1
00:00:04,630 --> 00:00:02,710
push to not feel fear slash not feel the

2
00:00:06,869 --> 00:00:04,640
ego quote unquote

3
00:00:08,629 --> 00:00:06,879
etc can be dangerous not because the

4
00:00:10,310 --> 00:00:08,639
lesson is false though it may be but

5
00:00:11,830 --> 00:00:10,320
because of the social pressure around

6
00:00:13,669 --> 00:00:11,840
those words in fact there's a

7
00:00:15,910 --> 00:00:13,679
universally frowned upon quality called

8
00:00:18,310 --> 00:00:15,920
being egotistical ego is embedded in it

9
00:00:20,470 --> 00:00:18,320
and also to be seen as retreating is the

10
00:00:22,390 --> 00:00:20,480
same as being seen as being defensive

11
00:00:24,630 --> 00:00:22,400
which is the same as being fearful which

12
00:00:26,470 --> 00:00:24,640
is seen as being lower on the spiritual

13
00:00:28,470 --> 00:00:26,480
hierarchy perhaps it's true that you

14
00:00:29,990 --> 00:00:28,480
should feel less disquieted about

15
00:00:32,310 --> 00:00:30,000
certain facts though perhaps your

16
00:00:35,590 --> 00:00:32,320
disqualitude is justified for where you

17
00:00:38,069 --> 00:00:35,600
are now and that's okay it's 100

18
00:00:39,830 --> 00:00:38,079
okay it took me about one year to

19
00:00:41,990 --> 00:00:39,840
realize that in fact perhaps more than

20
00:00:43,990 --> 00:00:42,000
one year took me over one year to

21
00:00:45,910 --> 00:00:44,000
realize that these spiritual lessons

22
00:00:47,750 --> 00:00:45,920
tend to not have the physics notion of

23
00:00:49,830 --> 00:00:47,760
path dependence perhaps you shouldn't

24
00:00:52,470 --> 00:00:49,840
uncritically adopt an eastern practice

25
00:00:54,310 --> 00:00:52,480
in a western world and vice versa or

26
00:00:56,069 --> 00:00:54,320
perhaps certain truths are dependent on

27
00:00:57,590 --> 00:00:56,079
the path that is where you are now

28
00:00:59,510 --> 00:00:57,600
rather than being independent of

29
00:01:01,910 --> 00:00:59,520
anything frankly it's embarrassing to

30
00:01:04,710 --> 00:01:01,920
reveal myself as so

31
00:01:05,910 --> 00:01:04,720
selfish and malicious and

32
00:01:08,550 --> 00:01:05,920
cowardly

33
00:01:11,109 --> 00:01:08,560
and to do so so unguardedly to carl

34
00:01:12,710 --> 00:01:11,119
fristen and even to you

35
00:01:18,310 --> 00:01:12,720
as i'm a

36
00:01:20,630 --> 00:01:18,320
at least currently i am but hopefully in

37
00:01:21,990 --> 00:01:20,640
the end it's a net benefit the dangers

38
00:01:24,390 --> 00:01:22,000
of going on to journeys of

39
00:01:26,710 --> 00:01:24,400
investigations into consciousness the

40
00:01:29,350 --> 00:01:26,720
self computational reality and even the

41
00:01:31,830 --> 00:01:29,360
phenomenon as it can lead to a spiral

42
00:01:34,550 --> 00:01:31,840
down some viciously dark paths that are

43
00:01:36,630 --> 00:01:34,560
at least ostensibly irrevocable while

44
00:01:39,030 --> 00:01:36,640
having certainty that your new founded

45
00:01:40,710 --> 00:01:39,040
conception of reality is indeed truthful

46
00:01:43,109 --> 00:01:40,720
while i don't like to state personal

47
00:01:45,270 --> 00:01:43,119
opinions because i generally don't have

48
00:01:47,510 --> 00:01:45,280
many strong views and the views that i

49
00:01:49,510 --> 00:01:47,520
do have are fickle changing week by week

50
00:01:51,830 --> 00:01:49,520
sometimes my current conclusion is that

51
00:01:53,670 --> 00:01:51,840
if it's not life affirming if it's not

52
00:01:55,350 --> 00:01:53,680
loving if it's not something you can

53
00:01:57,910 --> 00:01:55,360
tell others without others being

54
00:01:59,429 --> 00:01:57,920
horrified and frightened for you then

55
00:02:01,030 --> 00:01:59,439
use that as a sign that what you're

56
00:02:03,910 --> 00:02:01,040
experiencing or what you're feeling or

57
00:02:07,030 --> 00:02:03,920
what you think is not truthful please

58
00:02:09,350 --> 00:02:07,040
seek help don't suffer alone you're not

59
00:02:12,070 --> 00:02:09,360
alone and you can verify that simply by

60
00:02:13,750 --> 00:02:12,080
talking to as many people as you can if

61
00:02:15,270 --> 00:02:13,760
you're listening to this podcast then

62
00:02:17,990 --> 00:02:15,280
it's likely because you at least have

63
00:02:20,470 --> 00:02:18,000
some trust in it so trust that statement

64
00:02:22,630 --> 00:02:20,480
or at least test that statement out now

65
00:02:24,949 --> 00:02:22,640
on to the guest carl fristen is a

66
00:02:26,949 --> 00:02:24,959
professor of neuroscience at university

67
00:02:29,350 --> 00:02:26,959
college london and is one of the most

68
00:02:31,350 --> 00:02:29,360
cited individuals in his field he's also

69
00:02:32,790 --> 00:02:31,360
the inventor of the free energy

70
00:02:34,470 --> 00:02:32,800
principle and is someone who's made

71
00:02:36,550 --> 00:02:34,480
advances in the field of active

72
00:02:38,949 --> 00:02:36,560
inference that's matched almost by no

73
00:02:40,869 --> 00:02:38,959
one he may be one of the most insightful

74
00:02:42,390 --> 00:02:40,879
and brilliant people on the planet i've

75
00:02:45,110 --> 00:02:42,400
been lucky enough to speak to carl

76
00:02:46,949 --> 00:02:45,120
before in a substantive 4-hour podcast

77
00:02:49,190 --> 00:02:46,959
linked below which is best watched at

78
00:02:51,350 --> 00:02:49,200
least twice to properly understand this

79
00:02:52,790 --> 00:02:51,360
episode is an outlier in the toe channel

80
00:02:55,030 --> 00:02:52,800
because this will serve as an

81
00:02:56,470 --> 00:02:55,040
introduction to that episode as many

82
00:02:58,229 --> 00:02:56,480
after watching that we're still unclear

83
00:03:00,550 --> 00:02:58,239
as to what the free energy principle is

84
00:03:02,710 --> 00:03:00,560
if you feel like you're lost that's okay

85
00:03:04,390 --> 00:03:02,720
let that be a theme understanding is

86
00:03:06,149 --> 00:03:04,400
less like a light switch and more like a

87
00:03:07,589 --> 00:03:06,159
stove top that is it's not like all of a

88
00:03:09,110 --> 00:03:07,599
sudden you have an insight and then you

89
00:03:10,710 --> 00:03:09,120
understand some phenomenon you went from

90
00:03:13,030 --> 00:03:10,720
zero to one it's more like you get a

91
00:03:15,030 --> 00:03:13,040
gradual increase in understanding where

92
00:03:17,350 --> 00:03:15,040
a modicum here makes more sense and then

93
00:03:19,350 --> 00:03:17,360
a smidge in here as wheeler says and i

94
00:03:21,509 --> 00:03:19,360
like to quote the point isn't to drink

95
00:03:23,910 --> 00:03:21,519
from the fire hose but instead to get

96
00:03:25,750 --> 00:03:23,920
wet trusting that your conscious

97
00:03:28,070 --> 00:03:25,760
befuddlement is part of the process and

98
00:03:29,190 --> 00:03:28,080
unconsciously more is becoming clear

99
00:03:30,710 --> 00:03:29,200
click on the timestamp in the

100
00:03:33,190 --> 00:03:30,720
description if you'd like to skip this

101
00:03:35,509 --> 00:03:33,200
intro my name is kirchey mungle i'm a

102
00:03:37,990 --> 00:03:35,519
torontonian filmmaker with a background

103
00:03:40,470 --> 00:03:38,000
in mathematical physics dedicated to the

104
00:03:42,229 --> 00:03:40,480
explication of the variegated terrain of

105
00:03:44,149 --> 00:03:42,239
theories of everything from a

106
00:03:46,149 --> 00:03:44,159
theoretical physics perspective but as

107
00:03:47,910 --> 00:03:46,159
well as analyzing consciousness and

108
00:03:50,070 --> 00:03:47,920
seeing its potential connection to

109
00:03:51,990 --> 00:03:50,080
fundamental reality whatever that is

110
00:03:54,550 --> 00:03:52,000
essentially this channel is dedicated to

111
00:03:56,630 --> 00:03:54,560
exploring the under nature of reality

112
00:03:58,229 --> 00:03:56,640
the constitutional laws that govern it

113
00:04:00,550 --> 00:03:58,239
provided those laws exist at all and are

114
00:04:02,470 --> 00:04:00,560
even noble to us if you enjoy witnessing

115
00:04:05,270 --> 00:04:02,480
and engaging with others on the topics

116
00:04:07,509 --> 00:04:05,280
of psychology consciousness physics etc

117
00:04:09,110 --> 00:04:07,519
the channels themes then do consider

118
00:04:11,030 --> 00:04:09,120
going to the discord and the subreddit

119
00:04:12,869 --> 00:04:11,040
which are linked in the description

120
00:04:15,509 --> 00:04:12,879
there's also a link to the patreon that

121
00:04:17,430 --> 00:04:15,519
is patreon.com slash kirchheim uncle if

122
00:04:19,349 --> 00:04:17,440
you'd like to support this podcast as

123
00:04:21,270 --> 00:04:19,359
the patrons and the sponsors are the

124
00:04:23,590 --> 00:04:21,280
only reasons that i'm able to have

125
00:04:25,590 --> 00:04:23,600
podcasts of this quality and this depth

126
00:04:27,030 --> 00:04:25,600
given that i can do this now full time

127
00:04:29,189 --> 00:04:27,040
thanks to both the patrons and the

128
00:04:31,110 --> 00:04:29,199
sponsors support speaking of sponsors

129
00:04:32,870 --> 00:04:31,120
there are two the first sponsor is

130
00:04:34,310 --> 00:04:32,880
brilliant during the winter break i

131
00:04:36,150 --> 00:04:34,320
decided to brush up on some of the

132
00:04:38,310 --> 00:04:36,160
fundamentals of physics particularly

133
00:04:40,230 --> 00:04:38,320
with regard to information theory as i'd

134
00:04:41,909 --> 00:04:40,240
like to interview chiara marletto on

135
00:04:43,510 --> 00:04:41,919
constructor theory which is heavily

136
00:04:45,189 --> 00:04:43,520
based in information theory now

137
00:04:46,550 --> 00:04:45,199
information theory is predicated on

138
00:04:48,950 --> 00:04:46,560
entropy at least there's a fundamental

139
00:04:50,629 --> 00:04:48,960
formula for entropy so i ended up taking

140
00:04:52,710 --> 00:04:50,639
the brilliant course i challenged myself

141
00:04:54,710 --> 00:04:52,720
to do one lesson per day and i took the

142
00:04:55,909 --> 00:04:54,720
course's random variable distributions

143
00:04:58,230 --> 00:04:55,919
and knowledge

144
00:04:59,909 --> 00:04:58,240
uncertainty what i loved is that despite

145
00:05:01,510 --> 00:04:59,919
knowing the formula for entropy which is

146
00:05:03,430 --> 00:05:01,520
essentially hammered into you as an

147
00:05:05,749 --> 00:05:03,440
undergraduate it seems like it comes

148
00:05:07,510 --> 00:05:05,759
down from the sky arbitrarily and with

149
00:05:09,270 --> 00:05:07,520
brilliance for the first time i was able

150
00:05:10,710 --> 00:05:09,280
to see how the formula for entropy which

151
00:05:12,950 --> 00:05:10,720
you're seeing right now is actually

152
00:05:14,710 --> 00:05:12,960
extremely natural and it'd be strange to

153
00:05:16,550 --> 00:05:14,720
define it in any other manner there are

154
00:05:18,150 --> 00:05:16,560
plenty of courses and you can even learn

155
00:05:19,270 --> 00:05:18,160
group theory which is what's being

156
00:05:21,510 --> 00:05:19,280
referenced when you hear that the

157
00:05:23,830 --> 00:05:21,520
standard model is predicated on u1 cross

158
00:05:25,189 --> 00:05:23,840
su-2 cross su-3 those are league groups

159
00:05:26,629 --> 00:05:25,199
continuous lead groups visit

160
00:05:29,350 --> 00:05:26,639
brilliant.org

161
00:05:31,029 --> 00:05:29,360
tow t-o-e to get 20 off an annual

162
00:05:33,189 --> 00:05:31,039
subscription and i recommend that you

163
00:05:34,950 --> 00:05:33,199
don't stop before four lessons i think

164
00:05:36,710 --> 00:05:34,960
you'll be greatly surprised at the ease

165
00:05:38,629 --> 00:05:36,720
at which you can now comprehend subjects

166
00:05:41,350 --> 00:05:38,639
you previously had a difficult time

167
00:05:43,430 --> 00:05:41,360
grocking the second sponsor is algo now

168
00:05:45,430 --> 00:05:43,440
algo is an end-to-end supply chain

169
00:05:47,510 --> 00:05:45,440
optimization software company with

170
00:05:49,510 --> 00:05:47,520
software that helps business users

171
00:05:51,670 --> 00:05:49,520
optimize sales and operations planning

172
00:05:53,590 --> 00:05:51,680
to avoid stockouts reduce return and

173
00:05:55,990 --> 00:05:53,600
inventory write-downs while reducing

174
00:05:59,189 --> 00:05:56,000
inventory investment it's a supply chain

175
00:06:01,270 --> 00:05:59,199
ai that drives smart roi headed by amjad

176
00:06:03,670 --> 00:06:01,280
hussein who's been a huge supporter of

177
00:06:06,710 --> 00:06:03,680
this podcast since near its inception in

178
00:06:08,550 --> 00:06:06,720
fact amjad has his own podcast on ai and

179
00:06:10,469 --> 00:06:08,560
consciousness and business growth and if

180
00:06:12,390 --> 00:06:10,479
you'd like to support the toe podcast

181
00:06:14,070 --> 00:06:12,400
then visit the link in the description

182
00:06:16,070 --> 00:06:14,080
to see amjad's podcast because

183
00:06:18,550 --> 00:06:16,080
subscribing to him or at least visiting

184
00:06:20,469 --> 00:06:18,560
supports the toe podcast indirectly

185
00:06:22,070 --> 00:06:20,479
thank you and enjoy let's do an

186
00:06:23,990 --> 00:06:22,080
introduction once more so what is the

187
00:06:25,909 --> 00:06:24,000
free energy principle specifically

188
00:06:27,350 --> 00:06:25,919
specifically its relationship to a

189
00:06:29,029 --> 00:06:27,360
theory of everything as that's what this

190
00:06:31,510 --> 00:06:29,039
channel is about

191
00:06:32,390 --> 00:06:31,520
right um

192
00:06:34,150 --> 00:06:32,400
so

193
00:06:36,870 --> 00:06:34,160
my usual response here is do you want

194
00:06:38,070 --> 00:06:36,880
the the high road or the lo or or the

195
00:06:40,950 --> 00:06:38,080
low road

196
00:06:42,469 --> 00:06:40,960
um let's take the high road uh given

197
00:06:44,070 --> 00:06:42,479
that we're talking about theories of

198
00:06:45,110 --> 00:06:44,080
everything um

199
00:06:48,070 --> 00:06:45,120
so

200
00:06:49,749 --> 00:06:48,080
the free energy principle is a principle

201
00:06:53,029 --> 00:06:49,759
of least action

202
00:06:56,390 --> 00:06:53,039
so it's a method really it's put most

203
00:06:59,510 --> 00:06:56,400
simply a method that allows you to

204
00:07:03,270 --> 00:06:59,520
identify the most likely dynamics the

205
00:07:04,629 --> 00:07:03,280
most likely way a process will unfold

206
00:07:05,990 --> 00:07:04,639
what is the process that we're talking

207
00:07:07,670 --> 00:07:06,000
about here

208
00:07:09,909 --> 00:07:07,680
but we're talking about

209
00:07:11,589 --> 00:07:09,919
things like you and me particles or

210
00:07:13,990 --> 00:07:11,599
people

211
00:07:15,430 --> 00:07:14,000
so the first thing to note is that you

212
00:07:16,710 --> 00:07:15,440
can

213
00:07:18,629 --> 00:07:16,720
apply

214
00:07:20,790 --> 00:07:18,639
standard variational principles of least

215
00:07:22,790 --> 00:07:20,800
action to any system specifically a

216
00:07:25,110 --> 00:07:22,800
random dynamical system of the kind that

217
00:07:26,629 --> 00:07:25,120
could be described by say a launchvan

218
00:07:27,990 --> 00:07:26,639
equation that underwrites most of the

219
00:07:30,150 --> 00:07:28,000
physics we know

220
00:07:31,589 --> 00:07:30,160
but the special thing about the free

221
00:07:33,510 --> 00:07:31,599
energy principle

222
00:07:36,390 --> 00:07:33,520
is it tries to understand the

223
00:07:39,430 --> 00:07:36,400
relationship of something

224
00:07:41,909 --> 00:07:39,440
in relation to everything else

225
00:07:43,350 --> 00:07:41,919
so you're starting from the premise that

226
00:07:44,550 --> 00:07:43,360
something exists and then you have to

227
00:07:46,469 --> 00:07:44,560
think carefully

228
00:07:48,950 --> 00:07:46,479
your what is a thing and how do i

229
00:07:50,309 --> 00:07:48,960
distinguish it from something else so

230
00:07:52,230 --> 00:07:50,319
immediately you start you have to

231
00:07:55,749 --> 00:07:52,240
introduce a partition

232
00:07:58,070 --> 00:07:55,759
between the states of things

233
00:08:00,790 --> 00:07:58,080
that comprise states that are internal

234
00:08:02,550 --> 00:08:00,800
to the thing and states that are outside

235
00:08:03,589 --> 00:08:02,560
the thing and

236
00:08:05,990 --> 00:08:03,599
thereby

237
00:08:07,990 --> 00:08:06,000
in think carefully about how those

238
00:08:10,309 --> 00:08:08,000
states are coupled and what happens is

239
00:08:13,029 --> 00:08:10,319
that you have to introduce a further set

240
00:08:16,230 --> 00:08:13,039
of states known as blanket states

241
00:08:18,550 --> 00:08:16,240
that constitute a markov blanket that

242
00:08:20,309 --> 00:08:18,560
mediate a bi-directional exchange

243
00:08:23,270 --> 00:08:20,319
between the internal states and the

244
00:08:26,390 --> 00:08:23,280
external states so you've got this

245
00:08:27,510 --> 00:08:26,400
notion now of taking the states of any

246
00:08:29,270 --> 00:08:27,520
universe

247
00:08:30,469 --> 00:08:29,280
any system that you want to try and

248
00:08:33,029 --> 00:08:30,479
understand

249
00:08:36,469 --> 00:08:33,039
and partitioning it or carving it into

250
00:08:38,230 --> 00:08:36,479
four subsets internal external and then

251
00:08:40,630 --> 00:08:38,240
sensory active states that constitute

252
00:08:42,230 --> 00:08:40,640
the blanket states where the active

253
00:08:44,790 --> 00:08:42,240
states mediate the effects of the

254
00:08:46,150 --> 00:08:44,800
internal on the external on the inside

255
00:08:48,310 --> 00:08:46,160
on the outside

256
00:08:51,509 --> 00:08:48,320
and the sensory states conversely

257
00:08:53,190 --> 00:08:51,519
mediate the influences of the external

258
00:08:55,829 --> 00:08:53,200
on the internal

259
00:08:57,590 --> 00:08:55,839
so with this particular partition which

260
00:08:59,670 --> 00:08:57,600
you wouldn't need if you wanted to now

261
00:09:00,870 --> 00:08:59,680
just move on and develop

262
00:09:05,190 --> 00:09:00,880
um

263
00:09:08,710 --> 00:09:05,200
mechanics or lagrangian classical

264
00:09:10,470 --> 00:09:08,720
mechanics we're now in the special case

265
00:09:14,310 --> 00:09:10,480
of dealing with a system that has this

266
00:09:16,470 --> 00:09:14,320
partition and in this special case

267
00:09:19,269 --> 00:09:16,480
those conventional variational

268
00:09:24,230 --> 00:09:19,279
principles of least action

269
00:09:26,389 --> 00:09:24,240
translate into or can be used to

270
00:09:28,870 --> 00:09:26,399
associate the paths of least action the

271
00:09:31,829 --> 00:09:28,880
most likely paths that the system will

272
00:09:33,509 --> 00:09:31,839
take particularly the states of the

273
00:09:35,509 --> 00:09:33,519
particle or person that we're talking

274
00:09:37,269 --> 00:09:35,519
about

275
00:09:40,710 --> 00:09:37,279
that we can associate those paths at

276
00:09:43,590 --> 00:09:40,720
least action with paths that minimize

277
00:09:45,269 --> 00:09:43,600
something called variational free energy

278
00:09:47,269 --> 00:09:45,279
so that's basically it and then the

279
00:09:50,550 --> 00:09:47,279
story as well what's what's variation

280
00:09:54,470 --> 00:09:50,560
free energy well variational free energy

281
00:09:58,150 --> 00:09:54,480
uh is as with all actions in physics

282
00:09:59,190 --> 00:09:58,160
it's just a um a path integral or a time

283
00:10:03,829 --> 00:09:59,200
integral

284
00:10:06,470 --> 00:10:03,839
uh a sum an accumulation of a functional

285
00:10:09,670 --> 00:10:06,480
usually known as a lagrangian

286
00:10:11,829 --> 00:10:09,680
where in this instance the func the

287
00:10:13,829 --> 00:10:11,839
functional in question this variational

288
00:10:16,310 --> 00:10:13,839
free energy

289
00:10:20,710 --> 00:10:16,320
can be read as exactly the same quantity

290
00:10:25,030 --> 00:10:22,230
psychologists

291
00:10:27,269 --> 00:10:25,040
people trying to understand data

292
00:10:29,190 --> 00:10:27,279
would treat as

293
00:10:31,990 --> 00:10:29,200
a proxy for

294
00:10:33,590 --> 00:10:32,000
the evidence for some model of the world

295
00:10:35,430 --> 00:10:33,600
so this is known as bayesian model

296
00:10:37,750 --> 00:10:35,440
evidence

297
00:10:40,630 --> 00:10:37,760
statistically it's also known as a

298
00:10:41,509 --> 00:10:40,640
marginal likelihood the likelihood of

299
00:10:43,910 --> 00:10:41,519
these

300
00:10:46,710 --> 00:10:43,920
sensory data the sensory the path

301
00:10:48,389 --> 00:10:46,720
through the sensory states under a

302
00:10:50,550 --> 00:10:48,399
particular model

303
00:10:54,150 --> 00:10:50,560
that would try to explain

304
00:10:55,110 --> 00:10:54,160
those data

305
00:10:57,910 --> 00:10:55,120
so

306
00:11:00,470 --> 00:10:57,920
you've got this interpretation or

307
00:11:01,430 --> 00:11:00,480
certainly the fact that the functional

308
00:11:03,750 --> 00:11:01,440
form

309
00:11:05,590 --> 00:11:03,760
of this variation free energy that is

310
00:11:07,910 --> 00:11:05,600
being minimized by the paths of least

311
00:11:09,910 --> 00:11:07,920
action of the autonomous the active and

312
00:11:11,750 --> 00:11:09,920
the internal states of something a

313
00:11:14,710 --> 00:11:11,760
particle or a person

314
00:11:17,190 --> 00:11:14,720
now looks as if it is trying to minimize

315
00:11:19,670 --> 00:11:17,200
a quantity which provides an

316
00:11:21,190 --> 00:11:19,680
approximation or a negative

317
00:11:22,150 --> 00:11:21,200
approximation

318
00:11:24,630 --> 00:11:22,160
to

319
00:11:26,630 --> 00:11:24,640
marginal likelihood or moral evidencing

320
00:11:29,670 --> 00:11:26,640
so in other words it looks as if the

321
00:11:31,430 --> 00:11:29,680
most likely paths are trying to maximize

322
00:11:32,630 --> 00:11:31,440
model evidence

323
00:11:34,470 --> 00:11:32,640
so

324
00:11:36,150 --> 00:11:34,480
one can read that in many different ways

325
00:11:37,190 --> 00:11:36,160
one can read that

326
00:11:40,150 --> 00:11:37,200
and

327
00:11:43,829 --> 00:11:40,160
as jakob howie has read it

328
00:11:46,310 --> 00:11:43,839
a philosopher you're a scientist who

329
00:11:48,069 --> 00:11:46,320
has a commitment and understanding to

330
00:11:51,030 --> 00:11:48,079
this kind of formulation

331
00:11:52,949 --> 00:11:51,040
as self evidencing literally

332
00:11:54,230 --> 00:11:52,959
interacting with your

333
00:11:56,870 --> 00:11:54,240
environment

334
00:11:59,509 --> 00:11:56,880
actively inferring engaging with your

335
00:12:01,590 --> 00:11:59,519
experienced world your eco niche

336
00:12:03,190 --> 00:12:01,600
in a way that looks as if you're trying

337
00:12:05,030 --> 00:12:03,200
to solicit

338
00:12:07,670 --> 00:12:05,040
the most evidence

339
00:12:09,910 --> 00:12:07,680
for your model of that world

340
00:12:11,910 --> 00:12:09,920
this quantity is also found in machine

341
00:12:15,269 --> 00:12:11,920
learning it's known as an evidence lower

342
00:12:17,990 --> 00:12:15,279
bound i should um qualify or just

343
00:12:20,710 --> 00:12:18,000
clarify that in machine learning this

344
00:12:22,790 --> 00:12:20,720
free energy is the reverse of the free

345
00:12:24,870 --> 00:12:22,800
energy used in physics so in machine

346
00:12:28,389 --> 00:12:24,880
learning they try to maximize the

347
00:12:31,509 --> 00:12:28,399
negative free energy um aka an evidence

348
00:12:33,509 --> 00:12:31,519
lower bound or elbow um while in physics

349
00:12:35,430 --> 00:12:33,519
you're you're always articulating in

350
00:12:37,829 --> 00:12:35,440
terms of minimizing a potential energy

351
00:12:39,670 --> 00:12:37,839
or minimizing in this instance of

352
00:12:42,470 --> 00:12:39,680
variation free energy

353
00:12:44,470 --> 00:12:42,480
so you'll find that exactly

354
00:12:46,230 --> 00:12:44,480
the same functional the same quantity

355
00:12:47,269 --> 00:12:46,240
the same measure

356
00:12:55,590 --> 00:12:47,279
of

357
00:13:00,629 --> 00:12:55,600
sensorium was generated

358
00:13:02,470 --> 00:13:00,639
um in terms of um the the this um

359
00:13:06,629 --> 00:13:02,480
variational free energy

360
00:13:09,269 --> 00:13:06,639
that leads you to an interpretation of

361
00:13:10,710 --> 00:13:09,279
people particle persons

362
00:13:13,190 --> 00:13:10,720
inferring

363
00:13:14,949 --> 00:13:13,200
accumulating evidence making inferences

364
00:13:18,150 --> 00:13:14,959
about the states of the world that are

365
00:13:20,870 --> 00:13:18,160
generating their sensations so that in a

366
00:13:24,310 --> 00:13:20,880
nutshell or perhaps not on our channel

367
00:13:26,710 --> 00:13:24,320
is is a formal explanation of the free

368
00:13:27,670 --> 00:13:26,720
energy principle i did so slipping in

369
00:13:29,990 --> 00:13:27,680
the beginning

370
00:13:31,750 --> 00:13:30,000
very much like um

371
00:13:34,310 --> 00:13:31,760
things which you you know you may be

372
00:13:36,389 --> 00:13:34,320
familiar with when you were at school

373
00:13:37,670 --> 00:13:36,399
say hamilton's principle of station

374
00:13:39,110 --> 00:13:37,680
reaction

375
00:13:41,350 --> 00:13:39,120
um

376
00:13:44,949 --> 00:13:41,360
it is

377
00:13:47,670 --> 00:13:44,959
just a method you know it allows you

378
00:13:49,750 --> 00:13:47,680
to identify predict

379
00:13:52,870 --> 00:13:49,760
simulate engineer

380
00:13:54,310 --> 00:13:52,880
those trajectories as paths as dynamics

381
00:13:58,069 --> 00:13:54,320
those processes

382
00:13:58,870 --> 00:13:58,079
um that will um minimize

383
00:14:05,030 --> 00:13:58,880
the

384
00:14:07,269 --> 00:14:05,040
so what that means is that

385
00:14:09,030 --> 00:14:07,279
if you now understand

386
00:14:11,829 --> 00:14:09,040
or have at hand

387
00:14:12,949 --> 00:14:11,839
the functional form of your variation

388
00:14:15,670 --> 00:14:12,959
free energy

389
00:14:18,389 --> 00:14:15,680
that depends upon

390
00:14:19,990 --> 00:14:18,399
posterior beliefs or beliefs about the

391
00:14:22,150 --> 00:14:20,000
outside world

392
00:14:24,790 --> 00:14:22,160
under a generative model quite simply a

393
00:14:27,990 --> 00:14:24,800
probability density over

394
00:14:31,430 --> 00:14:28,000
the external causes of the

395
00:14:34,389 --> 00:14:31,440
consequences that are the sensory states

396
00:14:37,269 --> 00:14:34,399
and then you can use that methodology

397
00:14:41,430 --> 00:14:37,279
use that principle to start to simulate

398
00:14:44,550 --> 00:14:41,440
to build sentient artifacts that behave

399
00:14:47,189 --> 00:14:44,560
in the way that we assume under the free

400
00:14:48,629 --> 00:14:47,199
energy principle that you and i are

401
00:14:50,629 --> 00:14:48,639
behaving let's see if i can make a

402
00:14:52,629 --> 00:14:50,639
nutshell of the nutshell for some people

403
00:14:54,230 --> 00:14:52,639
who are more familiar with physics so

404
00:14:55,750 --> 00:14:54,240
