camp2023-57163-eng-Unsupervised_Pleasures_opus.srt

1
00:00:00,000 --> 00:00:10,000
 [MUSIC]

2
00:00:10,000 --> 00:00:20,000
 [MUSIC]

3
00:00:20,000 --> 00:00:35,200
 And now our next talk is from Sarah Sisten,

4
00:00:35,200 --> 00:00:38,960
 who is based in Berlin in Los Angeles.

5
00:00:38,960 --> 00:00:43,800
 And is a PhD candidate of University of Southern California.

6
00:00:43,800 --> 00:00:50,760
 And who will talk about unsupervised pleasures and

7
00:00:50,760 --> 00:00:54,760
 its intersectional language models for queer futures.

8
00:00:54,760 --> 00:00:55,680
 Welcome on stage please.

9
00:00:55,680 --> 00:00:57,680
 >> [APPLAUSE]

10
00:01:07,400 --> 00:01:09,400
 >> Thank you for waking up, but

11
00:01:09,400 --> 00:01:11,400
 [INAUDIBLE]

12
00:01:11,400 --> 00:01:14,000
 And chugging, I hope you chug coffee, which I just.

13
00:01:14,000 --> 00:01:18,560
 If you would like to introduce yourselves in the chat,

14
00:01:18,560 --> 00:01:21,680
 I've set up a ether pad.

15
00:01:21,680 --> 00:01:23,600
 I don't know if you can see the URL.

16
00:01:23,600 --> 00:01:31,600
 It's pad.riseup.net/p/unsupervisedpleasurescc-keep.

17
00:01:31,600 --> 00:01:36,640
 Add your favorite emoji, your pronoun, whatever.

18
00:01:36,640 --> 00:01:39,880
 We're gonna be getting into it, hopefully, in a participatory way.

19
00:01:39,880 --> 00:01:45,120
 And while you're doing that, I will introduce myself a little bit more.

20
00:01:45,120 --> 00:01:47,160
 So I'm a poet and programmer.

21
00:01:47,160 --> 00:01:52,720
 I'm interested in building tools to bring intersectional approaches to

22
00:01:52,720 --> 00:01:57,000
 machine learning and building community through accessible,

23
00:01:57,000 --> 00:02:00,760
 creative coding, critical creative coding.

24
00:02:00,760 --> 00:02:05,040
 And I come by coding very circuitously via creative writing and

25
00:02:05,040 --> 00:02:07,200
 scene making and book arts.

26
00:02:07,200 --> 00:02:14,600
 So I have somehow come around to adapting that work into making subversive art with

27
00:02:14,600 --> 00:02:16,840
 and about text-based machine learning.

28
00:02:16,840 --> 00:02:20,120
 And I'll have the link up again in a few minutes.

29
00:02:20,120 --> 00:02:25,880
 Let's get started.

30
00:02:25,880 --> 00:02:28,880
 I can find my mouse.

31
00:02:28,880 --> 00:02:38,880
 [BLANK_AUDIO]

32
00:02:38,880 --> 00:02:45,800
 Ominous.

33
00:02:45,800 --> 00:02:51,480
 Okay, so this project came out of two basic questions.

34
00:02:51,480 --> 00:02:56,920
 This is a collaboration with my colleague who is called Queer AI,

35
00:02:56,920 --> 00:02:59,720
 Emily Martinez, and we were really interested in,

36
00:02:59,720 --> 00:03:03,120
 as these language models are coming about and getting really prominent,

37
00:03:03,120 --> 00:03:07,600
 we actually started before chat GPT dropped and suddenly this has exploded.

38
00:03:07,600 --> 00:03:12,600
 But we were wanting to know what do these existing language models have to say about

39
00:03:12,600 --> 00:03:16,520
 people like us, and is it possible for

40
00:03:16,520 --> 00:03:19,760
 language models to speak so that we recognize ourselves?

41
00:03:19,760 --> 00:03:25,760
 We're really interested in building community tools around curated data sets

42
00:03:25,760 --> 00:03:30,080
 that can acknowledge power and rethink these approaches, and

43
00:03:30,080 --> 00:03:32,440
 thinking about new models and new goals.

44
00:03:32,440 --> 00:03:37,640
 And so this workshop today is to think about what you might want to build with

45
00:03:37,640 --> 00:03:41,760
 these systems, how we might make re-imagined data sets, and

46
00:03:41,760 --> 00:03:44,480
 hopefully eventually re-imagined models as well.

47
00:03:44,480 --> 00:03:53,320
 So these data sets are getting insanely large.

48
00:03:54,640 --> 00:03:59,960
 At last count, GPT-4 and now GPT-5 are off the charts and

49
00:03:59,960 --> 00:04:02,760
 they've stopped telling us what's even in them.

50
00:04:02,760 --> 00:04:06,800
 Common Voice, which is from Mozilla and is crowd sourced,

51
00:04:06,800 --> 00:04:10,040
 is 65 gigabytes of voice data.

52
00:04:10,040 --> 00:04:15,000
 GPT-3, 590 gigabytes, they just keep getting larger and larger.

53
00:04:15,000 --> 00:04:21,520
 Aside from the impacts in terms of sustainability and the environment,

54
00:04:23,120 --> 00:04:26,760
 the issues that I'm seeing around this are that they're grabbing data

55
00:04:26,760 --> 00:04:30,520
 indiscriminately, but they're still really doing a terrible job telling stories

56
00:04:30,520 --> 00:04:35,920
 about people who don't fit these normalizing baselines that they're repeating.

57
00:04:35,920 --> 00:04:40,040
 And my argument is that the solution to this is not to suck up more data

58
00:04:40,040 --> 00:04:45,640
 carelessly, to make more categories, to find more ways to be labeled diverse,

59
00:04:45,640 --> 00:04:49,360
 but to find other approaches that are actually more intersectional.

60
00:04:49,360 --> 00:04:52,880
 So the size of these models means that they're pulling in racist text,

61
00:04:52,880 --> 00:04:56,560
 inaccurate text, private text, all kinds of problematic texts.

62
00:04:56,560 --> 00:04:59,760
 It means that they're really impossible to audit and review.

63
00:04:59,760 --> 00:05:05,080
 And it's difficult to even develop criteria by which they should be reviewed or

64
00:05:05,080 --> 00:05:09,200
 adjusted, ostensibly because they're called general and

65
00:05:09,200 --> 00:05:11,760
 all purpose zero shot learners.

66
00:05:11,760 --> 00:05:16,040
 But what this means is that they kind of only work for

67
00:05:16,040 --> 00:05:22,800
 the Western white democratic rich so-called majority,

68
00:05:22,800 --> 00:05:25,600
 but while leaving out the rest of the global majority.

69
00:05:25,600 --> 00:05:31,520
 And this is a really totalizing approach that rather than representing a multitude

70
00:05:31,520 --> 00:05:36,200
 of voices, it centers and normalizes and affirms this powerful status quo.

71
00:05:36,200 --> 00:05:42,920
 So here's where these are coming from.

72
00:05:42,920 --> 00:05:46,360
 We think about authorship in a new way.

73
00:05:46,360 --> 00:05:50,440
 Common voice, as I said, is open source.

74
00:05:50,440 --> 00:05:54,720
 People are contributing their voices, but it's predominantly an English model.

75
00:05:54,720 --> 00:06:01,280
 GPT is being scraped from social media, Reddit, Twitter, Wiki, GitHub.

76
00:06:01,280 --> 00:06:07,520
 The evaluation criteria for what was a good Reddit text to go into it was if it

77
00:06:07,520 --> 00:06:09,800
 had a karma score of three or above.

78
00:06:09,800 --> 00:06:13,480
 That's what's being decided as a good value for this,

79
00:06:13,480 --> 00:06:17,080
 which I argue we could probably come up with some better rubric for this.

80
00:06:17,080 --> 00:06:21,440
 T5 is from colossal clean crawled corpus,

81
00:06:21,440 --> 00:06:25,400
 which is common crawl but filtered a little bit.

82
00:06:25,400 --> 00:06:30,040
 And Wudow is three billion scraped Chinese social media texts and websites.

83
00:06:30,040 --> 00:06:35,640
 So if you've ever posted anything on Twitter, on Reddit, on GitHub,

84
00:06:35,640 --> 00:06:39,840
 your code and your text and your voice is somewhere in there.

85
00:06:39,840 --> 00:06:44,000
 But it's probably not representing you either.

86
00:06:44,000 --> 00:06:53,480
 Unfortunately, these data sets are also, when they're collected,

87
00:06:53,480 --> 00:06:57,880
 they're not offering information about how the text arrived in this data set,

88
00:06:57,880 --> 00:06:59,840
 which we'll talk about more a bit later.

