WEBVTT - Why OpenAI Is A Terrible Business 0:00:02.440 --> 0:00:06.760 A media Hello, and welcome to Better Offline. I'm your 0:00:06.800 --> 0:00:20.439 host ed zich Tron. What as ever, please check the 0:00:20.480 --> 0:00:22.520 episode notes for links to the things that I'm talking 0:00:22.560 --> 0:00:26.880 about now. Frequent listeners know that I'm immensely skeptical of 0:00:26.920 --> 0:00:30.320 generative AI, both as a technology but also as a business. 0:00:30.680 --> 0:00:33.920 It's expensive, it's unreliable, it doesn't actually do that much, 0:00:34.280 --> 0:00:36.600 and the real world use cases range from kind of 0:00:36.640 --> 0:00:41.280 helpful to unhelpful to not really useful in any way, 0:00:41.320 --> 0:00:45.000 shape or form, and definitely not useful enough to justify 0:00:45.080 --> 0:00:48.320 hundreds of billions of dollars in spending. And I know 0:00:48.440 --> 0:00:50.720 somebody is gonna email and say, but ed, my mate's 0:00:50.760 --> 0:00:54.640 granny's uncle's dog uses chat GBT to brainstorm or something 0:00:54.680 --> 0:00:57.600 else equally flimsy. And I really need you to put 0:00:57.640 --> 0:01:01.280 all of this in context. Open AI loses five billion 0:01:01.400 --> 0:01:04.000 dollars a year, and basically every big tech company is 0:01:04.040 --> 0:01:06.800 blown past their emissions targets, with few signs that they'll 0:01:06.840 --> 0:01:08.760 ever bother to try and meet them again, all in 0:01:08.760 --> 0:01:12.240 pursuit of the piddliest or most convoluted use cases in history. 0:01:12.680 --> 0:01:16.560 It's a goddamn fast and that's why generative AI is 0:01:16.600 --> 0:01:19.280 such a regular topic on this show. I feel like 0:01:19.319 --> 0:01:22.280 the tech industry is experiencing a kind of collective madness, 0:01:22.280 --> 0:01:25.520 a delusion that's seeing companies like Microsoft and Google, but 0:01:25.680 --> 0:01:28.240 both their companies and our planet on a technology that 0:01:28.280 --> 0:01:33.000 continually fails to deliver on well anything I want to 0:01:33.200 --> 0:01:37.039 say here, And I actually wrote down this supposedly massive potential, 0:01:37.680 --> 0:01:40.520 but what was the potential? Ever? Can anyone actually tell 0:01:40.560 --> 0:01:44.320 me what this was meant to do? Other than agi? Anyway, 0:01:44.600 --> 0:01:47.520 Eventually I see this all falling apart, and I think 0:01:47.520 --> 0:01:49.840 there's going to be a clamorous event when everything does 0:01:49.880 --> 0:01:53.560 actually unravel. I return so often to Generative AI because 0:01:53.560 --> 0:01:57.200 I truly fear the damage that this confrontation will cause, 0:01:57.960 --> 0:02:00.120 and I'm concerned about the damage that's happening right now. 0:02:00.120 --> 0:02:03.360 If I'm honest, And while my tone's off in acerbic 0:02:03.360 --> 0:02:05.080 and I don't think you'd have it an any other way, 0:02:05.480 --> 0:02:08.840 I don't really take much joy in watching these things burn. Sure, 0:02:08.919 --> 0:02:11.639 I admit it's pretty funny thinking about satching a Delea 0:02:11.680 --> 0:02:13.800 getting fired or sunded up a shy getting fired, or 0:02:13.800 --> 0:02:18.440 both of them looking stupid, But it's not the same 0:02:18.520 --> 0:02:21.680 thing Miglee's kind of tempered. When I stop and think 0:02:21.680 --> 0:02:23.720 about what all of this means or will mean, I 0:02:23.720 --> 0:02:26.520 should say, for the companies trying to get funding right now, 0:02:26.760 --> 0:02:28.480 or for the companies that will try and raise money 0:02:28.520 --> 0:02:30.760 after the bubble bursts, and for the tech workers that 0:02:30.800 --> 0:02:33.440 will get laid off, either because Microsoft and Google want 0:02:33.440 --> 0:02:36.280 to put more money into data centers and GPUs, or 0:02:36.320 --> 0:02:38.680 because their generative AI bets have gone tits up and 0:02:38.720 --> 0:02:41.400 they need to save money some other way. And so 0:02:41.440 --> 0:02:44.560 I hope you'll forgive me for a somewhat more somber 0:02:44.600 --> 0:02:47.400 tone to this episode. I want to again make a 0:02:47.480 --> 0:02:50.280 clear headed analysis of where we are today