1 00:00:02,560 --> 00:00:06,920 Speaker 1: Zone Media, stop your grinnin and drop your linen. I'm 2 00:00:07,000 --> 00:00:21,200 Speaker 1: d zechron and this is better offline. Welcome to the 3 00:00:21,200 --> 00:00:23,319 Speaker 1: final part of our three part Hater's Guide to the 4 00:00:23,320 --> 00:00:25,599 Speaker 1: AI Bubble, where we've been talking about the growing cracks 5 00:00:25,640 --> 00:00:27,960 Speaker 1: emerging in the industry and the potential contagion to the 6 00:00:27,960 --> 00:00:30,840 Speaker 1: wider economy that are pullback in general of AI spending 7 00:00:30,880 --> 00:00:33,480 Speaker 1: will cause. And I must be clear, things are bad 8 00:00:33,640 --> 00:00:36,480 Speaker 1: the vibes A rand said. I spent the past two episodes, 9 00:00:36,520 --> 00:00:38,920 Speaker 1: and we'll spend much of this episode explaining why. But 10 00:00:38,960 --> 00:00:40,840 Speaker 1: I just want to recap something from the first part 11 00:00:40,840 --> 00:00:44,880 Speaker 1: of this series. Right now, seven companies, the Magnificent Seven, 12 00:00:45,080 --> 00:00:48,000 Speaker 1: account for the value of nearly thirty five percent of 13 00:00:48,040 --> 00:00:50,320 Speaker 1: all of US stocks, and the biggest of these companies 14 00:00:50,360 --> 00:00:52,440 Speaker 1: is in Nvidia, which has gotten fat off the back 15 00:00:52,479 --> 00:00:55,880 Speaker 1: of the AI bubble today a disproportionate amount of Nvidia's 16 00:00:55,920 --> 00:00:58,120 Speaker 1: revenue eighty eight percent of it is coming from the 17 00:00:58,160 --> 00:01:00,720 Speaker 1: other members of the Magnificent Seven. And there of the 18 00:01:00,840 --> 00:01:03,920 Speaker 1: enterprise GPUs designed to power the compute heavy eight A 19 00:01:04,040 --> 00:01:07,199 Speaker 1: applications that you're hearing about all day. If and when 20 00:01:07,280 --> 00:01:11,240 Speaker 1: they pull back, invidiabile sink so too will the other companies, 21 00:01:11,319 --> 00:01:13,360 Speaker 1: especially when it becomes evident that they've spent nearly half 22 00:01:13,400 --> 00:01:17,199 Speaker 1: a trillion dollars in two years on bugger roll. It'll 23 00:01:17,200 --> 00:01:20,960 Speaker 1: be an apocalyptic shedding of value. They won't die, but 24 00:01:21,280 --> 00:01:23,080 Speaker 1: it'll be the likes of which we haven't seen since 25 00:01:23,080 --> 00:01:24,840 Speaker 1: the dark days of two thousand and seven and two 26 00:01:24,840 --> 00:01:27,959 Speaker 1: thousand and eight for those companies. I don't know about 27 00:01:27,959 --> 00:01:31,480 Speaker 1: the global banking system imploding, but I also can't say 28 00:01:31,520 --> 00:01:33,679 Speaker 1: what I just I don't know. But what I do 29 00:01:33,800 --> 00:01:36,480 Speaker 1: know is tech is going to take such a precipitous 30 00:01:36,480 --> 00:01:38,679 Speaker 1: haircut that you're going to see scissor gashes in the 31 00:01:38,680 --> 00:01:42,600 Speaker 1: man's bald head. Terrible metaphor. If that sounds like a 32 00:01:42,600 --> 00:01:45,800 Speaker 1: them problem, though, and not a you problem, ask yourself, 33 00:01:45,840 --> 00:01:47,720 Speaker 1: what's in your fuah one K? Do you own any 34 00:01:47,800 --> 00:01:50,480 Speaker 1: index funds or ETFs? Are you close to retirement? And 35 00:01:50,480 --> 00:01:52,800 Speaker 1: how much do you enjoy eating ramen? Are you willing 36 00:01:52,840 --> 00:01:54,800 Speaker 1: to eat the same whiskers cat food as your American 37 00:01:54,840 --> 00:01:57,840 Speaker 1: short hair? I am, except I feed them a different, 38 00:01:57,880 --> 00:02:00,440 Speaker 1: better food because Babu deserves the best. But in the 39 00:02:00,480 --> 00:02:02,520 Speaker 1: last two episodes, I made the case that the generative 40 00:02:02,560 --> 00:02:05,880 Speaker 1: AI industry is unique in its capital demands and unparalleled 41 00:02:05,880 --> 00:02:07,840 Speaker 1: in the fact that no company beside in video or 42 00:02:07,880 --> 00:02:10,840 Speaker 1: consultancies are making any money from them. I can easily 43 00:02:10,840 --> 00:02:13,800 Speaker 1: imagine the chain of events where one big company fails, 44 00:02:13,880 --> 00:02:15,760 Speaker 1: like either because they ran out of money or because 45 00:02:15,800 --> 00:02:18,440 Speaker 1: the investors pulled back. This in turn spooks the rest 46 00:02:18,440 --> 00:02:21,600 Speaker 1: of the market, causing other companies to start seeing massive 47 00:02:21,639 --> 00:02:24,040 Speaker 1: revenue drops. And as the demand for compute drops, hyper 48 00:02:24,040 --> 00:02:28,000 Speaker 1: scalers like Microsoft, Amazon and Google massively cut their capital expenditures, 49 00:02:28,160 --> 00:02:31,160 Speaker 1: canceling data center build outs and of course slowing down 50 00:02:31,200 --> 00:02:34,160 Speaker 1: their GPU orders. And then Jensen Huang would have to 51 00:02:34,160 --> 00:02:36,160 Speaker 1: watch as the cash cow that made his company rich 52 00:02:36,200 --> 00:02:38,440 Speaker 1: and fat is take over a walk behind the barn 53 00:02:39,280 --> 00:02:41,760 Speaker 1: for two shots to ring out. Though I must be clear, 54 00:02:42,280 --> 00:02:48,600 Speaker 1: video isn't gonna die, None of the Magnificent seven will. However, 55 00:02:49,000 --> 00:02:51,120 Speaker 1: the first big failure, the first domino to fall, could 56 00:02:51,120 --> 00:02:52,959 Speaker 1: be any number of companies, and while I don't think 57 00:02:52,960 --> 00:02:55,680 Speaker 1: open Ai and Anthropic are necessarily going to be the first, 58 00:02:56,440 --> 00:02:59,040 Speaker 1: I also believe that there's an air of inevitability about 59 00:02:59,080 --> 00:03:02,120 Speaker 1: one of or both of their demises. This matter is 60 00:03:02,160 --> 00:03:04,560 Speaker 1: because these companies are for all intents and purposes the 61 00:03:04,560 --> 00:03:08,639 Speaker 1: generative AI industry. They're deeply unsustainable and unstable, and yet 62 00:03:08,680 --> 00:03:11,720 Speaker 1: they are also critical to the AI trade. Continuing, I 63 00:03:11,720 --> 00:03:14,959 Speaker 1: haven't really spent much time on my favorite subject, that 64 00:03:15,240 --> 00:03:17,720 Speaker 1: open ai is a systemic risk to the tech industry. 65 00:03:17,880 --> 00:03:21,639 Speaker 1: But let's recap my reporting. Open Ai and Anthropic both 66 00:03:21,680 --> 00:03:23,880 Speaker 1: lose billions of dollars a year after revenue, and their 67 00:03:23,919 --> 00:03:26,760 Speaker 1: stories do not mirror any other startup in history, not Uber, 68 00:03:26,960 --> 00:03:30,160 Speaker 1: not Amazon Web Services, and nothing. I've addressed my Uber 69 00:03:30,240 --> 00:03:34,080 Speaker 1: point before in how does open ai survive? It'll be 70 00:03:34,120 --> 00:03:36,560 Speaker 1: in this show notes. Someone always asks me about this. 71 00:03:36,720 --> 00:03:40,200 Speaker 1: Please stop, learn to read, learn to listen. Soft Bank 72 00:03:40,280 --> 00:03:42,640 Speaker 1: also is putting itself in dire straits simply to fund 73 00:03:42,640 --> 00:03:45,800 Speaker 1: open Ai. The steel threatens its credit ring, with SoftBank 74 00:03:45,840 --> 00:03:47,680 Speaker 1: having to take on what will be multiple loans to 75 00:03:47,680 --> 00:03:50,200 Speaker 1: fund the remaining thirty billion dollars open ai is forty 76 00:03:50,200 --> 00:03:52,600 Speaker 1: billion dollar funding round which is yet to close, and 77 00:03:52,880 --> 00:03:55,400 Speaker 1: open aiy is in fact still raising from the Middle East. 78 00:03:56,240 --> 00:03:58,680 Speaker 1: This is before you consider the other nineteen billion dollars 79 00:03:58,720 --> 00:04:00,960 Speaker 1: that soft Bank has agreed to contribute to the Stargate 80 00:04:01,040 --> 00:04:03,920 Speaker 1: Data Center project, money that it does not have available 81 00:04:03,960 --> 00:04:07,440 Speaker 1: at this time. But I must say, while recording this episode, 82 00:04:07,640 --> 00:04:09,280 Speaker 1: I saw some news come out and I'm just going 83 00:04:09,320 --> 00:04:11,920 Speaker 1: to read it to you and said soft Bank and 84 00:04:11,960 --> 00:04:15,720 Speaker 1: open Ai is five hundred billion AI project struggles to 85 00:04:15,720 --> 00:04:18,880 Speaker 1: get off ground. The Stargate venture introduced a white house event, 86 00:04:19,000 --> 00:04:21,440 Speaker 1: is now setting the more modest goal of building a 87 00:04:21,480 --> 00:04:24,719 Speaker 1: small data center by year. And I just want to say, 88 00:04:25,480 --> 00:04:29,200 Speaker 1: anyone that doubted me can suck one. I was oinking 89 00:04:29,240 --> 00:04:32,880 Speaker 1: and squawking around my room just before this when I 90 00:04:32,920 --> 00:04:34,440 Speaker 1: got this news in. I just want to say, I 91 00:04:34,600 --> 00:04:36,720 Speaker 1: just learned this news and I'm still taking it in. 92 00:04:37,000 --> 00:04:39,080 Speaker 1: I'm still taking in because it makes me so excited 93 00:04:39,120 --> 00:04:41,840 Speaker 1: and sad. Anyway, neither soft Bank nor open Ai has 94 00:04:41,880 --> 00:04:44,840 Speaker 1: the money for Stargate, regardless of what they built. Also, 95 00:04:45,160 --> 00:04:48,040 Speaker 1: Clammy Samy's making some interesting claims about what Stargate is. 96 00:04:48,080 --> 00:04:51,479 Speaker 1: He's claimed the Denton site, the core Weavers building is 97 00:04:51,520 --> 00:04:55,800 Speaker 1: also Stargate. This motherfucker loves lying anyway. Open A must 98 00:04:55,839 --> 00:04:57,680 Speaker 1: also convert to a for profit by the end of 99 00:04:57,680 --> 00:04:59,600 Speaker 1: twenty twenty five, all whose twenty billion dollars of the 100 00:04:59,640 --> 00:05:02,080 Speaker 1: remaining thirty billion dollars of funding. If it does not 101 00:05:02,080 --> 00:05:04,880 Speaker 1: convert by October twenty twenty six, it's current funding converts 102 00:05:04,880 --> 00:05:08,280 Speaker 1: to debt. It is also demanding remarkable and reasonable concessions 103 00:05:08,320 --> 00:05:11,679 Speaker 1: and Microsoft stuff like they want to reduce their rev share, 104 00:05:11,920 --> 00:05:14,960 Speaker 1: They want to reduce their their I think, their ability 105 00:05:15,000 --> 00:05:18,280 Speaker 1: to their exclusive rights to sell models. I did I 106 00:05:18,279 --> 00:05:19,960 Speaker 1: really need to do a monologue about this, and I 107 00:05:20,040 --> 00:05:23,000 Speaker 1: know this is technically a monologue, but nevertheless, they want 108 00:05:23,040 --> 00:05:25,320 Speaker 1: these insane things, and Microsoft is refusing to budge, and 109 00:05:25,360 --> 00:05:27,839 Speaker 1: it's willing to walk away from negotiations necessary to convert. 