SIO415: We Don't Live In A Simulation

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 415. I'm your host, Thomas Smith, and All done quite a bit in recent episodes on the simulation theory or the simulation argument. Two slightly different things, but the difference barely matters. But anyway, the the question of do we live in a simulation? And And some of these have been pretty, you know, pretty technical, pretty dense, both the philosophy, the physics, the computing. Eat that stuff up. I love it.

But here's what I want to do in the kind of grand finale here. You are someone who, I don't know, maybe you didn't follow every detail of those. Goal for this episode is for everyone, whether you were super in with detail of that and following right along with Brian and I, or if you were like, ah, this is a little bit my kind of area of expertise or something.

Either way, I want this to be the episode you can point to if you want to say, hey, Elon Musk, for example. If you're, I don't know, Grimes or somebody, hey Elon Musk, we don't live in a fucking simulation, you idiot, or something.

That is my goal here, because this, my view, from all we've done, from the work we've covered both already on the show and from just a ton of philosophy papers that I've read about this, why, is not warranted to worry about whether or not we live in a simulation. I don't think we do, or if we do, it's of a nature that makes it such that we really should just not worry about it.

I mean, to quote the clip I played you actually of Larry King with Neil deGrasse Tyson, where he's like, so what do I care? I don't know anything. You know, it's kind of that. So that's my goal here. I hope this is something you can tune into or point people to if you ever in that kind of conversation to say, here's a bunch of reasons why don't worry about that argument.

Argument so quickly, just so we're all on the same page, and just in case there's anybody who just coming to this episode alone, I'm going to give the you second version of the original Bostrom argument that I have since learned, by the way, has been modified and updated quite a bit. But the argument is really quite simple, as long as you remember all parts of it and the three different conclusions, at least again in the version one of the argument. Basically this.

If high quality, indistinguishable from reality, basically, simulations are possible, technologically mature post -human civilization would have such enormous computing power that they would be able to run countless, indeed who knows, millions, billions of ancestor simulations. So that would be simulations of basically how they got there, you, civilization thus far, which could be our civilization thus. So if that's us, that means all the way from who knows when, could be the Big Bang, to all theway to whenever the post -human thing is.

So running the foal, essentially the playing the the tape on the universe over again. Given substrate independence, which is the idea that you don't have to be made out of meat to have consciousness, that hypothetically, if you simulated all the components of someone's brain in a way, that that would become conscious, would have experience, which isn't a given, by the way, but it's something. And it's certainly something that some form of which I I think is true, but anyway, set that aside.

Given that lots of simulations, the fact that these simulations would contain people like us, experiencing entities like us, based on those, and this is the part where you got to be pretty precise, Bostrom says at least one of the following is true. The fraction of human level civilizations that reach that post -human stage is very close to zero. Two, the fraction of post -human civilizations that are interested in running ancestor simulations is very close to zero.

Or three, the faction of all people with our kind of experiences that are living in a simulation is very close to one. He goes a little further to say, if one is true, then we will most certainly go extinct before reaching post -humannity. If true, then there must be a strong convergence among the course of advanced civilizations so that virtually none contains any relatively wealthy individuals who desire to run an ancestor simulation and are free to do so.

And if three is true, then we almost certainly live in a simulation. That's what that one was. In Key point: in the Dark Forest of our current ignorance, it seems sensible to apportion one's credence roughly evenly between one, two, and three. So that's him saying, and I've seen video of him saying this too, you know, not knowing anything else, it's about 33% % likely of each.

Then he goes, I think, an increment further here by saying, the last sentence of his his original paper, paper is, unless we are now living in a simulation, our descendants will almost certainly never run an ancestor simulation. Now I want to go through why I don't think we live in a simulation. There's a quick distinction to be made here because the argument itself, the one you just listened to, totally allows for us not to be in a simulation.

It means that if we're not in a simulation, then either we will go extinct before reaching post -humanity or basically us and any other possible civilizations that tend to be advanced don't make simulations, ancestor simulations. And in and two, there is room for us to be in the base level reality. And so there's a first set of reasons not to worry about us living in a simulation that don't go against this argument. They don't invalidate Bostrom's argument.

I'm going to start there, and then I'm going to move into the, that actually kind, this argument is invalid some way, or some of the premises are wrong, those kind of things. So first off, within the argument, himself says, I was, one of his responses, I watched some video of him, and he'll respond to people's kind of surprise at his argument by saying, often they'll ask, so do you think we live in a simulation?

He'll say, well, no, I mean, between one, two, and three possibilities, so knowing nothing else, I mean, I think it's less than 50% % chance that we live in a simulation. That's how he puts it. Meaning, you know, 33% % chance, even he himself, at that time, now he's adjusted that. That's the first little point I want to make. And it's about Bayesian probability theory and sort of thing. Going to go in depth here, but basically speaking, nothing wrong with Bayesian probability.

