00:55:06,785
predefined Egypt
properties

1362
00:55:06,785 --> 00:55:08,285
and behavior primitives.

1363
00:55:08,285 --> 00:55:11,285
So the potential for

1364
00:55:11,285 --> 00:55:16,130
the phenotype
spaces. Okay.

1365
00:55:16,130 --> 00:55:17,825
Then we've got Geb,

1366
00:55:17,825 --> 00:55:19,610
which I already
pointed out was the

1367
00:55:19,610 --> 00:55:23,240
first to pass
Bedouin test.

1368
00:55:23,240 --> 00:55:25,430
It's another
population of agents

1369
00:55:25,430 --> 00:55:27,274
in a 2D space,

1370
00:55:27,274 --> 00:55:30,125
small number of
pre-defined behaviors.

1371
00:55:30,125 --> 00:55:32,210
Also behavior
is controlled

1372
00:55:32,210 --> 00:55:34,340
by artificial
neural networks.

1373
00:55:34,340 --> 00:55:35,779
But unlike Polly world,

1374
00:55:35,779 --> 00:55:36,890
there's no energy and

1375
00:55:36,890 --> 00:55:39,289
GAD agents are competing

1376
00:55:39,289 --> 00:55:42,155
for space on a
spatial grid.

1377
00:55:42,155 --> 00:55:45,350
As I said, it is first

1378
00:55:45,350 --> 00:55:46,879
a light system
that demonstrates

1379
00:55:46,879 --> 00:55:48,530
that class 3 behavior.

1380
00:55:48,530 --> 00:55:50,854
So good for Tannen.

1381
00:55:50,854 --> 00:55:52,909
Okay, what's next?

1382
00:55:52,909 --> 00:55:55,009
Wrapping it up.
Talked about

1383
00:55:55,009 --> 00:55:56,509
a lot of existing
models that

1384
00:55:56,509 --> 00:55:58,500
measure of
open-ended evolution

1385
00:55:58,500 --> 00:56:00,449
and some hypothesize
requirements.

1386
00:56:00,449 --> 00:56:01,830
And he accepted behavioral

1387
00:56:01,830 --> 00:56:04,679
hallmarks currently.

1388
00:56:04,679 --> 00:56:05,909
And this is I'm just going

1389
00:56:05,909 --> 00:56:07,770
to borrow from better.

1390
00:56:07,770 --> 00:56:10,619
Lewis appeared.

1391
00:56:10,619 --> 00:56:12,150
You've, you've
heard his name.

1392
00:56:12,150 --> 00:56:13,739
He's an important figure.

1393
00:56:13,739 --> 00:56:17,099
But trying to
say it's sort of

1394
00:56:17,099 --> 00:56:19,410
time for open-ended
evolution to

1395
00:56:19,410 --> 00:56:22,259
coalesce into a
proper discipline.

1396
00:56:22,259 --> 00:56:25,170
And in order to do that,
he's talking about,

1397
00:56:25,170 --> 00:56:28,965
you know, this Kuhnian
paradigms idea.

1398
00:56:28,965 --> 00:56:30,479
It needs to move into

1399
00:56:30,479 --> 00:56:32,280
a normal research program.

1400
00:56:32,280 --> 00:56:35,009
And so the way for
it to do that is

1401
00:56:35,009 --> 00:56:38,595
to collect around a set
of exemplary models.

1402
00:56:38,595 --> 00:56:41,040
And, and so
then, you know,

1403
00:56:41,040 --> 00:56:42,659
step one of this
is going to

1404
00:56:42,659 --> 00:56:44,880
be identify those
exemplary models and

1405
00:56:44,880 --> 00:56:47,699
measures and stop
the proliferation

1406
00:56:47,699 --> 00:56:49,335
of new models and
new measures.

1407
00:56:49,335 --> 00:56:53,819
I know Dr. likes to
make the joke about

1408
00:56:53,819 --> 00:56:56,310
academics would rather use

1409
00:56:56,310 --> 00:56:58,109
another academics
toothbrush

1410
00:56:58,109 --> 00:56:59,429
than their terminology.

1411
00:56:59,429 --> 00:57:01,394
But it's a
problem that we,

1412
00:57:01,394 --> 00:57:03,105
you know, we have to get,

1413
00:57:03,105 --> 00:57:05,100
we have to get
past because

1414
00:57:05,100 --> 00:57:08,324
start agreeing on
some core concepts.

