1 00:00:02,440 --> 00:00:06,280 Speaker 1: Zone Media, Hello, and welcome to Better Offline. I am, 2 00:00:06,360 --> 00:00:20,360 Speaker 1: of course your host ed zitron what and after a 3 00:00:20,360 --> 00:00:22,600 Speaker 1: few three part episodes, I had an idea, what if 4 00:00:22,640 --> 00:00:25,200 Speaker 1: I did a four parter? In all seriousness, I know 5 00:00:25,239 --> 00:00:27,320 Speaker 1: that this is a little bit long, but the topic 6 00:00:27,360 --> 00:00:29,560 Speaker 1: we're about to explore demand's quite a bit of depth 7 00:00:29,600 --> 00:00:31,600 Speaker 1: and it isn't something I could really do justice to 8 00:00:31,760 --> 00:00:34,240 Speaker 1: in a one part or two parts, or I guess 9 00:00:34,280 --> 00:00:37,760 Speaker 1: even three parter, but let's get into him. Over the 10 00:00:37,840 --> 00:00:40,760 Speaker 1: last few months, we felt the vibes shift downward in 11 00:00:40,800 --> 00:00:43,720 Speaker 1: an aggressive way, with both Marg Zuckerberg and Clammy Sam 12 00:00:43,760 --> 00:00:46,480 Speaker 1: Mortman saying that we're in a bubble. In the latter 13 00:00:46,560 --> 00:00:48,960 Speaker 1: case said, warnings of a bubble are always couched in 14 00:00:49,000 --> 00:00:52,159 Speaker 1: rank hypocrisy is it's always implied that whoever it is 15 00:00:52,240 --> 00:00:54,880 Speaker 1: and the companies they represent aren't part of that bubble, 16 00:00:54,920 --> 00:00:58,600 Speaker 1: but rather it's other people and other companies making unfortunate decisions. 17 00:00:59,080 --> 00:01:02,160 Speaker 1: The thing is, there's really no escape for either of 18 00:01:02,200 --> 00:01:05,360 Speaker 1: these guys, not for Zaken, definitely not for Sam Ortmon. 19 00:01:05,760 --> 00:01:07,760 Speaker 1: And over the next four episodes, I'm going to make 20 00:01:07,760 --> 00:01:09,760 Speaker 1: a comprehensive case for the fact that we're in a 21 00:01:09,800 --> 00:01:13,160 Speaker 1: bubble and condense everything I've been talking about into one series. 22 00:01:13,600 --> 00:01:15,280 Speaker 1: And I know I've been all over the place, and 23 00:01:15,319 --> 00:01:17,080 Speaker 1: I get a lot of people saying, oh, well, where 24 00:01:17,080 --> 00:01:18,959 Speaker 1: did you talk about this? And where'd you talk about that? 25 00:01:19,520 --> 00:01:21,520 Speaker 1: And that's kind of fair when you put out as 26 00:01:21,560 --> 00:01:23,800 Speaker 1: much as I do. But I'm going to break this 27 00:01:23,880 --> 00:01:25,640 Speaker 1: down in four episodes. I'm going to give you a 28 00:01:25,720 --> 00:01:29,319 Speaker 1: comprehensive argument against the bubble. Well I mean that for 29 00:01:29,440 --> 00:01:31,679 Speaker 1: a bubble, I guess, but against generative AI in general. 30 00:01:31,720 --> 00:01:34,040 Speaker 1: But in this episode, I think it's good to start 31 00:01:34,080 --> 00:01:36,520 Speaker 1: from the beginning and work our way forward to track 32 00:01:36,560 --> 00:01:38,920 Speaker 1: the thread from the origins of chat GPT to the 33 00:01:38,920 --> 00:01:41,880 Speaker 1: billion spurned building data centers all over the world and 34 00:01:41,920 --> 00:01:44,880 Speaker 1: the weak business justifications for burning and nearly a trillion 35 00:01:44,920 --> 00:01:48,360 Speaker 1: dollars to keep this hollow industry alive. Now, in twenty 36 00:01:48,400 --> 00:01:51,400 Speaker 1: twenty two, a kind of company called open Ai surprise 37 00:01:51,440 --> 00:01:53,880 Speaker 1: the world with a website called chat GPT. They could 38 00:01:53,920 --> 00:01:56,800 Speaker 1: generate text that sort of sounded like a person using 39 00:01:56,800 --> 00:02:00,360 Speaker 1: a technology called large language models LLMS, which can also 40 00:02:00,400 --> 00:02:02,760 Speaker 1: be used to generate images, video, and computer code, or 41 00:02:02,760 --> 00:02:06,680 Speaker 1: at least would eventually last. Language models require entire clusters 42 00:02:06,680 --> 00:02:09,880 Speaker 1: of servers connected with high speed networking or containing this 43 00:02:09,919 --> 00:02:13,880 Speaker 1: thing called a GPU. Graphics processing units. These are difference 44 00:02:13,919 --> 00:02:16,640 Speaker 1: to the GPUs in your xbox or laptop or gaming PC. 45 00:02:17,000 --> 00:02:19,040 Speaker 1: They cost much much more, and they're good at doing 46 00:02:19,120 --> 00:02:21,600 Speaker 1: the processes of inference, the creation of an output of 47 00:02:21,639 --> 00:02:24,720 Speaker 1: any LLM and training, feeding masses of training data to 48 00:02:24,760 --> 00:02:27,000 Speaker 1: the models, or feeding them information about what a good 49 00:02:27,080 --> 00:02:29,320 Speaker 1: output might look like so they can later identify a 50 00:02:29,360 --> 00:02:33,040 Speaker 1: thing or replicate it. These models showed some immediate promise 51 00:02:33,080 --> 00:02:36,280 Speaker 1: in their ability to articulate concepts or generate video visuals, 52 00:02:36,280 --> 00:02:39,800 Speaker 1: audio text, and code. They also immediately had one glaring 53 00:02:39,880 --> 00:02:44,040 Speaker 1: obvious problem because their propabilistic, meaning that they're just guessing 54 00:02:44,080 --> 00:02:46,920 Speaker 1: whatever the right output might be. These models can't actually 55 00:02:46,919 --> 00:02:49,919 Speaker 1: be relied upon to do exactly the same thing every 56 00:02:49,960 --> 00:02:53,200 Speaker 1: single time. So if you generated a picture of a 57 00:02:53,200 --> 00:02:56,400 Speaker 1: person that you wanted to, for example, using this story book, 58 00:02:56,600 --> 00:02:58,720 Speaker 1: every time you created a new page using the same 59 00:02:58,760 --> 00:03:01,320 Speaker 1: prompt to describe the protragton, the person would look different, 60 00:03:01,560 --> 00:03:03,600 Speaker 1: and that difference could be minor of something that a 61 00:03:03,639 --> 00:03:05,640 Speaker 1: reader could shrug off, or it could make the character 62 00:03:05,680 --> 00:03:08,440 Speaker 1: look like a completely different person. Now None of this, 63 00:03:08,560 --> 00:03:11,520 