It's sort of setting priors in a way. And this is, just again, a very surface level look at it, but you set prior probabilities, and then as evidence is collected, as data is collected, you adjust, adjust those probabilities accordingly. Simple enough. Can get pretty fancy and complex with it. Also exactly the kind of thing that I think someone like Elon Musk would overuse. There are a lot of people who do overuse this.

They get way too precise with Bayesian priors and all that in ways that it's just way too concise to be like real, in my opinion. It's way to illustrate this, is it's a little weird that there are three possibilities, because that gets us to that 33% % each. So if you were betting, like, do we or don't we live in a simulation, it actually really matters whether his argument came up with those three conclusions, or 10, or two.

Actually makes a big difference in the odds if you're just distributing the probabilities, uh, evenly, you, which makes me wonder, like, why is the argument organized in that way? Like, could you have subdivided one and two differently, or could, or you combine one and two, so that it's like, yeah, either advanced civilizations don't make simulations or we're in a simulation, like, could really combine them, and then it's like, oh shit, all of a sudden 50% chance.

Or for that matter, I could, could see separating it out further, and including whether or not simulations are even possible in, I could see it being, well, okay, one of the following is true. One, it's not actually possible to create simulations, to the fraction, you know, and then go on from there. And then all of a sudden you've got four conclusions, and then you're at 25% % each.

Doesn't feel a bit arbitrary, you know, it doesn't feel like, well, I'm just kind of dividing things up using language, or roughly speaking, using human logic. And yet somehow that's drastically changing I'm supposed to believe the chances we're in a simulation are. That can't be right, could it? It turns out there's a slightly more technical way to describe what's going on here. From, nothing wrong with the idea of setting that prior probability, if You you've got no other evidence.

From what I can tell, there are some different, different techniques in assigning Bayesian, think it depends on what you're evaluating. This is stats that's kind of beyond my knowledge, but I, read you a little on it, and it seems that there's different techniques, and which you choose might have to do with what data you expect to collect, essentially. And that wrong in principle, was saying, well, I don't really have any other reason to anything. So, so just divide out evenly, divide out that prior evenly.

If you are dividing out that prayer evenly, and then you are collecting no data, rely so on those probabilities as such, just you assigning a probability and then going, oh, look, look at this probability. Probability, if you've never, if you don't take any observations, if you don't take any measurements of something, and you can't, because the way this argument works, you really cannot do that, in my opinion.

I mean, there I, are little hypothetical ways, maybe someday we could move those priors around, but if all you've done is assigned a prior probability and then done nothing after that, well, that prior probability is really not worth anything. It's, it's worth almost nothing. So this first objection is just within that itself. Even if you accept everything about the argument, you could accept that number three is, you, the chances are like nil, or almost zero.

You could buy into Bostrom's argument and say, yeah, there's a virtually no chance that we're in a simulation. Even if I buy into his argument structure, I think that one and two are like ninety nine point nine nine percent likely, at one of those, and then the could simulation be point zero one. Totally valid within the argument.

It wouldn't refute the argument, it wouldn't be a problem necessarily with any of his logic or anything, but it's just a completely valid way to evaluate probabilities, or to at least assign probabilities, or at least as valid as anything else, because we don't have a way of running the data here. Came across a really fascinating paper written by some physicists that I will probably talk about it another time. It is super fascinating.

It has a lot to say, not just about this argument, but about another argument that I may talk about next time I feel like the audience is ready for some more philosophy. I don't know, I how much this resonates with anyone, so I'll probably take a break and see. But anyway, really fascinating paper, title of which is, "Are We Typical?" And the whole point of it is, you really just assume that we're typical.

That's a big part of this argument, of the simulation hypothesis, is, hey, there's way more simulated people than real people, you can assume, like, we're normal, and so we're likely to be simulated. Not really any guarantee that that's true, or that that's a valid way of looking, looking at us, that we've been randomly selected from some population of potential life forms in that way. Now that gets really, really complicated.

So that in itself is not where I wanted to rest this objection, but they did have some great examples of this prior probability thing that I just told you. Their paper was more about, like, in a lot of cosmological theories, theories, there are a lot of theories, because again, these are physicists writing this paper, there are a lot of theories that are consistent with the data.

So my understanding is, I'm not a physicist, of course, but my understanding is there, there are many models which might, the universe, and might be consistent with the universe. And oftentimes, there's not really a good way of picking between them. There's no, there's no additional, like, scientific measurement you can take, or, or at least so far that that has eliminated one or prioritized one or the other.

And so often physicists will argue based on other things, like whether or not it predicts us as a typical observer. So, like, if one theory predicts, hey, there's going to be a bunch of us, and the other theory predict, well, there's not going to that many of us, then some physicists, or I, actually, actually I have no concept of how prevalent this is, but according to this paper, a lot of people will say, well, definitely prioritize the one that predicts a lot of us, because, you, greater odds that we exist.

