1
00:00:02,480 --> 00:00:06,800
Speaker 1: Call Zone Media. Hi, I'm Edzeitron and welcome back to

2
00:00:06,880 --> 00:00:20,160
Speaker 1: Better Offline. As I've discussed in my last episode, there

3
00:00:20,200 --> 00:00:22,840
Speaker 1: are four intractable problems that are going to stop generative

4
00:00:22,880 --> 00:00:26,920
Speaker 1: AI from going much further. It's massive energy demands, its

5
00:00:26,960 --> 00:00:31,000
Speaker 1: massive computation demands. It's hallucinations when it authoritatively tells you

6
00:00:31,080 --> 00:00:34,040
Speaker 1: something that isn't true or makes horrible mistakes in images,

7
00:00:34,840 --> 00:00:37,040
Speaker 1: and the fact that these large language models have this

8
00:00:37,080 --> 00:00:41,080
Speaker 1: insatiable need for more training data. Yet I think what

9
00:00:41,200 --> 00:00:44,720
Speaker 1: might pop this bubble is a far simpler problem. Generative

10
00:00:44,720 --> 00:00:48,600
Speaker 1: AI simply does not deliver the magical automation that everybody

11
00:00:48,600 --> 00:00:52,080
Speaker 1: has been fantasizing about, and I don't think consumers or

12
00:00:52,200 --> 00:00:56,680
Speaker 1: enterprises are actually impressed. A year and a half after launch,

13
00:00:57,200 --> 00:00:59,600
Speaker 1: it seems like the kind of immediate and unquestioning in

14
00:00:59,640 --> 00:01:02,360
Speaker 1: fact youation with chat GBT and for that matter, or

15
00:01:02,400 --> 00:01:08,760
Speaker 1: other generative AIS has softened. Instead, there's this rising undercurrent

16
00:01:08,840 --> 00:01:12,760
Speaker 1: of apathy and mistrust and of course failure that's kind

17
00:01:12,760 --> 00:01:16,480
Speaker 1: of hard to ignore. In June twenty twenty three, traffic

18
00:01:16,480 --> 00:01:19,440
Speaker 1: to chat GPT's website, where people access the chat GPT

19
00:01:19,600 --> 00:01:22,200
Speaker 1: bot in a web browser fell for the first time

20
00:01:22,240 --> 00:01:25,160
Speaker 1: since launch, starting a trend that's continued for five of

21
00:01:25,160 --> 00:01:27,880
Speaker 1: the following eight months. According to data from similar Web,

22
00:01:28,720 --> 00:01:31,760
Speaker 1: people are becoming more aware of the technology's limitations, like

23
00:01:31,800 --> 00:01:35,839
Speaker 1: as I mentioned, hallucinations, which, as I note, is when

24
00:01:35,959 --> 00:01:39,560
Speaker 1: chat gpt confidently asserts things that aren't true, which can

25
00:01:39,600 --> 00:01:42,039
Speaker 1: be in writing, when it gives you an incorrect fact,

26
00:01:42,600 --> 00:01:44,759
Speaker 1: or in an image when it gives a dog eighteen

27
00:01:44,840 --> 00:01:49,280
Speaker 1: legs to make matters worse. According to data from data Ai,

28
00:01:49,360 --> 00:01:52,520
Speaker 1: which used to be known as Appanny, chat gbt's downloads

29
00:01:52,520 --> 00:01:54,760
Speaker 1: in iOS have begun to drop from a high of

30
00:01:54,880 --> 00:01:57,440
Speaker 1: just over seven hundred thousand a week to a plateau

31
00:01:57,440 --> 00:02:00,280
Speaker 1: of around four hundred and fifty thousand to five hundred

32
00:02:00,320 --> 00:02:04,639
Speaker 1: thousand a week since early twenty twenty three, which sounds

33
00:02:04,680 --> 00:02:07,480
Speaker 1: impressive until you hear that only seven point three five

34
00:02:07,560 --> 00:02:11,240
Speaker 1: percent of people who downloaded chatgbt in January twenty twenty

35
00:02:11,320 --> 00:02:14,200
Speaker 1: four actually used the app again thirty days after they

36
00:02:14,280 --> 00:02:18,000
Speaker 1: downloaded it, cratering from a high of twenty eight percent

37
00:02:18,040 --> 00:02:20,400
Speaker 1: a month after the app launched in June twenty twenty three.

38
00:02:21,480 --> 00:02:25,679
Speaker 1: In fact, things immediately appear to have fallen apart in

39
00:02:25,800 --> 00:02:28,280
Speaker 1: July twenty twenty three, only two months after launch, only

40
00:02:28,320 --> 00:02:31,200
Speaker 1: four point five nine percent of users opened the app

41
00:02:31,200 --> 00:02:34,680
Speaker 1: for a second time. Numbers like these tell the story

42
00:02:34,680 --> 00:02:38,320
Speaker 1: of a buzzy new application that isn't actually providing users

43
00:02:38,360 --> 00:02:41,840
Speaker 1: with much utility. I think the generative AI engine has

44
00:02:41,840 --> 00:02:45,919
Speaker 1: started to sputter for customers, for businesses, and indeed for

45
00:02:45,960 --> 00:02:49,120
Speaker 1: the startups that create them. That's bad news for any

46
00:02:49,160 --> 00:02:53,880
Speaker 1: industry that's yet to reach profitability or indeed sustainability, and

47
00:02:54,080 --> 00:02:57,360
Speaker 1: especially for generative AI, which remains relying on a kind

48
00:02:57,400 --> 00:03:01,000
Speaker 1: of an indefinite supply of cash to operate. Back in

49
00:03:01,040 --> 00:03:05,360
Speaker 1: April twenty twenty three, Dylan Patel, chief analyst at Semi Analysis,

50
00:03:05,400 --> 00:03:09,120
Speaker 1: calculated that GPT three, the previous generation of chech GPT

51
00:03:09,280 --> 00:03:12,600
Speaker 1: current ones known as GPT four, cost around seven hundred

52
00:03:12,639 --> 00:03:15,440
Speaker 1: thousand dollars a day to run. It's about twenty one

53
00:03:15,440 --> 00:03:17,680
Speaker 1: million dollars a month, or two hundred and fifty million

54
00:03:17,720 --> 00:03:21,240
Speaker 1: dollars a year. In October twenty twenty three, Richard Windsor,

55
00:03:21,440 --> 00:03:24,680
Speaker 1: the research director at large of Counterpoint Research, which is

56
00:03:24,720 --> 00:03:28,240
Speaker 1: one of the more reliable analyst houses, hypothesized that open

57
00:03:28,280 --> 00:03:30,960
Speaker 1: AI's monthly cash burn was in the region of one

58
00:03:31,000 --> 00:03:34,880
Speaker 1: point one billion dollars a month, based on them having

59
00:03:34,880 --> 00:03:37,440
Speaker 1: to raise thirteen billion dollars from Microsoft, most of it,

60
00:03:37,520 --> 00:03:40,720
Speaker 1: as I noted in credits for its Azure cloud computing

61
00:03:40,760 --> 00:03:44,000
Speaker 1: service to run their models. It could be more, it

62
00:03:44,000 --> 00:03:47,280
Speaker 1: could be less. As a private company, only investors and

63
00:03:47,320 --> 00:03:51,880
Speaker 1: other insiders can possibly know what's going on in open Ai. However,

64
00:03:52,040 --> 00:03:55,240
Speaker 1: four months later, Reuter's would report that open aye made

65
00:03:55,240 --> 00:03:57,600
Speaker 1: about two billion dollars in revenue in twenty twenty three,

66
00:03:57,800 --> 00:04:00,720
Speaker 1: a remarkable sum that much like every other story about

67
00:04:00,720 --> 00:04:05,000
Speaker 1: open ai, never mentions profit. In fact, I can't find

68
00:04:05,000 --> 00:04:07,880
Speaker 1: a single reporter that appears to have asked Sam Mormon

69
00:04:07,920 --> 00:04:11,280
Speaker 1: about how much profit open ai makes, only breathless hype

70
00:04:11,520 --> 00:04:15,680
Speaker 1: with no consideration of its sustainability. Even if open ai

71
00:04:15,840 --> 00:04:18,760
Speaker 1: burns a tenth of windsors estema about one hundred million

72
00:04:18,800 --> 00:04:22,200
Speaker 1: dollars a month, that's still far more money than they're making.

73
00:04:22,839 --> 00:04:25,000
Speaker 1: There's not a single story out there talking about them

74
00:04:25,000 --> 00:04:28,719
Speaker 1: making a profit, and I don't think they make in one.

75
00:04:28,960 --> 00:04:32,000
Speaker 1: Here's one thing we can be certain of. Though things

76
00:04:32,040 --> 00:04:37,080
Speaker 1: are getting more expensive, progress in generative AI means increasingly

77
00:04:37,120 --> 00:04:40,599
Speaker 1: complex models, and as I previously mentioned open AI's attempted

78
00:04:40,920 --> 00:04:45,000
Speaker 1: God damn it a Rakis model one, built specifically to

79
00:04:45,040 --> 00:04:48,359
Speaker 1: wow Microsoft by making chatgiputy more efficient, failed to actually

80
00:04:48,400 --> 00:04:51,719
Speaker 1: make it more efficient. Their attempts to make this a

81
00:04:51,760 --> 00:04:57,520
Speaker 1: better company are not working. Windsor the aforementioned analyst in

82
00:04:57,560 --> 00:05:00,400
Speaker 1: a separate blog also pointed out that there's nothing really

83
00:05:00,440 --> 00:05:05,560
Speaker 1: sticky about these companies. There's nothing stopping someone from switching from, say,

84
00:05:05,680 --> 00:05:09,760
Speaker 1: chat GPT to anthropics clawed two model. They're all trained

85
00:05:09,760 --> 00:05:13,280
Speaker 1: on similar data sets, and they all produce very similar answers.

86
00:05:13,680 --> 00:05:16,200
Speaker 1: And what one model might be better a one thing than another,

87
00:05:16,920 --> 00:05:22,480
Speaker 1: they're fundamentally very very similar. There's also nothing stopping someone

88
00:05:22,520 --> 00:05:27,120
Speaker 1: from simply giving up on generative AI altogether. It doesn't

89
00:05:27,160 --> 00:05:29,880
Speaker 1: seem to be the plug and play automation god that

90
00:05:29,920 --> 00:05:34,160
Speaker 1: everybody's been replicate to be, and judging by the plateauing

91
00:05:34,320 --> 00:05:37,719
Speaker 1: Chat GPT user numbers, I think that might already be happening.

