I’m not sure I understand, Bryan. Simulations are remarkably useful. If we think we understand the principles behind an event, we might build/program a simulation in order to see what happens. Star birth and death are often studied this way. If our simulation creates a similar object, we feel more confident in our theory.
Gaming, as I mentioned earlier, is another area where simulation has taken hold. Its similarity to actual events is small, but not nonexistent.
Boredom? Curiosity?
Our descendents will probably be simulations themselves, after all. Personally, I’d like to simulate a small planet and see what alternate evolutionary paths there might be. Plus, I’ve always fancied I’d make a great demigod…
Hm. I think not. Since there are only a few particle accelerators in the world, our Overlords only have to feed in the results from previously completed experiments to these virtual sites. Furthermore, if sufficient numbers of sub-modules (that is what they call us) do something outside the original program parameters, they can always slow down the clock speed until the Tech Department gets the proper patches in.
Incidentally, Bostrom explicitly notes that the simulation need not be comprehensive:
Again, however, my problem is it’s not clear that a simulation of 6 billion interacting consciousnesses in one large virtual system would be an economical way of answering any serious questions. Or providing entertainment for that matter. (Smaller simulations are a separate issue. Furthermore, as always, we cannot rule out solipsism.)
Then there is morality. Simulating warfare, for example, with actual conscious entities seems unethical and would likely be subject to challenge: I doubt whether it could secure approval from The Committee on the Use of Conscious Entities as Experimental Subjects.
Of course such a committee may not exist. Yet I would argue that in a state of the world where such activities are not effectively regulated, Option 1 would obtain: a species that cannot regulate complex evolutionary simulations would find itself extinguished centuries earlier by the relentlessly declining costs of weapons of mass destruction, continuing environmental degradation, and other problems requiring social coordination and regulation.
In other words, one small side effect of humanity’s survival would be that Option 3 does not obtain.
Hi all,
I agree with Bryan Elkers’ point. What could possibly be the motivating factor in creating such a simulation? By definition a posthuman civilisation would have collective knowledge far greater than ours so how could they possibly benefit from the experience?
Also, I agree with TVAA’s earlier comment with regard to the distinction between living in a simulation and living in real life. If we expand our definition of the simulation to include the universe and space-time, then it could equally be argued that everything we know as existence is a ‘simulation’ created by by some extradimensional creative force. Oh, hang on - this is starting to sound like religion…
Dave T.
You’re Hilary Putnam, aren’t you? C’mon, admit it!
I suspect you may have read him, since he is the one who most famously formulated his refutation of skepticism as addressing the notion that we are all in a computer simulation, rather than just “I am living in a computer simulation” like most people would phrase it. IIRC, it had something to do with a theory from his younger years about words and meaning (the famous Twin Earth Hypothesis about “water”), that words became connected to the thing they refer to by (basically) some guy pointing at the lake and saying “water”. So if NOBODY had ever encountered an instance of “water”, the word couldn’t mean anything. Thus “brain” and “vat” wouldn’t mean anything either if NOBODY had ever encountered them, so you couldn’t meaningfully say “everybody that ever lived has always been a brain in a vat.” And for some reason that only a postmodernist could fathom, establishing that conclusion was necessary to establish the rest of his balderdash antirealist musings.
This is why contemporary philosophy is stupid. Don’t even get me started on this “possible worlds” crap 
Well, this “simulation” idea is just a slightly-technologically-plausible variant on the idea that the universe is the long-running dream of a Hindu God, with a similarly unprovable premise.
Occam’s Razor being what it is, I’ll operate on the assumption that the universe is real until given extraordinary evidence to the contrary.
Excellent site, erl! I feel I should read the whole lot rather than just Bostrom’s paper to do it justice. For what it’s worth, my initial take on the idea:
Firstly, one should not assume that intelligent life could have come about at any time in the roughly 12 billion year life of the universe and yet, mysteriously, hasn’t lasted or left a visible mark. We are not too far past the point, give or take the odd billion years, when the probability of life-bearing planet formation became statistically significant as far as we know (and even that is a lot of guesswork!). For example, a planet so rich in such “exotic” elements (nucleosynthetically speaking) as silicon, zinc and iodine would almost certainly require the lifetimes of at least two stars before the planet-bearing sun was borne from the remains of their accompanying supernovae.
For this reason I believe premise (1) in the paper, that most civilisations must become extinct or else there would be some sign of them, need not be true even if the other premises are false. Given the evidence, it is entirely possible that we are genuinely the first civilisation to reach our stage of development.
(Of course, any simulator worth its salt would provide a convincing reason to believe that we were genuinely first-born. However, one could second-guess oneself at every turn in this regard thus forcing the scenario back to the Descartes’ Deceiver default, and even then we could be confident that some civilisation had to be the first one which lasted, and that there is no reason it should not have been ours.)
