1
00:00:02,400 --> 00:00:03,560
Speaker 1: Also media.

2
00:00:05,240 --> 00:00:07,800
Speaker 2: Hello and welcome to Better Offline. I'm your host Ed Zichron,

3
00:00:19,640 --> 00:00:21,920
Speaker 2: and we are recording here in beautiful New York City,

4
00:00:22,079 --> 00:00:24,040
Speaker 2: and I have a wonderful pair of guests today. I have,

5
00:00:24,360 --> 00:00:28,520
Speaker 2: of course, Alison Morrow from CNN Nightcap Newsletter and Edward

6
00:00:28,600 --> 00:00:32,560
Speaker 2: Anguiso Junior, the hater himself from the Tech Bubble newsletter.

7
00:00:32,800 --> 00:00:34,400
Speaker 2: Thank you so much for joining me today.

8
00:00:34,880 --> 00:00:35,760
Speaker 3: Great to be here as all.

9
00:00:36,680 --> 00:00:38,839
Speaker 2: So I think we should start with exactly what we

10
00:00:38,840 --> 00:00:41,000
Speaker 2: were just talking about. The open AI claims that they

11
00:00:41,040 --> 00:00:44,000
Speaker 2: have worked out what causes hallucinations. Alison, do you want

12
00:00:44,000 --> 00:00:44,519
Speaker 2: to go over this?

13
00:00:45,640 --> 00:00:47,479
Speaker 3: I should have read the paper a bit more carefully,

14
00:00:47,520 --> 00:00:50,280
Speaker 3: but I can you know. The highlights were getting digested

15
00:00:50,479 --> 00:00:54,640
Speaker 3: yesterday on X and Blue Sky, and it seems like

16
00:00:55,200 --> 00:00:58,240
Speaker 3: it's kind of the test taking problem of when you

17
00:00:58,360 --> 00:01:01,440
Speaker 3: encourage students when they're taking anderdized tests and you don't

18
00:01:01,480 --> 00:01:04,200
Speaker 3: know the answer, you guess, And that's exactly how these

19
00:01:04,280 --> 00:01:07,360
Speaker 3: models are trained, and they don't. You don't get a

20
00:01:07,400 --> 00:01:10,279
Speaker 3: point if you say, I don't know, come up with something,

21
00:01:10,360 --> 00:01:16,560
Speaker 3: and the models are meant to keep guessing until they

22
00:01:16,560 --> 00:01:19,520
Speaker 3: get something close to right. Right, So that's why you

23
00:01:19,560 --> 00:01:24,839
Speaker 3: get a lot of nonsense and hallucinations and open AI

24
00:01:25,000 --> 00:01:27,399
Speaker 3: at least in their reading of it, says oh, this

25
00:01:27,480 --> 00:01:30,200
Speaker 3: is a simple solution. We'll just encourage the models to

26
00:01:30,560 --> 00:01:34,080
Speaker 3: understand better when it's like a binary question and when

27
00:01:34,319 --> 00:01:36,280
Speaker 3: you can say I don't know the answer to that,

28
00:01:36,760 --> 00:01:37,360
Speaker 3: so we'll see.

29
00:01:38,000 --> 00:01:40,759
Speaker 2: I read the paper, albeit not today because my brain

30
00:01:40,840 --> 00:01:43,440
Speaker 2: just immediately read it as marketing copy. Just put that,

31
00:01:43,680 --> 00:01:47,560
Speaker 2: replaced it with another anime theme song. I went through

32
00:01:47,600 --> 00:01:49,680
Speaker 2: it and I was like, okay, so it's going to

33
00:01:49,720 --> 00:01:52,400
Speaker 2: encourage them to say I don't know. This feels like

34
00:01:52,480 --> 00:01:54,920
Speaker 2: a very flat view of what hallucinations are, though, because

35
00:01:55,000 --> 00:01:58,600
Speaker 2: hallucinations as people know them authoritatively stating something that isn't true,

36
00:01:58,880 --> 00:02:02,040
Speaker 2: but hallucinations in life coding model are. It will just

37
00:02:02,040 --> 00:02:04,680
Speaker 2: say yeah, I did that when it didn't. This is

38
00:02:04,800 --> 00:02:06,680
Speaker 2: very common you going through the Cursor and the Claude

39
00:02:06,680 --> 00:02:09,360
Speaker 2: code forums, you can see this or subredits at least,

40
00:02:09,560 --> 00:02:12,400
Speaker 2: and it's not that they say something that's true. They

41
00:02:12,400 --> 00:02:15,919
Speaker 2: don't know that something's not true also, so they might

42
00:02:16,000 --> 00:02:19,239
Speaker 2: say they don't know, and it's just very silly because

43
00:02:20,000 --> 00:02:22,520
Speaker 2: they claim that they're going to fix this problem with

44
00:02:22,600 --> 00:02:25,359
Speaker 2: this solution, but they've done years and hundreds of millions

45
00:02:25,360 --> 00:02:28,440
Speaker 2: of dollars of reinforcement learning. Do they unreinforce do they

46
00:02:28,480 --> 00:02:33,040
Speaker 2: reinforce it? More so that I don't know. I'm fucking

47
00:02:33,080 --> 00:02:34,799
Speaker 2: tired of these companies are going to be completely honest.

48
00:02:34,800 --> 00:02:36,440
Speaker 2: I realized this is kind of cliche for the show,

49
00:02:36,480 --> 00:02:38,920
Speaker 2: but it's the fact that these things get written up

50
00:02:38,919 --> 00:02:41,320
Speaker 2: as very serious things. US is saying, you, guys still

51
00:02:41,320 --> 00:02:43,360
Speaker 2: haven't worked this out. It's kind of frustrating to me.

52
00:02:44,400 --> 00:02:46,720
Speaker 2: And they don't seem to be the models. Some getting

53
00:02:46,760 --> 00:02:49,120
Speaker 2: bad dimension returns and all that, but this is the

54
00:02:49,120 --> 00:02:49,960
Speaker 2: best they've got.

55
00:02:50,280 --> 00:02:54,400
Speaker 4: I mean, does this kind of thrust feel like, you know,

56
00:02:54,400 --> 00:02:56,400
Speaker 4: do you guys feel like it's a downstream of the

57
00:02:56,480 --> 00:02:59,000
Speaker 4: attempts by some of these firms to say, oh, actually,

58
00:02:59,000 --> 00:03:03,120
Speaker 4: like if we dialor bad the sick a fancy, you know,

59
00:03:03,280 --> 00:03:05,520
Speaker 4: then we'll be able to have a much more engaging

60
00:03:05,600 --> 00:03:08,800
Speaker 4: consumer product. We'll be able to have it hallucinate less,

61
00:03:09,280 --> 00:03:11,040
Speaker 4: you know, and do psychosis less.

62
00:03:11,480 --> 00:03:11,680
Speaker 2: You know.

63
00:03:11,880 --> 00:03:13,639
Speaker 4: But does it do they feel linked in that way?

64
00:03:13,720 --> 00:03:16,440
Speaker 4: Does it just feel like a they you know, another

65
00:03:16,639 --> 00:03:17,760
Speaker 4: maybe another dead end.

66
00:03:18,080 --> 00:03:20,080
Speaker 2: I feel like it's just them trying to work shit

67
00:03:20,160 --> 00:03:22,760
Speaker 2: out and trying to like this also feels like a

68
00:03:22,840 --> 00:03:26,079
Speaker 2: very rudimentary answer that they probably already had. It doesn't.

69
00:03:26,720 --> 00:03:28,880
Speaker 2: Anytime someone comes up with an idea that I can

70
00:03:28,919 --> 00:03:31,200
Speaker 2: that's technical, I'm like, okay, mate, you cannot be cooking

71
00:03:31,240 --> 00:03:33,760
Speaker 2: with gas. This is this cannot be that good an idea.

72
00:03:33,800 --> 00:03:35,640
Speaker 2: And the thing is the sick of fancy problem I

73
00:03:35,640 --> 00:03:38,520
Speaker 2: don't think is solvable through solving hallucinations. The problem is

74
00:03:38,680 --> 00:03:41,880
Speaker 2: it should stop. It needs to just It should not

75
00:03:41,920 --> 00:03:44,440
Speaker 2: say I don't understand, it should say no. Actually, you

76
00:03:44,520 --> 00:03:46,680
Speaker 2: sound like you think I'm God King and that you

77
00:03:46,720 --> 00:03:48,480
Speaker 2: are God King, and that as God kings together we

78
00:03:48,480 --> 00:03:50,440
Speaker 2: will destroy the world. Yeah. I mean, I mean my

79
00:03:50,560 --> 00:03:51,200
Speaker 2: case is true.

80
00:03:51,280 --> 00:03:54,160
Speaker 4: But do you think like the emphasis or the attempt

81
00:03:54,200 --> 00:03:56,440
Speaker 4: to overcorrect on that is leading them to go down

82
00:03:56,440 --> 00:03:58,640
Speaker 4: solutions where they think, oh, like if we just like

83
00:03:58,760 --> 00:03:59,600
Speaker 4: this and that, then we.

84
00:03:59,800 --> 00:04:02,880
Speaker 2: Yeah, yes, yes, actually I think that that's right because

85
00:04:02,880 --> 00:04:04,360
Speaker 2: they have the I don't know if you saw the

86
00:04:04,400 --> 00:04:07,640
Speaker 2: Attorney general. Attorney's general from Delaware and California sent a

87
00:04:07,720 --> 00:04:11,000
Speaker 2: letter to open AI last week saying, hey, look, if

88
00:04:11,040 --> 00:04:13,520
Speaker 2: you need to fix these safety protocols, you need to

89
00:04:13,520 --> 00:04:15,440
Speaker 2: actually have them, because what you have right now doesn't work,

90
00:04:15,440 --> 00:04:18,400
Speaker 2: and we will block your nonprofit otherwise. I was really

91
00:04:18,440 --> 00:04:20,240
Speaker 2: happy to read that right up until the bit they

92
00:04:20,279 --> 00:04:23,480
Speaker 2: say we wish you well with your quest for AI dominance.

93
00:04:23,520 --> 00:04:26,680
Speaker 2: I'm just like, these are the fucking people protecting us

94
00:04:26,720 --> 00:04:28,800
Speaker 2: from the They're like, no, it's great, you want to

95
00:04:28,839 --> 00:04:32,359
Speaker 2: dominate everyone with AI. It's just you drove some people

96
00:04:32,400 --> 00:04:34,000
Speaker 2: to a murder suicide situation.

97
00:04:34,120 --> 00:04:36,880
Speaker 3: And wasn't that part of the problem with GPT five

98
00:04:36,960 --> 00:04:39,279
Speaker 3: as they tried to dial back the SICKA fancy and

99
00:04:39,320 --> 00:04:43,160
Speaker 3: then it took away the character and like the humanness

100
00:04:43,160 --> 00:04:46,240
Speaker 3: that people had gotten attached to in GPT four, yeah,

101
00:04:46,279 --> 00:04:49,280
Speaker 3: in earlier models, and like it all seems to come

102
00:04:49,320 --> 00:04:52,440
Speaker 3: back to open. AI doesn't realize that what it's selling

103
00:04:53,040 --> 00:04:57,600
Speaker 3: often is like a companion and a therapist, and it doesn't.

104
00:04:57,720 --> 00:05:00,000
Speaker 3: It reminds me of like Q tips. You're not supposed

105
00:05:00,040 --> 00:05:02,240
Speaker 3: but them in your ear right, but that's all anyone

106
00:05:02,320 --> 00:05:03,080
Speaker 3: uses qtubes.

107
00:05:03,360 --> 00:05:06,160
Speaker 2: I was gonna say they're not meant to do, of.

108
00:05:06,120 --> 00:05:09,359
Speaker 3: Course not, but like that's the consumer, that's how the

109
00:05:09,400 --> 00:05:13,440
Speaker 3: consumer has chosen to use this product. And they're saying like, well,

110
00:05:13,480 --> 00:05:15,680
Speaker 3: we don't condone that, we don't think it's like the

111
00:05:15,720 --> 00:05:18,240
Speaker 3: best use of our product, and you know, we know

112
00:05:18,320 --> 00:05:20,359
Speaker 3: better than the than the consumer, of course.

113
00:05:20,480 --> 00:05:22,760
Speaker 2: I actually I think it's one abstraction high, which is

114
00:05:22,800 --> 00:05:25,360
Speaker 2: I don't think they know what chat GPT is. I don't.

115
00:05:25,480 --> 00:05:28,800
Speaker 2: I did AI boaster piece came out last week, but

116
00:05:28,920 --> 00:05:30,880
Speaker 2: it was I had this whole thing where it was.

117
00:05:31,360 --> 00:05:33,840
Speaker 2: It's so they don't describe what chat GPT does. If

118
00:05:33,839 --> 00:05:36,200
Speaker 2: you're on the website, like it's like, yeah, analyzes data

119
00:05:36,240 --> 00:05:40,520
Speaker 2: and it's it's brainstorming and charts and stuff. The agent

120
00:05:40,560 --> 00:05:44,200
Speaker 2: does things too, Please buy it and then you try

121
00:05:44,200 --> 00:05:46,680
Speaker 2: and like actually look for any use cases and there's nothing.

122
00:05:46,760 --> 00:05:48,840
Speaker 2: I think that they're just guessing. But my favorite thing

123
00:05:48,839 --> 00:05:51,560
Speaker 2: I'm seeing, my favorite thing is that now people are

124
00:05:51,600 --> 00:05:54,560
Speaker 2: like GPT four h isn't the same because they brought

125
00:05:54,560 --> 00:05:57,159
Speaker 2: it back and now people are just freaking out. People like, no,

126
00:05:57,240 --> 00:06:00,359
Speaker 2: it's not the same, it's different. Somehow it's on I

127
00:06:00,360 --> 00:06:02,359
Speaker 2: honestly don't know if it's true. I just think that

128
00:06:02,400 --> 00:06:04,760
Speaker 2: they've entered gamer mode. This is just what. This is

129
00:06:04,760 --> 00:06:06,880
Speaker 2: how gamers react. It's like the gun isn't the same,

130
00:06:06,920 --> 00:06:10,040
Speaker 2: it's literally the same code. No, you've changed something, You've

131
00:06:10,360 --> 00:06:13,200
Speaker 2: I know, and it's what happens when you release an

132
00:06:13,240 --> 00:06:18,080
Speaker 2: imprecise glazing bot onto the world. It's also really funny

133
00:06:18,120 --> 00:06:20,839
Speaker 2: how literally any of these companies could have been open AI.

134
00:06:21,040 --> 00:06:23,200
Speaker 2: There doesn't seem to be anything special about this company

135
00:06:23,200 --> 00:06:24,880
Speaker 2: at all anymore or ever.

136
00:06:24,800 --> 00:06:27,720
Speaker 4: Well you know, I mean, not everybody can lose as

137
00:06:27,800 --> 00:06:32,360
Speaker 4: much money that feels a very special or have the

138
00:06:32,400 --> 00:06:33,840
Speaker 4: connection to you.

