WEBVTT - James van Geelen on His Viral AI Doom Scenario 0:00:02.720 --> 0:00:13.960 Bloomberg Audio Studios, Podcasts, radio News. 0:00:18.480 --> 0:00:21.280 Hello and welcome to another episode of The Odd Lots Podcast. 0:00:21.400 --> 0:00:22.760 I'm Tracy Alloway. 0:00:22.480 --> 0:00:23.599 And I'm Joe Wisenthal. 0:00:23.920 --> 0:00:25.320 Joe, we're in the media business. 0:00:25.320 --> 0:00:26.360 That's right, that's right. 0:00:26.520 --> 0:00:31.040 Have you ever had an article go viral unexpectedly viral? 0:00:31.720 --> 0:00:31.960 Yeah? 0:00:31.960 --> 0:00:35.680 I can't like trying to like remember like specifics, but yes, 0:00:35.800 --> 0:00:39.599 And it's one of those things typically where you're like 0:00:39.680 --> 0:00:41.760 really excited, like a lot of people are, you know, 0:00:41.800 --> 0:00:43.680 there's getting a lot of traction. Cool people are talking 0:00:43.680 --> 0:00:46.040 about this, and then it goes like multiple orders of 0:00:46.080 --> 0:00:48.400 magnitude bigger, and you're like, oh, this is like super 0:00:48.440 --> 0:00:51.680 weird and no context for what this is, and you're 0:00:51.760 --> 0:00:53.360 like sort of want to hide in your home and 0:00:53.400 --> 0:00:56.080 like close the laptop because then you sort of like 0:00:56.080 --> 0:00:57.640 make it all go away and stuff like that. 0:00:57.920 --> 0:01:00.480 Yeah, it's kind of like once you release it into 0:01:00.520 --> 0:01:02.639 the world, you don't actually have a lot of control 0:01:02.800 --> 0:01:04.760 over how people use it. And I think back to 0:01:05.080 --> 0:01:08.280 I wrote a piece about some investors trying to revive 0:01:08.400 --> 0:01:14.120 claims on Chinese imperial bonds, like antique Chinese imperial debt 0:01:14.200 --> 0:01:18.920 from the early nineteen hundreds and somehow this went absolutely 0:01:19.000 --> 0:01:22.080 viral in Hong Kong at the time of the pro 0:01:22.120 --> 0:01:25.440 democracy protest. So I would walk down the street and 0:01:25.480 --> 0:01:28.759 I would see these homemade banners that people had created 0:01:28.880 --> 0:01:33.480 saying that China owes the US like twenty billion in 0:01:33.560 --> 0:01:36.080 payments on old debts, And it was just so real, 0:01:37.120 --> 0:01:40.920 absolutely surreal, and like completely unexpected, because you wouldn't think 0:01:41.240 --> 0:01:44.039 that some like intricate debt story was suddenly going to 0:01:44.040 --> 0:01:46.480 become a pro democracy protest slogan. 0:01:46.959 --> 0:01:49.160 But the world works in mysterious ways. 0:01:49.480 --> 0:01:51.840 And speaking of the world working in mysterious ways, there 0:01:51.880 --> 0:01:55.240 is something that went viral this week. We are recording 0:01:55.360 --> 0:01:58.880 on February twenty seventh, and if you haven't heard of 0:01:58.880 --> 0:02:02.720 this particular thing, you have probably been living under the 0:02:02.800 --> 0:02:03.680 proverbial rock. 0:02:04.160 --> 0:02:08.520 Right So past Oddlot's guest James Van Galen, a co 0:02:08.600 --> 0:02:12.960 author to Peace on his sub stack Treny Research, talk 0:02:13.000 --> 0:02:15.720 about a potential AI doom scenario, which a lot of 0:02:15.800 --> 0:02:18.160 people talk about, and there's been a lot of talk 0:02:18.160 --> 0:02:22.320 about mass white color displacement as a possible thing that 0:02:22.360 --> 0:02:25.120 could happen as AI gets adopted, et cetera. But you know, 0:02:25.520 --> 0:02:28.399 we know that the market's been very skittish about this specifically, 0:02:28.440 --> 0:02:30.960 we've been seeing the software stock sell off all year, 0:02:30.960 --> 0:02:33.560 which we've talked about plenty on the podcast, and some 0:02:33.600 --> 0:02:36.640 of the private insurers and all this, and something about 0:02:36.639 --> 0:02:40.440 this moment and this particular piece. I think it came 0:02:40.480 --> 0:02:43.640 out on Sunday. Last Sunday landed with a sort of 0:02:43.680 --> 0:02:46.800 like unbelievable thud, and so it evidently started moving markets 0:02:46.840 --> 0:02:49.560 on Monday and then throughout the week. And this is 0:02:49.560 --> 0:02:52.120 the part that really flabbergassed me, was you see like 0:02:52.240 --> 0:02:57.519 all these banks and every economists, etcetera, like weighing in 0:02:57.639 --> 0:03:01.119 and many of the very critical and like securities which 0:03:01.120 --> 0:03:03.560 I didn't even know they like publish stuff because that's 0:03:03.560 --> 0:03:06.160 just a market maker, like they put out all this stuff. 