SIO414: So You Want To Simulate The Universe...

Transcript

This transcript was generated automatically and may contain errors — the audio is the record.

0:00:16 Thomas: And welcome to Serious Inquiries Only. This is episode 414. I'm your host Thomas Smith, and we are back at it with the simulation stuff. But now we're moving to a really fascinating area of it, which is Dr. Brian Gillis taking us through what would it actually take to simulate the universe? You sort of a given of some of these arguments.

Oh, yeah, you know, we have video games now, and we have 3D stuff, and we have virtual reality, and so, you know, Hop's given a jump to simulating our entire universe in a way that would be convincing to us. What is that, what would that actually take? Maybe it's easy, maybe it's not easy. Dr. Brian Gillis is going to take us through the physics and sort of the computing limits and stuff that would go into figuring out how plausible this is.

And I want to add, Brian did a lot of work on this, a lot of original work on this in trying to calculate this. So I hope you will stay tuned and listen for that really fascinating look into what it would take to simulate the universe. So you want to simulate a universe, do you? Well, find out what that would take after this auto ad that you can, of course, skip if you are a patron at patreon.com/seriouspod. Thank you so much to the patrons for making the show happen. And with that, let's figure out how how we're going to hack the mainframe here and simulate our entire universe. Dr. Brian Gillis, how are you doing today?

0:01:38 Bryan: Not so well. I'm unfortunately down with COVID here.

0:01:40 Thomas: Oh, sorry to hear that. I think that's the simulators. They're saying where they're punishing you for revealing their secrets. I think that's what it is.

0:01:41 Bryan: Yeah, that or somebody found a bug, just like back in World of Warcraft where there was actually accidentally a bug which let loose a pandemic.

0:01:50 Thomas: Yeah. Is that real?

0:01:56 Bryan: There was. It was quite a while ago. It's called the Corrupted Blood Incident, I think. I think basically there's this one boss that characters were meant to fight, which within the boss, you can only take like 10 or so people to fight it. And he would unleash a mini pandemic that you would have to fight to cure before it overwhelmed you and killed everyone. But there was accidentally a loophole in the code where normally it's supposed to be cleared when you leave. But if one of your pets gets the plague, you dismiss the pet. It vanishes, but its state is saved. You leave it and then take the pet out back, like you go back to the regular world, the pet still has the plague, and then it starts spreading to everyone. And so they, when players found this, they accidentally let loose a plague in the world, and it's very interesting sociologically what happened. Yeah, yeah.

0:02:33 Thomas: Oh, wow.

0:02:47 Thomas: That's so funny. Yeah, and then did you well, we don't have to get into that, but that's very interesting. So do you is it just like being poisoned in a game where you just die over time or something?

0:02:56 Bryan: It basically, well, it dies over time and it spreads, so you like, over time it deals, kills, damages, to you a huge amount, which would instantly kill lower level players. And also at the same, it would spread to anyone nearby. And what made it worse is there were non-playable characters in the game who were immune to dying.

0:03:18 Bryan: Yeah, they could get it, but they were immuneto dying, but they would still spread it.

0:03:00 Thomas: Right? Oh they could get it. Oh, funny. Cause I already in my mind, I was like, well, wouldn't everybody just die and it's fine. But it's like, we'll know if there's different timings and how people die. Sure. You reset and you're fine, but then there's still people around with the plague. And then I was thinking like, whoa, that'S a really interesting example of how disease works in general. You know, like that's so funny. But if you have those NPCs, thatS even worse.

0:03:39 Bryan: Yeah yeah. They're They're basically the asymptomatic carriers of this and that they have it, but they won't die from it, but they're going to spread it. And so a lot of the NPCs that got it were quest givers that everyone wanted to go to. So they would try to get it and then die. And it was really fascinating. It just had all the ingredients to it a surprisingly realistic simulation, because even in the outer world, players could heal it. And so some tried to set up like stations where they could cure it and try to cut it down.

0:03:45 Thomas: Yeah. Yeah. Wow. Yeah. Yeah. Then there's a vaccine and they're like, all right, uh, only for the meant healers that we have, because those are the essential workers, so we can only give the vaccine. Scene, wow, that is really fascinating.

0:04:18 Bryan: Yeah. Yeah, Yeah. You know, the one interesting most fact about it is they saw something in there, which they thought was unrealistic at the time, which was people who deliberately spread the plague. And it's people who were just making things worse for others with no gain on their own and they assumed, okay, well, that's only because it's a game, but then COVID came and you would see these videos here and there of somebody who was COVID positive going to the grocery store and licking everything, and it, no, no, those people exist in reality too.

0:04:21 Thomas: Right. Right. Uh, no one would ever do that yeah yeah Oh my gosh.

0:04:49 Thomas: Wow. Somebody I'm sure people wrote like academic papers on this event in World of Warcraft, right? I mean, that's got to be, that's so, so interesting.

0:04:50 Bryan: Oh, There must have been some, yeah. There are a ton of news articles, just search Corrupted Blood incident. You'll you'll get a ton, yeah.

0:04:56 Thomas: Yeah. Oh, I'm I'm going to that's Actually really interesting, and it's funny that like that mistake, whatever that was, yeah, it's a bug and all that, but like that created such an interesting scenario that it's like, wow, they almost could have done that on purpose, you know, just to make the game more interesting. You're never going to forget that. That's like an unforgettable game experience.

0:05:14 Bryan: Yeah. Well, eventually they, uh, did do it on purpose, they wanted to uh, do a zombie outbreak thing, which was inspired by this, and you know, they made it not quite as bad overall, so that it wouldn't overall disrupt game play, um, but they kind of themed it after the Corrupted Blood, using it as inspiration, preparation for a later zombie outbreak.

0:05:21 Thomas: Oh, that's funny. Oh wow.

0:05:38 Thomas: Yeah. That really cool. Well, all that is a great introduction to what is kind of like a, almost a, I guess a part three probably of the simulation discussion. Cause you wanted to take us through some of the, you know more like computational practical considerations of simulations. Is

0:05:40 Bryan: Yeah. Yes. And to catch you up on what we've talked about before, best to go back and listen to the episodes, but just in case it's been a bit of time, we talked about, like, the idea, is the universe a simulation, simulation, and, the the probability of it. And one of the big reasons that people used to believe it is that they say that if simulations happen, there will be many simulations. And the fact that there are many simulations means that the probabilityof us being in one of those simulations is much higher. And Now, this is a bit disputed. You But there is a very strong argument that this logic is valid, that the large number of simulations combined with a decent probability that simulations exist could lead to the conclusion that we're probably in a simulation. Now, now, what I want to go in there is the second part of this. What is the probability that these simulations exist at all? Because this is one of the things that this argument hinges on. Like, let's say it's impossible for simulations to exist.

Then, then in that case, it's obvious we aren't in one. Let's say, it is possible, but just barely, there's a one in a hundred chance that the entire universe will contain a total of one ancestor simulation. So we might exist in that or we might resist in reality. Well, a probability like that means, okay, we're probably in reality, but yeah, there's maybe a 1% % chance we're in that ancestor simulation, or it could be there's 50% % chance that there's trillions of ancestor

0:05:57 Thomas: Sure. Yeah. Yeah,

0:07:31 Bryan: Simulations or just a big mix of probabilities. So in to figure this out, we actually have to get a handle on what is the probability that these simulations exist at all. Intuitively, Thomas, what would you say? What's the chance that simulations like this either do exist or will exist in the universe.

0:07:56 Thomas: So I guess, are you saying, like, based on reality as we know it now, like, assuming, are You assuming we're the kind of the root reality, and then, or not, or does that matter?

