Imagine if you not only realized you were living in a simulation but that the simulation involved randomness (god playing dice) and you could somehow force backtracking with a simple button so that different values were chosen. Now, play the lottery and hit that button whenever you don't win; you could then surface an unlikely universe where you get really lucky.
Have you heard of Max Tegmark's quantum suicide argument? If you believe it sufficiently, well, buy a lottery ticket and cyanide, and really commit yourself to it :)
Let us for a moment, assume that the quantum suicide argument is correct. It is probably more likely that you will somehow survive the cyanide with life-long debilitating effects than for you to actually win the lottery.
Hah, that's very cool. I came up with this as an idea for a short story a couple of days ago (except slightly more drastically, as it involved the complete annihilation of humanity on a failure condition, allowing the practitioner to drag everyone else with him into his preferred outcome-world). I'm simultaneously pleased and displeased that someone's already given it a lot of thought.
It's unlikely to work unless the button is so reliable that you have a better chance of winning the lottery than the button/mechanism failing in some (possibly undetectable) way. It's the same argument against trying to use quantum immortality to force a miracle by killing yourself, or constructing a machine to kill you, unless some desirable event occurs. It's much more likely that your machine will fail, or you'll change your mind.
We could imagine that the simulation naturally backtracks given a paradox or dead end. So if the lottery is rigged, I would never get rich, and the simulation would just backtrack to where I would never decide to try this.
There is a theory which states that if ever anyone discovers exactly what the Universe is for and why it is here, it will instantly disappear and be replaced by something even more bizarre and inexplicable.
There is another theory which states that this has already happened.
And if it has? Well, then the statistical likelihood is that we're located somewhere in that chain of simulations within simulations. The alternative - that we're the first civilisation, in the first universe - is virtually (no pun intended) absurd.
wrong. there are multiple alternatives; we are the nth civilization in a real universe, we are the nth to the nth civilization in the multiverse, etc. Sloppy sensationalism and bad statistical reasoning in this article.
I've always felt there's something deeply fallacious -- or hopelessly wrapped up in human psychology -- about taking a representation of a virtual world that we might create and treating it as on par with the actual universe.
It's almost like saying "we're probably characters in a novel," since for every universe there are many novels. Or like that old proof that God exists because he has every desirable quality, including existence -- as if a hypothetical entity could be forced into being by sheer burden of how it is described.
When we talk and reason, we call something a universe to invoke all the general properties our universe has -- except existence, of course, because it would be silly if we could only talk about things that exist. Various things like books and simulations are physical representations of universes that don't exist. By virtue of our interpretation, they are universes nonetheless. When reading a book, we fill in assumptions and details from real life if the author gives us no reason to think otherwise. Tautologically, the world of the book is different from the real world in some limited and structured way, but not in any way that keeps it from being a world at all.
Physics seeks to explain the world exhaustively and reductionistically in terms of mechanisms, models, and mathematics -- basically, to boil it all down to the consequences that emerge from laws and equations that we can completely conceptualize. If the universe consisted just of electromagnetism, for example, then Maxwell's equations would be a "complete" description of the universe, and any simulation of them could be considered a universe. Everything about the universe, every general description that's true of it without reference to specific places, times, and things, would be reflected in the simulated universe as well, because science allows us to subsume it all into the more fundamental description given by the equations.
The problem is that a model of the universe is stil a model. Humans invented the notion that a thing and a description of a thing are equivalent.
If you want to convince anyone, you have to do better than just referring to your feelings. Plato felt the world of ideas was more real than the "real" world. Cosmologist Max Tegmark argues that we have no reason not to believe that the universe is anything more than pure math, and all formal systems exist in precisely the same way that our universe does. Philosopher David Lewis argues that the actual world is nothing but the possible world that we happen to be in. Representationalists believe that any simulation that duplicates the causal structure of conscious beings would have conscious beings in it. (A novel does not duplicate such a causal structure, so that's one place where your argument falls flat on its face, if indeed you were trying to make an argument.) Searle with his Chinese Box argument is not buying it.
The bottom line is that we just don't know, and are unlikely ever to know, but if one wants to argue intelligently about it, there's a huge amount of philosophical literature on just this topic.
On the other hand, your position would seem to deny human rights to future intelligent robots, which has serious ethical ramifications should you happen to be wrong.
There is a similar argument against AI. That a computer can't "feel" or experience any of the sensations we do. If you type 'emotional_state = "happy"', does that mean your computer feels happiness? Even if you have a complex simulation of neurons, if you zoom in, all you will see is little electrons bouncing through various logic gates and ending up in some kind of pattern.
But the thing is, the exact same thing is true for your brain too! If you zoom in on your brain you will find nothing more than electrical impulses and neurons that strengthen or weaken their connection to other neurons. If the neurons were suddenly replaced with little computers that performed exactly the same, you wouldn't feel any different. You would still experience emotions and feel sensations and have no idea any parts had been swapped under the hood, unless you looked.
I don't know why but this idea is confusing to me. This XKCD http://xkcd.com/505/ is a good example. My consciousness could be a bunch of static rocks and a man placing new rows based on simple rules. I find this thought disturbing, but there isn't anything fundamentally different between that, a computer simulation, and a world of a bunch of atoms hitting each other, following simple physical laws.
Different beings could exist in all three universes. And each would feel emotions and sensations and all that. And each would argue about how beings just like them in the other universes "don't really exist" like they do.
So how is this different from a character in a novel or a variable in a computer set to "happy"? I honestly don't know. The character in a book is completely static. Whatever is written in the pages, it isn't changing, it isn't taking input and producing output. But the rows of rocks are static too after all. They don't take any input from the outside world either. They change over a dimension of space, each new row one foot to the right of the last. As do the characters in a book. Each paragraph moves the characters forward through their own dimension of time. Maybe the process that creates the rows of rocks, the man following simple rules, is what makes it conscious. But the book is created by a writer following a process too, does that make it conscious?
The only thing I can think of is that the character in the book is far simpler than a human. Words that say "john is happy" is similar to setting a variable on your computer equal to "happy". Happiness in a human is far more complex, involving tons of interactions between neurons. But this answer isn't satisfactory to me. Complexity isn't what makes something intelligent or conscious after all.
> The only thing I can think of is that the character in the book is far simpler than a human. Words that say "john is happy" is similar to setting a variable on your computer equal to "happy". Happiness in a human is far more complex, involving tons of interactions between neurons. But this answer isn't satisfactory to me. Complexity isn't what makes something intelligent or conscious after all.
This is the difference between a running program and a short description of the program. The short description isn't runnable, and isn't running. The reason the program itself is doing what we want it to do is not because it's more complex than the short description (though it is), but because it's actually executing the algorithm. We could imagine more and more detailed descriptions of the program, specifying more and more closely how it works, and eventually it's so detailed that you could run the program from that description. At that point, the description is still not a running program, unless and until you run it. Being conscious is a process -- a running algorithm -- not a static description.
Well I'm not really claiming the book is sentient, just trying to come up with a criteria that excludes books and everything else except for people, without being completely arbitrary.
Anyways you can think of time as a dimension. Each moment of time exists, connected to the ones that came before and come after it. You can represent the every state that a turing machine goes through all at once. Just like those rows of rocks. What does "running" even mean in a universe where all points in time can exist at once?
But regardless of that, you can still say the book is "running" when it is being written or when it is being read. The characters don't just exist in the book, but literally in your mind. Though I still wouldn't call them conscious or even intelligent.
> What does "running" even mean in a universe where all points in time can exist at once?
If they all exist at once, then in what sense are they connected? Certainly not a causal sense, which I would argue matters quite a lot.
> The characters don't just exist in the book, but literally in your mind. Though I still wouldn't call them conscious or even intelligent.
If your mind was able to simulate the actual process of consciousness in the brain of the character being read about, then I would argue that the character could be conscious. Of course, brains don't have that much additional processing power. :)
"But the thing is, the exact same thing is true for your brain too!"
That is pure speculation. Just because we can observe similarities between two phenomenons it doesn't mean they are the same thing. You're erroneously assuming that the universe can only work in a way that fits our theories and models.