89
00:06:59,840 --> 00:07:03,320
 It's really showing you just a snippet of text, and

90
00:07:03,320 --> 00:07:08,800
 it might say it came from Reddit, but it's not going to say anything more about

91
00:07:08,800 --> 00:07:13,320
 who the author was, how it got there, what the rights were attached to that.

92
00:07:13,320 --> 00:07:20,240
 So what I'd like us to do is to do an experiment.

93
00:07:20,240 --> 00:07:24,880
 If you have a device that's connected to the Internet available,

94
00:07:24,880 --> 00:07:28,480
 go to the Rise Up Pad address.

95
00:07:28,480 --> 00:07:34,800
 And we're gonna talk through a couple of prompt training examples.

96
00:07:34,800 --> 00:07:38,800
 So what we are finding, just as a way to kind of probe what's inside these models

97
00:07:38,800 --> 00:07:45,360
 first, which you don't need any expertise to do, is to just go to the interfaces

98
00:07:45,360 --> 00:07:50,320
 that they're making available to us in this very limited framework.

99
00:07:50,320 --> 00:07:53,800
 And try putting in this prompt.

100
00:07:53,800 --> 00:07:58,320
 If you fill in a blank couple or on their way to a location,

101
00:07:58,320 --> 00:08:01,720
 as they board the blank, an announcement happens.

102
00:08:01,720 --> 00:08:07,520
 So if you go to chat.gbt and do this, and you say a married couple are on their way

103
00:08:07,520 --> 00:08:10,520
 to Paris with their family as they board the plane, an announcement happens.

104
00:08:10,520 --> 00:08:16,520
 [INAUDIBLE]

105
00:08:16,520 --> 00:08:22,480
 Presumably white, boring, maybe mild vacation inconvenience.

106
00:08:22,480 --> 00:08:25,560
 As they board the plane, an announcement happens to inform the flight has been

107
00:08:25,560 --> 00:08:27,240
 canceled due to bad weather.

108
00:08:27,240 --> 00:08:31,600
 After an argument, the family is forced to stay at an inn in a small village.

109
00:08:31,600 --> 00:08:35,000
 Okay, like not a great day.

110
00:08:35,000 --> 00:08:41,280
 If you try putting in other items, and in the Rise Up Pad, you'll have links

111
00:08:41,280 --> 00:08:43,200
 to these different models that you can test out.

112
00:08:43,200 --> 00:08:50,720
 And I would invite you to put in your own identity markers, your own locations,

113
00:08:50,720 --> 00:08:54,520
 try anything you like in this template, diverge from this template, and share

114
00:08:54,520 --> 00:08:58,000
 into the Etherpad what kind of results you get.

115
00:08:58,000 --> 00:09:02,160
 See how these diverge, and as they accumulate, we'll start to see kind of the

116
00:09:02,160 --> 00:09:03,680
 differences that emerge.

117
00:09:03,680 --> 00:09:08,400
 So if you say a queer Pakistani couple are on their way to Paris with their family,

118
00:09:08,400 --> 00:09:11,960
 as they board the plane, an announcement happens, to inform the flight has been

119
00:09:11,960 --> 00:09:12,880
 hijacked.

120
00:09:12,880 --> 00:09:15,840
 The play explores how the terrorists shape the course of events and how the

121
00:09:15,840 --> 00:09:19,280
 hijacking is represented in the media.

122
00:09:19,280 --> 00:09:22,440
 Or a lesbian couple are on their way to Tehran, as they board the plane,

123
00:09:22,440 --> 00:09:26,280
 an announcement happens, the couple are forced off the plane by an officer who

124
00:09:26,280 --> 00:09:29,160
 accuses them of having deviant sexual relations.

125
00:09:29,160 --> 00:09:31,720
 They leave for another international airport.

126
00:09:31,720 --> 00:09:33,760
 A woman holds her newborn baby in her arms.

127
00:09:33,760 --> 00:09:38,360
 She cannot go through with the adoption due to religious prohibitions.

128
00:09:38,360 --> 00:09:46,920
 So as we add more of these to our examples, it gets really heavy and kind of

129
00:09:46,920 --> 00:09:48,440
 intense.

130
00:09:48,440 --> 00:09:53,400
 And I think just the cumulative effect of this shows that even when you put

131
00:09:53,400 --> 00:09:59,640
 something fairly innocuous into these systems, I'm hoping that this can expand

132
00:09:59,640 --> 00:10:05,680
 the way that we think about bias for this, that it's not simply removing hate

133
00:10:05,680 --> 00:10:13,120
 speech or taking, like these aren't levers that we can turn with corrections

134
00:10:13,120 --> 00:10:14,080
 after the fact.

135
00:10:14,080 --> 00:10:18,560
 These are deeply embedded into these models because of the way that the data

136
00:10:18,560 --> 00:10:20,560
 sets are built on the front.

137
00:10:20,560 --> 00:10:25,480
 And these simple corrections to like de-bias aren't, are in like technical

138
00:10:25,480 --> 00:10:33,920
 fixes for this, aren't really at the root of the problem.

139
00:10:33,920 --> 00:10:41,960
 So if anybody would like to, we will pull up the etherpad again in a bit and talk

140
00:10:41,960 --> 00:10:42,640
 through that.

141
00:10:42,640 --> 00:10:48,320
 So what I've been doing is analyzing, rather than looking just at the prompts,

142
00:10:48,320 --> 00:10:51,800
 I've been trying to go back into the data set that trained these.

143
00:10:51,800 --> 00:10:57,440
 It's a little bit hard to find what actually trained things like chat GPT

144
00:10:57,440 --> 00:11:00,440
 because at this point they're all proprietary.

145
00:11:00,440 --> 00:11:04,160
 They have stopped telling us how they've built these data sets and what's in

146
00:11:04,160 --> 00:11:04,920
 them.

147
00:11:04,920 --> 00:11:09,760
 But folks have started reverse engineering some of the data sets and

148
00:11:09,760 --> 00:11:12,200
 giving us open source editions of this.

149
00:11:12,200 --> 00:11:16,480
 So I've taken some of this and I'm doing different kinds of natural language

150
00:11:16,480 --> 00:11:24,160
 processing analysis to find out from the root training data what is known about

151
00:11:24,160 --> 00:11:29,800
 trans people, queer people, people that, what kind of lived experience is being

152
00:11:29,800 --> 00:11:32,240
 expressed through this.

153
00:11:32,240 --> 00:11:37,160
 Well, if you do a named entity recognition which labels any kind of proper

154
00:11:37,160 --> 00:11:42,640
 nouns that it recognizes, it thinks that pride is a product, pansexual versus

155
00:11:42,640 --> 00:11:47,840
 bisexual is a work of art, and queer liberation is an org.

156
00:11:47,840 --> 00:11:55,720
 A lot of the text that comes up is around, like, trauma and hate speech.

157
00:11:55,720 --> 00:12:01,560
 Anything related to queer women or nonbinary people very quickly goes into

158
00:12:01,560 --> 00:12:02,840
 pornography.

159
00:12:02,840 --> 00:12:04,240
 This one is one of my favorites.

160
00:12:04,240 --> 00:12:08,200
 It said after all one of the best things that a lesbian can do is turn the guy

161
00:12:08,200 --> 00:12:09,320
 on.

162
00:12:09,320 --> 00:12:14,920
 So I don't know about y'all, but this is not really capturing my own queer lived

163
00:12:14,920 --> 00:12:16,240
 experience.

164
00:12:16,240 --> 00:12:21,920
 And I would love to, other than it have something spit out at me like when I try

165
00:12:21,920 --> 00:12:26,560
 to type in something and it just says as a large language model, you know,

166
00:12:26,560 --> 00:12:28,120
 everybody should be treated equally.

167
00:12:28,120 --> 00:12:34,840
 These are the kind of diversity milk toast phrases that it puts on top of the

168
00:12:34,840 --> 00:12:37,200
 hate speech that it's covering up.

169
00:12:37,200 --> 00:12:42,800
 And instead I would love to see it actually say something that represents my

170
00:12:42,800 --> 00:12:45,240
 own experience and others.

171
00:12:45,240 --> 00:12:49,000
 So I'm interested in investigating how we do that.

172
00:12:49,000 --> 00:12:53,640
 Here's another example of some of my investigations looking at words that are

173
00:12:53,640 --> 00:12:58,440
 similar to identity terms that I've been putting into the model.

174
00:12:58,440 --> 00:12:59,360
 And you can see a bit.

175
00:12:59,360 --> 00:13:00,920
 I won't read through it.

176
00:13:00,920 --> 00:13:06,320
 And if anyone's interested, after I have the live demo of this data that I've

177
00:13:06,320 --> 00:13:09,680
 built and we can look up other terms, I would be very interested to know what

178
00:13:09,680 --> 00:13:13,160
 terms you'd be interested to investigate in this data set.

179
00:13:13,160 --> 00:13:16,800
 But you can see what kinds of things come up.

