1
00:00:00,520 --> 00:00:03,720
Speaker 1: Hello the Internet, and welcome to season three, twenty six,

2
00:00:03,800 --> 00:00:08,200
Speaker 1: episode two of der Daily's Iday production of iHeartRadio.

3
00:00:08,640 --> 00:00:10,120
Speaker 2: This is a podcast where.

4
00:00:09,960 --> 00:00:12,920
Speaker 1: We take a deep dive into American share consciousness. And

5
00:00:12,960 --> 00:00:16,440
Speaker 1: it is Tuesday, February twenty seventh, twenty twenty four.

6
00:00:17,239 --> 00:00:20,880
Speaker 2: Oh yeah, it's I mean, what is that? That means?

7
00:00:20,920 --> 00:00:23,600
Speaker 2: It's Tuesday that we said twenty seventh? Is that where

8
00:00:23,640 --> 00:00:23,960
Speaker 2: we are?

9
00:00:24,120 --> 00:00:24,520
Speaker 1: That's right?

10
00:00:24,600 --> 00:00:29,480
Speaker 2: Seven two four. It's uh a Nausema Awareness Day. It's

11
00:00:29,600 --> 00:00:33,480
Speaker 2: National Retro Day, so I guess honor your see through

12
00:00:33,520 --> 00:00:39,200
Speaker 2: telephones and vinyl players. National Polar Bear Day, National Strawberry Day,

13
00:00:39,479 --> 00:00:43,040
Speaker 2: National Koalua Day, for all the for for all my

14
00:00:43,159 --> 00:00:45,920
Speaker 2: Kolua lovers out there, it's the day is yours. The

15
00:00:46,000 --> 00:00:47,120
Speaker 2: day is the white Russian?

16
00:00:47,360 --> 00:00:48,760
Speaker 1: Is Kolua white Russian? Or no?

17
00:00:48,880 --> 00:00:52,120
Speaker 2: That's Bailey's right, I think so, I don't know. It's

18
00:00:52,159 --> 00:00:55,520
Speaker 2: just much either way. It's just a way too much

19
00:00:55,560 --> 00:00:57,920
Speaker 2: sugar and you will have a bad time if that's

20
00:00:57,960 --> 00:00:58,480
Speaker 2: all you drink.

21
00:00:58,560 --> 00:01:01,920
Speaker 1: So yeah, that's right. Well, my name is Jack O'Brien aga.

22
00:01:02,480 --> 00:01:05,560
Speaker 1: Don't you wish your snack food were sweet like this?

23
00:01:06,360 --> 00:01:07,360
Speaker 2: Don't you want more?

24
00:01:07,440 --> 00:01:12,440
Speaker 1: Vague disk satisfiying bliss, don't you. That is courtesy of

25
00:01:12,480 --> 00:01:17,160
Speaker 1: Cleo Universe, in reference to the fact that some of

26
00:01:17,200 --> 00:01:21,280
Speaker 1: the greatest science of the late twentieth century was spent

27
00:01:21,920 --> 00:01:26,840
Speaker 1: focusing on how to get the perfect balance of mouth

28
00:01:27,000 --> 00:01:32,920
Speaker 1: feel and like enjoyment and dissatisfaction in nacho cheese doritos.

29
00:01:33,520 --> 00:01:36,680
Speaker 2: And yeah, gotta keep on popping them.

30
00:01:36,880 --> 00:01:42,160
Speaker 1: That's gotta gotta make them keep popping. That's uh, that's

31
00:01:42,200 --> 00:01:46,120
Speaker 1: who we are as a civilization. Unfortunately yet time and

32
00:01:46,160 --> 00:01:48,200
Speaker 1: I'm thrilled to be joined as always by my co

33
00:01:48,320 --> 00:01:49,800
Speaker 1: host mister Miles.

34
00:01:49,480 --> 00:01:55,360
Speaker 2: Graam, Miles Gray, Gag Boss, the Warren Bond rock Kinside Boss,

35
00:01:55,560 --> 00:01:58,600
Speaker 2: the Warrant Bondai rock kin Side. And that is the

36
00:01:58,680 --> 00:02:02,840
Speaker 2: reference to Senator Elizabeth Warren saying her dream blunt rotation

37
00:02:03,160 --> 00:02:07,680
Speaker 2: was just to smoke with the rock and my body

38
00:02:07,720 --> 00:02:10,840
Speaker 2: curled into itself and I became a stand and drifted

39
00:02:10,919 --> 00:02:13,600
Speaker 2: off into the sea. Yes, thank you to Laceroni for

40
00:02:13,639 --> 00:02:15,600
Speaker 2: that one, for letting us know that's what you're gonna

41
00:02:15,600 --> 00:02:17,440
Speaker 2: pass the duchy on the rock hand side.

42
00:02:17,760 --> 00:02:21,000
Speaker 1: On the rock hand side, I wonder what the rock

43
00:02:21,040 --> 00:02:25,239
Speaker 1: would be like, Hi, I feel like, probably pretty fun.

44
00:02:25,040 --> 00:02:29,000
Speaker 2: Because you know, he's like simmering conservative underneath the surface,

45
00:02:29,320 --> 00:02:31,600
Speaker 2: saying like I'm not good. It would just be like

46
00:02:31,600 --> 00:02:33,760
Speaker 2: getting high with Joe Rogue. It's like I feel like

47
00:02:33,800 --> 00:02:36,880
Speaker 2: it's you know, he'd say something weird and then being

48
00:02:36,919 --> 00:02:39,640
Speaker 2: like he'd probably like make an observation about your physique.

49
00:02:40,160 --> 00:02:42,800
Speaker 2: He'ah good, Dude's like you're right handed. Huh yeah, I

50
00:02:42,800 --> 00:02:44,640
Speaker 2: could tell man based on how your shoulders built. And

51
00:02:44,680 --> 00:02:50,480
Speaker 2: you're like, okay, going on, yeah, yeah.

52
00:02:50,520 --> 00:02:54,280
Speaker 1: Have some workout tips for you. I'm solicited. Don't need that, Miles.

53
00:02:54,400 --> 00:02:57,240
Speaker 1: We are thrilled to be joined by the hosts of

54
00:02:57,400 --> 00:03:02,000
Speaker 1: the Mystery AI Hype Theater three thousand podcast. It's doctor

55
00:03:02,080 --> 00:03:06,600
Speaker 1: Alex Hannah and professor Emily and Bendas.

56
00:03:06,639 --> 00:03:10,040
Speaker 2: Hello, Hello, welcome, Welcome to both of you.

57
00:03:10,560 --> 00:03:14,400
Speaker 1: Yeah, in addition to being podcast hosts, the highest honor

58
00:03:14,560 --> 00:03:18,600
Speaker 1: one can attain in our world, you both have some

59
00:03:18,639 --> 00:03:21,079
Speaker 1: pretty impressive credits, so I just wanted to go through

60
00:03:21,120 --> 00:03:23,760
Speaker 1: those as well. Of top Emily, you are a linguist

61
00:03:23,919 --> 00:03:28,000
Speaker 1: and professor at the University of Washington, where you are

62
00:03:28,360 --> 00:03:34,480
Speaker 1: director of the Computational Linguistics Laboratory. Alex you are director

63
00:03:34,600 --> 00:03:39,000
Speaker 1: of research at the Distributed AI Research Institute. You're both

64
00:03:39,040 --> 00:03:42,440
Speaker 1: widely published, you both received a number of academic awards,

65
00:03:42,440 --> 00:03:46,760
Speaker 1: you both have PhDs. What are you doing on our podcast?

66
00:03:47,400 --> 00:03:51,840
Speaker 1: Is how what you like? Our booker is really good

67
00:03:51,840 --> 00:03:54,960
Speaker 1: shout out to super producer Victor, But this is I

68
00:03:55,000 --> 00:03:56,240
Speaker 1: don't know what's going on here.

69
00:03:56,320 --> 00:03:59,400
Speaker 3: Yeah, we're excited to be here. And part of what's

70
00:03:59,440 --> 00:04:00,960
Speaker 3: going on is that, you know, we're talking to the

71
00:04:01,000 --> 00:04:04,280
Speaker 3: world about how AI, so called AI isn't all that right,

72
00:04:04,320 --> 00:04:06,200
Speaker 3: and so a chance to have this conversation is really

73
00:04:06,200 --> 00:04:07,040
Speaker 3: helpful and important.

74
00:04:07,240 --> 00:04:10,320
Speaker 1: We've been talking about that for a while, but you

75
00:04:10,360 --> 00:04:13,080
Speaker 1: are listening to your podcast really drove some things home.

76
00:04:13,120 --> 00:04:15,279
Speaker 1: So I'm very excited about this conversation.

77
00:04:15,680 --> 00:04:18,720
Speaker 2: Ar Yeah. Every time we have like we've had doctor

78
00:04:18,839 --> 00:04:20,960
Speaker 2: Kerrie mcinernie on before. We think it's been on your

79
00:04:20,960 --> 00:04:24,119
Speaker 2: show as well, every time we pick and uh juau

80
00:04:24,279 --> 00:04:27,360
Speaker 2: sayak out in New York. We were just always talking

81
00:04:27,400 --> 00:04:31,000
Speaker 2: about like our own evolution with how we felt about AI,

82
00:04:31,120 --> 00:04:33,880
Speaker 2: because before I was like, it's gonna take all the

83
00:04:34,000 --> 00:04:37,000
Speaker 2: writing jobs, and then people are like it's not. I'm like,

84
00:04:37,040 --> 00:04:39,640
Speaker 2: how do you know, Like here's some more information. I'm like,

85
00:04:40,040 --> 00:04:45,479
Speaker 2: it's vaporware to make money. So it's always nice, like, yeah,

86
00:04:45,520 --> 00:04:48,120
Speaker 2: it's always nice to kind of keep, you know, pushing

87
00:04:48,120 --> 00:04:51,240
Speaker 2: my own understanding along because I also encounter, you know,

88
00:04:51,240 --> 00:04:53,040
Speaker 2: a lot of friends and family who also kind of

89
00:04:53,080 --> 00:04:55,200
Speaker 2: have the same thing. There's still at to like, this

90
00:04:55,240 --> 00:04:57,760
Speaker 2: stuff looks like it's gonna change the world forever in

91
00:04:57,800 --> 00:05:00,520
Speaker 2: a way we'll never know. And yes, so it's always

92
00:05:00,560 --> 00:05:04,320
Speaker 2: great to have the input of actual learned experts on

93
00:05:04,360 --> 00:05:06,000
Speaker 2: the time learned expertise.

94
00:05:07,320 --> 00:05:10,400
Speaker 1: So we're going to get into all of that, but

95
00:05:10,520 --> 00:05:12,920
Speaker 1: before we do, we like to get to know our

96
00:05:12,960 --> 00:05:16,080
Speaker 1: guests a little bit better by asking you, what is

97
00:05:16,120 --> 00:05:20,800
Speaker 1: something from your search history that is revealing about who

98
00:05:20,839 --> 00:05:21,320
Speaker 1: you guys are.

99
00:05:21,680 --> 00:05:24,120
Speaker 4: I have an exciting one I can start, Hey, this

100
00:05:24,240 --> 00:05:28,919
Speaker 4: is Alex. Yeah, so I'm building a chicken coop because

101
00:05:29,320 --> 00:05:31,680
Speaker 4: my second job, rather than more one of my third

102
00:05:31,800 --> 00:05:36,000
Speaker 4: jobs is kind of being a little suburban farmer. And

103
00:05:36,120 --> 00:05:40,240
Speaker 4: so I'm getting some chicks delivered from a hatchery. And

104
00:05:40,800 --> 00:05:44,320
Speaker 4: they said you need to put this directly into the browder,

105
00:05:44,480 --> 00:05:47,720
Speaker 4: and I'm like, what's a brooder? And I had looked

106
00:05:47,760 --> 00:05:50,360
Speaker 4: up what a brooder is and then came up and

107
00:05:50,400 --> 00:05:53,440
Speaker 4: found that it can be anything from a rubber maide

108
00:05:53,480 --> 00:05:56,000
Speaker 4: tub where you put the chickens while they're very small,

109
00:05:56,880 --> 00:06:01,039
Speaker 4: until like a repurposed rabbit hut. So then I started

110
00:06:01,040 --> 00:06:04,839
Speaker 4: thinking all morning about how to build a chicken Breweder's.

111
00:06:05,080 --> 00:06:07,279
Speaker 2: And wait, just so, it's just like a like it's

112
00:06:07,279 --> 00:06:09,880
Speaker 2: like a pen, like a mini pen for the chien

113
00:06:10,160 --> 00:06:11,000
Speaker 2: to just kind of.

114
00:06:11,720 --> 00:06:13,800
Speaker 4: Yeah, it's like a little pen for the chicks, and

115
00:06:13,839 --> 00:06:15,760
Speaker 4: you put in a heat lamp, you know, because they

116
00:06:15,760 --> 00:06:18,760
Speaker 4: need to be warm. Small, and they don't. They have

117
00:06:18,800 --> 00:06:21,320
Speaker 4: only got that chicken fuzz. They don't have the chicken feathers.

118
00:06:21,600 --> 00:06:23,400
Speaker 2: Yeah, yeah, yeah, they.

119
00:06:23,279 --> 00:06:26,400
Speaker 1: Need to me this sounds like a job for a

120
00:06:26,680 --> 00:06:32,120
Speaker 1: shoe box. But that's probably too small, too small, small, Unfortunately,

121
00:06:32,160 --> 00:06:34,839
Speaker 1: that's what they're getting. If I'm building a chicken, you

122
00:06:34,839 --> 00:06:37,400
Speaker 1: can also what I got if you're in if you're

123
00:06:37,440 --> 00:06:39,440
Speaker 1: in a if you're in a rush, you can use

124
00:06:39,480 --> 00:06:41,440
Speaker 1: your bathtub if you have one of those, and put

125
00:06:42,360 --> 00:06:44,280
Speaker 1: some puppy pads. Yeah.

126
00:06:44,320 --> 00:06:47,200
Speaker 4: I learned a lot about this today as I was

127
00:06:47,279 --> 00:06:49,680
Speaker 4: just waiting for another meeting to start and just read

128
00:06:49,720 --> 00:06:50,200
Speaker 4: everything I.

129
00:06:50,200 --> 00:06:53,640
Speaker 1: Could nice and of course that chicken coop is gonna

130
00:06:53,680 --> 00:06:57,080
Speaker 1: come in handy free eggs during the robot apocalypse, right,

131
00:06:57,160 --> 00:07:00,359
Speaker 1: which Sam Altman has told me is coming. So I

132
00:07:00,400 --> 00:07:03,280
Speaker 1: have to assume that's also what you're thinking, right, right.

133
00:07:03,279 --> 00:07:06,440
Speaker 2: I mean she's prepping, she's prepping for the eventual DEMID. Yeah.

134
00:07:06,520 --> 00:07:07,440
Speaker 1: I think during.

135
00:07:07,240 --> 00:07:09,679
Speaker 4: COVID, I think there was a run on chicken eggs,

136
00:07:09,760 --> 00:07:11,920
Speaker 4: and then so I'd go to Costco and you couldn't

137
00:07:11,920 --> 00:07:15,560
Speaker 4: get the gross of chicken eggs, And then I'd go, hey,

138
00:07:15,600 --> 00:07:17,360
Speaker 4: and I went to my backyard.

139
00:07:20,680 --> 00:07:23,080
Speaker 2: Market got them for eight bucks an egg if you

140
00:07:23,120 --> 00:07:23,320
Speaker 2: want to.

141
00:07:23,480 --> 00:07:26,680
Speaker 1: Yeah, yeah, they are still really expensive. Emily, how about you?

142
00:07:26,760 --> 00:07:28,200
Speaker 1: What's something from your search history?

143
00:07:28,720 --> 00:07:30,040
Speaker 3: So I took a look, and it's a bunch of

144
00:07:30,040 --> 00:07:32,280
Speaker 3: boring stuff, like I can't remember it can't be bothered

145
00:07:32,280 --> 00:07:34,040
Speaker 3: to remember the website of a certain journal that I

146
00:07:34,080 --> 00:07:35,640
Speaker 3: was interested in. So I'm like searching the name of

147
00:07:35,640 --> 00:07:38,880
Speaker 3: the journal. But then like down below that we've been

148
00:07:38,960 --> 00:07:44,240
Speaker 3: watching for all mankind, which is this like alternate timeline

149
00:07:44,240 --> 00:07:47,520
Speaker 3: thing that that that's a what happens if the Soviets

150
00:07:47,560 --> 00:07:51,000
Speaker 3: landed on the moon first? Right, right, So it diverges

151
00:07:51,040 --> 00:07:54,840
Speaker 3: in the nineteen sixties, but it keeps referencing actual history.

152
00:07:54,880 --> 00:07:57,080
Speaker 3: So my search history is full of like, okay, so

153
00:07:57,120 --> 00:07:58,720
Speaker 3: when did the Vietnam War actually end?

154
00:07:58,920 --> 00:07:59,120
Speaker 1: Right?

155
00:07:59,240 --> 00:08:02,040
Speaker 3: And like know which Apollo mission did what. So it's

156
00:08:02,040 --> 00:08:03,840
Speaker 3: a bunch of queries like that, sort of comparing what's

157
00:08:03,840 --> 00:08:05,400
Speaker 3: in the show to actual history.

158
00:08:05,440 --> 00:08:07,720
Speaker 2: How does that line up based on your sort of

159
00:08:07,880 --> 00:08:09,560
Speaker 2: cursory research as you watch the show?

160
00:08:10,480 --> 00:08:14,360
Speaker 3: So interestingly, So there's there's a point where Ted Kennedy

161
00:08:14,600 --> 00:08:17,080
Speaker 3: is like in the background being talked about not going

162
00:08:17,120 --> 00:08:22,120
Speaker 3: to chap Equittic. Oh, and then an episode later he's president.

163
00:08:22,840 --> 00:08:23,080
Speaker 1: Wow.

164
00:08:24,360 --> 00:08:26,520
Speaker 3: So there's and I suspect there's way more of that

165
00:08:26,600 --> 00:08:28,280
Speaker 3: kind of stuff that I'm not catching right, right.

166
00:08:28,240 --> 00:08:30,120
Speaker 2: Right, just subtle things right right.

167
00:08:30,240 --> 00:08:33,480
Speaker 1: The media is like and Ted Kennedy missed a barbecue

168
00:08:33,480 --> 00:08:39,840
Speaker 1: this weekend in chap Equittic. That's that's super interesting. Yeah,

169
00:08:40,080 --> 00:08:43,640
Speaker 1: this is you know, third or fourth person who's mentioned

170
00:08:43,640 --> 00:08:46,880
Speaker 1: for all mankind to us, I think this is pushing

171
00:08:46,920 --> 00:08:49,839
Speaker 1: it over the threshold to where I have to I

172
00:08:49,920 --> 00:08:51,439
Speaker 1: have to watch this damn thing.

