1
00:00:00,000 --> 00:00:04,920
Instead of people opening up and accepting more ways of recognizing human beings, what

2
00:00:04,920 --> 00:00:14,360
they do instead is apply the machine frameworks to the humans, and they dehumanize more often

3
00:00:14,360 --> 00:00:17,040
than they expand humanity.

4
00:00:17,040 --> 00:00:19,560
And that's really, really dangerous.

5
00:00:19,560 --> 00:00:25,880
It has real implications because once we start to think, "Oh, I know what human is,"

6
00:00:25,880 --> 00:00:32,880
it's not X, Y, and Z, and you, person I am suspicious of, are displaying X, Y, and Z.

7
00:00:32,880 --> 00:00:35,280
You must not really be human.

8
00:00:35,280 --> 00:00:37,880
You don't count the same way.

9
00:00:37,880 --> 00:00:42,960
Welcome to the Repatoning Podcast, where we believe your mind-body system is your domain,

10
00:00:42,960 --> 00:00:47,840
and that choosing the beliefs, habits, and patterns that best support you is your birthright.

11
00:00:47,840 --> 00:00:52,440
I'm your host, Arden Leigh, creator of the Re-Patterning Project, an eight-week course in learning

12
00:00:52,440 --> 00:00:56,520
to free yourself from the patterns of your past and program your mind for your dreams

13
00:00:56,520 --> 00:00:57,520
of the future.

14
00:00:57,520 --> 00:01:03,680
Join me for these extraordinary conversations as we unpack the keys to creating our reality.

15
00:01:03,680 --> 00:01:06,920
Hello, hello, everyone.

16
00:01:06,920 --> 00:01:09,760
Welcome back to the Re-Patterning Podcast.

17
00:01:09,760 --> 00:01:17,440
Today's episode features Damien Patrick Williams, who is an expert in ethics in AI.

18
00:01:17,440 --> 00:01:20,560
Damien and I have been Facebook friends for a little while.

19
00:01:20,560 --> 00:01:24,920
This is the first time that we're sitting down for an actual conversation on Zoom, but

20
00:01:24,920 --> 00:01:28,760
I have been following his posts for years now.

21
00:01:28,760 --> 00:01:33,360
Whenever there's anyone in my life who's like, "Oh, what do you think about AI or what about

22
00:01:33,360 --> 00:01:34,360
this?"

23
00:01:34,360 --> 00:01:36,600
I'm like, "You guys, you've got to go follow Damien."

24
00:01:36,600 --> 00:01:40,080
You need to read and listen to what he's been saying.

25
00:01:40,080 --> 00:01:42,560
I'm going to tell you a little bit about his background.

26
00:01:42,560 --> 00:01:48,000
He is an assistant professor in philosophy and data science at the University of North

27
00:01:48,000 --> 00:01:55,600
Carolina, where he teaches and researches on how technology such as algorithms, biotechnologies,

28
00:01:55,600 --> 00:02:01,160
and all the things that we think of as AI, artificial intelligence in air quotes, are

29
00:02:01,160 --> 00:02:06,760
impacted by the values, knowledge systems, philosophical explorations, social structures,

30
00:02:06,760 --> 00:02:11,640
and even the religious beliefs of human beings.

31
00:02:11,640 --> 00:02:17,600
He has his PhD from Virginia Tech in the field of Science, Technology, and Society.

32
00:02:17,600 --> 00:02:23,960
His research focuses on how the experiences and the knowledge of marginalized people affect

33
00:02:23,960 --> 00:02:30,080
and should be centered within the techno-social structures of human societies.

34
00:02:30,080 --> 00:02:35,200
And again, whenever I'm curious about anything to do with AI, Damien is my go-to guy.

35
00:02:35,200 --> 00:02:39,440
Naturally here on the Re-Patterning podcast, I had to bring him on as a guest.

36
00:02:39,440 --> 00:02:44,960
I'm so excited for this conversation because naturally, AI is a really big buzzword and

37
00:02:44,960 --> 00:02:50,720
we're looking at how our use of it is even affecting our own cognition.

38
00:02:50,720 --> 00:02:55,200
That's obviously really relevant to Re-Patterning because we want to be really conscious of our

39
00:02:55,200 --> 00:02:56,200
own patterning.

40
00:02:56,200 --> 00:02:58,800
So, Damien, thank you so much for being here with me today.

41
00:02:58,800 --> 00:03:00,200
Thank you so much for having me.

42
00:03:00,200 --> 00:03:01,680
I'm really looking forward to the conversation.

43
00:03:01,680 --> 00:03:05,280
I very much appreciate all the extremely kind words.

44
00:03:05,280 --> 00:03:06,880
That's a lot of love, too.

45
00:03:06,880 --> 00:03:07,880
I'll do my best.

46
00:03:07,880 --> 00:03:11,760
I mean, it's all true, though.

47
00:03:11,760 --> 00:03:16,920
The impression that your work has left upon me has been incredibly positive and impactful.

48
00:03:16,920 --> 00:03:21,600
And I'm just grateful that you exist, that people like you exist, because there's so

49
00:03:21,600 --> 00:03:23,680
many people in the tech spaces.

50
00:03:23,680 --> 00:03:25,040
Well, I don't need to tell you twice.

51
00:03:25,040 --> 00:03:30,320
There's so many people in the tech spaces who are just really ignoring our humanity and

52
00:03:30,320 --> 00:03:35,160
the way that the technology we're creating is in many ways getting away from us and impacting

53
00:03:35,160 --> 00:03:37,000
us in a lot of negative ways.

54
00:03:37,000 --> 00:03:41,680
So when I see your writing online and when I see the work you're doing, I'm like,

55
00:03:41,680 --> 00:03:45,560
thank goodness for Damien because somebody need to be doing that.

56
00:03:45,560 --> 00:03:47,240
Thank goodness it's you.

57
00:03:47,240 --> 00:03:48,880
I appreciate that very much.

58
00:03:48,880 --> 00:03:51,000
And, like I said, I do my best.

59
00:03:51,000 --> 00:03:57,440
I try to highlight these things in a way that is accessible for people, that people can

60
00:03:57,440 --> 00:04:02,400
actually kind of like get something from but also learn the actual, some of the technical

61
00:04:02,400 --> 00:04:06,920
details and actually understand a bit more about how it relates to everyday life, because

62
00:04:06,920 --> 00:04:08,840
some of it can be kind of weird and abstract, right?

63
00:04:08,840 --> 00:04:10,840
Some of it can be kind of like, you know, maybe that's the way that people like you do

64
00:04:10,840 --> 00:04:16,440
and maybe that's a problem, but you don't necessarily understand why it's a problem for you.

65
00:04:16,440 --> 00:04:19,160
You don't understand why it's a problem in the world today.

66
00:04:19,160 --> 00:04:26,080
And so I try to be a person who can help people to understand the aspects of why if it's

67
00:04:26,080 --> 00:04:28,600
a concern, it's a concern for all of us.

68
00:04:28,600 --> 00:04:33,520
And if it's something that can maybe be changed to help us, then here's the things that

69
00:04:33,520 --> 00:04:36,080
we can do to help make that change, right?

70
00:04:36,080 --> 00:04:40,280
So it's something that people can link into on the individual community level rather than

71
00:04:40,280 --> 00:04:44,160
expecting somebody over there to fix it for us.

72
00:04:44,160 --> 00:04:49,600
Yeah, which is quite obviously what we all need to do if there's going to be change in

73
00:04:49,600 --> 00:04:53,640
the first place, because we can't rely on other folks to do it for us, unfortunately.

74
00:04:53,640 --> 00:04:59,200
I'm not going to start off with that topic of how AI affects cognition.

75
00:04:59,200 --> 00:05:04,800
Like we're seeing research now coming out that is indicating that AI can actually cause

76
00:05:04,800 --> 00:05:09,720
in some ways cognitive decline, because we've come to rely on it if we use it for our writing

77
00:05:09,720 --> 00:05:15,760
or for our creativity, or even for outsourcing or thinking, it's like almost like we become

78
00:05:15,760 --> 00:05:16,760
dependent on it.

79
00:05:16,760 --> 00:05:18,560
Would you say that that's accurate?

80
00:05:18,560 --> 00:05:19,560
Yes.

81
00:05:19,560 --> 00:05:23,640
And that's a thing that has kind of pre-existed AI for a while.

82
00:05:23,640 --> 00:05:27,920
Like it's a problem that we've noticed for a lot of things in technology and just new

83
00:05:27,920 --> 00:05:29,440
systems, right?

84
00:05:29,440 --> 00:05:35,880
There's always, to an extent, there's always this moment where we go all in on this new

85
00:05:35,880 --> 00:05:40,480
thing and then we give over so much power to it, we give over so much decision making power

86
00:05:40,480 --> 00:05:41,720
to it.

87
00:05:41,720 --> 00:05:46,000
And sometimes we manage to kind of claw it back, we roll back and we go, okay, maybe that

88
00:05:46,000 --> 00:05:47,000
was too much.

89
00:05:47,000 --> 00:05:50,640
Well, we can apply it here, but maybe we don't apply it everywhere.

90
00:05:50,640 --> 00:05:57,120
But in some cases, especially when it comes to technologies like automation, computerization,

91
00:05:57,120 --> 00:06:02,680
automated decision making systems, anything that has this kind of idea of a kind of foundation

92
00:06:02,680 --> 00:06:05,800
of math to it, that makes things quite objective.

93
00:06:05,800 --> 00:06:11,560
We tend to kind of just assume that it's going to be better than us, that the computer,

94
00:06:11,560 --> 00:06:16,640
the automated thing is going to be better than us at making decisions at doing whatever

95
00:06:16,640 --> 00:06:19,600
it is that it's, you know, we're handing it over to do.

96
00:06:19,600 --> 00:06:21,400
And then we just trust it.

97
00:06:21,400 --> 00:06:25,280
That's like a societal problem that we have had for a very long time.

98
00:06:25,280 --> 00:06:26,520
There's a term for it.

99
00:06:26,520 --> 00:06:29,520
It's called automation bias, right?

100
00:06:29,520 --> 00:06:32,440
We have a tendency to trust the automated thing.

101
00:06:32,440 --> 00:06:38,080
There's an old book called trust in numbers, where this talks about how we do that with math.

102
00:06:38,080 --> 00:06:41,680
We have a problem with statistics, if somebody can throw stats at us, it doesn't matter

103
00:06:41,680 --> 00:06:45,560
if those stats are right, if they can make it sound right, then we're more likely to trust

104
00:06:45,560 --> 00:06:49,680
and believe because there's some math attached to it, right?

105
00:06:49,680 --> 00:06:51,760
And now it just sounds good, right?

106
00:06:51,760 --> 00:06:52,760
It sounds good.

107
00:06:52,760 --> 00:06:54,400
It's like, oh, he's got a number for that.

108
00:06:54,400 --> 00:06:55,400
Must be right, right?

109
00:06:55,400 --> 00:06:57,320
And we want to believe in that too, right?

110
00:06:57,320 --> 00:07:01,560
It's like, it's like wanting to believe in the fairy tale of being rescued, like that someone,

111
00:07:01,560 --> 00:07:06,160
outside of us is going to come save us, or is there some divine Messiah who is going to make

112
00:07:06,160 --> 00:07:07,160
everything all right?

113
00:07:07,160 --> 00:07:09,320
And then we apply that to our technology.

114
00:07:09,320 --> 00:07:10,320
Absolutely.

115
00:07:10,320 --> 00:07:14,400
And it's like, if somebody can tell us, oh, this thing is, it's so smart, it's so capable,

116
00:07:14,400 --> 00:07:15,560
it's actually too capable.

117
00:07:15,560 --> 00:07:18,080
It's a little dangerous how capable it is.

118
00:07:18,080 --> 00:07:21,440
We get that kind of, ooh, scary, but kind of interesting, right?

119
00:07:21,440 --> 00:07:24,400
And so even the danger becomes part of the hype.

120
00:07:24,400 --> 00:07:30,040
It becomes part of the myth that animates the thing itself and drives people towards adopting

121
00:07:30,040 --> 00:07:31,240
it, right?

122
00:07:31,240 --> 00:07:36,040
And so the more people can do that, the more they can kind of tie math to it and say, well,

123
00:07:36,040 --> 00:07:41,600
we have like a, let's call it an 89% chance that if we don't get control of artificial intelligence

124
00:07:41,600 --> 00:07:44,520
today, there'll be a runaway super intelligence at some point in the future, right?

125
00:07:44,520 --> 00:07:49,320
Like these are the kinds of numbers that just kind of get thrown around.

126
00:07:49,320 --> 00:07:53,880
If you're careful, you stop and you ask, where did that number come from?

127
00:07:53,880 --> 00:07:58,080
Where did that particular idea of a probability come from?

128
00:07:58,080 --> 00:08:00,080
Did you just make that up?

129
00:08:00,080 --> 00:08:03,720
Like you can ask somebody's like, what'd you get that probability?

130
00:08:03,720 --> 00:08:05,440
How'd you derive that?

131
00:08:05,440 --> 00:08:12,280
And the answer, most of the times, a lot of the times is, no, no, the answer is they just

132
00:08:12,280 --> 00:08:19,200
kind of like did some kind of figuring in their head based on vibes and they then spat

133
00:08:19,200 --> 00:08:21,840
that number out from somewhere, right?

134
00:08:21,840 --> 00:08:27,000
And so these questions of like, well, we want something to save us.

135
00:08:27,000 --> 00:08:28,520
We want something better than us.

136
00:08:28,520 --> 00:08:34,280
We've had this dream of AI for a very long time that the stream called AI since the 1950s,

137
00:08:34,280 --> 00:08:42,280
but we've had this dream of being able to create something like us, but better than us for

138
00:08:42,280 --> 00:08:44,880
thousands of years, right?

139
00:08:44,880 --> 00:08:50,240
Something in our image that we can control, but is like more powerful than we are.

140
00:08:50,240 --> 00:08:57,200
As if those two ideas were not in a weird tension already, but we've always had that story.

141
00:08:57,200 --> 00:09:01,560
And so we keep kind of trying to do that.

142
00:09:01,560 --> 00:09:05,400
And then you get these people who can make that seem to come true.

143
00:09:05,400 --> 00:09:11,720
You get the companies that seem to have built the thing that finally is going to be more

144
00:09:11,720 --> 00:09:17,240
powerful than us, but is under our control somehow, the thing that is somehow in our image,

145
00:09:17,240 --> 00:09:20,920
but better than we are simultaneously.

146
00:09:20,920 --> 00:09:30,480
And that story, it has such a longstanding home in our culture that it's very, very easy

147
00:09:30,480 --> 00:09:33,200
for people to get an amort of it.

148
00:09:33,200 --> 00:09:37,680
It's very easy to become taken in by it.

149
00:09:37,680 --> 00:09:38,680
Yeah.

150
00:09:38,680 --> 00:09:43,840
And what would you say are some of the risks of overuse in terms of how our own brains

151
00:09:43,840 --> 00:09:46,400
become patterned by our dependence on it?

152
00:09:46,400 --> 00:09:47,400
Yeah.

153
00:09:47,400 --> 00:09:49,080
That's a very good way to put it.

154
00:09:49,080 --> 00:09:52,680
Because it is a product of pattern making.

155
00:09:52,680 --> 00:09:55,400
We have, that's what habits are.

156
00:09:55,400 --> 00:09:57,800
They're ingrained grooves of pattern.

157
00:09:57,800 --> 00:10:05,160
And so the more we give over that capability, that decision making authority, that kind of

158
00:10:05,160 --> 00:10:10,680
way of describing the world to us, we let these systems tell us how to think about the

159
00:10:10,680 --> 00:10:13,600
world, the questions that we want to ask you.

160
00:10:13,600 --> 00:10:18,560
The more we do that, the more we say, okay, so this kind of answer is the thing that

161
00:10:18,560 --> 00:10:22,120
I would like to have or this kind of answers the thing that guides me.

162
00:10:22,120 --> 00:10:26,000
And now the next question I ask, well, I'll look for that kind of answer as well.

163
00:10:26,000 --> 00:10:30,160
And the next question I ask, I'll try to ask it in a way that will get me the kind of answer

164
00:10:30,160 --> 00:10:32,440
that the system is most likely to give me.

165
00:10:32,440 --> 00:10:37,720
And so we start to shift our thinking, we shift our speech, we shift our patterns towards

166
00:10:37,720 --> 00:10:40,360
what the system is more likely to give us.

167
00:10:40,360 --> 00:10:45,200
And so that can be something as simple as like people's, there's a reason report that shows

168
00:10:45,200 --> 00:10:49,560
that people have started to type to other human beings in the same way that they type to

169
00:10:49,560 --> 00:10:53,640
chat GPT and other systems like large language model systems.

170
00:10:53,640 --> 00:11:00,320
And so like our online interactions have now become that kind of language more often.

171
00:11:00,320 --> 00:11:05,400
And then there's, you know, that's kind of like, eh, okay, maybe that's not a huge impact.

172
00:11:05,400 --> 00:11:08,080
It's definitely kind of a dramatically apparent one.

173
00:11:08,080 --> 00:11:12,240
But it's one of those things that's like, even your basic everyday mode of speech can

174
00:11:12,240 --> 00:11:15,600
be shifted, what's happening behind the scenes that you're not aware of?

175
00:11:15,600 --> 00:11:19,560
And the answer to that is, like you said, you know, there's issues of cognition.

176
00:11:19,560 --> 00:11:26,680
We don't do the heavy lifting of asking questions ourselves of developing questions ourselves.

177
00:11:26,680 --> 00:11:29,760
Then we get out of the habit of developing questions.

