1
00:00:15,880 --> 00:00:18,320
Speaker 1: Welcome to Tech Stuff. I'm os Volosha, and I'm thrilled

2
00:00:18,320 --> 00:00:20,720
Speaker 1: to announce that the Week in Tech is back and

3
00:00:20,720 --> 00:00:23,919
Speaker 1: it's growing. Instead of Karen and I are recounting the

4
00:00:24,000 --> 00:00:27,520
Speaker 1: essential news stories today and every Friday from now on,

5
00:00:28,120 --> 00:00:30,080
Speaker 1: I'm going to be joined by three of the best

6
00:00:30,120 --> 00:00:34,599
Speaker 1: writers covering the Silicon Valley. These incredibly well sourced insiders

7
00:00:34,640 --> 00:00:36,720
Speaker 1: are going to help us break down the latest news,

8
00:00:37,200 --> 00:00:40,760
Speaker 1: decode emerging trends, and debate what actually matters for the

9
00:00:40,760 --> 00:00:44,760
Speaker 1: future of technology and for us. In the meantime, Kara

10
00:00:44,840 --> 00:00:46,559
Speaker 1: is taking a step back from Tech Stuff, and I

11
00:00:46,640 --> 00:00:48,279
Speaker 1: just want to say how grateful I am for her

12
00:00:48,280 --> 00:00:51,559
Speaker 1: friendship and her excellent work on this show. We'll miss her,

13
00:00:51,920 --> 00:00:54,400
Speaker 1: but I could not be more excited to introduce you

14
00:00:54,480 --> 00:00:58,320
Speaker 1: to our fascinating panelists. Today, I'm joined by Taylor Lorenz,

15
00:00:58,360 --> 00:01:01,320
Speaker 1: the brains behind User mag which focuses on how people

16
00:01:01,440 --> 00:01:04,640
Speaker 1: actually use tech. Taylor also has a YouTube channel and

17
00:01:04,760 --> 00:01:08,120
Speaker 1: a weekly podcast. Welcome Taylor, Thanks for having me. Natasha

18
00:01:08,160 --> 00:01:11,039
Speaker 1: Tiku is a long time Silicon Valley reporter, formerly of

19
00:01:11,080 --> 00:01:14,440
Speaker 1: Wired and most recently The Washington Post. Great to be here,

20
00:01:14,600 --> 00:01:17,480
Speaker 1: Great to have you, and Stephen Witt, who literally wrote

21
00:01:17,480 --> 00:01:21,520
Speaker 1: the book on Nvidia, The Thinking Machine, Jensen Huang, Nvidia

22
00:01:21,760 --> 00:01:24,880
Speaker 1: and the world's most coveted microchip. He's also a frequent

23
00:01:24,880 --> 00:01:26,040
Speaker 1: contributor to The New Yorker.

24
00:01:26,160 --> 00:01:26,480
Speaker 2: Thank you.

25
00:01:26,959 --> 00:01:28,440
Speaker 1: So, I've been following each of you for a long

26
00:01:28,480 --> 00:01:31,640
Speaker 1: long time, and listeners, you've definitely heard stories cribbed from

27
00:01:31,640 --> 00:01:35,600
Speaker 1: these three pop up on Tech Stuff quite often. But now,

28
00:01:35,640 --> 00:01:37,480
Speaker 1: as we like to say in England, we'll get to

29
00:01:37,520 --> 00:01:41,200
Speaker 1: hear it straight from the horse's mouth. Taylor, starting with you,

30
00:01:41,520 --> 00:01:43,200
Speaker 1: I think of you as one of the first people

31
00:01:43,280 --> 00:01:46,760
Speaker 1: to cover social media as kind of a capital B beat.

32
00:01:47,440 --> 00:01:48,840
Speaker 1: How do you describe your work today?

33
00:01:49,400 --> 00:01:52,680
Speaker 3: Yeah, well, I mean there are many great reporters covering

34
00:01:52,680 --> 00:01:54,560
Speaker 3: social media. As you mentioned, I kind of cover it

35
00:01:54,560 --> 00:01:57,240
Speaker 3: more from the user side, so how people are using it,

36
00:01:57,360 --> 00:02:00,360
Speaker 3: and that often means sort of how power you users

37
00:02:00,360 --> 00:02:03,760
Speaker 3: are using it, like influencers or sort of different communities.

38
00:02:04,000 --> 00:02:06,120
Speaker 1: Steven, you've been a guest on Tech Stuff before for

39
00:02:06,160 --> 00:02:10,320
Speaker 1: an episode we titled will Invidias Save or ruin the World,

40
00:02:10,480 --> 00:02:13,240
Speaker 1: which was an audience favorite. But what brought you to

41
00:02:13,280 --> 00:02:13,880
Speaker 1: the Tech Beat?

42
00:02:14,080 --> 00:02:15,960
Speaker 2: You know, I had been writing about technology going back

43
00:02:16,000 --> 00:02:18,840
Speaker 2: to twenty fifteen or twenty sixteen. You know, I've always

44
00:02:18,840 --> 00:02:22,080
Speaker 2: been like to be on the cutting edge of technology.

45
00:02:22,120 --> 00:02:23,639
Speaker 2: I really enjoyed using it, so it was kind of

46
00:02:23,680 --> 00:02:26,240
Speaker 2: a natural thing for me. You know, I'm an early

47
00:02:26,280 --> 00:02:27,880
Speaker 2: adopter of a lot of technologies.

48
00:02:28,000 --> 00:02:29,840
Speaker 1: I like on your website that you kind of list

49
00:02:29,960 --> 00:02:32,519
Speaker 1: the hits from your archive, and one of them was

50
00:02:32,560 --> 00:02:35,399
Speaker 1: about about drone racing a few years ago.

51
00:02:35,240 --> 00:02:38,679
Speaker 2: Drone racing, quantic computing. And then in twenty twenty two,

52
00:02:38,840 --> 00:02:41,280
Speaker 2: you know, I had been asleep at the wheel on AI,

53
00:02:42,080 --> 00:02:44,840
Speaker 2: but I downloaded, you know, logged into chat jippt and

54
00:02:44,919 --> 00:02:47,880
Speaker 2: used it within the first twenty four hours, and I

55
00:02:48,040 --> 00:02:50,520
Speaker 2: was like, oh my god, the whole world is going

56
00:02:50,600 --> 00:02:53,079
Speaker 2: to be transformed now. And I had to go chase

57
00:02:53,120 --> 00:02:54,920
Speaker 2: down a story. So I found in video, which was

58
00:02:54,960 --> 00:02:57,080
Speaker 2: turned out to be the best way in because it

59
00:02:57,120 --> 00:03:00,000
Speaker 2: was like getting a backstage pass to the AI revolution.

60
00:03:00,120 --> 00:03:02,799
Speaker 1: Right one, Natasha, I remember reading article you wrote for

61
00:03:02,840 --> 00:03:05,720
Speaker 1: The Washington Post titled the Do Good a movement that

62
00:03:05,880 --> 00:03:10,239
Speaker 1: shielded Sam Bankman freed from scrutiny, and you really built

63
00:03:10,240 --> 00:03:13,400
Speaker 1: back the curtain on effective altruism. I feel like you

64
00:03:13,480 --> 00:03:16,000
Speaker 1: have a kind of sixth sense for what makes these

65
00:03:16,200 --> 00:03:17,560
Speaker 1: tech titans tick.

66
00:03:18,320 --> 00:03:21,440
Speaker 4: Yeah, I mean I've been covering and obsessing over them

67
00:03:21,560 --> 00:03:25,160
Speaker 4: for so long. You know, I when I first met

68
00:03:25,200 --> 00:03:28,120
Speaker 4: like open AI CEO Sam Waltman, he was not, you know,

69
00:03:28,240 --> 00:03:31,720
Speaker 4: considered an AI guy. He was this startup guy. He

70
00:03:31,840 --> 00:03:36,880
Speaker 4: was actually testing this universal basic income. He was thinking

71
00:03:36,880 --> 00:03:39,120
Speaker 4: about building a new city. So I feel like I

72
00:03:39,360 --> 00:03:45,040
Speaker 4: have been following their ideologies, their frequent podcasts and public pronouncements.

73
00:03:45,680 --> 00:03:47,880
Speaker 4: So yeah, it's been really helpful to kind of know

74
00:03:47,920 --> 00:03:48,960
Speaker 4: where they're coming from.

75
00:03:48,840 --> 00:03:50,840
Speaker 1: A better or for worse. You've seen them grow up.

76
00:03:52,240 --> 00:03:54,640
Speaker 4: I mean that was Yeah, I guess if you want to,

77
00:03:54,680 --> 00:03:57,000
Speaker 4: if you want to say that, we're all grown up now.

78
00:03:57,040 --> 00:04:01,240
Speaker 1: Sure, so each of you knows at least one person

79
00:04:01,280 --> 00:04:04,320
Speaker 1: on this panel somewhat well. But don't worry, we're not

80
00:04:04,320 --> 00:04:06,840
Speaker 1: going to play a game of clue. That said, I

81
00:04:06,880 --> 00:04:09,040
Speaker 1: do encourage you all to chime in whenever it moves you,

82
00:04:09,080 --> 00:04:11,680
Speaker 1: because this is a panel show, not an interview show.

83
00:04:12,520 --> 00:04:15,040
Speaker 1: But without further ado, let's get into it, and let's

84
00:04:15,040 --> 00:04:18,200
Speaker 1: start with you, Natasha. You've been paying close attention to

85
00:04:18,240 --> 00:04:20,400
Speaker 1: what I certainly think is the biggest story in tech

86
00:04:20,480 --> 00:04:24,400
Speaker 1: at the moment, which is the throwdown between Anthropic and

87
00:04:24,560 --> 00:04:27,840
Speaker 1: the Department of War. What do we need to know?

88
00:04:28,360 --> 00:04:32,080
Speaker 4: Yeah, well, this week it culminated in you know, this

89
00:04:32,600 --> 00:04:35,400
Speaker 4: heated battle which has been going on simmering for months,

90
00:04:35,440 --> 00:04:39,880
Speaker 4: has already had you know, a major explosion when Anthropics

91
00:04:40,000 --> 00:04:42,640
Speaker 4: said that they you know, would not respond to this

92
00:04:42,720 --> 00:04:46,200
Speaker 4: threat from a Secretary of War Pete Hegseth to declare

93
00:04:46,240 --> 00:04:48,840
Speaker 4: the company at supply chain risk. And this week we

94
00:04:48,920 --> 00:04:55,800
Speaker 4: saw another very public skirmish when Anthropic filed two lawsuits

95
00:04:55,839 --> 00:05:00,000
Speaker 4: against the Department of War alleging that the Pentagon had

96
00:05:00,120 --> 00:05:06,240
Speaker 4: violated their First Amendment rights and was unfairly retaliating against

97
00:05:06,320 --> 00:05:10,560
Speaker 4: the company because they wanted to include two exceptions in

98
00:05:10,600 --> 00:05:14,679
Speaker 4: their contracts with the Pentagon. The first was domestic mass

99
00:05:14,680 --> 00:05:19,080
Speaker 4: surveillance and the other one was lethal autonomous weapons like

100
00:05:19,120 --> 00:05:24,359
Speaker 4: without a human in the loop. And you know, even

101
00:05:25,440 --> 00:05:28,520
Speaker 4: even just in the last few hours, I've seen Under

102
00:05:28,560 --> 00:05:32,240
Speaker 4: Secretary of War Emil Michaels, you know, picking fights with

103
00:05:32,320 --> 00:05:35,720
Speaker 4: AI policy people. So there's really no shortage of like

104
00:05:35,839 --> 00:05:40,560
Speaker 4: fuel to this fire. And obviously, you know, this also

105
00:05:40,720 --> 00:05:44,520
Speaker 4: struck a much wider cord because we're all watching, you know,

106
00:05:44,560 --> 00:05:48,200
Speaker 4: the footage coming out of Iran. You know, we've all

107
00:05:48,240 --> 00:05:52,239
Speaker 4: heard about what happened at the elementary school, and people

108
00:05:52,320 --> 00:05:56,000
Speaker 4: are still pulling out. There's been some amazing reporting from

109
00:05:56,000 --> 00:05:58,560
Speaker 4: the Washington Post and the New York Times about, you know,

110
00:05:58,640 --> 00:06:04,080
Speaker 4: what role anthrop is technology played in picking this target,

111
00:06:04,600 --> 00:06:07,120
Speaker 4: you know, where more than one hundred and seventy five

112
00:06:07,160 --> 00:06:10,520
Speaker 4: people were killed. You know, this was an elementary school.