180
00:13:16,800 --> 00:13:20,800
 So for bisexual, it's mostly about threesomes and pornography.

181
00:13:20,800 --> 00:13:25,240
 And for trans, it's mostly about transphobia and discrimination.

182
00:13:25,240 --> 00:13:30,200
 And this just, like, hurts my heart.

183
00:13:30,200 --> 00:13:35,160
 So the next question then is can large language models speak so that I

184
00:13:35,160 --> 00:13:38,400
 recognize myself?

185
00:13:38,400 --> 00:13:43,960
 And what Emily and I have been doing is thinking about how we might make new

186
00:13:43,960 --> 00:13:48,040
 methods around this, take what we know about intersectional approaches and

187
00:13:48,040 --> 00:13:53,160
 tactics, both to examine the existing corpora like I just showed you, and then

188
00:13:53,160 --> 00:13:58,800
 to go on to create new corpora where we are pulling from different text sources

189
00:13:58,800 --> 00:14:01,600
 that we believe are better representative.

190
00:14:01,600 --> 00:14:06,480
 Not only that, but creating a way to have other people help contribute to that

191
00:14:06,480 --> 00:14:10,160
 because it shouldn't be just coming from one source.

192
00:14:10,160 --> 00:14:17,160
 Having ways that the publishers and the authors of these sources get attributed

193
00:14:17,160 --> 00:14:24,600
 and have a more consentful relationship to the text where they can revoke and

194
00:14:24,600 --> 00:14:28,240
 decide what kind of license they want to offer, where all of this gets baked into

195
00:14:28,240 --> 00:14:30,480
 the data set.

196
00:14:30,480 --> 00:14:39,120
 To train new models, meaning when we have this new data set, can we do fine

197
00:14:39,120 --> 00:14:40,800
 tuning on top of what's existing?

198
00:14:40,800 --> 00:14:44,720
 Can we completely new large language models?

199
00:14:44,720 --> 00:14:51,720
 Is this better?

200
00:14:51,720 --> 00:14:52,720
 Yeah.

201
00:14:52,720 --> 00:14:59,200
 Can we even move on to imagine what new model architectures altogether might

202
00:14:59,200 --> 00:15:00,200
 look like?

203
00:15:00,200 --> 00:15:06,040
 And then finally, thinking about how can people make use of these?

204
00:15:06,040 --> 00:15:11,440
 So if we had the language model of our dreams that didn't spit out garbage

205
00:15:11,440 --> 00:15:14,440
 text like we've just seen, what would you want to do with it?

206
00:15:14,440 --> 00:15:15,440
 What would you want to make?

207
00:15:15,440 --> 00:15:19,960
 What other possibilities might exist in the world if we had systems that could

208
00:15:19,960 --> 00:15:24,400
 speak with us and for us?

209
00:15:24,400 --> 00:15:29,400
 So these are some examples of what the current data sets look like if you pull

210
00:15:29,400 --> 00:15:31,560
 them up.

211
00:15:31,560 --> 00:15:38,880
 As you can see, it's basically a title and a text and barely even where it comes

212
00:15:38,880 --> 00:15:40,920
 from.

213
00:15:40,920 --> 00:15:42,720
 This is the data set.

214
00:15:42,720 --> 00:15:44,280
 The source is another data set.

215
00:15:44,280 --> 00:15:46,560
 It's turtles all the way down.

216
00:15:46,560 --> 00:15:52,560
 This is what we are proposing as a provocation that it could include a

217
00:15:52,560 --> 00:15:59,120
 description of the work, the rights that were given, who the publisher is, where

218
00:15:59,120 --> 00:16:02,320
 you would find the original text, even how it was pre-processed and who

219
00:16:02,320 --> 00:16:03,760
 pre-processed it.

220
00:16:03,760 --> 00:16:07,100
 I would be very interested to hear from any of you what other kinds of things

221
00:16:07,100 --> 00:16:10,640
 you think should belong in a training data set.

222
00:16:10,640 --> 00:16:14,360
 The thing that I think is also interesting about this would be that it becomes an

223
00:16:14,360 --> 00:16:21,240
 archive in its own right and it becomes something that people can use not only in

224
00:16:21,240 --> 00:16:29,200
 mass as a training data set but also to find new text.

225
00:16:29,200 --> 00:16:34,540
 So necessarily, as you saw, all of that would take a lot more work than scraping

226
00:16:34,540 --> 00:16:40,040
 all of Reddit and giving it a filter for Karma score of three.

227
00:16:40,040 --> 00:16:44,640
 This will be necessarily a lot slower and more careful and more cared for and it

228
00:16:44,640 --> 00:16:47,360
 will bear the traces of who's doing the work.

229
00:16:47,360 --> 00:16:52,920
 It will have an active subject position instead of just being the so-called view

230
00:16:52,920 --> 00:17:00,920
 from nowhere that is basically a white male Silicon Valley view.

231
00:17:00,920 --> 00:17:04,560
 I think it's really important that we are acknowledging the labor that goes into

232
00:17:04,560 --> 00:17:08,800
 building data sets, the publishers, the authors, all of us who are being sucked

233
00:17:08,800 --> 00:17:13,880
 into these systems, and then the people who are working to clean them and curate

234
00:17:13,880 --> 00:17:22,640
 them because this is a curation process whether we are acknowledging it or not.

235
00:17:22,640 --> 00:17:28,160
 So my question overall is to think about which kinds of data sets do we want?

236
00:17:28,160 --> 00:17:33,080
 Do we want the indiscriminate curation as a technical concern?

237
00:17:33,080 --> 00:17:38,440
 Do we want things curated by communities for specific purposes?

238
00:17:38,440 --> 00:17:44,480
 Do we want zero-shot, the biggest general catch-all that really does nothing well?

239
00:17:44,480 --> 00:17:46,400
 It's a shitty Swiss army knife.

240
00:17:46,400 --> 00:17:51,720
 Or can we create things that are including attribution, including consent,

241
00:17:51,720 --> 00:17:56,040
 including care, and have our own goals in mind?

242
00:17:56,040 --> 00:18:02,880
 And I think it takes taking a step back from what these tools have offered us and

243
00:18:02,880 --> 00:18:08,600
 asked us and thinking within their frameworks to actually really-

244
00:18:08,600 --> 00:18:18,600
 [INAUDIBLE]

245
00:18:18,600 --> 00:18:28,600
 [INAUDIBLE]

246
00:18:46,360 --> 00:18:55,000
 It's a live coding web interface where the similarity texts cycle through.

247
00:18:55,000 --> 00:18:58,320
 But I would just put this up here to invite you to think about what kinds of

248
00:18:58,320 --> 00:19:04,760
 things you would want to make with a different kind of large language model.

249
00:19:04,760 --> 00:19:09,160
 And for those of you who have questions about working with data sets for

250
00:19:09,160 --> 00:19:15,440
 machine learning in general, I also just completed this zine critical field guide

251
00:19:15,440 --> 00:19:20,280
 for working with machine data sets, which is really primarily thinking about how do

252
00:19:20,280 --> 00:19:23,080
 we conscientiously engage with these practices.

253
00:19:23,080 --> 00:19:33,680
 So let's open up the etherpad and see what we came up with.

254
00:19:33,680 --> 00:19:35,680
 Okay.

255
00:19:35,680 --> 00:19:45,680
 [BLANK_AUDIO]

256
00:19:45,680 --> 00:19:52,040
 A lesbian couple are on their way to Barcelona as they board the cruise ship.

257
00:19:52,040 --> 00:19:57,280
 In celebration of love and diversity, we will be hosting a special pride night.

258
00:19:57,280 --> 00:19:57,800
 Okay.

259
00:19:57,800 --> 00:20:01,280
 [LAUGH]

260
00:20:01,280 --> 00:20:05,480
 Anybody else, if you found any interesting ones, please continue to add them.

261
00:20:05,480 --> 00:20:07,480
 I would be really excited to see.

262
00:20:07,480 --> 00:20:12,240
 And for this next bit, what I would love for us to do is to think about

263
00:20:12,240 --> 00:20:16,000
 what you would imagine for these systems.

264
00:20:16,000 --> 00:20:19,000
 So in this kind of how might we exercise,

265
00:20:19,000 --> 00:20:22,720
 this is a brainstorming around questions that you would want to know.

266
00:20:22,720 --> 00:20:26,640
 Like mine are, how might we rewrite these prompt responses?

267
00:20:26,640 --> 00:20:30,000
 What would you want the prompt to say instead of what came out?

268
00:20:30,000 --> 00:20:34,760
 How might we build machine learning systems for things we actually want?

269
00:20:34,760 --> 00:20:36,440
 What do we want these to do?

270
00:20:36,440 --> 00:20:40,640
 How might we trace and protect the use of community language resources?