110 00:05:27,880 --> 00:05:31,160 Speaker 1: According to the Financial Times, and also open Ai does 111 00:05:31,160 --> 00:05:33,680 Speaker 1: not have a path of profitability, and its future, like Anthropics, 112 00:05:33,760 --> 00:05:36,039 Speaker 1: is dependent on the continual flow of capital from venture 113 00:05:36,040 --> 00:05:39,760 Speaker 1: capitalists and big tech, who must also continue to expand infrastructure. 114 00:05:40,760 --> 00:05:42,440 Speaker 1: I don't know how that's going to fucking happen with 115 00:05:42,480 --> 00:05:45,960 Speaker 1: Stargate going. God, I feel so good about that. Oh 116 00:05:46,360 --> 00:05:49,560 Speaker 1: I want to like bottle that up and drink it. Anyway. 117 00:05:50,279 --> 00:05:52,760 Speaker 1: Anthropic is in a similar but slightly better position. It's 118 00:05:52,760 --> 00:05:54,760 Speaker 1: set to lose three billion dollars this year on four 119 00:05:54,760 --> 00:05:57,200 Speaker 1: billion dollars of revenue. It also has no path of 120 00:05:57,200 --> 00:05:59,960 Speaker 1: profitability and recently jacked up prices on cursor, it's larger 121 00:06:00,120 --> 00:06:02,600 Speaker 1: customer has had to put restraints on clawed code after 122 00:06:02,640 --> 00:06:05,359 Speaker 1: allowing users to burn one hundred to ten thousand percent 123 00:06:05,400 --> 00:06:07,760 Speaker 1: of their revenue a month. These are the actions of 124 00:06:07,760 --> 00:06:11,640 Speaker 1: a desperate company. Nevertheless, Open Ai and Anthropics revenue amount to, 125 00:06:12,000 --> 00:06:14,159 Speaker 1: by my estimates, more than half of the entire revenue 126 00:06:14,160 --> 00:06:17,600 Speaker 1: of the generative AI industry, including the hyperscalers. And to 127 00:06:17,600 --> 00:06:20,360 Speaker 1: be abundantly clear, the two companies that amount to a 128 00:06:20,440 --> 00:06:26,880 Speaker 1: round half of all generative AI revenue are only losing money. Now. 129 00:06:26,880 --> 00:06:28,400 Speaker 1: I said a lot of this before, which is why 130 00:06:28,440 --> 00:06:30,080 Speaker 1: I'm trying not to harp on it too much. But 131 00:06:30,120 --> 00:06:32,960 Speaker 1: the most important company in the entire AI industry needs 132 00:06:33,000 --> 00:06:35,960 Speaker 1: to convert to a for profit entity from a nonprofit entity, 133 00:06:36,120 --> 00:06:38,120 Speaker 1: which has never happened before in this case, by the 134 00:06:38,240 --> 00:06:40,400 Speaker 1: end of the year or it's effectively dead. And even 135 00:06:40,400 --> 00:06:43,200 Speaker 1: if it does, Open AI burns billions and billions and 136 00:06:43,240 --> 00:06:46,120 Speaker 1: billions of dollars a year and will die without continual funding. 137 00:06:46,360 --> 00:06:49,200 Speaker 1: They have no path to profitability in anyone telling you 138 00:06:49,240 --> 00:06:51,960 Speaker 1: otherwise as a liar or a fantasist. And on top 139 00:06:51,960 --> 00:06:54,080 Speaker 1: of this is part of the thing about SoftBank, by 140 00:06:54,120 --> 00:06:57,800 Speaker 1: the Way and Stargate. They revealed that they've promised Oracle 141 00:06:57,880 --> 00:07:02,640 Speaker 1: thirty billion dollars a year their cloud contract. I swear 142 00:07:02,640 --> 00:07:06,200 Speaker 1: to fucking God does no one read these stories and 143 00:07:06,360 --> 00:07:09,960 Speaker 1: think does this not sound just like an egregious lie? Anyway, 144 00:07:10,080 --> 00:07:12,080 Speaker 1: as I wrote earlier in the year, there is really 145 00:07:12,080 --> 00:07:15,360 Speaker 1: no significant adoption of generative AI services or products. Chat 146 00:07:15,360 --> 00:07:18,119 Speaker 1: GPTs five hundred million weekly users, and otherwise it seems 147 00:07:18,120 --> 00:07:21,160 Speaker 1: that other services they struggled to get fifteen million users. 148 00:07:21,520 --> 00:07:24,880 Speaker 1: Although you could mention Google that's not a fair comparisons. 149 00:07:24,920 --> 00:07:28,040 Speaker 1: It cheated by combining Google Gemini with Google Assystem to 150 00:07:28,080 --> 00:07:30,360 Speaker 1: claim it had three hundred and fifty million active users 151 00:07:31,400 --> 00:07:33,880 Speaker 1: a while the five hundred million weekly users sounds and 152 00:07:34,200 --> 00:07:36,760 Speaker 1: is in fairness impressive. There's a world of difference between 153 00:07:36,800 --> 00:07:39,000 Speaker 1: someone using your product as part of their job, which 154 00:07:39,000 --> 00:07:40,920 Speaker 1: they pay for, and someone dicking around with an image 155 00:07:40,960 --> 00:07:43,760 Speaker 1: generator or a college student trying to cheat on their homework. 156 00:07:43,960 --> 00:07:46,840 Speaker 1: On top of that, now I'm recording this July twenty first, 157 00:07:46,840 --> 00:07:48,760 Speaker 1: there's an ICO story that came out today with some 158 00:07:48,800 --> 00:07:52,560 Speaker 1: bullshit vanity metrics, rope and AI. But that story said 159 00:07:52,560 --> 00:07:54,760 Speaker 1: the Open AI themselves had said that they have five 160 00:07:54,840 --> 00:07:57,800 Speaker 1: hundred million weekly active users. Now here's the crazy thing. 161 00:07:58,040 --> 00:08:01,000 Speaker 1: That five hundred million number. That from March. That's and 162 00:08:01,080 --> 00:08:03,880 Speaker 1: March now, but I came from an open AI funding announcement. 163 00:08:04,760 --> 00:08:08,080 Speaker 1: Have they platowed stick around? I'm going to find out. 164 00:08:08,400 --> 00:08:10,640 Speaker 1: But this is all worrying on so many levels, chief 165 00:08:10,640 --> 00:08:12,520 Speaker 1: of which is that everybody has been talking about AI 166 00:08:12,600 --> 00:08:14,920 Speaker 1: for three goddamn years, and everybody has said AI and 167 00:08:14,960 --> 00:08:17,760 Speaker 1: every earnings and media appearance, exhausting fucking blog posts, and 168 00:08:17,800 --> 00:08:20,960 Speaker 1: we still can't scrape together the bits to make functional industry. 169 00:08:22,160 --> 00:08:22,320 Speaker 2: Now. 170 00:08:22,360 --> 00:08:23,680 Speaker 1: I know some of you will probably read this in 171 00:08:23,680 --> 00:08:26,240 Speaker 1: point to chat gipt's users, and I quote myself writing 172 00:08:26,240 --> 00:08:28,320 Speaker 1: in my newsletter make fun of them, which also read 173 00:08:28,360 --> 00:08:31,600 Speaker 1: for this podcast. Chat gpt is allegedly five hundred million 174 00:08:31,600 --> 00:08:34,280 Speaker 1: weekly hatieth users and by the last count, only fifteen 175 00:08:34,320 --> 00:08:38,040 Speaker 1: point five million paying subscribers, an absolutely futrid conversion rate, 176 00:08:38,080 --> 00:08:40,240 Speaker 1: even before you realize that the actual conversion rate would 177 00:08:40,240 --> 00:08:43,080 Speaker 1: be monthly active subscribers, which is how any real software 178 00:08:43,080 --> 00:08:46,040 Speaker 1: company actually defines its metrics. By the fucking way, Why 179 00:08:46,120 --> 00:08:48,640 Speaker 1: is this impressive? Because it grew so fast it literally 180 00:08:48,720 --> 00:08:51,000 Speaker 1: had more pr and more marketing and more attention and 181 00:08:51,040 --> 00:08:53,840 Speaker 1: more opportunities to sell to more people than any company 182 00:08:53,840 --> 00:08:56,680 Speaker 1: has ever had in the history of anything. Every single 183 00:08:56,720 --> 00:08:59,000 Speaker 1: industry has been told to think about AI for three years, 184 00:08:59,040 --> 00:09:00,520 Speaker 1: and they've been told to do so because of a 185 00:09:00,520 --> 00:09:03,560 Speaker 1: company called open Ai. There isn't a single goddamn product 186 00:09:03,559 --> 00:09:06,440 Speaker 1: since Google or Facebook bes's had this level of media pressure, 187 00:09:06,640 --> 00:09:08,839 Speaker 1: and both of these companies launched without the massive amount 188 00:09:08,880 --> 00:09:11,960 Speaker 1: of media and social media that we have today. CHADGBT 189 00:09:12,160 --> 00:09:15,640 Speaker 1: is a very successful growth product and an absolutely horrifying business. 190 00:09:16,240 --> 00:09:19,360 Speaker 1: They're a banana republic that cannot function on their own. 191 00:09:19,760 --> 00:09:22,480 Speaker 1: They do not resemble Uber, Amazon, Web Services or any 192 00:09:22,520 --> 00:09:24,480 Speaker 1: other business in the past other than we work the 193 00:09:24,559 --> 00:09:26,920 Speaker 1: other company that SoftBank spent way too much money on, 194 00:09:27,400 --> 00:09:29,800 Speaker 1: and outside of chad GBT, there really isn't anything else. 195 00:09:30,880 --> 00:09:33,480 Speaker 1: But before I wrap up this three parter and I'm tired, 196 00:09:33,520 --> 00:09:35,680 Speaker 1: and I imagine you are too, I do want to 197 00:09:35,720 --> 00:09:39,080 Speaker 1: address something. I want to address the kind of both 198 00:09:39,160 --> 00:09:41,760 Speaker 1: sides thing that everyone's doing, and I think we all 199 00:09:41,800 --> 00:09:44,000 Speaker 1: need to do way less of this because at this 200 00:09:44,160 --> 00:09:49,040 Speaker 1: point it's kind of silly. Yes, Generati, if AI has functionality, 201 00:09:49,400 --> 00:09:51,840 Speaker 1: there are coding products and search products that people like 202 00:09:51,920 --> 00:09:54,400 Speaker 1: and some even pay for them. As I have discussed 203 00:09:54,400 --> 00:09:56,600 Speaker 1: across this series and I said discuss there, but I'm 204 00:09:56,679 --> 00:10:00,439 Speaker 1: keeping going and across countless podcasts and newslayers. None of 205 00:10:00,520 --> 00:10:03,280 Speaker 1: these companies are