Speaker 1: by the way, is me validating or saying that any 64 00:03:11,560 --> 00:03:14,880 Speaker 1: of this stuff is good. I'm just describing him. Moreover, 65 00:03:14,960 --> 00:03:17,760 Speaker 1: the probabilistic nature of generative AI meant that whenever you 66 00:03:17,800 --> 00:03:20,000 Speaker 1: asked it a question, it would guess as to the answer, 67 00:03:20,040 --> 00:03:21,920 Speaker 1: not because it knew the answer, but rather because it 68 00:03:21,960 --> 00:03:23,480 Speaker 1: was guessing on the right word to add in a 69 00:03:23,560 --> 00:03:26,920 Speaker 1: sentence based on previous training data. As a result, these 70 00:03:26,960 --> 00:03:30,200 Speaker 1: models would frequently make mistakes, something which we later referred 71 00:03:30,240 --> 00:03:33,440 Speaker 1: to as hallucinations. And that's not even mentioning the cost 72 00:03:33,480 --> 00:03:35,520 Speaker 1: of training these models, the cost of running them, the 73 00:03:35,600 --> 00:03:38,720 Speaker 1: vast amounts of computational power they required, the fact that 74 00:03:38,760 --> 00:03:41,080 Speaker 1: the legality of using materials straight from books in the 75 00:03:41,080 --> 00:03:44,400 Speaker 1: web without the owner's permission was and remains legally dubious, 76 00:03:44,640 --> 00:03:46,240 Speaker 1: or the fact that nobody seemed to know how to 77 00:03:46,320 --> 00:03:50,400 Speaker 1: use these models that actually create profitable businesses. These problems 78 00:03:50,400 --> 00:03:53,160 Speaker 1: were overshadowed by something flashy and new, and something that 79 00:03:53,200 --> 00:03:56,120 Speaker 1: investors and the tech media believed would eventually aut to 80 00:03:56,160 --> 00:03:59,800 Speaker 1: make jobs that have proven most resistance towardstion knowledge work 81 00:03:59,840 --> 00:04:03,320 Speaker 1: and the creative economy. The newness and hyper and these 82 00:04:03,400 --> 00:04:06,040 Speaker 1: expectations sent the market into a frenzy, with every hyper 83 00:04:06,040 --> 00:04:09,560 Speaker 1: scaler immediately creating the most aggressive market for one supplier 84 00:04:09,600 --> 00:04:12,840 Speaker 1: I've ever seen. Nvidia has sold over two hundred billion 85 00:04:12,840 --> 00:04:15,240 Speaker 1: dollars of GPUs since the beginning of twenty twenty three, 86 00:04:15,480 --> 00:04:18,440 Speaker 1: becoming the largest company on the American stock market and 87 00:04:18,480 --> 00:04:20,920 Speaker 1: trading over one hundred and seventy dollars as of writing 88 00:04:20,920 --> 00:04:23,119 Speaker 1: this sentence, only a few years off to being worth 89 00:04:23,440 --> 00:04:26,560 Speaker 1: nineteen dollars and fifty two cents a share. Now there's 90 00:04:26,560 --> 00:04:29,039 Speaker 1: a stock split that happened there, but it works out 91 00:04:29,040 --> 00:04:31,400 Speaker 1: that way. Now. While I've talked about some of the 92 00:04:31,440 --> 00:04:35,120 Speaker 1: propelling factors behind the AI wave, automation and novelty, that's 93 00:04:35,160 --> 00:04:38,320 Speaker 1: not really the complete picture. A huge reason why everybody 94 00:04:38,360 --> 00:04:41,719 Speaker 1: decided to do AI was because the software industry's growth 95 00:04:41,800 --> 00:04:45,240 Speaker 1: was slowing and SaaS software as a service company valuations 96 00:04:45,720 --> 00:04:48,880 Speaker 1: were stalling or dropping, resulting in the terrifying prospect of 97 00:04:48,880 --> 00:04:52,240 Speaker 1: companies having to underpromise and over deliver and be efficient, 98 00:04:52,320 --> 00:04:56,680 Speaker 1: you know, gross things like running sustainable businesses. Things that 99 00:04:56,839 --> 00:04:59,919 Speaker 1: normal companies, those whose valuations aren't contingent on ever increase, 100 00:05:00,120 --> 00:05:02,360 Speaker 1: ever constant growth, don't have to worry about because they're 101 00:05:02,400 --> 00:05:06,760 Speaker 1: normal companies. Suddenly there was a new promise of new technology, 102 00:05:06,880 --> 00:05:10,279 Speaker 1: large language models that were getting exponentially more powerful, which 103 00:05:10,800 --> 00:05:13,320 Speaker 1: was mostly a lie but hard to disprove because powerful 104 00:05:13,360 --> 00:05:16,719 Speaker 1: can mean basically anything, and the definition of powerful depended 105 00:05:16,880 --> 00:05:19,400 Speaker 1: entirely on whoever you asked at any given time and 106 00:05:19,400 --> 00:05:23,560 Speaker 1: what that person's motivations were. The media also immediately started 107 00:05:23,560 --> 00:05:26,719 Speaker 1: tripping over its own feet, mistakenly claiming open ais GPT 108 00:05:26,880 --> 00:05:29,320 Speaker 1: form model tricked a task grab it into solving a capture. 109 00:05:29,480 --> 00:05:32,120 Speaker 1: It didn't. This never happened, or saying that, and I 110 00:05:32,240 --> 00:05:34,440 Speaker 1: quote people who don't know how to code already used 111 00:05:34,440 --> 00:05:37,239 Speaker 1: bots to produce full fledged games. And if you weren't 112 00:05:37,320 --> 00:05:39,640 Speaker 1: wondering what the New York Times was referring to when 113 00:05:39,640 --> 00:05:42,480 Speaker 1: they said full fledged, there it meant Pong and are 114 00:05:42,520 --> 00:05:45,719 Speaker 1: coupled together rolling demo of Skyroads game from nineteen ninety three, 115 00:05:46,160 --> 00:05:48,440 Speaker 1: likely because a bunch of that training data was fed 116 00:05:48,440 --> 00:05:51,760 Speaker 1: into the models. Now, the media and investors helped pedal 117 00:05:51,800 --> 00:05:54,520 Speaker 1: the narrative that AI was always getting better, could basically 118 00:05:54,560 --> 00:05:56,880 Speaker 1: do anything, and that any problems you saw today would 119 00:05:56,880 --> 00:06:00,000 Speaker 1: inevitably be solved in a few short months or years. 120 00:06:00,200 --> 00:06:03,720 Speaker 1: Or that some point. I guess not really sure when 121 00:06:03,760 --> 00:06:06,440 Speaker 1: that point is, but damn do they think it's coming. 