92
00:05:39,279 --> 00:05:41,880
Speaker 1: It's also important to remember that while generative AI is

93
00:05:41,920 --> 00:05:46,160
Speaker 1: shiny and new artificially, intelligence is absolutely not, and over

94
00:05:46,240 --> 00:05:49,160
Speaker 1: the past decade it's found a number of homes from

95
00:05:49,240 --> 00:05:51,760
Speaker 1: expensive security apps that detect when a hacker is trying

96
00:05:51,800 --> 00:05:54,800
Speaker 1: to break into a corporate network. The spam filters proof

97
00:05:54,800 --> 00:05:58,279
Speaker 1: for reading tools like grammarly, Plenty of Things, even Siri

98
00:05:58,400 --> 00:06:01,560
Speaker 1: on your iPhone. In these context, AI is either a

99
00:06:01,640 --> 00:06:04,640
Speaker 1: small component of a larger product or something that directly

100
00:06:04,640 --> 00:06:08,400
Speaker 1: builds on human efforts. This stuff is actually valuable. AI

101
00:06:08,440 --> 00:06:10,960
Speaker 1: based spam filters are typically better than those reliant on

102
00:06:11,000 --> 00:06:13,919
Speaker 1: hand coded rules, for example, But it's also from a

103
00:06:13,960 --> 00:06:19,719
Speaker 1: marketing perspective, can of boring generative AIS A law is

104
00:06:19,760 --> 00:06:23,360
Speaker 1: that it can supplant humans either partially or entirely, producing

105
00:06:23,520 --> 00:06:26,760
Speaker 1: entire creative works that otherwise would have taken hours and

106
00:06:26,800 --> 00:06:30,960
Speaker 1: carried a real financial cost. But behind this glitzy technology

107
00:06:30,960 --> 00:06:34,760
Speaker 1: and media hype, the unspoken truth is that generative AI

108
00:06:34,920 --> 00:06:38,080
Speaker 1: holds way over the financial markets because it's regarded as

109
00:06:38,160 --> 00:06:41,120
Speaker 1: a tool to eliminate an entire swarth of jobs in

110
00:06:41,160 --> 00:06:44,880
Speaker 1: the creative and knowledge economies. It's a ghastly promise and

111
00:06:44,920 --> 00:06:48,880
Speaker 1: it underpins the vast market value of otherwise commercially unviable

112
00:06:48,880 --> 00:06:52,240
Speaker 1: generative AI companies like open Ai and Anthropic, and it's

113
00:06:52,240 --> 00:06:55,400
Speaker 1: what is driving I believe the multi billion dollar investments

114
00:06:55,600 --> 00:06:59,920
Speaker 1: we've seen from Microsoft, Amazon, and Google yeah, I see

115
00:07:00,000 --> 00:07:03,400
Speaker 1: no evidence of mass adoption of generative AI, and by

116
00:07:03,440 --> 00:07:06,359
Speaker 1: research suggests the enterprise adoption, which is the meat of

117
00:07:06,360 --> 00:07:09,640
Speaker 1: what would actually make these companies money, it just isn't there.

118
00:07:10,840 --> 00:07:13,800
Speaker 1: Deep within the earnings reports and the quotes of every

119
00:07:13,840 --> 00:07:16,840
Speaker 1: major cloud provider claiming that the AI revolution is here

120
00:07:17,080 --> 00:07:21,200
Speaker 1: is a deeply worrying trend. The AI revenue really isn't

121
00:07:21,200 --> 00:07:25,000
Speaker 1: contributing much to the bottom line outside of vacuous media coverage,

122
00:07:26,280 --> 00:07:28,000
Speaker 1: and I think the internal story is going to be

123
00:07:28,080 --> 00:07:39,560
Speaker 1: much bleaker. In early March, The Information published the story

124
00:07:39,560 --> 00:07:43,800
Speaker 1: about Amazon and Google tamping down GENERATIVEAI expectations, with these

125
00:07:43,840 --> 00:07:47,760
Speaker 1: companies dousing their salespeople's excitement about the capabilities of the

126
00:07:47,800 --> 00:07:51,000
Speaker 1: tech they're selling. A tech executive is quoted in the

127
00:07:51,080 --> 00:07:54,120
Speaker 1: article saying that customers are beginning to struggle with questions,

128
00:07:54,240 --> 00:07:57,840
Speaker 1: simple questions like is AI actually providing value? And how

129
00:07:57,880 --> 00:08:01,160
Speaker 1: do I evaluate how AI is doing? And a Gartner

130
00:08:01,160 --> 00:08:05,520
Speaker 1: analyst told Amazon Web Services sales staff that the AI

131
00:08:05,600 --> 00:08:08,280
Speaker 1: industry was at the peak of the hype cycle around

132
00:08:08,320 --> 00:08:13,000
Speaker 1: large language models and other generative AI, which is a

133
00:08:13,000 --> 00:08:16,560
Speaker 1: somewhat specific code for it's not going to get much

134
00:08:16,600 --> 00:08:21,760
Speaker 1: better anytime soon. This article confirms many of my suspicions

135
00:08:22,320 --> 00:08:26,320
Speaker 1: that and I quote the information here, other software companies

136
00:08:26,320 --> 00:08:29,320
Speaker 1: that have touted generative AI as a boon to enterprises

137
00:08:29,440 --> 00:08:33,040
Speaker 1: are still waiting for revenue to emerge, citing the example

138
00:08:33,080 --> 00:08:37,679
Speaker 1: of professional services firm KPMG buying forty seven thousand subscriptions

139
00:08:37,880 --> 00:08:41,360
Speaker 1: to Microsoft's Copilot AI at a significant discount on the

140
00:08:41,360 --> 00:08:46,920
Speaker 1: thirty dollars a c sticker price. Except KPMG bought the

141
00:08:46,960 --> 00:08:50,840
Speaker 1: subscriptions without really have engaged whether their employees actually got

142
00:08:50,880 --> 00:08:53,400
Speaker 1: anything out of it. They bought it, and I'm not

143
00:08:53,520 --> 00:08:57,679
Speaker 1: kidding you entirely so that if any KPMG customers ask

144
00:08:57,800 --> 00:09:03,559
Speaker 1: questions about AI, they'll be able to answer them. Was

145
00:09:03,640 --> 00:09:06,400
Speaker 1: so clearly in the bot Oh my god. Anyway, as

146
00:09:06,400 --> 00:09:09,600
Speaker 1: I've hinted, it's also not obvious how much AI actually

147
00:09:09,640 --> 00:09:13,120
Speaker 1: contributes to the bottom line. In Microsoft's Q four twenty

148
00:09:13,200 --> 00:09:16,720
Speaker 1: twenty three earnings report, gief financial officer Amy Hood reported

149
00:09:16,720 --> 00:09:19,559
Speaker 1: that six points of revenue growth in its Azureine Cloud

150
00:09:19,559 --> 00:09:23,400
Speaker 1: Services division was attributed to AI services. I went around

151
00:09:23,400 --> 00:09:25,920
Speaker 1: the web and I read every bloody article about their earnings.

152
00:09:26,400 --> 00:09:29,120
Speaker 1: I looked, and I looked and everyone was saying, Oh,

153
00:09:29,160 --> 00:09:31,080
Speaker 1: this is really good. I found someone who said it

154
00:09:31,120 --> 00:09:33,559
Speaker 1: was six percent of their revenue, and I went, that

155
00:09:33,679 --> 00:09:36,040
Speaker 1: sounds like complete bollocks to me. So I went and

156
00:09:36,080 --> 00:09:39,520
Speaker 1: spoke with Jordan Novette, who's covered Microsoft for many years.

157
00:09:39,520 --> 00:09:41,439
Speaker 1: He is a great cloud reporter over at CNBC, and

158
00:09:41,480 --> 00:09:45,160
Speaker 1: he actually covered Microsoft's earnings for the NBC itself, and

159
00:09:45,200 --> 00:09:49,000
Speaker 1: he confirmed that what this means is that AI contributed

160
00:09:49,160 --> 00:09:52,480
Speaker 1: six percent of the thirty percent of year over year

161
00:09:52,520 --> 00:09:56,840
Speaker 1: growth in Microsoft's as your Cloud services. That is a

162
00:09:56,880 --> 00:10:01,160
Speaker 1: percentage of a percentage. So by the way that that means,

163
00:10:01,200 --> 00:10:03,959
Speaker 1: so thirty percent growth year of year. So six percent

164
00:10:04,040 --> 00:10:06,960
Speaker 1: year of a year growth is from AI. Could be good,

165
00:10:07,040 --> 00:10:09,080
Speaker 1: but also all of the rest of it came not

166
00:10:09,120 --> 00:10:11,559
Speaker 1: from new products, just from the natural growth of the company.

167
00:10:13,320 --> 00:10:15,520
Speaker 1: It's unclear how much money that really is, but six

168
00:10:15,600 --> 00:10:18,280
Speaker 1: percent of the year over year growth isn't really that

169
00:10:18,400 --> 00:10:23,360
Speaker 1: exciting anyway elsewhere. Amazon CEO Andy Jasse, who took over

170
00:10:23,400 --> 00:10:26,120
Speaker 1: from Bezos a few years ago and was the chief

171
00:10:26,120 --> 00:10:29,360
Speaker 1: of Amazon Web Services, said that generative AI revenue was

172
00:10:29,400 --> 00:10:32,560
Speaker 1: still relatively small, but don't worry, you said it would

173
00:10:32,600 --> 00:10:34,600
Speaker 1: drive tens of billions of dollars of revenue over the

174
00:10:34,640 --> 00:10:38,319
Speaker 1: next several years, adding that virtually every consumer business Amazon

175
00:10:38,360 --> 00:10:42,160
Speaker 1: operated in already had or would have generative AI offerings.

176
00:10:42,720 --> 00:10:46,080
Speaker 1: Now they can just say that stuff. I really want

177
00:10:46,080 --> 00:10:47,520
Speaker 1: you to know. You can say what you want on

178
00:10:47,559 --> 00:10:50,520
Speaker 1: earning scores, as long as you're not just outright lying,

179
00:10:50,559 --> 00:10:52,679
Speaker 1: like saying we have one hundred billion dollars in cash,

180
00:10:52,679 --> 00:10:54,760
Speaker 1: but you have fifty dollars. That is a lie. You

181
00:10:54,800 --> 00:10:57,000
Speaker 1: can't do that. But you can be like iah At

182
00:10:57,080 --> 00:10:59,200
Speaker 1: We've got all sorts of AI and everything. Now it's

183
00:10:59,240 --> 00:11:02,319
Speaker 1: bloody magically. Take you don't take a look, but it's there.

184
00:11:02,320 --> 00:11:05,120
Speaker 1: I promise you. They can just say what they want.

185
00:11:06,320 --> 00:11:09,760
Speaker 1: But don't worry, they're not the only ones. Salesforce chief

186
00:11:09,760 --> 00:11:13,080
Speaker 1: financial officer Amy Weaver said in their most recent earning

187
00:11:13,120 --> 00:11:17,360
Speaker 1: score that Salesforce was not factoring in material contribution from

188
00:11:17,400 --> 00:11:21,240
Speaker 1: Salesforce's numerous AI products in its financial year twenty twenty five.

189
00:11:21,280 --> 00:11:26,079
Speaker 1: Guidance software company Adobe shares slid in their last earnings.

190
00:11:26,120 --> 00:11:28,680
Speaker 1: It's the company failed to generate meaningful revenue from its

191
00:11:28,679 --> 00:11:31,600
Speaker 1: masses of AI products, with analysts now worried about its

192
00:11:31,640 --> 00:11:36,000
Speaker 1: ability to actually monetize any of these generative products. Service

193
00:11:36,040 --> 00:11:38,720
Speaker 1: now claimed to its earnings that generative AI was meaningfully

194
00:11:38,720 --> 00:11:40,840
Speaker 1: contributed to its bottom line. Yet a story from The

195
00:11:40,840 --> 00:11:44,640
Speaker 1: Information quotes their chief financial officer as saying that from

196
00:11:44,640 --> 00:11:48,760
Speaker 1: a revenue contribution perspective, it's not going to be huge.