let's say you have a ball you want to

405
00:14:57,509 --> 00:14:55,760
know the dynamics of the ball you can

406
00:14:59,030 --> 00:14:57,519
understand this with newtonian force and

407
00:15:00,790 --> 00:14:59,040
that's useful for certain calculations

408
00:15:02,150 --> 00:15:00,800
or you can think of it as minimizing a

409
00:15:05,189 --> 00:15:02,160
lagrangian and then that actually

410
00:15:06,870 --> 00:15:05,199
provides a more useful for most cases

411
00:15:08,389 --> 00:15:06,880
way of calculating the trajectory of the

412
00:15:09,829 --> 00:15:08,399
ball so then

413
00:15:12,230 --> 00:15:09,839
right there when i said ball we've

414
00:15:13,990 --> 00:15:12,240
identified the ball now here you're not

415
00:15:15,350 --> 00:15:14,000
just limiting yourself to balls you're

416
00:15:16,710 --> 00:15:15,360
saying let's say cells instead of

417
00:15:18,550 --> 00:15:16,720
talking about people because we can

418
00:15:20,150 --> 00:15:18,560
break it down simply at least for me for

419
00:15:21,990 --> 00:15:20,160
myself i find it much more easy to think

420
00:15:23,590 --> 00:15:22,000
of a cell okay so then we want to know

421
00:15:26,550 --> 00:15:23,600
what are the dynamics of the cell how is

422
00:15:28,870 --> 00:15:26,560
the cell a single cell going to act

423
00:15:30,790 --> 00:15:28,880
okay firstly the first step would be

424
00:15:31,509 --> 00:15:30,800
identifying that cell

425
00:15:33,829 --> 00:15:31,519
and

426
00:15:35,829 --> 00:15:33,839
when you mention the word markov blanket

427
00:15:37,829 --> 00:15:35,839
that's a way of identifying what this

428
00:15:39,829 --> 00:15:37,839
object is because in order to identify

429
00:15:41,509 --> 00:15:39,839
what an object is it's often useful to

430
00:15:43,269 --> 00:15:41,519
identify what it isn't at the same time

431
00:15:45,750 --> 00:15:43,279
what it is and isn't so that markov

432
00:15:48,550 --> 00:15:45,760
blanket delimits the cell now is that

433
00:15:49,829 --> 00:15:48,560
correct so far absolutely yeah so you

434
00:15:52,230 --> 00:15:49,839
know i think the cell is a perfect

435
00:15:55,749 --> 00:15:52,240
example here you've got you know

436
00:15:57,670 --> 00:15:55,759
a cell is just defined by the boundary

437
00:15:58,949 --> 00:15:57,680
which would be the surface of the cell

438
00:16:00,710 --> 00:15:58,959
and that would essentially be the

439
00:16:01,670 --> 00:16:00,720
sensory states that i was talking about

440
00:16:04,389 --> 00:16:01,680
before

441
00:16:06,230 --> 00:16:04,399
the insides the intracellular states

442
00:16:08,629 --> 00:16:06,240
would be the internal states and then

443
00:16:10,550 --> 00:16:08,639
the milieu in which that cell lives

444
00:16:13,030 --> 00:16:10,560
which is usually in

445
00:16:15,749 --> 00:16:13,040
a bunch or a family of other cells uh

446
00:16:16,629 --> 00:16:15,759
will constitute all the external states

447
00:16:18,150 --> 00:16:16,639
so

448
00:16:20,470 --> 00:16:18,160
you've made the very first step just

449
00:16:22,550 --> 00:16:20,480
talking about a cell a thing

450
00:16:26,550 --> 00:16:22,560
you've defined it in terms of this

451
00:16:29,749 --> 00:16:26,560
partition into internal external and the

452
00:16:31,990 --> 00:16:29,759
uh the sensory states that bound and

453
00:16:33,509 --> 00:16:32,000
separate the inside from the outside you

454
00:16:36,150 --> 00:16:33,519
may be asking well where the active

455
00:16:38,069 --> 00:16:36,160
states come first cell i usually like to

456
00:16:40,230 --> 00:16:38,079
associate those with something called

457
00:16:43,269 --> 00:16:40,240
actin filaments that just lie underneath

458
00:16:46,310 --> 00:16:43,279
the surface the sensory states and

459
00:16:48,470 --> 00:16:46,320
actively push the cell surface say oh

460
00:16:51,350 --> 00:16:48,480
the philia and

461
00:16:54,870 --> 00:16:51,360
the um

462
00:16:57,749 --> 00:16:54,880
the the mechanisms that cause motion of

463
00:16:58,470 --> 00:16:57,759
the cell should it move around

464
00:17:00,230 --> 00:16:58,480
so

465
00:17:02,150 --> 00:17:00,240
those parts those states that are

466
00:17:03,990 --> 00:17:02,160
responsible for cellular motility so if

467
00:17:06,230 --> 00:17:04,000
we're now thinking about a moving cell

468
00:17:07,669 --> 00:17:06,240
or a swimming cell for example

469
00:17:09,750 --> 00:17:07,679
then you need to supplement that

470
00:17:12,309 --> 00:17:09,760
partition with with active states so

471
00:17:14,549 --> 00:17:12,319
that's absolutely the first step um and

472
00:17:17,189 --> 00:17:14,559
then you're going to go on and now

473
00:17:19,429 --> 00:17:17,199
translate the ball what's the ball in

474
00:17:20,630 --> 00:17:19,439
this in this metaphor

475
00:17:22,870 --> 00:17:20,640
you also mentioned the word

476
00:17:24,870 --> 00:17:22,880
bi-directional earlier and when i was

477
00:17:26,949 --> 00:17:24,880
learning about this the way that i like

478
00:17:28,549 --> 00:17:26,959
to understand this was that in computer

479
00:17:30,710 --> 00:17:28,559
science generally computer scientists

480
00:17:32,070 --> 00:17:30,720
think and can only think in terms of you

481
00:17:33,270 --> 00:17:32,080
have an input then there's a black box

482
00:17:34,630 --> 00:17:33,280
and you have an output and that's

483
00:17:36,150 --> 00:17:34,640
essentially computer science in a

484
00:17:37,430 --> 00:17:36,160
nutshell it's a bit more complicated

485
00:17:38,710 --> 00:17:37,440
than that but it's in a nutshell it's do

486
00:17:40,630 --> 00:17:38,720
you have an input you perform some

487
00:17:43,029 --> 00:17:40,640
calculation and you get an output the

488
00:17:45,110 --> 00:17:43,039
reason why you said bi-directional at

489
00:17:47,510 --> 00:17:45,120
least the way that i understand it so

490
00:17:49,350 --> 00:17:47,520
please correct this is it's because of

491
00:17:52,070 --> 00:17:49,360
actions in that model where there's

492
00:17:53,430 --> 00:17:52,080
input black box output that's generally

493
00:17:55,430 --> 00:17:53,440
if we want to use the term embody

494
00:17:57,669 --> 00:17:55,440
cognition more on the cognition side so

495
00:17:59,590 --> 00:17:57,679
you're just this black box where you're

496
00:18:01,430 --> 00:17:59,600
impressed with sensory information the

497
00:18:03,270 --> 00:18:01,440
input is your senses you're sensing all

498
00:18:04,630 --> 00:18:03,280
of what's happening and then this output

499
00:18:06,950 --> 00:18:04,640
now

500
00:18:09,029 --> 00:18:06,960
is the actions that generally a computer

501
00:18:10,390 --> 00:18:09,039
sits there but we're creatures in this

502
00:18:12,150 --> 00:18:10,400
world at least we think we're creatures

503
00:18:15,590 --> 00:18:12,160
in a physical world the black box has a

504
00:18:17,990 --> 00:18:15,600
model and then one way of maximizing or

505
00:18:20,549 --> 00:18:18,000
showing that your model is correct is by

506
00:18:22,549 --> 00:18:20,559
updating your model or it's by changing

507
00:18:24,390 --> 00:18:22,559
what you've sensed in order to match the

508
00:18:27,029 --> 00:18:24,400
model now to change what you've sensed

509
00:18:28,950 --> 00:18:27,039
is to act so it's useful to think of two

510
00:18:30,549 --> 00:18:28,960
black boxes so there's one that's

511
00:18:31,909 --> 00:18:30,559
inputting to the other and then the

512
00:18:33,590 --> 00:18:31,919
output goes into the other which is

513
00:18:35,990 --> 00:18:33,600
external this is the internal this is

514
00:18:37,510 --> 00:18:36,000
the cell it senses and then it acts upon

515
00:18:39,190 --> 00:18:37,520
the world and then that changes the

516
00:18:40,830 --> 00:18:39,200
world which then changes what you sent

517
00:18:43,270 --> 00:18:40,840
so is that the reason that you said

518
00:18:45,750 --> 00:18:43,280
bi-directional yeah no no i think that's

519
00:18:48,950 --> 00:18:45,760
a really important observation yeah

520
00:18:51,270 --> 00:18:48,960
so you know before we were talking about

521
00:18:53,510 --> 00:18:51,280
these states that separate the inside

522
00:18:55,510 --> 00:18:53,520
from the outside and i made a big thing

523
00:18:57,430 --> 00:18:55,520
that they comprise sensory and active

524
00:19:00,070 --> 00:18:57,440
states and that there is a

525
00:19:02,470 --> 00:19:00,080
a two-way traffic here and that two-way

526
00:19:03,830 --> 00:19:02,480
traffic instantiates exactly what you're

527
00:19:05,190 --> 00:19:03,840
talking about which is a circular

528
00:19:06,990 --> 00:19:05,200
causality

529
00:19:11,669 --> 00:19:07,000
so you know i use the word

530
00:19:13,430 --> 00:19:11,679
self-evidencing and described it as

531
00:19:14,230 --> 00:19:13,440
soliciting evidence

532
00:19:15,909 --> 00:19:14,240
for

533
00:19:18,310 --> 00:19:15,919
sensory data

534
00:19:20,789 --> 00:19:18,320
that it provides evidence for your model

535
00:19:23,110 --> 00:19:20,799
of what's going on but in soliciting

536
00:19:24,390 --> 00:19:23,120
this is something that a computer would

537
00:19:26,789 --> 00:19:24,400
not normally do

538
00:19:29,510 --> 00:19:26,799
a passive or a sessile artifact like a

539
00:19:32,070 --> 00:19:29,520
computer it would not be in charge of

540
00:19:35,190 --> 00:19:32,080
nor would it terribly worry or indeed be

541
00:19:37,350 --> 00:19:35,200
equipped with the capacity to select the

542
00:19:38,390 --> 00:19:37,360
inputs to select the data say had to

543
00:19:40,870 --> 00:19:38,400
analyze

544
00:19:43,350 --> 00:19:40,880
so just having a statistical package you

545
00:19:45,590 --> 00:19:43,360
give it some data and it produces an

546
00:19:47,430 --> 00:19:45,600
outcome say a p value or some kind of

547
00:19:50,070 --> 00:19:47,440
inference um

548
00:19:52,390 --> 00:19:50,080
so a lot of computing and a lot of

549
00:19:53,830 --> 00:19:52,400
machine learning you give it data and it

550
00:19:55,350 --> 00:19:53,840
provides a prediction or a

551
00:19:57,590 --> 00:19:55,360
classification

552
00:19:59,270 --> 00:19:57,600
that's not the kind of computational

553
00:20:01,830 --> 00:19:59,280
process that is implied by the free

554
00:20:05,710 --> 00:20:01,840
energy principle it's much more as you

555
00:20:08,710 --> 00:20:05,720
put it um inactive it has you know the

556
00:20:11,110 --> 00:20:08,720
aspirations to provide a mathematical

557
00:20:12,630 --> 00:20:11,120
image of the action perception cycle

558
00:20:16,950 --> 00:20:12,640
where crucially

559
00:20:21,350 --> 00:20:16,960
you have to go and decide what data

560
00:20:23,590 --> 00:20:21,360
to base your inferences on by

561
00:20:24,390 --> 00:20:23,600
acting upon the world

562
00:20:29,909 --> 00:20:24,400
so

563
00:20:31,830 --> 00:20:29,919
the thing that distinguishes this kind

564
00:20:33,750 --> 00:20:31,840
of formalism this calculus is bayesian

565
00:20:36,789 --> 00:20:33,760
calculus which is quintessentially

566
00:20:39,669 --> 00:20:36,799
inactive you know it's not just a

567
00:20:41,430 --> 00:20:39,679
um you know a good computer it can be

568
00:20:45,270 --> 00:20:41,440
made a good computer just by removing

569
00:20:47,350 --> 00:20:45,280
active states if you want to um but that

570
00:20:49,909 --> 00:20:47,360
you know that would be a limiting and a

571
00:20:51,909 --> 00:20:49,919
a rather um

572
00:20:55,750 --> 00:20:51,919
biophysically unrealizable and

573
00:20:58,950 --> 00:20:55,760
unsustainable kind of artifact

574
00:21:01,669 --> 00:20:58,960
the whole point here is that there is a

575
00:21:03,510 --> 00:21:01,679
reciprocal exchange with the world so

576
00:21:04,870 --> 00:21:03,520
your notion of having two black boxes

577
00:21:07,029 --> 00:21:04,880
talking to each other i think is a

578
00:21:08,230 --> 00:21:07,039
really nice notion um

579
00:21:09,909 --> 00:21:08,240
and indeed

580
00:21:11,350 --> 00:21:09,919
when

581
00:21:13,029 --> 00:21:11,360
applying

582
00:21:14,470 --> 00:21:13,039
this principle of least action the free

583
00:21:18,950 --> 00:21:14,480
energy principle

584
00:21:21,830 --> 00:21:18,960
to interesting scenarios um one ends up

585
00:21:25,350 --> 00:21:21,840
very often um putting two

586
00:21:27,669 --> 00:21:25,360
um sets of uh internal states each

587
00:21:29,990 --> 00:21:27,679
equipped with markov blankets talking to

588
00:21:33,430 --> 00:21:30,000
each other so the active states of one

589
00:21:35,990 --> 00:21:33,440
blanket or one particle or one person

590
00:21:39,190 --> 00:21:36,000
now constitute the sensory states of the

591
00:21:41,830 --> 00:21:39,200
other and vice versa so you've got now

592
00:21:46,230 --> 00:21:41,840
two systems trying to model each other

593
00:21:47,430 --> 00:21:46,240
and if their game is to effectively um

594
00:21:50,950 --> 00:21:47,440
provide

595
00:21:53,430 --> 00:21:50,960
or solicit the most evidence

596
00:21:55,350 --> 00:21:53,440
for their model of the rest of the world

597
00:21:57,669 --> 00:21:55,360
which is just another

598
00:21:59,669 --> 00:21:57,679
artifact very much like you who's doing

599
00:22:01,830 --> 00:21:59,679
trying to do exactly the same thing

600
00:22:03,029 --> 00:22:01,840
what you are you what will inevitably

601
00:22:05,750 --> 00:22:03,039
emerge

602
00:22:08,870 --> 00:22:05,760
from that joint minimization of free

603
00:22:11,510 --> 00:22:08,880
energy or the maximization of marginal

604
00:22:14,630 --> 00:22:11,520
likelihood is the kind of active

605
00:22:16,950 --> 00:22:14,640
engagement that renders say you and me

606
00:22:19,350 --> 00:22:16,960
as the two black boxes renders us

607
00:22:21,270 --> 00:22:19,360
mutually predictable so what we will do

608
00:22:23,990 --> 00:22:21,280
is we will come to share the same

609
00:22:25,909 --> 00:22:24,000
generative model the same narrative so

610
00:22:27,430 --> 00:22:25,919
that i can predict exactly what you're

611
00:22:29,350 --> 00:22:27,440
going to say next

612
00:22:31,270 --> 00:22:29,360
given that's the kind of thing that i

613
00:22:33,029 --> 00:22:31,280
would say because that's the sort of

614
00:22:35,590 --> 00:22:33,039
world that i'm trying to model

615
00:22:36,870 --> 00:22:35,600
so immediately you have this

616
00:22:37,669 --> 00:22:36,880
picture

617
00:22:44,950 --> 00:22:37,679
of

618
00:22:47,430 --> 00:22:44,960
variational free energy or the um

619
00:22:50,390 --> 00:22:47,440
the maximization of the joint marginal

620
00:22:52,830 --> 00:22:50,400
likelihood for shared models and one

621
00:22:55,110 --> 00:22:52,840
gets from that sort of quite compelling

622
00:22:57,430 --> 00:22:55,120
interesting um

623
00:22:59,029 --> 00:22:57,440
formulations of multi-agent games

624
00:23:03,190 --> 00:22:59,039
communication

625
00:23:05,590 --> 00:23:03,200
synchronization between uh con specifics

626
00:23:08,230 --> 00:23:05,600
that all rest upon

627
00:23:10,710 --> 00:23:08,240
trying to resolve uncertainty

628
00:23:13,190 --> 00:23:10,720
in the service of securing actively

629
00:23:15,270 --> 00:23:13,200
eliciting asking questions of the other

630
00:23:17,430 --> 00:23:15,280
black box you asking me questions me

631
00:23:19,590 --> 00:23:17,440
asking you questions

632
00:23:21,510 --> 00:23:19,600
just so i can understand what's going on

633
00:23:23,669 --> 00:23:21,520
out there where you are going on out

634
00:23:25,830 --> 00:23:23,679
there

635
00:23:28,230 --> 00:23:25,840
can any system let's say it's more than

636
00:23:29,510 --> 00:23:28,240
two so in the example that i gave and

637
00:23:31,270 --> 00:23:29,520
that we were talking about there's one

638
00:23:33,029 --> 00:23:31,280
cell and then the external world but

639
00:23:34,549 --> 00:23:33,039
obviously the world is much more complex

640
00:23:35,830 --> 00:23:34,559
than that and let's just imagine

641
00:23:38,149 --> 00:23:35,840
slightly more complicated there's two

642
00:23:40,230 --> 00:23:38,159
cells and then an external world

643
00:23:42,390 --> 00:23:40,240
is it always the case that

644
00:23:43,750 --> 00:23:42,400
from the perspective of one cell you may

645
00:23:45,270 --> 00:23:43,760
as well just

646
00:23:47,430 --> 00:23:45,280
treat the external world even though it

647
00:23:49,590 --> 00:23:47,440
comprises an external world plus another

648
00:23:52,230 --> 00:23:49,600
cell out there can you always treat that

649
00:23:54,310 --> 00:23:52,240
as its own singular black box such that

650
00:23:56,070 --> 00:23:54,320
we can always understand the dynamics as

651
00:23:58,070 --> 00:23:56,080
two black boxes or is there ever a need

652
00:24:00,230 --> 00:23:58,080
for a third a fourth a fifth

653
00:24:02,870 --> 00:24:00,240
i i think i think there probably would

654
00:24:05,110 --> 00:24:02,880
be a need to go if out there if the

655
00:24:07,110 --> 00:24:05,120
external states

656
00:24:10,549 --> 00:24:07,120
are themselves

657
00:24:12,710 --> 00:24:10,559
um partitioned into things so if my if

658
00:24:13,990 --> 00:24:12,720
my universe

659
00:24:17,269 --> 00:24:14,000
is

660
00:24:21,110 --> 00:24:17,279
composed of external states that can

661
00:24:24,470 --> 00:24:21,120
themselves be carved into natural kinds

662
00:24:27,029 --> 00:24:24,480
like other artifacts and objects and air

663
00:24:29,590 --> 00:24:27,039
planets liquids

664
00:24:31,990 --> 00:24:29,600
natural kinds as soon as you talk about

665
00:24:33,909 --> 00:24:32,000
a natural kind you're talking about a

666
00:24:35,510 --> 00:24:33,919
kind of thing as soon as there is a

667
00:24:37,750 --> 00:24:35,520
thing that's in play there's a mark of

668
00:24:39,830 --> 00:24:37,760
blankets so what that tells you is in

669
00:24:42,070 --> 00:24:39,840
principle it was possible to take

670
00:24:43,430 --> 00:24:42,080
all the states that constitute say an

671
00:24:44,789 --> 00:24:43,440
eco niche

672
00:24:46,390 --> 00:24:44,799
and this

673
00:24:49,269 --> 00:24:46,400
find identify

674
00:24:51,510 --> 00:24:49,279
the thing of interest say say me

675
00:24:52,230 --> 00:24:51,520
and my internal states and so and then i

676
00:24:53,990 --> 00:24:52,240
would

677
00:24:55,830 --> 00:24:54,000
identify my blanket states and then i

678
00:24:57,909 --> 00:24:55,840
got the rest of the states but now i can

679
00:25:00,549 --> 00:24:57,919
start again i can find another set of

680
00:25:02,390 --> 00:25:00,559
internal states and identify its blanket

681
00:25:04,710 --> 00:25:02,400
states and then take those states off

682
00:25:08,230 --> 00:25:04,720
the table and then start again and

683
00:25:11,750 --> 00:25:08,240
recursively tile all the external states

684
00:25:14,149 --> 00:25:11,760
so that now i've eliminated effectively

685
00:25:16,230 --> 00:25:14,159
external states and replaced them by the

686
00:25:17,510 --> 00:25:16,240
internal and the blanket states of all

687
00:25:19,350 --> 00:25:17,520
other things

688
00:25:21,990 --> 00:25:19,360
and in on that view

689
00:25:23,029 --> 00:25:22,000
what happens is that because

690
00:25:24,310 --> 00:25:23,039
there's a

691
00:25:25,830 --> 00:25:24,320
um

692
00:25:29,510 --> 00:25:25,840
a statistical

693
00:25:31,350 --> 00:25:29,520
insulation or separation of internal and

694
00:25:33,269 --> 00:25:31,360
external states by the blanket states

695
00:25:35,269 --> 00:25:33,279
that means i can never see your internal

696
00:25:36,870 --> 00:25:35,279
states i don't need to know your

697
00:25:39,110 --> 00:25:36,880
internal states

698
00:25:41,510 --> 00:25:39,120
because everything that is

699
00:25:43,830 --> 00:25:41,520
knowable about your internal states is

700
00:25:46,149 --> 00:25:43,840
on the surface on your blanket states

701
00:25:50,149 --> 00:25:46,159
what so now the picture that emerges is

702
00:25:50,950 --> 00:25:50,159
an ensemble of things natural kinds

703
00:25:53,909 --> 00:25:50,960
um

704
00:25:56,630 --> 00:25:53,919
that are coupled to each other

705
00:26:00,230 --> 00:25:56,640
via their blanket states

706
00:26:01,269 --> 00:26:00,240
no no no individual particle or person

707
00:26:02,149 --> 00:26:01,279
or thing

708
00:26:04,789 --> 00:26:02,159
um

709
00:26:06,710 --> 00:26:04,799
will ever be have access to the internal

710
00:26:08,870 --> 00:26:06,720
states of anything else

711
00:26:11,269 --> 00:26:08,880
but there will be

712
00:26:13,750 --> 00:26:11,279
coupling and influences and mediated by

713
00:26:15,029 --> 00:26:13,760
the blanket states the active states um

714
00:26:16,310 --> 00:26:15,039
and possibly the sensory states

715
00:26:18,549 --> 00:26:16,320
mathematically you can actually have the

716
00:26:20,549 --> 00:26:18,559
sensory states also influencing the

717
00:26:22,870 --> 00:26:20,559
outside but for simplicity let's let's

718
00:26:25,669 --> 00:26:22,880
assume that sensory states do not

719
00:26:27,350 --> 00:26:25,679
influence external states uh it is just

720
00:26:29,269 --> 00:26:27,360
the role of active states to actually

721
00:26:30,710 --> 00:26:29,279
couple back to the outside you mentioned

722
00:26:31,990 --> 00:26:30,720
something extremely interesting and i

723
00:26:34,390 --> 00:26:32,000
want to pick up on this you said that

724
00:26:36,149 --> 00:26:34,400
the sensory states can influence the

725
00:26:37,669 --> 00:26:36,159
external states at least that's not how

726
00:26:39,990 --> 00:26:37,679
i understand it so how can a sensory

727
00:26:42,870 --> 00:26:40,000
state influence an external state so i

728
00:26:46,310 --> 00:26:42,880
said that um mathematically it is that

729
00:26:48,310 --> 00:26:46,320
is possible um so if you if you think

730
00:26:51,510 --> 00:26:48,320
about um

731
00:26:53,669 --> 00:26:51,520
your from a pure from the point of view

732
00:26:56,950 --> 00:26:53,679
of um

733
00:26:58,950 --> 00:26:56,960
somebody who's trying to understand the

734
00:27:00,310 --> 00:26:58,960
statistical behavior the probabilistic

735
00:27:03,909 --> 00:27:00,320
behavior

736
00:27:07,110 --> 00:27:03,919
of a set of dynamics described as

737
00:27:09,190 --> 00:27:07,120
differential equations or your coupled

738
00:27:12,230 --> 00:27:09,200
systems

739
00:27:13,110 --> 00:27:12,240
then the name of the game really is to

740
00:27:16,390 --> 00:27:13,120
ask

741
00:27:19,590 --> 00:27:16,400
your what statistical dependencies that

742
00:27:22,390 --> 00:27:19,600
define a markov blanket specifically

743
00:27:24,070 --> 00:27:22,400
a markov blanket renders the internal

744
00:27:26,789 --> 00:27:24,080
states conditionally independent of the

745
00:27:29,190 --> 00:27:26,799
external states given the markov blanket

746
00:27:31,190 --> 00:27:29,200
so that's their role that's what that's

747
00:27:33,830 --> 00:27:31,200
what is implied

748
00:27:36,310 --> 00:27:33,840
by a separation of the inside from the

749
00:27:37,990 --> 00:27:36,320
outside that all you need to know

750
00:27:39,830 --> 00:27:38,000
is the blanket states because given the

751
00:27:42,549 --> 00:27:39,840
blanket states the inside and the

752
00:27:45,350 --> 00:27:42,559
outside are condition independent

753
00:27:49,029 --> 00:27:45,360
um so the question now is what kind of

754
00:27:49,990 --> 00:27:49,039
sparse coupling among all the states at

755
00:27:51,750 --> 00:27:50,000
hand

756
00:27:54,470 --> 00:27:51,760
give rise to

757
00:27:56,710 --> 00:27:54,480
that conditional independence that

758
00:27:57,909 --> 00:27:56,720
definitively specifies the markov

759
00:28:00,789 --> 00:27:57,919
blanket

760
00:28:03,669 --> 00:28:00,799
it so happens that all you need is the

761
00:28:06,710 --> 00:28:03,679
following rule

762
00:28:10,230 --> 00:28:06,720
internal states can only influence

763
00:28:12,710 --> 00:28:10,240
themselves and active states

764
00:28:15,909 --> 00:28:12,720
external states can only influence

765
00:28:18,870 --> 00:28:15,919
themselves and sensory states that's all

766
00:28:21,990 --> 00:28:18,880
you need so with that

767
00:28:24,710 --> 00:28:22,000
condition in place you now immediately

768
00:28:25,830 --> 00:28:24,720
have a markov blanket or a particular

769
00:28:27,510 --> 00:28:25,840
partition

770
00:28:29,430 --> 00:28:27,520
from that follows the conditional

771
00:28:31,669 --> 00:28:29,440
independence and therefore the

772
00:28:34,789 --> 00:28:31,679
interpretation of the

773
00:28:36,870 --> 00:28:34,799
um autonomous dynamics which is

774
00:28:38,630 --> 00:28:36,880
a description of the trajectories or the

775
00:28:41,029 --> 00:28:38,640
paths of the processes

776
00:28:42,549 --> 00:28:41,039
taken by the active and the internal

777
00:28:45,909 --> 00:28:42,559
states of a particular

778
00:28:47,350 --> 00:28:45,919
of a particle um

779
00:28:49,909 --> 00:28:47,360
so

780
00:28:51,990 --> 00:28:49,919
that means that it is actually

781
00:28:54,230 --> 00:28:52,000
mathematically allowable

782
00:28:56,230 --> 00:28:54,240
in the sense that you don't you do not

783
00:28:58,789 --> 00:28:56,240
destroy that

784
00:29:00,710 --> 00:28:58,799
that markov blanket or the statistical

785
00:29:02,549 --> 00:29:00,720
properties of that markov blanket it is

786
00:29:04,549 --> 00:29:02,559
allowable for the sensory states to

787
00:29:06,789 --> 00:29:04,559
influence uh active states and you may

788
00:29:08,470 --> 00:29:06,799
be thinking mom that's a bit silly you

789
00:29:12,230 --> 00:29:08,480
know there's no way in which my

790
00:29:15,190 --> 00:29:12,240
photoreceptors or my uh you know my the

791
00:29:16,310 --> 00:29:15,200
cochlear implant materially affect the

792
00:29:18,230 --> 00:29:16,320
outside

793
00:29:19,909 --> 00:29:18,240
i think that's absolutely true so for

794
00:29:20,789 --> 00:29:19,919
things like you and me

795
00:29:24,149 --> 00:29:20,799
um

796
00:29:26,710 --> 00:29:24,159
then you would not normally expect the

797
00:29:27,669 --> 00:29:26,720
sensory states to influence the external

798
00:29:29,990 --> 00:29:27,679
states

799
00:29:31,590 --> 00:29:30,000
but that's not the case for much simpler

800
00:29:32,630 --> 00:29:31,600
things

801
00:29:36,470 --> 00:29:32,640
when

802
00:29:39,669 --> 00:29:36,480
where

803
00:29:41,350 --> 00:29:39,679
most of the dynamical influences the

804
00:29:42,789 --> 00:29:41,360
coupling that we're talking about you'd

805
00:29:45,110 --> 00:29:42,799
write down in terms of the differential

806
00:29:46,070 --> 00:29:45,120
equations the launch of an equation

807
00:29:48,870 --> 00:29:46,080
um

808
00:29:50,470 --> 00:29:48,880
most of those influences are basically

809
00:29:53,590 --> 00:29:50,480
short range

810
00:29:55,669 --> 00:29:53,600
so you know we can preclude action at a