0:08:01 Bryan: Yeah, assuming we're either the root or the reality looks like ours. Because, you know, as we talked about before, if the simulation is vastly different from ours, we can't really intuit much about it. But if you look at a lot of the arguments that go on, they specifically state ancestor simulations, that they're making simulations of the universe, like itself.

0:08:10 Thomas: Okay, yeah, right, yeah. Okay, so assuming that, that does seem like like a pretty difficult question. I mean, for it to be a simulation, that's obviously so effective that we could be in it right now and not know. And therefore we would assume that, you in it has internal thoughts and feelings, you knows, is a sentient in that way. I suppose we could all be philosophical zombies, but I don't, I'm actually not sure if definitionally you yourself can be a philosophical zombie. I think the, I think philosophical zombies don't have internal experience, but you can't tell from the outside. So I guess I would not, I think, therefore, I know I'm not a philosophical zombie, I guess is what it would be.

0:09:00 Bryan: That that's Yes. Yeah. I think, actually, yes, I think you can't say that, yeah.

0:09:08 Thomas: And and so I think that would mean, I mean, it's got to be not only incredibly complex, that's got to be something intentionally done, you know, I would think, or most likely it would be something intentionally done, because if you're creating a simulation that's to be an ancestor simulation, like you're doing certain stipulations and And you're, you tuning those dials to make it happen in that way. And that means you're okay with this eventuality. You're okay. With creating a bunch of artificial life forms in that way. So I don't know. I could, I could see that being something that wouldn't happen. I could say, or like, I Could see it being possible that in universes in which people or in societies or whatever, in which people created this ability to make a simulation that rich. That that would also maybe come with some or ethics that would say, well, we shouldn't really do that. You know? So like, that's a possibility,

0:09:20 Bryan: Well, Yeah,

0:10:00 Thomas: I think. Um,

0:10:00 Bryan: Well, let's go with that. And that's actually one of the aspects I was thinking about is, would simulations not exist for moral reasons? Like, in in the previous discussion, we were talking about the Futurama episode on this. They faced that issue that once the simulation did exist, they said, no, it would be immoral to stop it, as that would be equivalent to killing them. But then that also means once it's created, you're obligated to keep it going. And so the only moral option, at least I see it, would be to never create in the first place, just have a simulation abortion, basically.

0:10:27 Thomas: Yeah.

0:10:36 Thomas: Or, or there could be one weirdly, like that might make sense too. Like that accidental creation that we see in that very episode that we watched. Like, I actually see that as like kind of plausible if the same way that, you know, oh, whoops, we might create AI that's sentient and not know it. And then it would be like, well, we can't kill it. You know, I could see it being a one -off in that way, possibly.

0:10:55 Bryan: Yeah. Yeah, you create the one, okay, we're never doing this again, but you already exist. We're not going to kill you. Yeah? The thing is, I don't think humans work that way. I think, like, look at the recent race with AI, or familiar with kind of the reasons the company OpenAI was founded?

0:10:59 Thomas: Yeah. Yeah.

0:11:15 Thomas: Yeah, a little bit because of the Sam Altman news and all that, you know. It was sort of founded under the idea that like, oh this could be dangerous, so we're going to make sure it's open source and all these rules to it for open AI. And And then that was all thrown out the window because of what happened recently.

0:11:17 Bryan: Yeah, yeah, yeah. Yeah, it was. But think basically it was founded under the argument that someone is going to do this eventually. It is very, very dangerous. So better that we do it than someone else does. And kind of the issue is everyone thinks they're the good person, and from an external perspective, it's very arguable if they are, and especially since now that we look at this kind of in retrospect, it's a lot more suspect whether these people actually were good, which I'll remind you, one of the founders of OpenAI was Elon Musk, who By the way, I, I looked it up, apparently Elon Musk is 99.99% sure he lives in a simulation. So yeah.

0:11:42 Thomas: Right. Yeah. Yeah. Yeah.

0:12:16 Bryan: Of the point I was making is that you, you can't really rely on human morals to keep them from going into questionably moral areas like creating a general AI. Because once we have that, we have to ask the questions of, okay, is it immoral to stop it? And people are heading that way anyways. And things like creating a simulation, I don't think we can trust human morals. They'll just use the logic, well, if I don't do it, somebody else will, and it's better I do it, and then they'll go ahead

0:12:24 Thomas: For him. Yeah yeah.

0:12:50 Thomas: Assuming there's some value to it. Because I do think that it would make it a little bit less likely if it was something that took this incredible ingenuity to do, and that is create an ancestor simulation that we're talking about, and yet it didn't really provide any value to anyone. I would think that if it provided no economic value or anything like that, I think it would be a little more likely that there's som worlds in which those people feel constrained. But if if it creates some sort of value, which you can imagine it would, yeah, I agree with you. That's like capitalism. Maybe capitalism is actually the one constant of the universe and there's no fucking getting, maybe we're in that hell that that's always going to, there's always going to be some capitalist in any alien civilization or different us in a different universe. And that will create these incentives.

0:12:52 Bryan: Yeah. Yeah. Yeah. Oh. Yeah. Yeah. And quite easy to argue that there may be some value, perhaps a very small, and may the value may be as simple as, oh cool, I managed to do it, and now, if the cost of doing it is less than that value, um, you can say there's probably going to be somebody who does go ahead and do that. And I think that's one of the things that leads people to believe we'll have simulations, is they expect the cost will eventually become negligible.

0:13:45 Thomas: Yeah. Yeah, yeah.

0:14:05 Bryan: And I think this is one of key places assumptions break down. So in computing, you've probably heard of Moore's law. Are you familiar with that? Yeah, it's It's basically

0:14:13 Thomas: Sure. Yeah it's kind of this this exponential increase the number of transistors we can put on something or something like that.

0:14:20 Bryan: 1970, 75, Moore, whatever his first name was, noticed that we're making exponential progress on transistors. We're basically doubling how many we can fit on a chip every two years. And since then, that has kept up in some ways due to a self -fulfilling prophecy as it became a target for people. So it held up to a very tight curve because of that. And it is crazy. And, you people look at this as a, you know, if you extract this into the future, exponential curves are ridiculous, like, as we found out in the recent pandemic, when we were seeing exponential growth of the coronavirus. And if you have exponential improvements in speed, then, like if it doubles in two years, then in 20 years, it's gone up by a factor of 1 ,024.

0:14:41 Thomas: Right. It's pretty crazy, though. It's amazing. Yeah. Yeah.

0:15:10 Thomas: Right.

0:15:10 Bryan: In 40 years, gone up by a vector of over a million. So it gets ridiculous. And on top of this, it isn't even just the processing speed. There's also another similar law called Cumi's law, which has noticed that the power efficiency of computation was roughly doubling every one and a half years. This one did notice a slowdown, that around the year 2000 and slowed to about every two and a half years, but you the two together, we're both getting things you as fast and twice as efficient every two years, and if you extrapolate those, yeah, it does look like there will be a point somewhere in the future where it's going to take a negligible amount of time to do a simulation and a negligable amount of energy to do it.

0:15:58 Thomas: Right. Yeah. Assuming exponential growth holds forever.

0:16:03 Bryan: And that's a big assumption. And that's where I think a lot of the people thinking about the simulations don't really consider it is how plausible is this exponential growth going to keep up? And honestly, even right now, there are a lot of people saying that, you okay, feels it like we're hitting the end of Moore's law. You know, there are a lot of people who will say different gates, some as early as 2025, just a year away. Some say it has already ended. Some say to last another 10 years. But the serious people in the industry are all saying it will end. We can't keep this up forever. We're hitting real physical constraints with this type of technology.

0:16:05 Thomas: Yeah.

0:16:44 Thomas: Yeah. There's got to be.