197
00:11:49,679 --> 00:11:51,400
Speaker 1: Going to be a bit honest and feeling a little

198
00:11:51,440 --> 00:11:55,440
Speaker 1: insane with this stuff. I feel crazy every time I

199
00:11:55,520 --> 00:11:59,920
Speaker 1: think about these stories, because elsewhere in the media, so

200
00:12:00,360 --> 00:12:03,960
Speaker 1: many people are saying how big and successful the generative

201
00:12:03,960 --> 00:12:07,600
Speaker 1: AI revolution is, and it's going to be Yeah. Every

202
00:12:07,600 --> 00:12:09,920
Speaker 1: time I look at the actual places where they write

203
00:12:09,960 --> 00:12:12,680
Speaker 1: down how much money it makes, any of the actual

204
00:12:12,720 --> 00:12:19,400
Speaker 1: signs of growth and significance and utility and adoption, it's

205
00:12:19,440 --> 00:12:22,640
Speaker 1: just not there. It's just breathless hype, with this kind

206
00:12:22,640 --> 00:12:27,720
Speaker 1: of whisper of stagnation and non existent adoption. And while

207
00:12:27,760 --> 00:12:31,200
Speaker 1: there are startups beginning to mind usefulness out of general AI,

208
00:12:31,520 --> 00:12:34,360
Speaker 1: and they do inside by automating internal queries and customer

209
00:12:34,360 --> 00:12:39,920
Speaker 1: sport questions, these are integrations rather than revolutions, and they're

210
00:12:39,920 --> 00:12:44,640
Speaker 1: far from the substance of a true movement. Maybe the

211
00:12:44,760 --> 00:12:47,280
Speaker 1: darker truth of the generative AI boom is that it's

212
00:12:47,320 --> 00:12:50,480
Speaker 1: a feature, not a product, and that these features might

213
00:12:50,520 --> 00:12:54,440
Speaker 1: be built entirely off the back of large language models

214
00:12:54,480 --> 00:12:59,319
Speaker 1: which are unsustainable to run, grow, or even make better.

215
00:13:00,480 --> 00:13:03,360
Speaker 1: What if AI only drives a couple percentage points of

216
00:13:03,400 --> 00:13:06,160
Speaker 1: real revenue growth of these companies. What if what we're

217
00:13:06,160 --> 00:13:10,640
Speaker 1: seeing today is the upper limit, not the beginning. Honestly,

218
00:13:10,679 --> 00:13:12,760
Speaker 1: I'm beginning to believe that a large part of the

219
00:13:12,760 --> 00:13:15,560
Speaker 1: AI boom is just hot air, and it's being pumped

220
00:13:15,640 --> 00:13:20,319
Speaker 1: up through a combination of executive bullshittery and very compliant

221
00:13:20,360 --> 00:13:23,280
Speaker 1: media that's so happy to write stories imagining what AI

222
00:13:23,360 --> 00:13:27,240
Speaker 1: can do, yet seems unable to check what it can

223
00:13:27,360 --> 00:13:33,160
Speaker 1: do or what it's doing. It's so weird. Now, there's

224
00:13:33,200 --> 00:13:35,360
Speaker 1: a bloke over at the Wall Street Journal called Chipcutter

225
00:13:35,360 --> 00:13:36,959
Speaker 1: who should really look into if you want to know

226
00:13:37,000 --> 00:13:38,840
Speaker 1: why your boss keeps asking you to go back to

227
00:13:38,840 --> 00:13:41,400
Speaker 1: the office. Wall Street Journal's Chip Cutter. He loves to

228
00:13:41,400 --> 00:13:44,400
Speaker 1: write things about how bosses are good and how returning

229
00:13:44,440 --> 00:13:46,760
Speaker 1: to the office is good. He wrote a piece in

230
00:13:46,840 --> 00:13:49,600
Speaker 1: March about how AI is being integrated into the office,

231
00:13:49,640 --> 00:13:51,719
Speaker 1: and most of it was just hundreds of words of

232
00:13:52,000 --> 00:13:54,880
Speaker 1: him guessing about what people might do but when he

233
00:13:54,920 --> 00:13:57,680
Speaker 1: gets to the bottom and he starts talking about companies

234
00:13:57,760 --> 00:14:02,040
Speaker 1: using it, it's almost entirely exact samples of people saying, yeah,

235
00:14:02,040 --> 00:14:04,000
Speaker 1: it makes too many mistakes for us to rely on it,

236
00:14:04,040 --> 00:14:07,400
Speaker 1: and we're just experimenting it. Elsewhere in the media, the

237
00:14:07,440 --> 00:14:10,000
Speaker 1: New York Times talked with Salesforce is head of AI

238
00:14:10,440 --> 00:14:14,599
Speaker 1: Clara's share and in this I think six hundred and

239
00:14:14,640 --> 00:14:18,120
Speaker 1: seven hundred word article, didn't really get to say much

240
00:14:18,160 --> 00:14:20,880
Speaker 1: of anything about AI or what their products do. All

241
00:14:20,920 --> 00:14:24,840
Speaker 1: she said was that the Einstein Trust layer handles data.

242
00:14:24,960 --> 00:14:26,840
Speaker 1: And you may think I'm being facetious here, that's all

243
00:14:26,880 --> 00:14:29,920
Speaker 1: she said about that, and then she added that it

244
00:14:29,960 --> 00:14:33,280
Speaker 1: would be transformational for jobs the way the internet was.

245
00:14:34,320 --> 00:14:38,120
Speaker 1: What what does that mean? Why am I reading this

246
00:14:38,160 --> 00:14:40,680
Speaker 1: in the newspaper? Why is this what I read in

247
00:14:40,720 --> 00:14:44,840
Speaker 1: the newspaper? How is this helping? I know I rant

248
00:14:44,880 --> 00:14:46,480
Speaker 1: a lot on this podcast, and I'm going to keep

249
00:14:46,480 --> 00:14:49,000
Speaker 1: doing it. You're stuck with me, all right, it's free, Okay,

250
00:14:49,200 --> 00:14:51,360
Speaker 1: you don't pay for this unless you do cooler zone media,

251
00:14:51,360 --> 00:14:54,640
Speaker 1: which you should pay for anyway. I know I'm ranting,

252
00:14:54,680 --> 00:14:58,280
Speaker 1: But the reason that this stuff really infuriates me is

253
00:14:58,480 --> 00:15:01,640
Speaker 1: it's misinformation on some level. I know it's kind of

254
00:15:01,720 --> 00:15:04,720
Speaker 1: dramatic to say, oh, they're misinforming people by suggesting that

255
00:15:04,760 --> 00:15:08,400
Speaker 1: AI can do stuff, but it is. It is misinformation.

256
00:15:08,680 --> 00:15:11,720
Speaker 1: When you're letting corporate executives go in the newspaper and

257
00:15:11,760 --> 00:15:14,680
Speaker 1: talk about how amazing their products will be without asking

258
00:15:14,720 --> 00:15:17,200
Speaker 1: them what they can do today, you're just giving them

259
00:15:17,240 --> 00:15:20,920
Speaker 1: free press. You're not giving them credit for stuff they've done.

260
00:15:21,040 --> 00:15:23,200
Speaker 1: You're giving them credit for things they're making up on

261
00:15:23,280 --> 00:15:26,520
Speaker 1: the spot. And when you do that, you make the

262
00:15:26,680 --> 00:15:30,800
Speaker 1: rich richer and the poor poorer. You centralize power in

263
00:15:30,800 --> 00:15:34,720
Speaker 1: the hands of assholes, people who are excited, people who

264
00:15:34,760 --> 00:15:39,160
Speaker 1: are borderline masturbatory, jumping around saying, oh God, I can't

265
00:15:39,200 --> 00:15:43,840
Speaker 1: wait till and replace humans with fucking computers. Good news

266
00:15:43,920 --> 00:15:45,920
Speaker 1: is they're not going to be able to, but that's

267
00:15:45,960 --> 00:15:47,920
Speaker 1: what they're excited about, and that's what they're getting media

268
00:15:47,960 --> 00:15:52,280
Speaker 1: coverage around. The media has been fooled, just like they

269
00:15:52,320 --> 00:15:56,400
Speaker 1: were with the metaverus, by this specious promise train of

270
00:15:56,440 --> 00:16:01,120
Speaker 1: the generative AI generation, and these worthless executives champions these

271
00:16:01,280 --> 00:16:05,560
Speaker 1: half truths and this magical thinking has spread far faster

272
00:16:05,840 --> 00:16:08,320
Speaker 1: due to the fact that AI actually exists and is

273
00:16:08,400 --> 00:16:11,760
Speaker 1: doing something, and it's actually much easier to imagine how

274
00:16:11,760 --> 00:16:15,000
Speaker 1: it might change our lives, unlike the metaverse. Even if

275
00:16:15,040 --> 00:16:17,640
Speaker 1: the way it might do so is somewhere between improbable

276
00:16:17,680 --> 00:16:22,840
Speaker 1: and impossible, it is easy to think about how my work.

277
00:16:23,320 --> 00:16:25,040
Speaker 1: You know, you could use an AI to or make

278
00:16:25,120 --> 00:16:28,960
Speaker 1: data entry or boring busy work. Surely all of this

279
00:16:29,040 --> 00:16:31,160
Speaker 1: you can automate, right, And when you use chat GPT

280
00:16:31,520 --> 00:16:33,760
Speaker 1: you can almost kind of, sort of somewhat see how

281
00:16:33,800 --> 00:16:36,840
Speaker 1: it might happen. Even if when you open up chat

282
00:16:36,880 --> 00:16:39,720
Speaker 1: GPT and try and make it do something, it's always

283
00:16:39,760 --> 00:16:42,560
Speaker 1: a bit off, never seems to quite do it. I'm

284
00:16:42,600 --> 00:16:44,560
Speaker 1: in a very in my day job, my PR firm.

285
00:16:44,720 --> 00:16:47,760
Speaker 1: I'm in a spreadsheet and document heavy business. Of all

286
00:16:47,800 --> 00:16:50,000
Speaker 1: the people who this could help, you think it would

287
00:16:50,000 --> 00:16:53,040
Speaker 1: be me A lot of my work is hey, all

288
00:16:53,080 --> 00:16:54,760
Speaker 1: of these things I need him in a spreadsheet. It

289
00:16:54,800 --> 00:16:59,480
Speaker 1: can't bloody do it. And I'm sure you listeners will

290
00:16:59,520 --> 00:17:02,200
Speaker 1: probably email me and say, oh, I've used CHATGPT for this,

291
00:17:02,560 --> 00:17:05,879
Speaker 1: don't care. I really mean that this thing is not

292
00:17:06,040 --> 00:17:08,080
Speaker 1: changing the world. And actually I think far more of

293
00:17:08,119 --> 00:17:09,920
Speaker 1: you have already shared. Thank you. By the way, easy.

