1
00:00:02,160 --> 00:00:08,119
Speaker 1: Zone Media. Hello you, it's your better offline monologue and

2
00:00:08,160 --> 00:00:17,920
Speaker 1: I'm your host ed ZiT trum. Now next week I'm

3
00:00:17,920 --> 00:00:19,400
Speaker 1: going to have a two part of that digs into

4
00:00:19,400 --> 00:00:21,840
Speaker 1: Microsoft Stata senter pull back an open ai is shaking

5
00:00:21,920 --> 00:00:25,200
Speaker 1: you funding situation. But this week's monologue focuses an open

6
00:00:25,200 --> 00:00:28,880
Speaker 1: AI's new model, GPT four point five. You may be

7
00:00:29,040 --> 00:00:31,880
Speaker 1: wondering what it does differently to GPT four oh, or

8
00:00:31,920 --> 00:00:34,159
Speaker 1: claudes on at three point seven, or any number of

9
00:00:34,200 --> 00:00:36,760
Speaker 1: other large language models, And if I'm honest, I have

10
00:00:36,800 --> 00:00:40,800
Speaker 1: absolutely no idea. Thankfully neither does Sam Ortman, CEO of

11
00:00:40,800 --> 00:00:43,480
Speaker 1: open Ai, who said, and I quote, the GPT four

12
00:00:43,479 --> 00:00:47,120
Speaker 1: point five was the first model that feels like talking

13
00:00:47,159 --> 00:00:51,560
Speaker 1: to a thoughtful person to him, which makes me wonder

14
00:00:51,680 --> 00:00:55,000
Speaker 1: what the other models have been like. So I went

15
00:00:55,040 --> 00:00:57,680
Speaker 1: back and look. So compare this to the launch of

16
00:00:57,720 --> 00:01:01,520
Speaker 1: GPT four oh, which Altman called open AI's best model ever,

17
00:01:01,640 --> 00:01:05,679
Speaker 1: saying that it was fast, smart, natively multimodal, referring to

18
00:01:05,720 --> 00:01:07,880
Speaker 1: the ability to accept text as well as audio and

19
00:01:07,959 --> 00:01:11,360
Speaker 1: video and photos as well, and available to all chat

20
00:01:11,400 --> 00:01:14,520
Speaker 1: GPT users, including on their free plan, adding that it

21
00:01:14,640 --> 00:01:18,400
Speaker 1: was a very good model, especially at Coding. By contrast,

22
00:01:18,440 --> 00:01:22,160
Speaker 1: Alltman summarized GPT four point five as a giant, expensive model,

23
00:01:22,480 --> 00:01:25,520
Speaker 1: one that required hundreds of thousands of GPUs to launch

24
00:01:25,520 --> 00:01:28,760
Speaker 1: beyond chat GPT pro. It started there. It's still not

25
00:01:28,880 --> 00:01:32,440
Speaker 1: as I record this out to plus users or free users. Now.

26
00:01:32,520 --> 00:01:35,160
Speaker 1: GPT pro of course is open AI is two hundred

27
00:01:35,160 --> 00:01:39,960
Speaker 1: dollars a month subscription, and it's unclear when plus, which

28
00:01:40,000 --> 00:01:41,640
Speaker 1: is the twenty buck a month one will get it,

29
00:01:41,680 --> 00:01:44,920
Speaker 1: but apparently it's in the next few days. Aortman also

30
00:01:44,920 --> 00:01:47,880
Speaker 1: added that GPT four point five isn't a reasoning model

31
00:01:47,960 --> 00:01:50,680
Speaker 1: and won't crush benchmarks on account of it being a

32
00:01:50,720 --> 00:01:53,480
Speaker 1: different kind of intelligence that has and all of these

33
00:01:53,520 --> 00:01:56,880
Speaker 1: are quotes magic to it that Sam Mortman had not

34
00:01:57,000 --> 00:02:02,360
Speaker 1: felt before. Yeah, just you know, shit's not doing well

35
00:02:02,800 --> 00:02:06,120
Speaker 1: when you have to just be like it's magic, It's

36
00:02:06,160 --> 00:02:09,240
Speaker 1: it's literally magic. I made magic. Now what does the

37
00:02:09,240 --> 00:02:13,519
Speaker 1: magic do? I'm really not sure. In fact, it's pretty

38
00:02:13,520 --> 00:02:15,960
Speaker 1: difficult to find exactly what it is that GPT four

39
00:02:16,000 --> 00:02:18,120
Speaker 1: point five does differently, or what it's good at, or

40
00:02:18,160 --> 00:02:21,240
Speaker 1: indeed really anything about him. BENJ. Edwards over At asked

41
00:02:21,240 --> 00:02:24,639
Speaker 1: Tenneker had one developer called it a lemon. GPT four

42
00:02:24,680 --> 00:02:27,960
Speaker 1: point five costs an incredible seventy five dollars per million

43
00:02:28,000 --> 00:02:30,880
Speaker 1: input tokens prompts and data pushed into a model, and

44
00:02:30,919 --> 00:02:33,840
Speaker 1: one hundred and fifty dollars per million output tokens is

45
00:02:33,840 --> 00:02:36,200
Speaker 1: in the thing it creates. A token is like zero

46
00:02:36,240 --> 00:02:39,920
Speaker 1: point seven. I think one token is maybe three words.