What are you suggesting? That the brain fundamentally is not a machine, a process of some kind? Even if everything we know about the universe was wrong, which could be true, but it is far, far, far from pure speculation. You might as well say it's pure speculation that gravity exists.
This shows that you can't just say "complexity makes something intelligent and conscious" by itself, you have to explain how it works, and "complexity" isn't an explanation, it's just a label. I agree.
But that's equally true of saying "complexity doesn't make something intelligent and conscious". All right, then what does? "Not complex" is a label just as much as "complex" is. Either way you haven't explained anything. That's what I was trying to get at.
Also, saying "complexity doesn't make something intelligent/conscious" is not the same as saying "something intelligent/conscious doesn't have to be complex", which is what I think the statement I originally responded to, taken in context, was really intended to mean. Do you really think intelligence and consciousness can be present in simple things like rocks? Might there not be a reason that the human brain is three pounds of complex matter, not three pounds of jello?
Yes, you are right, it is likely that anything worthy of being considered conscious would be fairly complex. (Though I believe it is possible to create an intelligence that is initially fairly simple. AIXI for example is the ultimate intelligence and could be created in a few dozen lines of code. It would just take nearly infinite computing power to do anything interesting and I wouldn't really consider it "conscious". Even the processes going on in your brain, at least the very basic function that makes us "intelligent", stripping away all the uninteresting stuff or additions to that, could probably be specified in a very small space. A lot of neurons may seem complicated, but you only need to understand the programming of a few, than just copy them a billion times. But all of this is beside the point.)
Anyways in the original context of where I said that, I was trying to come up with a satisfactory way to distinguish consciousness from non-consciousness. Complexity obviously isn't the way since lots of complex things are not conscious. It may be that all conscious things are also complex. But then complexity isn't the reason it's conscious, it just happens to correlate with it.
Setting a variable in a computer equal to "happy" obviously doesn't make the computer experience happiness. But then what possible sequence of commands and state changes would? If you accept that there is no possible sequence that would, then you have to accept that humans can not experience consciousness either. Because for all we know we are running in a computer too. Even if we are not, the process our brain follows easily could. The fact that it runs on atoms bouncing into each other and not electrons flowing through logic gates makes no difference.
AIXI for example is the ultimate intelligence and could be created in a few dozen lines of code.
Really? Show your work, please.
A lot of neurons may seem complicated, but you only need to understand the programming of a few, than just copy them a billion times.
This assumes that they are all "programmed" the same. Why do you think that must be the case? Also, you're leaving out all the chemical processes that contribute to brain function.
I was trying to come up with a satisfactory way to distinguish consciousness from non-consciousness. Complexity obviously isn't the way since lots of complex things are not conscious.
True.
It may be that all conscious things are also complex. But then complexity isn't the reason it's conscious, it just happens to correlate with it.
The connection might well be stronger than "just happens to correlate". Even if "complexity" by itself isn't the reason it's conscious, it might well be that consciousness requires a complex substrate.
Setting a variable in a computer equal to "happy" obviously doesn't make the computer experience happiness. But then what possible sequence of commands and state changes would?
Um, a much more complex sequence of commands and state changes?
If you accept that there is no possible sequence that would
I don't. I just think there is no simple sequence that would.
I really do believe that consciousness doesn't really have anything to do with intelligence. Even if humans are not simple. It's conceivable you could create something like a human with very simple machinery, just massively scaled up.
If that is too complex, you could create an even simpler algorithm which produces that. Maybe by creating a genetic algorithm which "evolves" human-like intelligence after billions of generations.
In both cases the end result may be complex, but the algorithm that creates it is not. Whether it be a network of billions of interconnected neurons that arise from simple programming of a few neurons, or a complex intelligence optimized by simple random mutation and selection.
This is the idea of emergence. That complex seeming behavior can "emerge" from simple rules and a simple process.
I really do believe that consciousness doesn't really have anything to do with intelligence.
The problem is that both of these terms are really too vague. What do you mean by "consciousness"? Some people claim insects are conscious; some claim bacteria are conscious; some even claim electrons are conscious.
Also, what do you mean by "intelligence"? By some definitions, the process of evolution by natural selection is "intelligent".
If by these terms, you mean "the properties humans have that we call consciousness and intelligence", then why do you think there is no connection between them? Does your own intelligence really have nothing to do with your consciousness? Or vice versa?
It's conceivable you could create something like a human with very simple machinery, just massively scaled up.
Conceivable, yes. Likely, I don't think so. But that's really a matter of opinion; it depends on how much of the complexity in the parts of the human brain (neurons, chemicals, etc.) is necessary for human consciousness or intelligence, and how much you think is just an artifact of the implementation, so to speak. We don't know enough about how the brain works yet to really judge.
the end result may be complex, but the algorithm that creates it is not.
In other words, the complexity is still there.
That complex seeming behavior can "emerge" from simple rules and a simple process.
Wait a minute--complex seeming? Or complex?
If you're trying to argue that the complexity may not be in the algorithm itself, but only in the actual working out of the algorithm in time and space, I'll buy that.
But if you're trying to argue that it isn't "really" complex because the algorithm is simple, I don't buy that. I never said the algorithm had to be complex; I just said there has to be complexity somewhere for there to be (human-level, to avoid all the definitional issues I raised above) consciousness and intelligence. That's perfectly consistent with the complexity ultimately being built out of simple parts--just a lot of them with a lot of interactions, so the complexity is in the interactions, not in the parts themselves.
(I think it's likely that the parts themselves have significant complexity too, as I said above, but even if that's true, there will still be some lower level, possibly much lower, where there are simple parts, just a huge, huge number of them with a lot of interactions.)
That was actually a typo. I meant to type "consciousness doesn't really have anything to do with complexity." In the sense that you could probably define an intelligent being like a human in very little space.
>In other words, the complexity is still there.
Well yes in a way. By the definitions used in information theory, complexity is just the amount of bits you need to accurately describe something in some language. By common usage complexity just means the amount of concepts you need to learn to understand something (about the same thing.) But it can also mean the number of moving parts a machine has, so to speak, which could be very large while simple enough to understand or write on paper. Like a computer display which has a few thousand pixels each of which is almost exactly the same.
I don't know if you would call a computer display complicated, at least not much moreso than the invidual pixels that make it up. The same is probably true of humans. Even if humans are fairly complicated, someday a connectionist approach or maybe a rough approximation of a human brain might succeed at creating something arguably conscious.
>If you're trying to argue that the complexity may not be in the algorithm itself, but only in the actual working out of the algorithm in time and space, I'll buy that.
This may be a much better way of describing what I just said above. So yeah.
There is definitely a difference between a description and a complete description, and, as another commenter pointed out, between a description and a simulation that embodies that description.
Consciousness is a property that we ascribe to ourselves and other entities. We don't experience it in others the way we experience it in ourselves, of course, but we infer it from the data available from interacting with them. They produce a pattern of signals that resonates deeply enough with us that we recognize them as like us in having goals, feelings, and access to the "human experience." For example, one of the most convincing things we could observe an AI do would be to make a novel observation about life, derived from current circumstances, that strikes us as deep and insightful, perhaps because it was hovering at the edge of our own awareness, or because it immediately activates mental patterns that we did not activate but are also meaningfully connected to the situation at hand for us.
It makes sense that if we were to copy the workings of the brain in sufficient detail in a machine, the machine would also be ascribed consciousness by human users. Perhaps we could say that it is an "artificial brain" in a literal sense, performing normal brain functions the way an artificial heart pumps blood.
In the case of the universe, it's at issue whether there is such a thing as a "complete description" of the universe, whether a simulation could be done, and then whether this simulation actually is a universe in some sense. Unlike brains, we don't regularly encounter other universes (actual, physical ones), ones which we don't live in and experience from the inside, but which we are forced to conclude are true manifestations that correspond in every important detail to our own nevertheless. A complete description of the universe by a future physicist (say) would start with a short list of all the types of information that go into describing a particular universe, like a list of type of particles and variables such as position and momentum. This list describes universes in general. Then we'd have to describe a particular universe, with a long, long, long list of all the individual particles and so on. Then we have to represent this information physically in some sort of computer, for the purpose of simulation, and perform the ongoing processing involved in running the simulation.