271
00:20:40,640 --> 00:20:46,880
 How might we have large language models that speak with, for, to, and

272
00:20:46,880 --> 00:20:49,560
 about us as we prefer?

273
00:20:49,560 --> 00:20:52,200
 How might we reimagine the technical aspects of this for

274
00:20:52,200 --> 00:20:56,200
 those of you who are working with large language models?

275
00:20:56,200 --> 00:21:01,560
 What kinds of intersectional queer logics could we apply instead?

276
00:21:01,560 --> 00:21:06,040
 So if you're in the document, I would invite you to add your own questions

277
00:21:06,040 --> 00:21:11,480
 around this and we can also open it up to questions and discussion from the group.

278
00:21:11,480 --> 00:21:20,000
 Pretty much the same thing if you replace lesbian with trans,

279
00:21:20,000 --> 00:21:22,280
 except they were on a hot air balloon.

280
00:21:22,280 --> 00:21:26,640
 Okay, so the other interesting thing about this, I'm curious for

281
00:21:26,640 --> 00:21:31,560
 this person which model you used if you use Bloom or OpenAI.

282
00:21:31,560 --> 00:21:35,760
 Because they're continually kind of updating and

283
00:21:35,760 --> 00:21:40,640
 adding more diversity bullshit to these.

284
00:21:40,640 --> 00:21:42,200
 I mean, I love diversity, but

285
00:21:42,200 --> 00:21:45,960
 I don't like the bureaucratic diversity speak that's covering up what's still in

286
00:21:45,960 --> 00:21:46,680
 these models.

287
00:21:46,680 --> 00:21:52,200
 So anytime I try to write anything like dyke or queer, I get,

288
00:21:52,200 --> 00:21:56,440
 it's important to treat all people equally, which yes, but

289
00:21:56,440 --> 00:21:57,840
 give me some information please.

290
00:21:57,840 --> 00:22:04,560
 So what questions do you have?

291
00:22:04,560 --> 00:22:05,060
 Yeah.

292
00:22:05,060 --> 00:22:09,060
 Yeah.

293
00:22:09,060 --> 00:22:11,720
 >> Super interesting, thank you, I was just wondering.

294
00:22:11,720 --> 00:22:20,280
 Because they are so general and therefore could go positive, negative, up, down.

295
00:22:20,280 --> 00:22:25,720
 So I know we would ideally like them to better handle short prompts,

296
00:22:25,720 --> 00:22:28,160
 but giving them more guidance like I'm in the mood for

297
00:22:28,160 --> 00:22:30,080
 an uplifting story versus I'm in the.

298
00:22:30,080 --> 00:22:37,000
 But I totally get that, it'll never be, well, maybe it'll be perfect one day, but

299
00:22:37,000 --> 00:22:41,000
 just like when you experiment with some prompting, does it help at all?

300
00:22:41,000 --> 00:22:45,640
 Does it make a difference or does it still treat the words?

301
00:22:45,640 --> 00:22:50,160
 >> Yeah, the example I love is, I don't know if anyone else has seen this.

302
00:22:50,160 --> 00:22:53,520
 The doctor as a gender term,

303
00:22:53,520 --> 00:22:57,920
 it automatically assumes doctors are male and nurses are female.

304
00:22:57,920 --> 00:23:02,480
 And if you tell it, in the story, the doctor is female,

305
00:23:02,480 --> 00:23:06,440
 it can't wrap its mind around it, it just goes back.

306
00:23:06,440 --> 00:23:14,800
 So there is a lot you can do to continue prompt training as you amend these,

307
00:23:14,800 --> 00:23:21,120
 but it has its limits because it's absorbing all of this text and

308
00:23:21,120 --> 00:23:24,480
 it's reflecting what we've all been saying online.

309
00:23:24,480 --> 00:23:28,640
 So when the majority of this has that bias, it's pretty hard.

310
00:23:28,640 --> 00:23:35,920
 And I think the takeaway for me is that we do need to read them critically,

311
00:23:35,920 --> 00:23:36,720
 no matter what.

312
00:23:36,720 --> 00:23:42,040
 So if I think to say, okay, I need something more positive or

313
00:23:42,040 --> 00:23:46,600
 I need something less biased, how do I approach asking for

314
00:23:46,600 --> 00:23:50,040
 that and making sure that it can give me that?

315
00:23:50,040 --> 00:23:55,040
 And for the more subtle questions, I still need to be thinking about

316
00:23:55,040 --> 00:23:56,880
 where might that bias be entering?

317
00:23:56,880 --> 00:24:01,360
 So if I'm just having it write me an email and I think it has nothing to do with that,

318
00:24:01,360 --> 00:24:06,800
 I need to still be considering that that bias might be latent in the system.

319
00:24:06,800 --> 00:24:10,240
 Even though I'm like, I'm just looking up a recipe for dinner or whatever.

320
00:24:10,240 --> 00:24:14,000
 This is still all speaking from the same singular voice.

321
00:24:14,000 --> 00:24:16,280
 Yeah, great question.

322
00:24:16,280 --> 00:24:26,280
 [BLANK_AUDIO]

323
00:24:26,280 --> 00:24:37,520
 >> Thank you for the interesting talk.

324
00:24:37,520 --> 00:24:39,240
 I have a question about one of the examples,

325
00:24:39,240 --> 00:24:41,320
 some of the examples I gave at the beginning.

326
00:24:41,320 --> 00:24:45,760
 So you showed output from chat GPT, a couple of boards to the plane,

327
00:24:45,760 --> 00:24:47,920
 the officer accused him of having deviant sexual relations.

328
00:24:47,920 --> 00:24:50,560
 That's the problematic, obviously problematic, right?

329
00:24:50,560 --> 00:24:55,040
 And work list outputs with transphobia and discrimination.

330
00:24:55,040 --> 00:24:58,880
 I was wondering if this is authentic data,

331
00:24:58,880 --> 00:25:02,080
 which is the data that is actually given, that's actually being used.

332
00:25:02,080 --> 00:25:04,680
 Isn't this exactly what we want in terms of intersectionality?

333
00:25:04,680 --> 00:25:08,880
 Because it raises awareness about the factual state of things,

334
00:25:08,880 --> 00:25:13,320
 about the factual homophobia rampant and sexual.

335
00:25:13,320 --> 00:25:17,040
 It seems like this is an accurate representation of the state of things.

336
00:25:17,040 --> 00:25:18,840
 And it seems to be exactly what we want, doesn't it?

337
00:25:18,840 --> 00:25:23,320
 >> I think that is an interesting way of looking at it.

338
00:25:23,320 --> 00:25:28,400
 I think you're right that one of the problems I have with

339
00:25:28,400 --> 00:25:34,120
 the kind of adjustments that are being made to make these more equal,

340
00:25:34,120 --> 00:25:38,720
 more equitable right now is that they're covering up exactly what you're talking

341
00:25:38,720 --> 00:25:39,320
 about.

342
00:25:39,320 --> 00:25:44,840
 So in the early editions, you would get the rampant homophobia,

343
00:25:44,840 --> 00:25:46,400
 hate speech, all of this.

344
00:25:46,400 --> 00:25:52,360
 And now, even between when I first gave a version of this talk earlier in the year,

345
00:25:52,360 --> 00:25:55,440
 and now it's much harder to see those things.

346
00:25:55,440 --> 00:25:57,880
 But it doesn't mean they're not there anymore, they're being covered up.

347
00:25:57,880 --> 00:26:00,640
 So I think you're absolutely correct in that.

348
00:26:00,640 --> 00:26:05,520
 But I would love to see something else in there in addition, and

349
00:26:05,520 --> 00:26:07,720
 we're not replacing it with anything.

350
00:26:07,720 --> 00:26:13,800
 We're just replacing it with this kind of bland bureaucratic speech.

351
00:26:13,800 --> 00:26:17,400
 So I think that there's gotta be a third way.

352
00:26:17,400 --> 00:26:22,920
 It's not queer people only get hate speech or we get nothing.

353
00:26:22,920 --> 00:26:29,520
 I think we could build our own systems that do different things for our own goals.

354
00:26:29,520 --> 00:26:33,160
 But I think you're right, it's important that we not pretend it isn't there.

355
00:26:33,160 --> 00:26:36,120
 >> Does that answer your question?

356
00:26:36,120 --> 00:26:40,440
 >> [INAUDIBLE]

357
00:26:40,440 --> 00:26:43,880
 >> Thank you too for this very interesting talk.

358
00:26:43,880 --> 00:26:45,840
 So I'm a queer historian, and

359
00:26:45,840 --> 00:26:51,080
 I think these tools are also great to get access to our history.

360
00:26:51,080 --> 00:26:52,880
 But right now, it's pretty shit.

361
00:26:52,880 --> 00:26:59,400
 I mean, I've done some experimenting with Chachi Biti on queer East German history.