47
00:02:40,160 --> 00:02:42,600
Speaker 1: Someone will get up my ass for this. Nevertheless, this

48
00:02:42,680 --> 00:02:45,240
Speaker 1: seems like a lot. It isn't when you're running a company.

49
00:02:45,600 --> 00:02:48,080
Speaker 1: And by the way, this is roughly three thousand percent

50
00:02:48,160 --> 00:02:51,080
Speaker 1: more expensive for input tokens and fifteen hundred percent more

51
00:02:51,120 --> 00:02:53,680
Speaker 1: expensive for output tokens than GPT four to zero for

52
00:02:53,800 --> 00:02:57,680
Speaker 1: results that open ai co founder Andredge Carpathy described as

53
00:02:57,800 --> 00:03:00,960
Speaker 1: a little bit better and awesome, but also not exactly

54
00:03:00,960 --> 00:03:04,200
Speaker 1: in ways that are trivial to point to. That translates

55
00:03:04,200 --> 00:03:06,960
Speaker 1: to it's a little bit better, but I can't really

56
00:03:07,000 --> 00:03:10,399
Speaker 1: tell you why. And yes, you're gonna hear me say

57
00:03:10,400 --> 00:03:13,120
Speaker 1: something similar in next week's episode, because the larger picture

58
00:03:13,160 --> 00:03:16,119
Speaker 1: for open ai right now is pretty fucking dire considering

59
00:03:16,120 --> 00:03:18,440
Speaker 1: their main backer, soft Bank, has to borrow billions of

60
00:03:18,440 --> 00:03:21,960
Speaker 1: dollars to fund them. Nevertheless, back to four point five

61
00:03:22,360 --> 00:03:24,639
Speaker 1: Since launch, which was for some reason, on the day

62
00:03:24,639 --> 00:03:27,760
Speaker 1: that Sam Ortman's child was being born in the hospital,

63
00:03:29,040 --> 00:03:32,280
Speaker 1: He's been posting some really weird shits since, though. A

64
00:03:32,280 --> 00:03:35,080
Speaker 1: few days after launch, Aortman claimed that GPT four point

65
00:03:35,080 --> 00:03:38,080
Speaker 1: five was the first time people had been emailing with

66
00:03:38,120 --> 00:03:41,440
Speaker 1: such passion, asking the open AI promised never to stop

67
00:03:41,480 --> 00:03:44,320
Speaker 1: offering a specific model or even replace it with an update,

68
00:03:44,600 --> 00:03:47,240
Speaker 1: at which point I assume everybody in the room started

69
00:03:47,280 --> 00:03:50,760
Speaker 1: clapping and they saluted Sam Mortman and said, thank you,

70
00:03:50,840 --> 00:03:53,280
Speaker 1: sir for making this happen. And by the way, what

71
00:03:53,360 --> 00:03:55,360
Speaker 1: I'm suggesting is that no one's ever done this, or

72
00:03:55,400 --> 00:03:58,560
Speaker 1: like one freak did, or maybe Aortman emailed it to himself,

73
00:03:59,080 --> 00:04:04,440
Speaker 1: who just shut up? Just your company burns five billion

74
00:04:04,480 --> 00:04:06,400
Speaker 1: dollars a year and the best you've got is this

75
00:04:06,480 --> 00:04:10,360
Speaker 1: warmed up dog shit about people marine todding you over

76
00:04:10,440 --> 00:04:13,400
Speaker 1: your model and never taking it away. Has opening I

77
00:04:13,520 --> 00:04:17,960
Speaker 1: ever even taken away a model? Jesus fucking these companies anyway.

