If AI Is Sentient Then So Is ‘Age of Empires II

0:00The claim that an LLM could be conscious or not is just inherently ludicrous because it would first require you to understand what consciousness is. And it's like whatever computer scientist at NVIDIA or OpenAI or Anthropic, which would make the claim, it's like, we've achieved consciousness internally. No, you didn't because you like probably don't even understand what that is. Even like the greatest minds on this issue don't fully understand what it is.

0:29Hello and welcome to the 404 Media Podcast, where we bring you unparalleled access to hidden worlds, both online and IRL. 404 Media is a journalist founded company and needs your support. To subscribe, go to 404media.co. As well as bonus content every single week, subscribers also get access to additional episodes where we respond to their best comments. gain access to that content at 404media.co. Also, remember to subscribe to our YouTube channel where you can watch all of our podcast episodes. Subscribe at youtube.com slash at 404media.co. I'm your host, Joseph. And with me are 404media co-founders, Sam Cole.

1:13Hello. Emmanuel Mayberg. Hello. And our regular contributor, Matthew Galt. Hello, hello. All right. So this first story, really, really fun, really, really interesting. The headline, which is probably also just going to be the title of the podcast, to be honest, is if AI is sentient, then so is Age of Empires 2. Before we get into this, what's everybody's experience of Age of Empires? Matthew, I'll ask you first as you wrote this, like, were you a big Age of Empires guy? I played Age of Empires, but I was not a big Age of Empires guy. In this space, I was much more of a StarCraft. There was a lot of StarCraft growing up that kind of filled the Age of Empires niche for me.

2:02Gotcha. Sam, were you into it? Not really. I mean, I was aware of it. I wasn't like Age of Empires head. I guess I was, I don't know. I was into Command and Conquer and stuff like that. I wasn't super into the world-building games. I'm kind of blown away by the lack of response to Age of Empires here. Emmanuel, please save us, surely. I've played it a bit. I agree with everyone else that it's like I've played other and I would argue better strategy games at the time. However, what I do remember about Age of Empires more than anything is that it had really amazing cheat codes. Do you remember this?

2:47Dude, yes. E equals MC squared to get the robot or the car, right? In Age of Empires 1. Yeah, there's the car that shoots lasers, which I always thought was very fun because you're playing like a historical game and there's like Iron Age, whatever, Spearman, and then you spawn a car that shoots lasers and melt them. I thought that was fun. I fucking loved it. I played so much Age of Empires 1 using the cheat codes often because I was a literal child. and very, very bad at the game. I would start it with, oh, you know, build a fishing hut and then do this. And then by the start of the second age, we're like, okay, it's time to bring out the robots, I think.

3:28And then Age of Empires 2, I played an absolute ton of as well. I might actually go back now because they've re-released some of them. They just remastered it, right? Like it's a little bit updated and nicer. Right. And that would be very interesting to see if the person we're going to talk about could build their goat-based LLM in the remaster of Age of Empires. But so we have this paper, which we're going to talk about. And the title is, If LLMs have human-like attributes, then so does Age of Empires 2. Obviously, this is surreal. It is supposed to annoy people, and I'm sure it did. It really did, yeah.

4:11It really did, yes, in a very, very good way. We'll get into the arguments, but who wrote it, first of all, Matthew? And what is their affiliation? Because that's pretty important. His name is Adrian De Winter, and he is an AI scientist working at Microsoft and the University of York. He's kind of a top-level applied scientist and researcher. This didn't come out of Microsoft. No, this is his own independent study, I would say. That's a good way of putting it. And he's got, if you go back and look at his website and other thing paper that he's written, and he has a substack version of this paper that kind of breaks it down into more, I guess, non-scientific language that's pretty entertaining and mixes some memes and such in there.

4:56He's got a couple different experiments and things that he's done, experiments, a couple different white papers that he's written about different topics that are pretty entertaining. A lot of absurdism kind of just all over everything. Yeah, and I'll just pull out this quote from the piece that he told you. He said, I have this tendency to dial up things to 11 when I really think I need to make a point. I should also note that absurdism is pretty standard in philosophy and theoretical computer science. I mean, that first bit of the quote is the energy we definitely need. Absolutely. And it's kind of interesting because he was very entertaining to talk to.

