1
00:00:00,320 --> 00:00:02,800
Speaker 1: If someone knocked on your door and was like, I'm

2
00:00:02,840 --> 00:00:05,160
Speaker 1: gonna sell you a potion that can cure all of

3
00:00:05,200 --> 00:00:08,920
Speaker 1: your problems, but it's just gonna cost everything you've ever owned,

4
00:00:08,960 --> 00:00:12,040
Speaker 1: including your firstborn child, you would be like, this is

5
00:00:12,080 --> 00:00:18,600
Speaker 1: a scam.

6
00:00:18,760 --> 00:00:20,480
Speaker 2: There are no Girls on the Internet. As a production

7
00:00:20,520 --> 00:00:28,040
Speaker 2: of iHeartRadio and unbost Creative, I'm Bridget Todd and this

8
00:00:28,120 --> 00:00:33,320
Speaker 2: is there are no Girls on the Internet. Few journalists

9
00:00:33,400 --> 00:00:36,080
Speaker 2: have covered AI with the depth and rigor of Karen

10
00:00:36,120 --> 00:00:40,280
Speaker 2: how Her book Empire of AI tells the story of

11
00:00:40,360 --> 00:00:43,680
Speaker 2: Open AI and Sam Altman not as someone trying to

12
00:00:43,720 --> 00:00:47,479
Speaker 2: start a tech company, but rather someone trying to start

13
00:00:47,520 --> 00:00:52,280
Speaker 2: an ever expanding empire. Now, with her new BBC podcast,

14
00:00:52,440 --> 00:00:56,360
Speaker 2: The Interface, Karen and her co hosts are pulling back

15
00:00:56,400 --> 00:01:00,240
Speaker 2: the curtain on how tech shapes everything from politics our

16
00:01:00,280 --> 00:01:04,880
Speaker 2: most personal lives. Karen's voice in tech is critical, but

17
00:01:04,959 --> 00:01:09,800
Speaker 2: it happened very much by accident. What brought you to

18
00:01:09,920 --> 00:01:13,600
Speaker 2: reporting on both open AI and then AI.

19
00:01:13,400 --> 00:01:20,720
Speaker 1: In general total accidents, both of them. So I had

20
00:01:20,760 --> 00:01:24,240
Speaker 1: an interesting route into journalism. I studied engineering in college

21
00:01:24,480 --> 00:01:26,680
Speaker 1: and thought I would work in the tech industry, which

22
00:01:26,680 --> 00:01:31,560
Speaker 1: I did when I graduated, and very quickly after experiencing

23
00:01:32,959 --> 00:01:36,880
Speaker 1: a little over a year of Silicon Valley, I realized

24
00:01:37,120 --> 00:01:37,880
Speaker 1: that it.

25
00:01:37,760 --> 00:01:38,440
Speaker 3: Was not for me.

26
00:01:39,440 --> 00:01:42,360
Speaker 1: I joined the tech industry at a time when in

27
00:01:42,560 --> 00:01:45,520
Speaker 1: it was twenty fifteen twenty sixteen era, when the what

28
00:01:45,640 --> 00:01:48,040
Speaker 1: was called the tech backlash or the tech lash was

29
00:01:48,120 --> 00:01:51,200
Speaker 1: just starting, because people were starting to realize how powerful

30
00:01:51,240 --> 00:01:54,200
Speaker 1: these companies were and how they were actually often undermining

31
00:01:54,200 --> 00:01:57,160
Speaker 1: the public interest. And so I thought I would have

32
00:01:58,440 --> 00:02:01,600
Speaker 1: I thought I needed to find another career, and so

33
00:02:02,440 --> 00:02:06,720
Speaker 1: on a whim, I pivoted to journalism because I was like,

34
00:02:06,880 --> 00:02:09,160
Speaker 1: the only other school I have is writing, so maybe

35
00:02:09,200 --> 00:02:12,239
Speaker 1: I can get a job in journalism. And I couldn't

36
00:02:12,320 --> 00:02:14,720
Speaker 1: find a job and what I wanted to report on,

37
00:02:14,760 --> 00:02:17,280
Speaker 1: which was the environment, because I had no experience in

38
00:02:17,320 --> 00:02:20,680
Speaker 1: journalism or in the environment. But I figured out that

39
00:02:20,720 --> 00:02:24,160
Speaker 1: I could parlay my tech experience working the technu stry

40
00:02:24,280 --> 00:02:28,240
Speaker 1: into reporting on tech, and so I started applying for

41
00:02:28,840 --> 00:02:33,440
Speaker 1: tech reporting jobs and the only job offer that I

42
00:02:33,480 --> 00:02:34,639
Speaker 1: got was to cover AI.

43
00:02:36,360 --> 00:02:39,360
Speaker 2: When Karen was assigned to cover AI, at first, she

44
00:02:39,520 --> 00:02:43,160
Speaker 2: wasn't thrilled. The landscape looked nothing like it does today,

45
00:02:43,480 --> 00:02:46,160
Speaker 2: none of the breathless hype, none of the alarm bells.

46
00:02:46,919 --> 00:02:50,440
Speaker 2: It was a different time, so Karen thought that AI

47
00:02:50,840 --> 00:02:52,680
Speaker 2: might be kind of a dug beat.

48
00:02:56,120 --> 00:02:59,960
Speaker 1: When I received that job offer, I actually was so disappointed,

49
00:03:00,160 --> 00:03:04,120
Speaker 1: like I thought that it would be the least interesting

50
00:03:04,880 --> 00:03:10,160
Speaker 1: job ever, and I almost didn't take it, like to

51
00:03:10,240 --> 00:03:11,760
Speaker 1: try and find another option.

52
00:03:11,919 --> 00:03:13,640
Speaker 3: But then I ended up giving it a try because

53
00:03:13,840 --> 00:03:14,240
Speaker 3: I just.

54
00:03:14,200 --> 00:03:19,400
Speaker 1: Needed something, and within months, just like two months, I

55
00:03:19,440 --> 00:03:21,800
Speaker 1: absolutely fell in love with the beat because I realized

56
00:03:21,840 --> 00:03:24,080
Speaker 1: that it was so much more than I had understood,

57
00:03:24,200 --> 00:03:28,120
Speaker 1: and it was an opportunity to explore every facet of

58
00:03:28,160 --> 00:03:33,359
Speaker 1: tech and society like I wanted to. And then because

59
00:03:33,400 --> 00:03:36,240
Speaker 1: I was covering AI for MIT Technology Review, which is

60
00:03:36,280 --> 00:03:39,720
Speaker 1: a very research at the time was a very research

61
00:03:39,720 --> 00:03:43,560
Speaker 1: focused publication. It was looking at the cutting edge stuff

62
00:03:43,600 --> 00:03:47,800
Speaker 1: happening in labs before commercialization potential, and AI at the

63
00:03:47,800 --> 00:03:53,480
Speaker 1: time was at that stage without a lot of commercial activity,

64
00:03:53,680 --> 00:03:57,440
Speaker 1: primarily being developed in academic labs or in corporate labs.

65
00:03:57,960 --> 00:04:00,440
Speaker 2: So how did you start covering open AIS pecifically?

66
00:04:01,200 --> 00:04:03,360
Speaker 1: Opening Eye came on my radar because it was one

67
00:04:03,400 --> 00:04:08,480
Speaker 1: of the research labs that build itself as having absolutely

68
00:04:08,520 --> 00:04:13,600
Speaker 1: no commercial interest, and because I was the junior AI

69
00:04:13,720 --> 00:04:17,200
Speaker 1: reporter on staff, and OpenEye was just important enough to

70
00:04:17,240 --> 00:04:20,280
Speaker 1: have a profile, but not so important to put the

71
00:04:20,360 --> 00:04:23,240
Speaker 1: senior AI reporter on it. I ended up being the

72
00:04:23,279 --> 00:04:26,080
Speaker 1: one assigned to profile open ai, and that's how I

73
00:04:26,160 --> 00:04:29,839
Speaker 1: ended up accidentally being the first journalist to ever profile.

74
00:04:29,480 --> 00:04:34,680
Speaker 2: Open It is so funny to take that walk down

75
00:04:34,720 --> 00:04:38,920
Speaker 2: memory lane and think about how different open Ai started

76
00:04:39,000 --> 00:04:43,880
Speaker 2: as because I had this almost the same trajectory of thinking, Oh,

77
00:04:43,960 --> 00:04:48,800
Speaker 2: this is an really academic, nonprofit organization. Probably don't need

78
00:04:48,839 --> 00:04:50,880
Speaker 2: to really like look too hard at what they're doing,

79
00:04:51,360 --> 00:04:55,359
Speaker 2: and that I mean, saying that now sounds absurd. You

80
00:04:55,400 --> 00:04:58,960
Speaker 2: know how quickly they've gone from oh, we're not a

81
00:04:58,960 --> 00:05:01,240
Speaker 2: commerc we're not we don't have commercial interest to we

82
00:05:01,279 --> 00:05:03,040
Speaker 2: are building an empire and we're going to take over

83
00:05:03,040 --> 00:05:03,440
Speaker 2: the world.

84
00:05:04,040 --> 00:05:06,400
Speaker 3: Yeah, yeah, no, it's it's super like.

85
00:05:06,760 --> 00:05:08,760
Speaker 1: I think at the time, I had a totally different

86
00:05:08,760 --> 00:05:13,640
Speaker 1: perception of open ai than I do now, so in

87
00:05:13,680 --> 00:05:16,760
Speaker 1: the sense that when I started profiling the company, it

88
00:05:16,880 --> 00:05:20,840
Speaker 1: was already clear that they were leaving behind the mission

89
00:05:20,880 --> 00:05:23,080
Speaker 1: of being a nonprofit, being open and being in the

90
00:05:23,080 --> 00:05:27,240
Speaker 1: public interest. But I had this impression that it started

91
00:05:27,240 --> 00:05:30,360
Speaker 1: that way, and then it was sort of corrupted along

92
00:05:30,400 --> 00:05:34,960
Speaker 1: the way by commercial interest. But in hindsight, after reporting

93
00:05:35,080 --> 00:05:39,320
Speaker 1: my book, I realized that actually there was a seed

94
00:05:39,360 --> 00:05:43,159
Speaker 1: of corruption from the very beginning within the company because

95
00:05:44,960 --> 00:05:48,640
Speaker 1: open ai set itself up to be a nonprofit, specifically

96
00:05:49,240 --> 00:05:52,559
Speaker 1: because it wanted to be the number one AI lab

97
00:05:52,720 --> 00:05:56,440
Speaker 1: and beat Google, which was at the time the dominant

98
00:05:56,480 --> 00:06:01,160
Speaker 1: AI player, and they, I think what happened is that

99
00:06:01,320 --> 00:06:03,800
Speaker 1: with this goal in mind of being number one and

100
00:06:04,000 --> 00:06:07,640
Speaker 1: dominating in this space, they realized that they couldn't compete

101
00:06:07,640 --> 00:06:11,760
Speaker 1: with Google on money because they simply didn't have that

102
00:06:12,040 --> 00:06:14,360
Speaker 1: as much like Google, being one of the richest companies

103
00:06:14,400 --> 00:06:17,760
Speaker 1: in the world, they could. Google could always outbid open

104
00:06:17,760 --> 00:06:21,680
Speaker 1: Ai on salaries and outspend open Ai on various things.

105
00:06:21,720 --> 00:06:25,279
Speaker 1: And so in order to recruit talent, which was the

106
00:06:25,279 --> 00:06:29,520
Speaker 1: first bottleneck that Opening Eye had to overcome, they were

107
00:06:29,600 --> 00:06:33,720
Speaker 1: able to instead appeal to researchers on a sense of

108
00:06:33,760 --> 00:06:38,200
Speaker 1: mission and purpose, which got allowed them to then, you know,

109
00:06:38,920 --> 00:06:43,000
Speaker 1: ask these researchers to take pay cuts and to consider

110
00:06:43,320 --> 00:06:47,080
Speaker 1: jumping ship from Google or from another lucrative job to

111
00:06:47,400 --> 00:06:49,760
Speaker 1: this more startup type environment.

112
00:06:50,360 --> 00:06:53,479
Speaker 3: And it was once they overcame.

113
00:06:53,040 --> 00:06:56,560
Speaker 1: That bottleneck of gathering up all of the researchers that

114
00:06:57,120 --> 00:07:00,440
Speaker 1: they then kind of started to slowly getting rid of

115
00:07:00,440 --> 00:07:03,240
Speaker 1: a nonprofit because it had lost its utility and their

116
00:07:03,279 --> 00:07:05,000
Speaker 1: bottleneck shifted to capital.

