WEBVTT - Make Fun Of Them, Pt. 2 0:00:02.800 --> 0:00:06.320 Zone Media. Hello and welcome to Better Offline. I'm your 0:00:06.320 --> 0:00:21.480 host ed Zeitron. Please buy our merchandise. I need money now. 0:00:21.480 --> 0:00:24.040 In the last episode, we talked about an unfortunate affliction 0:00:24.120 --> 0:00:26.880 affecting people in tech, where they're primed to spouting these inane, 0:00:26.920 --> 0:00:30.360 meaningless platitudes, and then the other affliction affecting people in 0:00:30.400 --> 0:00:32.880 the tech media, where they NodD along and say, huh, 0:00:32.920 --> 0:00:35.239 that makes sense to me. The thing is, if we 0:00:35.280 --> 0:00:38.080 actually care about tech, it's upon us to actually challenge 0:00:38.120 --> 0:00:40.800 these charlatans. We don't have to be mean or rude 0:00:40.880 --> 0:00:43.880 or harsh. We just have to say, what does that mean? 0:00:44.000 --> 0:00:46.680 Or what exactly do you mean? This is the pro 0:00:46.760 --> 0:00:49.440 tech position. It's not about being a hater or a 0:00:49.479 --> 0:00:52.879 cynic or whatever. It's an essential quality control mechanism that's 0:00:52.920 --> 0:00:56.360 been sorely lacking, especially in the AI bubble. The problem 0:00:56.440 --> 0:00:59.560 is that bullshit is a scarily effective mechanism by the 0:00:59.560 --> 0:01:02.160 shameless and the cynical, and it's managed to bamboos for 0:01:02.280 --> 0:01:04.480 so many, especially in the media, where the job is, 0:01:04.560 --> 0:01:07.280 at least in theory, to do the exact opposite of 0:01:07.319 --> 0:01:10.320 not reasonably and saying, oh this sounds good to me mate. 0:01:10.959 --> 0:01:13.880 Let's say the media, not the actually good people who 0:01:13.880 --> 0:01:16.760 I mentioned by naming the last episode, but slavish lickspills 0:01:16.800 --> 0:01:20.119 like Kevin Ruce and Casey new and actually want these 0:01:20.120 --> 0:01:22.600 companies to build powerful AI and believe they're smart enough 0:01:22.600 --> 0:01:26.039 to do so. Say that somehow, looking at their decaying finances, 0:01:26.040 --> 0:01:27.679 the lack of revenue, the lack of growth, and the 0:01:27.720 --> 0:01:30.120 remarkable lack of use cases, they still come out of 0:01:30.120 --> 0:01:32.880 it and say, sure, I think they're going to do this. 0:01:33.400 --> 0:01:36.200 The problem with bullshit like the Della's Words Seller Buffet 0:01:36.200 --> 0:01:38.480 and Altman's whatever the fuck he does is that it 0:01:38.560 --> 0:01:42.480 allows people, ostensibly smart, reasonable people to reach these conclusions 0:01:42.480 --> 0:01:47.280 without having to answer the silly little question of how, how, 0:01:47.720 --> 0:01:50.240 how how they're going to do it. Why haven't they 0:01:50.280 --> 0:01:52.120 done it yet? Why three years in are we still 0:01:52.120 --> 0:01:55.040 aren'table to describe what it is that chat GPT actually 0:01:55.080 --> 0:01:58.080 does and why we need it so badly? Take away 0:01:58.160 --> 0:02:00.400 how much money open ai makes for us, second, and 0:02:00.480 --> 0:02:04.560 indeed how much it loses. Does this product actually inspire 0:02:04.640 --> 0:02:08.280 anything in you? What is it that's actually magical about 0:02:08.280 --> 0:02:10.160 this other than the fact case so you get to 0:02:10.160 --> 0:02:12.760 hang around lots of parties and you too, Kevin Kevy boy, 0:02:12.919 --> 0:02:14.440 you get to hang around a lot of parties with 0:02:14.480 --> 0:02:16.720 a bunch of people sniff in their own farts, like 0:02:16.760 --> 0:02:19.680 that one episode of South Park with the priuses. And 0:02:19.760 --> 0:02:22.239 on a business level, what is it that I'm meant 0:02:22.280 --> 0:02:25.799 to be impressed by? Exactly? Open ai has allegedly hit 0:02:25.880 --> 0:02:28.680 ten billion dollars in annualized revenue, essentially the biggest month 0:02:28.720 --> 0:02:32.360 that can find multiplied by twelve, which is it's not 0:02:32.400 --> 0:02:36.360 actually that much really, considering that open ai is the 0:02:36.360 --> 0:02:39.120 most prominent software company in the world, with the biggest brand, 0:02:39.120 --> 0:02:43.000 and with the attention of the entirety of the world's media. 