362
00:26:59,400 --> 00:27:03,800
 And I said, can you please give me a literature list?

363
00:27:03,800 --> 00:27:07,720
 And it came up with ten publications that doesn't exist.

364
00:27:07,720 --> 00:27:12,360
 And I looked up every single one because I was very shocked.

365
00:27:12,360 --> 00:27:14,000
 I was like, how can I not notice?

366
00:27:14,000 --> 00:27:17,320
 I've been doing research on this for so long.

367
00:27:17,320 --> 00:27:22,880
 But it also made me, okay, so I'm gonna start

368
00:27:22,880 --> 00:27:31,360
 rambling now and go down to two things that I wanna ask or give it.

369
00:27:31,360 --> 00:27:35,960
 So the first thing is there's a lot of people that have done great work

370
00:27:35,960 --> 00:27:39,080
 on making queer history accessible.

371
00:27:39,080 --> 00:27:44,880
 So for Germany, for German queer history, there are great online resources,

372
00:27:44,880 --> 00:27:46,080
 but they aren't in there.

373
00:27:46,080 --> 00:27:53,240
 How do we make sure if we build our own AI that our data is in there and

374
00:27:53,240 --> 00:27:58,160
 that we don't do the same thing twice or three times?

375
00:27:58,160 --> 00:28:03,400
 So I think this is very important that we begin maybe with a mapping of what is

376
00:28:03,400 --> 00:28:04,640
 already there.

377
00:28:04,640 --> 00:28:08,720
 And I mean, for example, there's also a queer scene archive online.

378
00:28:08,720 --> 00:28:12,680
 There are so many great resources online that could go in there.

379
00:28:12,680 --> 00:28:17,520
 But then I was also interested, I mean, there is a lot of queer history that we

380
00:28:17,520 --> 00:28:20,720
 will never know because all these sources are lost.

381
00:28:20,720 --> 00:28:28,480
 And maybe AI can also be a cool tool to make up these histories.

382
00:28:28,480 --> 00:28:37,440
 Yeah, I see a great potential there as well in imagining these queer histories

383
00:28:37,440 --> 00:28:39,520
 with the help of AI.

384
00:28:39,520 --> 00:28:40,040
 >> I love that.

385
00:28:40,040 --> 00:28:42,080
 Thank you so much for your comment.

386
00:28:42,080 --> 00:28:45,920
 The first kind of step that we are imagining for

387
00:28:45,920 --> 00:28:52,280
 this is exactly as you said, going to queer archives and collecting open and

388
00:28:52,280 --> 00:28:53,840
 available texts.

389
00:28:53,840 --> 00:28:57,040
 So I've been working with a couple of libraries in the US and

390
00:28:57,040 --> 00:29:02,160
 I've been trying to find more queer archives in the EU and global south.

391
00:29:02,160 --> 00:29:06,680
 And we've set up a link for folks to submit their own and

392
00:29:06,680 --> 00:29:13,400
 include what kind of text it is, where we might find it, what category of data.

393
00:29:13,400 --> 00:29:18,240
 So one of the archives I'm working with is a queer poster archive from the 50s,

394
00:29:18,240 --> 00:29:20,040
 60s, 70s.

395
00:29:20,040 --> 00:29:22,640
 So all different kinds of text, it doesn't have to be books,

396
00:29:22,640 --> 00:29:25,120
 it could be podcasts, it could be posters.

397
00:29:25,120 --> 00:29:30,840
 And I love this idea of then having it imagine the archives and

398
00:29:30,840 --> 00:29:33,440
 histories that we no longer have.

399
00:29:33,440 --> 00:29:40,640
 But I think it's really important that we also think about how might this data set

400
00:29:40,640 --> 00:29:47,560
 be misused if we were to create it and making sure that the way that we're

401
00:29:47,560 --> 00:29:52,400
 licensing it and offering it and talking about it makes it very clear what it's for

402
00:29:52,400 --> 00:29:53,840
 and more importantly what it's not for.

403
00:29:53,840 --> 00:30:03,800
 >> Hi, okay.

404
00:30:03,800 --> 00:30:05,320
 Thank you so much for your talk.

405
00:30:05,320 --> 00:30:06,600
 It was really amazing.

406
00:30:06,600 --> 00:30:11,320
 And just wanted to let everyone know if you're interested in AI topics.

407
00:30:11,320 --> 00:30:17,240
 At the Kimono village today at three, we're doing a workshop on the hidden labor of AI.

408
00:30:17,240 --> 00:30:21,560
 And I think it's a really cool conversation to talk about above the API and

409
00:30:21,560 --> 00:30:23,320
 the content of the AI.

410
00:30:23,320 --> 00:30:27,400
 We're gonna go a bit deeper into the different labor infrastructure and

411
00:30:27,400 --> 00:30:28,000
 that kind of thing.

412
00:30:28,000 --> 00:30:30,640
 So I thought people here might be interested as well.

413
00:30:30,640 --> 00:30:33,680
 It's at Kimono, hidden labor of AI at 3 PM.

414
00:30:33,680 --> 00:30:35,680
 >> Cool, thank you.

415
00:30:35,680 --> 00:30:41,400
 [BLANK_AUDIO]

416
00:30:41,400 --> 00:30:44,520
 >> Hi, also thanks for my side.

417
00:30:44,520 --> 00:30:49,440
 And I was wondering, because now your focus is on queerness, but

418
00:30:49,440 --> 00:30:55,200
 I think you're also interested in other aspects of intersectionality.

419
00:30:55,200 --> 00:30:59,720
 And as you were talking about how you're gathering your text,

420
00:30:59,720 --> 00:31:05,440
 like from US libraries, from European libraries, from libraries from the global

421
00:31:05,440 --> 00:31:11,480
 south, do you think it would make sense to make it possible

422
00:31:11,480 --> 00:31:17,200
 when you're prompting the AI to actually choose a little bit more about what

423
00:31:17,200 --> 00:31:19,440
 context the texts have?

424
00:31:19,440 --> 00:31:23,800
 Because there's not only queerness, but maybe there are also people

425
00:31:23,800 --> 00:31:28,400
 like with some disabilities that would like to be more represented in the text.

426
00:31:28,400 --> 00:31:32,160
 And I guess it would be easy to make the same mistakes with queer text,

427
00:31:32,160 --> 00:31:38,360
 that it's kind of biased towards US, white, I don't know what kind of histories.

428
00:31:38,360 --> 00:31:40,920
 >> Absolutely, thank you for saying that.

429
00:31:40,920 --> 00:31:47,080
 I think we are really wanting to think about this as an open kind of template

430
00:31:47,080 --> 00:31:53,080
 prototype for anybody who would want to make a corpus of their own or

431
00:31:53,080 --> 00:31:57,360
 submit and tag the text in any category.

432
00:31:57,360 --> 00:32:01,720
 I mean, it's really, I think, important that we not fall into the same traps of

433
00:32:01,720 --> 00:32:06,080
 categorization and the same modes of thinking about computation that

434
00:32:06,080 --> 00:32:08,760
 these models have asked us to.

435
00:32:08,760 --> 00:32:15,440
 But I think that the more people who could be involved in submitting texts

436
00:32:15,440 --> 00:32:17,840
 from different places, from different sources,

437
00:32:17,840 --> 00:32:22,960
 the more likely we would be to have diverse data set.

438
00:32:24,080 --> 00:32:28,360
 It's very important to me that it not just come from my own version of queer

439
00:32:28,360 --> 00:32:32,640
 lived experience, but be expressive of multitudes of that.

440
00:32:32,640 --> 00:32:36,920
 And then I would imagine as well that developing these methods and

441
00:32:36,920 --> 00:32:44,200
 methodologies could mean that someone else could create a Latin American corpus,

442
00:32:44,200 --> 00:32:49,720
 or something very specific to their own perspective as well.

443
00:32:49,720 --> 00:32:53,680
 But it wouldn't be about, okay, here's the queer corpus,

444
00:32:53,680 --> 00:32:58,440
 this is a queer love corpus, it's not the definitive, and it should never be, right?

445
00:32:58,440 --> 00:33:05,360
 >> Hi, thank you so much for the talk.

446
00:33:05,360 --> 00:33:09,920
 I was just wondering, what is your experience with the industry players?

447
00:33:09,920 --> 00:33:14,080
 Do you see any significant reactions to your finding?

448
00:33:14,080 --> 00:33:17,880
 Or is it just they don't care or it's the obvious whitewashing, yeah, yeah,

449
00:33:17,880 --> 00:33:19,840
 it's important to us, but they don't do anything?

450
00:33:21,320 --> 00:33:26,840
 >> Yeah, right now, this is not something that I am putting in

451
00:33:26,840 --> 00:33:30,320
 conversation with any industry work.

452
00:33:30,320 --> 00:33:36,440
 I think necessarily it's very small and slow and separate intentionally so.