0:16:46 Bryan: With this kind of technology, Technology is a loophole that some people will say, well, another technology always comes around. Are you familiar with Ray Kurzweil?

0:16:53 Thomas: Yeah. I know that name.

0:16:57 Bryan: Yeah. One of the people talking about the magic of technological advancement. And he acknowledged that, yeah, there are going to be limits to any form of technology. But he claimed that, you know, this has happened throughout history. Every time one technology reached limit, a new type of technology came in. And if you read his books, he has nice graphs of these going back even before integrated circuits to cathode ray tubes. And so, yeah, it's always been doubling every two years. Well, I'll just be honest. No, it hasn't. Wasn't It like a human person came out of a cave and said, oh, I have one stick of information. And then two years later, they found a second stick. And then 2 years after that, they find 2 more. It's fit since 1965, in large part due to a self -fulfilling prophecy. But that, it is unreasonable. And even if it held up every time until now, that doesn't mean it's going to hold up into the future.

0:17:39 Thomas: Yeah. Yeah. Yeah.

0:17:56 Bryan: In fact, it turns out that with our understanding of physics, there are some very hard boundaries that we don't think any technology can get around.

0:18:05 Thomas: Right. Right. And this is kind of what we discussed on the entropy episode, right?

0:18:06 Bryan: Yeah, that Landauer's principle, which is there is a fundamental limit on the minimum amount of energy needed to make any irreversible computation. Now, there's a bit of a loophole that you can do a lot of reversible computations before you get up to something irreversable. But as long as you're making some irreversble change, which, you know, all of human experience is kind of based on irreversrible changes. That's how our minds work. Work. There's a limit to the energy there. That limit is ridiculously small, and we'll probably hit other problems before that. I talked about some problems like you have to worry about quantum tunneling of states within the computer. There are ways around these. One of the ways is that you can double the data. Instead of storing it in one place, you store it in two places, and then for each of those, you have some checksum of the data in each. And then every, let's say every hour you go and check, okay, is this checksum still valid? If you look at both and see, okay. Valid for both. Okay. We're good.

If you say, okay it's not valid for this one, but it is for the other. Okay, so that first one where it isn't valid, there's probably some tunneling going on over there. I'm going to assume the other one is still valid. Copy it over. And so that'll cost you some energy to do the checking and correcting. There's also a chance you'll come and say, uh -oh, both checksums are invalid. I got really unlucky here. All I can do is flag this data as invalid. Now, if you check it more frequently, there's less and less a chance they'll both be invalid, but it takes more energy to do it

0:18:19 Thomas: Mm hmm. Yeah. Yep. Right.

0:19:43 Bryan: That way because you're checking more frequently. If you want to get around the issues of quantum tunneling, you need to expend more energy, which kind of defeats the goals we're looking at here.

0:19:54 Thomas: I mean, there's gonna be some return on investment there, I'm sure some some optimum. But

0:19:57 Bryan: Yeah, currently, Yeah. We already do this with what's called RAID arrays for sensitive data that you don't want to lose. You double it up so that you can make sure you don'T lose it at some power cost. But the smaller the data storage gets, the more prone it is to loss. And actually, even just looking at data storage, thing one noticed I is that for a lot of my life, the size of a hard drive you could get with a new computer was increasing dramatically.

0:20:24 Bryan: It was another one of these things that seemed to be undergoing exponential growth. But recently, it seems to have plateaued in that most new computers are getting around maybe a one to four terabyte hard drive. And it's been that way for quite a while now.

0:20:00 Thomas: Mm hmm? Hmm. Yeah, yeah, it does. Yeah part of it is solid state. And so they're increasing, you quality of those hard drives. I

0:20:45 Bryan: Big part that they're shifting to a new technology, and do tend to be smaller capacity. They're improving, but again, not as much. But for, yeah, the mass storage, you really get a hard drive that's

0:20:55 Thomas: Felt like that too. Actually, I was looking at computers. And actually, I Was looking at maybe putting together a network attached storage. And I was like, this is really, this is where we are. I thought, I figured we'd have like some massive, you state drives that I could throw into a thing and make like a hundred terabytes, you know, and somehow it really does seem to have slowed down. I don't know if it's the pandemic influence at all or something else, but yeah, it's weird.

0:20:56 Bryan: Yeah.

0:21:12 Bryan: I It was even before that, I think. You're starting to hit the physical limits where quantum mechanics in particular is making it more and more difficult to store data and not have it be corrupted. That's one way we've got an issue. There's the data storage. Now, it turns out there are also a couple other limitations that are interesting. There is one which, on the subject of data storage, there are a couple different limitations on it. There are one that there's absolutely no way around, and another that maybe you can get around. So the hard limit is what's called the Bekenstein bound. Excuse me if I'm mispronouncing it, but it basically says that the amount of information you can store in a given amount of space is equal to the entropy of a black hole, which fills that space.

0:21:19 Thomas: Hmm. Oh, yeah.

0:22:09 Thomas: Oh wow. That's

0:22:11 Bryan: Black holes having having entropy, basically the argument is that holes have the maximum amount of entropy possible in a given amount of space. So in some ways, this is a reconciliation of entropy and information loss when things go into a black hole. Black provably can only have, I think, three properties, mass, spin, and electric charge. The exact proof of that is outside my wheelhouse, but that notably doesn't include entropy.

And so if you want to preserve the second law of thermodynamics, where entropy only ever goes up and you have things that go into a black black hole, that means you'll have things with different varieties of entropies falling in, some currently at low, some at high entropy. And the only way you can ensure that the entropy only ever stays the same or goes up is if the black hole has always the maximum entropy.

Because like if the blackhole had the minimum entropy, when things go in, well, that decreases entropy because they're going down to the minimum. Or if it's anything other than the maximum, if something with more than that that into it, it would still be decreasing. So black holes basically have the maximum entropy, and this gives you a calculation of how much information can fit in a given amount of space by the size of as a black hole that fits them. Now it turns out this is a very liberal limit.

0:22:11 Thomas: Interesting.

0:23:38 Thomas: Yeah. I was going to say that's gotta be not much of a limit, right?

0:23:38 Bryan: Um. Yeah. Because black holes are so ridiculously dense and have such high entropy. And practically speaking, um, Landauer's principle gives us a much stronger, uh, now, even though it's technically the amount of energy needed to make any irreversible computation. Also, you know, arguably, if you want to have the information there, you need to be able to access it some way.

0:24:04 Bryan: And it's very hard to imagine a way to access it and use it that doesn't make this irreversable computation in some way, and so you can use that to convert to a limit. Limit. And there are some people who even argue that the Landauer principle directly means that information is energy or is mass, since mass and energy are the same. This isn't an argument I buy personally, but I'm confident enough in it to say that it's completely false. Say that it means information is an energy, and you can't have the higher information state state without having this much more energy in the system. I think that at some level, if you want to access the limit, you do need to account for the energy constraints of Landauer's principles. Now, even this, I'll remind you, it is very small if you do the calculations.

0:24:04 Thomas: I Just gonna ask, has anyone ever used Landauer's limit to be like, okay, what's the maximum storage even theoretically possible in a, you know, four inch by three inch hard drive, something like that? That probably, insane, because that doesn't seem like much of a limit at all. Sure.

0:25:12 Bryan: So I did a lot of calculations myself in trying to figure out the plausibility of this. And I think I have a comparison there, the Landauer limit. You take the Boltzmann constant, which is about 1 times 10 to the minus 23, multiply it by the operating temperature, which, let's say we want to get that as low as possible, so we do three Kelvins, the cosmic microwave background. Okay, we've Got about, yeah, times 10 to the minus 23. Um, now, joule a is, you, one newton of force over one meter. So So that'd be lifting a kilogram of mass about 10 centimeters would be one joule. So you can kind of imagine, like, you have a one kilogram, or a two point something pound mass, in your hand. You lift it up 10 centimeters, or half a foot. It doesn't matter the specifics. And that by 10 to the power of 23, three. And that's the limit on the mass needed to store one computation. Now, of course, you want to convert energy to mass, you have to divide by the speed of light squared.