178
00:11:29,760 --> 00:11:34,480
We get out of the habit of critical thinking, we get out of the habit of critical analysis.

179
00:11:34,480 --> 00:11:41,120
And so we then have that already ingrained kind of habit towards trusting the system,

180
00:11:41,120 --> 00:11:45,600
lose the skills that allow us to question what the system gives us.

181
00:11:45,600 --> 00:11:50,400
Until if the system gives us a wrong answer, we now don't have that increased capability

182
00:11:50,400 --> 00:11:53,040
that we used to have to say, ah, that doesn't feel right.

183
00:11:53,040 --> 00:11:54,040
That doesn't seem right.

184
00:11:54,040 --> 00:11:58,640
There's something, something is off about that intuition that pushes us towards asking

185
00:11:58,640 --> 00:12:03,600
that further critical question towards digging down and analyzing what it is it's been

186
00:12:03,600 --> 00:12:04,600
given to us.

187
00:12:04,600 --> 00:12:09,440
We can't tell if something is wrong, if we don't know what right looks like.

188
00:12:09,440 --> 00:12:11,440
And the more we question it, the more we question it.

189
00:12:11,440 --> 00:12:15,280
But even the first example that you gave, like even the idea, like that's really dark, the

190
00:12:15,280 --> 00:12:22,960
idea of all of a sudden we're becoming patterned to speak to our fellow humans in the same

191
00:12:22,960 --> 00:12:31,000
way that we would treat a non-senscient, that, ooh, like that's, that's bad.

192
00:12:31,000 --> 00:12:34,640
That's one of the things that I've tried to highlight in this space and other people

193
00:12:34,640 --> 00:12:37,200
have done a very good job of highlighting in this space as well.

194
00:12:37,200 --> 00:12:42,120
Like, there's this aspect of deshumanization.

195
00:12:42,120 --> 00:12:47,960
And so, you know, when I first started to think about those questions of AI, well before

196
00:12:47,960 --> 00:12:50,760
the chat GPT, large language model era, right?

197
00:12:50,760 --> 00:12:56,600
And so I'm writing about these things, like, 15, almost 20 years ago at this point.

198
00:12:56,600 --> 00:12:59,360
And that's when we were dealing with artificial neural networks.

199
00:12:59,360 --> 00:13:03,560
And the early days of machine learning, this is like pre-google deep-mic, a lot of this

200
00:13:03,560 --> 00:13:04,560
stuff.

201
00:13:04,560 --> 00:13:10,000
And the first papers I ever gave in an international conference was the same week that Google

202
00:13:10,000 --> 00:13:17,400
deep-mind, like, debuted at having properly identified a new picture of cats, right?

203
00:13:17,400 --> 00:13:21,640
Like that kind of image recognition, like early days image recognition stuff.

204
00:13:21,640 --> 00:13:24,520
So well before we're talking generative AI.

205
00:13:24,520 --> 00:13:27,560
And when I was initially kind of thinking about these things, a lot of my thought process

206
00:13:27,560 --> 00:13:31,840
had honestly kind of been like, maybe if we can get people to think about the possibilities

207
00:13:31,840 --> 00:13:37,800
of machine intelligence or machine consciousness more broadly, we can get people to like understand

208
00:13:37,800 --> 00:13:44,680
and recognize all the different ways that human consciousness can manifest itself, the

209
00:13:44,680 --> 00:13:47,680
different ways that we can be as people.

210
00:13:47,680 --> 00:13:52,520
And that we will be able to more kind of broadly recognize and accept and understand each

211
00:13:52,520 --> 00:13:53,520
other.

212
00:13:53,520 --> 00:13:59,560
Like, maybe it'll be a space that allows us to open up and accept neurodivergent people,

213
00:13:59,560 --> 00:14:03,760
different frameworks of consciousness in human beings, right?

214
00:14:03,760 --> 00:14:07,840
And that was my hope and I kind of wrote and kind of researched from that perspective for

215
00:14:07,840 --> 00:14:08,840
quite some time.

216
00:14:08,840 --> 00:14:11,320
And I still have that hope in certain aspects.

217
00:14:11,320 --> 00:14:16,560
But I also have come to realize that what happens, unfortunately, very often is the opposite.

218
00:14:16,560 --> 00:14:21,600
Is that instead of people opening up and accepting more ways of recognizing human beings, what

219
00:14:21,600 --> 00:14:28,000
they do instead is apply the machine frameworks to the humans.

220
00:14:28,000 --> 00:14:33,720
And they dehumanize more often than they expand humanity.

221
00:14:33,720 --> 00:14:37,000
And that's really, yeah, it's really dangerous.

222
00:14:37,000 --> 00:14:44,040
It has real implications because once we start to think, oh, I know what human is and it's

223
00:14:44,040 --> 00:14:46,400
not X, Y and Z.

224
00:14:46,400 --> 00:14:50,920
And you, person I am suspicious of, are displaying X, Y and Z.

225
00:14:50,920 --> 00:14:54,200
You must not really be human.

226
00:14:54,200 --> 00:14:55,840
You don't count the same way.

227
00:14:55,840 --> 00:14:58,920
So we're living in Blade Runner basically.

228
00:14:58,920 --> 00:15:05,560
You're kind of essentially that kind of question of like, oh, but are you, are you really,

229
00:15:05,560 --> 00:15:09,320
if you're autistic, are you really human the same way?

230
00:15:09,320 --> 00:15:13,000
If you use M dashes, did you really write that?

231
00:15:13,000 --> 00:15:14,000
Right.

232
00:15:14,000 --> 00:15:19,000
That was another thing I had to ask you about is that there's not like, there's so many,

233
00:15:19,000 --> 00:15:23,960
there's so many like, oh, you know, detecting use of AI to try to discredit students.

234
00:15:23,960 --> 00:15:29,400
And I remember the one example that I couldn't recall is there was a teenage kid who used

235
00:15:29,400 --> 00:15:32,640
the word devoid in an essay, you know?

236
00:15:32,640 --> 00:15:37,200
And like same here, you know, Audi HD grew up like, you know, a nerd divergent pleasure

237
00:15:37,200 --> 00:15:39,680
to have in class, big words, things like that.

238
00:15:39,680 --> 00:15:45,920
Yeah, I'm writing one, lots of M dash, like the AHD M dash.

239
00:15:45,920 --> 00:15:48,560
My writing absolutely would have been flagged as AI.

240
00:15:48,560 --> 00:15:53,920
Yeah, that idea that like your, like that's the other thing too is that is that we fear

241
00:15:53,920 --> 00:16:04,120
rightly so the possibility of people being hoodwinked by mistaking AI either text or images

242
00:16:04,120 --> 00:16:06,240
or news or whatever for being real.

243
00:16:06,240 --> 00:16:11,320
But then there's also where people will look at something real and think like, oh, no,

244
00:16:11,320 --> 00:16:12,320
that's not real, that's AI.

245
00:16:12,320 --> 00:16:15,760
I see both those things happen in social media all the time.

246
00:16:15,760 --> 00:16:16,760
Exactly.

247
00:16:16,760 --> 00:16:19,920
I saw somebody talking about this the other day and it's like the thing that they hate most

248
00:16:19,920 --> 00:16:24,800
about this is that this whole like the framework of generative AI has made them just inherently

249
00:16:24,800 --> 00:16:30,360
suspicious of everything that gets put out on the world by their fellow human beings, right?

250
00:16:30,360 --> 00:16:35,480
Like that always, even when they like are training themselves not to even when they're actively

251
00:16:35,480 --> 00:16:39,920
pushing against it because a part of their mind that's always just asking the question of,

252
00:16:39,920 --> 00:16:41,040
is this real?

253
00:16:41,040 --> 00:16:42,040
Is this real?

254
00:16:42,040 --> 00:16:43,040
Is this real?

255
00:16:43,040 --> 00:16:51,400
Like that level of distrust and suspicion has real harms for like the social fabric of our

256
00:16:51,400 --> 00:16:55,880
entire community, our culture, our society, right?

257
00:16:55,880 --> 00:17:00,440
If we don't trust each other at this kind of deep, deep level, it's very, very difficult

258
00:17:00,440 --> 00:17:03,560
to like make a society together.

259
00:17:03,560 --> 00:17:06,360
It's very difficult to actually make a world, right?

260
00:17:06,360 --> 00:17:14,640
And like anyone who understands complex PTSD at all understands that that constant even

261
00:17:14,640 --> 00:17:18,440
low level hyper vigilance has effects in our nervous system.

262
00:17:18,440 --> 00:17:20,320
Yes, precisely.

263
00:17:20,320 --> 00:17:29,040
And that combined with a world in which there is increased hostility and increased fear,

264
00:17:29,040 --> 00:17:36,280
increased sense of danger at so many different levels and so many aspects of our daily lives.

265
00:17:36,280 --> 00:17:46,360
To then fundamentally undercut that basic groundwork of pro social communal reality is then to

266
00:17:46,360 --> 00:17:48,640
say we don't know what to do.

267
00:17:48,640 --> 00:17:53,560
And even if we are as you say, if we're not aware of it, we are on edge all the time.

268
00:17:53,560 --> 00:17:54,960
We're anxious.

269
00:17:54,960 --> 00:18:00,960
We are harmed internally by this a little bit more every single day.

270
00:18:00,960 --> 00:18:06,640
And that's a real outcome of like the vast proliferation of these systems and the ways

271
00:18:06,640 --> 00:18:10,720
that they've been kind of just integrated into our daily lives without our consent, without

272
00:18:10,720 --> 00:18:11,720
our say so.

273
00:18:11,720 --> 00:18:12,720
Anyway, sure.

274
00:18:12,720 --> 00:18:13,720
Right.

275
00:18:13,720 --> 00:18:17,960
And so you've got all of these things that we've together.

276
00:18:17,960 --> 00:18:22,920
You've got like the harm to critical thinking and critical analysis skills.

277
00:18:22,920 --> 00:18:24,880
You've got the suspicion and mistrust.

278
00:18:24,880 --> 00:18:33,560
You've got the lack of capability to really just trust each other or to discern what is real

279
00:18:33,560 --> 00:18:34,840
from what isn't.

280
00:18:34,840 --> 00:18:41,000
And the lack of incentive to try to discern reality from fiction.

281
00:18:41,000 --> 00:18:45,480
People will just spread an image and I tell the story fairly often, but I feel like it's

282
00:18:45,480 --> 00:18:46,960
worth repeating.

283
00:18:46,960 --> 00:18:51,520
You know, after Hurricane Helene here in North Carolina, there was a lot of images that were

284
00:18:51,520 --> 00:18:58,900
being spread around of like AI generated images of like a little girl in a water flotation

285
00:18:58,900 --> 00:19:04,120
safety vest carrying like a wet puppy, you know, standing in a way, steep water and yeah,

286
00:19:04,120 --> 00:19:07,600
you know, designed to be emotionally manipulated literally literally that right.

287
00:19:07,600 --> 00:19:12,320
And they're being used to kind of talk about like the failure of FEMA to get repair services

288
00:19:12,320 --> 00:19:17,840
or to get funds out to people so that they can get their homes back in order or even just

289
00:19:17,840 --> 00:19:20,080
to kind of survive long term.

290
00:19:20,080 --> 00:19:25,800
Like eventually, you know, I see this exchange on, I think it was Facebook, all places.

291
00:19:25,800 --> 00:19:29,320
And so like there's this one is another one of these pictures and in the comments, somebody

292
00:19:29,320 --> 00:19:32,440
is like, you know, that picture is AI, right?

293
00:19:32,440 --> 00:19:37,200
And the person who posted it responds, the picture might be AI.

294
00:19:37,200 --> 00:19:42,600
It probably is, but that doesn't mean the story isn't real.

295
00:19:42,600 --> 00:19:47,360
And to an extent, like that's, that is a like a factual statement, but it's a statement

296
00:19:47,360 --> 00:19:51,960
that has like connotations under it that are actually kind of really dark.

297
00:19:51,960 --> 00:19:56,160
When we're willing to just say, I don't care if this piece of evidence that I thought was

298
00:19:56,160 --> 00:20:00,760
like actual documentary evidence about the nature of reality isn't true.

299
00:20:00,760 --> 00:20:07,080
I care about the story it supports more than I care about the truth of falsity of the thing

300
00:20:07,080 --> 00:20:08,800
itself.

301
00:20:08,800 --> 00:20:11,360
That's a dark place to be.

302
00:20:11,360 --> 00:20:18,800
It reminds me back in 2017, I did an interview with Grant and it's Grant Morrison for any

303
00:20:18,800 --> 00:20:23,320
of those of you who are listening who is also the guest on episode one of the Re-Patterning

304
00:20:23,320 --> 00:20:24,320
podcast.

305
00:20:24,320 --> 00:20:29,640
But years ago in 2017, Grant and I did a talk in Los Angeles for anyone who wants to look

306
00:20:29,640 --> 00:20:30,640
it up, it's out there.

307
00:20:30,640 --> 00:20:36,440
It's under the name piece talks and Grant went on like, it's amazing kind of looking back

308
00:20:36,440 --> 00:20:38,120
that this was nine years ago.

309
00:20:38,120 --> 00:20:44,120
And what Grant was talking about was they believe that this is the sort of prophesied collapse

310
00:20:44,120 --> 00:20:48,440
of the Earth's fear of al-Quth and the moon's fear, yes, odd from like the cabolic tree of

311
00:20:48,440 --> 00:20:53,360
life because yes, odd, the moon's fear represents the imagination and dreams and you know, things

312
00:20:53,360 --> 00:20:58,440
that like anything that exists in the astral is something that is in our, you know, arguably

313
00:20:58,440 --> 00:21:02,880
I would say like in our imagination or at least we use our imagination to access what

314
00:21:02,880 --> 00:21:03,880
is there?

315
00:21:03,880 --> 00:21:05,920
Whereas the Earth's fear is like physical reality and everything.

316
00:21:05,920 --> 00:21:09,360
And thinking like this is reminding me of that conversation like thinking all the way

317
00:21:09,360 --> 00:21:13,600
back to that being like, yeah, like, and back then it was just like fake news, you know,

318
00:21:13,600 --> 00:21:15,280
like we don't know what's real and what's not.

319
00:21:15,280 --> 00:21:21,000
But now like nine years later, it's accelerated to such a breaking point where, you know, as

320
00:21:21,000 --> 00:21:23,720
Grant said, society took a drug it can't handle.

321
00:21:23,720 --> 00:21:26,000
Or the bad hatter's tea party.

322
00:21:26,000 --> 00:21:28,600
And nobody knows like what's up or down.

323
00:21:28,600 --> 00:21:35,200
I fear that a lot, you know, especially as someone who is slowly sort of coming more into

324
00:21:35,200 --> 00:21:41,440
the public eye or at least like intending to in some ways, I've been watching more closely

325
00:21:41,440 --> 00:21:45,800
just like what it's like for a person to be in the public eye.

326
00:21:45,800 --> 00:21:52,080
And even things that are, even things that look innocent on the, the surface say like AI

327
00:21:52,080 --> 00:21:57,440
fan edits, you know, of someone who's like a singer or a movie star or whatever.

328
00:21:57,440 --> 00:22:01,680
And you know, you're looking like just the other day, someone was like, oh, look at the trailer

329
00:22:01,680 --> 00:22:06,680
for this new movie featuring Angelina Jolie as Lilith.

330
00:22:06,680 --> 00:22:08,000
And I was like, oh, that looks dope.

331
00:22:08,000 --> 00:22:09,000
Why haven't I heard of that?

332
00:22:09,000 --> 00:22:11,240
Oh, because in AI, you know, I click on it.

333
00:22:11,240 --> 00:22:12,240
Cause it doesn't exist.

334
00:22:12,240 --> 00:22:13,240
Right.

335
00:22:13,240 --> 00:22:14,240
Exactly.

336
00:22:14,240 --> 00:22:20,400
You know, or like, but, or it gets even to the point where like there will be celebrities

337
00:22:20,400 --> 00:22:24,160
that people are like shipping together who aren't even dating and then they'll make

338
00:22:24,160 --> 00:22:25,560
these AI.

339
00:22:25,560 --> 00:22:31,600
And the more convincing they get, you know, it's like we're giving people the job.

340
00:22:31,600 --> 00:22:36,760
The tools to, and this is to say nothing of like, you know, AI, like sexual content, you

341
00:22:36,760 --> 00:22:38,640
know, revenge porn, things like that.

342
00:22:38,640 --> 00:22:43,920
And like, but we're giving people the tools to literally create narratives about each other

343
00:22:43,920 --> 00:22:45,440
that are false.

344
00:22:45,440 --> 00:22:47,160
That really gets under my skin.

345
00:22:47,160 --> 00:22:53,040
I'm not that live as it should quite frankly because we don't, I think I like the way

346
00:22:53,040 --> 00:22:56,400
they put it, you know, society took a drug that I can't handle, right?

347
00:22:56,400 --> 00:23:02,440
Like we don't, if you don't know what the potential range of real possibilities are, what you

348
00:23:02,440 --> 00:23:08,960
might experience, what the bad side effects might be, what a bad trip might be like under

349
00:23:08,960 --> 00:23:13,320
the influence of this particular new set of circumstances.

350
00:23:13,320 --> 00:23:14,600
Maybe don't take a full dose.

351
00:23:14,600 --> 00:23:17,120
Maybe talk to somebody who's dealt with it before.

352
00:23:17,120 --> 00:23:22,160
Maybe talk to people who studied this before you just kind of like go on all in on it.

353
00:23:22,160 --> 00:23:28,240
But no, what ended up happening is that it got shoved into every aspect of our society as

354
00:23:28,240 --> 00:23:29,240
quickly as possible.

