1
00:00:02,440 --> 00:00:06,760
Speaker 1: A media Hello, and welcome to Better Offline. I'm your

2
00:00:06,800 --> 00:00:20,439
Speaker 1: host ed zich Tron. What as ever, please check the

3
00:00:20,480 --> 00:00:22,520
Speaker 1: episode notes for links to the things that I'm talking

4
00:00:22,560 --> 00:00:26,880
Speaker 1: about now. Frequent listeners know that I'm immensely skeptical of

5
00:00:26,920 --> 00:00:30,320
Speaker 1: generative AI, both as a technology but also as a business.

6
00:00:30,680 --> 00:00:33,920
Speaker 1: It's expensive, it's unreliable, it doesn't actually do that much,

7
00:00:34,280 --> 00:00:36,600
Speaker 1: and the real world use cases range from kind of

8
00:00:36,640 --> 00:00:41,280
Speaker 1: helpful to unhelpful to not really useful in any way,

9
00:00:41,320 --> 00:00:45,000
Speaker 1: shape or form, and definitely not useful enough to justify

10
00:00:45,080 --> 00:00:48,320
Speaker 1: hundreds of billions of dollars in spending. And I know

11
00:00:48,440 --> 00:00:50,720
Speaker 1: somebody is gonna email and say, but ed, my mate's

12
00:00:50,760 --> 00:00:54,640
Speaker 1: granny's uncle's dog uses chat GBT to brainstorm or something

13
00:00:54,680 --> 00:00:57,600
Speaker 1: else equally flimsy. And I really need you to put

14
00:00:57,640 --> 00:01:01,280
Speaker 1: all of this in context. Open AI loses five billion

15
00:01:01,400 --> 00:01:04,000
Speaker 1: dollars a year, and basically every big tech company is

16
00:01:04,040 --> 00:01:06,800
Speaker 1: blown past their emissions targets, with few signs that they'll

17
00:01:06,840 --> 00:01:08,760
Speaker 1: ever bother to try and meet them again, all in

18
00:01:08,760 --> 00:01:12,240
Speaker 1: pursuit of the piddliest or most convoluted use cases in history.

19
00:01:12,680 --> 00:01:16,560
Speaker 1: It's a goddamn fast and that's why generative AI is

20
00:01:16,600 --> 00:01:19,280
Speaker 1: such a regular topic on this show. I feel like

21
00:01:19,319 --> 00:01:22,280
Speaker 1: the tech industry is experiencing a kind of collective madness,

22
00:01:22,280 --> 00:01:25,520
Speaker 1: a delusion that's seeing companies like Microsoft and Google, but

23
00:01:25,680 --> 00:01:28,240
Speaker 1: both their companies and our planet on a technology that

24
00:01:28,280 --> 00:01:33,000
Speaker 1: continually fails to deliver on well anything I want to

25
00:01:33,200 --> 00:01:37,039
Speaker 1: say here, And I actually wrote down this supposedly massive potential,

26
00:01:37,680 --> 00:01:40,520
Speaker 1: but what was the potential? Ever? Can anyone actually tell

27
00:01:40,560 --> 00:01:44,320
Speaker 1: me what this was meant to do? Other than agi? Anyway,

28
00:01:44,600 --> 00:01:47,520
Speaker 1: Eventually I see this all falling apart, and I think

29
00:01:47,520 --> 00:01:49,840
Speaker 1: there's going to be a clamorous event when everything does

30
00:01:49,880 --> 00:01:53,560
Speaker 1: actually unravel. I return so often to Generative AI because

31
00:01:53,560 --> 00:01:57,200
Speaker 1: I truly fear the damage that this confrontation will cause,

32
00:01:57,960 --> 00:02:00,120
Speaker 1: and I'm concerned about the damage that's happening right now.

33
00:02:00,120 --> 00:02:03,360
Speaker 1: If I'm honest, And while my tone's off in acerbic

34
00:02:03,360 --> 00:02:05,080
Speaker 1: and I don't think you'd have it an any other way,

35
00:02:05,480 --> 00:02:08,840
Speaker 1: I don't really take much joy in watching these things burn. Sure,

36
00:02:08,919 --> 00:02:11,639
Speaker 1: I admit it's pretty funny thinking about satching a Delea

37
00:02:11,680 --> 00:02:13,800
Speaker 1: getting fired or sunded up a shy getting fired, or

38
00:02:13,800 --> 00:02:18,440
Speaker 1: both of them looking stupid, But it's not the same

39
00:02:18,520 --> 00:02:21,680
Speaker 1: thing Miglee's kind of tempered. When I stop and think

40
00:02:21,680 --> 00:02:23,720
Speaker 1: about what all of this means or will mean, I

41
00:02:23,720 --> 00:02:26,520
Speaker 1: should say, for the companies trying to get funding right now,

42
00:02:26,760 --> 00:02:28,480
Speaker 1: or for the companies that will try and raise money

43
00:02:28,520 --> 00:02:30,760
Speaker 1: after the bubble bursts, and for the tech workers that

44
00:02:30,800 --> 00:02:33,440
Speaker 1: will get laid off, either because Microsoft and Google want

45
00:02:33,440 --> 00:02:36,280
Speaker 1: to put more money into data centers and GPUs, or

46
00:02:36,320 --> 00:02:38,680
Speaker 1: because their generative AI bets have gone tits up and

47
00:02:38,720 --> 00:02:41,400
Speaker 1: they need to save money some other way. And so

48
00:02:41,440 --> 00:02:44,560
Speaker 1: I hope you'll forgive me for a somewhat more somber

49
00:02:44,600 --> 00:02:47,400
Speaker 1: tone to this episode. I want to again make a

50
00:02:47,480 --> 00:02:50,280
Speaker 1: clear headed analysis of where we are today and where

51
00:02:50,320 --> 00:02:53,080
Speaker 1: I think we're going. And the reason I'm retreading this

52
00:02:53,120 --> 00:02:55,280
Speaker 1: path is because over the last month or so, we've

53
00:02:55,320 --> 00:02:58,600
Speaker 1: seen some really really alarming signs about the accelerating crisis

54
00:02:58,600 --> 00:03:01,600
Speaker 1: in generative AI, whispers that the party will soon come

55
00:03:01,639 --> 00:03:04,760
Speaker 1: to a halt. Now, you may remember a few episodes

56
00:03:04,800 --> 00:03:07,080
Speaker 1: back that I started talking about pale horses of the

57
00:03:07,120 --> 00:03:11,160
Speaker 1: AI apocalypse, signs that things were falling apart, like layoffs

58
00:03:11,200 --> 00:03:15,000
Speaker 1: at AI companies, price increases or decreases, internal discord at

59
00:03:15,040 --> 00:03:19,200
Speaker 1: these big AI companies, speciously impressive yet kind of flimsy

60
00:03:19,240 --> 00:03:23,240
Speaker 1: product announcements and so on and so forth. Since then,

61
00:03:23,440 --> 00:03:26,280
Speaker 1: I've kind of been proven right several times. Multiple pale

62
00:03:26,320 --> 00:03:29,120
Speaker 1: horses have emerged. There was a big, stupid magic trick

63
00:03:29,160 --> 00:03:31,519
Speaker 1: to impress investors in the form of open aised rush

64
00:03:31,600 --> 00:03:34,359
Speaker 1: launch of its one and that model, by the way,

65
00:03:34,440 --> 00:03:38,120
Speaker 1: was code named Strawberry. Rumored pricing increases which hat GBT

66
00:03:38,200 --> 00:03:40,800
Speaker 1: in the future. Now they're looking at forty four dollars

67
00:03:40,840 --> 00:03:43,080
Speaker 1: a month by twenty thirty, and they're looking at potentially

68
00:03:43,120 --> 00:03:46,000
Speaker 1: thousands for models in the future, and then layoffs at

69
00:03:46,000 --> 00:03:49,360
Speaker 1: Scale AI, which is one of the biggest players in

70
00:03:49,440 --> 00:03:52,000
Speaker 1: AI training data, if not the biggest. And these are

71
00:03:52,000 --> 00:03:55,200
Speaker 1: all signs that things are beginning to fall apart. And

72
00:03:55,240 --> 00:03:57,760
Speaker 1: I think it's important to explain how precarious things are

73
00:03:57,840 --> 00:04:00,720
Speaker 1: right now despite all of the money stashing around, and

74
00:04:00,760 --> 00:04:04,040
Speaker 1: how dangerous the power of magical thinking really is. I

75
00:04:04,080 --> 00:04:06,480
Speaker 1: want to express my concerns about the fragility of this

76
00:04:06,600 --> 00:04:10,040
Speaker 1: movement and the obsessiveness and directionless that brought us here,

77
00:04:10,680 --> 00:04:13,120
Speaker 1: and I want us all to kind of do better now.

78
00:04:13,160 --> 00:04:14,960
Speaker 1: Don't mean you the listener, but if you remember the

79
00:04:14,960 --> 00:04:17,920
Speaker 1: media that's written something vague about AI and you've kind

80
00:04:17,960 --> 00:04:21,440
Speaker 1: of carried water for Sam Mortman. It is time to stop.

81
00:04:21,480 --> 00:04:23,320
Speaker 1: It was already time to stop some time Aguary, I

82
00:04:23,360 --> 00:04:26,080
Speaker 1: should just be clear, but now really is the time

83
00:04:26,120 --> 00:04:29,160
Speaker 1: to stop. But this week's a two parter, and I'll

84
00:04:29,200 --> 00:04:32,960
Speaker 1: be explaining things well in two parts. Obviously. First of all,

85
00:04:32,960 --> 00:04:35,480
Speaker 1: we're going to talk about open ai, the most recent

86
00:04:35,520 --> 00:04:38,560
Speaker 1: funding round, and how bad their business is and exactly

87
00:04:38,600 --> 00:04:41,320
Speaker 1: how worried you should be about them now. The second

88
00:04:41,360 --> 00:04:43,359
Speaker 1: part is going to be about something I'm calling the

89
00:04:43,400 --> 00:04:46,440
Speaker 1: subprime AI crisis, where I see a growing bubble of

90
00:04:46,480 --> 00:04:51,200
Speaker 1: companies integrating generative AI at prices well heavily subsidized by

91
00:04:51,240 --> 00:04:54,039
Speaker 1: big tech and heavily discounted by companies like open ai

92
00:04:54,080 --> 00:04:57,560
Speaker 1: and Anthropic. Should either of these companies get desperate or

93
00:04:57,600 --> 00:04:59,839
Speaker 1: I don't know, actually want to make more money than

94
00:04:59,839 --> 00:05:02,280
Speaker 1: they spend, they're inevitably going to have to raise the

95
00:05:02,360 --> 00:05:05,240
Speaker 1: prices to match their actual costs, which will likely make

96
00:05:05,279 --> 00:05:09,880
Speaker 1: the existence of many AI startups completely untenable. You know,

97
00:05:10,320 --> 00:05:12,520
Speaker 1: kind of like when the teaser interest rates on subprime

98
00:05:12,560 --> 00:05:16,279
Speaker 1: mortgages that expired during the Financial crisis, and people suddenly

99
00:05:16,279 --> 00:05:20,000
Speaker 1: started losing their houses. In any case, these are going

100
00:05:20,040 --> 00:05:22,800
Speaker 1: to be meaningful, yeah, kind of brutal episodes, in part

101
00:05:22,839 --> 00:05:25,560
Speaker 1: because while it's enjoyable to watch big tech burn or

102
00:05:25,560 --> 00:05:28,520
Speaker 1: get embarrassed, there's a real human cost, like I said,

103
00:05:28,720 --> 00:05:31,320
Speaker 1: and the damage to our environment is already being done

104
00:05:31,560 --> 00:05:35,880
Speaker 1: and might not even stop when Generative AI does. And

105
00:05:35,920 --> 00:05:38,599
Speaker 1: whether Microsoft and Google and the other big generative AI

106
00:05:38,720 --> 00:05:41,840
Speaker 1: backers slowly wind down their positions or cannibalize their companies

107
00:05:41,880 --> 00:05:44,440
Speaker 1: to keep open AI and Anthropic alive, I'm convinced that

108
00:05:44,480 --> 00:05:46,120
Speaker 1: the end result is still going to be the same.