So since we exist, since we find ourselves here talking about this, then, you know, we should say that that's more likely. The point of this whole paper is actually that's not a valid step, really. It's not, it's not. It's on a justified step. It's really fascinating. Just wanted to read a couple quotes, not to get too into the weeds, but this, a lot to think about with along these lines.

So, quote from this paper, quote, "Consider two theories of the development of planet -based intelligent life, each grounded in the appropriate physics, chemistry, biology, and ecology. Theory A predicts that there are likely to be intelligent beings living in the atmosphere of Jupiter. Theory B predicts there are no such beings. Since Jupiter is much larger than Earth, Theory A predicts that today there are many more Jovians than humans.

Would we reject Theory A solely because humans would not then be typical of intelligent beings in our solar system? Would we use this theory to predict that there are no Jovian because that is the only way we could be typical? Such a conclusion seems absurd." Paragraph that I really liked here that gets to that prior probability thing.

0:13:42 Thomas: An assumption of typicality by assigning a low prior probability to theory A and a high prior probability to theory B. Alternatively, we favor theory B so that the Earth is more typical of the planets in the solar system. However, any a priori favoring of either theory is contrary to the standard scientific practice of assigning non-informative priors that do not pre-select conclusions before relevant data is obtained, quote. So yeah, that really kind up what I was saying.

That data, just looking at your priors, it doesn't, that's nothing. You know, it's kind of like there's no good reason to put much stock in those priors. So I thought that was fascinating. Now we've gotten through the first thing I was going to say of like 40. So I'll go try to go faster. Terms of more objections that are within the argument so don't disprove Bostrom or anything but do tend toward no we're not in the simulation.

I don't find his proposition one to be very compelling. Like the the idea that oh some for some reason if civilizations get to a suitable level of advancement, something takes them out, like the great filter kind of thing, like, oh, they blow themselves up or there's some technology. I don't really put that much stock in that one.

I don't really think that's anything that you could bank on in you're talking about some like truth about how the universe works, that that would happen in all cases or indeed, I don't really see why that's separated out as one. And if that's separated out, again, why don't you separate some other ones out so that the odds that we're in a simulation then decrease because of that prior probability.

But setting that aside, number two, it was the fraction of post-human civilizations that are interested in running ancestor simulations is very close to zero. Now, here's one that's worth talking about, because that to me, if simulations are possible in the ways that the argument requires. So granting that again, which I don't, we'll talk about later, but granting that for now, I don't really see why this is anything anyone would want to do, at least in the way that would be required for this argument to work.

So for this argument to work, in the future has to have, and Bostrom in his paper talks about planet-sized computers. Like we have harnessed entire planets in to accomplish this level of computing that's hypothetically, you in this paper here. What we're going to do with that is run like a real simulation makes conscious just thinking, feeling people, but then also kind of just runs back the tape on it.

So, you does say like it could be near enough to, so it doesn't have to be exactly ancestor simulation, you know, like not just literally word for word, everything that happened, but close enough that it creates people like us, I guess. But first off, I wonder how reasonable that is to even do. Do. This So borders on kind of the other objections that are around us not really being able to feasibly make these simulations, but this this, kind of borders both of those things.

I'll talk about it now. Does that work where you have the power and the understanding of physics and all that enough to make a realistic, again, again, go back to the Brian episode, the last Brian episode about the actual nuts and bolts of trying to make a simulation in this way. Difficult thing to do. Apparently requiring a planet-sized computer that we're not using for anything else. Like, first off, that would just be used for porn. Like let's be real. Oh, maybe this is the porn.

Actually, now that I think about it, maybe there's a porn thing. Okay. So someone is running an ancestor simulation for porn purposes. And now I'm realizing I am completely questioning my own argument here. I think that's maybe the best counter argument I've ever thought of to my own thing. So setting that aside. But for someone to do all that, make conscious creatures, it's one thing to make like a human, maybe. Like I could see someone making AI.

I could see someone making, you an AI kind of companion that turns out to be a thinking, feeling thing. I imagine people have interest in the future in trying to do that, but making a full universe of people that interact and have conscious experience for for the purpose of what, like if we're, if we have that level of amazing computing power, well, we kind of already know all the answers to, you questions that that would answer, you really know what incrementally that would gain you for just how unbelievably expensive that endeavor would be. Idea that we're going to just be able to do that.

So again, if we are, here's something I thought of that I haven't seen anywhere. So this could be a dumb thought or it could be a good thought. I don't know. Keeping quantum physics in mind and how it's possible we don't even live in a deterministic universe, possibly depending on your interpretation of quantum mechanics and all that, how are you creating an ancestor simulation at this level?