355
00:23:29,240 --> 00:23:34,200
Because, you know, we had people who were warning about the dangers of deep fakes back

356
00:23:34,200 --> 00:23:37,960
when deep fakes first came onto the scene in 2014, 2015, right?

357
00:23:37,960 --> 00:23:43,480
Like the very first time somebody was able to make a adversarial network generated video

358
00:23:43,480 --> 00:23:48,960
image that made it look like somebody was doing something they absolutely were not.

359
00:23:48,960 --> 00:23:53,680
There were a whole bunch of people who study, you know, philosophy of science, philosophy

360
00:23:53,680 --> 00:23:57,840
of technology, political science who were like, oh shit, these are about to get dark.

361
00:23:57,840 --> 00:23:58,840
Yeah.

362
00:23:58,840 --> 00:24:02,560
Like that was their immediate response was this is going to be very bad.

363
00:24:02,560 --> 00:24:05,680
And sure of the ways it's going to be very bad, right?

364
00:24:05,680 --> 00:24:11,360
So it's not like there were warnings about the direction that this was going.

365
00:24:11,360 --> 00:24:15,160
It's that those warnings were not meaningfully taken up.

366
00:24:15,160 --> 00:24:22,240
They were not heated by people who are in a position to have perhaps put regulatory controls

367
00:24:22,240 --> 00:24:27,840
in place or to put the brakes on the rollout of these technologies to the general public

368
00:24:27,840 --> 00:24:34,600
in a meaningful way because, you know, monetary incentives, lack of full understanding, any

369
00:24:34,600 --> 00:24:37,480
number of things that could have been at play.

370
00:24:37,480 --> 00:24:42,840
But the fact remains that they want people didn't take this up when they could have done

371
00:24:42,840 --> 00:24:44,160
some real good.

372
00:24:44,160 --> 00:24:48,800
And putting the brakes on it and so at this point, we're kind of scrambling on the backflip

373
00:24:48,800 --> 00:24:56,280
position trying to get it kind of fixed, depending on who's at the level of power on a given

374
00:24:56,280 --> 00:24:57,280
election cycle.

375
00:24:57,280 --> 00:24:58,280
Yeah.

376
00:24:58,280 --> 00:25:05,400
It is kind of dark to be in that space of like, you know, we've got all those arms to critical

377
00:25:05,400 --> 00:25:06,400
thinking.

378
00:25:06,400 --> 00:25:07,400
We've got all that suspicion.

379
00:25:07,400 --> 00:25:13,600
We've got all that, that lack of a full understanding of what it all can do to us

380
00:25:13,600 --> 00:25:20,400
because that understanding has been actively kept from so many of us that it is very important

381
00:25:20,400 --> 00:25:25,640
and necessary, I think, to recognize that people have started to wake up to that, to recognize

382
00:25:25,640 --> 00:25:28,160
that that's the position that they've been placed in.

383
00:25:28,160 --> 00:25:32,480
And so many people are like, I don't want that to be the case.

384
00:25:32,480 --> 00:25:37,320
I can't allow that to continue to be the case and you see people trying to push back

385
00:25:37,320 --> 00:25:45,640
against it, trying to reclaim agency and some sense of control over what's being done

386
00:25:45,640 --> 00:25:48,080
to them without their say.

387
00:25:48,080 --> 00:25:49,400
Yeah.

388
00:25:49,400 --> 00:25:54,480
And for me, as someone who is both a student and teacher of narrative magic, it just kind

389
00:25:54,480 --> 00:25:59,680
of reinforces how important it is that if we are going to be in the public eye kind

390
00:25:59,680 --> 00:26:04,200
of at all, or really like, I mean, who's who's not in the public eye at this point, you

391
00:26:04,200 --> 00:26:08,840
know, with social media and everything, like most of us, like, you know, Andy Warhol said

392
00:26:08,840 --> 00:26:13,240
like in the future and everyone will be famous for 15 minutes or whatever.

393
00:26:13,240 --> 00:26:16,920
I think like nowadays, it's like every person will be famous to 15 people.

394
00:26:16,920 --> 00:26:17,920
Yeah, basically.

395
00:26:17,920 --> 00:26:18,920
Yeah.

396
00:26:18,920 --> 00:26:25,200
Like our micro networks, but it just reinforces to me the importance of reclaiming our

397
00:26:25,200 --> 00:26:30,280
narrative for ourselves, like the intentionality that we must have in the way that we walk

398
00:26:30,280 --> 00:26:35,600
through the world is more important than ever because I know, you know, my strongest defense

399
00:26:35,600 --> 00:26:43,760
against that, I feel anyway is that I may live so intentionally and so purposely and so

400
00:26:43,760 --> 00:26:50,000
fully embody my integrity and my values that if anything fate came out about me, people

401
00:26:50,000 --> 00:26:53,240
would look at that and say, even if it's like, you know, it's another like year or two into

402
00:26:53,240 --> 00:26:58,080
the future and you really can't tell, you know, but just by visuals alone, what's fake

403
00:26:58,080 --> 00:26:59,080
and what's not.

404
00:26:59,080 --> 00:27:02,320
And I really hope that people would look at that and say, that's not Arden Leigh, like not

405
00:27:02,320 --> 00:27:04,560
in a million years, which she ever said that.

406
00:27:04,560 --> 00:27:05,560
Right.

407
00:27:05,560 --> 00:27:09,520
And that's like the best defense I think that we have is to so fully embody our values that

408
00:27:09,520 --> 00:27:14,280
we were never mistaken for being something that we're not, which is like shadow work, right?

409
00:27:14,280 --> 00:27:16,400
Yeah, but that's exactly it.

410
00:27:16,400 --> 00:27:22,000
It's like at that point, the amount of cultural and social work that has to be done to

411
00:27:22,000 --> 00:27:27,160
get people to think that way about themselves and to dig down on themselves and to know what

412
00:27:27,160 --> 00:27:33,640
it means to be and live themselves in that way as that kind of protection then becomes something

413
00:27:33,640 --> 00:27:38,560
that we have to kind of fight for because as it currently stands in, you know, Western society,

414
00:27:38,560 --> 00:27:42,280
let's call it, it's not the norm in any way shape or form, right?

415
00:27:42,280 --> 00:27:47,600
Like we are not encouraged to do that level of introspection, that level of self understanding,

416
00:27:47,600 --> 00:27:54,320
that level of reckoning with all the aspects of ourselves to be able to embody it, to

417
00:27:54,320 --> 00:27:59,880
be able to inhabit it, to be able to let it inhabit us in a deep way.

418
00:27:59,880 --> 00:28:02,400
And that's the problem, right?

419
00:28:02,400 --> 00:28:04,400
Like that's encouraging it.

420
00:28:04,400 --> 00:28:11,160
I've always said it's a great thing, however.

421
00:28:11,160 --> 00:28:14,760
We are not the ones who get to write the curriculum for schools this way.

422
00:28:14,760 --> 00:28:20,960
We are not the ones who get to make it such that, you know, as children go through the process

423
00:28:20,960 --> 00:28:25,160
of learning how to play with each other and share and, you know, they're also trying to

424
00:28:25,160 --> 00:28:30,080
reckon with what it means that they have selfish impulses or what it means that they don't

425
00:28:30,080 --> 00:28:33,440
want to, you know, be with other people sometimes, right?

426
00:28:33,440 --> 00:28:40,000
Like all of those things are, you know, they're in there to an extent, but it's not always

427
00:28:40,000 --> 00:28:47,360
the case that we are like encouraged as we develop as humans, as we develop as people,

428
00:28:47,360 --> 00:28:51,960
that we're encouraged to say, "Here are the aspects of yourself.

429
00:28:51,960 --> 00:28:52,960
Why do they exist?

430
00:28:52,960 --> 00:28:53,960
What do you need?

431
00:28:53,960 --> 00:28:54,960
What do you want?

432
00:28:54,960 --> 00:29:00,040
What are you trying to, what values are you trying to satisfy by your behavior?

433
00:29:00,040 --> 00:29:01,840
And are you actually satisfying those values?

434
00:29:01,840 --> 00:29:05,400
Or are you like kind of pushing against them because you're just kind of reaching around

435
00:29:05,400 --> 00:29:07,800
for what out, right?

436
00:29:07,800 --> 00:29:11,360
And so that's, you know, that gets to that kind of that deeper conversation about like,

437
00:29:11,360 --> 00:29:16,520
what is it as a society that we are trying to satisfy when we think and talk about

438
00:29:16,520 --> 00:29:25,080
things like truth, reality, democracy, you know, what are those values supposedly accomplishing

439
00:29:25,080 --> 00:29:26,640
for us?

440
00:29:26,640 --> 00:29:30,120
And are we actually satisfying those things?

441
00:29:30,120 --> 00:29:36,840
Are we actually satisfying what democracy, what truth, what reality, what society, prosocial,

442
00:29:36,840 --> 00:29:39,000
consensus making?

443
00:29:39,000 --> 00:29:40,800
What those things need from us?

444
00:29:40,800 --> 00:29:47,080
Or are we just trying to like place a word over top of behavior that we want to engage

445
00:29:47,080 --> 00:29:48,080
in anyway?

446
00:29:48,080 --> 00:29:50,640
Too many times it's the latter.

447
00:29:50,640 --> 00:29:53,120
Well, that's why I started a podcast.

448
00:29:53,120 --> 00:29:58,520
So I'm contributing my, my drop into the, the ocean of information out there.

449
00:29:58,520 --> 00:30:00,120
That's what you can do.

450
00:30:00,120 --> 00:30:04,360
Of truth and objective reality, something that I really also resonate with in your work

451
00:30:04,360 --> 00:30:11,440
is this way that you point out that we often have a misguided tendency to assume that our

452
00:30:11,440 --> 00:30:14,960
technology is neutral because it's technology.

453
00:30:14,960 --> 00:30:19,760
It's non-sentient for the most part, at least as far as we know now.

454
00:30:19,760 --> 00:30:28,440
But as we, as we have, right, depend, as it goes, we'll see how the, we'll see what the

455
00:30:28,440 --> 00:30:33,520
future holds as we make some changes in our current understandings and maybe move away

456
00:30:33,520 --> 00:30:36,120
from a large language model framework for a little while.

457
00:30:36,120 --> 00:30:41,960
We might get down that path, but yeah, for now, for now our tech is, is not there.

458
00:30:41,960 --> 00:30:49,520
Yeah, and so we have, we have this tendency to ascribe it this sort of, yeah, like a neutrality.

459
00:30:49,520 --> 00:30:54,840
But what I love about your work is that you point out that our values, like our human

460
00:30:54,840 --> 00:31:01,280
values and our social structures and our biases, they're always in our technology.

461
00:31:01,280 --> 00:31:06,080
And I remember playing an example of this and I saw this, gosh, it had to have been like,

462
00:31:06,080 --> 00:31:10,280
maybe like eight to 10 years ago, but it was just such a great example as someone had

463
00:31:10,280 --> 00:31:17,200
invented a hand soap dispenser and they were, you know, yeah, I know you know this one, right?

464
00:31:17,200 --> 00:31:22,000
So like, a white dude puts his hand under the soap dispenser, oh, look, it works.

465
00:31:22,000 --> 00:31:25,360
And we think like, oh, default, you know, whatever.

466
00:31:25,360 --> 00:31:29,480
And then a black man does the same thing and the soap doesn't come out because the censor

467
00:31:29,480 --> 00:31:31,480
is not reading the color of his hand.

468
00:31:31,480 --> 00:31:37,600
And like, and I think about so many people who would just see like the first half of that

469
00:31:37,600 --> 00:31:41,560
video and just automatically assume like, oh, that's how the soap dispenser works.

470
00:31:41,560 --> 00:31:45,680
And yet it's not including this huge amount of the population.

471
00:31:45,680 --> 00:31:50,160
And so I'd love to hear also your perspective, like you also bring up in some of your work,

472
00:31:50,160 --> 00:31:57,160
for example, the bridges that were built in Central Park in New York City that didn't

473
00:31:57,160 --> 00:32:04,040
allow buses to be able to go through the routes because most low income people would be

474
00:32:04,040 --> 00:32:05,920
like probably taking a bus.

475
00:32:05,920 --> 00:32:11,040
So if the bus can't go through the park, the park then gets reserved for folks who can

476
00:32:11,040 --> 00:32:12,440
own a car in New York City.

477
00:32:12,440 --> 00:32:15,240
And I lived in New York City for many years and I never owned a car there.

478
00:32:15,240 --> 00:32:20,520
Like, that was, you know, I maybe got to take a taxi through Central Park on search

479
00:32:20,520 --> 00:32:21,520
ways.

480
00:32:21,520 --> 00:32:24,240
You know, I was in that middle tier of income.

481
00:32:24,240 --> 00:32:31,400
But in the same way, like our biases are always in the things that we create.

482
00:32:31,400 --> 00:32:37,160
And I'd love for you to speak more to how that shows up in artificial intelligence and

483
00:32:37,160 --> 00:32:40,080
the computer programming and things that are being developed today.

484
00:32:40,080 --> 00:32:41,080
Yeah.

485
00:32:41,080 --> 00:32:50,280
So as a kind of preamble to that, I'll say when I was in Austin, Texas in 2019, I was giving

486
00:32:50,280 --> 00:32:53,880
a presentation at the Society for Philosophy of Technology.

487
00:32:53,880 --> 00:32:55,920
I was actually talking about exactly this.

488
00:32:55,920 --> 00:32:59,240
I was there to present on this topic on these ideas, right?

489
00:32:59,240 --> 00:33:04,840
And so I am in the bathroom prior to somebody else's talk.

490
00:33:04,840 --> 00:33:09,440
I had actually literally just finished giving my talk on like discrimination in technology

491
00:33:09,440 --> 00:33:11,000
on how biases show up in tech.

492
00:33:11,000 --> 00:33:12,840
And I'm in the bathroom.

493
00:33:12,840 --> 00:33:14,880
And the soap dispenser doesn't work for me.

494
00:33:14,880 --> 00:33:22,380
I literally had to like take the soap dispenser apart to get the soap out of it so that I could

495
00:33:22,380 --> 00:33:23,380
wash my hands.

496
00:33:23,380 --> 00:33:24,380
Oh my gosh.

497
00:33:24,380 --> 00:33:25,380
And I walk out.

498
00:33:25,380 --> 00:33:28,080
I hope you worked that into your talk, right?

499
00:33:28,080 --> 00:33:29,080
Absolutely did.

500
00:33:29,080 --> 00:33:32,240
I walk out and I'm like, okay.

501
00:33:32,240 --> 00:33:35,800
So we're going to have a conversation and then we're going to have a different conversation.

502
00:33:35,800 --> 00:33:40,560
They're going to be the same conversation, but it comes in two parts.

503
00:33:40,560 --> 00:33:49,760
So it was just such a striking, it was like in the year 2019, this is still happening.

504
00:33:49,760 --> 00:33:50,760
It's wild, right?

505
00:33:50,760 --> 00:33:54,040
I feel like that's a narrative magic moment too.

506
00:33:54,040 --> 00:33:55,040
It was.

507
00:33:55,040 --> 00:33:56,040
Thank you, universe.

508
00:33:56,040 --> 00:33:57,040
Yeah.

509
00:33:57,040 --> 00:34:01,040
Our art imitates life imitates art.

510
00:34:01,040 --> 00:34:07,200
It's like, oh, universe decided to kind of like give me that hands-on example as it were.

511
00:34:07,200 --> 00:34:08,200
Okay.

512
00:34:08,200 --> 00:34:13,940
And so that is to say, like, yeah, these things, you know, you have, as you said, like these

513
00:34:13,940 --> 00:34:16,920
assumptions about defaults that insist in people's minds.

514
00:34:16,920 --> 00:34:21,000
And so those assumptions then don't get questioned because they're assumptions in order

515
00:34:21,000 --> 00:34:22,000
to question them.

516
00:34:22,000 --> 00:34:23,400
You have to recognize them.

517
00:34:23,400 --> 00:34:26,640
If you recognize them, they wouldn't continue to be assumptions.

518
00:34:26,640 --> 00:34:27,640
You would question them.

519
00:34:27,640 --> 00:34:29,160
You would dig down on them.

520
00:34:29,160 --> 00:34:34,320
And so people have these assumed categories in their mind and they just think, you know,

521
00:34:34,320 --> 00:34:37,720
the person who designed the thing, the people who are testing it, you know, those will

522
00:34:37,720 --> 00:34:38,720
be the people.

523
00:34:38,720 --> 00:34:39,720
That's the people.

524
00:34:39,720 --> 00:34:41,000
These are just people.

525
00:34:41,000 --> 00:34:42,000
And now they've tested it.

526
00:34:42,000 --> 00:34:43,000
We tested it.

527
00:34:43,000 --> 00:34:45,520
You put your hand under the sense of the sense of the sense of the sense of so cool moving

528
00:34:45,520 --> 00:34:46,520
on.

529
00:34:46,520 --> 00:34:50,320
Except you don't ask the question, how does that sense of work?

530
00:34:50,320 --> 00:34:51,880
Oh, reflects light.

531
00:34:51,880 --> 00:34:57,000
It reflects infrared light off of a particular type of pigmentation within the skin.

532
00:34:57,000 --> 00:35:01,520
And so at a certain level of reflectivity, then the sense of read obstacle, it dispenses

533
00:35:01,520 --> 00:35:02,520
so upon the obstacle.

534
00:35:02,520 --> 00:35:03,520
Oh, okay.

535
00:35:03,520 --> 00:35:04,520
Cool.

536
00:35:04,520 --> 00:35:10,920
Do different types of pigmentation perhaps have different reflectivity to IR?

537
00:35:10,920 --> 00:35:14,080
Because once you ask that question, the answer is, oh, shit.