0:03:06.200 --> 0:03:08.760 So I responding to it or trying to take it out. 0:03:08.800 --> 0:03:11.520 It was a as a market story and a media story, 0:03:11.840 --> 0:03:12.519 a wild week. 0:03:12.919 --> 0:03:17.200 It has become the discourse dujore. There's actually a prediction. 0:03:16.880 --> 0:03:19.120 Market on it, which you were telling me about a 0:03:19.120 --> 0:03:19.880 few minutes ago. 0:03:19.960 --> 0:03:24.080 Like this thing has just become much bigger than the 0:03:24.120 --> 0:03:27.440 initial substack, which to me again says much more about 0:03:27.440 --> 0:03:30.360 the nervousness of the market and how little anyone actually 0:03:30.400 --> 0:03:33.080 knows about how AI is going to unfold at the 0:03:33.080 --> 0:03:35.440 moment that people are so keen to just like latch 0:03:35.560 --> 0:03:37.480 onto any scenario that comes out. 0:03:37.600 --> 0:03:39.960 I get these notes from like cell side or research 0:03:39.960 --> 0:03:41.960 shops and they're like, client has been asking us about 0:03:41.960 --> 0:03:45.120 the Satrini scenario, and it's just like, wow, this is wild, 0:03:45.280 --> 0:03:46.320 Like it's really like. 0:03:46.320 --> 0:03:46.640 A s right. 0:03:46.680 --> 0:03:49.160 People calling up Capital Economics being like I manage a 0:03:49.200 --> 0:03:51.520 portfolio of one hundred billion and I am concerned about 0:03:51.520 --> 0:03:54.640 a substack. Okay, well we should talk to the author 0:03:54.760 --> 0:03:56.640 of the substack, And as you said, we've had them 0:03:56.640 --> 0:03:59.760 on a number of times before, often talking about AI. 0:04:00.040 --> 0:04:02.520 It is, of course James fan Kielan, the founder of 0:04:02.560 --> 0:04:05.680 Satrini Research. So James, thanks so much for coming back 0:04:05.720 --> 0:04:06.280 on the podcast. 0:04:06.360 --> 0:04:07.080 Thanks for having me. 0:04:07.520 --> 0:04:11.760 Why don't we start with what Satrini Research actually is 0:04:12.640 --> 0:04:14.600 and what it is that you actually do in some 0:04:14.640 --> 0:04:18.040 of your other enterprises, because I think this has become 0:04:18.120 --> 0:04:21.479 also a source of confusion or at least interest for 0:04:21.880 --> 0:04:22.840 people who are reading this. 0:04:23.360 --> 0:04:27.599 Satrini Research is a pure investment research firm. We focus 0:04:27.680 --> 0:04:32.560 primarily on thematic equity and macro research. The progression of 0:04:32.600 --> 0:04:35.760 it was. I started it as a newsletter just speaking 0:04:35.839 --> 0:04:40.719 about stocks and bonds and whatever else. And as we 0:04:40.839 --> 0:04:43.040 had a kind of string of good calls, which you 0:04:43.120 --> 0:04:45.560 were kind enough to have us on with the GLP 0:04:45.680 --> 0:04:47.760 one early July twenty twenty three, I think. 0:04:47.720 --> 0:04:48.600 Yeah, that was a great call. 0:04:48.720 --> 0:04:51.400 Yeah, And the first piece we ever published was a 0:04:51.400 --> 0:04:54.200 piece that was very bullish on the AI infrastructure complex. 0:04:54.600 --> 0:04:57.680 So that's been an area that AI robotics has been 0:04:57.680 --> 0:04:59.880 a big area for us. In terms of thematic equity. 0:05:00.440 --> 0:05:03.560 We've kind of covered this winding road of bottlenecks in 0:05:03.680 --> 0:05:07.800 terms of optics, memory power, whatever else you can possibly 0:05:08.160 --> 0:05:11.520 allude to. We've probably covered from a what stories are 0:05:11.520 --> 0:05:14.400 people telling about the movements that are going on in stocks. 