5:33Actually, we were talking back and forth through emails and like a lot of, you know, you guys, a lot of subjects are kind of like, oh, rote and you get the basic standard stuff. There was a whimsy and chaos to him that I found very engaging and entertaining, even in like written communication. Nice. So I'm going to massively oversimplify, obviously, just to get us there. But basically, he built an LLM with the goats in Age of Empires to make this point we'll go on about. just drill down what did he actually do to build this LLM with goats in a way that listeners are going to be able to grasp without necessarily reading the paper, which is a fine line to draw right now.

6:17And there are visual aids in the paper and on his website and in our story that'll kind of help you see it if you need that. Essentially, what he did is he built a very, very basic, the kind of the underpinnings of the way of like a neural network in Age of Empires 2. The way you do that is you go to the sandbox mode, which is the thing that Age of Empires has that allows you to make your own maps and crucially like your own quests so you can set like conditional gates along the way. So like if a condition is met, then open this door, let something through. And what he did was basically built a bunch of very simple logic gates, literally like gates that would open and goats would move along a path and go through the gates and the way you kind of visualize it is like you've got you know like all computer code is broken down into ones and zeros and he had two different tracks one was for one and one was for zero and the goats moving along these tracks are like the signal carrier they're the bits and doing that he was able to kind of build out a very very basic like one bit perceptron which is like a sorting thing like a thing that takes binary code and like sorts it which is like the very very basic building block of all of the large language models that we use today but if you visualize it in your head you imagine like just these tracks with grass in them and the goats just move down the tracks and they hit a gate and the gate opens or closes and kind of sorts them into different areas.

8:03And that's very simply all that it is. Yeah, it's very similar, as you mentioned in the piece. It's very similar to what people make in Minecraft often. Using redstone. Yeah, and like, I mean, people have made some absolutely crazy shit in Minecraft, like fully functional computers, basically. But people aren't then going and saying, you know this is conscious or they're making a surreal point or anything they're doing it for other reasons so it makes this llm or like the foundation of an llm in um age of empires 2 how does this relate exactly to the llms we all know um and maybe don't use but we all know about them like claude or chat gpt like what is the link there or the argument he's trying to like rule between those.

8:52So essentially what he's trying to do is that he's it's very interesting because when I talked to him he was very careful to say that like it is not that these things are not conscious. It is that if we believe that they, if we think that they have sentient traits, then we must understand that the sentient traits that are presenting are also present in other things that have the same underlying technology. And he did that by picking the most absurd version of what the similar traits would be. The place that he kind of started in was, he reads a lot of these philosophical arguments in papers. And he started to notice that around like 50 % of the time, the people that are writing similar white papers just kind of take as given at the beginning of their research that an LLM is conscious or has some sort of sentient traits.

9:54And his point was like, if we're starting there, then something's already kind of broken. And we need to step back and realize that the reason we're reading these things as sentient is because we're interacting with them through a chat window. And there's something about that chat window that makes us anthropomorphize them. But if you pull away the chat window, which is just like the interaction layer for this technology, and you're just looking at a pure form of the tech that's underneath it, then you don't anthropomorphize it. Nobody's going to look at these goat tracks and say, oh my god, it's sentient.

10:31Even though it has many of the similar traits that these chat GPT and Claude do. Richard Dawkins is not going to sit down with Age of Empire 2 goat tracks and then write a lengthy op-ed about it. yeah and right at the beginning of your piece you throw to this atlantic essay which i mislead it's in my to read list my very long to read list but i really want to sit down at the weekend and go through it but they make the comparison of like i'm gonna butcher this but oh well if we think that llms you know are conscious we need to give that same sort of consideration to like microsoft word but you wouldn't do that because that would be stupid so it's very very similar sort of argument It's just here, the absurdity is obviously turned up to 11, as we say.

11:20Turning it into goats makes it... The point is to knock you out of that place where you're interacting with that chat window and anthropomorphizing the technology. And that's what he's trying to do. Right. And maybe I'll just read out a quote from them as well that he gave to you. The point of the paper is to formally show that we anthropomorphize too readily and that sometimes the claims we make with regards to LLM capabilities are too strong. It's not an easy task, given that human-like attributes is a bit of an abstract term. Emmanuel or Sam, do either of you edit this or is it Jason who's not here?

