Why AI Can’t Replace You with Cory Doctorow

This is a Headgun Podcast. Hello and welcome to Factually I'm Adam Conover. Thank you for joining me once again as I talk to an incredible thinker and get all those delicious juicy ideas out of their brain and into yours. This week we are bringing back one of the most legendary guests we ever have on this show. Fan favorite, favorite of mine, favorite of yours. And he is on the talk to us about AI. You know, the interesting thing about AI is that like many technologies, it's sold to us as being something that's going to make our lives better, right? It's going to help us do more. But the way it's being rolled out, doesn't it seem more like we are being asked to work for it? Like the tech companies instead of making things that empower us, now we're building a machine that we must obey or be punished. I mean, you can look at the way that the AI companies tell us we all need to get out of the way and prepare for a future where AI takes all of our jobs or we can see it in the way that companies like Amazon and Anthropic are buying up and ripping apart rare books to feed their AI models. Simply using humanities collective intelligence and recorded words to make their damn machines smarter and more powerful. I mean, even in my daily life, when I post on YouTube, I can't always tell if YouTube is a tool I'm using to bring audience to my ideas or if my ideas are actually just being fed into YouTube to help its algorithm become more powerful, to be the best algorithm it can be. The disorienting feeling to realize that maybe you're not in charge of your technology, maybe your technology is in charge of you. And at a time when more and more of us are feeling this way, I wanted to bring on one of my very favorite guests from the history of the show, one of the smartest tech writers and thinkers I know. His name, of course, is Cordy Doctorow. He's a prolific writer of novels and of nonfiction blog posts and honestly everywhere else across the internet. OG internet blogger, you might know him best for coining the term and concept in shittification, one of the most powerful ideas of the past couple years. Well, he is back with a brand new book and a brand new set of powerful ideas called The Reverse Centaur's Guide to Life After AI. This is an absolutely mind blowing conversation. I know you're going to love it. Before we get into it, I have a big announcement to make. We are launching as an exclusive feature for our Patreon subscribers, a new nonfiction book club where me and my incredible producer, Sam Roudman, one of the smartest, funniest people I know are going to be reading a new work of mind blowing nonfiction every couple of months and discussing it as a bonus exclusive podcast episode for our Patreon community. We are also going to be having a ton of other special guests on the show. We're going to discuss our thoughts about the book. We're going to have our community share, their thoughts we're going to bring on special guests. It's going to be a lot of fun and our first book is a doozy. It's Naomi Klein and Astra Taylor's new book, End Times Fascism and the Fight for a Living World. Naomi Klein is one of my favorite nonfiction authors. Her last book, Doppler Ganger, was one of my favorite books of the decade. I know this one is going to be a banger and I cannot wait to read it and discuss it with all of you. We would like to join us and get all of our other community features, bonus episodes of the podcast, every episode we put out ad free, head to patreon.com/atomconover. And now let's get to this week's interview with a wonderful Corey Doctoro. Corey, thank you so much for being on the show again. Adam, thank you for having me on the show again. It's a new month, so you have a new book out and what day that ends and why. It's just wonderful because whenever you have some new ideas out, it gives the rest of us something to talk about. It drives the conversation in all of us who are thinking critically about tech and our relationship with it. So tell me what you're thinking lately about AI. What is the reverse centaur from the title of your book? Yeah. Well, I'll answer both of those questions. One thing lately about AI, the thing that made me write this book was possibly the single most important miscalculation of my career, just that it got so sick of people insisting that I talked to them about AI, that I was like, I'm going to write a book, that explains why I think we should all talk and think less about AI. And of course, the punishment for that is now I have to talk about AI all the time. So reverse centaur, I think this is something that I'll be near and dear to your heart. It actually really relates to some of the work you did with the writer's strike. You will not be surprised to discover that this is not the first time that there's been a conflict between labor and capital when it comes to automation. And in the long history of automation, we've got some sort of theoretical ways of understanding it. And broadly, people who study this stuff will tell you that when labor drives automation, it is in service to improving quality. And when capital drives automation, it's in service to improving throughput. And that in particular, when capital has market power and they can make people accept worse goods, it's not just that they use automation to make more or something, they make automation to make more or something worse. Yeah. And they fire some workers and they produce something worse and they charge money for it and it sucks. And this brings us to centaurs and reverse centaurs. A centaur in automation theory is a person assisted by a machine. So like if you're using a spell checker, you're a centaur. If you are writing a bicycle, you even look like a centaur and you are a centaur. A reverse centaur is someone who is conscripted to serve as a peripheral to a machine. And we kind of know what this is like, right? So the metaphor here is that a centaur is like the head of a human, it's got all the discernment, it's got all the judgment, but you've got the body of the horse, it's strong and has endurance and fast. The reverse centaur is the reverse arrangement, right? The machine, the horse on top of the human, the horse is head on the human body, the bojack configuration. And this is the human taking orders from the machine, because there's some part of the machine, some part of the job the machine can do. And so the human is run at the maximum throughput that the machine can get out of them, right? The machine is being your boss here. And so the machine is faster than you and it's stronger than you and has more endurance than you. So it's not just using, it's generally it's using you up, right? It's not a coincidence that the most advanced warehouses the world has ever seen are Amazon warehouses and that they have three times the rate of injury of other warehouses, right? That's the consequence. And those are warehouses at which the people are in the service of robots, like the shelf comes up to you and all you are is a pair of hands, right? Yeah. And you know, no one's going to spend seven figures on automation at a warehouse and then go like, oh, let's run these machines slower than we need to. Yeah. You want to, you want to sweat that asset? So the reason this is relevant to AI is that there are these paradoxes in AI, don't tell us you've experienced them too, where you meet people who you know to be skilled practitioners of their trade and you know to be historically reliable narrators of their experience. And they go, oh my god, the stuff I'm doing with AI is amazing. And you know, it's kind of esoteric and you don't really understand, I know mathematicians who sit this, right? I am the son of a mathematician, which means that I know exactly how long qualify I am to assess when a mathematician tells me how good their job is and how good their work is, right? And then you meet people who are also skilled practitioners who are also historically reliable narrators. And they go, you would not believe the shit we're making with AI. Like, you know, I work in avionics. We are shoveling tech debt into the sky at a rate never before seen, never get on an airplane again. Yeah. And so here you have these two greatest people who like seem to be doing the same thing, they seem to be the same kinds of people, they seem to be using the same tools and they are these totally different experiences, how to understand it. And the answer is, they're not doing the same thing. One of them has had their boss fire 9/10 of their colleagues and said, use the AI to make up the difference, but sign your name at the bottom of every one of these outputs, really new, right? Daniel Davis calls that an accountability thing. And the other one, they're skilled practitioners. So they're grabbing an automation tool that has unreliable outputs and they're using it in a way that they deem, maximizes the reliability of it. It might be wrong, sometimes people are wrong, sometimes workers are wrong, but they're using it in a way that produces the best outputs they can. And they don't use it when they don't think it helps. And those people are excited. And I think it tells you that there are times when this collection of automation tools that in the absence of the bubble, we would just call plugins are useful and there's times when they're not useful. And what they aren't is so important that we should bet the world's economy on them and put all the workers in a woodchipper because it's only matter of time until the AI can do all of their jobs. >> Well, you have illuminated so much for me, I mean, just the dichotomy that you're talking about and people's experiences, I follow blogs that computer programmers, right? Like I have a couple of link sources where just computer engineers, their conversations with each other are still on blogs and I'll read their articles. And I have friends who work in startups and they're like, well, our programmers are doing so much, our engineers. I mean, it's going crazy, it's going great. And I'm like, yeah, but look at all these articles that all these other guys are writing about how all the techmen that it's creating and that we are destroying computer engineering as a discipline and da-da-da. Like, there's this really stark difference. And then, as you say, there's the people in my life who are using it to do interesting things, I think it's going to change everything. And then there's the people who, I literally took a friend out to dinner last night because she was just fired from her job at an enormous, one of the original tech companies that got real big. She was fired from her job where her job had been reduced over the last year to training an AI to do her job and other people's jobs. And she was like, I sold my soul to do this, and they laid me off anyway. And she was, you know, felt disgusting. She felt distraught over it. You know, we're going out for spaghetti and meatballs because she needs comfort food. >> Trying to your sorrows and carbs. >> Exactly. So, what you're