294
00:17:09,920 --> 00:17:12,080
Speaker 1: At Better off Line dot com. You can email me

295
00:17:12,160 --> 00:17:14,520
Speaker 1: your ideas and your angry comments, so you can go

296
00:17:14,560 --> 00:17:17,879
Speaker 1: on the reddit to complain. But the thing I'm hearing

297
00:17:17,920 --> 00:17:19,879
Speaker 1: for most people is, yeah, I've tried it and it

298
00:17:19,920 --> 00:17:22,920
Speaker 1: didn't do enough. I tried it and there were too

299
00:17:22,920 --> 00:17:26,639
Speaker 1: many mistakes. There was a Wall Street Journal article back

300
00:17:26,680 --> 00:17:31,680
Speaker 1: in February about how Amazon and Google were having trouble

301
00:17:31,760 --> 00:17:36,080
Speaker 1: selling AI services because well, when they went to sell

302
00:17:36,119 --> 00:17:39,720
Speaker 1: them these companies, these financial services companies in particular, they

303
00:17:39,720 --> 00:17:42,399
Speaker 1: were saying, yeah, but these hallucinations could actually get the

304
00:17:42,440 --> 00:17:47,159
Speaker 1: sec mad at us. And the answer that they had was, yeah,

305
00:17:47,480 --> 00:17:49,200
Speaker 1: what if we just made it so that the models

306
00:17:49,200 --> 00:17:53,639
Speaker 1: would sometimes say they don't know stuff. Every time you

307
00:17:53,720 --> 00:17:56,800
Speaker 1: get to a reckoning with AI where you want it

308
00:17:56,840 --> 00:18:00,800
Speaker 1: to be better, where you're like, hey, AI executive, how

309
00:18:00,800 --> 00:18:04,000
Speaker 1: will this actually be fixed these hallucination problems, for example,

310
00:18:04,160 --> 00:18:07,199
Speaker 1: they come up with the most meanly mouthship. And I

311
00:18:07,280 --> 00:18:09,720
Speaker 1: truly believe it's because there is no answer to these problems,

312
00:18:09,760 --> 00:18:13,480
Speaker 1: as I said in the previous episode, and I think

313
00:18:13,520 --> 00:18:16,360
Speaker 1: that's why I can't find any companies that have integrated

314
00:18:16,400 --> 00:18:18,639
Speaker 1: generative AI in a way that's truly improved their bottom

315
00:18:18,720 --> 00:18:22,440
Speaker 1: line other than Klana, which allows you to do zero

316
00:18:22,520 --> 00:18:25,720
Speaker 1: percent interest free loans on almost anything. It's a very

317
00:18:25,760 --> 00:18:29,600
Speaker 1: worrying company anyway. They claimed that their AI powered support

318
00:18:29,680 --> 00:18:34,000
Speaker 1: bot was estimated to drive a forty million dollar amount

319
00:18:34,080 --> 00:18:38,159
Speaker 1: in profit improvement in twenty twenty four, which does not,

320
00:18:38,280 --> 00:18:40,840
Speaker 1: by the way, despite it being troubleted by members of

321
00:18:40,880 --> 00:18:43,640
Speaker 1: the media, otherwise mean that they made forty million dollars

322
00:18:43,680 --> 00:18:47,040
Speaker 1: in profit. I actually can't find what profit improvement refers to.

323
00:18:48,400 --> 00:18:51,840
Speaker 1: And this is the classic AI boom story. By the way,

324
00:18:51,960 --> 00:18:57,120
Speaker 1: there's always this weird verbal judo going on where they're like, yeah, sir,

325
00:18:57,840 --> 00:19:00,800
Speaker 1: forty million dollars in profit improvement, upwards, downwards and side

326
00:19:00,800 --> 00:19:03,760
Speaker 1: to sidemark, it's really good, And I think it's just

327
00:19:03,880 --> 00:19:08,240
Speaker 1: headline grabbing. I think it's just buzz. And despite fears

328
00:19:08,320 --> 00:19:11,600
Speaker 1: to the contrary, AI doesn't appear to be replacing a

329
00:19:11,680 --> 00:19:14,200
Speaker 1: large amount of workers, and when it has, the results

330
00:19:14,240 --> 00:19:17,960
Speaker 1: have been pretty terrible, like when Microsoft replaced MSN dot

331
00:19:18,000 --> 00:19:21,040
Speaker 1: COM's editorial team with a series of AI bots that

332
00:19:21,080 --> 00:19:25,040
Speaker 1: have spread misinformation and conspiracy theories, things like Joe Biden

333
00:19:25,119 --> 00:19:29,440
Speaker 1: falling asleep. It's so weird. Interestingly, There was also a

334
00:19:29,480 --> 00:19:32,480
Speaker 1: study in Boston Consultancy Group, and just as a note,

335
00:19:32,640 --> 00:19:35,520
Speaker 1: if anyone would love the opportunity to just replace workers

336
00:19:35,520 --> 00:19:39,840
Speaker 1: with robots, it's BCG, mckenzi, Accentia. All these companies were

337
00:19:39,880 --> 00:19:43,920
Speaker 1: absolutely they would be giving, however much open ai wanted

338
00:19:43,960 --> 00:19:46,040
Speaker 1: to do that, and then they would charge fifty million

339
00:19:46,080 --> 00:19:47,840
Speaker 1: dollars for in integration that didn't work, which I guess

340
00:19:47,880 --> 00:19:51,000
Speaker 1: makes AI perfect them putting that aside. In a study

341
00:19:51,000 --> 00:19:54,439
Speaker 1: from BCG, they found the consultants that solved business problems

342
00:19:54,480 --> 00:19:57,520
Speaker 1: with open AI's GPT four model performed twenty three percent

343
00:19:57,920 --> 00:20:01,560
Speaker 1: worse than those who didn't use it, even when the

344
00:20:01,640 --> 00:20:04,800
Speaker 1: consultant was warned about the limitations of generative AI and

345
00:20:04,800 --> 00:20:10,920
Speaker 1: the risk of hallucinations. Yeah, really great stuff. To be clear,

346
00:20:11,400 --> 00:20:15,280
Speaker 1: I am not advocating for the replacement of workers with AI. However,

347
00:20:15,840 --> 00:20:18,400
Speaker 1: I'm saying that if it was actually capable of replacing

348
00:20:18,440 --> 00:20:22,120
Speaker 1: human outputs, even if it was even anywhere near doing so,

349
00:20:22,520 --> 00:20:25,480
Speaker 1: any number of these massive, horrifying firms would be doing

350
00:20:25,520 --> 00:20:27,920
Speaker 1: so at scale, and planning to do so more as

351
00:20:27,920 --> 00:20:31,440
Speaker 1: the models improve. They'd be fuddling cash right up. Open

352
00:20:31,480 --> 00:20:35,040
Speaker 1: ayes ass it would be incredible, but the reality of

353
00:20:35,080 --> 00:20:37,920
Speaker 1: the AI boom is kind of a little more boring.

354
00:20:39,000 --> 00:20:42,440
Speaker 1: It recently came out that Amazon's cashless just walk out

355
00:20:42,480 --> 00:20:45,199
Speaker 1: technology in some of their stores. You could walk in,

356
00:20:45,359 --> 00:20:47,320
Speaker 1: scan a QR code, and then you could just grab

357
00:20:47,320 --> 00:20:51,720
Speaker 1: your rouse tomato sauce and your condoms or whatever, your

358
00:20:51,760 --> 00:20:53,560
Speaker 1: weird magazines. I don't know what they sell, and there

359
00:20:53,560 --> 00:20:55,399
Speaker 1: I'm not giving any more money to Amazon than I

360
00:20:55,520 --> 00:20:59,159
Speaker 1: need to. Anyways, everyone thought, oh, it's just AI. You

361
00:20:59,160 --> 00:21:01,399
Speaker 1: could just walk in. It's the cameras would tell you

362
00:21:01,440 --> 00:21:04,320
Speaker 1: through computer vision what you had bought. It would be great.

363
00:21:04,560 --> 00:21:06,760
Speaker 1: Now it turns out that there's one thousand workers in

364
00:21:06,760 --> 00:21:11,680
Speaker 1: India that were monitoring these cameras and approving transactions. Worse still,

365
00:21:12,040 --> 00:21:14,760
Speaker 1: open AI used Kenyan workers who were paid less than

366
00:21:14,800 --> 00:21:19,159
Speaker 1: two dollars an hour to train chet GPT's outputs, and

367
00:21:19,200 --> 00:21:22,280
Speaker 1: they currently pay fifteen dollars an hour. I think for

368
00:21:22,320 --> 00:21:26,680
Speaker 1: American contractors, no benefits of course, you know, fuck workers, right,

369
00:21:27,119 --> 00:21:30,320
Speaker 1: that's the thing underneath this whole thing. It's just this

370
00:21:30,480 --> 00:21:33,680
Speaker 1: undercurrent of disrespect for human beings, and it pisses me off.

371
00:21:34,200 --> 00:21:36,120
Speaker 1: And I realized I'm pissed off about a lot of things.

372
00:21:36,119 --> 00:21:38,720
Speaker 1: You'd be listening for like an hour now half an

373
00:21:38,760 --> 00:21:42,920
Speaker 1: hour in this episode. Anyway, I'll keep going. But yeah,

374
00:21:43,000 --> 00:21:46,120
Speaker 1: like I said, if AI was changing things, if AI

375
00:21:46,280 --> 00:21:50,480
Speaker 1: was actually capable of replacing a person, it would have happened.

376
00:21:50,800 --> 00:21:53,360
Speaker 1: It would be happening right now, It'd be happening at scale.

377
00:21:53,600 --> 00:21:57,879
Speaker 1: It would be so much worse than things feel now, unless,

378
00:21:57,880 --> 00:22:01,720
Speaker 1: of course, it just wasn't possible. What if what we're

379
00:22:01,720 --> 00:22:04,760
Speaker 1: seeing today is not a glimpse of the future, but

380
00:22:04,840 --> 00:22:08,080
Speaker 1: actually the new terms of the present. What if generator

381
00:22:08,119 --> 00:22:11,199
Speaker 1: of AI isn't actually capable of doing much more than

382
00:22:11,200 --> 00:22:13,760
Speaker 1: what we're seeing today. What if there's not really a

383
00:22:13,840 --> 00:22:16,000
Speaker 1: clear timeline when it will actually be able to do more.

384
00:22:17,440 --> 00:22:21,080
Speaker 1: What if this entire hype cycle has been built, goosed,

385
00:22:21,600 --> 00:22:24,760
Speaker 1: and propped up by this compliant media, ready and willing

386
00:22:24,800 --> 00:22:29,000
Speaker 1: to take whatever these career embellishing bullshitters have to say.

387
00:22:29,920 --> 00:22:32,359
Speaker 1: What if this is just another metaverse, but with a

388
00:22:32,400 --> 00:22:36,360
Speaker 1: little bit more product. Every single time I've read about

389
00:22:36,400 --> 00:22:40,280
Speaker 1: the amazing things that artificial intelligence can do, I just

390
00:22:40,359 --> 00:22:42,800
Speaker 1: see somebody attempting to add fuel to a fire that's

391
00:22:42,800 --> 00:22:45,600
Speaker 1: close to going out. When the Wall Street Journals Joanna

392
00:22:45,640 --> 00:22:48,439
Speaker 1: Stern wrote about Sora open ayes yet to be released

393
00:22:48,520 --> 00:22:52,359
Speaker 1: video generating chatbot. She talked about how its photorealistic clips

394
00:22:52,359 --> 00:22:55,160
Speaker 1: were good enough to freak her out. And I get

395
00:22:55,200 --> 00:23:00,080
Speaker 1: it at first glance. These do look like people, these

396
00:23:00,119 --> 00:23:04,520
Speaker 1: images do. They look like something approaching a video. They

397
00:23:04,560 --> 00:23:09,439
Speaker 1: look almost real, kind of like text some chat GBT

398
00:23:09,640 --> 00:23:13,400
Speaker 1: is almost right or it's right fully, but it doesn't

399
00:23:13,880 --> 00:23:18,040
Speaker 1: feel right. But much like the rest of these outputs,

400
00:23:19,480 --> 00:23:21,600
Speaker 1: you look a little closer, and they have these weird

401
00:23:21,680 --> 00:23:24,240
Speaker 1: errors like cars disappearing in and out of the shot,

402
00:23:24,480 --> 00:23:27,840
Speaker 1: or a different car coming out from behind something, or

403
00:23:27,880 --> 00:23:32,600
Speaker 1: completely different images between frames. Are these strange, unrealistic moments

404
00:23:32,600 --> 00:23:37,439
Speaker 1: of lighting, and they're never much longer than thirty seconds. Stern,

405
00:23:37,520 --> 00:23:40,679
Speaker 1: who by the way, I deeply respect, isn't really afraid

406
00:23:40,720 --> 00:23:43,680
Speaker 1: of what Sura can do. But what would happen if

407
00:23:43,720 --> 00:23:46,600
Speaker 1: open ai was able to fix the hallucination problems that

408
00:23:46,720 --> 00:23:51,600
Speaker 1: makes these videos kind of unwatchable. Well, it's easy to

409
00:23:51,600 --> 00:23:54,119
Speaker 1: imagine tools like Saura could eventually play a role in

410
00:23:54,160 --> 00:23:58,639
Speaker 1: online disinformation campaigns, churning out like lifelike videos of politicians

411
00:23:58,680 --> 00:24:01,960
Speaker 1: saying or doing appalling things. We can all breathe a

412
00:24:02,000 --> 00:24:04,520
Speaker 1: sigh of relief in knowing that the videos themselves are

413
00:24:04,560 --> 00:24:07,960
Speaker 1: often so flawed. You can pretty much instantly see their

414
00:24:07,960 --> 00:24:11,320
Speaker 1: AI generated Also, SRA is not available to the public yet,

415
00:24:11,440 --> 00:24:13,199
Speaker 1: and I don't even know if it ever will be.