117
00:07:05,800 --> 00:07:07,279
Speaker 3: But the goal was always the same.

118
00:07:07,320 --> 00:07:10,000
Speaker 1: The goal was actually not let's create an AI lab

119
00:07:10,080 --> 00:07:14,600
Speaker 1: that is hugely open and transparent it to the public interest,

120
00:07:15,200 --> 00:07:18,760
Speaker 1: like priority number one, was always let's be number one

121
00:07:19,000 --> 00:07:23,480
Speaker 1: and dominate, And they shifted their tactics over time based

122
00:07:23,480 --> 00:07:26,520
Speaker 1: on what they needed in that moment.

123
00:07:27,160 --> 00:07:31,239
Speaker 4: Do you think some of those researchers later felt burned

124
00:07:31,320 --> 00:07:37,080
Speaker 4: or that they had been deceived after that process came

125
00:07:37,200 --> 00:07:39,920
Speaker 4: to more or less completion of transitioning away from the

126
00:07:39,920 --> 00:07:42,240
Speaker 4: nonprofit status that had initially attracted them.

127
00:07:42,720 --> 00:07:43,240
Speaker 3: Absolutely.

128
00:07:43,240 --> 00:07:45,200
Speaker 1: I think this was one of the most interesting things

129
00:07:45,920 --> 00:07:50,400
Speaker 1: for me when reporting the book and speaking with different

130
00:07:50,520 --> 00:07:52,720
Speaker 1: so many employees from different eras of the company, is that,

131
00:07:54,200 --> 00:07:57,200
Speaker 1: first of all, this is like having covered lots of

132
00:07:57,200 --> 00:08:00,880
Speaker 1: different tech companies, like I've written about Facebook, about Google,

133
00:08:01,000 --> 00:08:06,400
Speaker 1: about Microsoft, opening I is the only company where employees

134
00:08:06,440 --> 00:08:10,640
Speaker 1: cannot agree whether it is a company and like that

135
00:08:10,920 --> 00:08:14,800
Speaker 1: like it their opinion about whether it is a company,

136
00:08:14,840 --> 00:08:17,840
Speaker 1: and like what ultimately a stands for and what is

137
00:08:17,880 --> 00:08:22,400
Speaker 1: even the purpose of this organization is completely based on

138
00:08:22,600 --> 00:08:24,080
Speaker 1: when they started at the organization.

139
00:08:24,280 --> 00:08:27,760
Speaker 3: So early day employees they.

140
00:08:27,720 --> 00:08:31,160
Speaker 1: Still think of or thought of Open AI as a

141
00:08:31,200 --> 00:08:34,079
Speaker 1: nonprofit that just had to make some concessions and start

142
00:08:34,120 --> 00:08:38,319
Speaker 1: to look a little bit more company e. Whereas employees

143
00:08:38,320 --> 00:08:40,719
Speaker 1: that were joining after Opening I had already started the

144
00:08:40,720 --> 00:08:43,160
Speaker 1: for profit, had already raised a bunch of capital, and

145
00:08:43,200 --> 00:08:46,320
Speaker 1: were building commercial products. They simply saw it as just

146
00:08:46,400 --> 00:08:49,839
Speaker 1: another tech company, like any other tech company in Silicon Valley,

147
00:08:50,840 --> 00:08:54,200
Speaker 1: and this is like yeah, Like I had a funny

148
00:08:54,280 --> 00:08:58,520
Speaker 1: conversation with one of my fact checkers when she was

149
00:08:59,120 --> 00:09:01,760
Speaker 1: going through all the interviews, like she had the same

150
00:09:01,760 --> 00:09:05,920
Speaker 1: exact observation, and I was like, I feel so validated

151
00:09:05,960 --> 00:09:09,240
Speaker 1: because I thought I was going crazy, like interviewing these

152
00:09:09,280 --> 00:09:13,440
Speaker 1: people and being like, this isn't normal, right, Like usually

153
00:09:13,520 --> 00:09:16,200
Speaker 1: when you work for an entity, you should be able

154
00:09:16,200 --> 00:09:20,360
Speaker 1: to define whether or not the basics about that entity,

155
00:09:20,440 --> 00:09:23,280
Speaker 1: like whether it's a company or a nonprofit. But because

156
00:09:23,280 --> 00:09:26,199
Speaker 1: of Opening Eye's history, it's this. It was this prism

157
00:09:26,320 --> 00:09:28,680
Speaker 1: or this mirror that every single employee was holding up

158
00:09:28,679 --> 00:09:31,120
Speaker 1: to themselves and they were seeing something totally different.

159
00:09:31,800 --> 00:09:33,920
Speaker 2: So I listened to a lot of tech leaders speak

160
00:09:33,960 --> 00:09:36,920
Speaker 2: for the podcast, and generally I have a hard time

161
00:09:36,960 --> 00:09:40,120
Speaker 2: trusting any of them, but that is especially true for

162
00:09:40,200 --> 00:09:42,360
Speaker 2: sam Altman. This actually came up when I was working

163
00:09:42,400 --> 00:09:45,240
Speaker 2: on my own audiobook. I just kind of got the

164
00:09:45,320 --> 00:09:48,679
Speaker 2: sense that he is someone who will say whatever whenever,

165
00:09:48,800 --> 00:09:50,920
Speaker 2: and I don't think any of that is an accident either.

166
00:09:51,640 --> 00:09:54,880
Speaker 2: He is what I would call a flippery fish. How

167
00:09:54,880 --> 00:09:57,280
Speaker 2: did you nail him down for your book? Empire of Ai?

168
00:09:57,800 --> 00:10:00,200
Speaker 1: Sam Alman didn't agree to interview for the books, so

169
00:10:00,559 --> 00:10:03,800
Speaker 1: what I ended up doing was just listening to hours

170
00:10:03,840 --> 00:10:06,800
Speaker 1: and hours and hours and hours of footage of him

171
00:10:06,880 --> 00:10:09,320
Speaker 1: talking in various over the years.

172
00:10:10,200 --> 00:10:11,880
Speaker 3: That's kidding that in my opinion.

173
00:10:11,760 --> 00:10:16,960
Speaker 1: Which which ended up being actually a really great exercise,

174
00:10:17,120 --> 00:10:23,240
Speaker 1: because I realized that first of all, that he ships,

175
00:10:23,840 --> 00:10:26,800
Speaker 1: you know, like what he says over the years changes

176
00:10:26,880 --> 00:10:29,120
Speaker 1: a lot, because.

177
00:10:30,360 --> 00:10:32,319
Speaker 3: He will say what needs.

178
00:10:32,360 --> 00:10:34,559
Speaker 1: What he thinks his audience needs to hear in that moment,

179
00:10:34,679 --> 00:10:37,520
Speaker 1: and that will fluctuate based on you know, the zeitgeistuff

180
00:10:37,520 --> 00:10:41,440
Speaker 1: in the moment. But also I realized that he uses

181
00:10:41,800 --> 00:10:47,240
Speaker 1: very squishy language, even while taking a definitive tone to

182
00:10:47,360 --> 00:10:50,600
Speaker 1: say the things that he's saying so you'll see him

183
00:10:50,640 --> 00:10:56,040
Speaker 1: say statements like we believe that a lot of people

184
00:10:56,320 --> 00:10:59,000
Speaker 1: are going to like this, and.

185
00:11:00,600 --> 00:11:02,760
Speaker 3: Like and and and very.

186
00:11:02,520 --> 00:11:06,840
Speaker 1: Soon there will be more people that like this, you know,

187
00:11:07,120 --> 00:11:11,720
Speaker 1: like like he uses things that are unquantifiable, like a

188
00:11:11,760 --> 00:11:13,440
Speaker 1: lot of very soon.

189
00:11:15,840 --> 00:11:18,480
Speaker 3: Yeah, he rarely ever says anything.

190
00:11:18,200 --> 00:11:23,280
Speaker 1: Like majority, anything that could even have some measurable Oh,

191
00:11:24,240 --> 00:11:27,320
Speaker 1: majority means fifty one percent, so I can actually like

192
00:11:27,440 --> 00:11:31,240
Speaker 1: hold you to that number. Like, he never uses those terms,

193
00:11:31,400 --> 00:11:34,120
Speaker 1: he and he never uses specific values. He only ever

194
00:11:34,240 --> 00:11:36,280
Speaker 1: uses these like squishy things that are in the eye

195
00:11:36,320 --> 00:11:38,760
Speaker 1: of the beholder. And I think it's not a coincidence,

196
00:11:38,840 --> 00:11:41,600
Speaker 1: and that like opening eye then also became like an

197
00:11:41,840 --> 00:11:45,560
Speaker 1: entity that was viewed by different people based on, you know,

198
00:11:45,600 --> 00:11:46,559
Speaker 1: the eye of the beholder.

199
00:11:47,200 --> 00:11:48,880
Speaker 3: It's kind of how he operates.

200
00:11:49,360 --> 00:11:52,040
Speaker 4: I can't help but feel as an analogy for how

201
00:11:52,200 --> 00:11:55,400
Speaker 4: chatbots talk with people of you know, sort of painting

202
00:11:55,480 --> 00:12:00,360
Speaker 4: like a zeitgeisty picture but not actually saying anything with

203
00:12:00,920 --> 00:12:02,520
Speaker 4: specific nouns or verbs.

204
00:12:04,040 --> 00:12:04,720
Speaker 3: Yes, a lot of.

205
00:12:04,720 --> 00:12:09,360
Speaker 1: People have drawn this analogy between how Samuel operates and

206
00:12:09,360 --> 00:12:10,920
Speaker 1: how chat GPT is designed.

207
00:12:12,559 --> 00:12:15,280
Speaker 2: I definitely find myself getting caught in a trap of saying,

208
00:12:15,760 --> 00:12:18,560
Speaker 2: is this AI good or is this AI bad? And

209
00:12:18,600 --> 00:12:20,600
Speaker 2: you actually offer a much more helpful way to think

210
00:12:20,600 --> 00:12:24,880
Speaker 2: about that does this fortify or does this dismantle empire?

211
00:12:25,280 --> 00:12:28,679
Speaker 2: And I'm curious how you got to that framework because

212
00:12:28,679 --> 00:12:29,520
Speaker 2: it is so helpful.

213
00:12:31,200 --> 00:12:33,840
Speaker 1: Yeah, this is based on I have a good friend

214
00:12:33,880 --> 00:12:37,840
Speaker 1: of mine who's a researcher who graduated from Stanford.

215
00:12:37,440 --> 00:12:40,200
Speaker 3: Ria Collerie, and she gave this really.

216
00:12:39,960 --> 00:12:43,319
Speaker 1: Amazing talk in twenty nineteen that basically articulated and framed

217
00:12:43,400 --> 00:12:46,800
Speaker 1: this question. It was a talk that was given at

218
00:12:46,960 --> 00:12:50,200
Speaker 1: Europe's the Neural Information Processing Systems Conference, which is the

219
00:12:50,440 --> 00:12:55,280
Speaker 1: largest AI research conference that happens every year, like fifteen

220
00:12:55,320 --> 00:12:58,760
Speaker 1: thousand researchers to send AI researchers to send on one

221
00:12:58,880 --> 00:13:02,800
Speaker 1: city and take over the city for a week. And

222
00:13:03,160 --> 00:13:05,920
Speaker 1: she was giving this keynote at Queer in AI, which

223
00:13:06,000 --> 00:13:11,920
Speaker 1: is this organization that of AI researchers and AI professionals

224
00:13:12,440 --> 00:13:16,000
Speaker 1: that identify as queer and want to find community and

225
00:13:16,040 --> 00:13:18,679
Speaker 1: also reflect their queerness in their work.

226
00:13:19,480 --> 00:13:21,680
Speaker 3: And the key like in her.