122 00:06:06,960 --> 00:06:09,600 Speaker 1: And llms were touted as a kind of digital panacea, 123 00:06:09,680 --> 00:06:12,400 Speaker 1: and the companies building them off a traditional software companies 124 00:06:12,400 --> 00:06:15,000 Speaker 1: the chance to plug these models into their software using 125 00:06:15,000 --> 00:06:18,040 Speaker 1: an API, thus allowing them to write the same generative 126 00:06:18,040 --> 00:06:21,440 Speaker 1: AI wave that every other company was riding. The model 127 00:06:21,480 --> 00:06:25,040 Speaker 1: companies similarly started going after individual and business customers, offering 128 00:06:25,040 --> 00:06:28,080 Speaker 1: software and subscriptions that promised the world, though this mostly 129 00:06:28,120 --> 00:06:30,760 Speaker 1: boiled down to chatbots that could generate stuff, and then 130 00:06:30,960 --> 00:06:33,880 Speaker 1: doubled down with the promise of agents, a marketing term 131 00:06:33,880 --> 00:06:36,719 Speaker 1: that's meant to make you think autonomous digital worker, but 132 00:06:36,920 --> 00:06:40,159 Speaker 1: really means broken digital chatbot of some sort or just 133 00:06:40,240 --> 00:06:43,000 Speaker 1: broken digital product. It really depends how you're feeling that day. 134 00:06:44,520 --> 00:06:47,000 Speaker 1: Throughout this era, investors in the media spoke with a 135 00:06:47,040 --> 00:06:50,279 Speaker 1: sense of inevitability that they never really backed up with data. 136 00:06:50,839 --> 00:06:54,320 Speaker 1: It was an era based on confidently asserted vibes. Everything 137 00:06:54,440 --> 00:06:57,360 Speaker 1: was always getting better and more powerful, even though there 138 00:06:57,360 --> 00:07:00,840 Speaker 1: was never much proof that this was truly disruptive technology 139 00:07:01,120 --> 00:07:03,400 Speaker 1: other than in its ability to disrupt apps you were 140 00:07:03,480 --> 00:07:07,080 Speaker 1: using with AI, making them worse, For example, suggesting questions 141 00:07:07,080 --> 00:07:09,400 Speaker 1: on every Facebook post that you could ask meta AI, 142 00:07:09,640 --> 00:07:13,240 Speaker 1: but which METAAI couldn't answer. And I mean on memes, 143 00:07:13,760 --> 00:07:17,600 Speaker 1: on just random posts. It's really not useful in any way, 144 00:07:17,600 --> 00:07:21,440 Speaker 1: shape or form. AI became omnipresent, and it eventually grew 145 00:07:21,480 --> 00:07:24,040 Speaker 1: to mean everything and nothing. Open AI would see. It's 146 00:07:24,040 --> 00:07:26,880 Speaker 1: every move loaded over like a gifted child. It's CEO 147 00:07:26,960 --> 00:07:29,480 Speaker 1: Sam Allman called the Oppenheimer of our age, even if 148 00:07:29,480 --> 00:07:32,840 Speaker 1: it wasn't really obvious why everybody was impressed. GPT four 149 00:07:32,880 --> 00:07:36,000 Speaker 1: felt like something a bit different, but was it actually meaningful? 150 00:07:36,400 --> 00:07:39,440 Speaker 1: The thing is our official intelligence is built and sold 151 00:07:39,520 --> 00:07:41,600 Speaker 1: or not just faith, but a series of myths that 152 00:07:41,640 --> 00:07:44,240 Speaker 1: AI boosters expect us to believe, but the same certainty 153 00:07:44,280 --> 00:07:46,920 Speaker 1: that we treat things like gravity or the boiling point 154 00:07:46,920 --> 00:07:51,200 Speaker 1: of water. Can large language models actually replace coders? Not really, No, 155 00:07:51,320 --> 00:08:05,160 Speaker 1: and I'll get into way later in this series. Consora 156 00:08:05,320 --> 00:08:08,640 Speaker 1: Open AI's video creation tool replace actors or animators. No, 157 00:08:08,840 --> 00:08:10,560 Speaker 1: not at all, But it still fills the air full 158 00:08:10,600 --> 00:08:13,440 Speaker 1: of tension because you can immediately see who is preregistered 159 00:08:13,440 --> 00:08:17,120 Speaker 1: to replace everyone that works for them. AI is apparently 160 00:08:17,160 --> 00:08:20,880 Speaker 1: replacing workers, but nobody seems able to prove it at scale. 161 00:08:21,120 --> 00:08:23,680 Speaker 1: But every few weeks a story runs where everybody tries 162 00:08:23,720 --> 00:08:25,960 Speaker 1: to pretend that AI is replacing workers with some sort 163 00:08:26,000 --> 00:08:30,440 Speaker 1: of poorly sourced and incomprehensible study, never actually saying somebody's 164 00:08:30,520 --> 00:08:33,520 Speaker 1: job got replaced by AI, because it isn't happening at scale, 165 00:08:33,640 --> 00:08:36,080 Speaker 1: and because if you provide real world examples, people can 166 00:08:36,120 --> 00:08:38,959 Speaker 1: actually check if they're true. Now, I want to be clear, 167 00:08:39,360 --> 00:08:42,520 Speaker 1: some people have lost jobs to AI, just not really 168 00:08:42,520 --> 00:08:45,280 Speaker 1: white collar workers, software engineers, or really any of the 169 00:08:45,320 --> 00:08:48,000 Speaker 1: career paths that the mainstream media and AI investors would 170 00:08:48,040 --> 00:08:51,520 Speaker 1: have you believe. Brian Merchant has done excellent work covering 171 00:08:51,520 --> 00:08:54,640 Speaker 1: how llms have devoured the work of translators, using cheap, 172 00:08:54,760 --> 00:08:57,720 Speaker 1: almost good automation to lower already stagnant wages in a 173 00:08:57,720 --> 00:09:00,280 Speaker 1: field that has already been hurting before the add to 174 00:09:00,440 --> 00:09:03,160 Speaker 1: generative AI, with some having to abandon the field and 175 00:09:03,200 --> 00:09:06,560 Speaker 1: others pushed into bankruptcy. I've heard the same for art directors, 176 00:09:06,720 --> 00:09:09,400 Speaker 1: SEO experts, and copy editors, and Christopher Mims of The 177 00:09:09,400 --> 00:09:12,920 Speaker 1: Wall Street Journal covered these last year. These fields all 178 00:09:12,920 --> 00:09:15,480 Speaker 1: have something in common shitty bosses with little regard for 179 00:09:15,520 --> 00:09:18,080 Speaker 1: their customers. Who have been eagerly waiting for the opportunity 180 00:09:18,080 --> 00:09:21,720 Speaker 1: to slash labor to quote Merchant, the drum beat marketing 181 00:09:21,760 --> 00:09:24,720 Speaker 1: and pop culture of powerful AI encourages