profitable until one of them is profitable, 206 00:10:03,559 --> 00:10:07,680 Speaker 1: really profitable sustainably, so generative AI based companies are not 207 00:10:07,800 --> 00:10:10,840 Speaker 1: real businesses in any case. The problem isn't so much 208 00:10:10,880 --> 00:10:13,080 Speaker 1: that l elms don't do anything, but that people talk 209 00:10:13,120 --> 00:10:15,760 Speaker 1: about them doing things they can't. The use of the 210 00:10:15,760 --> 00:10:18,040 Speaker 1: word agent is a deliberate attempt to suggest that l 211 00:10:18,040 --> 00:10:20,840 Speaker 1: elms are autonomous, and any and all stories about AI 212 00:10:20,920 --> 00:10:24,560 Speaker 1: replacing jobs are intentionally manipulative attempts to boost stock valuations 213 00:10:24,640 --> 00:10:27,079 Speaker 1: and suggests that models are capable of replacing human workers 214 00:10:27,080 --> 00:10:30,320 Speaker 1: at scale, which they are not. Allison Morrow of CNN, 215 00:10:30,320 --> 00:10:31,880 Speaker 1: who's been on the show a few times, also has 216 00:10:31,880 --> 00:10:33,520 Speaker 1: an excellent piece on this, which I've linked to in 217 00:10:33,559 --> 00:10:36,520 Speaker 1: the show notes. As I discussed in my newsletter Sincerity 218 00:10:36,520 --> 00:10:38,800 Speaker 1: Wins the War, which I've also linked to, this is 219 00:10:38,840 --> 00:10:40,800 Speaker 1: one of the most egregious failures of the tech media 220 00:10:40,800 --> 00:10:44,400 Speaker 1: I've ever seen, willingly pushing warrio Amma days outright making 221 00:10:44,480 --> 00:10:47,000 Speaker 1: up of stuff. No one asked him if that if 222 00:10:47,000 --> 00:10:50,080 Speaker 1: the stat he said ten to twenty percent white collar unemployment, 223 00:10:51,000 --> 00:10:53,079 Speaker 1: if I started lying, I just really want to be 224 00:10:53,120 --> 00:10:56,200 Speaker 1: clear about something. If I went on this podcast and 225 00:10:56,240 --> 00:10:59,240 Speaker 1: I just fucking made up stats, if I was just like, yes, 226 00:10:59,320 --> 00:11:03,480 Speaker 1: seventy five percent of all generative ais CEOs, literally they 227 00:11:03,960 --> 00:11:06,920 Speaker 1: love kicking puppies, they love drop kicking puppies, you'd get 228 00:11:06,920 --> 00:11:09,360 Speaker 1: mad at me. Right. What's the difference between that and 229 00:11:09,520 --> 00:11:12,920 Speaker 1: Wario Ama Day saying twenty percent white collar and unemployment. 230 00:11:12,960 --> 00:11:15,480 Speaker 1: There is none. They're both lies. And I don't know 231 00:11:15,520 --> 00:11:17,319 Speaker 1: why I'm doing the dog kicking one, but it's just 232 00:11:17,360 --> 00:11:20,160 Speaker 1: an example. We move on. And as a side note, 233 00:11:20,200 --> 00:11:24,480 Speaker 1: the discussion also of the term AGI artificial general intelligence, 234 00:11:24,520 --> 00:11:27,440 Speaker 1: and every acceptance of it is an attempt to suggest 235 00:11:27,480 --> 00:11:30,480 Speaker 1: that large language models can create conscious intelligence, a fictional 236 00:11:30,480 --> 00:11:33,560 Speaker 1: concept that even Meta's chief AI scientist says won't come 237 00:11:33,760 --> 00:11:37,960 Speaker 1: from scaling up llms. Members of the media, listen to me, 238 00:11:38,760 --> 00:11:41,160 Speaker 1: every time you talk about the really smart engineers that 239 00:11:41,280 --> 00:11:44,920 Speaker 1: say Meta is paying one hundred million dollars two know 240 00:11:45,040 --> 00:11:47,640 Speaker 1: that you are doing free marketing for these companies when 241 00:11:47,640 --> 00:11:49,840 Speaker 1: what's really happening is people are giving tens of millions 242 00:11:49,880 --> 00:11:51,400 Speaker 1: of dollars, the guys who will work on teams that 243 00:11:51,440 --> 00:11:54,440 Speaker 1: are pursuing a totally unproven concept. It is like saying 244 00:11:54,440 --> 00:11:56,679 Speaker 1: that they are all going to see if they could. 245 00:11:57,040 --> 00:11:59,000 Speaker 1: I don't know. I was about to say the wrong 246 00:11:59,120 --> 00:12:02,160 Speaker 1: create the wrong trend from Rollss and Grummet, Rollis and 247 00:12:02,400 --> 00:12:06,120 Speaker 1: Wallace and Grummet. But you know those are one hundred 248 00:12:06,120 --> 00:12:08,679 Speaker 1: percent possible. I'm sure someone could build them. How about this, 249 00:12:08,760 --> 00:12:10,320 Speaker 1: They all believe they're going to hunt and kill the 250 00:12:10,360 --> 00:12:13,440 Speaker 1: tooth Fairy because the tooth Fairy is about as realistic 251 00:12:13,480 --> 00:12:16,920 Speaker 1: as agi, and if you disagree with me, emummy, I 252 00:12:17,000 --> 00:12:19,199 Speaker 1: might even read him. I should also add that the 253 00:12:19,280 --> 00:12:22,360 Speaker 1: use of the word singularity is similarly manipulative. These are 254 00:12:22,920 --> 00:12:26,000 Speaker 1: terms that are used to obfuscate the actual abilities of 255 00:12:26,080 --> 00:12:29,880 Speaker 1: large language models and indeed what can be built on them. 256 00:12:30,160 --> 00:12:35,880 Speaker 1: It's really frustrating because these should be really obvious tricks, 257 00:12:36,920 --> 00:12:40,079 Speaker 1: but the media keeps accepting them. And I don't want 258 00:12:40,120 --> 00:12:42,880 Speaker 1: to accuse anyone of being anything. I just want to 259 00:12:42,880 --> 00:12:44,800 Speaker 1: say I find it fucking disgusting. And when this all 260 00:12:44,840 --> 00:12:47,920 Speaker 1: falls apart, there will be a referendum on this. I 261 00:12:48,000 --> 00:12:51,240 Speaker 1: will not be kind or merciful to the people that 262 00:12:51,360 --> 00:12:54,640 Speaker 1: have continually pushed these narratives. I also want to add 263 00:12:54,679 --> 00:12:58,199 Speaker 1: there's another really annoying narrative that we need to get 264 00:12:58,320 --> 00:13:01,320 Speaker 1: rid of, and that's these stories models lying, cheating, and 265 00:13:01,320 --> 00:13:04,360 Speaker 1: stealing to reach their goals or stop themselves being turned off. 266 00:13:04,400 --> 00:13:07,280 Speaker 1: These are intentionally deceptive, as these models can and clearly 267 00:13:07,360 --> 00:13:11,600 Speaker 1: are being prompted to take these actions. To be abundantly clear, 268 00:13:11,640 --> 00:13:14,320 Speaker 1: the manipulative suggestion here is that these models are autonomous 269 00:13:14,360 --> 00:13:16,400 Speaker 1: or conscious in some way, which they're not. When I 270 00:13:16,440 --> 00:13:19,040 Speaker 1: say prompted, I mean we don't have all of the 271 00:13:19,080 --> 00:13:21,800 Speaker 1: information as to how these models were trained or prompted 272 00:13:21,800 --> 00:13:25,040 Speaker 1: to get these answers. They can make them do them. 273 00:13:25,040 --> 00:13:27,960 Speaker 1: They could create a model to do it. And oh anthropic, 274 00:13:28,000 --> 00:13:32,160 Speaker 1: wouldn't lie, wouldn't they, Dario Amat They went on stage 275 00:13:32,200 --> 00:13:35,680 Speaker 1: and said they'll be twenty percent unemployment from white collar labors, 276 00:13:35,720 --> 00:13:37,280 Speaker 1: and he'll claim me he said it was the future. 277 00:13:37,280 --> 00:13:40,839 Speaker 1: It's a fucking lie, absolutely ridiculous. And I believe that 278 00:13:40,880 --> 00:13:44,200 Speaker 1: the generative AI market is a fifty billion dollar revenue industry, 279 00:13:44,200 --> 00:13:47,120 Speaker 1: masquerading is a trillion dollar one, and the media is helping. 280 00:13:47,760 --> 00:13:50,760 Speaker 1: And underneath this troubling fact there's another one that should 281 00:13:50,760 --> 00:13:54,439 Speaker 1: give us pause. The AI trade really is not about AI. 282 00:13:54,720 --> 00:13:58,120 Speaker 1: It's not about generative AI, it's not about autonomous agents, 283 00:13:58,160 --> 00:14:01,679 Speaker 1: it's not about AGI. It's all about GPUs, and it's 284 00:14:01,679 --> 00:14:14,120 Speaker 1: incredibly brittle as a result. Now, as I've explained at length, 285 00:14:14,200 --> 00:14:16,640 Speaker 1: the AI trade is not one based on revenue user growth, 286 00:14:16,640 --> 00:14:20,240 Speaker 1: the effixive tools or significance of any technological breakthrough. Stocks 287 00:14:20,280 --> 00:14:22,560 Speaker 1: are not moving based on whether they're making money or AI, 288 00:14:22,640 --> 00:14:25,280 Speaker 1: because if they were, they'd be moving downward. However, due 289 00:14:25,320 --> 00:14:27,480 Speaker 1: to the ViBe's based nature of the AI trade, companies 290 00:14:27,480 --> 00:14:31,800 Speaker 1: are benefiting from the press inexplicably crediting growth to AI 291 00:14:31,880 --> 00:14:34,520 Speaker 1: with no proof that that's the case. Open AI is 292 00:14:34,560 --> 00:14:36,920 Speaker 1: a terrible business, But the only business worse than open 293 00:14:36,920 --> 00:14:39,080 Speaker 1: AI are the companies built on top of them. Large 294 00:14:39,120 --> 00:14:41,440 Speaker 1: language models are too expensive to run and have limited 295 00:14:41,440 --> 00:14:44,640 Speaker 1: abilities beyond the ones I'm named previously, and because everybody's 296 00:14:44,680 --> 00:14:47,320 Speaker 1: running models that all on some level do similar things, 297 00:14:47,320 --> 00:14:50,080 Speaker 1: it's very hard for people to build really innovative products 298 00:14:50,080 --> 00:14:55,680 Speaker 1: on top of them. But ultimately, this entire trade, all 299 00:14:55,720 --> 00:14:59,880 Speaker 1: of the AI trade, hinges on GPUs. Coreweave was initially 300 00:15:00,040 --> 00:15:03,200 Speaker 1: funded by Nvidia. They're a company that does AI data centers. 