0:02:43.160 --> 0:02:46.359 Open ai allegedly has five hundred million weekly active users 0:02:46.360 --> 0:02:49.320 on chat gpt and by the last count, only fifteen 0:02:49.360 --> 0:02:53.760 point five million paying subscribers, an absolutely putrid conversion rate, 0:02:53.800 --> 0:02:56.240 even before you realize that the actual conversion rate would 0:02:56.240 --> 0:02:59.880 be monthly active subscribers. That's how any real software come 0:03:00.120 --> 0:03:03.120 He actually defines its metrics by the fucking way. Why 0:03:03.160 --> 0:03:07.080 am I meant to be impressed? Why? Because chat gpt 0:03:07.280 --> 0:03:10.800 grew fast. It literally had more pr and more marketing 0:03:10.840 --> 0:03:13.560 and more attention and more opportunities to sell to more people. 0:03:13.560 --> 0:03:15.960 Than any company has ever had in the history of anything. 0:03:16.480 --> 0:03:19.600 Every single industry has been told to think about AI 0:03:19.720 --> 0:03:21.640 for three fucking years, and they've been told to do 0:03:21.720 --> 0:03:24.160 so because of a company called open Ai. There isn't 0:03:24.160 --> 0:03:27.240 a single goddamn product since Google or Facebook that has 0:03:27.320 --> 0:03:29.720 had this level of media pressure, and both of those 0:03:29.760 --> 0:03:32.600 companies launched without the massive amount of media and social 0:03:32.639 --> 0:03:36.480 media that we have today. Having literally everybody talking about 0:03:36.480 --> 0:03:39.040 your product all the time for years is pretty useful. 0:03:39.400 --> 0:03:42.360 Why isn't this company making more money? And why are 0:03:42.360 --> 0:03:46.040 we taking any of these people? Seriously? Mark Zuckerberg paid 0:03:46.080 --> 0:03:49.520 fourteen point three billion dollars to invest forty nine percent 0:03:49.600 --> 0:03:53.560 in but really to acquire Scale Ai, an AI data company, 0:03:53.560 --> 0:03:56.120 as a means of hiring its CEO, Alexander Wang to 0:03:56.760 --> 0:03:59.800 run his superintelligence team, and has been offering random Open 0:03:59.800 --> 0:04:02.880 air Eye employees one hundred million dollars to join Meta. 0:04:02.920 --> 0:04:05.720 And they also, by the way, thought about buying both 0:04:05.760 --> 0:04:09.800 AI search company Perplexity and general iv AI video company Runway, 0:04:09.800 --> 0:04:12.600 And they even tried to buy open ai co foundery 0:04:12.800 --> 0:04:15.920 Suits Caves pre product thirty two billion dollar valuation company 0:04:16.960 --> 0:04:21.000 Save Superintelligence, settling instead to hire its CEO and his 0:04:21.120 --> 0:04:25.040 venture fund. I just want to be clear superintelligence refers 0:04:25.080 --> 0:04:28.640 to a fictional concept. These people they may as well 0:04:28.680 --> 0:04:31.240 be saying they're gonna hunt and kill the fucking tooth Fairy. 0:04:31.520 --> 0:04:34.600 I feel like I'm going insane hundreds of millions, billions 0:04:34.640 --> 0:04:39.280 of dollars put into an idea that's fictional. It's like 0:04:39.360 --> 0:04:43.040 they're trying to make the Ninja turtles happen. Are they 0:04:43.040 --> 0:04:45.440 going to capture Santa Claus. Are they finally going to 0:04:45.520 --> 0:04:49.360 kill Slender Man? Are they going to find the Marrio brothers? 0:04:49.360 --> 0:04:52.120 Are we going to stop the Cooper Sam Altman? Is 0:04:52.160 --> 0:04:55.159 Mark Zuckerberg going to kill who is the bad guy 0:04:55.200 --> 0:04:58.160 from Sacrifice? Anyway, we'll get back to that later when 0:04:58.200 --> 0:05:01.280 you put aside the big number. These are the actions 0:05:01.320 --> 0:05:03.520 of a desperate dim wit with a failing product trying 0:05:03.520 --> 0:05:07.599 to buy his way to make generative AI into super intelligence. 0:05:08.440 --> 0:05:11.280 And by the way, just want to make it clear, 0:05:11.320 --> 0:05:15.600 the chief AI scientist, Yan Lakun they callin Yan Lacum, 0:05:16.400 --> 0:05:19.480 says it isn't going to work. Generative AI isn't going 0:05:19.520 --> 0:05:22.080 to make super intelligence market. Mark, You're going to listen 0:05:22.160 --> 0:05:25.200 to the people you fucking hired and just jack off 0:05:25.200 --> 0:05:27.559 and give people too much money. I'll take a hundred 0:05:27.600 --> 0:05:30.400 million dollars mate, while you're out there. But by assuming 0:05:30.400 --> 0:05:33.200 that there's some sort of grand strategy behind these moves beyond, 0:05:33.240 --> 0:05:35.760 if we get enough smart people together, something will happening. 