0:25:36 Thomas: Wow. Of course.

0:26:18 Bryan: And it does get very small. Is how much information do we need to store? Because we're saying, okay, it doesn't take much energy at this theoretical limit, but we might need to store a lot of information. Then again, maybe we won't be able to reach this limit. I think for my calculations, I can't find the value exactly here, but this turns to be about eight magnitudes lower than the mass of an electron, I believe I calculated. It's hard to imagine that we'll be able store a bit of information in anything less massive than an electron. It is already one Of the lightest particles we can deal with.

Maybe we could put it into an electron neutrino. That's, that's the lightest particle that has non-zero mass. Practically speaking, you storing one bit of information in an electron. Electrons can spin, and you can set it up so that the spin will always be up or always be down to some degree of uncertainty. But the Landauer our principle puts a limit many orders of magnitude below that. We've got a bit of room to work there. Think for going ahead, let's just talk about the question. Let's figure out, okay, we want to do a simulation of the universe. How much information do we need to store? And if we convert that, how does it compare against our

0:26:18 Thomas: Yeah. Yeah. Yeah. Oh.

0:27:43 Bryan: Limits? So like this episode of Futurama. Professor Frank has a simulation of the universe on a table in the Planet Express warehouse. You plausible is that that with any advancement in technology, he could do that? For this, we are going to break out Fermi estimation like we talked about in a couple previous episodes. So a reminder, Fermi summation is basically you take the factors that go together in something like this. Come up with your best estimates of them. Multiply them all together to get your estimate. And it works out better than you would think. Now, for something like this, we have some pretty good physical ideas of many aspects of this that we can put into it. So let's, let's go with your instincts, Thomas. We want to make a simulation of the the universe. What are the factors that scale with that that you can think of, or?

0:28:04 Thomas: That Scale with it, what do you mean? Oh.

0:28:40 Bryan: Or that it scales with, like, if this were twice as large, our simulation would need twice as much data. Yeah.

0:28:50 Thomas: Wow, yeah, I don't know, because right now I'm thinking, like, that one hand I feel like I've heard people say casually that it's like, well, you can never do it because, you know, you'd need for every bit of data in the real world you'd needed a bit of data or something to simulate it. So it ends up being, you just need a universe sized universe. That doesn't seem right to me, or maybe I'm mischaracterizing what I've heard, but that doesn't seems right to be the case. I mean, I don't know. I don't know. Doesn't seem to me because using, you know, mathematics and other things we've, we talked about in a previous episode, there got to be ways of shorthanding a lot of this stuff. I mean, the rules of physics, the laws of physics.

If you program them in, I think can save you a lot data, uh, in terms of what's going on, you But I could be wrong on that. I know. I one factor that makes sense to me, and I could be way off here, is how much you are able to distill the laws of physics and the universe down to defined equations.

I feel like the more you have ad hoc stuff that we quite make work with each other, I feel you'd need more power to simulate that, it seems to me. But

0:29:03 Bryan: Yeah.

0:29:51 Bryan: Let's say for the point of argument we've got physics figured out, we do have equations For it, um, actually, let's just go back a point that you said, well, for every thing in the the universe, you need something to simulate it. Thing we can think of kind of as a metaphor for that is emulation, which is basically having one type of computer pretend to be another. This is very often done to execute old software that was meant to run on a very specific type of computer that You don't easily have access to. This is most commonly done with video games since old consoles are out of production, but people still want to play those games, so they program their computer to emulate an old computer. And if you've ever looked into this at all, one thing you'll notice is that the new computer typically has to be vastly more powerful than the old computer to reasonably emulate it. Like, it needs to be about 10 years ahead of it, roughly speaking, speaking.

0:29:56 Thomas: Mm-hmm. Yeah.

0:30:52 Thomas: Interesting.

0:30:52 Bryan: It, because the cost of emulation is you just, there's a lot of inefficiencies into it, and my instincts say that if we're simulating the universe within the universe, it would kind of be like trying to simulate a Super Nintendo inside a Super Nintendo, that It's not going to work out well, there's going to be some amount of overhead we had. Yeah.

0:31:10 Thomas: Yeah, yeah. I suppose. I'm trying to think. Does that make sense? I mean I feel like if that were true like wouldn't wouldn't video games be just very inherently limited? I I don't not. I don't know.

0:31:25 Bryan: Let's go ahead with the universe, let's do our calculations. So the, let's say we're going to start off, we're gonna do a quote-unquote perfect simulation of the universe to figure out what the upper bound is for difficulty, and then we'll think from there how much we can save with various optimizations, and at That point we'll, we'll just have an idea of how big the problem is, how much we would need to optimize. First, for our Fermi estimation, we're multiplying things together. Well, the difficulty of the simulation, that's going to scale with the size of the universe, basically. If the universe is twice as large, has twice as much stuff in it, the simulation's going to need to be twice as long. Does that make sense?

0:31:46 Thomas: I'

0:32:11 Bryan: Probably simplest way to track everything in the universe, it be very inefficient to track every bit of space, say, whether something is there or not, particularly as the universe is mostly empty. It's going to be a lot more efficient to track every particle in the universo. So, probably the factor that makes the most sense for us to look at is the number of elementary particles in the univers. This is going to me one of the factors that goes into it. Now, unfortunately, this is a ridiculously high number, even if we limit it to the observable universe, and even if we don't consider light. Basically, the smallest number people estimate for the number of elementary particles in the observables universe is 10 to the power of 78.

0:32:17 Thomas: M

0:33:03 Bryan: Yeah. If you include light, which is the photons as well, it can go all the way up to around 10 to 89 at the upper estimates.

0:33:11 Thomas: Trying to Yeah. I mean, there's so many freaking photons. I thought so that wasn't included. So what what was that including, if not photons, electrons, neutrons, protons?

0:33:14 Bryan: Yeah, was including electrons, different types of quarks, anything fundamental, basically. So we're assuming that that we're not even treating, is the highest resolutions. We're not even treating protons individually since protons are made up of three different quarks, two up quarks one down quark.

0:33:28 Thomas: Yeah. Yeah. I was going to ask, like, how far down do we need to go in order to get an effective simulation?

0:33:40 Bryan: And that'll be one of the places that, yeah, we'll say, okay, maybe we can optimize that. Like we're not going to deal with quarks. That'll give us a factor of three, but 10 to the power of 89, or 10 to the power 78, divided by 3 is 10 to the power of 77 and a half. We're scratching the surface, but there's a lot we're going to need to scratch through. So, okay, that's one of the factors we need. Do we need to do with each particle? What can each particle be doing? Well, let's say we've got a particle. What do you want to know about it, Thomas?

0:34:11 Thomas: Oh, wow, what can we know about position and whatever the thing, the things that Heisenberg says we can know about it, I guess, right?

0:34:16 Bryan: Yeah, there we go. Position and momentum. Those are the ones that are most relevant to it. Momentum being speed times mass, but it's a bit more generalizable as photons can have different momentums, even if they're at the same speed, since speed times mass breaks down at the speed of light. So position and momentum. And then we've got three spatial dimensions. So you've got 3 dimensions of position and 3 dimensions of momentum for what it's doing. Next, we need to know what that particle is. Like, is that particle an electron, a photon, a down quark? What is it? Maybe we need to know something about...

0:34:36 Thomas: Right.

0:34:51 Thomas: So We need an Excel database. No, I'm just kidding, a sequel table that has all that information in it.