109
00:05:46,839 --> 00:05:48,880
Speaker 1: I fear tens of thousands of people will lose their

110
00:05:48,960 --> 00:05:51,160
Speaker 1: jobs and much of the tech industry will suffer as

111
00:05:51,200 --> 00:05:53,240
Speaker 1: they realize that the only thing that can grow forever

112
00:05:53,320 --> 00:05:56,400
Speaker 1: is a cancer. I'm going to paint you oblique picture

113
00:05:56,800 --> 00:05:58,880
Speaker 1: not just for the big AI players, but for tech

114
00:05:58,920 --> 00:06:01,400
Speaker 1: more widely, and for the people to work at these companies,

115
00:06:01,640 --> 00:06:03,640
Speaker 1: and tell you why I think the conclusion to this

116
00:06:03,760 --> 00:06:06,840
Speaker 1: sordid saga, as brutal and damaging as it will be,

117
00:06:07,080 --> 00:06:14,280
Speaker 1: is coming sooner than you think. Let's begin, as have

118
00:06:14,400 --> 00:06:17,280
Speaker 1: explained in agonizing detail in the past in my newsletter

119
00:06:17,440 --> 00:06:19,799
Speaker 1: Open AI will have to continue to raise more money

120
00:06:19,800 --> 00:06:23,679
Speaker 1: than any startup has ever raised in history, in perpetuity

121
00:06:24,000 --> 00:06:27,320
Speaker 1: just to survive. They're going to burn five billion dollars

122
00:06:27,320 --> 00:06:29,960
Speaker 1: this year, and yes, that's after they make their revenue

123
00:06:30,080 --> 00:06:32,039
Speaker 1: and costs are only set to increase over the next

124
00:06:32,040 --> 00:06:35,159
Speaker 1: few years they develop bigger models. Their business as it

125
00:06:35,240 --> 00:06:38,320
Speaker 1: stands is completely untenable, and I'm going to explain why

126
00:06:38,400 --> 00:06:42,640
Speaker 1: in a bit. Nevertheless, that's why open Ai, the ostensible

127
00:06:42,680 --> 00:06:46,000
Speaker 1: nonprofit that may soon become a for profit who bloody

128
00:06:46,040 --> 00:06:49,200
Speaker 1: knows at this point, raise the funding round at evaluation

129
00:06:49,279 --> 00:06:52,560
Speaker 1: of one hundred and fifty seven billion dollars, bringing in

130
00:06:52,600 --> 00:06:55,799
Speaker 1: a reported six point six billion dollars in cash, along

131
00:06:55,800 --> 00:06:58,320
Speaker 1: with opening a four billion dollar revolving credit line from

132
00:06:58,320 --> 00:07:01,160
Speaker 1: banks like JP morgan See and a bunch of others

133
00:07:01,160 --> 00:07:04,720
Speaker 1: that should know better after the acquisition of Twitter. Investors

134
00:07:04,720 --> 00:07:07,600
Speaker 1: in the funding ground included Josh Kushner's Thrive Capital in

135
00:07:07,720 --> 00:07:11,000
Speaker 1: Video and of course Microsoft, who were somehow undeterred by

136
00:07:11,000 --> 00:07:14,720
Speaker 1: the abrupt exiffs open ai cto mirror Marati days before

137
00:07:14,760 --> 00:07:18,120
Speaker 1: the round closed. Also joining the frame was soft Bank,

138
00:07:18,320 --> 00:07:20,680
Speaker 1: the Japanese investment company with a drag record of making

139
00:07:20,680 --> 00:07:23,400
Speaker 1: some of the worst bets in history, and they dumped

140
00:07:23,440 --> 00:07:27,440
Speaker 1: five hundred million dollars into open Ai, and that alone

141
00:07:27,520 --> 00:07:30,760
Speaker 1: should start getting people a little worried. You see, soft

142
00:07:30,800 --> 00:07:34,480
Speaker 1: Bank's founder, Masayoshi Son is what we in the business

143
00:07:34,520 --> 00:07:37,680
Speaker 1: call the dumbass. I realized this is a very broad

144
00:07:37,720 --> 00:07:41,480
Speaker 1: brush to paint somebody with. But let me explain. Soft

145
00:07:41,480 --> 00:07:44,680
Speaker 1: Bank's vision fund dumped sixteen billion dollars into we Work,

146
00:07:44,760 --> 00:07:46,480
Speaker 1: which by the end of last year, by the way,

147
00:07:46,560 --> 00:07:49,560
Speaker 1: was worth forty four million dollars, and nearly a billion

148
00:07:49,600 --> 00:07:52,080
Speaker 1: dollars into wire Card, which turned out to be an

149
00:07:52,080 --> 00:07:55,160
Speaker 1: elaborate fraud. With its leadership. Are the facing criminal charges

150
00:07:55,320 --> 00:07:59,280
Speaker 1: or becoming fugitives from justice in Russia. Soft Bank has

151
00:07:59,320 --> 00:08:02,760
Speaker 1: become effective synonymous with burning cash on terrible bets and

152
00:08:02,800 --> 00:08:05,880
Speaker 1: lost over thirty two billion dollars in the last three years.

153
00:08:07,080 --> 00:08:09,480
Speaker 1: The problem that open Ai faces is that they need

154
00:08:09,520 --> 00:08:12,280
Speaker 1: an absolute shit ton of money, and generally the only

155
00:08:12,320 --> 00:08:14,520
Speaker 1: companies that need that much are the ones that may

156
00:08:14,560 --> 00:08:18,240
Speaker 1: not have a sustainable business model. Open Ai demanded a

157
00:08:18,280 --> 00:08:21,880
Speaker 1: minimum investment of two hundred and fifty million dollars from investors,

158
00:08:22,080 --> 00:08:25,120
Speaker 1: and while regular American vcs might have that much, and

159
00:08:25,240 --> 00:08:27,960
Speaker 1: indeed had in the case of Costler and Kochume Management,

160
00:08:28,920 --> 00:08:30,840
Speaker 1: there aren't enough of them to fill out a six

161
00:08:30,880 --> 00:08:33,560
Speaker 1: point six billion dollar funding round. And because open Ai

162
00:08:34,000 --> 00:08:38,240
Speaker 1: needed that much, not wanted needed, their only real choices

163
00:08:38,240 --> 00:08:40,559
Speaker 1: were to go to Masayoshi Son, who's a man who

164
00:08:40,600 --> 00:08:43,360
Speaker 1: put tons of money into stupid things and believes he

165
00:08:43,440 --> 00:08:45,840
Speaker 1: was put on this earth to make digital superintelligence and

166
00:08:45,840 --> 00:08:47,880
Speaker 1: I have a link to that. By the way. It's insane,

167
00:08:48,920 --> 00:08:51,760
Speaker 1: all I mean. There's also another choice, you know, the

168
00:08:52,040 --> 00:08:55,880
Speaker 1: far dodgier investors in the Middle East m mg X,

169
00:08:56,200 --> 00:08:59,160
Speaker 1: one hundred billion dollar investment fund backed by the United

170
00:08:59,160 --> 00:09:03,040
Speaker 1: Arab Emirates to invest primarily in AI and semiconductor companies.

171
00:09:04,160 --> 00:09:06,800
Speaker 1: This should also be a big warning sign that things

172
00:09:06,800 --> 00:09:10,360
Speaker 1: are going poorly, because absolutely nobody raises from the UAE

173
00:09:10,480 --> 00:09:13,319
Speaker 1: of the SuDS because they want to. They're the place

174
00:09:13,320 --> 00:09:14,960
Speaker 1: you go if you need a lot of money and

175
00:09:15,000 --> 00:09:17,040
Speaker 1: you're not confident anybody else is going to give it

176
00:09:17,040 --> 00:09:20,240
Speaker 1: to you. Other aspects of this deal are also a

177
00:09:20,280 --> 00:09:22,840
Speaker 1: bit worrying, one of them being that one of the

178
00:09:22,880 --> 00:09:25,719
Speaker 1: founding partners of MGX is the Sovereign Wealth Fund of

179
00:09:25,760 --> 00:09:29,480
Speaker 1: Abu Dhabi, which brought in around five hundred million dollars

180
00:09:29,520 --> 00:09:33,319
Speaker 1: into Anthropic in twenty twenty three. Kind of a conflict

181
00:09:33,320 --> 00:09:35,360
Speaker 1: of interest. Do you think any of these people actually

182
00:09:35,440 --> 00:09:39,160
Speaker 1: goddamn care though they're all working to the same rot economy.

183
00:09:39,240 --> 00:09:43,319
Speaker 1: Nonsense fucking Anyway, as I mentioned previously, open Ai now

184
00:09:43,360 --> 00:09:45,480
Speaker 1: has access to that four billion dollars in debt from

185
00:09:45,520 --> 00:09:49,199
Speaker 1: banks in that revolving credit facility too, and that's also

186
00:09:49,280 --> 00:09:52,360
Speaker 1: not brilliant for a company that just burns cash. I

187
00:09:52,400 --> 00:09:54,680
Speaker 1: don't know what the interest rate on it is, but

188
00:09:54,720 --> 00:09:57,839
Speaker 1: I can't imagine it's great. Generally, revolving credit lines are

189
00:09:57,920 --> 00:10:02,960
Speaker 1: worse interest rates, So that's great, I guess. But open

190
00:10:03,000 --> 00:10:06,440
Speaker 1: ai is desperate. Fundraising comes down to one one thing,

191
00:10:06,559 --> 00:10:10,600
Speaker 1: one really obvious thing. They need so much money. They

192
00:10:10,679 --> 00:10:13,240
Speaker 1: needed like ten billion dollars here, and they kind of

193
00:10:13,280 --> 00:10:15,880
Speaker 1: got it in this wonky way. But they need that

194
00:10:15,960 --> 00:10:19,000
Speaker 1: money to survive because, like I said, they lost five

195
00:10:19,080 --> 00:10:22,440
Speaker 1: billion in twenty twenty four, and that figure is likely

196
00:10:22,480 --> 00:10:25,679
Speaker 1: to increase as more complex models demand more compute and

197
00:10:25,720 --> 00:10:29,520
Speaker 1: more training. Data and Anthropic CEO Dario ama Day predicted

198
00:10:29,520 --> 00:10:31,640
Speaker 1: that future models may cost as much as one hundred

199
00:10:31,640 --> 00:10:35,120
Speaker 1: billion dollars to train and open ai. By the way,

200
00:10:35,320 --> 00:10:37,640
Speaker 1: they may have just raised the biggest round in history.

201
00:10:37,960 --> 00:10:40,320
Speaker 1: They're going to have to raise another round again soon.

202
00:10:40,960 --> 00:10:45,959
Speaker 1: And I'll explain why. While open ai succeeded here, it

203
00:10:46,000 --> 00:10:48,920
Speaker 1: was only after a fairly arduous process where they'd already

204
00:10:48,960 --> 00:10:51,640
Speaker 1: tried to raise a hundred billion dollar valuation earlier in

205
00:10:51,679 --> 00:10:54,840
Speaker 1: the year, and that specifically turned off investors because of

206
00:10:54,880 --> 00:10:57,880
Speaker 1: the huge price tag. And there's this, by the way,

207
00:10:58,000 --> 00:11:00,640
Speaker 1: to quote the information, there's a growing concern over the

208
00:11:00,720 --> 00:11:04,920
Speaker 1: overvaluation of generative AI companies. Oh you fucking think do

209
00:11:04,920 --> 00:11:07,920
Speaker 1: you think that there's a concern here? Well, after this podcast,

210
00:11:07,920 --> 00:11:11,480
Speaker 1: you're going to be really, really concerned. I guarantee anyway

211
00:11:11,840 --> 00:11:14,200
Speaker 1: to get the round done, open ai committed to converting

212
00:11:14,280 --> 00:11:17,439
Speaker 1: itself from a nonprofit to a for profit entity. Failure

213
00:11:17,480 --> 00:11:19,440
Speaker 1: to do so within two years will see their funding

214
00:11:19,480 --> 00:11:21,920
Speaker 1: converted into debt, which by the way, will be the

215
00:11:21,960 --> 00:11:24,720
Speaker 1: kiss of death for a company that only ever loses money.