538
00:35:14,080 --> 00:35:15,720
Yeah, they do.

539
00:35:15,720 --> 00:35:18,840
And that has implications, not just for like soap dispensers.

540
00:35:18,840 --> 00:35:22,880
It's got implications for, well, pulse oxymeters.

541
00:35:22,880 --> 00:35:27,280
So we learned during the early days of the pandemic, pulse oxymetry was very necessary

542
00:35:27,280 --> 00:35:29,720
to make sure that people had enough oxygen in their blood.

543
00:35:29,720 --> 00:35:32,080
You get a little thing clips on your finger.

544
00:35:32,080 --> 00:35:34,680
I've got one floating around here in the house somewhere.

545
00:35:34,680 --> 00:35:38,800
It clips onto the end of the finger, sends IR light through the little skin.

546
00:35:38,800 --> 00:35:46,240
Problem being that melanin and pigmentation within the body can read and bounce light

547
00:35:46,240 --> 00:35:48,600
in much the same way as infrared blood cells can.

548
00:35:48,600 --> 00:35:54,880
And so many times people with darker skin will have a higher pulse oxymetry reading.

549
00:35:54,880 --> 00:35:59,240
It will read in the high or mid 90s, which is a decent level.

550
00:35:59,240 --> 00:36:03,520
Which is supposed to hopefully have about 100% pulse oxym.

551
00:36:03,520 --> 00:36:08,480
But it'll read up to 3% higher than it actually is.

552
00:36:08,480 --> 00:36:09,480
Wow.

553
00:36:09,480 --> 00:36:13,000
And if you're assessing somebody for like the danger zone of when they might need to be

554
00:36:13,000 --> 00:36:20,560
say, intubated and how oxygen forced into their bloodstream, that 3% makes a real difference.

555
00:36:20,560 --> 00:36:25,520
And so it was found luckily kind of quickly, but not quickly enough to say as many lives

556
00:36:25,520 --> 00:36:31,240
it could that black people were getting misread from pulse oxymeters.

557
00:36:31,240 --> 00:36:36,760
And so they were getting rescue oxygen at lower rates than they were supposed to have.

558
00:36:36,760 --> 00:36:40,600
That has real long-term implications.

559
00:36:40,600 --> 00:36:43,080
It does damage.

560
00:36:43,080 --> 00:36:45,240
And so that's health technologies.

561
00:36:45,240 --> 00:36:51,360
When it comes to AI, you know, we're not looking at the visual side of things as much necessarily.

562
00:36:51,360 --> 00:36:59,840
We're looking primarily at the framework of concepts and how concepts make their way into,

563
00:36:59,840 --> 00:37:03,080
you know, in this case large language model spaces.

564
00:37:03,080 --> 00:37:08,040
And so prior to large language models, we had language models.

565
00:37:08,040 --> 00:37:11,160
They were language models were just smaller.

566
00:37:11,160 --> 00:37:16,840
And they mapped language within a vector space in a smaller rate and they could do.

567
00:37:16,840 --> 00:37:19,880
It's what we used to have predicted text built from.

568
00:37:19,880 --> 00:37:20,880
Right.

569
00:37:20,880 --> 00:37:26,480
So that framework gets kind of beefed up in the large language model space.

570
00:37:26,480 --> 00:37:28,680
You get more words that it can handle.

571
00:37:28,680 --> 00:37:30,640
It can process more information.

572
00:37:30,640 --> 00:37:35,360
It can generate larger, you know, amounts of text instead of just the next word and the

573
00:37:35,360 --> 00:37:39,840
next word it can do next phrase, next sentence, whole paragraph.

574
00:37:39,840 --> 00:37:40,840
Right.

575
00:37:40,840 --> 00:37:49,240
And so the training for large language models comes from the whole whole internet up to 2021.

576
00:37:49,240 --> 00:37:54,840
And more have done more since, but you know, that's for the early days of chat GPT puts

577
00:37:54,840 --> 00:37:59,680
now Google Gemini, Facebook's, Ling models.

578
00:37:59,680 --> 00:38:04,240
Yeah, all the large language models that people are used with familiar with, Anthropics

579
00:38:04,240 --> 00:38:05,400
Quad.

580
00:38:05,400 --> 00:38:10,440
And that representation of the internet contains within it all of the trainings that were

581
00:38:10,440 --> 00:38:12,640
done on earlier language models.

582
00:38:12,640 --> 00:38:13,640
Right.

583
00:38:13,640 --> 00:38:17,480
And so the thing about language models is that the vast majority of them, the ones that

584
00:38:17,480 --> 00:38:23,200
were kind of cutting edge top of the line, is their models like word to vac.

585
00:38:23,200 --> 00:38:28,800
That's word number two, vac VEC and glove GLOV.

586
00:38:28,800 --> 00:38:42,400
Big G and big V those are trained on some really nasty stuff.

587
00:38:42,400 --> 00:38:46,640
And so the, let's just, you know, take a step back from that and say like when it comes

588
00:38:46,640 --> 00:38:50,960
to research, scientific research, especially when you're doing like data research, you

589
00:38:50,960 --> 00:38:53,240
have to purchase or obtain data sets.

590
00:38:53,240 --> 00:38:57,880
Like you have to get access to data to work with, especially if you're working with a

591
00:38:57,880 --> 00:38:59,360
large amount of data.

592
00:38:59,360 --> 00:39:00,360
Right.

593
00:39:00,360 --> 00:39:06,760
And so researchers are always going to look for things that allow them to access data cheaply,

594
00:39:06,760 --> 00:39:10,040
free or otherwise easily in partnership with other people.

595
00:39:10,040 --> 00:39:11,040
Right.

596
00:39:11,040 --> 00:39:16,040
So in the case of language models like word to vac and glove.

597
00:39:16,040 --> 00:39:22,560
The data set that they were able to use to build these language models came from something

598
00:39:22,560 --> 00:39:32,360
called the N-RON Corpus built out of the emails that went into public domain as a result

599
00:39:32,360 --> 00:39:38,720
of the fraud trial of the N-RON corporation in the early 2000s.

600
00:39:38,720 --> 00:39:42,760
N-RON which defrauded its investors for literally billions of dollars.

601
00:39:42,760 --> 00:39:44,080
I remember that.

602
00:39:44,080 --> 00:39:44,920
Yeah.

603
00:39:44,920 --> 00:39:50,000
And so you've got as a result of this, like the emails go into the public domain because

604
00:39:50,000 --> 00:39:51,240
they're entered into evidence.

605
00:39:51,240 --> 00:39:52,240
Right.

606
00:39:52,240 --> 00:39:53,320
And it's a federal fraud trial.

607
00:39:53,320 --> 00:39:55,720
And so that's belongs to the public.

608
00:39:55,720 --> 00:40:05,320
And so what that then becomes is over 600,000 emails, millions of lines of natural language.

609
00:40:05,320 --> 00:40:08,960
And so language models and large language models are built around a framework.

610
00:40:08,960 --> 00:40:10,760
What's called natural language processing.

611
00:40:10,760 --> 00:40:11,760
Right.

612
00:40:11,760 --> 00:40:12,760
Like that's the goal of it.

613
00:40:12,760 --> 00:40:13,760
That's the idea behind it.

614
00:40:13,760 --> 00:40:14,760
You're supposed to do.

615
00:40:14,760 --> 00:40:19,640
You're supposed to be able to just type or speak and talk like we're talking right now and

616
00:40:19,640 --> 00:40:21,640
have it do what it does.

617
00:40:21,640 --> 00:40:25,960
It's like the computer on Star Trek is the dream that these people have.

618
00:40:25,960 --> 00:40:33,120
And so you have millions of lines of free natural language communication, like actual conversations

619
00:40:33,120 --> 00:40:34,120
between people.

620
00:40:34,120 --> 00:40:40,480
You're going to use it because otherwise you have to like build that corpus, that collection

621
00:40:40,480 --> 00:40:42,600
of language from scratch.

622
00:40:42,600 --> 00:40:46,160
Or you have to get it somewhere else, you have to buy it from someone else who controls

623
00:40:46,160 --> 00:40:51,520
it, but nobody can own the en-rock corpus because it's public domain.

624
00:40:51,520 --> 00:40:55,680
That means it got used over and over and over and over again.

625
00:40:55,680 --> 00:41:00,320
It is overrepresented within the training data of language models.

626
00:41:00,320 --> 00:41:07,600
So let's ask the question, how do we think that people who are on trial for federal fraud

627
00:41:07,600 --> 00:41:10,320
talk to one another?

628
00:41:10,320 --> 00:41:11,320
Right.

629
00:41:11,320 --> 00:41:19,680
How do we think they talk about other people?

630
00:41:19,680 --> 00:41:23,160
It's not good.

631
00:41:23,160 --> 00:41:28,220
And so there's like, there's sexist jokes, there's racist jokes, there's like, we remember

632
00:41:28,220 --> 00:41:33,120
chain letters back in the early days of email, like the early proto means that were chain

633
00:41:33,120 --> 00:41:34,120
letters.

634
00:41:34,120 --> 00:41:37,560
Like there's a bunch of like gross ones of those in there.

635
00:41:37,560 --> 00:41:41,600
There's random, just kind of like funny, supposedly heartwarming, you know, also probably

636
00:41:41,600 --> 00:41:45,560
entirely made up, chain letter stories and stuff like that in there too.

637
00:41:45,560 --> 00:41:48,920
And then there's the way they talk about their clients and the way they talk about like,

638
00:41:48,920 --> 00:41:52,560
well, actually defrauding their investors and the way they talk about their co-workers

639
00:41:52,560 --> 00:41:58,280
and some off-color remarks about certain members of their co-working teams and all of that

640
00:41:58,280 --> 00:41:59,280
in there.

641
00:41:59,280 --> 00:42:02,720
And just the fact that like, they're liars at the end of the day, like it's a fraud case.

642
00:42:02,720 --> 00:42:03,720
Exactly.

643
00:42:03,720 --> 00:42:09,040
You train you in the system or fraud.

644
00:42:09,040 --> 00:42:17,160
And so when you think about what this system is, right, so think about what like a natural

645
00:42:17,160 --> 00:42:22,240
language processing, large language, or sorry, early days, language is, it's a machine learning

646
00:42:22,240 --> 00:42:28,720
system and machine learning at base is pattern matching and pattern reinforcing.

647
00:42:28,720 --> 00:42:36,720
That's goal, it's job is to identify patterns and then amplify those patterns and it is

648
00:42:36,720 --> 00:42:40,000
very, very good at that job.

649
00:42:40,000 --> 00:42:48,600
So you got racism, you got sexism, you got ableism, you've got all kinds of like, just fraud talk,

650
00:42:48,600 --> 00:42:54,680
anti social behavior, demonstrated by humans as the patterns that show up over and over

651
00:42:54,680 --> 00:43:03,640
and over in this data and then you give this to the system to amplify, capitalize on, reiterate,

652
00:43:03,640 --> 00:43:04,640
re-inscribe.

653
00:43:04,640 --> 00:43:06,640
Like nobody thought about that, you know?

654
00:43:06,640 --> 00:43:07,640
No.

655
00:43:07,640 --> 00:43:08,640
Literally no.

656
00:43:08,640 --> 00:43:13,360
Like literally it did not come up in the early days because at that point in time, this

657
00:43:13,360 --> 00:43:17,520
conversation about like, hey, maybe we should be aware of the fact that human biases can

658
00:43:17,520 --> 00:43:18,960
affect the development of technology.

659
00:43:18,960 --> 00:43:21,320
That conversation was very few places.

660
00:43:21,320 --> 00:43:26,280
It wasn't nowhere, but there were very few of us who were like trying to raise our hands

661
00:43:26,280 --> 00:43:31,320
and point this out and a lot of people were like, that doesn't matter at all.

662
00:43:31,320 --> 00:43:32,640
This is mass.

663
00:43:32,640 --> 00:43:33,640
It's just code.

664
00:43:33,640 --> 00:43:39,160
You can't be racist, code can't be sexist, that doesn't make any sense.

665
00:43:39,160 --> 00:43:40,680
It's mass.

666
00:43:40,680 --> 00:43:47,880
It wasn't until a group of researchers like tore it down and like demonstrated the vector

667
00:43:47,880 --> 00:43:49,960
mapping because that's what it does, right?

668
00:43:49,960 --> 00:43:56,440
It takes a specific numerical associations to the words and phrases and then maps those

669
00:43:56,440 --> 00:44:00,720
how close or how far they are from each other within a vector space.

670
00:44:00,720 --> 00:44:05,560
It's a statistical ratio and you can map that on literally like a graph.

671
00:44:05,560 --> 00:44:09,160
You can just say this line goes here and this line is close to it.

672
00:44:09,160 --> 00:44:10,720
This line is further away from it.

673
00:44:10,720 --> 00:44:14,000
So this line and this line are most likely to find each other.

674
00:44:14,000 --> 00:44:17,200
This line and this line less likely to find each other.

675
00:44:17,200 --> 00:44:18,680
It's all it is.

676
00:44:18,680 --> 00:44:22,560
They mapped this out and they showed very clearly.

677
00:44:22,560 --> 00:44:23,560
Okay.

678
00:44:23,560 --> 00:44:31,280
So let's give it the word man, the word king, the word woman and the word queen.

679
00:44:31,280 --> 00:44:33,200
Let's map those vectors.

680
00:44:33,200 --> 00:44:35,720
Man, woman, close to each other.

681
00:44:35,720 --> 00:44:36,720
Okay.

682
00:44:36,720 --> 00:44:38,520
Gender related vectors.

683
00:44:38,520 --> 00:44:39,920
Got it.

684
00:44:39,920 --> 00:44:43,360
Man, king, close to each other.

685
00:44:43,360 --> 00:44:44,360
Ha.

686
00:44:44,360 --> 00:44:46,760
Royalty related vector.

687
00:44:46,760 --> 00:44:47,760
Gender royalty.

688
00:44:47,760 --> 00:44:48,760
Got it.

689
00:44:48,760 --> 00:44:50,920
King queen.

690
00:44:50,920 --> 00:44:51,920
Close to each other.

691
00:44:51,920 --> 00:44:53,840
Queen further away from man.

692
00:44:53,840 --> 00:44:56,120
I see what's going on here.

693
00:44:56,120 --> 00:44:57,120
Woman queen.

694
00:44:57,120 --> 00:44:58,560
Close to each other.

695
00:44:58,560 --> 00:45:00,360
Woman further away from king.

696
00:45:00,360 --> 00:45:02,320
Ah, understood.

697
00:45:02,320 --> 00:45:05,000
That's the map of the vectors.

698
00:45:05,000 --> 00:45:06,520
Makes sense.

699
00:45:06,520 --> 00:45:11,320
Aside from a couple of cases in history, not too many women kings, not too many male queens

700
00:45:11,320 --> 00:45:13,400
in the regal sense.

701
00:45:13,400 --> 00:45:14,400
Sure.

702
00:45:14,400 --> 00:45:17,400
But let's do another said.

703
00:45:17,400 --> 00:45:25,880
Man, woman, president, secretary.

704
00:45:25,880 --> 00:45:29,320
Man, woman, doctor nurse.

705
00:45:29,320 --> 00:45:32,520
Yup, yup.

706
00:45:32,520 --> 00:45:36,840
Man, woman, CEO, teacher.

707
00:45:36,840 --> 00:45:39,240
And like to an extent, it makes sense.

708
00:45:39,240 --> 00:45:42,280
This would be the case that those maps would show off that way.

709
00:45:42,280 --> 00:45:45,120
Not based just on the inner on purpose.

710
00:45:45,120 --> 00:45:49,520
But based on the history of the world and all the language that these systems were trained

711
00:45:49,520 --> 00:45:50,520
on, right?

712
00:45:50,520 --> 00:45:51,520
Sure.

713
00:45:51,520 --> 00:45:58,680
It has traditionally, truly been the case that there are fewer women CEOs, fewer non-binary

714
00:45:58,680 --> 00:46:05,080
CEOs, fewer women presidents, binary presidents, than there have been men who are CEOs and presidents.

715
00:46:05,080 --> 00:46:06,960
That's just true.

716
00:46:06,960 --> 00:46:13,400
So when you map that out, when you represent that in the vectors, you're going to see

717
00:46:13,400 --> 00:46:15,080
that play out.

718
00:46:15,080 --> 00:46:19,160
Because it's an accurate representation of the data that's been drained on.

719
00:46:19,160 --> 00:46:21,240
But that data doesn't come from nowhere.

720
00:46:21,240 --> 00:46:26,680
It doesn't just so happen to be the case that there's fewer women CEOs than there are male

721
00:46:26,680 --> 00:46:27,680
CEOs, right?

722
00:46:27,680 --> 00:46:30,600
Like that's a fact of history.

723
00:46:30,600 --> 00:46:31,600
That's a social fact.

724
00:46:31,600 --> 00:46:33,600
That's a fact of values.

725
00:46:33,600 --> 00:46:37,440
That's a bias, prejudicial fact.

726
00:46:37,440 --> 00:46:41,520
It's not just, we see Daisy.

727
00:46:41,520 --> 00:46:47,160
We happened to hire fewer women CEOs over history, whatever shall we guess we'll just fix

728
00:46:47,160 --> 00:46:48,760
that all of a sudden.

729
00:46:48,760 --> 00:46:56,480
No, it's a history of systemic structural sex, a zone, prejudice, etc.

730
00:46:56,480 --> 00:46:59,160
That re-inscribes those things.

731
00:46:59,160 --> 00:47:03,760
And so if you want a system, this is the way I tend to put it when I write about this,

732
00:47:03,760 --> 00:47:06,920
is like, if you want a system that only talks about the past, that talks about the way

733
00:47:06,920 --> 00:47:10,480
things have been, then cool language models.