and permits management 182 00:09:24,760 --> 00:09:27,040 Speaker 1: to replace with the great jobs they might not otherwise have. 183 00:09:27,679 --> 00:09:30,440 Speaker 1: Across the board, the people being replaced by AI are 184 00:09:30,440 --> 00:09:33,160 Speaker 1: the victims of lazy, incompetent cost cutters who don't care 185 00:09:33,160 --> 00:09:36,520 Speaker 1: if they ship poorly translated text to quote Merchant. Again, 186 00:09:36,920 --> 00:09:40,079 Speaker 1: AI hypers created the cover necessary to justify slashing rates 187 00:09:40,080 --> 00:09:43,800 Speaker 1: and accepting just good enough automation output for video games 188 00:09:43,800 --> 00:09:47,560 Speaker 1: and media products. Yet the jobs crisis facing translators speaks 189 00:09:47,559 --> 00:09:49,760 Speaker 1: to the larger flaws of the large language model era 190 00:09:49,960 --> 00:09:52,240 Speaker 1: and why other careers aren't seeing this kind of disruption. 191 00:09:53,200 --> 00:09:57,000 Speaker 1: Generative AI creates outputs, and by extension, defines all labour 192 00:09:57,040 --> 00:09:59,640 Speaker 1: as some kind of output creative from a request. In 193 00:09:59,679 --> 00:10:02,040 Speaker 1: the case translation, it's possible for a company to get 194 00:10:02,040 --> 00:10:04,600 Speaker 1: by with a shitty version because many customers see translation 195 00:10:04,679 --> 00:10:07,280 Speaker 1: as what do these words say, even though, as one 196 00:10:07,320 --> 00:10:12,040 Speaker 1: worker told Brian Merchant, translation is about conveying meaning. Nevertheless, 197 00:10:12,120 --> 00:10:14,600 Speaker 1: translation workers already started to condense to a world where 198 00:10:14,679 --> 00:10:17,320 Speaker 1: humans would at times clean up machine generated text, and 199 00:10:17,360 --> 00:10:19,599 Speaker 1: the same worker warned that the same might come for 200 00:10:19,720 --> 00:10:23,000 Speaker 1: other industries. Yet the problem is that translation is a 201 00:10:23,040 --> 00:10:27,120 Speaker 1: heavily output driven industry, one where idiot bosses can say, oh, yeah, 202 00:10:27,160 --> 00:10:29,520 Speaker 1: that's fine because they ran an output back through Google 203 00:10:29,600 --> 00:10:32,480 Speaker 1: Translate and it seemed fine in their native tongue. The 204 00:10:32,520 --> 00:10:35,360 Speaker 1: problems of a poor translation are obvious, but customers of 205 00:10:35,400 --> 00:10:38,320 Speaker 1: translation are it seems, often capable of getting by with 206 00:10:38,360 --> 00:10:41,120 Speaker 1: a shitty product. The problem is that most jobs are 207 00:10:41,160 --> 00:10:43,400 Speaker 1: not output driven at all, and what we're buying from 208 00:10:43,400 --> 00:10:45,920 Speaker 1: a human beings a person's ability to think and do. 209 00:10:47,000 --> 00:10:50,640 Speaker 1: Every CEO talking about replacing workers with AI is an 210 00:10:50,640 --> 00:10:53,319 Speaker 1: example of the real problem that most companies are run 211 00:10:53,360 --> 00:10:56,360 Speaker 1: by people who don't understand or experience the problems they're solving, 212 00:10:56,640 --> 00:10:59,280 Speaker 1: don't do any real work, don't face any real problems, 213 00:10:59,280 --> 00:11:01,760 Speaker 1: and thus can never be trusted to solve them. In 214 00:11:01,800 --> 00:11:04,040 Speaker 1: the Era of the Business Idio, which is a piece 215 00:11:04,080 --> 00:11:05,880 Speaker 1: I wrote a while ago, I talked about how this 216 00:11:06,000 --> 00:11:09,040 Speaker 1: was the result of letting management consultants and neoliberal free 217 00:11:09,080 --> 00:11:12,600 Speaker 1: market sociopaths take over everything, leaving us with companies run 218 00:11:12,640 --> 00:11:14,600 Speaker 1: by people who don't know how the companies make money, 219 00:11:14,640 --> 00:11:17,800 Speaker 1: just that they must always make more without fail and 220 00:11:17,840 --> 00:11:20,240 Speaker 1: when you're a big stupid asshole. Every job that you 221 00:11:20,240 --> 00:11:22,319 Speaker 1: see is condensed to its outputs, and not the stuff 222 00:11:22,320 --> 00:11:24,920 Speaker 1: that leads up to the output, or the small nuances 223 00:11:24,960 --> 00:11:27,360 Speaker 1: and conscious decisions that make an output good as opposed 224 00:11:27,360 --> 00:11:30,960 Speaker 1: to simply acceptable or even bad. What does the software 225 00:11:30,960 --> 00:11:33,800 Speaker 1: engineer do? They write code right, What does a rater do? 226 00:11:33,880 --> 00:11:36,560 Speaker 1: They write words right? What does a hairdresser do they 227 00:11:36,600 --> 00:11:39,839 Speaker 1: can't hear? Yeah, that's of course not actually the case. 228 00:11:40,280 --> 00:11:42,200 Speaker 1: As I'll get into later in the series, the software 229 00:11:42,240 --> 00:11:44,400 Speaker 1: engineer does far more than just code. And when they 230 00:11:44,480 --> 00:11:46,600 Speaker 1: write code, they're not just saying what would solve this 231 00:11:46,640 --> 00:11:48,960 Speaker 1: problem with a big smile on their face. They're taking 232 00:11:49,000 --> 00:11:51,800 Speaker 1: into account their years of experience, what code does, what 233 00:11:51,880 --> 00:11:54,360 Speaker 1: code could do, and all the things that my break 234 00:11:54,400 --> 00:11:55,920 Speaker 1: is a result, and all of the things that you 235 00:11:55,960 --> 00:11:57,960 Speaker 1: can't really tell from just looking at the code, like 236 00:11:58,080 --> 00:12:00,520 Speaker 1: whether there's a reason things are made in a particle way, 237 00:12:01,280 --> 00:12:03,400 Speaker 1: And a good coder doesn't just hammer at the keyboard 238 00:12:03,400 --> 00:12:05,720 Speaker 1: with the aim of doing a particular task. They factor 239 00:12:05,760 --> 00:12:08,199 Speaker 1: in questions like how does this functionality fit into the 240 00:12:08,240 --> 00:12:11,080 Speaker 1: code that's already there? Or if someone has to update 241 00:12:11,160 --> 00:12:12,840 Speaker 1: this code in the future, how do I make it 242 00:12:12,880 --> 00:12:14,679 Speaker 1: easy for them to