Got a nice long action packed episode for you today. In what was originally going to be a two parter, Phil Torres and I talk about the simulation argument made popular by Nick Bostrom. I highly recommend you read the original paper and some of the other resources found here. In the second half, I make a lot of reference to the FAQ on that website, so please check it out. Specifically question 4.
Podcast: Play in new window | Download (Duration: 1:09:54 — 64.9MB)
Subscribe: RSS
I think you’re right that it’s circular. Or, some such shape.
The argument doesn’t work, even if we are in a simulation.
If i understood right, it was one of the following three must be true;
1)Our species dies out prior to reaching the tech required for such simulations.
2)We reach that level, but do not make simulations of conscious beings for ethical objections.
3)We are in a simulation.
3 Is basically that we create those simulations and so must be in one of them. It is where the problem lies. Pick it apart how you like, inductive reasoning is about crossing impossible divides.
It tries to rule out as an option; we aren’t in a simulation and yet create them in the future.
I’m going to relisten later and think some more, but as an analogy,
Say in the future, our technology gets so advanced we develop time travel.
Now, by the fact that we don’t have it now, as it would be brought back by the very nature of time travel, we can conclude that we will never develop time travel, of the sort that can go backwards.
That feels false to me too, for similar reasons perhaps.
“3 Is basically that we create those simulations and so must be in one of them. It is where the problem lies. Pick it apart how you like, inductive reasoning is about crossing impossible divides.”
I think this is a little off: if the “divide” you’re referring to here is epistemic certainty, then yes, induction fails. But this does not problematize the argument: probabilities are precisely what Bostrom offers, which is why he uses terms like “almost certainly.”
As for the first part of your comment, the argument states that if disjunct #1 is false and we reach a posthuman state, and if disjunct #2 is false and we run a large number of ancestral simulations, then it follows that we reach a posthuman state and run a large number of ancestral simulations. This does not quite get us to disjunct #3 as Bostrom states it. For this extra step, Bostrom employs some probabilistic reasoning: if we reach a posthuman state and run a large number of ancestral simulations (in which minds like ours exist), then the number of sims will vastly exceed the number of non-sims.
Now here is the tricky part that seems weird but is actually quite robust: first, we must be “agnostic” about whether we exist _in vivo_ or _in machina_. Why? Because a sufficiently high-resolution simulation (which is what we’re talking about) would be indistinguishable from the “real” world — unless of course our simulators decide to swoop down from the heavens to say Hi. This being said, the “indifference principle” tells us that when we lack independent reason for thinking that one option is more like than the others, we should distribute our probabilities evenly. So, if we do this, then we far more reason for believing that we’re sims than non-sims — i.e., simulating a large number of ancestral universes will constitute strong empirical evidence that we’re in a simulation.
Does that make sense? I wish I had explained it better in our discussion — I blame it on pre-election cognitive exhaustion compounded by some Trump-induced depression. 🙂
(Further questions/contentions are more than welcome.)
I think you cant use the indifference principle: it simply breaks down. Besides the ‘simulation’ we could be in one out of an infinite number of “non real” worlds contingent on any one out of an infinite number of ‘real realities’.
As opposed to that ,you suppose only one possibility for our current reality. Then you argue we must accept that we are in a non-real reality because of the indifference principle.
First I would argue that in case of infinities the indifference principle breaks down. And one could argue that (as Legolas’ Elephaunt) the infinity of non real possibilities ‘still only counts as one’.
I would only accept the simulation as an hypothesis. And then look for actual evidence for it. If you say there us no way for us to distinguish between a simulation and reality all you are saying there is essentially no difference. Then, we can use Occam’s rasor to reject the simulation hypothesis.
It also seems to be ignoring possibility #4 (or #5 if we count yours) that creating such simulations turns out to be impossible. I personally don’t see any reason why artificial minds couldn’t be created, but the argument seems to go a step further than even assuming that, instead assuming it is possible to simulate universes.
The argument does assume Morse’s Law to be true.
I would actually argue that it doesn’t. The argument works as long as computer progress continues, whether or not this occurs at an exponential or linear rate — or in “punctuated equilibrium” fashion according to which development occurs in spurts separated by long periods of stasis.
Sure, that’s fine. As long as there is progress toward more complexity.
For the computer power to be adequate, all sub-atomic particles must have all their properties correctly modeled for all of space-time that we can interact with. Creating the storage mechanism for that and the compute necessary to keep it going seems like we would need more atoms in the computer than the computer could be capable of modeling. Large saving could be gained with aggressive compression and large amounts of the universe being just estimated. both of these are possible, I guess. But the scale of this endeavor is so high that I would bet that Type II civilizations would only be able to model a small percent of the universe that we currently interact with. At a certain point, it becomes less complex to just build the experiment irl (like the computer built to calculate “the ultimate question”).
“. . . all sub-atomic particles must have all their properties correctly modeled for all of space-time that we can interact with. . . .”
Engineers of our own species on our own planet in our own time already know this is impossible and have gone past it. Such a brute-force approach is tantamount to creating a copy of a universe. Why bother? Look at the one you already have.
It is far, far more efficient to select and define particular scenarios (or “test vectors,” as the nerd contingent calls them) and simulate only those elements relevant to what the simulator wishes to look at. This is how computer chip designs are tested and verified. We already make stuff way too complex to simulate every imaginable permutation of.
As David pointed out, the God Geeks can stop the simulation, change the rules, move the scenery around, modify the scope and we’d never know the difference–unless they wanted us to–unless they wrote that into the simulation. If they do simulations using methodologies we have already mastered, but on a far grander scale, it’s not that unimaginable.
‘creating such simulations turns out to be impossible’ would be a case of Bostrom’s disjunct #2, while ‘we aren’t in a simulation and yet create them in the future’ is covered by #3.
I don’t really see your confusion here. This is a straightforward proof by contradiction, and is definitely not circular. https://en.wikipedia.org/wiki/Proof_by_contradiction
Me neither. Maybe Thomas could clarify his objection by stating what the ‘alternatives 4, 5, 6…’ he referred to might look like. I don’t see how any imaginable scenario could not be covered by Bostrom’s three possibilities…
Thomas there really is no difference between the robot AI, and the simulated AI. The consciousness of both are would be all software. Only the external stimuli would be different. One would interact with the “real world”, and the other would interact with a simulated world. If we were able to perfectly represent the the world using virtual reality, and put you in it, either by wiring you up, or uploading your brain to a computer, would you be any less conscious?
I was going to make the same point, but I thought it worth checking to see if it had already been made. I’m curious about what Thomas’s conception is of a robot brain that isn’t software running on a computer.
I think 3 takes account of the low (very very low) probability of us being the first reality and we end up simulating.. Someone has to be the first however unlikely it is. This seemed to be Thomas’s example of something outside of the three… I’m curious if Thomas can come up with something other than this that falls outside of the three options.