When we envision, oh yeah, I know we got the Sims, we got, you know, we've got video games and then, you know we go to full on ancestor simulation. Well, if you're simulating again, if, if you're stimulating enough to where we're in a simulation right now and everything we know about the world is simulated to have that level of detail, that means you are simulating as Brian laid out, like every fucking photon, quantum, quark, electron, neutron, all that shit. You're doing all of that.

How do you make that work so that it recreates us? Is this a dumb thought? I haven't seen this said anywhere, actually. But how do you do that? It ends up that we're entirely deterministic, then one thing that would be maybe interesting is perfectly recreating all of the past and then, like changing variables here and there. Okay, that actually would be really fascinating. But is that possible? Do we live in a completely deterministic universe to where you could say, yep, we know enough.

We have so much computing power that we can recreate exactly the state of the universe in whatever year so that we have that perfect starting point to know with certainty questions about our history.

Maybe but assuming that's not possible because I really don't know if it is and we are are dealing with an indeterminate fueled quantum madness how are you intervening simulation wise to make sure the right wave function is collapsed in the right way at the right times through the, you whatever billion years of the universe until we maybe came along, something like that. How are you guaranteeing all of that is leading to us even to a planet?

So say you get to planet Earth and then you have to somehow somehow guarantee that every potentially random or potentially outcome quantum level, possibly, the way it needs to so that whatever happened to create, so we came into existence happened. I don't even know. I mean, how would you even do that?

Brian, if you're listening, let me know if this is a good thought if not I'll tell everyone later but I I don't know how you would even do that that's why I could see like some kinds of simulations maybe that are like on a broad in the same way we try to simulate climate change at a very distant level we don't try to stimulate literally every molecule in that is orders of magnitude less complicated, like orders and orders and orders of magnitudes, less complicated than trying to simulate literally everything ever and all brain states and all that.

This is to say there's a lot within that number two proposition, was, to remind you again, the fraction of post-human civilizations that are interested in running ancestor simulations is close to zero.

Again least in terms of the way this argument needs to go I would put an enormous amount of probability in my Bayesian prior in proposition two there if I had to choose like a massive project that I don't even understand how it would work and yet we're like yeah I know people will want to just do that yeah I mean this oh maybe but in the way that would be required and not only that in one of the papers I read there a very very complicated mathematical paper that I'm gonna briefly mention but it's it's It's beyond my ken.

So I'm going to comment that it exists, but I can't like evaluate it per se. But I only bring it up now because there was a little aside that was like, hey, not for nothing, like out stepping of the math for a second. One this is such a bad idea. Like we had the capability to do this, this is such an immoral idea that I would literally take up arms in a war to stop people from doing this.

We have a guy writing a paper and he's not wrong. He's like, yeah, no, I'll be drafted into the future Terminator war if it's to stop someone from doing this, because think of how much suffering of conscious creatures that involves, you know, and that's in this as well. Like there could be a moral objection to wanting to do this. And I just loved it because this paper, this wasn't the topic of the paper at all.

This was just this guy like, hey, just like extracurricular here, you down the chalk, stopping the math. This idea is so fucking terrible that I would fight the robot war to stop it. You know, I love that. It was really funny, but that's true. Like willy nilly, you're just going to create the massive amount of suffering and knowing humans, what will stop it is because they'll many dogs die too?

0:22:47 Thomas: They'll be like, well, we're not going to stop it because, you know, the Holocaust happens again, but the dogs, there's a lot of puppies that are going to die. So yeah, you're right. Let's not do it.

0:22:57 Thomas: Massive reasons to doubt that we hit that number three proposition again, sticking within this argument. I want to touch on the, this can't remember if this means the argument is invalid, speaking, if one of its premises is invalid, the I argument istill valid. So I think arguments are valid as long as their premises logically would lead to their conclusions or whatever.

And if you say, well, one of your premises is wrong, then you say okay, so the argument, you obtain or whatever, but it's not like an invalid argument, like if the premise was true then the argument would still work. So anyway, that's not that important, but point is, the major, major caveat to all of this is what Brian and I talked about in the last episode, that Brian took us through just how fucking impossible it would be to run a simulation of the entire universe.

Now, I I could see reformatting the argument, and I think Bostrom even did reformat the argument to have like more possibilities, and one of them is this is just technologically impossible. Suppose if you reformatthe argument like that, then the argument still works, I suppose. It's just you're like, yeah, but this is the conclusion, not that we live in a simulation.

One of the key things I worried about in Brian's episode, that I think he very well, and also I wanted to add in another echoing voice to that, was, as to this, he goes into a lot of why, oh, you wouldn't have to simulate the whole universe. Like you could just simulate like a human, and then, you know, you'd have to be able to get all the brain states of all the humans.