227
00:13:21,640 --> 00:13:26,400
Speaker 1: Keynote she said, you know, it's so hard as an

228
00:13:26,440 --> 00:13:29,360
Speaker 1: as a queer AI researcher, Like she was talking with

229
00:13:29,440 --> 00:13:31,360
Speaker 1: like everyone in the room, like as queer AI research

230
00:13:31,440 --> 00:13:37,880
Speaker 1: is it's so hard sometimes to align like your work

231
00:13:38,080 --> 00:13:42,520
Speaker 1: with your identity because you're taught in the AI research

232
00:13:42,600 --> 00:13:45,680
Speaker 1: world that you should only ever be using your technical

233
00:13:45,720 --> 00:13:50,480
Speaker 1: mind and you should leave behind anything that suggests that

234
00:13:50,559 --> 00:13:53,319
Speaker 1: you are also like a person with a body that

235
00:13:53,559 --> 00:13:58,240
Speaker 1: exists in society. And so even as people within the

236
00:13:58,240 --> 00:14:01,439
Speaker 1: community feel that there are certain things about AI research

237
00:14:01,520 --> 00:14:04,920
Speaker 1: that don't feel quite right, like it often can enable surveillance,

238
00:14:05,000 --> 00:14:08,760
Speaker 1: it often can then police queer bodies and things like that,

239
00:14:08,760 --> 00:14:11,960
Speaker 1: that somehow this this needs to be separate from from

240
00:14:12,040 --> 00:14:15,680
Speaker 1: their mind. And part of the problem she diagnosed was

241
00:14:15,760 --> 00:14:20,320
Speaker 1: because of this very overly simplistic question of people just

242
00:14:20,360 --> 00:14:24,080
Speaker 1: asking like is AI good or is AI bad? And

243
00:14:24,360 --> 00:14:28,720
Speaker 1: so her provocation was, this is an impossible question to answer,

244
00:14:28,800 --> 00:14:30,720
Speaker 1: and that's part of the reason why we just like

245
00:14:31,280 --> 00:14:33,640
Speaker 1: end up not grappling with this, and we just when

246
00:14:33,680 --> 00:14:35,840
Speaker 1: we're an AI researcher, we put on a totally different

247
00:14:35,840 --> 00:14:37,960
Speaker 1: hat and then we take it off to like exist

248
00:14:38,040 --> 00:14:42,560
Speaker 1: in society. And the better question is does this particular

249
00:14:42,760 --> 00:14:50,080
Speaker 1: AI tool or practice or artifact shift power in ways

250
00:14:50,080 --> 00:14:52,560
Speaker 1: that consolidate it in the hands of the few, or

251
00:14:52,600 --> 00:14:54,640
Speaker 1: in ways that distribute it in the hands of the many,

252
00:14:54,760 --> 00:15:00,560
Speaker 1: and once you are more granular about that question, then

253
00:15:00,600 --> 00:15:03,240
Speaker 1: it becomes a lot easier to figure out, like on

254
00:15:03,280 --> 00:15:04,880
Speaker 1: a day to day level, and she was talking to

255
00:15:04,920 --> 00:15:07,920
Speaker 1: AI researchers, but I think this is applicable to literally everyone,

256
00:15:08,000 --> 00:15:10,200
Speaker 1: Like it becomes more clear on a data day level,

257
00:15:10,240 --> 00:15:14,320
Speaker 1: like whether you want to engage in certain types of

258
00:15:14,360 --> 00:15:18,840
Speaker 1: AI technologies, either as a producer of this technology, like

259
00:15:19,040 --> 00:15:22,800
Speaker 1: tech companies rope us in as data donors to the

260
00:15:22,840 --> 00:15:26,960
Speaker 1: training of these AI models, as as hosters of their

261
00:15:27,040 --> 00:15:28,880
Speaker 1: data centers and so on and so forth, or as

262
00:15:28,880 --> 00:15:33,440
Speaker 1: a consumer of these technologies. And I it's it's a

263
00:15:33,480 --> 00:15:35,200
Speaker 1: talk that has always stuck with me in a question

264
00:15:35,240 --> 00:15:38,200
Speaker 1: that's always stuck with me because it just makes it

265
00:15:38,480 --> 00:15:44,200
Speaker 1: so clear when for not just AI, any technology, like

266
00:15:44,440 --> 00:15:48,880
Speaker 1: whether like I personally want to engage with it or not,

267
00:15:49,080 --> 00:15:51,080
Speaker 1: or like whether other people should be engaging with a

268
00:15:51,080 --> 00:15:55,680
Speaker 1: certain technology or not, because it makes like it immediately

269
00:15:55,760 --> 00:15:59,440
Speaker 1: clarifies the idea of that the technologies role in society

270
00:15:59,480 --> 00:16:02,240
Speaker 1: is it ultimate like fortifying a box heer wroting it.

271
00:16:03,280 --> 00:16:05,080
Speaker 2: Karen is one of the experts that I spoke to

272
00:16:05,120 --> 00:16:08,480
Speaker 2: while researching my own audiobook Love at First Prompt, a

273
00:16:08,560 --> 00:16:11,360
Speaker 2: project that also took me deep into the world of

274
00:16:11,360 --> 00:16:16,080
Speaker 2: people using AI for companionship and intimacy. Many of them

275
00:16:16,200 --> 00:16:19,760
Speaker 2: told me that they felt like AI had genuinely helped them.

276
00:16:20,200 --> 00:16:23,160
Speaker 2: So how do we hold that alongside everything we know

277
00:16:23,240 --> 00:16:27,320
Speaker 2: about the risks, the threats to our democracy, our environment,

278
00:16:27,440 --> 00:16:28,000
Speaker 2: and beyond.

279
00:16:29,560 --> 00:16:31,840
Speaker 4: In this audio project, we've talked with a lot of

280
00:16:31,880 --> 00:16:36,000
Speaker 4: people who individuals who personally get a lot of value

281
00:16:36,120 --> 00:16:41,320
Speaker 4: from their chatbot companions in terms of self reflection, emotional support,

282
00:16:41,640 --> 00:16:45,440
Speaker 4: overcoming interpersonal challenges with other humans in their life, and

283
00:16:45,960 --> 00:16:49,000
Speaker 4: other areas. And it's really interesting to think about the

284
00:16:49,080 --> 00:16:52,400
Speaker 4: distinction between those individual benefits that they feel and then

285
00:16:52,440 --> 00:16:57,320
Speaker 4: the societal impacts and what it means for democracy. I guess,

286
00:16:57,800 --> 00:17:00,200
Speaker 4: given everything that you know about these systems and how

287
00:17:00,200 --> 00:17:04,760
Speaker 4: they're built in who profits, how do you how do

288
00:17:04,800 --> 00:17:07,240
Speaker 4: you think about that? You know, sort of balancing those

289
00:17:07,400 --> 00:17:09,679
Speaker 4: those different levels of people who feel that they are

290
00:17:10,600 --> 00:17:16,240
Speaker 4: individually benefit and individually benefiting from sharing these intimate parts

291
00:17:16,280 --> 00:17:20,080
Speaker 4: of their lives versus the sort of more societal level

292
00:17:20,359 --> 00:17:25,720
Speaker 4: impacts to privacy, democracy, power, all of that.

293
00:17:26,760 --> 00:17:28,399
Speaker 1: Yeah, I guess I wouldn't even frame it as a

294
00:17:28,440 --> 00:17:31,560
Speaker 1: trade off of like the individual benefiting, but then society

295
00:17:32,080 --> 00:17:33,400
Speaker 1: having these.

296
00:17:35,000 --> 00:17:37,280
Speaker 3: These negative externalities.

297
00:17:37,320 --> 00:17:41,600
Speaker 1: Like I think the individual themselves is also potentially going

298
00:17:41,600 --> 00:17:45,119
Speaker 1: to have negative impacts from their actions as well. Like

299
00:17:45,160 --> 00:17:48,720
Speaker 1: in the short term, when people are developing relationships with

300
00:17:48,760 --> 00:17:53,040
Speaker 1: these chatbots and then divulging super intimate information, maybe they

301
00:17:53,080 --> 00:17:56,879
Speaker 1: do in that moment feel like they gained something important

302
00:17:56,960 --> 00:18:02,640
Speaker 1: from that specific interaction, but they also lost an extraordinary

303
00:18:02,640 --> 00:18:04,879
Speaker 1: amount of privacy in that moment that could in the

304
00:18:04,880 --> 00:18:07,840
Speaker 1: long run come back to bite them, you know. Like

305
00:18:09,080 --> 00:18:12,959
Speaker 1: One of the things that I think is very distinctive

306
00:18:13,040 --> 00:18:16,639
Speaker 1: about these chatbots versus about something like Google Search is

307
00:18:16,680 --> 00:18:19,760
Speaker 1: like in the past, people were not uploading their medical

308
00:18:19,800 --> 00:18:23,800
Speaker 1: files to Google Search, but they do just seamlessly upload

309
00:18:23,800 --> 00:18:27,040
Speaker 1: their medical files to chattbut thinking that this is going

310
00:18:27,040 --> 00:18:29,119
Speaker 1: to ultimately give them some benefit in the short term,

311
00:18:29,640 --> 00:18:33,560
Speaker 1: and there's not really like any guardrails right now for

312
00:18:33,680 --> 00:18:36,400
Speaker 1: where that information is going to go. I mean, company

313
00:18:36,640 --> 00:18:40,680
Speaker 1: like and in fact Opening Ananthropic very recently rolled out

314
00:18:41,880 --> 00:18:46,240
Speaker 1: healthcare features to continue to encourage people to upload this stuff,

315
00:18:46,280 --> 00:18:49,199
Speaker 1: and they say in their advertisements, this is going to

316
00:18:49,240 --> 00:18:51,200
Speaker 1: be a place for you to upload all of your

317
00:18:51,240 --> 00:18:55,880
Speaker 1: medical records to our platforms, and for now they say,

318
00:18:56,320 --> 00:18:58,640
Speaker 1: we do not use any of this information for training.

319
00:18:58,840 --> 00:19:01,320
Speaker 1: We're going to cure it in a different way than

320
00:19:01,359 --> 00:19:04,639
Speaker 1: your other conversations. But this is completely based on trust

321
00:19:04,680 --> 00:19:08,080
Speaker 1: in these companies, and who's to say for that user

322
00:19:08,119 --> 00:19:11,240
Speaker 1: in the long run how those policies might change, and

323
00:19:11,280 --> 00:19:14,440
Speaker 1: then suddenly all of the intimate information that they've provided

324
00:19:14,720 --> 00:19:18,920
Speaker 1: is used against them.

325
00:19:19,080 --> 00:19:35,120
Speaker 2: Let's take a quick break at our back. Earlier this year,

326
00:19:35,240 --> 00:19:39,119
Speaker 2: open ai announced they'd be rolling out chatcheapt Health, a

327
00:19:39,160 --> 00:19:42,160
Speaker 2: new service that would encourage people to upload their medical

328
00:19:42,200 --> 00:19:45,919
Speaker 2: records and create an ecosystem for other health apps like

329
00:19:45,960 --> 00:19:50,040
Speaker 2: fitness trackers to interface with it. Open ai it's hyping

330
00:19:50,080 --> 00:19:52,800
Speaker 2: this up as a one stop shop for integrating all

331
00:19:52,880 --> 00:19:56,440
Speaker 2: data about your health. According to the company, more than

332
00:19:56,480 --> 00:20:00,240
Speaker 2: forty million people ask chatcheapt a healthcare related question every

333
00:20:00,240 --> 00:20:03,399
Speaker 2: single day, which amounts to more than five percent of

334
00:20:03,480 --> 00:20:07,960
Speaker 2: all global messages on the platform. So why not create

335
00:20:08,000 --> 00:20:11,080
Speaker 2: a dedicated tab for people's health questions and health needs?

336
00:20:11,560 --> 00:20:15,719
Speaker 2: What could go wrong? Right? I saw this woman on

337
00:20:15,800 --> 00:20:18,600
Speaker 2: threads talking about how she she had screenshot at a

338
00:20:18,600 --> 00:20:21,399
Speaker 2: message she had gotten from her doctor. Her cancer doctor.

339
00:20:21,640 --> 00:20:24,840
Speaker 2: That was just this dashed off sentence, right. It was like, oh,

340
00:20:24,920 --> 00:20:27,840
Speaker 2: scans came back, let's discuss next appointment. And she was like, oh,

341
00:20:27,840 --> 00:20:29,960
Speaker 2: my next appointment is in three months. Then it was

342
00:20:30,359 --> 00:20:32,800
Speaker 2: the image that she had she had put that put

343
00:20:32,880 --> 00:20:36,680
Speaker 2: her scans into chatjeb tea and Chatgebet was like, I'm

344
00:20:36,720 --> 00:20:40,000
Speaker 2: really sorry you're dealing with this. This can be really scary.