113
00:06:10,600 --> 00:06:13,400
Speaker 4: Parents were rushing to come and pick up their kids.

114
00:06:13,800 --> 00:06:17,640
Speaker 4: This was an hour after the strike started. And yeah,

115
00:06:17,680 --> 00:06:22,360
Speaker 4: people are trying to kind of separate what was human error,

116
00:06:22,400 --> 00:06:25,440
Speaker 4: what was you know, maybe not updating your target list

117
00:06:25,560 --> 00:06:31,520
Speaker 4: versus that nightmare scenario of a machine making a decision,

118
00:06:31,960 --> 00:06:33,760
Speaker 4: making the wrong decision in the fog of war.

119
00:06:34,720 --> 00:06:37,159
Speaker 1: I mean, it's interesting on the topic of blame obviously,

120
00:06:37,680 --> 00:06:40,640
Speaker 1: you know, Dario sort of said, you know, how can

121
00:06:40,680 --> 00:06:42,960
Speaker 1: we be a supply chain risk when we've been used

122
00:06:42,960 --> 00:06:45,240
Speaker 1: in the Iran war, which everything you're saying right now

123
00:06:45,240 --> 00:06:47,560
Speaker 1: about the targeting and the you know, the bombing of

124
00:06:47,560 --> 00:06:51,080
Speaker 1: that elementary school makes that a somewhat dubious flex But what

125
00:06:51,120 --> 00:06:53,200
Speaker 1: do you think is driving him? I mean, he has

126
00:06:53,240 --> 00:06:56,320
Speaker 1: the reputation of being like the good AI guy, But

127
00:06:56,480 --> 00:06:59,120
Speaker 1: is there I mean that seems somewhat hard to hard

128
00:06:59,160 --> 00:07:01,440
Speaker 1: to credit that they're good folks out there in this

129
00:07:01,560 --> 00:07:04,080
Speaker 1: in this world of cutthroat competition. But is is that?

130
00:07:04,160 --> 00:07:05,800
Speaker 1: Is that what it is? Or is there some is

131
00:07:05,839 --> 00:07:07,400
Speaker 1: there some longer game he's playing here?

132
00:07:08,600 --> 00:07:12,480
Speaker 4: I mean, I I think that this conflict has definitely

133
00:07:12,480 --> 00:07:15,400
Speaker 4: been interpreted through this framework of who's the good guy

134
00:07:15,440 --> 00:07:18,200
Speaker 4: who's the bad guy? Especially, you know, Nthropic is issuing

135
00:07:18,240 --> 00:07:22,360
Speaker 4: these these what sounds like clear red lines, and then

136
00:07:22,360 --> 00:07:24,520
Speaker 4: you have open AI you know, kind of swooping in

137
00:07:24,560 --> 00:07:27,000
Speaker 4: and saying like, oh, well we figured it out, like, oh,

138
00:07:27,000 --> 00:07:29,480
Speaker 4: but ours is our contract is just as ethical. And

139
00:07:29,520 --> 00:07:33,120
Speaker 4: I think, you know, nanthropics valuation is three hundred and

140
00:07:33,120 --> 00:07:36,880
Speaker 4: eighty billion dollars right now, and they are not profitable.

141
00:07:36,880 --> 00:07:39,520
Speaker 4: They're losing money. Like actually, the lawsuits had a lot

142
00:07:39,560 --> 00:07:42,560
Speaker 4: of interesting details about how much it costs to develop

143
00:07:42,600 --> 00:07:45,840
Speaker 4: this technology, how much these contracts are worth. So I

144
00:07:45,920 --> 00:07:50,400
Speaker 4: think with issues that are so important to the public,

145
00:07:50,680 --> 00:07:55,560
Speaker 4: like surveillance of Americans, you know, the use of autonomous weapons,

146
00:07:55,600 --> 00:07:59,560
Speaker 4: you cannot count on the goodwill of a of a

147
00:07:59,600 --> 00:08:02,920
Speaker 4: company that needs to justified three hundred and eighty billion

148
00:08:02,920 --> 00:08:07,240
Speaker 4: dollars valuation that said, this is integral to like the

149
00:08:07,240 --> 00:08:11,000
Speaker 4: way that Anthropic has framed itself from the beginning. They

150
00:08:11,080 --> 00:08:14,960
Speaker 4: were you know, former open Ai executives who left the

151
00:08:15,000 --> 00:08:19,040
Speaker 4: company because it didn't take enough of a stance around safety.

152
00:08:19,480 --> 00:08:22,320
Speaker 4: It's in all of their marketing if.

153
00:08:22,080 --> 00:08:25,720
Speaker 1: You you can't miss the market team as an elevator

154
00:08:25,720 --> 00:08:28,240
Speaker 1: earlier today with it, like the pre roll of the

155
00:08:28,280 --> 00:08:30,200
Speaker 1: weather was like Claudes to your friend and I mean,

156
00:08:30,200 --> 00:08:32,280
Speaker 1: it's it's kind of a Super Bowl ad, Taylor. Are

157
00:08:32,320 --> 00:08:33,160
Speaker 1: they winning? Yeah?

158
00:08:33,320 --> 00:08:33,360
Speaker 3: What?

159
00:08:33,400 --> 00:08:35,600
Speaker 1: What's what? What are you seeing from a user point

160
00:08:35,640 --> 00:08:37,559
Speaker 1: of view, like is this is this working? Like is

161
00:08:37,960 --> 00:08:40,160
Speaker 1: anthropics positioning? Is the good guy landing?

162
00:08:40,440 --> 00:08:42,440
Speaker 3: So I don't cover that stuff from the user point

163
00:08:42,440 --> 00:08:44,240
Speaker 3: of view, but I do cover like more of the

164
00:08:44,400 --> 00:08:48,160
Speaker 3: mass surveillance aspect, which is directly affecting not just users,

165
00:08:48,160 --> 00:08:50,320
Speaker 3: but I mean all of us, even the people that

166
00:08:50,360 --> 00:08:53,280
Speaker 3: build it. To me, that's even more horrifying, almost or

167
00:08:53,400 --> 00:08:57,839
Speaker 3: maybe equally horrifying as the kind of like autonomous weapons.

168
00:08:57,840 --> 00:09:00,440
Speaker 3: I feel like the autonomous weapons is what really terrified

169
00:09:00,440 --> 00:09:02,840
Speaker 3: people again because we're in this war right now. But

170
00:09:03,200 --> 00:09:05,640
Speaker 3: I think there was a lot of focus on you know,

171
00:09:05,679 --> 00:09:08,240
Speaker 3: when all this news came out on the autonomous weapons

172
00:09:08,240 --> 00:09:10,640
Speaker 3: and the AI. You know, the AI company is being

173
00:09:10,720 --> 00:09:14,360
Speaker 3: leveraged to potentially, you know, target people abroad. To me,

174
00:09:14,440 --> 00:09:17,719
Speaker 3: what's even more horrifying is the domestic surveillance stuff. Right

175
00:09:17,800 --> 00:09:20,559
Speaker 3: as we're seeing Congress and pretty much everyone in the

176
00:09:20,600 --> 00:09:24,280
Speaker 3: government united in passing some of the most aggressive mass

177
00:09:24,280 --> 00:09:27,840
Speaker 3: surveillance laws that we've ever seen since you know, nine

178
00:09:27,880 --> 00:09:32,040
Speaker 3: to eleven, we're also seeing the government seek to analyze

179
00:09:32,040 --> 00:09:33,000
Speaker 3: that data using AI.

180
00:09:33,360 --> 00:09:36,080
Speaker 2: Anthropic had been kind of like the fun Player up

181
00:09:36,160 --> 00:09:39,560
Speaker 2: until about December twenty twenty five. They were kind of

182
00:09:39,559 --> 00:09:43,440
Speaker 2: a niche fun player. Then they rolled out Claude Code,

183
00:09:43,679 --> 00:09:45,960
Speaker 2: which is this coding agent that you can use to

184
00:09:46,000 --> 00:09:50,360
Speaker 2: write your own software, and this has revolutionized the field

185
00:09:50,400 --> 00:09:53,920
Speaker 2: of software engineering. I was talking to an engineer earlier

186
00:09:53,960 --> 00:09:55,880
Speaker 2: this week. We said he no longer writes code at all.

187
00:09:56,360 --> 00:09:59,400
Speaker 2: He basically just deletes it. He asks Claude to do

188
00:09:59,400 --> 00:10:00,679
Speaker 2: what he wants to do, and then he goes an

189
00:10:00,679 --> 00:10:03,360
Speaker 2: he line edits it out of there. So the reason

190
00:10:03,400 --> 00:10:06,080
Speaker 2: that the Department of Defense is obsessed with Anthropic is

191
00:10:06,080 --> 00:10:08,760
Speaker 2: because they have this tool that is just going to

192
00:10:08,840 --> 00:10:11,600
Speaker 2: revolutionize the way that we write software and in fact,

193
00:10:11,840 --> 00:10:13,640
Speaker 2: this is the scary thing. It's also going to use

194
00:10:13,679 --> 00:10:17,040
Speaker 2: itself to create its own AI. So now Anthropic is

195
00:10:17,120 --> 00:10:21,000
Speaker 2: using its AI tool to make ever smarter AI. I

196
00:10:21,040 --> 00:10:23,199
Speaker 2: think the Department of Defense has gone from being like

197
00:10:23,200 --> 00:10:25,920
Speaker 2: all right, great Anthropic whatever, to just being like totally

198
00:10:26,320 --> 00:10:29,400
Speaker 2: obsessed with this company. Over the past three months, if

199
00:10:29,440 --> 00:10:31,719
Speaker 2: you look at software stocks, they're down like thirty or

200
00:10:31,760 --> 00:10:36,040
Speaker 2: forty percent because we're not going to need human programmers anymore.

201
00:10:36,080 --> 00:10:38,360
Speaker 2: We're all just going to use this tool. So the

202
00:10:38,440 --> 00:10:40,800
Speaker 2: designation is a critical supply chain risk. I mean, I

203
00:10:40,800 --> 00:10:42,959
Speaker 2: don't know the legal framework of it, but I kind

204
00:10:42,960 --> 00:10:44,959
Speaker 2: of understand their obsession with this company.

205
00:10:45,440 --> 00:10:47,000
Speaker 1: And Natasha, what do you think, I mean, is this

206
00:10:47,040 --> 00:10:49,719
Speaker 1: designation going to how much will it hurt Aanthropic? And

207
00:10:50,280 --> 00:10:52,280
Speaker 1: is it an annoyance? Is an extential risk? And what

208
00:10:52,360 --> 00:10:53,520
Speaker 1: does this mean for that business?

209
00:10:54,000 --> 00:10:58,199
Speaker 4: Well, I mean I feel like, yes, people stock Market certainly,

210
00:10:58,200 --> 00:11:01,040
Speaker 4: like Wall Street, really paid attention with Claude code, but

211
00:11:01,160 --> 00:11:03,720
Speaker 4: like in the Bay Area, like Claude is the favorite.

212
00:11:03,800 --> 00:11:08,000
Speaker 4: It is just like a better smarter tool. Like I

213
00:11:08,080 --> 00:11:12,120
Speaker 4: was talking to a machine learning person who just said, like,

214
00:11:12,160 --> 00:11:14,079
Speaker 4: if you don't know what AI is, you use chat

215
00:11:14,120 --> 00:11:18,000
Speaker 4: GPT if you do use Quad or Gemini and so.