So, this is a very popular paper, right? Thousands have read it, contributed ideas and/or synopses? Endless discussions, and nobody noticed the typo? That is sooooooo annoying. (I. Introduction, Paragragh 2, the second word is “form”. It should be “from”).
If I understand your objection correctly Thomas perhaps the argument is based on the assumption, given the age of the universe, that IF a simulation were possible it would already have been done. Hundred’s of billions of stars like our sun, which is only 4.5b years old have gone through their entire habitable life cycle in 14b+ years. I mean if it hasn’t happened yet it seems highly unlikely it ever will.
I should have read the FAQ #7 essentially made the same point.
It is so frustrating to (once AGAIN) hear Thomas and a guest talking past one another. The irony is, Thomas, who keeps saying “I’m probably wrong, but I can’t see why”, is actually RIGHT, and he can’t see why. It’s not that the argument is circular (not Bostrom’s, anyway). It’s that Torres is adding a very important, and fallacious, aspect to the argument that does not exist in the original. Bostrom never says, never never never, that option 3 means we must be in a simulation, and that is you know you are NOT (for argument’s sake), then option 3 is impossible. His argument only says it is extremely LIKELY. The two statements “We are not in a simulation” and “Option 3 is correct” are NOT a paradox, any more than a lottery winner is a paradox. Just because it is very unlikely that a given person should win the lottery, does not maked every lottery winner a paradox. It is frustrating to see this very very obvious point be missed by both proponents of this argument.
I don’t think that’s a satisfying answer. All three options say “extremely unlikely.” So if that were the way my objection were being dealt with, the argument would become meaningless since all three of those options could be compatible with it. Like I highly doubt that’s the way Bostrom would respond. I’m working on trying to express my issue with it more technically.
I think the problem with Tom’s objection that we know we’re not in a simulation is that it assumes we could possibly know that. How would you figure it out?
That’s not at all what my objection was
Well, more precisely, you had more than one objection. But you did make that one: at around minute 43, where you start assuming that we’re not in a simulation and then argue that it’s possible neither on the two remaining propositions are true.
“Bostrom never says, never never never, that option 3 means we must be in a simulation, and that is you know you are NOT (for argument’s sake), then option 3 is impossible.”
Perhaps I over-simplified for the purposes of exposition, but you are absolutely right that the argument is probabilistic. I’ll have to re-listen to my initial explanation of it, but I would be surprised if I wasn’t explicit about this. Alternatively, you may be confusing our discussions of the argument itself with our discussion of FAQuestion #4, which first assumes that disjunct #3 is absolutely true and then assumes that it’s absolutely false. The issue here isn’t about modal possibilities, but it is about non-probabilistic truth.
Maybe you could clarify some more?
Phil,
No, you did say this explicitly, but I do think you could have emphasized it more.
Ah, okay. Thanks for clarifying, and my apologies if it was insufficiently acknowledged!
Phil is correct here. However, he fails to emphasize the main idea of disjunct #3. Assume 1 and 2 are false. In that case there is just 1 “real” world and trillions upon trillions of simulated worlds with sentient beings. If there is no way to tell if you are actually in the real world and a simulated one, the chances you are in a simulated one far, far outweighs the possibility you are in the real one. To make this concrete, if there are 10^20 simulated world’s, and only 1 real one, the chances any particular bring finds itself in the real one is 1:10^20. It would extremely irrational to assume you’re in the real world.
If #1 and #2 are false, it doesn’t follow that there would be a particular number of simulations. There might only be enough power to run one.
I think part of the argument is that it’s taken as a given that a posthuman / very advanced civilization would find it pretty trivial to simulate beings and (parts of) universes.
If you don’t take that as a given then of course the whole thing breaks down. I think many people who have thought much about technology and the workings of computers and automation would come to the conclusion that it will at some point become pretty easy to simulate minds (if you’re not a Searlite).
To me, this argument has always been the Noah’s Ark of philosophy. They both make perfect sense if you ignore time and space.
You can’t represent something with less information than the thing you’re representing. So if you have a molecule in a simulation, that molecule has to be modeled by something. That something *has* to contain at least as much information as the thing it’s representing. So in order to model a room, you need to have enough bits to model every atom in that room. So basically, your simulator has to be at least as large as the thing it’s simulating. Forget modeling a universe.
Then you also have speed. If you are modeling a change, the change cannot happen faster than the thing modeling it can change. So any simulated particle will necessarily be slower than the thing modeling it. Forget about modeling billions of years.
You can make a traedoff though. You can sacrifice detail for speed, but that means that your model is losing information. Just like a video game, it would move quickly but lose detail.
So you have three options as I see it.
If you accept that our universe works as it does with fields and particles, in order for the model hypothesis to work, you have to accept last Thursdayism. The universe was created as-is, recently. Also, the universe cannot be as big as it is. Anything outside of a very small part of it cannot exist as we understand it.
You could go with a low-resolution universe, in which ours behaves like a video game where there aren’t really atoms making up the wall, it’s just a texture file applied to a modeled surface.
You could also go to hard solipsism, or shared hard solipsism. The universe exists only in your mind or in our minds only as a representation of what we understand about the world, but we can’t really do anything with it. It only appears that we can.
If you take the first tact, we would not have the time to create nested models. Someone outside of our universe would have to keep the lights on for a really long time, and the universe we could create, even if we had enough time, would be much smaller and slower than our own modeled one. Not to mention the fact that our universe would already be much smaller and slower than the one modeling us.
If you take the second two tacts, we could never create a nested universe, because either nothing outside of our minds actually exists, or the universe doesn’t behave in a way in which we can actually do anything with it.
Regardless, this isn’t how our universe appears to work. Our universe is old. Our universe is finely detailed. Our universe is big.
So as far as I can see, the argument only works if you ignore computation theory and the nature of our universe.
This is too narrow a view. It is only valid because it assumes our large scale values are actually large. It’s actually all scalar. Our speed of light, for instance, could be quite slow in the “real world”. Perhaps the size of our universe is contained in something we would consider tiny if we were in the “real world”. How could we ever know?
I thought I posted this as a response to David’s November 3, 2016 11:55 PM comment, but it ended up at the bottom, so I’m reposting it so my argument is complete. Not that it probably matters.
————-
That misses the entire point of the argument. Of course there’s some possible universe you could imagine that could simulate ours. But the power of the argument comes from the asserted fact that we could likely do this, and we know we exist, so we know it could have been done to us. It’s that bootstrapping that makes the argument work. Otherwise it’s no different than saying there could conceivably be a god who created our universe. Sure, there could be, but it’s not an argument.
I don’t think it misses the point, I think you miss the power of the counter to your post.
What makes you assert that our universe is big? That doesn’t follow by any stretch. We have a few extremely low resolution signals from things so far away in time and space that they could all be simulated by a current day laptop in real time.