12:02I'm just wondering what you guys think on this paper. It was Jason. I'll let philosopher Emmanuel speak. So a few, was it a month ago now, I wrote about this paper that came out of Google DeepMind, or at least one Google DeepMind researcher. But it was on Google DeepMind letterhead, which claimed that, and let me try to be accurate, that LLMs can never be conscious. crucially the paper argued that they still could be AGI but they can never be conscious and I don't even want to attempt to argue or like make the argument that the paper is making because it's very complicated and I'm wondering if Galt had the same experience when he was reporting this story which is like when you get down to it and you are trying to argue whether LLMs can be conscious or not, you're really...

13:12You've already lost. You get into the much bigger, more complicated, arguably impossible task of defining what consciousness is. Exactly. We haven't even done that after literal hundreds, if not millennia, of philosophical study. We haven't landed. Yeah, but it's like what you do understand, and obviously it's like we're journalists, we're laymen, we're not philosophers. We haven't researched this our entire lives, which is what many, many people have done for hundreds of years. But what you do understand is that once you dip your toe into that question, you realize that the claim that an LLM can be conscious or not is just inherently ludicrous because it would first require you to understand what consciousness is.

14:10and it's like whatever computer scientists at NVIDIA or OpenAI or Anthropic, which would make the claim, it's like, we've achieved consciousness internally. It's like, no, you didn't because you probably don't even understand what that is because even the greatest minds on this issue don't fully understand what it is. But certainly they disagree with you. There are hundreds of books arguing what the definition is, right? Right. Did you have that experience going through it? No, no, absolutely. And I really appreciated De Winter because he was very... One of the things that struck me when I was first reading the paper is the only place that the words artificial intelligence appear in the paper is when he's citing someone else that uses that phrase.

14:53He doesn't say it. He always calls them LLMs. And he's like, I want to be very precise about all of this. And he also brought up, speaking of this consciousness question specifically in Sentience, that study that was in science like a couple weeks ago where they found out that bees have like rudimentary problem solving capabilities and he's like you know we look at this and you know some people are shocked and they're like oh does this mean the bee is conscious he's like no it means that it has rudimentary problem solving capabilities and like it probably has some sort of awareness and interacts with the world in some way but like probably not in the way that we would think of as consciousness and like we do ourselves a disservice by kind of constantly attempting to throw this we tend to believe that like the way the humans view the world is prime I guess is maybe the way to put it and like that anything that we build should be kind of working to that model and like that's you know that's what we think of when we like the way that humans interact with the world is consciousness.

16:03When there's all these other different ways to interact with the world and solve problems that don't necessarily rise to that level, then you get into the, then you get into all the discussions of like, what is this consciousness? Does a dog have the same thing? Et cetera, et cetera. And like, that's when things start to fall apart. And people start to believe their LLMs are talking to them. Yeah, especially if you're trying to, as you mentioned in the piece, especially if you're trying to do it on a binary model, which is like conscious or not. And you're going to get into a real trap with that sort of approach.

16:37People should definitely go read the full article for a more technical breakdown. I think we link to the GitHub as well. Absolutely. But with that, we'll have a quick break. And then when we come back, we're going to talk about another of Matthew's stories about the fight around a data center or rather how a city basically stole land. I'm saying that. He doesn't use the word steal. We'll be right back. after this.

17:09We have at this point an embarrassment of riches in the number of stories we've done about data brokers, the companies that quietly compile your name, your home address, your phone number, your relatives, your daily habits, and then sell that file and a lot more to basically anyone with a credit card. Advertisers get it? Sure. Also, spammers, scammers, and the occasional person you'd really prefer didn't have your address. You never opted in? There are hundreds of these brokers and opting out of them one by one is effectively a part-time job. Honestly, I tried it felt more like a full-time job, which is what Incogni does for you.

17:46Just outsource it to Incogni. Incogni acts as your personal privacy advocate, identifying the brokers holding your information and handling the legal legwork of demanding that they delete it. Incogni manages the entire request and follow-up process on a recurring schedule because this data has a notorious way of reappearing. It's kind of like whack-a-mole. But Incogni can be your secret weapon. You set it up once and watch the verified removals stack up on a dashboard. The practical result, fewer spam calls, less junk mail, and a much smaller footprint for anyone trying to find or impersonate you.