216
00:11:25,440 --> 00:11:28,440
Speaker 1: The deal, per Reuter's, also includes provisions that would allow

217
00:11:28,440 --> 00:11:31,440
Speaker 1: investors to adjust the valuation or claw back their investment

218
00:11:31,559 --> 00:11:35,360
Speaker 1: should the transition from nonprofit to for profit fail. And

219
00:11:35,400 --> 00:11:39,800
Speaker 1: I mean this is worrying because I don't think there's

220
00:11:39,840 --> 00:11:42,880
Speaker 1: any historical precedence of a company, a nonprofit at least,

221
00:11:42,880 --> 00:11:45,640
Speaker 1: that had this much money then converted. And what complicates

222
00:11:45,679 --> 00:11:48,080
Speaker 1: that as well is converting a business of this scale

223
00:11:48,080 --> 00:11:51,880
Speaker 1: from a nonprofit to a for profit. Well, it involves

224
00:11:52,000 --> 00:11:55,560
Speaker 1: transferring assets, and because any assets previously donated to the

225
00:11:55,559 --> 00:11:59,320
Speaker 1: public benefit, as Alexander Reed, partner at Baker Holsteller, is

226
00:11:59,400 --> 00:12:02,000
Speaker 1: quote as saying The Wall Street Journal cannot be converted

227
00:12:02,080 --> 00:12:05,600
Speaker 1: into a private benefit without compensating the public for the loss.

228
00:12:06,360 --> 00:12:10,079
Speaker 1: Every asset, which includes things like patterns and other intellectual property,

229
00:12:10,280 --> 00:12:13,400
Speaker 1: would have to be paid for. Well, it's hard to

230
00:12:13,400 --> 00:12:15,640
Speaker 1: say for certain, it's likely this deal and the process

231
00:12:15,640 --> 00:12:19,079
Speaker 1: of untangling open ai from its nonprofit beginnings is really

232
00:12:19,120 --> 00:12:22,559
Speaker 1: only going to add to open AI's cash problem. And

233
00:12:22,720 --> 00:12:24,920
Speaker 1: it's not like that was a good problem to deal

234
00:12:24,960 --> 00:12:28,480
Speaker 1: with in the beginning. And now I will do a

235
00:12:28,559 --> 00:12:31,440
Speaker 1: thorough review and analysis of open ai because I want

236
00:12:31,440 --> 00:12:33,600
Speaker 1: you to see exactly how stupid things have become in

237
00:12:33,640 --> 00:12:36,960
Speaker 1: the tech industry and how ridiculous all of this writing

238
00:12:37,000 --> 00:12:39,760
Speaker 1: this boiled my blood, by the way, because when you

239
00:12:39,800 --> 00:12:43,560
Speaker 1: see how lossy this company is, when you really ingest

240
00:12:43,600 --> 00:12:46,200
Speaker 1: how bad open ai is and how stupid it is

241
00:12:46,240 --> 00:12:49,240
Speaker 1: that we've let this happen, and we meaning the tech industry.

242
00:12:50,160 --> 00:12:52,480
Speaker 1: You then will think on your own lives and be like, wow,

243
00:12:53,120 --> 00:12:57,520
Speaker 1: you think I could pay my rent like late, or

244
00:12:57,600 --> 00:13:00,040
Speaker 1: like ask a company to pay my rent because my

245
00:13:00,240 --> 00:13:03,520
Speaker 1: job involves me just burning piles of cash. No, you

246
00:13:03,520 --> 00:13:05,960
Speaker 1: wouldn't be able to lose money for a business. You

247
00:13:06,040 --> 00:13:09,840
Speaker 1: and I we have to operate normally, not sam Oltman,

248
00:13:09,920 --> 00:13:13,920
Speaker 1: though sam Oltman doesn't have to be normal. Sam Altman,

249
00:13:14,040 --> 00:13:16,679
Speaker 1: like all of the gilded assholes of the tech industry,

250
00:13:17,080 --> 00:13:19,600
Speaker 1: is allowed to live in this fantasy land based on

251
00:13:19,679 --> 00:13:23,480
Speaker 1: marketing high Now, before we go into it, though, I'd

252
00:13:23,520 --> 00:13:26,439
Speaker 1: like to enjoy one of the following lovingly crafted advertisements

253
00:13:26,440 --> 00:13:28,480
Speaker 1: that in no way seem out of place from anything

254
00:13:28,520 --> 00:13:44,680
Speaker 1: I'm saying. And we're back, So to put it bluntly,

255
00:13:45,040 --> 00:13:48,320
Speaker 1: open ai is an absolute dog of a company. On

256
00:13:48,360 --> 00:13:51,320
Speaker 1: September twenty seventh, The New York Times reported that open

257
00:13:51,360 --> 00:13:54,120
Speaker 1: ai would lose five billion dollars in twenty twenty four,

258
00:13:54,240 --> 00:13:56,720
Speaker 1: a number which the Information had estimated back in July,

259
00:13:57,240 --> 00:13:59,280
Speaker 1: and the open Ai expected to raise the price of

260
00:13:59,320 --> 00:14:02,400
Speaker 1: chat GPT plus its premium product to twenty two dollars

261
00:14:02,440 --> 00:14:04,240
Speaker 1: a month by the end of twenty twenty four and

262
00:14:04,280 --> 00:14:09,600
Speaker 1: are remarkable forty four dollars a month by twenty thirty. Interestingly,

263
00:14:09,920 --> 00:14:13,720
Speaker 1: and worryingly, the Times also confirmed another hypothesis of mine

264
00:14:14,000 --> 00:14:17,440
Speaker 1: that and I quote fundraising material also signaled that open

265
00:14:17,440 --> 00:14:19,760
Speaker 1: ai would need to continue to raise money over the

266
00:14:19,800 --> 00:14:22,400
Speaker 1: next year because its expenses grew in tandem with the

267
00:14:22,480 --> 00:14:25,920
Speaker 1: number of people using its products. In simpler terms, open

268
00:14:25,960 --> 00:14:28,440
Speaker 1: ai may have raised six point five or six point

269
00:14:28,520 --> 00:14:31,880
Speaker 1: six billion in funding literally a week ago, but they

270
00:14:31,920 --> 00:14:34,880
Speaker 1: need to raise more, probably the same amount or more

271
00:14:35,440 --> 00:14:38,120
Speaker 1: I'm going to say, within the next six months. And

272
00:14:38,160 --> 00:14:40,960
Speaker 1: The Times also reports that open ai is making internal

273
00:14:41,000 --> 00:14:44,560
Speaker 1: revenue estimates I would describe as this is a technical term,

274
00:14:44,680 --> 00:14:50,640
Speaker 1: absolutely fucking ridiculous. Open aiy's monthly revenue hit three hundred

275
00:14:50,720 --> 00:14:53,480
Speaker 1: million dollars in August, and the company expects to make

276
00:14:53,520 --> 00:14:57,040
Speaker 1: three point seven billion dollars this year. The company will,

277
00:14:57,040 --> 00:14:59,560
Speaker 1: as mentioned, loose five billion dollars anyway, and I'm going

278
00:14:59,600 --> 00:15:02,720
Speaker 1: to keep minding you of that. Yeah. Open ai says

279
00:15:02,720 --> 00:15:05,880
Speaker 1: that it expects to make eleven point six billion dollars

280
00:15:05,920 --> 00:15:09,200
Speaker 1: in twenty twenty five and an astonishing one hundred billion

281
00:15:09,200 --> 00:15:11,520
Speaker 1: dollars by twenty twenty nine, a statement that is so

282
00:15:11,600 --> 00:15:14,200
Speaker 1: egregious that I'm surprised it's not some kind of financial

283
00:15:14,240 --> 00:15:17,360
Speaker 1: crime to say it. For some context, by the way,

284
00:15:17,600 --> 00:15:20,360
Speaker 1: Microsoft makes about two hundred and fifty billion dollars a year,

285
00:15:20,400 --> 00:15:23,000
Speaker 1: Google about three hundred billion dollars a year, and Apple

286
00:15:23,000 --> 00:15:26,040
Speaker 1: about four hundred billion dollars a year. So yeah, I

287
00:15:26,040 --> 00:15:28,840
Speaker 1: guess by twenty thirty that's how big open ai. Well

288
00:15:28,880 --> 00:15:33,120
Speaker 1: you've well, you fucking talk. It drives me insane. And also,

289
00:15:33,160 --> 00:15:35,600
Speaker 1: by the way, all of those companies make more money

290
00:15:35,600 --> 00:15:39,280
Speaker 1: than they spend. Open ai, on the other hand, spends

291
00:15:39,320 --> 00:15:42,080
Speaker 1: two dollars and thirty five cents to make one dollar.

292
00:15:42,360 --> 00:15:44,840
Speaker 1: If you remember one thing from this podcast, it's that

293
00:15:45,640 --> 00:15:49,040
Speaker 1: every dollar they make they have to spend two dollars

294
00:15:49,120 --> 00:15:53,400
Speaker 1: and thirty five cents to get. It drives me insane. People,

295
00:15:54,160 --> 00:15:56,360
Speaker 1: we don't have to live in these conditions. How do

296
00:15:56,440 --> 00:15:58,360
Speaker 1: I do this? How do I get to burn money?

297
00:15:58,400 --> 00:16:02,840
Speaker 1: What the no? Seriously, open ai loses money on every

298
00:16:03,080 --> 00:16:07,080
Speaker 1: single transaction, every single time that somebody uses chet, GPT

299
00:16:07,160 --> 00:16:10,680
Speaker 1: or plus or connects one of their models, and while

300
00:16:10,720 --> 00:16:14,040
Speaker 1: it might make money selling premium subscriptions, I severely doubt

301
00:16:14,120 --> 00:16:17,320
Speaker 1: those subscriptions are turning a profit and they I really

302
00:16:17,320 --> 00:16:20,520
Speaker 1: do think losing money even on their power users. And

303
00:16:20,600 --> 00:16:23,160
Speaker 1: as I'll get into the next episode, I believe there's

304
00:16:23,160 --> 00:16:26,640
Speaker 1: also a SUBPRIMEI crisis brewing because open ai is API

305
00:16:26,720 --> 00:16:30,840
Speaker 1: services which let people integrate their models into products, are

306
00:16:30,880 --> 00:16:34,280
Speaker 1: currently priced to attract customers in scale, and increasing these

307
00:16:34,280 --> 00:16:36,840
Speaker 1: prices to match the actual cost will likely make this

308
00:16:36,920 --> 00:16:40,600
Speaker 1: product unsustainable for many businesses currently relying on these discount

309
00:16:40,640 --> 00:16:43,520
Speaker 1: of rates, and that's if they have any usage. But

310
00:16:43,600 --> 00:16:46,640
Speaker 1: I'll also get to that later. Now, if you anything

311
00:16:46,800 --> 00:16:49,200
Speaker 1: like me, you have somebody who's told you that open

312
00:16:49,240 --> 00:16:51,520
Speaker 1: ai is a growth business and that it will just

313
00:16:51,800 --> 00:16:54,720
Speaker 1: turn the knob to make it self profitable, much like

314
00:16:54,760 --> 00:16:57,920
Speaker 1: Amazon Web Services, which is Amazon's cloud computing product, did

315
00:16:58,000 --> 00:17:01,240
Speaker 1: years ago. Now, I just want to be clear. If

316
00:17:01,280 --> 00:17:03,800
Speaker 1: you're listening and you think that you're just wrong, you're

317
00:17:03,800 --> 00:17:06,639
Speaker 1: just completely wrong. You live in a fancyland. Wake the

318
00:17:06,680 --> 00:17:08,760
Speaker 1: fuck up. I'm sick and tired of hearing this crap.