Finally, we must argue that the simulation is a universe. Not just deserving of being called one, or having the properties typically associated with one, but a true instance of some class that previously had the universe we live in as its single member.
One interesting feature of this line of thought is its seemingly infinite ambition. What about smaller goals we could set for mankind -- might we ever engineer an artificial star? How about an artificial galaxy? We can try to beef up the likelihood of this ever happening by considering not just mankind but any other "intelligent beings" the universe might happen to contain, in the past, present, and future. However, we are again on the shaky ground of inventing a class from a singleton. All extraterrestrial "intelligent beings" ever conceived are just stories with small, structured deviations from the human template; beings we identify with because they, too, have intentions and goals and act in their ultimate self-interest for the purpose of self-preservation and just trying to make a go of it in this big ol' universe. Some of them, it's presumed, make detailed copies of natural things at an astronomical scale and then talk about whether the old thing and the new thing are members of the same class of thing or not.
I agree that it's irrelevant to us and our operation if our brains are made of billiard balls or quantum clouds, and the same seems true of the universe.
It is still interesting whether the universe is "computable" or not, that is, whether something meeting the current definition of a computer could simulate it, even in theory. ...
>For example, one of the most convincing things we could observe an AI do would be to make a novel observation about life, derived from current circumstances, that strikes us as deep and insightful, perhaps because it was hovering at the edge of our own awareness, or because it immediately activates mental patterns that we did not activate but are also meaningfully connected to the situation at hand for us.
I think you've come much closer than I to getting a definition of consciousness, but I still don't find it satisfactory. You could have a computer mimic a human without actually having anything resembling an inner "human experience". For example, a giant look up table with a preprogrammed response to any possible input. We have created extremely simple versions of this that are almost effective enough to pass the turing test. Just by gather a large number of human responses to chatbots. Though you could say the process that creates the giant look up table is intelligent, the table itself obviously is not.
There are AIs like AIXI which could be vastly intelligent without reasonably being considered conscious. They don't actually "think" so much as brute-force every possible solution to a given problem, or every possible explanation for a series of inputs, etc. Given enough computing power this would actually be effective. You could then ask it to mimic a conscious being and it could do so easily (given either a definition of consciousness or a set of observations about other conscious beings.) Maybe doing that requires the intelligence to actually simulate a conscious process somewhere along the line, so I don't know if that counts. It may be able to do so through mere, without having to actually run a full simulation. Providing a "deep insight" about life is nowhere near as complicated as a full human consciousness.
Maybe there is no such thing as consciousness. In the sense that there is no simple way to define it that is fully consistent with everything we want it to mean. That there are always going to be arbitrary seeming exceptions and gray areas. But that has deep and disturbing implications for morality. Maybe it's not relevant to our day to day life, but if we ever want to program a friendly artificial intelligence, we are screwed.
>I think it's silly or meaningless to say we possibly -- or probably! -- live in an "artificial" universe.
Well assuming that there are more simulated beings in all of existence, whatever that means, than there are non-simulated ones, it is very likely we are simulations. Though it's impossible to know what exists beyond our own universe. What is the distribution of universes? Do 50% of universes allow for infinite computing power, for example? What programming language do they use, which would determine how many bits a given universe would take to specify in that language, and therefore how often it gets simulated? How many universes even have anything close to intelligent beings?
An infinitely powerful computer would allow you to simulate every possible universe all at once, instantly (you would also be able to simulate every possible branch of non-deterministic universes, continuous variables, time travelling, and other things that people claim makes the universe "uncomputable.) And you could instantly do it an infinite number of times. If it were possible for such a thing to exist in a universe, that universe would contain every other possible universe. Including copies of itself and others universes with infinite computing power, which would contain more universes, etc.
So what is the top level universe? Why does something exist rather than nothing? Perhaps everything that can exist does exist. Maybe every computable program has been (is being?) run, including one that specifies our universe. That's the simplest possible explanation that explains the existence of the universe. Though I ...
Consciousness and intelligence: You're right that proving something is "conscious" does seem difficult or impossible. Making something that is merely "intelligent," on the other hand, may be easy, depending on how you define the word. I think that's because consciousness is something we experience in ourselves, looking inwards, marveling at what it means that we are a mind living in a body, while intelligence is something we ascribe to other things in varying degrees if they exhibit certain behaviors. It follows that for us to consider something else to be conscious, we have to have that same feeling as when we experience our own consciousness, but via identification with something else (a robot or AI). We have to be like, "Wow, consciousness," and then be like, "Oh, that's not me, that's him!" It's the same way for intelligence, but it's a much broader term because there are many kinds of intelligence. Animals exhibit a lot of intelligence, as do some computer systems. In both cases (consciousness and intelligence), I think making systems that we recognize as more intelligent and more conscious involves understanding the brain and replicating its various functionality (since it's brains that will be making the call; it's brains that encode the distinction; the rest of the universe doesn't care). As we learn more about consciousness, we may be able to identify it as a particular kind of intelligence. For example, it may basically be the small part of our brain that directs our attention and awareness from moment to moment and decides what computational tasks to take on, while the rest of the brain is basically a large set of elaborate coprocessors specialized for different types of computation.
Simulated universes: I still think this is all meaningless metaphor. It's not easy to articulate why, but I'll try.
Some lines of thought just combine ideas in deep ways, but don't actually tell you anything new. For example, Zeno's paradox -- to get from point A to point B, you have to first go halfway, then halfway again, and so on forever -- doesn't actually teach us that nothing can ever go from point A to point B. It just demonstrates a glitch in the metaphors we use to reason about time and space. Actually, it's a glitch in the metaphor of traversing space as achieving a goal. Normally, if reaching a particular goal involves an infinite to-do list of subgoals, we wouldn't expect to ever get to the end and actually achieve the goal. However, if the goal is traversing some infinitely subdividable interval of space, we are provided with a mechanism to generate an infinite list of subgoals that are presumably all involved in achieving the larger goal. It's all just mental gymnastics and the normal tools of intuition breaking down.
To take a closer example, consider this line of thinking: "If you pick a random marriage proposal, chances are it is an imagined one rather than an actual one, because for every actual marriage proposal there are on average two or more imagined ones that never came to pass." This strange kind of thinking also mashes together diverse intuitive concepts. There's the idea that every class of thing (like "marriage proposal") has an extension set, an ensemble of instances. There's the fiction that an "event" is a discrete thing that we could identify and count if we had to, like counting the number of thoughts I've had today ("a thought" is a singular noun after all). There's the concept from probability theory of picking a member at random from a set (with the presumption that this is a well-defined act). There's the idea that an imagined X is still an X, because we still call it one. Unicorns are still unicorns even if they don't happen to exist.
The most intellectual, abstract part of the brain has a tendency to focus on what could possibly be rather than what is. We can get so wrapped up in generalities that we forget that this is the universe -- what we see around us, not what we imagine or are capable of conceiving of. Statements like, "Perhaps every...
Anyway, your objection here isn't actually relevant as far as I can tell. It's a probabilistic argument about the number of real minds vs. the number of simulated minds, and the bulk of the "proof" is in demonstrating that, under certain assumptions, the number of simulated minds would be so high in comparison to the number of real minds that you'd almost be forced to conclude that you are simulated.
It really doesn't have anything to do with "universes inside universes."
...and the main flaw of this still remains: we don't know the probabilities of a very advanced civilization running a simulation because we can have no idea of whether this would be interesting for them to do, we can't predict what would their goals and motivations be. Out great-great-...-great-grandchildren will not just be super-smart versions of us, they will have concepts and goals we cannot even imagine! A simulation of us, depressing as it may be to think it this way, would be just as interesting for them as an ant-farm to a little boy - he'd get bored of it quickly because he's as far away evolutionary speaking from ants as these super-iteligent beings would be from us, and that wouldn't make for a large number of ant-farms simply because ant-farms will be boring and useless...
my bad, I actually read it some time ago, after someone sent me a "summary" of the reasoning, omitting this variant, but somehow the "summary" was stuck in my head and not the actual paper. indeed, Nick Bostrom actually thought everything through, including this line of thought.