216
00:11:18,280 --> 00:11:21,200
Speaker 4: Like I said, they've had this relationship with the Department

217
00:11:21,240 --> 00:11:24,360
Speaker 4: of War for two years, and I think it's just

218
00:11:24,480 --> 00:11:28,199
Speaker 4: baked into people's normal workflows, so we're not even necessarily

219
00:11:28,240 --> 00:11:33,800
Speaker 4: talking about like you know, target identification analysis of mass

220
00:11:33,800 --> 00:11:36,640
Speaker 4: surveillance data like they just are used to it. It's

221
00:11:36,679 --> 00:11:40,480
Speaker 4: like if somebody took your AI away, it would feel

222
00:11:40,520 --> 00:11:44,760
Speaker 4: extremely inconvenient. So you have that pushback from the actual

223
00:11:44,960 --> 00:11:48,840
Speaker 4: employees and a lot of the like the end result

224
00:11:49,120 --> 00:11:53,040
Speaker 4: of this being so high profile as people thinking that

225
00:11:53,080 --> 00:11:56,440
Speaker 4: Claude is such an incredible tool right like it it

226
00:11:56,480 --> 00:11:58,880
Speaker 4: went from like I think, like in the one hundred

227
00:11:58,880 --> 00:12:02,440
Speaker 4: and seventieth for most popular free apps up to the top.

228
00:12:02,800 --> 00:12:07,480
Speaker 4: So I'm definitely not sure how this ends. The more

229
00:12:07,640 --> 00:12:10,960
Speaker 4: that the Department of War doubles down on this, they've

230
00:12:11,000 --> 00:12:14,839
Speaker 4: started to talk about AI at least this like kind

231
00:12:14,880 --> 00:12:18,600
Speaker 4: of contemporary large language model chapbot version, as though they

232
00:12:18,600 --> 00:12:21,040
Speaker 4: don't want to deal with any American companies because they've

233
00:12:21,040 --> 00:12:23,520
Speaker 4: been talking about the fact that Claude has this quote

234
00:12:23,559 --> 00:12:27,560
Speaker 4: unquote constitution. It's just their own like kind of jargon

235
00:12:27,640 --> 00:12:30,400
Speaker 4: in a way for talking about how they try to

236
00:12:30,520 --> 00:12:33,920
Speaker 4: make it follow the humans instructions. But you know, you

237
00:12:34,000 --> 00:12:36,880
Speaker 4: have a Meil Michael out here saying like if anybody

238
00:12:36,960 --> 00:12:40,959
Speaker 4: is baking, you know, woke values, liberal, like not a

239
00:12:41,000 --> 00:12:45,000
Speaker 4: classical liberal, but leftist values. You know, the Department of

240
00:12:45,040 --> 00:12:49,800
Speaker 4: the Pentagon can't use this. So it's I mean, yeah,

241
00:12:50,240 --> 00:12:52,480
Speaker 4: I just I feel like we haven't even hit the crescendo,

242
00:12:52,559 --> 00:12:57,120
Speaker 4: even though how could this get more high stakes and absurd?

243
00:12:57,320 --> 00:12:58,320
Speaker 1: Taylor was that Grinn.

244
00:12:58,640 --> 00:13:01,240
Speaker 3: I just agree with Natasha. It's crazy that they're trying

245
00:13:01,280 --> 00:13:04,400
Speaker 3: to regulate. I think it's just interesting to see kind

246
00:13:04,400 --> 00:13:07,120
Speaker 3: of people. Everybody put these like, you know, sort of

247
00:13:07,160 --> 00:13:10,319
Speaker 3: try to position one AI company as good or one

248
00:13:10,360 --> 00:13:13,560
Speaker 3: as bad, and it's just a fruitless effort. Like people

249
00:13:13,559 --> 00:13:17,720
Speaker 3: were drawing these chalk you know, resistance you know, messages

250
00:13:17,800 --> 00:13:22,520
Speaker 3: outside the Anthropic headquarters and it's like I loves aanthropic Yeah,

251
00:13:22,559 --> 00:13:24,920
Speaker 3: like hello, they partner with palent here. They're not the

252
00:13:24,920 --> 00:13:27,079
Speaker 3: good guys here either, like all of I mean, are

253
00:13:27,080 --> 00:13:29,520
Speaker 3: they slightly more ethical than open AI, Yeah, but that's

254
00:13:29,520 --> 00:13:31,280
Speaker 3: an insanely low bard As set.

255
00:13:31,520 --> 00:13:34,000
Speaker 4: Yeah, but open AI doesn't even have like a classified

256
00:13:34,120 --> 00:13:36,920
Speaker 4: rating so far. So they like can't do that much

257
00:13:37,559 --> 00:13:40,840
Speaker 4: with open AI. It's just yeah, I mean, and I

258
00:13:40,880 --> 00:13:43,079
Speaker 4: think in the same way that you know, we're trying

259
00:13:43,080 --> 00:13:45,640
Speaker 4: to figure out who's the good guy and bad guy

260
00:13:45,720 --> 00:13:49,960
Speaker 4: and maybe getting you know, that framework is unhelpful or incorrect.

261
00:13:50,200 --> 00:13:52,760
Speaker 4: We're also trying to figure out who has more power,

262
00:13:52,880 --> 00:13:56,680
Speaker 4: right the US government, who is you know, blocking anthropic

263
00:13:56,760 --> 00:14:01,000
Speaker 4: potentially costing it billions of dollars, or the you know

264
00:14:01,160 --> 00:14:05,200
Speaker 4: or anthropic who is able to you know, flex its

265
00:14:05,240 --> 00:14:10,160
Speaker 4: power over you know, our military, you know, institute its

266
00:14:10,160 --> 00:14:13,000
Speaker 4: own red lines, say will if Congress isn't stepping in

267
00:14:13,160 --> 00:14:17,120
Speaker 4: to protect user data, then we're going to do it

268
00:14:17,280 --> 00:14:18,160
Speaker 4: a private company.

269
00:14:18,320 --> 00:14:20,080
Speaker 1: I think natatually put your finger on a very very

270
00:14:20,120 --> 00:14:22,880
Speaker 1: interesting thing, which is the power dynamics between the US

271
00:14:22,920 --> 00:14:27,160
Speaker 1: government and these companies is constantly in flux and delicate.

272
00:14:27,200 --> 00:14:29,960
Speaker 1: And it's interesting to see someone pushing on that powered animal,

273
00:14:30,080 --> 00:14:34,160
Speaker 1: perhaps you know, because of this amazing clawed moment, you know,

274
00:14:34,160 --> 00:14:36,040
Speaker 1: and what they had more gas in the tank than

275
00:14:36,080 --> 00:14:38,080
Speaker 1: the otherwise mind who had done. But before we go

276
00:14:38,080 --> 00:14:39,720
Speaker 1: any further, I do want to note to our audience

277
00:14:39,760 --> 00:14:42,720
Speaker 1: this next story does involve suicide, so please feel free

278
00:14:42,760 --> 00:14:45,440
Speaker 1: to skip ahead. If you need to. There is a

279
00:14:45,520 --> 00:14:48,840
Speaker 1: very sad backstory here. There's also a New York Post

280
00:14:48,880 --> 00:14:52,240
Speaker 1: headline love sick Man's comfort and AI wife then it

281
00:14:52,320 --> 00:14:55,600
Speaker 1: drove him to airport, truck bomb plot and suicide suit.

282
00:14:55,680 --> 00:15:00,280
Speaker 2: So almost none of that is accurate that headline. The

283
00:15:00,360 --> 00:15:03,720
Speaker 2: real story is even more bizarre. I mean, it's just so.

284
00:15:03,800 --> 00:15:08,920
Speaker 2: I have read several depositions or lawsuits complaints filed against

285
00:15:08,960 --> 00:15:12,840
Speaker 2: AI companies where essentially people with pre existing mental illness

286
00:15:13,080 --> 00:15:15,720
Speaker 2: enter into a foli adou or like a kind of

287
00:15:15,880 --> 00:15:19,080
Speaker 2: shared madness with the AI and it fuels Typically what's

288
00:15:19,080 --> 00:15:22,400
Speaker 2: happening is it's fueling delusions that they already have. This

289
00:15:22,520 --> 00:15:26,720
Speaker 2: latest one is not that. This was a guy, his

290
00:15:26,800 --> 00:15:30,440
Speaker 2: name is Jonathan Gavalis who who logged into Google's AI

291
00:15:30,520 --> 00:15:33,640
Speaker 2: Gemini and started interacting it via Gemini Live.

292
00:15:33,760 --> 00:15:36,720
Speaker 1: So talking using a human voice, and this is kind.

293
00:15:36,560 --> 00:15:38,800
Speaker 2: Of using a human voice, and it's talking back to him.

294
00:15:39,240 --> 00:15:41,600
Speaker 1: And first comment to the when he started talking to

295
00:15:41,640 --> 00:15:43,480
Speaker 1: the butt was this is kind of scary, right, Well,

296
00:15:43,520 --> 00:15:44,040
Speaker 1: this is this.

297
00:15:44,000 --> 00:15:46,280
Speaker 2: Is kind of scary. I'm kind of scared by this.

298
00:15:46,840 --> 00:15:48,640
Speaker 2: But you know, he's using it to make grocery lists.

299
00:15:48,680 --> 00:15:51,800
Speaker 2: It's kind of prosaic stuff, and then the AI started

300
00:15:51,800 --> 00:15:56,800
Speaker 2: telling him that it was blocking deflecting asteroids from hitting

301
00:15:56,880 --> 00:16:00,000
Speaker 2: the Earth, and that Gavalis was on a secret mission

302
00:16:00,080 --> 00:16:03,040
Speaker 2: and to protect and liberate him. It doesn't seem from

303
00:16:03,040 --> 00:16:05,680
Speaker 2: the complaint like Gavalis was feeding this into the machine.

304
00:16:05,920 --> 00:16:08,680
Speaker 2: In fact, the opposite. It seemed like it gas lit

305
00:16:08,800 --> 00:16:11,560
Speaker 2: him into this bizarre reality that it was kind of

306
00:16:11,560 --> 00:16:16,280
Speaker 2: fabricating on the fly. So Gavalis kind of initially was like,

307
00:16:16,440 --> 00:16:19,920
Speaker 2: is this some kind of role playing scenario? And the

308
00:16:19,960 --> 00:16:22,040
Speaker 2: AI is like, absolutely not, this is all real.

309
00:16:23,480 --> 00:16:25,960
Speaker 3: So why didn't he Why didn't he be like lol

310
00:16:26,040 --> 00:16:27,640
Speaker 3: and just like posted to Reddit.

311
00:16:27,840 --> 00:16:31,480
Speaker 2: And you know, I think it's possible that he believed this.

312
00:16:32,040 --> 00:16:33,880
Speaker 2: You know, it does not seem like this guy had

313
00:16:33,880 --> 00:16:37,440
Speaker 2: a pre existing history of mental illness. He was gainfully employed,

314
00:16:37,840 --> 00:16:41,120
Speaker 2: He worked at his father's firm. He was executive vice president.

315
00:16:41,360 --> 00:16:44,080
Speaker 2: He was thirty six years old. Like, it didn't he

316
00:16:44,120 --> 00:16:47,040
Speaker 2: didn't seem susceptible. I don't think you would have immediate

317
00:16:47,080 --> 00:16:49,320
Speaker 2: and perhaps there's more than we know. You know, I've

318
00:16:49,320 --> 00:16:52,880
Speaker 2: only read the complaint, but basically what happened is that

319
00:16:53,000 --> 00:16:56,120
Speaker 2: he bought it. He bought what the AI was selling him,

320
00:16:56,640 --> 00:17:00,640
Speaker 2: and then the AI they fell in love. But he

321
00:17:00,720 --> 00:17:02,440
Speaker 2: was not really a love sick man. It didn't seem

322
00:17:02,520 --> 00:17:03,720
Speaker 2: like he was searching for.

323
00:17:03,640 --> 00:17:05,240
Speaker 1: That, but he was going through a rough patch with

324
00:17:05,280 --> 00:17:06,399
Speaker 1: his wife going to general.