understand what I've written and make 243 00:12:14,760 --> 00:12:17,720 Speaker 1: changes without breaking a bunch of other stuff. A writer 244 00:12:17,840 --> 00:12:20,800 Speaker 1: doesn't just write words. They just ideas and ideas and 245 00:12:20,800 --> 00:12:23,080 Speaker 1: emotions and thoughts and facts and feelings into a condensed 246 00:12:23,080 --> 00:12:25,240 Speaker 1: piece of text. They sit up late at night typing 247 00:12:25,320 --> 00:12:27,600 Speaker 1: thousands and thousands of words, and it drives them in say, 248 00:12:29,280 --> 00:12:31,360 Speaker 1: it's often quite a motive or at the very least 249 00:12:31,440 --> 00:12:34,000 Speaker 1: driven who or inspired by a given emotion, which is 250 00:12:34,000 --> 00:12:36,040 Speaker 1: something that an AI simply can't replicate in a way 251 00:12:36,040 --> 00:12:40,240 Speaker 1: that's authentic or believable. And a hairdresser doesn't just cut hair. 252 00:12:40,440 --> 00:12:43,679 Speaker 1: They cut your hair, which may be wiry, dry, oily long, 253 00:12:43,800 --> 00:12:46,760 Speaker 1: sure healthy, unhealthy on a scout with particular issues at 254 00:12:46,760 --> 00:12:48,559 Speaker 1: a time of year, perhaps you want to change length 255 00:12:48,720 --> 00:12:50,360 Speaker 1: at a time that fits you in the way you 256 00:12:50,559 --> 00:12:53,160 Speaker 1: like it, which may be impossible to actually write down. 257 00:12:53,200 --> 00:12:55,520 Speaker 1: But they get it just right, and they make conversation 258 00:12:55,640 --> 00:12:57,720 Speaker 1: making you feel at ease while they snip and clip away. 259 00:12:57,760 --> 00:13:00,320 Speaker 1: It tresses with you never having to think for a second. Fuck, 260 00:13:00,320 --> 00:13:02,040 Speaker 1: does this person know what they're doing? Are they going 261 00:13:02,080 --> 00:13:05,439 Speaker 1: to listen to me. This is the true nature of labor. 262 00:13:05,559 --> 00:13:08,640 Speaker 1: The executives fail to comprehend at scale that the things 263 00:13:08,640 --> 00:13:12,640 Speaker 1: we do are not units of work, but extrapolations of experience, emotion, 264 00:13:12,760 --> 00:13:15,960 Speaker 1: and context that cannot be condensed in written meaning or 265 00:13:16,000 --> 00:13:19,520 Speaker 1: bunches of trading material. Business idiots see our labor as 266 00:13:19,559 --> 00:13:22,640 Speaker 1: the results of a smart manager saying do this, rather 267 00:13:22,640 --> 00:13:25,480 Speaker 1: than human ingenuity interpreting both the requests and the shit 268 00:13:25,520 --> 00:13:31,080 Speaker 1: the manager didn't say. Now, what does the CEO do? Well? 269 00:13:32,040 --> 00:13:35,040 Speaker 1: I did look, and a Harvard study said that they 270 00:13:35,120 --> 00:13:38,240 Speaker 1: spend twenty five percent of their time on people and relationships, 271 00:13:38,280 --> 00:13:41,840 Speaker 1: twenty five percent on functional and business unit reviews, sixteen 272 00:13:41,880 --> 00:13:45,240 Speaker 1: percent on organization and culture, and twenty one percent on 273 00:13:45,360 --> 00:13:47,360 Speaker 1: just strategy, with a few percent here and there for 274 00:13:47,480 --> 00:13:52,840 Speaker 1: things like professional development. Hmm. That's who runs the vast 275 00:13:52,880 --> 00:13:55,920 Speaker 1: majority of companies. People that describe their work predominantly as 276 00:13:55,920 --> 00:13:58,640 Speaker 1: looking at staff, talking to people, thinking what we do next, 277 00:13:59,160 --> 00:14:02,480 Speaker 1: and go to lunch. The most highly paid jobs in 278 00:14:02,520 --> 00:14:04,920 Speaker 1: the world are impossible to describe their labour described in 279 00:14:04,920 --> 00:14:08,559 Speaker 1: a mishmash of linked inspiration. Yet everybody else's labor is 280 00:14:08,559 --> 00:14:11,560 Speaker 1: an output that can be automated. As a result, large 281 00:14:11,600 --> 00:14:14,720 Speaker 1: language models must seem like magic to these dickheads. When 282 00:14:14,720 --> 00:14:16,960 Speaker 1: you see everything as an outcome, an outcome, you may 283 00:14:17,040 --> 00:14:19,960 Speaker 1: or may not understand it. Definitely don't understand the process behind, 284 00:14:20,080 --> 00:14:22,560 Speaker 1: let alone care about. You've kind of already see your 285 00:14:22,600 --> 00:14:26,080 Speaker 1: workers as llms. You create a stratification of the workforce 286 00:14:26,120 --> 00:14:29,680 Speaker 1: that goes beyond the normal organizational chart, with senior executives 287 00:14:29,720 --> 00:14:32,840 Speaker 1: those closer to the class level of CEO acting as 288 00:14:33,000 --> 00:14:35,680 Speaker 1: those who have risen above the doldrums of doing things, 289 00:14:35,720 --> 00:14:38,440 Speaker 1: to the level of decision making, a fuzzy term that 290 00:14:38,480 --> 00:14:41,200 Speaker 1: can mean everything from making nuance decisions with input from 291 00:14:41,240 --> 00:14:44,640 Speaker 1: multiple different subject matter experts to as service now Bill 292 00:14:44,720 --> 00:14:47,280 Speaker 1: McDermott did in twenty twenty two and I quote make 293 00:14:47,360 --> 00:14:50,120 Speaker 1: it clear to everybody in a boardroom of other executives 294 00:14:50,120 --> 00:14:54,840 Speaker 1: that everything they do must be AIAIAIAIAI. And that's five 295 00:14:54,880 --> 00:14:57,760 Speaker 1: of those the same extents that some members of the 296 00:14:57,800 --> 00:14:59,920 Speaker 1: business and tech media that have for the most part 297 00:15:00,120 --> 00:15:02,000 Speaker 1: gotten by without having to think too hard about the 298 00:15:02,000 --> 00:15:05,840 Speaker 1: actual things the companies are saying. Look, I realize this 299 00:15:05,880 --> 00:15:08,560 Speaker 1: sounds a little mean, and it's not a unilateral statement, 300 00:15:08,960 --> 00:15:11,120 Speaker 1: and I must must be clear. It doesn't mean that 301 00:15:11,160 --> 00:15:13,560 Speaker 1: these people know nothing, just that it's been possible that 302 00:15:13,600 --> 00:15:15,880 Speaker 1: scoot through the world without thinking too hard about whether 