416
00:24:14,240 --> 00:24:16,960
Speaker 1: You just need to look at the hands or the backgrounds.

417
00:24:17,640 --> 00:24:19,920
Speaker 1: Look at the people in the background of any AI

418
00:24:20,040 --> 00:24:25,320
Speaker 1: generated photo or video. They often contain too many fingers

419
00:24:25,359 --> 00:24:29,680
Speaker 1: so you can't see their faces. Or in Sora's videos,

420
00:24:29,760 --> 00:24:34,040
Speaker 1: their legs don't look right. It's so weird and I

421
00:24:34,200 --> 00:24:37,840
Speaker 1: don't I don't know how to put it perfectly, but

422
00:24:38,520 --> 00:24:43,320
Speaker 1: they don't feel human. Just to be clear, though Sora

423
00:24:43,520 --> 00:24:46,920
Speaker 1: is dead on arrival, no one actually has access to it.

424
00:24:46,920 --> 00:24:49,320
Speaker 1: It's unclear when it will come out. Every journalist that

425
00:24:49,320 --> 00:24:52,880
Speaker 1: has quote unquote use Sura has just given a prompt

426
00:24:52,920 --> 00:24:56,720
Speaker 1: to open ai to run. But there's also a very

427
00:24:56,760 --> 00:24:59,160
Speaker 1: obvious problem that kind of relates to something I mentioned

428
00:24:59,200 --> 00:25:02,680
Speaker 1: in the previous episode. Open AI and every generative AI company,

429
00:25:03,200 --> 00:25:06,240
Speaker 1: they're all dependent on high quality data to train the models,

430
00:25:06,600 --> 00:25:10,640
Speaker 1: and video data is so much larger, more complex, and

431
00:25:11,720 --> 00:25:15,080
Speaker 1: harder to find. There's less of it because it's visual media,

432
00:25:16,800 --> 00:25:20,800
Speaker 1: and it's just a much bigger, more complex model and

433
00:25:20,840 --> 00:25:25,320
Speaker 1: a much harder computational task to create video moving image

434
00:25:25,440 --> 00:25:29,320
Speaker 1: is it's actually kind of putting aside my anger about

435
00:25:29,359 --> 00:25:31,800
Speaker 1: Generative AI. It's amazing they've done even this, but to

436
00:25:31,840 --> 00:25:34,440
Speaker 1: be clear, as amazing as it might look, it isn't

437
00:25:34,560 --> 00:25:37,159
Speaker 1: enough to do anything. It's just a kind of a

438
00:25:37,200 --> 00:25:40,760
Speaker 1: do hickey. And this data is so much more complex

439
00:25:40,800 --> 00:25:43,439
Speaker 1: than the text based data that open ai is running

440
00:25:43,480 --> 00:25:47,719
Speaker 1: out of to make CHATGBT spit out words. Even if

441
00:25:47,760 --> 00:25:51,600
Speaker 1: there were enough data, there's pretty good reason why open

442
00:25:51,680 --> 00:25:55,119
Speaker 1: ai is coy about when they'll release the model. Like

443
00:25:55,160 --> 00:25:57,720
Speaker 1: I said, it's expensive and complex to run, and at

444
00:25:57,720 --> 00:26:01,119
Speaker 1: no point has anyone on how the fuck this actually

445
00:26:01,119 --> 00:26:06,200
Speaker 1: makes them any money, how they sell this. It's so weird,

446
00:26:06,800 --> 00:26:09,280
Speaker 1: to be clear, when you use surra it turns text

447
00:26:09,320 --> 00:26:12,000
Speaker 1: prompts into a video. You can't edit the video, you

448
00:26:12,040 --> 00:26:14,080
Speaker 1: can't change the video. The video is what the video

449
00:26:14,160 --> 00:26:17,480
Speaker 1: looks like. There's no way to make Sora make the

450
00:26:17,520 --> 00:26:21,560
Speaker 1: same thing multiple times, which makes the very basics of

451
00:26:22,040 --> 00:26:25,360
Speaker 1: making film, which is multiple angles of the same thing,

452
00:26:25,720 --> 00:26:30,800
Speaker 1: completely goddamn impossible. In fact, consistency between the same two

453
00:26:30,880 --> 00:26:34,840
Speaker 1: prompts is impossible from these models because they're all probabilistic.

454
00:26:36,480 --> 00:26:39,320
Speaker 1: We've recently seen some of the first quote unquote movies

455
00:26:39,359 --> 00:26:42,040
Speaker 1: made with sourra and the first one was called Airhead,

456
00:26:42,160 --> 00:26:45,040
Speaker 1: which is about a minute long. It's this man with

457
00:26:45,080 --> 00:26:47,840
Speaker 1: a balloon head walking around and it's got this it's

458
00:26:47,960 --> 00:26:51,960
Speaker 1: very twee. It sucks. It's just putting aside the aipart.

459
00:26:52,000 --> 00:26:54,920
Speaker 1: It's just crappy, and it's got a guy being like, yeah,

460
00:26:54,960 --> 00:26:57,600
Speaker 1: having a balloon head is difficult. Yeah, it's weird. I

461
00:26:57,640 --> 00:27:00,000
Speaker 1: hate having a balloon head. I hope I don't get popped.

462
00:27:00,240 --> 00:27:04,639
Speaker 1: It sucks. It's really bad filmmaking. But also each shot,

463
00:27:04,640 --> 00:27:07,159
Speaker 1: and there's multiple mbile shots of this guy with a

464
00:27:07,200 --> 00:27:10,080
Speaker 1: yellow balloon head looks completely different. It's a different balloon

465
00:27:10,119 --> 00:27:13,600
Speaker 1: every goddamn time. And it's so funny because you have

466
00:27:13,720 --> 00:27:16,920
Speaker 1: these guys on twittering like, oh my god, oh my god,

467
00:27:17,760 --> 00:27:20,040
Speaker 1: I am crying and pissing myself. This is the best

468
00:27:20,040 --> 00:27:25,639
Speaker 1: thing I've ever seen. But it isn't. It's so close

469
00:27:25,720 --> 00:27:29,200
Speaker 1: yet so far away, And the only reason it's impressive

470
00:27:29,320 --> 00:27:31,600
Speaker 1: is people are willing to sit there and say, but

471
00:27:31,680 --> 00:27:36,359
Speaker 1: what if it wasn't shit? But it is it really

472
00:27:36,400 --> 00:27:41,920
Speaker 1: is like every other generative output. It's superficially impressive, kind

473
00:27:41,920 --> 00:27:44,520
Speaker 1: of sort of lifelike, but once you look at it

474
00:27:44,560 --> 00:27:49,520
Speaker 1: for more than a moment, it's just flawed, terribly, irrevocably flawed.

475
00:27:51,200 --> 00:27:55,000
Speaker 1: It's time to wake up. We are not in the

476
00:27:55,040 --> 00:27:58,160
Speaker 1: early days of AI. We're decades in and we're approaching

477
00:27:58,200 --> 00:28:01,239
Speaker 1: the top of the s curve of innovation. There are

478
00:28:01,320 --> 00:28:05,480
Speaker 1: products being built, don't worry, but it's all things like

479
00:28:05,880 --> 00:28:09,080
Speaker 1: Claude Author, which creator Matt Schumer calls a chain of

480
00:28:09,119 --> 00:28:11,240
Speaker 1: AI systems that will write an entire book for you

481
00:28:11,359 --> 00:28:14,200
Speaker 1: in minutes, and I call a new kind of asshole

482
00:28:14,240 --> 00:28:18,959
Speaker 1: that can shit more than you'd ever believe. Generative AI

483
00:28:19,240 --> 00:28:21,960
Speaker 1: is the ugliest creation of the rot economy, and its

484
00:28:22,000 --> 00:28:24,639
Speaker 1: main selling point is that can generate a great deal

485
00:28:24,680 --> 00:28:29,679
Speaker 1: of passable material. Images generated from generative AI models like

486
00:28:29,800 --> 00:28:32,760
Speaker 1: open Ayes Dali all have the same kind of eerie

487
00:28:32,840 --> 00:28:35,840
Speaker 1: feel to them, as they're mostly trained on the same data,

488
00:28:35,880 --> 00:28:38,440
Speaker 1: some of it licensed from shutstock, some of it outright

489
00:28:38,480 --> 00:28:43,600
Speaker 1: plagiarized from hundreds of artists. Without sounding too wanky and philosophical,

490
00:28:44,240 --> 00:28:47,880
Speaker 1: everything created by generative AI feels soulless, and that's because

491
00:28:47,920 --> 00:28:50,640
Speaker 1: it is no matter how detailed the problem, no matter

492
00:28:50,640 --> 00:28:53,600
Speaker 1: how well trained the model, no matter how well intentioned

493
00:28:53,600 --> 00:28:57,680
Speaker 1: the person writing that prompt these are still mathematical solutions

494
00:28:57,720 --> 00:29:01,960
Speaker 1: to the emotional problem of creation. One cannot recreate the

495
00:29:02,000 --> 00:29:05,000
Speaker 1: subtle fuck ups and delightful little neurological errors that make

496
00:29:05,040 --> 00:29:08,120
Speaker 1: writing a book, or a newsletter or a podcast special.

497
00:29:08,720 --> 00:29:11,720
Speaker 1: While this podcast is admittedly trying to generate what I

498
00:29:11,800 --> 00:29:15,320
Speaker 1: believe AI might do in the future, it's not generative,

499
00:29:15,320 --> 00:29:18,000
Speaker 1: and it's not generated as a result of me mathematically

500
00:29:18,000 --> 00:29:22,120
Speaker 1: considering how likely an outcome is. My fury is not

501
00:29:22,280 --> 00:29:24,800
Speaker 1: generated by an algorithm telling me that this is the

502
00:29:24,880 --> 00:29:27,560
Speaker 1: right thing to be angry at. I'm pissed off because

503
00:29:27,560 --> 00:29:29,880
Speaker 1: I feel like we're all being like to and treated

504
00:29:29,920 --> 00:29:34,280
Speaker 1: like idiots. What makes things created by humans special isn't

505
00:29:34,280 --> 00:29:36,520
Speaker 1: doing the right thing or the best thing, but the

506
00:29:36,560 --> 00:29:40,160
Speaker 1: outputs that result in us fighting past are on imperfections

507
00:29:40,320 --> 00:29:43,320
Speaker 1: and maladies like the strep infection I've been fighting for

508
00:29:43,360 --> 00:29:47,680
Speaker 1: the last few days, and like, look, to my knowledge,

509
00:29:47,680 --> 00:29:50,239
Speaker 1: you can't give a generative AI strep throat. But if

510
00:29:50,280 --> 00:29:52,080
Speaker 1: I ever find out it's possible, I will make it

511
00:29:52,160 --> 00:30:08,320
Speaker 1: my damn mission to give it to chat GBT. All

512
00:30:08,400 --> 00:30:11,760
Speaker 1: of this hype is predicated on solving problems with artificial

513
00:30:11,800 --> 00:30:15,400
Speaker 1: intelligence models that are only getting worse, and open AI's

514
00:30:15,480 --> 00:30:19,280
Speaker 1: only answers to these problems are a combination of will

515
00:30:19,280 --> 00:30:21,960
Speaker 1: work it out eventually, trust me, and we need a

516
00:30:22,000 --> 00:30:25,920
Speaker 1: technological breakthrough in both chips and energy. That's why Sam

517
00:30:25,960 --> 00:30:29,000
Speaker 1: Moultman has been trying to raise seven trillion dollars and

518
00:30:29,040 --> 00:30:31,320
Speaker 1: that's not a mistake, by the way, to make a

519
00:30:31,360 --> 00:30:34,480
Speaker 1: new kind of AI chip, because there's no sign that

520
00:30:34,560 --> 00:30:38,800
Speaker 1: this or even future generations of chips will actually fix anything.