describing is you've put such a wonderful, like, image on it that it is, which people get to be the brain that, you know, now have a more powerful body, and which people have been sort of chained in service to some other brain that isn't theirs, and feel underneath it. And my guess would be that just like, today on, on the whole, if we were to count up, it would be mostly people in the second category, like, on planet Earth. There'd be more of them than in the first category. Do you agree with that? >> Yeah, I think that's right. And I think that that is, you know, I've alluded already to this bubble, this investment bubble. I think that is an epithelon almond on another bubble, right? That the sums raised for this technology are vastly greater than it can ever return. And I think that there are economic and psychological reasons that that money was raised. And that when you understand that, you start to understand why, instead of getting, I don't know, like, Sandia National Labs, working with the IPCC to build an $80 billion supercomputer to use some deep learning techniques to improve weather prediction. We got, like, Elon Musk building the world's most efficient money furnace so that you can draw sonic with giant tits. >> Yeah. >> And so while the way that we use AI, the way AI has been rolled out, is historically contingent, and that it might have been different, it's also understandable. It's not mysterious how we got here. So I can give you an example from, I think, one of the most promising uses of AI to understand why it looks like the way it looks the way it does if you're interested. >> So you know, I think that I had cancer. And actually, the week this book came out in June, I was told by radiologists at the Kaiser Hospital on Sunset that I'm cancer-free. >> Wonderful. >> It's great news. So I've been spending a lot of time last couple of years. And to be clear, I was always asymptomatic. I wasn't particularly worried except in that kind of, you know, middle of the night, like, what if they're all wrong? And it's worse than they say it is. It is cancer after all, you know, they say it's 95 percent of the time. It'll be fine. What if it's the 5 percent? >> You know, but probably it was okay. And certainly, a lot of people are worse. But I just spent a lot of time thinking about radiology. And it seems pretty clear from the literature that there's times when AI spot solid mass tumors that radiologists miss. So you know the Kaiser Hospital down on Sunset where I got treated next to the big church of Scientology, right? It's Los Angeles. >> Yeah. >> And so like, imagine that right now, you know, it's whatever, it's like 845 there. Imagine that there's like a salesman for open AI. It's like dodging past a Scientologist who wants to give him a personality test. And running up to the CEO of Kaiser's office, and he's got this pitch, he says, look, right now you've got 10 radiologists working at this facility. They're averaging about $300,000 a year. Tell you what, you give us another million dollars a year. We're going to sell you a chatbot. Now right now, those radiologists, they do a hundred x-rays a day. That chatbot's going to sit on their shoulder and a couple times a day it's going to tap on the shoulder and it's going to go take another look at that one. And sometimes they're going to save a life. I'd be very happy with that pitch. No one is investing in opening AI because they think it's going to make radiology more expensive. Right? The way radiology is being pitched right now is that you have the salesman. He runs past the Scientologist. He goes up to the CEO's office. He says, look, you've got 10 radiologists on the strength, $300,000 a year, $3 million in total, cut their wages in half. Give $9 out of 10 of them, $9 out of 10 of them. That's going to free up $2.7 million. You keep half of that. We give the other half to Sam Altman. The remaining radiologists signs all the reports. They're just going to get a report every two seconds under their nose. And they have to decide really quickly whether the AI got it wrong. And if they if they miss it, then we blame them for the person who dies because they fucked up. Right. That's the pitch. That's the reverse centaur pitch. And you know, that person is set up to fail because one thing human beings are not good at is remaining vigilant for things that don't happen very often. Right? The TSA, we have created the world's water bottle spawning as mother fuckers. The human race has ever seen because all they do all day is find water bottles. But you take a red team and you put it through the checkpoint with a fake gun in the luggage, 95 times out of 100, they miss that gun. Yeah. Because they're not you cannot keep neurons in train to spot patterns that do not occur. Yeah. And so the first one is the first arrangement, the central arrangement is one in which the cadence is being set by the human, the the quality of the output is improving. And we're not asking humans to do superhuman things. But it is more expensive. Right. And the second one is the one in which you produce vast surpluses that might repay the people who have contributed to the bubble and the bubble is gigantic. So when I wrote the book, 700 billion dollars have been spent worldwide on both training models and operating them as well as the hardware for them over about three or four years. Another trillion has been spent in the year since since I wrote the book. And if you believe Morgan Stanley, or not Morgan Stanley, oh yeah, no, it was Morgan Stanley who did, oh no, it was Darkner who did say that we would all be spending six hours a day in the metaverse by now. But if you believe that it's going to be 2.5 trillion this year by the end of the year. Right. Now, how much is the industry making 50 billion dollars a year? Yeah. They say it's 60, but like 10 of that is the money that Microsoft gave to OpenAI and OpenAI, I get back to Microsoft, speaking as someone who writes techno thrillers about accounting fraud to call that accounting fraud is to do violence to the noble accounting fraud. Right. It's just a it's not even a good scam. Right. So they're just like not making enough money. Yeah. And the assets won't last long enough for them to start making money. They're rolling over these assets like every five years. Actually, it's more like every three. They claim every five because the graphic cards go bad or they become out of date or whatever, right? Yeah. Yeah. Yeah. Yeah. And when they go out of date, you know, so how is Nvidia getting so much performance out of these chips every year so that like it's worth buying this year's chips instead of last year's chips? They are throwing away any kind of fidelity to backwards compatibility. They're just like whatever we can think of to get more performance, we're going to do it. Well, that means that you are seeing successive generations of video chips where one or more of the power consumption networking and heat dissipation characteristics are so different that they can't put them in the same racks. You can't even put them under the same HVAC systems. There are times where like to upgrade a version or two of video chips, you scrape the data center to the ground. Wow. Because cheaper to start over than to retrofit with it with these really radical different things because they really are dancing right on the edge of what humans are capable of doing. Now that is not something that is new like we could have done that forever, but it's crazy to do it because you are assuming that your customers are willing to tear down their data centers or you know what this is like right you've been you're an old gamer guy right so like every now and again they're like oh by the way the new graphics cards requires a new socket you have to buy a new computer for it and everyone's like mother fucker right but it is like they've reached the limit right they've squeezed out all of the power they can get out of those all graphics cards. That's not what they're doing here they're just like okay we could get like 1% more out of the graphics card by keeping the socket the same shape. We can get 8% if we throw away all the HVAC requirements I guess we're doing the 8% because that's how they are like one of the magnificent seven seven companies that are 35% of the S&P 500. Six of whom are losing money to the seventh one which is in video which is loaning them the money to lose to them and that have no past or profitability. Well that makes me feel good about my retirement account. Yeah yeah go along on laser tag arenas because it's one thing that we can totally use data centers for get get deep into the laser tag arena futures. Hey and and you know why don't they tell that to the angry people of the city council meeting. Hey yeah don't worry after the apocalypse this could be a roller rink you know and you know how they always say this is not investment advice this is investment advice sell your 401k by laser tag future. Don't do that. There's there's no investment advice on this show and look it's hard to predict what financial markets are going to do sure we've been talking about a crash for for a while now. But yeah the market can remain irrational longer than you can remain solvent but exactly anything that can't go on forever eventually stops. So those are the two holes. You know I've been really focusing on my health and fitness recently but when my trainer recommended a high protein diet the dread of changing up my food habits hit me like the 250 pound weight I was definitely benching believe me that's when a friend Rick. hunger root. You can think of hunger root as a personal shopper and meal planner rolled into one with over 50,000 shift-crafted recipes and over a thousand grocery items the variety never gets old. Many recipes are ready in just 15 minutes and there's always something new to try and you don't got to seek out the new stuff. Hunger root keeps getting more personalized by learning your tastes and dietary needs over time. It knows what you actually want and it keeps refining what you get. At this point, I genuinely don't think about grocery shopping or meal planning. It just kind of happens now and I've been able to get so many new protein-rich meals into my diet without putting in any extra time or effort. I can just come home and know that there's something to whip up. They'll be quick, delicious and personalized to me. So I know you're going to love hunger root as much as I do. For a limited time, you can get 40% off your first order plus get a free item in every box for life. Go to hunger root.com/factually and use code factually. That's hunger root.com/factually. You get 40% off your first box and a free item of your choice for life. I've been wearing glasses pretty much since I came out of the womb and I can count on one hand. The amount of times I've actually replaced them. Of course, I outgrew my womb glasses pretty quickly. Plus, they were covered in gunk. But since then, I've been rotating between the same couple pairs of outdated scratched up frames