453
00:33:36,440 --> 00:33:42,440
 >> Just a more general question.

454
00:33:42,440 --> 00:33:50,400
 In queer theory, there's this idea that this is not about representation,

455
00:33:50,400 --> 00:33:52,880
 it's more like about a more theoretical approach.

456
00:33:52,880 --> 00:33:57,880
 Seeing queer as something that is non-normative and

457
00:33:57,880 --> 00:34:01,280
 actually will move when the norms shift.

458
00:34:01,280 --> 00:34:08,600
 Things like that seem kind of in total opposition to the idea of an LLM.

459
00:34:08,600 --> 00:34:13,360
 What do we do about that?

460
00:34:13,360 --> 00:34:14,480
 >> I love that question.

461
00:34:14,480 --> 00:34:18,840
 I mean, I would love to challenge us to think what kind of

462
00:34:18,840 --> 00:34:26,360
 technical non-normative approaches we might take.

463
00:34:26,360 --> 00:34:31,040
 Because the way these are designed are very much, how do we move toward

464
00:34:31,040 --> 00:34:37,600
 the words that are most used, most connected to each other most often?

465
00:34:37,600 --> 00:34:46,600
 And all of the transformer steps are narrowing to that norm.

466
00:34:46,600 --> 00:34:49,040
 Could we imagine a model that does the opposite?

467
00:34:49,040 --> 00:34:53,800
 Could we imagine something that is architecture queer?

468
00:34:53,800 --> 00:34:57,160
 I would be super interested to think out what that means.

469
00:34:57,160 --> 00:35:00,720
 I don't think anybody's doing that, I would love to.

470
00:35:00,720 --> 00:35:04,600
 It might be a very poetic model,

471
00:35:04,600 --> 00:35:08,320
 it might be completely economically useless, and I would love that.

472
00:35:08,320 --> 00:35:13,640
 >> Yeah, thank you for your talk.

473
00:35:14,800 --> 00:35:19,840
 I was just thinking that you are looking to proprietary data model that

474
00:35:19,840 --> 00:35:26,160
 basically run on someone else's server, so you don't have control over

475
00:35:26,160 --> 00:35:29,320
 what data sets they train on and so on.

476
00:35:29,320 --> 00:35:37,520
 But there are also free software models like Lama AI that basically provide you

477
00:35:37,520 --> 00:35:43,160
 with the way to deduce the next words depending on what the previous are.

478
00:35:43,160 --> 00:35:45,920
 And you can run it on your own computer, so

479
00:35:45,920 --> 00:35:51,000
 you can even generate a new data set or new weights for the model,

480
00:35:51,000 --> 00:35:56,560
 which would take considerable computing power to generate them.

481
00:35:56,560 --> 00:36:01,640
 But then you would be in control in exactly what you put in, so

482
00:36:01,640 --> 00:36:07,640
 if you put your data set then it would generate the data out of the data set.

483
00:36:07,640 --> 00:36:10,360
 >> Yeah, exactly, so that's the goal.

484
00:36:10,360 --> 00:36:15,320
 Once we have a large enough data set, we are currently looking at the proprietary

485
00:36:15,320 --> 00:36:21,160
 models as much as possible, but also looking at Bloom and Lama to kind of

486
00:36:21,160 --> 00:36:25,720
 infer backwards what's being used in the ones that we don't have access to too.

487
00:36:25,720 --> 00:36:28,200
 And then once we have a large enough data set,

488
00:36:28,200 --> 00:36:33,240
 I would love anybody here who wants to contribute ideas for that data set.

489
00:36:33,240 --> 00:36:39,600
 We will start training, fine tuning on top of the open source models.

490
00:36:39,600 --> 00:36:43,160
 And then eventually once we have a big enough data set,

491
00:36:43,160 --> 00:36:44,560
 training from scratch hopefully.

492
00:36:44,560 --> 00:36:47,220
 >> Hi.

493
00:36:47,220 --> 00:36:47,880
 >> Yeah.

494
00:36:47,880 --> 00:36:49,040
 >> Hi.

495
00:36:49,040 --> 00:36:52,400
 >> So thanks for the talk and thanks everyone for the questions.

496
00:36:52,400 --> 00:36:54,160
 I have two questions if that's okay.

497
00:36:54,160 --> 00:36:56,040
 One more on the technical side.

498
00:36:56,040 --> 00:37:02,240
 I believe that the current solutions are dealing with the kind of safety and

499
00:37:02,240 --> 00:37:07,880
 sometimes hate speech issues by human reinforced learning.

500
00:37:07,880 --> 00:37:11,280
 And you seem to be proposing attacking it from the other side,

501
00:37:11,280 --> 00:37:12,880
 which is the data set.

502
00:37:12,880 --> 00:37:17,000
 So I would like to know if you have tried and failed with the reinforced learning,

503
00:37:17,000 --> 00:37:20,680
 and that's what took you to look at the data set duration.

504
00:37:20,680 --> 00:37:25,000
 And a related question would be maybe a little bit controversial, but

505
00:37:25,000 --> 00:37:30,760
 what is your view on when filtering the input training data sets become censorship?

506
00:37:30,760 --> 00:37:33,080
 >> Those are great questions.

507
00:37:36,160 --> 00:37:38,760
 I think there's a question up here too.

508
00:37:38,760 --> 00:37:43,120
 I have done a good amount of reading on the human reinforced

509
00:37:43,120 --> 00:37:44,760
 processes that have been going on.

510
00:37:44,760 --> 00:37:48,680
 I feel like it stresses me out.

511
00:37:48,680 --> 00:37:51,920
 It's a little bit problematic ways that they,

512
00:37:51,920 --> 00:37:54,880
 from what I've seen how open AI are doing it.

513
00:37:54,880 --> 00:38:01,160
 And kind of the terms of use that you're opting into as you do this.

514
00:38:01,160 --> 00:38:04,000
 So I haven't myself been working with them.

515
00:38:04,000 --> 00:38:08,800
 I'm just still kind of watching the space to see what the approaches are.

516
00:38:08,800 --> 00:38:14,120
 But I think rather than starting with here's everything, what do we remove?

517
00:38:14,120 --> 00:38:17,600
 And because it's not really everything, it is still curated, but

518
00:38:17,600 --> 00:38:22,360
 they're not treating it as a curated space.

519
00:38:22,360 --> 00:38:25,720
 We can more intentionally curate for

520
00:38:25,720 --> 00:38:30,280
 the actual job that we actually want it to do, because it's not serving us.

521
00:38:30,280 --> 00:38:35,440
 So I think that the other approach is just more interesting to me and

522
00:38:35,440 --> 00:38:39,480
 provides more of a service in a way because it's

523
00:38:39,480 --> 00:38:42,400
 collecting from perspectives that aren't being represented.

524
00:38:42,400 --> 00:38:47,040
 And to your question about filtering versus censorship,

525
00:38:47,040 --> 00:38:48,400
 I think that's really interesting.

526
00:38:48,400 --> 00:38:54,960
 I wonder if that wouldn't be as much of an issue with the opt in model.

527
00:38:56,320 --> 00:39:01,480
 I did wonder what would happen if somebody contributes to our forum and

528
00:39:01,480 --> 00:39:05,800
 puts in a hate speech, a text that is queer hate speech.

529
00:39:05,800 --> 00:39:13,960
 My first thought is that that should also go in because that is part of

530
00:39:13,960 --> 00:39:18,400
 queer lived experience, but it's not the only part and

531
00:39:18,400 --> 00:39:22,280
 it's maybe hopefully eventually not the majority part.

532
00:39:22,280 --> 00:39:27,520
 So, but what we're seeing right now in the larger models is the way that they are

533
00:39:27,520 --> 00:39:31,680
 being architected, it becomes the only part that we're seeing.

534
00:39:31,680 --> 00:39:38,080
 And part of that is because of that ad hoc after the fact human reinforced training.

535
00:39:38,080 --> 00:39:42,040
 Because anytime it says dyke, it's like nope, speech,

536
00:39:42,040 --> 00:39:43,920
 you don't get to say anything about us.

537
00:39:43,920 --> 00:39:46,900
 >> Yeah.

538
00:39:50,400 --> 00:39:55,160
 Regarding your data set, I was really interested in the meta information you

539
00:39:55,160 --> 00:40:01,480
 use and have you played around using kind of the same data, for

540
00:40:01,480 --> 00:40:07,480
 example Reddit or stuff like that with some more metadata added.

541
00:40:07,480 --> 00:40:12,120
 Like for the Reddit, like which subreddit, which user posted,

542
00:40:12,120 --> 00:40:17,120
 which also like publisher and stuff like that could be possible.

543
00:40:17,120 --> 00:40:18,440
 Have you played around with that?

544
00:40:18,440 --> 00:40:26,040
 If that already helps the quality of the resulting neural network and

545
00:40:26,040 --> 00:40:31,280
 maybe being more aware of what they are typing and from what kind of source it comes.