139
00:06:34,200 --> 00:06:36,280
Speaker 2: Oh my god. What I haven't heard from Maco's son

140
00:06:36,279 --> 00:06:39,480
Speaker 2: in the minute. We haven't had an announcement of tomorrow

141
00:06:39,640 --> 00:06:42,320
Speaker 2: or tomorrow is gonna be? I like that he bought

142
00:06:42,360 --> 00:06:45,039
Speaker 2: a he bought a Fox count Lab, an old Fox

143
00:06:45,080 --> 00:06:48,480
Speaker 2: con lab too turned into an AI server building place

144
00:06:48,520 --> 00:06:51,599
Speaker 2: in like Ohio. I think it is. It's like, mate,

145
00:06:51,839 --> 00:06:54,839
Speaker 2: what are you doing? Man? Are you okay? Someone needs

146
00:06:54,880 --> 00:06:57,279
Speaker 2: to check in on mass We need to hold space

147
00:06:57,320 --> 00:06:58,640
Speaker 2: for Masso Shei's son.

148
00:06:58,839 --> 00:06:59,560
Speaker 4: Yeah, bring them on.

149
00:07:00,640 --> 00:07:03,640
Speaker 2: I have asked him. I have genuinely emailed soft Banks

150
00:07:03,680 --> 00:07:07,520
Speaker 2: PR and I've emailed Anthropics PR for Dario and han't

151
00:07:07,560 --> 00:07:09,680
Speaker 2: heard back from either. I assume there's an email issue

152
00:07:10,000 --> 00:07:12,760
Speaker 2: because there's no other reason they wouldn't come on this show.

153
00:07:13,400 --> 00:07:16,640
Speaker 2: It's just it now. I what's fun to watch though

154
00:07:16,680 --> 00:07:18,920
Speaker 2: is considering how many times we've had like AI bubble

155
00:07:18,960 --> 00:07:22,040
Speaker 2: conversations on here, everyone seems to be kind of waking

156
00:07:22,120 --> 00:07:24,559
Speaker 2: up to it. It's kind of fun. Must be clear,

157
00:07:25,400 --> 00:07:28,280
Speaker 2: everyone here was early on this. I was new, actually

158
00:07:28,400 --> 00:07:30,680
Speaker 2: very early. You've been early on every day as early

159
00:07:30,680 --> 00:07:32,920
Speaker 2: as you No, I give you credit. You're one of

160
00:07:33,000 --> 00:07:35,360
Speaker 2: like the three people who, when the metaverse was happening,

161
00:07:36,080 --> 00:07:38,680
Speaker 2: was actually calling it out. So it's good. It's good

162
00:07:38,680 --> 00:07:41,920
Speaker 2: to see in here, but also insane that it's still going.

163
00:07:42,160 --> 00:07:44,320
Speaker 2: That's what I don't like. This open a one hundred

164
00:07:44,320 --> 00:07:47,280
Speaker 2: and fifteen billion spurned within by twenty twenty eight, I

165
00:07:47,280 --> 00:07:50,320
Speaker 2: think it is. I don't understand how people keep publishing

166
00:07:50,360 --> 00:07:52,200
Speaker 2: this and being like and that will happen.

167
00:07:52,360 --> 00:07:54,760
Speaker 3: I will say, to my relief, the metaverse went away

168
00:07:54,880 --> 00:07:57,640
Speaker 3: quickly because when it first was announced and I wrote

169
00:07:57,640 --> 00:07:59,280
Speaker 3: a piece just being like what.

170
00:07:59,600 --> 00:08:00,760
Speaker 2: Yeah, what is what?

171
00:08:01,840 --> 00:08:04,440
Speaker 3: And then like everything about it seemed really dumb in

172
00:08:04,480 --> 00:08:07,679
Speaker 3: every iteration coming after, and so I was like, oh, okay,

173
00:08:07,920 --> 00:08:10,400
Speaker 3: I wasn't crazy, but with AI and I know you

174
00:08:10,400 --> 00:08:14,600
Speaker 3: can relate to this, I feel crazy because the lack

175
00:08:14,680 --> 00:08:18,000
Speaker 3: of utility is still there and like the absurdity of

176
00:08:18,040 --> 00:08:22,200
Speaker 3: the investment is still there, and it does seem like that.

177
00:08:22,200 --> 00:08:24,160
Speaker 3: That's why I wrote about the vibe shitt like it

178
00:08:24,240 --> 00:08:27,360
Speaker 3: has been like going in Edzitron's favor in the last

179
00:08:27,360 --> 00:08:27,960
Speaker 3: few weeks.

180
00:08:28,560 --> 00:08:32,120
Speaker 2: Fucking vindication finally. But it's it's funny as well because

181
00:08:32,280 --> 00:08:36,079
Speaker 2: even with them, this open Ai story comes out and

182
00:08:36,120 --> 00:08:37,800
Speaker 2: people are still like, yeah, they're gonna burn one hundred

183
00:08:37,800 --> 00:08:41,000
Speaker 2: and fifteen billion dollars. People reported about a month ago

184
00:08:41,040 --> 00:08:43,280
Speaker 2: that they're going to spend thirty billion dollars a year

185
00:08:43,280 --> 00:08:48,480
Speaker 2: on Oracles starting twenty twenty eight. How how how? I

186
00:08:48,640 --> 00:08:51,600
Speaker 2: just That's what I don't understand. How these articles keep

187
00:08:51,600 --> 00:08:53,560
Speaker 2: coming out. And I understand that reporters have to report

188
00:08:53,760 --> 00:08:55,839
Speaker 2: the news and things they have discovered. I get it,

189
00:08:56,000 --> 00:08:57,600
Speaker 2: But can no one just be like and no one

190
00:08:57,679 --> 00:08:59,839
Speaker 2: has any idea where this money's coming from. No one,

191
00:09:00,240 --> 00:09:04,640
Speaker 2: thirty billion dollars a fucking year from Oracle for service

192
00:09:04,679 --> 00:09:06,200
Speaker 2: that are yet to be built, and all of this.

193
00:09:06,400 --> 00:09:08,240
Speaker 2: They don't need to worry about the fact that the

194
00:09:08,280 --> 00:09:11,079
Speaker 2: servers aren't actually built. In Applene Texas is not finished,

195
00:09:11,080 --> 00:09:13,760
Speaker 2: and the money doesn't exist, and it isn't obvious how

196
00:09:13,840 --> 00:09:16,040
Speaker 2: Oracle affords it they will have to take on debt

197
00:09:16,080 --> 00:09:19,120
Speaker 2: to build this, and Crusoe and primary digital Infrastructure have

198
00:09:19,160 --> 00:09:21,520
Speaker 2: already done that. And I mean, other than that, it

199
00:09:21,559 --> 00:09:24,439
Speaker 2: could happen any day. I just wonder if the media

200
00:09:24,559 --> 00:09:26,319
Speaker 2: is actually not prepared for this. And I don't mean

201
00:09:26,320 --> 00:09:29,400
Speaker 2: in a conspiratle way, I mean just in a is

202
00:09:29,440 --> 00:09:32,760
Speaker 2: the media actually set up for companies just lying or

203
00:09:32,880 --> 00:09:34,200
Speaker 2: just projecting.

204
00:09:34,559 --> 00:09:37,880
Speaker 4: I mean, I feel like it reminds me of kind

205
00:09:37,880 --> 00:09:40,480
Speaker 4: of the relationship that maybe our critics or you know,

206
00:09:40,559 --> 00:09:43,520
Speaker 4: our coverage might have for the medium, where there's not

207
00:09:43,600 --> 00:09:49,240
Speaker 4: an inherent antagonism or skepticism of claims that are being offered,

208
00:09:49,920 --> 00:09:54,320
Speaker 4: an assumption of good faith that continually gets betrayed or

209
00:09:54,400 --> 00:09:57,400
Speaker 4: you know, punished, but gets carried on with over and

210
00:09:57,440 --> 00:10:00,520
Speaker 4: over and over again. I feel like the AI bubble

211
00:10:01,360 --> 00:10:03,880
Speaker 4: discussions that we're seeing. Part of me also feels like

212
00:10:04,080 --> 00:10:06,720
Speaker 4: they are going to disappear the second we start to

213
00:10:06,800 --> 00:10:10,439
Speaker 4: see some of the firms maybe announce some favorable metrics,

214
00:10:10,480 --> 00:10:13,319
Speaker 4: even though, as we've been talking about for a long time, right,

215
00:10:13,360 --> 00:10:16,080
Speaker 4: revenues are not there, profits are not there now, the

216
00:10:16,120 --> 00:10:20,920
Speaker 4: burn is only increasing, right, and there's no way forward

217
00:10:22,000 --> 00:10:23,880
Speaker 4: in the short term. Right, you know that I can

218
00:10:23,920 --> 00:10:26,640
Speaker 4: think of where these companies start to actually do the

219
00:10:26,640 --> 00:10:28,120
Speaker 4: things that they're claiming they're going to be doing with

220
00:10:28,240 --> 00:10:31,440
Speaker 4: transforming the world. But I can see a scenario where

221
00:10:31,679 --> 00:10:34,360
Speaker 4: you know, someone has a favorable quarter for adoption even

222
00:10:34,400 --> 00:10:37,000
Speaker 4: though you know, we just I just saw yesterday Apollo

223
00:10:37,040 --> 00:10:40,640
Speaker 4: Global Management was talking about large firms are actually scaling

224
00:10:40,679 --> 00:10:45,000
Speaker 4: back AI adoption, which already wasn't even providing returns and

225
00:10:45,080 --> 00:10:46,800
Speaker 4: was hurting productivity in the first place.

226
00:10:46,880 --> 00:10:47,080
Speaker 1: Right.

227
00:10:47,160 --> 00:10:49,440
Speaker 2: It's and it's really weird as well, because there was

228
00:10:49,440 --> 00:10:52,040
Speaker 2: the MIT study where it's like they said, ninety five

229
00:10:52,080 --> 00:10:55,480
Speaker 2: percent of generative AI integrations don't have any return on investment.

230
00:10:55,640 --> 00:10:58,360
Speaker 2: There are some people critiquing that number, but something that

231
00:10:58,360 --> 00:10:59,840
Speaker 2: comes out of that study that I really like was

232
00:10:59,880 --> 00:11:03,120
Speaker 2: it was saying that enterprise adoption is high, but the

233
00:11:03,160 --> 00:11:06,480
Speaker 2: actual transformation is low because this shit doesn't work. And

234
00:11:06,600 --> 00:11:10,400
Speaker 2: it's it's so strange, and I think that the only

235
00:11:10,480 --> 00:11:13,080
Speaker 2: reason that things won't immediately unrebel with a good quarter

236
00:11:13,240 --> 00:11:16,559
Speaker 2: is because the media has chosen to follow a direction.

237
00:11:16,679 --> 00:11:20,120
Speaker 2: Now you've got the when you've got the Atlantic, the

238
00:11:20,200 --> 00:11:23,839
Speaker 2: goddamn Atlantic publishing a story saying yeah, it turns out

239
00:11:23,880 --> 00:11:26,080
Speaker 2: that AI isn't really doing much, but it's holding up

240
00:11:26,120 --> 00:11:29,480
Speaker 2: our economy. It's like, holy shit, the Atlantic's willing to

241
00:11:29,480 --> 00:11:32,760
Speaker 2: admit something that happened happened. I didn't even know this

242
00:11:32,880 --> 00:11:35,640
Speaker 2: was in there. I thought that they just wrote up

243
00:11:35,640 --> 00:11:37,559
Speaker 2: whatever was email to them by and dohold.

244
00:11:37,320 --> 00:11:39,080
Speaker 4: Your bread, they just published Mike Solana.

245
00:11:40,200 --> 00:11:43,520
Speaker 2: Yeah, actually I retract all my statements to But it's

246
00:11:43,760 --> 00:11:45,800
Speaker 2: I think with the media.

247
00:11:46,120 --> 00:11:49,720
Speaker 3: I see it in political reporting and business reporting in tech.

248
00:11:50,559 --> 00:11:55,640
Speaker 3: There's a deference to authority that I think American media,

249
00:11:55,720 --> 00:11:59,800
Speaker 3: but all media, have an issue with. And I think

250
00:11:59,840 --> 00:12:02,880
Speaker 3: that sort of speaks to the underlying economics of being

251
00:12:02,960 --> 00:12:06,840
Speaker 3: in media right now, where there's a general chill both

252
00:12:06,880 --> 00:12:12,240
Speaker 3: economically and politically. Reporters are worried about their bylines being

253
00:12:12,320 --> 00:12:14,679
Speaker 3: out there and getting stuff wrong. And I'm not saying

254
00:12:14,679 --> 00:12:16,640
Speaker 3: that that's an excuse, but I do think that is

255
00:12:16,679 --> 00:12:20,319
Speaker 3: an institutional mindset that has taken route, especially in the

256
00:12:20,400 --> 00:12:23,200
Speaker 3: last ten years, that's just become like really hard to

257
00:12:23,200 --> 00:12:26,400
Speaker 3: be a journalist and to do it right. But you

258
00:12:26,520 --> 00:12:29,199
Speaker 3: are starting to see that MIT report was so important

259
00:12:29,200 --> 00:12:34,040
Speaker 3: because it caused people on Wall Street authority figures to

260
00:12:34,480 --> 00:12:37,439
Speaker 3: say hmm, I don't know about this, And then that

261
00:12:37,480 --> 00:12:41,040
Speaker 3: got a lot of mainstream financial media to kind of

262
00:12:41,080 --> 00:12:44,680
Speaker 3: like do that questioning headline about AI that they maybe

263
00:12:44,720 --> 00:12:46,160
Speaker 3: wouldn't have done six months ago.

264
00:12:46,440 --> 00:12:48,440
Speaker 2: I do like that there's a rumor about the next

265
00:12:48,440 --> 00:12:52,080
Speaker 2: deep Seek model doing agents. They're not going to, but

266
00:12:52,200 --> 00:12:54,120
Speaker 2: even just like if that comes out and they even

267
00:12:54,120 --> 00:12:56,080
Speaker 2: claim they can, I think that might have a market

268
00:12:56,080 --> 00:12:58,439
Speaker 2: panic just because they'll be like, ah, China, I.

269
00:12:58,360 --> 00:13:00,800
Speaker 4: Mean yeah, I mean that's I. I actually think there's

270
00:13:00,840 --> 00:13:03,000
Speaker 4: something to that, right, because you know, we did agents here,

271
00:13:03,360 --> 00:13:06,760
Speaker 4: didn't do shed for sales powers really the market did

272
00:13:06,760 --> 00:13:07,439
Speaker 4: the market even?

273
00:13:07,480 --> 00:13:11,920
Speaker 2: Really? Actually and a wonderful story from this information open

274
00:13:11,960 --> 00:13:15,559
Speaker 2: AI in their in their whole projections through twenty twenty nine,

275
00:13:15,559 --> 00:13:17,600
Speaker 2: they've reduced the amount of money of the micromagents by

276
00:13:17,600 --> 00:13:23,880
Speaker 2: twenty six billion dollars. Why I want anyway? So, but yeah,

277
00:13:23,920 --> 00:13:27,200
Speaker 2: it's like that this deep Seek thing could inspire people

278
00:13:27,240 --> 00:13:30,319
Speaker 2: to get scared horribly because no one actually believes that

279
00:13:30,360 --> 00:13:33,400
Speaker 2: agents exist. Looks they don't, but they think they do,

280
00:13:33,440 --> 00:13:36,120
Speaker 2: but they will, but they won't. It's this I don't

281
00:13:36,120 --> 00:13:38,240
Speaker 2: think I've ever actually seen anything like this in tech.