and where 0:02:50.320 --> 0:02:53.080 I think we're going. And the reason I'm retreading this 0:02:53.120 --> 0:02:55.280 path is because over the last month or so, we've 0:02:55.320 --> 0:02:58.600 seen some really really alarming signs about the accelerating crisis 0:02:58.600 --> 0:03:01.600 in generative AI, whispers that the party will soon come 0:03:01.639 --> 0:03:04.760 to a halt. Now, you may remember a few episodes 0:03:04.800 --> 0:03:07.080 back that I started talking about pale horses of the 0:03:07.120 --> 0:03:11.160 AI apocalypse, signs that things were falling apart, like layoffs 0:03:11.200 --> 0:03:15.000 at AI companies, price increases or decreases, internal discord at 0:03:15.040 --> 0:03:19.200 these big AI companies, speciously impressive yet kind of flimsy 0:03:19.240 --> 0:03:23.240 product announcements and so on and so forth. Since then, 0:03:23.440 --> 0:03:26.280 I've kind of been proven right several times. Multiple pale 0:03:26.320 --> 0:03:29.120 horses have emerged. There was a big, stupid magic trick 0:03:29.160 --> 0:03:31.519 to impress investors in the form of open aised rush 0:03:31.600 --> 0:03:34.359 launch of its one and that model, by the way, 0:03:34.440 --> 0:03:38.120 was code named Strawberry. Rumored pricing increases which hat GBT 0:03:38.200 --> 0:03:40.800 in the future. Now they're looking at forty four dollars 0:03:40.840 --> 0:03:43.080 a month by twenty thirty, and they're looking at potentially 0:03:43.120 --> 0:03:46.000 thousands for models in the future, and then layoffs at 0:03:46.000 --> 0:03:49.360 Scale AI, which is one of the biggest players in 0:03:49.440 --> 0:03:52.000 AI training data, if not the biggest. And these are 0:03:52.000 --> 0:03:55.200 all signs that things are beginning to fall apart. And 0:03:55.240 --> 0:03:57.760 I think it's important to explain how precarious things are 0:03:57.840 --> 0:04:00.720 right now despite all of the money stashing around, and 0:04:00.760 --> 0:04:04.040 how dangerous the power of magical thinking really is. I 0:04:04.080 --> 0:04:06.480 want to express my concerns about the fragility of this 0:04:06.600 --> 0:04:10.040 movement and the obsessiveness and directionless that brought us here, 0:04:10.680 --> 0:04:13.120 and I want us all to kind of do better now. 0:04:13.160 --> 0:04:14.960 Don't mean you the listener, but if you remember the 0:04:14.960 --> 0:04:17.920 media that's written something vague about AI and you've kind 0:04:17.960 --> 0:04:21.440 of carried water for Sam Mortman. It is time to stop. 0:04:21.480 --> 0:04:23.320 It was already time to stop some time Aguary, I 0:04:23.360 --> 0:04:26.080 should just be clear, but now really is the time 0:04:26.120 --> 0:04:29.160 to stop. But this week's a two parter, and I'll 0:04:29.200 --> 0:04:32.960 be explaining things well in two parts. Obviously. First of all, 0:04:32.960 --> 0:04:35.480 we're going to talk about open ai, the most recent 0:04:35.520 --> 0:04:38.560 funding round, and how bad their business is and exactly 0:04:38.600 --> 0:04:41.320 how worried you should be about them now. The second 0:04:41.360 --> 0:04:43.359 part is going to be about something I'm calling the 0:04:43.400 --> 0:04:46.440 subprime AI crisis, where I see a growing bubble of 0:04:46.480 --> 0:04:51.200 companies integrating generative AI at prices well heavily subsidized by 0:04:51.240 --> 0:04:54.039 big tech and heavily discounted by companies like open ai 0:04:54.080 --> 0:04:57.560 and Anthropic. Should either of these companies get desperate or 0:04:57.600 --> 0:04:59.839 I don't know, actually want to make more money than 0:04:59.839 --> 0:05:02.280 they spend, they're inevitably going to have to raise the 0:05:02.360 --> 0:05:05.240 prices to match their actual costs, which will likely make 0:05:05.279 --> 0:05:09.880 the existence of many AI startups completely untenable. You know, 0:05:10.320 --> 0:05:12.520 kind of like when the teaser interest rates on subprime 0:05:12.560 --> 0:05:16.279 mortgages that expired during the Financial crisis, and people suddenly 0:05:16.279 --> 0:05:20.000 started losing their houses. In any case, these are going 0:05:20.040 --> 0:05:22.800 to be