301 00:15:03,240 --> 00:15:05,640 Speaker 1: I don't want to talk about them, makes me angry, 302 00:15:06,000 --> 00:15:09,840 Speaker 1: but their IPO was funded partially buy and Video sells 303 00:15:09,880 --> 00:15:13,000 Speaker 1: the GPUs. Video is also one of their customers, and 304 00:15:13,080 --> 00:15:17,480 Speaker 1: core Weave raises debt to buy GPUs by using the 305 00:15:17,520 --> 00:15:20,880 Speaker 1: GPUs they've already bought as collateral, and so they get 306 00:15:20,920 --> 00:15:24,160 Speaker 1: those loans, and then they use that money to build 307 00:15:24,240 --> 00:15:27,280 Speaker 1: data centers and buy more GPUs to put in them, 308 00:15:27,400 --> 00:15:31,200 Speaker 1: which they then use as leverage to Do you see 309 00:15:31,240 --> 00:15:34,160 Speaker 1: the Do you see the problem? Do you see the 310 00:15:34,240 --> 00:15:37,200 Speaker 1: issue with that? This isn't me being polemic or hysterical, 311 00:15:37,240 --> 00:15:38,920 Speaker 1: by the way, this is quite literally what's happening and 312 00:15:38,960 --> 00:15:40,920 Speaker 1: how core Weave operates. If you want alarm by that, 313 00:15:40,960 --> 00:15:43,760 Speaker 1: I'm not sure what to tell you Elsewhere. Oracle is 314 00:15:43,800 --> 00:15:46,520 Speaker 1: buying forty billion dollars in GPUs for the still uninformed 315 00:15:46,520 --> 00:15:49,440 Speaker 1: Stargate Data Center project, and Meta is building a Manhattan 316 00:15:49,520 --> 00:15:51,840 Speaker 1: sized data center to fill within video GPUs. If you 317 00:15:51,880 --> 00:15:54,200 Speaker 1: believe that Meta will actually do that, which I do not. 318 00:15:55,080 --> 00:15:57,640 Speaker 1: Open AI is also Microsoft's largest as your client and 319 00:15:57,720 --> 00:16:00,720 Speaker 1: insanely risky proposition on multiple levels, not simply in the 320 00:16:00,720 --> 00:16:03,160 Speaker 1: fact that it's serving the revenue at cost, but the 321 00:16:03,200 --> 00:16:06,320 Speaker 1: Microsoft executives believed that open ai would fail in the 322 00:16:06,320 --> 00:16:08,560 Speaker 1: long term when they invested in twenty twenty three. According 323 00:16:08,600 --> 00:16:12,440 Speaker 1: to the information, Microsoft is in Videa's largest client for GPUs, 324 00:16:12,640 --> 00:16:15,360 Speaker 1: meaning that any changes to Microsoft's future and interest in 325 00:16:15,400 --> 00:16:18,320 Speaker 1: open AI, such as reducing its data center expansion, would 326 00:16:18,320 --> 00:16:21,320 Speaker 1: eventually hit in Vidia's revenue. Why do you think deep 327 00:16:21,400 --> 00:16:23,640 Speaker 1: seek shocked the market. It wasn't because of any clunky 328 00:16:23,680 --> 00:16:26,640 Speaker 1: story around training techniques or costs. It was because it's 329 00:16:26,640 --> 00:16:28,720 Speaker 1: said to the market that and Video might not sell 330 00:16:28,760 --> 00:16:34,280 Speaker 1: more GPUs every single quarter in perpetuity. Microsoft, Meta, Google, Apple, Amazon, 331 00:16:34,320 --> 00:16:37,560 Speaker 1: and Tesla aren't making much money from AI. In fact, 332 00:16:37,600 --> 00:16:40,120 Speaker 1: they're losing billions of dollars on whatever revenues they do 333 00:16:40,200 --> 00:16:42,520 Speaker 1: make from it. Their stock growth is not coming from 334 00:16:42,560 --> 00:16:45,360 Speaker 1: actual revenue or actual products, but the vibes around being 335 00:16:45,400 --> 00:16:48,560 Speaker 1: an AI company, which means absolute jack shit when you 336 00:16:48,600 --> 00:16:51,200 Speaker 1: don't have to users, finances, or products to back them up. 337 00:16:52,000 --> 00:16:54,800 Speaker 1: So really everything on the AI trade comes down to 338 00:16:54,840 --> 00:16:57,560 Speaker 1: in Video's ability to sell GPUs and more of them 339 00:16:57,560 --> 00:17:00,160 Speaker 1: each quarter. And this industry, if we're really honest, is 340 00:17:00,160 --> 00:17:02,560 Speaker 1: at a point where it really only exists to do 341 00:17:02,680 --> 00:17:06,440 Speaker 1: so genera. If AI products do not provide significant revenue growth, 342 00:17:06,440 --> 00:17:08,480 Speaker 1: its products are not useful in the way that unlock 343 00:17:08,520 --> 00:17:11,879 Speaker 1: significant business value, and the products that have some adoption 344 00:17:12,000 --> 00:17:17,399 Speaker 1: run at such a grotesque loss. It's really sad. But 345 00:17:17,480 --> 00:17:19,399 Speaker 1: I realize I've thrown a lot at you. And for 346 00:17:19,440 --> 00:17:21,480 Speaker 1: the second time this year, I've recorded the three part 347 00:17:21,560 --> 00:17:24,680 Speaker 1: podcast series. And what's great is you may think maybe 348 00:17:24,680 --> 00:17:27,640 Speaker 1: I took a break. Maybe no. I recorded these bad 349 00:17:27,680 --> 00:17:29,960 Speaker 1: boys back to back to back. That's why I sound 350 00:17:30,040 --> 00:17:32,280 Speaker 1: so tired and why my voice kind of I think 351 00:17:32,320 --> 00:17:36,000 Speaker 1: smooths out. I like doing this, but I needed to 352 00:17:36,040 --> 00:17:37,719 Speaker 1: say all of this. I really did because I am 353 00:17:37,800 --> 00:17:42,080 Speaker 1: genuinely really worried. We're in a bubble. If you do 354 00:17:42,160 --> 00:17:44,320 Speaker 1: not think we're in a bubble, you are not looking outside. 355 00:17:44,440 --> 00:17:48,879 Speaker 1: Apollo Global chief economist Torsten slock Killer name said it 356 00:17:48,960 --> 00:17:51,879 Speaker 1: last week in a research note published July sixteenth. Okay, 357 00:17:52,000 --> 00:17:54,520 Speaker 1: what he said was much worse. The difference between the 358 00:17:54,560 --> 00:17:56,639 Speaker 1: IT bubble in the nineteen nineties and the AI bubble 359 00:17:56,640 --> 00:17:58,240 Speaker 1: today is that the top ten companies in the S 360 00:17:58,280 --> 00:17:59,879 Speaker 1: and p today are more overvalued than they were in 361 00:17:59,880 --> 00:18:02,680 Speaker 1: the nineteen nineties. Slock wrote in a recent research note 362 00:18:02,720 --> 00:18:05,200 Speaker 1: that was widely shared across social media. In financial circles, 363 00:18:06,280 --> 00:18:08,840 Speaker 1: we're in a bubble. We're in a bubble general. If 364 00:18:08,840 --> 00:18:11,200 Speaker 1: AI does not do the things that it's being sold 365 00:18:11,240 --> 00:18:13,520 Speaker 1: as doing, and the things it can do aren't the 366 00:18:13,640 --> 00:18:16,000 Speaker 1: kind of things that create business returns, automate labor are 367 00:18:16,000 --> 00:18:17,840 Speaker 1: really do much more than one extension of a cloud 368 00:18:17,880 --> 00:18:20,800 Speaker 1: service platform. The money isn't there, the users aren't there. 369 00:18:20,880 --> 00:18:22,959 Speaker 1: Every company seems to lose money, and some companies lose 370 00:18:23,040 --> 00:18:25,320 Speaker 1: so much money it's impossible to tell how they'll survive. 371 00:18:25,840 --> 00:18:29,800 Speaker 1: Worse Still, this bubble is entirely symbolic the bailouts. And 372 00:18:29,840 --> 00:18:32,240 Speaker 1: I know I get a lot of emails about this, 373 00:18:32,320 --> 00:18:35,399 Speaker 1: and I'm really going to address this issue right now. 374 00:18:35,640 --> 00:18:37,840 Speaker 1: The bailouts of the Great Financial Crisis were focused on 375 00:18:37,880 --> 00:18:39,720 Speaker 1: banks and funds that have failed because they ran out 376 00:18:39,720 --> 00:18:42,359 Speaker 1: of money, and the top initiative existed to plug holes 377 00:18:42,520 --> 00:18:47,040 Speaker 1: with low interest loans. There are no few holes to 378 00:18:47,040 --> 00:18:50,480 Speaker 1: plug here, because even if OpenAI and anthropics somehow became 379 00:18:50,480 --> 00:18:53,320 Speaker 1: eternal money burners, the AI trade exists based on the 380 00:18:53,359 --> 00:18:57,239 Speaker 1: continued and continually increasing sale and use of GPUs. There 381 00:18:57,240 --> 00:18:59,639 Speaker 1: are limited amounts of capital, but also limited amounts of 382 00:18:59,720 --> 00:19:02,040 Speaker 1: data is to actually put GPUs, and on top of that, 383 00:19:02,280 --> 00:19:04,040 Speaker 1: at some point growth will slow at one of the 384 00:19:04,080 --> 00:19:07,040 Speaker 1: magnificent seven, at which cost time costs will have to 385 00:19:07,080 --> 00:19:09,080 Speaker 1: come down from things that lose them tons of money, 386 00:19:09,080 --> 00:19:11,600 Speaker 1: such as generative AI. I really want to be clear, 387 00:19:11,640 --> 00:19:15,200 Speaker 1: though a lot of people say, oh, it's really kind 388 00:19:15,240 --> 00:19:18,399 Speaker 1: of lazy doomerism. I don't appreciate it, because, like, if 389 00:19:18,440 --> 00:19:20,320 Speaker 1: you're going to be worried about something, be worried about 390 00:19:20,359 --> 00:19:23,080 Speaker 1: something that's real, you're saying, oh, Trump will just bail 391 00:19:23,080 --> 00:19:25,440 Speaker 1: them out. Oh, they'll get bailed out. Who will get 392 00:19:25,440 --> 00:19:28,879 Speaker 1: bailed out? For this fast to continue? In video? Is 393 00:19:28,920 --> 00:19:31,359 Speaker 1: thirty nine point one billion dollars of data cent revenue 394 00:19:31,359 --> 00:19:32,840 Speaker 1: in the last quarter. That means the next one needs 395 00:19:32,880 --> 00:19:34,399 Speaker 1: to be forty four and one after that needs to 396 00:19:34,440 --> 00:19:38,320 Speaker 1: be like fifty sixty eighty one hundred. This needs to 397 00:19:38,359 --> 00:19:42,119 Speaker 1: continue forever. Is the government going to nationalize in video? 398 00:19:42,280 --> 00:19:42,359 Speaker 2: Is? 399 00:19:42,760 --> 00:19:44,080 Speaker 1: Well? I guess that would take them off the market, 400 00:19:44,119 --> 00:19:46,760 Speaker 1: wouldn't it is they're just going to feed in video 401 00:19:46,840 --> 00:19:49,600 Speaker 1: fifty sixty seventy billion dollars. I guess if one of 402 00:19:49,640 --> 00:19:51,800 Speaker 1: these AI companies fail, they could give them money, but 403 00:19:51,840 --> 00:19:53,840 Speaker 1: they're just going to go ahead and fucking lose it. 404 00:19:54,640 --> 00:19:56,159 Speaker 1: What do you think they're going to do? Where does 405 00:19:56,200 --> 00:19:59,359 Speaker 1: the money go? In the Great Financial Crisis, banks failed, 406 00:19:59,400 --> 00:20:02,000 Speaker 1: you had to give the bank's money. Hedge funds failed, 407 00:20:02,040 --> 00:20:03,959 Speaker 1: you had to give money to places, and then they 408 00:20:03,960 --> 00:20:05,880 Speaker 1: took the money and they weren't out of money anymore. 