173
00:08:51,840 --> 00:08:54,520
Speaker 3: It's enjoyable, but it's tense. I don't know anything worth

174
00:08:54,600 --> 00:08:58,240
Speaker 3: Like people in outer space really creeps me out because

175
00:08:58,280 --> 00:09:01,040
Speaker 3: like that that degree of like lowliness and sort of

176
00:09:01,240 --> 00:09:05,360
Speaker 3: lack of failsafes, you know, like when all the fail

177
00:09:05,400 --> 00:09:07,160
Speaker 3: says are there are the ones that you built and

178
00:09:07,240 --> 00:09:10,320
Speaker 3: beyond that, you know, your sol that's that's creepy.

179
00:09:10,400 --> 00:09:12,320
Speaker 2: It seems uncomfortable outer space.

180
00:09:12,640 --> 00:09:17,000
Speaker 1: Yeah, what is something, Alex that you think is underrated?

181
00:09:17,480 --> 00:09:19,560
Speaker 4: I was trying to think of something and the only

182
00:09:19,600 --> 00:09:22,360
Speaker 4: thing I could come up with this morning was the

183
00:09:22,480 --> 00:09:24,080
Speaker 4: Nintendo Virtual Boy.

184
00:09:26,240 --> 00:09:28,920
Speaker 2: The red goggles on a tripod.

185
00:09:28,400 --> 00:09:31,360
Speaker 4: The red goggles on a tripod, and I know, you know,

186
00:09:31,920 --> 00:09:35,160
Speaker 4: and whatever. Twenty years later, thirty years later, I think

187
00:09:35,200 --> 00:09:37,760
Speaker 4: at this point, you know, Apple has its Apple Vision

188
00:09:37,800 --> 00:09:41,679
Speaker 4: Pro that looks pretty much the same except you can

189
00:09:41,760 --> 00:09:44,720
Speaker 4: walk around with it right right right. All I remember

190
00:09:44,840 --> 00:09:49,000
Speaker 4: is that I really really really wanted a Nintendo Virtual

191
00:09:49,120 --> 00:09:53,640
Speaker 4: Boy when I was whatever ten or old. I was

192
00:09:53,679 --> 00:09:56,800
Speaker 4: in nineteen ninety five, and you know, but then I

193
00:09:56,880 --> 00:09:59,320
Speaker 4: was reading the Wikipedia page and it kind of said,

194
00:09:59,800 --> 00:10:02,400
Speaker 4: you know, know it it was gonna be people headaches,

195
00:10:02,440 --> 00:10:05,640
Speaker 4: and it was overpriced. And but I'm going to stand

196
00:10:05,720 --> 00:10:09,640
Speaker 4: by my claim that it is underrated. It was ahead

197
00:10:09,640 --> 00:10:12,920
Speaker 4: of its time and it was one of those products

198
00:10:12,960 --> 00:10:17,520
Speaker 4: that completely you know, I don't know, I think, you know,

199
00:10:17,600 --> 00:10:21,240
Speaker 4: if you could go back, maybe it was just maybe

200
00:10:21,400 --> 00:10:23,040
Speaker 4: I don't know, I don't know what they could have

201
00:10:23,080 --> 00:10:26,280
Speaker 4: done differently, but yeah, that's that's all I got.

202
00:10:26,800 --> 00:10:30,240
Speaker 2: Was hell, have you seen? Okay? So I was the

203
00:10:30,280 --> 00:10:33,040
Speaker 2: same way, like I used to subscribe to Nintendo Power

204
00:10:33,120 --> 00:10:35,280
Speaker 2: all that shit. It was such a nerdy game kid,

205
00:10:35,320 --> 00:10:36,960
Speaker 2: and when like those ads for it, it was like

206
00:10:37,000 --> 00:10:40,200
Speaker 2: this is the future. It blew my mind. My parents

207
00:10:40,200 --> 00:10:43,040
Speaker 2: obviously were like, that's a hell no for a month, right,

208
00:10:44,640 --> 00:10:45,880
Speaker 2: how silly you would look?

209
00:10:46,080 --> 00:10:48,280
Speaker 1: We don't yeah deal with or in public with that.

210
00:10:48,480 --> 00:10:50,800
Speaker 4: Yeah, it's also like a couple of grand wasn't It

211
00:10:50,840 --> 00:10:52,160
Speaker 4: was just ridiculously it.

212
00:10:52,200 --> 00:10:54,200
Speaker 2: Was something I think it was like a few It

213
00:10:54,240 --> 00:10:57,280
Speaker 2: was just something like more than like what a PlayStation cost,

214
00:10:57,480 --> 00:10:59,560
Speaker 2: which was sort of like the height of it or

215
00:10:59,600 --> 00:11:01,920
Speaker 2: something like. Anyway, I remember a kid in my school

216
00:11:02,240 --> 00:11:06,160
Speaker 2: had one, brought it to school and we lined up

217
00:11:06,200 --> 00:11:08,680
Speaker 2: to play it, and it was the most underwhelming experience

218
00:11:08,720 --> 00:11:13,400
Speaker 2: that like it broke because like there's nothing VR about this.

219
00:11:13,520 --> 00:11:16,840
Speaker 2: It's like lightly three D everything I feel like in

220
00:11:16,840 --> 00:11:20,320
Speaker 2: this like monochromatic, like red scale kind of graphic thing.

221
00:11:20,720 --> 00:11:24,840
Speaker 2: It just was really it was underwhelming, But I completely

222
00:11:24,880 --> 00:11:26,800
Speaker 2: follow the same path you had. Alex and being like

223
00:11:26,880 --> 00:11:28,280
Speaker 2: this is I need this?

224
00:11:28,280 --> 00:11:29,560
Speaker 4: This was the future?

225
00:11:29,880 --> 00:11:32,959
Speaker 2: Yeah yeah yeah, and yeah, like that's so funny. And

226
00:11:32,960 --> 00:11:35,160
Speaker 2: never even thought about the Vision pro as being like

227
00:11:35,240 --> 00:11:38,120
Speaker 2: the like the spiritual sequel.

228
00:11:38,600 --> 00:11:42,520
Speaker 4: Exactly, it's the direct descendant. Yeah, you know the Nintendo

229
00:11:43,360 --> 00:11:47,880
Speaker 4: Virtual Boy crawled, so the Apple Vision bro exactly right.

230
00:11:48,600 --> 00:11:51,600
Speaker 1: How about you, Emily, what is something you think is underrated?

231
00:11:52,320 --> 00:11:54,319
Speaker 3: So after listening to a couple of your episodes and

232
00:11:54,400 --> 00:11:56,959
Speaker 3: like all the nineties nostalgia, I have to say gen x.

233
00:11:57,600 --> 00:12:01,160
Speaker 2: Xx, Yeah, people I thought were the coolest.

234
00:12:01,760 --> 00:12:07,559
Speaker 3: We've got some monsters among us, you know, yeah, yeah,

235
00:12:07,600 --> 00:12:10,520
Speaker 3: but you know gen X were small. We're scrappy. I

236
00:12:10,559 --> 00:12:12,920
Speaker 3: saw someone a millennial that I know, online posting something

237
00:12:12,960 --> 00:12:14,640
Speaker 3: about how weird it is to talk to gen X

238
00:12:14,640 --> 00:12:18,560
Speaker 3: folks about their internet experiences are a back. Don't cite

239
00:12:18,640 --> 00:12:19,680
Speaker 3: the old magic to me.

240
00:12:22,760 --> 00:12:23,920
Speaker 2: I was there when.

241
00:12:23,760 --> 00:12:26,160
Speaker 4: It was created. Although I feel like I feel like

242
00:12:26,200 --> 00:12:29,320
Speaker 4: it's it is. It's a bit meta, Emily, because I

243
00:12:29,360 --> 00:12:32,680
Speaker 4: feel like gen xers are always going to say that

244
00:12:32,720 --> 00:12:36,360
Speaker 4: they're underrated, like they're underappreciated, and it's it's kind of

245
00:12:36,400 --> 00:12:36,960
Speaker 4: a class.

246
00:12:37,120 --> 00:12:38,880
Speaker 3: It's our whole identity exactly.

247
00:12:39,640 --> 00:12:40,280
Speaker 2: It is.

248
00:12:40,320 --> 00:12:43,400
Speaker 4: It is a referential Yeah, it's a class feature of

249
00:12:43,520 --> 00:12:47,800
Speaker 4: gen xers to say that they are unappreciated. And you know,

250
00:12:48,760 --> 00:12:51,719
Speaker 4: you children have to struggle with AOL dial up. You know,

251
00:12:52,040 --> 00:12:56,360
Speaker 4: how about this, you know, prior no internet or yeah, I.

252
00:12:56,280 --> 00:12:58,080
Speaker 3: Mean unless you you don't even know how to read

253
00:12:58,080 --> 00:12:58,400
Speaker 3: the math.

254
00:13:02,679 --> 00:13:04,840
Speaker 2: Yeah, yeah, I grew up in l A. I know

255
00:13:04,840 --> 00:13:07,120
Speaker 2: how to use a Thompson Guide or Thomas Guide.

256
00:13:07,200 --> 00:13:07,760
Speaker 1: That's that was.

257
00:13:07,880 --> 00:13:12,240
Speaker 2: That was our Google Maps before anything. But yeah, yeah, exactly.

258
00:13:12,400 --> 00:13:14,560
Speaker 2: But yeah, like I think about too, like as like

259
00:13:14,600 --> 00:13:16,839
Speaker 2: an older millennial, like gen X were all the people

260
00:13:16,840 --> 00:13:20,640
Speaker 2: I thought were the coolest people growing up, Like this

261
00:13:20,840 --> 00:13:22,600
Speaker 2: is like I want to be I want to be

262
00:13:22,640 --> 00:13:26,840
Speaker 2: a hacker. I want to be like them of oh yeah,

263
00:13:26,880 --> 00:13:28,319
Speaker 2: well hey, I don't know what are we going to

264
00:13:28,360 --> 00:13:31,120
Speaker 2: say about ourselves as millennials. I wonder I think we're

265
00:13:31,160 --> 00:13:34,080
Speaker 2: just like we're just dead inside on some level that

266
00:13:34,120 --> 00:13:36,600
Speaker 2: we're like, yeah, whatever, cool we are.

267
00:13:37,000 --> 00:13:42,000
Speaker 4: We're dead inside and we killed everything, right that the

268
00:13:42,040 --> 00:13:47,120
Speaker 4: stock you know, millennials are killing, you know, the napkin industry.

269
00:13:47,120 --> 00:13:50,480
Speaker 1: Right, but the toast is going to be so good, Yeah.

270
00:13:50,800 --> 00:13:54,079
Speaker 4: It's going to be amazing. We you know, we killed

271
00:13:54,360 --> 00:13:59,280
Speaker 4: housing somehow because we spent too much on avocado toast. Yes,

272
00:13:59,320 --> 00:14:04,280
Speaker 4: and you know, you know tumoric lattes or whatever, those

273
00:14:04,320 --> 00:14:06,600
Speaker 4: are good, Like they're good.

274
00:14:06,440 --> 00:14:08,959
Speaker 3: Though maybe maybe even underrated.

275
00:14:09,880 --> 00:14:12,319
Speaker 1: There will be two people will be proposing to each

276
00:14:12,320 --> 00:14:17,040
Speaker 1: other with rings made of turmeric and avocado instead of diamonds,

277
00:14:17,240 --> 00:14:19,680
Speaker 1: because you build the diamond industry.

278
00:14:19,360 --> 00:14:24,000
Speaker 2: Our new currency. Yes, yeah, but we always talk about that.

279
00:14:24,040 --> 00:14:28,400
Speaker 2: Framing is just basically millennials can't afford X thing, right,

280
00:14:28,640 --> 00:14:31,880
Speaker 2: we can't. We're not killing the diamond industry. We don't

281
00:14:31,960 --> 00:14:35,560
Speaker 2: have money for diamond industry. We don't have money for these.

282
00:14:35,400 --> 00:14:39,200
Speaker 1: Other millennials have this weird trend of living five to

283
00:14:39,360 --> 00:14:45,120
Speaker 1: an apartment because they love it. Yeah, what, Alex, what's

284
00:14:45,120 --> 00:14:46,240
Speaker 1: something you think is overrated?

285
00:14:47,560 --> 00:14:50,600
Speaker 4: Well, we're we're all that, We're all the AI stuff

286
00:14:50,640 --> 00:14:54,440
Speaker 4: and so you know we're both Emily and I are

287
00:14:54,440 --> 00:14:58,160
Speaker 4: gonna talk about parts of it. But image generators for sure.

288
00:14:58,800 --> 00:15:03,400
Speaker 4: Text the image generator, I mean people think they're very flashy,

289
00:15:03,520 --> 00:15:06,480
Speaker 4: but I mean it does a lot of copying. You know,

290
00:15:06,800 --> 00:15:09,600
Speaker 4: I was going to talk at a San Francisco public

291
00:15:09,640 --> 00:15:15,240
Speaker 4: library with Carlo Ortiz, who's a concept artist, and she

292
00:15:15,320 --> 00:15:17,360
Speaker 4: had all these examples in which you know, give it

293
00:15:17,400 --> 00:15:20,640
Speaker 4: a prompt and it would literally just copy kind of

294
00:15:21,720 --> 00:15:25,040
Speaker 4: a game art and then put it on the new

295
00:15:25,080 --> 00:15:27,520
Speaker 4: thing and have some kind of swivels around it. But

296
00:15:27,640 --> 00:15:30,840
Speaker 4: the stuff is just I mean, and it's also I mean,

297
00:15:30,880 --> 00:15:34,160
Speaker 4: the the kind of images the thing, it's this this

298
00:15:34,400 --> 00:15:39,240
Speaker 4: distinctive style that esthetically, apart from all the problems with

299
00:15:39,320 --> 00:15:42,440
Speaker 4: the stuff and the non consensual use and the data

300
00:15:42,480 --> 00:15:46,520
Speaker 4: theft and all the awful kinds of you know, non

301
00:15:46,560 --> 00:15:51,200
Speaker 4: consensual deep fake porn and the kind of like far

302
00:15:51,360 --> 00:15:54,680
Speaker 4: right imagery, like, the stuff is just ugly, Like it's

303
00:15:54,680 --> 00:15:57,520
Speaker 4: got this like every every person in it just looks sunken.

304
00:15:58,040 --> 00:16:00,680
Speaker 4: They look like they've just seen some shit, you know,

305
00:16:00,800 --> 00:16:01,480
Speaker 4: they've got some.

306
00:16:01,480 --> 00:16:04,920
Speaker 1: Showers like it's a mistare.

307
00:16:04,800 --> 00:16:08,360
Speaker 4: Yeah, just just glassy, and you're just you're just like

308
00:16:08,440 --> 00:16:09,840
Speaker 4: what is the appeal?

309
00:16:10,640 --> 00:16:10,880
Speaker 1: You know?

310
00:16:11,160 --> 00:16:13,680
Speaker 4: And and so I just yeah, I just I don't

311
00:16:13,760 --> 00:16:14,080
Speaker 4: like it.

312
00:16:14,560 --> 00:16:18,240
Speaker 2: Yeah, I think every so many things when people try

313
00:16:18,280 --> 00:16:20,480
Speaker 2: and do like real life, because I look at like

314
00:16:20,520 --> 00:16:23,320
Speaker 2: the mid Journey subreddit on Reddit to see like what

315
00:16:23,360 --> 00:16:25,360
Speaker 2: people are making, and that was sort of like my entermre,

316
00:16:25,440 --> 00:16:27,640
Speaker 2: Like wow, these like like when people get the prompt. Right,

317
00:16:27,680 --> 00:16:31,400
Speaker 2: it's interesting. They're like Simpsons but real life characters and

318
00:16:31,560 --> 00:16:36,080
Speaker 2: also from but from a Korean drama. Like okay, so

319
00:16:36,120 --> 00:16:39,600
Speaker 2: we're getting the Korean like the k drama irl Simpsons,

320
00:16:40,080 --> 00:16:42,440
Speaker 2: but all the the aesthetic like all looks like like

321
00:16:42,440 --> 00:16:45,520
Speaker 2: sort of like David la Chappelle's photography. It's weird. There's

322
00:16:45,560 --> 00:16:48,920
Speaker 2: like this hyper stylized look to it that feels very specific,

323
00:16:49,520 --> 00:16:51,640
Speaker 2: And I think the thing that interests me is sort

324
00:16:51,640 --> 00:16:54,520
Speaker 2: of like, as someone who's terrible at visual art, it's

325
00:16:54,600 --> 00:16:57,760
Speaker 2: like this way for someone with absolutely no talent in

326
00:16:57,800 --> 00:17:01,120
Speaker 2: that area to be like I summon this thing I'm

327
00:17:01,200 --> 00:17:04,000
Speaker 2: thinking of and you're like, fine, it's a bunch of copies.

328
00:17:04,000 --> 00:17:06,080
Speaker 2: But like I think that's the thing that most people

329
00:17:06,080 --> 00:17:10,000
Speaker 2: are like, oh cool, right, it made the thing.

330
00:17:10,359 --> 00:17:14,320
Speaker 4: It made the thing it made mediocre like art.

331
00:17:14,640 --> 00:17:19,480
Speaker 1: I had the thought AI is like a doesn't matter generator,

332
00:17:19,840 --> 00:17:23,000
Speaker 1: like Star Trek has the matter generator, and it is

333
00:17:23,119 --> 00:17:26,119
Speaker 1: like ship out a bunch of stuff that doesn't matter.

334
00:17:26,280 --> 00:17:29,000
Speaker 1: That's like uninspired, but it like matters to you for

335
00:17:29,080 --> 00:17:32,199
Speaker 1: a split second. You know, it's like, hey, wow, that

336
00:17:32,320 --> 00:17:34,240
Speaker 1: is that's weird that like that just came from a

337
00:17:34,280 --> 00:17:37,720
Speaker 1: text prompt, but it's it actually doesn't matter in any

338
00:17:37,960 --> 00:17:39,159
Speaker 1: long term sense.

339
00:17:39,240 --> 00:17:42,200
Speaker 4: Right, Yeah, that's I love that framing. I mean, what's

340
00:17:42,320 --> 00:17:46,000
Speaker 4: that famous that the the like image of the like

341
00:17:46,440 --> 00:17:48,399
Speaker 4: the diner, Yeah.