meaningful, yeah, kind of brutal episodes, in part 0:05:22.839 --> 0:05:25.560 because while it's enjoyable to watch big tech burn or 0:05:25.560 --> 0:05:28.520 get embarrassed, there's a real human cost, like I said, 0:05:28.720 --> 0:05:31.320 and the damage to our environment is already being done 0:05:31.560 --> 0:05:35.880 and might not even stop when Generative AI does. And 0:05:35.920 --> 0:05:38.599 whether Microsoft and Google and the other big generative AI 0:05:38.720 --> 0:05:41.840 backers slowly wind down their positions or cannibalize their companies 0:05:41.880 --> 0:05:44.440 to keep open AI and Anthropic alive, I'm convinced that 0:05:44.480 --> 0:05:46.120 the end result is still going to be the same. 0:05:46.839 --> 0:05:48.880 I fear tens of thousands of people will lose their 0:05:48.960 --> 0:05:51.160 jobs and much of the tech industry will suffer as 0:05:51.200 --> 0:05:53.240 they realize that the only thing that can grow forever 0:05:53.320 --> 0:05:56.400 is a cancer. I'm going to paint you oblique picture 0:05:56.800 --> 0:05:58.880 not just for the big AI players, but for tech 0:05:58.920 --> 0:06:01.400 more widely, and for the people to work at these companies, 0:06:01.640 --> 0:06:03.640 and tell you why I think the conclusion to this 0:06:03.760 --> 0:06:06.840 sordid saga, as brutal and damaging as it will be, 0:06:07.080 --> 0:06:14.280 is coming sooner than you think. Let's begin, as have 0:06:14.400 --> 0:06:17.280 explained in agonizing detail in the past in my newsletter 0:06:17.440 --> 0:06:19.799 Open AI will have to continue to raise more money 0:06:19.800 --> 0:06:23.679 than any startup has ever raised in history, in perpetuity 0:06:24.000 --> 0:06:27.320 just to survive. They're going to burn five billion dollars 0:06:27.320 --> 0:06:29.960 this year, and yes, that's after they make their revenue 0:06:30.080 --> 0:06:32.039 and costs are only set to increase over the next 0:06:32.040 --> 0:06:35.159 few years they develop bigger models. Their business as it 0:06:35.240 --> 0:06:38.320 stands is completely untenable, and I'm going to explain why 0:06:38.400 --> 0:06:42.640 in a bit. Nevertheless, that's why open Ai, the ostensible 0:06:42.680 --> 0:06:46.000 nonprofit that may soon become a for profit who bloody 0:06:46.040 --> 0:06:49.200 knows at this point, raise the funding round at evaluation 0:06:49.279 --> 0:06:52.560 of one hundred and fifty seven billion dollars, bringing in 0:06:52.600 --> 0:06:55.799 a reported six point six billion dollars in cash, along 0:06:55.800 --> 0:06:58.320 with opening a four billion dollar revolving credit line from 0:06:58.320 --> 0:07:01.160 banks like JP morgan See and a bunch of others 0:07:01.160 --> 0:07:04.720 that should know better after the acquisition of Twitter. Investors 0:07:04.720 --> 0:07:07.600 in the funding ground included Josh Kushner's Thrive Capital in 0:07:07.720 --> 0:07:11.000 Video and of course Microsoft, who were somehow undeterred by 0:07:11.000 --> 0:07:14.720 the abrupt exiffs open ai cto mirror Marati days before 0:07:14.760 --> 0:07:18.120 the round closed. Also joining the frame was soft Bank, 0:07:18.320 --> 0:07:20.680 the Japanese investment company with a drag record of making 0:07:20.680 --> 0:07:23.400 some of the worst bets in history, and they dumped 0:07:23.440 --> 0:07:27.440 five hundred million dollars into open Ai, and that alone 0:07:27.520 --> 0:07:30.760 should start getting people a little worried. You see, soft 0:07:30.800 --> 0:07:34.480 Bank's founder, Masayoshi Son is what we in the business 0:07:34.520 --> 0:07:37.680 call the dumbass. I realized this is a very broad 0:07:37.720 --> 0:07:41.480 brush to paint somebody with. But let me explain. Soft 0:07:41.480 --> 0:07:44.680 Bank's vision fund dumped sixteen billion dollars into we Work, 0:07:44.760 --> 0:07:46.480 which by the end of last year, by the way, 0:07:46.560 --> 0:07:49.560 was worth forty four million dollars, and nearly a billion 0:07:49.600 --> 0:07:52.080 dollars into wire Card, which turned out to be an 0:07:52.080 --> 0:07:55.160 elaborate fraud. With its leadership. Are the facing criminal charges 0:07:55.320 --> 0:07:59.280 or becoming fugitives from justice in Russia. Soft Bank has 0:07:59.320 --> 0:08:02.760 become effective synonymous with burning cash on terrible bets and 0:08:02.800 --> 0:08:05.880 lost over thirty two billion dollars in the last three years. 