The objection is totally relevant because you cannot calculate probabilities if you don't know the number of real universes, the probability of life arising from non-life, etc. I would argue, using that line of reasoning that it is more likely that we would be a dream of a sentient entity rather than a simulation of one if his probabilistic assumptions are taken as probable.
Dreams only last a few hours, so that's not really a possibility. If you are proposing we are the "dreams" of a sentient and extremely advanced machine, then that is hardly different from saying that we are just simulated.
And I'm pretty sure the numbers you mentioned--like the number of real universes, the probability of life arising from non-life--really don't come into it at all. Skimming over the paper, none of that stuff seems to come up at all. Again, this isn't about universes inside universes (the news article phrased the argument poorly).
Im not talking about universes within universes. Let's take our own universe as an example, and using probability as we know it in our own as holding in whatever 'real' universe(s) that may or may not simulate itself or some subset of itself. If sentient beings evolve (taking Bostroms postulate of that probability as reasonably high) and assuming they 'dream' if they are anything like us, as he does, and since we have (yet) to develop a technology that can simulate a universe as complex as ours, it's far more probable that sentients evolve and dream universes than evolve and simulate us. Even in our own "existence" (whatever that means) you have to agree that there have been far more dreams than games played. This follows probabilistically; not all people play games but almost every human that has ever lived has dreamed. Therefore it is far more probable that we are a dream, or passing thought, for that matter, than an intended simulation. Hinduism win.
A provoking question: if we found evidence that convinced us, beyond a reasonable doubt, that we were living in a computer simulation, would you change your behavior?
You have to live as if you have free will. You can't say, "I've considered all the possibilites, and have decided that there is no free will", because considerations and decisions are acts of free will themselves. Of course, that doesn't mean you have free will.
I find it interesting how people often cite determinism as incompatible with free will. Our minds don't violate the laws of physics, but they still make decisions. Similarly, a car can still be described as "driving", even if the physics behind it are deterministic.
Bostrom's postulates are exceedingly bad. His assumptions rely to much on the isotropy of properties of the one "real" existence and the simulated ones. I would expect, statistically, that there would be far more universes simulated that have properties that are nothing like the "real" one as technology progresses, i.e. there are more instances of "sonic the hedgehogs" and "super mario brothers 1" over the more "realistic" call of duty. It follows that if our universe is one of the more probable universes that our laws of logic and probability estimates are not calculated from probable postulates, thus our assumptions leading to any conclusion are probably in error.
> It follows that if our universe is one of the more probable universes that our laws of logic and probability estimates are not calculated from probable postulates, thus our assumptions leading to any conclusion are probably in error.
What 'postulates' would they be calculated from? Does it even make sense to speak of universes following different 'logics'?
> far more universes simulated that have properties that are nothing like the "real" one as technology progresses, i.e. there are more instances of "sonic the hedgehogs" and "super mario brothers 1"
Those seem like perfectly normal universes to me: they're fully describable by short programs executing over discrete binary states like a chunk of RAM. Where's the violation of logic or metaphysical oddities there?
The real universe is actually continuous rather than discrete at the macro-scale (ie: relativity). In addition, a "video-game universe" runs on a programmatic logic with discrete conditionals rather than our universe's equations that don't branch.
> The real universe is actually continuous rather than discrete at the macro-scale (ie: relativity).
But the 'real universe' appears discrete on the smallest level, and I don't believe physicists will ultimately settle on a theory in which two completely different sorts of theories take over at different regimes and the macro-scale theory is not reducible, even in principle, to derivations and predictions of the micro-scale theory. The quantum theories will eventually predict the relativity-scale results./
> In addition, a "video-game universe" runs on a programmatic logic with discrete conditionals rather than our universe's equations that don't branch.
How would you know? What governs quantum randomness?
What 'postulates' would they be calculated from? Does it even make sense to speak of universes following different 'logics'?
Exactly the point. What game do you know of that follows the laws of logic as we know them in the 'real' universe we habitate? It follows that it's nonsensical that we can reason ourself to reality if we are a simulation.
If we were created like this, wouldn't it be feasible to say that the creators could just patch this up so the experiment will appear to work but is indeed, rigged.
My guess is that there are other civilizations out there, but they're suffering from the exact same problems as us - long distances, huge energy requirements, lots of time. If we can't get out to the stars, I can see why plenty of civilizations would be having trouble.
First off: why would you assume that we could interact with them?
It appears to be phenomenally difficult to get from point to point in the galaxy/universe (see the recent trend in science-fiction to stick to a speed of light limited universe). If we come up with some fantastic energy source (see: ZPE), then I'll say that it's phenomenally slow so that no civilization would invest in doing it. If we come up with some civilization that would do so, I'd point out that it's phenomenally slow/difficult to send signals across universal distances. Rational folks think that non-solar-local interactions are probably not possible... Large brained rational folks think to look for solar-local artifacts of Kardashev-scale>1 civilizations (see: http://www.personal.psu.edu/jtw13/blogs/astrowright/2013/03/...).
Recall that early explorers met Native Americans and called them Indians because they had misinterpreted distances, didn't know all of the continents, etc. We have a history of overcoming increasingly large distances with improved transport. This does not mean that we shall always do so.
It may be depressing, but it may be the case that exponential curves in natural do not exist. We may live in a time where we understand that sigmoids rule the evolution of our world/civilizations...
(And no the plural of Malthus is not proof of an exponential curves in nature...)
The obvious point is, we may very well be the first, or 10th. And that's precisely why 'they' aren't anywhere.
And it's plausible that we're vastly underestimating how difficult it actually is to spread machines throughout the galaxy.
I don't happen to agree with the notion that we'd have to see evidence by now if there has been other life.
It's entirely plausible that a hundred other unique humanoids have existed on this planet, a billion or two years before we did, and that they all failed with minimum trace. We know for a fact that several variations have failed, so why not another 100? And if it's so difficult, then why not apply those odds to other planets?
In five billion years our planet, which is very hospitable to life, has managed to produce one currently existing 'intelligent' life form (that can barely fly to the moon). That means it's extraordinarily rare even on our planet. By most measures, humans barely managed to survive, and we got lucky to so far avoid a wipeout event.
The odds of life being generated in any given billion year time frame, on a planet that can possibly support it, and assuming the right conditions for intelligent life (eg that the planet doesn't get destroyed or altered by an asteroid etc), and on the reduction goes.
I think spacefaring life is exceptionally rare, and it only gets X amount of time to potentially exist before the solar system or galaxy or universe crushes it.
In short, I don't think there's enough time for intelligent life to span even a galaxy before being wiped out by nature. And why don't we see probes floating around? For the same reason Voyager will eventually be obliterated by rocks in space or gravity (and in the time scale of the milky way, it won't be a very long before that happens).
I don't think this is true. Are we assuming that every universe has an infinite number of civilizations? It's impossible to have uniform probability across an infinite set of discrete probabilities, because every outcome would have a probability of 0.
So with no other knowledge, it's slightly more likely that we are the first civilization than the 5 millionth.
Of course it's tricky to talk about the "first" civilization in the face of relativistic time, but I'm assuming that there is some way to have an ordering of all civilizations.
Assuming there are very many civilizations in our universe, the odds that we're the first are tremendously lower than the odds that we're not. I think that's the point.
> As opposed to us simply being some other civilization on one of the other billions of habitable planets in that same "first" universe?
Yes, but if we are in a simulation we can almost answer the Fine-tuned Universe Paradox too, two birds with one stone. That is nice for Occam's rasor reasons.
We know absolutely nothing beyond this universe. We barely even know much about this universe.
You can't meaningfully speculate about the nature of what (if anything) is dimensionally/physically/extra-universally "outside" of our universe.
We don't even know what that means, much less if it exists or what the number/nature of it would be.
Our universe may be a piece of candy floating on a chocolate river in another universe that flows eternally.
It's all just make believe. Make believe can be fun, but let's not pretend that we're doing anything more substantial than what three year olds do when talking about space robots.