342
00:17:48,280 --> 00:17:50,320
Speaker 1: The nighthawks. Is that what it's called?

343
00:17:50,480 --> 00:17:54,800
Speaker 4: Yeah, yeah, yeah, yeah yeah yeah. And someone had this

344
00:17:54,960 --> 00:17:58,880
Speaker 4: great thread. Well, no, it was a troll thread though,

345
00:17:59,000 --> 00:18:00,679
Speaker 4: you know what I'm talking about. There was like the

346
00:18:00,800 --> 00:18:04,119
Speaker 4: nighthawks that the image and and and then someone was

347
00:18:04,200 --> 00:18:07,120
Speaker 4: like I thought it, like, look at this composition. It's

348
00:18:07,119 --> 00:18:09,840
Speaker 4: so boring. What if they were happier? And it was

349
00:18:09,920 --> 00:18:12,000
Speaker 4: kind of like a troll thread. I think it was, Like.

350
00:18:12,560 --> 00:18:14,720
Speaker 3: I thought, I thought it was genuine. I thought that

351
00:18:14,840 --> 00:18:16,680
Speaker 3: they were really saying I'm making it better with AI.

352
00:18:17,119 --> 00:18:19,000
Speaker 4: I think it was. I think it was a troll thread.

353
00:18:19,080 --> 00:18:21,680
Speaker 4: It's like, why are these people so sad.

354
00:18:21,680 --> 00:18:22,919
Speaker 2: When if they were happy?

355
00:18:23,280 --> 00:18:26,639
Speaker 4: What if it was daytime? Yeah, it was. It was

356
00:18:26,680 --> 00:18:30,080
Speaker 4: sort of like that, like you know, and I thought

357
00:18:30,080 --> 00:18:33,480
Speaker 4: it was a troll thread. Maybe I'm granting posters like

358
00:18:33,520 --> 00:18:36,960
Speaker 4: a little too much like Grace here, but I think

359
00:18:37,000 --> 00:18:40,000
Speaker 4: it was like the idea of like I made it interesting.

360
00:18:40,119 --> 00:18:40,600
Speaker 4: Look at this?

361
00:18:40,760 --> 00:18:41,800
Speaker 2: Why is it so dark?

362
00:18:42,119 --> 00:18:45,399
Speaker 1: Yeah? Right? What if there was confetti.

363
00:18:46,240 --> 00:18:49,000
Speaker 2: I know, what if like a swagged out pope was

364
00:18:49,040 --> 00:18:53,920
Speaker 2: also sitting at.

365
00:18:54,280 --> 00:19:00,520
Speaker 1: So much can image, but the swinded out pope did though.

366
00:19:00,600 --> 00:19:03,119
Speaker 4: I thought it was real, and I was and then

367
00:19:03,160 --> 00:19:06,080
Speaker 4: I was like, this is such a cool But again, it's.

368
00:19:05,920 --> 00:19:08,240
Speaker 1: Not like chat GPT was like, you know it would

369
00:19:08,240 --> 00:19:10,920
Speaker 1: be funny. It was somebody was like, you know what

370
00:19:10,960 --> 00:19:13,639
Speaker 1: would be funny and gave that prompt to chat GBT

371
00:19:13,960 --> 00:19:17,440
Speaker 1: and then it ended up being But like that's that's

372
00:19:17,440 --> 00:19:19,600
Speaker 1: the thing that you guys talk about a lot, is

373
00:19:19,720 --> 00:19:23,159
Speaker 1: just the erasure of like people are like, well and

374
00:19:23,240 --> 00:19:26,520
Speaker 1: here's chat GBT doing this thing, and it's like that

375
00:19:26,800 --> 00:19:29,840
Speaker 1: somebody told it to do and then programmed it to do,

376
00:19:30,280 --> 00:19:33,520
Speaker 1: like it's doing a specific It is not a person

377
00:19:33,800 --> 00:19:34,520
Speaker 1: who is coming up.

378
00:19:35,440 --> 00:19:37,879
Speaker 2: It's like talking about like Picasso's paint brush, but just

379
00:19:37,920 --> 00:19:40,080
Speaker 2: as the paint brush, like look what the paint brush did,

380
00:19:40,600 --> 00:19:43,160
Speaker 2: Oh my god, dude, which is cool.

381
00:19:43,240 --> 00:19:47,040
Speaker 1: And then there's like a little like Buddhism there that

382
00:19:47,080 --> 00:19:50,320
Speaker 1: I appreciate, but I feel like that's not where it's

383
00:19:50,320 --> 00:19:53,880
Speaker 1: headed unfortunately. Emily, do you have something that you think

384
00:19:54,119 --> 00:19:55,000
Speaker 1: is overrated?

385
00:19:55,560 --> 00:19:57,640
Speaker 3: Yeah? I mean the short answer is large language models

386
00:19:57,640 --> 00:19:59,720
Speaker 3: are overrated, But I think we're going to get into

387
00:19:59,760 --> 00:20:02,600
Speaker 3: that we are probably, so I'm going to shift to

388
00:20:02,640 --> 00:20:04,680
Speaker 3: my secondary answer, which is actually to take a page

389
00:20:04,680 --> 00:20:08,359
Speaker 3: from Alex's work to say that scale is overrepresented. That

390
00:20:08,840 --> 00:20:11,040
Speaker 3: if the goal of taking something and scaling it to

391
00:20:11,080 --> 00:20:14,640
Speaker 3: millions of people is like the thing, the only folks

392
00:20:14,720 --> 00:20:18,120
Speaker 3: really benefiting from that are the capitalists behind it. Right,

393
00:20:18,160 --> 00:20:21,680
Speaker 3: the product is worse. The impact on the societies or

394
00:20:21,720 --> 00:20:24,080
Speaker 3: the communities that lost access to whatever their local solution

395
00:20:24,320 --> 00:20:27,119
Speaker 3: was is worse. Right, So scale is the thing that

396
00:20:27,160 --> 00:20:30,000
Speaker 3: so many people are chasing, especially in the Bay Area,

397
00:20:30,040 --> 00:20:32,840
Speaker 3: but also up here in Seattle, and it's.

398
00:20:32,960 --> 00:20:36,920
Speaker 2: Way over it all over in every place. Everyone will

399
00:20:36,960 --> 00:20:45,720
Speaker 2: hear that all the time, and like, yeah, yeah.

400
00:20:44,359 --> 00:20:46,840
Speaker 1: I worked on a website that had like really was

401
00:20:47,040 --> 00:20:49,959
Speaker 1: having really fast growth, and you know, it is just

402
00:20:50,000 --> 00:20:52,840
Speaker 1: based on like publishing these three articles a day and

403
00:20:52,880 --> 00:20:55,920
Speaker 1: really like focusing on making them as good as possible.

404
00:20:56,359 --> 00:21:01,400
Speaker 1: And unfortunately, once it got a lot of then the

405
00:21:01,600 --> 00:21:05,240
Speaker 1: executives came in and their first question was how do

406
00:21:05,280 --> 00:21:07,280
Speaker 1: we scale this? How do we scale this? Though, like,

407
00:21:07,280 --> 00:21:09,159
Speaker 1: how do we get a hundred articles a day? And

408
00:21:09,200 --> 00:21:12,199
Speaker 1: I was like, back in two thousand and ten. You know,

409
00:21:12,480 --> 00:21:17,440
Speaker 1: but it's just been how capitalism thinks about this from

410
00:21:17,480 --> 00:21:20,320
Speaker 1: the beginning. Yeah, and now they're doing it with movies.

411
00:21:20,359 --> 00:21:22,840
Speaker 1: They're like, who needs movies? Now I could make myself

412
00:21:22,920 --> 00:21:26,000
Speaker 1: the star of Oppenheimer. It's like, why would you want

413
00:21:26,040 --> 00:21:29,040
Speaker 1: to do what.

414
00:21:28,680 --> 00:21:31,120
Speaker 4: You seems like you missed the point of that movie.

415
00:21:31,520 --> 00:21:34,159
Speaker 2: Okay, I want to be destroyer of world.

416
00:21:34,640 --> 00:21:39,480
Speaker 1: Oh that may feel strong? Yeah, yeah, all right, Well

417
00:21:39,560 --> 00:21:41,439
Speaker 1: let's take a quick break and we'll come back and

418
00:21:41,440 --> 00:21:55,280
Speaker 1: get into it. We'll be right back, and we're back.

419
00:21:55,359 --> 00:21:58,960
Speaker 1: We're back and yeah, So, as we mentioned, you know,

420
00:21:59,000 --> 00:22:01,320
Speaker 1: we've done some episodes on how there's a lot of

421
00:22:01,359 --> 00:22:06,119
Speaker 1: hype around it. Your podcast really helped me understand just

422
00:22:06,160 --> 00:22:10,560
Speaker 1: how much of the current AI crazes really hype. There's

423
00:22:10,560 --> 00:22:15,640
Speaker 1: this one anecdote I just wanted to mention where a

424
00:22:15,680 --> 00:22:20,560
Speaker 1: physicist who works for a car company was brought into

425
00:22:20,600 --> 00:22:26,199
Speaker 1: a meeting to talk about hydraulic breaks and one of

426
00:22:26,240 --> 00:22:30,320
Speaker 1: the executives of the car company asked them if putting

427
00:22:30,359 --> 00:22:34,360
Speaker 1: some AI in the brakes wouldn't help them operate more smoothly,

428
00:22:34,880 --> 00:22:39,720
Speaker 1: which they were kind of confounded, And we're like, what

429
00:22:40,440 --> 00:22:42,720
Speaker 1: does what do you like, I guess that gets to

430
00:22:42,760 --> 00:22:44,360
Speaker 1: the big question I have a lot of the time

431
00:22:44,400 --> 00:22:46,560
Speaker 1: with AI hype in the media, which is, what do

432
00:22:46,600 --> 00:22:48,720
Speaker 1: these people think AI is?

433
00:22:49,320 --> 00:22:51,520
Speaker 3: It's magic fairy dust, But you actually made it slightly

434
00:22:51,520 --> 00:22:55,480
Speaker 3: more plausible by making it breaks. The story was pistons pistons, okay,

435
00:22:55,560 --> 00:23:02,560
Speaker 3: and the engineer was like, it's metal. There's no circus here.

436
00:23:02,600 --> 00:23:07,320
Speaker 1: Right, Yeah, But yeah, I mean, there are three kind

437
00:23:07,359 --> 00:23:09,720
Speaker 1: of big ideas that I got a new perspective on

438
00:23:10,119 --> 00:23:12,159
Speaker 1: from your show that I wanted to start with. The

439
00:23:12,720 --> 00:23:16,560
Speaker 1: first one is kind of this distinction between limited and

440
00:23:16,760 --> 00:23:21,959
Speaker 1: general intelligence or like general AI, which is something that

441
00:23:22,400 --> 00:23:25,119
Speaker 1: I've heard mentioned a lot in the past couple of

442
00:23:25,160 --> 00:23:30,600
Speaker 1: years in articles about chat GPT. They're basically trying to

443
00:23:30,680 --> 00:23:33,280
Speaker 1: use this as a distinction to say, look, we've known

444
00:23:34,280 --> 00:23:37,600
Speaker 1: that computers can beat a chess master for a while now,

445
00:23:38,200 --> 00:23:43,159
Speaker 1: but the distinction that we are trying to sell to

446
00:23:43,200 --> 00:23:47,359
Speaker 1: you is that this one is different because it's a

447
00:23:47,480 --> 00:23:51,560
Speaker 1: general intelligence. It can kind of reason its way into

448
00:23:51,680 --> 00:23:56,560
Speaker 1: doing lots of things. Yeah, and one of the details

449
00:23:56,600 --> 00:24:00,600
Speaker 1: I heard you point out on your show is that

450
00:24:00,680 --> 00:24:05,440
Speaker 1: it's still pretty limited, like it's still basically doing a

451
00:24:05,480 --> 00:24:09,920
Speaker 1: single thing pretty well, like that the chat GPT is

452
00:24:10,320 --> 00:24:13,159
Speaker 1: can you can you talk about that, like what is it?

453
00:24:13,480 --> 00:24:17,399
Speaker 1: What is that distinction and how kind of imaginary it is?

454
00:24:17,880 --> 00:24:20,399
Speaker 3: Yeah, So before Alice goes off on chess as a

455
00:24:20,440 --> 00:24:23,720
Speaker 3: thing in this and we'll get there, Alex, you're right

456
00:24:23,760 --> 00:24:25,880
Speaker 3: that chat GYPTA does one thing well. But the one

457
00:24:25,880 --> 00:24:29,360
Speaker 3: thing that's doing well is coming up with plausible sequences

458
00:24:29,400 --> 00:24:29,800
Speaker 3: of words.

459
00:24:30,119 --> 00:24:30,320
Speaker 2: Right.

460
00:24:30,400 --> 00:24:32,679
Speaker 3: The problem is that those plausible sequences of words can

461
00:24:32,680 --> 00:24:35,040
Speaker 3: be on any topic. So it looks like it is

462
00:24:35,080 --> 00:24:38,440
Speaker 3: doing well at playing trivia games and being a search

463
00:24:38,440 --> 00:24:42,760
Speaker 3: engine and writing legal documents and giving medical diagnoses, and

464
00:24:43,240 --> 00:24:45,560
Speaker 3: the less you know about the stuff you're asking it,

465
00:24:45,720 --> 00:24:48,359
Speaker 3: the more plausible and authoritative it sounds. But in fact,

466
00:24:48,400 --> 00:24:51,560
Speaker 3: underlyingly it's doing just one thing, predicting a plausible sequence

467
00:24:51,560 --> 00:24:52,800
Speaker 3: of words, right.

468
00:24:52,920 --> 00:24:53,280
Speaker 1: Yeah.

469
00:24:53,359 --> 00:24:56,040
Speaker 4: And the thing about chess is that it wasn't too

470
00:24:56,080 --> 00:25:00,800
Speaker 4: long ago that general intelligence meant chess playing, you know,

471
00:25:01,000 --> 00:25:04,080
Speaker 4: and so yeah, you had you had these you know,

472
00:25:04,200 --> 00:25:07,760
Speaker 4: during the Cold War, they would have these machines that

473
00:25:07,800 --> 00:25:10,000
Speaker 4: could play chess, and then that was supposed to be

474
00:25:10,080 --> 00:25:14,320
Speaker 4: a substitute for general AI. They thought that was general

475
00:25:14,359 --> 00:25:15,240
Speaker 4: intelligence and it was.

476
00:25:15,240 --> 00:25:15,800
Speaker 2: A big deal.

477
00:25:15,880 --> 00:25:20,600
Speaker 4: When you know, IBM Watson beat Gary Kasparov, and you

478
00:25:20,640 --> 00:25:26,920
Speaker 4: know IBM Watson was that deep people, Yes, yes, Watson

479
00:25:27,040 --> 00:25:30,400
Speaker 4: Watson won Jeopardy, Jeopardy, yeah it it beat it, beat

480
00:25:30,480 --> 00:25:34,000
Speaker 4: Ken Jennings and then so yeah, thank you. It was

481
00:25:34,040 --> 00:25:37,119
Speaker 4: another one those IBM. And so then that was the

482
00:25:37,240 --> 00:25:39,359
Speaker 4: kind of that was that was the bar. And then

483
00:25:39,359 --> 00:25:41,000
Speaker 4: they're like, well, Tess is now too easy.

484
00:25:41,040 --> 00:25:41,560
Speaker 2: We have to do.

485
00:25:42,040 --> 00:25:44,600
Speaker 4: We have to do go go, you know. And then

486
00:25:44,760 --> 00:25:46,800
Speaker 4: and then they got it into some real time strategy

487
00:25:46,800 --> 00:25:50,840
Speaker 4: games like like StarCraft and and Doda too, and so

488
00:25:51,440 --> 00:25:54,200
Speaker 4: you know, so there's always been this kind of thing

489
00:25:54,280 --> 00:25:57,520
Speaker 4: that's been a stand in for intelligence, and this has

490
00:25:57,720 --> 00:26:02,080
Speaker 4: been what the advertisement for, you know, for general intelligences.

491
00:26:02,480 --> 00:26:04,520
Speaker 4: And so in addition to what Emily's saying, it's really

492
00:26:04,560 --> 00:26:07,760
Speaker 4: good at doing this next word thing. I mean, it

493
00:26:08,160 --> 00:26:14,640
Speaker 4: seems to achieve some acceptable baseline for all these different tasks.

494
00:26:15,119 --> 00:26:20,080
Speaker 4: But then these tasks become like stand ins for general intelligence,

495
00:26:20,680 --> 00:26:23,080
Speaker 4: and it's really it really gives away the game when

496
00:26:23,960 --> 00:26:26,720
Speaker 4: you know, people like Mark Zuckerberg, you know, come out

497
00:26:26,720 --> 00:26:28,880
Speaker 4: and they say, well, Matt is going to work on AGI.

498
00:26:29,040 --> 00:26:31,679
Speaker 4: Now I don't really know what that is but you know,

499
00:26:31,720 --> 00:26:32,400
Speaker 4: we're going to do.

500
00:26:32,400 --> 00:26:35,240
Speaker 1: It, but you guys seem to like it as investors,

501
00:26:35,359 --> 00:26:38,000
Speaker 1: and so therefore that is what we're going to call

502
00:26:38,119 --> 00:26:38,840
Speaker 1: everything now.

503
00:26:39,400 --> 00:26:42,320
Speaker 4: Yeah, yeah, it's it's it's literally turning back towards the

504
00:26:42,359 --> 00:26:45,919
Speaker 4: machine and like dialing up the AGI nob and seeing

505
00:26:45,920 --> 00:26:48,840
Speaker 4: if the crowd, you know, since seems to tear at it.

506
00:26:49,000 --> 00:26:52,359
Speaker 1: Yeah, so I think Emily, you were saying that just

507
00:26:53,400 --> 00:26:56,720
Speaker 1: you were specifically saying that they would take an existing sentence,

508
00:26:56,880 --> 00:26:59,800
Speaker 1: like one of the training methods, take an existing sentence,

509
00:26:59,800 --> 00:27:03,040
Speaker 1: come rubble word, computer guesses what that word is and

510
00:27:03,080 --> 00:27:05,719
Speaker 1: then compares guests to what the actual word is, and

511
00:27:05,760 --> 00:27:08,680
Speaker 1: like it got better and better at doing that until

512
00:27:08,720 --> 00:27:11,240
Speaker 1: it seems like the computer is talking to you using

513
00:27:11,240 --> 00:27:14,960
Speaker 1: the language a human would use in conversation. Like just

514
00:27:15,000 --> 00:27:17,200
Speaker 1: hearing it put that way for some reason was like, oh,

515
00:27:17,320 --> 00:27:22,560
Speaker 1: it's so it's like just better. It's doing the same

516
00:27:22,600 --> 00:27:25,920
Speaker 1: thing that a chess computer does, but it's just more

517
00:27:26,000 --> 00:27:29,240
Speaker 1: broadly impressive to people who don't know how to play chess.