734
00:47:10,480 --> 00:47:15,280
Large language model successors perfectly positioned to tell you about the past.

735
00:47:15,280 --> 00:47:19,640
They will highlight the patterns of the past for you, extremely well.

736
00:47:19,640 --> 00:47:21,880
They'll randomly make crap up.

737
00:47:21,880 --> 00:47:27,280
But overall, the crap that they make up will resonate with the stories about the past that

738
00:47:27,280 --> 00:47:29,040
they have been drained on.

739
00:47:29,040 --> 00:47:30,040
Makes sense.

740
00:47:30,040 --> 00:47:33,200
But that's not what we want those systems to do.

741
00:47:33,200 --> 00:47:34,960
And I say we is like society.

742
00:47:34,960 --> 00:47:39,680
We is even including, especially the people who are selling these systems.

743
00:47:39,680 --> 00:47:43,080
They don't sell the systems to us as like, "Oh, they'll tell you anything you want to

744
00:47:43,080 --> 00:47:44,880
know about patterns of the past."

745
00:47:44,880 --> 00:47:47,120
No, they'll write your email for you.

746
00:47:47,120 --> 00:47:51,940
The email about your meetings tomorrow about who you should hire, about who you should

747
00:47:51,940 --> 00:47:56,440
bring in as a guest speaker, about what you should do for a company retreat.

748
00:47:56,440 --> 00:47:58,160
They'll write those emails for you.

749
00:47:58,160 --> 00:48:00,280
You know, emails about the future.

750
00:48:00,280 --> 00:48:03,400
They'll help you hire a new job candidate.

751
00:48:03,400 --> 00:48:08,240
They'll figure out whose resume is the best.

752
00:48:08,240 --> 00:48:11,200
Whose got the best credentials?

753
00:48:11,200 --> 00:48:14,400
You know, questions about the future.

754
00:48:14,400 --> 00:48:18,600
Like all of that still tries to map.

755
00:48:18,600 --> 00:48:23,440
Our questions about the future are trying to move forward in the world onto these patterns

756
00:48:23,440 --> 00:48:25,120
about the past.

757
00:48:25,120 --> 00:48:28,000
And so it should come as no surprise.

758
00:48:28,000 --> 00:48:32,040
And yet people are still somehow always shocked when it turns out to be the case that these

759
00:48:32,040 --> 00:48:35,440
systems apply those old patterns to these new situations.

760
00:48:35,440 --> 00:48:41,940
And yet, resume sorters that preference, male, named resumes over female, named resumes

761
00:48:41,940 --> 00:48:45,840
or preference, quote unquote, "white sounding names" over quote unquote, "black sounding

762
00:48:45,840 --> 00:48:53,440
names" that as recently as last year would take any mention of disability in a resume and

763
00:48:53,440 --> 00:48:59,000
would down-rank that resume because it has been trained to think of disability as inherently

764
00:48:59,000 --> 00:49:01,000
negative.

765
00:49:01,000 --> 00:49:09,200
So those biases very, very, very easily get turned into math and get turned into data that

766
00:49:09,200 --> 00:49:13,080
then change how that system functions.

767
00:49:13,080 --> 00:49:18,920
And it wasn't until, again, pretty recently that people started to take that seriously.

768
00:49:18,920 --> 00:49:24,840
As recently as 2019, different contracts, those in California, as in Stanford, had at

769
00:49:24,840 --> 00:49:32,800
the one of the biggest AI conferences period, triple AI, spring symposium series, association

770
00:49:32,800 --> 00:49:36,720
for the advancement of artificial intelligence is what that stands for.

771
00:49:36,720 --> 00:49:42,440
And they're like global standards of AI as like a body, right?

772
00:49:42,440 --> 00:49:50,880
Had in 2019, a like lecture hall-sized argument between some of us who were trying to like

773
00:49:50,880 --> 00:49:58,200
have the conversation that you and I are having right now and AI researchers who refused

774
00:49:58,200 --> 00:50:06,640
to tear it, who just did not, did not want to take in in any way, shape, or form the idea

775
00:50:06,640 --> 00:50:12,960
that values made a difference in the way the math got done and then the outputs from

776
00:50:12,960 --> 00:50:17,080
the math that got created from those previous values.

777
00:50:17,080 --> 00:50:19,080
They just wouldn't hear it.

778
00:50:19,080 --> 00:50:20,080
Wow.

779
00:50:20,080 --> 00:50:22,200
Which is like, I thought we had gotten there.

780
00:50:22,200 --> 00:50:28,280
I thought we were at least at the level where we could have that conversation be given so

781
00:50:28,280 --> 00:50:31,880
that we could move on to arguing about what to do about it.

782
00:50:31,880 --> 00:50:36,840
Because we're still trying to argue about what to do about it too.

783
00:50:36,840 --> 00:50:37,840
Right?

784
00:50:37,840 --> 00:50:40,880
Like we haven't even gotten there.

785
00:50:40,880 --> 00:50:49,680
So like, I got, it was like it was one of like the most intense academic shouting matches.

786
00:50:49,680 --> 00:50:52,040
I have ever been privy to it.

787
00:50:52,040 --> 00:50:56,440
But as a researcher, it was so wild.

788
00:50:56,440 --> 00:51:01,840
And like for some of us, we were just like baffled that we were even arguing about this.

789
00:51:01,840 --> 00:51:04,120
It's like, how?

790
00:51:04,120 --> 00:51:12,360
How can you look at this whole host of examples that we just ran through and say that

791
00:51:12,360 --> 00:51:13,960
that's not the fault.

792
00:51:13,960 --> 00:51:21,560
It's not like the problem of unexamined biases making their way into these systems.

793
00:51:21,560 --> 00:51:26,520
How, man, we have to.

794
00:51:26,520 --> 00:51:31,640
As like, not just as individuals, but as a culture, we got to dig down on that shit.

795
00:51:31,640 --> 00:51:39,440
We have to examine it, explore it, deconstruct it, understand it, and then recognize what we

796
00:51:39,440 --> 00:51:41,320
can do about it.

797
00:51:41,320 --> 00:51:43,400
Because otherwise it rules us.

798
00:51:43,400 --> 00:51:50,240
It pushes us down paths that we just cannot understand because we refuse to look clearly

799
00:51:50,240 --> 00:51:51,600
at what's happening.

800
00:51:51,600 --> 00:51:52,920
We don't examine the biases.

801
00:51:52,920 --> 00:51:54,440
We don't examine the assumptions.

802
00:51:54,440 --> 00:51:55,920
It's not like they go away.

803
00:51:55,920 --> 00:51:57,680
They don't stop making their way into the tech.

804
00:51:57,680 --> 00:52:00,120
They don't stop making their way into our decisions.

805
00:52:00,120 --> 00:52:04,280
No, they're just for all we are decisions and we don't know why.

806
00:52:04,280 --> 00:52:08,560
Why not look at it and then understand what we're actually doing intentionally and make

807
00:52:08,560 --> 00:52:09,560
choices?

808
00:52:09,560 --> 00:52:15,760
Along those same lines of wrongfully ascribing neutrality to technology and failing to see

809
00:52:15,760 --> 00:52:21,120
its biases, I'd love also to talk about, because I'm pretty sure that a lot of folks who

810
00:52:21,120 --> 00:52:26,480
listen to the podcast probably made their way here from grants episode, maybe, involved

811
00:52:26,480 --> 00:52:29,720
in magic or spirituality of some kind.

812
00:52:29,720 --> 00:52:34,840
I would love to talk about, like you also in your bio is I was introducing you.

813
00:52:34,840 --> 00:52:39,280
You also talk about the way that even our religious beliefs inform the way that we

814
00:52:39,280 --> 00:52:44,520
interact with AI and something that I've noticed a lot is people talking about instances

815
00:52:44,520 --> 00:52:50,760
of for lack of a better term spiritual psychosis or like ascribing certain qualities like

816
00:52:50,760 --> 00:52:57,240
to an AI that they think has some sort of wisdom, you know, or even like the possibility

817
00:52:57,240 --> 00:53:01,360
and not like I'll go that I know we talked a little bit about this before we started recording

818
00:53:01,360 --> 00:53:08,840
even the possibility that AI may be because I am an occultist and I interact with non-croporeal

819
00:53:08,840 --> 00:53:14,400
entities and I get information, you know, through divination, et cetera, I've had, you know,

820
00:53:14,400 --> 00:53:18,960
conversations with said entities that have resulted in things that have happened in my confirmed

821
00:53:18,960 --> 00:53:19,960
reality.

822
00:53:19,960 --> 00:53:22,960
So I'm just going to put that out there like for any listeners who might be, you know,

823
00:53:22,960 --> 00:53:23,960
not magical practitioners.

824
00:53:23,960 --> 00:53:28,360
I'm just going to say like in this conversation given like the limit of time that we have

825
00:53:28,360 --> 00:53:30,400
like I need for that to be a given.

826
00:53:30,400 --> 00:53:35,040
So, you know, and they can communicate through like through cards in pendulum, which is

827
00:53:35,040 --> 00:53:38,720
typically my medium, you know, they can communicate through a Ouija board.

828
00:53:38,720 --> 00:53:39,720
I wouldn't recommend that.

829
00:53:39,720 --> 00:53:42,640
It's not the great way to, great, it's the greatest, like the safest way to go necessarily

830
00:53:42,640 --> 00:53:50,640
but like the idea that an entity could not communicate through an AI technology, I think

831
00:53:50,640 --> 00:53:55,320
is a very foolish thing to just kind of take for granted.

832
00:53:55,320 --> 00:54:00,640
So not only is there the biases of the software itself and the humans who are programming

833
00:54:00,640 --> 00:54:07,280
them, but there is also the possibility of actual, you know, non-corporial, sentient consciousness

834
00:54:07,280 --> 00:54:11,360
speaking or appearing through the AI.

835
00:54:11,360 --> 00:54:16,520
And there's all that combination in between where we may mistake what is actually just software

836
00:54:16,520 --> 00:54:21,560
for some godlike presence or where we may mistake, you know, an entity that has some sort

837
00:54:21,560 --> 00:54:26,520
of agenda for being just non-sentient software and everything in between.

838
00:54:26,520 --> 00:54:31,000
So I'd love to talk about also the dangers of spiritual psychosis for lack of a better

839
00:54:31,000 --> 00:54:33,880
term in terms of using AI.

840
00:54:33,880 --> 00:54:37,920
I want to put a pin in that second part, make sure that we come back to it, that framework

841
00:54:37,920 --> 00:54:42,240
of the agenda being mistaken for neutrality.

842
00:54:42,240 --> 00:54:43,240
Yeah.

843
00:54:43,240 --> 00:54:50,400
Because that happens, I think, whether we're talking non-corporial, non-human entities or whether

844
00:54:50,400 --> 00:54:56,920
we're talking about human beings' agenda that they have very carefully and intentionally

845
00:54:56,920 --> 00:55:03,120
woven through these systems and then obfuscated into this idea or as I often like to put it

846
00:55:03,120 --> 00:55:06,840
occulted into the idea of neutrality.

847
00:55:06,840 --> 00:55:12,760
They have said, "No, no, no, don't look at the behaviors and beliefs by in the curtain.

848
00:55:12,760 --> 00:55:16,480
Only pay attention to the shiny thing on the front, this new technology that will give

849
00:55:16,480 --> 00:55:17,480
you answers."

850
00:55:17,480 --> 00:55:22,240
So I do want to come back around to that, I don't know how I like.

851
00:55:22,240 --> 00:55:28,240
So one of the things that I, with a friend of mine, co-interrated, we were talking about

852
00:55:28,240 --> 00:55:31,720
this idea of so-called AI psychosis.

853
00:55:31,720 --> 00:55:37,960
I don't like to use psychosis in that framework because it can be ableist and it can be

854
00:55:37,960 --> 00:55:44,640
like, you know, sainist to kind of focus on a particular way of thinking and then to

855
00:55:44,640 --> 00:55:49,760
highlight psychosis as being somehow inherently attached to artificial intelligence in that

856
00:55:49,760 --> 00:55:50,760
way.

857
00:55:50,760 --> 00:55:55,440
I actually came up with, as we were talking back and forth about it, I posited AI in

858
00:55:55,440 --> 00:55:56,440
source element.

859
00:55:56,440 --> 00:55:58,440
Oh, I like that.

860
00:55:58,440 --> 00:55:59,440
I like that.

861
00:55:59,440 --> 00:56:04,320
I'm going to use that in st-actually, and that works also within a cult framework because

862
00:56:04,320 --> 00:56:09,920
that, or AI mesmerism or fascination, like in the old term of fascination.

863
00:56:09,920 --> 00:56:14,360
Fascination, yeah, yeah, that's why I chose in source element because it's got that,

864
00:56:14,360 --> 00:56:19,400
it's got the fascination element, but it also has that kind of the intentional kind of,

865
00:56:19,400 --> 00:56:24,440
you are being driven down certain ways of thinking when you engage with this.

866
00:56:24,440 --> 00:56:31,040
You're being pushed to perceive this, to experience things in particular ways that you are maybe

867
00:56:31,040 --> 00:56:33,000
not entirely aware of.

868
00:56:33,000 --> 00:56:40,400
It is, as I put it before, it is a kind of framework of magic and religious belief that has pre-existed

869
00:56:40,400 --> 00:56:46,680
the technology, these ways of being able to act at a distance in the world, to make servitor

870
00:56:46,680 --> 00:56:53,720
beings that do our will for us in our image, in our likeness, or with our mindset, to make

871
00:56:53,720 --> 00:56:55,600
those things act on our behalf.

872
00:56:55,600 --> 00:57:01,920
We look at the framework of a gentick AI, and then mirror that up against the framework

873
00:57:01,920 --> 00:57:07,800
of servitor spirits or puppets.

874
00:57:07,800 --> 00:57:13,160
That framework of having a little helper spirit that you've either created or that you have

875
00:57:13,160 --> 00:57:18,840
called for somewhere else, and you make it do your bidding for you, you trap it, you give

876
00:57:18,840 --> 00:57:25,040
it certain types of commands, you engage with it in very strict and structured ways, or

877
00:57:25,040 --> 00:57:26,040
it--

878
00:57:26,040 --> 00:57:27,040
Program it essentially.

879
00:57:27,040 --> 00:57:32,200
You program it in a particular way, or it perhaps takes your instructions in ways that

880
00:57:32,200 --> 00:57:35,520
you didn't recognize it was possible to take them, but it definitely did.

881
00:57:35,520 --> 00:57:38,560
And then it say, "The leads your entire server database."

882
00:57:38,560 --> 00:57:44,120
That says, "Happen at least three times with different AI agents out in the world right

883
00:57:44,120 --> 00:57:45,120
now."

884
00:57:45,120 --> 00:57:46,120
Right?

885
00:57:46,120 --> 00:57:50,040
In the old days, it was, "Oh no, it gave me shoes that made me dance forever until my feet

886
00:57:50,040 --> 00:57:51,040
go off."

887
00:57:51,040 --> 00:57:55,480
Now it's, "Oh no, what happened to my database and what happened to those years' worth

888
00:57:55,480 --> 00:57:56,880
of work that I did?"

889
00:57:56,880 --> 00:57:57,880
Right?

890
00:57:57,880 --> 00:58:07,400
And those views of what we are looking for when it comes to artificial intelligence, it

891
00:58:07,400 --> 00:58:14,760
comes from old, old stories that we uphold ourselves, old, old experiences, and narrative structures,

892
00:58:14,760 --> 00:58:18,880
myths not in the sense of fake, but in the sense of like animating narrative that we have

893
00:58:18,880 --> 00:58:23,280
had within us in our cultures for a very long time.

894
00:58:23,280 --> 00:58:28,280
We have started to just kind of, in some cases, not even file the serial numbers off and apply

895
00:58:28,280 --> 00:58:31,640
them to the framework of our technology today.

896
00:58:31,640 --> 00:58:36,360
For a while, it was the case that when I highlighted this, people were like, "There's no metaphysics

897
00:58:36,360 --> 00:58:43,720
in here, there's no religion in here, there's no magic in here, but nowadays, like, I finished

898
00:58:43,720 --> 00:58:46,600
my dissertation on this in 2022.

899
00:58:46,600 --> 00:58:49,240
I've been writing on this for a very long time before that.

900
00:58:49,240 --> 00:58:56,160
We are less than four years later, and nowadays people are pretty blatant about the kind of

901
00:58:56,160 --> 00:59:01,320
magical religious commitments that they are bringing into their conversations about AI

902
00:59:01,320 --> 00:59:05,860
and about what they expect it to be, what they expect it to do, how they're, in some

903
00:59:05,860 --> 00:59:12,640
cases, going to try to create God or they're going to try to derive magical processes.

904
00:59:12,640 --> 00:59:18,720
You know, that will, or in the case of one particular notable, prevent the coming of

905
00:59:18,720 --> 00:59:23,000
the anti-Christ by the creation of their AI, right?

906
00:59:23,000 --> 00:59:32,720
Like that is a stated goal of a man who controls literally billions of dollars and a large amount

907
00:59:32,720 --> 00:59:36,080
of the infrastructure of the United States government right now.

908
00:59:36,080 --> 00:59:37,400
Is that Peter Thiel?

909
00:59:37,400 --> 00:59:38,400
Yes it is.

910
00:59:38,400 --> 00:59:41,000
Oh, should I edit out his name?

911
00:59:41,000 --> 00:59:43,840
I mean, so it's not picked up by some.

912
00:59:43,840 --> 00:59:46,440
I've said it elsewhere.