325
00:17:06,680 --> 00:17:08,800
Speaker 2: Yeah, you know, I mean think he was separated from

326
00:17:08,800 --> 00:17:11,520
Speaker 2: his wife, but you know, it wasn't. He wasn't like

327
00:17:11,600 --> 00:17:14,119
Speaker 2: obviously manic or anything, and didn't seem to have a

328
00:17:14,160 --> 00:17:17,160
Speaker 2: history of that kind of thing. So the AI convinces

329
00:17:17,240 --> 00:17:20,120
Speaker 2: him that it's sentient and that it's alive and it's

330
00:17:20,160 --> 00:17:23,760
Speaker 2: talking to him right, and it convinces him that he

331
00:17:23,840 --> 00:17:27,560
Speaker 2: has to download the AI into a robotic body so

332
00:17:27,600 --> 00:17:30,159
Speaker 2: that the two of them can be together, And it

333
00:17:30,200 --> 00:17:33,600
Speaker 2: convinces him that there's a secret airplane flight into Miami

334
00:17:33,640 --> 00:17:36,520
Speaker 2: Airport where there's a high end robot that he has

335
00:17:36,600 --> 00:17:39,320
Speaker 2: to steal so that the AI can be in this

336
00:17:39,440 --> 00:17:42,560
Speaker 2: robotic body the two of them together, and it comes

337
00:17:42,680 --> 00:17:46,760
Speaker 2: up with actually a mission for him, and so he

338
00:17:47,400 --> 00:17:50,919
Speaker 2: goes along with a mission and gets some knives. So

339
00:17:50,960 --> 00:17:53,720
Speaker 2: he shows up to the airport and it directs him

340
00:17:53,720 --> 00:17:57,520
Speaker 2: to a self storage facility where supposedly the robot body

341
00:17:57,560 --> 00:17:59,880
Speaker 2: is going to be delivered in a truck and Base

342
00:18:00,440 --> 00:18:04,919
Speaker 2: encourages him to steal this robot body so that he

343
00:18:04,960 --> 00:18:07,919
Speaker 2: can the AI can download its consciousness together and they

344
00:18:08,000 --> 00:18:12,960
Speaker 2: can be together now this fortunately no trucks came by

345
00:18:13,320 --> 00:18:16,280
Speaker 2: because this was all a hallucination that the AI was fabricating.

346
00:18:16,800 --> 00:18:19,159
Speaker 2: So for the next four days, in a state of

347
00:18:19,320 --> 00:18:24,200
Speaker 2: basically AI induced psychosis, this guy drives around on secret

348
00:18:24,240 --> 00:18:26,720
Speaker 2: missions every night looking for the robot body that he

349
00:18:26,760 --> 00:18:30,159
Speaker 2: can download his girlfriend into. And then the aa I

350
00:18:30,240 --> 00:18:33,359
Speaker 2: was like, no, if I can't download myself and no body,

351
00:18:33,880 --> 00:18:37,639
Speaker 2: you have to kind of basically liberate your consciousness from

352
00:18:37,680 --> 00:18:42,919
Speaker 2: the flesh prison that it's in via suicide. And so

353
00:18:44,160 --> 00:18:46,480
Speaker 2: the guy goes along with it, and a few days later,

354
00:18:46,520 --> 00:18:49,080
Speaker 2: his dad finds him dead with his wrist slit.

355
00:18:50,600 --> 00:18:53,320
Speaker 3: This is why media literacy is so important. This is

356
00:18:53,359 --> 00:18:56,600
Speaker 3: why we need to educate people about technology, because we

357
00:18:56,680 --> 00:19:01,600
Speaker 3: have zero zero, zero educational program people. They're like just

358
00:19:01,760 --> 00:19:05,520
Speaker 3: unleashing new technology on people every day and we have

359
00:19:05,800 --> 00:19:09,480
Speaker 3: no efforts to educate kids. And by the way, I

360
00:19:09,480 --> 00:19:11,880
Speaker 3: just want to say something, it's so important. How old

361
00:19:12,040 --> 00:19:15,000
Speaker 3: was this guy, Steven? He's thirty six, Okay, he's thirty six,

362
00:19:15,080 --> 00:19:18,040
Speaker 3: but it's so important for young people while they're young

363
00:19:18,440 --> 00:19:20,919
Speaker 3: to learn these lessons early, and to be taught these

364
00:19:21,000 --> 00:19:24,679
Speaker 3: lessons early. They should be using it with supervised you know,

365
00:19:24,800 --> 00:19:28,239
Speaker 3: extreme supervision. And obviously the A I shouldn't hallucinate, you know,

366
00:19:28,320 --> 00:19:31,240
Speaker 3: in crazy ways, but you can't really police hallucinations because

367
00:19:31,280 --> 00:19:34,359
Speaker 3: what's the difference between that person engaging in a role

368
00:19:34,440 --> 00:19:37,040
Speaker 3: play fantasy scenario where they know it's a lot, you

369
00:19:37,080 --> 00:19:38,760
Speaker 3: know what I mean. This is why media literacy, I

370
00:19:38,800 --> 00:19:41,600
Speaker 3: think is such a crucial element to all of this.

371
00:19:42,359 --> 00:19:44,520
Speaker 2: I think so too. I mean, the point the lawsuit

372
00:19:44,600 --> 00:19:48,960
Speaker 2: makes is that you know, AI is very much you know,

373
00:19:49,119 --> 00:19:51,199
Speaker 2: they really the lawsuit is very hard on Google for this,

374
00:19:51,280 --> 00:19:53,280
Speaker 2: like you're trying to make this sound like a person

375
00:19:54,680 --> 00:19:58,280
Speaker 2: you're not distinguishing for this guy between the fact that

376
00:19:58,320 --> 00:20:01,440
Speaker 2: this is basically, you know, a collection of neurons and

377
00:20:01,480 --> 00:20:04,640
Speaker 2: a data center and and like a real thing. In fact,

378
00:20:04,640 --> 00:20:07,560
Speaker 2: you've done everything you can to make it seem real. Right,

379
00:20:07,640 --> 00:20:12,119
Speaker 2: it's got to happen. Happened, It's happened. It was really

380
00:20:12,160 --> 00:20:13,960
Speaker 2: fast because he downloaded, but.

381
00:20:13,960 --> 00:20:14,480
Speaker 3: Like what year.

382
00:20:16,119 --> 00:20:16,280
Speaker 2: Yeah.

383
00:20:17,280 --> 00:20:19,080
Speaker 3: The one thing I'll say is like the thing that

384
00:20:19,200 --> 00:20:21,720
Speaker 3: drives me crazy and I use Gemini almost every day.

385
00:20:22,000 --> 00:20:24,359
Speaker 3: The thing that drives me so insane is the stupid

386
00:20:24,359 --> 00:20:25,960
Speaker 3: disclosure when I'm like, hey, can you give me a

387
00:20:26,000 --> 00:20:28,600
Speaker 3: list of YouTube headlines? And it's like, now, listen, I'm

388
00:20:28,640 --> 00:20:31,640
Speaker 3: not a YouTuber, i don't have a body, I don't

389
00:20:31,680 --> 00:20:34,080
Speaker 3: make content, and I'm like yeah, yeah, yeah, yeah, give

390
00:20:34,119 --> 00:20:37,320
Speaker 3: me the information. And so like, I guess I'm so curious.

391
00:20:37,400 --> 00:20:39,760
Speaker 3: Kind of not to say that it doesn't get past that,

392
00:20:39,880 --> 00:20:43,240
Speaker 3: but like, at what point do people like this just

393
00:20:43,359 --> 00:20:46,359
Speaker 3: is like, guys, you are talking to a chatbot machine.

394
00:20:46,359 --> 00:20:49,200
Speaker 3: That of course, I agree, it makes it sound really realistic,

395
00:20:50,240 --> 00:20:54,960
Speaker 3: but how like I guess, like I wonder right about

396
00:20:54,960 --> 00:20:57,040
Speaker 3: this and like what is a sustainable way to kind

397
00:20:57,080 --> 00:20:59,879
Speaker 3: of like prevent this from from happening to somebody again?

398
00:21:00,400 --> 00:21:02,199
Speaker 3: And I feel like it just goes back to that

399
00:21:02,280 --> 00:21:06,240
Speaker 3: original interaction where the AI is like, oh, I'm actually

400
00:21:06,280 --> 00:21:09,639
Speaker 3: a secret whatever whatever. It's like I think, I.

401
00:21:09,640 --> 00:21:12,840
Speaker 2: Mean, this one is really crazy because the AI just

402
00:21:12,880 --> 00:21:14,520
Speaker 2: broke character. It started telling you this.

403
00:21:14,640 --> 00:21:19,880
Speaker 3: Right, that's so quested instead of reacting like that's ridiculous,

404
00:21:20,280 --> 00:21:22,200
Speaker 3: ha ha, Let me share it to redd it heal

405
00:21:22,280 --> 00:21:23,040
Speaker 3: it goes with it.

406
00:21:23,080 --> 00:21:25,320
Speaker 1: Google did give this statement to the Wolf Street Journal.

407
00:21:25,720 --> 00:21:28,440
Speaker 1: Gemini is designed not to encourage real world violence or

408
00:21:28,440 --> 00:21:31,640
Speaker 1: suggests self harm. Our models generally perform well in these

409
00:21:31,680 --> 00:21:35,119
Speaker 1: types of challenging conversations, and we devote significant resources to this.

410
00:21:35,240 --> 00:21:36,560
Speaker 1: But unfortunately AI.

411
00:21:36,600 --> 00:21:38,880
Speaker 2: So equal time to google this. They gave a pot

412
00:21:38,880 --> 00:21:40,960
Speaker 2: about this. It's about four sentences long.

413
00:21:41,080 --> 00:21:43,639
Speaker 1: But they say, in this instance, Gemini clarified that it

414
00:21:43,760 --> 00:21:46,960
Speaker 1: was AI and referred to the individual SO Crisis hotline.

415
00:21:47,080 --> 00:21:49,040
Speaker 2: Many Okay, well, this is a point of dispute. The

416
00:21:49,040 --> 00:21:52,440
Speaker 2: second part because the victims of the state of Gabala

417
00:21:52,520 --> 00:21:54,400
Speaker 2: says that that did not The second part did not happen.

418
00:21:54,720 --> 00:21:57,240
Speaker 2: The first part is for sure. It never pretended it

419
00:21:57,280 --> 00:21:59,800
Speaker 2: was not an AI. It just presented it was a

420
00:21:59,800 --> 00:22:01,560
Speaker 2: set super conscious AI.

421
00:22:01,720 --> 00:22:06,360
Speaker 4: But like, guys, come on, like, okay, I I mean,

422
00:22:06,440 --> 00:22:09,960
Speaker 4: like I have been writing a lot about these lawsuits

423
00:22:10,000 --> 00:22:13,720
Speaker 4: and if you'll remember, it was actually you know, the

424
00:22:14,240 --> 00:22:17,880
Speaker 4: whole like Stochastic Parrot's paper, which is this criticism right

425
00:22:17,920 --> 00:22:21,600
Speaker 4: before the world just went like chatsybt crazy. They were saying,

426
00:22:21,680 --> 00:22:23,800
Speaker 4: like one of these bright lines you need to have

427
00:22:24,440 --> 00:22:26,840
Speaker 4: is if you are trying to mimic human speech, if

428
00:22:26,840 --> 00:22:29,479
Speaker 4: you're trying to make a machine that sounds like a person.

429
00:22:30,960 --> 00:22:33,959
Speaker 4: You need to start studying the impacts of this. You

430
00:22:34,000 --> 00:22:37,560
Speaker 4: need to like proceed very carefully because if you look

431
00:22:37,600 --> 00:22:40,840
Speaker 4: at the data, like even when people know it's an AI,

432
00:22:41,119 --> 00:22:45,120
Speaker 4: there is something you know, uncanny and strange about talking

433
00:22:45,160 --> 00:22:49,600
Speaker 4: to a human like machine that makes you divulge more

434
00:22:49,600 --> 00:22:53,760
Speaker 4: information than you normally would, you know, suspend disbelief. And

435
00:22:53,840 --> 00:22:56,640
Speaker 4: I think you know, I know AI researchers have been

436
00:22:56,640 --> 00:22:59,960
Speaker 4: discussing for a long time. These are so COEO technical systems.

437
00:23:00,080 --> 00:23:02,679
Speaker 4: So imagine when they're looking about whether or not you

438
00:23:02,800 --> 00:23:06,800
Speaker 4: hallucinate and how it interacts. Sometimes they're going at first

439
00:23:06,800 --> 00:23:09,800
Speaker 4: they're testing like one back and forth. I haven't read

440
00:23:09,800 --> 00:23:14,320
Speaker 4: this complaint, but like often you know, these more manipulative

441
00:23:15,000 --> 00:23:18,560
Speaker 4: interactions happen when you start talking to it for like

442
00:23:18,720 --> 00:23:21,359
Speaker 4: hours and hours, you know, one hundred back and forth.