1415
00:57:08,324 --> 00:57:11,970
So once all those
exemplary measures

1416
00:57:11,970 --> 00:57:13,530
and models are
identified that we

1417
00:57:13,530 --> 00:57:14,670
want to apply all of

1418
00:57:14,670 --> 00:57:17,580
those measures to all
of those models and

1419
00:57:17,580 --> 00:57:22,529
get a sort of
grid of output.

1420
00:57:22,529 --> 00:57:24,060
You know, what, how

1421
00:57:24,060 --> 00:57:26,159
can we compare these
things to each other?

1422
00:57:26,159 --> 00:57:28,679
We want to do
that to produce

1423
00:57:28,679 --> 00:57:31,199
a review paper of
open-ended evolution,

1424
00:57:31,199 --> 00:57:32,670
produce a tutorial on

1425
00:57:32,670 --> 00:57:34,290
open-end build
evolution and

1426
00:57:34,290 --> 00:57:36,359
eventually produce
a textbook

1427
00:57:36,359 --> 00:57:38,100
on it, open and evolution.

1428
00:57:38,100 --> 00:57:41,250
In order to get

1429
00:57:41,250 --> 00:57:43,665
this to be a, a
mature discipline.

1430
00:57:43,665 --> 00:57:47,159
Okay. So that's that,

1431
00:57:47,159 --> 00:57:48,405
that, that
concludes my talk.

1432
00:57:48,405 --> 00:57:50,025
I know I'm totally
out of time,

1433
00:57:50,025 --> 00:57:51,989
but if anybody wants
to stick around,

1434
00:57:51,989 --> 00:57:55,360
I'm here for
questions, comments.

1435
00:57:59,150 --> 00:58:01,605
I'm going to ask
you to stop sharing

1436
00:58:01,605 --> 00:58:03,240
just so we can see
each other again.

1437
00:58:03,240 --> 00:58:04,020
Okay.

1438
00:58:04,020 --> 00:58:08,100
And if we have to
reshare, that'll be fine.

1439
00:58:08,100 --> 00:58:09,584
But I do want to say

1440
00:58:09,584 --> 00:58:11,970
that I'm torn
about keeping

1441
00:58:11,970 --> 00:58:14,249
the mind the recorder
going because

1442
00:58:14,249 --> 00:58:15,390
some people might
have to leave

1443
00:58:15,390 --> 00:58:16,679
and want to know
what the Q and a.

1444
00:58:16,679 --> 00:58:18,480
It looks like. On
the other hand,

1445
00:58:18,480 --> 00:58:20,440
you don't really want,

1446
00:58:20,440 --> 00:58:22,714
I want to try
and experiment.

1447
00:58:22,714 --> 00:58:25,340
I want to do a poll.

1448
00:58:25,340 --> 00:58:26,690
And I want you
to do is use

1449
00:58:26,690 --> 00:58:28,820
the reactions on
the bottom right of

1450
00:58:28,820 --> 00:58:31,204
your screen to go

1451
00:58:31,204 --> 00:58:33,634
check if you think we
should keep it going.

1452
00:58:33,634 --> 00:58:34,909
And actually if
you think we

1453
00:58:34,909 --> 00:58:36,199
should stop the
recording and

1454
00:58:36,199 --> 00:58:37,985
make this beyond
recording,

1455
00:58:37,985 --> 00:58:39,124
I will demonstrate.

1456
00:58:39,124 --> 00:58:40,580
I'll hit my there we go.

1457
00:58:40,580 --> 00:58:41,540
You guys all understood.

1458
00:58:41,540 --> 00:58:41,929
Just fine.

1459
00:58:41,929 --> 00:58:44,609
I'm going to take
my vote back off.

1460
00:58:44,980 --> 00:58:47,720
I'm seeing all brain

1461
00:58:47,720 --> 00:58:49,160
for those who
bothered to vote,

1462
00:58:49,160 --> 00:58:50,585
so that means keep
the recording going.

1463
00:58:50,585 --> 00:58:51,529
Thank you.

1464
00:58:51,529 --> 00:58:54,329
Let's take questions now.

1465
00:58:57,040 --> 00:58:59,820
I know it was a lot.

1466
00:59:00,910 --> 00:59:03,489
That's an understatement.

1467
00:59:03,489 --> 00:59:05,820
I'll start with a comment.