282
00:13:38,240 --> 00:13:40,760
Speaker 2: I'm gonna be honest, it's worse than crypto, even worse

283
00:13:40,800 --> 00:13:42,720
Speaker 2: than just the general generative AI. I think is this

284
00:13:42,800 --> 00:13:45,719
Speaker 2: concept of agents though, because I saw some fucking thing

285
00:13:45,800 --> 00:13:49,440
Speaker 2: about some one political blog saying that Donald Trump would

286
00:13:49,440 --> 00:13:52,080
Speaker 2: do some sort of act he would do the I

287
00:13:52,120 --> 00:13:53,840
Speaker 2: forget the exact thing, but it would be an act

288
00:13:53,880 --> 00:13:57,839
Speaker 2: that would make AI copyright holders just handshit over to

289
00:13:57,880 --> 00:14:00,800
Speaker 2: AI due to us needing to beat China. And it

290
00:14:00,880 --> 00:14:04,119
Speaker 2: mentioned with you know, like, yeah, the growing agency capabilities

291
00:14:04,160 --> 00:14:06,280
Speaker 2: of AI. It's just what the fuck is that. I've

292
00:14:06,280 --> 00:14:08,199
Speaker 2: never seen a tech thing in my life now that

293
00:14:08,320 --> 00:14:10,600
Speaker 2: has not existed like this, and people talk about it

294
00:14:10,640 --> 00:14:11,200
Speaker 2: like it's real.

295
00:14:11,760 --> 00:14:14,280
Speaker 4: And I think also it's interesting to see the more

296
00:14:14,440 --> 00:14:18,080
Speaker 4: it doesn't manifest, the more some of the recommendations to

297
00:14:18,160 --> 00:14:21,360
Speaker 4: make them happen just sound more and more like these

298
00:14:21,360 --> 00:14:24,440
Speaker 4: are also things that might somehow bend the cost curve

299
00:14:24,440 --> 00:14:27,560
Speaker 4: in our direction. Maybe we should make the Internet unusable

300
00:14:27,600 --> 00:14:30,360
Speaker 4: to anything other than some of these programs. Yeah right,

301
00:14:30,840 --> 00:14:35,160
Speaker 4: And you know, I'm curious which one is going to

302
00:14:35,240 --> 00:14:38,640
Speaker 4: give out first, Like the really savvy ability a lot

303
00:14:38,640 --> 00:14:43,040
Speaker 4: of these firms have had in spinning, sputtering or you know,

304
00:14:43,320 --> 00:14:47,920
Speaker 4: spinning you know, a crisis that might drain the markets

305
00:14:48,000 --> 00:14:50,560
Speaker 4: or deter investors and to oh, actually we just need

306
00:14:50,600 --> 00:14:53,320
Speaker 4: even more capital, similar to what they did with deep Seek.

307
00:14:53,640 --> 00:14:57,120
Speaker 4: The solution is even more compute intensive models. You know,

308
00:14:57,160 --> 00:14:58,880
Speaker 4: whether they're going to be able to do that faster

309
00:14:59,000 --> 00:15:02,120
Speaker 4: than people wising up and saying, you know, maybe we

310
00:15:02,160 --> 00:15:05,960
Speaker 4: shouldn't misallocate trillions of dollars a capital over the next

311
00:15:05,960 --> 00:15:06,840
Speaker 4: few years towards this.

312
00:15:07,000 --> 00:15:09,880
Speaker 2: But this is the thing, though, I don't think there's

313
00:15:09,920 --> 00:15:13,400
Speaker 2: anything stopping this, because the suggested thing was the okay,

314
00:15:13,440 --> 00:15:16,040
Speaker 2: so all of these modial companies can steal from everyone,

315
00:15:16,160 --> 00:15:18,880
Speaker 2: which is what happened already even with this anthropic settlement.

316
00:15:18,960 --> 00:15:21,400
Speaker 2: It's not it's great people are getting three thousand dollars.

317
00:15:21,440 --> 00:15:22,880
Speaker 2: I love that, but.

318
00:15:22,680 --> 00:15:24,840
Speaker 4: It's also what the companies are offering that's similar to

319
00:15:24,840 --> 00:15:27,600
Speaker 4: what the your payout would have been if you said, yeah, right,

320
00:15:28,640 --> 00:15:31,040
Speaker 4: I didn't know that. I think like some publishers were offering,

321
00:15:31,360 --> 00:15:33,800
Speaker 4: or you know, trying to ask authors, Hey, if we

322
00:15:33,880 --> 00:15:36,440
Speaker 4: pay you this, would you allow your book to be

323
00:15:36,520 --> 00:15:39,320
Speaker 4: trained on would you allow us, you know, be put

324
00:15:39,360 --> 00:15:42,640
Speaker 4: into a data set. The payment that you're getting from

325
00:15:42,640 --> 00:15:45,120
Speaker 4: the settlement feels or reminds me of the amount of

326
00:15:45,120 --> 00:15:50,720
Speaker 4: money that people are being offered and those sorts of deals.

327
00:15:58,600 --> 00:16:02,480
Speaker 2: I think the thing is, Okay, they already steal everything. Well, okay,

328
00:16:02,520 --> 00:16:04,120
Speaker 2: we need to give them as much money as possible.

329
00:16:04,120 --> 00:16:06,320
Speaker 2: We've already done that. Are we just going to do

330
00:16:06,360 --> 00:16:08,120
Speaker 2: this forever? Because even if we do this for rever,

331
00:16:08,200 --> 00:16:11,040
Speaker 2: nothing's going to change, even if I'm completely wrong and

332
00:16:11,120 --> 00:16:15,360
Speaker 2: open ai keeps going another five years. Okay, so we're

333
00:16:15,400 --> 00:16:17,400
Speaker 2: just going to annihilate one hundred and fifteen billion dollars.

334
00:16:17,560 --> 00:16:19,840
Speaker 2: There is no there are no more things here like

335
00:16:20,320 --> 00:16:24,480
Speaker 2: they're projections open ays projections from the information their chart.

336
00:16:24,520 --> 00:16:26,200
Speaker 2: I don't even need to show you this because this

337
00:16:26,360 --> 00:16:29,680
Speaker 2: is just fan fiction at this point. There is in

338
00:16:29,720 --> 00:16:32,120
Speaker 2: twenty twenty six starting there is this growth of this

339
00:16:32,200 --> 00:16:36,560
Speaker 2: orange thing that is other revenue. Who knows what it is?

340
00:16:37,200 --> 00:16:39,360
Speaker 2: I don't know. Open Ai doesn't seem to And that's

341
00:16:39,360 --> 00:16:41,200
Speaker 2: really important because they're going to make what looks like

342
00:16:41,280 --> 00:16:45,480
Speaker 2: several billion dollars from this next year. What the fuck

343
00:16:45,600 --> 00:16:48,280
Speaker 2: is going Every time I look at this company, I

344
00:16:48,280 --> 00:16:50,520
Speaker 2: feel a little more insane because they've now lowered their

345
00:16:50,560 --> 00:16:53,600
Speaker 2: expectations of selling their access to their models by five

346
00:16:53,600 --> 00:16:56,240
Speaker 2: billion dollars over the next few years. What even is

347
00:16:56,280 --> 00:16:58,720
Speaker 2: open ai at this point? Is it just a wrap?

348
00:16:58,760 --> 00:17:01,600
Speaker 2: Have they become a wrapper company of their own models?

349
00:17:01,880 --> 00:17:05,359
Speaker 2: They're no better than Cursor. It's just it's so weird,

350
00:17:05,480 --> 00:17:07,359
Speaker 2: and I realize them kind of going in circles at

351
00:17:07,359 --> 00:17:12,520
Speaker 2: this point, but it's I even the metaverse, even Crypto,

352
00:17:12,760 --> 00:17:15,800
Speaker 2: even Crypto functioned. It was bad. It's still bad. It

353
00:17:15,880 --> 00:17:19,000
Speaker 2: is bad cloud software, but it still did the thing. Hey,

354
00:17:19,040 --> 00:17:22,480
Speaker 2: I doesn't even seem to be doing it, and they

355
00:17:22,480 --> 00:17:24,480
Speaker 2: need more money to prove that it can't do it.

356
00:17:24,520 --> 00:17:27,200
Speaker 2: And actually they don't have enough right now, but they're

357
00:17:27,240 --> 00:17:30,160
Speaker 2: going to need even more. I don't even know how

358
00:17:30,200 --> 00:17:32,440
Speaker 2: people are still taking this seriously because on top of that,

359
00:17:32,840 --> 00:17:36,879
Speaker 2: did you hear about the Microsoft negotiations over the nonprofit.

360
00:17:36,800 --> 00:17:37,520
Speaker 4: I've been hearing.

361
00:17:37,760 --> 00:17:41,199
Speaker 2: Well, they're delaying it to next year. They have to.

362
00:17:41,320 --> 00:17:42,760
Speaker 2: They need to convert by the end of the year

363
00:17:42,760 --> 00:17:45,480
Speaker 2: otherwise SoftBank comes their round in half. And everyone's just like, yeah,

364
00:17:45,480 --> 00:17:48,199
Speaker 2: I'll be fine, mate, it'll work it out. What the

365
00:17:48,240 --> 00:17:50,919
Speaker 2: fuck have you have either of you ever? I know

366
00:17:51,240 --> 00:17:53,400
Speaker 2: ed you covered Uber a lot. I don't even think

367
00:17:53,400 --> 00:17:55,000
Speaker 2: the economics match with that either.

368
00:17:55,200 --> 00:17:58,400
Speaker 4: No. I mean, you know, it's interesting because I think

369
00:17:59,480 --> 00:18:05,119
Speaker 4: Uber's strategy, central strategy from the beginning was, you know,

370
00:18:05,160 --> 00:18:08,040
Speaker 4: we have a few existing playbooks that we need to reference,

371
00:18:08,400 --> 00:18:12,199
Speaker 4: you know, the coke deregulation of the taxi industry in

372
00:18:12,280 --> 00:18:16,960
Speaker 4: the nineties, as well as Regula deregulatory campaigns that they led,

373
00:18:18,440 --> 00:18:22,440
Speaker 4: you know, in Seattle and historic campaigns in San Francisco.

374
00:18:22,960 --> 00:18:26,399
Speaker 4: There's a lot that we can reference. And if we

375
00:18:26,440 --> 00:18:30,480
Speaker 4: can figure out a way to bootstrap ourselves onto the

376
00:18:30,520 --> 00:18:36,400
Speaker 4: model and onto those previous histories of deregulation while delaying

377
00:18:36,680 --> 00:18:39,920
Speaker 4: scrutiny long enough our economics to actually, you know, get

378
00:18:39,960 --> 00:18:41,639
Speaker 4: to a profitable place, we'll get there, which is what

379
00:18:41,680 --> 00:18:44,480
Speaker 4: they did right. But even then, I mean, I feel

380
00:18:44,560 --> 00:18:47,280
Speaker 4: like from the beginning, as much as I hated a

381
00:18:47,320 --> 00:18:50,000
Speaker 4: lot of the coverage of Uber for years, the people

382
00:18:50,000 --> 00:18:53,199
Speaker 4: who were always correct about it, We're like the labor

383
00:18:53,920 --> 00:18:56,200
Speaker 4: reporters who had actually, you know, if you spend time

384
00:18:56,200 --> 00:18:58,480
Speaker 4: talking to the drivers, that will lead you to be

385
00:18:58,520 --> 00:19:00,600
Speaker 4: a little bit more interested in the you know, what

386
00:19:00,760 --> 00:19:04,399
Speaker 4: can justify the suffering behind this, and then you almost

387
00:19:04,440 --> 00:19:08,440
Speaker 4: always will see that there's no act. At that point,

388
00:19:08,480 --> 00:19:10,520
Speaker 4: there were no way the unit economics worked unless you

389
00:19:10,520 --> 00:19:11,320
Speaker 4: subsidize everything.

390
00:19:11,400 --> 00:19:11,560
Speaker 1: Yeah.

391
00:19:11,560 --> 00:19:11,960
Speaker 2: I feel like.

392
00:19:11,960 --> 00:19:14,960
Speaker 4: Similarly, there's something going on with with artificial intelligence firms

393
00:19:14,960 --> 00:19:19,080
Speaker 4: and the global AI value chain where if you start

394
00:19:19,160 --> 00:19:22,080
Speaker 4: with a labor analysis and you look at you know,

395
00:19:22,119 --> 00:19:25,359
Speaker 4: invisible workers or ghost workers that are integitical.

396
00:19:24,920 --> 00:19:27,520
Speaker 2: The Kenyan people training the models m M yeah.

397
00:19:27,440 --> 00:19:30,720
Speaker 4: Or labeling or you know, an any any laborer that's

398
00:19:30,840 --> 00:19:33,639
Speaker 4: out of sight, out of mind. You know, then the

399
00:19:34,080 --> 00:19:36,520
Speaker 4: you know, starting there and going up, it becomes hard

400
00:19:36,560 --> 00:19:40,239
Speaker 4: to ask, Okay, how are we supposed to allocate all

401
00:19:40,280 --> 00:19:44,040
Speaker 4: this capital towards a model that, as is right now

402
00:19:44,280 --> 00:19:46,200
Speaker 4: is cutting all these corners for costs and it's still

403
00:19:46,240 --> 00:19:48,399
Speaker 4: burning tens and tens and tens of billions of dollars

404
00:19:48,400 --> 00:19:50,040
Speaker 4: and it's asking for trillions of dollars more.

405
00:19:51,880 --> 00:19:52,439
Speaker 2: But I don't know.

406
00:19:52,480 --> 00:19:54,600
Speaker 4: I mean, part of my fear is that they are

407
00:19:54,640 --> 00:19:57,199
Speaker 4: successful in the way that Uber was, where if you

408
00:19:57,240 --> 00:20:00,520
Speaker 4: get enough buy in from military industrial comp well, you know,

409
00:20:00,600 --> 00:20:03,240
Speaker 4: for an AI's case, if you got enough buying from

410
00:20:03,359 --> 00:20:06,240
Speaker 4: military doustrial complex, if you get enough buy in from

411
00:20:06,440 --> 00:20:09,720
Speaker 4: UH you know, social programs and interfacing them and helping

412
00:20:09,760 --> 00:20:12,040
Speaker 4: cut cutting them or redirect traffic through them, if you

413
00:20:12,040 --> 00:20:14,040
Speaker 4: get enough buy in from other tech firms, if you

414
00:20:14,080 --> 00:20:16,600
Speaker 4: get rents from other startups they need to use and

415
00:20:16,680 --> 00:20:19,320
Speaker 4: get access to your product. And also if you graft

416
00:20:19,359 --> 00:20:23,199
Speaker 4: yourself onto everybody's daily interactions, daily lives, the way they

417
00:20:23,240 --> 00:20:25,480
Speaker 4: interface with the Internet, can you actually make it work?

418
00:20:25,560 --> 00:20:27,520
Speaker 4: Which is also another way to say, like, what if

419
00:20:27,560 --> 00:20:28,919
Speaker 4: you just become a massive parasite?

420
00:20:29,640 --> 00:20:33,000
Speaker 2: Funny, the funny and grim thing is is AI is

421
00:20:33,000 --> 00:20:36,280
Speaker 2: a terrible parasite. It's not good at it. It doesn't

422
00:20:36,320 --> 00:20:39,480
Speaker 2: like because Uber's success came from being able to graft

423
00:20:39,520 --> 00:20:43,480
Speaker 2: itself on through utility and subsidized pricing. That meant that

424
00:20:43,520 --> 00:20:46,320
Speaker 2: everyone used it and also cabs kind of fucking sucked. Yeah,

425
00:20:46,359 --> 00:20:47,600
Speaker 2: I mean, yeah, they sucked.