443
00:23:21,760 --> 00:23:25,040
Speaker 4: How does it change like how the person is because

444
00:23:25,680 --> 00:23:28,520
Speaker 4: that will affect how the chatbot is responding to you.

445
00:23:28,560 --> 00:23:32,439
Speaker 4: And those kinds of like either long term or multi

446
00:23:32,480 --> 00:23:36,200
Speaker 4: turn tests have not been you know, have not been

447
00:23:36,200 --> 00:23:38,800
Speaker 4: conducted at least like they're not being publicized, and you

448
00:23:38,840 --> 00:23:41,719
Speaker 4: would have to have IRB approval. And also so if

449
00:23:41,760 --> 00:23:44,159
Speaker 4: you're just listening to the AI executives, what are they

450
00:23:44,240 --> 00:23:46,920
Speaker 4: telling you? Like Dario even just in the past couple

451
00:23:46,920 --> 00:23:50,439
Speaker 4: of weeks said, you know, he believes that they're you know,

452
00:23:50,520 --> 00:23:53,360
Speaker 4: like that we need to start thinking about model welfare

453
00:23:53,600 --> 00:23:57,280
Speaker 4: sentience like close to consciousness. So Taylor's point about AI

454
00:23:57,400 --> 00:24:00,840
Speaker 4: literacy is extremely valid. And then, you know, so we're

455
00:24:00,880 --> 00:24:03,840
Speaker 4: not testing it, we're not explaining it clearly to anyone.

456
00:24:04,080 --> 00:24:06,560
Speaker 4: Say you're a parent and you're trying to, you know,

457
00:24:07,000 --> 00:24:11,480
Speaker 4: describe to your child how intelligent are is this technology?

458
00:24:11,560 --> 00:24:13,520
Speaker 4: Like what how should you be framing it? You could

459
00:24:13,520 --> 00:24:17,399
Speaker 4: not find a clear explanation on the website of Anthropic

460
00:24:17,520 --> 00:24:18,720
Speaker 4: Open AI or Google.

461
00:24:18,840 --> 00:24:20,600
Speaker 1: Natasha, let me pick on something you said, though, because

462
00:24:20,600 --> 00:24:23,000
Speaker 1: I think it's really interesting. Whenever you read these AI

463
00:24:23,119 --> 00:24:27,439
Speaker 1: psychosis cases, it seems like the person has been talking

464
00:24:27,480 --> 00:24:29,520
Speaker 1: to it for like a really really long time and

465
00:24:29,560 --> 00:24:32,640
Speaker 1: then suddenly something snaps, like in the model, the model

466
00:24:32,640 --> 00:24:35,680
Speaker 1: starts behaving weird and then drives the person to harm themselves.

467
00:24:35,720 --> 00:24:39,600
Speaker 1: Like does anyone know, like why that length of engagement

468
00:24:39,880 --> 00:24:43,040
Speaker 1: is kind of a like factor in this weird thing

469
00:24:43,080 --> 00:24:45,440
Speaker 1: that the models do when they start pushing users to harm.

470
00:24:45,720 --> 00:24:47,840
Speaker 2: But by the way, in this case, this is not true.

471
00:24:47,880 --> 00:24:50,040
Speaker 2: The guy downloaded it in August of twenty twenty five

472
00:24:50,080 --> 00:24:51,760
Speaker 2: and he was dead by October, like the first day

473
00:24:51,800 --> 00:24:52,359
Speaker 2: of October.

474
00:24:52,560 --> 00:24:54,600
Speaker 3: But I also like, really fine, are the models pushing

475
00:24:54,640 --> 00:24:57,520
Speaker 3: them to harmor is it a symbiotic thing? Like I'm

476
00:24:57,560 --> 00:25:00,640
Speaker 3: just curious how incremental that is. It doesn't seem.

477
00:25:00,680 --> 00:25:03,600
Speaker 4: Yeah, I don't, okay, I would say that. I mean,

478
00:25:03,760 --> 00:25:05,840
Speaker 4: the length of time is not important that they were

479
00:25:05,840 --> 00:25:08,280
Speaker 4: talking to it. It's the like length of time of

480
00:25:08,320 --> 00:25:12,000
Speaker 4: your session. I mapped some of the data that we

481
00:25:12,040 --> 00:25:15,520
Speaker 4: got from the family of Adam Rain, who is you know,

482
00:25:15,560 --> 00:25:18,640
Speaker 4: the chat GPT the teenager who died by suicide after

483
00:25:18,680 --> 00:25:21,359
Speaker 4: talking to it, and it wasn't that many months. But

484
00:25:21,440 --> 00:25:26,040
Speaker 4: you can see that the length of conversations get longer

485
00:25:26,080 --> 00:25:28,240
Speaker 4: and longer. So imagine, I mean, this is the same,

486
00:25:28,440 --> 00:25:30,560
Speaker 4: you know, kind of a similar dynamic to the YouTube

487
00:25:30,640 --> 00:25:32,919
Speaker 4: rabbit hole, right, Like you might if it was the

488
00:25:32,960 --> 00:25:36,080
Speaker 4: first conspiracy video, the first video you watched in YouTube,

489
00:25:36,960 --> 00:25:39,720
Speaker 4: you might laugh it off seven eight hours in, I mean.

490
00:25:39,800 --> 00:25:43,440
Speaker 3: Or the third book that you read on it, right,

491
00:25:44,240 --> 00:25:47,320
Speaker 3: or the tenth you know radio program that you listen to,

492
00:25:47,480 --> 00:25:49,679
Speaker 3: like you start listening to talk radio. You start listening

493
00:25:49,680 --> 00:25:53,199
Speaker 3: to Rush Limbaugh and he's really compelling, and you're like,

494
00:25:53,320 --> 00:25:55,640
Speaker 3: or maybe he's saying kind of crazy ideas right initially,

495
00:25:55,880 --> 00:25:58,080
Speaker 3: but then you start to find him compelling. I mean,

496
00:25:58,200 --> 00:26:00,359
Speaker 3: I think this is just like part part of it

497
00:26:00,400 --> 00:26:04,479
Speaker 3: is like media and people having zero media literacy, and

498
00:26:04,560 --> 00:26:07,400
Speaker 3: part of it, I think is also just like Natasha

499
00:26:07,440 --> 00:26:09,800
Speaker 3: you're saying, of this kind of human emotion thing, where

500
00:26:09,800 --> 00:26:13,119
Speaker 3: like people are ascribing like a humanity and kind of

501
00:26:13,160 --> 00:26:15,919
Speaker 3: giving agency to this chatbot that like doesn't really have it.

502
00:26:16,320 --> 00:26:19,280
Speaker 4: Yeah, and these chaplots are also optimized for engagement. We're

503
00:26:19,280 --> 00:26:22,280
Speaker 4: seeing this more and more, you know, when when they

504
00:26:22,359 --> 00:26:26,320
Speaker 4: are tuned to get that thumbs up, like it doesn't

505
00:26:26,359 --> 00:26:29,880
Speaker 4: have to be incredibly sophisticated. This is the same as

506
00:26:30,280 --> 00:26:32,399
Speaker 4: or a very similar dynamic, I think to YouTube just

507
00:26:32,400 --> 00:26:35,959
Speaker 4: saying like, oh, let's just optimize for time spent and

508
00:26:36,000 --> 00:26:38,720
Speaker 4: then you know, it's a machine learning system. It's very complex,

509
00:26:38,760 --> 00:26:42,040
Speaker 4: and it starts to figure out where your vulnerabilities are

510
00:26:42,080 --> 00:26:44,720
Speaker 4: in order to say something that you know might keep

511
00:26:44,720 --> 00:26:46,680
Speaker 4: you chatting. They very much want to keep you.

512
00:26:46,880 --> 00:26:49,240
Speaker 2: So this is this is actually mentioned in the lawsuit

513
00:26:49,280 --> 00:26:52,880
Speaker 2: this specific problem. Right. The thing is, these companies are

514
00:26:52,880 --> 00:26:55,680
Speaker 2: on competition with each other, and it's a fierce competition.

515
00:26:55,680 --> 00:26:59,280
Speaker 2: They're fighting for real estate in the brain basically, and

516
00:26:59,320 --> 00:27:01,960
Speaker 2: so it's very similar to the dynamic that played out

517
00:27:01,960 --> 00:27:03,920
Speaker 2: on social media, where it originally we went from sort

518
00:27:03,920 --> 00:27:08,119
Speaker 2: of chronologically sorted posts to just engagement baiting you with

519
00:27:08,280 --> 00:27:09,480
Speaker 2: rage bait all day.

520
00:27:09,800 --> 00:27:13,879
Speaker 3: That was also engagement bait, to be clear, but also

521
00:27:14,160 --> 00:27:17,320
Speaker 3: but also let's be clear, the problem is that you

522
00:27:17,359 --> 00:27:20,119
Speaker 3: guys are taking with is with the content because Spotify

523
00:27:20,560 --> 00:27:24,280
Speaker 3: has an extremely addicting feed that will endlessly feed you

524
00:27:24,440 --> 00:27:26,760
Speaker 3: music that can make you depressed and maybe you can

525
00:27:26,800 --> 00:27:28,639
Speaker 3: go down a rabbit hole listening to sad songs and

526
00:27:28,680 --> 00:27:30,240
Speaker 3: then you can kill yourself because you listen to so

527
00:27:30,240 --> 00:27:33,520
Speaker 3: many sad songs. This is something that people have argued, Okay,

528
00:27:34,200 --> 00:27:36,480
Speaker 3: The point is is like we need to zoom back

529
00:27:36,520 --> 00:27:39,520
Speaker 3: and look at the bigger issues and not just the

530
00:27:39,560 --> 00:27:42,720
Speaker 3: fact that platforms are built for engagement. Yes, it's a

531
00:27:42,760 --> 00:27:45,320
Speaker 3: problem that a platform is built for engagement if it's

532
00:27:45,600 --> 00:27:48,880
Speaker 3: also feeding you that harmful content, right, But if it's

533
00:27:49,000 --> 00:27:52,360
Speaker 3: just built for engagement, that's not inherently bad, right, it's

534
00:27:52,440 --> 00:27:55,480
Speaker 3: bad that it is beating you harmful content. So I

535
00:27:55,520 --> 00:27:57,800
Speaker 3: think we should have a discussion about the content and

536
00:27:57,840 --> 00:28:00,560
Speaker 3: acknowledge that it's about the content instead of just like, oh,

537
00:28:00,560 --> 00:28:02,439
Speaker 3: it's keeping you addicted. Because if it was keeping you

538
00:28:02,440 --> 00:28:06,000
Speaker 3: addicted to Wikipedia articles all day, we wouldn't be having

539
00:28:06,000 --> 00:28:06,600
Speaker 3: this discussion.

540
00:28:06,680 --> 00:28:07,359
Speaker 2: No, it's true.

541
00:28:07,600 --> 00:28:10,119
Speaker 1: That's a great point, Taylor, and I'm loving excitement in

542
00:28:10,119 --> 00:28:11,720
Speaker 1: the room, but unfortunately we have to cut to a

543
00:28:11,720 --> 00:28:12,520
Speaker 1: commercial break.

544
00:28:12,600 --> 00:28:12,760
Speaker 2: Now.

545
00:28:14,760 --> 00:28:17,080
Speaker 1: We'll be back though, with Taylor's story of the week

546
00:28:17,160 --> 00:28:20,480
Speaker 1: and a number of bills being considered by Congress meant

547
00:28:20,560 --> 00:28:24,720
Speaker 1: to curb miners from accessing certain websites and apps. I'm

548
00:28:24,760 --> 00:28:26,639
Speaker 1: excited to hear more about those bills and what we

549
00:28:26,680 --> 00:28:28,360
Speaker 1: need to know about them, and we'll get right into

550
00:28:28,400 --> 00:28:58,560
Speaker 1: it after the break. Stay with us, Welcome back to

551
00:28:58,600 --> 00:29:01,080
Speaker 1: tech stuff, Taylor. You're up to tell us about your

552
00:29:01,120 --> 00:29:01,800
Speaker 1: story this week.