811
00:29:56,549 --> 00:29:55,679
distance for example if you're a single

812
00:29:58,870 --> 00:29:56,559
cell

813
00:30:01,269 --> 00:29:58,880
you know unless you're next to me i

814
00:30:03,190 --> 00:30:01,279
don't really know what you're up to

815
00:30:04,870 --> 00:30:03,200
i'm having to exclude here all sorts of

816
00:30:07,269 --> 00:30:04,880
interesting things about long-range

817
00:30:08,870 --> 00:30:07,279
electrochemical communication and

818
00:30:12,710 --> 00:30:08,880
electrical gradients and the like but

819
00:30:15,830 --> 00:30:12,720
let's just take um a bag of cells um

820
00:30:16,870 --> 00:30:15,840
that um that just diffuse stuff locally

821
00:30:18,310 --> 00:30:16,880
they just

822
00:30:20,230 --> 00:30:18,320
they touch each other and that's the

823
00:30:22,549 --> 00:30:20,240
only way that they can that they can

824
00:30:25,350 --> 00:30:22,559
influence each other is by being

825
00:30:27,590 --> 00:30:25,360
proximate in some euclidean space so

826
00:30:29,990 --> 00:30:27,600
everything short range now in this

827
00:30:32,950 --> 00:30:30,000
situation you have a very different kind

828
00:30:35,590 --> 00:30:32,960
of structure so here the cell the

829
00:30:37,669 --> 00:30:35,600
sensory states are the cell surface and

830
00:30:39,750 --> 00:30:37,679
when these active states that lie

831
00:30:41,990 --> 00:30:39,760
underneath the cell surface

832
00:30:43,990 --> 00:30:42,000
change the environment they do so by

833
00:30:46,789 --> 00:30:44,000
pushing the cell surface

834
00:30:49,669 --> 00:30:46,799
into the external milieu and cause

835
00:30:51,669 --> 00:30:49,679
changes in the sensory states the

836
00:30:52,389 --> 00:30:51,679
surface states of other cells

837
00:30:55,510 --> 00:30:52,399
so

838
00:30:57,510 --> 00:30:55,520
being a cell is probably a good example

839
00:30:59,269 --> 00:30:57,520
of where your sensory states which just

840
00:31:01,990 --> 00:30:59,279
are your surface states

841
00:31:04,950 --> 00:31:02,000
um actually do all the heavy lifting in

842
00:31:06,230 --> 00:31:04,960
terms of changing the outside the

843
00:31:10,310 --> 00:31:06,240
spatial

844
00:31:11,909 --> 00:31:10,320
cells that you're co-inhabiting a

845
00:31:13,269 --> 00:31:11,919
particular organ

846
00:31:14,310 --> 00:31:13,279
for example

847
00:31:15,350 --> 00:31:14,320
so

848
00:31:17,590 --> 00:31:15,360
you know it's a little bit

849
00:31:20,950 --> 00:31:17,600
counter-intuitive um but i think a

850
00:31:23,350 --> 00:31:20,960
really interesting um sort of you know

851
00:31:25,669 --> 00:31:23,360
thought experiment um and actually a

852
00:31:26,630 --> 00:31:25,679
practical uh consideration when it comes

853
00:31:29,990 --> 00:31:26,640
to

854
00:31:32,870 --> 00:31:30,000
looking at the different ways in which

855
00:31:36,389 --> 00:31:32,880
insides couple to outsides and the

856
00:31:37,830 --> 00:31:36,399
nature of interactions between the

857
00:31:40,310 --> 00:31:37,840
internal states of something in the

858
00:31:41,990 --> 00:31:40,320
external states of something um and we

859
00:31:44,230 --> 00:31:42,000
sort of take it for granted that we can

860
00:31:47,110 --> 00:31:44,240
do action at a distance you know uh

861
00:31:49,269 --> 00:31:47,120
literally i can talk to you in in canada

862
00:31:50,950 --> 00:31:49,279
while you're talking while i'm in london

863
00:31:53,350 --> 00:31:50,960
um you know that

864
00:31:55,350 --> 00:31:53,360
that that is that is a beautiful example

865
00:31:58,149 --> 00:31:55,360
of this action to distance

866
00:32:00,470 --> 00:31:58,159
um that means that you know we have

867
00:32:02,950 --> 00:32:00,480
quite a special

868
00:32:05,990 --> 00:32:02,960
kind of coupling between our sensory

869
00:32:07,669 --> 00:32:06,000
states and our active states that is not

870
00:32:09,029 --> 00:32:07,679
um

871
00:32:11,830 --> 00:32:09,039
a gift

872
00:32:14,710 --> 00:32:11,840
of very simple um cellular-like

873
00:32:18,310 --> 00:32:14,720
structures or automata that can't

874
00:32:20,789 --> 00:32:18,320
interpret signals um or be subject to or

875
00:32:23,830 --> 00:32:20,799
be sensitive to

876
00:32:26,070 --> 00:32:23,840
influences uh at a distance just so you

877
00:32:28,070 --> 00:32:26,080
know or for people who are listening and

878
00:32:29,909 --> 00:32:28,080
if this wasn't obvious enough the free

879
00:32:32,070 --> 00:32:29,919
energy principle is famous for or

880
00:32:33,990 --> 00:32:32,080
perhaps infamous for being somewhat

881
00:32:35,830 --> 00:32:34,000
inscrutable for people who have a

882
00:32:38,230 --> 00:32:35,840
physics background or math background

883
00:32:39,990 --> 00:32:38,240
what would be the minimum prerequisites

884
00:32:41,750 --> 00:32:40,000
this is how i like to conceptualize the

885
00:32:43,430 --> 00:32:41,760
podcast before i interview someone i

886
00:32:44,710 --> 00:32:43,440
think about okay what are all the pieces

887
00:32:46,789 --> 00:32:44,720
of knowledge i need to know prior to

888
00:32:48,310 --> 00:32:46,799
interviewing them so obvious example is

889
00:32:49,830 --> 00:32:48,320
let's say qft you need to understand

890
00:32:51,750 --> 00:32:49,840
classical mechanics or invariants

891
00:32:52,950 --> 00:32:51,760
quantum mechanics perhaps bundle theory

892
00:32:55,590 --> 00:32:52,960
if you want to understand young mills

893
00:32:57,990 --> 00:32:55,600
and so on properly so what would be the

894
00:32:59,750 --> 00:32:58,000
prerequisites for someone to properly

895
00:33:01,590 --> 00:32:59,760
understand the free energy principle

896
00:33:03,269 --> 00:33:01,600
let's say if they have a math or physics

897
00:33:04,630 --> 00:33:03,279
background

898
00:33:07,110 --> 00:33:04,640
it wouldn't be

899
00:33:10,149 --> 00:33:07,120
it would be probably much less than

900
00:33:12,549 --> 00:33:10,159
you'd expect to encounter um a

901
00:33:13,350 --> 00:33:12,559
university i think you certainly need to

902
00:33:18,230 --> 00:33:13,360
know

903
00:33:19,509 --> 00:33:18,240
the basics of variational calculus and

904
00:33:22,310 --> 00:33:19,519
it would be really nice if you

905
00:33:23,590 --> 00:33:22,320
understood where a lagrangian com came

906
00:33:24,710 --> 00:33:23,600
from

907
00:33:30,070 --> 00:33:24,720
and

908
00:33:32,070 --> 00:33:30,080
you know basically you know exactly

909
00:33:34,710 --> 00:33:32,080
hamilton's principle of least action and

910
00:33:36,950 --> 00:33:34,720
how that inherits from path integrals of

911
00:33:39,909 --> 00:33:36,960
lagrangians and

912
00:33:41,990 --> 00:33:39,919
how that arises from um

913
00:33:45,190 --> 00:33:42,000
dynamical systems theory of a very

914
00:33:47,430 --> 00:33:45,200
simple sort where you can write down any

915
00:33:49,269 --> 00:33:47,440
random dynamical system as you know a

916
00:33:51,269 --> 00:33:49,279
stochastic differential equation or a

917
00:33:52,710 --> 00:33:51,279
large van equation

918
00:33:54,710 --> 00:33:52,720
i think

919
00:33:56,470 --> 00:33:54,720
that's probably quite sufficient you

920
00:33:59,190 --> 00:33:56,480
know what um

921
00:34:00,389 --> 00:33:59,200
what you would need i think um

922
00:34:03,590 --> 00:34:00,399
to

923
00:34:07,750 --> 00:34:03,600
appreciate all the interpretational

924
00:34:10,149 --> 00:34:07,760
richness afforded by what happens to

925
00:34:12,550 --> 00:34:10,159
the paths of least action when you have

926
00:34:14,069 --> 00:34:12,560
a mark off blanket you need to know a

927
00:34:15,510 --> 00:34:14,079
little bit of probability theory you

928
00:34:18,230 --> 00:34:15,520
need to know what a conditional

929
00:34:20,470 --> 00:34:18,240
distribution is you need to know uh the

930
00:34:23,270 --> 00:34:20,480
important you know what uh have an

931
00:34:25,349 --> 00:34:23,280
intuition as to what is meant by

932
00:34:27,190 --> 00:34:25,359
model evidence for example or marginal

933
00:34:28,790 --> 00:34:27,200
likelihood so

934
00:34:32,149 --> 00:34:28,800
you know whether

935
00:34:34,069 --> 00:34:32,159
i would imagine you know a week's um

936
00:34:36,550 --> 00:34:34,079
foraging on uh wikipedia you could

937
00:34:39,109 --> 00:34:36,560
probably acquire that if you know if you

938
00:34:40,470 --> 00:34:39,119
felt flu if you were fluent already and

939
00:34:42,950 --> 00:34:40,480
sort of um

940
00:34:45,349 --> 00:34:42,960
uh in path integral like formulations of

941
00:34:47,669 --> 00:34:45,359
the kind you can find on wikipedia so

942
00:34:49,589 --> 00:34:47,679
that you certainly wouldn't need

943
00:34:51,030 --> 00:34:49,599
you know what sort of bundles were you

944
00:34:52,149 --> 00:34:51,040
wouldn't need category theory you

945
00:34:54,149 --> 00:34:52,159
wouldn't need to know anything about

946
00:34:55,669 --> 00:34:54,159
gauge theories um

947
00:34:56,550 --> 00:34:55,679
you you know you wouldn't even you

948
00:34:57,990 --> 00:34:56,560
wouldn't even

949
00:35:01,430 --> 00:34:58,000
you need to know about lagrange

950
00:35:02,390 --> 00:35:01,440
multipliers yeah but you need to know um

951
00:35:05,510 --> 00:35:02,400
you know

952
00:35:08,790 --> 00:35:05,520
the basics of how to uh

953
00:35:10,950 --> 00:35:08,800
how how to understand um differential

954
00:35:12,870 --> 00:35:10,960
equations with you know with random

955
00:35:15,589 --> 00:35:12,880
fluctuations with a probabilistic or

956
00:35:18,230 --> 00:35:15,599
stochastic aspect to them and how you

957
00:35:20,470 --> 00:35:18,240
can understand paths of least action uh

958
00:35:22,150 --> 00:35:20,480
in these kinds of systems your example

959
00:35:24,790 --> 00:35:22,160
of wikipedia is

960
00:35:26,630 --> 00:35:24,800
is on point because the way that i

961
00:35:28,630 --> 00:35:26,640
conceptualized this interview this

962
00:35:31,829 --> 00:35:28,640
podcast and the way that we spoke about

963
00:35:33,589 --> 00:35:31,839
it prior to going on air via emails is

964
00:35:36,470 --> 00:35:33,599
it's going to be tendril-like or

965
00:35:37,829 --> 00:35:36,480
gossamer-like where you have these words

966
00:35:39,349 --> 00:35:37,839
i'm imagining what you're saying is like

967
00:35:41,670 --> 00:35:39,359
a wikipedia page and you say certain

968
00:35:43,910 --> 00:35:41,680
keywords which would be if you were a

969
00:35:45,589 --> 00:35:43,920
wikipedia page underlined in blue and

970
00:35:47,270 --> 00:35:45,599
worthy of their own investigation in and

971
00:35:50,390 --> 00:35:47,280
of themselves so some of these i'm going

972
00:35:51,990 --> 00:35:50,400
to pick out and interrupt you

973
00:35:53,270 --> 00:35:52,000
sometimes and i apologize for that like

974
00:35:55,750 --> 00:35:53,280
i did earlier

975
00:36:00,069 --> 00:35:55,760
in order to

976
00:36:02,230 --> 00:36:00,079
prerequisites because we had a four hour

977
00:36:03,510 --> 00:36:02,240
behemoth before and that was meant to

978
00:36:05,430 --> 00:36:03,520
serve as an introduction to the free

979
00:36:07,910 --> 00:36:05,440
energy principle however there were some

980
00:36:09,430 --> 00:36:07,920
people who were still uncertain as to

981
00:36:11,589 --> 00:36:09,440
exactly what the free energy principle

982
00:36:13,270 --> 00:36:11,599
is so this will serve as an introduction

983
00:36:15,030 --> 00:36:13,280
to that introduction

984
00:36:17,109 --> 00:36:15,040
okay

985
00:36:19,430 --> 00:36:17,119
you mentioned marginal likelihood slash

986
00:36:21,270 --> 00:36:19,440
model evidencing and i believe you see

987
00:36:23,030 --> 00:36:21,280
those two as the same and i'm unclear as

988
00:36:24,470 --> 00:36:23,040
to what any of those two are so can you

989
00:36:27,829 --> 00:36:24,480
give an explanation what is model

990
00:36:30,230 --> 00:36:27,839
evidencing marginal likelihood

991
00:36:33,109 --> 00:36:30,240
excellent question so this is where your

992
00:36:34,470 --> 00:36:33,119
you would need to be slightly fluent um

993
00:36:35,990 --> 00:36:34,480
in the

994
00:36:37,510 --> 00:36:36,000
the concepts and constructs that

995
00:36:39,030 --> 00:36:37,520
statisticians

996
00:36:41,670 --> 00:36:39,040
would bring to the table

997
00:36:45,190 --> 00:36:41,680
so um

998
00:36:46,230 --> 00:36:45,200
if i now um want to use probability

999
00:36:49,990 --> 00:36:46,240
theory

1000
00:36:52,790 --> 00:36:50,000
um to articulate the

1001
00:36:55,190 --> 00:36:52,800
the job of a statistician

1002
00:36:57,990 --> 00:36:55,200
trying to make sense of data

1003
00:36:59,750 --> 00:36:58,000
um and just know that that sense making

1004
00:37:01,990 --> 00:36:59,760
is basically the interpretation we're

1005
00:37:06,829 --> 00:37:02,000
putting on these paths of least action

1006
00:37:10,790 --> 00:37:06,839
um then you you you'd start off by by

1007
00:37:12,310 --> 00:37:10,800
um understanding the problem in terms of

1008
00:37:15,109 --> 00:37:12,320
um

1009
00:37:18,630 --> 00:37:15,119
estimating under uncertainty sort of

1010
00:37:21,270 --> 00:37:18,640
assimilating data whilst

1011
00:37:24,710 --> 00:37:21,280
quantifying and accommodating

1012
00:37:27,270 --> 00:37:24,720
uncertainty due to random effects so you

1013
00:37:28,150 --> 00:37:27,280
know the random effects here can be read

1014
00:37:30,790 --> 00:37:28,160
as

1015
00:37:32,870 --> 00:37:30,800
um the random fluctuations of the vena

1016
00:37:34,550 --> 00:37:32,880
fluctuations or the innovations in some

1017
00:37:36,310 --> 00:37:34,560
stochastic process from the point of

1018
00:37:38,950 --> 00:37:36,320
view of the physicist from the point of

1019
00:37:42,790 --> 00:37:38,960
view of statistician

1020
00:37:45,270 --> 00:37:42,800
those that randomness um is inherited as

1021
00:37:47,270 --> 00:37:45,280
random effects observation noise for

1022
00:37:50,390 --> 00:37:47,280
example would be one if you were dealing

1023
00:37:52,390 --> 00:37:50,400
with a state space model which is really

1024
00:37:54,790 --> 00:37:52,400
a model of a dynamical process as the

1025
00:37:57,589 --> 00:37:54,800
statistician might use to do data

1026
00:37:59,990 --> 00:37:57,599
simulation given some given some time

1027
00:38:04,230 --> 00:38:00,000
serious data there may be a distinction

1028
00:38:07,510 --> 00:38:04,240
between observation noise or

1029
00:38:10,069 --> 00:38:07,520
sensor noise and the noise or the random

1030
00:38:11,910 --> 00:38:10,079
fluctuations on the states that can't be

1031
00:38:15,109 --> 00:38:11,920
directly observed that would be known as

1032
00:38:16,870 --> 00:38:15,119
state noise or system noise for example

1033
00:38:18,230 --> 00:38:16,880
they both you're

1034
00:38:19,910 --> 00:38:18,240
describing the same thing but from the

1035
00:38:22,390 --> 00:38:19,920
point of view of the statistician these

1036
00:38:25,190 --> 00:38:22,400
these this is becomes noise or

1037
00:38:26,390 --> 00:38:25,200
uncertainty or random effects

1038
00:38:29,349 --> 00:38:26,400
so how do you deal with this but

1039
00:38:30,550 --> 00:38:29,359
basically um you're confronted with the

1040
00:38:34,069 --> 00:38:30,560
problem of

1041
00:38:35,109 --> 00:38:34,079
some data and you have a hypothesis or a

1042
00:38:38,550 --> 00:38:35,119
model

1043
00:38:41,109 --> 00:38:38,560
about how those data were generated

1044
00:38:43,589 --> 00:38:41,119
by some latent states

1045
00:38:47,510 --> 00:38:43,599
sometimes known as hidden states

1046
00:38:49,270 --> 00:38:47,520
and hidden um here i think uh

1047
00:38:51,190 --> 00:38:49,280
just as to try and connect it to the

1048
00:38:53,270 --> 00:38:51,200
markov blanket they're hidden behind the

1049
00:38:54,550 --> 00:38:53,280
mark of blankets so the statistician can

1050
00:38:57,030 --> 00:38:54,560
if you think of that being in the

1051
00:38:59,270 --> 00:38:57,040
internal states they're surrounded

1052
00:39:02,390 --> 00:38:59,280
they're enshrouded by a markov blanket

1053
00:39:03,829 --> 00:39:02,400
the data is impressing themselves on the

1054
00:39:05,990 --> 00:39:03,839
sensory veil

1055
00:39:07,510 --> 00:39:06,000
there's something on the outside that

1056
00:39:10,550 --> 00:39:07,520
are latent in the sense that they're

1057
00:39:12,150 --> 00:39:10,560
hidden behind this uh this veil so they

1058
00:39:13,270 --> 00:39:12,160
only have access to the data they don't

1059
00:39:16,230 --> 00:39:13,280
know how

1060
00:39:18,390 --> 00:39:16,240
the data were caused so for example they

1061
00:39:20,150 --> 00:39:18,400
might have a hypothesis that this drug

1062
00:39:23,030 --> 00:39:20,160
treatment

1063
00:39:23,870 --> 00:39:23,040
materially affected

1064
00:39:25,510 --> 00:39:23,880
some

1065
00:39:27,990 --> 00:39:25,520
pathophysiology

1066
00:39:30,950 --> 00:39:28,000
that reduced the symptoms that were

1067
00:39:33,270 --> 00:39:30,960
measurable or recordable by the doctor

1068
00:39:35,750 --> 00:39:33,280
but it's only the observable data that

1069
00:39:38,230 --> 00:39:35,760
the doctor can get from the patients

1070
00:39:40,790 --> 00:39:38,240
that the statistician has at hand now

1071
00:39:43,190 --> 00:39:40,800
the statistician now has to infer was

1072
00:39:44,630 --> 00:39:43,200
there a drug effect and what she will do

1073
00:39:47,190 --> 00:39:44,640
she will create

1074
00:39:49,270 --> 00:39:47,200
a generative model that

1075
00:39:52,550 --> 00:39:49,280
if there was a drug effect of a certain

1076
00:39:54,230 --> 00:39:52,560
size then i'd expect a separation of the

1077
00:39:57,109 --> 00:39:54,240
data for example

1078
00:39:59,589 --> 00:39:57,119
now that generative model

1079
00:40:01,829 --> 00:39:59,599
at the end of the day can always be

1080
00:40:03,109 --> 00:40:01,839
written down as a joint probability

1081
00:40:06,069 --> 00:40:03,119
density

1082
00:40:08,870 --> 00:40:06,079
um over the causes and consequences

1083
00:40:11,190 --> 00:40:08,880
where the causes are the latent or the

1084
00:40:13,589 --> 00:40:11,200
hidden states that you are trying to

1085
00:40:16,390 --> 00:40:13,599
estimate and the consequences of the

1086
00:40:18,710 --> 00:40:16,400
observable data that you have at hand

1087
00:40:21,589 --> 00:40:18,720
given that generative model

1088
00:40:23,190 --> 00:40:21,599
you can now estimate um

1089
00:40:25,349 --> 00:40:23,200
given some data

1090
00:40:27,670 --> 00:40:25,359
the most likely

1091
00:40:29,750 --> 00:40:27,680
hidden or latent states or causes of

1092
00:40:31,829 --> 00:40:29,760
that data basically the treatment effect

1093
00:40:33,670 --> 00:40:31,839
that you were hypothesized

1094
00:40:35,750 --> 00:40:33,680
so that's the posterior distribution

1095
00:40:38,470 --> 00:40:35,760
that's the probability

1096
00:40:40,950 --> 00:40:38,480
of some hidden state say the external

1097
00:40:44,790 --> 00:40:40,960
states the markov blanket given

1098
00:40:48,630 --> 00:40:44,800
conditioned upon the data

1099
00:40:50,870 --> 00:40:48,640
so that's the holy grail usually of um a

1100
00:40:51,670 --> 00:40:50,880
statistical analysis and you know what

1101
00:40:54,470 --> 00:40:51,680
your

1102
00:40:58,230 --> 00:40:54,480
most classical uh inference for example

1103
00:41:00,630 --> 00:40:58,240
those based on t-tests or f-tests um uh

1104
00:41:03,910 --> 00:41:00,640
right through to more sophisticated and

1105
00:41:04,950 --> 00:41:03,920
informed uh log-odd ratios such as base

1106
00:41:07,589 --> 00:41:04,960
factors

1107
00:41:09,510 --> 00:41:07,599
these are just scoring

1108
00:41:12,630 --> 00:41:09,520
the evidence

1109
00:41:15,349 --> 00:41:12,640
for an effective interest expressed in

1110
00:41:18,710 --> 00:41:15,359
this latent state space

1111
00:41:20,470 --> 00:41:18,720
in terms of the probability of one model

1112
00:41:23,510 --> 00:41:20,480
relative to another model

1113
00:41:25,270 --> 00:41:23,520
and usually in science that would be the

1114
00:41:27,910 --> 00:41:25,280
alternate hypothesis versus the null

1115
00:41:28,630 --> 00:41:27,920
hypothesis and then you can look at the

1116
00:41:44,630 --> 00:41:28,640
the

1117
00:41:45,510 --> 00:41:44,640
so to come to our example our example of

1118
00:41:48,230 --> 00:41:45,520
um

1119
00:41:51,190 --> 00:41:48,240
you know the statistician trying to

1120
00:41:53,750 --> 00:41:51,200
infer whether the drug has had a

1121
00:41:57,589 --> 00:41:53,760
remedial effect or not what she would do

1122
00:41:58,710 --> 00:41:57,599
is compute the probability of her data

1123
00:42:00,630 --> 00:41:58,720
with

1124
00:42:03,190 --> 00:42:00,640
an effect and then compute the

1125
00:42:05,589 --> 00:42:03,200
probability of her data if there was no

1126
00:42:07,510 --> 00:42:05,599
effect of the drug so the you know the

1127
00:42:09,829 --> 00:42:07,520
data sets from the treatment and the

1128
00:42:12,230 --> 00:42:09,839
non-treatment group were treated as from

1129
00:42:14,550 --> 00:42:12,240
a single group and then she would use

1130
00:42:16,950 --> 00:42:14,560
that to make an inference that there was

1131
00:42:18,630 --> 00:42:16,960
indeed an effect so

1132
00:42:19,829 --> 00:42:18,640
that probability the probability of the

1133
00:42:22,550 --> 00:42:19,839
data

1134
00:42:24,550 --> 00:42:22,560
given a particular model a specific

1135
00:42:26,870 --> 00:42:24,560
model be it the ultimate hypothesis on

1136
00:42:28,630 --> 00:42:26,880
the null hypothesis that's the marginal

1137
00:42:31,109 --> 00:42:28,640
likelihood so why is it called a

1138
00:42:32,870 --> 00:42:31,119
marginal likelihood well what you've

1139
00:42:34,870 --> 00:42:32,880
done is

1140
00:42:35,910 --> 00:42:34,880
you've estimated given your generative

1141
00:42:39,190 --> 00:42:35,920
model

1142
00:42:43,270 --> 00:42:39,200
the probability of the data

1143
00:42:45,190 --> 00:42:43,280
having marginalized or averaged away

1144
00:42:48,950 --> 00:42:45,200
the thing that you don't know which is

1145
00:42:49,829 --> 00:42:48,960
the effect size the treatment effect

1146
00:42:52,150 --> 00:42:49,839
so

1147
00:42:53,910 --> 00:42:52,160
the probability of the data

1148
00:42:57,829 --> 00:42:53,920
given a model

1149
00:43:00,870 --> 00:42:57,839
is the average probability of the data

1150
00:43:03,910 --> 00:43:00,880
given that model and the effect size

1151
00:43:05,430 --> 00:43:03,920
averaged under your beliefs about the

1152
00:43:08,069 --> 00:43:05,440
effect size

1153
00:43:09,829 --> 00:43:08,079
so it that's why it's called a marginal

1154
00:43:10,710 --> 00:43:09,839
a marginal likelihood

1155
00:43:13,349 --> 00:43:10,720
and

1156
00:43:16,069 --> 00:43:13,359
that's a great question because um

1157
00:43:18,550 --> 00:43:16,079
you know it actually um

1158
00:43:20,950 --> 00:43:18,560
leads one to um

1159
00:43:23,589 --> 00:43:20,960
the genesis of variational free energy

1160
00:43:25,430 --> 00:43:23,599
on one reading by uh in the context of

1161
00:43:27,670 --> 00:43:25,440
richard feynman's pathetic formulation

1162
00:43:31,190 --> 00:43:27,680
of quantum electrodynamics

1163
00:43:33,109 --> 00:43:31,200
he he had this marginalization problem

1164
00:43:36,550 --> 00:43:33,119
um so

1165
00:43:39,750 --> 00:43:36,560
he wanted to or one could read the

1166
00:43:43,030 --> 00:43:39,760
history as him wanting

1167
00:43:46,069 --> 00:43:43,040
to evaluate the marginal likelihood

1168
00:43:46,079 --> 00:43:50,950
paths of say small particles

1169
00:43:54,870 --> 00:43:53,109
which would necessarily necessarily

1170
00:43:57,829 --> 00:43:54,880
involve

1171
00:43:59,589 --> 00:43:57,839
computing or marginalizing

1172
00:44:00,870 --> 00:43:59,599
integrating of extremely high

1173
00:44:03,109 --> 00:44:00,880
dimensional

1174
00:44:04,950 --> 00:44:03,119
probability distributions

1175
00:44:07,829 --> 00:44:04,960
which is intractable

1176
00:44:10,230 --> 00:44:07,839
so in order to solve the marginalization

1177
00:44:12,150 --> 00:44:10,240
or the integration problem to ensure

1178
00:44:14,230 --> 00:44:12,160
everything sells to one effectively to

1179
00:44:16,309 --> 00:44:14,240
get your partition function

1180
00:44:18,150 --> 00:44:16,319
he introduced

1181
00:44:19,910 --> 00:44:18,160
the variation free energy

1182
00:44:22,790 --> 00:44:19,920
and that was a really clever move

1183
00:44:24,150 --> 00:44:22,800
because he converted what was an

1184
00:44:27,230 --> 00:44:24,160
intractable

1185
00:44:29,030 --> 00:44:27,240
and non-realizable and incomputable

1186
00:44:32,710 --> 00:44:29,040
marginalization

1187
00:44:34,950 --> 00:44:32,720
or integration problem into a tractable

1188
00:44:37,430 --> 00:44:34,960
optimization problem simply by

1189
00:44:40,150 --> 00:44:37,440
introducing this variational free energy

1190
00:44:41,829 --> 00:44:40,160
which was a band on the marginal

1191
00:44:44,230 --> 00:44:41,839
likelihood

1192
00:44:46,550 --> 00:44:44,240
so that's why it's called an elbow or an

1193
00:44:49,349 --> 00:44:46,560
evidence lower bound simply because it

1194
00:44:51,670 --> 00:44:49,359
now becomes a computable object or

1195
00:44:53,750 --> 00:44:51,680
mathematical functional

1196
00:44:56,470 --> 00:44:53,760
that you can evaluate given some data

1197
00:45:00,069 --> 00:44:56,480
and a generative model that is always in

1198
00:45:02,390 --> 00:45:00,079
machine learning lower than the log of

1199
00:45:04,390 --> 00:45:02,400
the marginal likelihood

1200
00:45:05,510 --> 00:45:04,400
in my world the physics world is the

1201
00:45:08,069 --> 00:45:05,520
other way around but let's stick with

1202
00:45:10,150 --> 00:45:08,079
the machine learning world so the

1203
00:45:11,990 --> 00:45:10,160
negative uh variational free energy of

1204
00:45:14,710 --> 00:45:12,000
the physicist is

1205
00:45:16,710 --> 00:45:14,720
always smaller than the thing you want

1206
00:45:19,750 --> 00:45:16,720
to actually evaluate so if you just

1207
00:45:21,829 --> 00:45:19,760
optimize by pushing up the elbow or the

1208
00:45:23,990 --> 00:45:21,839
variation free energy until you can't

1209
00:45:26,630 --> 00:45:24,000
make it any larger

1210
00:45:28,309 --> 00:45:26,640
and under the assumption that the band

1211
00:45:30,550 --> 00:45:28,319
the tightness of the bound is high

1212
00:45:31,829 --> 00:45:30,560
you've now got a good approximation to

1213
00:45:33,990 --> 00:45:31,839
what you were chasing which is the

1214
00:45:36,069 --> 00:45:34,000
marginal likelihood and once the

1215
00:45:38,069 --> 00:45:36,079
statistician has that marginal

1216
00:45:39,510 --> 00:45:38,079
likelihood she can repeat the process

1217
00:45:41,829 --> 00:45:39,520
for another model say the null

1218