0:05:36.000 --> 0:05:38.680 The media helps boost the powerfuls messaging and buoy their 0:05:38.680 --> 0:05:43.520 stock valuations. You are not educating anybody by humoring these goofballs. 0:05:43.720 --> 0:05:45.599 In fact, the right way to approach this would be 0:05:45.600 --> 0:05:49.560 to ask meta a very simple question. Why why does 0:05:49.560 --> 0:05:52.240 a multi trillion dollar market cap company with a near 0:05:52.279 --> 0:05:55.200 monopoly overall social media spend billions of dollars in what 0:05:55.240 --> 0:05:58.279 appears to be a totally irresponsible way? No, no, no, 0:05:58.279 --> 0:06:00.359 no need to do that, No need to think these 0:06:00.600 --> 0:06:04.320 these big thoughts that might make people uncomfortable. No no, no, 0:06:05.560 --> 0:06:08.760 we just need like ten or fifteen different articles suggesting 0:06:08.760 --> 0:06:11.000 that Mark Zuckerberger is a genius and we're watching him 0:06:11.000 --> 0:06:16.200 be a genius. Anyway, putting that aside, what exactly is 0:06:16.240 --> 0:06:18.640 the impressive part of generative AI again, I'm coming back 0:06:18.680 --> 0:06:21.600 to this the fucking code. Enough about the code. I'm 0:06:21.640 --> 0:06:23.320 tired of hearing about the code as where to god, 0:06:23.360 --> 0:06:25.560 you people think that being a software engineer is only 0:06:25.600 --> 0:06:28.200 fucking coding, and it's fine if you ship mediocre code. 0:06:28.400 --> 0:06:31.520 Is it bad? Code? Can't bring down entire organizations? What 0:06:31.680 --> 0:06:34.080 is it you think a software engineer does? Is all 0:06:34.080 --> 0:06:36.640 they do code? If you think the answer is yes, 0:06:36.760 --> 0:06:41.520 you are wrong. Human beings may make mistakes in writing code, 0:06:41.560 --> 0:06:43.359 but at least they know what a mistake looks like, 0:06:43.360 --> 0:06:46.200 which a generative AI does not. Because a generative AI 0:06:46.279 --> 0:06:48.560 doesn't know what anything is or anything at all, because 0:06:48.600 --> 0:06:52.400 it is a probabilistic model. Congratulations, you've made another way 0:06:52.839 --> 0:06:55.480 way in which software engineers can automate parts of their jobs. 0:06:55.560 --> 0:06:58.039 Stop being so fucking excited about the idea that people 0:06:58.080 --> 0:07:01.000 are going to lose their livelihoods. It's not and founded 0:07:01.040 --> 0:07:03.760 on absolutely nothing other than your adulation for the powerful. 0:07:04.400 --> 0:07:07.200 These models are dangerous and chaotic, built with little intention 0:07:07.279 --> 0:07:09.200 of regard for the future, just like the rest of 0:07:09.240 --> 0:07:12.080 big text products. Chat GPT would have been a much 0:07:12.080 --> 0:07:14.400 smaller deal if Google had any interest in tending Google 0:07:14.440 --> 0:07:16.880 Search into a product that truly answered the query, as 0:07:16.920 --> 0:07:19.640 opposed to generating more of them to show more impressions 0:07:19.640 --> 0:07:22.600 to advertisers. A nuance search engine that could look at 0:07:22.600 --> 0:07:25.600 a user's query and spit out a series of websites 0:07:25.640 --> 0:07:29.000 that might help answer a question, rather than just summarizing 0:07:29.000 --> 0:07:31.040 a few of them for an answer, or just giving 0:07:31.080 --> 0:07:34.160 you a series of SEO articles. And if you ever 0:07:34.240 --> 0:07:36.400 need proof that Google just doesn't know how to fucking 0:07:36.440 --> 0:07:39.360 innovate anymore, really look at those AI summaries. It's a 0:07:39.360 --> 0:07:42.000 product that both misunderstands search and why people use chat 0:07:42.040 --> 0:07:44.840 GPT as a search replacement in the first place. While 0:07:44.880 --> 0:07:48.120 open AI may summarize stuff to give an answer, it 0:07:48.160 --> 0:07:52.480 at least gives something approximating an answer answer, rather than 0:07:52.480 --> 0:07:54.520 a summary that feels like an absent tee parent trying 0:07:54.520 --> 0:07:56.160 to get rid of you and then throwing twenty bucks 0:07:56.160 --> 0:07:58.320 at you and the hopes you'll leave them alone. And 0:07:58.360 --> 0:08:00.960 even when it does answer shit, it does so in 0:08:01.000 --> 0:08:04.160 this very peculiar way and gets very obvious things wrong. 