1468
00:59:05,820 --> 00:59:09,570
I'm realizing that I
ended the talk by saying

1469
00:59:09,570 --> 00:59:10,860
we should stop producing

1470
00:59:10,860 --> 00:59:13,319
new models and
new measures.

1471
00:59:13,319 --> 00:59:14,880
I do hope to present

1472
00:59:14,880 --> 00:59:17,879
my own new model the
next time we do this.

1473
00:59:17,879 --> 00:59:22,110
So I have 1,

1474
00:59:22,110 --> 00:59:23,940
2, 3, 4, 5, 6, 7, 8,

1475
00:59:23,940 --> 00:59:26,160
9, 10 questions, comments.

1476
00:59:26,160 --> 00:59:27,884
I probably won't
use any of them.

1477
00:59:27,884 --> 00:59:30,270
But it certainly was a
great talk in that it

1478
00:59:30,270 --> 00:59:32,084
generates more
questions and

1479
00:59:32,084 --> 00:59:33,870
no doubt possibly answer.

1480
00:59:33,870 --> 00:59:36,959
Okay, Well, thanks.

1481
00:59:36,959 --> 00:59:41,339
First question. Well, I

1482
00:59:41,339 --> 00:59:43,799
was wondering if if
you think of hard,

1483
00:59:43,799 --> 00:59:46,134
the hard and soft
distinction.

1484
00:59:46,134 --> 00:59:48,230
Let's talk about healing

1485
00:59:48,230 --> 00:59:49,535
the problem with robots.

1486
00:59:49,535 --> 00:59:51,049
Robots don't
heal themselves,

1487
00:59:51,049 --> 00:59:54,950
and humans and creatures
in real life do it.

1488
00:59:54,950 --> 00:59:56,179
Whereas in software,

1489
00:59:56,179 --> 00:59:57,560
you can sort
of a batch and

1490
00:59:57,560 --> 01:00:02,000
maybe play with code
that can heal itself.

1491
01:00:02,000 --> 01:00:03,320
So I just wondered this is

1492
01:00:03,320 --> 01:00:04,489
the concept of healing

1493
01:00:04,489 --> 01:00:07,759
important that all
to open-endedness.

1494
01:00:07,759 --> 01:00:10,910
So I actually
think a vita in

1495
01:00:10,910 --> 01:00:14,750
particular is
designed in a way

1496
01:00:14,750 --> 01:00:18,230
that allows
those organisms

1497
01:00:18,230 --> 01:00:24,065
to kinda robustly
repair themselves.

1498
01:00:24,065 --> 01:00:26,780
And I actually want
to say gab also.

1499
01:00:26,780 --> 01:00:30,679
It's in, in the way
that it's designed.

1500
01:00:30,679 --> 01:00:32,944
It's neural can, trollers.

1501
01:00:32,944 --> 01:00:34,579
You could have this.

1502
01:00:34,579 --> 01:00:36,260
I don't know that
they actually

1503
01:00:36,260 --> 01:00:39,005
get damaged

1504
01:00:39,005 --> 01:00:41,329
from interaction with
the environment.

1505
01:00:41,329 --> 01:00:43,190
But there's at
least some sense

1506
01:00:43,190 --> 01:00:46,280
of how are they trying to

1507
01:00:46,280 --> 01:00:50,209
preserve their own
genetic identity

1508
01:00:50,209 --> 01:00:51,709
in the face of mutation?

1509
01:00:51,709 --> 01:00:53,809
I mean, that, that
sort of this idea

1510
01:00:53,809 --> 01:00:55,040
of evolution,

1511
01:00:55,040 --> 01:00:56,555
of evolvability

1512
01:00:56,555 --> 01:00:58,099
at least gets at
some of that.

1513
01:00:58,099 --> 01:00:59,119
And so like I say,

1514
01:00:59,119 --> 01:01:00,770
I don't know that
that's really feeling.

1515
01:01:00,770 --> 01:01:03,335
But this question, again,

1516
01:01:03,335 --> 01:01:07,129
how do I maintain my
identity in the face

1517
01:01:07,129 --> 01:01:11,060
of perturbations from
the environment or,

1518
01:01:11,060 --> 01:01:13,274
you know, mutations

1519
01:01:13,274 --> 01:01:14,760
that occur during
reproduction.

1520
01:01:14,760 --> 01:01:17,729
That's, that's a core
system science idea.

