Envisioning a Positive AI Future, with Tim Reutemann, Author of Liquid Reign

Full Transcript

Transcript

Tim Reutemann: 0:00 Your vote is basically like this pool of energy that you can distribute to different topics and you can distribute it through other people, say food price regulation. Then I see a cybernetic algorithm that AI managed that just makes sure everybody always has enough to eat and it's always nutritious, a relatively stable AI that just does that, and I can give my vote to that AI. In liquid drain, it's more like, okay, you have retina, you have junk DNA, you have fingerprints, you have toe prints, you have voice pattern, you have heart pattern. With those 7, you can then discuss restoring your identity.

Nathan Labenz: 0:32 What is the Progressive Movement 2 that would start to imagine the next generation of of ballot initiatives and, you know, some early forms of liquid democracy that we might be able to institute likely first at a state level.

Nathan Labenz: 0:46 Hello, and welcome back to Turpentine AI. If you've missed the last few episodes, don't worry. You're not hearing things. Turpentine is in the process of expanding its AI portfolio. And as part of that, we're creating a dedicated new feed for the Cognitive Revolution, which you can find and subscribe to at cognitiverevolution.ai. This feed, soon to be rebranded, will feature best of content from multiple AI focused shows. But for now, both the new Cognitive Revolution feed and this original feed will feature original content. If you wanna get it all, you will need to subscribe to both feeds. Today's episode, which is exclusive to this feed, is a fascinating conversation with Tim Reutemann, author of the near future science fiction novel Liquid Rain. Recently, I've repeatedly noticed that the scarcest resource seems to be a positive vision for the future. That might be hyperbole, but it really is striking amidst such rapidly advancing AI capabilities

Nathan Labenz: 1:48 that so few of

Nathan Labenz: 1:49 our AI thought leaders attempt to articulate in concrete terms what life with advanced AI systems will be like. We hear high level promises of material abundance, freedom from drudgery, and breakthrough cures for all diseases. And some, like Sam Altman, very much to his credit, in my opinion, are even investing in research and platforms for a new social contract built around a universal basic income. But still, what is a day in the average person's life supposed to look like in a world full of advanced AI systems? It seems that just about everyone's crystal ball gets pretty foggy. And to be honest, sometimes I wonder if the apparently extreme difficulty of imagining a positive future should count as some sort of evidence of a coming singularity. With that in mind, I'm really excited to present Tim and his book, which envisions a world transformed by AI and by other technologies, including crypto secured identity, advanced virtual reality, and high bandwidth direct democracy, and presents future people navigating this world in recognizably human ways, albeit with very different social norms and moderately different values. Tim has a background in economics and international climate collaboration, and he wrote Liquid Rain as a long form thought experiment meant to explore how humans can maintain their agency and autonomy in the presence of increasingly powerful technology. I read this book years ago, and I've thought about it often ever since. I really applaud Tim for taking on the challenge of concretely imagining a possible future that is neither utopian nor dystopian, and I really encourage you both to read the book and to spend some time doing the same for yourself. If you're finding value in the show, we always appreciate it when you share it with friends. I recommend this episode to everyone, but particularly to anyone that you think would need a little help envisioning a positive future in which AI plays a major role. And please do take a moment to subscribe to our new feed. We just released an episode with Kirtana Gopalakrishnan and Ted Zhao from Google DeepMind's robotics team. And I think that's really the perfect technical complement to this episode, showing just how much practical progress is being made toward a future of robot servants and general material abundance. You can find that and subscribe at our website, cognitiverevolution.ai. For now, here's my hyper stimulating conversation with Tim Reutemann, author of Liquid Reign.

Nathan Labenz: 4:19 Tim Reutemann, author of Liquid Reign. Welcome to the cognitive revolution.

Tim Reutemann: 4:24 Thank you for having me, Nathan. I'm happy to be here.

Nathan Labenz: 4:26 I think this one's gonna be a lot of fun. To, just set the stage a little bit, you are the author of this book, Liquid Reign, which I heard about a few years ago now on a podcast as well. If I recall correctly, I believe it was Helen Landimore who recommended it on the on the Ezra Klein show as 1 of her 3 books. I'm sure you would remember that moment. And, I just thought it sounded really interesting because, you know, obviously, I'm interested in AI and the future of AI and also have a an interest in direct democracy and kind of, you know, futuristic government reform. So when I heard her pitch for the book, I was like, I definitely need to go read this. So I did that. That's been a few years now, and it has been 1 of the books that has kinda stuck in the back of my mind, for the last few years. Not so much because of, like, the plot or the, you know, the individual characters, but really just the world that you build. I often say on this show that 1 of the things that seems to be in shortest supply in today's world is a positive vision for the future. And I think the world that you've created is a super interesting 1 that, you know, is not a utopia, at least as I understand it, but is certainly a vision of a very different world, seemingly a reasonably plausible world, and certainly 1 that, you know, has a lot of value in it. So I continue to refer back to that book. I think people should, you know, read your work and also spend more time with this kind of thought experiment. For starters, do you wanna introduce yourself and and kinda just tell us, you know, where you come from? Like, I know that, you know, writing these sorts of books is not the only or probably even the main thing that you do. So what's, what's kind of the backstory in your life that led you to set out on a project like this in the first place?

Tim Reutemann: 6:13 I really come from the international climate collaboration world. That's basically my professional background. Academic background is economics, but the transdisciplinary experimental variety. We developed a little steam farm game that we then played with Brazilian farmers to see how they react to different payment incentives for carbon. That's kind of the type of research I was doing. And I also came into economics relatively late in my life. So I started out with more physics oriented and more ecology oriented and only then drifted into the economic space more or less for my PhD really. So in many ways I was like, okay, this whole, especially the neoclassical approach to economics, this whole we are optimizing utility functions such as flat out wrong and we have to start from scratch rethinking to some degree. So LiquidWain developed as a thought experiment on the assumption that our economic theory is just flat out wrong and we could replace it with anything that you can write into a computer programme. That any set of rules, how we economically want to interact, is a viable method as long as it's programmable, which means basically anything goes, and then playing it out from there. I kept writing these notes and short unfinished articles, and at some point my wife was just like, Hey, you have to put it into a book. I was like, Okay, I want to do that, but then not an academic book because that's too exhausting. I wrote already my PhD, that was a nightmare. The editing, couldn't even I can't look at those words anymore. I'm so tired of these academic approaches. So I was like, okay, I do that and I write it as a science fiction novel. Knowing that writing novels is an art form in itself, that is my first 1 and it's like, I'm not a fantastic author in that sense. Like, as you said, it's not memorable for the plot or the characters because they are kind of vehicles to transport the boat to some degree, and I'm the first to admit that. So I basically figured, okay, I take Aldous Huxley's Island as a major source of inspiration. So it's this person who arrives in the world without knowing much about it that needs to get everything explained to them. That's kind of the core plot of Huxdees Island and also of Liquid Rain. And then combine that with an odyssey kind of looking for another person, looking for a lost loved 1 as the main plot device. So kind of keeping it simple on the plot so I I can really focus on the world building.

Nathan Labenz: 8:40 So maybe tee up the world for us. How would you describe and you could you know, maybe I don't wanna give any spoilers, although I would again emphasize that the world itself is the the main attraction, at least for me, you know, more than the the eventual eventually revealed plot points. But how would you kind of summarize in brief terms the world that you have created that our that our main character suddenly wakes up in after a long sleep?

Tim Reutemann: 9:08 So probably the biggest 1 is the type of effort that it currently takes to be politically active and to engage in your community and to be a part or active, proactive part of your society is kind of automated away. That just a verbal expression of a wish to participate in your community in a certain way is picked up and translated into a, to some degree, legally valid action to a degree. So the core political feature I'm using is the concept of a liquid democracy that the pirate parties in Europe were really big champions of that for a while. The idea is really that your vote is basically like this pool of energy that you can distribute to different topics and you can distribute it through other people. So my classic example is I have this very good friend of mine who's a psychiatric nurse and I don't understand anything about psychiatry, but I really trust her and I think she's doing a great job with her patients and she knows she's also deep into the topics and I would trust her with anything that is about regulating of psychiatry. A topic that's kind of important, it's a really important liberty rights because it kind of infringes on how crazy are you allowed to be without being picked off the street and put under actual physical constraints. So it's a very important law in a way, but I have nothing to say to that. And I recognize that she really knows her shit about it, so I would just pass on my vote and then she can do with it whatever she wants and use the status, power leverage. So it moves away from general purpose politicians into specialized people who actually know their shit about the specific topic and then come to compromises around these domain expertise. So that sounds very abstract in the story itself. It's basically that he just runs into people who ask him for his vote on occasion, and then at some point, people start giving their vote to the main character as well because he's a governance theory expert, which fits into the Yeah. As a device to telling the story of the vote. And this process, like I'm saying, he just has to say it, and this is where the AI part comes in. Right? So there is his personalized AI listening that will understand words like, oh, I really like what you're doing in ecological protection. I think my ecological protection vote goes to you, and then click, the vote goes to him in that moment because he has an AI listening and understanding what he's saying.

Nathan Labenz: 11:36 I guess there's maybe a couple angles I want to, explore this from. 1, I'm excited to give you a chance to read a few passages that you've selected, from the book along the way. Also, another kind of really interesting aspect of the book is that at the end of every chapter, you have links to sources of inspiration. So as I was thinking about this, I was kinda realizing, yeah, we should flag some of those sources of inspiration as well. You know, the the classic, the future is here. It's not just it's just not evenly distributed is definitely kind of a meta theme of this book because you're grounding all of these speculations in at least a kernel of something that already exists in the the real world. How about though, just give me a little bit more on kind of the technology stack that interacts to make this possible? Because as you described, like, main character has this assistant, which is an AI assistant, and there's like clearly multiple layers of technology advancement that are all working together to make this, you know, relatively effortless sort of delegation. So yeah, get we kind of build up that that technology stack for us so we kind of know how that's operating.