0:08:04.200 --> 0:08:08.800 I looked at the pricing for Claude the Anthropic. I 0:08:08.800 --> 0:08:10.760 looked up for what the price of claud code was, 0:08:10.800 --> 0:08:12.880 and it was like, yeah, sixty dollars. There's no sixty 0:08:12.920 --> 0:08:16.360 dollars plan that I could find on Anthropic, maybe one 0:08:16.400 --> 0:08:20.800 of the team's plans. Anyway, Google Search makes them like 0:08:20.800 --> 0:08:33.679 one hundred billion dollars a year. It's fucking insane. If 0:08:33.720 --> 0:08:37.240 Google Search truly evolved, chat GPT wouldn't really matter, because 0:08:37.240 --> 0:08:39.320 the idea of a machine that can theoretically answer a 0:08:39.400 --> 0:08:42.120 question is kind of why people use Google in the 0:08:42.120 --> 0:08:46.520 fucking first place. Why doesn't the state of Google dominate 0:08:46.600 --> 0:08:49.760 tech news just like how random ketamine fuel tweets from 0:08:49.760 --> 0:08:53.360 me long Musk do. Why aren't we collectively repulsed by 0:08:53.360 --> 0:08:56.280 Google as a company, and why aren't we collectively repulsed 0:08:56.280 --> 0:08:59.640 by open Ai? No matter how big chat GPT is, 0:08:59.679 --> 0:09:01.560 the fact that there's a product out there with hundreds 0:09:01.559 --> 0:09:04.280 of millions of users that constantly gets answers wrong is 0:09:04.400 --> 0:09:07.480 genuinely worrying for society. And that's before you get to 0:09:07.480 --> 0:09:10.280 the environmental damage. The fact it's trained its models on 0:09:10.440 --> 0:09:13.000 hundreds of millions of people's art and writing, and now 0:09:13.000 --> 0:09:15.840 I don't know the fact that it loses over it 0:09:15.880 --> 0:09:18.360 loses like billions of it's probably more like twelve billion 0:09:18.360 --> 0:09:21.480 dollars a year. It's planning to lose over one hundred 0:09:21.480 --> 0:09:24.640 billion dollars a year before becoming profitable, and it can't 0:09:24.640 --> 0:09:27.920 even they can't even explain how it become profitable. I'm 0:09:27.960 --> 0:09:31.240 trying to calm down, all right, I'm trying. I don't 0:09:31.240 --> 0:09:33.319 even I don't write in here to get pissed off. 0:09:33.360 --> 0:09:36.079 It's just when I think about it too much. Starting 0:09:36.120 --> 0:09:40.160 the music from Kill Bill anyway, But why are we 0:09:40.280 --> 0:09:43.959 not more horrified? Why are we not more forlorn that 0:09:44.000 --> 0:09:46.400 this is where hundreds of billions of dollars are being 0:09:46.480 --> 0:09:49.599 forced The most prominent company in the tech industry is 0:09:49.640 --> 0:09:52.120 an unstable monolith with a vague product that can only 0:09:52.120 --> 0:09:55.440 make ten billion dollars a year in revenue, not profit, 0:09:55.880 --> 0:09:58.600 as the very fabric of its existence is shoved down 0:09:58.640 --> 0:10:02.240 the throat of every executive the world at once. Also, 0:10:02.240 --> 0:10:04.480 if it's not fed by the way twenty to forty 0:10:04.520 --> 0:10:07.840 billion dollars a year, it will die. Give me a 0:10:07.880 --> 0:10:10.800 fucking break. I don't know. I sound pretty ornery. I 0:10:10.800 --> 0:10:12.760 get accused of being a hater or missing the grand 0:10:12.840 --> 0:10:15.160 mystery of this bullshit every few minutes by somebody with 0:10:15.200 --> 0:10:17.199 an AI avatar of a guy who looks like he's 0:10:17.240 --> 0:10:19.920 been banned from multiple branches are best Buy. I understand 0:10:19.920 --> 0:10:22.679 there are things that people do with large language models. 