426
00:20:47,640 --> 00:20:50,120
Speaker 4: And also transit in most city sucks, and I mean

427
00:20:50,280 --> 00:20:51,200
Speaker 4: it has only gone worse.

428
00:20:51,359 --> 00:20:55,520
Speaker 2: The inherent colonialism of most technology applies very good from

429
00:20:55,600 --> 00:20:58,080
Speaker 2: Karen Howe. Of course, Empire of Ai. Empires of Ai

430
00:20:58,200 --> 00:21:01,240
Speaker 2: did NEP's a great book, but in this one, it's

431
00:21:01,320 --> 00:21:03,800
Speaker 2: kind of shit colonialism as well, because they don't even

432
00:21:03,880 --> 00:21:06,520
Speaker 2: they haven't found a way to actually exploit in a

433
00:21:06,520 --> 00:21:08,560
Speaker 2: way that's profitable. They haven't found a way to use

434
00:21:08,640 --> 00:21:11,160
Speaker 2: human beings because the fundamental thing they want to do.

435
00:21:11,440 --> 00:21:13,840
Speaker 2: It's kind of like if Uber sometimes you got in

436
00:21:13,880 --> 00:21:15,560
Speaker 2: the car and you got out at the wrong place,

437
00:21:15,600 --> 00:21:18,399
Speaker 2: and I don't mean like in a different country, or

438
00:21:18,440 --> 00:21:21,320
Speaker 2: you got into the car and it just exploded sometimes

439
00:21:21,400 --> 00:21:23,840
Speaker 2: And I sound like I'm joking, but it really is

440
00:21:24,040 --> 00:21:27,280
Speaker 2: that bad. And on top of it, it's not replacing labor.

441
00:21:27,440 --> 00:21:29,480
Speaker 2: And it's also not the kind of tech that can

442
00:21:29,560 --> 00:21:33,360
Speaker 2: replace labor. So I it's my grand theory that they're

443
00:21:33,400 --> 00:21:35,560
Speaker 2: just playing the hits. They just they're trying in the

444
00:21:35,560 --> 00:21:38,560
Speaker 2: same way that you just eloquently put Uber played the

445
00:21:38,640 --> 00:21:40,800
Speaker 2: hits of his how we did deregulation, his how we

446
00:21:40,840 --> 00:21:42,800
Speaker 2: did growth, and this is how software grew in the past.

447
00:21:43,119 --> 00:21:44,760
Speaker 2: I think that AI is trying to do the same thing,

448
00:21:45,000 --> 00:21:47,160
Speaker 2: and it's bad at it. It's like watching a new

449
00:21:47,240 --> 00:21:50,240
Speaker 2: class of dipshit try and do what the more evil

450
00:21:50,240 --> 00:21:54,200
Speaker 2: dipshits of the past did and fail. In fact, these

451
00:21:54,240 --> 00:21:56,320
Speaker 2: lobbying groups lobbying for AI. I hear a lot of

452
00:21:56,320 --> 00:21:58,720
Speaker 2: people saying, oh, they're lobbying their lobby It's like what for.

453
00:21:59,240 --> 00:22:01,960
Speaker 2: Oh no, they're gonna build data centers everywhere. They already do.

454
00:22:01,960 --> 00:22:03,960
Speaker 2: They're gonna steal from it. They already do. They're trying

455
00:22:03,960 --> 00:22:07,120
Speaker 2: to replace, They're already trying everything that they that they

456
00:22:07,240 --> 00:22:10,879
Speaker 2: everyone's scared of. They already can do. Other than the

457
00:22:11,200 --> 00:22:14,720
Speaker 2: doing part, they can't do any of that. I'm sorry,

458
00:22:14,720 --> 00:22:17,320
Speaker 2: I just I just remembered as well yesterday trying to

459
00:22:17,359 --> 00:22:19,159
Speaker 2: read that because I went on the chat gpt pro

460
00:22:19,240 --> 00:22:22,280
Speaker 2: subreddit because I hate myself, and I was trying to

461
00:22:22,280 --> 00:22:24,720
Speaker 2: find someone who'd used the agent, and every post was

462
00:22:24,720 --> 00:22:27,240
Speaker 2: someone saying, anyone ever used the agent, You've got any tips?

463
00:22:27,240 --> 00:22:31,679
Speaker 2: And everyone's just it doesn't fucking work. It's broken. It's actually,

464
00:22:31,680 --> 00:22:34,359
Speaker 2: here's a good question, asn't ed have you? Can you

465
00:22:34,400 --> 00:22:37,160
Speaker 2: think of a company that's ever released something just completely

466
00:22:37,200 --> 00:22:40,919
Speaker 2: broken before? Because the meta us kind of worked. It

467
00:22:40,960 --> 00:22:43,439
Speaker 2: wasn't what they were promising, but it worked. It was

468
00:22:43,480 --> 00:22:45,400
Speaker 2: a virtual world ish.

469
00:22:45,800 --> 00:22:48,680
Speaker 3: I mean, the pharmaceutical industry has a nice long history

470
00:22:48,880 --> 00:22:54,440
Speaker 3: putting out quasi effective drugs that have all kinds of consequences.

471
00:22:54,520 --> 00:22:57,720
Speaker 3: And I can't remember who skeeted it about this a

472
00:22:57,720 --> 00:22:59,879
Speaker 3: few weeks ago. It was after one of the you know,

473
00:23:00,280 --> 00:23:03,560
Speaker 3: it's become such a genre of the journalism right now,

474
00:23:03,560 --> 00:23:07,200
Speaker 3: about AI, about like this man became delusional and had

475
00:23:07,200 --> 00:23:10,879
Speaker 3: a psychotic episode because of his Chatgypt relationship. It was

476
00:23:10,920 --> 00:23:13,240
Speaker 3: one of those going around and someone's skeeded about how

477
00:23:13,280 --> 00:23:16,119
Speaker 3: if there was if this is coming from a pharmaceutical company,

478
00:23:16,160 --> 00:23:19,560
Speaker 3: it would be recalled immediately. There are real regulations in

479
00:23:19,600 --> 00:23:22,800
Speaker 3: place that could actually claw that back and help save

480
00:23:22,880 --> 00:23:26,240
Speaker 3: people's lives. But there are no regulations around AI, so

481
00:23:26,520 --> 00:23:30,880
Speaker 3: we get Chatgypt gods and spiritual awakenings and all these

482
00:23:31,000 --> 00:23:32,040
Speaker 3: psychotic episodes.

483
00:23:32,080 --> 00:23:34,920
Speaker 2: I do think that that stuff is going to genuinely

484
00:23:34,920 --> 00:23:37,280
Speaker 2: be its downfall though, because right now it's burning more

485
00:23:37,280 --> 00:23:39,840
Speaker 2: money than anyone's ever barned before. And the most common

486
00:23:39,920 --> 00:23:42,280
Speaker 2: use case people can talk about is, yeah, it drove

487
00:23:42,320 --> 00:23:45,879
Speaker 2: that guy insane, that guy went crazy. That there are

488
00:23:45,960 --> 00:23:50,200
Speaker 2: children who's horrifying killing themselves because of this thing. That's

489
00:23:50,200 --> 00:23:53,720
Speaker 2: what it's getting known for. And otherwise it's like, yeah,

490
00:23:53,760 --> 00:23:57,679
Speaker 2: your most annoying friend loves this because that really it's them.

491
00:23:58,280 --> 00:24:00,400
Speaker 2: You love them, but they're like I learned about well listen,

492
00:24:00,480 --> 00:24:04,280
Speaker 2: chat JBT is like you didn't well you know on that.

493
00:24:04,920 --> 00:24:07,760
Speaker 4: Part of also part of my a fear I have

494
00:24:07,920 --> 00:24:11,560
Speaker 4: is I think similar to how when firms were rolling

495
00:24:11,600 --> 00:24:16,359
Speaker 4: out facial recognition surveillance and insisting that we need to

496
00:24:16,359 --> 00:24:21,240
Speaker 4: biometric surveillance to help keep city safe, community safe, products safe.

497
00:24:22,680 --> 00:24:26,160
Speaker 4: One angle that people used to attack it was, well,

498
00:24:26,200 --> 00:24:28,560
Speaker 4: you know, like the racial bias of these things will

499
00:24:28,560 --> 00:24:34,520
Speaker 4: allow them to you know, miss identify black or brown

500
00:24:34,560 --> 00:24:37,000
Speaker 4: people more often than not, and they might get arrested,

501
00:24:37,000 --> 00:24:39,000
Speaker 4: they might get targeted by the police, won't wear another

502
00:24:40,600 --> 00:24:42,000
Speaker 4: and that is why we should get rid of the

503
00:24:42,000 --> 00:24:44,520
Speaker 4: technology versus we should get rid of the technology. And

504
00:24:44,520 --> 00:24:47,520
Speaker 4: I think, like, I'm curious, you know, how it's going

505
00:24:47,600 --> 00:24:52,639
Speaker 4: to go, the concern about it inducing psychosis or inducing suicides,

506
00:24:52,720 --> 00:24:59,119
Speaker 4: because I could see a scenario easily where they patch

507
00:24:59,160 --> 00:25:01,880
Speaker 4: together something that like a fix, and it's not until

508
00:25:01,960 --> 00:25:04,800
Speaker 4: later a year, two or three after there's after people

509
00:25:04,800 --> 00:25:10,360
Speaker 4: are much more defended that other harms come to the foreground,

510
00:25:10,640 --> 00:25:14,160
Speaker 4: whereas we lose something God, I think not to say,

511
00:25:14,320 --> 00:25:16,480
Speaker 4: you know, not to say or marginalize the fact that

512
00:25:16,520 --> 00:25:19,680
Speaker 4: it has immense social costs or harm here, but it

513
00:25:20,080 --> 00:25:21,960
Speaker 4: does in some ways remind me of the way in

514
00:25:22,000 --> 00:25:26,480
Speaker 4: which that debate over facial recognition went and then they

515
00:25:26,600 --> 00:25:29,480
Speaker 4: you know, they solved, you know, with quotation marks the

516
00:25:29,560 --> 00:25:32,600
Speaker 4: racial bias problem, and now people have more or less

517
00:25:32,640 --> 00:25:36,719
Speaker 4: accepted that facial recognition is okay, actually long this it's

518
00:25:36,800 --> 00:25:37,760
Speaker 4: not racist.

519
00:25:37,440 --> 00:25:40,919
Speaker 2: And that's the thing. It's people say, this is a

520
00:25:40,920 --> 00:25:45,680
Speaker 2: white bloke. But it's like people really underplay how endemic

521
00:25:45,760 --> 00:25:48,880
Speaker 2: that racism is within all algorithms, you know, Compass, which

522
00:25:48,920 --> 00:25:52,160
Speaker 2: is like this very very very old algorithm for being

523
00:25:52,240 --> 00:25:55,600
Speaker 2: like it's basically minority report, both in the reference to

524
00:25:55,600 --> 00:25:58,120
Speaker 2: the thing and it reports and minorities in that it says, yeah,

525
00:25:58,160 --> 00:26:01,160
Speaker 2: this this person will likely have again, and that should

526
00:26:01,200 --> 00:26:03,960
Speaker 2: be and it isn't a unilateral the judge has to

527
00:26:03,960 --> 00:26:06,960
Speaker 2: take it. But what a surprise. It's often used to

528
00:26:07,680 --> 00:26:13,160
Speaker 2: black people to the jail systems because it's heavily biased

529
00:26:13,160 --> 00:26:17,360
Speaker 2: against them. And yeah, I somewhat fear lllm's doing similar

530
00:26:18,560 --> 00:26:21,240
Speaker 2: they're probably already doing. And I think that every algorithmic

531
00:26:21,320 --> 00:26:24,240
Speaker 2: system is inherently racist. There's not enough people running them

532
00:26:24,240 --> 00:26:26,920
Speaker 2: who actually fucking try. It's inherently biased against a woman.

533
00:26:27,320 --> 00:26:29,159
Speaker 2: I think there's also I wish I had this in

534
00:26:29,160 --> 00:26:30,959
Speaker 2: front of you. But there's also something about how like

535
00:26:31,880 --> 00:26:34,000
Speaker 2: more there were more fans of Generative AI who were

536
00:26:34,040 --> 00:26:36,680
Speaker 2: male than female, or it's just but.

537
00:26:36,640 --> 00:26:39,080
Speaker 4: Do you think it's possible that you know, they'll try

538
00:26:39,200 --> 00:26:42,439
Speaker 4: to say, oh, we can solve for the psychosist problem,

539
00:26:42,640 --> 00:26:46,200
Speaker 4: and then that will undermine a large The problem is

540
00:26:46,480 --> 00:26:48,080
Speaker 4: of the criticism.

541
00:26:48,240 --> 00:26:50,359
Speaker 2: How do you solve it? Because it is probably a

542
00:26:50,400 --> 00:26:53,640
Speaker 2: small it is probably a small scale problem. We actually

543
00:26:53,680 --> 00:26:55,800
Speaker 2: don't know, and it's not that these companies know or

544
00:26:55,840 --> 00:26:59,879
Speaker 2: will tell us. But nevertheless, each one is so horrifying.

545
00:27:00,960 --> 00:27:04,240
Speaker 2: This is horrib like the story in the Wall Street Journal,

546
00:27:05,040 --> 00:27:07,840
Speaker 2: it's Julie Jargon, another person there who wrote that where

547
00:27:07,880 --> 00:27:10,119
Speaker 2: it was like a murder suicide, an actual son of

548
00:27:10,160 --> 00:27:13,400
Speaker 2: Sam Mortman's situation. Whi's fucking terrifying that this is happening.

549
00:27:13,760 --> 00:27:17,080
Speaker 2: I don't know if you can completely solve that, because well,

550
00:27:17,119 --> 00:27:18,840
Speaker 2: it takes is one popping up again for them to

551
00:27:18,840 --> 00:27:22,680
Speaker 2: go fuck. And it's also not just a chat GPT problem.

552
00:27:22,760 --> 00:27:27,240
Speaker 2: There's this woman on TikTok who she has been saying

553
00:27:27,280 --> 00:27:30,040
Speaker 2: what Claude has been telling her, and it's like, ah,

554
00:27:30,080 --> 00:27:32,159
Speaker 2: this is giving me psychic visions. I think it is.

555
00:27:32,320 --> 00:27:35,719
Speaker 2: It's also an the ultimate grifter tool. It's just it's

556
00:27:35,920 --> 00:27:37,800
Speaker 2: that's why I think it's taken off so much on

557
00:27:37,840 --> 00:27:39,920
Speaker 2: social media as well. It's a tool that naturally fits

558
00:27:39,920 --> 00:27:43,840
Speaker 2: into the grifter's toolbox. I think that I actually have

559
00:27:44,119 --> 00:27:46,119
Speaker 2: similar fears that they will try and find ways to

560
00:27:46,160 --> 00:27:49,639
Speaker 2: handwave away from this if it was the only problem,

561
00:27:50,000 --> 00:27:52,200
Speaker 2: but they have so many problems. There have so many

562
00:27:52,280 --> 00:27:54,399
Speaker 2: problems at this point. But I do also think that

563
00:27:54,440 --> 00:27:58,600
Speaker 2: people need to remember how racism in algorithms is. Fairly,

564
00:27:59,000 --> 00:28:01,240
Speaker 2: it's in all of them. I mean, you remember Microsoft

565
00:28:01,240 --> 00:28:05,639
Speaker 2: Connect which literally couldn't see black people, which was a

566
00:28:05,720 --> 00:28:08,040
Speaker 2: joke in the show Better Off Ted if anyone watched that,

567
00:28:08,160 --> 00:28:12,199
Speaker 2: a great show. It's just it's insane the I mean, sadly,

568
00:28:12,520 --> 00:28:15,080
Speaker 2: it's very obvious why this keeps happening. It's because the

569
00:28:15,119 --> 00:28:17,560
Speaker 2: people are predominantly white and male, and it's just they

570
00:28:17,800 --> 00:28:20,000
Speaker 2: can't really and also you can't really fix this stuff

571
00:28:20,000 --> 00:28:23,160
Speaker 2: without intentionally building the data, which would require and spend

572
00:28:23,200 --> 00:28:24,840
Speaker 2: money on something they don't care about.