There is a Star Trek Voyager episode where three humanoids on a planet survived a natural disaster that destroyed their planet by going into stasis and having their brains stimulated by a computer simulation for over 19 years. In that time, the adaptive computer responded to their worst fears by turning the "fun" model world full of clowns and dancers into a circus nightmare. The virtual characters created to entertain the survivors turned into sinister clowns. The head clown, and the manifestation of fear itself, has full access to their brains and would adapt to circumvent any attempts by them to leave his world or terminate the stasis.
Are you sure? Is the universe described by an equation s=t*v, v=10 untrue unless we compute it? Does moving electrons on a CPU die suddenly give it meaning? Do scraps of graphite left on paper by a pencil? The simplest turing complete cellular automaton can be computed using ordinary rocks.
Are there no infinitely many natural numbers unless we observe them? And then finally is actually computing a universe described by a set and a step function the only way to make it real? Is it not "real" by simply being possible?
As for discreteness of time, in quantum mechanics changes in state are modeled by square matrices over complex numbers. GL(n, C) is complete - for every matrix M you also have M^1/2, representing half the state chance. It follows that our time, at conceptual level, is at least as "dense" as rational numbers.
“There is a theory which states that if ever anyone discovers exactly what the Universe is for and why it is here, it will instantly disappear and be replaced by something even more bizarre and inexplicable.
There is another theory which states that this has already happened.”
― Douglas Adams, The Restaurant at the End of the Universe
I think it's a bit silly to try to infer things about the computer that could be simulating our universe based on things internal to the simulation. Many of our fictions don't have the same laws of physics as we do.
It's a hypothesis - namely, that the entities running the simulation are trying to simulate their own universe to make deductions and discoveries about it.
Therefore, our universe would be a reasonable facsimile of there's in most respects.
Which has the implication that their may be limitations on their computational substrates which required workarounds or optimizations, or as the lattice QCD case points out - which require representing phenomenoma as discrete computational steps, rather then continuous analytical functions. It's a well known problem in all the computational disciplines, that changing the time-step of your simulation frequently leads to wildly different results then you expect.
Think about the speed of light and tell me how it makes _any_ sense that time slows down due to velocity. Yet that's exactly what EVE Online does to stop nodes crashing during heavy gameplay.
To the best of our knowledge, but that's kind of the point of this work isn't it? Look at something blown up on a large enough scale and see if the results show that it's actually continuous or if there's evidence that (in this case) there's a "real" lattice QCD grid size affecting things.
I understand the hypothesis. I am even aware of acausal game-theoretic arguments that people at our stage of technological development may be disproportionately likely to be simulated.
I didn't mean to say that this isn't an interesting line of inquiry, just that I don't think we can possibly consider the result conclusive either way. I think it's reasonable to re-weight one's beliefs to an extent based on the outcome, but not so far as to adopt one possibility as "a belief" and discard the other.
'Sense' is from the perspective of your limited (human) viewpoint. It makes no 'sense' to me that a civilization advanced enough to simulate a whole universe for the purpose of study would be stupid enough to expose such a bug/limitation that would expose their (possible) existence before we even built the first transistor.
If someone is running our universe as a simulation they still don't necessarily know of our existence. I have a hard time figuring out how you'd detect intelligent life or its byproducts in your simulation. Even if you do detect something alive, how do you interact with it in a non-destructive way? Also not trivial at all :) That "long-distance phone call" might be far less likely than us ever figuring out if we're indeed a simulation.
"Abstract. I think there's a small but non-negligible probability that humans or their descendants will create infinitely many new universes in a laboratory. Under weak assumptions, this would entail the creation of infinitely many sentient organisms. Many of those organisms would be small and short-lived, and their lives in the wild would often involve far more pain than happiness. Given the seriousness of suffering, I conclude that creating infinitely many universes would be infinitely bad."
I've always been interested by this sort of stuff (as I'm sure several people are -- especially the HN audience). It always brings up a bunch of interesting questions:
- If we're self aware of the simulation, would we be able (or ever want to) reverse engineer the fabric of which makes up our sim?
- Would there be infinite simulations? How far _up_ does the rabbit hole go?
- Another idea, and this is reaching, would be to consider the motivation behind running simulations like these. Are the "controllers" running simulations to try and find solutions to problems they are facing? Of course, this is the stuff right out of Scifi movies.
The problem with this theory (the universe is a simulation) is that it's a tautology. You cannot prove or disprove it, and for practical reasons it changes nothing if you assert it's true.
What is really being researched is whether the universe is computable, because that's something you can prove.
If they can prove the universe is computable and that we would (given the resources) create a simulation of our own, wouldn't that lend credence to the idea that we would be in a simulation as well?
I see what you mean though, it would be impossible (with our current understanding of ... everything) to actually determine is we're in a sim or not.
The point is, to determine whether "this" universe is a simulation or not, you would need to look after a special characteristic that differentiates it from a "parent" universe.
If you prove the universe is computable, that still doesn't give you any answer whether it's a simulation or not.
As it stands, the theory is a tautology, because that's just proving the universe is something that it is.
I've always thought that quantum theory provided some decent evidence for this. Instead of always computing every value for every result, the universe does so only when things are "observed", saving on computation.
No, it doesn't really work that way. A wavefunction is based upon a configuration space and you can't just neglect part of it or you'll get inaccurate results.
"The theory basically goes that any civilisation which could evolve to a 'post-human' stage would almost certainly learn to run simulations on the scale of a universe. And that given the size of reality - billions of worlds, around billions of suns - it is fairly likely that if this is possible, it has already happened."
That sounds distinctly like Anselm's ontological proof for the existence of God, and thus likely suffers from the same logical fallacies.
If you want to take on a non-strawman version of the argument, check out benhammer's links below.
That said, I do often feel that at the literal edges of our universe (in all the senses of that term), our uncertainties simply dominate everything and all of our guesses about the true nature of reality are dominated by that uncertainty. Something (or, depending, multiple somethings) must be true for some value of true, but I'm skeptical if we can discover it. If it popped up and simply flat-out told us what the truth was somehow (the Great Simulator writing the message in the stars [1]), we probably couldn't confirm it.
Well the most interesting part of the Matrix trilogy was when Neo used his powers in the 'real' world; specifically stopping the drones and being able to see despite losing his eyesight.
Totally. I expected the simulation within a simulation, turtles all the way down, plot twist.
My favorite theory was that all the humans were passengers on a generational seed ship traveling to the next star, kept in near stasis to keep them alive and mostly sane during the millennial transit.
That would have played nicely with The Architect's explanation that in the perfect simulated world, humans went insane. So The Architect devised worlds with struggles. But the iteration we observed got out of hand, due to The Oracle's meddling.
Yes, this. I take it that the amount of inverse entropy is always diminishing. In order to fully simulate a universe with the same amount of entropy^-1 we need at least as much as we had at the initial moment. But entropy^-1 will always decrease.
I imagine it would be like a computer game, where you only had to simulate the universe from the point of view of the user. This would fit nicely with the Copenhagen interpretation.
Yeah, I am thinking about the Trueman show, around him the simulation is very detailed and polished. The further he is away though from a part of the world the less simulation effort that is being made.
Anselm defined God as "that than which nothing greater can be conceived", and then argued that this being could exist in the mind. He suggested that, if the greatest possible being exists in the mind, it must also exist in reality. If it only exists in the mind, a greater being is possible—one which exists in the mind and in reality.
What are the specific parallels you see, beyond them both sounding specious?
The simulation argument says: if simulating minds is possible, and technologies eventually advance to the point that they can do it, a civilization would make lots of them. Thus, most minds are virtual, so your mind has a high base rate of being one.
Anselm's argument is: the best thing in the world would have to exist, since existence is better than non-existence. I define God as the best thing, so God must exist.
I don't quite see the parallel, but if you see something specific, I'd like to hear it.
210 comments
[ 3.7 ms ] story [ 235 ms ] thread(No, please, do NOT do this.)
There is a theory which states that if ever anyone discovers exactly what the Universe is for and why it is here, it will instantly disappear and be replaced by something even more bizarre and inexplicable.
There is another theory which states that this has already happened.
https://en.wikipedia.org/wiki/Slartibartfast
wrong. there are multiple alternatives; we are the nth civilization in a real universe, we are the nth to the nth civilization in the multiverse, etc. Sloppy sensationalism and bad statistical reasoning in this article.