345
00:20:40,080 --> 00:20:42,520
Speaker 2: I think your tests indicate da da da da. And

346
00:20:43,119 --> 00:20:46,960
Speaker 2: I thought, boy, is this company really exploiting people who

347
00:20:47,160 --> 00:20:51,879
Speaker 2: are fearful or vulnerable or scared or you know, up

348
00:20:51,920 --> 00:20:55,960
Speaker 2: against navigating a healthcare system that can sometimes be awful?

349
00:20:56,280 --> 00:21:01,080
Speaker 2: Are there ways that tech companies are offering people something

350
00:21:01,520 --> 00:21:03,639
Speaker 2: saying this is going to be helpful for you, and

351
00:21:03,680 --> 00:21:07,120
Speaker 2: really they're just benefiting from a bad system we're all

352
00:21:07,119 --> 00:21:07,960
Speaker 2: trying to navigate.

353
00:21:08,880 --> 00:21:11,399
Speaker 1: I've been thinking about this a lot with the rolling

354
00:21:11,400 --> 00:21:14,440
Speaker 1: out of the healthcare features, in particular, because I've talked

355
00:21:14,480 --> 00:21:16,960
Speaker 1: with so many people who have had exactly this experience,

356
00:21:17,000 --> 00:21:19,600
Speaker 1: like they get a diagnosis for themselves or for a

357
00:21:19,640 --> 00:21:22,600
Speaker 1: loved one, and all of a sudden, like everything feels

358
00:21:22,600 --> 00:21:28,360
Speaker 1: really overwhelming. And interestingly, I almost exclusively hear this among Americans.

359
00:21:28,760 --> 00:21:33,040
Speaker 1: This experience so one, Like it's also tied to the

360
00:21:33,080 --> 00:21:36,840
Speaker 1: fact that, like we have this healthcare system that is

361
00:21:36,880 --> 00:21:40,320
Speaker 1: just impenetrable and horrible to navigate and makes you feel

362
00:21:40,320 --> 00:21:43,840
Speaker 1: really isolated and makes you like start worrying about your

363
00:21:43,880 --> 00:21:48,280
Speaker 1: finances and everything. And there is like something to be

364
00:21:48,359 --> 00:21:52,800
Speaker 1: said that like, in this moment of great need, there

365
00:21:52,840 --> 00:21:55,119
Speaker 1: suddenly is this tool that appears that like helps you

366
00:21:55,200 --> 00:21:57,400
Speaker 1: sort through that and helps you navigate that, and and

367
00:21:57,480 --> 00:22:03,439
Speaker 1: like there's I could never say that these people should

368
00:22:03,720 --> 00:22:06,640
Speaker 1: not be using a tool that's that could make such

369
00:22:06,640 --> 00:22:11,200
Speaker 1: a convoluted process extremely at least a little bit more

370
00:22:13,720 --> 00:22:19,880
Speaker 1: helpful and navigable. But on like these companies know, that's

371
00:22:19,920 --> 00:22:24,000
Speaker 1: exactly what's happening. Like they are kind of tapping into

372
00:22:24,040 --> 00:22:28,000
Speaker 1: these moments of vulnerability to get you to develop more

373
00:22:28,080 --> 00:22:32,480
Speaker 1: dependence on their tools, and not just dependence in terms

374
00:22:32,480 --> 00:22:37,560
Speaker 1: of like uh, you know, just like the practical dependence,

375
00:22:37,560 --> 00:22:41,800
Speaker 1: but like emotional dependence as well, right, Like they specifically

376
00:22:41,840 --> 00:22:46,119
Speaker 1: design the tools to to pepper in those comments like

377
00:22:46,200 --> 00:22:49,040
Speaker 1: oh this is so hard, like let me let me

378
00:22:49,080 --> 00:22:51,520
Speaker 1: help you with that, let me I'm here for you,

379
00:22:52,040 --> 00:22:54,640
Speaker 1: what you know, and like those are all design decisions

380
00:22:54,680 --> 00:22:58,200
Speaker 1: that they layer in to make this a holistic part

381
00:22:58,200 --> 00:23:01,120
Speaker 1: of the experience. So I think the way that I

382
00:23:01,240 --> 00:23:03,280
Speaker 1: kind of fall on this issue is that, like, it's

383
00:23:03,320 --> 00:23:09,320
Speaker 1: not that these tools shouldn't exist, and clearly because the

384
00:23:09,440 --> 00:23:14,760
Speaker 1: numbers show that, like people just like there are a

385
00:23:14,800 --> 00:23:17,560
Speaker 1: lot of people that use these tools in this way.

386
00:23:17,680 --> 00:23:21,879
Speaker 1: So it's to me, like the solution is we should

387
00:23:21,920 --> 00:23:23,960
Speaker 1: have these tools, but they need to be developed safely,

388
00:23:24,080 --> 00:23:25,880
Speaker 1: and part of this is going to be that they

389
00:23:25,880 --> 00:23:29,439
Speaker 1: have to be regulated. Like most of the medical system

390
00:23:29,600 --> 00:23:34,000
Speaker 1: and most ways that people interface with the healthcare system

391
00:23:34,040 --> 00:23:36,960
Speaker 1: in general are heavily regulated, like the medications that you

392
00:23:37,000 --> 00:23:39,160
Speaker 1: take every time you go to the doctor, like all

393
00:23:39,160 --> 00:23:44,040
Speaker 1: of your exchanges are protected by the law. And this

394
00:23:44,200 --> 00:23:47,359
Speaker 1: is like a weird moment in which suddenly these companies

395
00:23:47,400 --> 00:23:51,040
Speaker 1: are offering many of the same services but without any

396
00:23:51,119 --> 00:23:55,119
Speaker 1: of the protections, and there of course also at the

397
00:23:55,119 --> 00:24:00,439
Speaker 1: same time like lobbying against getting those protections implemented. But yeah, like,

398
00:24:00,480 --> 00:24:02,560
Speaker 1: to me, the healthy balance that we should be trying

399
00:24:02,600 --> 00:24:05,640
Speaker 1: to move towards is that we have these tools available

400
00:24:05,680 --> 00:24:08,320
Speaker 1: for people to help them when they desperately need it,

401
00:24:08,400 --> 00:24:11,920
Speaker 1: but also in a way that is safe, where people

402
00:24:12,000 --> 00:24:14,800
Speaker 1: are also not getting harmed along the way.

403
00:24:15,520 --> 00:24:17,359
Speaker 2: That's such a good point, right, that so much of

404
00:24:17,359 --> 00:24:19,720
Speaker 2: the healthcare industry and the way that we experience it

405
00:24:19,760 --> 00:24:22,720
Speaker 2: is heavily regulated, and now all of a sudden, we've

406
00:24:22,760 --> 00:24:26,600
Speaker 2: got these completely unregulated AI companies and products they make

407
00:24:26,760 --> 00:24:30,720
Speaker 2: flooding that space. Like you said, people's high level of

408
00:24:30,760 --> 00:24:33,439
Speaker 2: interest using AI to help manage their health care needs

409
00:24:33,680 --> 00:24:35,920
Speaker 2: does suggest that it might be filling a need that

410
00:24:35,960 --> 00:24:39,200
Speaker 2: people have for more information about their health, more support.

411
00:24:39,640 --> 00:24:44,520
Speaker 2: But there are obviously so many huge risks. Risks to privacy,

412
00:24:44,760 --> 00:24:47,320
Speaker 2: risks of getting bad or just flied out incorrect or

413
00:24:47,320 --> 00:24:51,679
Speaker 2: even dangerous advice, risks of being exploited during a vulnerable moment,

414
00:24:51,760 --> 00:24:54,280
Speaker 2: you know, just to name a few. So I'm curious

415
00:24:54,320 --> 00:24:56,640
Speaker 2: in your mind, what would a better system look like?

416
00:24:56,640 --> 00:24:56,720
Speaker 3: Like?

417
00:24:56,760 --> 00:24:59,600
Speaker 2: How should regulators and lawmakers be thinking of the role

418
00:24:59,640 --> 00:25:00,800
Speaker 2: of AI and healthcare.

419
00:25:01,320 --> 00:25:03,720
Speaker 1: The way that I think about it is like, first

420
00:25:03,720 --> 00:25:07,160
Speaker 1: and foremost, we should be thinking much more broadly about

421
00:25:07,160 --> 00:25:09,480
Speaker 1: what constitutes AI regulation. I think most of the time

422
00:25:09,520 --> 00:25:13,680
Speaker 1: when people think about AI regulation, they're imagining just regulating

423
00:25:14,280 --> 00:25:19,440
Speaker 1: the applications once AI is developed and how it's allowed

424
00:25:19,480 --> 00:25:23,119
Speaker 1: to be used. I think that AI regulation needs to

425
00:25:23,119 --> 00:25:25,720
Speaker 1: be brought into also think about how AI, what kinds

426
00:25:25,720 --> 00:25:28,679
Speaker 1: of AI should be created in the first place. So

427
00:25:29,280 --> 00:25:32,439
Speaker 1: that means like we should be having more regulation on

428
00:25:32,480 --> 00:25:35,200
Speaker 1: the data that's allowed to go into these AI models,

429
00:25:35,359 --> 00:25:39,120
Speaker 1: on where data centers get developed for training these models,

430
00:25:39,280 --> 00:25:42,280
Speaker 1: how much energy and water are they allowed to use,

431
00:25:42,320 --> 00:25:44,119
Speaker 1: and how much are they allowed to hike up the

432
00:25:44,240 --> 00:25:49,000
Speaker 1: utility prices of customers while they're training these models. And

433
00:25:49,400 --> 00:25:52,800
Speaker 1: I also think that we should be regulating the applications

434
00:25:52,800 --> 00:25:56,280
Speaker 1: and that, you know, if they're going to enter into

435
00:25:56,320 --> 00:25:59,160
Speaker 1: the healthcare industry and people are going to start using

436
00:25:59,200 --> 00:26:03,280
Speaker 1: these tools as a therapist, I mean usually a human

437
00:26:03,320 --> 00:26:07,720
Speaker 1: therapist has to get licensed by a body and has

438
00:26:07,760 --> 00:26:13,479
Speaker 1: to be recognized as actually able to practice therapy. And

439
00:26:13,600 --> 00:26:17,879
Speaker 1: so that's another piece of like for that specific industry.

440
00:26:18,040 --> 00:26:21,760
Speaker 1: If AI companies are going to position their products as therapists,

441
00:26:22,560 --> 00:26:25,840
Speaker 1: they should be regulated just or they should they should

442
00:26:25,840 --> 00:26:30,000
Speaker 1: be licensed just as human therapists are licensed. So it

443
00:26:30,080 --> 00:26:33,159
Speaker 1: really does, I think, depend a case by case on

444
00:26:34,040 --> 00:26:39,000
Speaker 1: which you know the vast facet faces of AI that

445
00:26:39,040 --> 00:26:42,439
Speaker 1: we're talking about, But one of the things that I

446
00:26:42,480 --> 00:26:47,840
Speaker 1: think would cut across like every just as a baseline

447
00:26:47,840 --> 00:26:49,960
Speaker 1: we should be thinking about when it comes to a

448
00:26:50,119 --> 00:26:54,320
Speaker 1: regulation is we just generally need more transparency across the

449
00:26:54,440 --> 00:26:57,720
Speaker 1: entire age of element and deployment supply chain. Like we

450
00:26:57,920 --> 00:27:01,240
Speaker 1: currently don't know what day is being used to train

451
00:27:01,359 --> 00:27:04,040
Speaker 1: these models. We currently don't know the energy footprint of

452
00:27:04,080 --> 00:27:07,560
Speaker 1: these data centers. We often don't know when you're going

453
00:27:07,600 --> 00:27:10,119
Speaker 1: to the doctor's office whether or not AI is actually

454
00:27:10,119 --> 00:27:15,959
Speaker 1: being used on you because there aren't super robust disclosure

455
00:27:16,040 --> 00:27:19,960
Speaker 1: laws where doctors or anyone who's using AI has to

456
00:27:20,480 --> 00:27:23,040
Speaker 1: disclose that they are doing so on the person that

457
00:27:23,040 --> 00:27:25,679
Speaker 1: they're using it on. We also don't know, like if

458
00:27:25,720 --> 00:27:28,840
Speaker 1: you're at the doctor's office and they're using a specific

459
00:27:28,960 --> 00:27:33,600
Speaker 1: AI application, what's actually running in the background. Because there's

460
00:27:33,680 --> 00:27:36,760
Speaker 1: applications and then there's the models that power them, and

461
00:27:37,080 --> 00:27:39,399
Speaker 1: it could turn out that that model is run by

462
00:27:39,440 --> 00:27:42,600
Speaker 1: Google or by Microsoft or by Amazon or whatever it is,

463
00:27:43,320 --> 00:27:47,560
Speaker 1: and you currently don't have control over that as a

464
00:27:47,760 --> 00:27:51,560
Speaker 1: patient or just as a person existing in society and

465
00:27:51,920 --> 00:27:57,800
Speaker 1: in general, as like the first step for improving the

466
00:27:57,920 --> 00:28:03,800
Speaker 1: rights of users and citizens engaging in an high enabled society,

467
00:28:04,000 --> 00:28:05,720
Speaker 1: like transparency is number one.