319
00:17:09,160 --> 00:17:13,600
Speaker 1: Open ai is nothing like Amazon Web Services. First of all,

320
00:17:13,760 --> 00:17:18,399
Speaker 1: open ai owns none of its infrastructure as everything everything

321
00:17:18,520 --> 00:17:22,679
Speaker 1: is run on Microsoft's as Your Cloud. Secondly, Microsoft, also

322
00:17:22,800 --> 00:17:25,760
Speaker 1: as part of an open ai funding around from twenty nineteen,

323
00:17:26,040 --> 00:17:30,400
Speaker 1: has full access to all of open AI's pre agi research,

324
00:17:30,440 --> 00:17:32,440
Speaker 1: which is all of it, by the way, and full

325
00:17:32,480 --> 00:17:35,160
Speaker 1: license to sell and integrate all of their technology. They

326
00:17:35,160 --> 00:17:40,720
Speaker 1: have complete free reign over open aiy's intellectual property. Thirdly,

327
00:17:40,960 --> 00:17:44,000
Speaker 1: Amazon Web Services did not immediately start with a deficit

328
00:17:44,080 --> 00:17:46,600
Speaker 1: so great that it required raising more money than's ever

329
00:17:46,640 --> 00:17:49,840
Speaker 1: been raised in the history of tech. Where still open

330
00:17:49,880 --> 00:17:52,920
Speaker 1: AI's technology doesn't get cheaper as it scales. In fact,

331
00:17:53,040 --> 00:17:57,560
Speaker 1: it does the opposite. This is not like other businesses.

332
00:17:57,560 --> 00:18:00,719
Speaker 1: You want to compare this to stop saying this to people.

333
00:18:00,920 --> 00:18:04,439
Speaker 1: It's ridiculous. And I know why people say this, and

334
00:18:04,480 --> 00:18:06,439
Speaker 1: I am ranting, and you just have to forgive me.

335
00:18:06,840 --> 00:18:09,359
Speaker 1: I know that people don't want to believe that this

336
00:18:09,560 --> 00:18:12,280
Speaker 1: much money can be wrong. It can be. It is

337
00:18:12,320 --> 00:18:15,520
Speaker 1: currently being wrong. You are watching and listening to it

338
00:18:15,720 --> 00:18:18,639
Speaker 1: being wrong. This is not like Google Cloud. It is

339
00:18:18,680 --> 00:18:20,600
Speaker 1: not like the early days of the Internet. It is

340
00:18:20,600 --> 00:18:24,560
Speaker 1: not like Amazon Web Services. It's nothing like them at all.

341
00:18:25,320 --> 00:18:29,240
Speaker 1: There is no historical example like this, not even Uber

342
00:18:29,600 --> 00:18:32,520
Speaker 1: Uber at least had a business model. In the worst

343
00:18:32,600 --> 00:18:35,080
Speaker 1: year of its life, Uber lost I think six point

344
00:18:35,119 --> 00:18:38,320
Speaker 1: six billion dollars or something like that. So about one

345
00:18:38,320 --> 00:18:40,720
Speaker 1: point six billion dollars in open ai lost in twenty

346
00:18:40,760 --> 00:18:43,280
Speaker 1: twenty four. Right, you want to know what year that was?

347
00:18:43,480 --> 00:18:47,280
Speaker 1: Twenty twenty the year when people couldn't go outside, and

348
00:18:47,320 --> 00:18:49,680
Speaker 1: there are many years after they probably should have stayed indoors. Two,

349
00:18:49,800 --> 00:18:52,919
Speaker 1: But that's a different podcast. The point I'm making is,

350
00:18:53,119 --> 00:18:55,080
Speaker 1: if you've got a comfortable little thing in your soul

351
00:18:55,119 --> 00:18:57,480
Speaker 1: that's saying this is just like things I've seen before,

352
00:18:57,560 --> 00:19:01,119
Speaker 1: they'll turn it around. They absolutely And now I'm going

353
00:19:01,160 --> 00:19:04,199
Speaker 1: to break down exactly how untenable open ai is with

354
00:19:04,320 --> 00:19:09,000
Speaker 1: a few statements for starters. Open ai is trying to

355
00:19:09,040 --> 00:19:11,600
Speaker 1: hit eleven point six billion dollars of revenue by the

356
00:19:11,680 --> 00:19:13,720
Speaker 1: end of twenty twenty five, and to do that it

357
00:19:13,720 --> 00:19:16,560
Speaker 1: will have to triple its revenue and its current cost

358
00:19:16,640 --> 00:19:19,040
Speaker 1: of revenue, so two thirty five to make a dollar,

359
00:19:19,400 --> 00:19:22,520
Speaker 1: open ai will spend twenty seven billion dollars to hit

360
00:19:22,560 --> 00:19:26,280
Speaker 1: eleven point six billion dollars, even if open ai somehow

361
00:19:26,320 --> 00:19:29,080
Speaker 1: harves their costs, and by the way, in the next

362
00:19:29,160 --> 00:19:30,760
Speaker 1: year they're going to be building a brand new model,

363
00:19:30,760 --> 00:19:32,800
Speaker 1: which will cost them an absolute shit ton of money.

364
00:19:33,000 --> 00:19:35,440
Speaker 1: They will still lose two billion dollars in twenty twenty

365
00:19:35,520 --> 00:19:39,040
Speaker 1: five to hit these revenues. However, I must be clear,

366
00:19:39,720 --> 00:19:42,639
Speaker 1: these costs are going up. They need this company to

367
00:19:42,640 --> 00:19:45,440
Speaker 1: grow by three hundred percent and the only way they're

368
00:19:45,440 --> 00:19:47,960
Speaker 1: going to do that is by making a new model.

369
00:19:48,480 --> 00:19:50,960
Speaker 1: And the only thing that's really growing at this company

370
00:19:51,000 --> 00:19:53,720
Speaker 1: is they're free users, and those free users lose the

371
00:19:53,840 --> 00:19:57,760
Speaker 1: money every single time they use it. And even if

372
00:19:57,800 --> 00:20:00,800
Speaker 1: they added a two dollars price increase to chat GPT

373
00:20:00,880 --> 00:20:04,560
Speaker 1: plus and a similar price on their business plans like

374
00:20:04,640 --> 00:20:08,120
Speaker 1: teams and enterprise, these things aren't really going to significantly

375
00:20:08,200 --> 00:20:10,720
Speaker 1: move the needle without also adding a bunch of growth

376
00:20:10,720 --> 00:20:14,480
Speaker 1: on top. An open Aiy's latest GPT four model it

377
00:20:14,520 --> 00:20:17,320
Speaker 1: cost them one hundred million dollars to train, and more

378
00:20:17,320 --> 00:20:20,159
Speaker 1: complex models like the Orian one they're allegedly coming up

379
00:20:20,200 --> 00:20:22,320
Speaker 1: with and I imagine oh one which will get to

380
00:20:22,440 --> 00:20:24,960
Speaker 1: next episode, well, they're going to cost them an absolute

381
00:20:25,000 --> 00:20:27,760
Speaker 1: shit ter more to build. And also, by the way,

382
00:20:27,800 --> 00:20:30,800
Speaker 1: the information estimated back in July the open AI's training

383
00:20:30,800 --> 00:20:33,720
Speaker 1: costs would be about three billion dollars in twenty twenty four.

384
00:20:34,080 --> 00:20:37,760
Speaker 1: The costs are not coming down. People. Now here's another problem.

385
00:20:38,160 --> 00:20:41,040
Speaker 1: Buzzy tech companies need exciting things to get investors in,

386
00:20:41,080 --> 00:20:44,640
Speaker 1: customers jazzed and opening eye. Well, they haven't really had

387
00:20:44,720 --> 00:20:49,000
Speaker 1: anything exciting or truly important since the launch of GPT

388
00:20:49,160 --> 00:20:52,720
Speaker 1: three point five, and even their latest reasoning model, like

389
00:20:52,800 --> 00:20:54,479
Speaker 1: I said, I'll get to it next episode, it's not

390
00:20:54,480 --> 00:20:58,400
Speaker 1: been particularly impressive. That model, by the way, is much

391
00:20:58,440 --> 00:21:00,679
Speaker 1: more expensive to run and uses something called chain of

392
00:21:00,760 --> 00:21:04,800
Speaker 1: thought reasoning, which I'll get into. But nevertheless, it takes

393
00:21:04,840 --> 00:21:07,760
Speaker 1: more power to run because they're doing way more calculations

394
00:21:08,400 --> 00:21:11,000
Speaker 1: and they didn't even have a use case when they

395
00:21:11,000 --> 00:21:14,119
Speaker 1: announced it. What is going on with this company? But

396
00:21:14,240 --> 00:21:19,160
Speaker 1: putting that aside, open AI's products are also becoming increasingly commoditized,

397
00:21:19,200 --> 00:21:22,400
Speaker 1: with Google, Meta, Amazon, and even Microsoft building their own

398
00:21:22,480 --> 00:21:25,800
Speaker 1: large language models and other models to compete. Where still

399
00:21:25,920 --> 00:21:29,480
Speaker 1: these models are using effectively identical training data, which they're

400
00:21:29,520 --> 00:21:32,640
Speaker 1: also running out of, which makes their outputs and by extension,

401
00:21:32,680 --> 00:21:36,920
Speaker 1: the technology itself kind of similar. But most worrying of all,

402
00:21:37,160 --> 00:21:39,200
Speaker 1: and I'm really going to drum on this one later.

403
00:21:39,560 --> 00:21:42,280
Speaker 1: Open AI's cloud business, So the one where they connect

404
00:21:42,480 --> 00:21:44,960
Speaker 1: their models to other businesses and then the businesses sell

405
00:21:45,040 --> 00:21:50,880
Speaker 1: products to customers using those models technology. It's small, It's

406
00:21:50,960 --> 00:21:53,720
Speaker 1: really small. It's small to the point that it suggests

407
00:21:53,800 --> 00:21:57,000
Speaker 1: there's a fundamental weakness in the generative AI industry. It's

408
00:21:57,000 --> 00:21:59,919
Speaker 1: extremely worrying that the biggest player in the game, oh

409
00:22:00,000 --> 00:22:03,240
Speaker 1: only makes a billion dollars less than thirty percent of

410
00:22:03,240 --> 00:22:08,000
Speaker 1: its revenue from selling access to its supposedly innovative technology.

411
00:22:08,920 --> 00:22:13,199
Speaker 1: And fundamentally, I can't really find any compelling evidence that

412
00:22:13,240 --> 00:22:15,080
Speaker 1: suggests that open ai is going to be able to

413
00:22:15,119 --> 00:22:18,880
Speaker 1: sustain this growth. In fact, I really can't find any

414
00:22:18,960 --> 00:22:22,919
Speaker 1: historical comparison for this company. And I also feel like

415
00:22:23,000 --> 00:22:26,280
Speaker 1: open AI's growth is already stumbling. But if you don't

416
00:22:26,280 --> 00:22:28,800
Speaker 1: want to stumble through life bereft of joy or growth

417
00:22:28,880 --> 00:22:31,960
Speaker 1: or happiness, I really recommend you buy one of the

418
00:22:32,000 --> 00:22:34,439
Speaker 1: following things. And if you can't buy it, download it.

419
00:22:34,680 --> 00:22:37,399
Speaker 1: If you can't download it, go into a shop and

420
00:22:37,440 --> 00:22:39,879
Speaker 1: demand that they start stocking it. And if they'll just

421
00:22:39,960 --> 00:22:42,159
Speaker 1: give it to you, demand that you can give them money.

422
00:22:42,400 --> 00:22:58,119
Speaker 1: Listen to the advertisers. All right, we're back, So for

423
00:22:58,160 --> 00:23:00,160
Speaker 1: the next part of this episode, I'm going to begin

424
00:23:00,280 --> 00:23:02,960
Speaker 1: to exactly how open ai makes money, which is going

425
00:23:03,040 --> 00:23:05,640
Speaker 1: to require me to lay out its various businesses and products,

426
00:23:05,680 --> 00:23:07,800
Speaker 1: and it's going to have a lot of numbers. Trust me,

427
00:23:07,920 --> 00:23:12,200
Speaker 1: it's worth it, because this shit is insane. So, according

428
00:23:12,240 --> 00:23:14,760
Speaker 1: to the New York Times, open ai expects chat GPT

429
00:23:14,880 --> 00:23:17,760
Speaker 1: to make about two point seven billion dollars in revenue

430
00:23:17,760 --> 00:23:20,320
Speaker 1: in twenty twenty four, with an additional billion dollars in

431
00:23:20,359 --> 00:23:24,399
Speaker 1: revenue coming from other businesses using open AI's technology. Roughly

432
00:23:24,440 --> 00:23:27,679
Speaker 1: seventy three percent, or two point seven billion dollars of

433
00:23:27,720 --> 00:23:30,640
Speaker 1: open AI's revenue comes from selling premium versions of chat

434
00:23:30,680 --> 00:23:35,240
Speaker 1: jupt called plus Teams and Enterprise. Chat Jupt Plus is

435
00:23:35,280 --> 00:23:38,280
Speaker 1: marketed to individuals for twenty bucks a month, offering faster

436
00:23:38,400 --> 00:23:41,800
Speaker 1: response times, priority, access to new features, and capability is

437
00:23:41,840 --> 00:23:44,480
Speaker 1: not found in the free product, like image generation, which

438
00:23:44,520 --> 00:23:47,879
Speaker 1: you can get in so many different other places. Importantly,

439
00:23:48,080 --> 00:23:50,400
Speaker 1: open ai can use anything you'd do as training data

440
00:23:50,480 --> 00:23:54,359
Speaker 1: unless you explicitly opt out. Open Ai also sells access

441
00:23:54,400 --> 00:23:57,480
Speaker 1: to a team's product, which costs twenty five dollars a

442
00:23:57,560 --> 00:23:59,720
Speaker 1: user a month if pay annually, so three hundred bucks

443
00:23:59,760 --> 00:24:02,360
Speaker 1: a year per user and thirty dollars a user a month.