0:05:14.440 --> 0:05:15.920 I remember the last time that I was on odd 0:05:15.920 --> 0:05:19.640 lots it was about this massive Stargate data center build out. Yeah, 0:05:19.680 --> 0:05:22.359 and Joe was very surprised to see that Caterpillar was 0:05:22.720 --> 0:05:24.880 and I think very happy that the old economy was 0:05:24.880 --> 0:05:25.240 getting a. 0:05:26.880 --> 0:05:28.920 He's an old economy standards. 0:05:28.920 --> 0:05:31.440 And really that's what we've been doing for the past 0:05:31.440 --> 0:05:34.560 three years. I've built out the team, and this piece 0:05:34.680 --> 0:05:38.599 very much was just a response to what the market 0:05:38.600 --> 0:05:41.599 has done here today, which is bonds of ra allied 0:05:42.080 --> 0:05:44.839 software companies have gotten sold off, a lot of fintech 0:05:44.839 --> 0:05:48.039 companies have gotten sold off, Private equity has sold off, 0:05:48.360 --> 0:05:51.600 and we're always kind of looking for the cohesive narrative 0:05:51.720 --> 0:05:55.600 that can connect disparate market moves, and the pieces co 0:05:55.680 --> 0:05:58.880 author all up posed to me a question which was 0:05:59.360 --> 0:06:02.599 we've been folkocused on the bullishness surrounding AI infrastructure for 0:06:02.640 --> 0:06:06.360 a while and it's translated into this capability curve that 0:06:06.680 --> 0:06:09.680 is moving a lot faster than anyone could expect. If 0:06:09.720 --> 0:06:14.800 you imagine this exponential anologulorhythmic chart, it's just a diagonal line. 0:06:14.800 --> 0:06:16.920 It goes up into the right. People have been trying 0:06:16.960 --> 0:06:19.640 to put stigmoids or kind of level that curve off 0:06:19.680 --> 0:06:22.479 for a long time and it hasn't. So we basically 0:06:22.600 --> 0:06:25.919 drew that line out and said, what could be the 0:06:25.920 --> 0:06:29.640 implications of this happening. It's a scenario which we would 0:06:29.680 --> 0:06:34.800 ascribe maybe ten to fifteen percent towards and it comes 0:06:34.800 --> 0:06:37.560 from a place of that. Everybody talks about equity markets 0:06:37.560 --> 0:06:40.080 being forward looking, but really a lot more of what 0:06:40.120 --> 0:06:45.000 you see as people justifying historical moves with new narratives 0:06:45.000 --> 0:06:47.800 that they come up with afterwards. Very little of it 0:06:47.839 --> 0:06:51.080 is driven by let me think of potential future outcomes 0:06:51.920 --> 0:06:55.120 as an investor, which was the audience that this was 0:06:55.160 --> 0:06:57.760 meant to go out to. I feel a lot more 0:06:57.839 --> 0:07:00.919 comfortable when I can envision the bulk, the barecase, the 0:07:00.960 --> 0:07:04.159 base case, and the most uncomfortable that you can be 0:07:04.160 --> 0:07:07.040 as investors when you can't see the barecase at all. 0:07:07.200 --> 0:07:10.240 So every time that we get into a market that's 0:07:10.240 --> 0:07:12.840 similar to this, people start asking what if this time 0:07:12.920 --> 0:07:16.320 is different? And I guess the thing that this piece 0:07:16.400 --> 0:07:19.160 did differently was it asked what if this time is different? 0:07:19.200 --> 0:07:22.560 But not so much in a Skehiynex and Micron are 0:07:22.560 --> 0:07:25.040 going from price to book to price to earnings, but 0:07:25.200 --> 0:07:27.560 in a way where what if this time is different? 0:07:27.600 --> 0:07:31.120 Where the period of transition has to respond to a 0:07:31.280 --> 0:07:37.880 very very fast, accelerating capability curve, and you start from 0:07:37.920 --> 0:07:41.960 a place where there's a strong kind of historical precedent 0:07:42.440 --> 0:07:44.960 for the past century or or two centuries. Every time 0:07:44.960 --> 0:07:48.160 you've had a technological revolution, it's been great, it's been awesome. 