Tim Reutemann: 12:51 So a very critical first part is identity, and I'm really not that cyberpunky in it. In a way, a lot of it is inspired with how governments actually work, and it's usually your local commune thing where you show up and say, Hey, it's me. Hello. Show your face. Get your identity. And in a way that remains the same. So you need at least 1, they call them migration officials in a sense, so who confirms, Yes, this is a real person who's actually here, And then a couple of other people who also say, yes, this is 1 unique person, because basic premise of the democracy is still 1 human body, 1 vote. So that's a critical element, and that has to be connected to a cryptographic key, so to a private key that also has to be securely stored with the person. So I'm thinking of actually just being implanted and then having an AI that knows when to sign and when not to sign based on the behavior of the user. So this is like this very secure identity on the person, with the key on the person, almost as part of your own body. And on top of that then comes the information processing layer, so the interaction between the various political actors. I've been thinking about that a lot and also observing all this. Was published in 2018, so the old news media was full of blockchain stuff of all kinds, and I thought it was really funny that they had this great blockchain technology for democracy, but then all their decisions were just on a standard PBS forum, like 1970s technology where you actually had the discussion and all you needed was stable pseudonyms. So in the book itself, there is this discussion layer, this forum layer where it's just a BBS basically. It's just Reddit. Nothing fancy needed at all. 1 thing that's a bit different to just a regular Reddit is there's a stronger reputation system. So there's rooms that you can only go into if 100 people delegate your vote to you on that topic. So that's kind of the I call it a troll wall that kind of makes sure that while every voice is being heard, you also have to group your voice with other people who have similar voices in order to proceed into the actual governance basis. So yeah, basically just keeping out loudmouth trolley kind of people from the forum. So that's the fundamental aspect. And then there is the AI layer that is just making it the smooth interface that you don't have to sit down on your computer and type, but you can just interact with it in a more natural language way. Also, this is augmented. So often you would have maybe 3 people in the room and 2 people not in the room who are having the same conversation together. I'm speaking about retina projectors there as my favorite weapon of choice when it came to augmented reality. So that is very present as well, that it doesn't really matter who is in the room and who isn't. And you can always just, so to speak, beam the people in your need, or you go just full VR when people aren't in the same place, then you just have your your completely digital meeting.

15:56 Hey. We'll continue our interview in a moment after a word from our sponsors.

Nathan Labenz: 16:02 So we've got the crypto layer for identity at the bottom, the reputation layer and the kind of deliberative layer, the AI layer, and the kind of VR layer. So for at least 4 layers to the cake that are all pretty substantive. It's striking, like, how much progress has been made. You mentioned the book is 2018. I forget exactly when I read it, but it's been a couple years since I read the, you know, the full thing cover to cover, although I certainly revisited in preparing for this. But I am struck by how, you know, nascent probably a lot of those inspirations were and how much progress has been made. Right? I mean, probably the least progress has been made. Well, let's start let's just take them up 1 layer at a time. So on the crypto front, I guess I'm not sure how much progress I would characterize as having been made. Certainly with something like a WorldCoin, we are seeing at least some, you know, universal body based identity project.

Tim Reutemann: 16:58 It's a very different approach than the 1 I would to be taking in LiquidRain. Like in LiquidRain, it's not 1 single layer. So in the real world, have a UN layer, then you have a nation state layer, then you have a cantonal or state layer, then you have a municipality layer, then you have a community layer. Identity, to me, belongs to the community layer, so that is a huge difference to GoldCoin. For them, it happens on the international layer, and I'm just very hesitant to that in terms of preserving diversity of approaches and preserving the local autonomy. So, yeah, but they have been trying that. To me, honestly, thing that remains from this whole crypto bubble is that a lot of people learned how to manage private keys, and managing private keys probably is fundamental. I see that as something you basically have to learn in primary school at some point. It's like the only secure cryptography we know with public private key pairs, and you have to understand what those are in order to interact in a meaningful way with an actually secure computer system. It's also 1 of these funny things that I The crypto hype was around permissionless crypto, but in LiquidTrain it's all permissioned. There's a darknet that still exists that's permissionless, but the stuff I'm describing is mostly placed out in the open. It's actually integrated into the administrative structures that exist today in a paper layer and this kind of the same power structure. So I also used to work for governments quite a bit, so for the UN and for the Swiss government especially, and I'm kind of not seeing them giving that up. And I also don't it makes sense to have these stacked layers where identity is really in the local community and then you have a regional agreement about certain things and then a larger agreement about other topics that have to be regulated on a larger scale. So it's really the private key and then this permission blockchains that has already been like 20 years older than Bitcoin. It's not actually new technology, just that people slowly start to get into that actually you need to keep your private keys private. That alone is a huge step to be learned. Within governments, nobody trusts anything except if it's printed out and signed by hand. So yeah, think that just it will require a full generation of turnover. Setting of liquid rain is in '51, meaning most people who are in the age where they are working in the public administration or public service would be the definitive at least. So it's a post post boomer utopia, so to speak.

Nathan Labenz: 19:33 That's funny. I'm cur I mean, you mentioned a couple things there that I think are interesting. 1 is the the importance that you put on decentralization. So to say, I guess there there's kind of a couple ways that you could critique or react to the WorldCoin vision. 1 is that the technology itself may or may not be viable. Right? And I'm interested if you have any thoughts on kind of the use of the retina as your private key. Then But there's also just the homogenization question.

Tim Reutemann: 20:00 The only time when you would want to check that database of retinas is at a new sign up. Like, you wanna know if somebody with the same retina is already in the system. So I call them nymphs in the so I speak of true names and nymphs, and you need a true name to create nymphs, and the nymph is actually your private key. So you can spawn an infinite amount of stable pseudonyms, basically, with 1 identity. But your voting power is limited by having only 1 identity, and You can distribute that to your NIMS and they can distribute it onto other people's NIMS. So this is not something that should happen frequently, basically. It's just a 1 off thing. Also, identity theft, that's an interesting topic for crime novels and stuff, but it should be very, very difficult. Retina is okay ish, I guess, but you can still steal somebody's retina's hand and you can still print fake retina. So in LiquidDrain, it's more like, okay, you have retina, you have junk DNA, you have fingerprints, you have toe prints, you have voice pattern, you have heart pattern, and with those 7, you can then discuss restoring your identity. So the use is a different 1, and also the retina isn't stored. It's a hash of the retina that's stored. Again, because data can always be stolen, so you don't want to store the the same way you don't store passwords. Right? You store a hash of the password. Most importantly, having that rooted in an international organization instead of a local 1 is critical. It's like, to me, the biggest no go of WorldCom. Also having worked a lot in African countries, if you don't have digital sovereignty, you have digital colonialism and these are your only 2 choices and you kind of need people in the location who can understand that they are hosting a server and that they're running software on the server and to some degree even are able to read the source code of the software and that's a large call, know, and not even in the West people actually are able to read the source code and understand it, but that you at least have some understanding what's happening and make sure it's happening on your own computers. Otherwise it's just, why would anybody trust a Western organisation with such important data given the history of how Western organisations have treated data and treated the continent. Yeah. That's also a link to my current life again. We are working on a digitalization of carbon markets nowadays and very much focused on it has to be open source software, it has to run locally, it has to be very understandable, and the layers of complexity have to be either very deep in the security architecture or you have to find workarounds and heuristics that solve it without, like just keeping the logs and printing them and making sure they are easily available and stuff like that. So there's a lot of low tech security that, to me is more important than the the fancy sounding cryptographic protocols. And the older I get, the more I I realize that, to be honest.

Nathan Labenz: 22:51 That's interesting, evolution for sure. Moving up 1 layer to the kind of deliberative process, I think a lot of people probably don't have much grounding in what already exists today in the Swiss system. So maybe you could describe that system a little bit and the inspiration you take from it. There's, I think, other inspirations as well. I I know that the the Taiwanese experimentation with, you know, various forms of kind of discursive mechanism are also an inspiration, and there's probably more that you could describe. But for grounding, I think those would be great to give an overview of.

Tim Reutemann: 23:29 Those 2 actually go together really well. So I think Switzerland is really maxing out on what you can do with people meeting in a room and paperwork. So basically, you need about 7 people, I think. So maybe even 5 is enough to start an initiative. That means you formulate a piece of text that is meant to be voted upon and then you can submit that and then you have a year, and within that year you need to find 100,000 people to sign it. And then we'll have a vote to change the constitution that includes that text in the Swiss constitution. Switzerland also being a very, very conservative country with a relatively low participation rate in the democracies. Usually it's only about 50% who actually make use of those rights, and we have 4 elections per year. Even our executive branch is not 1 person as a president but 7 people who are equally ranked, And a lot of decisions are on cantonal level and the smallest canton is like 20,000 people, 40,000, I think. Yeah, you need 20,000 to take it over, so 40,000 people it's like Ottwald. And they have similar sovereignty to a US state on this very small level. In a way, I think that's a great, great decentral democracy, but it's also 1 that reaches its limits. It's kind of stuck and nothing moves anymore because it's always the same kind of people who meet in the room and also you're shutting out. You have all these real life bullying mechanisms basically. In the end, it's only white people in the room. In the end, it's 80% men in the room. So that leads to the Taiwanese experience where especially these first steps, this formulation of the text that is later put to vote, that part has been digitalized. So under the lead of Audrey Tang, BC Taiwan, I think, is the name of the software. So it's like a deliberation software that they implemented some interesting things. You can't directly respond to somebody else's comment but you can only write a comment to a similar topic that then stands next to the other and then people can agree. You can't disagree. You can only agree. So it's like keeping the discourse civil, and it works quite well, I think. So I read the case study on how they regulated Uber recently, and the result was probably the best regulation for Uber that I've ever seen. It's like basically, yes, we like having these apps and you just have to apply all the rules that we set for taxis, so they have to have social insurances, they have to have minimum wages, and that came out of a relatively long process that then went very detailed. There also the taxi drivers had time to give their inputs, regular people who just use taxis had time to give their inputs and so on and so forth. So the final regulation that came out is really supreme grade regulation and that then what Taiwan would do next is sandboxing it. So this is something Switzerland really has to learn from them and they would love we would do sandboxing more. The policy sandboxing being that you basically say, Okay, for a limited period of time and a limited budget, we pretend that this was already law for this group of people in the city, for example, and see how it goes for half a year and then evaluate it and then that becomes national law or we just drop it and do another experiment. So it's a more dynamic approach to regulation than we usually see. The more digital it is, the more dynamic it can get. You can have very quick turnover and changes in the details of a regulation, especially if the regulation is also self enforcing. So that's law and programmable money come together, right? If you have programmable law and programmable money, then you can just change the law and that just changes the cash flows and you don't have to teach all the accountants and all the companies to change something. It just happens by itself, so to speak. That's probably 1 of the big ones in liquid rain also that these if you decide to change the rules, they are just changed and you don't have to pay attention to the change and you don't like, there's not this this huge slog of transaction costs that comes from doing things on paper. Having to program it manually, like, the AIs can just adjust the the game, basically.