18:19For the most comprehensive protection, we recommend their unlimited plan, which includes custom removals. As journalists, we know that sensitive data doesn't just sit on broker websites. It ends up on public forums, random directories, blogs, news portals, data broker databases. So if you find your personal details exposed on a specific site, you simply send Incogni the URL and their team of privacy experts will step in to manually handle the corporate takedown request for you. Take control of your data footprint risk-free with a 30-day money-back guarantee. Go to incognite.com slash 404media and use code 404media for 60 % off an annual plan.

18:56That's code 404media at incognite.com slash 404media. I'm not really sure who decided that buying glasses should involve staring at a tiny product photo and somehow knowing whether those frames are going to look good on your face. Because every time I've looked for glasses online, that's always been the hardest part. I have a really wide head. So I always wonder, will they fit? Will they look ridiculous? Is this actually my style? I can't tell you the number of times that I've bought glasses that I've been really upset with and had to either return or honestly just eat the purchase. And that's why I've been so impressed with Warby Parker.

19:32What really sold me was their virtual try-on feature. You just use your phone's camera and you can see how different frames look on your face in real time. It sounds simple, but once you use it, you realize how much guesswork it removes from the process. These are my Warby Parker frames. I went with the Elio in Black Walnut. I really like them. I've had them for about a year now. I have Elio in another color as well. And they're basically... I grab whichever one I can find. And what I like most about them is that they feel premium without being overpriced. I've gotten a bunch of compliments on them and they've held up really well.

20:12The other thing I appreciate is that Warby Parker makes the whole experience really easy. Prescription glasses start at$95 and they also offer sunglasses like I have, contacts, online eye exams, and they even have more than 300 retail stores if you ever want an in-person opinion. And if you use vision insurance, well, they make it easy to check eligibility and apply benefits automatically. Honestly, once you buy from Warby Parker, you realize how much simpler buying glasses can be. I haven't bought sunglasses from anywhere else besides Warby Parker in the last, I don't know, five years. So I've been a longtime customer.

20:47I'm really happy. You should try out the virtual try-on feature because it saves you a lot of time and prevents you from second-guessing every frame that you look at. Right now, buy one pair of glasses and get 20 % off any additional pairs at warbyparker.com slash 404. That's 20 % off any additional pairs when you purchase one pair at warbyparker.com slash 404. One of the underrated benefits of having a good wardrobe is making fewer decisions. If you've got a handful of pieces that fit well, are comfortable, and go with everything, getting dressed takes about 30 seconds, especially in the summer.

21:25That's been quince for me ever since I started buying their t-shirts. I've also been reaching for the same linen shirt over and over again. Not because I don't have other options, but because it's lightweight, breathable, and just works whether I'm heading into the office, grabbing dinner, or traveling for the weekend. Quince's 100 % European linen shirts and pants start at just$34. And they're basically my summer uniform. I have some in short sleeve. I have some in long sleeve. I have a bunch of different colors. Sometimes I wear them with nothing else. Sometimes I wear them layered with a tee like this.

21:58Their tees are another of my favorites. They're soft enough to wear all day. And when the temperature drops a bit at night, their lightweight cotton sweaters are also a perfect layer. What's surprising most is the quality for the price. Quince keeps prices 50 % to 80 % lower than similar brands because they work directly with ethical factories and skip the middlemen. You're paying for premium materials instead of a luxury markup. And it's not just clothing anymore. Quince has become one of my go-to places for home goods, travel essentials, and everyday basics too. I have a Quince rug. I have a Quince duvet cover.

22:32I have two Quince towels. Whenever I need anything around my house or new clothes. I check Quince first. They come really quickly in the mail as well. So if you need something last minute, Quince is a great place to go to. Make your summer wardrobe easier. Go to quince.com slash 404media for free shipping on your order and 365 day returns. Now available in Canada too. That's quince.com slash 404media for free shipping and 365 day returns. quince.com slash 404media. If you sold somebody a loaded gun who you knew was in a vulnerable state and they shot themselves, I think it is murder. Just because you're using the internet doesn't mean you get away with murder.

23:18I'm Damon Fairless, host of Hunting Warhead. This season, I take you inside the business of suicide and the places desperate people go when they can't find what they need in the real world. Hunting the Suicide Salesman. Available now wherever you get your podcasts.