So actually, our universe is *just us*, on our planet. Nevermind that you don’t know if the universe goes beyond just you, the universe humans currently inhabit is a speck in the universe we assume exists. For beings that are super-intelligent multi-star civilizations, simulating billions of planets full of people like us would seem to be pretty easy.
That is, in real time *or even faster*. Why simulate in real time? If you’re simulating something pretty small, do it faster! And yeah, why not start the simulation last Thursday?
So the feasibility seems pretty straight forward, at least to me.
“. . . So actually, our universe is *just us*, on our planet. . . .”
Is that a given? I’m nearly as well versed in the Atheist liturgy as some of the others, so I didn’t know, but that makes for a much simpler discussion if you can forget about the crowd in Andromeda (the next galaxy over) who were simulating universes before our planet cooled.
I think you’re either over-estimating the power of laptops, or underestimating how much we pick up from space. But even if our planet is the only thing that is simulated, it’s still untenable.
It’s not as easy as some people seem to think to create a large, detailed simulation. People seem to think you can make something that’s bigger on the inside like the comment in the podcast of students running billions of universes on their laptops. You just can’t do that. Information entropy doesn’t let you.
So to simulate even just our planet, or even just a city, that behaves like our universe would be a monumental task. And the assumption that post-humans would spend the massive amount of resources turning planets into mini universe simulators is completely spurious. And if they are so post-human that limits of time and space become essentially meaningless to them, I don’t think we can speculate about what they would try to accomplish. They could just as easily be paperclip maximizers for all we know. To assume that we know they would make simulations and not just a few, but millions, and that that is the most likely scenario pushes credulity even further.
If our universe is solipsistic, then the argument almost seems to work. But if we are in a solipsistic universe, then that negates the premiss because then we are not beings capable of becoming post-humans that could create simulations. So in order for that argument to work, you have to assume your conclusion.
Besides, you would still have to actually have the conversation about solipsism. Just saying “Nevermind that you don’t know if the universe goes beyond just you” is to completely miss the weight of the subject.
I agree. I think the discussion here is suffering from a malady that effects almost all theist/atheist discussions: We are discussing at nearly infinite length that which we have neglected to define. Discussing is easy and fun. Rigorous definition of terms and conditions takes time and consensus and quickly becomes work.
I think the “simulation” has to be limited to the bubble of perception of earth-bound humans or else the conception of it is unmanageable, forget the execution of it. But is it to be limited to what we can conceive of right now? We have technology now that was flat unimaginable a hundred years ago. To rule out a simulation on the grounds that it is technologically impossible may be short sighted; there’s a pretty good chance it won’t remain impossible forever. Anyway . . .
@Joel you just don’t need to simulate all that. I don’t know if it’s that nobody read the original paper, maybe not. Even the FAQ addresses this.
You don’t need to simulate all the details of a full 13 billion year old universe in order to fully simulate the experiences of a few trillion people for thousands of years on a planet or two. You only need to actually simulate the things they are looking at.
Quantum effects literally make almost no difference to the simulation. Stars, distant galaxies? We get an extremely narrow band of data from those, and only when we have a telescope looking at it. Maybe my laptop was off by a couple orders of magnitude, but who cares? By definition we can’t be getting any more data than our entire computing capacity, and that’s pretty insignificant to a posthuman civilization. So *none of that* has to be simulated. Only the data we get has to be made convincing enough.
Same with microscopic effects. Cells don’t have to exist. Germs, microbes. None of that. Just simulate their effects, and then when someone looks close with a microscope? They just need to see a tiny patch, not a whole body of cells, just a few thousand maybe.
Chemical reactions don’t need to be simulated. Just their expected outcomes. You know, the way we perceive them.
I think you fail to grasp the massive amount of stuff that can be cut out of a simulation and still have it be convincing. Completely realistic. The most complicated thing to simulate is all the brains, and once you have that you just feed them a convincing reality generated on the fly.
Or you could just go read the paper, it’s much more detailed and convincing than me:
http://simulation-argument.com/simulation.html
I was mistaken in my comment about if you were the only being in the simulation. That doesn’t align with the simulation hypothesis. For the hypothesis (not the argument) to hold, many full civilizations have to be simulated such that the population of those outweighs the chance that you’re in the baseline reality.
We cannot not be in a simulation until we can explain existence. Until then, what we call Reality is real because we say it is real and we cannot prove that anything is any more real than that. We simulate Reality on a moment by moment basis simply by saying “this is real” and “that is not real” without being able to prove that what we call real exists outside of our perception. We simulate reality on this level by default without being aware that we’re doing it.
There can, of course, be levels of simulation above where we are here—wherever that is. Some folks in “Reality A” might be unconsciously “simulating” a local reality of their own minute-to-minute that just happens to include a planet-sized super computer on which they are consciously running a simulation of a world they made up (our world, Reality B) just to see what would happen. Regardless of what their Reality looks like, they are a lot like us in that they cannot explain their existence. They can, however, explain our existence: they made us up, on purpose, just ’cause—they know where we came from.
In my opinion we (royal rhetorical ‘we’) are always assuming a bedrock Reality that, even if it is really there cannot be explained. We want to know if we are that Reality or some lesser reality. Are we sure that the intrinsic nature of any reality level is of a higher order than that of any of the others? Maybe it’s a daisy chain made of nothing but daisies.
Besides, conscious awareness is overrated. A weather vane has a measure of consciousness. It can sense stimuli, respond and communicate. What else do you want? There is a paper, published in 1959, that is scary, if you ask me, called, “What the Frog’s Eye Tells the Frog’s Brain.” It’s scary because these four researcher dudes found that a frog is essentially a hardwired device that has more in common with an electromechanical washing machine (or a weather vane) than a human being. The animation supplied with the animate object we call a frog is disappointing to say the least—worse than South Park. Makes one wonder what other “living creatures” in our world are barely living at all. Maybe that’s because these things are a “simulation” and not really real, Maybe in bedrock Reality A, eight levels above us, there are philosopher frogs that are really deep once you get to know them. Hell, maybe the frogs a running the simulation.
http://pespmc1.vub.ac.be/books/whatthefrogseyetellsthefrogsbrain.pdf
I have two objections, here’s my first. Suppose we are in a simulation… well, the beings running that simulation will also use the argument to conclude that they are a simulation. And perhaps they indeed are, but the beings running *that* simulation will do the same, and so on. Every being, whether real or simulated, concludes that she is simulated.
So, assume there must exist a level of “true reality” somewhere (I hope that’s a safe assumption, i.e. it’s not just an infinite regress of simulations all the way down!). Anyway, the beings living in reality will use this argument to incorrectly conclude that they are simulated. Therefore the argument necessarily leads to an incorrect conclusion at least some of the time. That makes it much less powerful than something like a mathematical theorem, which would be required to be true in all circumstances and all times. Instead we’re in the realm of a probabilistic rule-of-thumb, which is interesting but not very satisfying.