because honestly, new glasses are expensive. That's why it was a total game changer when I discovered zenny. I ordered a new pair of classic browline prescription sunglasses that probably would have cost me upwards of $300 anywhere else. But on zenny, I paid I kid you not $58.85. That's $58.85. Zenny is an online I wear shop with stylish prescription glasses, sunglasses, and customizable lens options at prices that make it easy to own more than one pair all starting at under 30 bucks. You go to zenny.com, pick a frame, enter your prescription, customize your lenses, and they ship right to your door. No store visit, no pressure. At that price, you start asking, "Do I really need new glasses?" And you start asking, "Why wouldn't I just get a few?" You can get a work pair, a fun pair, a pair that only goes with one outfit. You don't have to justify it. Plus, zenny has a virtual try-on so you can see exactly how a frame looks on your face before you order. Now I wear my new sunglasses almost every day. And it's not that I'm not taking care of them, but it's nice to know I can replace them without checking my bank account, you know? So say goodbye to that scratched-up pair that's done its job. Upgrade today with zenny. Go to zenny.com/podcast and use code podcast15 for 15% off your first order, plus free shipping on all U.S. orders over $65. The style sell out, so don't sit on it, that zenny.com/podcast promo code podcast15. I love how most investment advice is just like fables, you know, little idioms. Going back to the radiology example, I think that's really illuminates a lot because you're showing how, look, that the technology is useless. It's interesting technology. There are situations in which it could accentuate or improve human decision-making, human ability, et cetera. That was obvious to me the first time I used it. Oh, there's definitely some cool potential here. But the way that it is being sold to the people who are buying it is the second version, right? It's the version where you can replace your labor with this and you get more for less. You get more money for less costs overall or you get more output for more costs overall, right? Yeah. But the reason people are so upset about this, the public is that we can tell in that scenario, we just get the less. Like there's no pitch to me, the cancer patient or you, the cancer patient. Hey, guess what? We replace 10 radiologists with one radiologist plus a big computer. Like we all know that's worse. We all know, look, one of my jokes I've been doing when I talk about this is like, what's the future of AI? It's you on a customer service thing going fuck, fuck, fuck, fuck, fuck, representative, representative, representative because it has been harder for you to get what you need out of the system because more people have been replaced with more systems. We've all experienced that is what you use if you don't care if the job is done well. Exactly. And that is evidently the case. And so it feels like the industry has had to sort of do this, pulling a rabbit out of a hat style of marketing where they're like, oh, it's the future. It could be imagine in this very big way. I mean, you watch commercials for and thropic is currently running commercials where it's just people going, could AI make me a better mother? What could the future of AI be? It's just, and then the tagline is keep asking questions. It's like, mother fucker, aren't you supposed to tell me what it's for? What's the price to me? Like, right, if we, if we keep reading these horses to run faster and faster, it's one of them going to get birth to a locomotive, right? Like if we teach the word guessing program enough words, well, it wake up. Yeah. So, you know, the, I mentioned the writer's strike before, and you just hit on something important about, about how people are being sold on this. So if you talk to a radiologist today, they will say, oh, we use AI in radiology, and we get to decide how it's going to be used because we have a lot of market power. There's way more jobs for radiologists than there are trained radiologists who are making good money. And the people who are making AI part of their pitch is, wouldn't it be nice if these mouthy, entitled workers who greet every one of your pronouncements with ego shattering confrontations with people who know how to do things, who explain to you that your idea is illegal, will kill people, will get everyone put in jail and is impossible. And you know, I've been a Hollywood screenwriter a little here and there, right? I know what writers get paid. Some of them get paid a lot of money. Most of them don't. It's not a big line item for the studios. Yeah. Right. Why were they so corny to replace writers? Yeah. With chatbots because you prompt a chatbot the way you give shitty notes to a writer's room. Right. Make me ET, but it's about a dog and there's a love interest and a car chase, which admittedly is just airbud. But if you, if you give those notes to a writer's room, everything stops. Well, like, you have to look on the floor for all the eyeballs that fill out of people's eye sockets when they will. Their eyes so hard of you that their eyes fell out of their face. They call you a fucking idiot. They tell you to go make a spreadsheet. Right. Now, you know, if you go to an AI and you give them this, this prompt, right? It will shit out a script. You can go to the smokehouse for, you know, two deep dish martinis and come back and you've got a movie. Now, the fact that that movie is probably terrible is, like, maybe a small price to pay for resolving the cognitive dissonance that arises from, like, telling yourself over and over again that you're in the driver's seat, but knowing that if you don't show up for work, everything is fine. And if the writers don't show up for work, there's no movie. And like, maybe you're in the back seat with a Fisher Price steering wheel. And AI is how you wire the Fisher Price steering wheel into the car's drivetrain. And so, you know, hospital administrators are not happy about the fact that radiologists and other medical, skilled medical professionals tell them that you can't do things more cheaply or faster or whatever. We all saw that documentary, the pit. And so, you know, this is an easy pitch to people who would like to tame their workers. And, you know, I don't think it's a coincidence that so much of the automation is going to take your job, talk is about programmers specifically because if there's one group of people who have had enormous labor power, despite not being unionized, it's tech workers. Right. Right. They were adding a million dollars a year on per engineer to the bottom line of every firm they went to work for. Yeah. Jerry Garcia's, Jerry Garcia's chef is making them lunch every day like all that shit. Yeah. They get all the good. Yeah. Massages, a surgeon who frees your egg so you can work through your fertile year. It's not because they like you, right? They're just scared because there's 10 other bosses at the factory gates who want to give you a job. Yeah. Right. And a million dollars walks at the door with you. And, you know, one of the things that the tech companies have figured out back to monopolies is that they can produce an inferior output. So I think Musk is really the innovator here, right? He goes into Twitter. He's like, what if I drastically reduced the reliability of Twitter, drastically reduced its moderation staff, but also like just its technical reliability. I had a lot more outages and a lot more slowdowns, a lot more weird bugs by firing a ton of engineers. I mean, you hogs are going to keep eating that slop, right? You're locked in. This is where your friends are. So long as you love your friends, more than you hate me, you're going to keep using Twitter, right? And he was right. I mean, people, I only got off it a couple months ago because that's where my audience was. Just like a lot of other creative professionals, right? It's just like this point where he finally made it worth less to me, but it took a while, right? So I think a lot of these companies are like, we're going to fire a ton of engineers and make a palpably worse product and say that it's AI and also like not have anyone say, hey, if you're firing workers, is that because you're doing worse, which is usually why you fire workers? And instead, you're like, we're doing so good, we fired our workers. Which like, how great is that for someone who has to answer to the capital markets? Yeah, right? And all of the-- this is related to this notion of market power and market dominance, because in the segment that there's a paradox about the way workers narrate their experience of using AI, there's a paradox about growth companies and monopolies, which is that as companies are growing, they have these incredible valuations, because a share in a company is a claim on its future earnings, and if a company is growing really fast, then those future earnings are going to be much larger than they are this year. So if you have two companies that are earning the same amount of money, but this one's growing and this one isn't, this one's shares are going to be worth way more, because next year it'll earn more money, you've got a share on that money, right? Now the corollary of this is that when you stop growing, because anything that can go on forever eventually stops, Google gets a 90% search market share, where are they going to get more people to search Google, when you stop growing the market wakes up and they go, hey, not only are you probably not going to grow anymore, but all the other people who are holding your stock are sleeping with one eye open and one hand over the cell button, and as soon as they think you're not going to grow anymore, they're going to sell, and my shares are going to go down. And when your stock is highly liquid because it's got a very high price earnings ratio, you can use it like cash to grow. You can buy other companies by giving them stock. Now stock is not like cash. You get cash from creditors or customers or investors. You get stocked by typing zeros in a spreadsheet, right? So like a mature company bidding against a growth stock, right? Like the growth stock is just bidding by pressing the zero key, and the mature company is like begging people for money, right? So like your power to grow is going to fall off a cliff once the market considers you to be a mature company, which means that even if I think you could be growing, if everyone else decides you're not, and they sell off your stock, well, then you're not going to grow anymore. And that means you're going to stagnate and shrink, and you'll be in acquisition target. And it's just like, I got to get out because they're all getting out. Meanwhile, if you own that company, right? If your Mark Zuckerberg and like your portfolio is full of Facebook stock, right? Or if you're one of the executives, you hired a stock. The day that happens, you become personally much poorer, right? Like we say, oh, infinite growth is the ideology of a tumor. Mark