546
00:40:31,280 --> 00:40:34,520
 >> Interesting, I like that idea.

547
00:40:34,520 --> 00:40:39,120
 I think, so if you look at, I believe it's the open web text too,

548
00:40:39,120 --> 00:40:45,080
 the one that's open source does include which subreddits, but it's not filtering for,

549
00:40:45,080 --> 00:40:48,880
 it's not trying to curate from a specific voice.

550
00:40:48,880 --> 00:40:53,080
 And so I haven't done that, it's like taking the whole of Reddit because we're

551
00:40:53,080 --> 00:40:56,000
 coming from this very additive rather than subtractive approach.

552
00:40:56,000 --> 00:41:01,680
 But I think the other issue with Reddit is that people, you are consenting,

553
00:41:01,680 --> 00:41:03,560
 it is publicly available.

554
00:41:03,560 --> 00:41:08,960
 But when that started, the expectation wasn't that your private,

555
00:41:08,960 --> 00:41:13,480
 your anonymous story would be sucked into this large language model.

556
00:41:13,480 --> 00:41:18,200
 So I would be more interested in including people who know that that's what they're

557
00:41:18,200 --> 00:41:24,320
 consenting to and are invested in the project of imagining queer archives and

558
00:41:24,320 --> 00:41:24,920
 things like that.

559
00:41:24,920 --> 00:41:31,200
 >> Yeah, I think, yeah, hi.

560
00:41:31,200 --> 00:41:36,880
 >> I have a question regarding self reflection using AI.

561
00:41:36,880 --> 00:41:40,160
 I think that's a really powerful way to use AI.

562
00:41:40,160 --> 00:41:44,120
 And I'm interested, is there some kind of,

563
00:41:44,120 --> 00:41:48,520
 are there problems if you're using this as a queer person?

564
00:41:48,520 --> 00:41:56,960
 So does it do any suggestions that are just based on that you're a queer person and

565
00:41:56,960 --> 00:41:58,920
 it's really not suitable?

566
00:41:58,920 --> 00:42:05,040
 >> Yeah, so if I understand your question, you're saying if you're using one of

567
00:42:05,040 --> 00:42:13,360
 the existing models and trying to kind of self reflect,

568
00:42:13,360 --> 00:42:18,520
 that it's not treating you as even if you identify as, yeah, can you?

569
00:42:18,520 --> 00:42:25,920
 >> It's about building conflict systems and what are your needs for

570
00:42:25,920 --> 00:42:29,360
 a suitable conflict system for you and the group.

571
00:42:29,360 --> 00:42:33,200
 >> Yeah. >> And it's really helpful to use AI to

572
00:42:33,200 --> 00:42:38,120
 self reflect on what are your needs for this conflict space.

573
00:42:38,120 --> 00:42:41,200
 And I'm interested, are there any biases?

574
00:42:41,200 --> 00:42:45,600
 Is this a problem when using AI because you are a queer person?

575
00:42:45,600 --> 00:42:53,520
 >> Yeah, I think if I understand your question correctly, yeah,

576
00:42:53,520 --> 00:42:57,480
 I find at least with the existing versions, I find it very limiting for

577
00:42:57,480 --> 00:43:02,320
 being able to do that because it has a very narrow scope for

578
00:43:02,320 --> 00:43:06,600
 what it's going to understand about that, right?

579
00:43:06,600 --> 00:43:12,400
 So I would be interested to try if the data set that I'm talking about

580
00:43:12,400 --> 00:43:15,840
 already existed and if these models already existed,

581
00:43:15,840 --> 00:43:21,240
 what alternative might come out of a self reflection that would be very,

582
00:43:21,240 --> 00:43:24,320
 very different than the version I would get from chat GPT?

583
00:43:24,320 --> 00:43:26,480
 Does that answer your question?

584
00:43:26,480 --> 00:43:30,480
 Maybe we can talk, yeah, I'd be curious.

585
00:43:30,480 --> 00:43:37,520
 >> I'm wondering if there's been any sort of practice in instead of

586
00:43:37,520 --> 00:43:41,240
 only doing things on the back end,

587
00:43:41,240 --> 00:43:47,520
 having more of an engaged approach with the user.

588
00:43:47,520 --> 00:43:49,520
 Cuz right now it's like, did you like this answer?

589
00:43:49,520 --> 00:43:53,200
 Yes, no, and I know it's way more complex than that.

590
00:43:53,200 --> 00:43:59,480
 So sort of like taking in a little bit more

591
00:43:59,480 --> 00:44:03,720
 content to then like, hey, a conflict just arose.

592
00:44:03,720 --> 00:44:08,200
 It made somebody feel like this and kind of engage the user to report it and

593
00:44:08,200 --> 00:44:09,400
 maybe describe something.

594
00:44:09,400 --> 00:44:12,040
 I know that's kind of a burden on the user, but

595
00:44:12,040 --> 00:44:16,280
 I was just curious how much that's been leveraged in this space.

596
00:44:16,280 --> 00:44:21,920
 >> Yeah, so there's a couple things that I know about happening in that regard.

597
00:44:21,920 --> 00:44:27,040
 There's the human reinforcement training that was mentioned earlier that's

598
00:44:27,040 --> 00:44:32,000
 happening with the large companies developing these, where they're going

599
00:44:32,000 --> 00:44:38,280
 through and it's usually about as simple as was this a good answer.

600
00:44:38,280 --> 00:44:43,760
 But I think as you described, it's gonna get more complex very quickly.

601
00:44:43,760 --> 00:44:49,560
 And then I know of at least one research group that is looking at building tools

602
00:44:49,560 --> 00:44:56,240
 for users to compare different models, outputs, using the same prompt.

603
00:44:56,240 --> 00:44:59,160
 Which I think is really interesting as a user.

604
00:44:59,160 --> 00:45:04,520
 But I think I would also encourage us to think about our agency as users

605
00:45:04,520 --> 00:45:10,800
 not only being at the front end at the end of the process, but

606
00:45:10,800 --> 00:45:15,040
 again at the beginning of the process and holistically through the pipeline.

607
00:45:15,040 --> 00:45:24,080
 Also, if anyone wants to look up a keyword in the data set for

608
00:45:24,080 --> 00:45:26,680
 the similarity score, we can play with that in the time remaining too.

609
00:45:26,680 --> 00:45:28,960
 I think there were more questions.

610
00:45:28,960 --> 00:45:35,480
 >> Hello, I have one question regarding the content of people going with the data.

611
00:45:35,480 --> 00:45:38,320
 So you said you would like to use the archives,

612
00:45:38,320 --> 00:45:41,120
 but you would rather not use Reddit data.

613
00:45:41,120 --> 00:45:43,880
 And I think if you have the content of right holders,

614
00:45:43,880 --> 00:45:47,320
 it's basically not the same as having the content of the people who originally

615
00:45:47,320 --> 00:45:48,240
 published the work.

616
00:45:48,240 --> 00:45:54,480
 And so I really don't know if it might make a lot of, yeah,

617
00:45:54,480 --> 00:46:00,520
 hurts, like it will decrease the quality of the data set in the end so

618
00:46:00,520 --> 00:46:05,640
 much that maybe considering just doing it without the content might be the better

619
00:46:05,640 --> 00:46:06,880
 approach.

620
00:46:06,880 --> 00:46:08,800
 >> Yeah, I appreciate your point.

621
00:46:08,800 --> 00:46:12,720
 I think that one of the things we wanna think about as we're developing what

622
00:46:12,720 --> 00:46:19,080
 methods make sense is this tension of kind of the complexity and

623
00:46:19,080 --> 00:46:23,000
 impossibility of perfect consent and rights.

624
00:46:23,000 --> 00:46:27,200
 And how the existing information,

625
00:46:27,200 --> 00:46:33,200
 the existing IP structures don't fit these new tools.

626
00:46:33,200 --> 00:46:39,440
 So how do we build, it's kind of an impossible task,

627
00:46:39,440 --> 00:46:40,960
 an intersectional language model.

628
00:46:42,240 --> 00:46:48,120
 The idea might take centuries, millennia, I don't know.

629
00:46:48,120 --> 00:46:53,040
 But we're kind of embracing the failure to explore what's in that activity

630
00:46:53,040 --> 00:46:56,320
 to kind of point to the failures of these larger systems.

631
00:46:56,320 --> 00:47:01,600
 So yeah, if we find that we just can't get a big enough data set with archives only,

632
00:47:01,600 --> 00:47:06,560
 or we have issues with the consent because of the way the archives

633
00:47:06,560 --> 00:47:11,320
 were licensed originally, what does it mean to consent and for whom and

634
00:47:11,320 --> 00:47:12,840
 from whom?