Exactly.
Yes, one out of, say, 10^20 beings (just a large number) who say they are in a simulation will be wrong. Why is that a problem? It would be very irrational for any of them to conclude they are not, even though they might actually not be in one. Therefore, assuming 1 and 2 are false, it would be very silly to conclude we were the one and only true reality.
But where are you getting that number of 10^20 simulations from? Couldn’t it be a smaller number like… 2? All we’re really saying with this argument is that we don’t know whether we are in a simulation, which is trivial.
Sure, that’s possible. I was just pulling some large number. The point is, once we were able to do it once, it seems unlikely we would stop at 2 or 200 or 20,000. Think about how many simulations we run just to predict the weather.
And while it is trivial that we don’t know, we are asking what is most likely give certain assumptions. If we assume disjunct 1 and 2 are false, then it follows probabilistically that we are now in a simulation.
Even if it was 2, each civilization would usually be correct to assume they are in a simulation. But I think it’s not much of a leap to assume there can be more than 2. I think that’s part of the premise (that it’s possible).
So if it turns out that making simulations is necessarily very difficult and thus we have only, say, 100,000 simulated people at any one time, does that mean that those simulations are not simulations based off of your Indifference distribution? The Indifference distribution here seems flawed, as well as other parts, because the probabilities involved are simply being created whole-cloth. It seems similar to the theists that argue for the existence of a dirty from probability, failing to note that we would need to have previous examples and data to generate a probability.
That aside, what do you think of how theists regularly use this argument to claim the existence of a deity? Granted that is obviously not its purpose but the only time I ever really hear this being discussed is when it is being used in that manner.
I’ve brought it up before in theistic discussions and most conclude it isn’t how they conceive of God. Mainly because would make the programmer a fallible being, not worthy of worship.
A being who can turn your reality, your very existence and all evidence that you ever existed OFF and go to bed may not be worthy of worship, but is as close to God as makes no difference.
Yeah, but you’d never know when that happened. Think about it. We could be on a backup tape that has been locked away for a thousand “years” and just been brought back online yesterday. And we wouldn’t have a clue. And while I agree it is as close to a god as we might get, I don’t think it actually qualifies.
Well, we’d have to define god, which seems to be the bugaboo of theist/atheist discussion. We talk for hours and hours, stretching into centuries at this point, without knowing for sure what it is we’re discussing.
Surely this leads to pointing out that the possibility that no being is worthy of worship and, at a minimum, any being that was wouldn’t accept worship in the first place.
Of course. Though it should be noted that a being being *intrinsically* worthy of worship is a contradiction. “Worth” (value) is based on desires and the ability of something to fulfill a desire. Perhaps one may believe worshiping a being fulfills some desire. In such a case, that being would be “worthy of worship”, but only to that individual.
Yes, the childish narcissism of demanding worship automatically disqualifies one from deserving it, which has always been a problem with the hairy thunderer Judeo-Christian god concept. Omnipotent? Maybe. Divine? Probably not.
What about a #4 where it is technically impossible to create such a simulation? Just to accurately simulate the solar system down to the smallest particles it seems to me the computer would require a hard drive at least the size of the solar system.
Sure, that’s possible and should probably be added to he disjuncts.
It’s also possible that our entire universe takes up what amounts to a thumb drive in the “real world”.
“It’s also possible that our entire universe takes up what amounts to a thumb drive in the “real world”.”
I suppose that’s possible if the “real world” had particles many orders of magnitude smaller than the simulation, because you can’t store zero’s, and ones in any less space than the smallest particles allow. Just imagine if it took only 1mb to accurately simulate a quark, then the real universe would have to have particles a million times smaller, and that would only result in a simulation of the solar system that required solar system sized storage. So actually, unless there are in fact particles in our universe smaller than quarks it would be impossible for us to simulate one.
Look, it’s all scalar. Perhaps it would take 1,000,000 of our particles to represent the smallest possible particle in he simulation. So what? The simulated people have no reference point to compare their quarks with “real” quarks. Perhaps our quarks are made up of millions of particles in the “real” world. We think the numbers of particles in the world is large, but “large” is subjective. Perhaps the real world has quadrillion times more particles in it. In such a case, simulating our world would be trivial.
“Perhaps the real world has quadrillion times more particles in it. In such a case, simulating our world would be trivial.”
I conceded that in my comment, The more important point I was making is it wouldn’t be possible to do in our universe. That being said St. Ralph suggested below a way it might be possible, but I may have problems with that as well.
That level of detail is unnecessary. To simulate Mike Paps you don’t have to simulate every molecule in every cell of his body, you just have to define his range of perception and his reactions to stimuli that he is able to perceive. I’m not saying that’s simple; it’s not. But it’s more doable than simulating every quark in the Universe.
One interesting concept is a zoom-able simulation in which resolution or detail is only generated if needed. If you only know of Mike Paps via his comments on Atheistically Speaking, and don’t know him personally, there’s a lot of detail that can be left out of a simulation of him. If you actually meet him sometime, the simulation will have to be kicked up a notch, but for now, it can be relatively low-res.
I’m skeptical in the feasibility of this, though I admit it is not impossible. It would have to work very hard to be able to predict where each being was going to look and adjust the resolution on the fly. Otherwise, people would be reporting very strange observations. We don’t hear or experience that ourselves. Granted the software may just “be that good”.
We actually already do this to a degree with CAD programs such as AutoCAD. In such a program, there exists, in a file, a mathematical description of an object of some sort. When a view is rendered of the entire object, say an automobile, you can see what it is very easily, but the available resolution limits the detail. When you command the program to zoom in on, say a tire, the program doesn’t just magnify a portion of the existing picture; it goes into the descriptor file and generates a new view of the object based on the data found there. In this new view you can’t see the whole automobile anymore, but you can read the lettering on the side of the tire and see individual pebbles lodged in the tread thereof. We do this already and we do it very well–as far as it goes.
As you said a few comments back, we simulated beings don’t necessarily have any concept of the passage of time at the simulator’s level. If we turn our heads and refocus our eyes, it could have taken their machinery days to re-render the new view we now “see,” but we’d never know it. Almost by definition there can be no correlation of our time with theirs. With no such thing as “real time,” all kinds of thing become possible–maybe ALL things. [fade to black as Outer Limits theme comes up gradually in the background]
Okay, you got me. 😉 Good point.
Almost every video game since video games existed already do that. You think they render the stuff behind the camera? Nope. It isn’t there. VR (in the sense of Oculus Rift) is a big deal because it has to render for each eye! So darn right they don’t render what’s behind you. Nor do they render what’s far away. Textures get downscaled, geometry gets lower detail. Some games make that obvious, but it doesn’t have to be so.
My argument exactly. Only what’s in front of your face has to be simulated in any detail.