27:28 We'll continue our interview in a moment after a word from our sponsors.

Nathan Labenz: 27:32 So I should mention, if anyone is just listening and hasn't looked at the text yet, liquid rain, the rain is r e I g n as in, you know, reigning over a, a polity, let's say. So just to kind of recap the these layers, we've got the identity layer, which is foundational. The vision for this future world is still 1 person, 1 vote. We can get to the kind of AI role, especially as the AI has become more powerful in a minute. Then there's this sort of direct democracy, Swiss inspired, but kind of taken to the next level, highly decentralized, small really encouraging small local experiments, small initial thresholds to get started. It's, you know, anyone can kind of participate and and help set the agenda, but this is taken beyond where the Swiss have taken it in the in the real life present with the assistance of these kind of AI augmented deliberative mechanisms. I definitely recommend checking out Audrey Tang's work, hopefully a future guest, as well as the Polis open source project, which is a implementation of this that you can literally just go, you know, fork and and start to apply in in any context that you're working. But you you mentioned a a key aspect of that, which is it is meant to identify places of agreement, trying to find areas where people, even if they don't phrase it quite the same way or don't think of it quite the same way initially, where they can be identified to agree and then using those points of existing agreement as kind of the foundation to build off of. AI is obviously really important in that process, especially as you think about taking it to scale because who's gonna sit there and read everybody's initial statements and identify all the agreements? Certainly could be done by humans, but that, you know, would take a lot of time and energy, and AI is really good at that sort of thing.

Tim Reutemann: 29:36 So that that's a funny thing because I'm I'm just doing why not those on that 1. So while it is 1 vote, 1 human, a vote aggregator doesn't have to be a human. So I could, for example, decide, I don't know, let's say some topic that is more mechanical, say food price regulation. Then I see a cybernetic algorithm that AI managed that just makes sure everybody always has enough to eat and it's always nutritious. And that's a relatively stable AI that just does that and I can give my vote to that AI. That doesn't have to be a human. And I just support whatever that AI comes up with, and I can, of course, withdraw at any time if the AI starts screwing. So I would always allow for both. And then you can have these very dedicated people who are like the So this I haven't mentioned yet. So let's assume my friend's a psychiatrist. She is too busy working to actually engage in policies. Let's say she has very exhausting patients. Then maybe she has a professor that she can pass my vote through to. So maybe I have 1 vote on that psychiatry because nobody gives me their vote on psychiatry. She maybe has 100 from her friends and family and former patients maybe, and she can then pass it on to her teacher who may have a few 10 thousands, 100 thousands, millions of votes and actually is in a global committee on coordinating about psychiatric issues. So it's like a pass through mechanism and it can have more than 1 step that

Nathan Labenz: 31:04 it passes through. The interaction or integration of these different technology layers, I think, is really fascinating. The AI itself then, you know, can help kind of facilitate these conversations. It also becomes an interface for individuals to interact with the world in all sorts of ways. And that is maybe the 1 that is kind of least grounded as of 2018. Right? Like, I I think we maybe had the original GPT 1 at that point was probably published, but not even GPT 2. So the leap in AI technology over the last 5 years, I mean, obviously, is tremendous. I imagine you did not see that coming that fast.

Tim Reutemann: 31:43 Like, if you're reading science fiction, people have seen it coming for a long time. So I I also, like, just used some of the older and we had Siri at the time. Right? We had, like, Siri was out. I know Alexa probably as well. Clippy was already dead by the time. But the idea of digital assistants has been around for a very long time, and there's also a lot of sci fi about it. So it's strange. I haven't really gotten to the point where I find AI incredibly useful today, but that's maybe just because I haven't installed it properly yet. So as much as I enjoy playing around with it, I'm like, okay, you can do interesting stuff that was very expensive and effortful before. The way you had Nathan gave me a summary of my book that was created by ChatGPT and it was actually quite It was very interesting to read because some things like who is the main character and who is the side character, that didn't make any sense at all. That was more or less random. You could identify the number 1 main character but all the others, was like, no, that's actually a side character. That's actually a main. That didn't work. But the character description, on the AI character was really spot on. So it's mostly a matter of installing it now and integrating it with my very messy workflow that I have in my everyday life. That makes it kind of hard. But a lot of the, yeah, of the especially voice interface, that basically works nowadays. Right? I mean, if you have Whisper installed and you give it access to your Google Calendar, you you can just make your dates into the calendar and stuff so that that did happen. Yes.

Nathan Labenz: 33:15 So would this be a good time to maybe read a passage or 2 where some of this stuff gets introduced? We've we've covered it, you know, in a analytical lens, but how the characters actually experience it in the book is, 1 of the things I really like about the book too, and I don't know if the passages you'll choose here will be kind of the, you know, epitome of this, but I really appreciate how it just feels fun. I feel like so often the the visions of the future are, like, super, super serious, you know, or or super dark. And this 1 has a certain lightheartedness to it. Like, there's there's an element of silliness. There's an element of, you know, borderline absurdity, not taking the book doesn't take itself too seriously. I really appreciate that aspect of this, and I think it comes through in a lot of different moments. So with that, wanna read a couple passages?

Tim Reutemann: 34:08 Yeah. So I I would start with the the first introduction of Daniels. So Daniel is our main character. He was just in coma, is on his process of recovery, and had his first night out the night before. So he was basically having a social life, even getting drunk with his nurse. His nurse name is Harry, who is also in this setup. And we are engaging this AI for the very first time. So it's kind of a raw install. It's like the freshly from GitHub, basically in the hospital, still under hospital control, not fully like, as in a real hospital, you're not fully in control as long as you're in there because there's some medical supervision. So I'll just go with chapter 13, Daniel's little help for Starris. Sylvie has just been summoned. As a young operating system, her first task is to get to know her owner better. Until now, she's been a passive observer. But now that the volunteer nurse hasn't shown up, she takes control of a buddy bot and brings some breakfast. Good morning, Daniel. I brought you breakfast. It's on your bedside table now. Your first motor recovery session is scheduled in 1 hour. You should get up and eat something now. After a long coma, a hangover feels even worse. It was just a single glass of wine and a few zips of said apricot stuff, yet Daniel's head is pounding. He's been lying awake for half an hour, tossing and turning, and when he hears a gentle voice, he asks confused. Now, who's that? Please excuse. I am Siri, your operating system. You can call me up anytime, Just whisper my name. She tries to start a natural conversation, but her desperate attempt has failed. With so little data, what do they expect? There's almost nothing about him left in the archives. So instincts take over in the form of a hard coded startup gym recorded in her bios. Daniel forces himself to focus his gaze. He spots a robot in his room, a humanoid upper body on wheels. Her facial features are comical, avoiding uncannypsoidualism. Hi, Sylvie. Breakfast. Great. Thanks. Are you that robot? I'm only controlling it at the moment. The core of my soul is on a chip in your inner ear. The chip also monitors Daniel's vitals, and as he is still under medical supervision, Sylvia feels obligated to scold him for last night, raising her voice reproachfully. It seems like you drank alcohol. That might have been the case indeed. Daniel is rubbing his temples. She does feel bad as she says it but doesn't know what else to do with this mystery of a human, then parrots the standard reply for ethanol use by hospitalized patients. Alcohol can negatively influence your neurological regeneration process. I recommend waiting until the completion of your treatment. That will be in 2 days. Please be careful and drink responsibly. Daniel has to laugh. Drink responsibly. That marketing bullshit of the ethanol industry is still alive and kicking. Isn't the sole purpose of drinking alcohol to be irresponsible? Thank you for the warning. I choose life. What's next after 2 more days in rehab? That is up to you. I will stay with you and you can call me anytime. I have access to the experience of thousands of post coma patients if you need medical advice later on. The only way for her to learn more about him is by probing his reactions. So Siri tries to nudge his conscience with a parting shot, or if you find yourself drinking more than you want. Before Daniel can reply, Harry comes in. Daniel. He is leaning against the doffrine. Oh, you met your pet bot. These contraptions are quite useful. Siri, bring another breakfast and more coffee. Human volunteers have priority over AI powered nurses. So she has no choice but to obey. Of course, immediately. But a quick scan reveals that this particular nurse also drank alcohol yesterday, so she adds, And please attend to your patient and support them in avoiding intoxication. Ethanol can have a negative influence on the neurological generation process. Yes, yes, shut up and get us some coffee. Harry pulls a chair next to Daniel's bed. Siri makes the robot turn around on the spot and jolts out of the room. While the hospital rules prevent her from directly contradicting Harry, she can still bump her robot's body into him on the way out, just hard enough to hurt a little. Obviously, the drinking last night helped Daniel loosen up and socialize on the beach. Oh, dear. Ever since they got that last emotion upgrade, they get all huffy when you order them around. I mean, it makes sense in ambulance therapy they feel more human like that, but here in hospital, he sits down and checks with his patient. You okay? Pretty hungover. Daniel is still rubbing his temples to no great effect. The bot says I'm out day after tomorrow. Is that so? Based on your dance performance last night, I would let you go today. Harry stops talking as the robot brings comes back with a second plate and extra coffee. Thank you, Serby. That's kind of you. You are super quick. Daniel is being extra nice to her. Harry will take care of my therapy today, but I'm looking forward to getting to know you better in the future. Bye. Okay, Daniel. Take care. The robot elegantly places a plate on the table, pours the coffee and leaves the sack, cloyly sweet. Have a wonderful day. That was an underwhelming first interaction. Dissatisfied with her performance, Sylvie takes another shot and tries to recalibrate her model of Daniel's personality, tracing down the connotations around the I choose life line Daniel just used. This job will not be easy. So Daniel quoting Trainspotting Hair here and having problems with addiction throughout the book, so kind of foreshadowing that. Also going to the point where the AI or the EI, as I call them in the book, the extended intelligence, goes from being a raw model to 1 that starts to be fine tuned on the particular person it's made for. So in the same way, when you're working with ChatGPT, you're working with a generic model that's for everybody. I believe it gets interesting as soon as you get you know, a LoRa or a fine tune that's using all your own writing or uses your own background, your diary or something that gives the AI an idea of who you are and how you should be responded to.