1521
01:01:17,729 --> 01:01:19,875
I think it does appear

1522
01:01:19,875 --> 01:01:24,420
in an open-end evolution.

1523
01:01:24,420 --> 01:01:26,789
But yeah, I don't
know that anybody

1524
01:01:26,789 --> 01:01:29,174
has, is looking.

1525
01:01:29,174 --> 01:01:31,830
I don't know that I've
seen a model that,

1526
01:01:31,830 --> 01:01:34,454
that really
explicitly represents

1527
01:01:34,454 --> 01:01:37,845
harm to a digital
organized.

1528
01:01:37,845 --> 01:01:40,020
Honestly, you don't see
very many models that

1529
01:01:40,020 --> 01:01:43,904
represents the development
of an organism IV.

1530
01:01:43,904 --> 01:01:49,289
They tend to be born
fully flat and you know,

1531
01:01:49,289 --> 01:01:51,419
that's, you know,
I think this

1532
01:01:51,419 --> 01:01:54,645
raises a very
basic question.

1533
01:01:54,645 --> 01:01:55,979
I think implicit in

1534
01:01:55,979 --> 01:01:58,484
this question has
a bigger question,

1535
01:01:58,484 --> 01:02:03,854
which is, these are all
very abstract models.

1536
01:02:03,854 --> 01:02:06,269
So in the back
of our minds,

1537
01:02:06,269 --> 01:02:07,410
we sort of think of

1538
01:02:07,410 --> 01:02:10,784
these models as
vaguely representing

1539
01:02:10,784 --> 01:02:12,960
biological evolution

1540
01:02:12,960 --> 01:02:15,674
and technological
evolution,

1541
01:02:15,674 --> 01:02:18,569
societal roles.

1542
01:02:18,569 --> 01:02:22,529
So we have
real-world examples

1543
01:02:22,529 --> 01:02:24,119
in the back of our mind.

1544
01:02:24,119 --> 01:02:27,000
And we have this
abstract model that is

1545
01:02:27,000 --> 01:02:29,729
related to these
real-world examples.

1546
01:02:29,729 --> 01:02:31,800
And the question is, what,

1547
01:02:31,800 --> 01:02:33,509
what do we insist that

1548
01:02:33,509 --> 01:02:35,399
the abstract model have in

1549
01:02:35,399 --> 01:02:39,035
it in order to
be reasonable?

1550
01:02:39,035 --> 01:02:41,269
And what do we forgo with?

1551
01:02:41,269 --> 01:02:43,489
So for example,
you, Shane,

1552
01:02:43,489 --> 01:02:46,594
you just said a lot
of these models,

1553
01:02:46,594 --> 01:02:48,830
these models have
no development

1554
01:02:48,830 --> 01:02:49,880
of the entity.

1555
01:02:49,880 --> 01:02:52,834
They, they just
born complete,

1556
01:02:52,834 --> 01:02:54,185
completely developed.

1557
01:02:54,185 --> 01:02:55,774
So the question is,

1558
01:02:55,774 --> 01:03:00,409
is that a reasonable
thing to not ask for in,

1559
01:03:00,409 --> 01:03:02,869
in an open-ended
evolutionary model,

1560
01:03:02,869 --> 01:03:04,100
or do we want that?

1561
01:03:04,100 --> 01:03:05,990
And Wayne basically asked,

1562
01:03:05,990 --> 01:03:08,149
do we want a
healing capacity?

1563
01:03:08,149 --> 01:03:09,799
So there are a
lot of and I

1564
01:03:09,799 --> 01:03:12,035
would ask like about echo.

1565
01:03:12,035 --> 01:03:14,689
If entities don't
interact with

1566
01:03:14,689 --> 01:03:18,745
other entities in some
direct interaction,

1567
01:03:18,745 --> 01:03:22,170
echo they do or they
do or their mother.

1568
01:03:22,170 --> 01:03:24,329
But tiara, they
don't. Eric?

1569
01:03:24,329 --> 01:03:26,085
Yeah. Yeah. Okay.

1570
01:03:26,085 --> 01:03:27,179
I don't. Okay.

1571
01:03:27,179 --> 01:03:28,770
So so I think one of

1572
01:03:28,770 --> 01:03:31,770
the one of the
ingredients in but

1573
01:03:31,770 --> 01:03:34,409
those formulation
of how to make

1574
01:03:34,409 --> 01:03:37,650
this solidify is to

1575
01:03:37,650 --> 01:03:39,540
agree upon what do

1576
01:03:39,540 --> 01:03:41,879
these models
have to contain?