573
00:28:25,680 --> 00:28:28,800
Speaker 3: And they don't really understand what they're doing when they

574
00:28:28,840 --> 00:28:31,719
Speaker 3: go in to tweak these models. They don't know how

575
00:28:32,240 --> 00:28:35,480
Speaker 3: overcorrecting or undercorrecting their being so they kind of have

576
00:28:35,560 --> 00:28:38,200
Speaker 3: to just try and then put it out in the

577
00:28:38,240 --> 00:28:41,280
Speaker 3: world and then wait for something bad to happen. It's funny.

578
00:28:41,360 --> 00:28:44,360
Speaker 3: It's not funny, it's extremely sad. In that journal story

579
00:28:44,360 --> 00:28:47,320
Speaker 3: that you referenced, which I read twice because I was

580
00:28:47,480 --> 00:28:51,440
Speaker 3: like horrified, and also the reporting was incredible, it was

581
00:28:51,440 --> 00:28:54,360
Speaker 3: really and they said it as this appears to be

582
00:28:54,400 --> 00:28:59,560
Speaker 3: the first instance of a murder resulting like we've seen

583
00:28:59,640 --> 00:29:03,040
Speaker 3: sue sides, but like this is a murder suicide. And

584
00:29:03,600 --> 00:29:09,400
Speaker 3: when OpenAI responded to the question about did the bot

585
00:29:09,480 --> 00:29:12,320
Speaker 3: ever respond to this guy who is clearly having a

586
00:29:12,320 --> 00:29:15,480
Speaker 3: delusional episode, Hey, you need to talk to a real

587
00:29:15,520 --> 00:29:17,480
Speaker 3: life therapist, you need to go to the hospital, you

588
00:29:17,520 --> 00:29:20,600
Speaker 3: need to seek help. And I think they declined to comment.

589
00:29:20,640 --> 00:29:23,160
Speaker 3: It was like a very evasive maneuver. But like ultimately

590
00:29:23,200 --> 00:29:26,400
Speaker 3: the journal had seen. The one time where the bot

591
00:29:26,600 --> 00:29:29,080
Speaker 3: said please go to the emergency room was when the guy,

592
00:29:29,600 --> 00:29:32,200
Speaker 3: the parent guy who was having paranoid delusions, said I

593
00:29:32,200 --> 00:29:34,600
Speaker 3: think my mom is trying to poison me, and the

594
00:29:34,640 --> 00:29:37,040
Speaker 3: bot said, if you think you've been poisoned, you should

595
00:29:37,040 --> 00:29:39,880
Speaker 3: go to the hospital and get your stomach pumped.

596
00:29:39,960 --> 00:29:42,960
Speaker 2: I also, I agree they don't know how to tweet

597
00:29:43,000 --> 00:29:45,960
Speaker 2: these things, but I must be clear worked in tech

598
00:29:46,040 --> 00:29:49,239
Speaker 2: for a long sixteen seventeen years now, and that's not

599
00:29:49,240 --> 00:29:52,760
Speaker 2: even including my game's journalism work. It is not hard

600
00:29:52,840 --> 00:29:55,880
Speaker 2: for them to just have a just a unilateral thing

601
00:29:55,920 --> 00:29:58,720
Speaker 2: of oh, you're talking like this, I'm going to stop.

602
00:29:59,160 --> 00:30:01,320
Speaker 2: I mean, topic just announced that they have a thing

603
00:30:01,320 --> 00:30:04,440
Speaker 2: that will cut a conversation, which is good. All of

604
00:30:04,480 --> 00:30:07,480
Speaker 2: them should do this, but it's they could if someone

605
00:30:07,600 --> 00:30:11,520
Speaker 2: and the whole thing people I've seen where it should be.

606
00:30:11,920 --> 00:30:14,680
Speaker 2: If you start talking like I'm going to do this,

607
00:30:14,800 --> 00:30:18,920
Speaker 2: I am becoming this, it should say, hey, you sound

608
00:30:18,960 --> 00:30:21,520
Speaker 2: like you're having a paranoid like I'm worried about you.

609
00:30:21,520 --> 00:30:23,320
Speaker 2: You should go and speak with it and just stop

610
00:30:23,400 --> 00:30:26,040
Speaker 2: working with them. People will say, well, the way they

611
00:30:26,080 --> 00:30:29,320
Speaker 2: get around that is by telling the chat GPT window,

612
00:30:29,480 --> 00:30:32,120
Speaker 2: Oh yeah, I'm writing a story. I don't know. Do

613
00:30:32,200 --> 00:30:34,640
Speaker 2: we need them to write a story about that? Do

614
00:30:34,720 --> 00:30:36,720
Speaker 2: we need? Of course? What is the answer, And the

615
00:30:36,760 --> 00:30:39,200
Speaker 2: answer is they don't give a rap fucking no one's mate,

616
00:30:39,320 --> 00:30:42,680
Speaker 2: I really, I genuinely think because it's really the same

617
00:30:42,720 --> 00:30:45,760
Speaker 2: thing with like same thing with social networks as well.

618
00:30:46,040 --> 00:30:49,520
Speaker 2: It's you don't ban every slur the moment someone says it.

619
00:30:49,600 --> 00:30:51,240
Speaker 2: But I don't know you have a thing that says, hey,

620
00:30:51,240 --> 00:30:53,760
Speaker 2: someone said a slur. Maybe take a quick look at

621
00:30:53,760 --> 00:30:56,000
Speaker 2: the slur and you could probably just ban that person,

622
00:30:56,040 --> 00:30:59,280
Speaker 2: because I'm guessing that most uses of the N word

623
00:30:59,280 --> 00:31:03,480
Speaker 2: on social media are not used as in culturally sensitive ways.

624
00:31:03,520 --> 00:31:05,440
Speaker 2: They're probably insanely racist.

625
00:31:06,040 --> 00:31:09,360
Speaker 5: You've just calt them down. It's like, well, we can't.

626
00:31:09,400 --> 00:31:11,880
Speaker 5: It's an issue of free speech. Fuck you, no, it's not.

627
00:31:12,120 --> 00:31:13,880
Speaker 5: It's an issue of free speech when a person can't

628
00:31:13,880 --> 00:31:18,560
Speaker 5: exist online without racism happening to them, right, and these models,

629
00:31:18,600 --> 00:31:20,720
Speaker 5: they could stop them. But I do think there's a

630
00:31:20,760 --> 00:31:23,240
Speaker 5: compelling argument of they really don't know what to do

631
00:31:23,560 --> 00:31:26,080
Speaker 5: that every time they touch it, something else breaks.

632
00:31:26,440 --> 00:31:29,000
Speaker 2: It's honestly, it's kind of the most egregious version of

633
00:31:29,040 --> 00:31:31,680
Speaker 2: the most common software problem, which is coding is really

634
00:31:31,680 --> 00:31:33,680
Speaker 2: fucking annoying, and we don't know how these work.

635
00:31:33,720 --> 00:31:36,440
Speaker 3: Also, and generalor AI is not going to fix your

636
00:31:36,440 --> 00:31:38,680
Speaker 3: coding problems, no matter how many times you tell us,

637
00:31:38,800 --> 00:31:42,320
Speaker 3: Sam Altman that a guy is just going to fix

638
00:31:42,360 --> 00:31:47,400
Speaker 3: everything for us.

639
00:31:59,720 --> 00:32:01,520
Speaker 2: That's actually be my favorite thing to do right now.

640
00:32:01,560 --> 00:32:04,040
Speaker 2: It's go on the R slash cursor, AR slash chat,

641
00:32:04,040 --> 00:32:07,200
Speaker 2: GPT pro R slash claude AI and just looking at

642
00:32:07,200 --> 00:32:10,120
Speaker 2: people complaining, and what they're complaining about is, hey, I

643
00:32:10,200 --> 00:32:13,920
Speaker 2: keep hitting rate limits, Hey I keep it keeps breaking things. Hey,

644
00:32:13,960 --> 00:32:16,280
Speaker 2: it's you get one guy every so often who says

645
00:32:16,800 --> 00:32:18,320
Speaker 2: this has changed my life, and then you see what

646
00:32:18,360 --> 00:32:20,720
Speaker 2: the responses being like, yeah, but fucked up all my

647
00:32:20,920 --> 00:32:24,040
Speaker 2: stuff really badly doesn't really work. And we have an

648
00:32:24,120 --> 00:32:26,960
Speaker 2: upcoming episode with a cult vult Voji about this where

649
00:32:26,960 --> 00:32:29,600
Speaker 2: it's like, the average software engineer is not just writing

650
00:32:29,640 --> 00:32:33,040
Speaker 2: code anyway, and so you this is also I think

651
00:32:33,400 --> 00:32:35,640
Speaker 2: this is actually really this is funny, This is a

652
00:32:35,680 --> 00:32:37,680
Speaker 2: good this is a good one to laugh about. So

653
00:32:37,760 --> 00:32:40,840
Speaker 2: their only real growth market right now is writing code.

654
00:32:41,840 --> 00:32:44,920
Speaker 2: The problem is writing code requires you to use reasoning models.

655
00:32:45,280 --> 00:32:48,959
Speaker 2: Reasoning models inherently burn more tokens. And the way they

656
00:32:49,000 --> 00:32:51,640
Speaker 2: burn tokens is because they're think thinking. They don't really think.

657
00:32:51,680 --> 00:32:53,720
Speaker 2: They look over what a prompt ask for and goes, okay,

658
00:32:53,720 --> 00:32:56,080
Speaker 2: what would be the steps to solve this? With code

659
00:32:56,080 --> 00:32:59,800
Speaker 2: that becomes so so complex, And the more models reason,

660
00:32:59,840 --> 00:33:03,640
Speaker 2: the more they hallucinate. So the very product that they

661
00:33:03,640 --> 00:33:06,120
Speaker 2: are building that is going to save them is also

662
00:33:06,240 --> 00:33:09,160
Speaker 2: the one that is going to burn more compute and

663
00:33:09,560 --> 00:33:12,360
Speaker 2: this is a rumor I've heard from a source that

664
00:33:12,440 --> 00:33:15,280
Speaker 2: like it can take like four to twelve GPUs for

665
00:33:15,400 --> 00:33:19,520
Speaker 2: one person's particular particularly rough coding, like a.

666
00:33:19,560 --> 00:33:21,600
Speaker 4: Refactory that's sustainable.

667
00:33:22,280 --> 00:33:24,840
Speaker 2: And that's for one of the smaller models as well.

668
00:33:25,200 --> 00:33:27,720
Speaker 2: That's for like four many which is a reasoning model.

669
00:33:27,840 --> 00:33:29,400
Speaker 2: It's like, what do you think the big ones are?

670
00:33:29,440 --> 00:33:34,200
Speaker 4: Like in the information they talk about open ai having

671
00:33:34,800 --> 00:33:39,000
Speaker 4: a new eighty billion dollars and eighty billion dollars in

672
00:33:39,120 --> 00:33:43,120
Speaker 4: casts that they spend yes over the next three Yeah,

673
00:33:43,360 --> 00:33:44,160
Speaker 4: it's like it's.

674
00:33:44,000 --> 00:33:45,960
Speaker 2: One hundred and fifteen by twenty twenty nine as well.

675
00:33:46,080 --> 00:33:48,480
Speaker 4: It's a good chunk of this come out of Oh,

676
00:33:48,560 --> 00:33:52,280
Speaker 4: it turns out that computer is incredibly expensive and we

677
00:33:52,440 --> 00:33:54,440
Speaker 4: want to center our business model around it.

678
00:33:54,600 --> 00:33:57,640
Speaker 2: I think it's that, and I think it's just they

679
00:33:57,680 --> 00:33:59,360
Speaker 2: don't know what else to do. It's kind of like

680
00:33:59,400 --> 00:34:01,520
Speaker 2: we're saying with the the uber model, they're playing the hits.

681
00:34:01,520 --> 00:34:03,280
Speaker 2: It's like, fuck, what did we do in the past.

682
00:34:03,320 --> 00:34:06,400
Speaker 2: We spent a lot of money ship what do we

683
00:34:06,440 --> 00:34:11,439
Speaker 2: buy GPUs? I guess what may we train more? They're

684
00:34:11,440 --> 00:34:13,560
Speaker 2: going to spend so much money on training and it's like,

685
00:34:13,800 --> 00:34:16,920
Speaker 2: to what end, your last model was a joke. This

686
00:34:16,960 --> 00:34:17,400
Speaker 2: is why it was.

687
00:34:17,440 --> 00:34:19,440
Speaker 4: It was really interesting to see that OpEd that came

688
00:34:19,480 --> 00:34:23,040
Speaker 4: from Eric Schmid and his research where he was you know,

689
00:34:23,360 --> 00:34:25,680
Speaker 4: Eric Smith is someone who was an architect of the

690
00:34:25,760 --> 00:34:30,200
Speaker 4: idea that offormacy of Google, pharmacy of Google, you know,

691
00:34:30,920 --> 00:34:35,239
Speaker 4: chairman of National Security, a group that was trying to

692
00:34:35,239 --> 00:34:40,520
Speaker 4: figure out how to merge artificial intelligence and too defence contractors,

693
00:34:40,600 --> 00:34:43,719
Speaker 4: and how to create a you know, foreign policy that

694
00:34:43,760 --> 00:34:46,799
Speaker 4: would allow America to dominate, you know, really to win

695
00:34:46,920 --> 00:34:50,080
Speaker 4: an arms race and AI arms race with China. And

696
00:34:50,480 --> 00:34:53,920
Speaker 4: he comes away saying the strategy I basically helped craft,

697
00:34:54,000 --> 00:34:56,120
Speaker 4: which was that we need to prioritize a GI so

698
00:34:56,160 --> 00:34:58,400
Speaker 4: that we can get prioritize a GI so that we

699
00:34:58,440 --> 00:35:03,840
Speaker 4: can get like a permanent lead to deter any potential rivals.

700
00:35:03,480 --> 00:35:06,960
Speaker 4: Is scaring everyone, you know, and it doesn't work. It's

701
00:35:06,960 --> 00:35:10,720
Speaker 4: a waste of capital, it's misallocating capital, it's it's imposing

702
00:35:10,760 --> 00:35:13,200
Speaker 4: all these harms. And if we look at the you know,

703
00:35:13,239 --> 00:35:15,719
Speaker 4: competitor that we're going up against against China, that by

704
00:35:15,760 --> 00:35:20,839
Speaker 4: abandoning the AGI pursuited instead prioritizing ways to figure out

705
00:35:20,840 --> 00:35:23,879
Speaker 4: how to experiment with it, integrate with it, build up

706
00:35:24,040 --> 00:35:27,319
Speaker 4: you know, build up practical use applications. You know, there's

707
00:35:27,440 --> 00:35:31,160
Speaker 4: a much more general public acceptance of it, willingness to

708
00:35:31,160 --> 00:35:34,319
Speaker 4: try it out, adopt it. And we're not seeing because

709
00:35:34,360 --> 00:35:37,080
Speaker 4: we're not trying to scale out these massive either monopolies

710
00:35:37,560 --> 00:35:41,319
Speaker 4: or one size fits all models. You see wider you know,

711
00:35:41,360 --> 00:35:44,400
Speaker 4: adoption and something that looks like it's a more sustainable model.