521
00:30:40,520 --> 00:30:44,920
Speaker 1: Generative AI's core problems, it's hallucinations, its massive energy, and

522
00:30:44,960 --> 00:30:49,520
Speaker 1: its massive unprofitable compute demands are not close to being solved.

523
00:30:50,360 --> 00:30:53,480
Speaker 1: I've now watched a frankly alarming amount of interviews with

524
00:30:53,520 --> 00:30:57,240
Speaker 1: both open AI CEO Sam Moultman and their CTO mirror Marati,

525
00:30:57,480 --> 00:31:00,800
Speaker 1: and every time they're saying the same specious empty talking points,

526
00:31:01,120 --> 00:31:04,080
Speaker 1: promising that in the future chap GPT will do this

527
00:31:04,120 --> 00:31:06,880
Speaker 1: and that as all evidence points to their models getting

528
00:31:06,880 --> 00:31:09,320
Speaker 1: worse and through the last years, by the way, they've

529
00:31:09,360 --> 00:31:12,240
Speaker 1: just said the same thing in every interview. They always

530
00:31:12,280 --> 00:31:17,080
Speaker 1: talk about chat GPT being like something or help creatives,

531
00:31:17,120 --> 00:31:19,280
Speaker 1: they never really say how, which just kind of weird.

532
00:31:20,240 --> 00:31:23,640
Speaker 1: But yeah, generative AI models they're expensive, they're compute intensive,

533
00:31:24,240 --> 00:31:27,080
Speaker 1: and they don't seem to provide obvious, tangible, mass market

534
00:31:27,160 --> 00:31:31,280
Speaker 1: use cases. Marathi in Oltmund's futures depend heavily on keeping

535
00:31:31,280 --> 00:31:34,000
Speaker 1: the world believing that development and improvement of these models

536
00:31:34,040 --> 00:31:38,720
Speaker 1: capabilities will continue at this rapacious pace of progress, even

537
00:31:38,800 --> 00:31:43,440
Speaker 1: though it's unquestionably slowed, with open AI even admitting themselves

538
00:31:43,520 --> 00:31:46,440
Speaker 1: that their latest model, GPT four, may actually be worse

539
00:31:46,560 --> 00:31:50,120
Speaker 1: at some tasks. I study from UC Berkeley last year

540
00:31:50,800 --> 00:31:53,200
Speaker 1: found that GPT four was actually worse at coding them

541
00:31:53,200 --> 00:31:57,200
Speaker 1: before and that chat GPT was at times refusing to

542
00:31:57,320 --> 00:32:04,040
Speaker 1: do certain tasks. Nobody wants to work anymore. Well, I

543
00:32:04,080 --> 00:32:07,120
Speaker 1: feel like I'm walking down the street telling people their

544
00:32:07,160 --> 00:32:10,520
Speaker 1: houses are on fire, only to be told to stop

545
00:32:10,520 --> 00:32:14,640
Speaker 1: insulting their new heating system. These models aren't intelligent. They're

546
00:32:14,640 --> 00:32:18,760
Speaker 1: mathematical behemoths generating the best guess on training, data and labeling,

547
00:32:19,400 --> 00:32:21,440
Speaker 1: and thus they don't really know what they're being asked

548
00:32:21,440 --> 00:32:26,280
Speaker 1: to do. You can't fix that. You can't fix hallucinations.

549
00:32:27,320 --> 00:32:31,600
Speaker 1: You can't just make these problems go away with more compute,

550
00:32:31,720 --> 00:32:35,800
Speaker 1: you can mitigate them. The current philosophy, by the way,

551
00:32:35,840 --> 00:32:37,719
Speaker 1: is that you can use another model to look at

552
00:32:37,760 --> 00:32:40,760
Speaker 1: another model's outputs, which, as I mentioned in the previous episodes,

553
00:32:40,840 --> 00:32:44,880
Speaker 1: is very silly. But seriously, everyone telling you hallucinations are

554
00:32:44,880 --> 00:32:47,520
Speaker 1: going away, look a little deeper and look at when

555
00:32:47,720 --> 00:32:50,560
Speaker 1: they actually failed to tell you how they were. It's

556
00:32:50,600 --> 00:32:53,760
Speaker 1: just very silly. Look. Every bit of excitement for this

557
00:32:53,800 --> 00:32:55,960
Speaker 1: technology right now is based on this idea of what

558
00:32:56,040 --> 00:32:59,120
Speaker 1: it might do, as I've said, and that quickly gets

559
00:32:59,120 --> 00:33:02,360
Speaker 1: conflated with what it could do, which allows Sam Mortman,

560
00:33:02,640 --> 00:33:04,880
Speaker 1: who by the way, is far more of a marketing

561
00:33:04,920 --> 00:33:08,640
Speaker 1: person than an engineer. His one startup, Looped was a failure.

562
00:33:09,320 --> 00:33:11,520
Speaker 1: He's failed upwards. He was in Why Combinat he did this.

563
00:33:11,520 --> 00:33:15,400
Speaker 1: It's actually ridiculous. He's so famous. All of this bullshit

564
00:33:15,480 --> 00:33:17,800
Speaker 1: it allows him to sell the dream of open AI,

565
00:33:18,240 --> 00:33:20,840
Speaker 1: and he's selling it based on the least specific promises

566
00:33:20,880 --> 00:33:23,440
Speaker 1: I've seen since Mark Zuckerberg said we'd live in our

567
00:33:23,480 --> 00:33:28,120
Speaker 1: bloody oculus headsets. And it's frustrating because this money and

568
00:33:28,160 --> 00:33:31,800
Speaker 1: this attention could go to important things. We have real

569
00:33:31,840 --> 00:33:36,720
Speaker 1: problems in society. I believe that Sam Moortman and pretty

570
00:33:36,800 --> 00:33:39,560
Speaker 1: much anyone in a position of power and influence in

571
00:33:39,600 --> 00:33:42,520
Speaker 1: the AI space has been tap dancing this entire time,

572
00:33:42,880 --> 00:33:45,800
Speaker 1: hoping that he could amass enough power and revenue that

573
00:33:45,880 --> 00:33:49,880
Speaker 1: his success would be inevitable. Yet I think his hype

574
00:33:49,880 --> 00:33:53,280
Speaker 1: campaign has been a little bit too successful, and it's

575
00:33:53,280 --> 00:33:56,240
Speaker 1: deeply specious, and he, along with the rest of the

576
00:33:56,320 --> 00:33:59,640
Speaker 1: AI industry, has found himself suddenly having to deliver a

577
00:33:59,640 --> 00:34:03,560
Speaker 1: future these not even close to developing. I am always

578
00:34:03,640 --> 00:34:06,640
Speaker 1: scared of automation taking our jobs. I think it's always

579
00:34:06,640 --> 00:34:08,759
Speaker 1: worth being scared of. But I don't think that's the

580
00:34:08,760 --> 00:34:11,960
Speaker 1: thing the tech industry is working on right now. I

581
00:34:12,000 --> 00:34:15,040
Speaker 1: don't think they're close, and I think there's something more

582
00:34:15,080 --> 00:34:18,880
Speaker 1: imminent to fear, and that thing is the bottom falling

583
00:34:18,920 --> 00:34:21,640
Speaker 1: out of generative AI as companies realize that the best

584
00:34:21,640 --> 00:34:23,840
Speaker 1: they're going to see is maybe a few digits of

585
00:34:23,840 --> 00:34:29,200
Speaker 1: profit growth. Companies like Nvidia, Google, Amazon, Snowflake, and Microsoft,

586
00:34:29,640 --> 00:34:32,719
Speaker 1: they have hundreds of billions of dollars of market capitalization

587
00:34:32,800 --> 00:34:35,520
Speaker 1: as well as expected revenue new growth tied into the

588
00:34:35,560 --> 00:34:39,040
Speaker 1: idea that everyone's going to integrate AI into everything, and

589
00:34:39,080 --> 00:34:42,759
Speaker 1: that they'll be doing more than they are today. You

590
00:34:42,760 --> 00:34:46,520
Speaker 1: can already see the desperation coming from these companies, like Microsoft,

591
00:34:46,600 --> 00:34:49,560
Speaker 1: for example, which in March effectively absorbed a company called

592
00:34:49,560 --> 00:34:53,200
Speaker 1: Inflection AI into itself, kind of an acquisition by Stealth.

593
00:34:53,719 --> 00:34:56,640
Speaker 1: Inflection AI is a public benefit company that portrays itself

594
00:34:56,640 --> 00:35:00,160
Speaker 1: as a nicer, gentler version of open AI. It's called product,

595
00:35:00,200 --> 00:35:04,160
Speaker 1: a chat GPT style chatbot touts its empathetic tone, its humor,

596
00:35:04,200 --> 00:35:07,640
Speaker 1: and its emotional awareness. Inflection was created in twenty twenty

597
00:35:07,680 --> 00:35:10,120
Speaker 1: two with an all star founding team that included Reid Hoffmann,

598
00:35:10,280 --> 00:35:12,920
Speaker 1: the founder of LinkedIn, and most of Fars Suleiman, the

599
00:35:12,920 --> 00:35:15,920
Speaker 1: British born co founder of deep Mind, which Google acquired

600
00:35:15,960 --> 00:35:19,399
Speaker 1: in twenty fourteen. In mid twenty twenty three, Microsoft took

601
00:35:19,400 --> 00:35:21,759
Speaker 1: part in a one point three billion dollar funding round

602
00:35:21,920 --> 00:35:24,400
Speaker 1: which saw the company acquire a significant stake in Inflection,

603
00:35:24,640 --> 00:35:29,359
Speaker 1: alongside other AI players like Nvidia. Inflection's core product has

604
00:35:29,400 --> 00:35:33,720
Speaker 1: the same inherent underlying issues as every other generative AI product, Hallucinations,

605
00:35:33,760 --> 00:35:36,920
Speaker 1: for example, but it has an accomplished team that has

606
00:35:36,960 --> 00:35:40,520
Speaker 1: taken a different approach to its competitors. Whereas chat GPT

607
00:35:40,640 --> 00:35:43,040
Speaker 1: and clored two tend to be, or at least aspire

608
00:35:43,080 --> 00:35:46,560
Speaker 1: to be, functional tools that provide information or complete tasks.

609
00:35:46,719 --> 00:35:49,120
Speaker 1: Inflections sought to make its product feel a bit more organic.