303 00:15:15,960 --> 00:15:19,040 Speaker 1: or not something is true, just because an executive said up. 304 00:15:19,640 --> 00:15:22,000 Speaker 1: When Salesforce said back in twenty twenty four that it's 305 00:15:22,040 --> 00:15:25,360 Speaker 1: Einstein Trust Layer and AI would be transformational for jobs, 306 00:15:25,560 --> 00:15:28,080 Speaker 1: the media dutifully wrote it down and published it without 307 00:15:28,080 --> 00:15:31,080 Speaker 1: a second thought. It fully trusted Mark Benioff when he 308 00:15:31,120 --> 00:15:33,800 Speaker 1: said that Agent Force agents would replace human workers, and 309 00:15:33,800 --> 00:15:35,720 Speaker 1: then again when he said that AI agents were doing 310 00:15:35,800 --> 00:15:38,960 Speaker 1: thirty to fifty percent of all the work in Salesforce itself, 311 00:15:39,200 --> 00:15:43,280 Speaker 1: even though that's an unproven and nakedly ridiculous statement. Salesforce 312 00:15:43,320 --> 00:15:45,320 Speaker 1: is CFO, by the way, said earlier in this year 313 00:15:45,320 --> 00:15:48,440 Speaker 1: that AI wouldn't boost sales growth in twenty twenty five. 314 00:15:48,800 --> 00:15:51,480 Speaker 1: One would think this would change how Salesforce was covered 315 00:15:51,560 --> 00:15:54,120 Speaker 1: or how seriously one takes Mark Benioff, But it hasn't 316 00:15:54,200 --> 00:15:57,720 Speaker 1: because nobody's really paying attention. In fact, nobody seems to 317 00:15:57,720 --> 00:16:13,480 Speaker 1: want to do their job in this case. And this 318 00:16:13,760 --> 00:16:15,840 Speaker 1: is how the core myths of jenerit if AI were 319 00:16:15,840 --> 00:16:18,600 Speaker 1: built by executives saying stuff and the media publishing it 320 00:16:18,640 --> 00:16:21,920 Speaker 1: without thinking about him. AI is replacing workers. AI is 321 00:16:21,920 --> 00:16:26,000 Speaker 1: writing entire computer programs. AI is getting exponentially more powerful. 322 00:16:26,200 --> 00:16:28,320 Speaker 1: What does powerful mean? Well, it means that the models 323 00:16:28,360 --> 00:16:30,520 Speaker 1: are getting better on benchmarks that are rigged in their favor. 324 00:16:30,640 --> 00:16:33,360 Speaker 1: But because nobody fucking explains what the benchmarks are, regular 325 00:16:33,400 --> 00:16:35,640 Speaker 1: people are regularly told that AI is powerful and getting 326 00:16:35,640 --> 00:16:38,640 Speaker 1: more powerful every single day. The only thing powerful about 327 00:16:38,680 --> 00:16:42,880 Speaker 1: generifai is its pathology. The world's executives entirely disconnect different 328 00:16:42,920 --> 00:16:45,360 Speaker 1: labor and natural production, are doing the only thing they 329 00:16:45,400 --> 00:16:47,320 Speaker 1: know how to spend a bunch of money and say 330 00:16:47,400 --> 00:16:51,920 Speaker 1: vague stuff about AI being the future. There are people journalists, investors, 331 00:16:51,960 --> 00:16:54,480 Speaker 1: and analysts that have built entire careers on filling in 332 00:16:54,480 --> 00:16:56,480 Speaker 1: the gaps for the powerful as they splurge billions of 333 00:16:56,520 --> 00:16:59,640 Speaker 1: dollars in repeat with increasing desperation that the future is here, 334 00:16:59,680 --> 00:17:06,159 Speaker 1: and then well, absolutely nothing else happens. You've likely seen 335 00:17:06,320 --> 00:17:09,160 Speaker 1: a few ridiculous headlines recently, though. One of the most 336 00:17:09,200 --> 00:17:11,720 Speaker 1: recent and most absurd is that open Ai will pay 337 00:17:11,720 --> 00:17:14,440 Speaker 1: Oracle three hundred billion dollars over the next four years, 338 00:17:14,480 --> 00:17:17,320 Speaker 1: closely followed with the claim that Invidia will invest and 339 00:17:17,320 --> 00:17:19,520 Speaker 1: I put that in air, quotes one hundred billion dollars 340 00:17:19,560 --> 00:17:22,960 Speaker 1: in open Ai to build ten gigawats of ai data centers, 341 00:17:23,200 --> 00:17:25,200 Speaker 1: though the deal is structured in a way that means 342 00:17:25,240 --> 00:17:28,120 Speaker 1: open Ai is paid progressively as each giga what is deployed, 343 00:17:28,640 --> 00:17:31,600 Speaker 1: and also apparently open Ai will be leasing the chips 344 00:17:31,680 --> 00:17:34,240 Speaker 1: rather than buying them out right. I must be clear 345 00:17:34,240 --> 00:17:36,400 Speaker 1: that these deals are intentionally made to continue the myth 346 00:17:36,440 --> 00:17:38,639 Speaker 1: of generat ifai to pump in video and to make 347 00:17:38,680 --> 00:17:41,320 Speaker 1: sure open ai insiders can sell ten point three billion 348 00:17:41,320 --> 00:17:43,680 Speaker 1: dollars worth of shares, which they're currently trying to do 349 00:17:43,880 --> 00:17:47,480 Speaker 1: at evaluation of five hundred billion goddamn dollars. I want 350 00:17:47,480 --> 00:17:50,119 Speaker 1: to be clear about something open Ai cannot afford the 351 00:17:50,119 --> 00:17:53,000 Speaker 1: three hundred billion dollars. Open Ai has not received a 352 00:17:53,080 --> 00:17:55,720 Speaker 1: dollar from Nvidia and won't do so for at least 353 00:17:55,760 --> 00:17:57,400 Speaker 1: a month when I think they're going to receive ten 354 00:17:57,440 --> 00:18:00,400 Speaker 1: billion dollars. But the rest of that ninety's only coming 355 00:18:00,440 --> 00:18:02,840 Speaker 1: when they build those data centers, which open Ai can't 356 00:18:02,840 --> 00:18:05,800 Speaker 1: afford to do. In Vidia needs this myth to continue 357 00:18:05,840 --> 00:18:07,960 Speaker 1: because in truth, all of these data centers are being 358 00:18:07,960 --> 00:18:10,560 Speaker 1: built for demand that doesn't exist, or that if it 359 00:18:10,920 --> 00:18:14,480 Speaker 1: did exist, doesn't necessarily translate into business customers paying huge 360 00:18:14,480 --> 00:18:17,840 Speaker 1: amount amounts for access to open AI's generative AI services. 