518
00:27:29,600 --> 00:27:32,040
Speaker 3: So it's it's not actually doing the same thing that

519
00:27:32,040 --> 00:27:35,280
Speaker 3: a chess computer does. So what Deep Blue did was

520
00:27:35,280 --> 00:27:39,000
Speaker 3: they just threw enough computation at to calculate it out,

521
00:27:39,080 --> 00:27:41,040
Speaker 3: what are the outcomes of these moves, what opens up,

522
00:27:41,080 --> 00:27:43,399
Speaker 3: what opens up? And you put enough calculation in for

523
00:27:43,520 --> 00:27:46,560
Speaker 3: chess that's doable with computers of the era of Deep

524
00:27:46,560 --> 00:27:48,800
Speaker 3: Blue Go. It turns out the search space is much

525
00:27:48,840 --> 00:27:52,760
Speaker 3: bigger so the Alpha Go. What they did was they

526
00:27:52,800 --> 00:27:55,399
Speaker 3: actually trained it more like a language model on the

527
00:27:55,440 --> 00:27:59,640
Speaker 3: sequences of moves in games from really highly rated go players.

528
00:28:00,200 --> 00:28:03,399
Speaker 3: Someone beat it recently by playing terribly and then it

529
00:28:03,440 --> 00:28:05,560
Speaker 3: was like, I don't know what to do, yeah, because

530
00:28:05,560 --> 00:28:06,600
Speaker 3: it was completely.

531
00:28:07,720 --> 00:28:10,240
Speaker 2: Yeah, that's why I usually lose at chess.

532
00:28:10,440 --> 00:28:15,719
Speaker 1: I'm so much better than everybody that the level confuses.

533
00:28:16,800 --> 00:28:18,880
Speaker 3: So there was a big deal about how GPT four

534
00:28:18,880 --> 00:28:20,600
Speaker 3: could play chess super well. And one of the things

535
00:28:20,640 --> 00:28:23,000
Speaker 3: about most of the models that are out there right

536
00:28:23,000 --> 00:28:25,080
Speaker 3: now is we know nothing about their training data or

537
00:28:25,200 --> 00:28:27,800
Speaker 3: very little, like the companies are closed on the training data,

538
00:28:28,280 --> 00:28:32,040
Speaker 3: even open AI right but people figured out that there's

539
00:28:32,080 --> 00:28:36,120
Speaker 3: actually an enormous corpus of highly rated chess players chess

540
00:28:36,160 --> 00:28:38,920
Speaker 3: games in the training data in the specific notation for

541
00:28:39,000 --> 00:28:41,560
Speaker 3: GPT four, So when it's playing chess, it is doing

542
00:28:41,920 --> 00:28:44,480
Speaker 3: predict a likely next token based on that part of

543
00:28:44,520 --> 00:28:47,000
Speaker 3: its training data. But that's not what Deep Blue is

544
00:28:47,080 --> 00:28:49,160
Speaker 3: up to. Deep Blue is a different strategy. And in

545
00:28:49,240 --> 00:28:51,360
Speaker 3: all of these cases, there's this idea that, like, well,

546
00:28:51,440 --> 00:28:53,360
Speaker 3: chess is something that smart people do, so if a

547
00:28:53,360 --> 00:28:56,320
Speaker 3: computer can play chess, that means we've built something really smart,

548
00:28:56,720 --> 00:28:58,680
Speaker 3: which is ridiculous. And I just want to be very

549
00:28:58,680 --> 00:29:00,560
Speaker 3: clear that Alex and I aren't saying there's some better

550
00:29:00,560 --> 00:29:05,040
Speaker 3: way to build AGI, right, We're saying there's better ways

551
00:29:05,120 --> 00:29:09,480
Speaker 3: to seek out effective automation in our lives.

552
00:29:10,000 --> 00:29:10,440
Speaker 1: I'm kind of.

553
00:29:10,400 --> 00:29:11,720
Speaker 3: Speaking for you there, Alex, but I think we're I

554
00:29:11,720 --> 00:29:13,000
Speaker 3: think we're in a great yeah.

555
00:29:14,000 --> 00:29:16,960
Speaker 4: Or rather, I mean, even if we want automation and

556
00:29:17,040 --> 00:29:19,400
Speaker 4: particular parts of our lives, you know, what is what

557
00:29:19,480 --> 00:29:22,000
Speaker 4: is the function of making chess robots?

558
00:29:22,080 --> 00:29:22,240
Speaker 1: Right?

559
00:29:22,920 --> 00:29:26,160
Speaker 4: And you know, insofar is building chess robots in the

560
00:29:26,240 --> 00:29:28,880
Speaker 4: Cold War? I mean a lot of it had to

561
00:29:28,880 --> 00:29:31,640
Speaker 4: do with you know, Cold War kind of scientific documen,

562
00:29:31,760 --> 00:29:35,000
Speaker 4: you know, the US versus the USSR and seeing if

563
00:29:35,040 --> 00:29:37,600
Speaker 4: we could do better, you know, and you know, chess

564
00:29:37,760 --> 00:29:40,560
Speaker 4: was one front getting to the moon. Speaking of the

565
00:29:40,600 --> 00:29:44,080
Speaker 4: show which has already completely evaded me that we were

566
00:29:44,080 --> 00:29:47,400
Speaker 4: talking about earlier for all mankind, for all mankind, thank you,

567
00:29:47,440 --> 00:29:49,560
Speaker 4: please cut me out, make me sound smart.

568
00:29:50,240 --> 00:29:52,280
Speaker 1: It's a tough title. It doesn't stick in the brain.

569
00:29:52,440 --> 00:29:55,080
Speaker 1: And they should have used AI to name that show.

570
00:29:55,160 --> 00:29:57,480
Speaker 1: Yeah exactly, Yeah.

571
00:29:57,200 --> 00:30:00,560
Speaker 3: Yeah, historical sexism that makes it stick, maybe makes it

572
00:30:00,600 --> 00:30:01,040
Speaker 3: a little.

573
00:30:00,880 --> 00:30:04,160
Speaker 4: Bit slipper, right, yes, yeah, so it's the same thing,

574
00:30:04,240 --> 00:30:07,000
Speaker 4: you know, get into the moon first. So I mean

575
00:30:07,080 --> 00:30:09,760
Speaker 4: this becomes a stand in, you know, and it's you

576
00:30:09,800 --> 00:30:12,040
Speaker 4: know a lot of it is frankly kind of showingism.

577
00:30:12,120 --> 00:30:14,080
Speaker 4: You know, mine can do it better than yours, and

578
00:30:14,920 --> 00:30:17,280
Speaker 4: you know, and and really getting to that and so

579
00:30:18,280 --> 00:30:20,960
Speaker 4: you know, this association with chess, as m least said,

580
00:30:21,000 --> 00:30:23,320
Speaker 4: of like this thing smart people can seem to do,

581
00:30:24,080 --> 00:30:28,719
Speaker 4: not things like this automation useful in these people's lives, right,

582
00:30:28,880 --> 00:30:31,560
Speaker 4: will it make these kind of runt tasks kind of

583
00:30:31,560 --> 00:30:34,320
Speaker 4: easier for people? And said we sort of jumped over

584
00:30:34,600 --> 00:30:38,240
Speaker 4: that completely to go, let's make this art kind of

585
00:30:38,280 --> 00:30:40,040
Speaker 4: thing and oh, it's going to put out a bunch

586
00:30:40,080 --> 00:30:43,040
Speaker 4: of you know, concept artists and writers. Even if it

587
00:30:43,040 --> 00:30:45,680
Speaker 4: doesn't work, well, you know, we're going to scare the

588
00:30:45,720 --> 00:30:48,640
Speaker 4: wits out of these people who do this for a living.

589
00:30:49,320 --> 00:30:52,240
Speaker 2: What right, like is there I feel like you know,

590
00:30:52,360 --> 00:30:55,160
Speaker 2: looking at Cees and just what happened at Cees. So

591
00:30:55,240 --> 00:30:58,920
Speaker 2: many companies like it's got AI like this thing this

592
00:30:59,000 --> 00:31:02,560
Speaker 2: refrigerators you AI. And I feel like for most people

593
00:31:02,560 --> 00:31:05,280
Speaker 2: who are just consumers are just kind of passively getting

594
00:31:05,280 --> 00:31:09,800
Speaker 2: this information. There's like a discrepancy between how it's being

595
00:31:09,840 --> 00:31:14,120
Speaker 2: marketed and what it is actually doing. Right yeah, and

596
00:31:14,720 --> 00:31:16,480
Speaker 2: can you And so part of me is like how

597
00:31:16,600 --> 00:31:20,440
Speaker 2: the like why are all these C suite maniacs suddenly

598
00:31:20,480 --> 00:31:22,840
Speaker 2: being like, dude, this is the future, Like you got

599
00:31:22,840 --> 00:31:24,840
Speaker 2: to get in on it. And I know that a

600
00:31:24,840 --> 00:31:26,520
Speaker 2: lot of it is hype, but can you sort of

601
00:31:26,560 --> 00:31:29,520
Speaker 2: like help me understand like the hypeline or pipeline to

602
00:31:29,840 --> 00:31:33,440
Speaker 2: how a company says ooh, this is what we do,

603
00:31:33,600 --> 00:31:36,520
Speaker 2: and then how that creeps up to the capitalists where

604
00:31:36,520 --> 00:31:39,200
Speaker 2: then they're like yeah, yeah, yeah, what what what and

605
00:31:39,240 --> 00:31:41,920
Speaker 2: then begin to say like make these proclamations about what

606
00:31:42,240 --> 00:31:44,160
Speaker 2: large language models are doing or how it's going to

607
00:31:44,240 --> 00:31:47,280
Speaker 2: revolutionize things. What sort of like because from your perspective,

608
00:31:47,640 --> 00:31:49,200
Speaker 2: can you just sort of be like just give us

609
00:31:49,240 --> 00:31:52,320
Speaker 2: the the unfiltered like this is how this like this

610
00:31:52,400 --> 00:31:56,120
Speaker 2: conversation is moving from these laboratories into places like Goldman Sachs.

611
00:31:56,480 --> 00:31:58,480
Speaker 4: Yeah, I mean, I think a lot of a lot

612
00:31:58,480 --> 00:32:04,440
Speaker 4: of hype seems to op rate via fomo. It's honestly, honestly,

613
00:32:04,440 --> 00:32:06,560
Speaker 4: if you wants it kind of reductive kind of way,

614
00:32:07,000 --> 00:32:10,480
Speaker 4: you need to get in on this at the ground level.

615
00:32:11,160 --> 00:32:13,920
Speaker 4: If you don't, you're going to be missing out on

616
00:32:14,320 --> 00:32:20,320
Speaker 4: huge opportunities, huge returns on investment, right. And one of

617
00:32:20,360 --> 00:32:22,920
Speaker 4: the ways I would say that AI hype is kind

618
00:32:22,960 --> 00:32:26,320
Speaker 4: of at least a bit qualitatively different then let's say

619
00:32:27,080 --> 00:32:30,360
Speaker 4: I don't know crypto or n F T S or yeah,

620
00:32:30,480 --> 00:32:35,840
Speaker 4: you know, you know DeFi is that you know, crypto

621
00:32:35,920 --> 00:32:38,360
Speaker 4: has kind of shown itself to be such a such

622
00:32:38,400 --> 00:32:42,600
Speaker 4: a you know, house of cards that you know it

623
00:32:42,640 --> 00:32:45,200
Speaker 4: doesn't it takes only a few things like you're Sam

624
00:32:45,280 --> 00:32:49,000
Speaker 4: Bankman free and going to jail, or you know, these

625
00:32:49,280 --> 00:32:53,240
Speaker 4: these huge kinds of scandals happening with with finance and

626
00:32:53,600 --> 00:32:57,800
Speaker 4: coin base, that you know, those things, and then it's

627
00:32:57,800 --> 00:33:01,240
Speaker 4: the wild fluctuation of bitcoin and other types of crypto

628
00:33:01,800 --> 00:33:03,880
Speaker 4: that at least you have some kind of proof of

629
00:33:03,960 --> 00:33:08,040
Speaker 4: concepts on this on this chat GPT where it seems

630
00:33:08,080 --> 00:33:12,040
Speaker 4: like this thing can do a lot, and so there's

631
00:33:12,160 --> 00:33:15,360
Speaker 4: enough kind of stability in that where folks say, yeah,

632
00:33:15,440 --> 00:33:17,320
Speaker 4: let's let's go ahead and get in on this. And

633
00:33:17,320 --> 00:33:20,120
Speaker 4: if we don't get it on the ground floor, we're

634
00:33:20,200 --> 00:33:25,040
Speaker 4: missing out on millions or billions of dollars in in ROI.

635
00:33:25,720 --> 00:33:29,160
Speaker 4: And so you know, this is you know, but the hype,

636
00:33:29,240 --> 00:33:31,560
Speaker 4: I mean, the hype is really at its base is

637
00:33:31,640 --> 00:33:34,080
Speaker 4: kind of fomo element of it, and it's and I

638
00:33:34,120 --> 00:33:38,080
Speaker 4: think it's just the the thing where, you know, it

639
00:33:39,160 --> 00:33:42,840
Speaker 4: obscures so much of what's going going on under the hood,

640
00:33:43,360 --> 00:33:45,520
Speaker 4: and folks just think it's magic. You know, you see

641
00:33:45,600 --> 00:33:48,880
Speaker 4: gen jen ai. People say we're putting gen ai in

642
00:33:48,960 --> 00:33:51,880
Speaker 4: this that that you could rub it on an engine.

643
00:33:53,040 --> 00:33:59,000
Speaker 4: When yeah, when when when AI itself I mean itself,

644
00:33:59,080 --> 00:34:02,240
Speaker 4: that kind of calling things AI, I mean it self

645
00:34:02,280 --> 00:34:05,920
Speaker 4: has this marketing quality to it. It's really you know,

646
00:34:06,360 --> 00:34:08,520
Speaker 4: but if you kind of look at the class of

647
00:34:08,560 --> 00:34:12,239
Speaker 4: technologies have been called AI since the inception of AI

648
00:34:12,400 --> 00:34:15,480
Speaker 4: and then mid nineteen fifties, it can be anything as

649
00:34:15,520 --> 00:34:20,480
Speaker 4: basic as a logistic regression, which is a very basic

650
00:34:20,600 --> 00:34:23,799
Speaker 4: statistical method that it's taught in stats. You know, one

651
00:34:23,840 --> 00:34:26,439
Speaker 4: A one or two A one too. You know, these

652
00:34:26,560 --> 00:34:30,000
Speaker 4: language models that we see open AI, howking, and so

653
00:34:30,560 --> 00:34:33,759
Speaker 4: you could say AI isn't anything I've got AI, I've

654
00:34:33,840 --> 00:34:36,319
Speaker 4: programmed by hand, I did the math by hand.

655
00:34:36,880 --> 00:34:39,839
Speaker 1: Ooh right, you know right? Yeah.

656
00:34:40,040 --> 00:34:41,480
Speaker 3: So I think what's part of what's going on here

657
00:34:41,560 --> 00:34:44,040
Speaker 3: is that because of the way chat GPT was set

658
00:34:44,080 --> 00:34:46,080
Speaker 3: up as a demo and then all these people did

659
00:34:46,239 --> 00:34:49,839
Speaker 3: free pr for open a sort of sharing their screencaps,

660
00:34:50,480 --> 00:34:52,680
Speaker 3: everybody has the sense that there's a machine that can

661
00:34:52,680 --> 00:34:55,800
Speaker 3: actually reason right now, and then you call anything else AI,

662
00:34:56,120 --> 00:34:59,399
Speaker 3: and that the lubiz factor from chat GPT is sort

663
00:34:59,400 --> 00:35:02,040
Speaker 3: of like makes the whole thing sparkle. There's a funny

664
00:35:02,040 --> 00:35:05,040
Speaker 3: story from this conference called NEUROPS, which is neural information

665
00:35:05,120 --> 00:35:08,440
Speaker 3: processing systems, not real neurons. These are the fake ones

666
00:35:08,520 --> 00:35:10,600
Speaker 3: that these things are built out of. It's like a

667
00:35:10,640 --> 00:35:13,280
Speaker 3: mathematical function that's sort of an approximation of a nineteen

668
00:35:13,320 --> 00:35:15,759
Speaker 3: forties notion of what a neuron is. And a few

669
00:35:15,840 --> 00:35:16,799
Speaker 3: years back, I want to say, like.

670
00:35:16,920 --> 00:35:20,720
Speaker 1: I sound impressed by the way.

671
00:35:19,560 --> 00:35:22,359
Speaker 3: Around twenty sixteen twenty seventeen, I think some folks at

672
00:35:22,560 --> 00:35:26,600
Speaker 3: NEUREPS did a parody. They put together this pitch deck

673
00:35:27,000 --> 00:35:29,520
Speaker 3: of a company that was going to build AGI and

674
00:35:29,560 --> 00:35:32,759
Speaker 3: it's ridiculous. Like the pitch deck is basically step three

675
00:35:33,000 --> 00:35:35,279
Speaker 3: it dot step three profit, Like there's nothing in there.

676
00:35:36,239 --> 00:35:39,680
Speaker 3: And this was just on the cusp of when AI

677
00:35:39,800 --> 00:35:41,840
Speaker 3: went from being sort of a joke like you wouldn't

678
00:35:41,880 --> 00:35:44,600
Speaker 3: call your work AI if you were serious, to where

679
00:35:44,680 --> 00:35:47,120
Speaker 3: we are now. And what these people didn't realize when

680
00:35:47,120 --> 00:35:49,160
Speaker 3: they were doing their pitch deck was that there were

681
00:35:49,160 --> 00:35:51,879
Speaker 3: some folks in the audience who had already stepped over

682
00:35:51,920 --> 00:35:54,719
Speaker 3: into that, like, you know, AI, true believer, We're going

683
00:35:54,760 --> 00:35:56,840
Speaker 3: to make a lot of money on this step. People

684
00:35:56,880 --> 00:36:00,000
Speaker 3: came up and offered them money for their fake company.

685
00:36:00,960 --> 00:36:01,560
Speaker 1: That's great.

686
00:36:02,080 --> 00:36:04,759
Speaker 2: Yeah, so it's say it's like the dot com boom. Yeah,

687
00:36:04,760 --> 00:36:06,600
Speaker 2: there's just like what's the company called some dot com?