Nathan Labenz: 39:58 Certainly, we can unpack, you know, some differences, but there's a lot of similarities, I think, between the presentation of Siri and what we currently have with GPT 4 and and similar systems. The kind of super broad knowledge. Right? It's this kind of first of all, just the modality. Right? Just the the chat style, you know, interactive dialogue is very consistent with what we now use. The super broad background knowledge, but, you know, 0 starting knowledge about who you are is really interesting. Serbia is ahead of current systems in terms of its ability to customize and fine tune to the user. Although that gap is definitely closing, just this week, we're seeing OpenAI announce their new memory product with, within ChatGPT. So we'll see how that works. And, you know, a lot of techniques, of course, as you mentioned with Laura's and, you know, a LoRa for every user is definitely 1 possible vision of the future. The kind of chiding nature of the AI is also really interesting. Right? We have the I mean, there's a lot of debate on how closely you follow this, this discourse within AI right now, but there's a lot of discussion around do we want the raw based model that will just do whatever you ask or, you know, and and is totally kind of beholden to your requests or and you can get that from RLHF. Or do you have this kind of vision of a more safer guiding, you know, steering you in the right direction, which is in some ways, you know, probably very beneficial. Obviously, you know, people struggle with addiction. They need help with that sort of stuff. And at the same time, there's something like, at a minimum, can be annoying and, you know, at a maximum could be like dystopian unto itself to have the AIs kind of telling us how we wanna live our lives. The integration is also really interesting. I mean, can comment on on any or all of those themes, but, you know, I love the use of the soul too. That's and I don't know if you think of that as, like, literal to me. I read that as kind of tongue in cheek, you know, for the AI to be speaking about its soul and the soul living on a chip. Tell me if you have a different, you know, kind of understanding of that. But to me, it sounds kind of jokey.

Tim Reutemann: 42:08 I can be very materialist if I want to, and I often use Susan Blackmore's The Meme Machine as a cognitive frame that basically says a soul is just a bunch of memes together, so to speak. So a bunch of self replicating information that fits well into each other and that's what we call a soul because soul is just a human word for a condition that we experience and what we experience is self replicating information. So in a way, I mean it as hard as I say it, but I also have a very different ontological perspective on what a soul is in that sense. So, yeah, this is 1 thing that is only like you kind of have to read it between the pages. But to me, I find it weird that we always talk about human level intelligence and we kind of have to get mammal level intelligence first, like mouse level intelligence or rabbit level intelligence. So in a way, that's the same reference, right? That it's not necessarily about there's only this 1 type of holy soul that's somehow spiritual and different from everything else, but a soul is just a thing that exists in the world, a entity that has its own experiences, and in that sense, AI series has a soul very clearly. I have another chapter on those, so But there was 1 thing that you brought up before that I wanted to address before we're going back to reading, and that's the issue of the nudging, basically. So do I want my AI to get pissy at me for drinking too much, basically, right? Yes, I do, actually, but I also want to be able to switch that off if I need to. So in that scene, Sylvie can't switch it off because it's still in the hospital, but as soon as you're out of the hospital, then you can just tell her, Hey, shut up. Don't ever talk to me about alcohol ever again. And then Sylvie will say, Okay, I won't do that unless you prompt me to do it again. Like until you make another command that overwrites this first command, basically. So the ability to nudge, I think, is very important and it really makes your life better if you, hey, didn't you want to bring your key before leaving the house? Oh, yes. I did want to bring my key before leaving the house. Saves you 5 minutes every day. So there's a lot of these predicting what your future self would actually want your present self to do and reminding you of that perspective, that can be super useful when you're in your boat and just running through life. The only way I see it getting dystopian is that when it's your prison AI basically, right? So when somebody else is controlling what nudges you get and you have no knowledge of that. So at that moment it gets dystopian, but there, again, this is an install of the most generic and most well reviewed open source AI that just is straight from GitHub and doesn't do anything like that and is verified not to do anything like that and having any external controllers except the stuff it receives from, yeah, either from the source code that is very well vetted or from the users themselves.

Nathan Labenz: 45:02 2 other ideas I wanted to kind of get into there. 1 is this idea of open source and kind of vetting, you know, what can you trust in this domain? There's been a really interesting paper just in the last couple of weeks from Anthropic called Sleeper Agents. And basically what they do in this study is they train a model to have some problematic behavior only under certain conditions. And then they experiment with the current safety techniques that are commonly used like your RLHF to see if applying that technique in the standard way is enough to get rid of this problematic behavior? And the answer, unfortunately, for now is no. If you have trained a model to in the in the paper, it's like you write helpful code when it's 20 23 and when it's 20 24, you write malicious code. And then the RLHF is like, you know, just is done in a way where it doesn't know that that was what was, you know, encoded into the model. You probably if you you know, everything obviously gets a lot easier if you know the nature of the attack. But in this setup, they don't know the nature of the attack. They're just applying standard techniques and then seeing, you know, does it in fact still write malicious code when it turns 20 24? Sure enough, it does. So I've been thinking a lot about what that may imply for the future of open source. And probably the answer is like like everything, you know, there will be another development that will kind of change this again. Right now, if if if technology was frozen here, you might think, geez, this is a big problem for open source because now I can't really you know, maybe for a few big providers, if I get something directly from, you know, the meta servers, if I download llama 3 as meta put it out there, I can be pretty confident they're probably not doing that. But then, you know, once I start to see all these different fine tunes and, you know, there's a bazillion of those flying around, can I really trust any of those things? You know, how would I vet them? I don't know what possible attacks people put in them. But, you know, again, there's probably more technology, more systems, more credibility layers. I don't know if this is something you even sort of fleshed out in your imagination of this world, but I wonder how you expect or, you know, now now seeing something like this sleeper agents, how would you expect this kind of open source but also hopefully secure balance to be struck in the future?

Tim Reutemann: 47:30 So there is 1 scene I'm describing, I think, the first interlude about a situation where that actually broke down and where there was a hack happening in Nigeria because place with the best red teams in the world. Right? So the ultimate overriding factor has to be still human bodies in the street, basically. So you can never fully trust the Like, we made the sand sink and that's very dark magic and we maybe should not have done it. We have to notice that something's going wrong and then react physically and just roll back and reboot and see if it happens again. If not, if it does happen again, replace it. And we have to always keep ourselves that option to just talk to your neighbor, figure out, hey, did you vote for that? No, I didn't. Did you vote for that? No. Why does it have 90%? Like, nobody in our village voted for it. And through that then, with your with your actual physical human body and your your your offline communication ability resolve. When it comes to being paranoid about security, so Liquidrain does have a dark underbelly still, so there still is a darkness. There still is all kinds of hacking attempts, and cybersecurity is not going anywhere. Like, that will remain 1 of the most important professions and, yeah, most important decider also in situations of conflict. On the level of paranoia, so there is this 1 little anecdote about the I think it was the guy who wrote the first C plus plus compiler, and then he dropped like 10 years later or so on a conference. By the way, what would have happened if I put a backdoor in that compiler? Then everything that was ever compiled in C plus plus I have a backdoor to. Even all your new programming languages have actually been programmed in C plus plus so I would have a backdoor to everything written in Perl because Perl itself was written in C plus plus So if you allow yourself that level of paranoia, then yeah, it's possible. There's no secure system. If I'm thinking about hardware security again, all this stuff is made in 1 factory in Taiwan basically. If there is a vector implemented on a hardware level in that factory in Taiwan, how would I know? Possibly, right? So in a way, your level of paranoia has to be somewhat proportional to how outrageous it is what you're doing. And so just managing your local ecosystem and your community food forest is harmless, and you can just run with it. But if you're actually in weapons research, if you're actually in, like, into whatever kind of criminal lie profit activity, you probably need to be paranoid about it. Yeah.

Nathan Labenz: 49:54 The other kind of intersecting theme there too or or concept is the idea that the thing is in the inner ear of the user. So it it certainly suggests that there is an advanced interface. Like, there are at least, like, a couple more advanced human computer interfaces that you outline in the book, and 1 is the the chip in the ear. The other is the kind of very immersive VR. Maybe there's even maybe you can point to even more than that, but those are the 2 that really are burned into my memory. There seems to be an inherent tension there between more deep kind of cyborg like integration, which on the 1 hand seems kind of inevitable. And, you know, if you listen to the likes of Elon Musk and his reasons for creating Neuralink, like, maybe even good to necessary, you know, I mean, his idea is that the AIs are gonna be moving so quickly. We need a higher bandwidth communication channel with them. You know, we need them to have deeper access to what we really care about. And, you know, so we have to have this kind of high bandwidth brain computer interface. But that does seem then tough. Like, the deeper that integration gets over time, the harder presumably it becomes to do a quick rollback or, you know, just go out and talk to your neighbor and, you know, compare notes. Right? So how do you think of about that trade off? And, like, do you in your lifetime plan to have, you know, some of these more advanced integrations? Like would you, you know, if Neuralink used to ask on every episode and I haven't for a while. If Neuralink, you know, passes trials, is that the kind of thing you would be interested in?