0:34:59 Bryan: Yeah. Yeah. Maybe we want to know its spin state, but I'm just going to put that aside right now. We've got enough. Got six parameters for where it is, what's it and one for what it is. Then if we want to calculate how much data we need to store that? Let's zoom in on one of those. Let's do the what is it question. How many bits of information do you need to store what it is? And obviously for this question you need to know well what are the number of possibilities?

0:35:22 Thomas: Yeah.

0:35:34 Thomas: What are the possible things, yeah.

0:35:36 Bryan: Yeah, so for a current standard model of physics there are, as I calculated it, 37 possibilities and more for a dark matter particle. I forget in my spreadsheet here if that includes it or not, but Basically, if we're converting to bits, it doesn't matter. You need eight bits of information to store what it is, as eight bits allows you 64 possibilities. So that allows you all 37 elementary particles, plus a bunch if you want to make some optimizations. Like you have, okay, one of these isn't elementary, this is the proton. It's a complex particle, but it's common enough that we're going to give it its own number here to save

0:35:36 Thomas: Yeah. Yeah. Hmm. Yeah, well,

0:36:21 Bryan: Information, so you might do something like that, yeah.

0:36:22 Thomas: That that's where I almost think my thing earlier, just just a minor point. And I could be wrong. That's where I think my think earlier could be a factor, because if we understood physics to some greater degree where it's like, oh, actually, this isn't a separate thing. It's actually just this thing under different circumstances or something, for example, like maybe that reduces the amount of information you need by that way. But I don't know, maybe just a thought.

0:36:41 Bryan: Yeah. So, are optimizations here for the calculations? We're going to keep going with the upper bound and then we'll talk about optimizations at the end, once we know how big the problem is, so

0:36:52 Thomas: Sure. Yeah. Once we turn this thing on. Yeah,

0:36:56 Bryan: Yeah, so first chunk, we're going to need eight bits of information for what is it. Um, next bit we're gonna need, okay. Okay, where is it? Length.

0:37:07 Thomas: Well, how do you do that? That's funny.

0:37:09 Bryan: Doing a full resolution, so that means it could be anywhere in the observable universe. Well, that's a number we can look up. The observable universe has a diameter of 93 billion light years. You can convert that to meters. You get 8.7 times 10 to the 26 meters. That'll give you the maximum size, but you want to be able to narrow down its position pretty well because like subatomic interactions are going to depend on that so you're going to need a resolution well probably the plank length is going to be a minimum

0:37:11 Thomas: Yeah? How many parsecs? Oh yeah. Yeah

0:37:45 Bryan: The minimum certainly isn't going to lower than that

0:37:45 Thomas: Well, isn't that the maximum resolution you could have is the, because that's the smallest like length, right? Yeah. Sounds

0:37:47 Bryan: Yeah smallest length and that is, you know, around 10 to the minus 35 meters, which means overall we're spanning 61 orders of magnitude between maximum and minimum. To describe length. Yeah. Now, you convert that into bits, a bit over 200 bits. So, you know, yeah, it's because we're doing exponential scaling either way, bit scale exponentially. So it's really not that bad. Now you can do the same thing for momentum. I'm just going to say it's the same number of bits because Because Heisenberg's uncertainty principle, the uncertainty of position is inverse to the uncertainty of momentum. So I'm just going to say it covers the same

0:38:03 Thomas: Like a lot. Okay. Yeah right. Sure, sure.

0:38:37 Bryan: We can use the same number of bits for it. Bits to describe where something is, what it's doing. You've got three length, three momentum, and what it is. So the length of momentum each time is about 1 ,200. What it is. That's only eight, so that's just loose change. We can even forget about that. Overall, you've got about 1 ,200 bits or one kilobit or an eighth of a kilobyte of information describe what one particle is, which is a lot in computer terms, the scaling of information means it isn't as bad as it could be.

0:39:05 Thomas: Crazy. Yeah. Yeah. Right.

0:39:15 Bryan: The problem is we're multiplying that by 10 to the power of 80.

0:39:20 Thomas: Yeah, that's gonna be a big database and

0:39:22 Bryan: Yeah. So that's when things explode.

0:39:25 Thomas: Then you accidentally do drop table universe, oh shit, several

0:39:30 Bryan: That's why you got to sanitize your inputs, Tom. It's sanitize their inputs. Okay, yeah, this is our first limit is just the data storage for a single snapshot of the universe. Now, if you want to evolve it and have it simulated, that's another problem, but we're just storing the current state of this. So, we're multiplying, like, 1200 you by, like case, maybe 10 to the 80 particles. Got, know, you, as I calculated out putting all the numbers in as best I could, I got, know 7 you .74 times 10 to 82 bits of information required. Now, then we convert that through Landauer's principle into joules. We get 2.22 times 10 to the 60 joules, so we're still at 60 orders of magnitude. Then, well, convert that into mass. Mass, so what's the mass? So about two and a half times 10 to the 43 kilograms. Putting that in terms, the mass of the Earth, let me look that up, is 6 times 10 to the 24. So 2.5 times 10 to the 43, we're 19 orders of magnitude greater than the mass of the earth.

0:39:33 Thomas: Okay.

0:40:43 Thomas: We're gonna need a little bigger earth, yeah.

0:40:46 Bryan: The information, we can think about optimizations, but it's really hard to imagine optimizing away 19 orders of magnitude and even then we're getting a computer as big as earth, yeah.

0:40:56 Thomas: Like Maybe we could wind zip it, but other than that, I got nothing, I Think

0:41:01 Bryan: Well Now, and a lot of these optimizations, you also have to grapple with the fact that, um, there are both lossless and lossy optimizations. Lossless are ones where you keep the data, um, nature of the data means that you can simplify things in some way to overall save yourself information. A good example of this is, let's say you have an average text document written English and just using only English letters, spaces, periods, con punctuation. Simple, now, way to encode that would be the ASCII character set, which uses six bits for each character.

If you were to try to optimize it, you could say, well, some of these characters are a lot more common than others, like the space. There are a of spaces in this document, that the letter E is pretty common, that the Q, that's not pretty common, the percentage sign, that that only shows up once in here. We like the carrot, why is is even in the care set?

So you look through all of those, and you can do this thing, I think it's called Hoff, I forget what it's called, but you can basically set up a different way of encoding all of these where ones that show up the most commonly have the shortest representations and the least common ones have the longest representations. The least common ones, a lot of them will take more than six bits, but they'll be less common. And the more common ones take less than six bits.

And then at the top of the file, you put in this little dictionary saying, saying, okay, for the rest of this file, we're using this specialized encoding. It, that way, you'll preserve all of the information. And because you know that this information is, well, specifically, this is low entropy information, that there's some amount of order to it, that some things are much more common than others, you can, on average, save space in this. But it's not phenomenal savings, you'll maybe save 50% of the space doing it that way.

0:41:09 Thomas: Right, which Like a lot, but when We're dealing With, yeah.

0:43:01 Bryan: And it's useful to do in a lot of cases. There are a lot file formats we use that are lossless encoding that for most situations will save you space. For image files, the way to store an image is just storing the data for every pixel in the image. And that's what's been a format. But there's also a very common format called PNG, which is lossless encoding, which has a way to say, OK, we see right here there's a big block of pixels that all have the same Values, so

0:43:03 Thomas: Right. Yeah. So

0:43:35 Bryan: We can simplify how we describe That

0:43:37 Thomas: Like if You saved A picture of just full, like white, then it's Going to be like, oh, if you saved an actual, you know, pixel by pixel file of it, think then it'd be a pretty defined size no matter what it was, right, but if you're able To use these kind of algorithms to say, well, no, it's pretty easy, just say the whole frigging thing is white, that's going to save you a lot of space.