0:10:22.720 --> 0:10:25.160 I am aware, but none of it matters because the 0:10:25.200 --> 0:10:27.400 way they're being discussed is like we're two steps away 0:10:27.400 --> 0:10:30.760 from digitally replacing hundreds of millions of people's jobs. The 0:10:30.800 --> 0:10:33.920 reality is far simpler. We have an industry that has 0:10:33.960 --> 0:10:37.439 spent nearly half a trillion dollars between its capital expenditures 0:10:37.480 --> 0:10:40.240 and venture capital funding to create an industry with the 0:10:40.240 --> 0:10:44.480 combined revenue of the fucking smart watch industry. What I'm 0:10:44.520 --> 0:10:47.480 talking about isn't inflammatory. In fact, it's far more deeply 0:10:47.559 --> 0:10:50.200 rooted in reality than those claiming the open ayes building 0:10:50.200 --> 0:10:52.839 the future or Kevin Ruce walking up on stage dressed 0:10:52.880 --> 0:10:56.240 like a fucking ring master. Anyone, if you're a listener 0:10:56.280 --> 0:10:58.880 who is at the hard Fork Live show, email me. 0:10:59.320 --> 0:11:01.760 Please email me and tell me what that was about. 0:11:01.920 --> 0:11:03.680 And if you don't know what I'm talking about, Kevin 0:11:03.760 --> 0:11:09.160 Rouse dress like a fucking circus ring master, this podcast shameful, man, shameful, 0:11:09.720 --> 0:11:14.960 Take a shower, take a walk, go outside, mate. But 0:11:15.040 --> 0:11:19.240 look like I said, what I'm saying sounds inflammatory, but 0:11:19.280 --> 0:11:22.360 it's not. If we add up the combined capital expenditures 0:11:22.360 --> 0:11:25.000 and projected AI revenues of the big four hyperscalers, we 0:11:25.080 --> 0:11:27.440 end up with roughly three hundred and twenty seven billion 0:11:27.440 --> 0:11:31.880 dollars in capital expenditures and only eighteen billion dollars in revenue. 0:11:31.960 --> 0:11:34.160 And that's not profit, by the way, I really do 0:11:34.280 --> 0:11:36.120 mean it. That's less than the projective revenue of the 0:11:36.120 --> 0:11:39.480 global smartwatch industry. But then someone smashes through my door 0:11:39.520 --> 0:11:43.040 and they go, ooh, what about open Ai. I've talked 0:11:43.040 --> 0:11:46.079 about this so much. So what open ai makes twelve 0:11:46.120 --> 0:11:48.880 point seven billion dollars this year but they lose like 0:11:48.960 --> 0:11:52.160 ten fourteen billion dollars. What does that mean to you? Exactly? 0:11:52.520 --> 0:11:54.560 What do you what are you gonna say the cost 0:11:54.559 --> 0:11:58.199 of inference is coming down? No? No, if you are 0:11:58.240 --> 0:12:01.160 someone who's saying the cost of inference is coming down, 0:12:01.480 --> 0:12:03.960 I need you to stop. You are wrong. You are wrong. 0:12:04.040 --> 0:12:06.880 You are wrong. I hate hearing this because you are 0:12:06.920 --> 0:12:10.160 so wrong. No, the cost that people are being charged 0:12:10.280 --> 0:12:12.600 is going down. We have no firm data on the 0:12:12.640 --> 0:12:15.079 actual costs of inference because the companies don't want to 0:12:15.120 --> 0:12:18.920 talk about it. And yes, they will absolutely lower prices 0:12:19.000 --> 0:12:22.520 to compete with other companies. The Information just reported the 0:12:22.559 --> 0:12:25.040 open Ai was doing this to compete with Microsoft a 0:12:25.040 --> 0:12:28.360 couple weeks back, and fucking open Ai reduced the price 0:12:28.400 --> 0:12:30.920 of their O through Reasoning Model one focused on code 0:12:30.920 --> 0:12:35.720 by eighty percent to compete with Claude four Opusmax, just 0:12:35.760 --> 0:12:38.320 that any of you could fucking look at. I know 0:12:38.360 --> 0:12:40.400 I'm getting mad at people who probably don't listen to 0:12:40.480 --> 0:12:44.400 the podcast, but one day they will. I'll make them anyway. 0:12:44.480 --> 0:12:46.679 Even if we add open AI's revenue to the part, 0:12:46.760 --> 0:12:49.480 we're at about thirty point seven billion dollars. If we 0:12:49.480 --> 0:12:52.000 had the supposed one billion in revenue from training data 0:12:52.000 --> 0:12:55.240 startup search three hundred billion dollars in annualized revenue from churing. 