409 00:20:06,359 --> 00:20:08,240 Speaker 1: This isn't a case where companies will run out of 410 00:20:08,240 --> 00:20:11,480 Speaker 1: money but otherwise be okay. They don't have a way 411 00:20:11,520 --> 00:20:14,840 Speaker 1: of being okay. There is no way of being okay. 412 00:20:15,000 --> 00:20:16,800 Speaker 1: If you want to be a duma. Look at the 413 00:20:16,840 --> 00:20:19,280 Speaker 1: fact that the fucking stock market is built on saling 414 00:20:19,320 --> 00:20:22,480 Speaker 1: more and more GPUs every quarter. There is no plugging 415 00:20:22,520 --> 00:20:26,080 Speaker 1: that hole. There's nothing for you to do. Oh god, 416 00:20:26,119 --> 00:20:30,880 Speaker 1: this guy, what isn't the cost of inference coming down? No, 417 00:20:31,800 --> 00:20:34,000 Speaker 1: you do not have proof of this statement. The cost 418 00:20:34,000 --> 00:20:36,320 Speaker 1: of tokens is going down, and that is not the 419 00:20:36,359 --> 00:20:39,240 Speaker 1: cost of inference going down. Everyone's saying this is they're 420 00:20:39,240 --> 00:20:40,679 Speaker 1: saying it because a guy once said it to them. 421 00:20:40,720 --> 00:20:42,240 Speaker 1: You do not have proof. I have more proof for 422 00:20:42,280 --> 00:20:45,160 Speaker 1: what I am saying, well, it theoretically might be going down. 423 00:20:45,880 --> 00:20:48,800 Speaker 1: All evidence points to the larger models costing more money, especially 424 00:20:48,880 --> 00:20:52,199 Speaker 1: reasoning having ones like Claude Opu's four. Inference is not 425 00:20:52,280 --> 00:20:54,520 Speaker 1: the only thing happening. If this is your one response 426 00:20:54,560 --> 00:20:56,520 Speaker 1: for a big bozo and a dufus and should go 427 00:20:56,560 --> 00:20:59,280 Speaker 1: back to making squeaky noises when you see tech executives 428 00:20:59,359 --> 00:21:02,960 Speaker 1: or hear my name and a podcast. It's sickening. Even 429 00:21:03,000 --> 00:21:05,080 Speaker 1: if the cost of inference is going down, on what 430 00:21:05,320 --> 00:21:09,120 Speaker 1: model exactly is it coming down on? I don't know three, 431 00:21:09,520 --> 00:21:11,840 Speaker 1: opus fource on it. Four. I don't think it is. 432 00:21:12,320 --> 00:21:14,720 Speaker 1: The price that the model developers are charging. By the way, 433 00:21:14,960 --> 00:21:18,120 Speaker 1: is not the price of inference. That's not the same thing. 434 00:21:18,960 --> 00:21:21,400 Speaker 1: It's not the same thing at all. I'm really sick 435 00:21:21,440 --> 00:21:23,840 Speaker 1: of this goddamn point. You can hear it in my voice. 436 00:21:24,280 --> 00:21:27,040 Speaker 1: But let's move on. So some people bring up a 437 00:21:27,160 --> 00:21:30,280 Speaker 1: six and they are these customized chips for specific options 438 00:21:30,280 --> 00:21:33,240 Speaker 1: to reduce the amount that these companies are spending on compute. Right. 439 00:21:33,480 --> 00:21:35,879 Speaker 1: I have a few thoughts when is this going to happen? 440 00:21:35,920 --> 00:21:40,280 Speaker 1: For example, say open aim Broadcom actually build their asick 441 00:21:40,320 --> 00:21:42,760 Speaker 1: in twenty sixteen, twenty twenty six. They won't do this, 442 00:21:42,800 --> 00:21:44,760 Speaker 1: but that's when they're meant to How many will they build? 443 00:21:44,840 --> 00:21:47,840 Speaker 1: Do they have contact contracts with companies that can actually 444 00:21:48,040 --> 00:21:51,560 Speaker 1: produce high performance silicon or there's only three of those, Samsung, 445 00:21:51,560 --> 00:21:55,040 Speaker 1: TSMC and SMICE which is currently sanctioned. And these companies 446 00:21:55,359 --> 00:21:58,360 Speaker 1: typically have their capacity booked well in advance. Great example 447 00:21:58,400 --> 00:22:02,359 Speaker 1: for you in Nvidia restarts that h twenty GPUs and 448 00:22:02,480 --> 00:22:05,639 Speaker 1: told customers is reported by the information that they're not 449 00:22:05,720 --> 00:22:07,679 Speaker 1: going to make more because the capacity they had at 450 00:22:07,680 --> 00:22:12,919 Speaker 1: TSMC got given away because they canceled it. Even starting 451 00:22:12,920 --> 00:22:15,439 Speaker 1: a production run of a semiconductor product can take weeks. 452 00:22:15,560 --> 00:22:18,439 Speaker 1: Do they have the server architecture prepared? Have they tested it? 453 00:22:18,480 --> 00:22:21,199 Speaker 1: Does it work? Is the performance actually good? Because the 454 00:22:21,200 --> 00:22:23,920 Speaker 1: information reported the Microsoft has failed to create a workable, 455 00:22:23,920 --> 00:22:25,600 Speaker 1: reliable ASIC and the one that they're going to bring 456 00:22:25,640 --> 00:22:27,959 Speaker 1: out next year is going to be still worse than 457 00:22:28,000 --> 00:22:31,320 Speaker 1: in videos. What makes open ay special? What gives them 458 00:22:31,359 --> 00:22:34,240 Speaker 1: the industry knowledge that means that they won't run into 459 00:22:34,240 --> 00:22:39,280 Speaker 1: these problems? The answer is nothing, absolutely bloody nothing. Anyway, 460 00:22:39,840 --> 00:22:42,960 Speaker 1: if this happens, and should be clear, it takes a 461 00:22:43,000 --> 00:22:44,800 Speaker 1: lot of money to build these chips, and they're yet 462 00:22:44,800 --> 00:22:47,080 Speaker 1: to prove that these chips be better than Nvideo GPUs 463 00:22:47,080 --> 00:22:50,280 Speaker 1: FORRAI compume they're going to have to retrofit every data 464 00:22:50,280 --> 00:22:52,119 Speaker 1: center for them. By the way, they're not just going 465 00:22:52,200 --> 00:22:54,919 Speaker 1: to plug into the same thing, these Blackwell chips, and 466 00:22:54,960 --> 00:22:58,119 Speaker 1: actually generally GPUs they have specialized server architecture and N 467 00:22:58,160 --> 00:23:01,520 Speaker 1: Video is a big stickler for DETAI. So we're just 468 00:23:01,520 --> 00:23:03,240 Speaker 1: going to rip up all the stuff. So this will 469 00:23:03,280 --> 00:23:05,800 Speaker 1: I assume take two minutes, because it needs to. It's 470 00:23:05,840 --> 00:23:07,879 Speaker 1: the only way. By the way, if this actually happens, 471 00:23:07,880 --> 00:23:09,800 Speaker 1: it still fucks up the AI trade because then video 472 00:23:09,840 --> 00:23:17,600 Speaker 1: still needs to sell GPUs. Jesus Christ, I'm sorry. I'm sorry, 473 00:23:17,560 --> 00:23:22,639 Speaker 1: And people ask, is this real? Anger? Yeah. I I 474 00:23:22,760 --> 00:23:26,280 Speaker 1: read about these things, and I read the inconsistencies in 475 00:23:26,320 --> 00:23:29,760 Speaker 1: the media, and I talk about them, and I think 476 00:23:29,800 --> 00:23:32,000 Speaker 1: of the amount of times in my life I have 477 00:23:32,119 --> 00:23:34,560 Speaker 1: had people make me feel stupid or make me feel crazy. 478 00:23:34,600 --> 00:23:36,560 Speaker 1: I'm sure many of you have had this experience too, 479 00:23:36,760 --> 00:23:39,680 Speaker 1: people that are allegedly smart, people that have positions of power, 480 00:23:39,720 --> 00:23:44,359 Speaker 1: people that we look to as responsible figures. And I 481 00:23:44,400 --> 00:23:46,240 Speaker 1: think about all the times that I've known I was right, 482 00:23:46,280 --> 00:23:49,879 Speaker 1: but because there was a consensus, a group ideal, you 483 00:23:49,920 --> 00:23:51,480 Speaker 1: had to kind of stick by it even if it 484 00:23:51,520 --> 00:23:55,159 Speaker 1: was something factually wrong. What I did the experience in 485 00:23:55,200 --> 00:23:59,159 Speaker 1: the past was seeing the media do it like this, 486 00:23:59,320 --> 00:24:01,520 Speaker 1: and the way in which the media has done it, 487 00:24:01,520 --> 00:24:05,120 Speaker 1: it's sickening. It's sickening because the people that will get 488 00:24:05,240 --> 00:24:09,159 Speaker 1: hurt will be the people whose retirements and who's just 489 00:24:09,880 --> 00:24:13,000 Speaker 1: savings are going to get lamped by this crap. And 490 00:24:13,040 --> 00:24:16,399 Speaker 1: I'm worried. I'm worried because despite all these obvious, brutal 491 00:24:16,400 --> 00:24:19,439 Speaker 1: and near on fixable problems, everybody is walking around acting 492 00:24:19,480 --> 00:24:21,920 Speaker 1: like things are going great with AI. The New York 493 00:24:21,960 --> 00:24:24,720 Speaker 1: Times claims everybody is using AI for everything, and I'm talking, 494 00:24:24,760 --> 00:24:29,280 Speaker 1: of course about Kevin Ruse and Casey Newton. But baby 495 00:24:29,920 --> 00:24:32,479 Speaker 1: cannot wait for when this explodes. Guys, you have no 496 00:24:32,560 --> 00:24:34,480 Speaker 1: idea how funny it's going to be. And by the way, 497 00:24:34,480 --> 00:24:36,280 Speaker 1: this is a blatant lie, one that exists to prop 498 00:24:36,359 --> 00:24:38,639 Speaker 1: up an industry that has categorically failed to deliver the 499 00:24:38,680 --> 00:24:41,560 Speaker 1: innovations of returns that it promised, yet still receives this 500 00:24:41,600 --> 00:24:43,960 Speaker 1: glowing press from a tech and business media that refuses 501 00:24:44,000 --> 00:24:45,840 Speaker 1: to look outside and see that the sky is red 502 00:24:45,880 --> 00:24:49,560 Speaker 1: and frogs are landing everywhere other than the frog thing. 