Secondly, I question the “value” a future society would place in these retroactive creations. Granted, their educational benefit would be enormous, as would the entertainment value associated with “immersing” oneself in the early 21st Century for an afternoon. (Note to person sharing my brain as I write this: Piss off back to your own bland, painless little life you smug, patronising voyeur!)
However, the morality of the program would, I hope, be examined by the finest minds of the time and found to be sorely lacking (as flowbark pointed out). Assuming that the suffering experienced by a simulation is every bit as real and undesirable as that experienced by the future “real” people, then suffering is being deliberately created on a massive scale, when all that is needed to end it is the flick of a switch, or a new and less painful life loaded from disc (or whatever). We would not make a five year old child climb chimneys merely to simulate the life of a child in Victorian England for our own edification. It would seem to me quite odd, although of course not impossible, if we were to make leaps and bounds in technology but become crippled in terms of our empathy and desire to see unnecessary suffering minimised.
(Incidentally: “No” they might counter, “those people you think have been created merely to suffer are illusory. The only lives we create are happy, fulfilling lives with a minimum of pain and suffering; the others are mindless, programmed puppets. Happy lives rather like yours, in fact!” But this is merely to tread further along the path of solipsism, and in any case even in my life there have been times when I wanted a pain-free disc to be loaded so much that I might physically attack the person who chose not to, should I ever meet them.)
Anyway, there’s my quick critique, but I’ll be thinking about this for a while! For the moment, excuse me…
LOAD simulation 452.8: “Zeus plays for Liverpool”
Ah, but we might as well be.
For any system X of interactions, there are an infinite number of possible “deeper” systems that could emulate it. They’re all equally “real” (and “unreal”) as far as X is concerned.
It would be possible for the sim-builders to incorporate fudged data into the simulation to cover the limitations of the system when we run particle accelerators.
However there are more ways than that of taking advantage of quantum proterties. These days a lot of semiconductor technology relies heavily on quantum mechanical effects. In order for the simulation not to have to simulate down to a quantum granularity they would have to be inputting fudged data all over the place to accommodate our use of solid state electronics. Even if they’re managing to simulate at these levels without a complete implementation of quantum mechanics (not impossible - solid state electronics don’t really probe the quantum level that deeply)
It’s true there are ways to bring the computational overhead of the simulation down (having areas like the inside of the Earth simulated only approximately etc) but we’re still looking at a computer that would be physically huge and consume an immense amount of energy and resources. I would suggest that this means that nobody would be running a simulation for idle curiosity. Any posthuman society that places any value on resources would have difficulty in building a simulation.
Of course, these aren’t necessarily arguments that we’re not living in a simulation. They do allow us to sketch out the shape of the people who could build such a simulation though. They’d have to be able to build computers at least approaching the size of a planet or be posessed of a radically different and more efficient method of computation. They would have to place little or no value on resources such as matter and energy (particularly energy) to be able to afford to run the system.
These conclusions then raise another question: If they can build a computer sufficient to simulate the Earth for us, why bother at all? The sim-builders probably live in a perfectly good universe already - why not just genetically engineer a few thousand humans, dump them on a terraformed planet and watch the action unfold from your majestic starships? The ability to simulate obviates the need to do so - you’ll already be powerful enough to create engineered environments using the raw materials in your own universe with all the advantages that brings (it’ll run in real-time, won’t comsume as much energy and you won’t have to keep getting the tech-support guys out of bed at 3am when the whole thing crashes becasue the guys at CERN just did something unexpected).
Of course, this still leaves the possibility that the simulation is being run to explore a different kind of universe. In this case we’re left with the conclusion that the sim-builders aren’t posthumans at all, but are a truly alien race exploring the consequences of different types of universe. This may go some way towards explaining why it would exist at all with the apparent ethical objections to running such a simulation containing sentient beings. They would not necessarily share our ethics.
Armilla, and others, I’m just a layman here and completely out of my field, but I’ve noticed a recurring theme throughout many of these posts.
Many of you seem to be positing that whatever mechanism is running the simulation that is this universe is beholden to the parameters of the physics defined within this program.
But a multiplayer game of Return to Castle Wolfenstein is perfectly capable of being created here on earth thousands of times a day, sometimes playing out scenarios entirely between artificial players. It does not really accurately portray the way the universe outside of the game actually works, and it uses an extremely rudimentary approximation of Newtonian physics.
Is it not possible to imagine a superreality which is infinitely more complex than this one, with infinitely larger computing resources, running on a completely different set of rules which can easily accomodate the subset which is our reality?
If I may be so bold as to answer for Armilla: Yes.
If the simulation is so good that we can’t tell it is a simulation then what could it be a simulation of?
This is circular pseudointellectual crap.
Dal Timgar
TVAA calls it right. That’s certainly a possibility we must entertain. I’m tempted to argue that it may even be a necessity.
If the universe were a simulation, the entities in charge of it may not be able to ascertain that we are sentient (the old Turing thing).