78
00:04:18,520 --> 00:04:21,400
Speaker 1: A few days later, Altman posted a conversation where he

79
00:04:21,480 --> 00:04:24,280
Speaker 1: asked GPT four point five if it believed it was real,

80
00:04:24,600 --> 00:04:26,560
Speaker 1: leading to a series of bullet points with things like

81
00:04:26,640 --> 00:04:29,400
Speaker 1: what do we mean by real only for GPT four

82
00:04:29,400 --> 00:04:31,719
Speaker 1: point five, saying that it believed that it was not

83
00:04:31,800 --> 00:04:35,159
Speaker 1: an independent consciousness, but rather a structured experience happening within

84
00:04:35,480 --> 00:04:38,680
Speaker 1: your consciousness, referring to Sam Altman, which is the kind

85
00:04:38,680 --> 00:04:40,919
Speaker 1: of shit that's only impressive if you're an imbecile or

86
00:04:40,920 --> 00:04:43,279
Speaker 1: so stoned you've texted date of your friends the question

87
00:04:43,520 --> 00:04:45,680
Speaker 1: what if the joker was Batman? And by the way,

88
00:04:45,720 --> 00:04:47,800
Speaker 1: the answer to that is called The Batman Who Laughs

89
00:04:47,839 --> 00:04:50,280
Speaker 1: and it's one of the worst comics ever written. If

90
00:04:50,279 --> 00:04:52,200
Speaker 1: you want to talk to me about DC Metal, please

91
00:04:52,240 --> 00:04:55,800
Speaker 1: email me. It's easy. That's e z or z. If

92
00:04:55,800 --> 00:04:59,240
Speaker 1: you're Canadian or British at Better Offline dot com I

93
00:04:59,320 --> 00:05:02,320
Speaker 1: a really if you are working for DC Comics right

94
00:05:02,320 --> 00:05:04,640
Speaker 1: now and you had anything to do with Death Metal

95
00:05:04,760 --> 00:05:06,920
Speaker 1: or The Batman Who Laughs, you and I have a grievance.

96
00:05:07,240 --> 00:05:09,920
Speaker 1: You and I need to talk. Sorry what this is

97
00:05:09,960 --> 00:05:13,680
Speaker 1: a tech pop us right back to open AI. More worryingly,

98
00:05:13,960 --> 00:05:17,520
Speaker 1: Samultman posted an idea for paid plans where your twenty

99
00:05:17,520 --> 00:05:20,919
Speaker 1: dollars plus subscription converts the credits you can use across

100
00:05:20,920 --> 00:05:24,080
Speaker 1: speech just like Deep Research O one. GPT four point five, Sora,

101
00:05:24,200 --> 00:05:26,800
Speaker 1: and so on, with no fixed limits per feature, and

102
00:05:26,839 --> 00:05:29,080
Speaker 1: you choose what you want. If you run out of credits,

103
00:05:29,360 --> 00:05:31,960
Speaker 1: you can buy more. This, to be clear, is an

104
00:05:32,000 --> 00:05:34,840
Speaker 1: attempt to raise prices without actually raising them by attempting

105
00:05:34,839 --> 00:05:38,240
Speaker 1: to limit usage of open AI's more expensive models. Chat,

106
00:05:38,279 --> 00:05:41,040
Speaker 1: GPT plus and other subscriptions give you a limit, for example,

107
00:05:41,080 --> 00:05:43,560
Speaker 1: a limit of eighty messages every three hours on GPT

108
00:05:43,680 --> 00:05:46,320
Speaker 1: four to oh, but using one doesn't limit your use

109
00:05:46,320 --> 00:05:49,679
Speaker 1: of other products. Here, open ai is trying to create

110
00:05:49,680 --> 00:05:52,040
Speaker 1: a rent seeking model where power users have to pay

111
00:05:52,200 --> 00:05:54,600
Speaker 1: for more credits if they want to use, say open

112
00:05:54,640 --> 00:05:56,720
Speaker 1: ai is more expensive models like Sora and O one,

113
00:05:56,880 --> 00:05:59,040
Speaker 1: and I imagine any situation like this will be one

114
00:05:59,040 --> 00:06:01,479
Speaker 1: where they hope that people simply won't use their credits

115
00:06:01,680 --> 00:06:03,520
Speaker 1: or overuse them and have to pay for top ups.

116
00:06:04,040 --> 00:06:06,839
Speaker 1: This is, of course all theoretical, but it heavily suggests

117
00:06:06,839 --> 00:06:09,920
Speaker 1: that open ai is getting desperate. And now the information

118
00:06:10,040 --> 00:06:12,880
Speaker 1: is reporting that open ai executives have told some investors

119
00:06:12,880 --> 00:06:16,200
Speaker 1: that they will be charging two thousand dollars per month

120
00:06:16,400 --> 00:06:19,080
Speaker 1: for their low end agent product. And yes that's a

121
00:06:19,160 --> 00:06:24,080
Speaker 1: quote sold to And again I quote high income knowledge

122
00:06:24,080 --> 00:06:27,560
Speaker 1: workers with supposed mid tier agents for software development costing

123
00:06:27,720 --> 00:06:31,400
Speaker 1: possibly ten thousand dollars a month, with supposed PhD level

124
00:06:31,440 --> 00:06:34,440
Speaker 1: research agents costing twenty thousand dollars a month, and I