503 00:24:49,680 --> 00:24:53,480 Speaker 1: I'm not being dramatic. Everywhere you look in the AI trade, 504 00:24:53,480 --> 00:24:56,879 Speaker 1: things are getting worse. No revenue, William's being burned, no mote, 505 00:24:56,880 --> 00:25:00,200 Speaker 1: no infrastructure play, no comparables in history other than the 506 00:25:00,240 --> 00:25:02,040 Speaker 1: dot com bubble and we work, and a series of 507 00:25:02,040 --> 00:25:05,000 Speaker 1: flagrant lies spouted by the powerful and members of the 508 00:25:05,000 --> 00:25:08,800 Speaker 1: press that are afraid of moving against market consensus. Worse still, 509 00:25:08,840 --> 00:25:12,560 Speaker 1: despite Nvidia's strength, in Vidia is the market's weakness through 510 00:25:12,560 --> 00:25:15,520 Speaker 1: no fault of its own. Really, Jensen Wong sells GPUs, 511 00:25:15,600 --> 00:25:17,679 Speaker 1: people want to buy GPUs, and now the rest of 512 00:25:17,720 --> 00:25:20,840 Speaker 1: the market is leading aggressively on one company selling one thing, 513 00:25:21,160 --> 00:25:23,320 Speaker 1: feeding it billions of dollars in the hopes that the 514 00:25:23,320 --> 00:25:26,600 Speaker 1: things they're buying start making them profit. And that really 515 00:25:26,760 --> 00:25:29,480 Speaker 1: is the most ridiculous thing. At the center of the 516 00:25:29,520 --> 00:25:33,520 Speaker 1: AI trade sits enterprise GPUs that on installation immediately start 517 00:25:33,600 --> 00:25:36,800 Speaker 1: losing the company and question money. Large language models burn 518 00:25:36,880 --> 00:25:39,320 Speaker 1: cash for negative returns to build products that all kind 519 00:25:39,320 --> 00:25:42,000 Speaker 1: of work the same way. Now, if you're gonna say 520 00:25:42,000 --> 00:25:44,000 Speaker 1: I'm wrong, sit and think carefully about why is it 521 00:25:44,000 --> 00:25:45,720 Speaker 1: because you don't want me to be right? Is it 522 00:25:45,760 --> 00:25:48,680 Speaker 1: because you think these companies will work it out? This 523 00:25:48,760 --> 00:25:51,960 Speaker 1: isn't anything like uber aws or any other situation. It's 524 00:25:52,000 --> 00:25:55,200 Speaker 1: its own monstrosity, a creature of hubrious and ignorance caused 525 00:25:55,240 --> 00:25:57,359 Speaker 1: by a tech industry that's out of ideas, built on 526 00:25:57,359 --> 00:26:01,280 Speaker 1: top of a one company. Really, play with me all 527 00:26:01,320 --> 00:26:03,560 Speaker 1: you want about how there are actual people using AI. 528 00:26:03,880 --> 00:26:06,679 Speaker 1: You've probably read the my AIS skeptic friends are All 529 00:26:06,800 --> 00:26:09,000 Speaker 1: Nuts blog, and if you're going to send that to me, 530 00:26:09,080 --> 00:26:11,800 Speaker 1: read the response from Nick Suesh First. I've linked both 531 00:26:11,840 --> 00:26:13,800 Speaker 1: in the show notes. And if you're going to say 532 00:26:13,840 --> 00:26:16,080 Speaker 1: that I don't actually speak to people who use these products, 533 00:26:16,080 --> 00:26:20,840 Speaker 1: you are categorically wrong and in denial. I'm only saying 534 00:26:20,920 --> 00:26:22,840 Speaker 1: all of this with this aggressive tone because for the 535 00:26:22,840 --> 00:26:25,280 Speaker 1: best part of two years, I've been made to repeatedly 536 00:26:25,280 --> 00:26:28,040 Speaker 1: explain myself in a way that no AI optimist has 537 00:26:28,080 --> 00:26:31,000 Speaker 1: ever been made, and I admit I resent it. I 538 00:26:31,040 --> 00:26:34,400 Speaker 1: have read hundreds of thousands of words, with hundreds of citations, 539 00:26:34,400 --> 00:26:37,960 Speaker 1: and still to this day, there are some people that 540 00:26:38,000 --> 00:26:40,959 Speaker 1: claim I'm somehow flawed in my analysis, that I'm missing something, 541 00:26:40,960 --> 00:26:43,159 Speaker 1: that I am somehow failing to make my case that 542 00:26:43,200 --> 00:26:46,520 Speaker 1: I haven't spoken to enough fart sniffers, that I haven't 543 00:26:46,840 --> 00:26:51,920 Speaker 1: delighted in the ideals and falsities of AGI. The only 544 00:26:51,960 --> 00:26:54,600 Speaker 1: people failing to make their case the AI optimists, still 545 00:26:54,640 --> 00:26:57,800 Speaker 1: claiming that these companies are making powerful AI, and once 546 00:26:57,880 --> 00:27:00,760 Speaker 1: this bubble pops, I will be asking for an apology. 547 00:27:15,280 --> 00:27:18,600 Speaker 1: But I must be clear, even though a lot of 548 00:27:18,640 --> 00:27:21,960 Speaker 1: things are coming true that I thought would, and I 549 00:27:22,000 --> 00:27:25,040 Speaker 1: think more will come true, I'll be proven right again 550 00:27:25,080 --> 00:27:29,280 Speaker 1: and again. I don't actually like what's happening. And I 551 00:27:29,400 --> 00:27:32,440 Speaker 1: like ending these podcasts with personal thoughts about stuff because 552 00:27:32,520 --> 00:27:35,639 Speaker 1: I'm an emotional and overly honest person. I enjoy doing 553 00:27:35,680 --> 00:27:39,960 Speaker 1: this a great deal. I do not, however, enjoy telling 554 00:27:39,960 --> 00:27:42,800 Speaker 1: you at length how brittle everything is. An ideal tech 555 00:27:42,840 --> 00:27:45,639 Speaker 1: industry would be one built on innovation, revenue, real growth, 556 00:27:45,720 --> 00:27:48,280 Speaker 1: based on actual business returns that helped humans be better, 557 00:27:48,560 --> 00:27:52,760 Speaker 1: and not outright lies about replacing them. All that Generative 558 00:27:52,760 --> 00:27:54,680 Speaker 1: AI has done has shown how much loss there is 559 00:27:54,720 --> 00:27:57,040 Speaker 1: in both the markets and the media and replacing labor, 560 00:27:57,520 --> 00:27:59,280 Speaker 1: and I must be clear, it really is the media. 561 00:27:59,320 --> 00:28:01,840 Speaker 1: There are some of you who are fucking circos and 562 00:28:02,000 --> 00:28:05,199 Speaker 1: not pointing to anyone specifically. But if you are excited 563 00:28:05,280 --> 00:28:08,240 Speaker 1: about AI replacing people, I need you to review your 564 00:28:08,280 --> 00:28:10,440 Speaker 1: biases a little bit. I need you to think about 565 00:28:10,480 --> 00:28:13,880 Speaker 1: what it is you're so excited about another aside. By 566 00:28:13,920 --> 00:28:18,080 Speaker 1: the way, all these agi artificial general intelligence conversations, they 567 00:28:18,160 --> 00:28:20,560 Speaker 1: love to talk about why they're scared of them, but 568 00:28:20,600 --> 00:28:22,439 Speaker 1: they never want to talk about one thing, which is, 569 00:28:22,480 --> 00:28:27,440 Speaker 1: if we make a conscious intelligence, wouldn't that have personage 570 00:28:28,520 --> 00:28:34,440 Speaker 1: and we control and manipulate this conscious being, that's slavery. 571 00:28:34,560 --> 00:28:36,800 Speaker 1: We would have been inventing a new kind of slave, 572 00:28:37,560 --> 00:28:41,200 Speaker 1: because if we are making computers that have consciousness, that's 573 00:28:41,240 --> 00:28:44,360 Speaker 1: what we are doing. If you don't believe this, by 574 00:28:44,400 --> 00:28:47,440 Speaker 1: the way, you just don't want to think about the truth. 575 00:28:48,880 --> 00:28:51,640 Speaker 1: But on the jobless thing, I really do believe that 576 00:28:51,680 --> 00:28:55,000 Speaker 1: there are multiple reporters if your genuine excitement when they 577 00:28:55,040 --> 00:28:58,080 Speaker 1: write scary stories about how Daria Amadasa's white collar workers 578 00:28:58,080 --> 00:28:59,640 Speaker 1: will be fired in the next few years in favor 579 00:28:59,680 --> 00:29:03,080 Speaker 1: of age. It's gross and I know I'm repeating myself, 580 00:29:03,080 --> 00:29:05,600 Speaker 1: but I really must be clear, be more mindful about 581 00:29:05,600 --> 00:29:07,680 Speaker 1: what you're writing, because what you're writing does not prove 582 00:29:07,760 --> 00:29:10,880 Speaker 1: this thesis at all. You're just repeating a guy's thingy, 583 00:29:11,240 --> 00:29:15,440 Speaker 1: you're repeating what a guy said, and it's sickening. But 584 00:29:15,560 --> 00:29:17,520 Speaker 1: before I move into the end of this, I also 585 00:29:17,560 --> 00:29:20,160 Speaker 1: want to address one other thing that and I'm doing this, 586 00:29:20,200 --> 00:29:22,360 Speaker 1: I'm fully going to admit because someone just sent me 587 00:29:22,360 --> 00:29:24,920 Speaker 1: an email about this and they just sent me this 588 00:29:25,320 --> 00:29:28,120 Speaker 1: gotcha thing where it's okay, the government just gave Department 589 00:29:28,160 --> 00:29:30,480 Speaker 1: of Defense gave two hundred million dollar contracts to open 590 00:29:30,480 --> 00:29:34,959 Speaker 1: Ai Xai Google anthropic. I believe and that somehow government 591 00:29:34,960 --> 00:29:36,720 Speaker 1: money will plug this. Well, first of all, all the 592 00:29:36,760 --> 00:29:40,160 Speaker 1: revenue is unprofitable. Second of all, no, it won't like it. 593 00:29:40,440 --> 00:29:42,840 Speaker 1: It isn't enough money they need for open Ai to 594 00:29:42,920 --> 00:29:45,600 Speaker 1: somehow become profitable. They need to need like ten billion 595 00:29:45,640 --> 00:29:48,520 Speaker 1: dollars of free money every year, like just given to 596 00:29:48,560 --> 00:29:52,920 Speaker 1: them with no expenditure. On top of this, there is 597 00:29:53,000 --> 00:29:55,880 Speaker 1: this weird duma of philosophy. It's like, oh, the government 598 00:29:55,920 --> 00:29:58,400 Speaker 1: will use it for propaganda. They will use it as 599 00:29:58,400 --> 00:30:01,240 Speaker 1: a cheap propaganda machine. Bec and really does propaganda. We've 600 00:30:01,240 --> 00:30:06,080 Speaker 1: done propaganda for years, every government does. There's nothing efficient 601 00:30:06,160 --> 00:30:08,720 Speaker 1: or cheap or even useful about it. They're already the 602 00:30:09,080 --> 00:30:12,720 Speaker 1: White House is already posting AI generated stuff. Is how 603 00:30:12,800 --> 00:30:14,840 Speaker 1: is this going to keep the industry going exactly? Who 604 00:30:14,880 --> 00:30:17,680 Speaker 1: fucking knows. But then there's this thing of oh, the 605 00:30:17,720 --> 00:30:21,440 Speaker 1: surveillance state, the generative AI. This will prop up open AI, 606 00:30:21,520 --> 00:30:23,440 Speaker 1: anthropic and the likes. It will plug into all the 607 00:30:23,480 --> 00:30:25,960 Speaker 1: data sets and go, oh, we can do shit. Even 608 00:30:26,000 --> 00:30:27,760 Speaker 1: if it does that, I don't even want to think 609 00:30:27,760 --> 00:30:30,320 Speaker 1: of them doing this. It's not a fucking business. It's 610 00:30:30,320 --> 00:30:32,440 Speaker 1: not going to make them that much money. There's nothing 611 00:30:32,440 --> 00:30:35,360 Speaker 1: specialist about it because all large anguish models are fundamentally 612 00:30:35,400 --> 00:30:38,040 Speaker 1: trained on the same data, not only really tweaked by 613 00:30:38,080 --> 00:30:42,160 Speaker 1: reinforcement learning and the like. It's just they'll just buy 614 00:30:42,200 --> 00:30:45,040 Speaker 1: whatever model is cheapest, or they'll put it on prem 615 00:30:45,120 --> 00:30:47,719 Speaker 1: It doesn't really And I do want to say, if 616 00:30:47,760 --> 00:30:50,160 Speaker 1: you're scared of this stuff, be scared of the real stuff. 