1577
01:03:41,879 --> 01:03:44,429
To be legitimate models,

1578
01:03:44,429 --> 01:03:46,935
to be legitimate
candidates

1579
01:03:46,935 --> 01:03:48,584
for open-ended evolution.

1580
01:03:48,584 --> 01:03:50,594
What they have to model,

1581
01:03:50,594 --> 01:03:53,415
and what don't they
have to model?

1582
01:03:53,415 --> 01:03:57,030
Yeah. I I actually
think you know, Taylor,

1583
01:03:57,030 --> 01:04:00,720
at least to me and in
his requirements for

1584
01:04:00,720 --> 01:04:02,970
open evolution and talking

1585
01:04:02,970 --> 01:04:03,989
about this idea of

1586
01:04:03,989 --> 01:04:06,000
a complex physical
environment and,

1587
01:04:06,000 --> 01:04:07,289
and more importantly,

1588
01:04:07,289 --> 01:04:09,209
the embeddedness
of an orgasm and

1589
01:04:09,209 --> 01:04:11,190
that environment is kind

1590
01:04:11,190 --> 01:04:12,330
of getting at
this idea that

1591
01:04:12,330 --> 01:04:14,249
maybe the reason
that we can't do

1592
01:04:14,249 --> 01:04:16,230
open-end evolution
and software.

1593
01:04:16,230 --> 01:04:17,339
The way that we
can do it in

1594
01:04:17,339 --> 01:04:19,829
the real world is that
the real-world is

1595
01:04:19,829 --> 01:04:23,655
just so much
physically richer.

1596
01:04:23,655 --> 01:04:25,770
There's just more,

1597
01:04:25,770 --> 01:04:29,219
there's more regularity
to exploit, right?

1598
01:04:29,219 --> 01:04:32,400
Whereas in the software
model, you know, you,

1599
01:04:32,400 --> 01:04:34,079
you have to explicitly

1600
01:04:34,079 --> 01:04:36,329
represent everything
that's possible.

1601
01:04:36,329 --> 01:04:38,655
Um, you know, and, and,

1602
01:04:38,655 --> 01:04:39,989
and I guess you'd say the

1603
01:04:39,989 --> 01:04:41,310
joke is kind of
if you wanted to

1604
01:04:41,310 --> 01:04:43,919
really do an open-end
evolution model,

1605
01:04:43,919 --> 01:04:46,319
you'd have to start
again from scratch,

1606
01:04:46,319 --> 01:04:47,939
creating a universe with

1607
01:04:47,939 --> 01:04:50,580
physical laws and
chemical laws.

1608
01:04:50,580 --> 01:04:52,920
And then sort of a
biochemistry laid on

1609
01:04:52,920 --> 01:04:55,920
that chemicals out
set of chemical laws.

1610
01:04:55,920 --> 01:05:00,690
And then stacking up the,

1611
01:05:00,690 --> 01:05:05,279
the sciences in
order to get that,

1612
01:05:05,279 --> 01:05:06,959
that richness
that we have.

1613
01:05:06,959 --> 01:05:08,220
You know, in
the real world.

1614
01:05:08,220 --> 01:05:10,364
And actually, I didn't

1615
01:05:10,364 --> 01:05:12,524
figure out how to
fit into this talk.

1616
01:05:12,524 --> 01:05:17,280
But you've got work
by campus and gooey

1617
01:05:17,280 --> 01:05:19,079
us who are talking about

1618
01:05:19,079 --> 01:05:22,574
this idea of full body.

1619
01:05:22,574 --> 01:05:24,525
And you've got
a bond guard

1620
01:05:24,525 --> 01:05:26,100
talking about embodiment.

1621
01:05:26,100 --> 01:05:27,870
And I guess this occurs
more in the sort of

1622
01:05:27,870 --> 01:05:31,484
hired a life robotics
side of things.

1623
01:05:31,484 --> 01:05:34,710
But that the,

1624
01:05:34,710 --> 01:05:37,769
the being in a body
and that body being in

1625
01:05:37,769 --> 01:05:41,069
the world creates a lot
of opportunities for

1626
01:05:41,069 --> 01:05:45,809
novelty and behavior
and for making use of,

1627
01:05:45,809 --> 01:05:47,490
again, regularity
and environment.