468
00:28:06,119 --> 00:28:08,240
Speaker 2: When you put it that way, it is really crazy

469
00:28:08,240 --> 00:28:12,639
Speaker 2: that we're tolerating this. Like I like it is. If

470
00:28:13,400 --> 00:28:15,600
Speaker 2: if we were to go back in time twenty years

471
00:28:15,640 --> 00:28:17,439
Speaker 2: and you were to explain to me what would be

472
00:28:17,520 --> 00:28:20,359
Speaker 2: commonplace just the way that you just did, I would

473
00:28:20,359 --> 00:28:22,679
Speaker 2: say no, that we wouldn't all, we wouldn't stand for that.

474
00:28:22,720 --> 00:28:25,560
Speaker 2: Nobody would volunteer or sign up for that. I don't

475
00:28:25,560 --> 00:28:29,000
Speaker 2: know sometimes when I hear what these tech the way

476
00:28:29,000 --> 00:28:31,760
Speaker 2: that these tech companies are framing that they'll be in

477
00:28:31,880 --> 00:28:35,400
Speaker 2: the most intimate aspects of our lives, and we would

478
00:28:35,520 --> 00:28:38,080
Speaker 2: get no transparency into way that that into the way

479
00:28:38,120 --> 00:28:39,760
Speaker 2: that that shows up and what that really looks like

480
00:28:39,800 --> 00:28:42,280
Speaker 2: and means. I it's it blows my mind, Like it's

481
00:28:42,280 --> 00:28:44,160
Speaker 2: it's really hard for me to believe that we have

482
00:28:44,360 --> 00:28:46,520
Speaker 2: signed up for this and that there's people who are like, oh,

483
00:28:46,560 --> 00:28:47,959
Speaker 2: and it's going to be great for you. We were

484
00:28:47,960 --> 00:28:48,840
Speaker 2: actually gonna love it.

485
00:28:49,440 --> 00:28:51,120
Speaker 1: Yeah, I mean, I think this is one of the

486
00:28:51,160 --> 00:28:54,400
Speaker 1: things that like this didn't happen overnight, right like Silicon

487
00:28:54,480 --> 00:28:57,960
Speaker 1: Valley in general, over the last ten years very slowly

488
00:28:58,120 --> 00:29:02,360
Speaker 1: built up our tolerance for this situation.

489
00:29:02,160 --> 00:29:05,600
Speaker 3: By building this narrative that like.

490
00:29:05,760 --> 00:29:09,360
Speaker 1: Privacy is over and everyone would rather trade their privacy

491
00:29:09,400 --> 00:29:14,040
Speaker 1: for convenience, and by making us accustomed to the idea

492
00:29:14,160 --> 00:29:19,160
Speaker 1: that like, we don't have any control over the fundamental

493
00:29:19,400 --> 00:29:22,240
Speaker 1: building blocks of our digital lives anymore, like the things

494
00:29:22,280 --> 00:29:24,960
Speaker 1: that give us information, the algorithms that sort through our

495
00:29:24,960 --> 00:29:31,360
Speaker 1: news seats. And so I think the AI conversation is

496
00:29:31,560 --> 00:29:36,200
Speaker 1: being installed on already like a very solid foundation of

497
00:29:36,360 --> 00:29:41,040
Speaker 1: ten years of Silicon Valley. Just eroding away our rights

498
00:29:41,080 --> 00:29:44,200
Speaker 1: and making us is like a frog and a boiling

499
00:29:44,240 --> 00:29:45,360
Speaker 1: pot metaphor.

500
00:29:45,560 --> 00:29:48,560
Speaker 2: Yeah, you know. You open the book with this stunning

501
00:29:48,680 --> 00:29:50,720
Speaker 2: quote from Sam Altman, I think his blog where he

502
00:29:50,800 --> 00:29:53,200
Speaker 2: essentially says that the best way to motivate people is

503
00:29:53,280 --> 00:29:55,040
Speaker 2: to build a religion, and the best way to do

504
00:29:55,080 --> 00:29:57,120
Speaker 2: that is to start a company. And first of all,

505
00:29:57,120 --> 00:29:59,760
Speaker 2: I feel like starting with that quote is like, Okay,

506
00:30:00,040 --> 00:30:03,040
Speaker 2: so here's where we're at, Like we're starting with him

507
00:30:03,120 --> 00:30:05,720
Speaker 2: talking about how he wants to start a religion. But

508
00:30:05,800 --> 00:30:09,800
Speaker 2: I think when you really parse through what that means

509
00:30:09,800 --> 00:30:12,120
Speaker 2: and you consider that open ay is building tools that

510
00:30:12,160 --> 00:30:15,520
Speaker 2: people use in these intimate parts of their life now,

511
00:30:15,560 --> 00:30:19,120
Speaker 2: for things like emotional support, intimate companionship, therapy, sex, all

512
00:30:19,200 --> 00:30:22,719
Speaker 2: of that stuff they're using these tools in moments of

513
00:30:23,040 --> 00:30:27,760
Speaker 2: spiritual or existential need and vulnerability. What does it mean

514
00:30:27,920 --> 00:30:31,080
Speaker 2: that they're being built and led by somebody who, as

515
00:30:31,120 --> 00:30:35,200
Speaker 2: you point out, will use this like mission driven language

516
00:30:35,280 --> 00:30:38,680
Speaker 2: as a strategic tool as opposed to like a genuine commitment.

517
00:30:38,720 --> 00:30:41,160
Speaker 2: What does it mean that somebody is at the helm

518
00:30:41,200 --> 00:30:44,360
Speaker 2: of all of these intimate things who you know, there

519
00:30:44,400 --> 00:30:49,360
Speaker 2: are plenty of concerns about his honesty and his approaches

520
00:30:49,400 --> 00:30:51,080
Speaker 2: to safety, Like what do we do with that?

521
00:30:52,080 --> 00:30:55,360
Speaker 1: I think for me there's an even bigger meta question,

522
00:30:55,440 --> 00:30:57,400
Speaker 1: which is just like what does it mean that we

523
00:30:57,440 --> 00:31:00,640
Speaker 1: allow one individual, regardless of their character, should do all

524
00:31:00,640 --> 00:31:02,840
Speaker 1: these things and have access to all these things, and

525
00:31:04,440 --> 00:31:06,840
Speaker 1: have like a window, an intimate window into so many

526
00:31:06,840 --> 00:31:10,080
Speaker 1: people's lives. Like even if we were to swap out

527
00:31:10,120 --> 00:31:12,360
Speaker 1: Sam Molmon for someone else, I still don't think this

528
00:31:12,480 --> 00:31:15,800
Speaker 1: setup is okay, Right, Like, there's just one person or

529
00:31:15,840 --> 00:31:18,160
Speaker 1: a small group of people that are running these companies,

530
00:31:18,800 --> 00:31:22,160
Speaker 1: and they can make sixty decisions in an hour That

531
00:31:22,320 --> 00:31:28,400
Speaker 1: fundamentally changes how billions of people are engaging with their

532
00:31:28,400 --> 00:31:31,080
Speaker 1: technologies on a day to day basis, which then affects

533
00:31:31,160 --> 00:31:34,920
Speaker 1: their lives and their work and their schools. And their healthcare.

534
00:31:35,720 --> 00:31:40,560
Speaker 1: And you know, it's just like it's like that is

535
00:31:40,680 --> 00:31:47,200
Speaker 1: just inherently unsound. That is inherently an undemocratic setup, and

536
00:31:47,520 --> 00:31:51,120
Speaker 1: we cannot have we cannot be talking about, you know,

537
00:31:51,640 --> 00:31:55,520
Speaker 1: existing in a democratic society where everyone has is supposed

538
00:31:55,560 --> 00:31:58,719
Speaker 1: to have agency and control over the future decisions and

539
00:31:58,760 --> 00:32:03,000
Speaker 1: be able to collectively self overned, when in reality governance.

540
00:32:02,600 --> 00:32:05,440
Speaker 3: Is just happening by a tiny group of people.

541
00:32:06,360 --> 00:32:10,680
Speaker 4: Yeah. I really appreciate how much your writing connects with democracy.

542
00:32:10,720 --> 00:32:13,160
Speaker 4: I feel like a lot of the uh, the experts

543
00:32:13,200 --> 00:32:15,840
Speaker 4: that we've talked with about this are uh experts in

544
00:32:16,040 --> 00:32:20,360
Speaker 4: different fields, but they've been you know, it's been a

545
00:32:20,360 --> 00:32:23,720
Speaker 4: lot of discussion about uh. A lot of the discussions

546
00:32:23,720 --> 00:32:26,480
Speaker 4: we've had have been about intimacy and relationships.

547
00:32:26,720 --> 00:32:27,560
Speaker 1: UH.

548
00:32:28,080 --> 00:32:31,920
Speaker 4: But as you point out, I think democracy and the

549
00:32:31,920 --> 00:32:38,640
Speaker 4: implications for democracy of this these rising AI companies is enormous.

550
00:32:39,520 --> 00:32:42,040
Speaker 2: Do you do you view that.

551
00:32:42,320 --> 00:32:47,720
Speaker 4: You know, democracy as like a casualty of these companies

552
00:32:48,000 --> 00:32:53,280
Speaker 4: building building empires, or do you think it's sort of

553
00:32:53,360 --> 00:32:57,720
Speaker 4: part and parcel of the same thing, undermining democracy and

554
00:32:58,800 --> 00:33:02,040
Speaker 4: you know, as part of this process of consolidating power

555
00:33:02,080 --> 00:33:02,520
Speaker 4: and money.

556
00:33:03,200 --> 00:33:05,200
Speaker 3: I think it's kind of a little bit of both.

557
00:33:05,880 --> 00:33:08,640
Speaker 1: I mean, there's certain players in Silicon Valley that have

558
00:33:08,760 --> 00:33:12,080
Speaker 1: been extremely explicit about how they just don't believe in

559
00:33:12,120 --> 00:33:15,000
Speaker 1: democracy and think that we need to move to a

560
00:33:15,040 --> 00:33:18,040
Speaker 1: different way of organizing society. Like Peter Teel has said

561
00:33:18,040 --> 00:33:21,200
Speaker 1: this very explicitly. He thinks that democracy is incompatible with

562
00:33:23,200 --> 00:33:28,880
Speaker 1: like a good society. And also there are other people

563
00:33:28,880 --> 00:33:31,600
Speaker 1: that I think have who have not been so explicit

564
00:33:31,760 --> 00:33:35,680
Speaker 1: about their desire to undermine democracy, but more in their

565
00:33:36,000 --> 00:33:41,720
Speaker 1: quest for you know, their quest for wanting to scale

566
00:33:42,080 --> 00:33:46,160
Speaker 1: monopolistic companies, their quest for wanting to build an AI

567
00:33:46,200 --> 00:33:52,160
Speaker 1: like God, their quest for wanting to become like a

568
00:33:52,200 --> 00:33:52,880
Speaker 1: great man of.

569
00:33:52,920 --> 00:33:56,200
Speaker 3: History like in that journey.

570
00:33:56,440 --> 00:33:59,800
Speaker 1: Yes, democracy also then becomes a casualty because in order

571
00:33:59,840 --> 00:34:03,520
Speaker 1: to do those things, what they're suggesting is consuming all

572
00:34:03,520 --> 00:34:06,520
Speaker 1: of the resources in the world and undermining everyone's agency

573
00:34:06,560 --> 00:34:11,959
Speaker 1: to get there. So that, yes, so you could say

574
00:34:11,960 --> 00:34:15,840
Speaker 1: that maybe therefore there's not actually a difference whether or

575
00:34:15,840 --> 00:34:19,080
Speaker 1: not they intended explicitly to undermine democracy or not, Like

576
00:34:21,160 --> 00:34:25,400
Speaker 1: that's the final destination of what they what their agendas are.