444
00:24:02,400 --> 00:24:04,840
Speaker 1: If paid monthly from this point on, your date is

445
00:24:04,880 --> 00:24:08,720
Speaker 1: excluded from that used to train open AI's models. By deform.

446
00:24:09,000 --> 00:24:12,360
Speaker 1: Open ai sells enterprise subscriptions that include an expanded context

447
00:24:12,400 --> 00:24:15,920
Speaker 1: window for longer prompts, meaning you can give more detailed instructions,

448
00:24:16,080 --> 00:24:20,080
Speaker 1: admin controls, and enhance support an ongoing account management. It

449
00:24:20,119 --> 00:24:22,480
Speaker 1: isn't clear how much this costs, but I found a

450
00:24:22,480 --> 00:24:24,520
Speaker 1: Reddit thread from a year ago that suggests it's about

451
00:24:24,560 --> 00:24:26,600
Speaker 1: sixty dollars a user a month, with a minimum of

452
00:24:26,640 --> 00:24:29,399
Speaker 1: one hundred and fifty seats on an annual contract. I

453
00:24:29,440 --> 00:24:32,160
Speaker 1: also believe open ai offers discounts for customers who buy

454
00:24:32,200 --> 00:24:35,480
Speaker 1: in bulk very standard software as a service business model.

455
00:24:37,800 --> 00:24:40,200
Speaker 1: A further billion dollars, or twenty seven percent of open

456
00:24:40,240 --> 00:24:43,159
Speaker 1: AI's revenue comes from open ai licensing its models and

457
00:24:43,200 --> 00:24:46,800
Speaker 1: services via its API. One thing you notice when you

458
00:24:46,840 --> 00:24:48,800
Speaker 1: look at this price page is that there's this huge

459
00:24:48,880 --> 00:24:51,840
Speaker 1: variety of models and APIs available, and that there's massive

460
00:24:51,840 --> 00:24:55,360
Speaker 1: amounts of variation in pricing too. I look into these

461
00:24:55,359 --> 00:24:57,240
Speaker 1: in depth in my newsletter, which by the way, is

462
00:24:57,240 --> 00:24:59,680
Speaker 1: called open ai is a bad business and it's worth

463
00:24:59,720 --> 00:25:01,879
Speaker 1: reading if you want to know, if not, just for

464
00:25:01,920 --> 00:25:04,639
Speaker 1: the fact that the pricing in its various services is

465
00:25:04,640 --> 00:25:07,000
Speaker 1: also super messy, but you don't need to know that

466
00:25:07,119 --> 00:25:10,560
Speaker 1: right now. Open Ai also makes about two hundred million

467
00:25:10,600 --> 00:25:13,000
Speaker 1: a year selling access to its models through its Microsoft

468
00:25:13,040 --> 00:25:16,199
Speaker 1: according to Bloomberg, where open Ai takes a twenty percent

469
00:25:16,240 --> 00:25:18,639
Speaker 1: cut of all revenue, and by the way, that's not

470
00:25:18,760 --> 00:25:21,480
Speaker 1: on top of the billion. This means that open ai

471
00:25:21,680 --> 00:25:23,959
Speaker 1: only makes about eight hundred million dollars a year by

472
00:25:24,000 --> 00:25:26,800
Speaker 1: selling access to its API. That two hundred million is

473
00:25:26,800 --> 00:25:31,000
Speaker 1: on top I want to add, how worrying this is

474
00:25:31,040 --> 00:25:34,040
Speaker 1: both open Ai and the larger generative AI market. If

475
00:25:34,080 --> 00:25:37,200
Speaker 1: open Ai, the most prominent name in all of generative ai,

476
00:25:37,880 --> 00:25:40,920
Speaker 1: is only making a billion dollars a year from selling

477
00:25:41,280 --> 00:25:44,160
Speaker 1: the shovels for the gold rush, what does that say

478
00:25:44,160 --> 00:25:47,000
Speaker 1: about the growth trajectory for this company all the actual

479
00:25:47,080 --> 00:25:50,480
Speaker 1: usage of generative AI products. That's the question for later

480
00:25:50,480 --> 00:25:54,360
Speaker 1: in the show. The first let's talk dollars. So as

481
00:25:54,400 --> 00:25:57,000
Speaker 1: it stands, open Ai makes the majority, like I said,

482
00:25:57,119 --> 00:25:59,760
Speaker 1: more than seventy percent of its revenue from selling access

483
00:26:00,040 --> 00:26:03,160
Speaker 1: to premium versions of chat GPT. A few weeks ago,

484
00:26:03,200 --> 00:26:06,080
Speaker 1: the Information Report that chat gpt plus had more than

485
00:26:06,119 --> 00:26:08,840
Speaker 1: ten million paying subscribers, and that it had one million

486
00:26:08,920 --> 00:26:12,399
Speaker 1: more than we're paying for higher priced plans for business teams.

487
00:26:14,040 --> 00:26:16,440
Speaker 1: As I've laid out previously, this means that open ai

488
00:26:16,520 --> 00:26:19,200
Speaker 1: is making about two hundred dollars million dollars a month

489
00:26:19,240 --> 00:26:22,879
Speaker 1: from consumer subscribers. But business teams is a very vague.

490
00:26:23,560 --> 00:26:25,560
Speaker 1: It isn't obvious what it means. It could mean the

491
00:26:25,600 --> 00:26:28,919
Speaker 1: team plan, it could mean the enterprise plan, and you

492
00:26:29,000 --> 00:26:32,760
Speaker 1: may think, well ed, couldn't you be horribly wrong? Couldn't

493
00:26:32,760 --> 00:26:35,840
Speaker 1: it be that they're actually ripping' they they got a

494
00:26:35,880 --> 00:26:38,920
Speaker 1: million enterprise subscribers. Wouldn't that be good? Wrong? Oh? The

495
00:26:39,000 --> 00:26:41,960
Speaker 1: New York Times has my back here. Based on their reporting,

496
00:26:42,119 --> 00:26:45,440
Speaker 1: we can actually get a little more specific. Chat gpt

497
00:26:45,480 --> 00:26:48,280
Speaker 1: plus is ten million customers, making open Ai around two

498
00:26:48,320 --> 00:26:51,240
Speaker 1: point four billion dollars a year. As ten million users

499
00:26:51,280 --> 00:26:54,480
Speaker 1: spending about twenty bucks a month, which equates two hundred million,

500
00:26:54,640 --> 00:26:57,920
Speaker 1: might apply by twelve two point four billion dollars. Dad

501
00:26:58,000 --> 00:27:01,439
Speaker 1: I swear I was listening in Mathson Times. This means,

502
00:27:01,440 --> 00:27:03,440
Speaker 1: by the way, that business users make up about three

503
00:27:03,520 --> 00:27:05,800
Speaker 1: hundred million dollars a year in revenue. Of twenty five

504
00:27:05,840 --> 00:27:09,400
Speaker 1: million dollars a month, which isn't really great. While ten

505
00:27:09,440 --> 00:27:12,880
Speaker 1: million paying customers might seem like a lot. Chat GPT

506
00:27:13,320 --> 00:27:15,960
Speaker 1: is effectively to generative AI what Google is to search.

507
00:27:16,240 --> 00:27:19,680
Speaker 1: Ten million people paying for this kind of table stakes,

508
00:27:19,880 --> 00:27:21,880
Speaker 1: and the idea that this company can triple that number

509
00:27:21,880 --> 00:27:25,720
Speaker 1: in a year is absolutely ridiculous. Open ai has been

510
00:27:25,840 --> 00:27:28,560
Speaker 1: covered by effectively every media outlet in the world. It's

511
00:27:28,600 --> 00:27:31,680
Speaker 1: mentioned in almost every single conversation about AI, even when

512
00:27:31,680 --> 00:27:33,840
Speaker 1: it's not about generative AAI, and has the backing and

513
00:27:33,880 --> 00:27:36,720
Speaker 1: marketing push of Microsoft and pretty much the entirety of

514
00:27:36,720 --> 00:27:40,919
Speaker 1: Silicon Valley. Chat GPT has over two hundred million weekly users,

515
00:27:41,119 --> 00:27:43,359
Speaker 1: and The New York Times reports that open ai has

516
00:27:43,440 --> 00:27:46,359
Speaker 1: and I quote, three hundred and fifty million people using

517
00:27:46,359 --> 00:27:49,560
Speaker 1: their services each month as of joom though it's unclear

518
00:27:49,640 --> 00:27:52,320
Speaker 1: if that includes those using the API. I would guess

519
00:27:52,320 --> 00:27:55,920
Speaker 1: that's chat GPT users collectively. This means that open ai,

520
00:27:56,160 --> 00:27:58,840
Speaker 1: the most popular company in the industry, in the most

521
00:27:58,880 --> 00:28:02,160
Speaker 1: prevalent industry in the industry, talked about to everyone and everywhere,

522
00:28:02,280 --> 00:28:04,720
Speaker 1: can only convert about three percent of its free users

523
00:28:04,760 --> 00:28:09,440
Speaker 1: into paying customers. Now, by the way, Eager listeners may go, well,

524
00:28:09,480 --> 00:28:14,680
Speaker 1: three twenty percent conversion, that's not bad, right, right, wrong. Generally,

525
00:28:14,760 --> 00:28:19,320
Speaker 1: free products don't cost this much to run. Chat GPT

526
00:28:19,480 --> 00:28:23,440
Speaker 1: itself is extremely expensive. Whether it's free or whether it's premium,

527
00:28:23,560 --> 00:28:26,760
Speaker 1: it's the same thing. So yeah, those three hundred and

528
00:28:26,800 --> 00:28:30,320
Speaker 1: fifty million people, they're more of like a parasite than anything.

529
00:28:30,920 --> 00:28:33,240
Speaker 1: The ninety seven percent who won't give them the credit card.