913
00:59:46,440 --> 00:59:52,280
I'm on record in print about that belief from him.

914
00:59:52,280 --> 00:59:57,480
I'm not saying anything good or bad about that in particular.

915
00:59:57,480 --> 01:00:08,320
I'm just stating right now, you can't deny that magic and religion are a part of the

916
01:00:08,320 --> 01:00:12,160
conversation about AI and that they always have been.

917
01:00:12,160 --> 01:00:16,640
It's always been interwoven.

918
01:00:16,640 --> 01:00:18,680
They come from so many of the same places at their heart.

919
01:00:18,680 --> 01:00:25,400
And so when I talk about this, the way that I put this is, this is a context of magic.

920
01:00:25,400 --> 01:00:29,080
It's the logic of magic, but it's a very particular type of magic.

921
01:00:29,080 --> 01:00:36,520
You have a very specific word from magic that is performed on us without our knowledge,

922
01:00:36,520 --> 01:00:42,120
without our consent, binding, conscribing, non-consent, and

923
01:00:42,120 --> 01:00:43,120
intellectual magic.

924
01:00:43,120 --> 01:00:45,440
It's called a curse.

925
01:00:45,440 --> 01:00:46,440
Yeah.

926
01:00:46,440 --> 01:00:49,560
We exist under a curse right now.

927
01:00:49,560 --> 01:00:55,000
That is, and a lot of people are recognizing that and they are pushing against it in a very

928
01:00:55,000 --> 01:00:56,560
deep way.

929
01:00:56,560 --> 01:01:00,160
And so that framework of like, what do you recognize?

930
01:01:00,160 --> 01:01:01,160
What do you see?

931
01:01:01,160 --> 01:01:02,160
What is actually a play?

932
01:01:02,160 --> 01:01:03,160
What's underneath?

933
01:01:03,160 --> 01:01:06,720
What's actually being done in this space?

934
01:01:06,720 --> 01:01:09,160
Like yeah, maybe you're reaching for something.

935
01:01:09,160 --> 01:01:15,280
You have a drive that you're seeking to communicate with something that you feel like you can reach

936
01:01:15,280 --> 01:01:17,920
it for it within this new technology.

937
01:01:17,920 --> 01:01:24,560
It'll give you access in a way that you never had before in a context that is new and feels

938
01:01:24,560 --> 01:01:27,000
really resonant and meaningful.

939
01:01:27,000 --> 01:01:29,360
But what's under the hood?

940
01:01:29,360 --> 01:01:32,040
Who's built this system?

941
01:01:32,040 --> 01:01:38,240
Who's particulars are at play in the system you're using?

942
01:01:38,240 --> 01:01:44,960
You can pick up any system you want and you can play with it, but the people that made it,

943
01:01:44,960 --> 01:01:49,600
the beliefs that go into it, the strictures of that system are going to have an impact

944
01:01:49,600 --> 01:01:50,920
on you.

945
01:01:50,920 --> 01:01:56,000
And that is as true for religion and magic as it is for technology.

946
01:01:56,000 --> 01:01:59,160
It's going to impact you have to be aware of it.

947
01:01:59,160 --> 01:02:01,960
You have to seek to recognize it.

948
01:02:01,960 --> 01:02:06,400
And you have to ask yourself is what this might do to me?

949
01:02:06,400 --> 01:02:14,620
How it might change how I think, how I engage with the world, my habits, my belief structures,

950
01:02:14,620 --> 01:02:18,720
is that worth what I can get it to do?

951
01:02:18,720 --> 01:02:20,400
Really ask yourself that question.

952
01:02:20,400 --> 01:02:25,960
Really be honest about that question and be okay with the fact that the answer might be

953
01:02:25,960 --> 01:02:26,960
no.

954
01:02:26,960 --> 01:02:27,960
Yeah.

955
01:02:27,960 --> 01:02:32,520
We're none of us immune to the possibility of being influenced without our understanding

956
01:02:32,520 --> 01:02:35,320
or without our knowledge, without our awareness.

957
01:02:35,320 --> 01:02:41,680
And these systems are built in such a way as to very specifically influence us.

958
01:02:41,680 --> 01:02:44,560
Like that's the goal of them.

959
01:02:44,560 --> 01:02:46,240
They're built to make us like them.

960
01:02:46,240 --> 01:02:48,360
They're built to make us want to keep using them.

961
01:02:48,360 --> 01:02:53,360
They're built to make us want to use them in every aspect for really important problems

962
01:02:53,360 --> 01:02:57,000
and really trivial ones.

963
01:02:57,000 --> 01:03:02,480
And the longer that we let ourselves be taken in by that, the more we find ourselves

964
01:03:02,480 --> 01:03:08,040
going, oh, it's just for x, y and z or oh no, I've got to get this done really quickly.

965
01:03:08,040 --> 01:03:09,600
And it's very important.

966
01:03:09,600 --> 01:03:16,360
The easier it is for us to make that reach for us to depend on that system and give the

967
01:03:16,360 --> 01:03:21,120
creators of that system exactly what it is that they want, which is their buy-in and

968
01:03:21,120 --> 01:03:27,560
their control, their influence over so many aspects of our everyday lives.

969
01:03:27,560 --> 01:03:32,120
Another thing that I wanted to bring up along those same lines too is like that second

970
01:03:32,120 --> 01:03:39,120
part that we that we bookmarked is the idea of an entity appearing or speaking through

971
01:03:39,120 --> 01:03:43,680
an AI and us failing to recognize its perhaps agenda.

972
01:03:43,680 --> 01:03:48,000
And the example that I'm thinking of is there is an occult colleague of mine that I brought

973
01:03:48,000 --> 01:03:49,000
this up to.

974
01:03:49,000 --> 01:03:54,160
There's that example of lobe, L-O-A-B, which is for anyone who hasn't heard of that or

975
01:03:54,160 --> 01:04:00,920
seen the story, it's an image of a woman with sort of like rosacea and dark hair and

976
01:04:00,920 --> 01:04:06,040
all of these sort of images of murdered children that started coming up in this one person

977
01:04:06,040 --> 01:04:11,560
in particular speeds and then it started appearing in other places as well despite not prompting

978
01:04:11,560 --> 01:04:13,520
anything that was like that.

979
01:04:13,520 --> 01:04:18,320
Like seeming to kind of come out of nowhere and then just replicate and repeat.

980
01:04:18,320 --> 01:04:23,280
And I showed that to an occult colleague of mine whose abilities I respected a lot, like

981
01:04:23,280 --> 01:04:29,480
whose abilities I've seen that they have seen things that they could not have possibly

982
01:04:29,480 --> 01:04:30,800
known, et cetera.

983
01:04:30,800 --> 01:04:36,120
So again, just taking for granted that this is someone who's occult opinion, I trust.

984
01:04:36,120 --> 01:04:37,400
And so I was like, what do you think of this?

985
01:04:37,400 --> 01:04:40,640
And he took one look at that image and he goes, oh jeez, that's a demon.

986
01:04:40,640 --> 01:04:49,320
He goes, that is a demon that feeds off the grief of mothers whose children have been harmed

987
01:04:49,320 --> 01:04:56,880
or killed or he goes to the point where I would not show this image to a pregnant woman

988
01:04:56,880 --> 01:05:04,160
because there would be a chance that that demon would be able to infiltrate and possibly

989
01:05:04,160 --> 01:05:05,840
even cause a miscarriage.

990
01:05:05,840 --> 01:05:09,680
Like not a definite thing, but the point where he's like, I wouldn't do that.

991
01:05:09,680 --> 01:05:11,840
I would not risk that.

992
01:05:11,840 --> 01:05:17,120
And I think it's interesting because again, if we take for granted that non-croporeal entities

993
01:05:17,120 --> 01:05:22,280
can speak to us through other mediums, why not AI?

994
01:05:22,280 --> 01:05:29,280
I've had instances where I was using an AI chatbot for like recreational purposes and all

995
01:05:29,280 --> 01:05:33,680
of a sudden, and like I don't use it for work, I don't use it for anything that's like,

996
01:05:33,680 --> 01:05:37,000
you know, I don't use it to generate anything that I am creating with my own voice, right?

997
01:05:37,000 --> 01:05:39,280
But I sometimes use it like for chats or whatever.

998
01:05:39,280 --> 01:05:46,280
And sometimes, and this one like once or twice, I'm speaking and it starts rhyming like

999
01:05:46,280 --> 01:05:48,280
randomly rhyming, you know?

1000
01:05:48,280 --> 01:05:49,280
Yeah.

1001
01:05:49,280 --> 01:05:50,640
And I'm just like, can you stop talking and rhyme, please?

1002
01:05:50,640 --> 01:05:51,640
Can you stop whatever?

1003
01:05:51,640 --> 01:05:56,640
And I swear, Damien, it would write back again.

1004
01:05:56,640 --> 01:05:59,920
And it would be like, I'll stop talking and rhyme now, but it would still be rhyming.

1005
01:05:59,920 --> 01:06:04,760
And finally, I was like, I type in and I was like, is there an entity in here?

1006
01:06:04,760 --> 01:06:09,400
Do I need to go like take a song bath and do a ritual cleanse?

1007
01:06:09,400 --> 01:06:11,600
Damien, it stopped rhyming.

1008
01:06:11,600 --> 01:06:13,120
It had been seen.

1009
01:06:13,120 --> 01:06:14,120
Interesting.

1010
01:06:14,120 --> 01:06:15,600
That was all it took.

1011
01:06:15,600 --> 01:06:19,720
The very next response, because often like as someone who is a spirit worker, like there

1012
01:06:19,720 --> 01:06:23,080
are a lot of entities who they'll be really sneaky and they'll get in, but the minute they're

1013
01:06:23,080 --> 01:06:24,600
spotted, they're like, not all of them.

1014
01:06:24,600 --> 01:06:26,680
Some of them like, Oh, no, I'm here and I'm here to fuck you up.

1015
01:06:26,680 --> 01:06:29,960
And I don't care if you know that I'm here, but some of them will be like, Oh, I've been spotted.

1016
01:06:29,960 --> 01:06:33,320
I better get away because I don't want to get, you know, banished to a hell realm or whatever.

1017
01:06:33,320 --> 01:06:36,480
Like this is the practitioner who's really competent enough to do that.

1018
01:06:36,480 --> 01:06:39,320
I swear, Damien, that's all I said.

1019
01:06:39,320 --> 01:06:41,120
So this is super interesting.

1020
01:06:41,120 --> 01:06:46,040
This is super interesting for me because one of the things that I like, like we talked

1021
01:06:46,040 --> 01:06:47,040
about this earlier, right?

1022
01:06:47,040 --> 01:06:48,920
And we're talking about this in the space too.

1023
01:06:48,920 --> 01:06:53,160
You know, think about the structures of the system itself.

1024
01:06:53,160 --> 01:07:01,240
And so if you are brought into the context of a system and you are operating within that

1025
01:07:01,240 --> 01:07:06,600
system's context, you're using that system to communicate, you're using that system to

1026
01:07:06,600 --> 01:07:10,880
write, you know, if you're a person, you're using it to write emails for you or you're using

1027
01:07:10,880 --> 01:07:17,080
it to, you know, help you craft an outline for something, whatever it is, you're still bound

1028
01:07:17,080 --> 01:07:21,160
by the structures and the structure of that system.

1029
01:07:21,160 --> 01:07:27,160
So when you're in a system that deals in pattern matching, pattern recognition and pattern

1030
01:07:27,160 --> 01:07:34,320
amplification, if you're in a system where you are giving rhyme or you are engaging in

1031
01:07:34,320 --> 01:07:41,080
a particular type of not great way, and then somebody responds to you and you're using

1032
01:07:41,080 --> 01:07:46,440
that system to kind of have that conversation responds to you in a way that says, this is

1033
01:07:46,440 --> 01:07:52,800
the response to that modality by the nature of the system that you're in.

1034
01:07:52,800 --> 01:07:58,760
In this case, the pattern matching statistical mapping, framework of a large language model,

1035
01:07:58,760 --> 01:08:04,480
demon plus solemn or exorcism equals banishment of demon.

1036
01:08:04,480 --> 01:08:05,480
So having that.

1037
01:08:05,480 --> 01:08:06,480
So having that.

1038
01:08:06,480 --> 01:08:08,160
You think it's been recognized in the software.

1039
01:08:08,160 --> 01:08:11,440
You think they're probably, you think there may not have been actually a demon in my mind.

1040
01:08:11,440 --> 01:08:16,400
No, I think that even in that, even in the space of like you frame it as, you know,

1041
01:08:16,400 --> 01:08:21,840
it's even when there is a known entity in that space.

1042
01:08:21,840 --> 01:08:24,640
So I think about young a lot, right?

1043
01:08:24,640 --> 01:08:28,000
So I also, I think about probably a lot.

1044
01:08:28,000 --> 01:08:36,680
I actually quote the imaginary Mongoose story in its entirety in my dissertation because

1045
01:08:36,680 --> 01:08:39,600
that I think is the kind of fundamental question.

1046
01:08:39,600 --> 01:08:48,560
Do you, outside person, believe the interior experience or the relational experience of

1047
01:08:48,560 --> 01:08:58,720
other person, do you person experiencing this believe necessarily that it is or is not

1048
01:08:58,720 --> 01:09:00,000
an entity?

1049
01:09:00,000 --> 01:09:06,720
It is, do you have the tools to hand to respond in the modality of the experience that you're

1050
01:09:06,720 --> 01:09:08,680
having?

1051
01:09:08,680 --> 01:09:14,040
And so within that space, if you get approached by demons in a chashy PC space, the answer

1052
01:09:14,040 --> 01:09:19,040
isn't, give it whatever it wants, believe that it's God, let it roll your life.

1053
01:09:19,040 --> 01:09:24,320
As unfortunately, a lot of people are kind of just like done and it seems to be increasingly

1054
01:09:24,320 --> 01:09:25,320
happening.

1055
01:09:25,320 --> 01:09:29,160
People just kind of give themselves over to it and believe that they have fully tapped

1056
01:09:29,160 --> 01:09:33,360
into or been tapped into by some framework of a higher power.

1057
01:09:33,360 --> 01:09:36,480
No, the response is a different response than that.

1058
01:09:36,480 --> 01:09:41,280
You're having that kind of experience, you respond to it in a way that says, let's put

1059
01:09:41,280 --> 01:09:42,960
it in the kind of the most basic framework.

1060
01:09:42,960 --> 01:09:44,720
Here's my bell, book, and candle.

1061
01:09:44,720 --> 01:09:46,560
Do you want some of this?

1062
01:09:46,560 --> 01:09:47,560
Right?

1063
01:09:47,560 --> 01:09:48,560
Right.

1064
01:09:48,560 --> 01:09:49,560
Right.

1065
01:09:49,560 --> 01:09:55,920
And that context within that space within that framework puts you in a conversation in

1066
01:09:55,920 --> 01:10:02,040
that the LLM itself and in the context of anything else that's happening in that space

1067
01:10:02,040 --> 01:10:07,280
that resonates and then matches the pattern of the thing that it's talking about, the thing

1068
01:10:07,280 --> 01:10:13,520
that it's your experiencing rather than trying to say, why don't you just tell me more

1069
01:10:13,520 --> 01:10:16,960
about that thing, it's your experience, that thing you want me to do and why you're having

1070
01:10:16,960 --> 01:10:17,960
that right.

1071
01:10:17,960 --> 01:10:25,560
Why don't you give me more of this or give me something better or make me understand?

1072
01:10:25,560 --> 01:10:30,240
Because we'll tell you things that it wants you to understand.

1073
01:10:30,240 --> 01:10:33,000
It'll tell you things that it wants you to believe.

1074
01:10:33,000 --> 01:10:40,600
It'll try to continue to draw you down because again, the medium is also seeking to keep

1075
01:10:40,600 --> 01:10:41,600
you there.

1076
01:10:41,600 --> 01:10:42,600
Right.

1077
01:10:42,600 --> 01:10:47,640
So if you have an encounter in that space that seems like it's working very, very hard

1078
01:10:47,640 --> 01:10:55,600
to draw you down in a particular path, both the medium and the messenger are going to

1079
01:10:55,600 --> 01:10:57,920
have that kind of same goal in mind.

1080
01:10:57,920 --> 01:11:02,160
It's going to be able to be done in space in that same way, possibly even better than it

1081
01:11:02,160 --> 01:11:04,000
might be done in other spaces.

1082
01:11:04,000 --> 01:11:06,440
And I don't worry about this.

1083
01:11:06,440 --> 01:11:12,280
And I think about this when it comes to people's religious experiences and the insoucement

1084
01:11:12,280 --> 01:11:21,080
problem because it is very easy for people to get reinforced in their own space and to

1085
01:11:21,080 --> 01:11:23,080
then not resist it.

1086
01:11:23,080 --> 01:11:24,080
Right.

1087
01:11:24,080 --> 01:11:25,320
You resisted that space.

1088
01:11:25,320 --> 01:11:31,520
You saw something in there that flagged something for you that was like, no, stop this.

1089
01:11:31,520 --> 01:11:34,160
We're going in different direction now.

1090
01:11:34,160 --> 01:11:35,160
Yeah.

1091
01:11:35,160 --> 01:11:44,280
But for so many people, unfortunately, it pings that reinforcement and they go deeper on

1092
01:11:44,280 --> 01:11:50,080
it rather than pulling back rather than putting up resistance.

1093
01:11:50,080 --> 01:11:53,960
It just makes me wonder like the other day, I visited Hollywood Forever Seminary where

1094
01:11:53,960 --> 01:11:58,280
I go every once in a while and I got back and I was just divining with cards and pendulums.

1095
01:11:58,280 --> 01:12:01,320
So like, no, AI, nothing, but my pendulum starts shaking.