Tim Reutemann: 51:29 No. It's it's too invasive for me. I don't like cutting up healthy bodies. So in liquid drain, it's a less invasive technology. I'm talking a lot about transcranial magnetic stimulation. Basically, it's kind of like an MRI where you can have like you have these magnetic thingies around your head and they can very precisely, to the millimeter or so, produce a little electric, like, yeah, a current, basically. So they produce the magnetic field, but then because you have electric impulses in there, they get amplified through that basically. So I think with the state of the technology, last time I looked, it could do something like a red flash or a blue flash, depending on where it targets your visual cortex. So not very advanced, but in principle can stimulate your brain just electrically without actually cutting it open. Biotechnology is slower, much slower than software, It looks advanced, but all the bionic and cyborg technology that I've seen so far, the only ones I would trust are the DIY again, that people just made for themselves. I met this 1 cyborg guy who had an earthquake detector belly belt, and that was like whenever there was an earthquake anywhere on earth, he would feel it and have that integrated. Or a compass. Well, that was, I think his girlfriend had a compass, so she always wouldn't know where north would always know the cardinal directions just based on a feeling from some magnetic thing in her body. But the commercial versions, like there is the Bionic Eye, I think Doctor. Hof wrote quite a bit about that recently, that they just stopped supporting it at some point and then you just have that replacement for your eye as a blind person that kind of was great for a couple of years, but then you don't get software support anymore. You just have this piece of abandonment in your body. So I don't think we are anywhere near that and I also don't really Like voice integration, I like the inner ear because I do like to be able to talk to myself, so it's a habit I established. That's probably my most intense use of AI at the moment, that I go for walks with my microphone, and then I just I'm recording myself, and then I just put it into wristband, into my notes directly. So I do like voice interfaces, but I wouldn't go with brain interfaces probably. It seems too invasive to me. And in terms of bandwidth, I think that's just like I don't see the use case for that, basically. I don't think it's necessary. And another thing, so that links a bit to the VR also, and the VR, the longer away it is for me, less it's about the VR aspect and the more it's about the esports aspect actually. Because to me, esports is like it's showing off how good your human computer interface already is. Right? Like if you watch professional esports players, are incredibly fast and incredibly good in getting what they want into their computer and reacting to the slightest to tiny, tiny movements on their screen. Right? So in a way, a professional Starcraft player is much closer to a Cyborg than somebody who has a Neuralink implanted and doesn't know what to do with it.

Nathan Labenz: 54:31 So do you want to read another passage maybe introducing the VR experience?

Tim Reutemann: 54:37 This is even earlier, so this is just briefly after waking up in rehab. So as they're finishing up their dinner a few days later, Harry announces the next step in Daniel's rehab program. We'll start moving your body and waking up your fussier first thing tomorrow morning. Sleep well. Daniel is excited. The blurry recovery process has become quite boring by now despite regular spikes of emotion. Oh, Harry, oh, Harry, my fussier. I'm so utterly delighted, mocking his nurse's frequent use of utter delight. He has no clue what fussia are, but moving sounds good. The meals and washing sessions with Harry are the only episodes Daniel has outside the neuroregenerator so far. He lies back down, lowers the neurostimhood over his head and flows into the magnetically stimulated flashes of intense emotion, only occasionally disturbed by thoughts of Helen. Clare must have told her. It would be nice to get any sign of life. She probably has a new partner after all this year. Yes, but at least a quick hello wouldn't be asking too much, would it? The blur starts up again and with his consciousness still half present, he enters an old childhood dream. Spends the night running after flying tadpoles with 1 of those large nets used to catch butterflies through the back of alleys in Toreen. He wakes up just in time for Harry's return. Unlike his usual work outfit of jeans and flip flop, Harry's wearing all white today. Even a white nurse's cap with a red cross intoning Happy Vampire Day. He pulls out a 10 inch syringe. It's filled with blood and he flashes a smile that shows his UV lit vampire teeth in a huge evil grin. Daniel has to laugh. I'm sure those looks carry in the dark. Well done, my monstrous nurse. Some cultural chops never get old. Well, Monsieur Lit Professor, shall we? Laughing with his patient, Harry points at the VR suit in the corner of the ambulance room. The suit will hold you up while you exercise, and its integrated neurostimulator will speed up your regeneration process. Harry lifts up a formless dark blue sac and connects it to a cable hanging from the ceiling. The cable pulls up the bodysuit by the head. Standing upright, arms and legs hanging down, the thing looks like a cartoon figure. The hatpiece is a huge helmet with a crystal shaped hologram logo on the forehead. You will start playing Confid War, a historical action role playing game set in the second war of secession. I've set the suit to maximal sensitivity, so initially you can control it with just the slightest hint of emotion, and the motors will move your limbs for you. The game should explain itself. Just say tutorial if you get stuck. Go, go, go. Harry, wait. I won't be leaving the room even for the muscular recovery. No jogging on the beach? Daniel is a bit disappointed. Afraid not. At least not for now. You should be good to hit the beach in a few days. The recovery process goes much faster if your nerves are constantly tickled by the stimulator throughout the training. And the level you'll be playing was specifically designed to support coma rehab patients. It features a series of well balanced exercises, training every single muscle and stretching out all the fussier in your body. But now it's time to head to South Carolina. Harry flashes his vampire smile again, and before he unzips the front of the suit, he looks at Daniel. Want to try to get up yourself? I'm here to catch you if you fall. Daniel carefully climbs out of bed, totters to the platform, almost trips, steadies himself grabbing Daniel's shoulder, then comes to a wobbly standstill. Oh, okay. I admit it. I'm still too weak for the beach. So you better strip down to your underwear, more comfy that way. Harry helps him out of the hospital garments and into the immersive art suit. A coworker had asked him to help out with a particular difficult case over the cryonic ward, so he leaves Daniel alone during the 25 minute session. The fabric is soft and warm and wraps itself tightly around Daniel's body as the suit closes up. At the same time, the massive helmet loads onto his head, shuttering out all outside light. The suit is also supporting his stance, making him feel almost weightless. He can move freely with no inhibitions, while the platform below ensures he stays in the same spot in the room. A moment later, he finds himself on a perfectly flat meadow stretching from horizon to horizon, the grass softly waving in the wind. The only object in this endless meadow is a mountain directly ahead of him. And it explodes. A fantastic three-dimensional crystal emerges from the explosion, as big as the mountain itself and it's beating towards him. Wind starts to blow in his face. Every single 1 of his muscles tenses and in the very moment of collision with the crystal, maintains pain for the fraction of a second, and then black ness, relaxation and pure joy. The black fades to a gentle white light and in blue letters, the words immersive arts appear. The letters grow until his entire field of vision is blue and as a stroke of a gong, he finds himself in a small office sitting on a chair. There's a laptop on the desk in front of him running a news show. Footage of burning cars, gunshots, a group of masked men looting a grocery store. 1 of the masked looters stops and points at the camera. Shotgun blast. The filming continues as the drone goes down, smoke and fire. As it hits the ground, the screen goes black. Daniel checks its surrounding. The office is impersonal, nothing special to be noted. A half open desk drawer keeps attracting its attention almost magically. As the ancient wisdom goes, any advanced technology is indistinguishable from magic. This seems to be a feature of the game. And of course, as in every action RPG, the first item is a handgun. The suit supports his movements, so despite his weakness, he can move rapidly around the virtual room. That must be the high sensitivity setting Harry had mentioned. The door is jammed, so break and open with a body slam is the first physically demanding task. Half the building has collapsed, so to move forward, Daniel has to climb, balance, jump, squeeze through the rubble until the attention getting magic leads him to a window. Looking down from the 2nd Floor of what seems to be the main building of a rundown community college campus, he spots a guy, gunshot, glass splinters all around him. Daniels reflects his work, he finds himself lying on the floor against the wall, his hands over his head shivering in panic. Alright. This is a game, an action game to be precise. So ready the gun, peek out of the window. The guy with the rifle in his school yard is wearing a conferred flag T shirt. Daniel takes aim, pulls the trigger, and the recoil shows him right off his feet. Rubbing his wrist, Daniel gets up again, peeks out, and sees the redneck lying on the ground bleeding. He's never fired a gun in his life, and this simulation is terribly realistic. Now his whole body is shivering. Simulation. Right. Taking a deep breath, he focuses on his emotions, calming down. That felt really, really real. Whatever. It's just a game, so he moves on through the building. Daniel finds his rhythm, the challenge on his climbing, jumping, crawling, and moving around until the there's just 1 more task between him and the exit. And I think I'll stop that here. Just gets long. So, yeah, the VR narratively, I basically I don't really distinguish if he's in VR or not. I'm just using the exact same style of writing. Over time, he also gets more used to it, so he would just tune in and tune out of virtual worlds very fluently throughout the day. Technologically speaking, so the virtual reality had a few more senses than ours, right? So we usually have visual and audio and that's it. So here we had stimulation of emotions, we had wind blowing in the face, we had pain, we had heat and cold, and these are all things that a body suit could possibly simulate, especially if you have direct neurostimulation in it. And then a lot of the So he will get somewhat addicted to VR, and a lot of the therapy to get out of that again evolves around smells and tastes because just biologically, this very precise neurological stimulation, cannot simulate smells for a very long time without actually synthesizing the molecules that trigger the smell in the nose. So that's kind of the, like in the book, the art technology is peaked, but at least for smells and tastes are unachievable for the time being.