577
00:34:26,080 --> 00:34:29,040
Speaker 2: Yeah, to that point, you've talked about how open AI

578
00:34:29,360 --> 00:34:33,680
Speaker 2: their core insight is just the concept of more, more data,

579
00:34:34,000 --> 00:34:38,879
Speaker 2: more resources. But when I think about them pivoting more

580
00:34:38,920 --> 00:34:41,480
Speaker 2: into like an intimacy or emotional support space, which I

581
00:34:41,480 --> 00:34:43,399
Speaker 2: know that they're aware that that's how people are using

582
00:34:43,400 --> 00:34:46,920
Speaker 2: their product, it seems like those are spaces that require

583
00:34:47,080 --> 00:34:52,239
Speaker 2: the opposite of more right context care, specificity, nuance. These

584
00:34:52,280 --> 00:34:55,160
Speaker 2: things that are that are so kind of they're they're

585
00:34:55,239 --> 00:34:57,120
Speaker 2: at odds with the concept of just like more and

586
00:34:57,120 --> 00:34:59,719
Speaker 2: more and more all the time. Is there a tension there?

587
00:35:00,520 --> 00:35:04,040
Speaker 1: It's funny because I don't think from the AI researchers perspective,

588
00:35:04,040 --> 00:35:06,560
Speaker 1: there's attention. They would say that in order to have

589
00:35:06,760 --> 00:35:11,280
Speaker 1: more to deliver more high context care.

590
00:35:12,600 --> 00:35:14,320
Speaker 3: Via chatbot, they.

591
00:35:14,200 --> 00:35:15,840
Speaker 1: Just need more and more and more and more of

592
00:35:15,920 --> 00:35:20,560
Speaker 1: your intimate data. So so like for them, it's not

593
00:35:20,920 --> 00:35:26,160
Speaker 1: attention and that's why they are motivated to collect more

594
00:35:26,200 --> 00:35:27,560
Speaker 1: and more and more of the stuff.

595
00:35:30,200 --> 00:35:33,640
Speaker 3: But yeah, I mean, like, to me, what thettension is.

596
00:35:33,320 --> 00:35:36,520
Speaker 1: Is that there is a certain worldview that undergrads the

597
00:35:36,560 --> 00:35:40,399
Speaker 1: development of these systems that suggests that everything in our

598
00:35:40,440 --> 00:35:46,240
Speaker 1: existence can ultimately be quantified through data and can ultimately be.

599
00:35:48,320 --> 00:35:50,480
Speaker 3: Mediated through technical systems.

600
00:35:51,120 --> 00:35:57,480
Speaker 1: And that worldview really devalues human to human interaction, really

601
00:35:57,520 --> 00:36:05,600
Speaker 1: devalues the soft social, emotional element of relationships and society,

602
00:36:06,440 --> 00:36:10,160
Speaker 1: and that is the tension for me, is like they're

603
00:36:10,280 --> 00:36:14,839
Speaker 1: trying to suggest that there will be an AI god

604
00:36:14,880 --> 00:36:19,319
Speaker 1: that they can developed purely through quantitative methods, purely through

605
00:36:19,360 --> 00:36:23,840
Speaker 1: consuming data in technical systems, that is somehow going to

606
00:36:23,880 --> 00:36:28,440
Speaker 1: manage to replace all of the facets of emotional social

607
00:36:29,200 --> 00:36:35,560
Speaker 1: human relationships that have you know, been driving society for millennia.

608
00:36:36,360 --> 00:36:39,680
Speaker 2: And this just goes back to my ultimate personal orientation

609
00:36:39,800 --> 00:36:42,279
Speaker 2: around tech, which is that I think that it is

610
00:36:42,320 --> 00:36:48,800
Speaker 2: being led and driven by people who devalue emotional labor,

611
00:36:49,000 --> 00:36:52,799
Speaker 2: you know, soft skills. Like we've had conversations for a

612
00:36:52,840 --> 00:36:55,960
Speaker 2: really long time about how the on living that matters

613
00:36:56,000 --> 00:37:00,239
Speaker 2: are these like hard skills, quantifiable skills. And there are

614
00:37:00,239 --> 00:37:02,440
Speaker 2: people who are trying who were meant to trust are

615
00:37:02,440 --> 00:37:04,799
Speaker 2: going to build the kind of world that we want

616
00:37:04,880 --> 00:37:08,640
Speaker 2: to live in using their technology, who have never valued

617
00:37:08,640 --> 00:37:11,239
Speaker 2: these things, and that they're building technology that sort of

618
00:37:11,800 --> 00:37:14,880
Speaker 2: argues that you don't even really need them, when we

619
00:37:15,000 --> 00:37:17,520
Speaker 2: know that we do need care work and emotional work.

620
00:37:17,520 --> 00:37:19,799
Speaker 2: It's worked that they devalue and do not respect. So

621
00:37:19,920 --> 00:37:22,080
Speaker 2: why then would I put my trust in them to

622
00:37:22,200 --> 00:37:24,320
Speaker 2: use technology to build the kind of world that anybody

623
00:37:24,360 --> 00:37:25,480
Speaker 2: might want to actually live in.

624
00:37:26,560 --> 00:37:29,600
Speaker 1: Yeah, and the thing is like they also actually value

625
00:37:29,640 --> 00:37:35,080
Speaker 1: these things without explicitly recognized, Like they actually value it

626
00:37:35,200 --> 00:37:39,080
Speaker 1: more than they articulate, because, for example, all these tech

627
00:37:39,080 --> 00:37:42,160
Speaker 1: companies have policies that require their employees to come into

628
00:37:42,160 --> 00:37:46,000
Speaker 1: the office, Like, why would you require your employees to

629
00:37:46,120 --> 00:37:48,920
Speaker 1: be in the same room with one another working face

630
00:37:48,960 --> 00:37:52,640
Speaker 1: to face unless you believed that there is genuine value

631
00:37:52,800 --> 00:37:57,040
Speaker 1: to that interaction that you could not get from being remote.

632
00:37:57,719 --> 00:38:00,080
Speaker 1: It's like kind of like the greatest irony is that

633
00:38:01,040 --> 00:38:05,359
Speaker 1: they actually design their companies and the way that they

634
00:38:05,400 --> 00:38:10,040
Speaker 1: work and the way that they make decisions in their

635
00:38:10,040 --> 00:38:13,920
Speaker 1: lives in direct opposition to the things that they say.

636
00:38:16,080 --> 00:38:19,120
Speaker 1: And I think it also, you know, like that the

637
00:38:19,200 --> 00:38:21,440
Speaker 1: Neurop's conference that I was talking about, like fifteen thousand

638
00:38:21,440 --> 00:38:24,319
Speaker 1: people descending into one city just to have face to

639
00:38:24,400 --> 00:38:25,200
Speaker 1: face time.

640
00:38:25,080 --> 00:38:25,760
Speaker 3: With one another.

641
00:38:26,280 --> 00:38:28,839
Speaker 1: And these are the people that are building these technologies

642
00:38:28,880 --> 00:38:31,880
Speaker 1: that they then suggest will be replacing all of that

643
00:38:31,960 --> 00:38:37,759
Speaker 1: human interaction. So yeah, it's interesting. I think it's one

644
00:38:37,760 --> 00:38:40,359
Speaker 1: of those situations where you have to like see what

645
00:38:40,400 --> 00:38:47,120
Speaker 1: they do rather than listen to what they say. More.

646
00:38:47,120 --> 00:39:00,640
Speaker 2: After a quick break, let's get right back into it.

647
00:39:02,560 --> 00:39:06,719
Speaker 4: So Altman has talked a lot about the movie Her

648
00:39:07,200 --> 00:39:08,920
Speaker 4: and you know you mentioned that you listened to so

649
00:39:09,040 --> 00:39:11,440
Speaker 4: much footage of him talking, so you're probably one of

650
00:39:11,480 --> 00:39:14,319
Speaker 4: the world's foremost experts in how much he has talked

651
00:39:14,320 --> 00:39:19,919
Speaker 4: about that movie. Like, he's invoked it many times to

652
00:39:19,920 --> 00:39:24,160
Speaker 4: describe what he thinks chat GPT could be, but then

653
00:39:24,320 --> 00:39:27,560
Speaker 4: in recent times has seeming like kind of walked back

654
00:39:28,040 --> 00:39:31,600
Speaker 4: the extent to which he wants chat GPT to be

655
00:39:32,239 --> 00:39:37,920
Speaker 4: involved in intimate relationships. But then they roll out erotic content,

656
00:39:38,120 --> 00:39:42,120
Speaker 4: and so what do you think explains this whiplash?

657
00:39:42,760 --> 00:39:45,799
Speaker 1: At the end of the day, Opening Eye is now

658
00:39:45,840 --> 00:39:51,160
Speaker 1: a business, so even as it is still motivated very

659
00:39:51,239 --> 00:39:57,520
Speaker 1: much ideologically by a quest to build there's so called

660
00:39:57,600 --> 00:40:00,640
Speaker 1: artificial genre intelligence, they also we just need to make

661
00:40:00,680 --> 00:40:05,360
Speaker 1: money now, and they are losing an extraordinary amount of

662
00:40:05,400 --> 00:40:08,000
Speaker 1: money right now. The last figure that I saw was

663
00:40:08,040 --> 00:40:11,759
Speaker 1: they had committed to one point four trillion dollars of

664
00:40:12,239 --> 00:40:15,960
Speaker 1: debt for building out the data center infrastructure that they

665
00:40:16,000 --> 00:40:19,480
Speaker 1: need for training and delivering the next generations of their

666
00:40:19,480 --> 00:40:25,680
Speaker 1: GPT models. So they're trying everything to monetize in order

667
00:40:25,719 --> 00:40:29,600
Speaker 1: to plug that hole. That's why they're rolling out ads.

668
00:40:29,960 --> 00:40:34,759
Speaker 1: That's why they launched a whole slew of different products

669
00:40:34,880 --> 00:40:39,239
Speaker 1: last year, like a new web browser and agents, and

670
00:40:39,840 --> 00:40:45,359
Speaker 1: they're trying shopping integrated shopping experience and Chatbut they are

671
00:40:45,960 --> 00:40:48,920
Speaker 1: they've announced that they are going to build some kind

672
00:40:48,920 --> 00:40:53,320
Speaker 1: of hardware device. They're just spraying the market with various

673
00:40:53,320 --> 00:40:56,520
Speaker 1: different ideas, and a core part of being able to

674
00:40:56,560 --> 00:41:02,080
Speaker 1: then make all of these into into revenue generation engines

675
00:41:02,200 --> 00:41:06,480
Speaker 1: is users. They need to continue having more users, and

676
00:41:06,520 --> 00:41:09,440
Speaker 1: they need to have those users spend more time on

677
00:41:09,480 --> 00:41:15,560
Speaker 1: the platform. And the truth is Open Eyes actually losing

678
00:41:15,640 --> 00:41:20,319
Speaker 1: market share right now because there are other competitors like Anthropic,

679
00:41:20,520 --> 00:41:22,840
Speaker 1: like Google that are starting to eat their lunch.

680
00:41:23,040 --> 00:41:26,000
Speaker 3: And so as they experience all of.

681
00:41:26,040 --> 00:41:30,400
Speaker 1: These pressures from competitors and they experience just like the

682
00:41:30,440 --> 00:41:34,600
Speaker 1: dire need for more cash flow, they keep flip flopping

683
00:41:34,640 --> 00:41:38,640
Speaker 1: on decisions to try and figure out what is going

684
00:41:38,760 --> 00:41:44,320
Speaker 1: to make their product more engaging, more sticky, and more addictive.

685
00:41:45,280 --> 00:41:48,160
Speaker 2: To that end, I mean, you've talked about how AI

686
00:41:48,560 --> 00:41:52,919
Speaker 2: is essentially showing us that like surveillance capitalism moving into

687
00:41:52,960 --> 00:41:55,360
Speaker 2: its most extreme form, And as I've been reporting my

688
00:41:55,400 --> 00:41:58,279
Speaker 2: own book, like I'm curious now that we're at a

689
00:41:58,280 --> 00:42:01,520
Speaker 2: place where it's in people's emotion lives, their relationships or

690
00:42:01,520 --> 00:42:05,960
Speaker 2: mental health struggles. They're putting ads in the tatbet is

691
00:42:06,160 --> 00:42:09,440
Speaker 2: have we just like morphed into a into like I

692
00:42:09,440 --> 00:42:11,920
Speaker 2: would say final form, But I don't think it will

693
00:42:11,960 --> 00:42:15,560
Speaker 2: be the final form of just the most extreme version

694
00:42:15,560 --> 00:42:19,040
Speaker 2: of ai nable tourveillance capitalism pretty much.