0:07:48.480 --> 0:07:51.320 And you see that when you go from ninety five 0:07:51.320 --> 0:07:54.040 percent of the population working in agriculture to five percent 0:07:54.080 --> 0:07:56.480 of the population and you create all these amazing jobs, 0:07:56.840 --> 0:08:00.120 but it happens over a period of fifty years. Now 0:08:00.480 --> 0:08:03.160 we have this capability curve where you go from two 0:08:03.200 --> 0:08:05.760 minutes agents are capable of two minutes of autonomy on 0:08:05.880 --> 0:08:08.960 intellectually complex tasks, and now depending on who you ask, 0:08:08.960 --> 0:08:11.920 it's eight to sixteen hours. And that's happened in two years. 0:08:12.160 --> 0:08:15.000 That is an exponential curve. What happens when we get 0:08:15.040 --> 0:08:17.360 to multi day. You know what happens when we get 0:08:17.360 --> 0:08:20.080 to multi week. And really the core of this is 0:08:20.160 --> 0:08:23.640 if this capability curve continues being as fast an expansion 0:08:23.640 --> 0:08:26.640 as it is, what does the world look like. There 0:08:26.680 --> 0:08:28.840 are a lot of very good reasons why that capability 0:08:28.840 --> 0:08:31.120 curve could level off, but that is the core of 0:08:31.160 --> 0:08:31.640 the argument. 0:08:31.800 --> 0:08:34.040 I do think that's just like an important sort of 0:08:34.400 --> 0:08:38.400 level set for people here, which is that the progress 0:08:38.480 --> 0:08:42.320 that we've seen since chet GPT came out whenever that 0:08:42.559 --> 0:08:45.960 was late twenty twenty two, has exceeded all of the 0:08:46.000 --> 0:08:49.560 expectations of everyone who's working on it at the time, 0:08:49.640 --> 0:08:52.320 including the people who are in the space and the 0:08:52.360 --> 0:08:54.840 most bullish and like the true believers. And there are 0:08:54.920 --> 0:08:57.560 various like measures and stuff, but you know, you mentioned 0:08:57.600 --> 0:09:00.000 the length of time you know that it could replic 0:09:00.320 --> 0:09:03.560 a human focused on stuff, like all the people like 0:09:03.600 --> 0:09:06.040 they made like these bets, right, and there were even 0:09:06.080 --> 0:09:09.719 prediction markets on their capabilities, and so like, as you say, like, 0:09:09.880 --> 0:09:13.199 it seems very plausible that the gains will level out 0:09:13.280 --> 0:09:18.200 in some way, or that perhaps simple computer tasks don't 0:09:18.240 --> 0:09:21.160 actually replace a lot of white color work because there's 0:09:21.200 --> 0:09:22.920 more to white color work than what could be done 0:09:22.960 --> 0:09:25.920 on a computer, including personality and all kinds of stuff. 0:09:26.440 --> 0:09:28.600 All of that seems very plausible and I probably even 0:09:28.720 --> 0:09:30.920 buy some of that. But this point that you make, 0:09:31.080 --> 0:09:34.920 it's like, yeah, sure, but it is still proving very fast. 0:09:34.600 --> 0:09:39.560 And it's something where the overall trend of the cost 0:09:39.679 --> 0:09:44.640 of inference per cognitive task has gone down so significantly, 0:09:44.800 --> 0:09:48.360 maybe depending on the forecast, ten to thirty times over 0:09:48.400 --> 0:09:52.120 the past year, and a task that was uneconomical in 0:09:52.320 --> 0:09:55.280 the first quarter twenty six might cross that threshold in 0:09:55.320 --> 0:09:58.880 the third quarter. And the other interesting thing is this 0:09:59.000 --> 0:10:03.000 capability gap where AI is capable of a lot of 0:10:03.040 --> 0:10:05.719 things and a lot of people don't know that it's 0:10:05.720 --> 0:10:08.160 capable of that, right, So is it about the capability 0:10:08.160 --> 0:10:10.360 improving or is it about people becoming more familiar with 0:10:10.400 --> 0:10:14.960 that and as AI infrastructure. It's been a great trade 0:10:15.040 --> 0:10:16.920 and it continues to stay tight, and I think the 0:10:16.960 --> 0:10:20.600 best rebuttal to this piece has been well, I think 0:10:21.280 --> 0:10:23.800 Gavin Baker made this point, which is the world is 0:10:23.840 --> 0:10:27.600 short on Watson waivers, and that's true, absolutely true. But 0:10:28.080 --> 0:10:32.079 technological revolutions are volatile, right, Improvements come from places that 0:10:32.120 --> 0:10:35.320 