00:45:44,790 --> 00:45:41,839
hypothesis with no treatment effect and

1219
00:45:46,790 --> 00:45:44,800
then she can take the ratio of the free

1220
00:45:48,870 --> 00:45:46,800
energies the variation free energies and

1221
00:45:50,790 --> 00:45:48,880
start to compare the models using base

1222
00:45:54,390 --> 00:45:50,800
factors and then go back to the doctors

1223
00:45:56,470 --> 00:45:54,400
say look um i am

1224
00:45:58,630 --> 00:45:56,480
pretty confident that that drug really

1225
00:46:00,950 --> 00:45:58,640
did have an effect and i can say that

1226
00:46:03,750 --> 00:46:00,960
because the evidence in your data

1227
00:46:06,950 --> 00:46:03,760
suggests that the marginal likelihood of

1228
00:46:08,790 --> 00:46:06,960
a hypothesis or a model that included an

1229
00:46:09,630 --> 00:46:08,800
effective drug

1230
00:46:12,230 --> 00:46:09,640
was

1231
00:46:14,309 --> 00:46:12,240
52.6 times greater than the marginal

1232
00:46:15,829 --> 00:46:14,319
likelihood if i assume there had been no

1233
00:46:17,430 --> 00:46:15,839
uh no change

1234
00:46:20,630 --> 00:46:17,440
so that's where the notion of a marginal

1235
00:46:22,950 --> 00:46:20,640
likelihood comes from um so the marginal

1236
00:46:24,710 --> 00:46:22,960
marginal likelihood just is the evidence

1237
00:46:29,750 --> 00:46:24,720
for a model it just is

1238
00:46:31,750 --> 00:46:29,760
um the um the probability of some data

1239
00:46:33,750 --> 00:46:31,760
you know in the present context you know

1240
00:46:36,470 --> 00:46:33,760
sensory or from the quality of the free

1241
00:46:38,309 --> 00:46:36,480
energy principle the sensory states uh

1242
00:46:39,270 --> 00:46:38,319
or the sensory trajectory of say time

1243
00:46:40,630 --> 00:46:39,280
series

1244
00:46:43,270 --> 00:46:40,640
um

1245
00:46:45,670 --> 00:46:43,280
given a particular model marginalizing

1246
00:46:47,270 --> 00:46:45,680
away all your beliefs about the

1247
00:46:51,349 --> 00:46:47,280
particulars effect size and the

1248
00:46:53,589 --> 00:46:51,359
parameters and all all the um quantities

1249
00:46:57,829 --> 00:46:53,599
that were responsible for generating

1250
00:47:01,190 --> 00:47:00,069
under the structure afforded by by the

1251
00:47:03,430 --> 00:47:01,200
model

1252
00:47:05,670 --> 00:47:03,440
is that is that clear i mean you know

1253
00:47:07,510 --> 00:47:05,680
these are the fundaments of um

1254
00:47:10,390 --> 00:47:07,520
you know machine learning data

1255
00:47:11,349 --> 00:47:10,400
simulations system identification

1256
00:47:12,630 --> 00:47:11,359
and

1257
00:47:16,470 --> 00:47:12,640
you know

1258
00:47:17,270 --> 00:47:16,480
you could actually argue this is one way

1259
00:47:20,309 --> 00:47:17,280
of

1260
00:47:22,829 --> 00:47:20,319
nuancing the scientific um process not

1261
00:47:25,510 --> 00:47:22,839
quite in using the rhetoric

1262
00:47:28,630 --> 00:47:25,520
of um karl popper but certainly has this

1263
00:47:29,589 --> 00:47:28,640
poparian aspect to him that

1264
00:47:32,790 --> 00:47:29,599
every

1265
00:47:35,349 --> 00:47:32,800
bit of scientific inquiry

1266
00:47:38,470 --> 00:47:35,359
including the way we live our lives is

1267
00:47:40,470 --> 00:47:38,480
just about hypothesis testing

1268
00:47:42,790 --> 00:47:40,480
and we just generate

1269
00:47:45,109 --> 00:47:42,800
alternative explanations for the

1270
00:47:46,470 --> 00:47:45,119
evidence at hand and then we try to

1271
00:47:49,589 --> 00:47:46,480
evaluate the evidence for this

1272
00:47:51,270 --> 00:47:49,599
hypothesis relative that hypothesis

1273
00:47:53,589 --> 00:47:51,280
evaluate the marginal likelihood of the

1274
00:47:55,670 --> 00:47:53,599
data at hand for this hypothesis versus

1275
00:47:58,069 --> 00:47:55,680
that hypothesis this model versus that

1276
00:47:59,910 --> 00:47:58,079
model and then we usually commit to the

1277
00:48:03,349 --> 00:47:59,920
model with the greatest evidence this

1278
00:48:06,550 --> 00:48:03,359
just is evidence-based scientific uh

1279
00:48:08,790 --> 00:48:06,560
progress uh um you're having

1280
00:48:10,550 --> 00:48:08,800
committed to or selected

1281
00:48:13,670 --> 00:48:10,560
technically known as bayesian model

1282
00:48:16,150 --> 00:48:13,680
selection uh this hypothesis you then

1283
00:48:17,670 --> 00:48:16,160
think about okay what's the next space

1284
00:48:20,630 --> 00:48:17,680
of hypotheses or what's the next

1285
00:48:22,870 --> 00:48:20,640
portfolio of ideas or models

1286
00:48:24,870 --> 00:48:22,880
that i want to now go and gather

1287
00:48:26,710 --> 00:48:24,880
evidence for

1288
00:48:29,589 --> 00:48:26,720
i think this analogy is quite nice i'm

1289
00:48:33,030 --> 00:48:29,599
glad you've let me down this explanation

1290
00:48:35,030 --> 00:48:33,040
because because a good scientist then

1291
00:48:38,069 --> 00:48:35,040
the good one will generate new

1292
00:48:40,470 --> 00:48:38,079
hypotheses and then the problem

1293
00:48:42,230 --> 00:48:40,480
contending the scientist

1294
00:48:44,630 --> 00:48:42,240
who's been who's just chasing these

1295
00:48:47,670 --> 00:48:44,640
marginal likelihoods or um model

1296
00:48:50,710 --> 00:48:47,680
evidences for her models of of her

1297
00:48:53,270 --> 00:48:50,720
fielding uh of inquiry

1298
00:48:56,549 --> 00:48:53,280
is to design a good experiment

1299
00:48:58,870 --> 00:48:56,559
so here we we we come back to the you

1300
00:48:59,589 --> 00:48:58,880
know the other side of the coin

1301
00:49:01,670 --> 00:48:59,599
that

1302
00:49:04,150 --> 00:49:01,680
takes us um

1303
00:49:05,990 --> 00:49:04,160
beyond machine learning as a black box

1304
00:49:08,390 --> 00:49:06,000
sort of you know give it some inputs

1305
00:49:10,309 --> 00:49:08,400
give it an output we now come back to

1306
00:49:12,790 --> 00:49:10,319
the design of our experiments how we

1307
00:49:16,630 --> 00:49:12,800
actively secure and solicit

1308
00:49:19,030 --> 00:49:16,640
the evidence that will best disambiguate

1309
00:49:20,950 --> 00:49:19,040
among our hypotheses and i repeat

1310
00:49:23,109 --> 00:49:20,960
usually when you're writing scientific

1311
00:49:25,030 --> 00:49:23,119
papers it's a null hypothesis there's an

1312
00:49:27,430 --> 00:49:25,040
alternate hypothesis but in real life

1313
00:49:29,829 --> 00:49:27,440
and in real scientific thinking people

1314
00:49:32,549 --> 00:49:29,839
you normally have eight hypotheses or a

1315
00:49:35,750 --> 00:49:32,559
hundred hypotheses um

1316
00:49:38,870 --> 00:49:35,760
and they're scoring all of them um

1317
00:49:41,349 --> 00:49:38,880
by evaluating the marginal likelihood of

1318
00:49:43,030 --> 00:49:41,359
the data that they have carefully

1319
00:49:45,910 --> 00:49:43,040
solicited by

1320
00:49:48,870 --> 00:49:45,920
bayes optimal experimental design to

1321
00:49:49,990 --> 00:49:48,880
maximally disambiguate in terms of the

1322
00:49:53,109 --> 00:49:50,000
marginal likelihood of the model

1323
00:49:53,910 --> 00:49:53,119
evidence between all of their hypotheses

1324
00:49:55,510 --> 00:49:53,920
so

1325
00:49:59,270 --> 00:49:55,520
that um

1326
00:50:01,670 --> 00:49:59,280
that sort of brings another aspect of

1327
00:50:03,990 --> 00:50:01,680
probability theory to the table

1328
00:50:05,190 --> 00:50:04,000
um which would be an important

1329
00:50:07,510 --> 00:50:05,200
complement

1330
00:50:09,349 --> 00:50:07,520
for somebody wanting to understand

1331
00:50:12,230 --> 00:50:09,359
bayesian statistics

1332
00:50:14,069 --> 00:50:12,240
the bayes aspect of course is is that as

1333
00:50:16,390 --> 00:50:14,079
soon as you start to talk about

1334
00:50:18,470 --> 00:50:16,400
conditional distributions of the kind

1335
00:50:19,910 --> 00:50:18,480
that are marginal likelihood so it's

1336
00:50:22,390 --> 00:50:19,920
conditional in the sense that it's a

1337
00:50:25,030 --> 00:50:22,400
probability of some data conditioned

1338
00:50:27,750 --> 00:50:25,040
upon a model or given a model having

1339
00:50:29,030 --> 00:50:27,760
marginalized away all the parameters of

1340
00:50:31,990 --> 00:50:29,040
of that model

1341
00:50:34,549 --> 00:50:32,000
um and you know building a post area the

1342
00:50:36,069 --> 00:50:34,559
probability of my

1343
00:50:38,309 --> 00:50:36,079
hidden states

1344
00:50:41,190 --> 00:50:38,319
the parameters of my model

1345
00:50:42,470 --> 00:50:41,200
given the data falls out of bayes rule

1346
00:50:44,630 --> 00:50:42,480
so we're talking about

1347
00:50:47,109 --> 00:50:44,640
bayesian inference here and just

1348
00:50:49,670 --> 00:50:47,119
straightforward probability theory uh

1349
00:50:51,670 --> 00:50:49,680
that can be articulated in terms in

1350
00:50:54,470 --> 00:50:51,680
terms of bayes rule but there's another

1351
00:50:56,150 --> 00:50:54,480
really important application of bayes

1352
00:50:58,230 --> 00:50:56,160
rule

1353
00:50:59,589 --> 00:50:58,240
which speaks to this sort of inactive

1354
00:51:02,790 --> 00:50:59,599
situated

1355
00:51:05,430 --> 00:51:02,800
statistician where statistician now

1356
00:51:08,390 --> 00:51:05,440
is not simply in the service of the

1357
00:51:11,510 --> 00:51:08,400
doctor supplying her with data and

1358
00:51:14,390 --> 00:51:11,520
asking a particular question

1359
00:51:16,710 --> 00:51:14,400
now the statistician is in charge of

1360
00:51:18,950 --> 00:51:16,720
designing the perfect experiment that

1361
00:51:20,710 --> 00:51:18,960
gives her the data that allows her to

1362
00:51:23,750 --> 00:51:20,720
make the most efficient

1363
00:51:25,349 --> 00:51:23,760
and statistically powerful

1364
00:51:27,750 --> 00:51:25,359
inference about whether there was a

1365
00:51:29,670 --> 00:51:27,760
treatment effect and that's the

1366
00:51:31,829 --> 00:51:29,680
that's the problem of optimal uh

1367
00:51:34,870 --> 00:51:31,839
experimental design there is an answer

1368
00:51:37,750 --> 00:51:34,880
to that it's called the principle of

1369
00:51:40,150 --> 00:51:37,760
basal bays optimal design uh first

1370
00:51:41,589 --> 00:51:40,160
articulated by people like um uh

1371
00:51:45,030 --> 00:51:41,599
lindelin the

1372
00:51:47,589 --> 00:51:45,040
in the 1950s and then re-articulated by

1373
00:51:49,190 --> 00:51:47,599
people like david mckay in active

1374
00:51:51,510 --> 00:51:49,200
learning

1375
00:51:53,270 --> 00:51:51,520
and it is if you like the other side of

1376
00:51:55,990 --> 00:51:53,280
the bayesian coin

1377
00:51:57,990 --> 00:51:56,000
when it comes to um

1378
00:52:00,390 --> 00:51:58,000
um

1379
00:52:04,309 --> 00:52:00,400
bayesian assimilation of statistical

1380
00:52:06,470 --> 00:52:04,319
assimilation of data and inference um

1381
00:52:08,150 --> 00:52:06,480
and one could also argue it's the other

1382
00:52:09,910 --> 00:52:08,160
side of the coin when it comes to things

1383
00:52:12,870 --> 00:52:09,920
like basic decisions theory as well but

1384
00:52:14,870 --> 00:52:12,880
that that's a slight distraction uh i

1385
00:52:19,109 --> 00:52:14,880
think we just stick with the big idea

1386
00:52:20,790 --> 00:52:19,119
here that you know there are good

1387
00:52:22,470 --> 00:52:20,800
let's look at it like this if it is the

1388
00:52:25,190 --> 00:52:22,480
case that the marginal likelihood is

1389
00:52:26,790 --> 00:52:25,200
just the probability of some data

1390
00:52:28,390 --> 00:52:26,800
given a model

1391
00:52:30,870 --> 00:52:28,400
and your job

1392
00:52:34,390 --> 00:52:30,880
is to evaluate

1393
00:52:37,589 --> 00:52:34,400
that quantity and usually optimize the

1394
00:52:39,990 --> 00:52:37,599
um optimize the uh the marginal

1395
00:52:42,069 --> 00:52:40,000
likelihood or maximize the um the

1396
00:52:43,910 --> 00:52:42,079
evidence for your model you can either

1397
00:52:45,270 --> 00:52:43,920
change the model or you can change the

1398
00:52:47,270 --> 00:52:45,280
data

1399
00:52:50,150 --> 00:52:47,280
and that means that you've got sort of

1400
00:52:51,910 --> 00:52:50,160
two uh two problems to solve as it were

1401
00:52:54,950 --> 00:52:51,920
if you cast this as an optimality

1402
00:52:56,470 --> 00:52:54,960
problem as a maximization of an evidence

1403
00:52:58,230 --> 00:52:56,480
lower bound

1404
00:53:00,309 --> 00:52:58,240
or a minimization of variation free

1405
00:53:01,589 --> 00:53:00,319
energy from a from a feminist point of

1406
00:53:03,430 --> 00:53:01,599
view

1407
00:53:06,549 --> 00:53:03,440
and that's i think comes back to your

1408
00:53:09,190 --> 00:53:06,559
your original uh point about this of two

1409
00:53:10,150 --> 00:53:09,200
black boxes two computers inferring each

1410
00:53:11,510 --> 00:53:10,160
other

1411
00:53:12,950 --> 00:53:11,520
um

1412
00:53:15,990 --> 00:53:12,960
you know where

1413
00:53:17,349 --> 00:53:16,000
there are uh i can either change my mind

1414
00:53:19,510 --> 00:53:17,359
about you

1415
00:53:21,349 --> 00:53:19,520
uh or i can try and change you by

1416
00:53:23,750 --> 00:53:21,359
queering you asking a question

1417
00:53:25,990 --> 00:53:23,760
performing an experiment upon you to get

1418
00:53:28,549 --> 00:53:26,000
some better um

1419
00:53:30,549 --> 00:53:28,559
better data from you that resolves my

1420
00:53:32,549 --> 00:53:30,559
uncertainty about you

1421
00:53:34,150 --> 00:53:32,559
is there a relationship between changing

1422
00:53:37,349 --> 00:53:34,160
your evidence versus changing your model

1423
00:53:39,589 --> 00:53:37,359
and the psychoanalytic repression so

1424
00:53:42,390 --> 00:53:39,599
repression would be saying you know what

1425
00:53:44,150 --> 00:53:42,400
that data doesn't exist let me ignore it

1426
00:53:46,069 --> 00:53:44,160
versus the

1427
00:53:48,230 --> 00:53:46,079
perhaps a cognitive behavioral therapy

1428
00:53:50,230 --> 00:53:48,240
technique would be to say no go out and

1429
00:53:51,829 --> 00:53:50,240
perhaps update your model of it view it

1430
00:53:53,750 --> 00:53:51,839
differently view the maybe it's our

1431
00:53:56,230 --> 00:53:53,760
arachnophobia for example i think it's a

1432
00:53:58,470 --> 00:53:56,240
beautiful analogy and you know not

1433
00:54:01,910 --> 00:53:58,480
probably beyond an analogy there are

1434
00:54:06,549 --> 00:54:01,920
people in computational psychiatry who

1435
00:54:08,950 --> 00:54:06,559
think exactly like that um and you know

1436
00:54:11,109 --> 00:54:08,960
you can find articles trends in

1437
00:54:12,710 --> 00:54:11,119
cognitive science reviews of the car you

1438
00:54:14,470 --> 00:54:12,720
know something to deal with you know how

1439
00:54:16,710 --> 00:54:14,480
do you how do you resolve cognitive

1440
00:54:20,470 --> 00:54:16,720
dissonance how do you understand wishful

1441
00:54:23,030 --> 00:54:20,480
thinking um so examples

1442
00:54:27,190 --> 00:54:23,040
i think that's a perfect example of

1443
00:54:29,510 --> 00:54:27,200
this way of maximizing the evidence for

1444
00:54:31,190 --> 00:54:29,520
your models of the world in that case it

1445
00:54:32,710 --> 00:54:31,200
sounds like at least the colloquial

1446
00:54:34,710 --> 00:54:32,720
understanding maybe in the west or maybe

1447
00:54:35,910 --> 00:54:34,720
throughout the world i'm unsure is that

1448
00:54:37,190 --> 00:54:35,920
you should update your model you

1449
00:54:38,470 --> 00:54:37,200
shouldn't try to change the evidence to

1450
00:54:40,230 --> 00:54:38,480
match your model you should update your

1451
00:54:41,990 --> 00:54:40,240
model and in fact there are some

1452
00:54:43,510 --> 00:54:42,000
political implications i'm sure people

1453
00:54:46,230 --> 00:54:43,520
shout that on the streets across various

1454
00:54:48,390 --> 00:54:46,240
political divides so is this an argument

1455
00:54:50,309 --> 00:54:48,400
to say that updating the model is

1456
00:54:51,910 --> 00:54:50,319
actually the one that should be primary

1457
00:54:53,510 --> 00:54:51,920
because the way that i understand from

1458
00:54:55,510 --> 00:54:53,520
how you've explained it is that both are

1459
00:54:57,030 --> 00:54:55,520
equally valid however when i look across

1460
00:54:58,549 --> 00:54:57,040
the world and i look at how we think

1461
00:55:00,230 --> 00:54:58,559
just in terms of common sense

1462
00:55:01,589 --> 00:55:00,240
psychological advice at least to us in

1463
00:55:04,230 --> 00:55:01,599
the 21st century it sounds like you

1464
00:55:06,309 --> 00:55:04,240
should update your model um well but i

1465
00:55:07,990 --> 00:55:06,319
think you're right to say that you know

1466
00:55:10,309 --> 00:55:08,000
it would be um

1467
00:55:12,309 --> 00:55:10,319
disingenuous from a mathematical point

1468
00:55:14,150 --> 00:55:12,319
of view to privilege the model versus

1469
00:55:16,790 --> 00:55:14,160
the data and

1470
00:55:20,790 --> 00:55:18,470
so how does one choose between those

1471
00:55:22,630 --> 00:55:20,800
mathematically how does one i don't i i

1472
00:55:25,109 --> 00:55:22,640
i don't think you can in the sense you

1473
00:55:27,910 --> 00:55:25,119
know you you can't you can't spit a coin

1474
00:55:32,549 --> 00:55:27,920
a coin has two two sides to it i mean

1475
00:55:35,030 --> 00:55:32,559
um the whole point of this inactivist um

1476
00:55:38,790 --> 00:55:35,040
take on sense-making

1477
00:55:41,190 --> 00:55:38,800
um is the it you know you you have to

1478
00:55:43,109 --> 00:55:41,200
optimally go and listen to the right

1479
00:55:46,309 --> 00:55:43,119
kind of news talk to the right kind of

1480
00:55:48,390 --> 00:55:46,319
people um in order to solicit the right

1481
00:55:50,069 --> 00:55:48,400
kind of data that's going to maximize

1482
00:55:51,750 --> 00:55:50,079
the evidence for your model whilst at

1483
00:55:54,150 --> 00:55:51,760
the same time you are updating your

1484
00:55:56,549 --> 00:55:54,160
model um sometimes known as basically

1485
00:55:58,470 --> 00:55:56,559
palpitating um

1486
00:56:00,630 --> 00:55:58,480
or certain bayesian model selection for

1487
00:56:02,870 --> 00:56:00,640
example so the two things go hand in

1488
00:56:05,829 --> 00:56:02,880
hand i think one way of resolving or

1489
00:56:08,390 --> 00:56:05,839
answering your question is just to

1490
00:56:12,549 --> 00:56:08,400
say in a political um

1491
00:56:18,230 --> 00:56:14,789
perhaps even sort of

1492
00:56:19,430 --> 00:56:18,240
communication science uh sense or uh or

1493
00:56:22,390 --> 00:56:19,440
at least

1494
00:56:25,349 --> 00:56:22,400
the social aspects of it um what you're

1495
00:56:27,270 --> 00:56:25,359
what you're saying is that um

1496
00:56:30,230 --> 00:56:27,280
there is now

1497
00:56:32,470 --> 00:56:30,240
once you acknowledge that you also are

1498
00:56:33,349 --> 00:56:32,480
in charge of getting the data

1499
00:56:34,630 --> 00:56:33,359
um

1500
00:56:39,910 --> 00:56:34,640
then

1501
00:56:42,789 --> 00:56:39,920
problems of how much weight you afford

1502
00:56:44,470 --> 00:56:42,799
to that data versus that data and where

1503
00:56:47,349 --> 00:56:44,480
you get your data from which news

1504
00:56:49,349 --> 00:56:47,359
channels do you uh do um

1505
00:56:52,150 --> 00:56:49,359
forage

1506
00:56:53,990 --> 00:56:52,160
who do you listen to who do you trust

1507
00:56:55,190 --> 00:56:54,000
who do you ignore

1508
00:56:57,190 --> 00:56:55,200
um

1509
00:57:00,309 --> 00:56:57,200
and we do this all the time we do this

1510
00:57:02,470 --> 00:57:00,319
in a very skillful way so personally

1511
00:57:04,150 --> 00:57:02,480
in terms of deploying

1512
00:57:06,549 --> 00:57:04,160
our attention

1513
00:57:09,829 --> 00:57:06,559
so if you're an engineer

1514
00:57:11,990 --> 00:57:09,839
and you are thinking about

1515
00:57:14,309 --> 00:57:12,000
assimilating visual information using a

1516
00:57:17,430 --> 00:57:14,319
kalman filter for example

1517
00:57:20,230 --> 00:57:17,440
then one of the most artful challenges

1518
00:57:22,230 --> 00:57:20,240
or art for resolutions that the cameron

1519
00:57:25,270 --> 00:57:22,240
filter brings to the table is to get the

1520
00:57:26,549 --> 00:57:25,280
kalman gain right how much weight do you

1521
00:57:28,710 --> 00:57:26,559
afford

1522
00:57:33,030 --> 00:57:28,720
the prediction errors that supply the up

1523
00:57:35,430 --> 00:57:33,040
to term of a basin or a kalman filter

1524
00:57:38,870 --> 00:57:35,440
and we we do that on the fly all the

1525
00:57:40,630 --> 00:57:38,880
time right down to the level of

1526
00:57:43,109 --> 00:57:40,640
turning on and off

1527
00:57:44,069 --> 00:57:43,119
the attention or the precision or the

1528
00:57:45,750 --> 00:57:44,079
gain

1529
00:57:48,309 --> 00:57:45,760
um afforded

1530
00:57:52,150 --> 00:57:48,319
visual information so my favorite

1531
00:57:54,789 --> 00:57:52,160
example is um psychotic eye movements um

1532
00:57:57,030 --> 00:57:54,799
so just to demonstrate that you know

1533
00:57:59,349 --> 00:57:57,040
we could you know one can

1534
00:58:02,710 --> 00:57:59,359
do this experiment at home you can just

1535
00:58:05,430 --> 00:58:02,720
look um from one side of your visual

1536
00:58:07,750 --> 00:58:05,440
scene to the other side and back and

1537
00:58:09,670 --> 00:58:07,760
forth and back and forth with psychotic

1538
00:58:11,829 --> 00:58:09,680
eye movements

1539
00:58:13,670 --> 00:58:11,839
and you will not perceive the world

1540
00:58:16,309 --> 00:58:13,680
jumping around

1541
00:58:18,789 --> 00:58:16,319
however if you reproduce the same visual

1542
00:58:21,109 --> 00:58:18,799
impressions just by palpating your eye

1543
00:58:22,710 --> 00:58:21,119
in the spirit of hellholes

1544
00:58:23,430 --> 00:58:22,720
what's that palpatine what does that

1545
00:58:24,870 --> 00:58:23,440
mean

1546
00:58:26,309 --> 00:58:24,880
yeah if you well you can literally do

1547
00:58:29,750 --> 00:58:26,319
this now if you just take your finger

1548
00:58:32,150 --> 00:58:29,760
and place it gently on the outer

1549
00:58:34,950 --> 00:58:32,160
aspect of your eyeball and just gently

1550
00:58:36,789 --> 00:58:34,960
press in until you see the world shift

1551
00:58:37,510 --> 00:58:36,799
around and move around

1552
00:58:40,230 --> 00:58:37,520
so

1553
00:58:43,349 --> 00:58:40,240
you can see visual motion

1554
00:58:45,030 --> 00:58:43,359
that is a consequence of optic flow in

1555
00:58:47,270 --> 00:58:45,040
introduced in this instance by you

1556
00:58:49,829 --> 00:58:47,280
pushing your retina

1557
00:58:50,870 --> 00:58:49,839
translating your retina with the input

1558
00:58:53,270 --> 00:58:50,880
fixed

1559
00:58:54,470 --> 00:58:53,280
um and that is basically exactly the

1560
00:58:57,510 --> 00:58:54,480
same from the point of view of the

1561
00:59:00,309 --> 00:58:57,520
retina the back of your eyeball as

1562
00:59:01,990 --> 00:59:00,319
moving the iso

1563
00:59:07,270 --> 00:59:02,000
with the

1564
00:59:08,870 --> 00:59:07,280
you only saw

1565
00:59:11,750 --> 00:59:08,880
motion of the world

1566
00:59:14,309 --> 00:59:11,760
when you were moving your eyeball

1567
00:59:17,270 --> 00:59:14,319
when you move your eyes in the normal

1568
00:59:20,069 --> 00:59:17,280
way you don't see that motion

1569
00:59:21,349 --> 00:59:20,079
and the reason is that your brain

1570
00:59:22,789 --> 00:59:21,359
ignores

1571
00:59:25,109 --> 00:59:22,799
visual input

1572
00:59:27,270 --> 00:59:25,119
during eye movement and that's called

1573
00:59:29,349 --> 00:59:27,280
psychotic suppression so i think it's a

1574
00:59:30,390 --> 00:59:29,359
beautiful example of the fact that every

1575
00:59:33,109 --> 00:59:30,400
moment

1576
00:59:35,510 --> 00:59:33,119
we are selectively ignoring

1577
00:59:36,710 --> 00:59:35,520
information that is not newsworthy we

1578
00:59:38,870 --> 00:59:36,720
are selecting it we don't know we're

1579
00:59:42,150 --> 00:59:38,880
doing it but we do it all the time of

1580
00:59:44,870 --> 00:59:42,160
course you know at a much more abstract

1581
00:59:47,670 --> 00:59:44,880
and epistemic level exactly the same

1582
00:59:49,750 --> 00:59:47,680
principles apply when we choose you know

1583
00:59:52,549 --> 00:59:49,760
which newspapers which blog which

1584
00:59:54,549 --> 00:59:52,559
twitter accounts to follow um yeah which

1585
00:59:56,710 --> 00:59:54,559
friends to talk to which universes to go

1586
01:00:00,069 --> 00:59:56,720
to um is it going to be a scholarpedia

1587
01:00:02,630 --> 01:00:00,079
or a wikipedia no all of these decisions

1588
01:00:05,270 --> 01:00:02,640
you know that we make in terms of

1589
01:00:08,870 --> 01:00:05,280
securing the right kind of informative

1590
01:00:11,270 --> 01:00:08,880
precise data that will um literally have

1591
01:00:13,190 --> 01:00:11,280
a salience in the sense that it complies

1592
01:00:16,630 --> 01:00:13,200
with the principles of optimum bayesian

1593
01:00:19,270 --> 01:00:16,640
design in the sense that it will resolve

1594
01:00:20,549 --> 01:00:19,280
the most uncertainty about what i don't

1595
01:00:25,030 --> 01:00:20,559
know

1596
01:00:27,750 --> 01:00:25,040
this sort of um

1597
01:00:29,270 --> 01:00:27,760
being in charge of the data

1598
01:00:32,630 --> 01:00:29,280
even before thinking about the

1599
01:00:35,430 --> 01:00:32,640
consequences of your belief updating and

1600
01:00:37,030 --> 01:00:35,440
you know you're getting uh just knowing

1601
01:00:39,829 --> 01:00:37,040
that you have to write that and of

1602
01:00:41,990 --> 01:00:39,839
course the you know the um

1603
01:00:44,150 --> 01:00:42,000
from the from a socio-political point of

1604
01:00:46,150 --> 01:00:44,160
view what we're talking about now is how

1605
01:00:46,870 --> 01:00:46,160
to disambiguate between fake news and

1606
01:00:49,109 --> 01:00:46,880
true

1607
01:00:50,549 --> 01:00:49,119
news how to

1608
01:00:57,270 --> 01:00:50,559
um

1609
01:01:01,109 --> 01:00:57,280
us being isolationist or precludes as

1610
01:01:03,910 --> 01:01:01,119
succumbing to um cult like

1611
01:01:06,630 --> 01:01:03,920
ideologies um you know precludes

1612
01:01:08,630 --> 01:01:06,640
fundamentalism of a dysfunctional sort

1613
01:01:11,030 --> 01:01:08,640
or why is it dysfunctional well usually

1614
01:01:12,470 --> 01:01:11,040
because that kind of fundamentalism does

1615
01:01:15,589 --> 01:01:12,480
not speak to

1616
01:01:17,430 --> 01:01:15,599
and does not um minimize that joint

1617
01:01:20,069 --> 01:01:17,440
predictability of that joint free energy

1618
01:01:21,589 --> 01:01:20,079
that we were talking about um

1619
01:01:24,710 --> 01:01:21,599
several minutes ago

1620
01:01:26,230 --> 01:01:24,720
so what what you will find is that

1621
01:01:28,870 --> 01:01:26,240
[Music]