688
00:36:07,239 --> 00:36:11,399
Speaker 1: They're ready for the next thing, right, they've like kind

689
00:36:11,400 --> 00:36:15,160
Speaker 1: of stalled out post iPhone, and they're like ready for

690
00:36:15,280 --> 00:36:19,279
Speaker 1: the next big thing, whether it's here or not. And

691
00:36:19,600 --> 00:36:23,480
Speaker 1: so yeah, the g whiz afication or like the g

692
00:36:23,600 --> 00:36:28,040
Speaker 1: whiz results from open AI and chat GPT I feel

693
00:36:28,120 --> 00:36:30,800
Speaker 1: like they were like good enough, let's get this thing going,

694
00:36:30,960 --> 00:36:34,440
Speaker 1: Let's get this machine cranking out. The idea of describing

695
00:36:35,040 --> 00:36:39,759
Speaker 1: like the mistakes that chat GPT makes as hallucinations. Like

696
00:36:39,800 --> 00:36:43,640
Speaker 1: the marketing that's being done across the board here is

697
00:36:44,320 --> 00:36:45,279
Speaker 1: pretty impressive.

698
00:36:45,600 --> 00:36:49,040
Speaker 3: The fact that CHATGPT even uses the word I right,

699
00:36:49,440 --> 00:36:52,920
Speaker 3: that's a programming choice. Yeah, could have been otherwise and

700
00:36:52,960 --> 00:36:54,800
Speaker 3: would have made things much much clearer.

701
00:36:55,040 --> 00:36:59,120
Speaker 1: Right, Yeah, chat GPT is trained on the entire Internet.

702
00:36:59,440 --> 00:36:59,960
Speaker 3: Not true.

703
00:37:00,120 --> 00:37:05,000
Speaker 1: Yeah, that was something that I accepted when it first started.

704
00:37:05,000 --> 00:37:08,560
Speaker 1: I'm like, well, this thing hallucinates and its brain is

705
00:37:08,600 --> 00:37:10,840
Speaker 1: the entire Internet in what could go wrong?

706
00:37:11,040 --> 00:37:13,080
Speaker 3: So the thing about the phrase entire internet is that

707
00:37:13,400 --> 00:37:14,840
Speaker 3: the Internet is not a thing where you can go

708
00:37:14,920 --> 00:37:18,799
Speaker 3: somewhere and download, right. It is distributed, and so if

709
00:37:18,840 --> 00:37:20,840
Speaker 3: you are going to collect information from the Internet, you

710
00:37:20,840 --> 00:37:22,960
Speaker 3: have to decide which URLs you're going to visit, and

711
00:37:23,000 --> 00:37:25,360
Speaker 3: they aren't all listed in one place. So there's no

712
00:37:25,400 --> 00:37:28,400
Speaker 3: way to say the entire Internet. There's some choices happening already.

713
00:37:28,600 --> 00:37:28,880
Speaker 2: Yeah.

714
00:37:28,920 --> 00:37:31,120
Speaker 4: And what they mean when they say the entire Internet,

715
00:37:31,239 --> 00:37:34,400
Speaker 4: I mean most of the time, they mean most of

716
00:37:34,440 --> 00:37:38,600
Speaker 4: the English available Internet, And you mean, and although they've

717
00:37:38,600 --> 00:37:40,839
Speaker 4: got some, but then much of the work that has

718
00:37:40,880 --> 00:37:47,200
Speaker 4: been translated has been actually machine translated from English into

719
00:37:47,239 --> 00:37:49,040
Speaker 4: another language. So, for instance, you know, a bunch of

720
00:37:49,080 --> 00:37:52,799
Speaker 4: the URLs that they have are from the Google Japanese

721
00:37:52,880 --> 00:37:57,080
Speaker 4: patents and common crawl and the Dodge paper. Yeah, a

722
00:37:57,080 --> 00:38:00,000
Speaker 4: lot of it's been machine translated from English into Japanese

723
00:38:00,120 --> 00:38:03,800
Speaker 4: use and it's just available through like Google's Japanese patent site.

724
00:38:04,280 --> 00:38:06,400
Speaker 4: And so a bunch of this stuff is you know,

725
00:38:06,600 --> 00:38:08,040
Speaker 4: just just loaded in and.

726
00:38:07,960 --> 00:38:09,480
Speaker 2: It's and so and so.

727
00:38:09,520 --> 00:38:12,319
Speaker 4: I mentioned this data set common crawl, which isn't you know,

728
00:38:12,960 --> 00:38:17,040
Speaker 4: in the GPT two paper I think they cite it,

729
00:38:17,160 --> 00:38:19,520
Speaker 4: or it's the TPT three paper. It's I think it's

730
00:38:19,520 --> 00:38:21,839
Speaker 4: a GPT three P pair And they said, yeah, it's

731
00:38:21,920 --> 00:38:24,160
Speaker 4: the common crawl. Is this data set that they say

732
00:38:24,200 --> 00:38:26,799
Speaker 4: they use, which is this kind of this weird like

733
00:38:26,920 --> 00:38:29,239
Speaker 4: nonprofit I was collecting a bunch of data that could

734
00:38:29,280 --> 00:38:33,080
Speaker 4: be useful for some limited scientific inquiry, but then it

735
00:38:33,160 --> 00:38:36,440
Speaker 4: became now they've completely rebranded themselves. They're saying, we have

736
00:38:36,480 --> 00:38:41,120
Speaker 4: the data that you know, fuels large language models and Mozilla,

737
00:38:41,200 --> 00:38:45,200
Speaker 4: the Mozilla Foundation folks that also the corporation makes the

738
00:38:45,239 --> 00:38:49,720
Speaker 4: web browser Firefox. They did this report on common crawl

739
00:38:49,760 --> 00:38:52,920
Speaker 4: and kind of what's behind it too, and and find

740
00:38:53,160 --> 00:38:56,399
Speaker 4: you know, the weird idiosyncrasies of it. But yeah, when

741
00:38:56,400 --> 00:38:58,879
Speaker 4: they say the whole internet, I mean yeah, it's these

742
00:38:59,000 --> 00:39:02,239
Speaker 4: currated data sets. You can't get this whole internet. That's

743
00:39:02,360 --> 00:39:03,920
Speaker 4: that's just that's just a falsity.

744
00:39:04,520 --> 00:39:06,359
Speaker 2: But for our listeners, you can get it from Jack

745
00:39:06,400 --> 00:39:07,960
Speaker 2: and I. We do have that file.

746
00:39:08,040 --> 00:39:12,600
Speaker 4: So if you're it's just it's just one large ZIP file.

747
00:39:12,960 --> 00:39:17,680
Speaker 2: Yeah, yeah, exactly. Yeah, it's on pirate Bay. Check it out.

748
00:39:17,760 --> 00:39:21,520
Speaker 1: Yeah yeah, and we read it every morning to prepare

749
00:39:21,560 --> 00:39:22,120
Speaker 1: for this show.

750
00:39:22,320 --> 00:39:24,680
Speaker 2: Just the whole thing to credit and downloaded.

751
00:39:24,800 --> 00:39:27,319
Speaker 1: Yeah. Uh, let's take a quick break and we'll be

752
00:39:27,600 --> 00:39:41,600
Speaker 1: right back. And we're back. So yeah, just in terms

753
00:39:41,680 --> 00:39:45,200
Speaker 1: of juxtaposing, like what these companies are saying and the

754
00:39:45,239 --> 00:39:50,920
Speaker 1: mainstream media is buying versus what is actually happening. So

755
00:39:51,200 --> 00:39:54,680
Speaker 1: I was reminded of Sam Altman telling The New Yorker

756
00:39:55,320 --> 00:39:59,440
Speaker 1: that he keeps like a bag with cyanide capsules ready

757
00:39:59,440 --> 00:40:03,680
Speaker 1: to go in case of the AI apocalypse. So you know,

758
00:40:03,880 --> 00:40:08,120
Speaker 1: he is an expert who gets billions of dollars richer

759
00:40:08,160 --> 00:40:12,600
Speaker 1: if people think his technology is so powerful that he's

760
00:40:12,640 --> 00:40:15,840
Speaker 1: like freaked out by it. So, but that that's something

761
00:40:16,440 --> 00:40:20,440
Speaker 1: that I don't know I've seen reprinted in like long

762
00:40:20,560 --> 00:40:23,960
Speaker 1: articles about the danger of AI. And then there's this

763
00:40:24,120 --> 00:40:29,920
Speaker 1: other like more real world trend where like you you

764
00:40:29,960 --> 00:40:33,680
Speaker 1: talk about a Google engineer who admitted they're not going

765
00:40:33,719 --> 00:40:37,680
Speaker 1: to use any large language model as part of their

766
00:40:37,719 --> 00:40:39,360
Speaker 1: families healthcare journey.

767
00:40:39,760 --> 00:40:42,080
Speaker 4: Oh that was a Google That was a Google senior

768
00:40:42,200 --> 00:40:50,080
Speaker 4: vice president, not a senior of Google VP. Greg Corrado

769
00:40:50,719 --> 00:40:53,480
Speaker 4: is one of the heads of Google Health itself.

770
00:40:55,040 --> 00:40:58,640
Speaker 1: There is yeah, and there's also this story from your

771
00:40:58,640 --> 00:41:01,480
Speaker 1: show has a fresh Health segment at the end where

772
00:41:01,520 --> 00:41:06,160
Speaker 1: you talk about just saying examples headlines that.

773
00:41:06,080 --> 00:41:07,719
Speaker 2: Are just fresh.

774
00:41:08,160 --> 00:41:09,200
Speaker 1: Yeah, it's just fresh out.

775
00:41:09,200 --> 00:41:10,360
Speaker 4: It's new versions of.

776
00:41:11,840 --> 00:41:12,360
Speaker 2: Yeah.

777
00:41:12,440 --> 00:41:15,759
Speaker 1: The one about Duo Lingo from a recent episode where

778
00:41:15,760 --> 00:41:21,320
Speaker 1: they're getting rid of human translators, firing them, cutting the workforce,

779
00:41:21,640 --> 00:41:26,160
Speaker 1: replacing them with translation AI, even though the technology isn't

780
00:41:26,160 --> 00:41:30,440
Speaker 1: there yet. But the point that you were making on

781
00:41:30,480 --> 00:41:33,160
Speaker 1: the show is that they're willing to go forward with

782
00:41:33,200 --> 00:41:37,840
Speaker 1: that because the user base won't notice the difference until

783
00:41:37,880 --> 00:41:41,080
Speaker 1: they're in Spain and need to get to a hospital

784
00:41:41,280 --> 00:41:46,000
Speaker 1: and asking for the biblioteca. You know, like it's they

785
00:41:46,040 --> 00:41:49,200
Speaker 1: are specifically an audience who is not going to know

786
00:41:49,920 --> 00:41:54,520
Speaker 1: how bad the product that they're getting is because they're

787
00:41:54,600 --> 00:41:56,960
Speaker 1: just like not in a in a position to know

788
00:41:57,000 --> 00:42:00,319
Speaker 1: that by the nature of the product. And so just

789
00:42:00,400 --> 00:42:06,000
Speaker 1: this distinction between being hyped to the mainstream media and

790
00:42:06,120 --> 00:42:09,320
Speaker 1: like these long read like New York or Atlantic articles

791
00:42:09,680 --> 00:42:12,680
Speaker 1: as this is a future that we should be scared

792
00:42:12,719 --> 00:42:15,880
Speaker 1: of because like it's going to become self aware and

793
00:42:15,960 --> 00:42:19,640
Speaker 1: Sam Altman is freaked out, and then what it's actually doing,

794
00:42:20,120 --> 00:42:25,120
Speaker 1: which is just making everything shittier around us, is I

795
00:42:25,160 --> 00:42:29,160
Speaker 1: think a big kind of chunk that I took away

796
00:42:29,160 --> 00:42:31,480
Speaker 1: from your show that was just like, oh, yeah, that

797
00:42:31,520 --> 00:42:34,560
Speaker 1: makes way more sense. That feels much more likely to

798
00:42:34,600 --> 00:42:37,400
Speaker 1: be how this thing progresses.

799
00:42:38,200 --> 00:42:40,959
Speaker 3: Yeah, So the AI doumerism, which is when Altman says

800
00:42:41,160 --> 00:42:43,160
Speaker 3: I've got my bug out bag and my sinanite capsules

801
00:42:43,160 --> 00:42:46,200
Speaker 3: in case the robot apocalypse comes, or when Jeff Hinton,

802
00:42:46,840 --> 00:42:51,200
Speaker 3: who's credited has a Turing Award for his work on

803
00:42:51,680 --> 00:42:55,160
Speaker 3: the specific kind of computer program that's us see Gablo

804
00:42:55,200 --> 00:42:58,840
Speaker 3: statistics that's behind these large language models. He's now concerned

805
00:42:58,880 --> 00:43:00,880
Speaker 3: that it's on track to because I mean, smarter than us,

806
00:43:00,880 --> 00:43:03,799
Speaker 3: and it's gonna like these piles of linear algebra are

807
00:43:03,800 --> 00:43:07,680
Speaker 3: not going to combust into consciousness. And anytime someone is like,

808
00:43:07,920 --> 00:43:10,920
Speaker 3: you know, pointing to that boogeyman, what they're doing is

809
00:43:10,960 --> 00:43:13,560
Speaker 3: they're basically hiding the actions of the people in the

810
00:43:13,560 --> 00:43:16,520
Speaker 3: picture who are using it to do things like make

811
00:43:16,560 --> 00:43:20,279
Speaker 3: a shitty language learning product because it's cheaper to do

812
00:43:20,320 --> 00:43:21,759
Speaker 3: it that way than to pay the people with the

813
00:43:21,760 --> 00:43:23,520
Speaker 3: actual expertise to do the translations.

814
00:43:23,800 --> 00:43:27,000
Speaker 4: Right, yeah, yeah, And it's just leading just to I mean,

815
00:43:27,239 --> 00:43:29,120
Speaker 4: I like this kind of thought on this, I mean

816
00:43:29,160 --> 00:43:32,120
Speaker 4: this kind of process. And Corey Doctor has got this

817
00:43:32,600 --> 00:43:36,000
Speaker 4: kind of sister concept of AI with hype, which is

818
00:43:36,480 --> 00:43:40,960
Speaker 4: in shitification, which I think that the Luistic Society America

819
00:43:41,000 --> 00:43:42,919
Speaker 4: it was their their overall.

820
00:43:43,160 --> 00:43:46,680
Speaker 3: Yeah, American dialect Society Dialects, Yeah, picked it as the

821
00:43:46,719 --> 00:43:48,920
Speaker 3: overall word of the year for twenty twenty three. But

822
00:43:48,960 --> 00:43:51,680
Speaker 3: in shtification is something very specific, right, It's not just

823
00:43:51,760 --> 00:43:55,839
Speaker 3: like we've now we're now swimming in AI extruded junk, right,

824
00:43:56,000 --> 00:43:59,960
Speaker 3: AI and quotes as always, It's something more the company

825
00:44:00,239 --> 00:44:03,000
Speaker 3: create a platform that you sort of get lured in

826
00:44:03,040 --> 00:44:05,880
Speaker 3: because initially it's really useful for So this is like

827
00:44:05,920 --> 00:44:08,920
Speaker 3: you think about how Amazon was great for finding products

828
00:44:08,960 --> 00:44:11,080
Speaker 3: sucked for your local brick and mortar businesses. But as

829
00:44:11,080 --> 00:44:13,400
Speaker 3: a consumer it was super convenient because you could find things.

830
00:44:13,840 --> 00:44:17,600
Speaker 3: But then the companies basically turn around and they extract

831
00:44:17,719 --> 00:44:19,960
Speaker 3: all of the value that the customer is getting out

832
00:44:20,000 --> 00:44:24,040
Speaker 3: of it, and then they turn around to the other parties.

833
00:44:24,040 --> 00:44:25,560
Speaker 3: There are the people trying to sell things, and they

834
00:44:25,600 --> 00:44:27,600
Speaker 3: extract value out of them. So you start off with

835
00:44:27,640 --> 00:44:30,440
Speaker 3: this thing that's initially quite useful and usable, and then

836
00:44:30,440 --> 00:44:33,279
Speaker 3: it getsified in the name of making profits for the platform.

837
00:44:33,520 --> 00:44:35,759
Speaker 3: And that's like the specific thing about in certification.

838
00:44:36,640 --> 00:44:39,920
Speaker 4: And I would say there's some kinds of processes you

839
00:44:39,960 --> 00:44:42,239
Speaker 4: can think about indification, the kind of idea that you

840
00:44:42,280 --> 00:44:44,880
Speaker 4: have to rush to a certain kind of market, that

841
00:44:44,960 --> 00:44:47,920
Speaker 4: you have a monopoly on this kind of thing, And

842
00:44:48,000 --> 00:44:50,280
Speaker 4: I guess yeah. I mean, the thing about large language

843
00:44:50,280 --> 00:44:52,520
Speaker 4: models I think that we get we get on and

844
00:44:52,560 --> 00:44:55,600
Speaker 4: talk a lot about, is that large language models are

845
00:44:55,640 --> 00:44:59,759
Speaker 4: kind of born shitty with content. So it's not like

846
00:44:59,760 --> 00:45:03,160
Speaker 4: the platform started and that platform monopoly led to this

847
00:45:03,280 --> 00:45:07,200
Speaker 4: kind of process of incertification. It's more like you decided

848
00:45:07,239 --> 00:45:10,640
Speaker 4: to make a tool that is a really good word

849
00:45:10,680 --> 00:45:14,120
Speaker 4: prediction machine, and you use it for a substitute for

850
00:45:14,239 --> 00:45:16,960
Speaker 4: places in which people are meant to be speaking kind

851
00:45:17,000 --> 00:45:19,040
Speaker 4: of with their own tone, with their own voice, with

852
00:45:19,160 --> 00:45:22,160
Speaker 4: their own forms and their own putting, so they're on

853
00:45:22,280 --> 00:45:25,239
Speaker 4: expertise and then and so it's kind of yeah, it's

854
00:45:25,280 --> 00:45:28,960
Speaker 4: it's bored and shitty, and so you know this, this

855
00:45:29,040 --> 00:45:30,719
Speaker 4: kind of thing I think is really helpful. It makes

856
00:45:30,760 --> 00:45:33,680
Speaker 4: me think of kind of a thing I think. I

857
00:45:33,719 --> 00:45:36,040
Speaker 4: saw a few times on Twitter where people are like, well,

858
00:45:36,040 --> 00:45:38,160
Speaker 4: if you're an expert in any of these fields and

859
00:45:38,200 --> 00:45:41,279
Speaker 4: you read content by large language models, you actually know

860
00:45:41,480 --> 00:45:45,680
Speaker 4: anything about this, you're going to know that it's absolute bullshit, right,

861
00:45:46,280 --> 00:45:49,760
Speaker 4: you know, if you ask it to write you a short,

862
00:45:49,880 --> 00:45:55,120
Speaker 4: you know, treatise on I don't know, sociological methodology, something

863
00:45:55,400 --> 00:45:58,160
Speaker 4: specific that I know a little bit about, it's going

864
00:45:58,239 --> 00:46:01,960
Speaker 4: to be absolute bullshit, right, But good enough to computer

865
00:46:02,040 --> 00:46:05,960
Speaker 4: engineers and you know, higher ups at these companies.