you don't really expect them to and I think you 0:10:35.360 --> 0:10:38.480 can't fully underwrite the idea that there aren't algorithmic improvements 0:10:38.559 --> 0:10:42.520 or there aren't improvements to the compute infrastructure. So we 0:10:42.559 --> 0:10:45.840 should look at Okay, if this capability curve continues improving, 0:10:46.480 --> 0:10:50.400 what are the downstream impacts there? And has the financial 0:10:50.440 --> 0:10:52.760 system ever been stress tested for a scenario like this, 0:10:53.360 --> 0:10:55.520 Because even if it takes five years, even if it 0:10:55.559 --> 0:10:58.800 takes seven years, eventually we will get there. And that's 0:10:58.880 --> 0:11:00.840 not a bearish take, it's a very bullish take. I 0:11:00.840 --> 0:11:03.840 think that there will be great opportunities that arise because 0:11:03.840 --> 0:11:06.720 of AI, but that's not to say that there won't 0:11:06.760 --> 0:11:09.240 be a period of transition. And the faster that it comes, 0:11:09.280 --> 0:11:14.000 the more aggressive that transition is. And I think the 0:11:14.040 --> 0:11:17.080 point of The piece really was to get comfortable with 0:11:17.640 --> 0:11:20.880 what monitoring that looks like. And I'll just make the 0:11:20.920 --> 0:11:23.880 point that the piece also starts out with an SMP 0:11:24.040 --> 0:11:27.480 that goes to eight thousand, because AI infrastructure is a 0:11:27.559 --> 0:11:30.400 very bullish trade that makes up a lot of the index, 0:11:30.440 --> 0:11:33.800 and that's a very strong and very momentum having trade 0:11:33.880 --> 0:11:37.440 right now. And it ends with the reminder that it's 0:11:37.440 --> 0:11:40.679 still February twenty twenty six. But in the middle of it, 0:11:40.679 --> 0:11:44.000 it says, how do we kind of get comfortable with 0:11:44.520 --> 0:11:49.880 the non immediacy of the replacement If a company decides 0:11:50.120 --> 0:11:53.040 whether they're doing it because AI's gotten better or because 0:11:53.679 --> 0:11:56.040 the market likes it when they cut jobs. 0:11:55.920 --> 0:11:59.120 What is You're already through saw with the blog last night. 0:11:59.000 --> 0:12:01.280 And you can argue whether that's because of AI or 0:12:01.280 --> 0:12:05.080 whether that's because of over hiring during COVID. But Caines 0:12:05.160 --> 0:12:07.319 said that by the end of the century we'd have 0:12:07.440 --> 0:12:10.760 a fifteen hour work week, and he was wrong, and 0:12:11.040 --> 0:12:12.480 there's a lot of exit you have to kind of 0:12:12.480 --> 0:12:15.839 look at why he was wrong. There are a few explanations. 0:12:15.920 --> 0:12:18.320 David Graeber says that we just kind of created all 0:12:18.360 --> 0:12:20.680 these bulk jobs. This is the title of the book. 0:12:20.679 --> 0:12:21.720 I'm not cursing. 0:12:21.800 --> 0:12:23.360 People have said worse on this podcast. 0:12:24.200 --> 0:12:27.200 The other explanation is that human wants and desires you 0:12:27.200 --> 0:12:30.760 can't really model for, and we will create whatever we 0:12:30.840 --> 0:12:34.720 need to fill that. At the same time, that required 0:12:35.320 --> 0:12:38.520 mechanisms by which humans kind of are involved in the 0:12:38.559 --> 0:12:41.960 process of making those machines better. It's kind of not 0:12:42.000 --> 0:12:47.640 necessarily in every scenario concurrent with the idea of a 0:12:47.720 --> 0:12:51.360 piece of software that has the ability for a recursive improvement. 