695
00:42:20,480 --> 00:42:23,080
Speaker 1: I mean, you see this all the time with the

696
00:42:23,120 --> 00:42:29,040
Speaker 1: announcements that companies are making, where like over the last year,

697
00:42:29,239 --> 00:42:31,759
Speaker 1: maybe year and a half, there were these slew of

698
00:42:31,760 --> 00:42:35,640
Speaker 1: announcements from various companies where they just said we will

699
00:42:35,680 --> 00:42:38,359
Speaker 1: now be training on your data unless you opt out.

700
00:42:38,719 --> 00:42:41,520
Speaker 1: Like Zoom made an announcement saying we're now going to

701
00:42:41,560 --> 00:42:43,239
Speaker 1: train on your video calls. Of course, there was a

702
00:42:43,360 --> 00:42:46,919
Speaker 1: huge backlash so that didn't end up happening. But then

703
00:42:47,040 --> 00:42:49,960
Speaker 1: like Facebook announced, oh, we're changing all our settings, so

704
00:42:50,000 --> 00:42:52,320
Speaker 1: we're going to train on everything now, including your public

705
00:42:52,320 --> 00:42:55,719
Speaker 1: Instagram posts. LinkedIn made that announcement like we're going to

706
00:42:55,800 --> 00:42:58,600
Speaker 1: train on your LinkedIn posts and unless you opt out,

707
00:42:59,080 --> 00:43:03,720
Speaker 1: and like that is to me, the most obvious sign

708
00:43:04,120 --> 00:43:08,840
Speaker 1: of the surveillance capitalism creep. Like before they were already

709
00:43:09,600 --> 00:43:11,839
Speaker 1: monetizing off of the data that we were leaving on

710
00:43:11,880 --> 00:43:15,040
Speaker 1: their these digital trails that we were leaving online. But

711
00:43:15,239 --> 00:43:18,040
Speaker 1: now they're just becoming like so much more explicit about

712
00:43:18,080 --> 00:43:22,560
Speaker 1: like any last vestige of data that they have not

713
00:43:22,719 --> 00:43:26,160
Speaker 1: yet used and train like fed into their models and

714
00:43:26,840 --> 00:43:30,320
Speaker 1: tried to like ring money out of they are now grabbing.

715
00:43:31,160 --> 00:43:34,920
Speaker 1: And it is now become incumbent on the user to

716
00:43:35,200 --> 00:43:37,680
Speaker 1: track all of these updates and then go into the

717
00:43:37,719 --> 00:43:39,680
Speaker 1: settings and like find them and turn them all off.

718
00:43:41,040 --> 00:43:43,680
Speaker 1: And that's just the tip of the iceberg for the

719
00:43:43,760 --> 00:43:47,520
Speaker 1: much broader expansion of surveillance capitalism that's happening beneath the surface.

720
00:43:47,560 --> 00:43:54,360
Speaker 2: Now I see AI is this inherently extractive dynamic where

721
00:43:55,200 --> 00:43:57,760
Speaker 2: everything is being taken, whether it's our data, our privacy,

722
00:43:58,239 --> 00:44:02,359
Speaker 2: and I just like that in some ways is fundamentally

723
00:44:02,440 --> 00:44:07,520
Speaker 2: at odds with then trusting those same companies to automate care,

724
00:44:07,760 --> 00:44:11,480
Speaker 2: automate connection, automate intimacy. And one of the ways that

725
00:44:11,520 --> 00:44:13,560
Speaker 2: I really see that is in some of the folks

726
00:44:13,600 --> 00:44:18,000
Speaker 2: that you've interviewed, like content moderators, in Kenya who suffered

727
00:44:18,120 --> 00:44:22,759
Speaker 2: like pretty intense psychological harm filtering out violent content, and

728
00:44:22,800 --> 00:44:25,279
Speaker 2: so it's just hard for me not to see this

729
00:44:25,440 --> 00:44:31,400
Speaker 2: as a fundamentally extractive, exploitative dynamic, but then a dynamic

730
00:44:31,440 --> 00:44:33,360
Speaker 2: where people are telling us, oh, you're going to be

731
00:44:33,400 --> 00:44:37,640
Speaker 2: able to trust this to, you know, create a better world.

732
00:44:37,719 --> 00:44:39,239
Speaker 2: How could you create a better world with a tool

733
00:44:39,280 --> 00:44:42,560
Speaker 2: that fundamentally might have exploitation and extraction at the heart.

734
00:44:43,480 --> 00:44:45,440
Speaker 3: One of my theories for why these companies have been

735
00:44:45,480 --> 00:44:46,759
Speaker 3: so successful.

736
00:44:46,560 --> 00:44:52,239
Speaker 1: The Star is because they have very successfully hid a

737
00:44:52,320 --> 00:44:56,759
Speaker 1: lot of the exploitation and extraction. Like the you know,

738
00:44:56,800 --> 00:44:59,120
Speaker 1: the AI industry in general, or the tech industry in general.

739
00:44:59,160 --> 00:45:01,560
Speaker 1: They use a lot of you femisms in the way

740
00:45:01,560 --> 00:45:05,239
Speaker 1: that they talk about AI and what they're building, you know,

741
00:45:05,280 --> 00:45:09,680
Speaker 1: like data centers are the cloud. It's like an ethereal

742
00:45:09,960 --> 00:45:14,040
Speaker 1: thing that exists in the sky, not giant, sweaty computers

743
00:45:14,080 --> 00:45:17,839
Speaker 1: that are sucking up an enormous amount of energy. And

744
00:45:18,120 --> 00:45:21,920
Speaker 1: they talk about how autonomous their agents and their models are,

745
00:45:22,440 --> 00:45:25,200
Speaker 1: not actually acknowledging the fact that it's built on the

746
00:45:25,239 --> 00:45:28,080
Speaker 1: backs of tens of thousands of contract workers that live

747
00:45:28,120 --> 00:45:31,799
Speaker 1: in places like Kenya and so yeah, like I think

748
00:45:31,920 --> 00:45:37,320
Speaker 1: they create this veneer of magic, of mysticism of the

749
00:45:37,880 --> 00:45:43,080
Speaker 1: fact that this technology somehow falls from the heavens and

750
00:45:43,120 --> 00:45:47,000
Speaker 1: that therefore then benefit from from the fact that a

751
00:45:47,040 --> 00:45:49,480
Speaker 1: lot of consumers just don't realize when they're engagement these

752
00:45:49,520 --> 00:45:53,080
Speaker 1: so like they just see the kind of clean interface,

753
00:45:53,800 --> 00:45:57,680
Speaker 1: and those consumers just evaluate the technology based on oh,

754
00:45:57,760 --> 00:45:59,520
Speaker 1: like that was helpful, So like, let me use it

755
00:45:59,560 --> 00:46:03,200
Speaker 1: again without actually considering all of the things that went

756
00:46:03,320 --> 00:46:07,040
Speaker 1: into building the technology. That then reveal the logic of

757
00:46:07,080 --> 00:46:10,120
Speaker 1: the industry and what that industry will then how that

758
00:46:10,160 --> 00:46:13,120
Speaker 1: industry will continue to apply that logic to your data

759
00:46:13,160 --> 00:46:14,320
Speaker 1: and your life.

760
00:46:15,480 --> 00:46:18,200
Speaker 2: Do you think that this is just another iteration of

761
00:46:19,160 --> 00:46:23,400
Speaker 2: wealthier countries using because there are so many examples like

762
00:46:23,440 --> 00:46:26,040
Speaker 2: the best fashion industry, or there are just so many

763
00:46:26,080 --> 00:46:29,680
Speaker 2: examples of industries where we're just trained in the West

764
00:46:29,680 --> 00:46:32,839
Speaker 2: to just take and use and not think about what

765
00:46:32,920 --> 00:46:35,879
Speaker 2: the labor and the people that went into this thing

766
00:46:35,920 --> 00:46:38,839
Speaker 2: that we are taking and using. So I'm curious when

767
00:46:38,840 --> 00:46:42,120
Speaker 2: it comes to how some of the use cases that

768
00:46:42,160 --> 00:46:44,520
Speaker 2: people are using AI for, is this the same thing

769
00:46:44,560 --> 00:46:47,600
Speaker 2: where people are just taking and using without thinking about

770
00:46:47,640 --> 00:46:49,280
Speaker 2: who's on the other end doing the labor.

771
00:46:49,920 --> 00:46:50,919
Speaker 3: Yeah, one hundred percent.

772
00:46:51,000 --> 00:46:56,120
Speaker 1: I think AI, the AI industry does engage in very

773
00:46:56,360 --> 00:47:00,480
Speaker 1: It's it's part of a long history of extra exploitationing

774
00:47:00,520 --> 00:47:03,440
Speaker 1: extraction of many different industries. But I think the difference

775
00:47:03,680 --> 00:47:08,799
Speaker 1: is that with something like fashion or with food, it's

776
00:47:08,880 --> 00:47:11,799
Speaker 1: much more obvious to the consumer that there is a

777
00:47:11,800 --> 00:47:15,680
Speaker 1: supply chain that exists because it is a physical object

778
00:47:15,760 --> 00:47:17,880
Speaker 1: that you hold in your hand, and so you know

779
00:47:18,080 --> 00:47:22,080
Speaker 1: that at some point there were materials that were fed

780
00:47:22,200 --> 00:47:25,080
Speaker 1: into building this object, and there were people involved in

781
00:47:25,200 --> 00:47:29,759
Speaker 1: laboring to create this object, whereas AI as a digital thing,

782
00:47:30,040 --> 00:47:32,800
Speaker 1: like a lot of people still don't have that connection

783
00:47:33,080 --> 00:47:37,400
Speaker 1: that this digital thing actually exists physically in the world

784
00:47:37,480 --> 00:47:40,640
Speaker 1: and also requires human labor. And I think the other

785
00:47:40,719 --> 00:47:44,320
Speaker 1: thing is, like with something like fashion, you know people

786
00:47:44,360 --> 00:47:47,080
Speaker 1: are the industry is not telling you that they are

787
00:47:47,120 --> 00:47:49,000
Speaker 1: creating God Like.

788
00:47:49,000 --> 00:47:54,359
Speaker 3: They're not tell they're not selling you this over the

789
00:47:54,520 --> 00:47:59,080
Speaker 3: top religious narrative about why you need to buy their clothing.

790
00:47:59,840 --> 00:48:02,560
Speaker 1: And that's what the AI industry is doing, right, Like

791
00:48:02,600 --> 00:48:07,680
Speaker 1: they are not just hiding the exploitation extraction, but they're

792
00:48:07,680 --> 00:48:12,719
Speaker 1: also packaging it in this different like this like crazy

793
00:48:12,800 --> 00:48:16,800
Speaker 1: rhetoric where they say that if you engage and allow

794
00:48:16,840 --> 00:48:20,160
Speaker 1: them to build these tools, they're not just tools, in

795
00:48:20,239 --> 00:48:23,279
Speaker 1: fact that they're going to bring us civilization two point zero,

796
00:48:23,320 --> 00:48:26,000
Speaker 1: solve all of our problems, bring abundance to the whole world,

797
00:48:26,120 --> 00:48:26,960
Speaker 1: and everything is.

798
00:48:26,920 --> 00:48:34,440
Speaker 3: Going to be amazing. And I think that then accelerates.

799
00:48:33,719 --> 00:48:38,600
Speaker 1: The exploitation and extraction because it justifies it. So even

800
00:48:38,640 --> 00:48:42,360
Speaker 1: when the exploitation or the extraction is revealed, the people

801
00:48:42,360 --> 00:48:46,160
Speaker 1: who benefit from this like magical or want to benefit

802
00:48:46,200 --> 00:48:48,680
Speaker 1: from this magical oracle that exists at the end of

803
00:48:48,719 --> 00:48:52,839
Speaker 1: the journey, are willing to then accept the fact that

804
00:48:53,600 --> 00:48:56,560
Speaker 1: maybe there is some exploitation extraction that needs to happen

805
00:48:56,600 --> 00:48:57,120
Speaker 1: along the way.