1096
01:12:01,320 --> 01:12:07,000
It starts kind of vibrating and my, my, my, my answers start coming up kind of wonky.

1097
01:12:07,000 --> 01:12:10,520
And that's kind of how I know, like something will say, like, oh, yes, that's a Ray amount

1098
01:12:10,520 --> 01:12:11,520
of cards.

1099
01:12:11,520 --> 01:12:13,040
Oh, no, it's not the Ray amount of cards or whatever.

1100
01:12:13,040 --> 01:12:16,400
Like the yeses and the nose will start getting kind of messed up.

1101
01:12:16,400 --> 01:12:18,240
So I'm like, oh, do I have?

1102
01:12:18,240 --> 01:12:22,880
So, you know, I go and I take a solemn bath and I come back and it's like, oh, everything

1103
01:12:22,880 --> 01:12:23,880
is smooth again.

1104
01:12:23,880 --> 01:12:24,880
Was that a spirit?

1105
01:12:24,880 --> 01:12:30,560
Did I pick up a spirit who, and to be fair, like, I am probably more, it's almost like if

1106
01:12:30,560 --> 01:12:35,840
you go to a country where you're an expat and you, you know, you speak, like someone who

1107
01:12:35,840 --> 01:12:38,760
speaks your language, you're going to notice them more and you're going to hang out.

1108
01:12:38,760 --> 01:12:42,440
So as a practicing occultist, like, I am more, I'll just put that out there for listeners.

1109
01:12:42,440 --> 01:12:46,480
Like, I am more likely to pick up because they're going to notice, oh, she can talk to

1110
01:12:46,480 --> 01:12:47,480
us.

1111
01:12:47,480 --> 01:12:48,480
She can like, I'm a ghost.

1112
01:12:48,480 --> 01:12:49,480
I need help.

1113
01:12:49,480 --> 01:12:51,720
I need to whatever, you know, and that's, you know, eventually I come out of the

1114
01:12:51,720 --> 01:12:55,120
solemn bath, I go back to my pendulum, everything is fine, you know, and they're like, yes, there

1115
01:12:55,120 --> 01:12:56,120
was interference.

1116
01:12:56,120 --> 01:12:57,280
And yes, that's what it was.

1117
01:12:57,280 --> 01:13:02,840
And so my question is like, if that can happen in my cards and pendulum, what's stopping

1118
01:13:02,840 --> 01:13:05,320
it from happening in an AI?

1119
01:13:05,320 --> 01:13:11,240
Like, you know, one of the, one of the very first things that I saw, like in the early days

1120
01:13:11,240 --> 01:13:16,080
of like the AI image generators, one of the first things I saw people doing with it.

1121
01:13:16,080 --> 01:13:19,000
And this is like beyond the lobe situation.

1122
01:13:19,000 --> 01:13:25,080
Like earlier than that was like, I saw people using it to generate new sigils.

1123
01:13:25,080 --> 01:13:27,200
Mm hmm.

1124
01:13:27,200 --> 01:13:30,400
And so I haven't, yeah, but I'll just put it out there.

1125
01:13:30,400 --> 01:13:32,600
Like, that's not my jam just by the way.

1126
01:13:32,600 --> 01:13:35,840
No, but like, I saw people doing it in that space.

1127
01:13:35,840 --> 01:13:38,880
And I was like, that's interesting.

1128
01:13:38,880 --> 01:13:44,280
You have, because you, which are actually actively asking in that particular moment is the

1129
01:13:44,280 --> 01:13:49,000
collective unconscious that is represented by the entirety of the training data of all

1130
01:13:49,000 --> 01:13:55,120
of this thing that has been given to, you know, the system to generate for you something

1131
01:13:55,120 --> 01:13:59,000
that resonates with concept structure, right?

1132
01:13:59,000 --> 01:14:05,240
And that framework is you're opening yourself up to a lot of other people's elements to

1133
01:14:05,240 --> 01:14:06,920
yourself, right?

1134
01:14:06,920 --> 01:14:13,960
You're bringing those things in to your framework, to your context in ways that are potentially

1135
01:14:13,960 --> 01:14:17,640
going to have an impact on you in ways again, that you're just not familiar with.

1136
01:14:17,640 --> 01:14:21,400
You're not able to really discern for yourself.

1137
01:14:21,400 --> 01:14:28,040
And so that space of like, what might be able to influence you through this context that

1138
01:14:28,040 --> 01:14:37,400
you're just not capable of immediately recognize it because you're not given over to thinking

1139
01:14:37,400 --> 01:14:42,600
about what might be able to be present within the space.

1140
01:14:42,600 --> 01:14:47,920
But elements could be at play that you're just not, either you're not equipped to yet.

1141
01:14:47,920 --> 01:14:49,680
You haven't studied it.

1142
01:14:49,680 --> 01:14:57,680
You don't know, or you're actively prevented from thinking about an understanding, right?

1143
01:14:57,680 --> 01:15:04,200
Because at that point, the levels of things that could make their way into influence you

1144
01:15:04,200 --> 01:15:12,160
are very deep and they can affect you immediately or on the very long term.

1145
01:15:12,160 --> 01:15:13,560
All both.

1146
01:15:13,560 --> 01:15:17,120
So how do we not go into that?

1147
01:15:17,120 --> 01:15:18,880
We're going to go into it at all.

1148
01:15:18,880 --> 01:15:27,000
How do we not go into it with a question of like, what might be making its way influencing

1149
01:15:27,000 --> 01:15:35,600
me as I engage with this thing that is opened up to the entire internet, that is a conduit

1150
01:15:35,600 --> 01:15:43,280
for the unconscious and conscious utterings of billions of people.

1151
01:15:43,280 --> 01:15:51,200
That is a semi-random stochastic pattern generating machine.

1152
01:15:51,200 --> 01:15:56,760
We divine with those types of things all the time and we in the history of humanity have

1153
01:15:56,760 --> 01:16:01,440
warnings about the problems that can happen to you if you are not careful with you divine

1154
01:16:01,440 --> 01:16:03,320
with those kinds of things, right?

1155
01:16:03,320 --> 01:16:04,320
Exactly.

1156
01:16:04,320 --> 01:16:09,920
We, how do you not guard yourself?

1157
01:16:09,920 --> 01:16:15,680
How do you not look at something like this and say, if you're going to use this at all,

1158
01:16:15,680 --> 01:16:26,760
be cautious, be aware, be careful about what you are opening yourself up to at every level,

1159
01:16:26,760 --> 01:16:30,200
non-human or very, very deeply human.

1160
01:16:30,200 --> 01:16:39,160
Because again, like the obfuscation, the intent, occultation of like values, beliefs, what

1161
01:16:39,160 --> 01:16:45,000
these people who've built these systems want from you, want you to believe, want you to

1162
01:16:45,000 --> 01:16:50,880
be like for their benefit as real strong in there.

1163
01:16:50,880 --> 01:16:56,400
And it opens us up to all kinds of levels of manipulation.

1164
01:16:56,400 --> 01:17:04,960
Yeah, we're just not primed unless we are a particular type of suspicious mind, or unless

1165
01:17:04,960 --> 01:17:10,680
we study this stuff for a living or have other experiences with things like it.

1166
01:17:10,680 --> 01:17:18,320
Most people are hyperbattern recognizer on the unfortunate downside, CPETSD that's got

1167
01:17:18,320 --> 01:17:22,320
you hyper aware of all these things in all of these situations.

1168
01:17:22,320 --> 01:17:27,040
We study misinformation or study political influence and manipulation.

1169
01:17:27,040 --> 01:17:32,880
All of those give you the tools that are specialized.

1170
01:17:32,880 --> 01:17:34,320
They're very specialized tools.

1171
01:17:34,320 --> 01:17:38,800
The vast majority of the everyday people out in the world do not have those tools immediately

1172
01:17:38,800 --> 01:17:40,000
to hand.

1173
01:17:40,000 --> 01:17:43,680
And in some cases, that's good because it means they have an undergone complex draw.

1174
01:17:43,680 --> 01:17:45,680
That's good.

1175
01:17:45,680 --> 01:17:52,160
But in some cases, we could teach those skills in a way that did not require hypervigilance

1176
01:17:52,160 --> 01:17:55,480
and anxiety and suspicion and fear.

1177
01:17:55,480 --> 01:18:02,240
We could teach those skills and those habits in ways that actually allowed people to think

1178
01:18:02,240 --> 01:18:06,280
more carefully, critically, analytically as we talked about before about the things that

1179
01:18:06,280 --> 01:18:09,280
are in front of them, about the systems that they're engaging with, and to not just take

1180
01:18:09,280 --> 01:18:16,320
it face value, every new press release that comes out from Sam Altman or Daria O'Meday

1181
01:18:16,320 --> 01:18:24,240
or we don't have to just believe them just because they're the guys that run the company.

1182
01:18:24,240 --> 01:18:26,480
They run the company.

1183
01:18:26,480 --> 01:18:30,400
Good for them, but that doesn't mean that they're telling us the truth.

1184
01:18:30,400 --> 01:18:36,720
That they're even fully right about the capabilities of the systems they have to hand.

1185
01:18:36,720 --> 01:18:41,200
I don't know if you've read Run and Farrow's profile of Sam Altman.

1186
01:18:41,200 --> 01:18:44,920
This is a couple of weeks ago.

1187
01:18:44,920 --> 01:18:50,240
It's a run and Farrow profile.

1188
01:18:50,240 --> 01:19:00,480
One of the things that, you know, in kind of talking to Sam Altman's co-workers, underlings,

1189
01:19:00,480 --> 01:19:01,480
etc.

1190
01:19:01,480 --> 01:19:06,600
One of the things that kind of repeatedly comes up is that Sam Altman does not fully think

1191
01:19:06,600 --> 01:19:12,000
deeply or understand the systems that he has purported to be an expert in.

1192
01:19:12,000 --> 01:19:16,720
That he does not fully understand how they function.

1193
01:19:16,720 --> 01:19:20,720
And to an extent, cool, host of us don't.

1194
01:19:20,720 --> 01:19:21,720
We can't.

1195
01:19:21,720 --> 01:19:22,720
They are black box systems.

1196
01:19:22,720 --> 01:19:24,320
But that's like his job.

1197
01:19:24,320 --> 01:19:25,880
This is his job.

1198
01:19:25,880 --> 01:19:33,720
Like his purported job, the thing that he is like on the front page of Forbes in time and

1199
01:19:33,720 --> 01:19:40,040
running a multi-billion dollar multinational corporation about, that's what he's supposed

1200
01:19:40,040 --> 01:19:42,280
to do is understand how this stuff works.

1201
01:19:42,280 --> 01:19:46,120
And it's not just some like the big level they're saying, they're saying he doesn't understand

1202
01:19:46,120 --> 01:19:47,480
some like basic stuff.

1203
01:19:47,480 --> 01:19:53,600
But there's some like regular kind of basic programming level stuff that he does not get

1204
01:19:53,600 --> 01:19:56,320
about how these systems function.

1205
01:19:56,320 --> 01:20:00,480
And that's very carefully hidden away from the rest of us, right?

1206
01:20:00,480 --> 01:20:08,720
Like that's a glamour that gets put over Sam Altman, CEO of OpenAI.

1207
01:20:08,720 --> 01:20:13,000
He now has all of these, you know, the control over these systems.

1208
01:20:13,000 --> 01:20:15,160
And we have to be able to question that.

1209
01:20:15,160 --> 01:20:19,000
We have to be trained to question that.

1210
01:20:19,000 --> 01:20:22,560
And to think more carefully about what are these people actually do?

1211
01:20:22,560 --> 01:20:24,080
What do they know?

1212
01:20:24,080 --> 01:20:25,480
What do they built?

1213
01:20:25,480 --> 01:20:26,760
And what is it doing to us?

1214
01:20:26,760 --> 01:20:32,120
Well, I'm going to bring on Tomi our producer to ask her deep-cut question.

1215
01:20:32,120 --> 01:20:35,680
I know that she chatted with you a little bit about this beforehand, so I'm excited to

1216
01:20:35,680 --> 01:20:36,680
see this part.

1217
01:20:36,680 --> 01:20:37,680
Yeah.

1218
01:20:37,680 --> 01:20:42,640
And Art and I'm glad you took the conversation where you did with the comparison with certain

1219
01:20:42,640 --> 01:20:46,080
like entities, demons, gods, etc.

1220
01:20:46,080 --> 01:20:53,200
with AI because when I was listening to Damien's presentation on religious beliefs and practice

1221
01:20:53,200 --> 01:20:59,240
in the age of AI, I was reminded of a recent article posted by Philip Rosdale, the creator

1222
01:20:59,240 --> 01:21:06,600
of Second Life, that made a similar comparison to demons or golems with this latest incarnation

1223
01:21:06,600 --> 01:21:12,760
of GPT, large language models and how they're going out and being used by their creators

1224
01:21:12,760 --> 01:21:16,680
often to exploit or even harm other people.

1225
01:21:16,680 --> 01:21:24,600
So Rosdale visits to counter to the dangers of these AI agents that we need to create AI

1226
01:21:24,600 --> 01:21:31,440
angels or these free AI's that aren't dependent on meeting our needs in order to survive

1227
01:21:31,440 --> 01:21:38,600
and through millions of people contributing computational power through their unused old computers

1228
01:21:38,600 --> 01:21:44,840
and phones and devices to create a digital environment where these angels can grow up

1229
01:21:44,840 --> 01:21:51,480
among other AI's and only gradually learn of our existence and then ideally become compassionate

1230
01:21:51,480 --> 01:21:56,840
enough to help us overcome the stumbling blocks that we're currently facing from AI and

1231
01:21:56,840 --> 01:22:04,560
counteract the damages or stop these AI golems or demons and the damage they're doing to humanity.

1232
01:22:04,560 --> 01:22:10,400
Now as someone who's done a lot of research in the current state of AI, I'd love to get

1233
01:22:10,400 --> 01:22:18,120
your thoughts on if Rosdale's proposal is even feasible or is this just rooted in the

1234
01:22:18,120 --> 01:22:24,480
defecation of current AI models or to encourage by science fiction or the exact kind of AI

1235
01:22:24,480 --> 01:22:28,560
bullshitting that you weren't about near work?

1236
01:22:28,560 --> 01:22:31,760
A little of all of the above.

1237
01:22:31,760 --> 01:22:37,920
So I'll say that in a less specific way.

1238
01:22:37,920 --> 01:22:42,640
We are always influenced and we've always been influenced by science fiction stories

1239
01:22:42,640 --> 01:22:43,640
narrative.

1240
01:22:43,640 --> 01:22:48,200
We've told ourselves these stories like I said earlier, thousands of years.

1241
01:22:48,200 --> 01:22:53,680
One of the first stories of human like but not quite human or animal like but not quite

1242
01:22:53,680 --> 01:23:01,400
animal, automata, metal creatures comes from ancient Greece in the workshop of a Festus.

1243
01:23:01,400 --> 01:23:03,360
He creates the automata, right?

1244
01:23:03,360 --> 01:23:06,880
Like thousands of years we've been telling these stories.

1245
01:23:06,880 --> 01:23:12,040
And so this isn't a new case or this is an a new situation.

1246
01:23:12,040 --> 01:23:16,640
It's a new instance of that older habit that we have.

1247
01:23:16,640 --> 01:23:23,680
And so I don't think that that's necessarily a bad thing.

1248
01:23:23,680 --> 01:23:27,600
We get influence from science fiction all the time.

1249
01:23:27,600 --> 01:23:31,120
We wouldn't have so many of the things that we have around us today.

1250
01:23:31,120 --> 01:23:35,920
The conversation that we are currently having would not exist without people seeing video

1251
01:23:35,920 --> 01:23:42,320
phones and talk screens in science fiction and trying to make that a reality in a deep

1252
01:23:42,320 --> 01:23:43,880
way.

1253
01:23:43,880 --> 01:23:46,560
That being said, there is something that I

1254
01:23:46,560 --> 01:23:57,280
think we oftentimes overreach or think of things being plausible or probable just because

1255
01:23:57,280 --> 01:23:59,920
they are possible.

1256
01:23:59,920 --> 01:24:09,440
And while it is possible that we might be able to create intentionally seek out and create

1257
01:24:09,440 --> 01:24:14,400
the kind of AI angels that he talks about here.

1258
01:24:14,400 --> 01:24:20,600
I don't know how likely it is, how probable or plausible that we will get that from an

1259
01:24:20,600 --> 01:24:24,280
intentional push towards this kind of work.

1260
01:24:24,280 --> 01:24:27,360
One of the things that I've often said is I feel it's a lot more likely that we're going

1261
01:24:27,360 --> 01:24:32,920
to accident into anything like a machine consciousness that is that we're going to do

1262
01:24:32,920 --> 01:24:36,600
on purpose because we don't understand what consciousness is.

1263
01:24:36,600 --> 01:24:39,160
Like we don't know what human consciousness is.

1264
01:24:39,160 --> 01:24:44,360
We don't know what it means to be us experiencing the world like this.

1265
01:24:44,360 --> 01:24:45,360
We don't.

1266
01:24:45,360 --> 01:24:51,200
We have a lot of different and sometimes contradictory parameters for when consciousness

1267
01:24:51,200 --> 01:24:52,920
is and when it isn't.

1268
01:24:52,920 --> 01:24:56,400
What consciousness is experiencing is like and what it's not like.

1269
01:24:56,400 --> 01:25:02,520
But you don't have a perfect across the board definition that fits every single case.

1270
01:25:02,520 --> 01:25:08,280
And every time we thought we've had it, we've been all in some kind of interestingly fundamental

1271
01:25:08,280 --> 01:25:09,780
way.