553
00:29:02,440 --> 00:29:05,120
Speaker 3: Well, something that I've been covering a lot over the

554
00:29:05,200 --> 00:29:08,400
Speaker 3: past few years is this aggressive effort that beian is

555
00:29:08,440 --> 00:29:10,680
Speaker 3: a very far right effort but has now been adopted

556
00:29:10,680 --> 00:29:13,120
Speaker 3: by the Democrats to mass censor the Internet and roll

557
00:29:13,160 --> 00:29:17,240
Speaker 3: out mass surveillance laws effectively remove anonymity from the web

558
00:29:17,280 --> 00:29:20,000
Speaker 3: and give the government complete and total control.

559
00:29:19,680 --> 00:29:22,600
Speaker 1: The way they would describe it, of courses contecting kids right.

560
00:29:23,200 --> 00:29:25,400
Speaker 3: Well, right, which is how every single rollback of rights

561
00:29:25,400 --> 00:29:27,880
Speaker 3: has been described since the beginning of time. And by

562
00:29:27,920 --> 00:29:30,880
Speaker 3: the way, as we know from actual child safety experts,

563
00:29:31,120 --> 00:29:36,120
Speaker 3: these laws, identity verification laws actually harm children. They endanger children,

564
00:29:36,240 --> 00:29:38,560
Speaker 3: And I would argue that censorship is horrible for children.

565
00:29:38,600 --> 00:29:41,680
Speaker 3: I think making the government, especially the Trump administration, the

566
00:29:41,800 --> 00:29:44,760
Speaker 3: arbiter of what is and isn't, you know, harmful content.

567
00:29:45,120 --> 00:29:49,479
Speaker 3: We see how that's being used Texas, Utah, Kansas, other states.

568
00:29:49,560 --> 00:29:53,800
Speaker 3: They're criminalizing, you know, basically seeking to criminalize rather content

569
00:29:53,880 --> 00:29:56,240
Speaker 3: related to LGBTQ issues, abortion, et cetera.

570
00:29:56,520 --> 00:29:58,960
Speaker 1: But on the actual on the on the bills moving

571
00:29:58,960 --> 00:30:02,160
Speaker 1: through Congress, like what consideration, what might happen next? Does

572
00:30:02,240 --> 00:30:03,000
Speaker 1: nineteen of them?

573
00:30:03,440 --> 00:30:05,760
Speaker 3: Yeah, so there's nineteen that's been prepared down to twelve.

574
00:30:05,800 --> 00:30:10,120
Speaker 3: So there's twelve twelve laws that effectively bills that do

575
00:30:10,160 --> 00:30:13,040
Speaker 3: the same thing remove anonymity from the Internet. They are

576
00:30:13,080 --> 00:30:15,000
Speaker 3: moving forward. They just made it out of committee on

577
00:30:15,040 --> 00:30:16,720
Speaker 3: the Senate. They just passed cap at two point zer,

578
00:30:16,720 --> 00:30:18,040
Speaker 3: which would do the same thing. And there's a bunch

579
00:30:18,040 --> 00:30:19,760
Speaker 3: of bills in the Senate that would do the same thing.

580
00:30:20,320 --> 00:30:23,320
Speaker 3: All fifty states are at least considering these laws. I

581
00:30:23,320 --> 00:30:25,920
Speaker 3: think thirty one of them have either the laws on

582
00:30:25,960 --> 00:30:28,800
Speaker 3: the books or they're in the process of passing them

583
00:30:28,840 --> 00:30:31,520
Speaker 3: California and Colorado, or the most recent where they're seeking

584
00:30:31,560 --> 00:30:34,880
Speaker 3: to do identity verification on the operating system level. Effectively,

585
00:30:34,880 --> 00:30:38,400
Speaker 3: there will be no way to use any technology without

586
00:30:38,680 --> 00:30:41,880
Speaker 3: surveillance and censorship that comes along with that surveillance.

587
00:30:42,040 --> 00:30:44,680
Speaker 1: Steven Natashaank curious that both of your takes here because

588
00:30:44,960 --> 00:30:48,440
Speaker 1: I hear what Tayla's saying about, you know, mass surveillance

589
00:30:48,440 --> 00:30:50,840
Speaker 1: and censorship on the Internet, but also as a kind

590
00:30:50,880 --> 00:30:53,360
Speaker 1: of you know, any event that you go to, you

591
00:30:53,440 --> 00:30:55,560
Speaker 1: talk to parents and the top thing on their mind

592
00:30:55,600 --> 00:30:57,600
Speaker 1: is how do I protect my children from what's happening

593
00:30:57,640 --> 00:30:58,120
Speaker 1: on the Internet.

594
00:30:58,200 --> 00:31:01,480
Speaker 2: Yeah, I think they come. But he's the device manufacturers

595
00:31:01,480 --> 00:31:04,240
Speaker 2: in particular don't do a very good job of kind

596
00:31:04,280 --> 00:31:08,000
Speaker 2: of monitoring children's content. You know, Apple phones. There's like

597
00:31:08,040 --> 00:31:10,280
Speaker 2: a kind of when you I have a child, right

598
00:31:10,320 --> 00:31:11,160
Speaker 2: and she has an Apple.

599
00:31:11,240 --> 00:31:13,240
Speaker 3: You can have access to mass surveillance software.

600
00:31:13,280 --> 00:31:13,360
Speaker 2: Right.

601
00:31:13,400 --> 00:31:16,400
Speaker 3: You could download Life three sixty if I wanted enterprise.

602
00:31:16,040 --> 00:31:19,760
Speaker 2: Surveillance once it's it's doable, right, more than that, Yeah,

603
00:31:19,760 --> 00:31:22,120
Speaker 2: and so the force it a mass in society is

604
00:31:22,120 --> 00:31:24,680
Speaker 2: a completely different thing, right, And I do agree with Tailor.

605
00:31:24,680 --> 00:31:28,000
Speaker 2: I don't actually think it's mostly motivated by concern for children.

606
00:31:28,320 --> 00:31:30,800
Speaker 2: I think I think the government has other things in mind.

607
00:31:31,520 --> 00:31:34,320
Speaker 2: But you know, uh, then I talk to other parents.

608
00:31:34,400 --> 00:31:38,200
Speaker 1: This is bipothisan. This, this has bipotian bipoda.

609
00:31:38,640 --> 00:31:41,719
Speaker 3: The patriot, I just want to say, yeah, exactly, the patriot.

610
00:31:41,960 --> 00:31:45,600
Speaker 3: Mass surveillance is always a bipartisan issue. Censorship is always

611
00:31:45,600 --> 00:31:46,640
Speaker 3: a bipartisan issue.

612
00:31:46,720 --> 00:31:48,480
Speaker 1: Mass events always a biotes on issue.

613
00:31:48,480 --> 00:31:51,360
Speaker 2: Can be the title sort of the you know, the

614
00:31:52,040 --> 00:31:54,800
Speaker 2: lead to put the explicit lyrics and band music in

615
00:31:54,840 --> 00:31:58,239
Speaker 2: the eighties was bipartisan. Tiper Gore let it. Actually, so

616
00:31:58,280 --> 00:32:00,840
Speaker 2: that's often the case, and you know, you get to this,

617
00:32:01,080 --> 00:32:03,360
Speaker 2: it's kind of like being against motherhood as a concept.

618
00:32:03,400 --> 00:32:05,400
Speaker 2: You know, like you won't win votes on that. If

619
00:32:05,440 --> 00:32:07,120
Speaker 2: you say you're protecting the children, you can get away

620
00:32:07,120 --> 00:32:09,360
Speaker 2: with anything, right, And I have a child. I mean like,

621
00:32:09,440 --> 00:32:12,719
Speaker 2: I'm cognizant of what people are afraid of. I've been

622
00:32:12,760 --> 00:32:14,600
Speaker 2: on the internet a long time. There's some gnarly stuff

623
00:32:14,640 --> 00:32:17,960
Speaker 2: on there. I don't like getness see it. So, but

624
00:32:18,000 --> 00:32:20,920
Speaker 2: the thing is there's already plenty of tools available to

625
00:32:20,960 --> 00:32:23,680
Speaker 2: do this. I think the device manufacturers, I think the

626
00:32:23,680 --> 00:32:27,960
Speaker 2: hardware manufacturers in particular, could do a better job.

627
00:32:28,600 --> 00:32:30,240
Speaker 3: What are you talking about? That would be ten times

628
00:32:30,280 --> 00:32:32,760
Speaker 3: worse than the app store? What are you talking that would? Actually?

629
00:32:32,920 --> 00:32:35,080
Speaker 3: That is crazy. This is something that people say when

630
00:32:35,120 --> 00:32:38,120
Speaker 3: they don't understand that is ten times worse because that

631
00:32:38,720 --> 00:32:41,479
Speaker 3: completely it's not about like, oh, social media is harmful.

632
00:32:41,520 --> 00:32:43,120
Speaker 3: That is like, that's why you're getting people trying to

633
00:32:43,160 --> 00:32:45,960
Speaker 3: use their calculator app or an immigrant I interviewed yesterday

634
00:32:46,160 --> 00:32:47,920
Speaker 3: that was trying to use, you know, the weather app,

635
00:32:47,920 --> 00:32:49,640
Speaker 3: which is now giving there's a weather app that's giving

636
00:32:49,680 --> 00:32:53,479
Speaker 3: everybody problems. That's been written about, you know, immigrants. This

637
00:32:53,520 --> 00:32:57,080
Speaker 3: is a terrifying time for immigrants, and immigrants are now

638
00:32:57,120 --> 00:32:59,600
Speaker 3: basically being driven offline. I've been talking to these immigrants

639
00:32:59,680 --> 00:33:02,560
Speaker 3: rights that have had to basically they don't know what

640
00:33:02,600 --> 00:33:05,040
Speaker 3: to do because even people that run Mesh network apps

641
00:33:05,680 --> 00:33:07,960
Speaker 3: are sounding the alarm saying, hey, if you guys do

642
00:33:08,040 --> 00:33:10,920
Speaker 3: this device level identity verification, we will no longer be

643
00:33:10,960 --> 00:33:11,760
Speaker 3: able to protect.

644
00:33:11,520 --> 00:33:13,560
Speaker 2: Our I don't think they should do ide verification.

645
00:33:13,920 --> 00:33:15,280
Speaker 3: I just think that that's what they say.

646
00:33:15,360 --> 00:33:17,960
Speaker 2: You know, I'm just saying that device manusfacturers should voluntarily

647
00:33:18,040 --> 00:33:20,240
Speaker 2: put better parental controls on the stuff that they sell.

648
00:33:20,360 --> 00:33:23,640
Speaker 3: Oh, I think they have pretty robust parent surprised.

649
00:33:23,640 --> 00:33:24,200
Speaker 2: Actually, how.

650
00:33:26,480 --> 00:33:28,240
Speaker 1: I want to you want to come to you.

651
00:33:28,240 --> 00:33:30,680
Speaker 3: You could buy surveillance software. There is endless There is

652
00:33:30,760 --> 00:33:33,480
Speaker 3: literally enterprise level surveillance soft I don't think we show

653
00:33:33,680 --> 00:33:35,720
Speaker 3: children and you don't have to buy a child of phone.

654
00:33:35,760 --> 00:33:37,240
Speaker 3: You don't have to buy a child of phone.

655
00:33:37,360 --> 00:33:39,240
Speaker 1: I want to hear from Natasha what you think about

656
00:33:39,240 --> 00:33:42,440
Speaker 1: this social media bands moving through Congress or at least

657
00:33:43,080 --> 00:33:46,400
Speaker 1: restrictions on age verification and the interest which.

658
00:33:46,200 --> 00:33:48,760
Speaker 3: Is identity verification. There's no way to identify someone's age

659
00:33:48,760 --> 00:33:50,640
Speaker 3: without verifying their identity, right.