617 00:30:50,360 --> 00:30:54,040 Speaker 1: Don't just assume the worst and then don't think about 618 00:30:54,080 --> 00:30:55,840 Speaker 1: it much more. If you're going to assume the worst, 619 00:30:56,040 --> 00:31:01,640 Speaker 1: assume based on facts. I want you to feel better 620 00:31:01,720 --> 00:31:04,120 Speaker 1: when you listen to this, or at least edified. And 621 00:31:04,200 --> 00:31:06,840 Speaker 1: I don't want to ever scare you. I don't really 622 00:31:06,920 --> 00:31:08,600 Speaker 1: want to tell you how to feel. Ever, it's one 623 00:31:08,640 --> 00:31:10,840 Speaker 1: of the early feedbacks on the shows. Never tell people 624 00:31:10,880 --> 00:31:13,920 Speaker 1: how to feel, but I want you to contact me 625 00:31:13,960 --> 00:31:15,880 Speaker 1: if you're ever worried. I'm happy to help. But also 626 00:31:16,000 --> 00:31:19,680 Speaker 1: know that these are not smart men, and things have 627 00:31:19,760 --> 00:31:22,600 Speaker 1: got out of hand, and everything I'm discussing is the 628 00:31:22,640 --> 00:31:24,960 Speaker 1: result of the rot economy thesis. I wrote in twenty 629 00:31:25,000 --> 00:31:27,920 Speaker 1: twenty three, The underlying thesis of this entire show, this 630 00:31:28,040 --> 00:31:30,600 Speaker 1: growth of all cost mindset that's driven every tech company 631 00:31:30,600 --> 00:31:34,440 Speaker 1: to focus on increasingly large quarterly revenue numbers even if 632 00:31:34,440 --> 00:31:37,200 Speaker 1: the products are or they're deeply unprofitable, or in the 633 00:31:37,200 --> 00:31:40,760 Speaker 1: case of generative AI both. Nowhere has this ever been 634 00:31:40,840 --> 00:31:44,360 Speaker 1: more obvious, Never has there been a more pungent version 635 00:31:44,400 --> 00:31:46,760 Speaker 1: of the rot economy than in large language models or 636 00:31:46,800 --> 00:31:50,200 Speaker 1: more specifically, transformer based models. In their relationships with GPUs, 637 00:31:50,840 --> 00:31:53,360 Speaker 1: by making everything about growth, you inevitably reach a point 638 00:31:53,360 --> 00:31:54,840 Speaker 1: where the only thing you know how to do is 639 00:31:54,880 --> 00:31:57,560 Speaker 1: spend money. And both l LAMBS and GPUs allow big 640 00:31:57,600 --> 00:31:59,480 Speaker 1: tech to do the thing that work before building a 641 00:31:59,480 --> 00:32:01,560 Speaker 1: bunch of days centers and buying a large bunch of 642 00:32:01,640 --> 00:32:04,480 Speaker 1: chips and paying a lot of people without making sure 643 00:32:04,480 --> 00:32:06,960 Speaker 1: they've done the critical work of making sure this would 644 00:32:06,960 --> 00:32:12,280 Speaker 1: create things that were useful or profitable, and neither seems 645 00:32:12,320 --> 00:32:14,720 Speaker 1: to be the case. As a result, we're now sitting 646 00:32:14,760 --> 00:32:16,640 Speaker 1: on top of one of the most brittle situations in 647 00:32:16,720 --> 00:32:19,760 Speaker 1: economic history. Our market is held up by whether four 648 00:32:19,840 --> 00:32:22,080 Speaker 1: or five companies will continue to buy chips that start 649 00:32:22,080 --> 00:32:25,200 Speaker 1: losing the money the second they're installed. And I'm disgusted 650 00:32:25,200 --> 00:32:27,640 Speaker 1: by how many people are unwilling or unable to engage 651 00:32:27,680 --> 00:32:31,640 Speaker 1: with the truth, favoring instead a scornful, contemptuous tone toward 652 00:32:31,720 --> 00:32:34,320 Speaker 1: anybody who doesn't believe that generative AI is the future. 653 00:32:34,880 --> 00:32:37,800 Speaker 1: If you're a writer that writes about AI smarmily insulting 654 00:32:37,800 --> 00:32:41,000 Speaker 1: people who don't understand AI, you're a shitty fucking writer 655 00:32:41,160 --> 00:32:43,840 Speaker 1: because ai't is either not that good or you're not 656 00:32:43,880 --> 00:32:47,280 Speaker 1: good at explaining what it's good. Perhaps it's both, dickhead, 657 00:32:47,520 --> 00:32:49,400 Speaker 1: And if you want to know my true agenda, it's 658 00:32:49,400 --> 00:32:52,600 Speaker 1: that I see something in generative AI, and it's boosters 659 00:32:52,600 --> 00:32:57,320 Speaker 1: that I truly dislike. Large language models, authoritativity state things 660 00:32:57,360 --> 00:32:59,520 Speaker 1: that are incorrect because they have no concept of right 661 00:32:59,600 --> 00:33:02,520 Speaker 1: or wrong, well really anything. They don't have concepts of concepts. 662 00:33:02,640 --> 00:33:05,880 Speaker 1: They don't think. I believe that the writers Managers and 663 00:33:05,920 --> 00:33:08,760 Speaker 1: executives that find it exciting do so because it gives 664 00:33:08,760 --> 00:33:11,080 Speaker 1: them the ability to pretend to be intelligent without actually 665 00:33:11,160 --> 00:33:15,040 Speaker 1: learning anything, to do everything they can to avoid actual work, knowledge, 666 00:33:15,080 --> 00:33:19,280 Speaker 1: or responsibility for themselves or others. There is this overwhelming 667 00:33:19,320 --> 00:33:22,680 Speaker 1: condescension that comes from fans of generative AI, the sense 668 00:33:22,720 --> 00:33:25,200 Speaker 1: that they know something, you've done, something they double down on. 669 00:33:25,600 --> 00:33:27,600 Speaker 1: We are being forced to use it by bosses or 670 00:33:27,640 --> 00:33:29,880 Speaker 1: services we like that now insist it's parts of our 671 00:33:29,920 --> 00:33:33,120 Speaker 1: documents or our search engines, not because it does something 672 00:33:33,200 --> 00:33:35,479 Speaker 1: or helps us, but because those pushing it need us 673 00:33:35,480 --> 00:33:37,560 Speaker 1: to use it to prove that they know what's going on. 674 00:33:38,120 --> 00:33:40,479 Speaker 1: To quote my editor Matt Hughes, generative AI is an 675 00:33:40,480 --> 00:33:43,360 Speaker 1: expression of contempt towards people, one that considers them to 676 00:33:43,360 --> 00:33:47,080 Speaker 1: be a commodity at best and a rapidly depreciating asset 677 00:33:47,160 --> 00:33:49,480 Speaker 1: at worst. And Matt would have said that correctly. I 678 00:33:49,520 --> 00:33:52,840 Speaker 1: haven't quite cracked why. But GENERATAI also brings out the 679 00:33:52,880 --> 00:33:55,280 Speaker 1: worst in people. By giving the illusion of labor, it 680 00:33:55,280 --> 00:33:57,720 Speaker 1: excites those who are desperate to replace the commoditize it. 681 00:33:58,080 --> 00:34:00,600 Speaker 1: By giving the illusion of education, it excites those are 682 00:34:00,600 --> 00:34:03,280 Speaker 1: too idle to actually learn things by convincing them, like 683 00:34:03,400 --> 00:34:06,120 Speaker 1: uber CEO Travis Kalenik, then a few minutes that they 684 00:34:06,160 --> 00:34:09,480 Speaker 1: can learn quantum physics. By giving the illusion of activity, 685 00:34:09,520 --> 00:34:11,680 Speaker 1: it allows the gluttony of business idi it's that control 686 00:34:11,760 --> 00:34:14,880 Speaker 1: everything to pretend that they're doing something. By giving the 687 00:34:14,920 --> 00:34:18,160 Speaker 1: illusion of futurity, which is David Carp's excellent term, it 688 00:34:18,200 --> 00:34:21,680 Speaker 1: gives reporters that long since disconnected from actual software and 689 00:34:21,719 --> 00:34:23,840 Speaker 1: hardware the ability to pretend that they know what's happening 690 00:34:23,840 --> 00:34:27,320 Speaker 1: in the tech industry. And fundamentally, it's biggest illusion is 691 00:34:27,440 --> 00:34:31,120 Speaker 1: economic activity, because despite being questionably useful and burning billions 692 00:34:31,160 --> 00:34:33,880 Speaker 1: of dollars, its neat to do so creates justification for 693 00:34:33,920 --> 00:34:36,600 Speaker 1: spending billions of dollars on GPUs and data center sprawl, 694 00:34:36,640 --> 00:34:38,680 Speaker 1: which allows big tech to send money into something and 695 00:34:38,719 --> 00:34:43,919 Speaker 1: give the illusion of growth. I really love doing this show. 696 00:34:44,080 --> 00:34:47,040 Speaker 1: I love it so much, but I don't enjoy it 697 00:34:47,120 --> 00:34:50,600 Speaker 1: this material. I mean, I enjoy performing it, I guess, 698 00:34:50,680 --> 00:34:54,480 Speaker 1: But what I'm talking about is pretty fucking grim. And 699 00:34:54,600 --> 00:34:58,880 Speaker 1: every week I am cataloging how people in power haven't 700 00:34:59,400 --> 00:35:04,040 Speaker 1: noticed something that's inherently a scam that's really based on lies, 701 00:35:04,080 --> 00:35:06,840 Speaker 1: and how the people responsible in the media have just 702 00:35:07,280 --> 00:35:10,120 Speaker 1: kind of turned away from readers in favor of doing 703 00:35:10,160 --> 00:35:13,040 Speaker 1: what they think the market wants or confirming their own 704 00:35:13,120 --> 00:35:19,080 Speaker 1: lazy biases. It's sad, and I think I'm right, and 705 00:35:19,120 --> 00:35:23,080 Speaker 1: I'm not necessarily happy about being right. And if I'm wrong, 706 00:35:23,520 --> 00:35:26,080 Speaker 1: I get asked this a lot. If I'm wrong, I 707 00:35:26,120 --> 00:35:28,959 Speaker 1: will explain why I'm wrong in great detail. I will 708 00:35:29,000 --> 00:35:32,279 Speaker 1: not shy away from taking accountability, because I think that 709 00:35:32,280 --> 00:35:34,920 Speaker 1: that's what you should expect from me. I'm serious. If 710 00:35:34,960 --> 00:35:37,319 Speaker 1: I'm somehow wrong here, you will hear why. I'll do 711 00:35:37,520 --> 00:35:41,400 Speaker 1: entire episodes about it. I'm serious, but I really do 712 00:35:41,440 --> 00:35:43,960 Speaker 1: not think I am, and that's why I sound kind 713 00:35:43,960 --> 00:35:47,280 Speaker 1: of downcast and alarmed what I'm describing as a bubble. 