But he seems to think once we have those planet sized computers, that that's easy, which may maybe that's true, I don't actually know. He says, like, you wouldn't need to simulate all the universe and all that. You just need to do it such that, like, whatever any humans paying attention to kind of works properly. But then, and I asked Brian this, too, and he went through some of why he didn't think that worked.

And there's another person who had a great article and video about this that I think makes a lot of sense, is someone named Dr. Sabine Hassenfelder. Let me just read parts of this, because I think it echoes that point that Brian made very well.

"Bostrom therefore has to assume that it's somehow possible to not care much about the details in some parts of the world where no one is currently looking and just fill them in, in case someone looks." Oh, and I want to step out, because also in Bostrom's paper, he says, you know, there are some things that would have to be simulated all the time, like, you knows, computers. So he thinks that, that was kind of confusing to me.

He acknowledges in the paper, like, yeah, no, you'd have to simulate computers all the time, cause those are kind of like, you would notice if computers were like not being simulated part time. It seems to to be what he's saying, but like that wouldn't apply to everything. And Dr. Hassenfelder, I think, thinks that would apply to anything. So, continuing: "Again, though, he doesn't explain how this is supposed to work. What kind of computer code can actually do that?

What algorithm can identify conscious subsystems and their intention and then quickly fill in the required information without ever producing an observable inconsistency? A much more difficult issue than Bostrom seems to appreciate. You cannot in general just throw away physical processes on short distances and still get the long distances right." Climate makes sense to me. I I understand how this would work, that you would just not have to spend any resources simulating all the stuff in the universe until you need it.

You know, like right now, I can look around and look outside, I can I can get a telescope, look at a star, I can, you, we have have places where it would make all kinds of scientific observations of stuff, and that all works. Like the science all checks out very well, consistently, day after day after day after way. How in the world is that being accomplished by only simulating stuff that you're looking at or that anyone's looking at?

In mind too that people look at like everything all the time, so if you're really simulating the full universe, the full, and actually I'm just realizing too, it's necessary to Bostrom's argument that people, is being simulated, because then that's how the, the number of simulated people outnumbers the regular people, because you can't just say, oh, well, maybe it's just me, maybe, maybe they're just simulating me, me, and then, you know, nobody else, and so that saves a lot of computing power.

Well, then if that's true, then there's no way I outnumber the number of real people. So therefore, these simulations have to be complete, like completely everybody, all people. And I do not see any possible way that you could save significant computing power on that. And given that, and giving what Brian said about the full computing capability you would need to pull this off, that seems downright impossible to me. I strongly, strongly view the premise of this argument as flawed.

Don't think we're getting anything like ancestor simulations. We get some simulations? There even be simulations that have like a conscious, thinking, feeling artificial intelligence? Yeah, I think there could be. It be an entire globe history of billions and billions of humans such that, you this argument would work? Way. Been thinking about it. And the more I think about it, and I'll separate out, this is just my thoughts, I almost think that quote unquote ancestor simulations, simulations, might be a contradiction in terms.

Is a major reason why I don't think this argument works. What I just said, and given what Brian told us, and what this Dr. Hassenfelder said as well, and other people said, that you can't just like fake it on the simulated universe, you can't just like, oh, we won't do it it over here while no one's looking, like that doesn't work.

So, assuming you have to simulate each and every fucking thing, like to 80, whatever Brian said, just like insane numbers of photons, Higgs bosons, whatever, whatever else is the quarks, all that stuff, that's true, if, as again, we covered on the Brian episode, the most efficient possible possible way to simulate an electron is an electron, like a real electron. And the only current way to do any of that is with a shitload of electrons, meaning whatever you're simulating, definitionally, you must have way greater resources then.

So like whatever universe that I am going to simulate has to, I think, like definition, be less complex than the universe I inhabit. Therefore, the idea of an ancestor simulation, I think, might be paradoxical, because then you're necessarily having to reduce the world you're simulating significantly. However you have to reduce, use, that means you're not creating a true ancestor simulation. I that's actually important, this is not just nitpicking.

I thinkthat's very important, because a lot of the the punch effect of this argument is like, yeah, we're going to have simulations upon simulations upon simulations, and all that.

And if it's actually true that you need vastly greater resources to simulate something than the thing itself, so the base level reality, is it's, is not a true ancestor of the simulated reality, and if those are different, then the whole argument breaks down, because the only reason we can can make claims about like the nature of what the simulations will be like is by doing that little trick of, hey, it's an ancestor.

So that's why, if we develop the ability to run simulations and therefore everyone else did or whatever, you know, like making those claims based on what we do is because of that ancestor part. If actually it's fucking impossible to do that, and each time you simulate, sure, maybe you can make a simulation, but it's orders of magnitude less complex than your reality in some way, then you're not making an ancestor simulation.