Zuckerberg doesn't want Facebook to grow forever for ideological reasons. It just doesn't want to be 95% poorer. No one wants to be 95% poorer, right? And so he wants endless growth. And when you have everyone in the world using Facebook and there's 4 billion Facebook users, Facebook and Insta users, right? Or when 90% of the world is searching on Google, where's your growth going to come from? Well, maybe it comes from like rating another company, right? Maybe the pivot to video would have worked and Facebook could have become YouTube or G+ could have worked and Google could have become Facebook. It didn't. So then they sort of making up markets, right? Oh, we're going to do the metaverse, right? T's or whatever. Like obviously those don't exist, but like on the plus side if they don't exist and you just made them up, no one can tell you what they're worth. We're going to go to Mars, like we're going to go to Mars, whatever, right? And then you get AI. And AI is the latest one. It's kind of for all the marbles. The sums involved are much larger. Facebook lost 61 billion dollars on the metaverse. They lost 150 billion dollars over the last three years on AI and they're on track to lose another 150 billion dollars here on AI. It's the money losing us thing the human race is ever done. And why are capital allocators giving them all this money? Well, partly I think billionaires are sort of like bosses, I mean, by definition, billionaires are bosses. They deal with people as statistical abstractions, especially if you work in social media or you run a large services company like Amazon, you know, people are just big statistical artifacts. They're not individuals. You turn nods on the back end of Google search and billions of people change their behavior, right? And so people aren't really real. They're kind of fantasms. Plus like you have to hurt people to be a billionaire. And if you really think other people hurt the way you hurt, you couldn't really do it. Like if Jeff Bezos thought that an Amazon driver needing to pee, like really needing to pee felt the way it did when he really needs to pee. He wouldn't make some pee in bottles, right? Like Jeff Bezos needs to pee as well. He's 10 years older than me. I need to pee all the time. So he needs to pee all the time. That guy knows what it's like to need to pee. And, you know, and Elon Musk, he's, you know, taking so much ketamine that he needs to pee all the time, right? So he knows what it's like, right? These people, these people understand, right? And yet it doesn't feel the same way that like you could not be on F.D. and Island if those girls were as real to you as your own daughter. Yeah, right? These people can't be real. And so you have this one group of capital allocators for whom everyone's a phantasm member anyway. The difference between a chop-on to person just blurs, right? And like you don't think what anyone else does is hard anyways, because they're phantasm. How hard could it be to clean a toilet or to pack a package at an Amazon warehouse or to drive a van? It can't be that hard. Otherwise, this phantasm couldn't do it. So of course a chop-off can do it. But then there's another group of operators, capital allocators who may not believe that. They may have a more realistic view, but they know that bosses would really like to fire workers, right? And they're like, I don't care if it can do your job. I care if a salesman can convince your boss to fire you on a place with a chop-off, whether or not the chop-off can do your job. Like, you don't have to believe that peptides will make you a love god to believe that Andrew take and sell a lot of peptides and invest in his peptide business, right? And you don't have to believe that AI can do your job to believe that bosses are credulous doles who are infinitely horny to replace mouthy workers with pliable chop-offs. And so we get all this capital allocated to this thing that from a computer science perspective, perfectly interesting, respectable, and fine, and even useful. I'm going to go out on a limb here and say, statistical inference isn't bad. Like, measuring what happened before and what followed it in order to make guesses about what will happen in the future is fine. And do we need to do it all theory-free the way the LLMs do? Because the way we used to make AI is we used to have symbolic AI, these things called world models where you would like try and actually describe using software, how the world works. So you'd like actually make a model of the world and you make little symbols and then you'd have this machine learning system manipulate the symbols. Like the way deep blue works we're actually going to teach it, yeah, like deep blue. Exactly. So chess is a really interesting example, because what they did with LLMs, the breakthrough with LLMs, 15, 20, 15 years ago, 12, 12, 15 years ago, which was really quite exciting as they said, if we just stopped trying to explain the world to the computer, we just gave it tons of training data and we said, find the correlates. Now, correlation is not causation, but causes and effects are correlated. And there's lots of things where we don't understand the causal relationship we still get to use them. There's tons of medicine. We don't know what's method of action. We just know that there's a set of symptoms that respond usually to this molecule and we'd like to know how it works. But if you're the person who's not sick anymore, you probably don't give a shit. And if we can discover those, we can observe more causal relationships that we can explain. And so if we can just like use the big data engine to find a bunch of correlations and then do experiments where we say, okay, I'm going to move this input and see if this output changes. Then I'm going to infer a causal relationship. And whenever I want X to happen, I'm going to do Y, right? Well, that worked actually way better than we thought it would. This is the thing that made everyone excited about AI. Everyone was like, oh, that'll kind of work, but it's going to go off the rails a lot. And it went off the rails a lot less often. And I think we got surprised. I'm certainly surprised by how much more of the world can be manipulated, explained, and replicated without a theory of how things work just by understanding correlations without any causal understanding. Interesting, super interesting. So here's chess. When you ask an LLM to play chess, it doesn't have a world model. It just knows that statistically if the following moves have happened, that the next moves are the ones that are most likely, an LLM will put a chess piece on a square that already has a chess piece that it's not allowed to occupy. Right. LLM can't play chess. We built, it's not just that like deep blue play chess. We built electro mechanical computers out of vacuum tubes and switches that played a valid game of chess. Right. And an LLM has like, I don't even know the name for the numbers of how many times more computing it has. Like we're talking about like half-tillions, septillions of times more computing, and it plays a worse game of chess. Now you can use an LLM coding assistant to write what is now the best chess playing computers we have. They were built with LLM's as coding assistants with human beings directing their effort. And so, here you go. Right. If you want to do theory-free inference, there's times where a skill programmer can make it really high quality software if they're in charge. But there are also tasks where it's not just grossly inefficient. Like using half-tillions of times more computing to play chess is just dumb. Right. But also, it sucks. Yeah. And I don't like the people involved with this. The people who are driving this bubble are so convinced that if you just do enough-- computing. You can just throw away all investigations into causality and just like solve everything. Yeah. And, you know, Musk is talking about like building a Dyson sphere and capturing all the photons the sun and that's to do this, right? And like on the one hand, that's nuts, right? On the other hand, the one thing he's right about here is that might be the amount of energy you need to actually get reliable theory-free inference for all these complex causal relationships. I think one of the reasons you're right, the CEOs like it, is it makes them, it turns everything into such an easy math problem for them. It turns it into more input equals more output and that's all I have to think about. If I get more of X, I get more of Y and all I have to do is make this number go up, then that number go up. I mean, I was talking to my friend Robert Zachini on a podcast that recently came out and we talked about the connection between, you know, crypto and AI, what do they have in common? What if computer just is money? What if? It doesn't matter what the computer does. You don't need anyone skilled to run the computer. You don't have to think about what it does, just the more computer, the more money. Also, what if financial value can be created with a labor? What if you can have a world without people where instead of you needing to motivate a bunch of mule-ish people to do a thing, you just press go on the money on the line, go up machine and money emerges. And man, this is how MBAs think, right? This is like what they teach you in business school. Forget about the dirt and muck of the world, right? Work on a level of abstraction and, you know, once you, and that's the logically a lot of tech industry was built on, right? Oh, what if we, you know, disrupt this industry and build a long tail of blah, blah, blah, and that, you know, it's like you can sort of forget about the real world as it is. You enter a level of abstraction and you're like, oh, more of this means more of that. And AI is that logic taken to the nth degree where just everything becomes computer. And I'll tell you, you know, the verdict is in, right? There have been big meta-analyses on the impact of MBAs on businesses because there are a bunch of businesses that are family businesses that after a certain number of generations known to the family wants to take them on a higher professional manager inevitably with an MBA. And so you can see, right, like a practitioner versus a finance, trained finance manager, how they perform. The companies aren't more profitable. The share of money going to management goes up and the share of money going to labor goes down. That is the most reliable, the single, most reliable impact of an MBA back to disciplining labor, right? That is the project of the MBA. Yeah. And it's the project of AI. Now there's another connection between AI and crypto, though, that I want to just tease out here briefly. Please let's get back to labor, but tell me this. Yeah. So Moore's Law, right? We were able to make chips faster, right? Every 18 months they doubled in speed for a long time. Anything we can't go on forever eventually stops. We start to reach feature sizes. You can think of a chip as like a little maze that you fire electrons down, like balls going down a