712
00:35:44,440 --> 00:35:46,360
Speaker 4: Are we going to follow it? Probably not, of course not.

713
00:35:46,680 --> 00:35:49,760
Speaker 2: What I love about chasing China as well is China

714
00:35:49,800 --> 00:35:51,879
Speaker 2: has had stories for like a year where it's like, yeah,

715
00:35:51,920 --> 00:35:54,480
Speaker 2: we have a bunch of unused GPU compute. Yes, well

716
00:35:54,680 --> 00:35:57,400
Speaker 2: we're massively overbuilt. The Joseph Sia I think it was

717
00:35:57,480 --> 00:36:01,040
Speaker 2: the Chinese billionaires said, yeah, it's a bubble. We haven't

718
00:36:01,040 --> 00:36:03,000
Speaker 2: really a gp of bubble, and America is just like

719
00:36:03,000 --> 00:36:04,719
Speaker 2: we can need to fucking copy it, we need to

720
00:36:04,760 --> 00:36:07,319
Speaker 2: beat them. We're great, We're gonna run our economy into

721
00:36:07,360 --> 00:36:09,000
Speaker 2: the ground.

722
00:36:09,200 --> 00:36:11,160
Speaker 4: China, it's like we're saying we're gonna confu them, And

723
00:36:11,160 --> 00:36:14,040
Speaker 4: what is it that we're actually doing. We're prioritizing developing

724
00:36:14,200 --> 00:36:16,879
Speaker 4: artificial intelligence that has like a question mark consumer use.

725
00:36:17,280 --> 00:36:20,360
Speaker 4: That's going to be used in you know, killing machines

726
00:36:20,360 --> 00:36:24,440
Speaker 4: and drones maybe and for surveillance purpose. Yeah, that's not

727
00:36:24,480 --> 00:36:26,640
Speaker 4: even generator of a I. But that's where the actual

728
00:36:27,000 --> 00:36:30,520
Speaker 4: excitement for any sort of artificial intelligence future is. And

729
00:36:30,560 --> 00:36:33,520
Speaker 4: this is you know, the generative AI stuff. It's talked

730
00:36:33,560 --> 00:36:36,040
Speaker 4: about as if it is the future, the transformative future

731
00:36:36,080 --> 00:36:39,880
Speaker 4: of artificial intelligence. In reality, it's just the actual you know,

732
00:36:39,920 --> 00:36:43,719
Speaker 4: the actual interest excitement capital is gonna I think go

733
00:36:43,800 --> 00:36:45,400
Speaker 4: back to the center of gravity, which is like, how

734
00:36:45,400 --> 00:36:48,799
Speaker 4: do we just figure out the shiniest, the most fearsome weaponry.

735
00:36:49,000 --> 00:36:51,879
Speaker 2: But I think that what's weird about this is I've

736
00:36:52,360 --> 00:36:55,239
Speaker 2: I don't think we've had a bubble that spreads so

737
00:36:55,440 --> 00:36:58,880
Speaker 2: far into consumers' hearts. I'm not saying it's as bad

738
00:36:58,920 --> 00:37:01,640
Speaker 2: as the housing bubble consumer software. If we go back

739
00:37:01,640 --> 00:37:03,720
Speaker 2: to the dot com boom, I think it's like forty

740
00:37:03,719 --> 00:37:05,719
Speaker 2: five percent of Americans had access to the Internet. It

741
00:37:05,760 --> 00:37:08,600
Speaker 2: was relatively small in comparison, though the massive other investment

742
00:37:08,600 --> 00:37:12,160
Speaker 2: in fiber happened. But I don't think people realize what

743
00:37:12,200 --> 00:37:14,480
Speaker 2: they see to the chat GPT may not exist in

744
00:37:14,520 --> 00:37:16,720
Speaker 2: a year or two at least not in the same

745
00:37:16,760 --> 00:37:18,959
Speaker 2: way it's going to be. So you're already seeing week

746
00:37:19,040 --> 00:37:24,040
Speaker 2: long weight rate limits on anthropics. Claude, like, do people

747
00:37:24,080 --> 00:37:27,200
Speaker 2: not realize that this could have I guess they don't realize,

748
00:37:27,320 --> 00:37:29,719
Speaker 2: and I don't. I think that there's going to be

749
00:37:29,719 --> 00:37:32,399
Speaker 2: a big dunce mask off. There are so many people

750
00:37:32,400 --> 00:37:34,800
Speaker 2: who have fallen behind this. I mean not to bridge

751
00:37:34,840 --> 00:37:37,279
Speaker 2: too aggressively into this, but there is a story in

752
00:37:37,280 --> 00:37:39,160
Speaker 2: the Wall Street Journal that I shared with you, of course,

753
00:37:39,200 --> 00:37:42,279
Speaker 2: about this movie called Critters with a Z or a

754
00:37:42,360 --> 00:37:46,200
Speaker 2: Z from my Canadian UK listeners, where open Ai will

755
00:37:46,239 --> 00:37:48,799
Speaker 2: be providing the compute and the tech to do a

756
00:37:48,880 --> 00:37:51,680
Speaker 2: movie called Creators with a budget of less than thirty

757
00:37:51,680 --> 00:37:54,480
Speaker 2: million dollars. Though it's not obvious whether open ai and

758
00:37:54,520 --> 00:37:56,840
Speaker 2: their computer is part of that. But it's the weirdest

759
00:37:56,840 --> 00:37:58,440
Speaker 2: shit in the world. Alison, you would bring this up

760
00:37:58,440 --> 00:38:00,680
Speaker 2: that like they're still using a bunch of humans.

761
00:38:01,360 --> 00:38:01,640
Speaker 1: Yeah.

762
00:38:01,920 --> 00:38:04,000
Speaker 3: So I was reading the same story and I haven't

763
00:38:04,040 --> 00:38:06,399
Speaker 3: done any This came out this morning, so I haven't

764
00:38:06,440 --> 00:38:08,120
Speaker 3: done my own reporting on it, but I will say

765
00:38:08,320 --> 00:38:11,280
Speaker 3: from the story I read, it seems like they're hiring

766
00:38:11,400 --> 00:38:16,080
Speaker 3: two different animation studios with artists and writers working on

767
00:38:16,120 --> 00:38:19,640
Speaker 3: the script. They're hiring human actors to voice the characters,

768
00:38:19,880 --> 00:38:24,520
Speaker 3: and then some mystery X amount of the movie will

769
00:38:24,560 --> 00:38:27,239
Speaker 3: be put together with AI. And I honestly don't know

770
00:38:27,840 --> 00:38:31,400
Speaker 3: how different that is from a regular Pixar or DreamWorks

771
00:38:31,440 --> 00:38:35,680
Speaker 3: animation process, but it seems like there's When I first

772
00:38:35,719 --> 00:38:38,400
Speaker 3: saw the you know, the teaser image is very cute,

773
00:38:38,480 --> 00:38:40,440
Speaker 3: and I was like, oh God, They're like, this is

774
00:38:40,600 --> 00:38:42,960
Speaker 3: gonna be some AI propaganda and it's gonna be very

775
00:38:43,040 --> 00:38:46,520
Speaker 3: cute and hard for me to refute. But actually it's

776
00:38:46,560 --> 00:38:49,319
Speaker 3: just a human made movie, it seems, and with an

777
00:38:49,360 --> 00:38:50,399
Speaker 3: extra computer help.

778
00:38:50,520 --> 00:38:52,600
Speaker 2: And this picture I'm holding up. Of course we're listening

779
00:38:52,640 --> 00:38:54,759
Speaker 2: to a podcast that you can all see this. It's

780
00:38:54,840 --> 00:38:57,400
Speaker 2: just this generic blue furry creature.

781
00:38:57,760 --> 00:38:59,480
Speaker 3: It looks like an extra from Monsters, Inc.

782
00:38:59,560 --> 00:39:02,799
Speaker 2: It really does. And what it's not a copyright. Yeah,

783
00:39:02,800 --> 00:39:05,799
Speaker 2: it's the same thing, it's different. But what's funny with that?

784
00:39:05,880 --> 00:39:07,920
Speaker 2: As well? As I was mentioned this as a lead

785
00:39:07,960 --> 00:39:11,720
Speaker 2: in it's that thirty million dollar thing. If that doesn't

786
00:39:11,760 --> 00:39:15,520
Speaker 2: include open eyes compute, it probably costs the same as

787
00:39:15,560 --> 00:39:19,440
Speaker 2: a pixel movie. Because you're still like, actually, three D

788
00:39:19,520 --> 00:39:22,120
Speaker 2: animation is one of the few other GPU use cases,

789
00:39:22,320 --> 00:39:24,959
Speaker 2: so really it's just a different thing running. It'll be funny.

790
00:39:24,960 --> 00:39:27,120
Speaker 4: Also if they save money because they don't do any marketing,

791
00:39:27,160 --> 00:39:28,680
Speaker 4: and they're like, see how cheap it is if you

792
00:39:28,719 --> 00:39:31,319
Speaker 4: don't have as a movie at all.

793
00:39:31,960 --> 00:39:34,560
Speaker 3: I think they might be getting around some Hollywood unions.

794
00:39:34,760 --> 00:39:38,560
Speaker 2: Yeah, oh really, they're going completely overseas too.

795
00:39:39,719 --> 00:39:41,600
Speaker 3: I'd have to check. Don't quote me on it, but

796
00:39:41,760 --> 00:39:43,359
Speaker 3: I think I know they were using at least one

797
00:39:43,400 --> 00:39:47,960
Speaker 3: overseas animation studio, so they're probably saving a lot on

798
00:39:48,000 --> 00:39:51,760
Speaker 3: the animation process by not paying animators. I would guess.

799
00:39:51,800 --> 00:39:54,080
Speaker 2: It's so cool. And also another fact from the story

800
00:39:54,120 --> 00:39:55,640
Speaker 2: is we don't know how long the piece will be,

801
00:39:56,560 --> 00:39:59,440
Speaker 2: and it's if it's like five minutes long, I'm so sorry,

802
00:39:59,480 --> 00:40:01,600
Speaker 2: come on, feature length, feature length.

803
00:40:01,920 --> 00:40:04,920
Speaker 4: I should make it as long as the Silent Napoleon film.

804
00:40:05,120 --> 00:40:08,440
Speaker 2: Which one six hours. Yeah, I actually love that they

805
00:40:08,480 --> 00:40:10,879
Speaker 2: should be forced to it. I don't know how they're

806
00:40:10,920 --> 00:40:12,399
Speaker 2: going to do a feature long movie, because I don't

807
00:40:12,440 --> 00:40:14,920
Speaker 2: know if you've cursed yourself by looking at the AI

808
00:40:15,000 --> 00:40:17,600
Speaker 2: generated movies that people try every so often, one pops

809
00:40:17,680 --> 00:40:20,920
Speaker 2: up on Twitter where it's it'll be Yeah, I made

810
00:40:20,960 --> 00:40:22,799
Speaker 2: this entire thing in AI, and you look at the

811
00:40:22,800 --> 00:40:24,839
Speaker 2: front and it's like a different fucking thing each time

812
00:40:25,160 --> 00:40:29,400
Speaker 2: that balloon boy one, different size balloon, different color balloon.

813
00:40:29,640 --> 00:40:32,640
Speaker 2: You read the stories about the balloon Boy one, it's like, yeah,

814
00:40:33,160 --> 00:40:34,959
Speaker 2: they kept putting a face in the balloon. We don't

815
00:40:34,960 --> 00:40:37,880
Speaker 2: know why. I just I and I know I have

816
00:40:37,920 --> 00:40:39,480
Speaker 2: a good amount of film and TV people who listen

817
00:40:39,560 --> 00:40:42,319
Speaker 2: who are quite anxious about this. This doesn't scare me

818
00:40:42,400 --> 00:40:45,640
Speaker 2: because they're very vague about the details. Every other big

819
00:40:45,680 --> 00:40:49,399
Speaker 2: tech innovation, I even other than the metaverse, I guess

820
00:40:49,480 --> 00:40:51,239
Speaker 2: they usually like to show you behind the curtain a

821
00:40:51,239 --> 00:40:53,000
Speaker 2: little bit and like talk up that it. There'd be

822
00:40:53,000 --> 00:40:56,200
Speaker 2: a big, splashy story and like MIT Technology Review or

823
00:40:56,200 --> 00:40:57,920
Speaker 2: something like that, being like all New York Times be like,

824
00:40:57,920 --> 00:40:59,759
Speaker 2: oh look at this, look at that, look at all things,

825
00:40:59,800 --> 00:41:02,280
Speaker 2: and the like, Yeah, we're just using some people somewhere

826
00:41:02,360 --> 00:41:05,200
Speaker 2: in a place and they will make it. And in

827
00:41:05,239 --> 00:41:07,600
Speaker 2: the War Street Journal story as well, they showed sketches

828
00:41:08,040 --> 00:41:11,880
Speaker 2: that would then be turned into AI. I just this

829
00:41:12,040 --> 00:41:15,720
Speaker 2: feels like a death rattle far more than something terribly scary,

830
00:41:15,760 --> 00:41:18,040
Speaker 2: And I understand film TP people are likely a bit scared,

831
00:41:18,040 --> 00:41:20,919
Speaker 2: but it's like they're using out of the country studios.

832
00:41:21,320 --> 00:41:23,920
Speaker 2: They're try that. Of course, I just assume they're skipping

833
00:41:24,000 --> 00:41:26,440
Speaker 2: union stuff because this is all they do. It's like

834
00:41:26,680 --> 00:41:28,960
Speaker 2: this is the best they can squeak out years in

835
00:41:29,160 --> 00:41:32,760
Speaker 2: fucking how is this? And it's like a boring looking

836
00:41:33,160 --> 00:41:36,640
Speaker 2: children's thing, I guess with a name from two thousand

837
00:41:36,680 --> 00:41:37,040
Speaker 2: and one.

838
00:41:37,360 --> 00:41:40,279
Speaker 3: Oh, it does have the producers or writers who worked

839
00:41:40,320 --> 00:41:43,680
Speaker 3: on Paddington and Peru apparently, So what a.

840
00:41:43,640 --> 00:41:46,399
Speaker 2: Movie about a criminal a sequel to a movie about

841
00:41:46,400 --> 00:41:50,360
Speaker 2: a criminal who unfairly attacked you? Grun No, sorry, I

842
00:41:50,360 --> 00:41:51,879
Speaker 2: mean then this is the question of.

843
00:41:51,880 --> 00:41:55,400
Speaker 4: You know where in the Uber analogy is this is

844
00:41:55,560 --> 00:42:00,360
Speaker 4: this you know Uber's failed expansions where they tried different

845
00:42:00,400 --> 00:42:04,440
Speaker 4: models overseas, or is this Uber returning home where they

846
00:42:04,480 --> 00:42:07,400
Speaker 4: take the lessons from overseas or they use those overseas

847
00:42:07,400 --> 00:42:09,879
Speaker 4: things to buy them a bit more time to then

848
00:42:10,000 --> 00:42:11,480
Speaker 4: subsidize operations.