1628
01:05:47,490 --> 01:05:49,890
That's just not possible

1629
01:05:49,890 --> 01:05:52,364
in the same way in
a software system.

1630
01:05:52,364 --> 01:05:57,420
Because it's
like, you know,

1631
01:05:57,420 --> 01:05:59,070
that sort of thought
experiment of

1632
01:05:59,070 --> 01:06:01,499
what is a paperclip
for, right?

1633
01:06:01,499 --> 01:06:03,269
And so you can, you know,

1634
01:06:03,269 --> 01:06:05,070
in your software system
you can say, oh,

1635
01:06:05,070 --> 01:06:08,984
the paperclip is for
connecting paper.

1636
01:06:08,984 --> 01:06:10,860
But it's difficult in,

1637
01:06:10,860 --> 01:06:12,959
in software to
say well that,

1638
01:06:12,959 --> 01:06:15,120
that paperclip
can also be,

1639
01:06:15,120 --> 01:06:16,545
you can unbend it.

1640
01:06:16,545 --> 01:06:17,790
You can, there's, I mean,

1641
01:06:17,790 --> 01:06:20,895
it has physical
properties that are

1642
01:06:20,895 --> 01:06:23,039
manipulable and
ways that you

1643
01:06:23,039 --> 01:06:25,619
might not imagine
from the beginning.

1644
01:06:25,619 --> 01:06:27,239
And I will say that

1645
01:06:27,239 --> 01:06:29,715
is at least that's
a big part of,

1646
01:06:29,715 --> 01:06:33,434
of what is guiding me.

1647
01:06:33,434 --> 01:06:34,740
And the way that
I've tried to design

1648
01:06:34,740 --> 01:06:36,569
my model is to make
something like

1649
01:06:36,569 --> 01:06:39,899
that possible to
be represented,

1650
01:06:39,899 --> 01:06:41,624
at least a very
abstract way.

1651
01:06:41,624 --> 01:06:43,710
And so again,
hopefully I get

1652
01:06:43,710 --> 01:06:44,549
to talk to you guys about

1653
01:06:44,549 --> 01:06:46,605
that sometime
in the future.

1654
01:06:46,605 --> 01:06:50,820
Soon. Soon, but not yet.

1655
01:06:50,820 --> 01:06:52,200
Well else has a question.

1656
01:06:52,200 --> 01:06:54,400
I want to give
others a chance.

1657
01:06:59,870 --> 01:07:04,019
Your your beard it
or I don't hear you.

1658
01:07:04,019 --> 01:07:05,519
We don't hear you.

1659
01:07:05,519 --> 01:07:15,155
Bronson. Can you
hear me now?

1660
01:07:15,155 --> 01:07:16,969
Yeah, I got you. I lose

1661
01:07:16,969 --> 01:07:20,089
my microphone on
my headphones.

1662
01:07:20,089 --> 01:07:21,410
But thank you.

1663
01:07:21,410 --> 01:07:23,825
Said there's a lot of
information to process,

1664
01:07:23,825 --> 01:07:24,649
but I can think of it

1665
01:07:24,649 --> 01:07:27,900
from a dynamical
perspectives.

1666
01:07:27,970 --> 01:07:31,670
And you were talking about

1667
01:07:31,670 --> 01:07:34,039
the fitness function and

1668
01:07:34,039 --> 01:07:35,300
optimization when it comes

1669
01:07:35,300 --> 01:07:37,715
to the non open-ended.

1670
01:07:37,715 --> 01:07:40,099
And certain things can
get kind of stuck and

1671
01:07:40,099 --> 01:07:42,065
basins of attraction
and whatnot.

1672
01:07:42,065 --> 01:07:44,480
This is a pretty way
with open-ended that

1673
01:07:44,480 --> 01:07:46,819
we can kind of see
that all reflect that.

1674
01:07:46,819 --> 01:07:47,869
Or is that
important for an

1675
01:07:47,869 --> 01:07:49,790
open-ended basins
of attraction?

1676
01:07:49,790 --> 01:07:52,699
Are there, is
there actually,

1677
01:07:52,699 --> 01:07:54,349
I want to say
there's a paper

1678
01:07:54,349 --> 01:08:03,949
by I'm not able

1679
01:08:03,949 --> 01:08:04,970
to find out the
top my head.