0:43:38 Bryan: Yeah. Yeah. Yeah. And with lossless signaling, you do get benefits that way, now there is a computational overhead that whenever you do something coding like this, it takes more effort to work with it because it's More complicated to decode it, it, particularly with PNG. You just display it pixel by pixel. You need to do some interpretation as you go through. But with the computers as we have them, that's trivial. If we're doing a universe simulation, it could be an issue.

0:44:08 Thomas: Yeah. Yeah, cool.

0:44:26 Bryan: We could maybe go a step beyond that and say, okay, we're never going to make this work, even with lossless compression, compression, we need to go into lossy compression. So in the image example, this would be like going from PNG to JPEG, which is a type of lossy compression. And JPEG actually relies on Fourier transforms for how it works, like we talked about previously, it that basically does a Fourier transform and preserves the, you, scale aspects of the image and throws away all of the fine detail that we don't tend to care about most Of the time. So for natural images, like you just go out into the world and take a picture, then, and compress it with JPEG, the fine retail is not usually that relevant to us and we can throw away. So JPEGs works really well for natural image. There's a difference, but you don't really care about the difference. But if you do a JPEG of, say, a text document on a computer, this is when you'll notice the difference.

You'll see these random black spots around letters, just the very unnatural image with the very sharp borders aren't well modeled with a Fourier transformation that throws away the higher frequency information.

You have to give up something when you're doing that, but you get huge space savings, um, you don't save a power of 10 to the 19, you might get down like one order of magnitude with the JPEG, um, maybe two depending on how you compress it, there are different levels to, you've got to do a lot of compression.

And even then, we're facing other problems with data storage in that it's not just the we need to store it, but where we want to put it, because mass tends to gravitate to itself. And if you've got too much mass, it'll collapse in on itself and potentially

0:45:04 Thomas: Hmm. Yeah

0:46:17 Bryan: Cause A black hole, so

0:46:17 Thomas: All our data. Oh, no. It's hole ate my, homework.

0:46:19 Bryan: Yeah. So realistically happens in Futurama is the professor puts the simulation of the entire universe on the table in the warehouse and then the entire solar system immediately collapses into a black hole, yeah.

0:46:34 Thomas: The solar system, of the reality, reality. Oh yeah.

0:46:37 Bryan: Their Reality Yeah.

0:46:38 Thomas: Because it's like, hypothetically would have to be so dense in order to simulate that much, you're saying? It's

0:46:43 Bryan: And It depends on exactly the calculations you're using, but even the case scenario, the black hole radius for containing all the information for the entire universe, as we calculated it, is about 3 times 10 to the 16 meters, which, compared to the diameter of the earth, is about 1.3 times 10 to the 7, or about yeah, million meters, so that's 10 orders of magnitude more than the diameter of the year, so we're not just collapsing earth into a black hole, but a good chunk of the solar system as well, yeah, that I imagine people are thinking is, well, when people talk about doing these simulations, we're not simulating the entire universe.

0:47:30 Bryan: We're just going to simulate Earth, definitely a good way to cut it down. Still do end up with problems here, but the problems are less. So issue we get to is information storage is just one of the hurdles. The next hurdle is actually running the simulation. You can't just have it sit there in a snapshot. You have to actually do things with it. And it turns out there are physical limits to this as well, so physical limit to this is called Bremerman's limit, which gives the maximum computations that can be run per second based on the mass, or equivalently based on the energy of a computer, and

0:47:28 Thomas: Yeah. Maybe just earth or something. Yeah, man.

0:48:14 Thomas: I got to go into the figuring out limits of stuff business. I feel like that's a good way to earn your name. You know, I'm just going to come up with, okay, there's a limit to how many bananas you can eat in a day. And it's, it's the Thomasman limit. And there's, there it's like 30. I don't know. Start putting me in Wikipedia now. Okay. Was that useful? Does that help you?

0:48:17 Bryan: Yeah. Yeah. Yeah. There we go. Yeah. Okay. Yeah, yeah, yeah. It simply the reason this limit exists is, you know, if you go down to quantum mechanics, if you want to do a computation, you to need take one wave function, which is things are waves at the quantum mechanic level, and let let it evolve into a different function, which is distinguishable, just roughly speaking. And in order for it to evolve, you the rate know, at which it evolves depends on the energy of it. The more energy, the faster it evolves.

So you put it through a lot of computations, be with all of this, and you get a limit. And the limit you get, so let's say we turn the entire Earth into a computer, every bit bit of mass working towards it, and it's operating exactly at this limit, the best we can do is 10 to the 75 bitwise computations per second, a lot. But remember that one snapshot of the simulation is going to be around 10 to the 80 particles. It's going to to take us maybe 10 to the 5 seconds, that is 10 ,000 seconds or a few hours to simulate everything in our entire universal snapshot. Bitwise computation. This isn't even going through all of the complex physics involved.

0:48:36 Thomas: Hmm.

0:49:24 Thomas: Don't know what that means.

0:49:51 Thomas: Yeah, hmm.

0:49:52 Bryan: Yeah. Our universal computation, we're running into a big problem here. Now, I wanted to do some calculations with and figure out, okay, but let's just say we just want to do Earth here. We face a couple problems. Biggest the problem is that Earth is a lot denser than the rest of the universe. And by a lot I mean 30 orders of magnitude. This is the thing with astrophysics, you talk about orders of magnet to all the time, actually I don't even know if we define it. Basically, an order of magnitude is you multiply it by 10, and Then two, orders of magnitudes, multiply by 10 again, so 30 orders of magnitude, you multiply by ten thirty times, or it's a million times a million times a million times a millions times a millions, yeah.

0:50:27 Thomas: Yeah it's a lot. So a hundred. Yeah, yeah,

0:50:41 Bryan: So the Earth is a lot denser than the universe. So we're going to have to do a much more accurate simulation of it. And, you know, we're putting all of the Earth into it, are going to run into massive problems in

0:50:40 Thomas: It's a lot, I don' know why, something about this is bugging me and I don't know what it is. Maybe it's just that, like, if the, if this is all true, it doesn't seem like just video games would be impossible. I mean, there's some level at which we're able to simulate stuff now, and yeah,

0:51:10 Bryan: Well, they're massively simplified, remember. And

0:51:12 Thomas: I know, know, I know, they're massively sympathetic, I'm just, it just feels like the scale at which you're talking seems so much more difficult than you would think, just, and obviously could just be intuition is terrible, but is there, is there something missing? Like, I isn't, isn't there some way in which you, you don't need to be simulating every single particle at all times, you can, you know, you can take shortcuts, or you can use math, or you consider, is there's something, something missing there? Hmm.

0:51:37 Bryan: Yeah, Yeah, you definitely can take shortcuts. You can make a simulation that is a lot blurrier, then you have to worry about the people inside noticing it. I'll give you one example here, which video games can't do. It's just beyond the level of video games currently, and that is fur, particularly cat fur or hair. I'm not sure if it's technically fur or care. Um, cats have fur that is numerically a hundred times as dense as human hair. Um, it is you fine, and you know, if you have a cat and you look at the way the light goes through it, you see that the light does very interesting things going through it. But if you go into any video game, like, I think like a year or two ago there was this game Stray that came out starring a cat. And you can try to get up close, look at the fur on that cat and see how it just doesn't hold up

0:52:01 Thomas: This doesn't

0:52:32 Bryan: To reality. Imagine that you're in a simulation and you want to make sure that they never notice it. You're either going to have to run it at maximum resolution all the time, or worry that somebody is going to take a magnifying glass and look at the hairs on their cat, which is something anybody could do.