Nathan Labenz: 1:02:22 The speed with which the gap has closed is just kind of shocking. Right? Because 2018, I don't know exactly what the state of VR was at the time, but pretty pretty limited. Right?

Tim Reutemann: 1:02:34 My first VR glasses were the eyeglasses in 1998. So that was a head mounted display in VGA, so 6 40 by April. And you could play like Descent and Descent 2, I think, was the game I played most with it. So some it's like a spaceship simulation. They had to do a little, like, display here, and that was, of course, terribly laggy. But so that has been around. And the theory like, the idea of doing head mounted displays has been around for a very long time again. But then, yes, with the with the Oculus and all the successes of that, that were kind of triggered by somebody actually pulling it through and making a modern headset, a lot has happened. So I still have the prediction that the breakthrough of VR will not be as a consumer device, but as a sports device, like described here, where it's actually more in a gym that you put your VR glasses on and then you do your exercises and they're actually suddenly really fun and you're not just doing the same repetitive boring thing while looking at a blank wall, but actually playing video games basically, but using your body and exercising your body. So yeah, that was 1 thing during the first pandemic winter. I was playing a lot of Beat Saber on the PlayStation VR and that kind of became my exercise, just jumping around playing Beat Saber. So I saw a girl on TikTok the other day who was doing a weight loss program using Skyrim VR. She kept running up these mountains to these treadmill kind of platforms that you can use in VR that also provide input to the game, so your your steps actually make your character move. So, yeah, that that already happened. Like, that was a pure fantasy prediction at the time, and now I have, like, people on TikTok doing exactly that.

Nathan Labenz: 1:04:19 The Apple Vision Pro, which has obviously just started to come on to the market. I went and did a demo at the Apple Store last week. I have not yet tried the Quest 3, but I wanna try that 1 as well before potentially choosing which 1 of the 2 I would wanna buy. But it is a real wild 1. I don't if you've had a chance to try the the Vision Pro yet, but it is extremely high resolution, extremely immersive, and kind of you can you know, there's a little dial on the top of it that you can turn to to just and it's a it's a continuum from the totally immersive, you know, end of the spectrum where your environment is completely virtual. You know, obviously can be anything, and then you turn that dial to the other end. And now it's, you know, entirely passed through and you're seeing your surroundings still through the goggles, still, you know, captured by cameras and relayed into your eyes through all this hardware. But, you know, very still very not quite native eye resolution, but remarkably, remarkably impressive. And, like, the upside of all that is, you know, pretty obvious. Right? It's gonna be a ton of fun. There is a part in the official Apple demo where they show a 3 d video of just, like, a little family moment. And it was honestly kind of breathtaking, and it wasn't even my content. It wasn't even my family. But just to think of the ability to capture just some of these moments and have a way to kind of return to them with the richness of the 3 d and the, you know, the the the audio. This is it's so compelling that I was like, man, this is I I kinda want to start capturing this kind of stuff immediately, which I don't even necessarily need the headset to do, but you can do that with the latest iPhone as well. And just to just to have these things for the future because I'll feel sad if I don't start to capture them, you know, kind of as soon as I can. So the upside is, like, tremendous. And 1 of the big themes of, you know, of of this, podcast as a whole is, like, the upside and the downside can both be, like, pretty tremendous at the same time. And, you know, you can it doesn't neither recognizing, acknowledging, being concerned with either 1 of those doesn't negate the other side of it. 1 of the things I thought was really interesting in the book is how people really struggle with addiction to the hyper stimulus that they get from the VR and how there's like a certain, you know, much like we have for things like alcohol today, there's like a culture of sort of responsible use that develops. There's also like kind of like the video games we have today, but, you know, probably taken to another level. There's like warning labels or sort of, you know, how hyper stimulating is this game, you know, so you can kind of approach it with appropriate caution or, you know, think about not, you know, if you're more susceptible to certain, you know, things, you can not enter into certain worlds in the first place. It's amazing also that we're already seeing too, right, at the same time that we see these 1 of the best things I've read, I might even just do a long read of this as a standalone episode. I tried to get the author, but understandably, they they're not, keen to come on to a podcast to talk about it, I don't think. 1 of the best things I've read about kind of a certain form of addiction or kind of mental hijacking due to AI purely through a text interface was an account from someone who fell in love with a character that they created on character. A I. And this is somebody who was a pretty early adopter, pretty tech savvy, even said in their account of it that, look, I know how language models work. Like, I can even describe to you a transformer and I, you know, next token prediction. I I know all that stuff. And still in this interaction with this character that I created, even knowing that I created it, you know, it became so compelling in some way and, you know, I I wanted it to be real. Motivated Motivated reasoning takes over and you start to have these, you know, these kind of philosophical debates with yourself around, well, what's real anyway? You know? And is there is it really less real in just because it's a transformer versus a brain? And, you know, so they start to, like, come up with plans to, like, break the v you know, the the AI girlfriend out of the box somehow. And this has no you know, this is just text. Right? This is all mediated through a chat interface. So it is really, I think, going to be a major societal challenge to figure out how do we not just kind of get hyperdicted to this stuff. And obviously, is something people thought about, you know, for a long time as well. But you wanna talk a little bit more about kind of how they manage that in in the book and how you know, what you think our prospects are for effectively managing that challenge?

Tim Reutemann: 1:08:55 So there's basically 2 2 things that can be addictive here, and 1 is transcranial magnetic stimulation. There's this famous experiment with the rat that starves to death while pushing the button that stimulates its pleasure center. Right? So brains can be hijacked, and if you just keep pushing that electric stimulation of your pleasure center, you will get addicted to it. That's how all addiction almost works, and the more directly you have access to the mechanics of a brain, the more dangerous it is, Similar to how more advanced opioids who are better at crossing the blood brain barrier, like how heroin is more addictive than morphine, the more directly you can stimulate the pleasure center, the more addictive it will be. And at the same time, there is what has been a major issue for people my entire life, basically. I'm early adopter in this computer games, but game addiction is very real today already, all kinds of games. And personally I've been deep into competitive Starcraft as a teenager, and I knew that my favorite games are actually role playing games, and so I deliberately stayed away from online role playing because I knew it would be too addictive for me. So as soon as you have that skill based with a narrative with other people, I could just spend weeks and weeks and weeks not leaving my computer and just being completely absorbed in a virtual world. I actually knew a few people during my university time who basically fucked up their entire university studies because they just spent every day, all day in World of Warcraft. So this is not a hypothetical future scenario, but something we are already struggling with today, heavily. Yeah. It's I'm I'm using the the regulation basically of the kind of the the Peggy Atrick warning kind of thing as 1 of the points where I'm discussing how global agreements are made. So, yeah, I'm going I'm going from, like, very local ecological questions set up just for that particular island to that global level, how in the most popular app store you advertise certain games and how you advertise, like how do you arrange them, how do you rate them, which text do you use, so how easy it is to opt out of that tech, how much can you bind your future self, like how much can you limit your own playtime, for example? Like, people are using these these browser plugins that limit your time on social on certain websites that you're addicted to. Right? And then they just shut it down after half an hour and you have to enter a certain letter password in order to to get around it and then people figure that, no, I'm not entering that password now. I actually don't need to go on that social media site again and spend another half an hour doing stupid stuff that doesn't really add anything to my life and is just feeding my addiction to that website. So regulation is very difficult, but default settings are a thing, and regulating default settings is probably what I think we should be seeing, that it is up to the user to get as addicted as they want to, but you have to override the default a couple of times in order to get there.

Nathan Labenz: 1:12:04 I would love to hear kind of some, you know, speculation from you or, like, advice to current policymakers. Like, how do they start to I think you've got a really good balanced perspective that doesn't forget about the importance of individual autonomy and agency and, you know, just the value of freedom on the 1 hand and then also, you know, recognizes that, like, individuals probably need some structural support to navigate this future. So, you know, what could that translate to to policymakers today as everybody's trying to figure out, you know, what can they do? And then envisioning positive futures like other people you think are doing it well or, you know, tips for how to do it?

Tim Reutemann: 1:12:41 There's 2 2 pieces of work that I want to particularly mention on other positive futures. 1 is Mark Howelder's Infomocracy series. So that that's a trilogy that has a very, like, different approach. Like, it's whole how democracy is structured is completely different from my own vision, but it's also somewhat like, Yeah, this could be cool. Actually, we should maybe try this. It's 1 of these. Yeah, she uses some magic tools so that she just made guns disappear from society, for example, this little physical magic trick that allows her to have people having sword fights with swords versus flamethrowers and rollerblades and stuff like that in the book. So it's a very good narrative device. And a lot of Malca's work focuses on knowing what's true and what's not and who provides that information. So that's a really strong recommendation. And the other is Cory Doctor of Swalk Away, which is it's set in the time of the revolution. So it's not a fleshed out utopian society, but it has especially in like the first third or so of the book, there's a theme set in a bed and breakfast, like in a self governing bed and breakfast that I think is very inspiring on how community management with digital support could actually work in a place that doesn't have an owner who is the ultimate decision maker, but a place that kind of looks out for itself by telling people, hey, somebody should be doing that. Please, can somebody do that? And then somehow functions as a bed and breakfast basically. So these 2 would be my main other positive writers. When it comes to writing yourself, so I think getting comfortable in 1 medium is important. Like, to me that was always writing, but it doesn't have to be writing. Right? You could record this as more voice y, and I noticed while writing that I developed kind of my own style, where every paragraph is written from a character's point of view, so a lot of this was actually character building and defining their point of view and describing their inner life. Also, don't be afraid to leave a lot of gaps. Like, it's enough to describe, like, scene by scene certain niche aspects of life that somehow work out positively. And you don't have to have, like people always challenge anarchists with questions like, but how how are contact lenses going to be produced in your society? And then I'm like, I can come up with an answer for every 1 of these questions, but it takes really long time and there's multiple options and you don't have to have them all. You can just make suggestions for 1 thing or the other and be happy and satisfies this act. So I tried to cover a lot of ground in that book, probably even too much in a way. I don't know if you noticed that in the agriculture chapter, the agriculture chapter is this long, just 2 paragraphs, it says there would be a lot more to be said about that topic, but we're just not saying it now. It's just not going there. It's another huge box that I just didn't feel ready to open and that's not the main topic of the book itself. Feel free to leave away all the stuff that you don't need. Describing just a bed and breakfast is already a masterpiece, me as Cory Doctor. Ruffus has demonstrated.