0:08:07.080 --> 0:08:09.480 The problem that open Ai faces is that they need 0:08:09.520 --> 0:08:12.280 an absolute shit ton of money, and generally the only 0:08:12.320 --> 0:08:14.520 companies that need that much are the ones that may 0:08:14.560 --> 0:08:18.240 not have a sustainable business model. Open Ai demanded a 0:08:18.280 --> 0:08:21.880 minimum investment of two hundred and fifty million dollars from investors, 0:08:22.080 --> 0:08:25.120 and while regular American vcs might have that much, and 0:08:25.240 --> 0:08:27.960 indeed had in the case of Costler and Kochume Management, 0:08:28.920 --> 0:08:30.840 there aren't enough of them to fill out a six 0:08:30.880 --> 0:08:33.560 point six billion dollar funding round. And because open Ai 0:08:34.000 --> 0:08:38.240 needed that much, not wanted needed, their only real choices 0:08:38.240 --> 0:08:40.559 were to go to Masayoshi Son, who's a man who 0:08:40.600 --> 0:08:43.360 put tons of money into stupid things and believes he 0:08:43.440 --> 0:08:45.840 was put on this earth to make digital superintelligence and 0:08:45.840 --> 0:08:47.880 I have a link to that. By the way. It's insane, 0:08:48.920 --> 0:08:51.760 all I mean. There's also another choice, you know, the 0:08:52.040 --> 0:08:55.880 far dodgier investors in the Middle East m mg X, 0:08:56.200 --> 0:08:59.160 one hundred billion dollar investment fund backed by the United 0:08:59.160 --> 0:09:03.040 Arab Emirates to invest primarily in AI and semiconductor companies. 0:09:04.160 --> 0:09:06.800 This should also be a big warning sign that things 0:09:06.800 --> 0:09:10.360 are going poorly, because absolutely nobody raises from the UAE 0:09:10.480 --> 0:09:13.319 of the SuDS because they want to. They're the place 0:09:13.320 --> 0:09:14.960 you go if you need a lot of money and 0:09:15.000 --> 0:09:17.040 you're not confident anybody else is going to give it 0:09:17.040 --> 0:09:20.240 to you. Other aspects of this deal are also a 0:09:20.280 --> 0:09:22.840 bit worrying, one of them being that one of the 0:09:22.880 --> 0:09:25.719 founding partners of MGX is the Sovereign Wealth Fund of 0:09:25.760 --> 0:09:29.480 Abu Dhabi, which brought in around five hundred million dollars 0:09:29.520 --> 0:09:33.319 into Anthropic in twenty twenty three. Kind of a conflict 0:09:33.320 --> 0:09:35.360 of interest. Do you think any of these people actually 0:09:35.440 --> 0:09:39.160 goddamn care though they're all working to the same rot economy. 0:09:39.240 --> 0:09:43.319 Nonsense fucking Anyway, as I mentioned previously, open Ai now 0:09:43.360 --> 0:09:45.480 has access to that four billion dollars in debt from 0:09:45.520 --> 0:09:49.199 banks in that revolving credit facility too, and that's also 0:09:49.280 --> 0:09:52.360 not brilliant for a company that just burns cash. I 0:09:52.400 --> 0:09:54.680 don't know what the interest rate on it is, but 0:09:54.720 --> 0:09:57.839 I can't imagine it's great. Generally, revolving credit lines are 0:09:57.920 --> 0:10:02.960 worse interest rates, So that's great, I guess. But open 0:10:03.000 --> 0:10:06.440 ai is desperate. Fundraising comes down to one one thing, 0:10:06.559 --> 0:10:10.600 one really obvious thing. They need so much money. They 0:10:10.679 --> 0:10:13.240 needed like ten billion dollars here, and they kind of 0:10:13.280 --> 0:10:15.880 got it in this wonky way. But they need that 0:10:15.960 --> 0:10:19.000 money to survive because, like I said, they lost five 0:10:19.080 --> 0:10:22.440 billion in twenty twenty four, and that figure is likely 0:10:22.480 --> 0:10:25.679 to increase as more complex models demand more compute and 0:10:25.720 --> 0:10:29.520 more training. Data and Anthropic CEO Dario ama Day predicted 0:10:29.520 --> 0:10:31.640 that future models may cost as much as one hundred 0:10:31.640 --> 0:10:35.120 billion dollars to train and open ai. By the way, 0:10:35.320 --> 0:10:37.640 they may have just raised the biggest round in history. 