1622
01:01:32,069 --> 01:01:28,880
that dialectic between changing your

1623
01:01:34,870 --> 01:01:32,079
mind on the basis of some evidence and

1624
01:01:37,670 --> 01:01:34,880
changing the evidence by uh changing

1625
01:01:39,910 --> 01:01:37,680
where you look for that evidence that

1626
01:01:41,109 --> 01:01:39,920
that balance is at the heart of action

1627
01:01:43,430 --> 01:01:41,119
and perception

1628
01:01:44,950 --> 01:01:43,440
and indeed you you could argue the all

1629
01:01:47,990 --> 01:01:44,960
of action

1630
01:01:50,470 --> 01:01:48,000
is simply in the service of gathering

1631
01:01:52,950 --> 01:01:50,480
better data what do you mean by better

1632
01:01:55,910 --> 01:01:52,960
those data that was that minimize your

1633
01:01:58,150 --> 01:01:55,920
surprise minimize your expected surprise

1634
01:02:00,549 --> 01:01:58,160
or your uncertainty resolving

1635
01:02:03,109 --> 01:02:00,559
uncertainty about the way you think the

1636
01:02:05,750 --> 01:02:03,119
world works and more particularly how

1637
01:02:07,190 --> 01:02:05,760
you work in that in that world this

1638
01:02:09,030 --> 01:02:07,200
sounds like a deeper issue than just a

1639
01:02:12,549 --> 01:02:09,040
political problem it sounds like an

1640
01:02:14,950 --> 01:02:12,559
issue for the process of suggesting that

1641
01:02:16,789 --> 01:02:14,960
the conclusions that one has come to was

1642
01:02:19,109 --> 01:02:16,799
from a dispassionate assessment of the

1643
01:02:20,630 --> 01:02:19,119
evidence so you hear people say this

1644
01:02:22,309 --> 01:02:20,640
generally people who like to think of

1645
01:02:24,309 --> 01:02:22,319
themselves as extremely rational people

1646
01:02:26,710 --> 01:02:24,319
who've come to their conclusions because

1647
01:02:29,030 --> 01:02:26,720
they have objectively analyzed some

1648
01:02:30,710 --> 01:02:29,040
situation it sounds like that is a

1649
01:02:32,069 --> 01:02:30,720
doomed project because we're all

1650
01:02:34,710 --> 01:02:32,079
sampling different data without

1651
01:02:36,390 --> 01:02:34,720
realizing it now is that the case or no

1652
01:02:38,789 --> 01:02:36,400
yes no i think that's absolutely right

1653
01:02:40,789 --> 01:02:38,799
um and i don't think there's anything

1654
01:02:42,470 --> 01:02:40,799
bad about it i mean that's just the way

1655
01:02:44,069 --> 01:02:42,480
things are so how does one overcome that

1656
01:02:47,430 --> 01:02:44,079
what what do we do how is it this the

1657
01:02:51,190 --> 01:02:47,440
scientific process seems to work

1658
01:02:53,510 --> 01:02:51,200
um well it works because um uh

1659
01:02:56,470 --> 01:02:53,520
the the operational definition of

1660
01:02:58,870 --> 01:02:56,480
working simply is maximizing the

1661
01:03:00,549 --> 01:02:58,880
marginal likelihood so your liberty to

1662
01:03:03,030 --> 01:03:00,559
actually maximize the margin likely by

1663
01:03:06,710 --> 01:03:03,040
either changing the data or changing to

1664
01:03:08,549 --> 01:03:06,720
changing the model um so notice that in

1665
01:03:10,230 --> 01:03:08,559
this instance

1666
01:03:11,510 --> 01:03:10,240
you're not going to make any false

1667
01:03:14,150 --> 01:03:11,520
inferences

1668
01:03:16,789 --> 01:03:14,160
you know provided that all the data you

1669
01:03:18,630 --> 01:03:16,799
solicit is always very consistent with

1670
01:03:21,270 --> 01:03:18,640
your model you you won't need to change

1671
01:03:23,029 --> 01:03:21,280
your mind um if you've got the wrong

1672
01:03:25,029 --> 01:03:23,039
model the data will provide this

1673
01:03:27,190 --> 01:03:25,039
confirmatory evidence and and you may

1674
01:03:28,630 --> 01:03:27,200
well to try and get data from another

1675
01:03:30,390 --> 01:03:28,640
source and if you can do that and

1676
01:03:32,630 --> 01:03:30,400
continue engaging until your model is

1677
01:03:35,270 --> 01:03:32,640
fit for purpose what you have basically

1678
01:03:37,510 --> 01:03:35,280
done is find a niche for yourself where

1679
01:03:39,109 --> 01:03:37,520
your ideology your conceptions your

1680
01:03:41,670 --> 01:03:39,119
understanding of the world the world the

1681
01:03:43,670 --> 01:03:41,680
way the world work is perfectly fine so

1682
01:03:46,390 --> 01:03:43,680
you've found a culturally

1683
01:03:49,510 --> 01:03:46,400
or possibly a physiological eco niche or

1684
01:03:50,950 --> 01:03:49,520
an information eco niche that suits you

1685
01:03:53,430 --> 01:03:50,960
and you will have a high adaptive

1686
01:03:54,789 --> 01:03:53,440
fitness and survive and persist in that

1687
01:03:56,789 --> 01:03:54,799
eco niche

1688
01:04:01,670 --> 01:03:56,799
if you can't you move to another one and

1689
01:04:04,230 --> 01:04:01,680
so now we have a model of speciation um

1690
01:04:07,589 --> 01:04:04,240
we have a model of

1691
01:04:10,069 --> 01:04:07,599
shared narratives shared convictions

1692
01:04:11,990 --> 01:04:10,079
that can be communicated so supply and

1693
01:04:12,950 --> 01:04:12,000
teach each other

1694
01:04:15,029 --> 01:04:12,960
that

1695
01:04:17,430 --> 01:04:15,039
is self-assembling

1696
01:04:19,190 --> 01:04:17,440
and um

1697
01:04:20,710 --> 01:04:19,200
self-constructing

1698
01:04:24,230 --> 01:04:20,720
co-created

1699
01:04:25,589 --> 01:04:24,240
simply because you are looking for

1700
01:04:27,109 --> 01:04:25,599
at the same time you're looking for

1701
01:04:29,670 --> 01:04:27,119
information that resolves the

1702
01:04:31,589 --> 01:04:29,680
uncertainty about your models of how the

1703
01:04:34,069 --> 01:04:31,599
world works

1704
01:04:36,950 --> 01:04:34,079
which makes it look as if from the

1705
01:04:40,710 --> 01:04:36,960
outside you're basically changing the

1706
01:04:43,589 --> 01:04:40,720
world to make your hypotheses come true

1707
01:04:46,150 --> 01:04:43,599
and that of course is just another way

1708
01:04:47,589 --> 01:04:46,160
of describing action so if we come away

1709
01:04:49,750 --> 01:04:47,599
from the sort of

1710
01:04:51,910 --> 01:04:49,760
the exchange of information as a

1711
01:04:54,230 --> 01:04:51,920
scientist or somebody who reads the news

1712
01:04:55,270 --> 01:04:54,240
or somebody who's trying to do a phd or

1713
01:04:57,589 --> 01:04:55,280
um

1714
01:05:00,309 --> 01:04:57,599
indeed just you know make sense of uh

1715
01:05:02,230 --> 01:05:00,319
recent political events uh if we can

1716
01:05:06,630 --> 01:05:02,240
bring it back down to

1717
01:05:09,430 --> 01:05:06,640
simple things like um moving your eyes

1718
01:05:12,950 --> 01:05:11,660
in a sense that is you

1719
01:05:17,510 --> 01:05:12,960
[Music]