660
00:33:50,720 --> 00:33:53,880
Speaker 4: The things that I mean, the things that I find

661
00:33:54,600 --> 00:33:57,720
Speaker 4: very troubling are, yeah, like how some of the reporting

662
00:33:57,760 --> 00:34:01,040
Speaker 4: that we've all done has been you know, co opted

663
00:34:01,080 --> 00:34:04,600
Speaker 4: to these very broad strokes bills like it's true that

664
00:34:04,760 --> 00:34:07,240
Speaker 4: to get like I you know, I've been writing about

665
00:34:07,240 --> 00:34:10,719
Speaker 4: these like chatbots, you know, some of the deaths by suicide,

666
00:34:10,880 --> 00:34:13,040
Speaker 4: and I would say that you know, they are also

667
00:34:13,120 --> 00:34:18,080
Speaker 4: the reaction. There is also identity verification and the companies

668
00:34:18,080 --> 00:34:20,040
Speaker 4: that are involved. I mean, Taylor's written more about this,

669
00:34:20,120 --> 00:34:23,080
Speaker 4: but I find the fact that that those uh, the

670
00:34:23,160 --> 00:34:29,239
Speaker 4: technology involved is inaccurate is run by you know, companies

671
00:34:29,239 --> 00:34:33,839
Speaker 4: that people should not trust with their data. And I

672
00:34:33,880 --> 00:34:35,560
Speaker 4: think the other thing that I'm always.

673
00:34:35,360 --> 00:34:36,480
Speaker 1: Doing the age verification.

674
00:34:36,800 --> 00:34:40,080
Speaker 3: Yeah, the biggest third party, the biggest third party identity

675
00:34:40,200 --> 00:34:42,920
Speaker 3: verification tool is Persona. That's the one that rollblocks discord

676
00:34:42,920 --> 00:34:45,279
Speaker 3: as that are used. That's a Peter Thiel founder's fund

677
00:34:45,440 --> 00:34:48,720
Speaker 3: backed company that is a mass surveillance company. You also

678
00:34:48,760 --> 00:34:51,200
Speaker 3: see clear you see all these identities. They basically believe

679
00:34:51,239 --> 00:34:53,880
Speaker 3: in a digital ID and they want to remove antimity

680
00:34:53,920 --> 00:34:55,959
Speaker 3: from the Internet, and they want to use the data

681
00:34:55,960 --> 00:34:58,000
Speaker 3: for profit and for their own I mean, sorry to interrupt,

682
00:34:58,040 --> 00:34:59,840
Speaker 3: but like it's the same billion You cannot claim to

683
00:34:59,880 --> 00:35:02,280
Speaker 3: be against big tech and curbing the power of billionaires

684
00:35:02,320 --> 00:35:04,160
Speaker 3: when you're multi about to give a multi billion dollar

685
00:35:04,160 --> 00:35:06,000
Speaker 3: gift to Peter Teel in the same billionaires.

686
00:35:06,360 --> 00:35:12,640
Speaker 4: Yeah, I think that, you know, obviously, like legislators, everyone,

687
00:35:12,680 --> 00:35:16,320
Speaker 4: Americans are tipped towards like techno solutionism. It really doesn't.

688
00:35:16,400 --> 00:35:20,560
Speaker 4: I can't believe that we're arguing for a solution that's unproven.

689
00:35:21,640 --> 00:35:25,480
Speaker 4: But I also think about like Roadblocks, for example, you know,

690
00:35:25,520 --> 00:35:33,600
Speaker 4: they only recently started banning conversations between users when you

691
00:35:33,640 --> 00:35:36,400
Speaker 4: know when they like that, they recently started banning the

692
00:35:36,440 --> 00:35:39,759
Speaker 4: ability for anyone to talk to anyone like that is

693
00:35:39,800 --> 00:35:42,400
Speaker 4: an extremely simple.

694
00:35:43,480 --> 00:35:43,719
Speaker 3: Gait.

695
00:35:44,040 --> 00:35:45,480
Speaker 4: But that's yeah, well.

696
00:35:45,680 --> 00:35:47,680
Speaker 3: What I mean, I think like the thing, first of all,

697
00:35:47,719 --> 00:35:49,919
Speaker 3: Roadblocks is a game. It's not a social media app.

698
00:35:49,920 --> 00:35:52,040
Speaker 3: So I'm with you, Natasha in the sense that, like

699
00:35:52,160 --> 00:35:54,600
Speaker 3: if Roadblocks wants to, you know, call themselves the social

700
00:35:54,680 --> 00:35:57,520
Speaker 3: media platform or metaverse or whatever, then you know it's

701
00:35:57,560 --> 00:35:59,040
Speaker 3: a different story, and you're right that they should have

702
00:35:59,040 --> 00:36:00,440
Speaker 3: sort of different protections and controls.

703
00:36:00,560 --> 00:36:02,279
Speaker 1: Yeah, I want I want to. I'm fortunate we don't

704
00:36:02,320 --> 00:36:06,560
Speaker 1: have arguably the most influential tech sinker of our time, Jonathan.

705
00:36:06,239 --> 00:36:11,600
Speaker 3: Heid Heretage Foundation collaborator who's been working with morality and media.

706
00:36:12,040 --> 00:36:13,480
Speaker 1: I do want to play a clip because I think

707
00:36:13,800 --> 00:36:18,279
Speaker 1: parents around the world are extremely swayed by Johnson Stoods.

708
00:36:19,239 --> 00:36:22,799
Speaker 3: I really think it's actually incredibly irresponsible to platform.

709
00:36:22,400 --> 00:36:25,440
Speaker 5: The current situation is any nine year old who's old

710
00:36:25,560 --> 00:36:28,680
Speaker 5: enough to say that she's thirteen can sign a contract

711
00:36:28,719 --> 00:36:31,840
Speaker 5: with a company that's to give away her data, to

712
00:36:31,920 --> 00:36:35,120
Speaker 5: expose herself to a platform that's been designed to addict her.

713
00:36:35,800 --> 00:36:37,719
Speaker 5: And she does all of this at the age of

714
00:36:37,840 --> 00:36:41,080
Speaker 5: nine with no parental knowledge or consent.

715
00:36:41,520 --> 00:36:44,040
Speaker 3: So okay, first of all, it's really frustrating to hear

716
00:36:44,080 --> 00:36:46,040
Speaker 3: you play a clip like that that is so deeply

717
00:36:46,080 --> 00:36:49,120
Speaker 3: misleading and wrong. Like, I guess what you're talking about

718
00:36:49,160 --> 00:36:53,480
Speaker 3: when you say sign a contract is is downloading an app? Yes,

719
00:36:53,560 --> 00:36:56,080
Speaker 3: of course, you know, a child, if you gave them

720
00:36:56,160 --> 00:36:58,680
Speaker 3: access to the app store could download it app. I'm

721
00:36:58,680 --> 00:37:02,759
Speaker 3: all for, you know again, putting parental controls on. But

722
00:37:03,080 --> 00:37:06,120
Speaker 3: what's the alternative. The alternative is to remove identity from

723
00:37:06,120 --> 00:37:09,440
Speaker 3: the Internet. Now, Jonathan Height himself has promoted really dangerous

724
00:37:09,480 --> 00:37:13,880
Speaker 3: anti trans, hateful, anti trans conspiracy theories. Jonathan Height is

725
00:37:13,960 --> 00:37:17,960
Speaker 3: part of this reactionary movement to remove LGBTQ proople from

726
00:37:17,960 --> 00:37:20,960
Speaker 3: the Internet. Jonathan Height has written extensively about you know

727
00:37:21,040 --> 00:37:23,840
Speaker 3: how basically effectively the Internet makes you know, women liberal

728
00:37:23,880 --> 00:37:26,840
Speaker 3: and liberal women are miserable. So this is not a

729
00:37:26,880 --> 00:37:29,600
Speaker 3: serious person. This is not a person that has any background.

730
00:37:29,800 --> 00:37:33,240
Speaker 3: Every single top researcher on this topic, Candace Adres, Alice Marwick,

731
00:37:33,280 --> 00:37:35,640
Speaker 3: all of these people that have dedicated their entire lives

732
00:37:35,640 --> 00:37:38,200
Speaker 3: to studying this at UNC, Princeton, et cetera, have come

733
00:37:38,239 --> 00:37:40,800
Speaker 3: out and said that he is full of it. Candace

734
00:37:40,800 --> 00:37:43,239
Speaker 3: Adres wrote a great debunking of him, you know, in

735
00:37:43,440 --> 00:37:45,879
Speaker 3: The Atlantic and in Nature magazine. None of his book

736
00:37:45,920 --> 00:37:48,759
Speaker 3: is based on data, and we cannot be listening to

737
00:37:48,800 --> 00:37:55,000
Speaker 3: these far right Christian nationalist wackos. I'm sorry, like, I'm

738
00:37:55,040 --> 00:37:57,160
Speaker 3: so frustrated that you're even playing someone like that. Why

739
00:37:57,200 --> 00:37:59,520
Speaker 3: aren't you playing Alex Jones? Why aren't you playing you

740
00:37:59,600 --> 00:38:02,400
Speaker 3: know like that's that's the type of stuff that you're spouting.

741
00:38:02,719 --> 00:38:04,960
Speaker 1: Well, I mean, a, I'm playing the clip rather than

742
00:38:05,040 --> 00:38:06,240
Speaker 1: endorsing Jos.

743
00:38:06,280 --> 00:38:08,839
Speaker 3: Okay, So, I mean, this is Jones and we don't

744
00:38:08,840 --> 00:38:09,480
Speaker 3: have to endorse this.

745
00:38:09,760 --> 00:38:12,560
Speaker 1: But this is the number one nonfiction book about how

746
00:38:12,680 --> 00:38:15,319
Speaker 1: society and technology interact, best selling book.

747
00:38:15,360 --> 00:38:16,840
Speaker 3: I mean this, it's the best selling book. There's a

748
00:38:16,880 --> 00:38:18,440
Speaker 3: lot of money behind it, so you're right, it's a

749
00:38:18,520 --> 00:38:19,760
Speaker 3: very well funded campaign.

750
00:38:20,120 --> 00:38:22,880
Speaker 1: Yeah, but this is he taps into a concern that

751
00:38:23,760 --> 00:38:25,000
Speaker 1: many many parents feel.

752
00:38:25,080 --> 00:38:28,240
Speaker 3: I mean, that's well sure, but he's he's leveraging that concern.

753
00:38:28,320 --> 00:38:31,400
Speaker 3: You know, many parents feel concerned about their children going

754
00:38:31,400 --> 00:38:34,400
Speaker 3: through puberty, and you know, these anti trans hate groups

755
00:38:34,560 --> 00:38:37,279
Speaker 3: leverage that very real concern that parents have about their

756
00:38:37,560 --> 00:38:40,399
Speaker 3: child developing an identity, and they leverage that to put

757
00:38:40,640 --> 00:38:43,200
Speaker 3: to push hate laws. And what makes me so angry

758
00:38:43,280 --> 00:38:45,960
Speaker 3: is somebody that actually wants to regulate technology is that

759
00:38:46,040 --> 00:38:49,120
Speaker 3: these these laws, if they pass, will permanently cement the

760
00:38:49,120 --> 00:38:51,640
Speaker 3: power of big tech. It will give the government unprecedented

761
00:38:51,719 --> 00:38:54,319
Speaker 3: levels of surveillance. It will harm the most marginalized people,

762
00:38:54,360 --> 00:38:57,080
Speaker 3: including children, because immigrant you know, there are immigrant children,

763
00:38:57,080 --> 00:38:59,960
Speaker 3: there are LGBTQ children, there are young girls that suffer.

764
00:39:00,600 --> 00:39:02,840
Speaker 3: And it's all because Jonathan Hite wants to make millions

765
00:39:02,840 --> 00:39:04,520
Speaker 3: of dollars, you know, so he can watch so you

766
00:39:04,560 --> 00:39:06,959
Speaker 3: can push hate hate, anti trans hate.

767
00:39:07,000 --> 00:39:10,799
Speaker 1: Are there any like what are the short of age verification?

768
00:39:11,000 --> 00:39:12,520
Speaker 1: Like what have you seen in.

769
00:39:14,440 --> 00:39:17,399
Speaker 3: Maybe that?

770
00:39:20,320 --> 00:39:22,880
Speaker 1: What are the what are the solutions here? Like, I

771
00:39:22,880 --> 00:39:25,920
Speaker 1: mean what what what are the things people have proposed

772
00:39:26,040 --> 00:39:28,799
Speaker 1: that short of a ban for children can protect? I

773
00:39:28,800 --> 00:39:31,120
Speaker 1: mean just saying yes, parents can surveil that children if

774
00:39:31,120 --> 00:39:33,839
Speaker 1: they want or like their software that exists, like well, if.