0:12:55.640 --> 0:12:58.559 We optimistically assume that Perplexity will have one hundred million 0:12:58.800 --> 0:13:01.560 arr up than thirty four million in twenty twenty four, 0:13:01.600 --> 0:13:04.840 but they lost sixty four million, and say they make 0:13:04.840 --> 0:13:07.520 one hundred million in twenty twenty five, and assume that 0:13:07.600 --> 0:13:11.640 any sphere which makes cursor that their five hundred million 0:13:11.720 --> 0:13:15.000 dollar run rate stays consistent through twenty twenty five, even 0:13:15.040 --> 0:13:17.160 though they just had to completely change their pricing, because 0:13:17.559 --> 0:13:19.920 because open ai and Anthropic joked on the prices, we 0:13:20.000 --> 0:13:29.560 are at carry the two about thirty two point seven billion. Hmm. 0:13:30.360 --> 0:13:32.760 That's not good. But I'm not being fair, am I. 0:13:32.760 --> 0:13:35.200 I include many of the names from the Information's Generator 0:13:35.280 --> 0:13:38.040 AI database. I'm stubborn, so I made a point of 0:13:38.080 --> 0:13:39.680 adding them all up and ended up with a total 0:13:39.720 --> 0:13:42.360 of less than thirty nine billion dollars of total revenue 0:13:42.360 --> 0:13:45.880 in the entire generator of AI industry. Jesus fucking Christ, 0:13:47.240 --> 0:13:51.800 Fuck God, damn it. This this is what we've been 0:13:51.840 --> 0:13:55.559 doing for three years. If you're missed, blanket fan, this 0:13:55.600 --> 0:13:59.560 is my Kodak printer moment. Why why did you do it? 0:13:59.640 --> 0:14:02.440 According to the Information Generative, AI companies have raised more 0:14:02.440 --> 0:14:04.760 than eighteen point eight billion dollars in the first quarter 0:14:04.800 --> 0:14:07.839 of twenty twenty five, after Vis's invested twenty one billion 0:14:07.880 --> 0:14:09.920 dollars in Q four twenty twenty four and four billion 0:14:09.960 --> 0:14:12.280 dollars in Q three twenty twenty four, for grand total 0:14:12.280 --> 0:14:15.000 of forty three point eight billion, or a total of 0:14:15.080 --> 0:14:18.319 three hundred and seventy point eight billion dollars of investment 0:14:18.679 --> 0:14:21.640 and capital expenditures for an industry that, despite being the 0:14:21.720 --> 0:14:24.200 single most talked about thing on the planet, cannot even 0:14:24.440 --> 0:14:26.480 create a tenth of the dollars it requires to make 0:14:26.560 --> 0:14:30.720 it function. These companies are predominantly unprofitable, perpetually searching for 0:14:30.760 --> 0:14:33.000 product market fit and even when they find it seeming 0:14:33.040 --> 0:14:36.600 capable of generating revenue numbers that remotely justify their valuations. 0:14:36.800 --> 0:14:38.960 And if I'm honest, I think the truly radical position 0:14:39.040 --> 0:14:41.160 here is the one taken by most tech reporters that 0:14:41.200 --> 0:14:43.480 would rather take the lazy position of well, Uber lost 0:14:43.520 --> 0:14:45.880 a lot of money than think for two seconds about 0:14:45.880 --> 0:14:48.520 whether we're all being sold a line of shit. What 0:14:48.560 --> 0:14:50.960 we're watching is a mountain of waste perpetuated by the 0:14:51.040 --> 0:14:54.400 least charming failed sons of our generation. Nobody should be 0:14:54.400 --> 0:14:57.000 giving sach In the Della or Sam Ortman a glossy profile. 0:14:57.120 --> 0:15:00.000 They should be asking direct, brutal questions, much like Joanna's 0:15:00.000 --> 0:15:03.240 Stern just did of Craig Federighi, who had absolutely fucking 0:15:03.320 --> 0:15:06.240 nothing to share about why Apple Intelligence sucked because he's 0:15:06.240 --> 0:15:09.080 never been pushed like this. Put aside the money for 0:15:09.120 --> 0:15:12.520 a second, to be honest, These men are pathetic, unimpressive, uninventive, 0:15:12.520 --> 0:15:16.280 and dreadfully, dreadfully boring anthropics Warrio Ama Day and Open 0:15:16.280 --> 0:15:18.600 Ai is Clammy. Sam Mortman a farmer in common with 0:15:18.640 --> 0:15:22.120 televangelist Joe Olstein than the level out with Steve Jobs 0:15:22.200 --> 0:15:24.240 or any number of people that have actually invented things, 0:15:24.240 --> 0:15:26.560 and I know about Steve Jobs. And they got that 0:15:26.680 --> 0:15:29.520 way because we took them seriously instead of saying, wait, 0:15:29.560 --> 0:15:33.800 what do you mean? What do you mean? What does 0:15:33.840 --> 0:15:35.760 that mean? To a single one of their wrongheaded o 0:15:35.880 --> 0:15:39.360 fish and dim witted hype ERPs, it's boring. I'm terribly 0:15:39.400 --> 0:15:41.680 horribly bored. And if you're interested in this ship, I'm 0:15:41.720 --> 0:15:45.280 genuinely