It's almost like saying "we're probably characters in a novel," since for every universe there are many novels. Or like that old proof that God exists because he has every desirable quality, including existence -- as if a hypothetical entity could be forced into being by sheer burden of how it is described.
When we talk and reason, we call something a universe to invoke all the general properties our universe has -- except existence, of course, because it would be silly if we could only talk about things that exist. Various things like books and simulations are physical representations of universes that don't exist. By virtue of our interpretation, they are universes nonetheless. When reading a book, we fill in assumptions and details from real life if the author gives us no reason to think otherwise. Tautologically, the world of the book is different from the real world in some limited and structured way, but not in any way that keeps it from being a world at all.
Physics seeks to explain the world exhaustively and reductionistically in terms of mechanisms, models, and mathematics -- basically, to boil it all down to the consequences that emerge from laws and equations that we can completely conceptualize. If the universe consisted just of electromagnetism, for example, then Maxwell's equations would be a "complete" description of the universe, and any simulation of them could be considered a universe. Everything about the universe, every general description that's true of it without reference to specific places, times, and things, would be reflected in the simulated universe as well, because science allows us to subsume it all into the more fundamental description given by the equations.
The problem is that a model of the universe is stil a model. Humans invented the notion that a thing and a description of a thing are equivalent.
The bottom line is that we just don't know, and are unlikely ever to know, but if one wants to argue intelligently about it, there's a huge amount of philosophical literature on just this topic.
On the other hand, your position would seem to deny human rights to future intelligent robots, which has serious ethical ramifications should you happen to be wrong.
But the thing is, the exact same thing is true for your brain too! If you zoom in on your brain you will find nothing more than electrical impulses and neurons that strengthen or weaken their connection to other neurons. If the neurons were suddenly replaced with little computers that performed exactly the same, you wouldn't feel any different. You would still experience emotions and feel sensations and have no idea any parts had been swapped under the hood, unless you looked.
I don't know why but this idea is confusing to me. This XKCD http://xkcd.com/505/ is a good example. My consciousness could be a bunch of static rocks and a man placing new rows based on simple rules. I find this thought disturbing, but there isn't anything fundamentally different between that, a computer simulation, and a world of a bunch of atoms hitting each other, following simple physical laws.
Different beings could exist in all three universes. And each would feel emotions and sensations and all that. And each would argue about how beings just like them in the other universes "don't really exist" like they do.
So how is this different from a character in a novel or a variable in a computer set to "happy"? I honestly don't know. The character in a book is completely static. Whatever is written in the pages, it isn't changing, it isn't taking input and producing output. But the rows of rocks are static too after all. They don't take any input from the outside world either. They change over a dimension of space, each new row one foot to the right of the last. As do the characters in a book. Each paragraph moves the characters forward through their own dimension of time. Maybe the process that creates the rows of rocks, the man following simple rules, is what makes it conscious. But the book is created by a writer following a process too, does that make it conscious?
The only thing I can think of is that the character in the book is far simpler than a human. Words that say "john is happy" is similar to setting a variable on your computer equal to "happy". Happiness in a human is far more complex, involving tons of interactions between neurons. But this answer isn't satisfactory to me. Complexity isn't what makes something intelligent or conscious after all.
This is the difference between a running program and a short description of the program. The short description isn't runnable, and isn't running. The reason the program itself is doing what we want it to do is not because it's more complex than the short description (though it is), but because it's actually executing the algorithm. We could imagine more and more detailed descriptions of the program, specifying more and more closely how it works, and eventually it's so detailed that you could run the program from that description. At that point, the description is still not a running program, unless and until you run it. Being conscious is a process -- a running algorithm -- not a static description.
Anyways you can think of time as a dimension. Each moment of time exists, connected to the ones that came before and come after it. You can represent the every state that a turing machine goes through all at once. Just like those rows of rocks. What does "running" even mean in a universe where all points in time can exist at once?
But regardless of that, you can still say the book is "running" when it is being written or when it is being read. The characters don't just exist in the book, but literally in your mind. Though I still wouldn't call them conscious or even intelligent.
If they all exist at once, then in what sense are they connected? Certainly not a causal sense, which I would argue matters quite a lot.
> The characters don't just exist in the book, but literally in your mind. Though I still wouldn't call them conscious or even intelligent.
If your mind was able to simulate the actual process of consciousness in the brain of the character being read about, then I would argue that the character could be conscious. Of course, brains don't have that much additional processing power. :)
That is pure speculation. Just because we can observe similarities between two phenomenons it doesn't mean they are the same thing. You're erroneously assuming that the universe can only work in a way that fits our theories and models.
Why not?
But that's equally true of saying "complexity doesn't make something intelligent and conscious". All right, then what does? "Not complex" is a label just as much as "complex" is. Either way you haven't explained anything. That's what I was trying to get at.
Also, saying "complexity doesn't make something intelligent/conscious" is not the same as saying "something intelligent/conscious doesn't have to be complex", which is what I think the statement I originally responded to, taken in context, was really intended to mean. Do you really think intelligence and consciousness can be present in simple things like rocks? Might there not be a reason that the human brain is three pounds of complex matter, not three pounds of jello?
Anyways in the original context of where I said that, I was trying to come up with a satisfactory way to distinguish consciousness from non-consciousness. Complexity obviously isn't the way since lots of complex things are not conscious. It may be that all conscious things are also complex. But then complexity isn't the reason it's conscious, it just happens to correlate with it.
Setting a variable in a computer equal to "happy" obviously doesn't make the computer experience happiness. But then what possible sequence of commands and state changes would? If you accept that there is no possible sequence that would, then you have to accept that humans can not experience consciousness either. Because for all we know we are running in a computer too. Even if we are not, the process our brain follows easily could. The fact that it runs on atoms bouncing into each other and not electrons flowing through logic gates makes no difference.
Really? Show your work, please.
A lot of neurons may seem complicated, but you only need to understand the programming of a few, than just copy them a billion times.
This assumes that they are all "programmed" the same. Why do you think that must be the case? Also, you're leaving out all the chemical processes that contribute to brain function.
I was trying to come up with a satisfactory way to distinguish consciousness from non-consciousness. Complexity obviously isn't the way since lots of complex things are not conscious.
True.
It may be that all conscious things are also complex. But then complexity isn't the reason it's conscious, it just happens to correlate with it.
The connection might well be stronger than "just happens to correlate". Even if "complexity" by itself isn't the reason it's conscious, it might well be that consciousness requires a complex substrate.
Setting a variable in a computer equal to "happy" obviously doesn't make the computer experience happiness. But then what possible sequence of commands and state changes would?
Um, a much more complex sequence of commands and state changes?
If you accept that there is no possible sequence that would
I don't. I just think there is no simple sequence that would.
If that is too complex, you could create an even simpler algorithm which produces that. Maybe by creating a genetic algorithm which "evolves" human-like intelligence after billions of generations.
In both cases the end result may be complex, but the algorithm that creates it is not. Whether it be a network of billions of interconnected neurons that arise from simple programming of a few neurons, or a complex intelligence optimized by simple random mutation and selection.
This is the idea of emergence. That complex seeming behavior can "emerge" from simple rules and a simple process.
The problem is that both of these terms are really too vague. What do you mean by "consciousness"? Some people claim insects are conscious; some claim bacteria are conscious; some even claim electrons are conscious.
Also, what do you mean by "intelligence"? By some definitions, the process of evolution by natural selection is "intelligent".
If by these terms, you mean "the properties humans have that we call consciousness and intelligence", then why do you think there is no connection between them? Does your own intelligence really have nothing to do with your consciousness? Or vice versa?
It's conceivable you could create something like a human with very simple machinery, just massively scaled up.
Conceivable, yes. Likely, I don't think so. But that's really a matter of opinion; it depends on how much of the complexity in the parts of the human brain (neurons, chemicals, etc.) is necessary for human consciousness or intelligence, and how much you think is just an artifact of the implementation, so to speak. We don't know enough about how the brain works yet to really judge.
the end result may be complex, but the algorithm that creates it is not.