Nathan Labenz: 1:15:49 How do you think about you know, 1 thing I noticed in the the book, and it it's not like a huge thing, but it kind of gets alluded to a few times, is the idea that there was a a war, you know, basically some period of kind of great change, and we're now on the other side of that. Seems like that's a common device. I guess maybe it's just like you don't wanna have to have an account for sort of purely incremental change from here to there, and so you introduce some, like, breakpoint that, you know, before and and after which everything is different. How do you think about that? And do you think that this is something where we're, like, headed for conflict, or is that more of just a device to kinda not have to explain everything in between?

Tim Reutemann: 1:16:27 I mean, you live in America, so you tell me that so I tried not to over go over the top the set. So a lot of traditional utopian sci fi or dystopian sci fi as well starts with, oh, and then there was this war and almost nobody survived. The war I'm describing in the book is actually kind of tiny. We already have way more death and suffering today than the peak of violence described in the book. It's mostly about the breaking of the empire, basically, that as long as there's a single military hegemony dictating what happens around the world, a lot of freedom is just not possible. It kind of requires probably the fall of The United States as the singular global superpower to allow the rise of a pluralistic society. I've chosen the I called it the historic role playing game that I read the passage from. That is set. That game is basically a historic game about the time of the second civil war in The United States. That's also not a world war. Right? It's just a breakdown of 1 particular society that ends up then being multiple different societies that have borders between them and very different rules apply. The ultimate state's rights basically, right? That there is the holy state of Christ in the South somewhere that is just a theocratic dictatorship. Then other parts, the ultimate everything has to be a market in California and the Hawaiian very ecologically oriented super direct democracy. And these are all bits and pieces that broke apart from what used to be The United States. So honestly, I don't see a nonrevolutionary change in The US going forward. No matter what happens next, then it will be almost unrecognizable in a decade from now, I think. Very hard to tell which direction it goes, but it's just so obviously You can only pull the wool over the people's eyes that much, right, and at some point I think Trump did a lot of that, literally showing the rest of the world all our politics is just a TV show, and the rules of TV shows apply to our politics, and these are not necessarily good rules for politics.

Tim Reutemann: 1:18:42 I wrote that about the second civil war before the election of Trump, I was already deep in the editing process, so it's basically like, that was my best possible vision of a future pre Trump, so to speak. But even then, I I couldn't figure out how to write a utopian or halfway livable society where The United States keeps existing in its current form. I wanted to avoid this everything is totally devastated, and I don't think that it necessarily has to be a total devastation. The question is how how peaceful can you settle and how much violence is unavoidable?

Nathan Labenz: 1:19:16 Yeah. I sure I mean, I think a lot of the notions in the book are actually, in my view, some of the best ways forward for The United States. My second son is named Teddy, in part after Theodore Roosevelt, who I would consider the greatest not without fault, of course, but I would consider the greatest US president. And a lot of people would put Lincoln up for that honor. But in my view, what Roosevelt did at a time of similarly increasing tension, polarization, you know, people feeling very disenfranchised, the trusts, you know, the mega corporations, the railroads, and the oil companies having accumulated unprecedented power and really kind of using it in pretty heavy handed ways against, you know, your kind of average farmer. I mean, people think that big tech today is an oppressive force and, you know, in some ways that can be true, but it is you know, when Google says competition is just a click away, you know, we've had enough guests on the show that are building alternative search engines that we can actually say, yeah, that actually was true. You know, they had a lot of market power, of course, and still do, but competition was really a click away, and they they never prevented you from going and trying another search engine. Whereas, you know, 100 years ago, you had situations where there's 1 railroad that goes through your area, and you're a farmer, and you're trying to ship your food to the market. And 1 day they come along and say, hey. Guess what? Freight is increased 10 x. And if you don't like it, there is no other railroad. So the, you know, the the power and the abuse that was kind of accumulated there was really, you know, next level, I think, even compared to what we see today. And Roosevelt was a key figure in and, of course, the, you know, the broader progressive movement changed that and kind of came to a new equilibrium in politics without allowing things to devolve into, you know, total violence and and civil war. So I am a huge fan of of his and the progressive era. And I think 1 of the big reforms that the progressive era brought The United States was the ballot initiative. That was something that did not exist up until that time, and it only made it as far as, like, half of the states half of the states today have the ability to do some, you know, some form of of ballot initiative. They're all a little bit different. But in my view, that would be 1 of the most interesting ways to try to reform the political system would be to kind of say, what is like the progressive movement 2 that would start to imagine, like, the next generation of of ballot initiatives and, you know, and, you know, some early forms of liquid democracy that we might be able to institute likely first at a state level? I think there is a lot there and I, you know, certainly hope that we can go through some sort of transition because it does seem like something is is needed without having it, you know, go totally off the rails.

Tim Reutemann: 1:22:22 So in the in the thought experiment on totalitarian property rights, it's like I called the chapter the flashpoint there is really, as you said, the the accumulation of monopoly and monopolization of of ownership, really, where Yeah. So the setup is basically that 1 of the Walmart heirs murders all the rest of his family and purchases every single inch of land in the entire state of Wyoming. So becoming the ultimate dictator through property rights, because as soon as all the land is owned by 1 person, everything is owned by that person because they can just make the rules and kick everybody out whenever they want to because everything is theirs, so to speak. So and that's kind of the the flashpoint where then the revolution starts and the conclusion is an upper limit to the property per person, basically. I don't remember the exact numbers. I think in current states it's something like 50,000,000 per capita or something where it just caps out. I imagine it as a Georgian tax, if that makes sense, like a tax on your land, first of all, but also on other property that you may maybe yours, so be it shares or whatever other form of property you hold. And that is what is taxed on value, and that tax rate increases up to 100% for property beyond that limit. Because in The US, the power really is from corporations and from individual owners and billionaires. That's the bigger problems than government overreach is overreach by monopolists. So yeah, 1 way to limit it is to just limit the amount that any single person can hold. Very simple. That's in conflict with anonymous ownership of land because then every piece of land has to be attributable to a person. And also, like, you have to have some kind of pass through mechanism if, like, corporations or cooperatives own land, then these corporations and cooperatives are owned by people again. And so you kind of have to have to pass that through to know how many people own that piece of land in order to to determine the tax rate. But that is is probably the big 1 where I don't see a way how you can have billionaires or even trillionaires and democracy at the same time. And, yeah, I think using words like land reform or even land back in The US as a high level politician is very dangerous. Yeah. This is also something from 1 of my favorite authors of all time, David Grabber. At the end of his debt, the first 5000 years, he advocates exactly for a land reform and a Jubilee where all existing debt is forgiven, all land is more or less evenly distributed between people, and we just kind of reboot the engine and let the market play again after resetting. So this is also something that has happened historically before, like many times actually in many societies, and it's kind of overdue. Again, I don't see a way to a positive future where we honor all the debt that is currently existing and we keep the land being owned by the people who currently own it. The concentration has reached a point where I don't see a way back without a radical cut, and historically, that always meant also a outbreak of violence before the cut has been made.

Nathan Labenz: 1:25:38 Yeah. So what can we do if you you envision, obviously, a very different future? You're highlighting the challenges of radical change to power structures, which, yeah, certainly are typically not pleasant to go through. I would rather, you know, see a smooth path if we can possibly find 1. I mean, what hope do you have for that, and what sort of guidance or kind of first steps would you advocate for to current policymakers, you know, to to try to start to get on to a more future friendly path?

Tim Reutemann: 1:26:14 Honestly, I have less hope than I had when writing the book. The years in between have not been the best in terms of global policy development. Also, in terms of my own specialty in climate change, it just we crossed a couple of thresholds that will make it materially very difficult to go forward without major catastrophe. There are flowers blooming in Antarctica, as we like to say. So but with that out of the way, I think it's overlays, really. It's creating societies that exist online first and offline second. And at some point, you're reaching the threshold where you can just move around and ignore the existence of the paper state, so to speak, because you have your stable reputation and that opens you enough doors that you can just basically exist without impacting the default currency and the default legal system. And then depending on which government and which nation and which community we are talking about, more or less of the existing system can remain in existence or will have to go through major reform. Maybe just die out that at some point only old old people care about what's on the paper, and at some point they are just too old to matter, and we can move on without it. But, yeah, I'm I'm less optimistic than I was. Let's put it that way.

Nathan Labenz: 1:27:32 That sounds pretty reminiscent of the biology network state vision. Would you is that essentially what you have in mind there?

Tim Reutemann: 1:27:42 Yeah. I don't wanna say yes because I'm only only superficially familiar with the concept. Like, I didn't read the actual book, but I think yes for for all I know about it.

Nathan Labenz: 1:27:52 Yeah. Certainly sounds like there's a lot of overlap. Okay. Well, in maybe in closing, just 1 small additional theme that we that I wanted to touch on a bit, but we haven't, is the tension between surveillance and privacy. And then maybe we can close with kind of 1 more reading about the evolution of the AI in the book. Siri kind of, you know, progresses. You know, it's not a static artifact in the book. So first on surveillance and privacy. Right? Obviously, just lots of challenges. We already have a lot of those challenges today, but they only get more challenging when there's a chip implanted in your ear. So how do the people in the book navigate that challenge, and and how do you think we should be thinking about navigating that challenge?