610
00:35:49,719 --> 00:35:52,359
Speaker 1: For Microsoft, the appeal was obvious. It has so much

611
00:35:52,440 --> 00:35:55,279
Speaker 1: riding on its AI ambitions, but in terms of money

612
00:35:55,320 --> 00:35:57,680
Speaker 1: spent as well as its share price, that it can't

613
00:35:57,680 --> 00:36:01,000
Speaker 1: really afford to appear stagnant or worse as though it

614
00:36:01,040 --> 00:36:04,360
Speaker 1: made a bad bet. Acquiring Inflection would help it maintain

615
00:36:04,400 --> 00:36:07,600
Speaker 1: its image, especially with idiot Wall Street analysts. But here's

616
00:36:07,600 --> 00:36:11,439
Speaker 1: the problem. Microsoft already holds a massive stake in Open AI,

617
00:36:11,680 --> 00:36:14,840
Speaker 1: and regulators both in America and Europe are wary of

618
00:36:14,880 --> 00:36:20,080
Speaker 1: market consolidation. Acquiring Inflection it'd give them a little too

619
00:36:20,160 --> 00:36:25,400
Speaker 1: much scrutiny. So Microsoft took a third, nastier path. Instead

620
00:36:25,440 --> 00:36:27,880
Speaker 1: of buying the company, it bought the employees, with Suliman

621
00:36:28,000 --> 00:36:30,239
Speaker 1: and the majority of his coworkers jumping ship to found

622
00:36:30,280 --> 00:36:35,520
Speaker 1: Microsoft's new AI division. It secured the talent a subsequent

623
00:36:35,560 --> 00:36:38,960
Speaker 1: six hundred and fifty million dollar licensing deal, yet another

624
00:36:39,000 --> 00:36:44,239
Speaker 1: example of Microsoft basically paying itself, and then gave that

625
00:36:44,280 --> 00:36:46,359
Speaker 1: deal to the shell of Inflection. You know, the one

626
00:36:46,360 --> 00:36:48,960
Speaker 1: without any of the staff left, giving it access to

627
00:36:49,000 --> 00:36:53,359
Speaker 1: the company's tech its IP and there's nothing regulators could

628
00:36:53,360 --> 00:36:55,960
Speaker 1: do to stop him. To be clear, Microsoft is in

629
00:36:55,960 --> 00:36:58,040
Speaker 1: a position where it could easily absorb the shock wave

630
00:36:58,080 --> 00:37:00,919
Speaker 1: of a potential AI bubble burst. It still prints money

631
00:37:00,920 --> 00:37:04,040
Speaker 1: from its other business units like office and cloud computing,

632
00:37:04,120 --> 00:37:08,000
Speaker 1: Microsoft Windows, and the Xbox gaming system, and the same

633
00:37:08,080 --> 00:37:10,200
Speaker 1: is true for the other big names like Google and Nvidia.

634
00:37:10,680 --> 00:37:13,680
Speaker 1: They're well insulated for any slowdown in AI investment or

635
00:37:13,680 --> 00:37:18,760
Speaker 1: from a growing apathy towards AI enterprise customers. I will note, however,

636
00:37:19,560 --> 00:37:23,120
Speaker 1: these massive investments in data centers, if they're all for nought,

637
00:37:23,920 --> 00:37:27,919
Speaker 1: you will see a form of crash. I can't say

638
00:37:27,920 --> 00:37:30,759
Speaker 1: the same for startups, though other companies aren't going to

639
00:37:30,760 --> 00:37:34,600
Speaker 1: be so lucky. Stability AAI, the developer of stable diffusion,

640
00:37:34,680 --> 00:37:37,400
Speaker 1: a generative AI that can produce images from written prompts

641
00:37:37,880 --> 00:37:40,440
Speaker 1: innovative for the time, is perhaps the canary in the

642
00:37:40,440 --> 00:37:44,200
Speaker 1: coal mine of PKI. Stability AI rode the same waves

643
00:37:44,239 --> 00:37:46,799
Speaker 1: as Open AI, especially in twenty twenty three, but now

644
00:37:46,800 --> 00:37:49,800
Speaker 1: that money is tighter and skepticism is higher, it's struggling

645
00:37:49,880 --> 00:37:53,120
Speaker 1: to stay aflow. Although the company raised one hundred million dollars.

646
00:37:53,160 --> 00:37:55,799
Speaker 1: In early twenty twenty three, it burned through nearly eight

647
00:37:55,840 --> 00:37:58,360
Speaker 1: million dollars a month, and in a recent attempt to

648
00:37:58,440 --> 00:38:02,759
Speaker 1: raise further cash they fail. The company routinely missed payroll and,

649
00:38:02,760 --> 00:38:05,399
Speaker 1: according to Forbes, a master sizeable debt with the US

650
00:38:05,480 --> 00:38:08,640
Speaker 1: tax authorities that culminated in threats to seize the company's assets.

651
00:38:09,400 --> 00:38:11,920
Speaker 1: They owed debts to Amazon, Google and core Weave, and

652
00:38:12,000 --> 00:38:16,920
Speaker 1: each compute provider that specializes in AI applications. With negligible

653
00:38:16,960 --> 00:38:20,319
Speaker 1: revenues and rapid cash burn combined with no obvious way

654
00:38:20,360 --> 00:38:23,640
Speaker 1: to monetize the product, stability Ai is now in turmoil,

655
00:38:23,760 --> 00:38:26,400
Speaker 1: with its key talent leaving the company in March, followed

656
00:38:26,400 --> 00:38:30,040
Speaker 1: by the company's CEO and co founder, Amount Mustak. Its

657
00:38:30,040 --> 00:38:32,960
Speaker 1: ongoing existence is in question, with The Financial Times writing

658
00:38:33,000 --> 00:38:35,759
Speaker 1: in March that the company's future, despite once being seen

659
00:38:35,760 --> 00:38:38,680
Speaker 1: as among the world's most promising startups, is in doubt.

660
00:38:40,200 --> 00:38:42,360
Speaker 1: While it would be fair to say that stability Aii

661
00:38:42,800 --> 00:38:46,640
Speaker 1: was unique at its internal turmoil, its external pressures, the

662
00:38:46,680 --> 00:38:49,320
Speaker 1: ability you or lack thereof, to monetize an expensive product,

663
00:38:49,440 --> 00:38:52,200
Speaker 1: and its reliance on external funding to survive much more

664
00:38:52,239 --> 00:38:56,040
Speaker 1: common across the industry. Its survival depended on investors believing

665
00:38:56,040 --> 00:38:58,919
Speaker 1: in a lofty future for AI, where it's integrated into

666
00:38:58,960 --> 00:39:01,319
Speaker 1: every facet of our lives and it plays a role

667
00:39:01,360 --> 00:39:03,920
Speaker 1: in almost every industry, which, of course we now know

668
00:39:04,000 --> 00:39:08,160
Speaker 1: it doesn't. While that belief hasn't been shattered, or at

669
00:39:08,200 --> 00:39:11,319
Speaker 1: least not yet, it's fair to say that expectations and

670
00:39:11,360 --> 00:39:15,279
Speaker 1: aspirations are increasingly tempered after reaching the apex of the

671
00:39:15,320 --> 00:39:18,440
Speaker 1: AI pissfest. The tech industry is getting a hangover, and

672
00:39:18,560 --> 00:39:22,800
Speaker 1: companies like Stability can't survive the headache. But to be clear,

673
00:39:23,160 --> 00:39:25,720
Speaker 1: I am not excited for the AI bubble to pop,

674
00:39:26,440 --> 00:39:28,120
Speaker 1: and on some level, as weird as it sounds, I

675
00:39:28,160 --> 00:39:31,600
Speaker 1: kind of hope it doesn't. Once it bursts, the AI

676
00:39:31,680 --> 00:39:34,520
Speaker 1: bubble will hit far more than the venture capitalists that

677
00:39:34,560 --> 00:39:38,280
Speaker 1: propped it up. This hype cycle has driven the global

678
00:39:38,320 --> 00:39:41,800
Speaker 1: stock markets to their best first quarter in five years,

679
00:39:42,280 --> 00:39:45,120
Speaker 1: and once the markets fully turn on the companies that

680
00:39:45,520 --> 00:39:48,640
Speaker 1: falsely promised an AI revolution, it's going to lead to

681
00:39:48,680 --> 00:39:52,040
Speaker 1: a massive market downturn and another merciless round of layoffs

682
00:39:52,160 --> 00:39:55,320
Speaker 1: throughout the tech sector, led by Microsoft, Google and Amazon.

683
00:39:56,120 --> 00:39:59,120
Speaker 1: This will in turn suppressed tech labor and flood the

684
00:39:59,160 --> 00:40:01,800
Speaker 1: market with techtile. It's going to suck for everyone involved

685
00:40:01,840 --> 00:40:05,560
Speaker 1: in software. A market crash led by the tech industry

686
00:40:05,600 --> 00:40:09,280
Speaker 1: will only hurt innovation, further draining the already low amounts

687
00:40:09,360 --> 00:40:12,120
Speaker 1: going into the hands of venture capitalists that control the

688
00:40:12,160 --> 00:40:16,480
Speaker 1: dollars going into new startups, and once again, the entire

689
00:40:16,560 --> 00:40:19,520
Speaker 1: industry will suffer because people don't want to build new

690
00:40:19,560 --> 00:40:22,360
Speaker 1: things or try new ideas. No, they want to fund

691
00:40:22,400 --> 00:40:25,560
Speaker 1: the same people doing the same things or similar things

692
00:40:25,600 --> 00:40:27,839
Speaker 1: again and again because it feels good to be part

693
00:40:27,840 --> 00:40:31,480
Speaker 1: of a consensus. Even if you're wrong. Silicon Valley will

694
00:40:31,480 --> 00:40:35,120
Speaker 1: continually fail to innovate at scale until it learns to

695
00:40:35,120 --> 00:40:38,440
Speaker 1: build real things again, things that people actually use, and

696
00:40:38,480 --> 00:40:41,719
Speaker 1: things to actually do something. I don't know if I

697
00:40:41,719 --> 00:40:43,919
Speaker 1: want to be right or wrong here. If I'm wrong,

698
00:40:44,080 --> 00:40:47,200
Speaker 1: generative AI could replace millions of people's jobs, something that

699
00:40:47,239 --> 00:40:49,480
Speaker 1: far too many people in the media are excited about,

700
00:40:49,800 --> 00:40:52,640
Speaker 1: despite the fact that the media is the first industry

701
00:40:52,640 --> 00:40:56,359
Speaker 1: that open AI kind of wants to automate. If I'm right,

702
00:40:56,800 --> 00:40:59,080
Speaker 1: we're going to face a dot com bubble s town

703
00:40:59,200 --> 00:41:01,600
Speaker 1: turn in tech, one that's far worse than what we

704
00:41:01,640 --> 00:41:05,120
Speaker 1: saw in the last few years. In any case, I

705
00:41:05,160 --> 00:41:07,319
Speaker 1: do wish the tech industry would get their heads out

706
00:41:07,320 --> 00:41:11,080
Speaker 1: of their asses. I'm tired. I'm tired of watching tech

707
00:41:11,120 --> 00:41:13,960
Speaker 1: firms life their teeth about the future that will live

708
00:41:14,000 --> 00:41:16,400
Speaker 1: in the metaverse, that our future will be decentralized and

709
00:41:16,440 --> 00:41:19,040
Speaker 1: paid for in cryptocurrency, and that our world will be

710
00:41:19,080 --> 00:41:23,359
Speaker 1: automated with chatbots. I truly think that these companies think

711
00:41:23,440 --> 00:41:26,719
Speaker 1: regular people are stupid, which is why Microsoft put out

712
00:41:26,719 --> 00:41:29,640
Speaker 1: a minute long Super Bowl commercial for their Copilot AI

713
00:41:29,880 --> 00:41:32,600
Speaker 1: that featured several prompts like write the code for my

714
00:41:32,640 --> 00:41:35,160
Speaker 1: three D open world game that don't actually do anything

715
00:41:35,280 --> 00:41:38,000
Speaker 1: that prompt I just mentioned. Go type it into Copilot.