maze. Eventually we get the paths in the maze are so small that electrons tunnel through them because of quantum effects, which is like we're losing the ability to make those traces any smaller or any closer. And so when Moore's Law taps out, the firms need to continue to show gains and they go from serial computing to parallel computing. So this is when you see the explosion in like GPUs with multiple with multiple slower powered CPUs on them. And if you were like prior to that moment, if you wanted to work on a problem that was a perform better one is parallelized, right? A lot of graphics problems, a lot of gaming problems, machine learning, crypto. You were like you had the shitty end of the stick because no one was putting a lot of energy into improving parallel computing until that moment. Whereas if you were just like someone, if you had like a serial computing application and you like it was like oh it's kind of too slow to make it work right now, you could just like sit down and wait for 18 months for the chips to get twice as fast and then it would work, right? Like you know you were like you were born on third base, you were like a boomer who bought a house and then just suddenly you were a billionaire, right? And then it changed, right? Then it changed and all the energy got put into parallelization. And there are classes of problems that are more readily paralyzed. And so first of all, those problems became the problems that we solved a lot. It was the graphics revolution. I was a silicon graphics contractor at one point when we were doing tons of super parallel parallel graphics applications. And people started to look for parallel applications they could do because if you had a parallel application, even if you couldn't do it this year and 18 months, you'd be able to do it. And so you remember when crypto started, you couldn't buy graphics cards for your gamers, couldn't buy graphics cards, right? Because they bought them all for crypto. Yeah. And now you can't get crypto graphics cards because they're buying them all for AI, right? This is just like finding these like new applications for parallel compute. And that's like that's like what binds them together under the skin. Yeah, that's really fascinating. I also remember so vividly, like a decade ago, every year, like Nvidia would have an amazing new graphics card that like could do like they were the advancements they were making in video game graphics cards for the consumer. We're like pretty amazing right before the crypto. Like, oh my god, this card is like $100 and it does 10 times as much as the one from last year. What the hell? Incredible. And like I have not seen that news story on Kotaku.com for a while, right? Like the advancements before the consumer has stopped because they found a much more profitable use for this very specific kind of parallel computing. Let's talk about labor. You know, you were talking about these companies need to be infinite growth machines. That is what their current value is staked upon. That's what Mark Zuckerberg's current value is staked upon. He's feeling feeling around for this. What's the new market that he's what's the new oil well that he's going to tap into the new deep reservoir? Oh, it's going to be the metaverse. Oh, Elon says it's going to Mars. Then they figure out all at once. Oh, it's the entire labor market. That's what we're going to drain. And so Wall Street is like, oh, yeah, the entire labor market like, oh, the entire output of all of humanity. Yeah, that is a lot of value. If you could provide that for cheaper, then yeah, you would have an infinite or not an infinite, but a vast, vast, vast growth opportunity that'll last you, you know, a number of years into the future. So that's what they're that's what they're going after. I want to go back to the point that you made about the writer's room, though, because I think that's a really interesting dynamic because, you know, I used to get notes, right? And sort of the dynamic between the executive and the and the note taker is, is the executive gives the notes. And then we in the writer's room goes, oh, God, they're so fucking stupid. Like they, okay, I, what you do is you have to find the note behind the note. You have to go, okay, they have a problem, but their problem isn't what they think it is because they're not writers. They don't know what the problem is, but they have a bad feeling. So what if I were an executive? Oh, they probably don't, they probably don't understand XYZ. So we actually need to make a different change than they think they have to make. And my conversations with them as a showrunner would always be, and we had very nice executives, that worked with very many, but you'd go, hey, so here's what we're going to do. Like I know what you said, we're actually that's not going to work, but let me tell you what we'll solve your problem. And if the conversation goes well, they go, okay, great, that sounds wonderful. Thank you. The point being the people in charge like they don't know what their employees do because they don't do the thing their employees do because their work is specialized. And so they should not be allowed to replace all of us because they don't know what we do on a mass scale. The other thing that comes to mind is I used to tell people when the writers goes on strike, oh, they all want to replace you with AI. And I talked to reporters, they'd say, well, aren't you going to be replaced by AI? Aren't you worried about this? And I would say, well, look, ultimately, AI can't do a writer's job because the job of a writer is not to output text. The job of a writer is to be yelled at. The job of a writer is to take the phone call and make the person on the other and feel better and go talk to the director and go talk to the actor and be a human in the system. There has to be a human in the system. What you're evoking is like what I would end up as a writer is the reverse centaur, right? Is my only job is to get yelled at. The boss knows the AI's homework. Yeah, exactly. The boss knows on at the end of the day, hey, I do need someone to yell at. The human on the bottom knows, well, the boss is actually kind of dumb and they need someone to solve their problem. And so you need that human as interface, right? But you know less than them, then then you used to have. And you know, one of the things that's interesting about the writer's strike. So there's two important things, facts about the writer's strike. I actually think the writer's strike is the most important thing that has happened in the AI labor story because it's the only group of workers who beat AI. Yeah. And the first important thing about the writer's strike is they got the right to use AI in the writers room if they wanted to, but without any cuts to their pay or head count. Now, like if you're writing on Young and the Restless, right? You got this giant Bible of all the things that have happened over however many years. You might want to feed every script that has ever been shot into an LLM. So you can say, Hey, what was this character like, you know, when did it, like how many times they got an angry at this person, or they ever had beef with this person? Or like, does there ever been like a romance with it? You don't remember, right? And there are like continuity people. Yeah, that is the one. That is the one's job. Like on Star Trek, famously, that's the one's job, right? But they don't remember either, right? Like there's just so much lore, right? If you're, and particularly if you can pull that into the like the bins and the shots and so you can say like, what's the time code, right? Like can we watch that? Can we pull that scene and watch it? And you can just do it in a writer's room. Like no one is going to, no one's going to go through scene by scene and write down the emotional beats of every interaction between every character. But I know I'm going to do it. And if it was wrong, who cares, right? Like if you say, find me all the time, so and so, as Picard was vulnerable, and it shows you a bit where he wasn't. You're not like, oh god, my whole day is wasted. You know, napsed, right? Right. That's fine. Those are low cost errors that the machine might make. They're low salient errors. And so if you're a writer and you want that, I don't think I want it. I don't write a lot of series, but I don't think I want it for my series. I kind of like the way that they come back to me impressionistically and then when I'm working on a new volume and then I go back, but if you want it, it's fine. Lots where every writer has a different way of writing and if you've ever hung out with other writers, well, you've been in a writer's room. So you hang out with other writers and you're like, what the fuck are you doing? How do you write like that? That's crazy. Every writer who's ever explained to me how they wrote, I was like, are you nuts? And then I explain how I write and they're like, are you nuts? Because that's how we all write. Fine. Right? Okay. So you've got a tool that you can use in this way. Whatever. That's fine. The other thing about the writer's strike that's really, really important is that they didn't go on strike against the AI companies and they didn't go on strike as like 1099 employees. They went on strike against the studios and they went on strike as freelancers who are in a multi-employer bargaining arrangement that is the weakest form of sectoral bargaining, which is when all the workers in a sector are bargain with all the employers in the sector. And because of that, they were able to win. And the reason that's important is that right now, the media companies that wanted to fire those writers and replace them with chat pods are suing the AI companies for violating copyright. A lot of writers are like, okay, I'm on their side. Yeah, yeah. And the thing is, we've expanded copyright over and over again for nearly 50 years starting with the 1976 act. In fact, it is 50 years starting with the 1976 act. Last longer, covers more uses, has higher penalties, there's more kinds of works that covers. Every one of these rights that was given to writers was immediately written into the contracts as something that our employers took away from us. Right? It's not like when your copyright on your works went from 50 to 70 years, you didn't suddenly get a contract that said you are contracting to give us this rights for 50 years and we'll come back in 50 and pay you again for the next 20. Your contract now just right, you are giving us 70 years for exactly the same amount of money. So, same thing happened once we, it wasn't really a formal copyright expansion, but it was like a judicial expansion, a copyright that basically said if you sample, you clear. Right? So there's now an affirmative right to control sampling, right? Well, that just became an aspect of every musician's contract. Right. And so now you have the circumstance where you know, 70% of the sound recordings are owned by three companies. So if you want to sample, it's going to be owned by one of