curious why, especially if you're a reporter, because right now, 0:15:45.320 --> 0:15:48.960 the innovation happening in AI is at best further mutations 0:15:48.960 --> 0:15:51.720 of the software as a service business model, providing far 0:15:51.800 --> 0:15:55.720 less value than previous innovations that are calamitous cost reasoning 0:15:55.720 --> 0:15:58.040 models don't even reason. It's proven by an Apple paper 0:15:58.160 --> 0:16:00.480 released a few weeks ago, and agents as a sceptu 0:16:00.600 --> 0:16:04.320 fucked because large language models are inherently unreliable. And yes, 0:16:04.760 --> 0:16:07.680 a study out of fucking Salesforce found that agents began 0:16:07.720 --> 0:16:10.400 to break down when given multi step tasks, such as 0:16:10.560 --> 0:16:13.400 any task that you'd want to have an agent automating. 0:16:28.440 --> 0:16:33.000 But but, but but but I have one radical suggestion, 0:16:34.040 --> 0:16:36.320 let's start making fun of them. Let's start making fun 0:16:36.360 --> 0:16:39.400 of these people. They're not charming, they're not building anything 0:16:39.600 --> 0:16:42.120 they've scootered long a massing billions of dollars, promising the 0:16:42.120 --> 0:16:44.440 world and delivering you a hell is shit. They deserve 0:16:44.480 --> 0:16:47.800 our derision or at the very least, our deep unerring suspicion, 0:16:48.000 --> 0:16:51.480 if not for what they've done, for what they've not done. 0:16:51.520 --> 0:16:54.440 Sam Ltman is nowhere near delivering a functioning agent, let 0:16:54.440 --> 0:16:57.720 alone anything approaching intelligence, and really only has one skill 0:16:57.920 --> 0:16:59.880 making other companies risk a bunch of money on this 0:17:00.080 --> 0:17:03.200 you bid fucking ideas. No, really, he convinced Oracle to 0:17:03.200 --> 0:17:05.240 buy forty billion dollars of then video chips to put 0:17:05.240 --> 0:17:08.119 in the Abilene takes a Stargate data center, despite the 0:17:08.119 --> 0:17:10.440 fact that the Stargate organization is yet to be formed, 0:17:10.440 --> 0:17:13.320 as reported by The Information, SoftBank and Microsoft pay for 0:17:13.359 --> 0:17:15.040 all of open AI's bills, and the media does his 0:17:15.160 --> 0:17:18.600 marketing for him. Open Ai is, as I have said before, 0:17:18.920 --> 0:17:22.080 a banana republic. It requires the media and the markets 0:17:22.080 --> 0:17:24.119 to make up why it should exist. It requires other 0:17:24.160 --> 0:17:26.720 companies to pump it full of money and build its infrastructure, 0:17:26.880 --> 0:17:29.280 and it doesn't even make products that matter. While Sam 0:17:29.280 --> 0:17:32.600 Altman constantly talks about all the other exciting shit that 0:17:32.640 --> 0:17:35.280 people will build that never seems to get built. You 0:17:35.320 --> 0:17:37.679 can keep honking off about how it will build the 0:17:37.760 --> 0:17:40.600 API that will power the future. But if that's the case, 0:17:40.640 --> 0:17:44.199 where's the fucking future? Exactly where is it? What am 0:17:44.240 --> 0:17:48.080 I looking at here? Where's the economic activity? Where's the productivity? 0:17:48.119 --> 0:17:50.760 The return sucked? The costs are too high? Why am 0:17:50.840 --> 0:17:54.399 I the radical person saying this? This entire situation is 0:17:54.480 --> 0:17:58.200 goddamn ridiculous and incomparable waste, even if it's somehow went 0:17:58.280 --> 0:18:00.639 to the green. For the horrendous amount of capital and 0:18:00.680 --> 0:18:03.080 generative AI to make sense, the industry would have to 0:18:03.119 --> 0:18:07.040 have more revenue than the smartphone and enterprise SaaS market combined, 0:18:07.119 --> 0:18:09.800 rather than less than half of the mobile gaming industry. 0:18:10.440 --> 0:18:14.359 Sacha Nadella, sam Orman, Wario Ammada, Tim Cook, Andy Jesse. 0:18:14.680 --> 0:18:16.920 They deserve to be laughed at, mocked, or at least 0:18:17.080 --> 0:18:21.040 very heavily interrogated because their combined might has produced no 0:18:21.160 --> 0:18:24.119 exciting or interesting products outside of, at best, what will 0:18:24.160 --> 0:18:27.520 amount to a productivity upgrade for integrated development environments and 0:18:27.560 --> 0:18:29.359 faster ways to throw out code that may or may 0:18:29.440 --> 0:18:32.960 not be reliable. These things aren't nothing, but they're nowhere 0:18:32.960 --> 0:18:35.560 near the something that we've been promised, So I put 0:18:35.600 --> 0:18:38.480 it to you, dear listener, why are we taking them seriously? 