0:52:32 Thomas: Work huh, yeah, well, I mean, that's assuming they know what to look. I mean, couldn't you just say, well, what if the real reality is actually way more intricate than this reality and we don't know the difference? Yeah.

0:52:59 Bryan: But then you're kind of breaking the assumption of all being done with ancestor simulations, which is, yeah,

0:53:05 Thomas: No, that's a good point, which is, I think, another reason why I don't think this is very likely. I think this cut, I think this does sort of cut to my, uh, with the simulation argument itself, we're gonna need a lot of cat fur, yeah.

0:53:11 Bryan: Yeah. Like one thing, like, let's just focus on cat fur. Let's say you're doing an ancestor simulation. Yeah, but your limit is that you can't make cats as realistic as they are in the real world. So those people are going to be patting cats and their enjoyment of it will be slightly different because the fur isn't as fine.

0:53:21 Thomas: Yeah, no, one Cecil are gonna notice, you know, yeah.

0:53:37 Bryan: So yeah. People, you know, the fur is not as soft. Maybe they don't bond as closely with cats. Those small things add up. It's one of these things. What's what we call chaos mathematically, that small changes over time compound into big changes.

0:53:42 Thomas: Right, right, yeah, I remember Jeff Goldblum seductively taught me about it in Park, I'll never forget it, how

0:53:58 Bryan: Yeah, thinking about simulations, I think the way they try to get around this is saying, well, okay, you don't always run it at, um, the maximum resolution, you just zoom in when you need to, which of course raises a big question of how do you know when you need to yeah And you know, you, we can take some inspiration, uh, video games, like, for, for instance, one of the things video games do is that things are only visible when the player's camera is pointing at them. They go invisible as soon as they're outside the camera. Then maybe you have to do a bit if there's anything that can reflect. So, you know, you do things like that. But, you, a simulation entirely around one person, or do you do it around the entirety of humanity? Humanity, you to put in extra overhead to monitor all of the potential

0:54:13 Thomas: Do you calculate when you need to, yeah, yeah.

0:54:46 Thomas: Including

0:54:46 Bryan: Observers, and you also have to think about all of the secondary effects like Here's one way that a lot of video games fail is trying to handle reflections of something on water, that very often if you look at things reflected on water, you'll only see things reflected if you can also see the non-reflected version of them. Let's say you're looking across a lake at a mountain. If you zoom down at the lake, the mountain will often disappear in the video game because it only renders the mountain if it's on screen for you. But if you

0:54:48 Thomas: Cameras instruments yeah everything, yeah, huh.

0:55:22 Bryan: Turn back and look at the mountain, you'll see it suddenly appear copied into the water.

0:55:26 Thomas: That's interesting, right.

0:55:27 Bryan: You're doing a simulation, you have, have to be better than that, and a lot of games do do it, uh, they'll render not only the primary things you see but reflections. But then again, like, you can do experiments in our world, you can set up an infinite series of mirrors, and there, there's never any, any that you run out of, uh, at a certain point, um, simulation would have to, like, one of the things people say in response to simulation theory is if we are in a simulation, there must be some way to break it, and there are a lot of things we do that when you think about it would probably break a simulation, or we would notice the issue, from cat fur to two mirrors on the opposite side of the same room, a ridiculously easy setup, yeah,

0:55:47 Thomas: What if we set up a bunch of mirrors and cats, just like mirrors and rats and mirrors and boots and cats and boots of cats, and then all of a sudden bam universe explodes or just Stops

0:56:17 Bryan: Yeah. Yeah. The whole thing grinds to a halt as it's trying to simulate infinite images in a hall of mirrors, putting all of this together, you can go into the math, show how hard it is, you know, it's kind of impossible to quantify how much you need to simplify it. But I think my gut is saying that the limit of simulating the entire universe, or even just simulating, is so much higher than the amount of room we have to work with, even at the theoretical physical limits, which I'll remind you, we aren't even close to getting, and probably won't, because we'll hit other quantum effects along the way. We'll probably come 10 orders of magnitude shy. Um,

0:56:21 Thomas: Yeah hmm interesting shoot, that's

0:57:01 Bryan: So my gut says if we were in a simulation, we would be able to spot errors of it. It almost impossible that we would know, at least if it's in quote-unquote ancestor simulation. If it's a more complicated world simulating ours, yeah, possibly, would you, wouldn't, no, they might have the room to simulate it, yeah.

0:57:02 Thomas: Interesting yeah, I mean, maybe they're multi -dimensional and that, that opens up a whole, you know, I guess, more than three-dimensional, but yeah, that would open up a lot of exponents, I think, the orders of, of magnitude. But yeah, no, it makes sense.

0:57:23 Bryan: Yeah, what it's like, as you said before, like, simulating us could be like us simulating a 2D Mario game. We limit things down enough, um, then physics that within that game is the reality to them, but we've just set up a physics that is much simplified compared to what we deal with. So maybe our physics is much simplified compared to whoever's simulating us, yeah.

0:57:52 Thomas: I do bristle at any mention of we would notice. I feel like that definitionally isn't correct. Like what would happen is we would observe some phenomenon and we'd be like, wow, I guess we better explain this scientifically. And then we either would or we wouldn't be able to. And I don't think, you know, Now, I guess it would be if we keep getting inconsistent breaks in the laws of physics, maybe, right? I mean, yeah,

0:57:57 Bryan: Yeah, I I think I think that's probably it, and maybe the way to look at it is, are the laws of physics consistent across all scales and all positions of the universe? And also, another thing is, um, chaos occur even in things we aren't looking at, like we talked about turbulence in a previous episode. Turbulence is chaos manifesting, a gas building up into these little eddies, these little swirls in them, and you can see that happening even when you aren't working at the microscopic scale. So if they wanted to simplify it, they would have to have some model of turbulence that mimics it, but then this is also going to to have to happen, you, everywhere in the universe. Like massive nebulae out, outside in space are going to have their own turbulence which has to be properly simulated, and then we would have to be able to see the results of it. So they're, they simplify, they're gonna have to have a nearly perfect model of how to simplify it, or else we're gonna notice something is going wrong, that things don't quite add up.

0:59:16 Thomas: That's, a strong case. I don't think we're going to be able to do this project. It's coming in over budget. Yeah, interesting.

0:59:19 Bryan: Yeah. I looked into to it a bit, um, you know, there are some people, um, saying that what we'd have to do is like take multiple probes, send them out into space a far distance apart, because the assumption would be that a simulation would have to be high resolution at one point, but then if it's high resolution, it can't handle a lot of space. So like, let's say they're doing an ancestor simulation of Earth, but we want to simulate a couple of probes going out to a ridiculous distance and do some high resolution test then. That's surprisingly achievable, you know, given like long distance probes. We would have to wait for a while for them to get far out enough. It could be done. So if you really wanted to, you could set up an experiment like that to test it out.

1:00:04 Thomas: I feel like Elon's going to do that. It's exactly the kind of waste of time that Elon Musk would do.

1:00:07 Bryan: Yeah. Yeah, as I said, he thinks we're in a simulation, so he would like to prove it, it, I'm sure, uh,

1:00:13 Thomas: There any way to do some middle-out compression or something? Is there anything to like, what if we switch it to like, instead of simulating this whole outward universe, we're just controlling all the sensory inputs of a human or something, or we Boltzman brain it does that save us anything? I guess we still have to know, we have to have a consistent, maybe if it's just one guy there's one brain, but I guess if you have to have a consistent interactive universe outside, I guess that probably still still doesn't save you. You have to know at all times what's going to be thrown at those senses. Yeah.