866
00:46:06,400 --> 00:46:06,800
Speaker 1: Yeah.

867
00:46:06,880 --> 00:46:09,320
Speaker 3: So here's an example. The other day, I came across

868
00:46:09,360 --> 00:46:11,960
Speaker 3: an article that supposedly quoted me out of this publication

869
00:46:12,000 --> 00:46:14,640
Speaker 3: from India called Bihar Praba, and I'm like, I never

870
00:46:14,840 --> 00:46:16,919
Speaker 3: said those words. I could see how someone might think

871
00:46:16,960 --> 00:46:19,160
Speaker 3: I would. And then I searched my emails, like no,

872
00:46:19,239 --> 00:46:21,719
Speaker 3: I never corresponded with any journalist at this outfit. So

873
00:46:21,760 --> 00:46:24,600
Speaker 3: I wrote them, I said, fabricated quote. Please take it

874
00:46:24,640 --> 00:46:28,080
Speaker 3: down in printer retraction, which they did, and they wrote

875
00:46:28,080 --> 00:46:30,160
Speaker 3: back and said, oh yeah, that actually we prompted some

876
00:46:30,239 --> 00:46:31,280
Speaker 3: large language models.

877
00:46:31,080 --> 00:46:33,680
Speaker 1: To create that before us and posted yes, because it

878
00:46:33,760 --> 00:46:37,640
Speaker 1: seems like you might have and the large language models

879
00:46:38,040 --> 00:46:41,120
Speaker 1: they don't like that. That's isn't that? What hallucinations are

880
00:46:41,160 --> 00:46:43,440
Speaker 1: a lot of the time is just the large language

881
00:46:43,440 --> 00:46:47,480
Speaker 1: model making up stuff it seems like is what the

882
00:46:47,520 --> 00:46:49,960
Speaker 1: person wants them to say exactly.

883
00:46:50,040 --> 00:46:52,759
Speaker 3: But here's the whole thing. Every single thing output from

884
00:46:52,760 --> 00:46:55,240
Speaker 3: a large language model has that property. It's just sometimes

885
00:46:55,280 --> 00:46:55,840
Speaker 3: it seems.

886
00:46:55,640 --> 00:46:57,759
Speaker 1: To make sense. The whole thing is trying to do

887
00:46:57,840 --> 00:47:00,239
Speaker 1: a trick of like predict I figured out what you

888
00:47:00,239 --> 00:47:02,800
Speaker 1: wanted me to say, ha ha, But it's like, well,

889
00:47:02,920 --> 00:47:06,360
Speaker 1: but what I wanted you to say is not always

890
00:47:06,600 --> 00:47:10,040
Speaker 1: that's not how I want my questions answered that. That's

891
00:47:10,080 --> 00:47:16,000
Speaker 1: actually a wildly flawed way of coming up like answering people.

892
00:47:16,200 --> 00:47:19,040
Speaker 1: It's definitely something that I do in my day to

893
00:47:19,120 --> 00:47:22,520
Speaker 1: day life because I'm scared of conflict and a people pleaser.

894
00:47:22,680 --> 00:47:26,279
Speaker 1: But that's not I'm not a good scientific instrument for

895
00:47:26,320 --> 00:47:28,400
Speaker 1: that reason, you know, Like, but you.

896
00:47:28,960 --> 00:47:32,839
Speaker 4: Just got you just an avoidant attachment style, which as

897
00:47:32,880 --> 00:47:34,440
Speaker 4: a as another avoidance.

898
00:47:36,239 --> 00:47:40,640
Speaker 1: They've just made me a scientific model that's terrible.

899
00:47:41,080 --> 00:47:42,440
Speaker 4: Like you, I can't believe this.

900
00:47:44,719 --> 00:47:46,319
Speaker 2: I was like, I don't know if you saw that

901
00:47:46,760 --> 00:47:49,320
Speaker 2: headline where Tyler Perry was like, I was going to

902
00:47:49,520 --> 00:47:53,120
Speaker 2: open an eight hundred million dollar studio, but I stopped

903
00:47:53,160 --> 00:47:57,120
Speaker 2: the second I saw what Sora, this video generative AI

904
00:47:57,239 --> 00:48:00,480
Speaker 2: could do, and I realized we're in trouble. Said quote,

905
00:48:00,560 --> 00:48:03,960
Speaker 2: I had no idea until I saw recently the demonstrations

906
00:48:04,000 --> 00:48:06,640
Speaker 2: of what it's able to do. It's shocking to me.

907
00:48:07,160 --> 00:48:09,920
Speaker 2: And he's basically saying, He's like, you could make up

908
00:48:09,960 --> 00:48:12,880
Speaker 2: a pilot and save millions of dollars. This is going

909
00:48:12,920 --> 00:48:15,640
Speaker 2: to have all kinds of ramifications. That feels like quite,

910
00:48:15,719 --> 00:48:19,120
Speaker 2: that feels like half like just ignorance because this person's like,

911
00:48:19,160 --> 00:48:22,000
Speaker 2: oh my god, look like total wow factor but also

912
00:48:22,080 --> 00:48:25,640
Speaker 2: maybe hype. But I'm also curious from your perspective, what

913
00:48:25,640 --> 00:48:29,279
Speaker 2: what are the actual dangers that we're facing that you know,

914
00:48:29,320 --> 00:48:31,640
Speaker 2: because right now I think everything is just all about

915
00:48:31,960 --> 00:48:33,560
Speaker 2: these are the jobs it's going to take. I think

916
00:48:33,600 --> 00:48:36,840
Speaker 2: in the l ll M episode where the LM predicted

917
00:48:37,239 --> 00:48:40,239
Speaker 2: what that what the potential of llms were and the

918
00:48:40,360 --> 00:48:45,000
Speaker 2: jobs that it could take in a very paper was Yeah,

919
00:48:45,320 --> 00:48:47,480
Speaker 2: where it's like, huh, you know, like just sort of

920
00:48:47,480 --> 00:48:50,920
Speaker 2: the unethical nature of how even these companies are doing

921
00:48:51,040 --> 00:48:53,600
Speaker 2: research and creating data to support this.

922
00:48:53,920 --> 00:48:56,360
Speaker 1: Can we just stop? Can you just stop and explain

923
00:48:56,440 --> 00:48:58,719
Speaker 1: exactly what the methodology of that paper was? No?

924
00:48:58,840 --> 00:49:00,719
Speaker 2: Please, I will. I will allow the experts to do it,

925
00:49:00,800 --> 00:49:04,799
Speaker 2: because it's it's it's absolutely bonkers to hear because any

926
00:49:04,840 --> 00:49:07,600
Speaker 2: person who's like been at any like tried to look

927
00:49:07,640 --> 00:49:10,000
Speaker 2: at a study or something and you look at methodology like.

928
00:49:11,560 --> 00:49:13,880
Speaker 3: So methodology is a very kind term.

929
00:49:13,680 --> 00:49:18,920
Speaker 2: For what was. Yeah, truly truly, So, Alex.

930
00:49:18,760 --> 00:49:20,160
Speaker 3: Do you want to summarize real quick what was?

931
00:49:20,200 --> 00:49:20,279
Speaker 1: Then?

932
00:49:20,280 --> 00:49:21,719
Speaker 3: There was two different things we were looking at. It

933
00:49:21,920 --> 00:49:23,560
Speaker 3: was something that came out of Open AI and something

934
00:49:23,560 --> 00:49:27,760
Speaker 3: from Goldman Sachs and the Golden Sacks one was silly,

935
00:49:27,880 --> 00:49:29,759
Speaker 3: but not as silly as the Open Eye one.

936
00:49:30,280 --> 00:49:33,160
Speaker 4: Yeah, I mean getting in the detail. And I went

937
00:49:33,239 --> 00:49:35,840
Speaker 4: through this and I puzzle this paper. I like poked

938
00:49:35,840 --> 00:49:38,560
Speaker 4: my friend as a labor sociologist, I'm like, what the

939
00:49:38,600 --> 00:49:42,080
Speaker 4: hell is going on here? And you know, okay, so

940
00:49:42,160 --> 00:49:44,160
Speaker 4: there's this kind of metric that you can use to

941
00:49:44,239 --> 00:49:47,320
Speaker 4: judge how hard a task is that the government collects,

942
00:49:48,120 --> 00:49:50,359
Speaker 4: and there's this kind of job classification. They rid him

943
00:49:50,400 --> 00:49:54,120
Speaker 4: from you know, one to seven effectively. And so what

944
00:49:54,160 --> 00:49:57,360
Speaker 4: Goldman Sachs said was, well, basically, anything from one to

945
00:49:57,440 --> 00:50:00,000
Speaker 4: four probably a machine can do. And you're like, okay,

946
00:50:00,080 --> 00:50:04,319
Speaker 4: that's kind of silly. That's huge assumptions there. I understand though,

947
00:50:04,360 --> 00:50:06,520
Speaker 4: as a researcher you have to make some assumptions when

948
00:50:06,560 --> 00:50:09,600
Speaker 4: you don't have great data. But what Opening Eye did

949
00:50:09,719 --> 00:50:16,000
Speaker 4: is that they asked two entities what like how well

950
00:50:16,400 --> 00:50:19,680
Speaker 4: like what could be automated? They asked one other open

951
00:50:19,800 --> 00:50:22,719
Speaker 4: AI employees, hey, what jobs do you think could be automated?

952
00:50:23,040 --> 00:50:26,840
Speaker 4: Already hilarious because you know they're they're you know, pretty

953
00:50:27,360 --> 00:50:29,719
Speaker 4: doing those jobs. Yeah, they're not doing those doubts. They're

954
00:50:29,719 --> 00:50:32,680
Speaker 4: pretty primed to think that their technology is great. And

955
00:50:32,719 --> 00:50:36,480
Speaker 4: then they asked DPT for itself. They prompted and say, hey,

956
00:50:36,520 --> 00:50:40,319
Speaker 4: can we automate these jobs? So you know, and so.

957
00:50:40,360 --> 00:50:42,240
Speaker 1: You'll never guess what the answer was.

958
00:50:42,400 --> 00:50:45,200
Speaker 4: You'll never guess. You'll never guess.

959
00:50:45,360 --> 00:50:48,320
Speaker 3: They took the output of that as if it were data,

960
00:50:48,400 --> 00:50:50,680
Speaker 3: and then like you know, these ridiculous graphs and blah

961
00:50:50,680 --> 00:50:52,959
Speaker 3: blah blah. It's just like the whole thing is is

962
00:50:52,960 --> 00:50:54,120
Speaker 3: is fantasy.

963
00:50:54,080 --> 00:50:57,680
Speaker 2: Right, So is the danger there just sort of this

964
00:50:57,800 --> 00:51:01,080
Speaker 2: reliance Like I guess, so if we're classifying the sort

965
00:51:01,120 --> 00:51:04,360
Speaker 2: of threats to our sense of like how information is

966
00:51:04,400 --> 00:51:07,560
Speaker 2: distributed or what is real or what is hype or whatever,

967
00:51:07,960 --> 00:51:10,440
Speaker 2: or if they're they're actually taking jobs, what, what is

968
00:51:10,480 --> 00:51:12,680
Speaker 2: something that I think people that people actually need to

969
00:51:12,680 --> 00:51:15,759
Speaker 2: be aware of or to sort of prepare themselves for

970
00:51:15,880 --> 00:51:18,040
Speaker 2: how this is going to disrupt things in a way

971
00:51:18,040 --> 00:51:19,880
Speaker 2: that you know, isn't necessarily the end of the world,

972
00:51:19,920 --> 00:51:22,360
Speaker 2: but definitely changing things for the for the worst.

973
00:51:22,640 --> 00:51:24,560
Speaker 3: There's a whole bunch of things, and the one that

974
00:51:24,600 --> 00:51:26,920
Speaker 3: I'm sort of most going on about these days is

975
00:51:26,920 --> 00:51:29,320
Speaker 3: the threats to the information ecosystem, So I want to

976
00:51:29,360 --> 00:51:32,640
Speaker 3: talk about that. But there's also things like automated decision

977
00:51:32,719 --> 00:51:35,399
Speaker 3: systems being used by governments to decide who gets social

978
00:51:35,480 --> 00:51:39,280
Speaker 3: benefits and who gets them yanked right, things like doctor

979
00:51:39,320 --> 00:51:41,480
Speaker 3: Joybul and WHINNI worked with a group of people in

980
00:51:41,800 --> 00:51:43,279
Speaker 3: I want to say in New York City who were

981
00:51:43,320 --> 00:51:48,359
Speaker 3: pushing back against having a face recognition system installed as

982
00:51:48,400 --> 00:51:50,920
Speaker 3: their like entry system, So they were going to be

983
00:51:51,000 --> 00:51:55,320
Speaker 3: continuously surveilled because the company who owned the house that

984
00:51:55,360 --> 00:51:56,839
Speaker 3: they lived in, or maybe it was I'm not sure

985
00:51:56,840 --> 00:51:57,239
Speaker 3: if it was going.

986
00:51:57,200 --> 00:51:59,680
Speaker 4: To be apartment apartment complex, yes, yeah.

987
00:51:59,760 --> 00:52:04,120
Speaker 3: Yeah, wanted to basically use biometrics as a way to

988
00:52:04,719 --> 00:52:07,680
Speaker 3: have them gain entry to their own homes where they lived.

989
00:52:07,880 --> 00:52:11,040
Speaker 3: So there's dangerous of lots of lots of different kinds.

990
00:52:11,719 --> 00:52:13,399
Speaker 3: The one that's maybe closest to what we've been talking

991
00:52:13,440 --> 00:52:16,200
Speaker 3: about though, is these dangerous to the information ecosystem where

992
00:52:16,280 --> 00:52:19,319
Speaker 3: you now have the synthetic media, like that news article

993
00:52:19,320 --> 00:52:21,239
Speaker 3: I was talking about before being posted as if it

994
00:52:21,280 --> 00:52:24,680
Speaker 3: were real, without any watermarks, without anything that either a

995
00:52:24,680 --> 00:52:26,319
Speaker 3: person can look at and say, I know that's fake,

996
00:52:26,600 --> 00:52:28,279
Speaker 3: Like I knew because it was my name, and I

997
00:52:28,320 --> 00:52:30,960
Speaker 3: knew I didn't say that thing, right, but somebody else

998
00:52:31,000 --> 00:52:34,320
Speaker 3: wouldn't have. And there's also not a machine readable watermarking

999
00:52:34,360 --> 00:52:36,000
Speaker 3: in there, so you can just filter it out and

1000
00:52:36,040 --> 00:52:38,440
Speaker 3: this stuff goes and goes. So there was Oh a

1001
00:52:38,480 --> 00:52:43,400
Speaker 3: few months ago, someone posted a fake mushroom foraging guide

1002
00:52:43,760 --> 00:52:45,680
Speaker 3: that they had created using a large language model to

1003
00:52:45,800 --> 00:52:48,520
Speaker 3: Amazon as a self published book so that it's just

1004
00:52:48,600 --> 00:52:51,920
Speaker 3: up there as a book. And coming back around to Sora,

1005
00:52:52,560 --> 00:52:55,799
Speaker 3: those videos like they look impressive at first glance, but

1006
00:52:55,840 --> 00:52:58,160
Speaker 3: then just like Alex was saying about the art having

1007
00:52:58,160 --> 00:53:01,160
Speaker 3: this sort of facile style to it, there's similar things

1008
00:53:01,160 --> 00:53:04,480
Speaker 3: going on in the videos, but still like it should

1009
00:53:04,480 --> 00:53:06,759
Speaker 3: be watermarked, it should be obvious that you're looking at

1010
00:53:06,800 --> 00:53:09,040
Speaker 3: something fake. And what open ai has done is they've

1011
00:53:09,040 --> 00:53:11,279
Speaker 3: put this tiny, little faint thing down the lower right

1012
00:53:11,280 --> 00:53:14,399
Speaker 3: hand corner that looks like some sort of readout from

1013
00:53:14,680 --> 00:53:17,000
Speaker 3: a sensor going up and down, and then it swirls

1014
00:53:17,000 --> 00:53:20,759
Speaker 3: into the open ai logo and it's faint, and it's

1015
00:53:20,760 --> 00:53:23,080
Speaker 3: in the same spot that Twitter puts the button for

1016
00:53:23,120 --> 00:53:25,160
Speaker 3: turning the sound on and off. So it's hidden by that.

1017
00:53:25,200 --> 00:53:28,719
Speaker 3: If you're seeing these things on Twitter, and it's completely opaque, right,

1018
00:53:28,719 --> 00:53:30,359
Speaker 3: if you are not in the know, if you don't

1019
00:53:30,360 --> 00:53:31,920
Speaker 3: know what open ai is, if you don't know that

1020
00:53:32,000 --> 00:53:34,919
Speaker 3: fake videos might exist, that doesn't tell you anything, right.

1021
00:53:35,200 --> 00:53:37,000
Speaker 3: So these are the things I'm worried about.

1022
00:53:37,239 --> 00:53:39,719
Speaker 4: And the things I'm really worried about is, you know,

1023
00:53:39,800 --> 00:53:44,960
Speaker 4: these things doing a pretty terrible job at producing written

1024
00:53:45,000 --> 00:53:51,160
Speaker 4: content and videos and images. And so it's not that

1025
00:53:51,280 --> 00:53:55,040
Speaker 4: they could replace a human person, but it just takes

1026
00:53:55,360 --> 00:53:59,399
Speaker 4: a boss or a VP to think that it's good enough. Right,

1027
00:53:59,440 --> 00:54:04,920
Speaker 4: and then this replaces whole classes of occupation. So again

1028
00:54:05,600 --> 00:54:08,480
Speaker 4: talking to Carl Ortiz, who's a concept artist. She's done

1029
00:54:08,480 --> 00:54:12,239
Speaker 4: work for Marvel and and DC and huge studios and

1030
00:54:12,520 --> 00:54:17,800
Speaker 4: Magic Togathering and you know, and she's basically saying, after

1031
00:54:18,239 --> 00:54:23,560
Speaker 4: mid journey stability, AI produces this incredibly crappy content. Jobs

1032
00:54:23,600 --> 00:54:27,960
Speaker 4: for concept artists have really dried up. They've gone and

1033
00:54:28,040 --> 00:54:30,279
Speaker 4: it's really hard. And I mean, especially for folks who

1034
00:54:30,280 --> 00:54:32,799
Speaker 4: are just breaking into the industry. You might just be

1035
00:54:32,840 --> 00:54:35,600
Speaker 4: trying to get there their you know, their work out

1036
00:54:35,640 --> 00:54:39,680
Speaker 4: there for entry level jobs. They can't find anything right now.