those three companies. And if I can unpack the point you're making about the 50 to 70 years, like in the writer's guild, unlike a lot of other writers, when a book author writes a book, they own the copy right, right? But when a screenwriter writes a screenplay, we actually sell the copyright to the company. And part of like the lore of the writer's guild is, hey, that's why we get health insurance, right? We give up the copyright, we get a pension of health insurance as a result. That was a battle that was fought 70 years ago. That was the deal we decided to take. It's a pretty good one. All things considered, right? I agree. The industry worked because no company is going to invest $100 million in shooting a screenplay that they don't own on the copyright, too. But the copyright under law is 50 years long. Once it's extended to 70 years, they're not giving us an extra 30% of health insurance, right? It's just that's right. Now the thing they got are more value. Right. Exactly. Right. And they, and they, so they lock things up for longer, right? So like if you wanted to, you know, you can make a Mickey Mouse cartoon now because Mickey is finally into the public domain, but if you wanted to go into business against Disney by making Mickey Mouse cartoons, the fact that copyright last longer has made it harder to compete with Disney, which means that there's fewer places to sell your work. And it's a harder market to enter. There's more barriers to entry. And back to sampling, right? Once we created a sampling right for musicians, we created a world in which musicians who wanted to sample paid about $500 per sample to a record company that just gave it to their investors. They didn't give it to a musician. So every musician got poorer because musicians got a new right. Right. So here's what I promised will happen to. If we create an affirmative right to control AI training and, and sidebar here, if you think that copyright absolutely prohibits training, you don't understand how copyright works. It's very much an open question. Those activities are like making transient copies of works. That's how we get search engines, analyzing works. It's not illegal to count things in works. It's not a copyright matter publishing the council works, also not illegal, right? So maybe you can make a copyright law that makes it illegal. But like, it's definitely not like a slam dunk, oh my god, training works is illegal. Anthropic is not giving writers a settlement because they're training on books. They're doing it because they pirate books. That is illegal. So if we create that right, either through judicial interpretation or statute, every one of the contracts that we sign with the five labels, the five publishers rather, the four studios, the three labels, the two companies that control the apps and the one company that controls ebooks and audio books, every one of those contracts is going to suddenly read, you assign to us in perpetuity all of your rights to train A.I.'s and David Zotslav, the guy who wanted to have A.I.'s replace his writers will go and sell your copyrights to Anthropic to train with. And the only thing that will be different is that David Zotslav will be richer after you are put out of a job when they replace your superior writing with inferior chapots. And so this argument about who should get the biggest mouthful when they eat us up is such a stupid argument to be having when there's this other argument, which is like, what if we had sectoral bargaining, I think that is unique to the Hollywood Guilds, and only in this weakest form, and was outlawed in 1947 with the Tavht Hartley Act, which like, if we're going to make a new law, don't make it the copyright law, your boss wants, make it the sectoral bargaining right, every other worker in America wants. No worker in America is like, oh, if only copyright was stronger, I'd get more money for working the cash registers at a Wendy's, right? Whereas they all want sectoral bargaining, even if they don't know it yet, if you explain it to them, they'd be like, fuck yeah, I want sectoral bargaining. Every cat, we solve the Uber problem. If every cow driver is in a sectoral bargaining unit with every other cow driver, whether they're driving right, sure, not, wouldn't matter if their employees are not employees, right? You don't work for Disney when you write a Disney cartoon, you've got, you know, Adam Countover, I'll see or whatever, right? And so, you know, that is like the path we should be going down, right? If we're going to get a new right, it should be in solidarity with every other worker, and then class enmity with our bosses. We shouldn't be like campaigning for the right to find a new way to be angry about how our works are being used. You should be fighting for the right for health insurance and money and mortgages and prices for our kids, you know? And not like running around and saying, oh, scraping the whole internet is theft. Jesus Christ, if you want to know what the internet looked like before Trump stolen and gave it to his chud buddies, you better hope we can scrape it, right? If you want to know like whether companies are doing discriminatory pricing, you better hope we can scrape all their prices. If you want to know whether Musken and Zuckerberg are downranking content critical of the regime and their own businesses, and filling the void with things that are sort of conspiratorial or bullshit or, you know, manipulative, you better hope we can scrape all those feeds that are cramming into billions of eyeballs, right? Scraping is how we hold companies to account. Now, if you crash a server by scraping it, you're a dickhead and you should be in trouble. Scraping isn't theft. Right. Scraping is good. That's such a good point, because so often people, you know, when they're railing against AI on blue sky or wherever it is, we'll say, well, yeah, it's the scraping is the theft. But actually, we all love scraping in a lot of ways. We all love the internet archive, right, which nobody wants that to be illegal. You could make the argument. A lot of people could, I think plenty of people have, hey, this is stealing from us. But we all like that because we all like the goal. We all like the outcome. Yeah, yeah. It's good for society. What we don't like about AI is it is also scraping and drawing some conclusions from the scraping and like mucking up the data and producing something new and novel, which is what we like when it's used for other purposes. We don't like the end result. We don't like the goal of it. And so the laws that we make should be about the goal and about how it hurts people. Right. We should have, so like the established religion of late-stage capitalism is the worship of property. And so we can only talk about things in property terms, right? Right. Like, if they could, they would rebrand murderous theft of life, right? the most valuable things we have in the world are not are priced because they are priced. They are too important to put prices on, right? People, the most important one. We have these rights regimes that are capable of splitting these seemingly impossible Gordian knots, right? Like when can someone train a model on your work and when shouldn't they? That are like built around things like privacy and human rights and labor rights. And again, those are like complex regimes and they're such a thing as a bad human rights regime, right? You know, I've seen cops use anti-hate speech laws to prevent people from criticizing cops. That's a bad human rights regime, right? But like human rights is a framework that is designed to take account of these conflicting equities and resolve them in ways that property rights are absolutely not suited to. And you know, the getting like corraled into viewing yourself as a creative professional, viewing yourself not as a worker, but as like an LLC with an MFA, right? Like a small business person bargaining with other small business, with large businesses and small businesses. It's a trap, right? We've made copyright bigger and broader for 45, 50 years. The media companies that acquire our copyrights from us are richer than they have ever been and we are poorer than we have ever been, right? Like creating more copyrights. It's like giving your bully to a school kid extra lunch money, right? There isn't an amount of lunch money that's gonna get that kid lunch. You're just giving money to the bullies. You give that kid enough lunch money. They'll start running international campaigns saying, think of those hungry children. They need more lunch money and they'll take that too, right? Like maybe in a big competitive market having these tradable rights would be useful, right? But like under concentrated markets, like a labor market, then we need labor rights. Yeah. The art partisans, like our coalition partners in fighting for copyright are our bosses and our coalition partners in fighting for broader labor rights and a sectoral bargaining right is really the only one that we need is every worker and not our bosses. Damn, you are correct. And in fact, if you look at one of the times that some of the more company friendly Hollywood unions will team up with the companies. It's when they're fighting for copyright, right? It's like you wouldn't steal a car. You wouldn't steal a movie era, you know, was when you had some of those more, you know, company friendly forces on the side of labor go, hey, all they're stealing from us, there's, but really what are they doing there? They're protecting the right of the company to have the copyright. The worker has already given up the copyright. And I literally just had a conversation with a friend the other day about the violence that copyright of melody and rhythm has done to music creativity versus, this is such an interesting little side note that was explained to me by somebody who works in games. In video games, game mechanics are not copyrightable because they're actually patentable, right? 'Cause they're algorithms rather than, you know, ideas and it's harder to get a patent than it is to get a copyright. And so most companies don't bother patenting their game mechanics. Thus, all game designers are able to borrow game mechanics from each other pretty freely. And thus we get, you know, Balotro, which is built on 40 years of design of other games. - Yeah, rogue likes and poker games. I'm a stupid Balotro addict. I sold all the basic decks and I'm in the second tier now. And like, it's so hard that I just, I'm like, why am I still doing this? 