0:10:37.960 --> 0:10:40.320 They're going to have to raise another round again soon. 0:10:40.960 --> 0:10:45.959 And I'll explain why. While open ai succeeded here, it 0:10:46.000 --> 0:10:48.920 was only after a fairly arduous process where they'd already 0:10:48.960 --> 0:10:51.640 tried to raise a hundred billion dollar valuation earlier in 0:10:51.679 --> 0:10:54.840 the year, and that specifically turned off investors because of 0:10:54.880 --> 0:10:57.880 the huge price tag. And there's this, by the way, 0:10:58.000 --> 0:11:00.640 to quote the information, there's a growing concern over the 0:11:00.720 --> 0:11:04.920 overvaluation of generative AI companies. Oh you fucking think do 0:11:04.920 --> 0:11:07.920 you think that there's a concern here? Well, after this podcast, 0:11:07.920 --> 0:11:11.480 you're going to be really, really concerned. I guarantee anyway 0:11:11.840 --> 0:11:14.200 to get the round done, open ai committed to converting 0:11:14.280 --> 0:11:17.439 itself from a nonprofit to a for profit entity. Failure 0:11:17.480 --> 0:11:19.440 to do so within two years will see their funding 0:11:19.480 --> 0:11:21.920 converted into debt, which by the way, will be the 0:11:21.960 --> 0:11:24.720 kiss of death for a company that only ever loses money. 0:11:25.440 --> 0:11:28.440 The deal, per Reuter's, also includes provisions that would allow 0:11:28.440 --> 0:11:31.440 investors to adjust the valuation or claw back their investment 0:11:31.559 --> 0:11:35.360 should the transition from nonprofit to for profit fail. And 0:11:35.400 --> 0:11:39.800 I mean this is worrying because I don't think there's 0:11:39.840 --> 0:11:42.880 any historical precedence of a company, a nonprofit at least, 0:11:42.880 --> 0:11:45.640 that had this much money then converted. And what complicates 0:11:45.679 --> 0:11:48.080 that as well is converting a business of this scale 0:11:48.080 --> 0:11:51.880 from a nonprofit to a for profit. Well, it involves 0:11:52.000 --> 0:11:55.560 transferring assets, and because any assets previously donated to the 0:11:55.559 --> 0:11:59.320 public benefit, as Alexander Reed, partner at Baker Holsteller, is 0:11:59.400 --> 0:12:02.000 quote as saying The Wall Street Journal cannot be converted 0:12:02.080 --> 0:12:05.600 into a private benefit without compensating the public for the loss. 0:12:06.360 --> 0:12:10.079 Every asset, which includes things like patterns and other intellectual property, 0:12:10.280 --> 0:12:13.400 would have to be paid for. Well, it's hard to 0:12:13.400 --> 0:12:15.640 say for certain, it's likely this deal and the process 0:12:15.640 --> 0:12:19.079 of untangling open ai from its nonprofit beginnings is really 0:12:19.120 --> 0:12:22.559 only going to add to open AI's cash problem. And 0:12:22.720 --> 0:12:24.920 it's not like that was a good problem to deal 0:12:24.960 --> 0:12:28.480 with in the beginning. And now I will do a 0:12:28.559 --> 0:12:31.440 thorough review and analysis of open ai because I want 0:12:31.440 --> 0:12:33.600 you to see exactly how stupid things have become in 0:12:33.640 --> 0:12:36.960 the tech industry and how ridiculous all of this writing 0:12:37.000 --> 0:12:39.760 this boiled my blood, by the way, because when you 0:12:39.800 --> 0:12:43.560 see how lossy this company is, when you really ingest 0:12:43.600 --> 0:12:46.200 how bad open ai is and how stupid it is 0:12:46.240 --> 0:12:49.240 that we've let this happen, and we meaning the tech industry. 0:12:50.160 --> 0:12:52.480 You then will think on your own lives and be like, wow, 0:12:53.120 --> 0:12:57.520 you think I could pay my rent like late, or 0:12:57.600 --> 0:13:00.040 like ask a company to pay my rent because my 0:13:00.240 --> 0:13:03.520 job involves me just burning piles of cash. No, you 0:13:03.520 --> 0:13:05.960 wouldn't be able to lose money for a business. You 0:13:06.040 --> 0:13:09.840 and I we have to operate normally, not sam Oltman, 0:13:09.920 --> 0:13:13.920 though sam Oltman doesn't have to be normal. Sam Altman, 0:13:14.040 --> 0:13:16.679 like all of the gilded assholes of the tech industry, 0:13:17.080 --> 0:13:19.600 is allowed to live in this fantasy land based on 0:13:19.679 --> 0:13:23.480 marketing high Now, before we go into it, though, I'd 0:13:23.520 --> 0:13:26.439 like to enjoy one of the following lovingly crafted advertisements 0:13:26.440 --> 0:13:28.480 that in no way seem out of place from anything 0:13:28.520 --> 0:13:44.680 I'm saying. And we're back, So to put it bluntly, 0:13:45.040 --> 0:13:48.320 open ai is an absolute dog of a company. On 0:13:48.360 --> 0:13:51.320 September twenty seventh, The New York Times reported that open 0:13:51.360 --> 0:13:54.120 ai would lose five billion dollars in twenty twenty four, 0:13:54.240 --> 0:13:56.720 a number which the Information had estimated back in July, 0:13:57.240 --> 0:13:59.280 and the open Ai expected to raise the price of 0:13:59.320 --> 0:14:02.400 chat GPT plus its premium product to twenty two dollars 0:14:02.440 --> 0:14:04.240 a month by the end of twenty twenty four and 0:14:04.280 --> 0:14:09.600 are remarkable forty four dollars a month by twenty thirty. Interestingly, 0:14:09.920 --> 0:14:13.720 and worryingly, the Times also confirmed another hypothesis of mine 0:14:14.000 --> 0:14:17.440 that and I quote fundraising material also signaled that open 0:14:17.440 --> 0:14:19.760 ai would need to continue to raise money over the 0:14:19.800 --> 0:14:22.400 next year because its expenses grew in tandem with the 0:14:22.480 --> 0:14:25.920 number of people using its products. In simpler terms, open 0:14:25.960 --> 0:14:28.440 ai may have raised six point five or six point 0:14:28.520 --> 0:14:31.880 six billion in funding literally a week ago, but they 0:14:31.920 --> 0:14:34.880 need to raise more, probably the same amount or more 0:14:35.440 --> 0:14:38.120 I'm going to say, within the next six months. And 0:14:38.160 --> 0:14:40.960 The Times also reports that open ai is making internal 0:14:41.000 --> 0:14:44.560 revenue estimates I would describe as this is a technical term, 0:14:44.680 --> 0:14:50.640 absolutely fucking ridiculous. Open aiy's monthly revenue hit three hundred 0:14:50.720 --> 0:14:53.480 million dollars in August, and the company expects to make 0:14:53.520 --> 0:14:57.040 three point seven billion dollars this year. The company will, 0:14:57.040 --> 0:14:59.560 as mentioned, loose five billion dollars anyway, and I'm going 0:14:59.600 --> 0:15:02.720 to keep minding you of that. Yeah. Open ai says 0:15:02.720 --> 0:15:05.880 that it expects to make eleven point six billion dollars 0:15:05.920 --> 0:15:09.200 in twenty twenty five and an astonishing one hundred billion 0:15:09.200 --> 0:15:11.520 dollars by twenty twenty nine, a statement that is so 0:15:11.600 --> 0:15:14.200 egregious that I'm surprised it's not some kind of financial 0:15:14.240 --> 0:15:17.360 crime to say it. For some context, by the way, 0:15:17.600 --> 0:15:20.360 Microsoft makes about two hundred and fifty billion dollars a year, 0:15:20.400 --> 0:15:23.000 Google about three hundred billion dollars a year, and Apple 0:15:23.000 --> 0:15:26.040 about four hundred billion dollars a year. So yeah, I 0:15:26.040 --> 0:15:28.840 guess by twenty thirty that's how big open ai. Well 0:15:28.880 --> 0:15:33.120 you've well, you fucking talk. It drives me insane. And also, 0:15:33.160 --> 0:15:35.600 by the way, all of those companies make more money 0:15:35.600 --> 0:15:39.280 than they spend. Open ai, on the other hand, spends 0:15:39.320 --> 0:15:42.080 two dollars and thirty five cents to make one dollar. 0:15:42.360 --> 0:15:44.840 If you remember one thing from this podcast, it's that 0:15:45.640 --> 0:15:49.040 every dollar they make they have to spend two dollars 0:15:49.120 --> 0:15:53.400 and thirty five cents to get. It drives me insane. People, 0:15:54.160 --> 0:15:56.360 we don't have to live in these conditions. How do 0:15:56.440 --> 0:15:58.360 I do this? How do I get to burn money? 