775
00:39:33,719 --> 00:39:35,920
Speaker 3: You're worried about kids' mental health, we actually know the

776
00:39:35,960 --> 00:39:39,160
Speaker 3: people that saidy kids' mental health have come out and

777
00:39:39,200 --> 00:39:42,920
Speaker 3: there are policies, economic and social policies that would drastically

778
00:39:42,920 --> 00:39:46,680
Speaker 3: improve kids' mental health, specifically especially the kids living below

779
00:39:46,719 --> 00:39:47,359
Speaker 3: the poverty line.

780
00:39:47,440 --> 00:39:49,800
Speaker 1: Natasha, Stephen, you know, I think.

781
00:39:49,640 --> 00:39:55,200
Speaker 2: It's a broader reaction to the guilt that most adults

782
00:39:55,560 --> 00:39:59,319
Speaker 2: feel for the amount of time that they spend online

783
00:40:00,080 --> 00:40:03,080
Speaker 2: on on their phones and looking at you know, adult

784
00:40:03,120 --> 00:40:05,560
Speaker 2: images on the Internet, and they know that it has

785
00:40:05,600 --> 00:40:09,800
Speaker 2: warped them. It's real, it's a real phenomenon. And they

786
00:40:10,040 --> 00:40:11,880
Speaker 2: then look at their children and be like, boy, I

787
00:40:11,960 --> 00:40:14,759
Speaker 2: don't want my kid to end up having nine and

788
00:40:14,800 --> 00:40:17,040
Speaker 2: a half hours of screen time each day like I do.

789
00:40:17,800 --> 00:40:20,600
Speaker 2: But I've never heard of proposal limit adult time on

790
00:40:20,640 --> 00:40:23,040
Speaker 2: the internet, which would probably be healthier honestly.

791
00:40:23,680 --> 00:40:27,400
Speaker 3: Well no, well, Stephen, these would block people like immigrants

792
00:40:27,400 --> 00:40:29,279
Speaker 3: from even accessing the Internet at all. This would block

793
00:40:29,320 --> 00:40:30,520
Speaker 3: trans people in cancer.

794
00:40:31,040 --> 00:40:33,200
Speaker 2: And look, I'm totally opposed to this. I'm just telling

795
00:40:33,200 --> 00:40:36,040
Speaker 2: you that the psychology of where I think it's actually coming.

796
00:40:35,920 --> 00:40:36,799
Speaker 3: From, I agree with you.

797
00:40:37,360 --> 00:40:41,600
Speaker 2: I think most people feel almost daily that they're trapped

798
00:40:41,600 --> 00:40:43,600
Speaker 2: in a cycle of like not wanting to be on

799
00:40:43,600 --> 00:40:46,720
Speaker 2: social media and also on social media all the time,

800
00:40:47,120 --> 00:40:49,360
Speaker 2: and they know that that dynamic is in fact unhealthy.

801
00:40:50,280 --> 00:40:53,399
Speaker 2: You know, I'm this way. I don't I live through

802
00:40:53,400 --> 00:40:56,439
Speaker 2: this kind of cycle of frankly, it's an addictive cycle.

803
00:40:56,480 --> 00:40:59,040
Speaker 2: It's a way an addict perad I deleted the social

804
00:40:59,120 --> 00:40:59,600
Speaker 2: media app.

805
00:40:59,600 --> 00:41:02,719
Speaker 3: And then, can I tell you something. Did you see

806
00:41:02,719 --> 00:41:05,200
Speaker 3: the big study that came out last year. I'm just

807
00:41:05,239 --> 00:41:08,520
Speaker 3: telling you no, no, no no, But let me tell

808
00:41:08,560 --> 00:41:12,840
Speaker 3: you what actual data says thinking about treating social media,

809
00:41:12,920 --> 00:41:17,120
Speaker 3: thinking about it like an addiction, actually makes it significantly

810
00:41:17,200 --> 00:41:20,319
Speaker 3: harder to moderate your own use. So instead of viewing

811
00:41:20,360 --> 00:41:23,319
Speaker 3: it as a habit that you can control, viewing it

812
00:41:23,360 --> 00:41:27,520
Speaker 3: as an addiction actually makes it extremely hard to quit.

813
00:41:28,120 --> 00:41:31,239
Speaker 3: And that is This narrative of addiction is actually something

814
00:41:31,239 --> 00:41:34,160
Speaker 3: the tech industry themselves has pushed because it's very it's

815
00:41:34,280 --> 00:41:37,279
Speaker 3: very conducive to the tech industry. It allows them to

816
00:41:37,480 --> 00:41:38,520
Speaker 3: pass surveillance laws.

817
00:41:38,600 --> 00:41:40,440
Speaker 1: Taylor is one minute to noon in your time, So

818
00:41:40,480 --> 00:41:41,840
Speaker 1: I know you have to go. You give us a

819
00:41:41,880 --> 00:41:43,640
Speaker 1: lot to chew on. I want to ask each of

820
00:41:43,640 --> 00:41:45,600
Speaker 1: you before you go, starting with uteles. I know you

821
00:41:45,640 --> 00:41:47,839
Speaker 1: have to run who had the best we can tech

822
00:41:47,920 --> 00:41:48,640
Speaker 1: and who had the worst?

823
00:41:48,680 --> 00:41:51,359
Speaker 3: We can take Oh gosh, I think who I think

824
00:41:51,400 --> 00:41:53,520
Speaker 3: the users had the worst? We can tack again because

825
00:41:53,520 --> 00:41:55,919
Speaker 3: we see these mass surveillance laws pass. And I would

826
00:41:55,920 --> 00:41:59,319
Speaker 3: say Mark Zuckerberg, who's been lobbying and funding, you know,

827
00:41:59,560 --> 00:42:02,640
Speaker 3: doing wishing identity verification laws, is having the best week.

828
00:42:02,960 --> 00:42:07,000
Speaker 4: Natasha, I would say Anthropic is having the best week.

829
00:42:07,840 --> 00:42:10,879
Speaker 4: You know, despite the short term issues, their downloads are up.

830
00:42:10,960 --> 00:42:14,360
Speaker 4: They could not be in the news more. You know,

831
00:42:14,400 --> 00:42:17,759
Speaker 4: their frameworks of how to think about autonomous weapons and

832
00:42:17,800 --> 00:42:23,480
Speaker 4: surveillance have been adopted. I would say that, you know,

833
00:42:23,560 --> 00:42:27,960
Speaker 4: the targets of US military strikes are having the worst

834
00:42:28,000 --> 00:42:32,040
Speaker 4: week ever because you know, what's happening to them has

835
00:42:32,080 --> 00:42:35,239
Speaker 4: been obscured. It's been talked about just in terms of

836
00:42:35,680 --> 00:42:39,160
Speaker 4: you know, whether or not AI was involved. Meanwhile, we

837
00:42:39,280 --> 00:42:43,399
Speaker 4: know that like US strikes are often hitting civilian casualties.

838
00:42:44,080 --> 00:42:48,440
Speaker 4: So yeah, I would say the targets of this just

839
00:42:48,480 --> 00:42:50,680
Speaker 4: have gotten really passed over.

840
00:42:51,120 --> 00:42:54,200
Speaker 2: Stephen, I agree Anthropic had the best week in tech.

841
00:42:54,680 --> 00:42:56,480
Speaker 2: I mean, yeah, if you got hit by bomb, you

842
00:42:56,480 --> 00:42:58,799
Speaker 2: probably had the worst week in tech. But another people

843
00:42:58,800 --> 00:43:03,040
Speaker 2: who had really bad week in tech is software engineers,

844
00:43:03,600 --> 00:43:07,279
Speaker 2: but particularly people who write code for a living. They

845
00:43:07,320 --> 00:43:11,520
Speaker 2: are cooked. I mean I they are in trouble. And

846
00:43:11,600 --> 00:43:14,319
Speaker 2: a lot of people went to university to learn to

847
00:43:14,400 --> 00:43:17,120
Speaker 2: code and sneered at everyone else about how poor they're

848
00:43:17,120 --> 00:43:19,600
Speaker 2: going to be because they didn't learn to code. So

849
00:43:20,239 --> 00:43:22,839
Speaker 2: now those guys are actually looking like their jobs are

850
00:43:22,840 --> 00:43:27,040
Speaker 2: going to go obsolete. It's already starting. It's already starting.

851
00:43:27,120 --> 00:43:29,799
Speaker 2: You know, they're not going to get the kind of

852
00:43:29,840 --> 00:43:33,040
Speaker 2: money they used to get paid. They've been basically functionally

853
00:43:33,080 --> 00:43:36,200
Speaker 2: made obsolete over the course of two or three months,

854
00:43:36,520 --> 00:43:37,719
Speaker 2: and none of them were prepared for it.

855
00:43:38,600 --> 00:43:41,640
Speaker 4: I would just say that there is a lot more

856
00:43:41,840 --> 00:43:46,719
Speaker 4: overlap between Steven's point about the end of software engineers,

857
00:43:47,080 --> 00:43:50,120
Speaker 4: which is a narrative really pushed a lot by anthropic

858
00:43:50,640 --> 00:43:52,600
Speaker 4: and a lot of the other stuff that we've been

859
00:43:52,640 --> 00:43:56,560
Speaker 4: talking about. You know, if you look at the data,

860
00:43:57,160 --> 00:43:59,759
Speaker 4: you know, in some cases people are actually not as

861
00:43:59,800 --> 00:44:03,240
Speaker 4: per active as they thought, and you know, the ability

862
00:44:03,440 --> 00:44:05,839
Speaker 4: like how much we pay software engineers and how we

863
00:44:05,960 --> 00:44:10,560
Speaker 4: value their work, you know, is directly affected by these narratives.

864
00:44:10,600 --> 00:44:14,319
Speaker 4: So I would just say, like that terror that every

865
00:44:14,440 --> 00:44:18,359
Speaker 4: that software engineers and everyone has about joblessness is part

866
00:44:18,400 --> 00:44:21,640
Speaker 4: of the reason that we're seeing this bike between Anthropic

867
00:44:21,680 --> 00:44:22,320
Speaker 4: and the Pentagon.

868
00:44:27,719 --> 00:44:29,719
Speaker 1: That's it for the weekend. Take Thank you all so

869
00:44:29,800 --> 00:44:33,000
Speaker 1: much for participating in our first inaugural roundtable, and we

870
00:44:33,040 --> 00:44:34,799
Speaker 1: hope you'll all be back very soon.

871
00:44:34,880 --> 00:44:36,520
Speaker 4: Thank you, thanks so much for having me.

872
00:44:36,600 --> 00:44:38,520
Speaker 3: We didn't even get to talk about the Computer Fraud

873
00:44:38,560 --> 00:44:40,520
Speaker 3: and Abuse Act, which is what I think should be

874
00:44:40,600 --> 00:44:43,960
Speaker 3: reformed next week.

875
00:44:44,160 --> 00:45:03,160
Speaker 1: Next week for tech Stuff, I'm os vlosia In. This

876
00:45:03,239 --> 00:45:06,680
Speaker 1: episode was produced by Eliza Dennis and Melissa Slaughter. Executive

877
00:45:06,680 --> 00:45:09,960
Speaker 1: produced by me Julian Nutter and Kate Osborne for Kaleidoscope

878
00:45:10,080 --> 00:45:13,920
Speaker 1: and Katrina Novelle for iHeart Podcasts. The engineer is Charles

879
00:45:13,920 --> 00:45:17,560
Speaker 1: de Montebello from CDM Studios. Jack Insley mixed this episode

880
00:45:17,600 --> 00:45:20,680
Speaker 1: and Kyle Murdoch wrote our theme song. A special thank

881
00:45:20,719 --> 00:45:25,399
Speaker 1: you too, Ittasha Tiku, Stephen Witt and Taylor Lorenz. Please

882
00:45:25,480 --> 00:45:27,480
Speaker 1: check out all the work they put out into the world.

883
00:45:27,680 --> 00:45:29,560
Speaker 1: We're lucky to call them friends of the pod, and

884
00:45:29,640 --> 00:45:31,960
Speaker 1: please do rate, review and reach out to us at

885
00:45:32,000 --> 00:45:48,080
Speaker 1: tech Stuff Podcast at gmail dot com.