“One interesting concept is a zoom-able simulation in which resolution or detail is only generated if needed.”
Your suggestions make a lot of sense, but could it be done so perfectly we wouldn’t be able to determine we were living in a simulation? I suppose we could be simulated in such a way that we don’t actually have free will, and would be incapable of noticing discrepancies that might tip us off.
How would Mario know he was in a computer game? Maybe he would not. He would not know his blocky, 2-d world is not the way the real world is. His mind may not be able to even comprehend our world any more than we can really comprehend a 4-d cube.
I think “free will” is only going to get you in trouble as “libertarian free will” may not be a real think in our world.
“I think “free will” is only going to get you in trouble as “libertarian free will” may not be a real think in our world.”
Yeah I wanted to clarify that I don’t believe we have free will. I simply meant that the creator of the simulator could have intentionally restricted it, much like Mario’s. If Mario actually had consciousness, and his will wasn’t intentionally restricted who knows what he might be able to deduce about his universe.
That’s essentially covered by #1. We fie out before it is possible to do is no different from our perspective
One important thing that this episode did not cover: is there any possible way to determine whether we are living in a simulation?
That answer is actually yes. There are a number of things we can look for that would provide launch evidence.
For one, as far as we know, true random number generation is impossible for computers. Our computers use very complex algorithms to generate pseudo-random numbers. If we were to one day discover a property that while “should” be random, actually always adheres to an algorithm, that would be strong evidence.
There may also be evidence in high-energy cosmic rays if we observe they travel on a grid as opposed to completely analog. There are currently experiments designed to detect this anamoly.
It’s not that hard to generate a pseudo-random number sequence that’s thousands of years long at the top speed of a supper computer. Encryption algorithms rely on them. It might be more possible to find evidence of filtering or modification of apparent “randomity.” Scientists, depending on what they are working on, often want the equivalent of a bucket of sand in which the grains of sand are oriented completely at random, that is, no pattern of grain orientation can be detected. But . . . they only want sand. If you leave randomity up to honest-to-God Pure Chance, your sand is going to have twigs and bottle caps and used band-aids and old dead jellyfish in it. There’s random and then there’s random. If we don’t find an old dead jellyfish once in a while, it might be evidence that someone has been running our sand through a sieve.
Granted, even if we were in a simulation, we may never find pseudo-random numbers. My only point was that *if we did*, that would be strong evidence in favor of it.
What might the analog of a dead jellyfish actually look like to us? What kind of thing did you have in mind? I am thinking of the 8 8’s in in a row in pi. Is it that kind of thing? And how could we detect a lack of that?
It would be proving a negative–probably impossible. I was thinking more like how when they get the Infinite Improbability Drive really ticking over in The Hitch Hiker’s Guide To the Galaxy, sperm whales and bowls of petunias spontaneously pop into existence in outer space above a planet. The fact that stuff like that never happens in real life should be a clue.
To say that *never* happens, you would have to be monitoring every planet in the universe for all time.
Precisely.
That misses the entire point of the argument. Of course there’s some possible universe you could imagine that could simulate ours. But the power of the argument comes from the asserted fact that we could likely do this, and we know we exist, so we know it could have been done to us. It’s that bootstrapping that makes the argument work. Otherwise it’s no different than saying there could conceivably be a god who created our universe. Sure, there could be, but it’s not an argument.
But let’s say we build a simulation that is only 1/1000th our size. Those people would still conclude that their universe is vast. They might say “it would take a computer the size of the [their] universe to simulate it”.
Just like you are right now. My point is, saying a computer would have to be vast to simulate all the quarks and everything isn’t a good argument against the SA.
I like to think I managed to follow the twists and turns in that discussion.
I like the point that was made about the parallel to the argument from evil, though I think that creates one of the biggest problems!
Given that the argument sets out :
1. Humanity goes extinct before we reach post human epoch.
2. We reach a stage where we have the capability of creating sims but decide not to.
3. We create ancestral sims ergo we are (likely) in a simulation
Given that we recognise the problem with a being, (omnibenevolent or not), creating a reality in which the inhabitants undergo emence suffering, I don’t see why we would ever get to the third option in the argument.
Secondly, we are currently able to run simulations that don’t necessitate fully sentient sim characters to produce whatever result is desired. Sure, if it becomes possible it may be desirable to produce a detailed estimation of how certain civilisations lived millennia ago for example. However, would it require an entire universe to be simulated, or just enough of a world to examine whatever the focus of the sim would be?
There were a tone of other issues I thought about during the discussion but typically they’ve evaporated from my mind at this point!
Ooh that was one!
I was fine with the reasoning that we could evemtually be able to simulate a mind in a computer. What I couldn’t get past was, at what point is it justified to suspect that these simulated minds would (or could) reach a conscious or self aware state? Not only that but what would be the benefit of creating consciousness in a simulated world?
Right I’m sure there are other things I’ve forgotten but that’ll do for this TLDR!
I agree with Thomas that parts of it seem circular, particularly at one point where Phil basically made a “the bible is true because the bible says it” type comment concerning point 3 and us living in a simulation.
I do find this type thought experiment, philosophical discussion to be slightlty pointless (not the right word) at times. As in what do we actually learn regarding the true nature of our reality by discussing the not very well defined parameters of a “possible” alternative reality!
Anywho I look forward to Thomas’ follow up episode if and when he does one. Hopefully by then I’ll have a better understanding of the simulation arguement.
What do we learn? Not much. Without anything to confirm, it is as useless as solipsism. But if we can discover a way to confirm (or disconfirm), it would at least tell us something very interesting about reality. And isn’t that enough? Isn’t that what much of science is about anyway?
Isn’t that why we’re limited to thinking about it, because we cannot confirm it! Yes science is about modelling and trying to explain our reality, it should by its nature be verifiable. This is not that, it’s playing clever logic games, connecting lots of sometimes tenuous dots and reaching a hypothetical conclusion!
Say we could confirm that we are in fact just part of a simulated reality. What do we then do with that information? Would we act any differently in our day to day lives?
Ouch. While the flaw in your logic is obvious to me… it does not follow that I will have any more success pointing at it than Phil did… but here goes:
The argument is carefully crafted so that all possible universes fit into three buckets:
1. We will go extinct before being able to run many simulations.
2. We will never be able to run simulations. (might be technical reasons, or moral, or other.)
3. We will run simulations.
I think that the above is unnecessarily complex, and can be reduced to 2:
1. We will not run simulations.
2. We will run simulations.
But reducing to two possibilities is unsatisfying because we naturally ask “why would we not run simulations?” So, let try to have out cake and eat it to:
1. We will not run simulations.
a. We may go extinct due to “the great filter.”
b. We may choose not to run simulations due to moral concerns.
c. Simulations on that scale may not be possible.
2. We will run simulations.
If you follow this so far, lets switch to a second question: How can we know if we are in option 1 or option 2? Hard to say. Lets table this question, and first ask:
What are the difference in the option 1 and 2 universes?