1037
00:54:39,800 --> 00:54:43,319
Speaker 4: And so imagine what that's going to replace, what that's

1038
00:54:43,360 --> 00:54:45,560
Speaker 4: going to encroach on, right, I mean, that's kind of

1039
00:54:45,640 --> 00:54:48,840
Speaker 4: unique thing about kind of creative fields and coding fields

1040
00:54:48,840 --> 00:54:52,239
Speaker 4: and and and whatnot. And then I would say this, yeah,

1041
00:54:52,280 --> 00:54:56,320
Speaker 4: this this automated decision making kind of in government and hiring.

1042
00:54:56,360 --> 00:55:00,759
Speaker 4: I mean, those are those are you know, deathly you know, airfyed, right,

1043
00:55:00,800 --> 00:55:03,279
Speaker 4: And I mean this is already being deployed. I mean

1044
00:55:03,320 --> 00:55:06,400
Speaker 4: at the US Mexico border. There's kind of massive. The

1045
00:55:06,440 --> 00:55:09,640
Speaker 4: Markup actually just put out this interview with David Moss,

1046
00:55:09,680 --> 00:55:13,760
Speaker 4: who's at the Electronic Freedom Foundation, No either the Electronic

1047
00:55:13,760 --> 00:55:16,000
Speaker 4: Freedom Refession. I think he's at the ACLU. I have

1048
00:55:16,040 --> 00:55:18,840
Speaker 4: to look this up, but it's about basically a survey

1049
00:55:19,040 --> 00:55:23,160
Speaker 4: of like surveillance technology that's not the southern border, and

1050
00:55:23,520 --> 00:55:26,560
Speaker 4: Dave Moss is at EFF. The Markdown Putt put a

1051
00:55:26,560 --> 00:55:28,760
Speaker 4: published something about with him. It was like a virtual

1052
00:55:28,800 --> 00:55:32,520
Speaker 4: reality tour of servants technology or something wild like that.

1053
00:55:32,840 --> 00:55:34,880
Speaker 3: I was gonna say. There's also things like shot Spotter,

1054
00:55:35,200 --> 00:55:37,239
Speaker 3: which reports to be able to detect what a gun

1055
00:55:37,280 --> 00:55:39,560
Speaker 3: has been fired. And this has been deployed by police

1056
00:55:39,560 --> 00:55:42,319
Speaker 3: departments all over the country, and there's you know, there's

1057
00:55:42,360 --> 00:55:44,560
Speaker 3: no transparency into how it was evaluatedor how it even

1058
00:55:44,600 --> 00:55:46,200
Speaker 3: works or why you should believe it works. And so

1059
00:55:46,200 --> 00:55:49,760
Speaker 3: what we have is a system that tells the cops

1060
00:55:49,760 --> 00:55:52,359
Speaker 3: that they're running into an active shooter situation, which is

1061
00:55:52,480 --> 00:55:55,680
Speaker 3: definitely a recipe for reducing police violence.

1062
00:55:55,440 --> 00:56:00,319
Speaker 2: Right right, yeah, right, our tide Microsoft, Yeah, and that

1063
00:56:00,360 --> 00:56:02,920
Speaker 2: there is a reason an investigation that that showed that

1064
00:56:03,000 --> 00:56:07,040
Speaker 2: it's always almost always used in neighborhoods.

1065
00:56:07,239 --> 00:56:10,600
Speaker 4: Yeah, communisic, the communities of color. Yeah, the Wired the

1066
00:56:10,640 --> 00:56:12,799
Speaker 4: Wired piece on that basically.

1067
00:56:13,000 --> 00:56:13,239
Speaker 1: Yeah.

1068
00:56:13,239 --> 00:56:15,480
Speaker 4: I think they said about what seventy percent of the

1069
00:56:15,520 --> 00:56:17,480
Speaker 4: census tracts something ridiculous like that.

1070
00:56:17,600 --> 00:56:21,400
Speaker 3: Yeah, Yeah, it's absurd. And then there's things like EPIC,

1071
00:56:21,480 --> 00:56:25,400
Speaker 3: which is a healthcare electronic health records company, is partnering

1072
00:56:25,440 --> 00:56:28,760
Speaker 3: with Microsoft to push the synthetic text into so many

1073
00:56:28,800 --> 00:56:32,400
Speaker 3: different systems, so that you're going to get like reports

1074
00:56:32,440 --> 00:56:35,520
Speaker 3: in your patient records that were automatically generated and then

1075
00:56:35,520 --> 00:56:37,600
Speaker 3: supposedly checked by a doctor who doesn't have time to

1076
00:56:37,680 --> 00:56:40,439
Speaker 3: check it thoroughly, right, and they're going to be doing

1077
00:56:40,480 --> 00:56:44,120
Speaker 3: things like, you know, randomly putting information about BMI and

1078
00:56:44,120 --> 00:56:47,480
Speaker 3: when it's not even relevant, or misgendering people over the place,

1079
00:56:47,719 --> 00:56:49,919
Speaker 3: or you know, this kind of stuff is going to hurt.

1080
00:56:50,200 --> 00:56:50,560
Speaker 1: Yeah.

1081
00:56:50,640 --> 00:56:53,520
Speaker 3: And I want to add to the issue of like

1082
00:56:53,880 --> 00:56:57,239
Speaker 3: entry level jobs drying up for people who do for example, illustration,

1083
00:56:57,840 --> 00:57:00,000
Speaker 3: doctor joybol and Winni points out that we're getting what's

1084
00:57:00,040 --> 00:57:04,720
Speaker 3: called an apprentice gap, where those positions where the easy

1085
00:57:04,719 --> 00:57:06,920
Speaker 3: stuff gets automated. And I don't think this is just

1086
00:57:06,960 --> 00:57:08,960
Speaker 3: in creative fields. But the positions where you're doing the

1087
00:57:08,960 --> 00:57:10,759
Speaker 3: easy stuff and you're learning how to master it and

1088
00:57:10,800 --> 00:57:13,400
Speaker 3: you are working alongside someone who's doing the harder stuff.

1089
00:57:13,800 --> 00:57:16,840
Speaker 3: If that gets replaced by automation, then it becomes harder

1090
00:57:16,880 --> 00:57:19,880
Speaker 3: to train the people to do the stuff that requires expertise.

1091
00:57:19,800 --> 00:57:22,840
Speaker 1: Right, Yeah, And it's easier to do in creative fields

1092
00:57:22,880 --> 00:57:28,000
Speaker 1: because there's such a just inability for executives to you know,

1093
00:57:28,400 --> 00:57:31,160
Speaker 1: like they don't know it. They've never known anything about

1094
00:57:31,200 --> 00:57:33,959
Speaker 1: like what what is quality creative work? So I feel

1095
00:57:34,000 --> 00:57:35,840
Speaker 1: like it's much easier for them to just be like, yeah,

1096
00:57:35,880 --> 00:57:38,600
Speaker 1: get rid of that, and well, yeah, like the way

1097
00:57:38,640 --> 00:57:41,840
Speaker 1: that we'll find out about that that isn't working is

1098
00:57:41,960 --> 00:57:45,600
Speaker 1: the quality of the creative output will be far worse.

1099
00:57:46,160 --> 00:57:48,720
Speaker 4: Right, I mean what I mean, all those folks really

1100
00:57:48,880 --> 00:57:52,560
Speaker 4: tend to care about our you know, content and engagement metrics.

1101
00:57:52,680 --> 00:57:55,600
Speaker 4: You know, you can't actually have something that's kind of

1102
00:57:55,680 --> 00:57:57,680
Speaker 4: known for quality or creativity.

1103
00:57:57,760 --> 00:57:58,000
Speaker 2: Right.

1104
00:57:58,600 --> 00:58:00,440
Speaker 4: It does remind me a little bit about kind of

1105
00:58:00,760 --> 00:58:05,680
Speaker 4: you know, the first rebels against automation, the Luodites, and

1106
00:58:06,040 --> 00:58:07,600
Speaker 4: you know, the kind of the kind of way in

1107
00:58:07,640 --> 00:58:10,919
Speaker 4: which they did have this apprentice guilt system in which

1108
00:58:10,960 --> 00:58:14,000
Speaker 4: they trained for you know, a decade or so before

1109
00:58:14,080 --> 00:58:16,640
Speaker 4: they could you know, perhaps go ahead and open their

1110
00:58:16,640 --> 00:58:19,720
Speaker 4: own shop in a way that you know, those folks

1111
00:58:19,760 --> 00:58:23,640
Speaker 4: were replaced by these these water frames that were they

1112
00:58:23,680 --> 00:58:26,480
Speaker 4: called water frames because they were produced by hydraulic power.

1113
00:58:26,920 --> 00:58:31,280
Speaker 4: But then uh, you know, effectively powered by unskilled people

1114
00:58:31,320 --> 00:58:36,320
Speaker 4: and by unskilled usually children doing incredibly dangerous work. But yeah,

1115
00:58:36,440 --> 00:58:38,840
Speaker 4: folks that have been training for this, the apprentices were

1116
00:58:39,760 --> 00:58:41,160
Speaker 4: incredibly steaming mad.

1117
00:58:41,760 --> 00:58:46,160
Speaker 1: Yeah, and then we made their name a synonym for like.

1118
00:58:48,560 --> 00:58:48,800
Speaker 2: Hater.

1119
00:58:49,200 --> 00:58:55,040
Speaker 4: Yeah, there's been some nice efforts to reclaim them by

1120
00:58:55,280 --> 00:58:58,280
Speaker 4: Brian Merchant and yeah, some folks.

1121
00:58:58,440 --> 00:58:58,800
Speaker 2: Yeah.

1122
00:58:58,840 --> 00:59:01,920
Speaker 1: So just in the comparison to Crypto, it feels like

1123
00:59:03,040 --> 00:59:08,640
Speaker 1: the adoption here, the hype cycle here is more widespread

1124
00:59:08,640 --> 00:59:13,480
Speaker 1: than Crypto. Like with Crypto, there was that moment where

1125
00:59:13,520 --> 00:59:19,480
Speaker 1: we saw the person behind the curtain who was you know,

1126
00:59:19,600 --> 00:59:23,640
Speaker 1: the people that there was the fall of Crypto. With AI,

1127
00:59:23,760 --> 00:59:28,480
Speaker 1: I don't feel like there is any incentive for anyone

1128
00:59:28,560 --> 00:59:32,480
Speaker 1: to any of the stakeholders. I guess I got it

1129
00:59:32,600 --> 00:59:36,280
Speaker 1: involved to yeah, to just come out and be like, yeah,

1130
00:59:36,280 --> 00:59:39,520
Speaker 1: it was bullshit. You know, there's just so much buy

1131
00:59:39,560 --> 00:59:43,640
Speaker 1: in across the board where I guess we've already talked

1132
00:59:43,640 --> 00:59:45,720
Speaker 1: about where you see this going, But is there Do

1133
00:59:45,760 --> 00:59:50,160
Speaker 1: you think there's any hope for this getting kind of

1134
00:59:50,400 --> 00:59:54,560
Speaker 1: found out, the truth catching on, or do you think

1135
00:59:54,560 --> 00:59:57,280
Speaker 1: it's just going to have to be one hundred years

1136
00:59:57,600 --> 01:00:02,520
Speaker 1: from now when somebody changes their mind about AI haters

1137
01:00:02,520 --> 01:00:05,680
Speaker 1: and it's like, actually they were onto something the way

1138
01:00:05,720 --> 01:00:06,760
Speaker 1: we are about light eites.

1139
01:00:07,280 --> 01:00:09,520
Speaker 3: We're still trying. That's what we're up to with the podcast,

1140
01:00:09,640 --> 01:00:12,160
Speaker 3: right a lot of our other public scholarship. There was

1141
01:00:12,200 --> 01:00:16,080
Speaker 3: an interesting moment last week when chat gpt sort of

1142
01:00:16,080 --> 01:00:19,440
Speaker 3: glitched and was out putting sense. Yes, I mean it

1143
01:00:19,480 --> 01:00:21,840
Speaker 3: was it actually Spanglish or people just calling it because

1144
01:00:21,840 --> 01:00:22,800
Speaker 3: it had some Spanish in it.

1145
01:00:22,800 --> 01:00:24,800
Speaker 4: It had some, It wasn't all I mean it was

1146
01:00:24,840 --> 01:00:27,360
Speaker 4: doing some spangless stuff. With this stuff is just even

1147
01:00:27,920 --> 01:00:32,040
Speaker 4: more inscrutable than usual. It was just completely nonsense.

1148
01:00:32,120 --> 01:00:36,280
Speaker 3: Yeah, yeah, And that the sort of open eye statement

1149
01:00:36,280 --> 01:00:39,640
Speaker 3: about what went wrong basically described it for what it is.

1150
01:00:40,120 --> 01:00:41,920
Speaker 3: They had to like say something other than what they

1151
01:00:42,000 --> 01:00:43,720
Speaker 3: usually say. And then there was a whole thing with

1152
01:00:44,560 --> 01:00:48,400
Speaker 3: Google's image generator, which you know the so the baseline

1153
01:00:48,440 --> 01:00:52,080
Speaker 3: thing is super duper biased and like makes everybody look

1154
01:00:52,280 --> 01:00:56,880
Speaker 3: like white people. And there's this wonderful thread on Medium,

1155
01:00:57,360 --> 01:00:58,480
Speaker 3: and here I'm doing the thing wre I can't think

1156
01:00:58,480 --> 01:00:59,960
Speaker 3: of a person's name, but it's a wonderful post on

1157
01:01:00,080 --> 01:01:03,000
Speaker 3: Medium where someone goes through this thread of pictures that

1158
01:01:03,000 --> 01:01:04,880
Speaker 3: I think we're initially on Reddit, where someone asked one

1159
01:01:04,920 --> 01:01:09,200
Speaker 3: of these generators to create warriors from different cultures taking

1160
01:01:09,200 --> 01:01:12,360
Speaker 3: a selfie and they're all doing the American selfie smile.

1161
01:01:13,080 --> 01:01:16,600
Speaker 3: And so does this really uncanny Valley thing going on

1162
01:01:16,640 --> 01:01:18,960
Speaker 3: there where these people from all these different cultures are

1163
01:01:19,160 --> 01:01:23,080
Speaker 3: posing the way Americans pose, so huge bias in these things. Underlyingly,

1164
01:01:23,320 --> 01:01:25,440
Speaker 3: Google has some kind of a patch that basically whatever

1165
01:01:25,480 --> 01:01:27,960
Speaker 3: prompt you put in, they add some additional prompting to

1166
01:01:28,000 --> 01:01:30,840
Speaker 3: make the people look diverse. And then last week or

1167
01:01:30,880 --> 01:01:33,120
Speaker 3: so someone figured out that if you asked it to

1168
01:01:33,200 --> 01:01:35,960
Speaker 3: show you a painting of Nazi soldiers, they're not all

1169
01:01:35,960 --> 01:01:41,760
Speaker 3: white because of this patch, right right, So yeah, So

1170
01:01:42,040 --> 01:01:46,360
Speaker 3: Google's backpedaling of that I think was ridiculous. There was

1171
01:01:46,360 --> 01:01:50,160
Speaker 3: some statement in there about how they certainly don't intend

1172
01:01:50,240 --> 01:01:53,800
Speaker 3: for their system to put out historically inaccurate images. I'm like,

1173
01:01:54,240 --> 01:01:57,400
Speaker 3: what the hell, it's a fake image, like there's no

1174
01:01:57,480 --> 01:02:02,600
Speaker 3: accuracy there, no matter what. These mishaps maybe sort of

1175
01:02:02,680 --> 01:02:05,520
Speaker 3: pulled back the curtain for a broader group of people.

1176
01:02:05,640 --> 01:02:07,360
Speaker 3: There's some true believers out there who are not going

1177
01:02:07,400 --> 01:02:09,080
Speaker 3: to be reached. Sure, but I think it may have

1178
01:02:09,120 --> 01:02:11,040
Speaker 3: helped for some slice of society.

1179
01:02:11,240 --> 01:02:13,200
Speaker 2: Yeah, I know some people who are like, how do

1180
01:02:13,280 --> 01:02:18,920
Speaker 2: we know that Reinhardt Heydrich didn't have dreadlocks? Clear?

1181
01:02:19,280 --> 01:02:19,520
Speaker 1: Wow?

1182
01:02:19,680 --> 01:02:20,880
Speaker 2: But okay, sure, go off.

1183
01:02:21,000 --> 01:02:23,439
Speaker 4: Yeah, And I mean I think it's this is such

1184
01:02:23,440 --> 01:02:26,680
Speaker 4: an interesting question, right, is like where does when is

1185
01:02:26,680 --> 01:02:28,240
Speaker 4: the AI bubble going to pop?

1186
01:02:28,320 --> 01:02:28,560
Speaker 3: Right?

1187
01:02:28,640 --> 01:02:32,360
Speaker 4: And I mean, in some ways it feels like, you know,

1188
01:02:32,360 --> 01:02:34,480
Speaker 4: we're where, I mean, as much as we can do.

1189
01:02:34,560 --> 01:02:37,240
Speaker 4: We're kind of prodding it, right, you know, and seeing

1190
01:02:37,720 --> 01:02:40,680
Speaker 4: you know. And one thing that I off handedly said

1191
01:02:40,680 --> 01:02:44,400
Speaker 4: one one time and Emily loves is ridicule as praxis.

1192
01:02:44,680 --> 01:02:47,480
Speaker 4: So yeah, one thing you know, and I will in

1193
01:02:47,560 --> 01:02:50,880
Speaker 4: a turn of one of Emily's great quotes, is you know,

1194
01:02:51,480 --> 01:02:53,600
Speaker 4: oh gosh, I'm going to mess it up right now.

1195
01:02:53,680 --> 01:02:56,840
Speaker 4: It's I and I. It's it's the refuse to be impressed.