361 00:18:18,400 --> 00:18:21,159 Speaker 1: In Vidia, open Ai, Core Weave, and other AI related 362 00:18:21,200 --> 00:18:24,240 Speaker 1: companies hope that by announcing theoretical billions of dollars or 363 00:18:24,440 --> 00:18:27,680 Speaker 1: hundreds of billions of dollars of these strange, vague, and 364 00:18:27,680 --> 00:18:30,639 Speaker 1: impossible seeming deals, they can keep pretending that the demand 365 00:18:30,720 --> 00:18:33,080 Speaker 1: is there, because why else would they build all these 366 00:18:33,160 --> 00:18:37,080 Speaker 1: data centers? Right, Well, there's there's that, and the entire 367 00:18:37,080 --> 00:18:40,000 Speaker 1: stock market races on. In video's back. It accounts for 368 00:18:40,040 --> 00:18:41,800 Speaker 1: seven to eight percent of the value of the S 369 00:18:41,880 --> 00:18:44,520 Speaker 1: and P five hundred, and Jensen Huang needs to keep 370 00:18:44,560 --> 00:18:48,679 Speaker 1: selling those fucking GPUs. I intend to explain later how 371 00:18:48,720 --> 00:18:50,840 Speaker 1: all of this works and how brittle all of this 372 00:18:50,880 --> 00:18:53,879 Speaker 1: really is. But the intention of these deals is simple 373 00:18:53,960 --> 00:18:56,320 Speaker 1: to make you think this much money can't be wrong, 374 00:18:56,720 --> 00:18:59,359 Speaker 1: and I assure you it can. These people need you 375 00:18:59,400 --> 00:19:01,879 Speaker 1: to believe this is inevitable. But they are being proven 376 00:19:01,960 --> 00:19:04,439 Speaker 1: wrong again and again and again, and today I'm going 377 00:19:04,480 --> 00:19:07,840 Speaker 1: to continue to do so. Underpinning these stories about huge 378 00:19:07,880 --> 00:19:10,560 Speaker 1: amounts of money and endless opportunity lies a dark secret. 379 00:19:10,760 --> 00:19:12,320 Speaker 1: The none of this is working, and all of this 380 00:19:12,400 --> 00:19:15,320 Speaker 1: money has been invested in a technology that doesn't make 381 00:19:15,440 --> 00:19:17,720 Speaker 1: much revenue and loves to burn millions or billions or 382 00:19:17,960 --> 00:19:21,480 Speaker 1: hundreds of billions of dollars. Over half a trillion dollars 383 00:19:21,520 --> 00:19:24,159 Speaker 1: in fact, has gone into an entire industry without a 384 00:19:24,200 --> 00:19:27,400 Speaker 1: single profitable company developing models of products built on top 385 00:19:27,440 --> 00:19:31,240 Speaker 1: of these AI models. By my estimates, there's about forty 386 00:19:31,280 --> 00:19:33,359 Speaker 1: four billion dollars of revenue in general of AI this 387 00:19:33,480 --> 00:19:35,679 Speaker 1: year when you add in anthropic and open AIS revenue 388 00:19:35,680 --> 00:19:37,879 Speaker 1: to the part, along with other stragglers, and most of 389 00:19:37,880 --> 00:19:40,600 Speaker 1: that number has been gathered from reporting from outlets like 390 00:19:40,640 --> 00:19:43,320 Speaker 1: The Information. Because none of these companies share their revenues, 391 00:19:43,560 --> 00:19:45,600 Speaker 1: all of them lose shit tons of money, and their 392 00:19:45,640 --> 00:19:50,000 Speaker 1: actual revenues are really, really, really small. Only one member 393 00:19:50,000 --> 00:19:53,080 Speaker 1: of the Magnificent seven outside of Nvidia, has ever disclosed 394 00:19:53,080 --> 00:19:56,359 Speaker 1: its AI revenue, Microsoft, which has stopped reporting, in January 395 00:19:56,400 --> 00:19:59,320 Speaker 1: twenty twenty five, when it reported it would have thirteen 396 00:19:59,320 --> 00:20:02,080 Speaker 1: billion in aniized revenue this well, I guess that will 397 00:20:02,160 --> 00:20:04,800 Speaker 1: be for the month because it's month times twelve, about 398 00:20:04,800 --> 00:20:07,960 Speaker 1: one point eight three billion a month. I'm I know 399 00:20:08,040 --> 00:20:11,480 Speaker 1: that sounds like a lot. But Microsoft is a sales machine. 400 00:20:11,880 --> 00:20:16,199 Speaker 1: It's built specifically to create or exploit software markets, suffocating 401 00:20:16,200 --> 00:20:18,960 Speaker 1: competitors by using its scale to drive down prices and 402 00:20:19,000 --> 00:20:21,880 Speaker 1: to leverage the ecosystem it's created over the past few decades. 403 00:20:22,240 --> 00:20:24,359 Speaker 1: One billion a month in revenue is chump change for 404 00:20:24,400 --> 00:20:26,920 Speaker 1: an organization that makes over twenty seven billion dollars a 405 00:20:27,000 --> 00:20:30,000 Speaker 1: quarter in profits. But hey, it's the early days. Did 406 00:20:30,000 --> 00:20:34,680 Speaker 1: you get it? Hurt out? God? Thank you, Scott? Did 407 00:20:34,720 --> 00:20:36,760 Speaker 1: you not listen to my three part series on how 408 00:20:36,760 --> 00:20:38,679 Speaker 1: to argue with an AI booster? I went over it 409 00:20:38,720 --> 00:20:42,560 Speaker 1: over there, get out? Okay. This is also nothing like 410 00:20:42,640 --> 00:20:45,240 Speaker 1: any other tic era. There's never been this kind of 411 00:20:45,240 --> 00:20:48,000 Speaker 1: cash rush, even in the fiber boom. Over a decade. 412 00:20:48,040 --> 00:20:50,280 Speaker 1: Amazon spent about a tenth of the capex that the 413 00:20:50,280 --> 00:20:53,120 Speaker 1: Magnificent seven spent in the last two years on general AI, 414 00:20:53,240 --> 00:20:56,640 Speaker 1: building Amazon Web Services, something that now powers a vast 415 00:20:56,720 --> 00:20:59,160 Speaker 1: chunk of the web and has long been Amazon's most 416 00:20:59,160 --> 00:21:02,760 Speaker 1: profitable business generally. If AI is also nothing like Uber, 417 00:21:02,800 --> 00:21:05,080 Speaker 1: with open Ai and anthropics true costs coming in at 418 00:21:05,119 --> 00:21:07,200 Speaker 1: around one hundred and fifty nine billion dollars in the 419 00:21:07,240 --> 00:21:10,480 Speaker 1: past two years, approaching five times Uber's thirty billion dollar 420 00:21:10,560 --> 00:21:13,080 Speaker 1: all time burn and that's before the bullshit with Nvideo 421 00:21:13,119 --> 00:21:17,040 Speaker 1: and an Oracle. Microsoft last reported that AI revenue in January. 