Option 1 universes look just like the real “top level” option 2 universes. Option 1 universes look just like option 2 simulation universes. (This is part of the premise of the thought experiment. You may not like this part… as I do not.) The inhabitants of option 1 have the odds of being non-sims of exactly 1:1. The inhabitants of option 2 have the odds of being non- sims of 1:(whatever the number of run simulations).
We are not quite ready to figure out what option is true… we have to address your question of the “Mario brother’s universe.” This is out simply on the premises of the thought experiment. We are postulating and advanced civilization running “real universe level high resolution simulations” to learn about the universe. Perhaps with the idea of teaching school children about how evolution happened. “Today, class, we are going to use our ‘my first universe simulators’ to see how intelligent life is inevitable on an Earth like rock.” *This* is where (arguably) this thought experiment fails. I think it is trivial to say that “real universe level high resolution simulations” are more expensive than lower resolution simulations. And like Mario in the “Mario brother’s universe,” the inhabitants on such a universe are unlikely to see how the resolution could be higher… or what that even means. Much like the thought experiment of trying to describe 3 dimensions to a purely 2 dimensional creature. It cant be done except with abstract math. (Try to understand what a 4 dimensional cube is.) Therefor if we live in option 2… while it is possible that we are in a “real universe level high resolution simulations”, it is more likely that we are in a “Mario brother’s universe.”
But lets go back to the thought experiment as described. We are not going to deal with the much more prolific “Mario brother’s universes.” If we live in a world that we can expect to *ever* run “real universe level high resolution simulations”, then it follows that we are more likely to be sims than non-sims due to the odds above. The infinite arguments will come in talking about whether we can avoid killing ourselves before then and if such computer power is possible. Things like moral concerns are (in my option) red herrings for the same reason that the first human genetic experiments have already happened despite them being morally questionable at best.
Taking the extra step of asserting that “if it were true that we are non-sims, then if follows that we are in option 1” is unnecessarily confusing… but true. It is true because of the combination of the odds and the fact that option 1 and 2 are “all possible universes.” If you think it does not, then put your mind at ease with a small edit:
1. We will not run simulations.
a. We may go extinct due to “the great filter.”
b. We may choose not to run simulations due to moral concerns.
c. Simulations on that scale may not be possible.
d. Other.
2. We will run simulations.
Sure, maybe we have not thought of other options. So what?
I think you did a good job of spelling this out, however I think you made some leaps that don’t follow.
You said:
“The inhabitants of option 1 have the odds of being non-sims of exactly 1:1”
That is actually NOT the case. The inhabitants of option 1 can also be in a simulation. Because option one includes (in your example) a and b, they’re both realities where simulations are possible. If simulations are possible, then there’s no reason you can’t be in a simulation where people decide not to do more simulations. Hell, maybe that’s a way to avoid the infinite recursion of universes, always make sure your inhabitants go extinct or don’t want to create deeper sims.
I suppose part of the argument is that if we have the existence proof of simulations (i.e., option 3 where we definitely DO create them), then we can have high confidence we’re in one. Options 1 and 2 don’t mean you’re NOT in a simulation, they just mean you don’t have any special reason to believe it.
“That is actually NOT the case. The inhabitants of option 1 can also be in a simulation.”
I think this is a good point. In this case, we have a new decision tree:
1. We will not run simulations.
1.a. We may go extinct due to “the great filter.”
1.b. We may choose not to run simulations due to moral concerns.
1.c. We may choose not to run simulations due to other concerns.
1.d. Simulations on that scale may not be possible.
1.e. Other.
2. We will run simulations.
I think this is flawed as I will restate: “We” decided that the power to do human genetic experiments should not be used… yet someone did it in China… and it will happen again. Given a long enough time line someone with do anything that humanity is capable of. I am having trouble thinking of a reason that we would choose not to run sims if we were able to. I personally think it is immoral, but I dont think that would stop other people.
While this makes things messy for the thought experiment, I think the point is still valid that: ‘If we think that we will ever be able to create a “real universe level high resolution simulation”, we should realize that it is overwhelmingly statistically likely that we are sims.’
My second objection is that, from what I know about computer construction, I don’t believe that we will ever be able to create computers powerful enough to simulate entire universes. I guess this puts us into a slightly modified scenario 2: it’s not that we choose not to create simulations, it’s that we are unable to do so.
Here’s how I back that up. Transistors are getting smaller but pretty soon they will hit a hard limit when they are comprised of just a few atoms each. At that point, Moore’s Law will necessarily stop. Okay, you say just create larger computers; but then you find that you can’t communicate fast enough between the different parts due to the speed of light. Already the time it takes for light to travel across a 1cm-wide CPU is preventing us from ramping up clock speeds. The CPU in my 2015 MacBook Pro runs at 2.3 GHz, barely any faster than 10 years ago.
Maybe quantum computers will save us… but the physicists I know seem doubtful about whether they will ever be possible. Something about the qubits needing to be protected from entanglement, not just while the computer is running but until the end of the universe! Otherwise the entanglement travels back in time and invalidates all of the answers given by the computer?! Yeah quantum physics is a mindfuck, but suffice to say that practical QC is by no means inevitable in our future.
The thing your and similar objections miss is that there is no reason to simulate an actual whole universe. The “whole universe” can literally be one person with all simulated inputs. Or it could be a planet where all the stars appear to follow the same rules but actually are just specks of light, or pictures on telescopes when you look closer. You don’t have to simulate anything close to all the atoms in a universe to have a fully convincing simulation.
And that’s all not to mention that the universe simulating ours doesn’t have to be the same size as ours even appears to be. It can be vastly larger. It also doesn’t need to follow the same laws of physics (Mario doesn’t follow our laws of physics, for example).
I understand that we don’t have to simulate every unobserved atom in the universe. But even providing the total sensory input for “only” 7 billion humans is a massive computational effort. Not to mention keeping track of objects while they are not being observed so that, for example, Arcturus doesn’t suddenly move to the southern hemisphere.
And you’re right we can’t say much about the complexity of universes that may be “above” us in any simulation hierarchy, but Bostrom’s argument rests on us, in *this* universe, either simulating universes or becoming extinct. I don’t believe we can simulate a universe remotely as large or complex as ours, even with shortcuts, so at the very least this puts us at the bottom of the simulation stack.
We don’t miss the missing pieces. You (the Simulators) don’t have to simulate anything but our perception of our universe because, outside of that, by definition, we don’t know the difference and can’t know the difference. When we go to a whoop-tee-doo CGI movie like “Dr. Strange,” in setting up the story they don’t follow every character from birth through grade school and high school and college, etc., etc. to the present. And they don’t follow every character to breakfast and to the bathroom and wherever else. There is a lot of presumed information missing, yet at the end of the movie most of us feel that we’ve been told a pretty complete story, like we’ve shared a pretty real experience. Of course, in the case of a movie that depends a lot on the skill of the movie makers–the Simulators.