1196
01:02:57,520 --> 01:03:00,160
Speaker 4: Uh oh yeah, resist, the urge to be resist? Is

1197
01:03:00,160 --> 01:03:02,280
Speaker 4: this year to be impressed. It's much it's much better

1198
01:03:02,280 --> 01:03:05,000
Speaker 4: when when when she says it, and it's and it's

1199
01:03:05,080 --> 01:03:07,240
Speaker 4: and it's kind of the idea of like some of

1200
01:03:07,240 --> 01:03:09,960
Speaker 4: it is a bit of the sheene right, But it

1201
01:03:10,000 --> 01:03:13,120
Speaker 4: feels like at some point, you know, if in the

1202
01:03:13,200 --> 01:03:16,439
Speaker 4: kind of operations of these things, you know enough, there'll

1203
01:03:16,480 --> 01:03:20,080
Speaker 4: be enough buy in, especially with automation, that's hard to

1204
01:03:20,560 --> 01:03:25,040
Speaker 4: reverse a lot of the automation without just a huge fight.

1205
01:03:25,200 --> 01:03:27,600
Speaker 4: And so I mean, I think something that helps our

1206
01:03:28,280 --> 01:03:32,200
Speaker 4: you know, worker led efforts, you know, and one of

1207
01:03:32,240 --> 01:03:34,760
Speaker 4: the most awesome things that we've seen this is something

1208
01:03:34,800 --> 01:03:41,520
Speaker 4: a scholar, Blair a tear Frost, calls AI counter governance,

1209
01:03:42,200 --> 01:03:45,960
Speaker 4: she calls and one example she is is the w

1210
01:03:46,080 --> 01:03:50,400
Speaker 4: GA strike and how folks struck for one hundred and

1211
01:03:50,440 --> 01:03:54,400
Speaker 4: forty eight days and after strike they not only got

1212
01:03:54,480 --> 01:03:59,280
Speaker 4: a bunch of new kinds of guarantees for minimum pay

1213
01:03:59,440 --> 01:04:01,920
Speaker 4: when it came to streaming and the residuals they get

1214
01:04:01,960 --> 01:04:06,320
Speaker 4: for it, they also, you know, have to basically be

1215
01:04:06,400 --> 01:04:09,800
Speaker 4: informed if any AI is being used in the writer's room.

1216
01:04:10,360 --> 01:04:13,920
Speaker 4: They can't they can't be forced to edit or re

1217
01:04:14,160 --> 01:04:17,480
Speaker 4: edit or rewrite AI generated content. They have to be

1218
01:04:17,600 --> 01:04:20,480
Speaker 4: everything has to be basically above board. I mean, isn't

1219
01:04:20,480 --> 01:04:22,800
Speaker 4: as far as it could have gone and banned it

1220
01:04:22,840 --> 01:04:25,560
Speaker 4: out right right, But if there's any use of it

1221
01:04:25,600 --> 01:04:28,560
Speaker 4: has to be disclosed and you can't bring in these

1222
01:04:28,600 --> 01:04:31,640
Speaker 4: tools to hold you know, whole cloth to place the

1223
01:04:31,680 --> 01:04:32,360
Speaker 4: writer's room.

1224
01:04:32,800 --> 01:04:36,240
Speaker 1: Yeah. Well, Alex and Emily, it has been such a

1225
01:04:36,240 --> 01:04:37,880
Speaker 1: pleasure having you on the show. I feel like you

1226
01:04:37,920 --> 01:04:43,919
Speaker 1: could keep talking about this hours for days. But thank

1227
01:04:43,960 --> 01:04:46,000
Speaker 1: you so much for coming on. And yeah, I would

1228
01:04:46,000 --> 01:04:48,720
Speaker 1: love to have you back. Where can people find you?

1229
01:04:49,040 --> 01:04:51,320
Speaker 1: Follow you here? Your podcast all that good stuff.

1230
01:04:51,560 --> 01:04:54,280
Speaker 3: Yeah, so the podcast is called Mystery AI Hype Theater

1231
01:04:54,400 --> 01:04:57,560
Speaker 3: three thousand. Then you can find it anywhere find podcasts

1232
01:04:57,560 --> 01:05:00,600
Speaker 3: are streamed, but also on peer two if you prefer

1233
01:05:00,680 --> 01:05:02,880
Speaker 3: that kind of content as video. So we start off

1234
01:05:02,880 --> 01:05:05,680
Speaker 3: this Twitch and then show up that way. I'm on

1235
01:05:06,160 --> 01:05:08,680
Speaker 3: all the socials. I tend to be Emily M. Bender

1236
01:05:08,880 --> 01:05:11,760
Speaker 3: so Twitter or blue sky on Macedon. I'm Emily M.

1237
01:05:11,800 --> 01:05:15,160
Speaker 3: Bender at what are we dare institute that social and

1238
01:05:15,200 --> 01:05:17,880
Speaker 3: I'm very findable on the web, very googleable at the

1239
01:05:17,960 --> 01:05:19,920
Speaker 3: University of Washington. How about you, Alex?

1240
01:05:20,720 --> 01:05:23,840
Speaker 4: Yeah, and if you want to chech catch the twitch stream.

1241
01:05:23,920 --> 01:05:29,560
Speaker 4: It's Twitch, dot tv, slash dare, undiscore Institute, usually on Mondays,

1242
01:05:29,640 --> 01:05:33,240
Speaker 4: usually this time slot. So next week we'll be here. Wait,

1243
01:05:33,720 --> 01:05:35,520
Speaker 4: I forget, we'll be here eventually.

1244
01:05:35,760 --> 01:05:37,640
Speaker 3: Yes, today's to today's Tuesday though in.

1245
01:05:39,160 --> 01:05:42,080
Speaker 4: That's right, that's right, Yeah, it's actually not so Mondays.

1246
01:05:42,480 --> 01:05:47,200
Speaker 4: Check us out. You can catch me at Twitter and

1247
01:05:47,200 --> 01:05:50,560
Speaker 4: Blue Sky Alex Hannah no h at the end. And

1248
01:05:50,640 --> 01:05:55,480
Speaker 4: then yeah, our mast on server, dare, hyphencommunity, dot social,

1249
01:05:56,640 --> 01:05:59,320
Speaker 4: what else? I think those are all the socials we have?

1250
01:06:00,160 --> 01:06:04,120
Speaker 1: Yeah, amazing. Is there a work of media that you've

1251
01:06:04,160 --> 01:06:04,760
Speaker 1: been enjoying?

1252
01:06:05,200 --> 01:06:09,400
Speaker 3: So I am super excited about doctor Joeyblamuni's book Unmasking AI,

1253
01:06:09,800 --> 01:06:13,400
Speaker 3: came out last Halloween. It is a beautiful sort of

1254
01:06:13,440 --> 01:06:18,040
Speaker 3: combination memoir slash introduction into the world of how so

1255
01:06:18,120 --> 01:06:20,960
Speaker 3: called AI works, how policymaking is done around it, how

1256
01:06:21,000 --> 01:06:23,800
Speaker 3: activism is done around it. And I actually got to

1257
01:06:23,960 --> 01:06:26,360
Speaker 3: interview her here in Seattle at the University of Washington

1258
01:06:26,440 --> 01:06:27,680
Speaker 3: last week, which was amazing.

1259
01:06:28,240 --> 01:06:30,360
Speaker 1: Nice. How about you, Alex?

1260
01:06:31,040 --> 01:06:36,440
Speaker 4: Yeah, I'd recommend the podcast Worlds Beyond Number and it

1261
01:06:36,520 --> 01:06:41,880
Speaker 4: is a Dungeons and Dragons live play, actual play podcast

1262
01:06:42,320 --> 01:06:46,400
Speaker 4: that is put Together by Brendan Lee Mulligan, Abria Engar,

1263
01:06:46,640 --> 01:06:51,600
Speaker 4: Lou Wilson, and Erica is talking about writing that is

1264
01:06:51,640 --> 01:06:56,320
Speaker 4: not generated by AI, but generated by some very talented

1265
01:06:56,640 --> 01:06:59,400
Speaker 4: improv comedians and storytellers themselves.

1266
01:07:00,440 --> 01:07:02,160
Speaker 1: Acale, How do you scale?

1267
01:07:02,200 --> 01:07:06,240
Speaker 4: Oh my god, it is actually surprisingly I mean that

1268
01:07:06,280 --> 01:07:10,720
Speaker 4: podcast is actually I think the single most funded thing

1269
01:07:10,800 --> 01:07:13,280
Speaker 4: on Patreon. Birth a bunch of records for it. I

1270
01:07:13,440 --> 01:07:18,680
Speaker 4: had about forty thousand subscribers. People want good content, Yeah,

1271
01:07:18,840 --> 01:07:23,800
Speaker 4: I'm humans high art, you know. People want this vibrant storytelling,

1272
01:07:24,400 --> 01:07:28,439
Speaker 4: amazing creation, you know, and so the more we can

1273
01:07:28,560 --> 01:07:30,000
Speaker 4: we can plug that stuff the better.

1274
01:07:30,400 --> 01:07:33,520
Speaker 1: By the way, you can also find Mystery AI Hype

1275
01:07:33,880 --> 01:07:37,400
Speaker 1: Theater three thousand. Like where podcasts are? So in case

1276
01:07:37,520 --> 01:07:41,760
Speaker 1: you don't do Twitch. Yeah, like I was listening you know,

1277
01:07:42,120 --> 01:07:43,160
Speaker 1: Spotify or whatever.

1278
01:07:43,240 --> 01:07:43,920
Speaker 2: You weren't on Twitch.

1279
01:07:44,640 --> 01:07:46,720
Speaker 1: I wasn't on Twitch this weekend. I was taking it.

1280
01:07:46,800 --> 01:07:48,960
Speaker 1: I was taking a break. You know. It's just like

1281
01:07:49,040 --> 01:07:50,560
Speaker 1: too much, too much Twitch.

1282
01:07:50,680 --> 01:07:51,640
Speaker 2: Couldn't get a hyperion.

1283
01:07:51,920 --> 01:07:54,680
Speaker 1: That's right? Okay, all right, Miles, where can people find you?

1284
01:07:54,720 --> 01:07:56,560
Speaker 1: What's working media you've been enjoying.

1285
01:07:56,720 --> 01:07:59,800
Speaker 2: At the at base platforms at Miles of Gray also

1286
01:08:00,000 --> 01:08:02,080
Speaker 2: if you like the NBA, you can hear Jack and

1287
01:08:02,120 --> 01:08:04,160
Speaker 2: I just go on and on about that on our

1288
01:08:04,240 --> 01:08:07,080
Speaker 2: other podcast, Miles and Jack Got Mad Moosties. And if

1289
01:08:07,080 --> 01:08:10,000
Speaker 2: you like trash reality TV like me, I like I

1290
01:08:10,080 --> 01:08:12,600
Speaker 2: comfort myself with it, check me out on four to

1291
01:08:12,600 --> 01:08:16,760
Speaker 2: twenty Day Fiance where we talk about ninety day Fiance tweet.

1292
01:08:16,840 --> 01:08:19,880
Speaker 2: I like, this is just kind of it's a historical day.

1293
01:08:19,920 --> 01:08:22,160
Speaker 2: I would have to say for the Internet. We missed

1294
01:08:22,160 --> 01:08:27,000
Speaker 2: an anniversary on Sunday. Apparently the very famous clip of

1295
01:08:27,080 --> 01:08:31,439
Speaker 2: Peter Weber, the bowler who won and came out with

1296
01:08:31,600 --> 01:08:36,120
Speaker 2: the iconic phrase who do you think you are? I Am?

1297
01:08:36,640 --> 01:08:39,040
Speaker 2: That had its twelfth anniversary and I just will to

1298
01:08:39,040 --> 01:08:41,120
Speaker 2: play that audio day because it's one of our favorites.

1299
01:08:42,000 --> 01:08:47,799
Speaker 2: Here we go, are you?

1300
01:08:49,400 --> 01:08:52,000
Speaker 1: Who are all?

1301
01:08:52,160 --> 01:08:56,360
Speaker 2: Right? Who do you think you are? Who do you?

1302
01:08:56,400 --> 01:08:56,519
Speaker 1: Oh?

1303
01:08:56,760 --> 01:08:57,080
Speaker 2: Man?

1304
01:08:57,479 --> 01:09:00,400
Speaker 1: The whole thing is good. Also, the run up to

1305
01:09:00,479 --> 01:09:02,920
Speaker 1: that is great. That's right. I did it.

1306
01:09:03,040 --> 01:09:03,639
Speaker 2: I did it?

1307
01:09:03,840 --> 01:09:04,760
Speaker 1: God damn it?

1308
01:09:05,840 --> 01:09:12,720
Speaker 4: What I yeah? I play an amateur sport. And just

1309
01:09:12,920 --> 01:09:16,360
Speaker 4: like the exhilaration you get and you just say anything.

1310
01:09:16,120 --> 01:09:17,560
Speaker 2: Yeah, yeah, I know.

1311
01:09:17,760 --> 01:09:20,120
Speaker 4: That's I love that clip.

1312
01:09:20,200 --> 01:09:22,840
Speaker 2: I respect I think you are? I It's like yeah,

1313
01:09:22,880 --> 01:09:26,080
Speaker 2: you've you're you've glory fried your brain circuits and now

1314
01:09:26,120 --> 01:09:29,599
Speaker 2: that's what you're saying. And the other one from at

1315
01:09:29,640 --> 01:09:32,000
Speaker 2: the Dark Prowler. It's a screenshot of an Uber notification

1316
01:09:32,040 --> 01:09:34,879
Speaker 2: on the phone. It says, young Stroker, the body Snatcher

1317
01:09:34,920 --> 01:09:37,120
Speaker 2: will be joining you along your route. You're still on

1318
01:09:37,240 --> 01:09:39,200
Speaker 2: schedule to arrive by seven fifty four.

1319
01:09:41,200 --> 01:09:49,280
Speaker 1: Okay, by the way, chicken nuggets glory fried. Thanks, Yeah,

1320
01:09:49,320 --> 01:09:53,000
Speaker 1: that'd be great. Oh yeah you Oh tweet I've been enjoying.

1321
01:09:53,080 --> 01:09:58,400
Speaker 1: Caitlyn at Kitlyn at Kate Holes tweeted overnight oats sounds

1322
01:09:58,439 --> 01:10:05,640
Speaker 1: like the name of a race horse who's sucks. That

1323
01:10:05,720 --> 01:10:07,880
Speaker 1: was an old one that was apparently September twenty fifth,

1324
01:10:07,960 --> 01:10:08,760
Speaker 1: twenty twenty two.

1325
01:10:09,040 --> 01:10:10,120
Speaker 2: Good one, Go banger.

1326
01:10:10,479 --> 01:10:14,000
Speaker 1: You can find me on Twitter at Jack Underscore O'Brien.

1327
01:10:14,040 --> 01:10:16,639
Speaker 1: You can find us on Twitter at daily Zeitgeist. We're

1328
01:10:16,640 --> 01:10:19,479
Speaker 1: at the Daily Zeitgeist on Instagram. We have a Facebook

1329
01:10:19,479 --> 01:10:22,639
Speaker 1: fan page on a website Daily zeikegeist dot com where

1330
01:10:22,640 --> 01:10:26,040
Speaker 1: you post our episode and our footnote. We link off

1331
01:10:26,040 --> 01:10:28,400
Speaker 1: to the information that we talked about in today's episode,

1332
01:10:28,640 --> 01:10:30,759
Speaker 1: as well as a song that we think you might enjoy.

1333
01:10:30,800 --> 01:10:33,639
Speaker 1: Miles is there a song that you think people might enjoy.

1334
01:10:33,880 --> 01:10:37,280
Speaker 2: This is just a nice little instrumental clip, not clip,

1335
01:10:37,280 --> 01:10:40,679
Speaker 2: a full on track from a London based producer goes

1336
01:10:40,680 --> 01:10:44,400
Speaker 2: by Vegan veg y N and the track is called

1337
01:10:44,520 --> 01:10:48,360
Speaker 2: a Dream goes On Forever and it's a super dreamy track.

1338
01:10:48,400 --> 01:10:51,000
Speaker 2: It's just kind of an interesting production. So maybe you know,

1339
01:10:51,280 --> 01:10:54,000
Speaker 2: maybe you know, just check out some mid journey you

1340
01:10:54,040 --> 01:10:57,439
Speaker 2: know images, some sore AI videos and just blast this

1341
01:10:57,560 --> 01:11:00,320
Speaker 2: in your headphones. Just kind of go on for a dream,

1342
01:11:00,320 --> 01:11:00,960
Speaker 2: goes for.

1343
01:11:00,920 --> 01:11:05,439
Speaker 3: The music and leave the synthetic media.

1344
01:11:04,280 --> 01:11:06,960
Speaker 1: Just fully embraced into the music, close your eyes and

1345
01:11:07,240 --> 01:11:10,839
Speaker 1: actually the mid journey images that your brain will produce.

1346
01:11:12,040 --> 01:11:14,880
Speaker 1: I've found that like at night when I sleep, sometimes

1347
01:11:14,960 --> 01:11:17,360
Speaker 1: my brain produces Sora images.

1348
01:11:17,800 --> 01:11:19,519
Speaker 2: This is like you sound like you sound like a

1349
01:11:19,600 --> 01:11:22,160
Speaker 2: youth pastor talking to kids about God. Or it's like

1350
01:11:22,200 --> 01:11:24,720
Speaker 2: you know where it's really the AI is actually up

1351
01:11:24,760 --> 01:11:25,400
Speaker 2: in here kids?

1352
01:11:25,479 --> 01:11:27,280
Speaker 1: Yeah, want to get.

1353
01:11:28,880 --> 01:11:32,560
Speaker 4: Yeah yeah yeah. The real training data we had is

1354
01:11:33,040 --> 01:11:33,559
Speaker 4: the word of.

1355
01:11:33,520 --> 01:11:35,800
Speaker 1: God, exactly exactly what I need.

1356
01:11:36,200 --> 01:11:38,280
Speaker 2: The Daily Zeitgeist is a production of by Heart Radio.

1357
01:11:38,280 --> 01:11:40,160
Speaker 1: For more podcasts from my Heart Radio, visit the iHeart

1358
01:11:40,240 --> 01:11:42,000
Speaker 1: Radio ap Apple podcast or wherever you listen to your

1359
01:11:42,000 --> 01:11:43,519
Speaker 1: favorite shows, that it is going to do it for

1360
01:11:43,680 --> 01:11:46,880
Speaker 1: us this morning. We are back this afternoon to tell

1361
01:11:46,920 --> 01:11:49,240
Speaker 1: you what is trending, and we'll talk to you all then.

1362
01:11:49,560 --> 01:11:52,639
Speaker 2: Bye bye,