0:12:52.400 --> 0:12:54.959 This isn't to say that tomorrow every single company in 0:12:55.040 --> 0:12:58.480 large enterprise goes out and replaces half their workforce, but 0:12:59.200 --> 0:13:01.680 you do have to take a holistic picture, which is 0:13:02.520 --> 0:13:05.240 everybody in venture capital has been talking about who's going 0:13:05.280 --> 0:13:07.400 to be the first one person unicorn because of a 0:13:07.440 --> 0:13:09.960 gendic AI. I don't know if we're there yet. I 0:13:09.960 --> 0:13:12.319 haven't really kept on top of that, but that does 0:13:12.360 --> 0:13:16.360 seem like something plausible to me, and I think one 0:13:16.360 --> 0:13:20.439 of the better lines of the Citadel Securities counter argument, yeah, 0:13:20.760 --> 0:13:25.199 was recursive capability doesn't imply recursive adoption. That's extremely true. 0:13:25.679 --> 0:13:28.840 The S curve framework, though, is kind of describing the 0:13:28.840 --> 0:13:32.240 wrong variable, and it's a variable that's really important when 0:13:32.280 --> 0:13:35.920 you don't just have incumbents adopting, but you have startups 0:13:35.920 --> 0:13:39.760 threatening and that variable is not necessarily breadth of adoption. 0:13:39.920 --> 0:13:43.720 It's intensity of adoption and capability of adoption. So you 0:13:43.840 --> 0:13:48.400 might have a flattening out S curve, and the seats 0:13:48.400 --> 0:13:50.840 that you've already enabled with these AI tools are just 0:13:50.920 --> 0:13:55.480 constantly getting better, and so that is you know. The 0:13:55.760 --> 0:13:57.920 other thing is the S curve is very kind of 0:13:58.880 --> 0:14:02.199 related to consumer adoption of new technologies. And what I 0:14:02.240 --> 0:14:04.920 would ask is was there an S curve for the 0:14:05.000 --> 0:14:09.880 adoption of spell check? Everybody already had a PC, everybody 0:14:09.920 --> 0:14:13.160 already had, you know, word processing software. It was kind 0:14:13.160 --> 0:14:14.839 of added as a feature. There are a lot of 0:14:14.920 --> 0:14:17.280 people in the world today that have no clue how 0:14:17.280 --> 0:14:20.640 to use SHATGPT, that are using AI every single day. 0:14:20.720 --> 0:14:23.360 It's probably what is going to recommend you this podcast. 0:14:23.640 --> 0:14:27.280 It's probably what is making these decisions of what items 0:14:27.280 --> 0:14:30.400 you see when you go on Amazon. So if these 0:14:30.520 --> 0:14:33.840 gentic capabilities are introduced as features to a technology that 0:14:33.880 --> 0:14:37.280 everyone has already adopted, you have to adjust your model 0:14:37.280 --> 0:14:37.480 for that. 0:14:53.560 --> 0:14:55.320 I have so many things to say about this, but 0:14:55.440 --> 0:14:58.560 first of all, there's something very dystopian about living in 0:14:58.600 --> 0:15:01.880 a world where like the upside is, well, we have 0:15:01.920 --> 0:15:04.480 a lot of bulk jobs in existence already, and so 0:15:04.760 --> 0:15:07.520 maybe some of those bold jobs will continue to exist 0:15:07.560 --> 0:15:10.400 even with AI. But the other thing is, like the 0:15:10.520 --> 0:15:14.960 self reinforcing nature of AI seems really important to me 0:15:15.120 --> 0:15:18.240 in the sense that, as you pointed out, James, like, 0:15:18.280 --> 0:15:20.560 it's not necessarily that people have to go out and 0:15:20.680 --> 0:15:25.040 find these new capabilities themselves. It's that the technology itself 0:15:25.080 --> 0:15:28.320 that they're may be already using just to substitute search 0:15:28.440 --> 0:15:32.040 or something like that, can do it on their behalf. 0:15:32.120 --> 0:15:34.480 And so you just get this feedback cycle where like 0:15:35.200 --> 0:15:38.760 one AI thing creates new AI things, and it just 0:15:38.880 --> 0:15:40.440 builds and builds on itself. 0:15:40.720 --> 0:15:42.800 I really would be remiss if I didn't say this again, 0:15:42.920 --> 0:15:44.840 which is a lesson that I've learned over the past 0:15:45.160 --> 0:15:48.000 five days, that you can put something in all caps, 0:15:48.080 --> 0:15:50.200 you can bould it, and people will still not read it. 0:15:50.400 --> 0:15:53.440 But maybe this is different because I'm speaking my base 0:15:53.520 --> 0:15:55.480 case is probably a lot closer to a lot of 0:15:55.520 --> 0:16:00.200 the people rebutting this article than the article itself. The 0:16:00.240 --> 0:16:04.560 point of this really was to explore what the bear 0:16:04.680 --> 0:16:07.640 case is if we continue to have a very bullish 0:16:07.960 --> 0:16:11.560 world in AI infrastructure. I think that any investor that 0:16:11.640 --> 0:16:15.200 reads it and thinks, and you know, disagrees with half 0:16:15.240 --> 0:16:17.680 of the things that we say, maybe agrees with half 0:16:17.720 --> 0:16:21.320 of it and forms a more nuanced understanding of what 0:16:21.440 --> 0:16:23.960 to watch out for. That's kind of our job. 0:16:24.240 --> 0:16:26.600 So this is important, and people who haven't read the 0:16:26.600 --> 0:16:29.480 piece should know that, Like right up front, you do 0:16:29.560 --> 0:16:31.440 say this, You say this piece is not a forecast. 