530
00:28:33,320 --> 00:28:36,600
Speaker 1: It's not great. And I think that lack of conversion

531
00:28:36,720 --> 00:28:39,760
Speaker 1: might be because chat GPT and chat GPT Plus are

532
00:28:39,760 --> 00:28:43,120
Speaker 1: really similar products. Chat GPT Plus that you use the

533
00:28:43,120 --> 00:28:45,760
Speaker 1: product more often and have access to numer models, but

534
00:28:45,840 --> 00:28:48,560
Speaker 1: there's no obvious new thing you can do. As a result,

535
00:28:48,880 --> 00:28:51,880
Speaker 1: you can't really upsell to plus unless you found someone

536
00:28:51,920 --> 00:28:54,760
Speaker 1: who's already worked out the use case for themselves. And

537
00:28:54,840 --> 00:28:58,200
Speaker 1: open Ai remains piss poor at actually marketing this product,

538
00:28:58,640 --> 00:29:01,120
Speaker 1: mostly because of the limitations of what a large language

539
00:29:01,120 --> 00:29:04,240
Speaker 1: model can do. As a result, most people diddling with

540
00:29:04,320 --> 00:29:06,280
Speaker 1: chat GBT will get what they need to out the

541
00:29:06,320 --> 00:29:09,600
Speaker 1: free version. I'd also argue that those willing to pay

542
00:29:09,640 --> 00:29:11,960
Speaker 1: for a Plus subscription are more likely to use the

543
00:29:12,000 --> 00:29:15,280
Speaker 1: platform way way more than free users, and because every

544
00:29:15,320 --> 00:29:18,760
Speaker 1: prompt on chat, GBT loses the company money. It's reasonable

545
00:29:18,760 --> 00:29:20,560
Speaker 1: to believe that paying users would be far more of

546
00:29:20,560 --> 00:29:22,960
Speaker 1: a burden on the system. While there's a chance to

547
00:29:23,000 --> 00:29:24,600
Speaker 1: open an eye, could have a chunk of users that

548
00:29:24,600 --> 00:29:27,600
Speaker 1: aren't particularly active. One cannot run a business based on

549
00:29:27,640 --> 00:29:30,560
Speaker 1: selling stuff you hope that people won't use. And no, no,

550
00:29:30,640 --> 00:29:33,800
Speaker 1: it's nothing like a gym or insurance. Stop it. When

551
00:29:33,800 --> 00:29:36,640
Speaker 1: I wrote the newsletter and I said that, I got

552
00:29:36,640 --> 00:29:38,800
Speaker 1: so many people who thought they were the cleverest people

553
00:29:38,800 --> 00:29:41,440
Speaker 1: at all school. Oh, it's like a gym. It's like, sure,

554
00:29:41,480 --> 00:29:44,719
Speaker 1: it's not the same. Stop it. You're not clever. No

555
00:29:44,720 --> 00:29:47,840
Speaker 1: one's impressed, And I should be clear. There's also a

556
00:29:47,880 --> 00:29:51,880
Speaker 1: reason why enterprise customers are generally more desirable than private individuals.

557
00:29:52,480 --> 00:29:55,920
Speaker 1: As with any other consumer centric subscription product, regular customers

558
00:29:55,920 --> 00:29:57,800
Speaker 1: are far more likely to cut their spending when they

559
00:29:58,160 --> 00:30:00,600
Speaker 1: don't feel like they're getting value from the product, or

560
00:30:00,840 --> 00:30:04,280
Speaker 1: when the household budget demands it, or there's economic problems.

561
00:30:04,440 --> 00:30:06,280
Speaker 1: Just look at Netflix. They were the biggest name while

562
00:30:06,320 --> 00:30:08,200
Speaker 1: they still are the biggest name in streaming, and they

563
00:30:08,240 --> 00:30:10,880
Speaker 1: lost a million customers in twenty twenty two because of

564
00:30:10,920 --> 00:30:13,920
Speaker 1: the cost of living crisis. These are real things that

565
00:30:14,000 --> 00:30:17,040
Speaker 1: will happen to open ai if they're not already happening.

566
00:30:17,680 --> 00:30:20,960
Speaker 1: Chat GPT plus is likely for many people, kind of

567
00:30:20,960 --> 00:30:23,960
Speaker 1: a lifestyle product, and the problem is that when people

568
00:30:24,000 --> 00:30:27,120
Speaker 1: lose their jobs, or inflation hikes happen, or cheaper things

569
00:30:27,160 --> 00:30:29,800
Speaker 1: come along that do much the same thing, these products

570
00:30:29,840 --> 00:30:32,040
Speaker 1: are the first to get slashed from the budget. And

571
00:30:32,080 --> 00:30:36,200
Speaker 1: that's before any arbitrary or silly or desperate price increase

572
00:30:36,240 --> 00:30:39,960
Speaker 1: is done by a company that isn't the good business. Honestly,

573
00:30:40,000 --> 00:30:42,360
Speaker 1: it's kind of remarkable that open ai found ten million

574
00:30:42,440 --> 00:30:45,840
Speaker 1: people to actually pay for chat gpt, But how do

575
00:30:45,880 --> 00:30:48,320
Speaker 1: you grow that to twenty million or forty million people.

576
00:30:48,400 --> 00:30:50,800
Speaker 1: I think they need like thirty million by the end

577
00:30:50,840 --> 00:30:55,040
Speaker 1: of twenty twenty five. How does that happen? Very worrying anyway.

578
00:30:55,080 --> 00:30:57,320
Speaker 1: At present, open ai makes about two hundred and twenty

579
00:30:57,320 --> 00:30:59,600
Speaker 1: five million dollars a month. That's two point seven billion

580
00:30:59,600 --> 00:31:03,360
Speaker 1: a year. By selling these premium subscriptions. To hit eleven

581
00:31:03,400 --> 00:31:06,120
Speaker 1: point six billion dollars in twenty twenty five, open ai

582
00:31:06,200 --> 00:31:09,080
Speaker 1: would have to increase revenue from chat GPT customers by

583
00:31:09,120 --> 00:31:12,520
Speaker 1: three hundred and ten percent. If we consider and forgive me.

584
00:31:12,560 --> 00:31:14,240
Speaker 1: I'm going to do some maths at you. The current

585
00:31:14,320 --> 00:31:17,560
Speaker 1: ratio of plus subscriptions, the teams and enterprise subscriptions that

586
00:31:17,720 --> 00:31:21,200
Speaker 1: eighty eight point eight nine percent is GPT plus versus

587
00:31:21,280 --> 00:31:24,080
Speaker 1: the business ones at eleven point eleven percent. Open Ai

588
00:31:24,160 --> 00:31:27,720
Speaker 1: would need to find eighteen point two nine million paying users,

589
00:31:27,800 --> 00:31:29,520
Speaker 1: and that's at the new price point of twenty two

590
00:31:29,560 --> 00:31:33,280
Speaker 1: bucks a month, while also retaining every single one of

591
00:31:33,320 --> 00:31:37,240
Speaker 1: the current subscribers to Chat GPT plus, who would also

592
00:31:37,320 --> 00:31:39,920
Speaker 1: need to renew at the same price point to hit

593
00:31:40,000 --> 00:31:42,440
Speaker 1: seven point four billion dollars or about six hundred and

594
00:31:42,480 --> 00:31:45,040
Speaker 1: sixteen million dollars a month. It would also have to

595
00:31:45,040 --> 00:31:47,480
Speaker 1: make an additional nine hundred and thirty three million dollars

596
00:31:47,480 --> 00:31:50,240
Speaker 1: in revenue from its business or enterprice clients, which again

597
00:31:50,520 --> 00:31:53,120
Speaker 1: would require open Ai to more than triple their users

598
00:31:53,160 --> 00:31:56,840
Speaker 1: and retain the current ones to triple these users. To

599
00:31:56,880 --> 00:32:00,080
Speaker 1: actually do this, Chat GPT has to meaningfully change, It

600
00:32:00,120 --> 00:32:04,000
Speaker 1: has to do so soon, or disclose multiple meaningful, powerful

601
00:32:04,120 --> 00:32:06,920
Speaker 1: use cases that are so impressive that eighteen million new

602
00:32:06,960 --> 00:32:09,040
Speaker 1: people agreed to give them twenty two dollars a month.

603
00:32:09,840 --> 00:32:13,640
Speaker 1: That's an incredible, and I would say insane goal and

604
00:32:13,680 --> 00:32:16,160
Speaker 1: one that I do not think this company is capable

605
00:32:16,160 --> 00:32:20,360
Speaker 1: of achieving. It would require making chat GPT something far

606
00:32:20,400 --> 00:32:22,520
Speaker 1: more compelling than it already is, in ways that I

607
00:32:22,560 --> 00:32:26,160
Speaker 1: can't even imagine. Just as the company lost multiple members

608
00:32:26,200 --> 00:32:30,360
Speaker 1: of its senior leadership team, it doesn't look good. There

609
00:32:30,360 --> 00:32:35,200
Speaker 1: needs to be a meaningful change. Now. I know I've

610
00:32:35,200 --> 00:32:37,440
Speaker 1: already given you a lot of worrying things about open

611
00:32:37,480 --> 00:32:42,440
Speaker 1: aiye about their very top heavy, subscription driven business, Why

612
00:32:42,480 --> 00:32:44,760
Speaker 1: it isn't going to grow, why they need to grow

613
00:32:44,840 --> 00:32:46,880
Speaker 1: it to grow faster than any one's ever grown, why

614
00:32:46,880 --> 00:32:49,200
Speaker 1: they need to stop losing so much money. But there's

615
00:32:49,200 --> 00:32:52,760
Speaker 1: something else to be worried about. There's something even flimsier

616
00:32:53,000 --> 00:32:58,240
Speaker 1: about this business, and it's actually something that's disastrous. It's disastrous.

617
00:32:58,280 --> 00:33:02,200
Speaker 1: I'm serious. It is simply disastrous. How little of open

618
00:33:02,240 --> 00:33:05,320
Speaker 1: AI's revenue comes from providing other companies that means to

619
00:33:05,360 --> 00:33:09,680
Speaker 1: integrate generative AI into their systems. It is astonishingly bad.

620
00:33:09,880 --> 00:33:13,600
Speaker 1: I cannot be clear enough here. Before I wrote this script,

621
00:33:14,040 --> 00:33:17,280
Speaker 1: before I wrote the newsletter that led to it, I believed,

622
00:33:17,440 --> 00:33:20,160
Speaker 1: in my heart of hearts that how OpenAI made their

623
00:33:20,200 --> 00:33:24,080
Speaker 1: money was from providing access to their models. That makes sense, right,

624
00:33:24,720 --> 00:33:27,760
Speaker 1: because this company's talked about everywhere. Their technology is meant

625
00:33:27,760 --> 00:33:30,360
Speaker 1: to be innovative. Everyone's meant to use it, right, that

626
00:33:30,400 --> 00:33:33,680
Speaker 1: makes sense. Surely most of their money would come from

627
00:33:33,840 --> 00:33:37,080
Speaker 1: people getting in on the revolution and then integrating the

628
00:33:37,080 --> 00:33:43,560
Speaker 1: revolution into their products, and then people using the revolutionary thing. Right. Wrong.

629
00:33:44,160 --> 00:33:47,240
Speaker 1: First of all, if open ai only makes a billion

630
00:33:47,320 --> 00:33:50,320
Speaker 1: dollars a year selling API access and thus let you

631
00:33:50,400 --> 00:33:53,000
Speaker 1: integrate the models into your products, it suggests that even

632
00:33:53,040 --> 00:33:57,040
Speaker 1: the biggest company in generative ai cannot find enough customers

633
00:33:57,040 --> 00:34:00,440
Speaker 1: to make its business viable. Secondly, it's yes, there's a

634
00:34:00,480 --> 00:34:04,880
Speaker 1: remarkably small amount of demand for GENERATIVEAI integrations. Or consider

635
00:34:04,920 --> 00:34:07,880
Speaker 1: another way that the companies connecting to open ai aren't

636
00:34:07,920 --> 00:34:11,560
Speaker 1: really making it very much money at all. If the

637
00:34:11,560 --> 00:34:15,520
Speaker 1: GENERATIVEAI revolution were really here, surely open ai, the household

638
00:34:15,600 --> 00:34:18,240
Speaker 1: name in large language models, would be rolling in cash

639
00:34:18,239 --> 00:34:21,160
Speaker 1: from their ABI business, not leaving it at less than

640
00:34:21,200 --> 00:34:24,919
Speaker 1: thirty percent of their and your revenue. This isn't the case.

641
00:34:25,080 --> 00:34:28,680
Speaker 1: It's so strange. So at the twenty twenty four open

642
00:34:28,680 --> 00:34:31,359
Speaker 1: Ai Dev Day event, it's developer conference, which took place

643
00:34:31,360 --> 00:34:34,320
Speaker 1: in October, first, the company said that over three million

644
00:34:34,360 --> 00:34:37,759
Speaker 1: developers are building apps using open AI's infrastructure. That works

645
00:34:37,800 --> 00:34:40,440
Speaker 1: out to about three hundred and thirty three dollars a developer,

646
00:34:40,719 --> 00:34:42,960
Speaker 1: and that's far less money than other companies which make

647
00:34:43,000 --> 00:34:46,640
Speaker 1: money by providing a service through their API actually make. Twilio,

648
00:34:46,760 --> 00:34:49,840
Speaker 1: a company that makes its money sending SMS, messages and

649
00:34:49,840 --> 00:34:53,319
Speaker 1: push notifications for companies, over the past quarter, made about

650
00:34:53,320 --> 00:34:56,879
Speaker 1: a billion dollars in revenue. That's what OpenAI made from

651
00:34:56,880 --> 00:34:59,760
Speaker 1: renting out its models and APIs over the past year.