I think if we look at what we know from direct personal experience compared to what we “know” via presumption, deduction, induction, “reading between the lines” and “filling in the blanks,” we find that 99% of our world is made of “file footage” that we arrange as required to make a sensible perception of a reality that we construct in response to precious little direct stimulation. There are supposedly people who remember every blasted thing that ever happened in their presence in their lifetime. Some are said to be geniuses, others appear to be tormented souls, but I’m not one of these alleged people and I don’t know any of them. To simulate me all you have to do is provide a memory of having heard of such folks.
I say there’s no where near as much to simulate as we think there is.
This books explains it in mind numbing detail:
http://www.nytimes.com/books/98/05/03/reviews/980503.03johnst.html
I question the subtitle of the review, “The brain creates a picture of reality that we mistake for the actual thing.” In my opinion, there’s a non-zero chance that the picture of reality that the brain creates IS the actual thing–or as close as anything comes to being “actual.”
Well, I think you’re thinking too small. Our “universe” currently consists of a single planet among trillions. There is so much space out there. It should be possible even using today’s computing technology to simulate just one planet’s entire reality including all the brains (the biggest part) and a simulated reality for them to exist in, given that you’re dealing with some civilization that can make use of all the planets and essentially unlimited energy from a star.
But that doesn’t matter. It’s one of the *givens* of the argument that a brain is relatively easy to simulate with a small computer. Section III of http://simulation-argument.com/simulation.html outlines this part. Given a range of estimates of what it is possible to compute with a given volume of matter, you can take the low end of the estimate and it’s still trivial to simulate our entire history of human minds in a computer that weighs a kilogram.
If you look over section III of that paper, are there any number or estimates, as conservatively as you want to interpret them, which you dispute? If not, the hardest part of simulating our “universe” is simulating the brains in it. And given the argument asserts that can be done with a pretty small computer, it should be easy to simulate billions of our universe for a posthuman civilization.
Basically this seems to boil down to
1 we will never create universe simulations
Or
2 we will.
And these options seem to be ascribed equal probability.
How is this different to
1 the universe was not created by a god.
Or
2 the universe was created by a God to look exactly like it does and to fool us into thinking #1 is more likely.
Then giving both options a 50/50 chance.
I’m probably wrong here, mind. Would appreciate someone smarter showing me the difference.
You’re not wrong, that’s roughly what the argument says. They don’t have equal probability, the probability is unknown. But if we determine it’s possible to do option 2 (because we do it), then we should have high confidence we are in a simulation. And if, as many do, we suspect option 2 is possible even though we aren’t there now, we should still have some suspicion that we are in a simulation.
That’s the point. Not that options 1-3 have equal probability. The argument simply groups reality into 3 different possibilities. And all it says is that if possibility 3 happens, we are probably in a sim.
Your analogy doesn’t quite work because we can’t get any evidence for option 2. The reason the original option 2 (er, 3) holds is that the premise is simulations are real and we make them.
Two things:
1. How is this not turtles all the way up? Phil and Thomas were talking about A and B and C, but why wouldn’t it go on for infinity using Phil’s logic?. And infinite processing seems unlikely.
2. How is it not clear that a heap is any disorganised collection of element where at least one element is not in contact with the surface.
Having been a life-long proponent of the Turtles All the Way Down theory, I laughed out loud when I read your question, “How is this not turtles all the way up?” My sentiments exactly.
But something else that occurred to me is that we must be talking levels of Reality here, not necessarily universes. We can assume that no one on Earth who listens to Atheistically Speaking is successfully simulating subordinate universes right now, but that doesn’t mean that no one else in this galaxy, or in one of the 100 billion other galaxies we are vaguely aware of in this universe, isn’t doing so. Someone in THIS universe could be simulating another universe without us having any knowledge of it or ever being capable of having any knowledge of it.
.
Does that change the discussion at all, if we can’t even be sure whether or not “we” are capable of such simulations or are already conducting them? Who is “we?”
– – – – – – – – – – – – – –
Your definition of a heap is elegantly accurate and adequate.
So a heap could consist of 4 grains of sand? That doesn’t sound quite right. The problem is, “heap” is an imprecise collection noun. The same issue arises when we ask “when did mankind become conscious?” “Consciousness” is not well-defined, so it’s kind of impossible to say with precision. The best we can do is give a range “around this timeframe.”
Two grains of sand, actually, if they were regularly shaped enough for one to reliably sit atop the other. I know what you mean, though. When Jed Clampett spoke of “a heap o’ grits” he was definitely talking about more than a spoonful.
I dunno. I’d consider two grains of (flat) sand a stack – even if they were not placed (or stacked). And to David’s point, I would consider 4 grains that happened to land so that 1 was supported by the other three a heap. The smallest heap, but a heap nonetheless. Please don’t ask when that heap becomes a mountain.
Consciousness is different to me because I think the complexity of the question defies a simple answer – too many variables. That said, I think if you wanted to agree to define consciousness as an awareness of your surroundings, then you pick that point at which a cell (nerve cell perhaps) allowed an organism to know something about its surroundings. But then some folks won’t like that use of the word “aware”, and fuzziness ensues.
It isn’t turtles all the way up because the argument doesn’t say that in scenario 3 you ARE in a simulation. It says that in scenario 3 you are vastly most likely to be in a simulation. In scenario 3, there is precisely 1 of the however many realities that is not simulated. So it could be you, or one above you, etc. But probably not you.
In reference to scenario 3, the last paragraph of section V says the the probability of being in a sim approaches 1.
Section III seems to refer to one civilization. So, a civilization (B) in scenario 3 creates a simulated civilization (C) also creates C1, C2, C3, etc.
Why isn’t civilization B itself subject to the same argument so that there is likely B1, B2, B3, etc created by some Civilization A. And then the same applies to A. Turtles all the way up…
Wouldn’t it be special pleading to say there’s only one B civilization?
Yes, B is subject to the same rules, so they would also conclude that there should be a B1 B2, etc. That doesn’t mean it’s turtles all the way up, and no special pleading exists anywhere.
As the argument goes, there is no point where you can conclude that if you make a bunch of simulations, that you’re not also in one. However, “approaches 1” as you quoted, does not mean “equals 1”. *Somebody* who makes that assumption is wrong.
It’s like saying, a rational person would assume that they will not win the lottery. But of all the people who assume that, someone is eventually wrong. It doesn’t mean the reasoning isn’t sound.
The difference seems to be that the odds of winning the lottery are nowhere near 0 in the number line Bostrom is working with. They’re only 1 in 300 million or so. Given the asymptotic numbers that Bostrom is claiming and accepting option 3, let’s assume we didn’t win the sim lottery and we are in fact sims. Then why wouldn’t the same conditions apply to our simulators? Which implies they’re in a simulation, and up the turtle ladder we go…
No, no no no you can’t just say “approaches 1” *equals* 1. That’s not what approaches means. It gets arbitrarily close to 1, but cannot actually *equal* 1.