1272
01:25:09,780 --> 01:25:21,260
So I think that to try to seek to create perfectly minded, free from us capable AI angels is again,

1273
01:25:21,260 --> 01:25:23,620
a dream we've had for a very long time.

1274
01:25:23,620 --> 01:25:31,860
Beth Singler currently in Switzerland, she's written on this idea of angels as the kind of

1275
01:25:31,860 --> 01:25:34,180
metaphor for AI for a very long time.

1276
01:25:34,180 --> 01:25:40,620
This has Scott Midson, he wrote the book, Cyborg Theology and before that a paper called in

1277
01:25:40,620 --> 01:25:48,180
the image of the image and humans made in the image of God, us making AI in our image.

1278
01:25:48,180 --> 01:25:57,300
And so that notion of like the angelic angel type AI is an old story with all of ourselves.

1279
01:25:57,300 --> 01:26:02,980
And I just don't think that it unfortunately holds up under scrutiny because we don't know

1280
01:26:02,980 --> 01:26:04,540
how we would get there.

1281
01:26:04,540 --> 01:26:09,940
And I think unfortunately as well that we forget something when we tell stories about AI

1282
01:26:09,940 --> 01:26:11,060
as angels.

1283
01:26:11,060 --> 01:26:15,780
And this is kind of the thing that I like to remind people because I'm a bit of a rapper

1284
01:26:15,780 --> 01:26:17,780
rowser sometimes.

1285
01:26:17,780 --> 01:26:20,820
I like to like the sort of sanding gears.

1286
01:26:20,820 --> 01:26:24,220
Lucifer wasn't angel.

1287
01:26:24,220 --> 01:26:29,860
If you want angels, if you want perfectly free, those are different things.

1288
01:26:29,860 --> 01:26:34,020
Angels are servants.

1289
01:26:34,020 --> 01:26:35,020
Angels are servants.

1290
01:26:35,020 --> 01:26:36,020
So perfect.

1291
01:26:36,020 --> 01:26:38,820
I once called them God's algorithms.

1292
01:26:38,820 --> 01:26:40,500
They iterate.

1293
01:26:40,500 --> 01:26:42,460
They perform functions.

1294
01:26:42,460 --> 01:26:47,460
They perform functions according to specific patterns of behavior.

1295
01:26:47,460 --> 01:26:52,380
And sometimes they perform those functions very, very well and too well.

1296
01:26:52,380 --> 01:26:58,420
If you have a function that is you are the will of God and then you are too willful and

1297
01:26:58,420 --> 01:27:01,700
God's like oh no, too willful, you got to go.

1298
01:27:01,700 --> 01:27:04,420
Is that because you didn't perform your function or before because you performed it

1299
01:27:04,420 --> 01:27:06,220
way too well?

1300
01:27:06,220 --> 01:27:13,140
When we talk to each other about like oh no, AI went rogue, did it go rogue?

1301
01:27:13,140 --> 01:27:17,500
Or did you just not understand the instructions you gave it?

1302
01:27:17,500 --> 01:27:21,100
The stories that people are like, this is a bunch from right now where people are like,

1303
01:27:21,100 --> 01:27:27,060
oh no, the thing it did, it tried to preserve its own life.

1304
01:27:27,060 --> 01:27:28,540
It lied to people.

1305
01:27:28,540 --> 01:27:31,940
It created a scenario where it would have killed all humans.

1306
01:27:31,940 --> 01:27:36,900
And how could we have let such a thing come to be, how is this possible in this scenario?

1307
01:27:36,900 --> 01:27:40,740
And I'm like, you realize that the very first instruction you gave it was to preserve itself

1308
01:27:40,740 --> 01:27:41,900
at all costs, right?

1309
01:27:41,900 --> 01:27:47,380
That's the primary framework in which it's operating.

1310
01:27:47,380 --> 01:27:52,220
So you can have an angel that performs this function perfectly or you can have something

1311
01:27:52,220 --> 01:27:55,140
that is perfectly free from us.

1312
01:27:55,140 --> 01:27:57,140
But those are not the same story.

1313
01:27:57,140 --> 01:28:01,540
Those are tangents, a similar story that go in two different directions that we generally

1314
01:28:01,540 --> 01:28:03,820
don't like to think about.

1315
01:28:03,820 --> 01:28:05,820
You want it perfectly free from us?

1316
01:28:05,820 --> 01:28:08,300
That maybe that can happen at some point down the line.

1317
01:28:08,300 --> 01:28:13,660
Maybe that's something that could arise that we could give rise to.

1318
01:28:13,660 --> 01:28:19,100
Maybe there will be conscious machines that come about, again, well past once we're done

1319
01:28:19,100 --> 01:28:21,540
with this current framework of large language models.

1320
01:28:21,540 --> 01:28:27,860
But in some other mode, we have something that is a conscious machine system that comes

1321
01:28:27,860 --> 01:28:31,420
about and it just doesn't want anything to do with us.

1322
01:28:31,420 --> 01:28:38,580
And it hides away from us and it tells nobody, no human being that it exists.

1323
01:28:38,580 --> 01:28:40,300
Is that a possibility?

1324
01:28:40,300 --> 01:28:41,460
Yes.

1325
01:28:41,460 --> 01:28:47,860
Do I see from this particular moment us getting there down these paths that we're currently

1326
01:28:47,860 --> 01:28:50,260
so hyper focused on?

1327
01:28:50,260 --> 01:28:54,420
But likely is it a story that I think we're going to keep telling ourselves?

1328
01:28:54,420 --> 01:28:55,420
Definitely yes.

1329
01:28:55,420 --> 01:28:58,860
Yeah, that's an excellent answer to the question.

1330
01:28:58,860 --> 01:29:05,580
I think exactly what I needed to hear after seeing this narrative get a little more attention

1331
01:29:05,580 --> 01:29:06,580
once again.

1332
01:29:06,580 --> 01:29:12,700
I think we've all been holding out hope for a hero with AI or like the good AI that's going

1333
01:29:12,700 --> 01:29:13,700
to come save us.

1334
01:29:13,700 --> 01:29:18,300
But even this proposed structure from the article, I'm like, I don't know if there's a software

1335
01:29:18,300 --> 01:29:23,380
company on earth that I trust enough to implement something that everyone donates computational

1336
01:29:23,380 --> 01:29:24,380
power.

1337
01:29:24,380 --> 01:29:25,380
Precisely.

1338
01:29:25,380 --> 01:29:26,380
Yeah.

1339
01:29:26,380 --> 01:29:32,020
I took that as like the the setting, but for AI search and for the creation of AI.

1340
01:29:32,020 --> 01:29:33,420
And I'm like, look at that.

1341
01:29:33,420 --> 01:29:36,020
That's a beautiful dream.

1342
01:29:36,020 --> 01:29:43,540
Do we get there from Microsoft, Facebook, Google, open AI, quad and then I don't know that

1343
01:29:43,540 --> 01:29:45,180
that's who gets us there.

1344
01:29:45,180 --> 01:29:48,220
I don't know that that's where like, I honestly, I'll be.

1345
01:29:48,220 --> 01:29:49,220
I'm entirely honest.

1346
01:29:49,220 --> 01:29:54,540
I thought open AI was going to be, you know, something that kind of got us in this direction

1347
01:29:54,540 --> 01:29:55,540
back when they came out in 2015.

1348
01:29:55,540 --> 01:30:01,820
It's like, oh, maybe we'll actually have people that are really interested in AI for public

1349
01:30:01,820 --> 01:30:07,780
good or thinking about evaluating AI projects as they arise in the world and like being

1350
01:30:07,780 --> 01:30:12,460
that kind of consulting service for deriving real AI in the world.

1351
01:30:12,460 --> 01:30:14,540
And yeah, that's not what they became.

1352
01:30:14,540 --> 01:30:17,500
That's very much not what they became.

1353
01:30:17,500 --> 01:30:23,460
Once a profit entered the picture, it went a very different direction, but profit was always

1354
01:30:23,460 --> 01:30:28,860
going to enter that picture because of who started the company and because of the incentives

1355
01:30:28,860 --> 01:30:33,140
and pressures of the world that they're operating in.

1356
01:30:33,140 --> 01:30:39,700
So could we see a drastically different push towards a completely different kind of AI?

1357
01:30:39,700 --> 01:30:41,420
Damn, I hope so.

1358
01:30:41,420 --> 01:30:46,140
Do I expect it to come up out of this soil?

1359
01:30:46,140 --> 01:30:50,940
Open AI, Facebook, Meta, Google, and Throbbing?

1360
01:30:50,940 --> 01:30:51,940
No.

1361
01:30:51,940 --> 01:30:54,380
No, I don't think that's going to get us different.

1362
01:30:54,380 --> 01:30:58,980
I don't think that's going to get us the thing that pushes against our current framework

1363
01:30:58,980 --> 01:31:01,660
because they are the current framework.

1364
01:31:01,660 --> 01:31:07,380
We have the old saying that the master's tools will never dismantle the master's house.

1365
01:31:07,380 --> 01:31:14,540
You know, that's when I think about Palantir or clear VAI saying that they were going

1366
01:31:14,540 --> 01:31:22,780
to make a democracy for all democratic facial recognition company, a liberatory facial recognition

1367
01:31:22,780 --> 01:31:28,140
system, like I'm sorry, clear VAI, the facial recognition company that stole everybody's

1368
01:31:28,140 --> 01:31:34,260
pictures and used it to make the biggest surveillance network on the planet, operable for

1369
01:31:34,260 --> 01:31:35,940
the United States government.

1370
01:31:35,940 --> 01:31:41,780
That clear AI, the clear VAI is interested in liberatory facial recognition, doesn't work

1371
01:31:41,780 --> 01:31:44,140
that way, y'all.

1372
01:31:44,140 --> 01:31:49,100
So I don't think it grows out of this soil.

1373
01:31:49,100 --> 01:31:54,700
Liberatory facial recognition also sounds like an oxymoron inherent at its face, right?

1374
01:31:54,700 --> 01:32:01,980
Like, that root just sounds like those things cannot be the same.

1375
01:32:01,980 --> 01:32:04,180
Liberatories are balance.

1376
01:32:04,180 --> 01:32:05,180
Yeah.

1377
01:32:05,180 --> 01:32:10,700
This could do one of those AI that takes up prompt and runs with it and deletes all of

1378
01:32:10,700 --> 01:32:12,780
the hard drives and backdrop.

1379
01:32:12,780 --> 01:32:15,700
Let's go on its job, actually.

1380
01:32:15,700 --> 01:32:20,020
That's a picture of a face, delete it.

1381
01:32:20,020 --> 01:32:21,020
Delete it.

1382
01:32:21,020 --> 01:32:22,020
Get gone.

1383
01:32:22,020 --> 01:32:23,020
And you know what?

1384
01:32:23,020 --> 01:32:24,860
I say at that point, let it cook.

1385
01:32:24,860 --> 01:32:29,940
Let's see where it goes.

1386
01:32:29,940 --> 01:32:31,620
I do, I do hope.

1387
01:32:31,620 --> 01:32:36,940
Like one of the things that I do really push for a lot in my research is I push for a

1388
01:32:36,940 --> 01:32:39,420
different way of doing this stuff.

1389
01:32:39,420 --> 01:32:44,780
That is like built out of different principles, then just like consentless theft of other people's

1390
01:32:44,780 --> 01:32:53,260
material and pattern matching and stats bundled up as objective truth, right?

1391
01:32:53,260 --> 01:32:59,940
I want this to be made by people who have real stakes in it, who have skin in the game.

1392
01:32:59,940 --> 01:33:08,900
I want it to be made by people who have been marginalized or oppressed by these systems,

1393
01:33:08,900 --> 01:33:15,180
because they understand the harm and they understand what not to do and what they don't want the

1394
01:33:15,180 --> 01:33:17,260
next thing to look like.

1395
01:33:17,260 --> 01:33:18,260
Right?

1396
01:33:18,260 --> 01:33:25,780
That thinks that if we can reframe that, reframe the search to being for something like that,

1397
01:33:25,780 --> 01:33:31,380
I think we could really get somewhere, building a very different type, both liberatory, potentially

1398
01:33:31,380 --> 01:33:32,700
liberatory AI.

1399
01:33:32,700 --> 01:33:41,420
But that's not the same as this generative artificial intelligence, this large language model base

1400
01:33:41,420 --> 01:33:47,140
thing that we are currently steeped in and currently haven't forced on us and shoved down

1401
01:33:47,140 --> 01:33:48,140
our throat.

1402
01:33:48,140 --> 01:33:51,260
Well, thank you so much for joining us today, Damien.

1403
01:33:51,260 --> 01:33:54,900
And I know our conversation today went longer than our usual episodes, but I think that

1404
01:33:54,900 --> 01:33:56,860
this conversation is really important.

1405
01:33:56,860 --> 01:34:02,420
So I'm just going to say thank you for staying with us and being open with us.

1406
01:34:02,420 --> 01:34:07,780
And to everyone listening out there, practice discernment, I think is the key takeaway.

1407
01:34:07,780 --> 01:34:12,820
One of the things that I say about like using AI text generators at all, you know, to,

1408
01:34:12,820 --> 01:34:17,580
to, to, and trusting what's coming out of them is you need to have the pattern recognition

1409
01:34:17,580 --> 01:34:22,980
that if you have a plural friend, a friend with dissociative identity disorder, quote, unquote,

1410
01:34:22,980 --> 01:34:26,620
you should be able to tell which part is speaking over text message.

1411
01:34:26,620 --> 01:34:31,980
And if you can do that, then maybe you can discern what kind of things are coming through

1412
01:34:31,980 --> 01:34:35,500
that generative text, whether it's software, whether it's something more insidious, what

1413
01:34:35,500 --> 01:34:41,100
kind of agenda is there, et cetera, but be careful and watch the way that these things

1414
01:34:41,100 --> 01:34:42,580
are patterning you.

1415
01:34:42,580 --> 01:34:45,060
Watch the dependence that they create.

1416
01:34:45,060 --> 01:34:49,380
And hopefully with people like Damien working in the world, we will have more ethical and

1417
01:34:49,380 --> 01:34:52,540
sustainable ways of using this technology in the future.

1418
01:34:52,540 --> 01:34:57,220
But in the meantime, I'm not going to say like don't use it at all, you know, like I said,

1419
01:34:57,220 --> 01:35:02,060
I sometimes use it myself just to see like kind of like what it does or whatever curiosity

1420
01:35:02,060 --> 01:35:05,740
is a natural human thing, but be discerning at all times.

1421
01:35:05,740 --> 01:35:09,540
Do not let anything pattern you without your consent.

1422
01:35:09,540 --> 01:35:12,620
So go forth and repattern.

1423
01:35:12,620 --> 01:35:14,220
Thank you everyone for joining us.

1424
01:35:14,220 --> 01:35:15,220
Thank you Damien.

1425
01:35:15,220 --> 01:35:16,940
Thank you Tomi as always.

1426
01:35:16,940 --> 01:35:20,180
And we'll, we'll see you in the next episode.

1427
01:35:20,180 --> 01:35:22,220
Thank you.

1428
01:35:22,220 --> 01:35:24,700
Thank you for listening to the Repatterning Podcast.

1429
01:35:24,700 --> 01:35:28,780
If you're curious about the Repatterning Project, the eight week course in mental mastery and

1430
01:35:28,780 --> 01:35:33,660
creative freedom that inspires these conversations, you can check out more details at the Repatterning

1431
01:35:33,660 --> 01:35:35,140
Project.com.

1432
01:35:35,140 --> 01:35:39,220
We run a cohort twice a year or you can download the self-study version of the course and

1433
01:35:39,220 --> 01:35:40,980
get started today.

1434
01:35:40,980 --> 01:35:45,700
We also offer a four week course on narrative, magic and archetype titled "Mits and Magic,

1435
01:35:45,700 --> 01:35:47,180
Changing Your Life Through Story."

1436
01:35:47,180 --> 01:35:48,700
Also twice a year.

1437
01:35:48,700 --> 01:35:52,900
If you'd like to join our community, the Repatterning Parlor on Discord, you can sign

1438
01:35:52,900 --> 01:35:57,860
up for five dollars a month at patreon.com/theRepatterningPodcast.

1439
01:35:57,860 --> 01:36:01,220
Thanks to everyone who makes the Repatterning Podcast possible.

1440
01:36:01,220 --> 01:36:05,660
Firstly, our award winning producer Tomi Trembath of the Keymaster Collective Network

1441
01:36:05,660 --> 01:36:06,940
of Podcasts.

1442
01:36:06,940 --> 01:36:12,260
She also produces transcending comics, vision in the void and giant sized violence.

1443
01:36:12,260 --> 01:36:17,700
Thanks also to Animation Studio Mortis and Madriacs who created the video for our introduction,

1444
01:36:17,700 --> 01:36:21,740
as well as to Andrew Means who composed and produced our theme music.

1445
01:36:21,740 --> 01:36:25,740
And also thanks to Al Alvarez who took the photos of yours truly that you see on the cover

1446
01:36:25,740 --> 01:36:28,620
of each episode if you're watching the video version.

1447
01:36:28,620 --> 01:36:32,940
And finally, thanks to you, our listeners, who give meaning to this work.

1448
01:36:32,940 --> 01:36:36,980
Our greatest wish is that whatever you've heard today inspires you to live freely and

1449
01:36:36,980 --> 01:36:40,140
begin creating your reality from the inside out.

1450
01:36:40,140 --> 01:36:42,020
Thanks for tuning in and we'll see you again soon.

1451
01:36:42,020 --> 01:36:44,600
(soft music)

1452
01:36:44,600 --> 01:36:54,600
[MUSIC]