849
00:42:11,560 --> 00:42:15,120
Speaker 2: Here is my comparison, and this is the drone deliveries.

850
00:42:15,800 --> 00:42:18,360
Speaker 2: This is the drone deliveries. It's the Amazon drone deliveries.

851
00:42:18,400 --> 00:42:20,520
Speaker 2: Great job, Casey and Kevin several years ago talk about

852
00:42:20,520 --> 00:42:23,960
Speaker 2: the Amazon drone deliveries and never fucking happened. Mate. It's

853
00:42:24,400 --> 00:42:26,319
Speaker 2: hilarious as well because it is the same thing. It's like,

854
00:42:26,560 --> 00:42:29,279
Speaker 2: we cobbled together this. It sucked. It took so much money.

855
00:42:29,320 --> 00:42:32,080
Speaker 2: It's horribly inefficient. It sucks. We hate it, you hate it,

856
00:42:32,080 --> 00:42:33,880
Speaker 2: the customers hate it. We hate we hate doing this,

857
00:42:34,000 --> 00:42:37,280
Speaker 2: but we did it. M M toada And it's okay,

858
00:42:37,560 --> 00:42:41,080
Speaker 2: Well you sure prove that, do you? Wort Alison. It's like, yeah,

859
00:42:41,080 --> 00:42:42,959
Speaker 2: we use the power of AI to hire a bunch

860
00:42:43,000 --> 00:42:45,000
Speaker 2: of people to do all the real work, because you

861
00:42:45,040 --> 00:42:47,880
Speaker 2: can't trust this to work. It does not work.

862
00:42:47,960 --> 00:42:50,160
Speaker 3: When I saw a headline for an AI movie, I

863
00:42:50,280 --> 00:42:52,759
Speaker 3: was like, it's it's gonna be awful. Yeah, it's gonna

864
00:42:52,840 --> 00:42:54,520
Speaker 3: like writing a movie is hard.

865
00:42:54,600 --> 00:42:56,600
Speaker 2: Well wait a minute. Also, this is the other thing

866
00:42:56,640 --> 00:42:59,160
Speaker 2: of Oh my god, how are they gonna lip sync

867
00:42:59,239 --> 00:43:02,640
Speaker 2: this shit? How do you live sing this? You can't generate,

868
00:43:03,040 --> 00:43:05,919
Speaker 2: you can't generate the same frame. How are they gonna

869
00:43:06,120 --> 00:43:08,440
Speaker 2: Are they gonna go in and post edit it with humans?

870
00:43:08,480 --> 00:43:12,120
Speaker 2: I assume at this point how much you actually relying

871
00:43:12,160 --> 00:43:13,239
Speaker 2: on AI for.

872
00:43:13,880 --> 00:43:15,120
Speaker 3: It's very unclear.

873
00:43:15,880 --> 00:43:18,120
Speaker 2: That's just it all fits. It feels like being at

874
00:43:18,120 --> 00:43:19,880
Speaker 2: a party where everyone's pissed themselves.

875
00:43:20,680 --> 00:43:25,520
Speaker 3: It does feel like some next level like young propaganda. Yeah,

876
00:43:25,560 --> 00:43:28,960
Speaker 3: like if they can get kids to enjoy whatever this

877
00:43:29,520 --> 00:43:32,920
Speaker 3: monstrous movie is going to be, then maybe there's like

878
00:43:33,239 --> 00:43:36,600
Speaker 3: a longer term brand play for open Ai as a

879
00:43:36,800 --> 00:43:38,920
Speaker 3: warm and cuddly but safe for children.

880
00:43:39,040 --> 00:43:42,400
Speaker 2: The thing is French and Korean companies have already already

881
00:43:42,400 --> 00:43:45,200
Speaker 2: been doing slot based three D shows. I don't mean

882
00:43:45,760 --> 00:43:49,240
Speaker 2: the famous one from the career K pop demon Hutness,

883
00:43:49,239 --> 00:43:51,759
Speaker 2: which is apparently very good. I've not watched it, and

884
00:43:51,920 --> 00:43:54,399
Speaker 2: please don't kill me. I'm not attacking that. I'm talking

885
00:43:54,400 --> 00:43:56,920
Speaker 2: about there is a gluttony of like very cheap kids

886
00:43:56,960 --> 00:43:58,880
Speaker 2: three D kids shows, and they've been around for decades

887
00:43:59,000 --> 00:44:01,399
Speaker 2: because you can do this on the cheap. Now you've

888
00:44:01,440 --> 00:44:04,920
Speaker 2: already another thing where the Uber model made sense as

889
00:44:04,960 --> 00:44:07,320
Speaker 2: long as you didn't count the costs, which is, yeah,

890
00:44:07,360 --> 00:44:10,000
Speaker 2: this is a way of getting people around that become

891
00:44:10,040 --> 00:44:13,200
Speaker 2: dependent on because it's useful. This is like we have

892
00:44:13,280 --> 00:44:17,360
Speaker 2: found an extremely expensive and annoying way to do something

893
00:44:17,560 --> 00:44:19,640
Speaker 2: that we already have a cheap alternative to do. It's

894
00:44:19,680 --> 00:44:22,120
Speaker 2: not like there was a cheap, a cheap, reliable cab

895
00:44:22,200 --> 00:44:24,799
Speaker 2: service that Uber replaced there was a slow shit cab

896
00:44:24,880 --> 00:44:28,160
Speaker 2: service that Uber replaced everywhere, and it's like, is it

897
00:44:28,239 --> 00:44:30,880
Speaker 2: a good company? Is it horrible to work us? Yes?

898
00:44:30,920 --> 00:44:33,640
Speaker 2: But does it work? Yes? This is we're going to

899
00:44:33,680 --> 00:44:36,520
Speaker 2: automate everything with the power of AI, other than labor,

900
00:44:37,920 --> 00:44:39,000
Speaker 2: other than stuff.

901
00:44:40,040 --> 00:44:42,960
Speaker 3: That's where the AI story starts to overlap again with Crypto,

902
00:44:43,239 --> 00:44:48,120
Speaker 3: where at least with Uber you understand what you're getting

903
00:44:48,160 --> 00:44:51,160
Speaker 3: as a consumer, and then with AI you're kind of like,

904
00:44:51,480 --> 00:44:53,640
Speaker 3: I don't really know what this is. I don't know

905
00:44:53,680 --> 00:44:56,800
Speaker 3: what problem it's solving. It's like a solution in search

906
00:44:56,880 --> 00:45:01,279
Speaker 3: of a problem. And that was Crypto's same bag. It's

907
00:45:01,400 --> 00:45:04,760
Speaker 3: just like, oh, we invented this cool new alternative money system.

908
00:45:05,160 --> 00:45:08,680
Speaker 2: Why the thing is with Crypto is they always had

909
00:45:08,719 --> 00:45:10,960
Speaker 2: a plan, which sucks. I really should have seen it coming.

910
00:45:11,040 --> 00:45:13,120
Speaker 2: I was not smart enough at the time. It was

911
00:45:13,120 --> 00:45:15,320
Speaker 2: they always wanted to just get embedded in the financial

912
00:45:15,360 --> 00:45:18,279
Speaker 2: system and then just turn the funny money into real money. Hey,

913
00:45:18,280 --> 00:45:19,960
Speaker 2: I doesn't have that. There is no way to turn

914
00:45:20,000 --> 00:45:22,759
Speaker 2: this into new You can't just generate new money. That's

915
00:45:22,800 --> 00:45:25,640
Speaker 2: what Crypto did and it fucking sucks. And by the way,

916
00:45:25,800 --> 00:45:28,000
Speaker 2: the next Crypto crash is going to wash out some

917
00:45:28,120 --> 00:45:31,040
Speaker 2: real like it's gonna really fuck people up. I don't

918
00:45:31,080 --> 00:45:34,560
Speaker 2: think people realize that SBF two, who at this point

919
00:45:34,600 --> 00:45:37,200
Speaker 2: might just be SBF Like if he just gets pardoned

920
00:45:37,200 --> 00:45:39,440
Speaker 2: and comes back, Honestly, if he comes back and does

921
00:45:39,440 --> 00:45:41,960
Speaker 2: it again, no one, no one can complain.

922
00:45:42,040 --> 00:45:43,560
Speaker 4: I'm going to the last clove. I've got to go

923
00:45:43,600 --> 00:45:46,799
Speaker 4: in the hyperbolic time chamber. I joined the fight just

924
00:45:46,840 --> 00:45:48,479
Speaker 4: so I can put him in cuffs.

925
00:45:48,480 --> 00:45:51,080
Speaker 2: You're going to put Sam Backman free back in cuffs, Yes,

926
00:45:51,719 --> 00:45:55,239
Speaker 2: Sam Altman free would be good. It's it's just I

927
00:45:55,239 --> 00:45:57,239
Speaker 2: don't see an end point for this. I don't see

928
00:45:57,360 --> 00:46:00,360
Speaker 2: everyone's that, even the boosters at this point, they're like,

929
00:46:00,360 --> 00:46:03,920
Speaker 2: and then it will be powerful. When how what are

930
00:46:03,920 --> 00:46:06,239
Speaker 2: you seeing that even tells you this. I don't even

931
00:46:06,280 --> 00:46:07,360
Speaker 2: want to fight. Just tell me.

932
00:46:07,680 --> 00:46:10,840
Speaker 3: I do think there's so there's just so much money

933
00:46:11,480 --> 00:46:13,799
Speaker 3: behind it, and there's so many people who've invested. I

934
00:46:13,840 --> 00:46:17,120
Speaker 3: was listening to a VC guy get interviewed on the

935
00:46:17,120 --> 00:46:19,400
Speaker 3: Odd Lots podcast, and I can't remember his name, so

936
00:46:19,480 --> 00:46:23,440
Speaker 3: I apologize, But he was talking about how all these founders,

937
00:46:23,520 --> 00:46:26,400
Speaker 3: like all these smaller startups that are getting in on

938
00:46:26,440 --> 00:46:29,839
Speaker 3: the AI game. All these founders have kind of been

939
00:46:29,920 --> 00:46:32,560
Speaker 3: raised with this idea of Silicon Valley and what it

940
00:46:32,600 --> 00:46:35,880
Speaker 3: will bring you, and it's life changing amounts of wealth

941
00:46:36,280 --> 00:46:38,520
Speaker 3: and when you have enough people, and like the vcs

942
00:46:38,520 --> 00:46:41,319
Speaker 3: are part of that, the actual tech startups are part

943
00:46:41,360 --> 00:46:45,040
Speaker 3: of that. Stanford and like kind of the whole ethos

944
00:46:45,080 --> 00:46:48,440
Speaker 3: of the of the valley is like, if you just

945
00:46:48,560 --> 00:46:51,279
Speaker 3: keep going and work hard enough, you can have generational

946
00:46:51,320 --> 00:46:54,239
Speaker 3: wealth and that is a very powerful force. And I

947
00:46:54,280 --> 00:46:57,280
Speaker 3: feel like I think we're going to be seeing AI

948
00:46:57,560 --> 00:47:00,439
Speaker 3: kind of hype last longer than we have and other

949
00:47:00,560 --> 00:47:05,680
Speaker 3: previous bubbles and tech cycles, in part because the potential

950
00:47:05,719 --> 00:47:09,200
Speaker 3: for the wealth is outstanding and it's like nothing we've

951
00:47:09,200 --> 00:47:09,680
Speaker 3: ever seen.

952
00:47:09,760 --> 00:47:12,480
Speaker 2: But that's the thing. You're completely right, except AI has

953
00:47:12,480 --> 00:47:14,440
Speaker 2: one problem, which is all the companies lose a shit

954
00:47:14,520 --> 00:47:16,680
Speaker 2: ton of money and no one's selling, no one is

955
00:47:16,680 --> 00:47:19,680
Speaker 2: buying that. There's been like three acquisitions. There's one to AMD,

956
00:47:19,840 --> 00:47:23,120
Speaker 2: one to end Video, one to a public company called Nice,

957
00:47:23,680 --> 00:47:26,040
Speaker 2: which sold the customer It was an AI cognitive I

958
00:47:26,080 --> 00:47:28,120
Speaker 2: think they were called. It was like an AI customer

959
00:47:28,160 --> 00:47:30,640
Speaker 2: service thing. Either never really seem that good anyway, But

960
00:47:31,360 --> 00:47:32,960
Speaker 2: that whole thing is true. And I think that that's

961
00:47:32,960 --> 00:47:36,440
Speaker 2: what people and I think that the myth of you

962
00:47:36,440 --> 00:47:38,600
Speaker 2: can just use AI to spin up a startup quickly

963
00:47:38,640 --> 00:47:41,520
Speaker 2: as well has kind of gone into has kind of

964
00:47:41,560 --> 00:47:44,520
Speaker 2: fueled that myth us as well. But the problem is

965
00:47:44,920 --> 00:47:48,239
Speaker 2: this is so different because the whole point of silicon Bag,

966
00:47:48,280 --> 00:47:50,160
Speaker 2: the whole thing where you can just move there and

967
00:47:50,160 --> 00:47:53,480
Speaker 2: start a startup, is because it didn't cost ruinous amounts

968
00:47:53,480 --> 00:47:56,280
Speaker 2: of money to start one. You didn't get three million

969
00:47:56,320 --> 00:47:58,560
Speaker 2: dollars from a VC and expect to spend two and

970
00:47:58,560 --> 00:48:02,040
Speaker 2: a half million of that on compute. You were like, Okay,

971
00:48:02,080 --> 00:48:05,360
Speaker 2: we're gonna we're gonna have to bootstrap a little bit further.

972
00:48:05,440 --> 00:48:07,279
Speaker 2: We've just got a little bit of venture capital. We're

973
00:48:07,280 --> 00:48:10,160
Speaker 2: gonna go this far. This is like every step of

974
00:48:10,200 --> 00:48:13,160
Speaker 2: the way, this cost increases massively. It used to be

975
00:48:13,200 --> 00:48:16,480
Speaker 2: it was sales and marketing and just people. AI is

976
00:48:16,680 --> 00:48:20,839
Speaker 2: people plus compute plus marketing plus this plus that I think,

977
00:48:21,080 --> 00:48:23,920
Speaker 2: you know, perplexity. The AI search engine spent one hundred

978
00:48:23,920 --> 00:48:26,120
Speaker 2: and sixty four percent of their revenue in twenty twenty

979
00:48:26,120 --> 00:48:29,160
Speaker 2: four just on compute and aws like, it's like, this

980
00:48:29,280 --> 00:48:34,279
Speaker 2: is not this whole generational wealthing. I fully agree it's

981
00:48:34,320 --> 00:48:37,200
Speaker 2: what they're using to sell it. I just don't think

982
00:48:37,239 --> 00:48:41,319
Speaker 2: it's gonna work. And it's scary because this could have

983
00:48:41,520 --> 00:48:43,600
Speaker 2: the wide thing, and I really haven't talked about this enough.