806
00:48:57,800 --> 00:49:00,200
Speaker 2: That's such a good way to put it. And know,

807
00:49:01,960 --> 00:49:06,319
Speaker 2: like I may buy fast fashion from Sheenne, the head

808
00:49:06,360 --> 00:49:08,880
Speaker 2: of Sheene is not on television talking about why I

809
00:49:08,920 --> 00:49:12,719
Speaker 2: need to be personally invested in him being successful, like

810
00:49:12,880 --> 00:49:15,600
Speaker 2: the empire that he's building, I'm not. It's such a

811
00:49:15,600 --> 00:49:18,160
Speaker 2: good point that like it's kind of crazy when you

812
00:49:18,200 --> 00:49:19,239
Speaker 2: think about it.

813
00:49:19,360 --> 00:49:24,400
Speaker 1: Yeah, yeah, yeah, no, it's it's it's it's really nuts.

814
00:49:24,400 --> 00:49:28,360
Speaker 1: And I think sometimes I use this analogy that like

815
00:49:29,440 --> 00:49:31,799
Speaker 1: if we were like back in the Middle Ages and

816
00:49:31,840 --> 00:49:34,319
Speaker 1: someone knocked on your door and was like, I'm going

817
00:49:34,400 --> 00:49:36,560
Speaker 1: to sell you a potion that can cure all of

818
00:49:36,600 --> 00:49:40,280
Speaker 1: your problems, but it's just gonna cost everything you've ever owned,

819
00:49:40,320 --> 00:49:44,319
Speaker 1: including your firstborn child. You would be like, this is

820
00:49:44,360 --> 00:49:49,239
Speaker 1: a scam. I no, And that's what these companies do.

821
00:49:49,520 --> 00:49:53,279
Speaker 1: And yet like everyone was like, Okay.

822
00:49:53,080 --> 00:49:54,560
Speaker 3: Take my firstborn child.

823
00:49:56,760 --> 00:50:00,799
Speaker 1: And there's just something about like the modernity and the

824
00:50:00,880 --> 00:50:04,840
Speaker 1: sexiness and like the advanced technical aspect of AI that

825
00:50:04,960 --> 00:50:10,520
Speaker 1: makes much harder to recognize that that's actually like what

826
00:50:10,560 --> 00:50:12,879
Speaker 1: these companies are saying to us.

827
00:50:13,280 --> 00:50:16,320
Speaker 2: Karen, Empire of AI has been such a huge success,

828
00:50:16,400 --> 00:50:19,880
Speaker 2: you're now launching The Interface with BBC. Tell us how

829
00:50:19,960 --> 00:50:22,400
Speaker 2: that came about, And I'm curious if it's a natural

830
00:50:22,440 --> 00:50:26,240
Speaker 2: progression of the work that you explored in your book. Yeah.

831
00:50:26,280 --> 00:50:29,719
Speaker 1: So The Interface is a podcast that is going to

832
00:50:29,760 --> 00:50:32,760
Speaker 1: be a weekly show with me and two other hosts,

833
00:50:32,840 --> 00:50:39,440
Speaker 1: Thomas Jermaine and Nikki Wolf, both also longtime tech investigative reporters,

834
00:50:40,239 --> 00:50:42,840
Speaker 1: and we are going to be talking each week about

835
00:50:42,880 --> 00:50:48,440
Speaker 1: different topics that intersect with tech, but are not exclusively

836
00:50:48,440 --> 00:50:52,120
Speaker 1: about tech, because our thesis is that tech is the

837
00:50:52,440 --> 00:50:57,160
Speaker 1: dominant driving force that's rewiring your world, and in order

838
00:50:57,200 --> 00:50:59,600
Speaker 1: to really understand all the things that are happening in

839
00:50:59,640 --> 00:51:03,359
Speaker 1: the world, whether it's politics, geopolitics, see environment, or just

840
00:51:03,640 --> 00:51:06,239
Speaker 1: the crazy things happening on your phone, like, you have

841
00:51:06,320 --> 00:51:10,160
Speaker 1: to have a grasp of technology. And we want to

842
00:51:10,200 --> 00:51:13,680
Speaker 1: make this show as broad as possible. It's not meant

843
00:51:13,719 --> 00:51:16,880
Speaker 1: for just people that love tech or just people that

844
00:51:17,000 --> 00:51:21,440
Speaker 1: understand tech, Like we really really want everyone to feel

845
00:51:21,480 --> 00:51:24,279
Speaker 1: like this is a show for them, and it very

846
00:51:24,320 --> 00:51:26,640
Speaker 1: much is an extension of the work that I was

847
00:51:27,000 --> 00:51:28,359
Speaker 1: that I did with Empire BAI.

848
00:51:28,760 --> 00:51:29,560
Speaker 3: A lot of my.

849
00:51:31,040 --> 00:51:36,400
Speaker 1: Thesis in Empire AI is twofold one that like, we

850
00:51:36,440 --> 00:51:39,600
Speaker 1: need to hold these tech companies accountable by revealing the

851
00:51:39,640 --> 00:51:42,759
Speaker 1: degree of power that they have within our lives. But

852
00:51:42,880 --> 00:51:47,200
Speaker 1: secondly that there is opportunity to change this that actually

853
00:51:47,360 --> 00:51:51,160
Speaker 1: every single individual has agency to shape the way that

854
00:51:51,200 --> 00:51:53,760
Speaker 1: technology is going to be developed in the future.

855
00:51:53,760 --> 00:51:56,880
Speaker 3: And so this is going to be a huge theme

856
00:51:57,000 --> 00:51:57,440
Speaker 3: of the.

857
00:51:57,440 --> 00:52:01,120
Speaker 1: Show where we really our goal is to provide people

858
00:52:01,200 --> 00:52:05,359
Speaker 1: with that sense of agency, with that the feeling of

859
00:52:05,400 --> 00:52:09,680
Speaker 1: being informed such that they can then make better choices

860
00:52:09,719 --> 00:52:12,239
Speaker 1: that work for them and for their lives and help

861
00:52:12,320 --> 00:52:16,440
Speaker 1: them rehabilitate their tech, their relationship with technology, because at

862
00:52:16,440 --> 00:52:17,560
Speaker 1: the end of the day, I think a lot of

863
00:52:17,560 --> 00:52:21,839
Speaker 1: people in this moment have frustrating relationships with their technology.

864
00:52:22,360 --> 00:52:26,520
Speaker 1: They are frustrated that they're doom scrolling all the time,

865
00:52:26,719 --> 00:52:30,040
Speaker 1: they are worried about their kids and their experiences online,

866
00:52:30,640 --> 00:52:34,000
Speaker 1: and they feel this kind of restlessness or this like

867
00:52:34,080 --> 00:52:37,200
Speaker 1: hopelessness in the face of all of that, And we

868
00:52:37,280 --> 00:52:40,880
Speaker 1: really want to return back that sense of control in

869
00:52:40,960 --> 00:52:41,840
Speaker 1: people's lives.

870
00:52:42,239 --> 00:52:44,360
Speaker 2: I cannot wait to listen. This is sort of the

871
00:52:44,400 --> 00:52:47,000
Speaker 2: orientation that I feel in tech as well, that tech

872
00:52:47,160 --> 00:52:49,920
Speaker 2: is For so long, it's sort of been like, oh,

873
00:52:50,040 --> 00:52:52,000
Speaker 2: tech is this thing, and then all the other things

874
00:52:52,000 --> 00:52:53,880
Speaker 2: are over here. And then if you don't think of

875
00:52:53,880 --> 00:52:56,319
Speaker 2: yourself as a techi, you two now you feel like, oh,

876
00:52:56,320 --> 00:52:58,880
Speaker 2: what do I know about this? I'm no engineer. Meanwhile,

877
00:52:58,920 --> 00:53:01,920
Speaker 2: technology intersects pretty much every issue that plays out in

878
00:53:01,920 --> 00:53:04,880
Speaker 2: our lives, whether it's gender justice, racial justice, the environment,

879
00:53:04,960 --> 00:53:07,480
Speaker 2: like all of these issues are also tech issues, and

880
00:53:07,719 --> 00:53:10,680
Speaker 2: they impact and intersect with tech so directly that we

881
00:53:10,719 --> 00:53:13,440
Speaker 2: have to start telling a truer story about how that

882
00:53:13,560 --> 00:53:16,560
Speaker 2: shows up and help people feel like they do have

883
00:53:16,680 --> 00:53:18,320
Speaker 2: that agency that I think is so important.

884
00:53:18,880 --> 00:53:22,120
Speaker 3: Absolutely. Yeah, It's like a huge, huge.

885
00:53:21,880 --> 00:53:25,759
Speaker 1: Core driving thesis for me is like we one percent

886
00:53:26,200 --> 00:53:28,920
Speaker 1: have agency, and it's like the number one way that

887
00:53:29,360 --> 00:53:32,480
Speaker 1: we can all kind of resist the narratives that Silicon

888
00:53:32,560 --> 00:53:35,400
Speaker 1: Valley has fed us is by reclaiming that agency.

889
00:53:36,280 --> 00:53:37,960
Speaker 2: How can folks find the interface?

890
00:53:38,920 --> 00:53:43,120
Speaker 1: It will be available everywhere that people listen to their podcast.

891
00:53:43,239 --> 00:53:45,279
Speaker 1: It will also be a visualized podcast, So we're going

892
00:53:45,320 --> 00:53:49,280
Speaker 1: to be posting videos on YouTube and so yeah, people

893
00:53:49,320 --> 00:53:52,920
Speaker 1: can follow us across all those different platforms as well

894
00:53:52,920 --> 00:53:56,040
Speaker 1: as each of the three hosts individually on our social

895
00:53:56,040 --> 00:53:56,840
Speaker 1: media platforms.

896
00:53:57,080 --> 00:54:00,000
Speaker 2: Y'all please follow Karen. She is one of the most

897
00:54:00,080 --> 00:54:02,600
Speaker 2: asking people on the fascinating people on the planet. We

898
00:54:02,760 --> 00:54:05,400
Speaker 2: my producer Mike, and I were every other comment in

899
00:54:05,400 --> 00:54:07,000
Speaker 2: the doc is like, oh she's too interesting, Oh she

900
00:54:07,080 --> 00:54:08,839
Speaker 2: fuck everything you say. I'm like, oh, we could ask

901
00:54:08,840 --> 00:54:11,759
Speaker 2: a million different follow ups. So truly, thank you for

902
00:54:11,840 --> 00:54:14,719
Speaker 2: your work. We're such huge fans. Y'all please listen to

903
00:54:14,719 --> 00:54:16,080
Speaker 2: this podcast. It's going to be a banger.

904
00:54:17,000 --> 00:54:19,480
Speaker 1: Thank you so much, Bridget, and thank you so much, Mike.

905
00:54:19,520 --> 00:54:21,440
Speaker 1: It was really awesome to speak with you both.

906
00:54:26,560 --> 00:54:28,640
Speaker 2: Got a story about an interesting thing in tech. I

907
00:54:28,680 --> 00:54:30,520
Speaker 2: just want to say hi. You can read us at

908
00:54:30,560 --> 00:54:33,280
Speaker 2: Hello at tegody dot com. You can also find transcripts

909
00:54:33,280 --> 00:54:35,719
Speaker 2: for today's episode at tengody dot com. There Are No

910
00:54:35,800 --> 00:54:37,879
Speaker 2: Girls on the Internet was created by me Bridget Todd.

911
00:54:38,239 --> 00:54:41,680
Speaker 2: It's a production of iHeartRadio and Unbossed Creative Jonathan Strickland

912
00:54:41,719 --> 00:54:44,440
Speaker 2: is our executive producer. Tarry Harrison is our producer and

913
00:54:44,520 --> 00:54:48,200
Speaker 2: sound engineer. Michael Almato is our contributing producer. I'm your host,

914
00:54:48,280 --> 00:54:51,080
Speaker 2: Bridget Todd. If you want to help us grow, rate

915
00:54:51,120 --> 00:54:54,839
Speaker 2: and review us on Apple Podcasts. For more podcasts from iHeartRadio,

916
00:54:54,960 --> 00:54:57,279
Speaker 2: check out the iHeartRadio app, Apple Podcasts, or wherever you

917
00:54:57,320 --> 00:54:58,239
Speaker 2: get your podcasts.

918
00:55:00,280 --> 00:55:08,560
Speaker 1: Aha Usha M.