23:47All right. This is another one, as I said, that Matthew wrote. And the headline is, A farmer donated land to turn into a park. The city is building a massive data center instead. You wrote this a little while ago, Matthew, and I edited it and I really wanted to save it. for a slot on the pod. So we're talking about it now so we can give it the proper attention it deserves. I'm just going to read out the lead of the article, first of all, so people see where we're going. And I think it just fully captures the issue here. So it says, Almost 30 years ago, a farming family deeded land to the city of Taylor in Texas on one condition.

24:32The city use it for a public park. for the nominal fee of$10. The farmers granted the 87 acres to a public trust in 1999. Taylor, the city, then sold it to Blueprint, a data center developer, for$10 million in 2025. Now, the land that was supposed to belong to the community will become 135 ,000 square foot data center. That, yep, kind of sums it up. Obviously, there is a lot more to it than that. Let's sort of go chronologically, or at least the order of the article. So who is Pamela Griffin? She's sort of one of the main people you spoke to for this piece, who's connected to this story. So Pamela Griffin is a woman who lives in Taylor, who owns a house that abuts the field where this data center is being built, currently being built right now.

25:30She has lived there. her and her family have lived there for generations. Basically, the story that she told me is that her grandmother purchased a bunch of land in that area of Taylor several generations ago. Back in a time where in Texas, cities were like segregated, like legally segregated. And there were only, if you were buying land and you were black and Griffin's family's black, you had to buy it in a certain part of town and so this was the site of Taylor that's like literally right next to the railroad tracks and like next to a power substation and her grandmother and then subsequent generations of the family basically bought a street in this little division and the family has lived there for you know several generations now and she's kind of the woman at the center of all of this like the fighting back and the raising issues and that sort of thing so that the later groundwork and the family owns this land as you say in this area where they were essentially forced to then a person called mr bland comes in what is pamela's memory of mr bland who is this person sort of what happens when he comes into the picture.

26:58So I'll back up just a little bit. Yeah. So Pamela told me that she had no idea what a data center was until people were canvassing after... The city went through the normal thing that happens with these... This is like a 20 ,000-person town. And it went through all the normal things that we're hearing that happen with data centers. Like developers come, make people sign NDAs, or the city council sign NDAs, and it kind of like happens in a rush. The concerns of the citizens aren't listened to, et cetera, et cetera. Kind of see this playbook over and over. Activists, the concern about the data center start going to the neighborhood where it's going to be like built next to, which is where Pamela lives, start knocking on doors.

27:41Pamela opens the door. She's never, she hadn't heard this thing was being built, didn't know what a data center was. They start looking into it with her and her family and they're like, this is terrible. This is bad. And Pamela has a memory from when she was a child. She's like, wait, isn't that Mr. Bland's land? Isn't it supposed to be a park? And she remembers this conversation between Mr. Bland and her father. And Mr. Bland and the Bland family owned the land where this data center is being built. And they were farmers. And Pamela and her family would play kind of in that field where the data center is being built and had for a bunch of years.

28:24And the farmer would be out there on his tractor and he would see them and he would wave and he was friendly and kind. And she remembers when she was a child, her father having a conversation with Mr. Bland. And Mr. Bland saying, hey, my family, we're selling and we're leaving. We're going to move somewhere else. I know that this area and this land is important to your family and that your children need a place to play. So when I leave, I'm going to leave it to the city for them to turn into a park. And it's one of these kinds of things that's like this half-remembered childhood conversation. And could there be any truth to this?

29:04Who knows? And spoiler, it turns out that she was correct, that this actually had happened. Yeah, and we'll get to that. So Pamela sees this, has this memory. what happens next because you see documents which corroborate this account what are those documents how did they come about and what did you see when you saw them so it's all in the public record the Blands leave in 1999 and they deed 87 acres to a public trust that specifically, I can't remember the exact name of the public trust, but it's a trust that specifically develops parks in Texas. They deed it to them for a nominal fee of$10, which is like a normal, like if you're going to do a real estate transaction for the lowest possible amount, it's like$10.

30:02Yeah, like a lawsuit for$1. Exactly. It's one of those kinds of situations. And it's a very short deed. It's a very short grant that says the Bland family is deeding this land to this foundation specifically so it can be built and turned into a public park. It's like very clear. It's just a couple sentences. It doesn't get turned into a park. Yes. That trust a few years after they leave in 1999, then turns around and sells it or deeds it to the county again for$10 and they kind of copy-paste that same language again. With the condition. With the condition. The county then, a few years after that, sells it to the city.