0:18:38.920 --> 0:18:41.560 What is there to take seriously other than their ability 0:18:41.560 --> 0:18:43.960 to force stuff on people and make money doing so. 0:18:45.119 --> 0:18:47.920 And I want to ask you a question, how do 0:18:48.000 --> 0:18:50.040 they manage to keep doing this? They always seem to 0:18:50.080 --> 0:18:53.399 find new growth every single quarter, every single quarter, without fail. 0:18:53.880 --> 0:18:55.639 Is it because they keep coming up with ideas? Or 0:18:55.640 --> 0:18:57.320 is it because they keep coming up with new ideas 0:18:57.359 --> 0:18:59.480 to get more money a vastly different choice that involves 0:18:59.480 --> 0:19:02.119 increasing the products or making them worse so that they 0:19:02.160 --> 0:19:05.000 can show you more ads. My positions aren't radical, and 0:19:05.040 --> 0:19:07.679 if you believe they are, your deverence to power disgusts me. 0:19:08.320 --> 0:19:10.840 In any case, I want to end this episode with 0:19:10.920 --> 0:19:13.720 something a little more inspirational, because I believe things can 0:19:13.840 --> 0:19:17.720 change when regular people feel stronger and more capable. I 0:19:17.760 --> 0:19:19.879 want you to know that you are fully capable of 0:19:20.000 --> 0:19:23.720 understanding all of this. I don't care if you think 0:19:23.760 --> 0:19:26.320 you're not a numbers person or you don't really get business. 0:19:26.840 --> 0:19:29.359 I don't have a single iota of economics training, and 0:19:29.440 --> 0:19:33.280 everything you've ever heard me say or read me right 0:19:34.160 --> 0:19:36.119 has been something I've had to learn, and I really 0:19:36.160 --> 0:19:38.800 mean that I was a lay person right up until 0:19:38.840 --> 0:19:40.520 the time I learned the stuff, and I be kind 0:19:40.520 --> 0:19:43.440 of stuff, knowah, just like you can be. The tech industry, 0:19:43.520 --> 0:19:48.119 the finance industry, the entire mechanisms of capital want you 0:19:48.200 --> 0:19:50.680 to believe that everything they do is magical and complex, 0:19:50.840 --> 0:19:53.639 when it's all far more obvious than you believe. You 0:19:53.680 --> 0:19:56.760 don't have to understand the entire fundamentals of finance to 0:19:56.800 --> 0:19:59.879 know how venture capital works. They buy percentages of come 0:20:00.000 --> 0:20:02.560 copanies at evaluation that they hope is much lower than 0:20:02.600 --> 0:20:04.199 the company would be worth in the future when they 0:20:04.280 --> 0:20:06.760 sell or go public. You don't need to be technical 0:20:06.760 --> 0:20:09.240 to know that large language models generate a response based 0:20:09.240 --> 0:20:12.000 on billions of pieces of training data and by guessing 0:20:12.040 --> 0:20:15.080 at what the next bit of text or thing might 0:20:15.160 --> 0:20:17.000 be in a line should be based on based on 0:20:17.040 --> 0:20:20.760 what they've seen in the model previously. These people love 0:20:20.840 --> 0:20:23.159 to say things like ah, but didn't see and present 0:20:23.280 --> 0:20:26.119 some anecdote, when no anecdote will ever defeat the basics 0:20:26.119 --> 0:20:28.439 of your business does not make enough money, the software 0:20:28.480 --> 0:20:29.960 does not do the things you claim it's meant to, 0:20:29.960 --> 0:20:32.879 and you have no path to profitability. They can yamor 0:20:32.960 --> 0:20:35.520 at you. All they want about lots of people using CHATGPT, 0:20:35.640 --> 0:20:37.840 but that doesn't change the fact that chat GPT just 0:20:37.960 --> 0:20:40.520 isn't that revolutionary, and their only play here is to 0:20:40.520 --> 0:20:42.880 make you feel stupid, rather than actually showing you why 0:20:42.880 --> 0:20:46.080 it's so fucking revolutionary. This is the argument of a 0:20:46.119 --> 0:20:50.040 manipulator and a coward, and you're above such things. You 0:20:50.080 --> 0:20:52.240