1720
01:05:18,789 --> 01:05:17,520
changing the data to match your model

1721
01:05:19,910 --> 01:05:18,799
so

1722
01:05:21,670 --> 01:05:19,920
if you

1723
01:05:23,510 --> 01:05:21,680
so this falls out of the free energy

1724
01:05:26,069 --> 01:05:23,520
principle because you know if it's the

1725
01:05:30,309 --> 01:05:26,079
case that the autonomous paths the

1726
01:05:32,470 --> 01:05:30,319
dynamics that we um um that um

1727
01:05:35,270 --> 01:05:32,480
that constitute are

1728
01:05:37,430 --> 01:05:35,280
our active engagement with the world if

1729
01:05:39,670 --> 01:05:37,440
they're trying to minimize free energy

1730
01:05:42,630 --> 01:05:39,680
and we in this instance minimize free

1731
01:05:45,670 --> 01:05:42,640
energy as a prediction error or surprise

1732
01:05:47,430 --> 01:05:45,680
um remembering that it's the negative of

1733
01:05:48,870 --> 01:05:47,440
the marginal likelihood

1734
01:05:49,670 --> 01:05:48,880
it's something

1735
01:05:52,309 --> 01:05:49,680
if

1736
01:05:53,990 --> 01:05:52,319
the probability of these data

1737
01:05:55,750 --> 01:05:54,000
given my understanding of my

1738
01:05:57,029 --> 01:05:55,760
construction of the world at this point

1739
01:05:59,510 --> 01:05:57,039
in time

1740
01:06:01,910 --> 01:05:59,520
is very very low

1741
01:06:04,230 --> 01:06:01,920
then it will have a very very small

1742
01:06:06,470 --> 01:06:04,240
marginal likelihood or have a very very

1743
01:06:08,390 --> 01:06:06,480
high free energy and therefore it will

1744
01:06:09,270 --> 01:06:08,400
be surprising given my model of the

1745
01:06:12,230 --> 01:06:09,280
world

1746
01:06:14,470 --> 01:06:12,240
some people like to articulate the other

1747
01:06:16,950 --> 01:06:14,480
that surprise in terms of a prediction

1748
01:06:19,109 --> 01:06:16,960
what i predicted was going to happen

1749
01:06:20,950 --> 01:06:19,119
is nothing like what i actually sense

1750
01:06:22,710 --> 01:06:20,960
the data i actually got

1751
01:06:23,750 --> 01:06:22,720
so

1752
01:06:25,109 --> 01:06:23,760
imagine

1753
01:06:26,470 --> 01:06:25,119
how um

1754
01:06:27,270 --> 01:06:26,480
imagine now

1755
01:06:32,069 --> 01:06:27,280
you

1756
01:06:35,349 --> 01:06:32,079
the autonomous states that include the

1757
01:06:37,349 --> 01:06:35,359
active states are just in the service of

1758
01:06:39,270 --> 01:06:37,359
minimizing prediction error

1759
01:06:41,990 --> 01:06:39,280
how we would you understand that in

1760
01:06:43,750 --> 01:06:42,000
terms of elemental movement and behavior

1761
01:06:47,109 --> 01:06:43,760
and one way of understanding that is

1762
01:06:49,190 --> 01:06:47,119
that basically your models of say how

1763
01:06:51,750 --> 01:06:49,200
you're going to move are actually

1764
01:06:54,230 --> 01:06:51,760
realized by creating data that is

1765
01:06:56,950 --> 01:06:54,240
consistent with your predictions

1766
01:06:58,829 --> 01:06:56,960
so i have in my head

1767
01:07:01,829 --> 01:06:58,839
a model a

1768
01:07:03,750 --> 01:07:01,839
hypothesis the notion that i'm going to

1769
01:07:04,470 --> 01:07:03,760
raise my arm

1770
01:07:07,750 --> 01:07:04,480
and

1771
01:07:09,990 --> 01:07:07,760
all that's happening in that

1772
01:07:11,510 --> 01:07:10,000
notion is that that's my belief about

1773
01:07:14,309 --> 01:07:11,520
state in the world

1774
01:07:17,589 --> 01:07:14,319
and i'm telling my spinal cord and my

1775
01:07:19,829 --> 01:07:17,599
musculature and my muscles

1776
01:07:22,630 --> 01:07:19,839
that's how i expect the world to be and

1777
01:07:24,230 --> 01:07:22,640
then my reflexes are driven by the

1778
01:07:25,270 --> 01:07:24,240
prediction errors

1779
01:07:27,029 --> 01:07:25,280
driven

1780
01:07:29,349 --> 01:07:27,039
by the free energy gradients literally

1781
01:07:32,789 --> 01:07:29,359
like the ball rolling down the hill

1782
01:07:34,789 --> 01:07:32,799
um to realize to change the world so it

1783
01:07:36,069 --> 01:07:34,799
supplies the kind of data that i

1784
01:07:39,349 --> 01:07:36,079
predicted

1785
01:07:40,789 --> 01:07:39,359
i just wanted to mention that um

1786
01:07:43,589 --> 01:07:40,799
that notion

1787
01:07:46,710 --> 01:07:43,599
of movement

1788
01:07:49,109 --> 01:07:46,720
as simply a realization

1789
01:07:51,510 --> 01:07:49,119
of predicted

1790
01:07:54,069 --> 01:07:51,520
endpoints

1791
01:07:54,829 --> 01:07:54,079
intentions to move

1792
01:07:58,069 --> 01:07:54,839
is

1793
01:08:00,470 --> 01:07:58,079
physiologically um fully licensed by

1794
01:08:02,789 --> 01:08:00,480
things like equilibrium point hypothesis

1795
01:08:06,549 --> 01:08:02,799
in motor control and motor physiology

1796
01:08:08,630 --> 01:08:06,559
but it's it goes back even further to um

1797
01:08:12,150 --> 01:08:08,640
ideomotor theory so this was in the

1798
01:08:13,990 --> 01:08:12,160
victorian age how people understood

1799
01:08:17,189 --> 01:08:14,000
movement that basically it was a

1800
01:08:19,749 --> 01:08:17,199
physical manifestation of beliefs about

1801
01:08:21,590 --> 01:08:19,759
the way my body should be posed or what

1802
01:08:24,789 --> 01:08:21,600
my body should be doing

1803
01:08:27,349 --> 01:08:24,799
so to me that is if you like an

1804
01:08:28,390 --> 01:08:27,359
elemental and quite fundamental example

1805
01:08:31,829 --> 01:08:28,400
of

1806
01:08:34,470 --> 01:08:31,839
the model

1807
01:08:38,390 --> 01:08:34,480
so if you changed your model in the

1808
01:08:41,189 --> 01:08:38,400
context of i'm going to stand up now

1809
01:08:43,110 --> 01:08:41,199
and you looked at the data and nothing

1810
01:08:45,269 --> 01:08:43,120
is happening you would immediately

1811
01:08:46,950 --> 01:08:45,279
change your mind and say oh i'm not

1812
01:08:48,630 --> 01:08:46,960
standing up

1813
01:08:50,789 --> 01:08:48,640
what would that look like that would

1814
01:08:53,269 --> 01:08:50,799
look like parkinson's disease

1815
01:08:55,110 --> 01:08:53,279
that would be a failure to properly

1816
01:08:57,590 --> 01:08:55,120
attenuate the evidence that i'm not

1817
01:09:00,149 --> 01:08:57,600
moving when i strongly believe that i am

1818
01:09:02,709 --> 01:09:00,159
about to move

1819
01:09:03,430 --> 01:09:02,719
and of course you can understand now

1820
01:09:08,149 --> 01:09:03,440
the

1821
01:09:12,630 --> 01:09:09,749
disease

1822
01:09:14,070 --> 01:09:12,640
that's due to a failure of

1823
01:09:15,749 --> 01:09:14,080
neuronal message passing and

1824
01:09:18,470 --> 01:09:15,759
particularly the weight or the gain

1825
01:09:20,309 --> 01:09:18,480
afforded proprioceptive information in

1826
01:09:21,910 --> 01:09:20,319
that light it's a failure of active

1827
01:09:23,110 --> 01:09:21,920
influences a failure of the active

1828
01:09:27,990 --> 01:09:23,120
states

1829
01:09:28,870 --> 01:09:28,000
to um probably uh respond to

1830
01:09:34,630 --> 01:09:28,880
um

1831
01:09:37,510 --> 01:09:34,640
evidence at hand

1832
01:09:38,950 --> 01:09:37,520
so this is a lovely example of ignoring

1833
01:09:41,990 --> 01:09:38,960
selectively ignoring just like the

1834
01:09:43,990 --> 01:09:42,000
satanic suppression with the eyes

1835
01:09:45,829 --> 01:09:44,000
that enables movement if you didn't do

1836
01:09:48,390 --> 01:09:45,839
that we couldn't move so what i'm trying

1837
01:09:50,070 --> 01:09:48,400
to say is that you know right from the

1838
01:09:52,709 --> 01:09:50,080
the inception of these kinds of this

1839
01:09:55,910 --> 01:09:52,719
kind of thinking in the inactive domain

1840
01:09:58,550 --> 01:09:55,920
uh in terms of ideomotive theory

1841
01:10:01,669 --> 01:09:58,560
the uh there is a there is i think a

1842
01:10:02,390 --> 01:10:01,679
fundamental importance of um

1843
01:10:03,430 --> 01:10:02,400
your

1844
01:10:05,270 --> 01:10:03,440
um

1845
01:10:07,430 --> 01:10:05,280
of the other side of the coin from

1846
01:10:10,709 --> 01:10:07,440
changing your mind which is basically

1847
01:10:13,910 --> 01:10:10,719
changing the data uh upon which you make

1848
01:10:18,390 --> 01:10:15,830
which causes you to change your mind

1849
01:10:19,590 --> 01:10:18,400
through through sense-making

1850
01:10:21,910 --> 01:10:19,600
and without that we wouldn't have

1851
01:10:24,709 --> 01:10:21,920
actions so you can almost say that

1852
01:10:26,870 --> 01:10:24,719
action and perception are the two sides

1853
01:10:29,110 --> 01:10:26,880
of the same coin the perception is

1854
01:10:31,430 --> 01:10:29,120
changing your mind by changing your

1855
01:10:33,830 --> 01:10:31,440
model to maximize the marginal

1856
01:10:36,470 --> 01:10:33,840
likelihood given a particular model

1857
01:10:38,870 --> 01:10:36,480
action just is changing the data by

1858
01:10:41,110 --> 01:10:38,880
acting upon the world to maximize the

1859
01:10:42,790 --> 01:10:41,120
marginal likelihood basically maximizing

1860
01:10:45,830 --> 01:10:42,800
changing the data so the probability of

1861
01:10:47,830 --> 01:10:45,840
the data given the model is so

1862
01:10:49,669 --> 01:10:47,840
active inference or that coupling of

1863
01:10:51,189 --> 01:10:49,679
action perception

1864
01:10:54,390 --> 01:10:51,199
is essentially

1865
01:10:57,830 --> 01:10:54,400
trying to realize your beliefs about or

1866
01:10:59,270 --> 01:10:57,840
your models of your lived world and if

1867
01:11:01,910 --> 01:10:59,280
you're successful in doing that you've

1868
01:11:04,310 --> 01:11:01,920
got the right eco niche you will have a

1869
01:11:05,910 --> 01:11:04,320
high marginal likelihood you know from

1870
01:11:07,270 --> 01:11:05,920
an evolutionary perspective you will

1871
01:11:08,630 --> 01:11:07,280
have um

1872
01:11:12,390 --> 01:11:08,640
a

1873
01:11:13,590 --> 01:11:12,400
perspective you will be uh people will

1874
01:11:15,669 --> 01:11:13,600
listen to you

1875
01:11:19,189 --> 01:11:15,679
and teach you your job into these

1876
01:11:22,310 --> 01:11:19,199
podcasts suggesting you you do indeed

1877
01:11:25,270 --> 01:11:22,320
solicit that kind of information uh that

1878
01:11:28,630 --> 01:11:25,280
causes people to you know to go to your

1879
01:11:30,310 --> 01:11:28,640
channel to attend to your stuff um

1880
01:11:31,510 --> 01:11:30,320
simply because they they think they're

1881
01:11:34,149 --> 01:11:31,520
gonna get good stuff and they're gonna

1882
01:11:37,270 --> 01:11:34,159
get uncertainty resolving stuff um so

1883
01:11:39,350 --> 01:11:37,280
this this this i think is is is uh puts

1884
01:11:41,590 --> 01:11:39,360
the balance back on the uh on getting

1885
01:11:42,550 --> 01:11:41,600
the data right or the data data foraging

1886
01:11:44,149 --> 01:11:42,560
right

1887
01:11:46,470 --> 01:11:44,159
okay well perhaps let me undermine my

1888
01:11:47,270 --> 01:11:46,480
own credibility here

1889
01:11:49,590 --> 01:11:47,280
so

1890
01:11:51,830 --> 01:11:49,600
a few weeks ago and i think about a year

1891
01:11:54,229 --> 01:11:51,840
ago or so we talked and i had some

1892
01:11:55,990 --> 01:11:54,239
experience about and i was worried and

1893
01:11:57,270 --> 01:11:56,000
you said kurt don't be worried don't i

1894
01:11:59,510 --> 01:11:57,280
had some experience where i thought i

1895
01:12:01,189 --> 01:11:59,520
heard a voice it was at a in a

1896
01:12:03,270 --> 01:12:01,199
hypnagogic state so i was just about to

1897
01:12:05,030 --> 01:12:03,280
fall asleep when

1898
01:12:06,149 --> 01:12:05,040
many people

1899
01:12:08,310 --> 01:12:06,159
that shouldn't be a state that i

1900
01:12:10,070 --> 01:12:08,320
attribute much credence to but i

1901
01:12:12,310 --> 01:12:10,080
remember being afraid because i heard my

1902
01:12:14,149 --> 01:12:12,320
wife either say okay or just some one

1903
01:12:16,070 --> 01:12:14,159
syllable word and then i was so worried

1904
01:12:17,910 --> 01:12:16,080
and i worked myself up into anxiety

1905
01:12:18,709 --> 01:12:17,920
thinking do i have schizophrenia oh my

1906
01:12:21,110 --> 01:12:18,719
gosh

1907
01:12:23,430 --> 01:12:21,120
there was no other voices since then but

1908
01:12:24,630 --> 01:12:23,440
i was extremely anxious

1909
01:12:26,070 --> 01:12:24,640
and then you told me don't worry and

1910
01:12:28,470 --> 01:12:26,080
even kurt even if there is there's

1911
01:12:30,630 --> 01:12:28,480
medication and many people go through

1912
01:12:31,510 --> 01:12:30,640
psychosis and or psychotic breaks and so

1913
01:12:33,430 --> 01:12:31,520
on

1914
01:12:35,669 --> 01:12:33,440
but luckily i didn't luckily there was

1915
01:12:37,990 --> 01:12:35,679
nothing wrong with me and your words did

1916
01:12:39,590 --> 01:12:38,000
calm me so thank you for that however

1917
01:12:41,510 --> 01:12:39,600
and coincidentally

1918
01:12:43,510 --> 01:12:41,520
perhaps not but coincidentally a few

1919
01:12:45,350 --> 01:12:43,520
weeks ago

1920
01:12:46,950 --> 01:12:45,360
i had a strange experience carl and i

1921
01:12:48,790 --> 01:12:46,960
wasn't sure if i was gonna bring this up

1922
01:12:49,990 --> 01:12:48,800
on the podcast for one reason because i

1923
01:12:51,030 --> 01:12:50,000
don't want you to think that i'm someone

1924
01:12:53,270 --> 01:12:51,040
who's

1925
01:12:55,830 --> 01:12:53,280
beset with psychological issues but on

1926
01:12:57,830 --> 01:12:55,840
the other side i am driven by a

1927
01:12:59,830 --> 01:12:57,840
filmmaking adage which says that the

1928
01:13:01,830 --> 01:12:59,840
more personal the pain the more widely

1929
01:13:03,669 --> 01:13:01,840
applicable it is and the opposite is

1930
01:13:05,590 --> 01:13:03,679
also false the opposite is false so if

1931
01:13:07,990 --> 01:13:05,600
you try to speak generally you end up

1932
01:13:10,390 --> 01:13:08,000
speaking to no one

1933
01:13:12,070 --> 01:13:10,400
what happened a couple weeks ago

1934
01:13:13,990 --> 01:13:12,080
was i had this

1935
01:13:16,070 --> 01:13:14,000
sense and it was so and even to talk

1936
01:13:19,350 --> 01:13:16,080
about it now is a bit frightening i had

1937
01:13:22,310 --> 01:13:19,360
this sense that i was like oh shoot

1938
01:13:24,709 --> 01:13:22,320
solipsism is everything in my head just

1939
01:13:26,550 --> 01:13:24,719
that thought and i've because i'm on

1940
01:13:29,430 --> 01:13:26,560
this podcast and i've entertained many

1941
01:13:31,189 --> 01:13:29,440
ideas it's extremely taxing to do so

1942
01:13:32,470 --> 01:13:31,199
because i try to take different people's

1943
01:13:33,910 --> 01:13:32,480
points of views

1944
01:13:35,830 --> 01:13:33,920
seriously

1945
01:13:37,270 --> 01:13:35,840
and perhaps i shouldn't but i feel like

1946
01:13:39,750 --> 01:13:37,280
in order to give them a fair shake i

1947
01:13:43,189 --> 01:13:39,760
need to embody them in some way shape or

1948
01:13:45,110 --> 01:13:43,199
form so if they say well all is mind for

1949
01:13:46,950 --> 01:13:45,120
example that's the idealists

1950
01:13:48,550 --> 01:13:46,960
i'm trying to imagine what would that be

1951
01:13:50,070 --> 01:13:48,560
like and if someone's a materialist well

1952
01:13:51,350 --> 01:13:50,080
what does it mean that this was dead and

1953
01:13:53,270 --> 01:13:51,360
it somehow emerged that we have

1954
01:13:55,270 --> 01:13:53,280
consciousness either way the point of

1955
01:13:57,030 --> 01:13:55,280
that is to say i've been feeling like

1956
01:13:59,189 --> 01:13:57,040
for almost a year and a half now just in

1957
01:14:01,270 --> 01:13:59,199
a void not sure what is true and what's

1958
01:14:04,630 --> 01:14:01,280
not true that's fine i can deal with

1959
01:14:06,950 --> 01:14:04,640
that except that that led me to this

1960
01:14:08,470 --> 01:14:06,960
spiraling of such

1961
01:14:10,790 --> 01:14:08,480
it was probably the most terrifying

1962
01:14:11,990 --> 01:14:10,800
experience i've ever had in my life

1963
01:14:13,590 --> 01:14:12,000
and it took me weeks and i'm still

1964
01:14:14,630 --> 01:14:13,600
recovering from it where i felt like

1965
01:14:17,189 --> 01:14:14,640
shoot

1966
01:14:19,590 --> 01:14:17,199
am i all that exists and then just that

1967
01:14:21,669 --> 01:14:19,600
thought alone i'm like i don't want to

1968
01:14:22,870 --> 01:14:21,679
believe that that could be the case and

1969
01:14:25,830 --> 01:14:22,880
i remember looking at my wife and

1970
01:14:28,070 --> 01:14:25,840
thinking is my wife even real how

1971
01:14:30,390 --> 01:14:28,080
much of an imbecile and how ungrateful

1972
01:14:32,310 --> 01:14:30,400
and how foolish is it and how

1973
01:14:34,390 --> 01:14:32,320
that i would even think that i feel bad

1974
01:14:36,470 --> 01:14:34,400
that i would even entertain that thought

1975
01:14:37,830 --> 01:14:36,480
oh my gosh i'm scared i don't want to go

1976
01:14:40,149 --> 01:14:37,840
to the

1977
01:14:42,229 --> 01:14:40,159
hospital because i felt like if i was to

1978
01:14:44,870 --> 01:14:42,239
truly accept that thought i could i was

1979
01:14:46,950 --> 01:14:44,880
on the brink of just losing it into an

1980
01:14:49,830 --> 01:14:46,960
insensitive spiral

1981
01:14:51,830 --> 01:14:49,840
i became extremely scared of myself and

1982
01:14:53,430 --> 01:14:51,840
i think that

1983
01:14:54,630 --> 01:14:53,440
i'm the type of person i am an anxious

1984
01:14:55,510 --> 01:14:54,640
person

1985
01:14:57,430 --> 01:14:55,520
and i

1986
01:15:00,149 --> 01:14:57,440
am the type of person that tends to

1987
01:15:02,630 --> 01:15:00,159
obsess over thoughts

1988
01:15:05,189 --> 01:15:02,640
so then i started to obsess over this

1989
01:15:09,430 --> 01:15:05,199
and for the past couple weeks i was so

1990
01:15:11,750 --> 01:15:09,440
scared so scared of my own mind so

1991
01:15:13,350 --> 01:15:11,760
almost like a hypochondriac for mental

1992
01:15:15,030 --> 01:15:13,360
disorders where i didn't want to even

1993
01:15:16,950 --> 01:15:15,040
look up what's the definition of

1994
01:15:18,709 --> 01:15:16,960
schizophrenia or psychosis was because i

1995
01:15:20,070 --> 01:15:18,719
was afraid that i would see the signs

1996
01:15:21,430 --> 01:15:20,080
everywhere

1997
01:15:24,070 --> 01:15:21,440
for example

1998
01:15:25,830 --> 01:15:24,080
a few days ago or about a week ago i

1999
01:15:28,149 --> 01:15:25,840
remember generally we all speak to

2000
01:15:30,310 --> 01:15:28,159
ourselves in our own heads and sometimes

2001
01:15:32,470 --> 01:15:30,320
i use the phrase so i don't like that or

2002
01:15:33,669 --> 01:15:32,480
i like that i use the word i in my own

2003
01:15:35,030 --> 01:15:33,679
internal monologue but sometimes the

2004
01:15:36,630 --> 01:15:35,040
word yeah you can do this you can do

2005
01:15:38,470 --> 01:15:36,640
this so sometimes you use i and

2006
01:15:40,149 --> 01:15:38,480
sometimes they use you

2007
01:15:42,870 --> 01:15:40,159
and i imagine we all do that it's

2008
01:15:45,669 --> 01:15:42,880
colloquially okay to exchange those two

2009
01:15:47,590 --> 01:15:45,679
however then i thought oh shoot

2010
01:15:49,510 --> 01:15:47,600
what if i felt distance from my thought

2011
01:15:50,870 --> 01:15:49,520
what if the one that's saying you is a

2012
01:15:53,910 --> 01:15:50,880
thought speaking to me like a

2013
01:15:55,590 --> 01:15:53,920
schizophrenic may hear oh my gosh oh my

2014
01:15:58,149 --> 01:15:55,600
gosh i don't want to hear this and i was

2015
01:15:59,830 --> 01:15:58,159
so afraid and i felt derealized and

2016
01:16:01,830 --> 01:15:59,840
depersonalized when you mentioned

2017
01:16:04,550 --> 01:16:01,840
parkinson's earlier that the difference

2018
01:16:06,709 --> 01:16:04,560
between the model and then the data it

2019
01:16:08,070 --> 01:16:06,719
that didn't strike me as a

2020
01:16:09,830 --> 01:16:08,080
symptom of parkinson's the way you were

2021
01:16:11,990 --> 01:16:09,840
describing it struck me as a symptom of

2022
01:16:13,430 --> 01:16:12,000
derealization

2023
01:16:15,510 --> 01:16:13,440
anyway i'm bringing this all up and i

2024
01:16:17,270 --> 01:16:15,520
actually feel

2025
01:16:18,870 --> 01:16:17,280
good about speaking about it it somehow

2026
01:16:23,030 --> 01:16:18,880
feels

2027
01:16:24,390 --> 01:16:23,040
curative to even voice these concerns

2028
01:16:53,830 --> 01:16:24,400
i

2029
01:16:55,830 --> 01:16:53,840
that some people who

2030
01:16:57,669 --> 01:16:55,840
view yosha box ideas that this all an

2031
01:16:58,950 --> 01:16:57,679
illusion or this all simulated

2032
01:17:01,430 --> 01:16:58,960
consciousness some people become

2033
01:17:03,110 --> 01:17:01,440
suicidal that's actually extremely

2034
01:17:04,790 --> 01:17:03,120
terrifying luckily i'm not suicidal

2035
01:17:06,630 --> 01:17:04,800
though i'm scared that i

2036
01:17:08,790 --> 01:17:06,640
could because i've heard that other

2037
01:17:09,910 --> 01:17:08,800
people could i'm not sure if my problem

2038
01:17:11,990 --> 01:17:09,920
is just

2039
01:17:13,669 --> 01:17:12,000
obsessive thoughts

2040
01:17:16,229 --> 01:17:13,679
or if it was indicative of something

2041
01:17:18,950 --> 01:17:16,239
else and i was so afraid and it's it's

2042
01:17:19,669 --> 01:17:18,960
such a scary feeling carl to be afraid

2043
01:17:22,870 --> 01:17:19,679
of

2044
01:17:24,630 --> 01:17:22,880
one's self to not even be able to sleep

2045
01:17:27,430 --> 01:17:24,640
at night because you're afraid of your

2046
01:17:29,430 --> 01:17:27,440
own thoughts and what that may lead to

2047
01:17:30,950 --> 01:17:29,440
so does any of that make sense and i

2048
01:17:34,149 --> 01:17:30,960
know that sounds like it's completely

2049
01:17:37,669 --> 01:17:34,159
off field perhaps it's not no

2050
01:17:39,590 --> 01:17:37,679
no no i think i think it follows on

2051
01:17:41,590 --> 01:17:39,600
very gracefully but but you've taken it

2052
01:17:43,750 --> 01:17:41,600
right now to um

2053
01:17:46,070 --> 01:17:43,760
some of the most important aspects of

2054
01:17:47,669 --> 01:17:46,080
modeling um which we all contend with

2055
01:17:50,070 --> 01:17:47,679
which of course is is

2056
01:17:51,030 --> 01:17:50,080
um i mean i mentioned before your

2057
01:17:54,070 --> 01:17:51,040
getting

2058
01:17:55,910 --> 01:17:54,080
models for our um

2059
01:17:57,510 --> 01:17:55,920
our lived worlds but of course the most

2060
01:17:59,910 --> 01:17:57,520
important part

2061
01:18:01,669 --> 01:17:59,920
when we're a baby of the world is our

2062
01:18:02,790 --> 01:18:01,679
body so the first thing we have to do is

2063
01:18:07,189 --> 01:18:02,800
to build

2064
01:18:10,070 --> 01:18:07,199
a model uh of our body and um work out

2065
01:18:12,790 --> 01:18:10,080
you know what things that we are able to

2066
01:18:15,350 --> 01:18:12,800
move um you know hence motor babbling

2067
01:18:18,550 --> 01:18:17,270
rattling um

2068
01:18:21,430 --> 01:18:18,560
probably more important than that of

2069
01:18:22,870 --> 01:18:21,440
course is um exchanges with the mother

2070
01:18:25,030 --> 01:18:22,880
um and

2071
01:18:27,910 --> 01:18:25,040
developing a sense that

2072
01:18:29,990 --> 01:18:27,920
mother is an object and crucially an

2073
01:18:32,149 --> 01:18:30,000
object that is separate from me

2074
01:18:34,950 --> 01:18:32,159
and that's quite a skillful move and it

2075
01:18:37,350 --> 01:18:34,960
takes a lot of my belief updating and

2076
01:18:39,750 --> 01:18:37,360
changing your models and building

2077
01:18:42,310 --> 01:18:39,760
a coherent um

2078
01:18:44,390 --> 01:18:42,320
explanation for all your sensations both

2079
01:18:45,990 --> 01:18:44,400
the political touch or the

2080
01:18:48,470 --> 01:18:46,000
and the physiological interceptic

2081
01:18:50,229 --> 01:18:48,480
consequences of being subtle for example

2082
01:18:51,350 --> 01:18:50,239
and being cared for

2083
01:18:54,229 --> 01:18:51,360
um

2084
01:18:57,189 --> 01:18:54,239
and i'm taking this route to the key

2085
01:18:59,110 --> 01:18:57,199
thing which is basically models of self

2086
01:19:01,030 --> 01:18:59,120
and selfhood

2087
01:19:03,510 --> 01:19:01,040
so

2088
01:19:05,830 --> 01:19:03,520
if if we are

2089
01:19:08,630 --> 01:19:05,840
coming back to this picture of

2090
01:19:12,430 --> 01:19:08,640
some mathematical universe that i've now

2091
01:19:15,030 --> 01:19:12,440
tiled with multiple markov blankets

2092
01:19:16,630 --> 01:19:15,040
encapsulating or defining lots of

2093
01:19:18,630 --> 01:19:16,640
particles

2094
01:19:21,990 --> 01:19:18,640
and basically we

2095
01:19:25,430 --> 01:19:22,000
inhabit a universe that is constituted

2096
01:19:27,189 --> 01:19:25,440
by things like us and in initially uh

2097
01:19:29,590 --> 01:19:27,199
you know mum and dad and brothers and

2098
01:19:31,430 --> 01:19:29,600
sisters and then you know

2099
01:19:35,270 --> 01:19:31,440
peers at school

2100
01:19:37,189 --> 01:19:35,280
that means that our models of our lived

2101
01:19:40,070 --> 01:19:37,199
or experienced world

2102
01:19:42,070 --> 01:19:40,080
have to have models of other in it which

2103
01:19:43,669 --> 01:19:42,080
means you need to be able to

2104
01:19:47,270 --> 01:19:43,679
disambiguate

2105
01:19:49,590 --> 01:19:47,280
between sensations caused by others and

2106
01:19:51,110 --> 01:19:49,600
sensations that you caused as a

2107
01:19:51,990 --> 01:19:51,120
fundamental

2108
01:19:54,470 --> 01:19:52,000
um

2109
01:19:56,470 --> 01:19:54,480
structural aspect of the models you're

2110
01:19:57,830 --> 01:19:56,480
going to bring to the table just to live

2111
01:19:59,910 --> 01:19:57,840
in a world

2112
01:20:01,830 --> 01:19:59,920
which is constituted by creatures like

2113
01:20:08,149 --> 01:20:01,840
you

2114
01:20:10,149 --> 01:20:08,159
something quite special um

2115
01:20:11,110 --> 01:20:10,159
which

2116
01:20:14,870 --> 01:20:11,120
is

2117
01:20:17,510 --> 01:20:14,880
a model that encompasses selfhood

2118
01:20:18,630 --> 01:20:17,520
so you wouldn't need this if you

2119
01:20:21,830 --> 01:20:18,640
um

2120
01:20:24,709 --> 01:20:21,840
lived in a world that did not involve an

2121
01:20:27,270 --> 01:20:24,719
exchange with other con specifics

2122
01:20:29,510 --> 01:20:27,280
but because we do live your our world is

2123
01:20:31,270 --> 01:20:29,520
constituted by other things like us

2124
01:20:33,750 --> 01:20:31,280
we need to be able to contextualize and

2125
01:20:36,229 --> 01:20:33,760
make the inference i did that or you did

2126
01:20:38,630 --> 01:20:36,239
that or it's my turn or it's your turn

2127
01:20:40,470 --> 01:20:38,640
or you've got that intentional stance

2128
01:20:41,669 --> 01:20:40,480
and you know i've got this intentional

2129
01:20:43,990 --> 01:20:41,679
stance

2130
01:20:46,229 --> 01:20:44,000
so just having as part of the generative

2131
01:20:50,070 --> 01:20:46,239
model of the hypothesis your the

2132
01:20:53,030 --> 01:20:50,080
hypothesis that i am me and you and you

2133
01:20:56,470 --> 01:20:53,040
is something which characterizes um one

2134
01:20:59,189 --> 01:20:56,480
would imagine very high life forms

2135
01:21:01,910 --> 01:20:59,199
now we go even higher

2136
01:21:03,270 --> 01:21:01,920
um basically philosophers and people

2137
01:21:06,229 --> 01:21:03,280
like you

2138
01:21:09,910 --> 01:21:06,239
so if you've got if you have

2139
01:21:11,590 --> 01:21:09,920
if you spend your life worrying about um

2140
01:21:12,950 --> 01:21:11,600
what is the self does the self even

2141
01:21:15,270 --> 01:21:12,960
exist

2142
01:21:17,590 --> 01:21:15,280
most precisely yeah well i mean that

2143
01:21:20,870 --> 01:21:17,600
that's that that seems to be the the

2144
01:21:24,550 --> 01:21:20,880
essence of your existential angst and

2145
01:21:28,229 --> 01:21:24,560
all its attendant um intraceptive and

2146
01:21:29,830 --> 01:21:28,239
emotional um consequences uh you know do

2147
01:21:31,830 --> 01:21:29,840
i exist

2148
01:21:33,950 --> 01:21:31,840
and i think really usefully you brought

2149
01:21:36,709 --> 01:21:33,960
to uh the table the notions of

2150
01:21:38,790 --> 01:21:36,719
depersonalization and derealization if

2151
01:21:40,229 --> 01:21:38,800
this is for physicists then can you

2152
01:21:41,669 --> 01:21:40,239
explain to your audience what those

2153
01:21:43,270 --> 01:21:41,679
things are because i think that it would

2154
01:21:45,750 --> 01:21:43,280
be really useful for people that do not

2155
01:21:48,229 --> 01:21:45,760
know what periodically well to be quite

2156
01:21:50,550 --> 01:21:48,239
frank carl i didn't want to look up this

2157
01:21:52,790 --> 01:21:50,560
exact definition because i know that i'm

2158
01:21:55,430 --> 01:21:52,800
a valid tutenarian and i would just

2159
01:21:57,510 --> 01:21:55,440
obsess over myself having these issues

2160
01:21:58,870 --> 01:21:57,520
so i can give an explanation of my

2161
01:22:01,270 --> 01:21:58,880
experience and then perhaps you can

2162
01:22:03,030 --> 01:22:01,280
delineate what the psychopathology is

2163
01:22:04,870 --> 01:22:03,040
okay i'll make an aside first some

2164
01:22:06,790 --> 01:22:04,880
people say that what we need to do is

2165
01:22:08,870 --> 01:22:06,800
realize that the self is an illusion you

2166
01:22:11,110 --> 01:22:08,880
can hear this in the more eastern end

2167
01:22:13,270 --> 01:22:11,120
and i know that many people who watch

2168
01:22:14,790 --> 01:22:13,280
this channel also feel like they take

2169
01:22:16,310 --> 01:22:14,800
these people seriously like i took them

2170
01:22:17,830 --> 01:22:16,320
seriously and i don't know if that's

2171
01:22:19,510 --> 01:22:17,840
exactly true

2172
01:22:21,430 --> 01:22:19,520
not whether or not the truth of the self

2173
01:22:23,669 --> 01:22:21,440
being an illusion is true but whether or

2174
01:22:25,750 --> 01:22:23,679
not that is indeed useful for everyone

2175
01:22:27,910 --> 01:22:25,760
it may not be it may be at least for me

2176
01:22:30,070 --> 01:22:27,920
it was extremely harmful and perhaps

2177
01:22:31,910 --> 01:22:30,080
still is and it may be dependent upon

2178
01:22:33,990 --> 01:22:31,920
where you are and psychological

2179
01:22:36,149 --> 01:22:34,000
development spiritual development and

2180
01:22:39,430 --> 01:22:36,159
sometimes these lessons that are told to

2181
01:22:41,350 --> 01:22:39,440
us by philosophers or spiritual gurus or

2182
01:22:43,910 --> 01:22:41,360
people who seem like spiritual gurus

2183
01:22:45,910 --> 01:22:43,920
perhaps they should be taken with such a

2184
01:22:47,669 --> 01:22:45,920
grain of salt and i've come to the

2185
01:22:49,669 --> 01:22:47,679
conclusion that if it's not life

2186
01:22:52,070 --> 01:22:49,679
affirming then perhaps don't follow that

2187
01:22:54,790 --> 01:22:52,080
perhaps use that as a clue that that's a

2188
01:22:57,110 --> 01:22:54,800
pathological path and don't do anything

2189
01:22:58,310 --> 01:22:57,120
that can be drastic don't give up

2190
01:22:59,510 --> 01:22:58,320
because i felt like i was in a state

2191
01:23:01,270 --> 01:22:59,520
where i could

2192
01:23:03,350 --> 01:23:01,280
snap and feel like well this is all a

2193
01:23:05,830 --> 01:23:03,360
dream i remember feeling looking at my

2194
01:23:07,270 --> 01:23:05,840
arms and feeling like am i even behind

2195
01:23:08,870 --> 01:23:07,280
like right now i'm feeling great and

2196
01:23:11,030 --> 01:23:08,880
sorry i don't mean i seem like it but

2197
01:23:12,470 --> 01:23:11,040
i'm speaking with you i feel safe i love

2198
01:23:14,950 --> 01:23:12,480
my wife i've actually whatever i'm

2199
01:23:16,790 --> 01:23:14,960
feeling happy right now and i feel like

2200
01:23:18,830 --> 01:23:16,800
i would a month ago which is i'm behind

2201
01:23:21,110 --> 01:23:18,840
my eyes and this is a person

2202
01:23:23,990 --> 01:23:21,120
speaking a week two weeks ago three

2203
01:23:25,350 --> 01:23:24,000
weeks ago i was feeling like

2204
01:23:26,870 --> 01:23:25,360
all of that could be a model i could

2205
01:23:29,030 --> 01:23:26,880
emulate anyone else's head i could be

2206
01:23:30,350 --> 01:23:29,040
anyone else this could all be in my head

2207
01:23:33,030 --> 01:23:30,360
so

2208
01:23:35,110 --> 01:23:33,040
debilitating destabilizing

2209
01:23:37,189 --> 01:23:35,120
walking around feeling like is this a

2210
01:23:38,470 --> 01:23:37,199
dream constantly checking my fingers

2211
01:23:39,990 --> 01:23:38,480
because that's one of the signs you can

2212
01:23:41,350 --> 01:23:40,000
check your for lucid dreaming how many

2213
01:23:43,030 --> 01:23:41,360
fingers do you have and then also

2214
01:23:45,110 --> 01:23:43,040
feeling bad like why the heck am i

2215
01:23:46,070 --> 01:23:45,120
thinking this is a dream shoot what if

2216
01:23:47,350 --> 01:23:46,080
it is

2217
01:23:48,470 --> 01:23:47,360
just

2218
01:23:54,390 --> 01:23:48,480
spiraling

2219
01:23:56,310 --> 01:23:54,400
okay that's what i had slash still have

2220
01:23:57,910 --> 01:23:56,320
to some degree though not right now i

2221
01:23:59,430 --> 01:23:57,920
wouldn't say that i'm over it because it

2222
01:24:01,350 --> 01:23:59,440
was such a recent experience i'm not

2223
01:24:03,830 --> 01:24:01,360
going to say that i've solved it i feel

2224
01:24:05,590 --> 01:24:03,840
like it's tapering off luckily anyway

2225
01:24:07,270 --> 01:24:05,600
that's what i had i would characterize

2226
01:24:08,950 --> 01:24:07,280
that as derealization in the sense that

2227
01:24:10,629 --> 01:24:08,960
i didn't feel like it was real i don't

2228
01:24:11,750 --> 01:24:10,639
know if it's depersonalization because i

2229
01:24:13,669 --> 01:24:11,760
don't know the difference between those

2230
01:24:15,590 --> 01:24:13,679
two nor dissociative identity disorders

2231
01:24:16,470 --> 01:24:15,600
so i don't know the differences between

2232
01:24:18,709 --> 01:24:16,480
those

2233
01:24:20,629 --> 01:24:18,719
well it's good it's nice you bring all

2234
01:24:22,470 --> 01:24:20,639
those three

2235
01:24:24,229 --> 01:24:22,480
their phenomenology is very very closely