In other words, the complexity is still there.
That complex seeming behavior can "emerge" from simple rules and a simple process.
Wait a minute--complex seeming? Or complex?
If you're trying to argue that the complexity may not be in the algorithm itself, but only in the actual working out of the algorithm in time and space, I'll buy that.
But if you're trying to argue that it isn't "really" complex because the algorithm is simple, I don't buy that. I never said the algorithm had to be complex; I just said there has to be complexity somewhere for there to be (human-level, to avoid all the definitional issues I raised above) consciousness and intelligence. That's perfectly consistent with the complexity ultimately being built out of simple parts--just a lot of them with a lot of interactions, so the complexity is in the interactions, not in the parts themselves.
(I think it's likely that the parts themselves have significant complexity too, as I said above, but even if that's true, there will still be some lower level, possibly much lower, where there are simple parts, just a huge, huge number of them with a lot of interactions.)
>In other words, the complexity is still there.
Well yes in a way. By the definitions used in information theory, complexity is just the amount of bits you need to accurately describe something in some language. By common usage complexity just means the amount of concepts you need to learn to understand something (about the same thing.) But it can also mean the number of moving parts a machine has, so to speak, which could be very large while simple enough to understand or write on paper. Like a computer display which has a few thousand pixels each of which is almost exactly the same.
I don't know if you would call a computer display complicated, at least not much moreso than the invidual pixels that make it up. The same is probably true of humans. Even if humans are fairly complicated, someday a connectionist approach or maybe a rough approximation of a human brain might succeed at creating something arguably conscious.
>If you're trying to argue that the complexity may not be in the algorithm itself, but only in the actual working out of the algorithm in time and space, I'll buy that.
This may be a much better way of describing what I just said above. So yeah.
Consciousness is a property that we ascribe to ourselves and other entities. We don't experience it in others the way we experience it in ourselves, of course, but we infer it from the data available from interacting with them. They produce a pattern of signals that resonates deeply enough with us that we recognize them as like us in having goals, feelings, and access to the "human experience." For example, one of the most convincing things we could observe an AI do would be to make a novel observation about life, derived from current circumstances, that strikes us as deep and insightful, perhaps because it was hovering at the edge of our own awareness, or because it immediately activates mental patterns that we did not activate but are also meaningfully connected to the situation at hand for us.
It makes sense that if we were to copy the workings of the brain in sufficient detail in a machine, the machine would also be ascribed consciousness by human users. Perhaps we could say that it is an "artificial brain" in a literal sense, performing normal brain functions the way an artificial heart pumps blood.
In the case of the universe, it's at issue whether there is such a thing as a "complete description" of the universe, whether a simulation could be done, and then whether this simulation actually is a universe in some sense. Unlike brains, we don't regularly encounter other universes (actual, physical ones), ones which we don't live in and experience from the inside, but which we are forced to conclude are true manifestations that correspond in every important detail to our own nevertheless. A complete description of the universe by a future physicist (say) would start with a short list of all the types of information that go into describing a particular universe, like a list of type of particles and variables such as position and momentum. This list describes universes in general. Then we'd have to describe a particular universe, with a long, long, long list of all the individual particles and so on. Then we have to represent this information physically in some sort of computer, for the purpose of simulation, and perform the ongoing processing involved in running the simulation.
Finally, we must argue that the simulation is a universe. Not just deserving of being called one, or having the properties typically associated with one, but a true instance of some class that previously had the universe we live in as its single member.
One interesting feature of this line of thought is its seemingly infinite ambition. What about smaller goals we could set for mankind -- might we ever engineer an artificial star? How about an artificial galaxy? We can try to beef up the likelihood of this ever happening by considering not just mankind but any other "intelligent beings" the universe might happen to contain, in the past, present, and future. However, we are again on the shaky ground of inventing a class from a singleton. All extraterrestrial "intelligent beings" ever conceived are just stories with small, structured deviations from the human template; beings we identify with because they, too, have intentions and goals and act in their ultimate self-interest for the purpose of self-preservation and just trying to make a go of it in this big ol' universe. Some of them, it's presumed, make detailed copies of natural things at an astronomical scale and then talk about whether the old thing and the new thing are members of the same class of thing or not.
I agree that it's irrelevant to us and our operation if our brains are made of billiard balls or quantum clouds, and the same seems true of the universe.
It is still interesting whether the universe is "computable" or not, that is, whether something meeting the current definition of a computer could simulate it, even in theory. ...
I think you've come much closer than I to getting a definition of consciousness, but I still don't find it satisfactory. You could have a computer mimic a human without actually having anything resembling an inner "human experience". For example, a giant look up table with a preprogrammed response to any possible input. We have created extremely simple versions of this that are almost effective enough to pass the turing test. Just by gather a large number of human responses to chatbots. Though you could say the process that creates the giant look up table is intelligent, the table itself obviously is not.
There are AIs like AIXI which could be vastly intelligent without reasonably being considered conscious. They don't actually "think" so much as brute-force every possible solution to a given problem, or every possible explanation for a series of inputs, etc. Given enough computing power this would actually be effective. You could then ask it to mimic a conscious being and it could do so easily (given either a definition of consciousness or a set of observations about other conscious beings.) Maybe doing that requires the intelligence to actually simulate a conscious process somewhere along the line, so I don't know if that counts. It may be able to do so through mere, without having to actually run a full simulation. Providing a "deep insight" about life is nowhere near as complicated as a full human consciousness.
Maybe there is no such thing as consciousness. In the sense that there is no simple way to define it that is fully consistent with everything we want it to mean. That there are always going to be arbitrary seeming exceptions and gray areas. But that has deep and disturbing implications for morality. Maybe it's not relevant to our day to day life, but if we ever want to program a friendly artificial intelligence, we are screwed.
>I think it's silly or meaningless to say we possibly -- or probably! -- live in an "artificial" universe.
Well assuming that there are more simulated beings in all of existence, whatever that means, than there are non-simulated ones, it is very likely we are simulations. Though it's impossible to know what exists beyond our own universe. What is the distribution of universes? Do 50% of universes allow for infinite computing power, for example? What programming language do they use, which would determine how many bits a given universe would take to specify in that language, and therefore how often it gets simulated? How many universes even have anything close to intelligent beings?
An infinitely powerful computer would allow you to simulate every possible universe all at once, instantly (you would also be able to simulate every possible branch of non-deterministic universes, continuous variables, time travelling, and other things that people claim makes the universe "uncomputable.) And you could instantly do it an infinite number of times. If it were possible for such a thing to exist in a universe, that universe would contain every other possible universe. Including copies of itself and others universes with infinite computing power, which would contain more universes, etc.
(Very interesting short story that explains the implications of this http://qntm.org/responsibility)
So what is the top level universe? Why does something exist rather than nothing? Perhaps everything that can exist does exist. Maybe every computable program has been (is being?) run, including one that specifies our universe. That's the simplest possible explanation that explains the existence of the universe. Though I ...
Simulated universes: I still think this is all meaningless metaphor. It's not easy to articulate why, but I'll try.
Some lines of thought just combine ideas in deep ways, but don't actually tell you anything new. For example, Zeno's paradox -- to get from point A to point B, you have to first go halfway, then halfway again, and so on forever -- doesn't actually teach us that nothing can ever go from point A to point B. It just demonstrates a glitch in the metaphors we use to reason about time and space. Actually, it's a glitch in the metaphor of traversing space as achieving a goal. Normally, if reaching a particular goal involves an infinite to-do list of subgoals, we wouldn't expect to ever get to the end and actually achieve the goal. However, if the goal is traversing some infinitely subdividable interval of space, we are provided with a mechanism to generate an infinite list of subgoals that are presumably all involved in achieving the larger goal. It's all just mental gymnastics and the normal tools of intuition breaking down.
To take a closer example, consider this line of thinking: "If you pick a random marriage proposal, chances are it is an imagined one rather than an actual one, because for every actual marriage proposal there are on average two or more imagined ones that never came to pass." This strange kind of thinking also mashes together diverse intuitive concepts. There's the idea that every class of thing (like "marriage proposal") has an extension set, an ensemble of instances. There's the fiction that an "event" is a discrete thing that we could identify and count if we had to, like counting the number of thoughts I've had today ("a thought" is a singular noun after all). There's the concept from probability theory of picking a member at random from a set (with the presumption that this is a well-defined act). There's the idea that an imagined X is still an X, because we still call it one. Unicorns are still unicorns even if they don't happen to exist.