1680
01:08:04,970 --> 01:08:06,260
I thought I was just

1681
01:08:06,260 --> 01:08:08,150
looking at it, but
now I've lost it.

1682
01:08:08,150 --> 01:08:12,109
But yeah, they're
having some Edwards

1683
01:08:12,109 --> 01:08:13,669
to ground this stuff

1684
01:08:13,669 --> 01:08:16,505
back in dynamical systems.

1685
01:08:16,505 --> 01:08:20,764
And looking at basins
of attraction.

1686
01:08:20,764 --> 01:08:29,009
That's very general kind
of systems concepts.

1687
01:08:29,650 --> 01:08:33,440
But I don't know,
like I said, there's,

1688
01:08:33,440 --> 01:08:39,054
there is a real insularity
in this community.

1689
01:08:39,054 --> 01:08:40,530
And again, it's like a lot

1690
01:08:40,530 --> 01:08:41,609
of the open-end evolution

1691
01:08:41,609 --> 01:08:42,900
were references

1692
01:08:42,900 --> 01:08:45,269
the other open-ended
Ocean work.

1693
01:08:45,269 --> 01:08:47,130
And you just don't
see a lot of

1694
01:08:47,130 --> 01:08:51,240
contacts with even
some of this stuff

1695
01:08:51,240 --> 01:08:56,535
on self-organization
and dynamical systems

1696
01:08:56,535 --> 01:09:00,150
and things that
you might expect.

1697
01:09:00,150 --> 01:09:05,625
I found this, There's
a guy hi land,

1698
01:09:05,625 --> 01:09:07,200
who writes about chemical

1699
01:09:07,200 --> 01:09:08,999
organization theory and

1700
01:09:08,999 --> 01:09:10,335
has this whole relational

1701
01:09:10,335 --> 01:09:12,539
agency ontology
that I find very

1702
01:09:12,539 --> 01:09:16,155
interesting and
very applicable

1703
01:09:16,155 --> 01:09:18,390
to what's being talked

1704
01:09:18,390 --> 01:09:19,589
about and this

1705
01:09:19,589 --> 01:09:21,165
open-ended evolution
literature.

1706
01:09:21,165 --> 01:09:22,814
He doesn't appear
to be aware

1707
01:09:22,814 --> 01:09:23,879
of the open-end evolution

1708
01:09:23,879 --> 01:09:25,080
of literature
and they don't

1709
01:09:25,080 --> 01:09:26,910
really appear to
be aware him.

1710
01:09:26,910 --> 01:09:31,124
So maybe your job chain?

1711
01:09:31,124 --> 01:09:32,009
Yeah.

1712
01:09:32,009 --> 01:09:33,989
Mm-hm.

1713
01:09:33,989 --> 01:09:35,040
Yeah.

1714
01:09:35,040 --> 01:09:38,295
That's our
colleague because

1715
01:09:38,295 --> 01:09:39,435
I don't think on
the call that

1716
01:09:39,435 --> 01:09:42,240
Rajesh is a good person

1717
01:09:42,240 --> 01:09:44,925
now who is familiar
with the last stuff.

1718
01:09:44,925 --> 01:09:47,340
And sorry,

1719
01:09:47,340 --> 01:09:49,410
I can't answer that
question better, I think.

1720
01:09:49,410 --> 01:09:50,609
Oh, that's okay.

1721
01:09:50,609 --> 01:09:52,004
It sounds like
the open-ended

1722
01:09:52,004 --> 01:09:53,789
needs to apply some
of their own ideas

1723
01:09:53,789 --> 01:09:55,379
to be more open

1724
01:09:55,379 --> 01:09:58,155
and gotta reach
out a little bit.

1725
01:09:58,155 --> 01:10:00,969
Thank you. Yeah.

1726
01:10:02,900 --> 01:10:04,740
I'm going to go ahead and

1727
01:10:04,740 --> 01:10:06,030
turn the recorder off.

1728
01:10:06,030 --> 01:10:08,190
I like to encourage
ongoing conversations,

1729
01:10:08,190 --> 01:10:10,139
but I think it's
reasonable to record for

1730
01:10:10,139 --> 01:10:11,910
the record at this point

1731
01:10:11,910 --> 01:10:13,589
with a few questions
being answered.

1732
01:10:13,589 --> 01:10:14,640
So I'm going to go ahead

1733
01:10:14,640 --> 01:10:18,939
and stop the recording.