2236
01:24:25,510 --> 01:24:24,239
related but you know

2237
01:24:29,510 --> 01:24:25,520
the way it described that that was

2238
01:24:35,669 --> 01:24:33,350
either um a de-realization um

2239
01:24:38,950 --> 01:24:35,679
which people normally described uh

2240
01:24:41,669 --> 01:24:38,960
describe as they're looking um they are

2241
01:24:44,070 --> 01:24:41,679
sensing um something that doesn't quite

2242
01:24:46,310 --> 01:24:44,080
seem real that you know the outside

2243
01:24:48,950 --> 01:24:46,320
world seems like it's a movie it's not

2244
01:24:51,910 --> 01:24:48,960
it's not it's lost that tangibility

2245
01:24:53,189 --> 01:24:51,920
there's no um there's no grip that can

2246
01:24:55,510 --> 01:24:53,199
be attained

2247
01:24:57,510 --> 01:24:55,520
that reassures you that stuff is

2248
01:25:00,149 --> 01:24:57,520
actually out there the other the

2249
01:25:03,110 --> 01:25:00,159
converse of course is i'm um i'm no

2250
01:25:04,870 --> 01:25:03,120
longer a person um and that could i

2251
01:25:07,669 --> 01:25:04,880
think he'd be even more frightening and

2252
01:25:11,510 --> 01:25:07,679
both of these i think speak to um

2253
01:25:14,470 --> 01:25:11,520
speak to um this dissociation that comes

2254
01:25:17,510 --> 01:25:14,480
with a dissolution of selfhood or at

2255
01:25:23,430 --> 01:25:20,149
construction of explanations for the

2256
01:25:25,590 --> 01:25:23,440
self in a lived world

2257
01:25:28,390 --> 01:25:25,600
you used a phrase which

2258
01:25:30,149 --> 01:25:28,400
i i think was was quite uh pertinent

2259
01:25:32,229 --> 01:25:30,159
here which is you know

2260
01:25:34,950 --> 01:25:32,239
one should not pursue this kind of

2261
01:25:38,149 --> 01:25:34,960
thinking and self-exploration and

2262
01:25:41,510 --> 01:25:38,159
introspection unless it is um

2263
01:25:44,070 --> 01:25:41,520
affirmatory and and self-affirmatory and

2264
01:25:46,310 --> 01:25:44,080
in a sense that's exactly from the sort

2265
01:25:48,550 --> 01:25:46,320
of mathematical and dry perspective of

2266
01:25:50,390 --> 01:25:48,560
the free energy principle that just is

2267
01:25:53,189 --> 01:25:50,400
maximizing model evidence this is

2268
01:25:55,110 --> 01:25:53,199
self-evidencing it is you know if you

2269
01:25:56,870 --> 01:25:55,120
can affirm

2270
01:25:59,189 --> 01:25:56,880
through all the right kind of

2271
01:26:01,110 --> 01:25:59,199
interactions be the emotional social

2272
01:26:04,310 --> 01:26:01,120
sexual um

2273
01:26:06,550 --> 01:26:04,320
intellectual um the the this you've got

2274
01:26:08,709 --> 01:26:06,560
the right kind of model of your world

2275
01:26:10,709 --> 01:26:08,719
which is constituted by people like your

2276
01:26:13,270 --> 01:26:10,719
interviewees and your wife and all your

2277
01:26:17,110 --> 01:26:13,280
colleagues and then

2278
01:26:20,950 --> 01:26:17,120
affirming that model simply is securing

2279
01:26:24,709 --> 01:26:20,960
um evidence for that good model of your

2280
01:26:26,550 --> 01:26:24,719
rich and uncultured world that just is

2281
01:26:28,310 --> 01:26:26,560
that's that

2282
01:26:31,189 --> 01:26:28,320
that sort of um

2283
01:26:32,149 --> 01:26:31,199
affirmatory uh process

2284
01:26:33,750 --> 01:26:32,159
however

2285
01:26:36,070 --> 01:26:33,760
there are dangers

2286
01:26:38,149 --> 01:26:36,080
that you um that lurk in the shadows

2287
01:26:39,669 --> 01:26:38,159
which you've clearly contended with to

2288
01:26:42,310 --> 01:26:39,679
do that properly

2289
01:26:44,550 --> 01:26:42,320
to be able to secure the right kind of

2290
01:26:48,830 --> 01:26:44,560
evidence that you are

2291
01:26:51,270 --> 01:26:48,840
living in a world of positive

2292
01:26:52,709 --> 01:26:51,280
um constructive

2293
01:26:53,830 --> 01:26:52,719
nourishing

2294
01:26:56,149 --> 01:26:53,840
enabling

2295
01:26:56,950 --> 01:26:56,159
human beings others other things like

2296
01:27:04,310 --> 01:26:56,960
you

2297
01:27:05,189 --> 01:27:04,320
entertain the idea

2298
01:27:09,750 --> 01:27:05,199
that

2299
01:27:11,750 --> 01:27:09,760
there is there is self and other um and

2300
01:27:14,149 --> 01:27:11,760
simply uh you know to engage in turn

2301
01:27:17,590 --> 01:27:14,159
taking you need to have this um this

2302
01:27:21,189 --> 01:27:17,600
explicit part of your charity model um

2303
01:27:27,189 --> 01:27:25,270
in so doing you will also have um

2304
01:27:29,270 --> 01:27:27,199
contextualize as part of your junior

2305
01:27:31,669 --> 01:27:29,280
model itself in a number of different

2306
01:27:33,590 --> 01:27:31,679
states uh and you need to have remember

2307
01:27:36,310 --> 01:27:33,600
before we talk about the good scientists

2308
01:27:38,709 --> 01:27:36,320
um having a portfolio of hypotheses

2309
01:27:40,470 --> 01:27:38,719
we're generating a new set of hypotheses

2310
01:27:43,350 --> 01:27:40,480
so that we can now gather evidence of

2311
01:27:45,270 --> 01:27:43,360
these hypotheses so that's part of uh

2312
01:27:47,750 --> 01:27:45,280
what you are doing i mean you're you

2313
01:27:48,629 --> 01:27:47,760
know in in doing these podcasts you are

2314
01:27:49,830 --> 01:27:48,639
on

2315
01:27:52,629 --> 01:27:49,840
a journey

2316
01:27:54,870 --> 01:27:52,639
of um developing different hypotheses

2317
01:27:57,030 --> 01:27:54,880
about ways of being

2318
01:27:59,270 --> 01:27:57,040
the the danger though when it comes to

2319
01:28:01,110 --> 01:27:59,280
different ways of

2320
01:28:02,390 --> 01:28:01,120
me as a self

2321
01:28:05,189 --> 01:28:02,400
existing

2322
01:28:06,149 --> 01:28:05,199
is that you can sometimes

2323
01:28:08,709 --> 01:28:06,159
lose

2324
01:28:10,950 --> 01:28:08,719
entertain the null hypothesis that i am

2325
01:28:13,910 --> 01:28:10,960
not a self or i do not exist in a

2326
01:28:15,910 --> 01:28:13,920
sentient fashion in a in the fashion of

2327
01:28:18,310 --> 01:28:15,920
having particular kinds of qualitative

2328
01:28:20,149 --> 01:28:18,320
experiences this is a perfectly viable

2329
01:28:22,390 --> 01:28:20,159
hypothesis it can be had in a

2330
01:28:23,990 --> 01:28:22,400
non-emotional context when you meet

2331
01:28:25,590 --> 01:28:24,000
people like philosophers entertain

2332
01:28:27,910 --> 01:28:25,600
things like the you know the zombie

2333
01:28:29,910 --> 01:28:27,920
hypothesis or brain avatar thought

2334
01:28:32,790 --> 01:28:29,920
experiments these are all really

2335
01:28:35,430 --> 01:28:32,800
interesting interesting instances

2336
01:28:37,669 --> 01:28:35,440
of a very high order life form

2337
01:28:40,229 --> 01:28:37,679
thinking about what would have been like

2338
01:28:42,310 --> 01:28:40,239
if i was maybe without me hood if i was

2339
01:28:44,790 --> 01:28:42,320
a zombie what would it be like if it

2340
01:28:46,790 --> 01:28:44,800
wasn't actually me talking

2341
01:28:49,110 --> 01:28:46,800
what would it be like if there was no

2342
01:28:51,189 --> 01:28:49,120
reality it was all just me all of these

2343
01:28:54,790 --> 01:28:51,199
alternative ways of being

2344
01:28:56,790 --> 01:28:54,800
that speak to selfhood or it's negation

2345
01:28:58,629 --> 01:28:56,800
or it's nihilism

2346
01:29:00,790 --> 01:28:58,639
now become plausible hypotheses for

2347
01:29:03,590 --> 01:29:00,800
which you have to search for evidence

2348
01:29:04,950 --> 01:29:03,600
and of course if you are just doing this

2349
01:29:06,229 --> 01:29:04,960
in your head and ruminating and

2350
01:29:08,709 --> 01:29:06,239
introspecting there's going to be very

2351
01:29:11,669 --> 01:29:08,719
little evidence available to you

2352
01:29:14,870 --> 01:29:11,679
to disaffirm

2353
01:29:16,709 --> 01:29:14,880
your hypothesis perhaps i'm not me um if

2354
01:29:18,390 --> 01:29:16,719
you don't actually engage with somebody

2355
01:29:19,669 --> 01:29:18,400
else there is no evidence

2356
01:29:22,149 --> 01:29:19,679
at hand

2357
01:29:25,990 --> 01:29:22,159
for to refute their hypothesis

2358
01:29:28,310 --> 01:29:26,000
so on the one hand that is a remarkable

2359
01:29:31,430 --> 01:29:28,320
capacity the very fact you can have this

2360
01:29:33,430 --> 01:29:31,440
hypothesis in your head i am not me uh

2361
01:29:35,430 --> 01:29:33,440
or it is not me having these thoughts it

2362
01:29:36,870 --> 01:29:35,440
is not me having these quantitative

2363
01:29:38,790 --> 01:29:36,880
experiences

2364
01:29:41,510 --> 01:29:38,800
it's quite remarkable it's a very very

2365
01:29:44,149 --> 01:29:41,520
high level um ability which which i

2366
01:29:47,910 --> 01:29:44,159
would imagine where you you and i are

2367
01:29:49,669 --> 01:29:47,920
the only kinds of um sentient artifacts

2368
01:29:51,990 --> 01:29:49,679
in the universe that can entertain that

2369
01:29:54,709 --> 01:29:52,000
kind of counterfactual hypothesis which

2370
01:29:56,310 --> 01:29:54,719
is necessary to explore different models

2371
01:29:58,790 --> 01:29:56,320
it's necessary indeed to actually just

2372
01:30:00,390 --> 01:29:58,800
behave and have plans you have to have

2373
01:30:03,030 --> 01:30:00,400
counterfactual hypothesis about the

2374
01:30:06,870 --> 01:30:03,040
future and different ways of being

2375
01:30:07,910 --> 01:30:06,880
because it comes with um it comes with a

2376
01:30:10,709 --> 01:30:07,920
price

2377
01:30:12,950 --> 01:30:10,719
if sometimes you

2378
01:30:15,990 --> 01:30:12,960
entertain a hypothesis for which there

2379
01:30:18,709 --> 01:30:16,000
is no evidence on hand that could

2380
01:30:21,189 --> 01:30:18,719
possibly refute it

2381
01:30:23,350 --> 01:30:21,199
and certainly if you adopt a particular

2382
01:30:25,669 --> 01:30:23,360
behavioral strategy or engagement with

2383
01:30:27,590 --> 01:30:25,679
the world that precludes

2384
01:30:30,310 --> 01:30:27,600
soliciting evidence against the

2385
01:30:34,790 --> 01:30:30,320
hypothesis that you are say you

2386
01:30:36,470 --> 01:30:34,800
um so examples of that would be

2387
01:30:39,590 --> 01:30:36,480
things like obsessional compulsive

2388
01:30:42,070 --> 01:30:39,600
disorder or agrophobia or

2389
01:30:45,669 --> 01:30:42,080
your rumination disorders

2390
01:30:47,430 --> 01:30:45,679
where you physically um or possibly even

2391
01:30:49,510 --> 01:30:47,440
prosocially

2392
01:30:51,030 --> 01:30:49,520
prevent yourself from acquiring any

2393
01:30:52,470 --> 01:30:51,040
further information that would allow you

2394
01:30:54,390 --> 01:30:52,480
to change your mind

2395
01:30:55,830 --> 01:30:54,400
and of course once you realize that's a

2396
01:30:57,510 --> 01:30:55,840
possibility

2397
01:31:00,709 --> 01:30:57,520
you um

2398
01:31:02,550 --> 01:31:00,719
you can always or you can also get into

2399
01:31:05,110 --> 01:31:02,560
the negatively balanced emotional

2400
01:31:08,870 --> 01:31:05,120
aspects of entertaining these hypotheses

2401
01:31:11,350 --> 01:31:08,880
so if i've got schizophrenia

2402
01:31:13,430 --> 01:31:11,360
that now offers a very plausible

2403
01:31:14,950 --> 01:31:13,440
hypothesis for these dissociative

2404
01:31:16,870 --> 01:31:14,960
experiences

2405
01:31:19,189 --> 01:31:16,880
why am i having these dissociative

2406
01:31:21,430 --> 01:31:19,199
experiences well one very plausible

2407
01:31:23,750 --> 01:31:21,440
hypothesis i am having i have a

2408
01:31:25,750 --> 01:31:23,760
dissociative order or i have a

2409
01:31:27,990 --> 01:31:25,760
disorganized thought disorder that

2410
01:31:30,629 --> 01:31:28,000
characterizes schizophrenia so this now

2411
01:31:32,629 --> 01:31:30,639
becomes a very plausible hypothesis

2412
01:31:34,950 --> 01:31:32,639
how would you go secure evidence against

2413
01:31:37,430 --> 01:31:34,960
that um well it becomes very difficult

2414
01:31:38,390 --> 01:31:37,440
because the whole prop point of having

2415
01:31:40,310 --> 01:31:38,400
these

2416
01:31:43,590 --> 01:31:40,320
very high order counter factual

2417
01:31:45,910 --> 01:31:43,600
hypotheses or models at hand to explain

2418
01:31:48,550 --> 01:31:45,920
your experienced world

2419
01:31:49,669 --> 01:31:48,560
um is that the the

2420
01:31:54,070 --> 01:31:49,679
they're very

2421
01:31:55,990 --> 01:31:54,080
and it's very difficult um uh if you

2422
01:31:57,669 --> 01:31:56,000
have got schizophrenia you're not going

2423
01:31:58,870 --> 01:31:57,679
to be able to trust any information

2424
01:32:01,669 --> 01:31:58,880
anywhere because they're correct

2425
01:32:04,310 --> 01:32:01,679
illusion hallucinatory so it's one of

2426
01:32:07,430 --> 01:32:04,320
these wonderful um self-fulfilling

2427
01:32:10,310 --> 01:32:07,440
hypotheses that are very difficult to uh

2428
01:32:12,149 --> 01:32:10,320
to dismantle um and you see the same

2429
01:32:14,550 --> 01:32:12,159
phenomenology

2430
01:32:17,110 --> 01:32:14,560
in in many instances

2431
01:32:19,590 --> 01:32:17,120
you mentioned behavior therapy before

2432
01:32:22,070 --> 01:32:19,600
and um in a way that doesn't have this

2433
01:32:24,229 --> 01:32:22,080
quite the same horrible existential

2434
01:32:27,510 --> 01:32:24,239
angst that you clearly experienced you

2435
01:32:29,189 --> 01:32:27,520
can actually construe um many phobias

2436
01:32:31,750 --> 01:32:29,199
for example

2437
01:32:34,149 --> 01:32:31,760
of the same kind in the sense that if

2438
01:32:35,990 --> 01:32:34,159
you are very frightened of spiders

2439
01:32:38,950 --> 01:32:36,000
you're very unlikely

2440
01:32:40,709 --> 01:32:38,960
to actually go and solicit the evidence

2441
01:32:44,229 --> 01:32:40,719
that would

2442
01:32:46,149 --> 01:32:44,239
hypothesis that spiders are not

2443
01:32:48,790 --> 01:32:46,159
frightening because you just don't go

2444
01:32:51,669 --> 01:32:48,800
near spiders so this is the kind of

2445
01:32:54,550 --> 01:32:51,679
hypothesis that maintains itself because

2446
01:32:56,870 --> 01:32:54,560
it's undermined the way that you acted

2447
01:33:00,390 --> 01:32:56,880
this list of evidence for your beliefs

2448
01:33:01,669 --> 01:33:00,400
but for you of course you you know

2449
01:33:03,750 --> 01:33:01,679
you know it's not a question of being

2450
01:33:06,310 --> 01:33:03,760
frightened of spiders it's a much more

2451
01:33:07,910 --> 01:33:06,320
existentially deep and worrisome notion

2452
01:33:10,310 --> 01:33:07,920
that in fact you're not you or you're

2453
01:33:13,270 --> 01:33:10,320
not even yourself um and you're not even

2454
01:33:16,470 --> 01:33:13,280
perceiving reality and these kinds of

2455
01:33:19,189 --> 01:33:16,480
hypotheses that come along qualified by

2456
01:33:23,830 --> 01:33:19,199
or associated with um plausible

2457
01:33:25,350 --> 01:33:23,840
explanations i'm going mad um um

2458
01:33:27,990 --> 01:33:25,360
are very difficult to refute in the

2459
01:33:30,870 --> 01:33:28,000
moment especially if you if you ruminate

2460
01:33:32,629 --> 01:33:30,880
on them um they will go away as soon as

2461
01:33:33,990 --> 01:33:32,639
you start re-engaging with with other

2462
01:33:36,310 --> 01:33:34,000
people

2463
01:33:37,590 --> 01:33:36,320
that's what i found to be the case carl

2464
01:33:38,950 --> 01:33:37,600
is that the more i talk about it the

2465
01:33:41,669 --> 01:33:38,960
more i'm speaking with you than just

2466
01:33:43,910 --> 01:33:41,679
speaking with people and not being so in

2467
01:33:46,310 --> 01:33:43,920
my head almost instantaneously it starts

2468
01:33:47,750 --> 01:33:46,320
to dissipate and also just so you know

2469
01:33:50,390 --> 01:33:47,760
there's a danger for the people who

2470
01:33:53,910 --> 01:33:50,400
watch this channel that it's seen as a

2471
01:33:56,950 --> 01:33:53,920
spiritual triumphant state to be egoless

2472
01:33:58,709 --> 01:33:56,960
to have ego death so then just so you

2473
01:34:01,030 --> 01:33:58,719
know when i was feeling that i was then

2474
01:34:03,110 --> 01:34:01,040
telling myself but am i supposed to have

2475
01:34:05,110 --> 01:34:03,120
ego death is this what ego death is like

2476
01:34:06,790 --> 01:34:05,120
and now i feel bad am i just not strong

2477
01:34:08,390 --> 01:34:06,800
enough as a person which then i feel

2478
01:34:10,390 --> 01:34:08,400
like i'm less on the social hierarchy

2479
01:34:12,390 --> 01:34:10,400
because i'm just a coward who cannot

2480
01:34:14,149 --> 01:34:12,400
deal with

2481
01:34:15,750 --> 01:34:14,159
who isn't psychologically strong enough

2482
01:34:17,750 --> 01:34:15,760
to deal with potential ego death if that

2483
01:34:19,270 --> 01:34:17,760
was even ego death

2484
01:34:21,030 --> 01:34:19,280
so that those are there's so many

2485
01:34:22,790 --> 01:34:21,040
dangers in this and i didn't realize

2486
01:34:25,669 --> 01:34:22,800
this i had this insight was in that

2487
01:34:27,910 --> 01:34:25,679
episode or experience that kurt kurkert

2488
01:34:29,430 --> 01:34:27,920
kurt something was like i was talking to

2489
01:34:31,510 --> 01:34:29,440
myself or something was talking to me

2490
01:34:33,830 --> 01:34:31,520
and telling me kurt you've been thinking

2491
01:34:35,350 --> 01:34:33,840
way too much this is you're extremely

2492
01:34:36,870 --> 01:34:35,360
analytical i never would have thought of

2493
01:34:39,510 --> 01:34:36,880
myself as someone who's too analytical

2494
01:34:41,750 --> 01:34:39,520
before it was clear to me that i was too

2495
01:34:42,790 --> 01:34:41,760
analytical and there's nothing wrong

2496
01:34:43,990 --> 01:34:42,800
with being analytical there's nothing

2497
01:34:46,629 --> 01:34:44,000
wrong with thinking but you've been

2498
01:34:49,189 --> 01:34:46,639
thinking a bit too much feel more just

2499
01:34:51,270 --> 01:34:49,199
ground yourself here engage in this

2500
01:34:53,189 --> 01:34:51,280
world more

2501
01:34:55,030 --> 01:34:53,199
and and there is also this feeling that

2502
01:34:57,030 --> 01:34:55,040
i don't love myself which is a strange

2503
01:34:59,270 --> 01:34:57,040
feeling carl that i don't have that i'm

2504
01:35:00,709 --> 01:34:59,280
a bit hard on myself and i would have

2505
01:35:02,470 --> 01:35:00,719
never said that before i would have

2506
01:35:04,229 --> 01:35:02,480
thought i'm not hard enough i need to

2507
01:35:06,070 --> 01:35:04,239
study harder i need to prepare for these

2508
01:35:09,189 --> 01:35:06,080
podcasts harder i need to release more i

2509
01:35:10,950 --> 01:35:09,199
need to learn more push push push push

2510
01:35:13,350 --> 01:35:10,960
it was only then that i felt like kurt

2511
01:35:14,390 --> 01:35:13,360
that's been dangerous man like just

2512
01:35:16,870 --> 01:35:14,400
relax

2513
01:35:18,229 --> 01:35:16,880
and and don't worry too much and feeling

2514
01:35:20,790 --> 01:35:18,239
keep talking about this somewhat

2515
01:35:22,149 --> 01:35:20,800
publicly is a bit embarrassing because i

2516
01:35:23,510 --> 01:35:22,159
do feel like

2517
01:35:24,870 --> 01:35:23,520
by saying this like there's still that

2518
01:35:26,229 --> 01:35:24,880
part of me that feels like am i saying

2519
01:35:28,070 --> 01:35:26,239
this to give my way out because i'm

2520
01:35:29,669 --> 01:35:28,080
truly lazy and i want to say this so i

2521
01:35:31,910 --> 01:35:29,679
get the sympathy of for example you

2522
01:35:33,910 --> 01:35:31,920
maybe the audience to say kurt no no

2523
01:35:36,149 --> 01:35:33,920
you're doing great so then i can get i

2524
01:35:39,270 --> 01:35:36,159
can feel there's so much self analyzing

2525
01:35:41,430 --> 01:35:39,280
that's going on right here carl

2526
01:35:44,709 --> 01:35:41,440
anyway i felt like kurt you were

2527
01:35:47,750 --> 01:35:44,719
thinking too much just relax perhaps

2528
01:35:49,910 --> 01:35:47,760
don't engage as deeply as you have been

2529
01:35:51,910 --> 01:35:49,920
with these ideas don't entertain them as

2530
01:35:53,669 --> 01:35:51,920
much you maybe you're in a fragile state

2531
01:35:56,149 --> 01:35:53,679
right now you could do that later right

2532
01:35:57,830 --> 01:35:56,159
now take it back that's what came over

2533
01:35:59,750 --> 01:35:57,840
me and also this lesson that i don't

2534
01:36:01,189 --> 01:35:59,760
have enough self-love or self-acceptance

2535
01:36:02,629 --> 01:36:01,199
last night i looked up i was using a

2536
01:36:04,550 --> 01:36:02,639
rubber band for my thoughts like

2537
01:36:05,830 --> 01:36:04,560
self-administered aversive conditioning

2538
01:36:07,189 --> 01:36:05,840
when you snap yourself if you're

2539
01:36:08,470 --> 01:36:07,199
performing a bad habit so i thought

2540
01:36:10,149 --> 01:36:08,480
maybe i could do that with my thoughts

2541
01:36:11,990 --> 01:36:10,159
so anytime that i feel like that that

2542
01:36:14,070 --> 01:36:12,000
would encourage my mind to not feel like

2543
01:36:15,590 --> 01:36:14,080
that i looked it up and it said that

2544
01:36:17,990 --> 01:36:15,600
works for physical habits but not for

2545
01:36:21,510 --> 01:36:18,000
mental habits so stop that and what

2546
01:36:22,390 --> 01:36:21,520
works is act i believe acceptance

2547
01:36:24,709 --> 01:36:22,400
and

2548
01:36:27,430 --> 01:36:24,719
commitment therapy which is a cbt

2549
01:36:29,270 --> 01:36:27,440
technique so then last night i was

2550
01:36:31,189 --> 01:36:29,280
telling myself you know what try that

2551
01:36:32,629 --> 01:36:31,199
out accept these thoughts and i'm not

2552
01:36:34,870 --> 01:36:32,639
exactly sure what it means to accept

2553
01:36:37,430 --> 01:36:34,880
them i don't precisely think it means

2554
01:36:39,830 --> 01:36:37,440
entertain them but at least meet them

2555
01:36:41,990 --> 01:36:39,840
not with horror and dread and avoidance

2556
01:36:43,350 --> 01:36:42,000
but just say that's okay man you think

2557
01:36:45,430 --> 01:36:43,360
like that some people feel like that

2558
01:36:47,189 --> 01:36:45,440
sometimes so i started doing that and

2559
01:36:49,109 --> 01:36:47,199
last night was the first time in the

2560
01:36:52,390 --> 01:36:49,119
three weeks since i've had this that i

2561
01:36:54,229 --> 01:36:52,400
felt absolutely like my regular self

2562
01:36:56,629 --> 01:36:54,239
almost from five minutes of this quote

2563
01:36:57,990 --> 01:36:56,639
unquote acceptance therapy

2564
01:36:59,910 --> 01:36:58,000
and then i woke up this morning felt

2565
01:37:02,550 --> 01:36:59,920
like my regular self like i'm behind my

2566
01:37:03,990 --> 01:37:02,560
eyes i have i'm in this world that was

2567
01:37:06,310 --> 01:37:04,000
fascinating

2568
01:37:07,510 --> 01:37:06,320
okay so i said quite a few statements

2569
01:37:09,189 --> 01:37:07,520
right there i'm sure there's plenty you

2570
01:37:11,030 --> 01:37:09,199
want to comment on what i'm also

2571
01:37:12,709 --> 01:37:11,040
interested in for you to put a pin in

2572
01:37:14,790 --> 01:37:12,719
your hat is what the heck does this

2573
01:37:16,390 --> 01:37:14,800
self-acceptance self-love if that seemed

2574
01:37:18,870 --> 01:37:16,400
the same mean in terms of the free

2575
01:37:22,229 --> 01:37:18,880
energy principle okay so i basically

2576
01:37:24,070 --> 01:37:22,239
just poured myself out there

2577
01:37:25,830 --> 01:37:24,080
there you go sorry carl for you to pick

2578
01:37:27,350 --> 01:37:25,840
up all those pieces i know it's a huge

2579
01:37:28,390 --> 01:37:27,360
responsibility i trust you you're

2580
01:37:30,470 --> 01:37:28,400
someone who studies this you're

2581
01:37:32,870 --> 01:37:30,480
extremely extremely bright and you have

2582
01:37:34,550 --> 01:37:32,880
such a wealth of knowledge and i do feel

2583
01:37:37,270 --> 01:37:34,560
like what i'm going through is not

2584
01:37:38,870 --> 01:37:37,280
something terribly unique i do feel like

2585
01:37:41,030 --> 01:37:38,880
this is something that many people who

2586
01:37:42,870 --> 01:37:41,040
are on this journey of understanding the

2587
01:37:45,350 --> 01:37:42,880
world and consciousness in their place

2588
01:37:47,270 --> 01:37:45,360
go through this perhaps at some point

2589
01:37:49,430 --> 01:37:47,280
i think i was trying to emphasize that

2590
01:37:51,189 --> 01:37:49,440
by saying that you know it is a gift

2591
01:37:54,390 --> 01:37:51,199
that you're able to have these

2592
01:37:56,709 --> 01:37:54,400
existential crises um your there are

2593
01:37:59,430 --> 01:37:56,719
dangers that lurk which you've clearly

2594
01:38:01,030 --> 01:37:59,440
encountered but at the end of the day

2595
01:38:03,270 --> 01:38:01,040
the very fact that you can entertain

2596
01:38:06,070 --> 01:38:03,280
these counterfactuals

2597
01:38:08,870 --> 01:38:06,080
is you know quite remarkable you know

2598
01:38:10,470 --> 01:38:08,880
to my mind it's it would be the highest

2599
01:38:13,510 --> 01:38:10,480
expression of the human condition just

2600
01:38:15,510 --> 01:38:13,520
to consider we are those alternative

2601
01:38:18,070 --> 01:38:15,520
hypotheses where bits of our humanity

2602
01:38:20,390 --> 01:38:18,080
are just not there anymore or have a a

2603
01:38:22,709 --> 01:38:20,400
different disposition or relationship to

2604
01:38:24,229 --> 01:38:22,719
reality or indeed other people but lots

2605
01:38:25,270 --> 01:38:24,239
of things you said it made a lot of

2606
01:38:27,750 --> 01:38:25,280
sense

2607
01:38:29,990 --> 01:38:27,760
from simple things like you know

2608
01:38:32,310 --> 01:38:30,000
soliciting reassurance from others that

2609
01:38:35,109 --> 01:38:32,320
you don't have to do

2610
01:38:37,830 --> 01:38:35,119
any more work in terms of preparing

2611
01:38:41,270 --> 01:38:37,840
yourself for being a good a good uh

2612
01:38:44,550 --> 01:38:41,280
expounder of ideas or or interviewer or

2613
01:38:46,310 --> 01:38:44,560
book writer or or whatever and you know

2614
01:38:49,270 --> 01:38:46,320
that's exactly what you should be doing

2615
01:38:52,070 --> 01:38:49,280
that's that's exactly securing evidence

2616
01:38:53,109 --> 01:38:52,080
for yourself models of an affirmatory

2617
01:38:55,590 --> 01:38:53,119
sort

2618
01:38:57,510 --> 01:38:55,600
and the more you do that then the more

2619
01:38:59,750 --> 01:38:57,520
the more that model will

2620
01:39:02,310 --> 01:38:59,760
be fit for purpose and you will have the

2621
01:39:04,629 --> 01:39:02,320
right kind of marginal maximizing your

2622
01:39:06,070 --> 01:39:04,639
marginal likelihood but

2623
01:39:07,910 --> 01:39:06,080
you're in a difficult position because

2624
01:39:09,510 --> 01:39:07,920
your job is actually to explore other

2625
01:39:11,270 --> 01:39:09,520
ways of making sense of things and other

2626
01:39:13,750 --> 01:39:11,280
ways of thinking things like you're

2627
01:39:17,350 --> 01:39:13,760
compelled to explore alternative

2628
01:39:19,990 --> 01:39:17,360
hypotheses um and you know you you've

2629
01:39:21,830 --> 01:39:20,000
described um

2630
01:39:22,870 --> 01:39:21,840
the you know the potential

2631
01:39:30,629 --> 01:39:22,880
um

2632
01:39:32,310 --> 01:39:30,639
can't now remember my nightmares but i

2633
01:39:34,310 --> 01:39:32,320
imagine that all of us do go through

2634
01:39:35,990 --> 01:39:34,320
this and we have nightmares as children

2635
01:39:38,870 --> 01:39:36,000
i think you stop having true nightmares

2636
01:39:40,709 --> 01:39:38,880
in your 30s and 40s but uh you know

2637
01:39:43,189 --> 01:39:40,719
that sort of you know before you've

2638
01:39:43,990 --> 01:39:43,199
really got that grip and maintain that

2639
01:39:44,950 --> 01:39:44,000
grip

2640
01:39:46,870 --> 01:39:44,960
on

2641
01:39:50,070 --> 01:39:46,880
self-hood and

2642
01:39:52,709 --> 01:39:50,080
you in your in your lived uh in your

2643
01:39:54,550 --> 01:39:52,719
lived world as part of that and engage

2644
01:39:57,189 --> 01:39:54,560
with that world before you've got there

2645
01:39:57,990 --> 01:39:57,199
you know it must be absolutely awful and

2646
01:40:00,470 --> 01:39:58,000
you know

2647
01:40:02,149 --> 01:40:00,480
nihilistic to to occasionally lose that

2648
01:40:03,669 --> 01:40:02,159
and then you know worrying that you're

2649
01:40:06,070 --> 01:40:03,679
never going to get it back it does

2650
01:40:07,510 --> 01:40:06,080
remind me of certain um

2651
01:40:09,669 --> 01:40:07,520
some

2652
01:40:11,430 --> 01:40:09,679
drug-induced states you know i i can

2653
01:40:12,950 --> 01:40:11,440
imagine that you're the kind of person

2654
01:40:15,830 --> 01:40:12,960
who would be very afraid of having a

2655
01:40:17,510 --> 01:40:15,840
very bad trip um and i think very

2656
01:40:20,070 --> 01:40:17,520
frightened of having one

2657
01:40:22,310 --> 01:40:20,080
yeah yeah so that is correct and luckily

2658
01:40:23,510 --> 01:40:22,320
i wasn't well what i imagined what i was

2659
01:40:25,830 --> 01:40:23,520
going through would

2660
01:40:27,910 --> 01:40:25,840
feels like how someone would feel on a

2661
01:40:28,709 --> 01:40:27,920
on a terrible psychedelic trip

2662
01:40:30,149 --> 01:40:28,719
yes

2663
01:40:32,229 --> 01:40:30,159
um and it's interesting you bring up

2664
01:40:33,510 --> 01:40:32,239
psychedelics because of course you know

2665
01:40:35,830 --> 01:40:33,520
um

2666
01:40:38,149 --> 01:40:35,840
there is a move

2667
01:40:40,390 --> 01:40:38,159
in sort of pharmacologically assisted

2668
01:40:45,109 --> 01:40:40,400
talking therapists you know to to use

2669
01:40:47,189 --> 01:40:45,119
psychedelics simply um not to aspire to

2670
01:40:48,870 --> 01:40:47,199
i don't know what ego death means but i

2671
01:40:51,750 --> 01:40:48,880
i i can have a guess at it from the

2672
01:40:53,510 --> 01:40:51,760
point of view of meditation and

2673
01:40:54,709 --> 01:40:53,520
mindfulness

2674
01:40:57,109 --> 01:40:54,719
but there's a really interesting

2675
01:41:00,070 --> 01:40:57,119
connection between

2676
01:41:02,229 --> 01:41:00,080
the use of psychedelics

2677
01:41:03,109 --> 01:41:02,239
and the aspiration

2678
01:41:07,270 --> 01:41:03,119
of

2679
01:41:09,590 --> 01:41:07,280
many um meditation-like practices um

2680
01:41:13,030 --> 01:41:09,600
that would i think subsume the intent

2681
01:41:15,669 --> 01:41:13,040
internal attention states um of um of

2682
01:41:18,149 --> 01:41:15,679
meditation and you know current

2683
01:41:20,470 --> 01:41:18,159
practices of mindfulness which is really

2684
01:41:22,870 --> 01:41:20,480
to try and redirect your attention

2685
01:41:24,470 --> 01:41:22,880
to the sensorium and usually intercepted

2686
01:41:25,669 --> 01:41:24,480
parts of it so it's like breathing for

2687
01:41:27,430 --> 01:41:25,679
example

2688
01:41:31,430 --> 01:41:27,440
so this is the exact opposite of what

2689
01:41:34,149 --> 01:41:32,709
deploying

2690
01:41:36,950 --> 01:41:34,159
that attention

2691
01:41:38,790 --> 01:41:36,960
all that um um

2692
01:41:40,870 --> 01:41:38,800
game control we're talking about in

2693
01:41:43,350 --> 01:41:40,880
terms of selecting

2694
01:41:47,270 --> 01:41:43,360
what kind of information sensory

2695
01:41:48,790 --> 01:41:47,280
information sensory states to engage and

2696
01:41:51,669 --> 01:41:48,800
um

2697
01:41:53,990 --> 01:41:51,679
determine your belief updating

2698
01:41:56,550 --> 01:41:54,000
so putting it putting all that attention

2699
01:41:59,590 --> 01:41:56,560
out to the sensory side of your deep

2700
01:42:01,590 --> 01:41:59,600
hierarchical models your your constructs

2701
01:42:03,590 --> 01:42:01,600
uh what you're doing is the opposite you

2702
01:42:06,709 --> 01:42:03,600
you you've been actually wandering

2703
01:42:10,390 --> 01:42:06,719
around at the deeper the highest levels

2704
01:42:12,629 --> 01:42:10,400
of these hierarchical models

2705
01:42:13,910 --> 01:42:12,639
sorry so in back when we were initially

2706
01:42:15,590 --> 01:42:13,920
talking and it said there's a black box

2707
01:42:17,669 --> 01:42:15,600
the input output and then there are two

2708
01:42:19,430 --> 01:42:17,679
black boxes it's as if you're saying pay

2709
01:42:20,709 --> 01:42:19,440
attention to the sensory forget about

2710
01:42:22,470 --> 01:42:20,719
the action we haven't talked much about

2711
01:42:24,709 --> 01:42:22,480
the action though obviously as in body

2712
01:42:26,070 --> 01:42:24,719
creatures sensory in action are tight

2713
01:42:28,070 --> 01:42:26,080
it's as if what i've been doing was

2714
01:42:29,590 --> 01:42:28,080
staying within the box and creating my

2715
01:42:31,590 --> 01:42:29,600
little sensations and actions within

2716
01:42:33,270 --> 01:42:31,600
there as little loops and you're saying

2717
01:42:34,310 --> 01:42:33,280
pay attention to the senses that come

2718
01:42:36,310 --> 01:42:34,320
from the outside is that what you're

2719
01:42:39,270 --> 01:42:36,320
saying or yes well no i'm not saying you

2720
01:42:40,709 --> 01:42:39,280
should do you you you will do

2721
01:42:42,470 --> 01:42:40,719
everything uh

2722
01:42:45,910 --> 01:42:42,480
in the right way and under the free

2723
01:42:47,750 --> 01:42:45,920
energy principle what i'm saying is that

2724
01:42:50,229 --> 01:42:47,760
what i'm just saying is that the

2725
01:42:51,669 --> 01:42:50,239
skilled practitioners of mindfulness and

2726
01:42:54,070 --> 01:42:51,679
meditation

2727
01:42:55,189 --> 01:42:54,080
i would imagine of the kind that would

2728
01:42:57,109 --> 01:42:55,199
lead to

2729
01:42:59,510 --> 01:42:57,119
ego death or

2730
01:43:01,990 --> 01:42:59,520
i'm not quite sure what that means um

2731
01:43:06,950 --> 01:43:03,669
would

2732
01:43:08,790 --> 01:43:06,960
employ exactly the same kind of

2733
01:43:11,590 --> 01:43:08,800
internally mediated and sometimes if

2734
01:43:12,629 --> 01:43:11,600
you're very skilled volitionally called

2735
01:43:14,550 --> 01:43:12,639
forth

2736
01:43:16,790 --> 01:43:14,560
mechanisms

2737
01:43:20,240 --> 01:43:16,800
that are actually um

2738
01:43:21,750 --> 01:43:20,250
targeted by psychedelics

2739
01:43:23,109 --> 01:43:21,760
[Music]