Furthermore, whatever shortcuts you have to take, Brian talked about cat hair, you know, but a serious thing, like, whatever short cuts you have to take, maybe that affects consciousness, you have a shortcut quantum physics or physics or whatever, then maybe you're not simulating reality to the granularity required to make conscious thinking beings, and therefore the whole argument breaks down at that point.

Could still be true that you and I right now are in a simulation, but if we were, the parent reality, like the real reality, would be incomprehensible. I think it would have to be, like I said, maybe another dimension, there'd, you have to be some way that the computing power would be so much more powerful than what we can even imagine here that it would not be the same as our, like you couldn't call it an ancestor simulation, you couldn't call us an ancestor simulation, we would have to be some simplification in some way, that the argument, that's also when I thought of the whole, how would you make sure everything went the way it would need to go create a true ancestor simulation before?

Think that's a good point, I'm very curious at Brian's view, maybe there's some obvious way that, that's the stupid point, could be, but I don't know, like how, if you are truly simulating everything to that level of granularity, you have to somehow make sure every quantum event, everything went the way it needed to, to a true ancestor simulation.

So it's kind of those reasons all combined that I think ancestor simulations as a concept might be paradoxical and impossible, at last in the way you need them for this argument to work. You need them to be indistinguishable, virtually, from real reality for argument to work, I, I don't think that's at all true, be impossible, possible.

One more way that thought I of that too, is I, a little debate on this online and saw people saying, oh, if you could simulate the universe, then that universe could definitely simulate a universe, because you've simulated everything to the degree it would need, you've already given it all the computing power or power, power, you know, energy, whatever, that it would need to simulate a universe, which is reason why I think, definitionally, that means you can't be simulating our full universe, it's impossible, wouldn't that just create infinite power, wouldn't violate something.

So if I can, if we could somehow simulate a university that is our entire universe, and within that universe, well then they could simulate a universal that's our entire university, and then they consider, and you could go forever, wherever, where are you getting all that computing power?

If I start doing that with like virtual machines, so I can run a virtual Windows on my Mac, if, let's say, then I could like run a virtual Mac on that Windows, that, that isn't going to fucking work very well, you're going to get multiplicity disease, you're gonna, you, you know, the clones of the clones of the clones, you're drastically reducing quality there. And pretend that continues forever.

So like, I'm going to simulate Windows that's going to stimulate Mac, that's gonna simulate Windows, that's gunna simulate Mac, it's just going to to come a crawl. It, but what if didn't? Like, would that be possible? That would mean you could just create infinite computing power just by simulating things, you knows? Because like, in addition to simulating the entire universe, other stuff is happening in that universe. So like, you've ended up with, mathematically, you're creating your universe plus more.

This is just me thinking about it, but that feels like it would have to violate some conservation of something, maybe Brian could tell me. So for that reason, again, it would half to degrade each time, and how much is it degrading? And I would think quite a lot, enough to where you're not really making ancestor simulations in the sense needed for this argument.

Other way, this is a major one I'm excited to talk about: if you tuned into this entire series, you've heard me say that I think this argument just has to be circular, that we could be using what we do to then say we're in a simulation or not, that's been my intuition on this. I was excited to find that there's a philosopher who had that same concern, but wrote it out in a way that's, you more, better.

0:34:39 Thomas: And really cool. That's why I really enjoy philosophy. I love thinking about these things. I love thinking about problems and coming up with objections. And then checking like, okay, did anyone else think of that and write something on it? And either they did. And that's cool. 'Cause it's like, oh yeah, that's validating. Like, I feel like, uh, you know, my instincts were, were good there, or they didn't. And it's like, well, I probably am wrong for some reason.

You know, there's gotta be a reason I'm wrong, or like, one in a billion, I thought of something original, but that's, I don't put very much credence to that. So I love that process. It's just really fun for me. This one was really satisfying to read because, like, I talked about this simulation argument on the show a long time ago, and I had the same objection, and I didn't find this objection written out then.

So this is like years in the making of, ah, I knew it. I knew this has to be circular. This comes from just Brian Eggleston is the name. This was referenced in the Wikipedia, but I just had missed it because it's like one little sentence. The Wikipedia sources to a Stanford website that's no longer available. So it's just a Wayback Machine.

So I don't know if this was like a class, I think it might've been a class paper or something, but this is cited in the Wikipedia is, like, yeah, this is a source. So, I, maybe they wrote it somewhere else later. But anyway, this Brian Eggleston seems to be a philosopher, but I actually have no idea what level of eminence or accomplishment or whatever, but it doesn't really matter, I guess. He points out the problem here, and he does it with the more formal probability, whatever crap that brain kind of blocks out. So I won't go through that.