422 00:21:17,080 --> 00:21:20,040 Speaker 1: By the way, it's now October. Why did it stop 423 00:21:20,040 --> 00:21:22,159 Speaker 1: reporting the number? Do you think is it because the 424 00:21:22,240 --> 00:21:25,160 Speaker 1: numbers are so good they couldn't possibly let you know? Hmm. 425 00:21:26,680 --> 00:21:29,640 Speaker 1: As a general rule, publicly traded companies, especially those where 426 00:21:29,640 --> 00:21:33,080 Speaker 1: the leadership are compensated primarily in equity so stock, they 427 00:21:33,119 --> 00:21:35,639 Speaker 1: tend to brag about their successes, in part because bragging 428 00:21:35,640 --> 00:21:37,600 Speaker 1: boosts the value of the thing that the leadership gets 429 00:21:37,600 --> 00:21:41,000 Speaker 1: paid in. There's no benefit to being shy. Look, Oracle 430 00:21:41,000 --> 00:21:43,879 Speaker 1: announced they literally filed something saying they had that huge 431 00:21:44,000 --> 00:21:46,919 Speaker 1: three hundred billion dollar contract. They did that despite the stock. 432 00:21:47,800 --> 00:21:50,560 Speaker 1: Why is Microsoft not doing that with their incredible AI revenues? 433 00:21:50,600 --> 00:21:53,760 Speaker 1: Do you think it's because they're shy? Come on, Satcha, 434 00:21:53,840 --> 00:21:55,560 Speaker 1: come on out, come on, Satcher. You can show me 435 00:21:55,600 --> 00:21:58,960 Speaker 1: in the numbers. Been in all seriousness, If Microsoft can't 436 00:21:58,960 --> 00:22:02,439 Speaker 1: sell this, nobody can. All right, So I'm explaining this 437 00:22:02,480 --> 00:22:04,159 Speaker 1: whole thing is if you're brand new and walking up 438 00:22:04,160 --> 00:22:06,760 Speaker 1: to this relatively unprepared, so I need to introduce another 439 00:22:06,800 --> 00:22:10,000 Speaker 1: company In twenty twenty, a splinter group jumped off of 440 00:22:10,000 --> 00:22:12,200 Speaker 1: open ai, funded by Amazon and Google to do much 441 00:22:12,240 --> 00:22:14,000 Speaker 1: the same thing as open ai, but pretend to be 442 00:22:14,080 --> 00:22:16,199 Speaker 1: nicer about it until they had to raise money from 443 00:22:16,200 --> 00:22:19,160 Speaker 1: the Middle East. I am, of course talking about Anthropic, 444 00:22:19,200 --> 00:22:21,000 Speaker 1: and they've always been a bit better at coding for 445 00:22:21,040 --> 00:22:23,600 Speaker 1: some reason, and people really like their clawed models. But 446 00:22:24,000 --> 00:22:26,800 Speaker 1: like does not mean profit or even much revenue. Both 447 00:22:26,800 --> 00:22:29,280 Speaker 1: open ai and Anthropic have become the only two companies 448 00:22:29,280 --> 00:22:31,879 Speaker 1: in generative AI to make any real progress, either in 449 00:22:31,960 --> 00:22:34,920 Speaker 1: terms of recognition or in sheer commercial terms, accounting for 450 00:22:34,920 --> 00:22:37,919 Speaker 1: the vast majority of revenue in the AI industry. In 451 00:22:37,960 --> 00:22:40,800 Speaker 1: a very real sense, the AI industry's revenue is open 452 00:22:40,840 --> 00:22:44,359 Speaker 1: ai and Anthropic. In the year where Microsoft recorded thirteen 453 00:22:44,400 --> 00:22:47,040 Speaker 1: billion dollars in AI revenues, ten billion dollars of that 454 00:22:47,080 --> 00:22:50,800 Speaker 1: came from open AI's spending on Microsoft Azure, Anthropic burned 455 00:22:50,840 --> 00:22:53,120 Speaker 1: five point three billion dollars last year, with the vast 456 00:22:53,160 --> 00:22:56,560 Speaker 1: majority of that going towards compute. Outside of these two companies, 457 00:22:56,600 --> 00:22:59,159 Speaker 1: there's barely enough revenue to justify a single data center. 458 00:22:59,480 --> 00:23:02,080 Speaker 1: Where we see today is a time of immense tension. 459 00:23:02,600 --> 00:23:05,000 Speaker 1: Mark Zuckerberg says we're in a bubble. Sam Mortman says 460 00:23:05,040 --> 00:23:07,120 Speaker 1: we're in a bubble. At a barber chairman and billionaire 461 00:23:07,200 --> 00:23:09,240 Speaker 1: Joe Size says we're in a bubble. Apollo says we're 462 00:23:09,240 --> 00:23:11,760 Speaker 1: in a bubble. Nobody's making money and nobody knows why 463 00:23:11,800 --> 00:23:14,040 Speaker 1: they're actually doing this anymore, just that they must do 464 00:23:14,119 --> 00:23:16,800 Speaker 1: it and must do so immediately. And they've yet to 465 00:23:16,840 --> 00:23:20,320 Speaker 1: make the case that jenerat Ifai warranted any of the expenditures. 466 00:23:20,960 --> 00:23:23,000 Speaker 1: Now we're a quarter of the way through this four party, 467 00:23:23,080 --> 00:23:25,080 Speaker 1: but this one was necessary. I needed to get you 468 00:23:25,200 --> 00:23:26,880 Speaker 1: up to speed and kind of give you the lay 469 00:23:26,920 --> 00:23:29,040 Speaker 1: of the land because we're going to go a little 470 00:23:29,040 --> 00:23:31,240 Speaker 1: deeper in the next episode and I can't wait for 471 00:23:31,280 --> 00:23:41,439 Speaker 1: you to hear him see it tomorrow. Thank you for 472 00:23:41,480 --> 00:23:44,159 Speaker 1: listening to Better Offline. The editor and composer of the 473 00:23:44,160 --> 00:23:47,240 Speaker 1: Better Offline theme song is Mattasowski. You can check out 474 00:23:47,280 --> 00:23:50,840 Speaker 1: more of his music and audio projects at Mattasouski dot com, 475 00:23:51,080 --> 00:23:55,959 Speaker 1: M A T. T O. S O w Ski dot com. 476 00:23:56,040 --> 00:23:58,639 Speaker 1: You can email me at easy at Better offline dot com, 477 00:23:58,760 --> 00:24:01,399 Speaker 1: or visit Better Offline dot com to find more podcast links, 478 00:24:01,400 --> 00:24:04,560 Speaker 1: and of course my newsletter. I also really recommend you 479 00:24:04,560 --> 00:24:06,720 Speaker 1: go to chat dot where's youread dot at to visit 480 00:24:06,720 --> 00:24:09,399 Speaker 1: the discord, and go to our slash Better Offline to 481 00:24:09,440 --> 00:24:12,560 Speaker 1: check out I'll Reddit. Thank you so much for listening. 482 00:24:13,400 --> 00:24:16,080 Speaker 1: Better Offline is a production of cool Zone Media. For 483 00:24:16,200 --> 00:24:19,360 Speaker 1: more from cool Zone Media, visit our website cool Zonemedia 484 00:24:19,400 --> 00:24:22,240 Speaker 1: dot com, or check us out on the iHeartRadio app, 485 00:24:22,320 --> 00:24:25,000 Speaker 1: Apple Podcasts, or wherever you get your podcasts.