984
00:48:43,920 --> 00:48:46,120
Speaker 2: The wider problem is as well, is all of these

985
00:48:46,160 --> 00:48:48,320
Speaker 2: people who went to Silicon Valley raised all this money

986
00:48:48,719 --> 00:48:51,879
Speaker 2: or have pretty much raised to sell companies that will

987
00:48:51,880 --> 00:48:53,920
Speaker 2: never sell, that they can never take public because they've

988
00:48:53,880 --> 00:48:56,200
Speaker 2: barn too much money, they don't really have great user

989
00:48:56,200 --> 00:48:59,359
Speaker 2: bases because they don't really have those, and so they're

990
00:48:59,360 --> 00:49:01,000
Speaker 2: just gonna sit there. And then you've got a bunch

991
00:49:01,040 --> 00:49:03,280
Speaker 2: of VC money tied up in that that will never exit,

992
00:49:03,320 --> 00:49:05,840
Speaker 2: and a bunch of limited partner money that will never exit.

993
00:49:06,800 --> 00:49:09,800
Speaker 2: I think that there is an entirely separate bubble building

994
00:49:10,960 --> 00:49:13,359
Speaker 2: that when that bust is going to, the depression within

995
00:49:13,480 --> 00:49:16,240
Speaker 2: Silicon value is going to be insane. It's already pretty gnolly,

996
00:49:16,560 --> 00:49:18,359
Speaker 2: but I think it's like thirty three percent of bench

997
00:49:18,440 --> 00:49:22,760
Speaker 2: capital when it's a AI last year. It's like, eventually

998
00:49:22,760 --> 00:49:25,040
Speaker 2: people are going to realize there's no exit for anyone.

999
00:49:25,120 --> 00:49:27,799
Speaker 2: And I don't know what that does. I mean, it

1000
00:49:27,840 --> 00:49:30,600
Speaker 2: will piss off limited partners. The money that comes to

1001
00:49:30,680 --> 00:49:32,120
Speaker 2: VCS is just not going to be there.

1002
00:49:32,480 --> 00:49:36,160
Speaker 4: Well, so then that's the question, right, because venture capital

1003
00:49:36,320 --> 00:49:42,640
Speaker 4: encourages on one level overvaluing because you need to figure

1004
00:49:42,640 --> 00:49:44,680
Speaker 4: out a way to make more money than what you

1005
00:49:44,760 --> 00:49:47,840
Speaker 4: put in on the exit within acquisition or some merger. Yeah,

1006
00:49:47,880 --> 00:49:51,880
Speaker 4: but on another level, you're also working within a network

1007
00:49:51,920 --> 00:49:54,839
Speaker 4: trying to enrich yourself and your friends, or trying to

1008
00:49:54,880 --> 00:50:00,600
Speaker 4: build the infrastructure for future startups, portfolio options that you

1009
00:50:00,600 --> 00:50:03,480
Speaker 4: and your friends make to come in and make money, whether.

1010
00:50:03,320 --> 00:50:06,160
Speaker 2: You're building a platform the other people can invest in bits.

1011
00:50:05,960 --> 00:50:09,280
Speaker 4: Of and so you know, on one level, I really,

1012
00:50:09,320 --> 00:50:11,879
Speaker 4: I really do. I agree that there's not really much

1013
00:50:11,920 --> 00:50:15,239
Speaker 4: of an exit ramp if there's actually no revenue and

1014
00:50:15,320 --> 00:50:17,600
Speaker 4: no profits. But then also i'd be curious, like do

1015
00:50:17,640 --> 00:50:19,480
Speaker 4: you think they're going to try to ram these things

1016
00:50:19,480 --> 00:50:21,239
Speaker 4: through similar to like what we saw with Corey Weve

1017
00:50:21,280 --> 00:50:23,279
Speaker 4: right where you know, like you you talked, I think

1018
00:50:23,400 --> 00:50:27,920
Speaker 4: extensively about ways in which the financials there do not

1019
00:50:27,960 --> 00:50:30,759
Speaker 4: actually make sense if you're interested in a company that

1020
00:50:30,840 --> 00:50:32,600
Speaker 4: actually has the capital to do what it's going to

1021
00:50:32,640 --> 00:50:35,480
Speaker 4: say is going to do, which is provide gpu compute

1022
00:50:35,520 --> 00:50:38,720
Speaker 4: to everybody, and even though it has such a central

1023
00:50:38,840 --> 00:50:43,280
Speaker 4: role in this ecosystem. It can't make profits that are

1024
00:50:43,560 --> 00:50:45,960
Speaker 4: you know, that justify the capital that's getting it has

1025
00:50:46,080 --> 00:50:49,680
Speaker 4: odious and burd burden some debt that should be a

1026
00:50:49,800 --> 00:50:53,359
Speaker 4: massive red flag and it might be you know, round tripping, right, yeah,

1027
00:50:53,520 --> 00:50:55,719
Speaker 4: but this is supposed to be like the darling of

1028
00:50:55,760 --> 00:50:58,399
Speaker 4: the sector and it and it got pushed through. Part

1029
00:50:58,440 --> 00:50:59,880
Speaker 4: of me feels like because of.

1030
00:51:00,080 --> 00:51:05,800
Speaker 2: Of I mean Invation and Video and Magneto and Magnetar

1031
00:51:05,880 --> 00:51:08,920
Speaker 2: Capital of course famous for the CDOs. Yeah, right, start

1032
00:51:08,920 --> 00:51:12,279
Speaker 2: they're back. But with core Weave, they pushed it through

1033
00:51:12,280 --> 00:51:14,720
Speaker 2: onto the markets. But that doesn't mean it can't die.

1034
00:51:14,760 --> 00:51:15,960
Speaker 4: Right, Well, so that's the thing.

1035
00:51:16,040 --> 00:51:17,200
Speaker 2: I do you think.

1036
00:51:17,040 --> 00:51:19,359
Speaker 4: That it's possible that they'd be successful in pushing it

1037
00:51:19,360 --> 00:51:21,400
Speaker 4: onto markets but it dies because I do, Yes, I

1038
00:51:21,400 --> 00:51:22,960
Speaker 4: feel like there will will definitely be a lot of

1039
00:51:22,960 --> 00:51:25,840
Speaker 4: investment incineration, But I also do think we're gonna have

1040
00:51:25,920 --> 00:51:27,600
Speaker 4: bags dumped on everybody.

1041
00:51:27,600 --> 00:51:29,480
Speaker 2: I think you could do it with something like core

1042
00:51:29,560 --> 00:51:32,840
Speaker 2: Weave and Lambda, which is another situation where Nvidia is

1043
00:51:32,840 --> 00:51:36,000
Speaker 2: the customer invested and also sells them the GPUs, which

1044
00:51:36,040 --> 00:51:38,520
Speaker 2: they then use as collateral to buy more GPUs using

1045
00:51:38,520 --> 00:51:41,960
Speaker 2: debt which is so good. You'll notice that there are

1046
00:51:42,000 --> 00:51:45,320
Speaker 2: no software companies going public. There's no software AI companies

1047
00:51:45,360 --> 00:51:48,000
Speaker 2: going public. Everyone thinks that open ai goes public here

1048
00:51:48,040 --> 00:51:50,400
Speaker 2: and oh they've got the market's gonna if they can

1049
00:51:50,440 --> 00:51:52,440
Speaker 2: even convert, the markets are gonna eat them for dinner.

1050
00:51:52,680 --> 00:51:57,799
Speaker 2: Oh yeah, we're gonna burn bazillions of dollars forever. No,

1051
00:51:58,000 --> 00:52:00,799
Speaker 2: they the markets didn't like Core. Weav that Core We've

1052
00:52:00,800 --> 00:52:04,160
Speaker 2: wouldn't have gone public had Nvidia not put more money in.

1053
00:52:05,120 --> 00:52:07,239
Speaker 2: Lambda is probably gonna be exactly the same, if they

1054
00:52:07,280 --> 00:52:10,399
Speaker 2: even make it. You won't see software companies because that's

1055
00:52:10,400 --> 00:52:14,800
Speaker 2: the other thing Core We've had, albeit bunches of debt assets,

1056
00:52:15,120 --> 00:52:18,919
Speaker 2: they have data centers kind of through Core Scientific God,

1057
00:52:18,920 --> 00:52:21,960
Speaker 2: I hate these fucking companies, but they don't they have

1058
00:52:22,080 --> 00:52:25,319
Speaker 2: things that they can point to in relationships, even open Ai.

1059
00:52:25,520 --> 00:52:28,840
Speaker 2: That's the thing with them, they don't. They barely have assets.

1060
00:52:29,160 --> 00:52:32,240
Speaker 2: Oracle is building their data center in Abilene with Cruso.

1061
00:52:32,600 --> 00:52:34,480
Speaker 2: They don't own any of the GPU. They have a

1062
00:52:34,480 --> 00:52:37,200
Speaker 2: few GPUs I think for research I've heard, but Microsoft

1063
00:52:37,239 --> 00:52:40,239
Speaker 2: owns most of their their infrastructure. They don't own their

1064
00:52:40,360 --> 00:52:43,319
Speaker 2: R and D well they do, but Microsoft also has

1065
00:52:43,320 --> 00:52:45,919
Speaker 2: access to that their intellectual property. Same deal. So it's

1066
00:52:45,960 --> 00:52:49,160
Speaker 2: like what actual value doesn't AI start up have? People

1067
00:52:49,200 --> 00:52:51,040
Speaker 2: always say oh, they're getting the data. They get the

1068
00:52:51,120 --> 00:52:53,080
Speaker 2: data so that the data will tell them. It's like

1069
00:52:53,120 --> 00:52:56,160
Speaker 2: what it's all that these horrible stories about, like oh,

1070
00:52:56,200 --> 00:52:58,759
Speaker 2: Dozea's got an LLM, they're doing this one. What's the

1071
00:52:58,920 --> 00:53:01,440
Speaker 2: end point? It's scaring, don't get me wrong, But and

1072
00:53:01,480 --> 00:53:05,520
Speaker 2: then what and there never is one? And I hope someone,

1073
00:53:05,640 --> 00:53:08,080
Speaker 2: I hope a AI software company goes public. I want

1074
00:53:08,080 --> 00:53:09,680
Speaker 2: to see this so bad. I want to see you

1075
00:53:09,719 --> 00:53:12,640
Speaker 2: have any idea. If you give me the open AI books,

1076
00:53:12,680 --> 00:53:15,760
Speaker 2: the anthropic books, you become the official homie better offline.

1077
00:53:15,760 --> 00:53:18,360
Speaker 2: I'll mention you on every episode. Get me these books.

1078
00:53:18,360 --> 00:53:19,920
Speaker 2: But because I think all of them are going to

1079
00:53:19,960 --> 00:53:22,279
Speaker 2: be like a dog's dinner, there's I've actually looked at

1080
00:53:22,280 --> 00:53:25,840
Speaker 2: the markets an uber uber by comparison. They did burnish

1081
00:53:25,800 --> 00:53:27,439
Speaker 2: it to the money. It's it's like twenty five billion

1082
00:53:27,480 --> 00:53:30,120
Speaker 2: between twenty nineteen twenty twenty two. A lot of that

1083
00:53:30,200 --> 00:53:32,000
Speaker 2: was on sales and R and D. It's pretty much

1084
00:53:32,000 --> 00:53:34,160
Speaker 2: group on. I think also the R and D with

1085
00:53:34,160 --> 00:53:37,799
Speaker 2: the autonomous cars, but separate problem but it's like, there

1086
00:53:37,840 --> 00:53:40,160
Speaker 2: wasn't a I can't find an example of someone that

1087
00:53:40,239 --> 00:53:45,799
Speaker 2: just annihilated fuel unless it's like planes. And I think

1088
00:53:45,840 --> 00:53:49,960
Speaker 2: we've established the use case for planes by now clear

1089
00:53:51,000 --> 00:53:54,360
Speaker 2: god Old. It's just it's all very frustrating. But you

1090
00:53:54,360 --> 00:53:55,719
Speaker 2: know what, I think I'm gonna call it there. I

1091
00:53:55,719 --> 00:53:58,200
Speaker 2: think we've had a good conversation. Allison. Where can people

1092
00:53:58,239 --> 00:53:59,240
Speaker 2: find you?

1093
00:53:59,239 --> 00:54:02,960
Speaker 3: You can find me on Blue Sky at amorrow or

1094
00:54:03,560 --> 00:54:06,840
Speaker 3: on CNN dot com slash nightcap ed.

1095
00:54:07,320 --> 00:54:09,680
Speaker 4: You can find me on Twitter at Big Black Jacobin.

1096
00:54:10,320 --> 00:54:13,120
Speaker 4: You can find me on Blue Sky on Edward and

1097
00:54:13,160 --> 00:54:16,200
Speaker 4: Guss Junior, and on substack at the Tech Bubble.

1098
00:54:16,680 --> 00:54:19,040
Speaker 2: And you can find me of course at Google dot com.

1099
00:54:19,320 --> 00:54:21,959
Speaker 2: Just type in prabagar Ragavan you'll find me. I pop

1100
00:54:22,040 --> 00:54:24,480
Speaker 2: right up. That's all me. Thank you so much for

1101
00:54:24,560 --> 00:54:27,359
Speaker 2: listening to everyone. This is my episodees are coming out

1102
00:54:27,400 --> 00:54:29,399
Speaker 2: in a weird order because I'm recording this knowing there's

1103
00:54:29,400 --> 00:54:31,120
Speaker 2: a three part of this week, but this will come

1104
00:54:31,120 --> 00:54:32,800
Speaker 2: out with a monologue of some sort. Thank you so

1105
00:54:32,880 --> 00:54:34,840
Speaker 2: much for listening everyone, of course, but he thank you

1106
00:54:34,840 --> 00:54:37,160
Speaker 2: for producing here out in New York City. And yeah,

1107
00:54:37,280 --> 00:54:38,319
Speaker 2: thanks everyone.

1108
00:54:46,360 --> 00:54:48,040
Speaker 6: Thank you for listening to Better Offline.

1109
00:54:48,160 --> 00:54:50,600
Speaker 2: The editor and composer of the Better Offline theme song

1110
00:54:50,680 --> 00:54:53,319
Speaker 2: is Matasowski. You can check out more of his music

1111
00:54:53,360 --> 00:54:56,279
Speaker 2: and audio projects at Matasowski dot com.

1112
00:54:56,440 --> 00:54:59,839
Speaker 6: M A T T O S O W S K I.

1113
00:55:01,400 --> 00:55:03,680
Speaker 2: You can email me at easy at better offline dot

1114
00:55:03,719 --> 00:55:05,960
Speaker 2: com or visit better offline dot com to find more

1115
00:55:05,960 --> 00:55:09,359
Speaker 2: podcast links and of course, my newsletter. I also really

1116
00:55:09,400 --> 00:55:11,680
Speaker 2: recommend you go to chat dot where's youreaed dot at

1117
00:55:11,680 --> 00:55:14,200
Speaker 2: to visit the discord, and go to our slash.

1118
00:55:13,880 --> 00:55:15,760
Speaker 6: Better Offline to check out our reddit.

1119
00:55:16,520 --> 00:55:17,880
Speaker 2: Thank you so much for listening.

1120
00:55:18,719 --> 00:55:21,400
Speaker 1: Better Offline is a production of cool Zone Media. For

1121
00:55:21,520 --> 00:55:24,680
Speaker 1: more from cool Zone Media, visit our website cool Zonemedia

1122
00:55:24,760 --> 00:55:27,560
Speaker 1: dot com, or check us out on the iHeartRadio app,

1123
00:55:27,640 --> 00:55:29,000
Speaker 1: Apple Podcasts, or wherever

1124
00:55:29,080 --> 00:55:30,320
Speaker 3: You get your podcasts.