125
00:06:34,440 --> 00:06:36,960
Speaker 1: will tell you the PhDs I know would probably do

126
00:06:37,000 --> 00:06:39,560
Speaker 1: it for half, and they'd even work for an annoying

127
00:06:39,560 --> 00:06:42,560
Speaker 1: asshole like Sam Altman. Now you may wonder what any

128
00:06:42,600 --> 00:06:44,640
Speaker 1: of these things do, and the answer is that neither

129
00:06:44,839 --> 00:06:47,200
Speaker 1: I nor the Information know. As of right now. The

130
00:06:47,240 --> 00:06:50,680
Speaker 1: only operational or agent, open ai has his operator, open

131
00:06:50,720 --> 00:06:53,800
Speaker 1: AI's agent that sometimes successfully uses a web browser search

132
00:06:53,839 --> 00:06:57,240
Speaker 1: for something in minutes, which would usually take you seconds.

133
00:06:57,920 --> 00:07:00,240
Speaker 1: The Information attempted to suggest that the two thousand a

134
00:07:00,240 --> 00:07:03,159
Speaker 1: month agent would be some sort of thing that could

135
00:07:03,240 --> 00:07:05,839
Speaker 1: sort through and rank sales lead But I'm sorry, do

136
00:07:05,920 --> 00:07:08,040
Speaker 1: I really have to read this shit with a straight face?

137
00:07:08,120 --> 00:07:10,560
Speaker 1: Twenty thousand dollars for a PhD level agent? What the

138
00:07:10,600 --> 00:07:13,040
Speaker 1: fuck does that mean? What would it do? Why do

139
00:07:13,120 --> 00:07:15,960
Speaker 1: these companies? I get emails every week having to justify

140
00:07:16,080 --> 00:07:19,920
Speaker 1: my fucking cynicism, But these shittheads, they're allowed to just

141
00:07:20,160 --> 00:07:22,080
Speaker 1: make up stuff and it leak it to the information

142
00:07:22,160 --> 00:07:24,560
Speaker 1: The Information publishes. It were all meant to be impressed.

143
00:07:24,600 --> 00:07:28,960
Speaker 1: What the fucking what the fuck I'm allowed to rent

144
00:07:28,960 --> 00:07:31,000
Speaker 1: on these? They're allowing me to rant on these. It's

145
00:07:31,120 --> 00:07:33,760
Speaker 1: just it sickens me. I have had this week at

146
00:07:33,840 --> 00:07:37,240
Speaker 1: least five people email me and be like, well ed,

147
00:07:37,280 --> 00:07:39,760
Speaker 1: what would it take to change your mind about this stuff?

148
00:07:39,800 --> 00:07:42,320
Speaker 1: Why do I have to fucking do it? Why do

149
00:07:42,400 --> 00:07:45,680
Speaker 1: I The multi billion dollar companies do a dogshit job

150
00:07:45,720 --> 00:07:48,920
Speaker 1: of actually explaining this stuff or selling it. They lose

151
00:07:48,960 --> 00:07:50,880
Speaker 1: billions of dollars. But I'm the guy who has to

152
00:07:50,960 --> 00:07:54,200
Speaker 1: justify myself. Oh well, I'll keep doing it. Nevertheless, NES's

153
00:07:54,240 --> 00:07:56,000
Speaker 1: old At the bottom of this article's a far more

154
00:07:56,040 --> 00:07:59,560
Speaker 1: obvious payale horse. Open ai is planning to charge twenty

155
00:07:59,600 --> 00:08:02,400
Speaker 1: percent to thirty percent of pro customers the two hundred

156
00:08:02,400 --> 00:08:04,880
Speaker 1: dollars a month subscription that loses the money every time,

157
00:08:05,200 --> 00:08:07,880
Speaker 1: a higher price because of how many research queries they're

158
00:08:07,920 --> 00:08:11,160
Speaker 1: doing with Altmann, according to the information, suggesting some sort

159
00:08:11,200 --> 00:08:14,000
Speaker 1: of hey guessoir a la carte or pay as you

160
00:08:14,040 --> 00:08:18,520
Speaker 1: go approach. I want to be clear about something. This

161
00:08:18,640 --> 00:08:21,040
Speaker 1: is not a company that's cooking. This is not a

162
00:08:21,080 --> 00:08:25,120
Speaker 1: company that's worked out anything. Open Aiye is unprofitable, unsustainable,

163
00:08:25,160 --> 00:08:27,920
Speaker 1: and deeply, deeply lost. These are the actions of a

164
00:08:28,000 --> 00:08:32,080
Speaker 1: desperate company run by a desperate man. You have only

165
00:08:32,120 --> 00:08:34,320
Speaker 1: Sam Morman had a thoughtful friend to talk about all

166
00:08:34,360 --> 00:08:35,800
Speaker 1: these problems too,