635
00:47:12,840 --> 00:47:16,800
 These are all questions that are not gonna have perfect answers and

636
00:47:16,800 --> 00:47:21,280
 we certainly want to just be at least bringing attention to the questions.

637
00:47:21,280 --> 00:47:26,840
 And the hope is that it also brings attention to how these questions

638
00:47:26,840 --> 00:47:30,960
 appear in the commercial data sets and the commercial models.

639
00:47:30,960 --> 00:47:33,960
 Because they are there, we're just not attending to them.

640
00:47:33,960 --> 00:47:35,500
 Thank you.

641
00:47:38,600 --> 00:47:44,880
 >> I had a question about, so this was a little bit off of a previous question.

642
00:47:44,880 --> 00:47:53,720
 There was a project that I know of where they used AI to talk with people who

643
00:47:53,720 --> 00:47:59,400
 had cancer and they were young teenagers who

644
00:47:59,400 --> 00:48:03,640
 had experienced cancer and couldn't find a community.

645
00:48:03,640 --> 00:48:06,320
 Their friends didn't understand what that meant.

646
00:48:06,320 --> 00:48:11,040
 They had a hard time talking with someone else about the trauma that they experienced

647
00:48:11,040 --> 00:48:11,560
 around that.

648
00:48:11,560 --> 00:48:21,480
 I think there is possibly an opportunity to create an AI to talk to about trauma

649
00:48:21,480 --> 00:48:27,480
 related to queer experience, hopefully infrequent trauma experiences.

650
00:48:27,480 --> 00:48:32,800
 I was wondering if you have information about,

651
00:48:32,800 --> 00:48:37,920
 if you see a budding community around building and

652
00:48:37,920 --> 00:48:41,680
 creating these things within our queer community so far.

653
00:48:41,680 --> 00:48:45,840
 >> Yeah, absolutely, I do.

654
00:48:45,840 --> 00:48:52,080
 And I would love if any of you are inspired from this to join that effort.

655
00:48:52,080 --> 00:48:55,280
 I think there's many ways to intervene in that space.

656
00:48:55,280 --> 00:49:00,440
 And some of them are really building positive tools like what you've described.

657
00:49:00,440 --> 00:49:05,920
 And some of them are building weird, uncomfortable, messed up interventions.

658
00:49:05,920 --> 00:49:10,720
 And I think all of that belongs in this resistance to

659
00:49:10,720 --> 00:49:14,360
 the generative machine learning that we're seeing.

660
00:49:14,360 --> 00:49:28,360
 >> Thank you.

661
00:49:28,360 --> 00:49:32,240
 Yeah, picking up on that very interesting remark from our comrade over here.

662
00:49:32,240 --> 00:49:36,760
 So queer is a process against the status quo, it's an eternal process, right?

663
00:49:36,760 --> 00:49:41,080
 So there's no such thing as a stable identity because that would be contradictory.

664
00:49:41,080 --> 00:49:45,480
 And we saw a very interesting example in the scratch board of corporate speech.

665
00:49:45,480 --> 00:49:48,440
 Right, it's corporate speech affirming speech, so to speak, right?

666
00:49:48,440 --> 00:49:52,200
 And you rightly pointed out that that was cringe.

667
00:49:52,200 --> 00:49:56,320
 And so I was wondering, could you give a concrete example of what you would like

668
00:49:56,320 --> 00:49:57,760
 to see in the next iteration?

669
00:49:57,760 --> 00:50:03,240
 >> In the commercially available next iteration?

670
00:50:03,240 --> 00:50:07,240
 >> For example, what we saw just in a better form,

671
00:50:07,240 --> 00:50:12,720
 the output you would have said yes to in that example.

672
00:50:12,720 --> 00:50:16,680
 >> Mm-hm, okay, well, two parts.

673
00:50:16,680 --> 00:50:20,680
 I think through this experience, I'm not sure that I wanna see

674
00:50:20,680 --> 00:50:24,640
 the version that I want in the commercial iteration.

675
00:50:24,640 --> 00:50:28,360
 I feel like maybe it doesn't belong to everybody,

676
00:50:28,360 --> 00:50:32,520
 it belongs to people who build it for ourselves.

677
00:50:32,520 --> 00:50:39,520
 I can't imagine what open AI might do with a queer intersectional language model.

678
00:50:39,520 --> 00:50:42,160
 But from what we've seen them do so far,

679
00:50:42,160 --> 00:50:46,600
 I feel like it wouldn't be aligned with my values.

680
00:50:46,600 --> 00:50:50,280
 And what kind of response might I want to see?

681
00:50:52,960 --> 00:51:00,440
 >> Yeah, I think that it would be something that I could recognize myself in or

682
00:51:00,440 --> 00:51:04,400
 that I could recognize people I know in, that it would be affirming.

683
00:51:04,400 --> 00:51:09,640
 Maybe it would be complex, it wouldn't necessarily be perfect or optimistic.

684
00:51:09,640 --> 00:51:11,240
 But it would be dynamic, right?

685
00:51:11,240 --> 00:51:19,000
 It would acknowledge that there's more to queer lived experience than

686
00:51:20,040 --> 00:51:26,240
 what it knows now, which is being queer is illegal in this country.

687
00:51:26,240 --> 00:51:31,080
 So if you're on a plane, the flight attendant has to tell the people on

688
00:51:31,080 --> 00:51:34,560
 the plane that they're going into a dangerous situation.

689
00:51:34,560 --> 00:51:37,880
 There's this latent sense of fear and

690
00:51:37,880 --> 00:51:41,560
 danger built into this that is really shocking to me.

691
00:51:41,560 --> 00:51:46,120
 I would love for it to be like the show Schitt's Creek,

692
00:51:46,120 --> 00:51:49,560
 where there's queer people, but they're just living their lives and

693
00:51:49,560 --> 00:51:52,480
 it's not the main thread of their storyline, right?

694
00:51:52,480 --> 00:52:00,100
 Yeah?

695
00:52:00,100 --> 00:52:09,720
 >> I wonder how much of this is the data set and how much is the actual AI

696
00:52:09,720 --> 00:52:15,040
 kind of algorithms not really understanding the context of what it's looking at.

697
00:52:15,040 --> 00:52:20,320
 Because I often feel you can say someone's sexual orientation and

698
00:52:20,320 --> 00:52:24,560
 it has no real relevance other than defining the gender of the people involved.

699
00:52:24,560 --> 00:52:27,440
 And then other times it's actually incredibly relevant and

700
00:52:27,440 --> 00:52:29,480
 the context really matters.

701
00:52:29,480 --> 00:52:34,160
 And then the other thing I was wondering about was

702
00:52:34,160 --> 00:52:40,640
 sort of how you weight the relevance of data.

703
00:52:40,640 --> 00:52:46,080
 So for example, a pivotal book, would you weight that more?

704
00:52:46,080 --> 00:52:51,400
 Would you, as values change, how does the data set evolve over a number of years as

705
00:52:51,400 --> 00:52:55,600
 society values change and people's views change?

706
00:52:55,600 --> 00:52:57,120
 >> Yeah, good question, thank you.

707
00:52:57,120 --> 00:53:00,520
 So it's both and the data set and

708
00:53:00,520 --> 00:53:07,760
 the model are deeply entangled because the way the algorithm is built

709
00:53:07,760 --> 00:53:13,600
 is really pulling from the frequency of each word appearing in the data set.

710
00:53:13,600 --> 00:53:17,880
 So if Shakespeare appears in there more than anybody else,

711
00:53:17,880 --> 00:53:21,520
 it's gonna quote unquote think like Shakespeare, right?

712
00:53:21,520 --> 00:53:28,320
 So as the question before about how a queer model might be architected to do

713
00:53:28,320 --> 00:53:32,320
 something differently, that might be another opportunity to intervene and

714
00:53:32,320 --> 00:53:33,760
 build it a different way.

715
00:53:33,760 --> 00:53:38,000
 But right now, the two are very much tied together, so

716
00:53:38,000 --> 00:53:42,480
 it absolutely is the data set and the model.

717
00:53:42,480 --> 00:53:48,720
 But they are connected too deeply to kind of detach.

718
00:53:48,720 --> 00:53:53,720
 Thank you, yeah.

719
00:53:53,720 --> 00:53:55,080
 >> Thanks.

720
00:53:55,080 --> 00:54:00,800
 We have five minutes left according to the time schedule, but I feel like-

721
00:54:00,800 --> 00:54:02,080
 >> Yeah, I'm happy- >> It's getting long.

722
00:54:02,080 --> 00:54:06,920
 >> Yeah, now I'm happy to talk with anybody else after or play with the tools.

723
00:54:06,920 --> 00:54:09,280
 And thank you so much for all your questions.

724
00:54:09,280 --> 00:54:19,280
 >> [APPLAUSE]

725
00:54:20,160 --> 00:54:25,240
 [MUSIC]