716
00:41:38,120 --> 00:41:40,160
Speaker 1: It will give you a guide to coding a game

717
00:41:40,200 --> 00:41:44,040
Speaker 1: no code created. Also in the commercial, he types in

718
00:41:44,400 --> 00:41:47,960
Speaker 1: classic truck shop called Pauls. But none of these image

719
00:41:48,000 --> 00:41:49,920
Speaker 1: generators can actually do words, so it just looks like

720
00:41:51,280 --> 00:41:54,400
Speaker 1: go and do it. Trust me. It's funny. But every

721
00:41:54,480 --> 00:41:59,600
Speaker 1: time that these big tech booms happen, every time they say, oh,

722
00:41:59,600 --> 00:42:01,560
Speaker 1: we're going to live in the metaverse and oh we're

723
00:42:01,560 --> 00:42:05,640
Speaker 1: going to be able to automate everything, every time they lie,

724
00:42:06,120 --> 00:42:09,680
Speaker 1: the world turns against the tech industry, and this particular

725
00:42:09,719 --> 00:42:13,040
Speaker 1: boom is so craven in its falsehoods that I think

726
00:42:13,080 --> 00:42:17,120
Speaker 1: it'll have a dramatic chilling effect on tech valuations if

727
00:42:17,120 --> 00:42:20,839
Speaker 1: the bubble pops quite as severely as I expect. And

728
00:42:20,920 --> 00:42:24,080
Speaker 1: Sam Mortman desperately needs you to believe the bubble won't pop.

729
00:42:24,360 --> 00:42:28,160
Speaker 1: He needs you to believe that generative AI will be essential, inevitable,

730
00:42:28,160 --> 00:42:31,799
Speaker 1: and intractable, because if you don't, you'll suddenly realize that

731
00:42:31,920 --> 00:42:35,200
Speaker 1: trillions of dollars in market capitalization and revenue are being

732
00:42:35,239 --> 00:42:39,680
Speaker 1: blown on something it's kind of mediocre. If you focus

733
00:42:39,719 --> 00:42:42,520
Speaker 1: on the present, what open AI's technology can do today

734
00:42:42,760 --> 00:42:45,279
Speaker 1: and will likely do for some time, you see in

735
00:42:45,400 --> 00:42:50,560
Speaker 1: terrifying clarity that generative AI isn't really a society altering technology.

736
00:42:50,840 --> 00:42:54,400
Speaker 1: It's just another form of efficiency driving cloud compute software

737
00:42:54,640 --> 00:42:59,200
Speaker 1: that benefits kind of a small amount of people. If

738
00:42:59,239 --> 00:43:02,800
Speaker 1: you stop saying things like AI could do or AI

739
00:43:02,920 --> 00:43:06,920
Speaker 1: will do, you have to start asking what AI can do,

740
00:43:07,080 --> 00:43:10,239
Speaker 1: and the answer is not that much and probably not

741
00:43:10,239 --> 00:43:13,040
Speaker 1: that much more. In the future. Surra is not going

742
00:43:13,080 --> 00:43:16,319
Speaker 1: to generate entire movies. It's going to continue making horrifying

743
00:43:16,920 --> 00:43:19,839
Speaker 1: human adjacent creatures that walk like the atats from star

744
00:43:19,920 --> 00:43:25,239
Speaker 1: Wars and cartoons that look remarkably like SpongeBob SquarePants Chat

745
00:43:25,280 --> 00:43:27,480
Speaker 1: GPT isn't going to run your business because it can

746
00:43:27,520 --> 00:43:30,280
Speaker 1: barely output a spreadsheet without fucking up the basic numbers,

747
00:43:30,560 --> 00:43:32,560
Speaker 1: if it even understands what you're asking it to do

748
00:43:32,640 --> 00:43:37,120
Speaker 1: in the first place. I think that AI has maybe

749
00:43:37,200 --> 00:43:41,320
Speaker 1: three quarters to prove itself worthwhile before the apocalypse really arrives.

750
00:43:42,480 --> 00:43:44,200
Speaker 1: When it does, you're going to see it first in

751
00:43:44,239 --> 00:43:47,520
Speaker 1: the real infrastructure companies, starting with Nvideo, who's grown to

752
00:43:47,560 --> 00:43:50,480
Speaker 1: about two trillion dollars in market capitalization because of the

753
00:43:50,520 --> 00:43:52,439
Speaker 1: chips they make, which are pretty much the only ones

754
00:43:52,440 --> 00:43:55,160
Speaker 1: that can power the AI revolution. There are other companies

755
00:43:55,160 --> 00:43:57,520
Speaker 1: that AMD in Micron, but m Video is the one

756
00:43:57,520 --> 00:44:00,480
Speaker 1: that's really grown. If you watch any of their notes,

757
00:44:00,880 --> 00:44:04,560
Speaker 1: they're insane. They're just full of fan fiction. Once Nvidia

758
00:44:04,640 --> 00:44:08,080
Speaker 1: starts to see growth slow, and Oracle in particular Oracle

759
00:44:08,760 --> 00:44:11,640
Speaker 1: massive data center company, massive data based company as well

760
00:44:11,840 --> 00:44:14,960
Speaker 1: one of the largest customers Microsoft building data sentence for them.

761
00:44:15,440 --> 00:44:18,600
Speaker 1: Once that starts slowing down, that's when you should start worrying.

762
00:44:19,360 --> 00:44:22,759
Speaker 1: But the real pain's going to come for Amazon, Microsoft

763
00:44:22,800 --> 00:44:25,840
Speaker 1: and Google when it's clear that there's not really that

764
00:44:25,920 --> 00:44:31,040
Speaker 1: much revenue going into their clouds. Once that happens, once

765
00:44:31,040 --> 00:44:34,120
Speaker 1: you start seeing Jim Kramer on CNBC saying I don't

766
00:44:34,120 --> 00:44:36,600
Speaker 1: think the AI boom is here, despite having saying it

767
00:44:36,719 --> 00:44:40,919
Speaker 1: just was, that's when things get nasty and the knock

768
00:44:40,920 --> 00:44:43,960
Speaker 1: on effects will be horrible. It's going to be genuinely painful,

769
00:44:44,000 --> 00:44:46,399
Speaker 1: worse than we've seen the last few years. And it's

770
00:44:46,440 --> 00:44:49,640
Speaker 1: all a result of the same problem. It's all a

771
00:44:49,680 --> 00:44:52,239
Speaker 1: result of the growth of all cost tech economy. When

772
00:44:52,280 --> 00:44:54,799
Speaker 1: things are made to expand, when things are made to

773
00:44:54,880 --> 00:44:58,799
Speaker 1: build more rather than build better, When you're building solutions

774
00:44:58,840 --> 00:45:03,560
Speaker 1: to use compute power to sell cloud computing services rather

775
00:45:03,600 --> 00:45:08,200
Speaker 1: than helping real people make their lives better. Tech is

776
00:45:08,239 --> 00:45:11,520
Speaker 1: not building for real people anymore. And the AI revolution,

777
00:45:11,680 --> 00:45:15,600
Speaker 1: despite its spacious hype, is not really for us. It's

778
00:45:15,600 --> 00:45:18,320
Speaker 1: not for you and me. It's for people at Satching

779
00:45:18,360 --> 00:45:21,760
Speaker 1: the Della of Microsoft to claim that they've increased growth

780
00:45:21,800 --> 00:45:25,279
Speaker 1: by twenty percent. It's for people at sam Altman to

781
00:45:25,320 --> 00:45:28,600
Speaker 1: buy another fucking Porsche. It's so that these people can

782
00:45:28,640 --> 00:45:31,960
Speaker 1: feel important and be rich, rather than improving society at all.

783
00:45:33,239 --> 00:45:35,680
Speaker 1: Maybe I'm wrong, Maybe all of this is the future,

784
00:45:35,680 --> 00:45:39,840
Speaker 1: maybe everything will be automated, but I don't see the signs.

785
00:45:41,160 --> 00:45:45,400
Speaker 1: This doesn't feel much different to the metaverse. There's a product,

786
00:45:45,640 --> 00:45:48,239
Speaker 1: but in the end, what's it really do? Just like

787
00:45:48,320 --> 00:45:51,000
Speaker 1: the metaverse, I don't think many people are really using it.

788
00:45:51,560 --> 00:45:55,480
Speaker 1: All signs point to this being an empty bubble. And

789
00:45:55,520 --> 00:45:57,279
Speaker 1: I'm sure you're sick of this too. I'm sure that

790
00:45:57,280 --> 00:45:59,800
Speaker 1: you're sick of the tech industry telling you the futures

791
00:45:59,840 --> 00:46:02,480
Speaker 1: here when it's the present and it fucking sucks. And

792
00:46:02,520 --> 00:46:05,880
Speaker 1: I'm swearing a lot, and I'm angry, but I'm justified

793
00:46:05,920 --> 00:46:07,560
Speaker 1: in this anger off feel and I'm not telling you

794
00:46:07,640 --> 00:46:09,759
Speaker 1: how to think. And I've heard from some of you saying, oh,

795
00:46:09,760 --> 00:46:11,640
Speaker 1: don't tell me how to think, and I agree. I agree.

796
00:46:11,640 --> 00:46:13,680
Speaker 1: I'm not here to tell you to be angry about anything.

797
00:46:13,920 --> 00:46:16,280
Speaker 1: But I want to give you at least my truth,

798
00:46:16,320 --> 00:46:18,800
Speaker 1: and I want to give you what I see is happening,

799
00:46:18,880 --> 00:46:20,920
Speaker 1: because I don't feel like enough people are doing that

800
00:46:20,960 --> 00:46:23,680
Speaker 1: in the tech industry. And that's what better Offline is

801
00:46:23,719 --> 00:46:26,520
Speaker 1: going to continue to be. I really appreciate you listening.

802
00:46:26,920 --> 00:46:28,680
Speaker 1: It's been about a month month and a half since

803
00:46:28,719 --> 00:46:30,759
Speaker 1: we started. It's only going to get better from here.

804
00:46:31,000 --> 00:46:43,279
Speaker 1: Thank you, thank you for listening to Better Offline. The

805
00:46:43,400 --> 00:46:45,640
Speaker 1: editor and composer of the Better Offline theme song, It

806
00:46:45,680 --> 00:46:48,319
Speaker 1: is Mattersowski. You can check out more of his music

807
00:46:48,360 --> 00:46:52,000
Speaker 1: and audio projects at Mattasowski dot com, M A T

808
00:46:52,000 --> 00:46:56,440
Speaker 1: T O. S O w Ski dot com. You can

809
00:46:56,480 --> 00:46:58,920
Speaker 1: email me at easy at Better Offline dot com, or

810
00:46:59,000 --> 00:47:01,839
Speaker 1: check out Better Offline to find my newsletter and more

811
00:47:01,840 --> 00:47:04,280
Speaker 1: links to this podcast. Thank you so much for listening.

812
00:47:05,440 --> 00:47:08,120
Speaker 1: Better Offline is a production of cool Zone Media. For

813
00:47:08,239 --> 00:47:11,400
Speaker 1: more from cool Zone Media, visit our website cool Zonemedia

814
00:47:11,440 --> 00:47:14,319
Speaker 1: dot com, or check us out on the iHeartRadio app,

815
00:47:14,360 --> 00:47:16,799
Speaker 1: Apple Podcasts, or wherever you get your podcasts.