2740
01:43:25,590 --> 01:43:23,119
require you

2741
01:43:28,709 --> 01:43:25,600
now to do all your belief updating and

2742
01:43:31,030 --> 01:43:28,719
evidence assimilation and sense making

2743
01:43:33,830 --> 01:43:31,040
at the much more the input level of of

2744
01:43:35,189 --> 01:43:33,840
the black box not at the if you like the

2745
01:43:37,109 --> 01:43:35,199
highest level which you could think of

2746
01:43:38,629 --> 01:43:37,119
as an output in terms of selecting the

2747
01:43:40,629 --> 01:43:38,639
plans and what am i going to do next and

2748
01:43:43,750 --> 01:43:40,639
what kind of person am i narratives that

2749
01:43:45,270 --> 01:43:43,760
we give ourselves that contextualize

2750
01:43:47,189 --> 01:43:45,280
the way that we behave and the other

2751
01:43:49,030 --> 01:43:47,199
things that we do

2752
01:43:50,470 --> 01:43:49,040
so you know i think there's

2753
01:43:52,950 --> 01:43:50,480
mechanistically a really interesting

2754
01:43:54,229 --> 01:43:52,960
connection between the notion of you

2755
01:43:56,310 --> 01:43:54,239
ruminating

2756
01:44:00,310 --> 01:43:56,320
going in circles around your head

2757
01:44:02,310 --> 01:44:00,320
exploring every darker and darker

2758
01:44:04,709 --> 01:44:02,320
corners of your mind and different

2759
01:44:06,709 --> 01:44:04,719
hypotheses for which there's very little

2760
01:44:08,629 --> 01:44:06,719
evidence available simply because you're

2761
01:44:10,950 --> 01:44:08,639
not attending to the censorium you're

2762
01:44:13,270 --> 01:44:10,960
not talking to anybody

2763
01:44:15,270 --> 01:44:13,280
you're not for example you're not you're

2764
01:44:17,109 --> 01:44:15,280
not focusing on your breathing

2765
01:44:29,109 --> 01:44:17,119
and indeed you try and supplement that

2766
01:44:34,709 --> 01:44:31,350
you probably are um skilled in

2767
01:44:36,149 --> 01:44:34,719
meditation um but that kind i think the

2768
01:44:38,149 --> 01:44:36,159
objective of

2769
01:44:40,550 --> 01:44:38,159
becoming a skilled practitioner of

2770
01:44:42,629 --> 01:44:40,560
mindfulness or meditation is simply to

2771
01:44:45,030 --> 01:44:42,639
get some volitional control over that

2772
01:44:47,350 --> 01:44:45,040
sort of attention

2773
01:44:50,310 --> 01:44:47,360
that is paid to these deeper

2774
01:44:53,270 --> 01:44:50,320
machinations versus attention that is

2775
01:44:57,270 --> 01:44:53,280
paid to the censorium that underwrites

2776
01:45:01,590 --> 01:44:57,280
the psychedelic aspect the um the

2777
01:45:03,590 --> 01:45:01,600
the allure of just sensory uh patterns

2778
01:45:05,189 --> 01:45:03,600
um

2779
01:45:07,350 --> 01:45:05,199
the sensations

2780
01:45:09,990 --> 01:45:07,360
um and textures that you get you know

2781
01:45:13,109 --> 01:45:10,000
when taking psychedelics so you know

2782
01:45:14,950 --> 01:45:13,119
getting control of the balance um

2783
01:45:16,229 --> 01:45:14,960
so it's not the fact you're attending to

2784
01:45:17,510 --> 01:45:16,239
your breathing which is important is the

2785
01:45:19,990 --> 01:45:17,520
fact that you could volitionally

2786
01:45:23,109 --> 01:45:20,000
redirect your attention away from

2787
01:45:26,550 --> 01:45:23,119
selfhood um

2788
01:45:29,109 --> 01:45:26,560
so but you seem to have got control um

2789
01:45:33,189 --> 01:45:29,119
by this acceptance and it strikes me

2790
01:45:36,470 --> 01:45:33,199
that you know in the absence of um

2791
01:45:38,790 --> 01:45:36,480
feelings of an emotional sort that were

2792
01:45:41,270 --> 01:45:38,800
articulated in your description in terms

2793
01:45:42,870 --> 01:45:41,280
of you know i'm an awful person i'm not

2794
01:45:53,750 --> 01:45:42,880
fit for purpose i i

2795
01:45:57,990 --> 01:45:56,629
but quite emotional constructs effective

2796
01:46:00,149 --> 01:45:58,000
valence

2797
01:46:02,629 --> 01:46:00,159
aspects they are also hypotheses they're

2798
01:46:04,550 --> 01:46:02,639
also explanations so the idea is that

2799
01:46:07,109 --> 01:46:04,560
you know anything that you can talk

2800
01:46:08,709 --> 01:46:07,119
about has to be part of an explanation

2801
01:46:11,109 --> 01:46:08,719
for particular ways of being that you

2802
01:46:12,950 --> 01:46:11,119
know i can be a worthless person or this

2803
01:46:15,750 --> 01:46:12,960
is i am a worthless person i'm not with

2804
01:46:16,870 --> 01:46:15,760
this person uh you know and you since uh

2805
01:46:18,550 --> 01:46:16,880
i am

2806
01:46:20,870 --> 01:46:18,560
in a state of anxiety i'm not in the

2807
01:46:24,149 --> 01:46:20,880
strength of anxiety we have to recognize

2808
01:46:27,669 --> 01:46:24,159
when we are anxious anxious is just a

2809
01:46:29,830 --> 01:46:27,679
state of being um which is necessarily

2810
01:46:32,390 --> 01:46:29,840
called for in certain situations that

2811
01:46:33,669 --> 01:46:32,400
rebalance this attention or precision

2812
01:46:37,030 --> 01:46:33,679
weighting in

2813
01:46:39,750 --> 01:46:37,040
computational psychiatry um

2814
01:46:41,430 --> 01:46:39,760
to make your belief updating fit for

2815
01:46:44,229 --> 01:46:41,440
purpose in this particular in this

2816
01:46:45,510 --> 01:46:44,239
particular context and what i'm trying

2817
01:46:48,470 --> 01:46:45,520
to get to

2818
01:46:50,390 --> 01:46:48,480
is why acceptance might have worked

2819
01:46:53,109 --> 01:46:50,400
because i imagine that if you were

2820
01:46:56,070 --> 01:46:53,119
suddenly found yourself ruminating and

2821
01:46:58,870 --> 01:46:56,080
locked into those ruminations

2822
01:47:01,270 --> 01:46:58,880
then very much akin to someone on a bad

2823
01:47:03,669 --> 01:47:01,280
trip who thinks ah i am going to be

2824
01:47:06,470 --> 01:47:03,679
locked in this forever this is going to

2825
01:47:07,669 --> 01:47:06,480
be my uncertain state of being for

2826
01:47:09,350 --> 01:47:07,679
eternity

2827
01:47:12,070 --> 01:47:09,360
because there is no reality that's how

2828
01:47:14,950 --> 01:47:12,080
it feels sorry how i felt yep

2829
01:47:17,189 --> 01:47:14,960
that um that feeling

2830
01:47:18,790 --> 01:47:17,199
will well if that is true i must be

2831
01:47:20,229 --> 01:47:18,800
feeling anxious you are then going to

2832
01:47:25,669 --> 01:47:20,239
look for evidence that you're feeling

2833
01:47:27,750 --> 01:47:25,679
anxious uh and because you are anxious

2834
01:47:30,390 --> 01:47:27,760
you will experience

2835
01:47:33,510 --> 01:47:30,400
certain cardio acceleration certain uh

2836
01:47:34,870 --> 01:47:33,520
intraceptive um flight or flight like

2837
01:47:37,350 --> 01:47:34,880
responses

2838
01:47:38,629 --> 01:47:37,360
and that will supply evidence that you

2839
01:47:41,350 --> 01:47:38,639
are anxious

2840
01:47:45,109 --> 01:47:41,360
and then that becomes evidence yes i am

2841
01:47:47,430 --> 01:47:45,119
right i am now in a state of nihilism

2842
01:47:49,350 --> 01:47:47,440
i do not exist and i should be anxious

2843
01:47:52,709 --> 01:47:49,360
about that and yes i am anxious so you

2844
01:47:54,950 --> 01:47:52,719
get into a vicious circle so this is a

2845
01:47:57,669 --> 01:47:54,960
sort of uh good old-fashioned cognitive

2846
01:48:00,470 --> 01:47:57,679
behavioral explanation for things like

2847
01:48:03,910 --> 01:48:00,480
panic attacks that you know uh you quite

2848
01:48:06,229 --> 01:48:03,920
reasonably in an entirely basical way

2849
01:48:07,910 --> 01:48:06,239
make sense of intraceptive bodily gut

2850
01:48:09,669 --> 01:48:07,920
feelings literally

2851
01:48:12,950 --> 01:48:09,679
you make sense of that

2852
01:48:14,709 --> 01:48:12,960
with the hypothesis ah i must be in an

2853
01:48:16,070 --> 01:48:14,719
anxious state of being

2854
01:48:18,070 --> 01:48:16,080
and that

2855
01:48:20,070 --> 01:48:18,080
conclusion that hypothesis gen then

2856
01:48:22,149 --> 01:48:20,080
generates autonomic actions that are

2857
01:48:24,070 --> 01:48:22,159
realized reflexively in the way we

2858
01:48:26,709 --> 01:48:24,080
talked about when we're looking at the

2859
01:48:28,070 --> 01:48:26,719
parkinson's idea um and that will be

2860
01:48:31,750 --> 01:48:28,080
reflected in terms of cardiac

2861
01:48:33,669 --> 01:48:31,760
acceleration uh neuroendocrine releases

2862
01:48:35,750 --> 01:48:33,679
into your body and your body will change

2863
01:48:37,910 --> 01:48:35,760
and your body will supply signals

2864
01:48:40,229 --> 01:48:37,920
interceptive evidence yes i am anxious

2865
01:48:42,790 --> 01:48:40,239
so it's again a self-fulfilling prophecy

2866
01:48:45,830 --> 01:48:42,800
in exactly the same way the idea motor

2867
01:48:48,950 --> 01:48:45,840
theory means that raising my hand is

2868
01:48:51,510 --> 01:48:48,960
just a realization of my beliefs but i

2869
01:48:53,590 --> 01:48:51,520
can literally raise my levels of anxiety

2870
01:48:55,270 --> 01:48:53,600
in exactly the same way but invisibly

2871
01:48:57,030 --> 01:48:55,280
from the outside but not when i can

2872
01:48:58,070 --> 01:48:57,040
sense my own body

2873
01:49:00,470 --> 01:48:58,080
so

2874
01:49:02,629 --> 01:49:00,480
if you got yourself caught up into this

2875
01:49:03,430 --> 01:49:02,639
joint hypothesis

2876
01:49:26,709 --> 01:49:03,440
i

2877
01:49:29,910 --> 01:49:26,719
you know

2878
01:49:30,709 --> 01:49:29,920
in a dark and nihilistic

2879
01:49:34,629 --> 01:49:30,719
place

2880
01:49:36,709 --> 01:49:34,639
could be no reality

2881
01:49:38,390 --> 01:49:36,719
that's a hypothesis

2882
01:49:39,830 --> 01:49:38,400
if you can just wait

2883
01:49:42,950 --> 01:49:39,840
for the evidence

2884
01:49:44,950 --> 01:49:42,960
that that hypothesis is incorrect

2885
01:49:47,109 --> 01:49:44,960
simply by letting your body calm down so

2886
01:49:48,950 --> 01:49:47,119
there is no further evidence of any

2887
01:49:51,590 --> 01:49:48,960
anxiety

2888
01:49:53,830 --> 01:49:51,600
then you can find secure even from your

2889
01:49:56,550 --> 01:49:53,840
own body without talking to somebody you

2890
01:49:58,070 --> 01:49:56,560
can secure evidence are the hypothesis

2891
01:50:00,470 --> 01:49:58,080
that i am in a state of internal

2892
01:50:03,350 --> 01:50:00,480
nihilism and anxiety is a silly

2893
01:50:05,510 --> 01:50:03,360
hypothesis because the evidence refutes

2894
01:50:06,709 --> 01:50:05,520
it i can now feel my

2895
01:50:10,310 --> 01:50:06,719
cur

2896
01:50:11,910 --> 01:50:10,320
not possibly personally but you you will

2897
01:50:13,030 --> 01:50:11,920
synthesize all your interceptive

2898
01:50:14,790 --> 01:50:13,040
feelings

2899
01:50:16,070 --> 01:50:14,800
um in a way

2900
01:50:21,350 --> 01:50:16,080
that

2901
01:50:23,590 --> 01:50:21,360
now i am a deeply nihilistic um anxious

2902
01:50:25,750 --> 01:50:23,600
state i think that just is the

2903
01:50:27,750 --> 01:50:25,760
motivation for the acceptance it is

2904
01:50:30,149 --> 01:50:27,760
pushing through

2905
01:50:31,189 --> 01:50:30,159
to get to a state of mind and the state

2906
01:50:33,430 --> 01:50:31,199
of body

2907
01:50:36,070 --> 01:50:33,440
literally you know realizing you're good

2908
01:50:37,990 --> 01:50:36,080
and making sense of your gut feelings

2909
01:50:41,430 --> 01:50:38,000
so that you can now refute the joint

2910
01:50:43,109 --> 01:50:41,440
hypothesis that i am uh in a state of

2911
01:50:44,870 --> 01:50:43,119
nihilism and i should be jolly

2912
01:50:45,990 --> 01:50:44,880
frightened of others

2913
01:50:47,430 --> 01:50:46,000
because there's no evidence you're

2914
01:50:51,189 --> 01:50:47,440
frightened anymore

2915
01:50:53,189 --> 01:50:51,199
so this was just a piece of um

2916
01:50:56,310 --> 01:50:53,199
hypothesis building

2917
01:50:57,669 --> 01:50:56,320
of you know a very sophisticated and um

2918
01:50:59,430 --> 01:50:57,679
uh

2919
01:51:02,629 --> 01:50:59,440
philosophical source you know the

2920
01:51:05,430 --> 01:51:02,639
highest level of exploring alternative

2921
01:51:09,350 --> 01:51:05,440
ways of relating to the world you now

2922
01:51:13,430 --> 01:51:09,360
have i think a very useful insight um

2923
01:51:15,669 --> 01:51:13,440
into what into the gift of retaining

2924
01:51:17,510 --> 01:51:15,679
selfhood that we i think we all we all

2925
01:51:19,430 --> 01:51:17,520
take it for granted but it's quite a

2926
01:51:20,390 --> 01:51:19,440
fragile thing you know

2927
01:51:21,750 --> 01:51:20,400
luckily

2928
01:51:24,149 --> 01:51:21,760
um you know

2929
01:51:27,350 --> 01:51:24,159
most of us get through the day if not

2930
01:51:30,070 --> 01:51:27,360
hopefully most of our lives you know by

2931
01:51:31,910 --> 01:51:30,080
retaining that grip and just you know

2932
01:51:33,990 --> 01:51:31,920
continue securing evidence that this is

2933
01:51:36,310 --> 01:51:34,000
the right hypothesis i'm a person i am

2934
01:51:37,750 --> 01:51:36,320
me i am functional this is you know this

2935
01:51:39,830 --> 01:51:37,760
is the way i meant to be and these are

2936
01:51:42,470 --> 01:51:39,840
the kinds of things i do and the kinds

2937
01:51:44,790 --> 01:51:42,480
of people i talk to but that is such a

2938
01:51:47,910 --> 01:51:44,800
fragile self-assembled and has to be

2939
01:51:50,709 --> 01:51:47,920
maintained hypothesis that you don't

2940
01:51:53,189 --> 01:51:50,719
you know that fragility i think only is

2941
01:51:54,870 --> 01:51:53,199
revealed occasionally and only to some

2942
01:51:58,229 --> 01:51:54,880
people when they have the alternative

2943
01:52:00,790 --> 01:51:58,239
hypothesis that you know i am not me or

2944
01:52:06,229 --> 01:52:00,800
my your me is not quite as functional as

2945
01:52:09,510 --> 01:52:08,310
i am only me and there is no reality out

2946
01:52:11,589 --> 01:52:09,520
there

2947
01:52:13,830 --> 01:52:11,599
how is it that you prevent yourself from

2948
01:52:15,910 --> 01:52:13,840
getting into existential crises when it

2949
01:52:18,390 --> 01:52:15,920
seems at least to me from the outside

2950
01:52:20,070 --> 01:52:18,400
that you study similar not theories per

2951
01:52:22,709 --> 01:52:20,080
se but the free energy principle is

2952
01:52:24,870 --> 01:52:22,719
about existence and is about selfhood so

2953
01:52:26,870 --> 01:52:24,880
is your model so strong carl is your

2954
01:52:29,270 --> 01:52:26,880
model already so strong that you can

2955
01:52:31,270 --> 01:52:29,280
entertain these without worry that

2956
01:52:33,669 --> 01:52:31,280
you'll break yours or without that being

2957
01:52:35,350 --> 01:52:33,679
a potential possibility or is it because

2958
01:52:37,189 --> 01:52:35,360
you purposefully analyze it

2959
01:52:38,550 --> 01:52:37,199
dispassionately analytically

2960
01:52:40,229 --> 01:52:38,560
intellectually so that you're not

2961
01:52:41,830 --> 01:52:40,239
identifying with what you're writing

2962
01:52:43,030 --> 01:52:41,840
down on a piece of paper how is it that

2963
01:52:44,950 --> 01:52:43,040
you prevent yourself or is it

2964
01:52:46,470 --> 01:52:44,960
constitutional like by predilection by

2965
01:52:48,470 --> 01:52:46,480
personality you're not neurotic you're

2966
01:52:50,470 --> 01:52:48,480
not as neurotic as i am um

2967
01:52:51,970 --> 01:52:50,480
i think there's there's probably truth

2968
01:52:53,030 --> 01:52:51,980
to all that and um

2969
01:52:55,589 --> 01:52:53,040
[Music]

2970
01:52:57,669 --> 01:52:55,599
and by being neurotic um you know that's

2971
01:52:59,350 --> 01:52:57,679
an entirely your neurosis are

2972
01:53:01,910 --> 01:52:59,360
absolutely essential to get through life

2973
01:53:03,109 --> 01:53:01,920
um you know i have a particular set of

2974
01:53:05,750 --> 01:53:03,119
neurologies

2975
01:53:06,390 --> 01:53:05,760
um which you know i treasure and trying

2976
01:53:10,830 --> 01:53:06,400
to

2977
01:53:12,870 --> 01:53:10,840
admit it in the big five model

2978
01:53:15,510 --> 01:53:12,880
yes um

2979
01:53:17,350 --> 01:53:15,520
but i think you're also right that um

2980
01:53:19,189 --> 01:53:17,360
some people might call this mentalizing

2981
01:53:21,189 --> 01:53:19,199
i think what you were doing is that you

2982
01:53:23,189 --> 01:53:21,199
were

2983
01:53:26,149 --> 01:53:23,199
in a

2984
01:53:27,830 --> 01:53:26,159
raw enthusiastic and creative way we're

2985
01:53:30,390 --> 01:53:27,840
just exploring different hypotheses

2986
01:53:35,189 --> 01:53:30,400
about the way that you yourself should

2987
01:53:36,790 --> 01:53:35,199
it exist um um relates to um you know

2988
01:53:40,470 --> 01:53:36,800
the rest of your

2989
01:53:43,430 --> 01:53:40,480
mind and and the outside world but for

2990
01:53:47,990 --> 01:53:43,440
you we got conflated with um

2991
01:53:48,709 --> 01:53:48,000
emotional states like anxiety um

2992
01:53:54,310 --> 01:53:48,719
and

2993
01:53:57,270 --> 01:53:54,320
states of being um

2994
01:53:59,430 --> 01:53:57,280
so if you want to in your words be more

2995
01:54:01,910 --> 01:53:59,440
dispassionate about this kind of mental

2996
01:54:05,109 --> 01:54:01,920
exploration on the inside then you have

2997
01:54:07,669 --> 01:54:05,119
to um you you you have to train yourself

2998
01:54:10,149 --> 01:54:07,679
not to conflate the two one way of doing

2999
01:54:13,990 --> 01:54:10,159
that is because to mentalize

3000
01:54:15,589 --> 01:54:14,000
so mentalizing is usually taken as a way

3001
01:54:18,229 --> 01:54:15,599
of if you like

3002
01:54:19,910 --> 01:54:18,239
disentangling what we what we would

3003
01:54:21,990 --> 01:54:19,920
normally quite functionally put together

3004
01:54:23,430 --> 01:54:22,000
certain states of being

3005
01:54:26,229 --> 01:54:23,440
of a sort of

3006
01:54:27,189 --> 01:54:26,239
intellectual sort or you need a physical

3007
01:54:31,109 --> 01:54:27,199
sort

3008
01:54:35,030 --> 01:54:31,119
along with some

3009
01:54:40,390 --> 01:54:36,950
aspect simply because that's the way the

3010
01:54:41,750 --> 01:54:40,400
world is so if you'll say um um very

3011
01:54:44,070 --> 01:54:41,760
late at night

3012
01:54:45,990 --> 01:54:44,080
um you're normally gonna feel fatigued

3013
01:54:49,030 --> 01:54:46,000
it's like you know uh slightly less

3014
01:54:51,109 --> 01:54:49,040
robust you know in in a way

3015
01:54:52,550 --> 01:54:51,119
physically now it doesn't have to be

3016
01:54:54,629 --> 01:54:52,560
like that but that's the way that

3017
01:54:55,830 --> 01:54:54,639
normally the world works and that's the

3018
01:54:56,790 --> 01:54:55,840
way that you've learned to model

3019
01:54:58,070 --> 01:54:56,800
yourself

3020
01:55:01,030 --> 01:54:58,080
and

3021
01:55:03,030 --> 01:55:01,040
if you train to be a professional say

3022
01:55:05,750 --> 01:55:03,040
like an airline pilot trying to land a

3023
01:55:08,709 --> 01:55:05,760
plane at night or like a psychiatrist

3024
01:55:10,950 --> 01:55:08,719
who has to deal with these um has to see

3025
01:55:14,070 --> 01:55:10,960
people going through

3026
01:55:20,550 --> 01:55:17,669
emotional or valence or effective um

3027
01:55:22,790 --> 01:55:20,560
episodes whether it's uh sort of um

3028
01:55:27,510 --> 01:55:22,800
effective in the sense of uh you know

3029
01:55:29,750 --> 01:55:27,520
mania or depression or whether it's um

3030
01:55:32,470 --> 01:55:29,760
the the you know the co-morbidity

3031
01:55:35,109 --> 01:55:32,480
associated with psychotic states uh say

3032
01:55:38,550 --> 01:55:35,119
schizoaffective schizophrenia uh sorry

3033
01:55:40,470 --> 01:55:38,560
schizoaffective disorders um what's that

3034
01:55:44,950 --> 01:55:40,480
it's a mixture of um

3035
01:55:46,790 --> 01:55:44,960
um abnormal mood um with having

3036
01:55:49,510 --> 01:55:46,800
psychosis of the kind you'd associate

3037
01:55:51,669 --> 01:55:49,520
with schizophrenia um so it's having an

3038
01:55:53,430 --> 01:55:51,679
emotional aspect to

3039
01:55:55,750 --> 01:55:53,440
say delusions and hallucinations for

3040
01:55:57,030 --> 01:55:55,760
example oh i thought that those those

3041
01:55:59,589 --> 01:55:57,040
don't go hand in hand so some people

3042
01:56:00,870 --> 01:55:59,599
have delusions but they're okay with it

3043
01:56:02,790 --> 01:56:00,880
yes yes

3044
01:56:05,109 --> 01:56:02,800
some people have hallucinations and and

3045
01:56:06,550 --> 01:56:05,119
would not uh would not even seek medical

3046
01:56:07,990 --> 01:56:06,560
help or their family and friends would

3047
01:56:08,950 --> 01:56:08,000
not not seek

3048
01:56:10,870 --> 01:56:08,960
help

3049
01:56:14,550 --> 01:56:10,880
so there are people out there i guess my

3050
01:56:16,870 --> 01:56:14,560
point you know um here is that you can

3051
01:56:19,510 --> 01:56:16,880
mentalize it and build professional

3052
01:56:21,270 --> 01:56:19,520
defenses against the emotional aspects

3053
01:56:23,510 --> 01:56:21,280
and i think that's what i've done

3054
01:56:26,830 --> 01:56:23,520
so i think about these things as a

3055
01:56:29,910 --> 01:56:26,840
doctor or a mathematician not as a

3056
01:56:33,030 --> 01:56:29,920
person and if you want uh if you want an

3057
01:56:35,830 --> 01:56:33,040
honest answer the other trick uh is to

3058
01:56:38,310 --> 01:56:35,840
is is to uh not use psychedelics but you

3059
01:56:42,950 --> 01:56:38,320
should you should try smoking so smoking

3060
01:56:46,870 --> 01:56:44,550
or cigarettes

3061
01:56:49,030 --> 01:56:46,880
or i actually smoke a pipe but

3062
01:56:50,629 --> 01:56:49,040
cigarettes will do in an emergency so

3063
01:56:52,790 --> 01:56:50,639
there are certain neurotron

3064
01:56:55,589 --> 01:56:52,800
manipulations of neurotransmitters the

3065
01:56:57,589 --> 01:56:55,599
first time a doctor has recommended

3066
01:56:59,189 --> 01:56:57,599
yes

3067
01:57:00,870 --> 01:56:59,199
well i hope we have a select audience

3068
01:57:02,470 --> 01:57:00,880
for this

3069
01:57:03,669 --> 01:57:02,480
well yeah

3070
01:57:06,149 --> 01:57:03,679
continue could you finish up your

3071
01:57:07,589 --> 01:57:06,159
thought i'm sorry

3072
01:57:09,189 --> 01:57:07,599
you're saying the smoking of the pipe

3073
01:57:10,870 --> 01:57:09,199
does what because i'm not the type of i

3074
01:57:12,870 --> 01:57:10,880
don't smoke and nor do i want to smoke

3075
01:57:15,430 --> 01:57:12,880
but you're saying that the smoking does

3076
01:57:18,310 --> 01:57:15,440
what was the purpose of the smoking

3077
01:57:20,310 --> 01:57:18,320
it's just a way of using um nuancing

3078
01:57:22,709 --> 01:57:20,320
your neurotransmitters that control that

3079
01:57:24,950 --> 01:57:22,719
synaptic gain or the gain and the weight

3080
01:57:28,390 --> 01:57:24,960
view afford different uh different parts

3081
01:57:29,910 --> 01:57:28,400
of your um belief updating so it has the

3082
01:57:31,830 --> 01:57:29,920
opposite effect of psychedelics

3083
01:57:33,990 --> 01:57:31,840
basically so

3084
01:57:36,310 --> 01:57:34,000
that's why people like tea and you know

3085
01:57:39,589 --> 01:57:36,320
have a cup of tea come that you'll stay

3086
01:57:41,669 --> 01:57:39,599
calm and uh and carry on wherever it is

3087
01:57:43,990 --> 01:57:41,679
nice cup of tea cigarette and cup of

3088
01:57:46,390 --> 01:57:44,000
coffee yeah that'll sort you out and

3089
01:57:48,790 --> 01:57:46,400
it's because all of these um drugs the

3090
01:57:50,070 --> 01:57:48,800
caffeine the nicotine they all act on

3091
01:57:53,510 --> 01:57:50,080
the same

3092
01:57:55,189 --> 01:57:53,520
kind of um receptors um

3093
01:57:57,910 --> 01:57:55,199
at different levels in your neuronal

3094
01:57:59,990 --> 01:57:57,920
hierarchy um that are responsible for

3095
01:58:01,830 --> 01:58:00,000
setting that gain control and gaiting

3096
01:58:03,669 --> 01:58:01,840
doing the intentional and the the

3097
01:58:05,589 --> 01:58:03,679
attenuation aspects that we're talking

3098
01:58:08,149 --> 01:58:05,599
about so it's just a way of getting

3099
01:58:09,830 --> 01:58:08,159
physical control over your hierarchical

3100
01:58:11,750 --> 01:58:09,840
belief updating

3101
01:58:12,790 --> 01:58:11,760
a bit like you're about

3102
01:58:14,070 --> 01:58:12,800
i would have thought that the caffeine

3103
01:58:17,510 --> 01:58:14,080
would have the opposite effect of

3104
01:58:24,070 --> 01:58:20,870
not not necessarily no um you know

3105
01:58:27,109 --> 01:58:24,080
you know if well it clearly depends um

3106
01:58:28,470 --> 01:58:27,119
if you're very agitated um

3107
01:58:30,310 --> 01:58:28,480
then um

3108
01:58:33,589 --> 01:58:30,320
that could certainly be the case but if

3109
01:58:34,550 --> 01:58:33,599
you're if if you are

3110
01:58:36,229 --> 01:58:34,560
worried

3111
01:58:39,830 --> 01:58:36,239
that you might be anxious and in fact

3112
01:58:43,669 --> 01:58:39,840
you're not then desensitizing certain um

3113
01:58:45,430 --> 01:58:43,679
certain interceptive sensations um

3114
01:58:46,870 --> 01:58:45,440
that register that in fact you're not

3115
01:58:48,629 --> 01:58:46,880
terribly you know your heart isn't

3116
01:58:50,870 --> 01:58:48,639
pounding that'll actually calm you down

3117
01:58:52,709 --> 01:58:50,880
a little bit uh if it goes too far then

3118
01:58:55,270 --> 01:58:52,719
beta blockers will be an another nice

3119
01:58:57,270 --> 01:58:55,280
example of that's what i used actually

3120
01:58:59,350 --> 01:58:57,280
all right oh well there you go so what

3121
01:59:02,149 --> 01:58:59,360
i'm trying to say is sometimes

3122
01:59:04,470 --> 01:59:02,159
with skillful use the use of a tobacco

3123
01:59:06,790 --> 01:59:04,480
pipe can have the same kind of effects

3124
01:59:08,550 --> 01:59:06,800
as a beta blocker

3125
01:59:10,070 --> 01:59:08,560
that's all i would say all right all

3126
01:59:11,830 --> 01:59:10,080
right professor

3127
01:59:13,430 --> 01:59:11,840
again thank you so much much safe to

3128
01:59:15,430 --> 01:59:13,440
stick with beta blockers that you won't

3129
01:59:16,629 --> 01:59:15,440
get mouth cancer or anything so so if

3130
01:59:17,990 --> 01:59:16,639
you could

3131
01:59:19,350 --> 01:59:18,000
i found the beta blockers work and

3132
01:59:21,270 --> 01:59:19,360
there's also some evidence of fear

3133
01:59:22,790 --> 01:59:21,280
extinction have you heard of that right

3134
01:59:24,229 --> 01:59:22,800
there's this uh professor named merrell

3135
01:59:25,830 --> 01:59:24,239
kenditt who i'm going to speak to at

3136
01:59:27,510 --> 01:59:25,840
some point she

3137
01:59:29,750 --> 01:59:27,520
takes people through phobias where she

3138
01:59:32,470 --> 01:59:29,760
gets them to be extremely frightened

3139
01:59:34,070 --> 01:59:32,480
she actually makes them almost manic or

3140
01:59:35,910 --> 01:59:34,080
panicking at least and then she gives

3141
01:59:37,430 --> 01:59:35,920
them a beta blocker and then there's

3142
01:59:39,109 --> 01:59:37,440
they still feel fear that day it's only

3143
01:59:41,669 --> 01:59:39,119
the next day when there's memory

3144
01:59:43,350 --> 01:59:41,679
reconsolidation and the emotional tags

3145
01:59:45,109 --> 01:59:43,360
are removed because of the beta blocker

3146
01:59:46,950 --> 01:59:45,119
so you have to have a good night's sleep

3147
01:59:50,629 --> 01:59:46,960
and then the fear seems to dissipate

3148
01:59:52,149 --> 01:59:50,639
drastically for incurable ptsd i was

3149
01:59:54,229 --> 01:59:52,159
trying to do that to myself so when i

3150
01:59:55,990 --> 01:59:54,239
was feeling some fear i would take a

3151
01:59:57,750 --> 01:59:56,000
beta blocker and then i would feel the

3152
01:59:59,910 --> 01:59:57,760
fear less still wouldn't be zero but

3153
02:00:01,990 --> 01:59:59,920
then thinking okay great maybe this will

3154
02:00:04,550 --> 02:00:02,000
be like my own intervention of a fear

3155
02:00:07,350 --> 02:00:04,560
extinction technique

3156
02:00:09,030 --> 02:00:07,360
yes i i i was that's a very clever idea

3157
02:00:10,310 --> 02:00:09,040
actually to combine uh beta blockers

3158
02:00:12,629 --> 02:00:10,320
with uh what

3159
02:00:15,030 --> 02:00:12,639
extinction or some people call flooding

3160
02:00:16,950 --> 02:00:15,040
so that the uh so that that's exactly

3161
02:00:19,270 --> 02:00:16,960
the kind of way i was um

3162
02:00:21,990 --> 02:00:19,280
uh counseling you you you would use the

3163
02:00:28,950 --> 02:00:22,000
people who smoke you smoking to moderate

3164
02:00:32,709 --> 02:00:30,470
thank you so much i

3165
02:00:34,310 --> 02:00:32,719
will talk again and perhaps the next

3166
02:00:35,750 --> 02:00:34,320
time we speak there were some questions

3167
02:00:38,229 --> 02:00:35,760
that i didn't even get to from michael

3168
02:00:40,390 --> 02:00:38,239
levin from professor norman weinberger

3169
02:00:42,070 --> 02:00:40,400
who's a mathematician and and there's

3170
02:00:44,390 --> 02:00:42,080
quite a few okay

3171
02:00:46,070 --> 02:00:44,400
carl i kept you on for so long and this

3172
02:00:48,149 --> 02:00:46,080
is gone

3173
02:00:49,270 --> 02:00:48,159
well this didn't i didn't intend for it

3174
02:00:51,589 --> 02:00:49,280
to go in this direction but i'm

3175
02:00:53,830 --> 02:00:51,599
extremely glad it has and maybe my

3176
02:00:56,390 --> 02:00:53,840
episode could serve and your advice

3177
02:00:58,870 --> 02:00:56,400
that's a plus the advice can serve as a

3178
02:01:00,709 --> 02:00:58,880
cautionary tale and as well as some

3179
02:01:02,550 --> 02:01:00,719
practical advice as to what to do and

3180
02:01:04,709 --> 02:01:02,560
how to avoid situations like this when

3181
02:01:07,589 --> 02:01:04,719
exploring topics like this thank you so

3182
02:01:12,229 --> 02:01:07,599
much professor i appreciate

3183
02:01:15,910 --> 02:01:14,149
the podcast is now finished if you'd

3184
02:01:19,030 --> 02:01:15,920
like to support conversations like this

3185
02:01:23,189 --> 02:01:19,040
then do consider going to patreon.com

3186
02:01:25,750 --> 02:01:23,199
c-u-r-t j-a-i-m-u-n

3187
02:01:27,589 --> 02:01:25,760
that is kurt jaimungle it's support from

3188
02:01:29,430 --> 02:01:27,599
the patrons and from the sponsors that