0:15:58.400 --> 0:16:02.840 What the no? Seriously, open ai loses money on every 0:16:03.080 --> 0:16:07.080 single transaction, every single time that somebody uses chet, GPT 0:16:07.160 --> 0:16:10.680 or plus or connects one of their models, and while 0:16:10.720 --> 0:16:14.040 it might make money selling premium subscriptions, I severely doubt 0:16:14.120 --> 0:16:17.320 those subscriptions are turning a profit and they I really 0:16:17.320 --> 0:16:20.520 do think losing money even on their power users. And 0:16:20.600 --> 0:16:23.160 as I'll get into the next episode, I believe there's 0:16:23.160 --> 0:16:26.640 also a SUBPRIMEI crisis brewing because open ai is API 0:16:26.720 --> 0:16:30.840 services which let people integrate their models into products, are 0:16:30.880 --> 0:16:34.280 currently priced to attract customers in scale, and increasing these 0:16:34.280 --> 0:16:36.840 prices to match the actual cost will likely make this 0:16:36.920 --> 0:16:40.600 product unsustainable for many businesses currently relying on these discount 0:16:40.640 --> 0:16:43.520 of rates, and that's if they have any usage. But 0:16:43.600 --> 0:16:46.640 I'll also get to that later. Now, if you anything 0:16:46.800 --> 0:16:49.200 like me, you have somebody who's told you that open 0:16:49.240 --> 0:16:51.520 ai is a growth business and that it will just 0:16:51.800 --> 0:16:54.720 turn the knob to make it self profitable, much like 0:16:54.760 --> 0:16:57.920 Amazon Web Services, which is Amazon's cloud computing product, did 0:16:58.000 --> 0:17:01.240 years ago. Now, I just want to be clear. If 0:17:01.280 --> 0:17:03.800 you're listening and you think that you're just wrong, you're 0:17:03.800 --> 0:17:06.639 just completely wrong. You live in a fancyland. Wake the 0:17:06.680 --> 0:17:08.760 fuck up. I'm sick and tired of hearing this crap. 0:17:09.160 --> 0:17:13.600 Open ai is nothing like Amazon Web Services. First of all, 0:17:13.760 --> 0:17:18.399 open ai owns none of its infrastructure as everything everything 0:17:18.520 --> 0:17:22.679 is run on Microsoft's as Your Cloud. Secondly, Microsoft, also 0:17:22.800 --> 0:17:25.760 as part of an open ai funding around from twenty nineteen, 0:17:26.040 --> 0:17:30.400 has full access to all of open AI's pre agi research, 0:17:30.440 --> 0:17:32.440 which is all of it, by the way, and full 0:17:32.480 --> 0:17:35.160 license to sell and integrate all of their technology. They 0:17:35.160 --> 0:17:40.720 have complete free reign over open aiy's intellectual property. Thirdly, 0:17:40.960 --> 0:17:44.000 Amazon Web Services did not immediately start with a deficit 0:17:44.080 --> 0:17:46.600 so great that it required raising more money than's ever 0:17:46.640 --> 0:17:49.840 been raised in the history of tech. Where still open 0:17:49.880 --> 0:17:52.920 AI's technology doesn't get cheaper as it scales. In fact, 0:17:53.040 --> 0:17:57.560 it does the opposite. This is not like other businesses. 0:17:57.560 --> 0:18:00.719 You want to compare this to stop saying this to people. 0:18:00.920 --> 0:18:04.439 It's ridiculous. And I know why people say this, and 0:18:04.480 --> 0:18:06.439 I am ranting, and you just have to forgive me. 0:18:06.840 --> 0:18:09.359 I know that people don't want to believe that this 0:18:09.560 --> 0:18:12.280 much money can be wrong. It can be. It is 0:18:12.320 --> 0:18:15.520 currently being wrong. You are watching and listening to it 0:18:15.720 --> 0:18:18.639 being wrong. This is not like Google Cloud. It is 0:18:18.680 --> 0:18:20.600 not like the early days of the Internet. It is 0:18:20.600 --> 0:18:24.560 not like Amazon Web Services. It's nothing like them at all. 0:18:25.320 --> 0:18:29.240 There is no historical example like this, not even Uber 0:18:29.600 --> 0:18:32.520 Uber at least had a business model. In the worst 0:18:32.600 --> 0:18:35.080 year of its life, Uber lost I think six point 0:18:35.119 --> 0:18:38.320 six billion dollars or something like that. So about one 0:18:38.320 --> 0:18:40.720 point six billion dollars in open ai lost in twenty 0:18:40.760 --> 0:18:43.280 twenty four. Right, you want to know what year that was? 0:18:43.480 --> 0:18:47.280 Twenty twenty the year when people couldn't go outside, and 0:18:47.320 --> 0:18:49.680 there are many years after they probably should have stayed indoors. Two, 0:18:49.800 --> 0:18:52.919 But that's a different podcast. The point I'm making is, 0:18:53.119 --> 0:18:55.080 if you've got a comfortable little thing in your soul 0:18:55.119 --> 0:18:57.480 that's saying