0:36:06 Thomas: I'll try to put it in just the plain plain English parts and not the like math parts. Says, "However, obviously we cannot count individuals from simulations that we ourselves run, because these simulated individuals don't contribute to the possibility that we are in a simulated universe, since we know for sure that we aren't them, since we created them. In fact, the only simulated individuals that can contribute to the probability that we are living in a simulated universe are individuals that we haven't and will not create.

In other words, only individuals that aren't from our universe or from universes that we might eventually simulate can be counted, as these are the only individuals for which the principle of indifference holds." Is important because it changes the expectation of simulated individuals that Bostrom is trying to calculate.

"The probability that at least one civilization reaches the ability to run simulations is equal to the probability that with the potential to reach such an ability exists times the probability that the civilization actually manages to reach the ability." And math, map. It was okay to assume that this equaled one, because we knew at least one world, our own, exists with the possibility of running simulations someday. Day. This allowed us to reduce the expectation of simulated people to math.

"However, because we can't count our world toward the expectation of simulated people, if we want to maintain the principle of indifference, the proposition must must become the proposition that a world other than our own exists, in which a civilization has the potential for achieving the ability to run simulations." Thus, the expectation of the number of simulated people becomes a new equation.

Up, he says, "But now that is simply the prior probability that we place on the existence of a world other than our own." If you didn't catch what that punchline is, it's basically, once he does that and kind of maths out the math there, it cancels out to the probability that we're in a simulation is more the probability that we think simulations exist outside of our world or whatever. You know, like it just reduces to like, oh, that's not really saying anything.

Before it seemed like it was saying this really interesting thing, but now it's just saying, well, the chances of simulation are the chancesof simulation, essentially. So it was very satisfying to see someone actually did write that out in a way that, you they crushed the actual, not numbers, but variables or whatever. But that's always been the thing.

At the core of this argument is, in whatever reality we're in, the relationship of the number of simulated people like us to real people like us, its heart, that's what this argument boils down to. Thinks if simulations are possible, then essentially simulated individuals will vastly outnumber the regular ones, assuming there's not some other reason why that just doesn't hold up to a statistical certainty.

Then that brings up another thought I had, which is, well, there a lot of ways that the number of people like us, again, that's what he says, and that's a problematic thing, but we'll leave it for now. The number of People Like Us could still be kind of greater than the simulated people, even given this argument, right? Because like, let's picture a universe, and our universe, we know it's vast.

Say in whatever the, you real universe is, that there's tons of alien life everywhere, because there's, there's billions and billions and millions of stars. Let's say there's just tons of civilizations that contain people roughly like us, maybe, and let's further say that even if one of those develops the ability to simulate ancestor simulations, that's that civilization exists for a very long time time before that, such that the people who came before the simulations were possible are huge, huge numbers.

In other words, it would mess with that ratio. Like even in a universe where simulations are possible and somebody does some simulating, well, maybe it takes quite a lot to outnumber the real people, depending on how those simulations go. So that was my thought. And then I came across something I didn't know existed, which is Bostrom's revised this argument a couple of times. It's actually really hard to find. It's not on the Wikipedia, really.

Like there's reference to it, but the links all go to the same website. So I had kept just reading that same website, because every link goes to that. But it turns out there's other areas in this kind of badly made website, or if not not badly, just very sparsely made, that contain his responses to like new objections. He has revised the argument. In my view, he's revised it to the point that it's really not quite as effective as it's been taken to be.

One of the revisions kind of involves the thing that I just said. It's slightly more technically accurate than what I said, which is, these these civilizations would be doing ancestor simulations, again, taking that on board, that that's possible. What, there's, say, two civilizations in the universe, just for simplicity. In one of them, there's like a kajillion humans like us, like people like us. And never develops the ability to create simulations.

Meanwhile, there's another, the one other civilization, assume they do achieve the abilityto create ancestor simulations, but they do so at a way earlier stage. So there haven't been as many lifetimes. So when you do the math in that universe, all the stuff would still hold for his argument. This is why it's a problem, because all the stuff still holds, but the conclusions don't. So all the things would still hold. So that would mean that are possible. That would mean civilization created one, etc., etc. But just by the math of that universe, it would still be true that the majority of people will have been real people, because the other civilization that never made the simulations never got to that that level had more overall life forms, and the civilization that did get that to that level had fewer, and so when they ran their ancestor simulations, it didn't actually add up, the mess of the numbers don't work, basically, and that's a problem that Bostrom identified.

And then what he does is something really bizarre to me, and this is where things get too technical, I'm not going going to try to go through it, but titles the paper "A Patch to the Argument," and he offers two patches to fix this problem. I don't understand that. So I can, I can summarize kind of what they are, which is, well, you can just assume like these certain ratios.