Saying it’s simulations all the way up is entirely nonsensical, and nobody serious is arguing that.
Apologies John, seems we have gone so deep in this conversation that I can’t reply in the normal fashion.
Please note that I very carefully *didn’t* say that “approaches 1 equals 1.” My point was that according to Bostrom, you apparently have a MUCH greater chance of winning the lottery (probably every week) than not being in a simulation.
Why is it nonsensical to say that, assuming we are in a simulation, that the folks simulating us wouldn’t be bound by the same conditions that Bostrom has laid out for us?
It isn’t nonsensical to say the people simulating us could also be in a simulation. There’s a good chance of it. I don’t think Bostrom goes over the probability distribution as you go up the sim-chain.
I suppose your exact words, in the comment I replied to, aren’t nonsensical. The part I would quibble with is “implies they’re in a simulation…”, because it doesn’t. It merely implies they’re *probably* in a simulation, and probably their simulators, etc etc. That leaves open the fact that the simulations have to stop somewhere.
But, maybe I misread your comments altogether. It sounded (to me) like you were saying that it really is simulations all the way up. It can’t be that. However, the *logic* of the simulation argument does indeed hold all the way up. It’s just that at some point the argument is wrong.
The implication being that whatever civilization is the “real” one, also should assume they’re in a simulation for the same reason. It’s just that by chance they’re wrong.
” . . . The implication being that whatever civilization is the ‘real’ one, also should assume they’re in a simulation for the same reason. It’s just that by chance they’re wrong.”
So we could be real, but we wouldn’t know it and would, in fact, disbelieve it even if it were the case. Sounds like you don’t get much reality with your reality. Human existence is pretty much wrapped around the axle for good (or ever or whatever).
It’s either “turtles all the way up” (i.e. we are simulated, and our simulators are simulated, etc etc to infinity) OR the argument causes real beings to incorrectly conclude that they are simulated.
Yes, it causes exactly one level (of billions, assumed) to incorrectly conclude it is simulated. If, as the argument asserts, there are millions or trillions of simulations, that approaches 100% accuracy.
Ugg.. I’ve heard this type of logic loop from intelligent design and young earth apologist, though at least this is a little more nuanced.
I think this thought experiment (I really dislike using “hypothesis” or “empirical evidence” for these arguments) has the trouble of replacing the belief of something bigger than us and filling the “god hole” in that there has to be larger/higher beings than us.
I have the same brain block as Thomas, and I guess I think my idea that I know we have cars, but I cannot assume that other alien civilization have cars or ever had cars or even there are higher alien civilizations.
For laughs, how about these.
4. We are not living in a simulation, because a simulation has no empirical or logical reason to allow a simulate to discuss their simulation.
5. We are in simulation, because philosophy logic often gets into a repeating loop that looks likes a program glitch loop one might find in a simulation. See number 4.
I think you nailed it. I bet the Simulators didn’t expect to get nailed by a simulate.
Neither of those (your 4 and 5) make any sense or have any reason to be believed. That’s the main difference between this argument and any theists belief in god.
It is not a logic loop. It’s sound, there is just some mental block that you’re not getting past, as you said.
How about, what part of the argument doesn’t make sense to you? Or maybe the problem is you don’t know how a logical argument is laid out so this one seems to fantastical?
Keep in mind, the *premise* or given of the argument is that conscious experience CAN be simulated. Seems believable to me, but if you disagree, then the argument fails.
Also keep in mind, the argument makes a case for how much computing power is necessary to simulate a human brain and how much is possible in a given lump of matter. If you disagree with its conclusions, the argument won’t work for you.
If the above two things are believable to you, then what can your objection to the argument possibly be? There is nothing circular as Thomas kept saying.
Another way to look at it is this: if we some day DO create simulations of entire civilizations (say, many of them), then we know at that point it’s possible. And the number of simulated beings far outweighs the number of “real” beings. If that’s the case, you would be expressing a bias in reasoning if you concluded that you’re not in a simulation. You’d be giving special weight to your reality that it doesn’t really deserve because you have no evidence to conclude your reality is any different than the ones you’ve simulated.
This is just how logic works. It’s just a little bit mind bending because it’s applied to yourself and your very consciousness. But it’s the same kind of logic we use to assume physics is the same on Earth as it is in some distant galaxy. We assume we are not a special case.
A key thing some people seem to miss is that the hypothesis doesn’t assert that you are in a simulation with probability 1, it says the probability gets close to 1 in the limit. There has to be exactly one exception, and of course it could be us. Just like we could be the only life in the universe, but not likely given what we know.
I think the biggest objections people have fall into a couple categories:
1) They disagree with the givens of the argument. That is, that minds can be simulated. Or how *much* computing is necessary for a mind, or how much computing can be done with a lump of matter. The argument rests on specific ranges of possibilities for those things, and if you don’t buy the premise you won’t buy the argument.
Another variation of this is in vastly overestimating what needs to be simulated for our reality to be all in a sim. They get into quantum mechanics and all the simulations of chemical interactions, etc. And the vast universe, it’s just so big! But none of that matters. Once you are simulating conscious minds, all you need to actively simulate is the inputs to the minds. You can put whatever you want there. All you have to simulate is what those minds actually observe, which is a negligible portion of the universe.
2) They don’t buy the logic leap that if many simulations can be created, then we are probably in one.
2 is what Thomas seemed stuck on. However, it isn’t some kind of logical fallacy, it’s an extrapolation just like almost all of science is. For example, how do we know we are in a galaxy, or what our galaxy looks like? Well, we don’t. We can’t, because we’re in it. But we can look around us at other galaxies and conclude ours probably looks like that. Or why do we suspect there is other life out there? We don’t have any evidence for it, all we have is logical leaps from the fact that we are life and the laws of physics *probably* work the same way everywhere else and there are trillions of chances for life to arise out there.
Anyway, this is pretty much the same kind of leap, the hard part to wrap your brain around is that you’re applying it to your own conscious experience. Which *feels* different but it isn’t. If trillions of simulations of entire civilizations becomes possible, then what reasoning could you use to say that your reality is different from the reality in the simulation? By design (lets say) it isn’t different.
I’m trying to think of a different argument that has a similar structure but I can’t come up with a perfect one. Maybe it would help illuminate whether the simulation argument is sound or not. Some people (including Thomas in the cast) have tried to come up with some (or just asserted they exist, i.e. Thomas’ “scenarios 4, 5, 6”), but they aren’t very convincing. They miss some crucial point.Thomas never outlined a single alternative to scenarios 1-3 even though he said he could imagine many of them. It might help if someone could suggest alternatives.