652
00:35:01,080 --> 00:35:03,680
Speaker 1: Twilio also made roughly four billion dollars over the last

653
00:35:03,760 --> 00:35:06,480
Speaker 1: four quarters, which is more than open AI's projected revenue

654
00:35:06,640 --> 00:35:09,800
Speaker 1: for the entirety of twenty twenty four. As I mentioned before,

655
00:35:10,000 --> 00:35:12,399
Speaker 1: open ai makes two hundred million dollars of its one

656
00:35:12,640 --> 00:35:16,120
Speaker 1: billion dollar revenue from Microsoft reselling its models a twenty

657
00:35:16,120 --> 00:35:18,800
Speaker 1: percent cut. That suggests that Microsoft two is making about

658
00:35:19,160 --> 00:35:23,680
Speaker 1: a billion dollars from open aised models. So in the

659
00:35:23,719 --> 00:35:25,920
Speaker 1: event that open ai and Microsoft are making about a

660
00:35:25,920 --> 00:35:29,080
Speaker 1: billion dollars in annualized revenue by providing access to open

661
00:35:29,120 --> 00:35:32,480
Speaker 1: AI's models. Again, these are estimates based on current growth trajectories.

662
00:35:33,360 --> 00:35:36,480
Speaker 1: It suggests that there's only two billion dollars in annual

663
00:35:36,480 --> 00:35:39,680
Speaker 1: revenue coming from both of these companies combined, and that's

664
00:35:39,800 --> 00:35:45,520
Speaker 1: without wait is open ai making less money from open

665
00:35:45,600 --> 00:35:50,520
Speaker 1: AI's models than Microsoft is. Oh my god, this business

666
00:35:50,600 --> 00:35:54,120
Speaker 1: is a stinker. And this also suggests that generative AI

667
00:35:54,280 --> 00:35:58,480
Speaker 1: as a technology doesn't have a product market fit. According

668
00:35:58,480 --> 00:36:00,719
Speaker 1: to a survey by Andrews and Hory It's a VC

669
00:36:00,920 --> 00:36:03,600
Speaker 1: earlier in the year and I quote, the twenty twenty

670
00:36:03,600 --> 00:36:06,400
Speaker 1: three market share of closed source models was estimated at

671
00:36:06,400 --> 00:36:08,880
Speaker 1: eighty to ninety percent, with the majority of that share

672
00:36:09,000 --> 00:36:12,440
Speaker 1: going to open ai and open source ones would include

673
00:36:12,440 --> 00:36:15,240
Speaker 1: things like metas Lama model, which is kind of open source.

674
00:36:15,719 --> 00:36:19,280
Speaker 1: That's a whole other thing anyway. Another survey from IoT

675
00:36:19,360 --> 00:36:22,200
Speaker 1: Analytics published late last year suggested that the number might

676
00:36:22,200 --> 00:36:24,239
Speaker 1: look a little different, with thirty nine percent of the

677
00:36:24,239 --> 00:36:26,600
Speaker 1: market share going to open Ai and thirty percent going

678
00:36:26,600 --> 00:36:30,080
Speaker 1: to Microsoft. Assuming that the latter numbers are true or

679
00:36:30,120 --> 00:36:33,680
Speaker 1: even close to true, this suggests that the generative AI

680
00:36:33,840 --> 00:36:38,040
Speaker 1: market is really small. If open Ai, which dominates with

681
00:36:38,160 --> 00:36:40,560
Speaker 1: I'd wager about thirty percent of the market, is only

682
00:36:40,600 --> 00:36:42,560
Speaker 1: making a billion dollars a year from selling access to

683
00:36:42,600 --> 00:36:45,760
Speaker 1: its models at this stage. In this massive hype bubble.

684
00:36:46,120 --> 00:36:48,640
Speaker 1: There might not even be ten billion dollars of annual

685
00:36:48,640 --> 00:36:53,200
Speaker 1: revenue from companies integrating generative AI into their products. That's tiny,

686
00:36:53,880 --> 00:36:56,359
Speaker 1: and this should be where all the money is. If

687
00:36:56,360 --> 00:36:58,960
Speaker 1: this stuff is the future of everything, why is the

688
00:36:59,000 --> 00:37:03,120
Speaker 1: revenue stream so painfully weak. Open ai is making so

689
00:37:03,280 --> 00:37:06,160
Speaker 1: little selling access to their models, and it suggests that

690
00:37:06,440 --> 00:37:09,440
Speaker 1: despite this hype cycle, there either is an interest in

691
00:37:09,440 --> 00:37:12,759
Speaker 1: integrating these products from developers, or when these integrations are

692
00:37:12,760 --> 00:37:16,880
Speaker 1: actually in there, consumers aren't really into them or using them. Remember,

693
00:37:17,080 --> 00:37:19,880
Speaker 1: these products are charged on usage, and so it's possible

694
00:37:19,960 --> 00:37:22,200
Speaker 1: for generative AI to be integrated into a service but

695
00:37:22,280 --> 00:37:25,000
Speaker 1: not actually drive much revenue for open Ai as a

696
00:37:25,000 --> 00:37:28,239
Speaker 1: result of users not really caring. While chat gpt as

697
00:37:28,280 --> 00:37:31,680
Speaker 1: brand recognition, companies integrating open AI's models into their products

698
00:37:31,719 --> 00:37:33,840
Speaker 1: are far more indicative of the long term health of

699
00:37:33,880 --> 00:37:37,359
Speaker 1: both the company and the industry itself. Because if open

700
00:37:37,400 --> 00:37:39,680
Speaker 1: ai can't convince people to integrate and use this shit,

701
00:37:39,719 --> 00:37:42,719
Speaker 1: do you think others are succeeding? I mean, think about it.

702
00:37:43,000 --> 00:37:45,400
Speaker 1: Where have you seen some weird chatbot up here in

703
00:37:45,440 --> 00:37:48,799
Speaker 1: your life? Where have you seen an LM poked into

704
00:37:48,840 --> 00:37:51,759
Speaker 1: something you use? Have you been like, oh good, I

705
00:37:51,800 --> 00:37:54,160
Speaker 1: can't wait, or have you just kind of tried to

706
00:37:54,200 --> 00:37:56,920
Speaker 1: ignore it. Maybe you've interfaced with it. It sucks. There

707
00:37:56,960 --> 00:37:59,480
Speaker 1: are some decent products, like there are email clients that

708
00:37:59,480 --> 00:38:03,560
Speaker 1: summarize emails Microsoft teams Apparently people really like the summarization.

709
00:38:03,680 --> 00:38:06,920
Speaker 1: But we are meant to be describing the future here.

710
00:38:06,960 --> 00:38:10,000
Speaker 1: We're meant to be describing something exciting and sexy. Not huh.

711
00:38:10,360 --> 00:38:14,200
Speaker 1: This summarized the meeting for me interesting. But looking at

712
00:38:14,239 --> 00:38:16,799
Speaker 1: these numbers, it's hard to imagine how open ai will

713
00:38:16,840 --> 00:38:19,200
Speaker 1: more than triple revenue in the next fifteen months to

714
00:38:19,280 --> 00:38:23,600
Speaker 1: hit eleven point six billion dollars in sales. Furthermore, at

715
00:38:23,640 --> 00:38:26,319
Speaker 1: its current burn ray, open ai is currently spending, like

716
00:38:26,360 --> 00:38:28,319
Speaker 1: I said, two dollars and thirty five cents to make

717
00:38:28,320 --> 00:38:31,040
Speaker 1: a buck, meaning that eleven point six billion dollars in

718
00:38:31,080 --> 00:38:33,760
Speaker 1: revenue could cost as much as twenty seven billion dollars

719
00:38:33,800 --> 00:38:37,520
Speaker 1: to actually make. And as I previously mentioned, what costs

720
00:38:37,520 --> 00:38:41,040
Speaker 1: could foreseeably come down, All signs point to them increasing.

721
00:38:41,840 --> 00:38:44,440
Speaker 1: GPT four costs one hundred million dollars to train, and

722
00:38:44,520 --> 00:38:47,000
Speaker 1: more complex future models will cost hundreds of millions or

723
00:38:47,040 --> 00:38:50,960
Speaker 1: billions to train, as I mentioned the Information said that

724
00:38:51,080 --> 00:38:54,239
Speaker 1: training costs were would look like three billion dollars in

725
00:38:54,280 --> 00:38:56,600
Speaker 1: twenty twenty four, and I think it's fair to assume

726
00:38:56,640 --> 00:38:58,440
Speaker 1: that the new models are just as costly, if not

727
00:38:58,520 --> 00:39:02,160
Speaker 1: more so. Ye there's something deliciously ironic about all of

728
00:39:02,200 --> 00:39:04,680
Speaker 1: this that, despite a clear lack of user interest in

729
00:39:04,719 --> 00:39:08,239
Speaker 1: generative AI, open Ahi's global marketing push has succeeded in

730
00:39:08,360 --> 00:39:11,080
Speaker 1: making lots of people intrigued enough to try a completely

731
00:39:11,080 --> 00:39:14,479
Speaker 1: free service that only loses the money. While it's usually

732
00:39:14,520 --> 00:39:16,719
Speaker 1: great news that a product has three hundred or more

733
00:39:16,840 --> 00:39:20,800
Speaker 1: million free users, every book time somebody uses a service

734
00:39:21,080 --> 00:39:24,120
Speaker 1: like chet GPT, it loses the company money. And in

735
00:39:24,160 --> 00:39:27,000
Speaker 1: the case of chet GPT, good lord, they must be

736
00:39:27,040 --> 00:39:30,319
Speaker 1: losing so much. The Information estimated in July the open

737
00:39:30,360 --> 00:39:33,480
Speaker 1: Ai will spend around four billion dollars in server costs

738
00:39:33,480 --> 00:39:35,879
Speaker 1: in twenty twenty four to run chat GPT and host

739
00:39:35,960 --> 00:39:39,480
Speaker 1: other companies running their services using GPT in its other models,

740
00:39:39,760 --> 00:39:43,120
Speaker 1: effectively meaning that every dollar of revenue is immediately eaten

741
00:39:43,120 --> 00:39:45,440
Speaker 1: by the costs of acquiring it, and that's before your

742
00:39:45,440 --> 00:39:47,920
Speaker 1: factor in more than fifteen hundred people, as many as

743
00:39:47,960 --> 00:39:50,920
Speaker 1: seventeen hundred people now work at open Ai, which is

744
00:39:50,960 --> 00:39:54,320
Speaker 1: another one point five billion dollars or more in costs,

745
00:39:55,320 --> 00:39:57,279
Speaker 1: and other costs, by the way, are on top of that,

746
00:39:57,320 --> 00:40:01,000
Speaker 1: they've got real estate taxes, stock grunts. This is all

747
00:40:01,160 --> 00:40:05,080
Speaker 1: very bad and while open ay could potentially reduce costs,

748
00:40:05,600 --> 00:40:09,000
Speaker 1: they've shown no proof that they can. And they tried once,

749
00:40:09,120 --> 00:40:12,680
Speaker 1: well at least one got out, the so called iraqis

750
00:40:12,800 --> 00:40:14,719
Speaker 1: model from last year that they tried to show that

751
00:40:14,800 --> 00:40:17,160
Speaker 1: was meant to be more efficient if failed to launch.

752
00:40:17,440 --> 00:40:20,399
Speaker 1: That's not a good sign, is it. But there's one

753
00:40:20,520 --> 00:40:24,120
Speaker 1: more problem. There's still more problems, and this one, well,

754
00:40:24,160 --> 00:40:27,279
Speaker 1: this one was caused by their fundraising. You see, they

755
00:40:27,360 --> 00:40:29,839
Speaker 1: just raised six point five six point six billion dollars

756
00:40:29,880 --> 00:40:32,640
Speaker 1: in capital one hundred and fifty seven billion dollar valuation.