0:16:31.840 --> 0:16:35.040 This is a possible scenario and how it could go, 0:16:35.320 --> 0:16:37.840 and we want to get into some of the details. 0:16:38.240 --> 0:16:41.320 But you know, one counter argument to sort of the 0:16:41.560 --> 0:16:45.880 idea of macro economic doom or financial crisis or whatever, 0:16:46.080 --> 0:16:50.080 is okay if you have AI and it's driving incredible 0:16:50.080 --> 0:16:54.480 productivity gains, If it's very disinflationary and so forth, if 0:16:54.600 --> 0:16:57.560 some people are becoming fabulously wealthy and part of this 0:16:57.560 --> 0:17:01.360 big redistribution that would happen, well, then the government has 0:17:01.400 --> 0:17:04.000 a lot more fiscal capacity to stabilize this. Right, then 0:17:04.000 --> 0:17:05.760 the government can spend a lot of money. Rates have 0:17:05.800 --> 0:17:09.800 come down, they can counteract the disinflation, not totally, unlike 0:17:09.840 --> 0:17:13.119 perhaps COVID would be like a great example, but it 0:17:13.160 --> 0:17:15.320 strikes me as like, well, if we're ever going to 0:17:15.359 --> 0:17:18.520 have a government that's thinking about these things proactively, that 0:17:18.720 --> 0:17:21.080 strikes me as a good reason to write them out. 0:17:21.119 --> 0:17:24.399 And it's notable like many of the executives at the 0:17:24.400 --> 0:17:28.200 top AI labs they talk about exactly this. In fact, 0:17:28.240 --> 0:17:31.040 it seems like they're pleading almost with the government to 0:17:31.480 --> 0:17:34.040 take this war seriously, because if we're going to have 0:17:34.119 --> 0:17:37.840 this big disruption and redistribution, we're going to have to 0:17:37.880 --> 0:17:41.080 start thinking about what are the fiscal mechanisms to counter 0:17:41.119 --> 0:17:42.040 it out. 0:17:41.640 --> 0:17:45.520 One hundred percent, I think that it's something where it's 0:17:45.640 --> 0:17:48.280 perfectly fine and good to say that the government will 0:17:48.359 --> 0:17:51.879 be able to deal with it, but it's probably better 0:17:51.960 --> 0:17:55.520 to formulate a framework in which the government is more 0:17:55.600 --> 0:17:57.879 able to do that. And in order to do that, 0:17:57.920 --> 0:17:59.639 you kind of have to have an idea of what 0:17:59.680 --> 0:18:02.880 to keep track of, and I can say that in 0:18:03.720 --> 0:18:07.720 the discourse that I've seen, I don't think that there's 0:18:07.840 --> 0:18:12.080 a very strong kind of data collection on this. Specifically, 0:18:12.440 --> 0:18:15.640 one of the big rebuttals has been that software job 0:18:15.680 --> 0:18:18.640 postings have gone up eleven percent year over year. Those 0:18:18.720 --> 0:18:22.640 job postings include AI and machine learning engineers, so you're 0:18:22.680 --> 0:18:26.960 really seeing a composition shift where these new you know, 0:18:27.040 --> 0:18:31.119 AI engineers are coming in and they're creating software that 0:18:31.240 --> 0:18:35.120 will improve itself. And when it comes to the government response, 0:18:35.480 --> 0:18:38.840 Jolts doesn't really speak about composition. In my opinion, there's 0:18:38.840 --> 0:18:41.760 not a great amount of data on white collar specifically, 0:18:42.200 --> 0:18:45.400 And yeah, it was, it was. It was almost worrying 0:18:45.440 --> 0:18:49.199 in itself to see this reaction where we write this 0:18:49.320 --> 0:18:52.160 article that's kind of saying what I think most people 0:18:52.200 --> 