You can assume that like, A, maybe that won't happen, or B, he does this kind of what seems like a bit of a trick to me, where he says, well, I, this is not even worth going into, but he does a thing to say basically, at where we are, like you and I right now, relative to the invention of one the megahertz processor, processor, that time difference means that we're close enough that people like us would still majority live in the the simulator.

To me, it seems a little tricky. I don't know for sure, because it's pretty complex. Right away, just even at the that we're patching the argument to make it work, then to me, that sounds like, well, then you're just trying to make the argument work. You're not, I don't need a patch for the argument. I'm just going to patch it up. I'm going to work. That just means the argument doesn't work, and you need to add in more

0:43:33 Thomas: And that's, that's how the arguments presented in the Wikipedia, under the updated section, there's like seven caveats now. So once again, divide up 100% chance by seven or whatever, and now you're getting even lower just from that prima facie probability assignment or prior assignment. Worse than that, this is where that paper I mentioned earlier comes in, that's so complicated I can't understand it.

The one where the guy's like, "I would take up arms against this, that someone's gonna make simulations." But that paper purports to show that when you crunch the numbers on Bostrom's new version with a patch, essentially, that it says way less than it did before. And it relies on, this is according to this paper, Dustin Weir, titled "Correcting Errors in the Bostrum-Kulisicki Simulation Simulation Arguments." I want to emphasize again, this is too complex for me to evaluate whether it holds up.

I just want to present that it's here, and it is also where that little blurb I read before was. So that was funny. But what this paper says, if I kind of read you the conclusion, here's Where the example parameter settings turned out to rely on inconsistent interpretation of vague magnitude terms. Been a little complex, but what this paper is saying, I read through it. It has so much math I don't get, but it's saying that like once Bostrom had to patch the argument in the way that I just described earlier, the time he got to all those modifications, he's relying on words that are like very likely, extremely small, almost certainly, and these are quotes, astronomically large.

And this paper says, hey, when I actually like figured out what all that would needs to be, I hiding a lot, unintentionally or intentionally, I don't know, but he's, he's hiding a in those vague terms. How much is very likely versus how much is almost certainty versus how how much is extremely small? This person is saying when I, when I coded all this out, like mathematically, the argument is nowhere near as powerful as it used to be because of all this.

And makes some sense to me, but after but after reading all this, uh, again, I can't verify that that's true, but just wanted to point out that that is something that is a contention of the person or people who wrote this paper. So honestly, there you have it. Know that wasn't like a listicle, whatever version. It wasn't a very short like top seven reasons this doesn't work. I know, but it's complicated stuff.

Basically, not only within the argument are there lots of reasons why the part where we're in a simulation could be extremely low probability and still the argument could be sound. There are also genuine reasons to question the of the premise that we could even make something that would make sense to call an ancestor simulation or that those are even possible. Even setting that aside, then I moved on to problems with the argument that make it seem possibly invalid.

And that has to do with this, how it's circular and also that mathematical problem that I don't totally understand with it. So all that is to say, well, I don't think we live in a fucking simulation, man. I just don't think we do. Or if we do, it's real reality is vastly more complex than we could fathom. Then at that point, when does it become indistinguishable from like, yeah, okay, well maybe that's what we would think anyway. You know, you know what I mean?

At what point does that like something like either a multiverse or like a deity or something? You know? At at what point, what is that even adding to say? Okay, it could be a simulation, but the real reality is so much more complex or capable of apparently vastly more computing power than we are at the very least. So therefore it'd be kind of unrecognizable to us. We couldn't really comprehend. Well, then we're kind of where we are anyway, where it's like, yeah, all right.

I I don't know how exactly we got here. Maybe you're religious. I'm I'm not, but you maybe it's some kind of multiverse thing. Maybe it's whatever, but essentially where's that getting us now? If we've changed the argument much, that it's barely gets you anything now, what is it really even doing anymore? So hope that brings you some small comfort because I think we live in a simulation, man. I just don't.

So next time Elon Musk smokes a bowl and tries to tell you how he's certain 99.99% sure, uh, full of shit, I I don't think that's valid. I think that Bostrom is a very smart person who has now trimmed away his argument so that it's still logically valid, but it says not much in my view. Idea of an ancestor simulation, I think is impossible. Actually, I'm going to, this is my opinion. I really don't think that's possible. So all those reasons and more. Congratulations.

We are the winners of a real reality. We are a new the proud owners of the base level reality in my view is what I think, or if we're not there, it's so impossible conceive of what the base reality is that doesn't matter. So there you go, folks, I hope you enjoyed that. I certainly did. I learned a ton. I really didn't know all that stuff that Brian brought us about the computational power needed and all that.

And there is so much other philosophy here that I got into that I might want to bring up at a later date. Probably probably going to give us a philosophy break for the near term here, but that was a lot of fun. I you enjoyed it half as much as I did, and I'll see you next time. Thank you so much for listening.