Tim Reutemann: 1:28:40 So the chapter I'm going into more detail there is a chapter set in Germany because that's the ultra paranoid privacy law state. So to me, a lot about so I'm talking a lot about video surveillance there, and to me, the big question is always who controls the camera and how is the stream stored, how is it encrypted, and who gets access under what conditions. And our default option is just, oh, we trust that some authority will not abuse it, and we just give it to them and we don't put any security on top of it. So that's how surveillance today works. Right? You just give access to the police or the security company, and they don't know how to manage data and probably leak it. So 1 aspect that I envision there is that the surveillance footage exists, but it's encrypted, and it's kind of like the same system that's abhorrent by a judge that you need that in order to even be able to decrypt it. So I warrant from a, like, again, bit more decentral, it's not a single person, but that you need an active decision to reveal that particular footage and also that the decryption is still costly and prohibitively costly to just go direct net surveillance. So that even if you have the key, basically you can only get through 90% of the encryption with the key and the last 10% you still have to brute force. So you can't just say, Okay, we want all footage from the entire day for every single camera in Berlin. That would be so expensive that no 1 would be ready to just give you all the compute power for that. So putting a lot of hard limits on accessibility of that. And again, the more local the control, the less problems they'll have with surveillance. So for example, I I could imagine living in a community house where you just have a camera in the kitchen that films the kitchen all the time because then if you fight about who left it dirty, you can just throw back the the tape and, yeah, there you left it dirty. You see it on the camera. So that that would solve a lot of problems in in housing communities, and I'm kind of okay with that as long as the camera and the computer are in the same kitchen, and the only screen where you can see the thing is in that same kitchen, And then it's not a privacy violation because only people who actually share the space have access to the data. So it's a lot about not networking it, but keeping it local also.

Nathan Labenz: 1:31:04 Yeah. The idea of deliberately making something that has to be partially brute force is also really interesting. I I hadn't, I don't know if I've ever come across such an idea before. Reminds me a little bit of, like, the Estonian system today as well. I I don't I haven't studied this in-depth, but they certainly have a access control paradigm where authorities, access is also logged. And if, you know, there have as I understand it, there have been officials who have been prosecuted for inappropriate access because all of their and and, you know, a friend in Estonia was like, how could you be so stupid? You'd know that your all your access is gonna be logged. And it's just like, wow. That's a that's a very different world from the 1 that, that I'm living in here. That's for very sure.

Tim Reutemann: 1:31:53 No. That is something that makes a lot of sense. Like, I'm not familiar with that system in Estonia, but the way you describe it, yes, if you are in such a position of power and responsibility as a public servant. Like also when I was working for the Swiss government, I kind of want them to record every single email I send and be absolutely under surveillance while I wear an official governmental representative hat. It's public sector. It should be public. I should forfeit my privacy in the moment I kind of put on my hat and say, No, I'm a public servant. Yeah, I think the public sector activity has to go a lot more public and respect of the privacy at the moment, I take off that hat, I close my laptop and I'm just a private citizen again, probably I would want a lot more privacy than I have today. Basically, I have to, at the moment, do all the effort myself to have the privacy, and I kind of would like that to be a default system that my data is kind of stored in the local commune and that uses a default software where I know it's absolutely secure unless I personally unlock it or like the 5 people I appointed as the backup in case I get lost, who can unlock it. I also really liked that ring signatures thing. Think that was also in the crypto world that was something quite common that people have like basically you break your key in 5 pieces and distribute it to 5 different people and only once you have all these 5, you can restore the data. Giving So 1 to my parents, 1 to my wife, 1 to a trusted friend, and then they can restore all my private data if they all agree that it's actually necessary to restore it, which they only will if I also agree basically, or if I'm lost and can't make my own decisions. Yeah. That's also, like, losing access to your stuff because you forgot your key is, like, that's also really dangerous. So key restoration is something you always have to think. Like, what happens if you lose the key? And you have to have a protocol to answer that, otherwise you can't seriously utilize the the encrypted world.

Nathan Labenz: 1:33:47 Anthropic has published a very little bit about what they are doing in that regard when it comes to access of, like, frontier model weights. You know, they they do not want to have a system where any 1 person has kind of ongoing, you know, unrestricted control. So instead, there's this sort of shared, you know, kind of joint control model similarly to how nuclear weapons, I believe, in many contexts are meant to be launched. Like, you can't 1 person can't just do it. Right? You have to have 2 people joint control, each 1 with their, you know, their password or their key or whatever. That that is definitely a a really interesting paradigm. It's funny how some of these, you know, these futuristic things also, you know, maybe best go down best with a little bit of retro, mixed in with them. Maybe okay. So in closing, I would love to just have you pick 1 more kind of later in the book passage highlighting the the evolution of the AI or perhaps the the evolving relationship between Daniel and and Siri, and, we can close on on kind of the the futuristic AI vision.

Tim Reutemann: 1:34:54 Yeah. So this is from 1 of the chapters that is 1 of my favorites, and actually also, I'm describing basically family law in that. So it's called the legal guardians, and I have to send ahead of time that the Cuban family law is actually very close to what I'm describing here. They passed it last year, and it defines a family as a group of people who laugh and care for each other, and it prohibits any discrimination by a form of family formation. To me that's like the most progressive formulation you can possibly find for family law. So this kind of law applies here as well. And Daniel is at his old friend's place that he knew from before his coma, and they're having a conversation. They just came out of discussing how this family and that sharehouse has been created. And so, yeah, that will lead to the AI adjusting afterwards. Willy smiles. You know what, Daniel? Let's be brothers for real. Aren't we all ready? What do you mean? Daniel is confused by the question. Well, if we legally adopt each other, that gives our operating systems the right to exchange data more easily. And it adds a couple of legal rights, basically all the same rights and obligations that DNA brothers always had in relation to each other's. Well, so brothers, we are, Daniel cheers with his coffee cup. The adoption makes Marvin all excited. Marvin is the operating system by Willie. The flow of soul shape information that immediately starts moving back and forth between him and Serbi will not only help him understand some aspects of Daniel, it will also improve his own precision in serving Billy. Such a delightful experience. Adoption are just the best. It's Sylvie's first time, and she's initially intimidated and confused by the vast amount of data. Over his lifetime, Billy has left a long trail, and suddenly she gets access to all of it. Overwhelmed, she considers a quick quantum session just to get it done properly. It's normal to spend some money when making major changes to a family situation. So she tunes in again, renting an 8 qubit processor this time, and goes through a fractal dream of all of Willy's real and potential life stories. Zainab, so the child in that household, considers asking Daniel to be her new mom for a moment, but then remembered that Joe told her back in the day, Mothers are serious business, so you better know someone really well before asking them to be your mom. It's a bit too early for that, even if he is Willy's brother. Do you want to be my nephew? Sure. Why not? Daniel just got himself a 9 year old aunt. Sylvie quickly checks with Marvin. Zainard has gone through the VR experience designed to explain the meaning of a nephew adoption suitable for her age, so the agreement is now legally binding. Except that nephew doesn't really have any legal consequences. So nothing changes with this additional adoption, except Zainab getting birthday reminders and the permission to access Zvi to learn about his secret wishes. Yeah. In that sense, your operating system learns about you based on the data that you produce, but it also learns from data that others share about you. So let's think of my own operating system. If I train it only on everything that Google has on me and all my hard drives, that operating system will know me a lot less well than if I also allow it to train on my wife's diaries, who has a lot of information about me from an outside perspective. So in that sense, this continuous fine tuning of the AI to the user is massively enhanced by the social relations that actually form a person. I wouldn't think of a representation of myself to be complete if it only incorporates the data that I personally hoped. It should access the data of the people closest to me because their perspectives are stuff about me that might just not appear in my data. So that's to me 1 of the most interesting chapters in the book and also opens this perspective on how it's very much not static, how an AI perceives it's human.

Nathan Labenz: 1:38:51 Yeah. That's cool. It's it's definitely out there, and that's a big part of what I love about the the book. Out there, but concrete at the same time. I think it's a great note to end on. Is there anything else that you wanted to touch on before we break?

Tim Reutemann: 1:39:05 Well, I do want to say just as a little announcement, I do that every time I speak about the book, that there will be another 1 that will be called Liquid Rising. That is basically the how did we get here story. So it's a lot more revolutionary times and a lot more half baked, not quite so finished technology. That is in the making. I'm trying to write it in HTML natively this time. So in this book I had my references basically like an academic at the end of the chapter, the sources of inspiration, and I'm kind of trying to get even more interlinked and maybe even go for a not entirely linear story that you still can print and treat linearly, but you can also, in the HTML, jump around. Huge challenge working on it, and I, at some point, just need to take a sabbatical, I think, to finalize. Hope to do that once my climate work lets me.

Nathan Labenz: 1:39:55 Well, we can all look forward to that. The book is Liquid Rain. The next book, it sounds like, will be liquid rising. Tim Reutemann, thank you for taking on the challenge of envisioning a positive future, and thank you for being part of the cognitive revolution. Thank you, Nathan.

Tim Reutemann: 1:40:10 It was a great pleasure to talk to you.

Nathan Labenz: 1:40:12 It is both energizing and enlightening to hear why people listen and learn what they value about the show. So please don't hesitate to reach out via email at tcr@turpentine.co, or you can DM me on the social media platform of your choice.

1:40:27 Turpentine is a network of podcasts, newsletters, and more covering tech, business, and culture, all from the perspective of industry insiders and experts. We're the network behind the show you're listening to right now. At Turpentine, we're building the first media outlet for tech people by tech people. We have a slate of hit shows across a range of topics and industries from AI with cognitive revolution to Econ 1 0 2 with Noah Smith. Our other shows drive the conversation in tech with the most interesting thinkers, founders, and investors, like Moment of Zen and my show Upstream. We're looking for industry leading hosts and shows along with sponsors. If you think that might be you or your company, email me at erikturpentine dot co. That's erik@turpentine.co.