0:18:55.760 are thinking. We're putting trillions of dollars at the white 0:18:55.760 --> 0:18:58.760 collar productivity machine, and oh that might, you know, have 0:18:58.840 --> 0:19:01.600 some level of disruption, and I get it. The thing 0:19:01.600 --> 0:19:03.800 that I'm very thankful to a lot of the rebuttals 0:19:03.800 --> 0:19:06.200 for is that they've reminded people that it's twenty twenty six, 0:19:06.440 --> 0:19:08.280 which we tried to do three times in the piece, 0:19:08.320 --> 0:19:11.760 but apparently we're not successful. Thank you to everyone that 0:19:11.600 --> 0:19:15.560 that made sure that this isn't like a spinout, crazy whatever. 0:19:16.119 --> 0:19:19.240 But the worrying side is, well, everyone seems very very 0:19:19.240 --> 0:19:22.080 comfortable that this is all going to be okay, and 0:19:22.160 --> 0:19:24.879 I think that that reasonably. I'm also a student of 0:19:24.920 --> 0:19:28.000 financial history. That reasonably comes from when you look back 0:19:28.040 --> 0:19:29.480 at the past and you say, well, we had this 0:19:29.520 --> 0:19:32.919 industrial revolution and it was amazing, and we've had mechanization 0:19:32.960 --> 0:19:34.960 and it was amazing, and we've had the Internet and 0:19:35.000 --> 0:19:37.359 it was amazing, and it created all these jobs that 0:19:37.359 --> 0:19:40.880 we couldn't have possibly foreseen beforehand. And you're looking at 0:19:40.880 --> 0:19:44.600 that from one hundred or more years in the future. 0:19:45.200 --> 0:19:48.040 We have the term Luddite because of the fact that 0:19:48.720 --> 0:19:55.440 the transition was so abrupt and marked that people were 0:19:55.440 --> 0:19:58.160 moved to physical violence. Right, we don't want that to happen. 0:19:58.400 --> 0:20:01.440 The transitions do occur in the faster that this happens. 0:20:01.760 --> 0:20:03.600 If this were going to happen over the next twenty 0:20:03.960 --> 0:20:06.560 or thirty years, fine, you know that that's going to 0:20:06.560 --> 0:20:09.280 be great. Everything's going to be awesome. I think that 0:20:09.320 --> 0:20:13.080 the real time frame is closer to five to fifteen, 0:20:13.560 --> 0:20:17.199 and obviously this piece extrapolates where it's three years. We 0:20:17.240 --> 0:20:20.520 should be prepared for anything, because the government isn't going 0:20:20.560 --> 0:20:24.520 to accurately forecast technological advancement, but they can accurately forecast 0:20:24.560 --> 0:20:26.439 what they should watch and what the best policy response 0:20:26.440 --> 0:20:26.800 would be. 0:20:27.040 --> 0:20:27.679 Yeah, this's the thing. 0:20:27.680 --> 0:20:30.040 The Luddites were like ultimately on the wrong side of 0:20:30.119 --> 0:20:33.440 history in terms of thinking that resistance to new technology 0:20:33.480 --> 0:20:35.680 would actually matter. But that doesn't mean that there wasn't 0:20:35.960 --> 0:20:39.680 major resistance and disruption on the way, that it. 0:20:39.520 --> 0:20:43.240 Wasn't absolutely awful. Yeah, no, exactly right from their perspective, 0:20:43.240 --> 0:20:44.400 from their lives exactly. 0:20:44.680 --> 0:20:48.159 You know, you mentioned software job openings still rising, and 0:20:48.160 --> 0:20:51.280 one of the reasons that's able to happen is because 0:20:51.320 --> 0:20:55.160 we still have a financial system that up until relatively recently, 0:20:55.200 --> 0:20:59.879 has been very comfortable with extending credit to software companies. 0:21:00.040 --> 0:21:04.040 And there's obviously a reflexivity between the financial system, the market, 0:21:04.320 --> 0:21:07.080 and the real economy. And you dig into that in 0:21:07.119 --> 0:21:09.240 your piece as well, And this is the part of 0:21:09.280 --> 0:21:11.960 it that I actually found the most interesting, where you 0:21:12.040 --> 0:21:16.879 describe how AI could actually and the disruptive effects of 0:21:16.920 --> 0:21:21.200 AI could actually end up becoming problematic, especially for private capital. 0:21:21.240 --> 0:21:23.840