Mangetout, good point. One that I would have taken paragraphs to make. So thanks for the brevity. More than just the soul of wit. Many a night I’ve spent discussing how we could know if the characters we play in video games are conscious. What’s funny is that if the character stopped responding intuitively to the controls, we blame the controller or reset the game. :eek:
Armilla, this is why I’m not quite sure if the limitation imposed upon us by accepting the paper you linked get us any closer to ruling out three. It certainly does indicate a better grasp of a limit on “practical similarity” in terms of us making a simulation.
Another possibility would be that the owners of the simulation could be simply unaware of our existence, or unaware of the existence of organic life at all - if the model is all about the collision of galaxies, they might need it to be realistic in terms of the stuff the galaxies are made of, but unless they peered really closely (which they might not bother doing), they might totally overlook the insignificant chemical interactions that occur on the surface of some of the planets.
Yet another possibility is that the calculations that describe the simulation in whcih our universe resides are entirely incidental to some other process being simulated - that everything we see is the result of mere sloppy rounding.
The pivotal argument that Bostrom uses is that the simulations are simulating the ancestors of the simulation builders. In this way he can draw inferences about the number of “real” humans and the number of simulated humans.
Bostrom’s calculations do not apply in any way to simulations which are not created by humans or which do not simulate the progression of the simulation buiders’ ancestors. The probability of us existing in a simulation run by inscruitable alien beings cannot be addressed by Bostrom’s work.
Bostrom’s assertions that the simulation need not be detailed are misleading. It does not address the issue that he relies on the number of simulated ancestors being roughly the same as the number of real ancestors, implying that the simulated humans will approach the level of technology of the simulation builders.
An accurate simulation of all ancestors up to the point at which the simulation machine is built would necessarily include the simulated humans discovering and beginning to use the technology which is used to build the simulation hardware. If this is not the case then it is not an ancestor simulation and no inferences can be drawn on relative population sizes.
I see two possibilities for the structure of the simulation:
(1)
The simulation is not complete. Some level of the physical structure of the universe is not fully implemented. In this case there will come a point in the ancestor simulation at which it will break down and diverge from reality. This is the point at which the simulated humans would discover and begin to use the principles on which the simulation machine works. Therefore some proportion of the ancestors are either not simulated, or the simulation diverges from reality so that no inferences can be drawn on relative populations over that interval.
(2)
The simulation is complete, allowing the simulated ancestors to complete all technological steps up to and including building their own simulation machine. In this case the simulation hardware is subject to Wolpert’s theories and would cosume a prodigious amount of resources and run slower than real time.
Case (1) will tend to create a disparity between the number of real ancestors and the number of simulated ancestors. Specifically it means there would be less simulated ancestors than real ones, thus causing Bostrom’s equation to predict a much lower chance of us currently being in a simulation. The actual disparity is unknown and, at the moment, unknowable. Depending on the development path of technology the number of ancestors not simulated may be much larger than the number that can be simulated leading to a near zero probability of us being inside a simulation.
Case (2) implies that the simulation is a) costly and b) slower than real time. This limits the number of simulations that could possibly be run, again pushing Bostrom’s probabilities down.
Essentially, as the efficiency of the simulations goes up so does the level of “fudged” reality. Many such fudged simulations could be run, but would only have value in simulating a subset of all ancestors. As the efficiency goes down and the simulation detail approaches that of reality the set of ancestors that can be simulated goes up but the number of simulations that can be run goes down.
This introduces enough unknowns into Bostrom’s equations that I don’t believe we can draw any conclusions from his line of reasoning.
Armilla, great stuff. However,
I’m ambivalent about this point. On one hand, I believe you say this because you’re saying that in order for the population of simulated beings to so greatly outnumber the number of real beings, the simulations’ technology would have to be great enough to enable such an amount of life (divided by the number of simulations running). That sounds right. On the other hand, the less complex the world (technology-wise), the more times the same simulation could be implimented. So the question we face here is whether or not the technology that enables higher populations (and the computational power necessary to accept the increased populations) is a significant task when compared to simply calculating out the world around the humans—say, a humanless Earth. Is there a significant overhead? Or is the computational power most necessary for the laws of physics themselves, with any additional fudging for humans being almost irrelevant?
After the paper you’ve presented, and the comments you’ve made, I am inclined to agree.
Great post Armilla.
Regarding option (1), I can imagine a situation where our Overlords have a partial simulation that cannot be penetrated by 21st century thought (in nonspeculative ways).
However, I agree that if the Overlords are interested in tracing their entire technological evolution, they have a problem: I would think that there would have to be some point in the timestream where a substantial fraction of simulations become self-aware in a substantive sense: in your words they, "discover and begin to use the principles on which the simulation machine works. "
At that point patches would no longer work and we are left with option 2. (My only point here is this point of discovery could conceivably be far in the future.)
Anyway, that was a terrific taxonomy that you set up.
ERIS:
---- Many a night I’ve spent discussing how we could know if the characters we play in video games are conscious. What’s funny is that if the character stopped responding intuitively to the controls, we blame the controller or reset the game.
Yes. That is why it is safest to be extremely predictable.