'Cause I never win and I still do it. But and it's not just that, it's not just that their patentable patents are harder to get. It's also like most game mechanics don't rise to the level of patentability because they're not novel, they're not non-obvious to skill practitioner of the art. And so like, they'd be very thin and narrow patent claims that would arise out of game mechanics. There's very few that I think rise to that level. Maybe like the Tetris one. - Well, one that's fast and we're getting a little far afield, but I just have to drop this little tidbit because I love it. There was the, about 10 years ago, the shadow of Mordor games for PlayStation and Xbox where it was a Lord of the Rings game where they had the nemesis system. It was the system of like, you know, orcs who would remember that you had killed them and would come back later. It was really cool and novel. People loved it. And I said to a friend who works and gives, hey, why didn't we get any more of that? And they said, that was one of the only companies that patented it. - Oh, wow. - They patented it. And as a result, it's been gone for 10 years. Like no one has played this. - Well, another 10 in the patent will expire. - And so it's the only last 20 years. - And then we'll let everyone like that. Do you play left for dead? Only very a little bit. - So they had this Hail Mary mechanic where only when the odds were really, really against you, right? Like the hoarder was coming over the hill and you run out of ammo and your health was down to one bar. They would suddenly jack up the odds from like one and a million to one and a hundred of you getting an epic drop that would let you survive it. And like if you know that feeling when you're in a game where it's like, oh, I've gotten into a cul-de-sac. I can't win. I'm just going to restart the game. That was the moment of which you could get a Hail Mary. It was the closest I'd ever seen to, you know, like a dungeon master, like a good dungeon master and like a TTRPG. If you come up with like a really cool way of killing a dragon, even though it's like, way out of your league, the dungeon master will go, okay, roll two natural 20s and we'll give it to you. And you roll it and most of the time, you won't get it, right? But like every now and again, you get your two natural 20s. It's like, oh shit. - Yeah. - Right? I did this epic. The end of the setting. - So much, yeah. - Oh my God, it's amazing. And so left for dead really come. Anyway, we've got the gun off of you, but I'm married to the first woman to play video games internationally. My wife played Quake for England. And so I'm not a giant gamer, but I live with a giant gamer. - Wow, okay, look, people have to forgive us. Two big nerds got together. We got started geeking out about video games. So let's bring it home. Look, we're in this situation as a workforce as a community of human beings where our entire economy is premised on the idea that we're all going to be fired. And the few of us who remain are going to be turned into reverse centaurs where we will be hitched to the machine and et cetera, et cetera. Is it possible for us to stop this? I mean, the power of capital is so large, you say we need sectoral bargaining laws. I would say on top of that, we need like mass organization, which is even harder to do than getting a law passed, right? Like the writer's guild doesn't just have laws that protect us. We have, we've organized people very laboriously for 90 years. And hey, we only have to organize 12,000 people. We don't have to organize, you know, a couple of billions. - Sure, yeah. - So it's one of the hardest things to do is organize people that way. So, you know, it's all well and good to sit here and say, "Hey, we need to pass this kind of law or that." What can we do on a practical level or are we all sort of, you know, doomed to have things get worse in this particular way that we all sense that they're about to. - So I think it's a different calamity we're hurtling towards 'cause the thing is, we haven't really gone through this much, but the AI companies are not gonna become profitable. They are incinerating money at a rate never seen. And they don't have a path to profitability. They're trying desperately to offload the responsibility for coming up with bails of $100 bills to stick in the money for us onto us. I don't know if I've heard you cover Elon Musk's IPO yet. - I don't think so, actually. - I don't know. - It's quite gnarly, but what I'll say briefly, is the way it's IPO. - Yeah, he got a suite. So SpaceX is basically like it's the outer shell of a matrioshka that whose inside shells are two wildly unprofitable companies, Twitter and Brock, two of the biggest money losers like anyone's ever seen. And then SpaceX is like a beltway bandit that's got a marginally profitable business, putting rockets into space for Uncle Zucker. And so mostly this is like a play to offload these AI stocks. And they got the stock exchange that they floated on, the Bernie Madoff's Exchange NASDAQ to give them a sweetheart deal where they debuted as part of the stock index, which normally wouldn't happen. And the reason that matters is that if you buy an index fund, which is like what you were told in high school, like take your 401k and put it in an index fund, your fund manager is legally required to buy everything in the index at any price. That's like what an index fund is, right? So like 5% of the index of SpaceX, they've got to buy, you know, top up their ownership levels to 5% and they have to do it at any price. Musk was also able to restrict the number of shares that the company sold and remove restrictions from the number of shares that insiders could sell. So those shares that were being bought at any price with your retirement savings were being bought from like Emirates and Silicon Valley billionaires. So that's what happened to your retirement savings. And this is what the other AI companies want to do. And it kind of doesn't matter except like to the people we're gonna go broke because the companies are still gonna fail. And so the problem right now is that we are going to replace a lot of workers who do things that me doing with chapots that do those jobs badly. We're gonna turn off those servers. No one's gonna be able to do those jobs because the workers who used to do those jobs will become discouraged, they'll have retrained, they'll have retired. We will just vaporize the process knowledge that makes the sector go. So we've been talking a lot about IP. You think about IP as anything you write down and hang up in a filing cabinet and sell with the company. So we value it, right? 'Cause it's like one of the other things. assets of the company. But there's another kind of intangible knowledge called process knowledge. It's sometimes called tacit knowledge, but that usually refers to what one worker knows. The process knowledge is the knowledge that's spread out across all the workers. You know, you like are working at the machine and you know that you're only supposed to put the feedstock into certain pace. That's process knowledge. You also know that if you like put it into fast and it jams, and the person who can unjam is on vacation. But the person who used to have that job, you can call them. Like, this is really important. This makes business is profitable. This is why this is the argument for why the show The Pit is good when people talk about why is this show made in law process knowledge? Why does the show have to be made in Los Angeles, right? I watch this show and I'm like, that is a hundred years of specialization in different fields. You got the best cinematographer. You got the best person making the prosthetics. You got the best writers and they all live in the same place. And a lot of what Hollywood has been doing is getting away from that process. Oh, just go to Georgia and just hire some. The Alpha Street, right? Pretty great people in Georgia. But they don't have that community of practice. They don't know. They don't have that. That's smeared out. So give an example of like how important this is. There's a company that makes chips. They had a ton of chips that came off the line that failed QA. They could only be run about half their nominal clock speed. Person in charge of that run knew that one of their customers had a faction that had been arguing for making a lower end laptop. So called up the person in charge of that faction and said, Hey, would you like a lot of really cheap slower chips? Like I said, yeah, I can totally convince my boss to let me do a product launch if we can get that key input really cheaply. That product was called the Mac Neo was the most successful launch apples had in 10 years. It's so successful. They used up all the dud chips and now they're making new dud chips to go in the MacBook Neo, right? I hadn't heard this story about the MacBook Neo. Yeah. You don't rebuild process knowledge overnight, even if you even if you can hire skilled people who've been trained in gone to university or you bring them in from overseas, they just don't know how to they don't know all this stuff that is the Germans call it finger schmitzing to full, the fingertip feeling, right? The feeling of of like if you're holding a ball and it's like balance on your fingertips and you can like make it go this way or that way and spin it. You know, that's the feeling that you get when you're working in a crew of people in an industry of people in a sector of people who all know a little bit about how everyone else works and can make stuff happen. So that is going to be vaporized. And we're going to spend generations rebuilding it. Yeah. We're going to do it at a time when a third of the stock market has been vaporized. And we know how to do exactly one thing when the stock market crashes. It's austerity. And we know that austerity drives people into the arms of fascists. And among all the things that fascists talk about, talk about fascists, they're really bad at doing things, especially complicated things, right? Hence the fact that everything in America is now made of E. coli, right? And we're going to be trying to rebuild all of these companies at a time when we are going to be run by the dumbest shards imaginable. And that is the thing that really scares me. And we're going to be doing it against a backdrop of heightening climate emergency, which is going to make all of this said, it's going to be playing on the hardest possible level. It's going to make my bilateral challenges look trivial by comparison. And that's my concern, which is why I want to pop the AI bubble as quickly as possible, which is why I think we need to center our critique on the material factors driving the bubble. We have to not accidentally help them sell AI by going along with it when they say AI can do my job. And that's bad. We need to make sure we're saying, AI can't do my job. And if you like the thing I make, you're going to be really sad when my boss fires me, replaces me with the AI. We are class allies against my boss, who wants you to have an inferior output. And for me to have no job, yeah, right? That's like how we attack the bubble. People hate AI, but their analysis is like, it's a little all over the map, right? Like, you know, the reason that they're doing deep fake porn and commercial illustrations is not because those are sectors that have high wage bills. No language on earth contains the phrases riches of porn star, right? Or as rich as a taxi driver, like self driving cabs, have you seen what Uber drivers make? Right? These are just