757
00:40:33,239 --> 00:40:35,480
Speaker 1: This means that all future rounds have to be at

758
00:40:35,520 --> 00:40:39,279
Speaker 1: that valuation or higher. A lower valuation, which is called

759
00:40:39,280 --> 00:40:42,600
Speaker 1: the down round, would make current investors quite pissy and

760
00:40:42,640 --> 00:40:44,480
Speaker 1: send a very loud signal to the market that the

761
00:40:44,480 --> 00:40:47,200
Speaker 1: company is having trouble because nobody thinks they're worth what

762
00:40:47,200 --> 00:40:50,400
Speaker 1: they used to be, which in turn would overwhelmingly suggest

763
00:40:50,480 --> 00:40:52,719
Speaker 1: open AI's only way to survive is to raise its

764
00:40:52,800 --> 00:40:56,680
Speaker 1: next valuation at a two hundred billion dollar valuation and

765
00:40:56,719 --> 00:40:59,759
Speaker 1: require yet another giant raise. I would say at least

766
00:40:59,800 --> 00:41:05,439
Speaker 1: ten billion. For context, the biggest IPO valuation in US

767
00:41:05,440 --> 00:41:08,600
Speaker 1: corporate history was Alababa, which debuted on the NASC with

768
00:41:08,680 --> 00:41:11,200
Speaker 1: a market cap of nearly one hundred and seventy billion dollars.

769
00:41:11,680 --> 00:41:14,240
Speaker 1: That figure is more than double the runner up Facebook,

770
00:41:14,280 --> 00:41:16,560
Speaker 1: which had a value of eighty one billion dollars. And

771
00:41:16,600 --> 00:41:18,719
Speaker 1: I want to give you some history on that one too.

772
00:41:19,040 --> 00:41:22,120
Speaker 1: Facebook's initial OPO was really bad because people were concerned

773
00:41:22,160 --> 00:41:26,680
Speaker 1: about mobile users and not monetizing them. Well, they won

774
00:41:26,760 --> 00:41:30,360
Speaker 1: that one, but the market's pulverized Facebook for that. Do

775
00:41:30,360 --> 00:41:32,400
Speaker 1: you think that they're going to be Oh yeah, so

776
00:41:32,520 --> 00:41:35,000
Speaker 1: your company loses five or more billion dollars a year

777
00:41:35,000 --> 00:41:37,480
Speaker 1: and you have no path of the profitability. Yes, put

778
00:41:37,480 --> 00:41:40,080
Speaker 1: you up on a Nasdaq sounds perfect. Fuck that. No,

779
00:41:40,280 --> 00:41:42,360
Speaker 1: they're not going to do that. How is this company

780
00:41:42,400 --> 00:41:44,440
Speaker 1: going to do? How is this company going to IPO?

781
00:41:45,320 --> 00:41:48,680
Speaker 1: And yet there's more problems too. I don't know how

782
00:41:48,719 --> 00:41:51,640
Speaker 1: openay is meant to convert itself from its weird nonprofit

783
00:41:51,680 --> 00:41:54,440
Speaker 1: structure into a for profit company. I don't know if

784
00:41:54,440 --> 00:41:58,400
Speaker 1: it's possible, it won't be easy. And what's crazier is

785
00:41:58,800 --> 00:42:00,640
Speaker 1: and there's so many little things like this in this

786
00:42:00,719 --> 00:42:02,319
Speaker 1: story where you're like, people are going to look back

787
00:42:02,360 --> 00:42:04,880
Speaker 1: on this and think, Wow, we were goddamn stupid. At

788
00:42:04,920 --> 00:42:06,960
Speaker 1: least I hope they did. But open ai has two

789
00:42:07,040 --> 00:42:10,640
Speaker 1: years from close to convert from a nonprofit to a

790
00:42:10,680 --> 00:42:13,920
Speaker 1: for profit or their funding will convert into debt. I

791
00:42:13,920 --> 00:42:19,640
Speaker 1: think it's six to nine percent interest. Hey, this isn't good. Also,

792
00:42:19,680 --> 00:42:21,560
Speaker 1: at some point open ai is going to have to

793
00:42:21,600 --> 00:42:23,440
Speaker 1: work out a way to go public, like I said,

794
00:42:24,480 --> 00:42:28,160
Speaker 1: because otherwise, why did people invest? Why did people bother?

795
00:42:28,920 --> 00:42:32,040
Speaker 1: They could do future secondary sales, And at some point,

796
00:42:32,080 --> 00:42:33,520
Speaker 1: if that's all they're going to be able to do

797
00:42:33,560 --> 00:42:38,160
Speaker 1: in secondary sales referring to selling private stock to another individual, well,

798
00:42:38,160 --> 00:42:40,359
Speaker 1: I mean at that point it's a Ponzi scheme. You're

799
00:42:40,400 --> 00:42:45,000
Speaker 1: just pulling in money to hand out money to other people. Eventually,

800
00:42:45,280 --> 00:42:48,239
Speaker 1: it's not good these This company has no past to profitability.

801
00:42:48,520 --> 00:42:51,759
Speaker 1: It doesn't have one. And regardless of what happens with

802
00:42:51,800 --> 00:42:54,760
Speaker 1: everything I'm mentioning, they still need to raise more funds.

803
00:42:55,160 --> 00:42:59,000
Speaker 1: Chat GPT's free version is an actual poison on open

804
00:42:59,040 --> 00:43:02,200
Speaker 1: AI's system. It's a marketing channel that burns billions of

805
00:43:02,239 --> 00:43:04,399
Speaker 1: dollars to introduce people to a product that only ten

806
00:43:04,440 --> 00:43:07,880
Speaker 1: million people will actually pay for. And open AI's future

807
00:43:07,960 --> 00:43:11,160
Speaker 1: depends largely on its ability to continue convincing people to

808
00:43:11,280 --> 00:43:14,319
Speaker 1: use it. So how does this continue? How does open

809
00:43:14,360 --> 00:43:18,880
Speaker 1: ai survive? I don't think it can. Open ai is

810
00:43:18,920 --> 00:43:21,759
Speaker 1: a disaster in the making, and behind it, it's a nastier,

811
00:43:21,880 --> 00:43:25,360
Speaker 1: shittier disaster, a lack of fundamental strength in the generative

812
00:43:25,360 --> 00:43:28,759
Speaker 1: AI market writ Large. If open ai can only make

813
00:43:28,800 --> 00:43:32,200
Speaker 1: a billion dollars as the leader in this market, and

814
00:43:32,239 --> 00:43:34,840
Speaker 1: really only eight hundred million because the other two hundred

815
00:43:34,840 --> 00:43:38,000
Speaker 1: million comes from Microsoft, it suggests that there's not really

816
00:43:38,040 --> 00:43:41,400
Speaker 1: developer or user interest in generative AI products writ Large.

817
00:43:42,320 --> 00:43:46,080
Speaker 1: Perhaps it's the hallucination problem, where it authoritatively states something

818
00:43:46,080 --> 00:43:48,759
Speaker 1: that isn't true. Well, maybe it's just that generative AI

819
00:43:48,880 --> 00:43:53,319
Speaker 1: isn't something that produces interesting interactions with a user. While

820
00:43:53,320 --> 00:43:55,520
Speaker 1: you could argue that somebody could work out a really

821
00:43:55,600 --> 00:43:58,920
Speaker 1: cool product, it's time to ask why Amazon, Google, Meta,

822
00:43:58,960 --> 00:44:02,560
Speaker 1: open Ai, Apple and Microsoft have failed to make one

823
00:44:03,520 --> 00:44:07,760
Speaker 1: one in the last two years? Where is it? Where's

824
00:44:07,760 --> 00:44:11,600
Speaker 1: the killer app? And no, it's not early days. Shut up.

825
00:44:12,000 --> 00:44:14,759
Speaker 1: More money than anything has ever raised has gone into

826
00:44:14,840 --> 00:44:18,160
Speaker 1: these companies. More attention has gone into these companies than

827
00:44:18,200 --> 00:44:23,920
Speaker 1: any other hyperscalar movement ever, even the metaverse. Where's the product? Man?

828
00:44:24,640 --> 00:44:28,000
Speaker 1: Where is it? And while this bubble can continue coasting

829
00:44:28,000 --> 00:44:30,560
Speaker 1: for a little while longer, nothing about the open ai

830
00:44:30,680 --> 00:44:33,719
Speaker 1: story looks good. This company's lost. They're bleeding money with

831
00:44:33,840 --> 00:44:37,120
Speaker 1: every single interaction with the customer, and they're flogging serve

832
00:44:37,440 --> 00:44:40,680
Speaker 1: software that's the best kind of useful and worst actively

833
00:44:40,719 --> 00:44:45,359
Speaker 1: harmful to the environment. Unless something significantly changes, like a

834
00:44:45,400 --> 00:44:49,719
Speaker 1: massive scientific breakthrough in energy or compute efficiency, I just

835
00:44:49,760 --> 00:44:52,240
Speaker 1: don't see how open ai makes it two more years.

836
00:44:52,880 --> 00:44:55,760
Speaker 1: And worse still, if my hypothesis about the wider market

837
00:44:55,840 --> 00:44:59,520
Speaker 1: is true, there might just not be a viable business

838
00:44:59,560 --> 00:45:02,480
Speaker 1: in make and selling large language models. While Meta and

839
00:45:02,520 --> 00:45:04,760
Speaker 1: open Ai might be able to claim hundreds of millions

840
00:45:04,760 --> 00:45:07,920
Speaker 1: of users on these services, I don't see any evidence

841
00:45:07,960 --> 00:45:10,560
Speaker 1: that these people will make them any money, and in fact,

842
00:45:10,600 --> 00:45:13,440
Speaker 1: I only see evidence they'll lose it. And if I'm right,

843
00:45:13,920 --> 00:45:16,719
Speaker 1: we're watching vcs in companies like Microsoft burn tens of

844
00:45:16,719 --> 00:45:19,600
Speaker 1: billions of dollars to power the next generation of products

845
00:45:20,680 --> 00:45:24,200
Speaker 1: and nobody really gives a shit about And in the

846
00:45:24,239 --> 00:45:26,720
Speaker 1: next episode, I'm going to go into the much bigger problem,

847
00:45:26,840 --> 00:45:30,719
Speaker 1: the subprime AI crisis, where everyone building on these platforms

848
00:45:30,760 --> 00:45:33,319
Speaker 1: is inevitably going to get rug pulled when Open AI

849
00:45:33,400 --> 00:45:36,640
Speaker 1: and Anthropic and other companies raise their prices, because if

850
00:45:36,680 --> 00:45:38,920
Speaker 1: they don't raise their prices, how are they going to

851
00:45:38,920 --> 00:45:41,920
Speaker 1: be able to afford to keep going. I don't bloody no,

852
00:45:42,080 --> 00:45:52,239
Speaker 1: but we'll get into it next episode. Thank you for

853
00:45:52,280 --> 00:45:54,960
Speaker 1: listening to Better Offline. The editor and composer of the

854
00:45:54,960 --> 00:45:58,080
Speaker 1: Better Offline theme song is Matosowski. You can check out

855
00:45:58,080 --> 00:46:01,759
Speaker 1: more of his music and audio projects Matttersolski dot com,

856
00:46:01,920 --> 00:46:05,440
Speaker 1: M A T T O s O W s Ki

857
00:46:05,640 --> 00:46:08,520
Speaker 1: dot com. You can email me at easy at Better

858
00:46:08,560 --> 00:46:11,000
Speaker 1: offline dot com or visit Better Offline dot com to

859
00:46:11,040 --> 00:46:13,880
Speaker 1: find more podcast links and of course my newsletter. I

860
00:46:13,920 --> 00:46:16,640
Speaker 1: also really recommend you go to chat dot Where's youreaed

861
00:46:16,719 --> 00:46:18,960
Speaker 1: dot at to visit the discord, and go to our

862
00:46:19,040 --> 00:46:22,359
Speaker 1: slash Better Offline to check out our reddit. Thank you

863
00:46:22,400 --> 00:46:23,360
Speaker 1: so much for listening.

864
00:46:24,200 --> 00:46:26,880
Speaker 2: Better Offline is a production of Cool Zone Media. For

865
00:46:27,000 --> 00:46:30,160
Speaker 2: more from Cool Zone Media, visit our website cool zonemedia

866
00:46:30,239 --> 00:46:33,080
Speaker 2: dot com, or check us out on the iHeartRadio app,

867
00:46:33,120 --> 00:46:35,800
Speaker 2: Apple Podcasts, or wherever you get your podcasts.