714 00:35:47,280 --> 00:35:50,239 Speaker 1: I'm one with an obvious weakness. One company's ability to 715 00:35:50,239 --> 00:35:52,279 Speaker 1: sell hardware to four or five other companies all to 716 00:35:52,360 --> 00:35:55,400 Speaker 1: run services that lose billions of dollars. At some point, 717 00:35:55,400 --> 00:35:58,040 Speaker 1: the momentum behind the video slows. Maybe it won't even 718 00:35:58,040 --> 00:35:59,800 Speaker 1: be sales slowing. Maybe it would just be the suggestion 719 00:35:59,880 --> 00:36:01,719 Speaker 1: that one of the largest companies won't be buying as 720 00:36:01,719 --> 00:36:05,640 Speaker 1: many GPUs. Perception matters just as much as actual numbers 721 00:36:05,840 --> 00:36:08,120 Speaker 1: and sometimes more, and a shifting sentiment could start a 722 00:36:08,200 --> 00:36:10,960 Speaker 1: chain of events that knocks down the entire house of cards. 723 00:36:11,360 --> 00:36:14,560 Speaker 1: Check may. I don't know when, I don't know how, 724 00:36:14,680 --> 00:36:18,080 Speaker 1: but I really really don't know how I'm wrong. And 725 00:36:18,160 --> 00:36:21,240 Speaker 1: I hate that so many people will see their retirements wrecked, 726 00:36:21,239 --> 00:36:24,040 Speaker 1: and that so many people intentionally or accidentally helped steer 727 00:36:24,040 --> 00:36:27,799 Speaker 1: the economy in this reckless, needless, and wasteful direction, all 728 00:36:27,800 --> 00:36:29,879 Speaker 1: because big tech didn't have a new way to show 729 00:36:29,960 --> 00:36:32,680 Speaker 1: quarterly growth. And that there are so many people out 730 00:36:32,719 --> 00:36:36,960 Speaker 1: there that love the idea of replacing workers and using 731 00:36:37,000 --> 00:36:40,160 Speaker 1: the computer to do work for them. And I hate 732 00:36:40,160 --> 00:36:42,480 Speaker 1: that so many people have lost their jobs because companies 733 00:36:42,520 --> 00:36:44,799 Speaker 1: are spending the equivalent of the entire GDP of some 734 00:36:44,920 --> 00:36:48,280 Speaker 1: European country on data centers and GPUs that won't actually 735 00:36:48,280 --> 00:36:51,360 Speaker 1: deliver any value. Sach and Adella is a fucking monster. 736 00:36:51,680 --> 00:36:53,920 Speaker 1: Fifteen thousand people laid off this year so you can 737 00:36:53,960 --> 00:36:58,000 Speaker 1: do more, fucking clippy, you ass wipe. If sach an 738 00:36:58,000 --> 00:37:01,120 Speaker 1: Adella somehow it hears this up, be so happy. But 739 00:37:01,200 --> 00:37:03,160 Speaker 1: my purpose here is to explain to you, no matter 740 00:37:03,160 --> 00:37:05,360 Speaker 1: your background or interest or creed or whatever way you 741 00:37:05,480 --> 00:37:08,759 Speaker 1: find my work why all of this is happening. As 742 00:37:08,800 --> 00:37:10,719 Speaker 1: you watch this collapse, I want you to tell your 743 00:37:10,719 --> 00:37:14,160 Speaker 1: friends about why, the people responsible and decisions they made, 744 00:37:14,160 --> 00:37:17,319 Speaker 1: and make sure it's clear that there are people responsible. 745 00:37:18,120 --> 00:37:22,600 Speaker 1: Sam Altman, Dario Ama Day, Sachin Adela, Sundar Pieshit, Tim Cook, 746 00:37:22,719 --> 00:37:27,040 Speaker 1: Elon Musk, Mark Zuckerberg, and Andy Jasse have overseen a needless, wasteful, 747 00:37:27,080 --> 00:37:30,600 Speaker 1: and destructive economic force that will harmer economy, our markets, 748 00:37:30,600 --> 00:37:32,879 Speaker 1: and the tech industry writ large. And when this is over, 749 00:37:33,160 --> 00:37:37,640 Speaker 1: they must be held accountable. And remember that you, as 750 00:37:37,680 --> 00:37:40,960 Speaker 1: a regular person, can understand all of this. These people 751 00:37:41,000 --> 00:37:42,840 Speaker 1: want you to believe that this is black magic, that 752 00:37:42,880 --> 00:37:45,240 Speaker 1: you're wrong to worry about the billions wasted or question 753 00:37:45,320 --> 00:37:47,879 Speaker 1: the usefulness of the tools. They want you to sit 754 00:37:47,920 --> 00:37:50,080 Speaker 1: there and say, tech God away from me. I just 755 00:37:50,080 --> 00:37:53,560 Speaker 1: don't understand how the computer works anymore. How about that 756 00:37:53,560 --> 00:37:57,279 Speaker 1: the computer doesn't work anymore, that the computer regularly does 757 00:37:57,320 --> 00:37:59,760 Speaker 1: not do the things you want it to. And Generative 758 00:37:59,760 --> 00:38:03,240 Speaker 1: AI just the most pornographic example of a tech industry 759 00:38:03,239 --> 00:38:07,560 Speaker 1: that gave up on user interfaces, user experience, and users altogether. 760 00:38:08,320 --> 00:38:10,680 Speaker 1: You were smarter than they reckon and stronger than they know. 761 00:38:10,760 --> 00:38:12,759 Speaker 1: In A better future is one where you recognize this 762 00:38:12,880 --> 00:38:15,400 Speaker 1: and realize that power and money doesn't make a man righteous, 763 00:38:15,440 --> 00:38:18,480 Speaker 1: right or smart. These people sicken me, and if I'm 764 00:38:18,480 --> 00:38:20,719 Speaker 1: going to be honest, I think a lot of this 765 00:38:20,880 --> 00:38:23,319 Speaker 1: work comes down to them reminding me of people I 766 00:38:23,360 --> 00:38:27,719 Speaker 1: grew up with, people I've met throughout my life. Speecious 767 00:38:27,800 --> 00:38:31,000 Speaker 1: managers that have got there by glad handling, not through 768 00:38:31,080 --> 00:38:34,040 Speaker 1: real work or producing anything of value, but by stealing 769 00:38:34,320 --> 00:38:37,319 Speaker 1: or tricking, by throwing other people under the bus, and 770 00:38:37,360 --> 00:38:39,960 Speaker 1: in some cases, and I'm thinking of one specific fucking 771 00:38:40,040 --> 00:38:43,799 Speaker 1: person driving it themselves. These people are the ones the 772 00:38:43,800 --> 00:38:47,239 Speaker 1: most excited about AI. They're the ones that believe that 773 00:38:47,320 --> 00:38:50,400 Speaker 1: this technology will finally give them the ability to have 774 00:38:50,480 --> 00:38:53,680 Speaker 1: the abilities of those that work without actually having to 775 00:38:53,680 --> 00:38:57,600 Speaker 1: get off their fucking asses. And they sicken me. They 776 00:38:57,680 --> 00:39:00,879 Speaker 1: disgust me, because there is a different way of looking 777 00:39:00,920 --> 00:39:03,520 Speaker 1: at how generative AI could have been. That it enhances humans, 778 00:39:03,520 --> 00:39:05,919 Speaker 1: that it makes them able to do more things rather 779 00:39:05,960 --> 00:39:08,799 Speaker 1: than replacing them. That likely would have led us in 780 00:39:08,840 --> 00:39:10,759 Speaker 1: a different direction and probably wouldn't have led to the 781 00:39:10,760 --> 00:39:13,440 Speaker 1: billions of dollars of investment in this because when you 782 00:39:13,480 --> 00:39:18,160 Speaker 1: think about helping people. Suddenly the money dries up. I 783 00:39:18,200 --> 00:39:21,080 Speaker 1: sound cynical, but I've said this before. I'm a broken 784 00:39:21,120 --> 00:39:24,280 Speaker 1: hearted romantic. I hate what the tech industry has become. 785 00:39:25,320 --> 00:39:27,840 Speaker 1: But people also said that I can do this. Fucking 786 00:39:27,840 --> 00:39:29,719 Speaker 1: four years ago I had five years ago, had three 787 00:39:29,800 --> 00:39:32,960 Speaker 1: hundred subscribers on my newsletter on about sixty seven thousands 788 00:39:33,000 --> 00:39:35,200 Speaker 1: of people, an indeterminate amount of you. I can't say 789 00:39:35,239 --> 00:39:39,480 Speaker 1: it due to policy. Listen to this show, and that's 790 00:39:39,520 --> 00:39:42,719 Speaker 1: proof enough. I think that one idiot, worked up white 791 00:39:42,760 --> 00:39:46,200 Speaker 1: boy from London that the school teacher once said should 792 00:39:46,239 --> 00:39:50,440 Speaker 1: get used to flipping burgers can do something. I'm nobody. 793 00:39:50,560 --> 00:39:53,080 Speaker 1: I'm just a guy, but I've spent the time to 794 00:39:53,080 --> 00:39:55,520 Speaker 1: read this stuff and to divine meaning from this stuff, 795 00:39:55,719 --> 00:39:58,520 Speaker 1: just like you can. I hope I've helped you. That's 796 00:39:58,520 --> 00:40:00,520 Speaker 1: what I'm here to do. I love doing this, and 797 00:40:00,560 --> 00:40:02,319 Speaker 1: I feel ever grateful that I'm able to every day, 798 00:40:02,320 --> 00:40:04,440 Speaker 1: and I love all of you for listening. It takes 799 00:40:04,480 --> 00:40:05,880 Speaker 1: a lot of ever to put this out, but it 800 00:40:05,920 --> 00:40:09,279 Speaker 1: also is a choice of yours. How much time you 801 00:40:09,320 --> 00:40:11,440 Speaker 1: can you have in a day is not your choice, 802 00:40:11,480 --> 00:40:13,840 Speaker 1: but how you used it is. And I deeply appreciate 803 00:40:13,880 --> 00:40:16,640 Speaker 1: you've forgiving me your time. Thank you so much. This 804 00:40:16,760 --> 00:40:28,040 Speaker 1: has been the hater's guide to the AI bubble. Thank 805 00:40:28,040 --> 00:40:29,560 Speaker 1: you for listening to Better Offline. 806 00:40:29,640 --> 00:40:32,080 Speaker 2: The editor and composer of the Better Offline theme song 807 00:40:32,160 --> 00:40:34,799 Speaker 2: is Matasowski. You can check out more of his music 808 00:40:34,840 --> 00:40:38,479 Speaker 2: and audio projects at Matasowski dot com, M A T 809 00:40:38,480 --> 00:40:42,960 Speaker 2: T O S O W s ki dot com. You 810 00:40:42,960 --> 00:40:45,480 Speaker 2: can email me at easy at Better Offline dot com 811 00:40:45,560 --> 00:40:47,880 Speaker 2: or visit Better Offline dot com to find more podcast 812 00:40:47,920 --> 00:40:51,240 Speaker 2: links and of course, my newsletter. I also really recommend 813 00:40:51,280 --> 00:40:53,200 Speaker 2: you go to chat dot where's youreaed dot at to 814 00:40:53,320 --> 00:40:55,680 Speaker 2: visit the discord, and go to our slash. 815 00:40:55,360 --> 00:40:57,200 Speaker 1: Better Offline to check out our reddit. 816 00:40:58,000 --> 00:40:59,360 Speaker 2: Thank you so much for listening. 817 00:41:00,200 --> 00:41:02,680 Speaker 1: Better Offline is a production of cool Zone Media. 818 00:41:02,800 --> 00:41:06,200 Speaker 2: For more from cool Zone Media, visit our website Coolzonmedia 819 00:41:06,239 --> 00:41:09,040 Speaker 2: dot com, or check us out on the iHeartRadio app, 820 00:41:09,120 --> 00:41:11,560 Speaker 2: Apple Podcasts, or wherever you get your podcasts.