this is just like things I've seen before, 0:18:57.560 --> 0:19:01.119 they'll turn it around. They absolutely And now I'm going 0:19:01.160 --> 0:19:04.199 to break down exactly how untenable open ai is with 0:19:04.320 --> 0:19:09.000 a few statements for starters. Open ai is trying to 0:19:09.040 --> 0:19:11.600 hit eleven point six billion dollars of revenue by the 0:19:11.680 --> 0:19:13.720 end of twenty twenty five, and to do that it 0:19:13.720 --> 0:19:16.560 will have to triple its revenue and its current cost 0:19:16.640 --> 0:19:19.040 of revenue, so two thirty five to make a dollar, 0:19:19.400 --> 0:19:22.520 open ai will spend twenty seven billion dollars to hit 0:19:22.560 --> 0:19:26.280 eleven point six billion dollars, even if open ai somehow 0:19:26.320 --> 0:19:29.080 harves their costs, and by the way, in the next 0:19:29.160 --> 0:19:30.760 year they're going to be building a brand new model, 0:19:30.760 --> 0:19:32.800 which will cost them an absolute shit ton of money. 0:19:33.000 --> 0:19:35.440 They will still lose two billion dollars in twenty twenty 0:19:35.520 --> 0:19:39.040 five to hit these revenues. However, I must be clear, 0:19:39.720 --> 0:19:42.639 these costs are going up. They need this company to 0:19:42.640 --> 0:19:45.440 grow by three hundred percent and the only way they're 0:19:45.440 --> 0:19:47.960 going to do that is by making a new model. 0:19:48.480 --> 0:19:50.960 And the only thing that's really growing at this company 0:19:51.000 --> 0:19:53.720 is they're free users, and those free users lose the 0:19:53.840 --> 0:19:57.760 money every single time they use it. And even if 0:19:57.800 --> 0:20:00.800 they added a two dollars price increase to chat GPT 0:20:00.880 --> 0:20:04.560 plus and a similar price on their business plans like 0:20:04.640 --> 0:20:08.120 teams and enterprise, these things aren't really going to significantly 0:20:08.200 --> 0:20:10.720 move the needle without also adding a bunch of growth 0:20:10.720 --> 0:20:14.480 on top. An open Aiy's latest GPT four model it 0:20:14.520 --> 0:20:17.320 cost them one hundred million dollars to train, and more 0:20:17.320 --> 0:20:20.159 complex models like the Orian one they're allegedly coming up 0:20:20.200 --> 0:20:22.320 with and I imagine oh one which will get to 0:20:22.440 --> 0:20:24.960 next episode, well, they're going to cost them an absolute 0:20:25.000 --> 0:20:27.760 shit ter more to build. And also, by the way, 0:20:27.800 --> 0:20:30.800 the information estimated back in July the open AI's training 0:20:30.800 --> 0:20:33.720 costs would be about three billion dollars in twenty twenty four. 0:20:34.080 --> 0:20:37.760 The costs are not coming down. People. Now here's another problem. 0:20:38.160 --> 0:20:41.040 Buzzy tech companies need exciting things to get investors in, 0:20:41.080 --> 0:20:44.640 customers jazzed and opening eye. Well, they haven't really had 0:20:44.720 --> 0:20:49.000 anything exciting or truly important since the launch of GPT 0:20:49.160 --> 0:20:52.720 three point five, and even their latest reasoning model, like 0:20:52.800 --> 0:20:54.479 I said, I'll get to it next episode, it's not 0:20:54.480 --> 0:20:58.400 been particularly impressive. That model, by the way, is much 0:20:58.440 --> 0:21:00.679 more expensive to run and uses something called chain of 0:21:00.760 --> 0:21:04.800 thought reasoning, which I'll get into. But nevertheless, it takes 0:21:04.840 --> 0:21:07.760 more power to run because they're doing way more calculations 0:21:08.400 --> 0:21:11.000 and they didn't even have a use case when they 0:21:11.000 --> 0:21:14.119 announced it. What is going on with this company? But 0:21:14.240 --> 0:21:19.160 putting that aside, open AI's products are also becoming increasingly commoditized, 0:21:19.200 --> 0:21:22.400 with Google, Meta, Amazon, and even Microsoft building their own 0:21:22.480 --> 0:21:25.800 large language models and other models to compete. Where still 0:21:25.920 --> 0:21:29.480 these models are using effectively identical training data, which they're 0:21:29.520 --> 0:21:32.640 also running out of, which makes their outputs and by extension, 0:21:32.680 --> 0:21:36.920