1:00:25 Bryan: So we're Boltzmann brains, brains in a jar, we're in the matrix yeah yeah yeah Then you'd still have to calculate, you know, how practical is it to do a simulation of one brain with everything it goes into and make it consistent, yeah, have to have some idea of the world external to that mind. Well, it's basically the Truman Show problem, that you're, you're trying to simulate one person's life, but they had to construct a whole artificial town for it. If we're doing it by simulation, we just need to simulate that, which is still a lot external to that one person. So, yeah, hard to say.

1:01:06 Thomas: Mm hmm. Yeah, I guess it doesn't save us anything. I don't know.

1:01:20 Bryan: At some point, I think there are economies of scale, because if you're simulating one person and they experience like like 200 other people who live in the same small town, well, wouldn't it be more efficient to just simulate all 200 of those people, but 200 in the Same Simulation rather than just have one real person and 200 that are simulated? That seems a better economy of scale.

1:01:40 Thomas: Yeah yeah This really, I'm surprised that this seems this impossible based on what you're telling me and that people are still entertaining this. I mean, I hadn't gotten into a lot of the details of why this would be so difficult, cult, um, so I kind of just thought, well, you know, my thing earlier where it's like maybe if their reality is is equivalent to the mario to us reality, then maybe it'd be possible, but it does kind of change the um uh hypothesis, and and as we noted the last the time that um experiment does account for this, so that would just answer or that, like, okay, I guess the odds that we create, that anyone creates a simulation, low. So, so it's not as though this breaks that argument just to give all due credit to that argument. It's like that argument says either we can't make simulations or if we can, we're probably in one essentially. And so I suppose with everything you're telling me, it feels like, I don't know, maybe we can'T ever make simulations to this, to and to this complexity?

1:02:18 Bryan: Yeah, that's at least what I see. Now, of course, this is just me doing my calculations here. Most the time I'm on it, I'm giving you the idea of other people's work. This is one of the times that I did calculations on my own. This hasn't been peer -reviewed, so there could be some place I made a mistake. Haven't found anybody else trying to do this. I searched around a bit, but it could be somebody else was doing it. You certainly a possibility I made a mistake here. And, you know, if any of the listeners want to check the numbers, I can post up the the spreadsheet I use to do the work.

1:02:57 Thomas: Oh, yeah.

1:03:12 Thomas: Grade Brian's quiz, everyone. This

1:03:15 Bryan: You can, yeah.

1:03:15 Thomas: Is your final SIO exam. I

1:03:19 Bryan: Look there, peer review my paper. There we go.

1:03:19 Thomas: Say Yeah. That's interesting. I mean, you can almost, if people had tried to answer this, but if you are saying people haven't done this level of calculation, shit, maybe we should peer review your paper and publish it

1:03:29 Bryan: Yeah, well, not that I can find. They probably have, but well, I don't know. It was one of those things that, it's hard to find There's things. A lot of places people could have done it. And often what happens in science is people will do back of the envelope calculus and say, oh no, that's ridiculous and never bother publishing it.

1:03:30 Thomas: Yeah yeah

1:03:46 Bryan: I searched around for publications on this and couldn't find anybody trying to do this. Possible I just missed it. Um, it, again, this is a bit outside of my experience, what I usually do. So I don't the best keywords to search. I'm missing something. But Um. Yeah, the best I can tell is that the math says we are a ridiculous amount away from being able to do a simulation, even assuming we can hit the physical limits of it. And the physical limits themselves are unlikely that we can reach.

1:03:45 Thomas: Yeah, could be right Yeah, so forget everything, we keep zooming down and down and we went to the universe, then the earth and then maybe a brain or whatever, what if we just want to simulate one fucking atom or something or a couple of atoms, I'm, now I'm trying, I'm struggling, maybe that thing that I had heard was correct, was which is, it does, does take more, by the time you factor in everything you need to do it, almost takes as much to simulate things as the things itself, but is that right or not? Well,

1:04:19 Bryan: Yeah. Yeah. In a lot of cases, I'd say, yes, like, let's just take one electron. If we set things up, we want to store all the information on an electron. We need, what, 1 ,200 bits of information, and each bit requires at minimum one electron to store it. So we're using a thousand -some electrons to describe where one electron is. Alternatively, yeah, we just put one electron.

1:05:04 Thomas: Shoot when you put it that way, you could have just started There is

1:05:07 Bryan: Yeah, the simplest simulation of the universe is the universe itself

1:05:10 Thomas: The universe yeah No, that makes sense, yeah. Don't more people talk about this? That's just what I'm wondering. I feel like this has taken, taken ahold of so many Silicon Valley brains. And I feel like Silicon Valley brains are the same brains that should figure this shit out. You know, they're, they're the ones doing the computing.

1:05:11 Bryan: Oh I don't know, it could be that people just aren't familiar with things outside their specialties and a lot of this is getting into, you know, deep into physics and Particle physics, quantum physics, general relativity, it's covering all angles of physics. And just because somebody is good with computers doesn't mean they have that expertise, and people will very often overestimate their expertise at things outside of their specialty. And that's one of the things I'm cognizant of too, as you know, even within physics, a lot of these things aren't my specialty as well. I'm pretty confident with the way I put things, but I do want to leave that caveat for the

1:05:34 Thomas: Particle physics or something. Yeah.

1:05:54 Thomas: Yeah. -hmm.

1:06:06 Bryan: Listeners that I haven't put the rigor into this that I would have if I were, uh, it, um, I, I've put enough into, yeah yeah

1:06:11 Thomas: Well, shit, we should publish. Let's do it. I feel like this is like, people need to know this, damn it. We're not in a simulation according to this. I don't know. Or there's some, something you're not thinking of. There's some way to optimize something or some way that think technology will change or limit. Yeah, I don't know. It just seems pretty crazy, though.

1:06:19 Bryan: Yeah The optimization is a thing that's just very hard to quantify. Like, like I I said a while back, we need to optimize like 19 orders of magnitude. Now, maybe there is a way to do that, it's hard to say. And it's specifically hard to to say that it's impossible, it's just that is one massive hurdle you need to overcome.

1:06:50 Thomas: Shit, we don'T THINK WE LIVE IN A SIMULATION, BRIAN. THAT'S KIND OF MY POSITION NOW. YOU 'VE CONVINCED ME.

1:06:55 Bryan: That's kind of what I've thought and

1:06:56 Thomas: YEAH. ALL RIGHT. WELL, SHIT. MAYBE WE BETTER TREAT EACH OTHER KINDLY AND TRY TO MAKE THE WORLD BETTER THAN, I GUESS, ELON.

1:07:02 Bryan: Well I think we should do that even if we are in a simulation

1:07:04 Thomas: I KNOW. I'M JUST TRYING TO GET THROUGH THIS SPECIFICALLY, Elon Musk. I feel bit of a solipsist or whatever. I don't think he thinks other people are real. I get that feeling. So, hey, Elon, start using your money that you're dwindling fortune to help others maybe because of this. That's my message to him. All right. I know he's a listener. He's a big fan.

1:07:04 Bryan: Yeah yeah

1:07:26 Thomas: No, Dr. Brian Gillis, thanks so much for taking us through this. Really really fascinating. And I will be curious to hear if the listeners have any, have feedback on it. You know, if anybody fails you on this quiz, I don't, I, I doubt it, but please write in if you have any observations on this or anything, I'd be really curious to hear them.

1:07:22 Bryan: We go yeah, oh You're very welcome Thomas all Right, thank again to everyone for listening and for your future feedback.

1:07:47 Thomas: You want to plug in anything or

1:07:48 Bryan: As before, I think I'll just plug again for anybody who wants to support Ukraine and their defense against Russia. They've now set up a website for their donation hub. You can just either search for United24 or the website is u24 .ua .gov. there's a good place to give your money.

1:08:06 Thomas: Alrighty. Excellent. Well, until next time.

1:08:10 Bryan: Right. Thank you very much.