demos. These are demos to make you go well if they can do this, they can do anything, right? Now it sucks. It's gross. It's horrible. It's cruel. It's evil. But we need to point out to people and we're like, well, pretty soon, they're going to replace all the cab drivers with with self driving cars. It's like, well, how much money do you think those drivers make? Which is a cost to run those models. Yeah. Right? Yeah. They're going to sell you $100 bills for a dollar a piece. How long do you think they can keep that up? Yeah. That's what I've, you know, I talked once to an executive from a ride hailing company. I've told this story maybe on the show before. But he was, he was making the argument about, oh, well, you know, we, every time we put a self driving car in the road, we're able to let go of this many drivers. And oh, that's a real problem. Like, oh, we have to think about the labor thing, you know, and I was like, well, you're going to replace someone who you're paying less than minimum wage to provide their own car, their own insurance, their own fuel. You're going to replace them with, you know, a high tech self driving car that you own and operate, you live, Uber, whatever the company is. You're telling me you're going to shift your business model to that. How does that make money? And the guy said, you know, no, here's what's going to happen. Here's what we all think is going to happen. It's going to be what Elon says, which is the consumers are all going to buy self driving cars. And then they're going to lease them to us for, you know, three hours a day, you know, you're going to let your car drive around and pick people up. And I wear and tear isn't a thing. Yeah, I was like, fuck, no, we're not. You think you think I'm going to, you think some somebody that you need to use condom in the car seat? There you go. Yeah, exactly. Of your $100,000, you know, self driving BMW. Oh, no, people love their fucking cars. They like to look at them sitting in the garage, you know, plus, plus geometry hates cars. Yeah, like geometry just totally hates cars. And you cannot solve the traffic problem by adding more cars. This is not correct. And so I think you raise a great point. Like we see the way most driving around, which by the way, we don't know how often some poor fucking India is having to, you know, like tap in to make it go around a cone. They don't tell us that information. But we see them driving around and we go, okay, that's great. Oh wow, they're going to replace us all with that soon. It's propaganda. If they can do that, how long until it's the radiologists? How long until it's teachers? I mean, teachers like the one application it should never be used for because if you stipulate as the AI companies insist that the AI needs a human who can discern from when it's right from when it's wrong because it has these defects that they whimsically call errors. Like the one person who definitionally can't tell an error from a non-error in the output of a chatbot is a student learning a subject. If they understood the subject well enough to discern errors from truth, they wouldn't be students. They would be experts. So what? So we need to call out like the overinflated claim. This is what I would say to people on the picket line again. There are plenty of people in the writers guild to say the chatbot is going to be able to do my job. It's going to be able to write a screenplay and like part of our job in leadership was to shake them and say, no, it cannot do your job. The starting point of your negotiation has to be I am valuable. I have value that the computer does not have. You need to recognize that in yourself first. Otherwise, you're giving the power to the computer before you've begun. You don't start from a position of powerlessness. You start from a position of the world is made of people. People need other people. I do things that that, you know, the boss's computer cannot do. The boss needs me. How do I use my power in order to, you know, protect my position? But I go further because every time you do a box pop with some streamer or whatever, who comes up and says, tell me about the writers right here like they can do my job. Every investor who sees that is like, well, I guess I'm putting all my money in there. Right. Right. Right. You don't help stocks, windlers, raise money for their pyramid scheme by insisting that they're selling a valuable product. Or you will end up as the reverse centaur you will to the machine, right? Oh, look at you putting a button on it. Is there any other call to action you give folks, Corey? I mean, my call to action is always the same, which is electronic frontier foundation is the most important, most effective digital rights group in the world. I've worked there for 25 years. We're on the right side of this. We're on the right side of all of it. We understand that we don't want to give power to large firms. We don't want to give powers to governments. We don't want to fuse governments with large firms. And we want to help you take the legitimate, righteous anger you feel and understand where the soft underbellies are, what targets you should strike at, theFF.org, member sport and nonprofit. And it's the most important thing I do. I've loved the FF for decades and I'm, it's such a wonderful thing that it still exists. The ACLU for the internet or the, whatever you want to call it, it's like, it's a, it's a, it's a, we can't be the, for the internet because A, the ACLU has an internet division. And B, the person who used to run it is now running aFF. That's just retired. It's, it's that important of an organization, perhaps even more so. It is. It's foundational rights organization. And Corey, where are all the wonderful. places that people can get your books. 'Cause you have the most interesting URL for people to get books than anybody. You don't send people to Amazon. You got your own whole fucking system worked out. - Well, first of all, you can just get them anywhere. - Yeah. - And like, shopping is in politics. Support it in a pet in bookstores 'cause you love them not because you think it's gonna hurt Jeff Bees' also ability to buy peanut-shaped rockets. Just like, get the books where you wanna get the books. - I love you, man. - I love the books first 'cause you love them. Don't think shopping is politics. Don't mistake shopping for politics. You even have good takes about you are aware to buy my books. That's how fun it is to talk to you. - And if you wanna get them from me, I am unique among writers for the big five. I write for McMillan for several of its imprints, most recently for our stress and churoo. If you wanna buy the books from me, I uniquely sell them without digital rights managers, is e-books and audio books. Sell them as without digital rights manager. It means you can play them on any player. You can share them, you can put them on new devices. They're not locked to any device and they're sold without any contract. So there's no terms of service. The terms of service are do what copyright allows. Don't do what copyright prohibits. Lend it, put it in your estate, give it away. Those are all things copyright allows. Thank goodness it does 'cause that's how I became someone who loves books. When you buy it for me, I do get about double the royalty, but it's, I mean, it's money, but it's not like a gigantic amount of money. Mostly it's just like raw sales that helps 'cause that determines the size of the events and how much publicity my publisher's put on it. The new book is called Reverse Center is Guide to Life After AI. The previous one was called in Shedification. The next one's a middle grade graphic novel from first second of the McMillan imprint called Unauthorized Bread. Then there's another book called The Post-American Internet Next Summer and then a graphic novel of in Shedification. I have a newsletter called Pluralistic which you can get a Pluralistic.net. It is Creative Commons Licensed CC Attribution which means you can use it commercially, you can sell it, you can give it away, you can adapt it, you can make a screenplay out of it. You don't have to give me any money, you don't have to ask my permission. I publish about six essays a week, four to six essays a week there, along with weird collages I make at a public domain materials. That's my new vocation is making collages. - And yet somehow you have the time to spend an hour and a half talking to my ass on a podcast and you're doing all that. I don't, I don't, I work when I'm anxious. So the Trump years have been very good to me. - That's great. - And just what's the URL for the books? - Just go to like, just put them into a search engine. - Got it. - From my shop. - For your own crap hound. - Sorry, I beg your pardon, craphound.com/shop. I got a domain back in the ladies. And I registered one base that's the title of the first story I ever sold, craphound.com. There's that great Saturday night live bit from the early dot com era where it's like, we at First National Bank do not hop on the technology trends. We make sure that things are seasoned and here to stay before we entrust your precious money to it. However, this does mean that by the time we got on the web, all the good domains were taken, which is why First National Bank can be found at www.clompiness.phard. (laughing) - Corey, I love talking to you so much and I'm so glad you write so many books 'cause if you don't get to have me back soon. So thank you so much for coming on, man. - All right, thanks, man. - Oh my God, thank you once again to Corey for coming on the show. I know you love that conversation. Of course, you can pick up a copy of his book at our special bookshop FactuallyPod.com/books or at craphound, his own store. You heard the URL he's set it for you. If you want to support this show and all the episodes we bring you week in, week out, in fact, if you want to join our new online community book club where we're reading Naomi Klein and Astrotailer's new book main times fascism, you can do that at patreon.com/adimcono for five bucks a month. Gets you all of our community features, our bonus podcast episodes, the book club, our community discord. You can join a community of smart, curious people just like you. But for 15 bucks a month, I'll read your name in the damn credits this week. I want to thank MF Dakota, Tevis, Virginia Beckenbauer, Rebecca Champa, and Christina Coronta. Let me just choose some other names at random here. I want to thank Quinn M. Enoch, Patrick Ryan, Shannon J. Lane, and Matt Clawson. Thank you so much for being members of our community. If you would like to join them, head to patreon.com/adimconover. I also want to thank my producers, Sam Raudman, Tony Wilson, and Mandy Schmidt, my engineer, Sam Rogitch. Everybody here at Headgun for making the show possible. Oh, if you want to see my tour dates and stuff like that, you can head to adamconover.net. I'm putting up some new ones soon. And in the meantime, I'll see you next week on Factually. That was a Headgun podcast. That was a Headgun podcast.