The most intellectual, abstract part of the brain has a tendency to focus on what could possibly be rather than what is. We can get so wrapped up in generalities that we forget that this is the universe -- what we see around us, not what we imagine or are capable of conceiving of. Statements like, "Perhaps every...
http://www.simulation-argument.com/simulation.pdf
Anyway, your objection here isn't actually relevant as far as I can tell. It's a probabilistic argument about the number of real minds vs. the number of simulated minds, and the bulk of the "proof" is in demonstrating that, under certain assumptions, the number of simulated minds would be so high in comparison to the number of real minds that you'd almost be forced to conclude that you are simulated.
It really doesn't have anything to do with "universes inside universes."
And I'm pretty sure the numbers you mentioned--like the number of real universes, the probability of life arising from non-life--really don't come into it at all. Skimming over the paper, none of that stuff seems to come up at all. Again, this isn't about universes inside universes (the news article phrased the argument poorly).
A provoking question: if we found evidence that convinced us, beyond a reasonable doubt, that we were living in a computer simulation, would you change your behavior?
I treat free will as an assumption:
- If we are, in fact, predestined, assuming I have free will doesn't hurt or matter.
- If we do have free will, I'm applying that free will to make decisions based on a correct assumption
- If we do have free will and I had assumed we were predestined, this could have worse results. At the very least, it is unlikely to improve outcomes
Not to mention you were predestined to assume so.
What 'postulates' would they be calculated from? Does it even make sense to speak of universes following different 'logics'?
> far more universes simulated that have properties that are nothing like the "real" one as technology progresses, i.e. there are more instances of "sonic the hedgehogs" and "super mario brothers 1"
Those seem like perfectly normal universes to me: they're fully describable by short programs executing over discrete binary states like a chunk of RAM. Where's the violation of logic or metaphysical oddities there?
But the 'real universe' appears discrete on the smallest level, and I don't believe physicists will ultimately settle on a theory in which two completely different sorts of theories take over at different regimes and the macro-scale theory is not reducible, even in principle, to derivations and predictions of the micro-scale theory. The quantum theories will eventually predict the relativity-scale results./
> In addition, a "video-game universe" runs on a programmatic logic with discrete conditionals rather than our universe's equations that don't branch.
How would you know? What governs quantum randomness?
Exactly the point. What game do you know of that follows the laws of logic as we know them in the 'real' universe we habitate? It follows that it's nonsensical that we can reason ourself to reality if we are a simulation.
Would? If they could, they might, or they might be pleased we figured it out :-P
As opposed to us simply being some other civilization on one of the other billions of habitable planets in that same "first" universe?
My guess is that there are other civilizations out there, but they're suffering from the exact same problems as us - long distances, huge energy requirements, lots of time. If we can't get out to the stars, I can see why plenty of civilizations would be having trouble.
It appears to be phenomenally difficult to get from point to point in the galaxy/universe (see the recent trend in science-fiction to stick to a speed of light limited universe). If we come up with some fantastic energy source (see: ZPE), then I'll say that it's phenomenally slow so that no civilization would invest in doing it. If we come up with some civilization that would do so, I'd point out that it's phenomenally slow/difficult to send signals across universal distances. Rational folks think that non-solar-local interactions are probably not possible... Large brained rational folks think to look for solar-local artifacts of Kardashev-scale>1 civilizations (see: http://www.personal.psu.edu/jtw13/blogs/astrowright/2013/03/...).
Recall that early explorers met Native Americans and called them Indians because they had misinterpreted distances, didn't know all of the continents, etc. We have a history of overcoming increasingly large distances with improved transport. This does not mean that we shall always do so.
It may be depressing, but it may be the case that exponential curves in natural do not exist. We may live in a time where we understand that sigmoids rule the evolution of our world/civilizations...
(And no the plural of Malthus is not proof of an exponential curves in nature...)
The obvious point is, we may very well be the first, or 10th. And that's precisely why 'they' aren't anywhere.
And it's plausible that we're vastly underestimating how difficult it actually is to spread machines throughout the galaxy.
I don't happen to agree with the notion that we'd have to see evidence by now if there has been other life.
It's entirely plausible that a hundred other unique humanoids have existed on this planet, a billion or two years before we did, and that they all failed with minimum trace. We know for a fact that several variations have failed, so why not another 100? And if it's so difficult, then why not apply those odds to other planets?
In five billion years our planet, which is very hospitable to life, has managed to produce one currently existing 'intelligent' life form (that can barely fly to the moon). That means it's extraordinarily rare even on our planet. By most measures, humans barely managed to survive, and we got lucky to so far avoid a wipeout event.
The odds of life being generated in any given billion year time frame, on a planet that can possibly support it, and assuming the right conditions for intelligent life (eg that the planet doesn't get destroyed or altered by an asteroid etc), and on the reduction goes.
I think spacefaring life is exceptionally rare, and it only gets X amount of time to potentially exist before the solar system or galaxy or universe crushes it.
In short, I don't think there's enough time for intelligent life to span even a galaxy before being wiped out by nature. And why don't we see probes floating around? For the same reason Voyager will eventually be obliterated by rocks in space or gravity (and in the time scale of the milky way, it won't be a very long before that happens).
So with no other knowledge, it's slightly more likely that we are the first civilization than the 5 millionth.
Of course it's tricky to talk about the "first" civilization in the face of relativistic time, but I'm assuming that there is some way to have an ordering of all civilizations.
Yes, but if we are in a simulation we can almost answer the Fine-tuned Universe Paradox too, two birds with one stone. That is nice for Occam's rasor reasons.
You can't meaningfully speculate about the nature of what (if anything) is dimensionally/physically/extra-universally "outside" of our universe.
We don't even know what that means, much less if it exists or what the number/nature of it would be.
Our universe may be a piece of candy floating on a chocolate river in another universe that flows eternally.
It's all just make believe. Make believe can be fun, but let's not pretend that we're doing anything more substantial than what three year olds do when talking about space robots.
"Physicists to Test if Universe is Computable"
I think the question is whether the compution is discrete (i.e., cheats) or is everywhere continuous.
Are there no infinitely many natural numbers unless we observe them? And then finally is actually computing a universe described by a set and a step function the only way to make it real? Is it not "real" by simply being possible?
As for discreteness of time, in quantum mechanics changes in state are modeled by square matrices over complex numbers. GL(n, C) is complete - for every matrix M you also have M^1/2, representing half the state chance. It follows that our time, at conceptual level, is at least as "dense" as rational numbers.
Therefore, our universe would be a reasonable facsimile of there's in most respects.
Which has the implication that their may be limitations on their computational substrates which required workarounds or optimizations, or as the lattice QCD case points out - which require representing phenomenoma as discrete computational steps, rather then continuous analytical functions. It's a well known problem in all the computational disciplines, that changing the time-step of your simulation frequently leads to wildly different results then you expect.
Think about the speed of light and tell me how it makes _any_ sense that time slows down due to velocity. Yet that's exactly what EVE Online does to stop nodes crashing during heavy gameplay.
Would you take that as evidence that it does not computate discretely?
I didn't mean to say that this isn't an interesting line of inquiry, just that I don't think we can possibly consider the result conclusive either way. I think it's reasonable to re-weight one's beliefs to an extent based on the outcome, but not so far as to adopt one possibility as "a belief" and discard the other.
http://www.utilitarian-essays.com/lab-universes.html
"Abstract. I think there's a small but non-negligible probability that humans or their descendants will create infinitely many new universes in a laboratory. Under weak assumptions, this would entail the creation of infinitely many sentient organisms. Many of those organisms would be small and short-lived, and their lives in the wild would often involve far more pain than happiness. Given the seriousness of suffering, I conclude that creating infinitely many universes would be infinitely bad."
- If we're self aware of the simulation, would we be able (or ever want to) reverse engineer the fabric of which makes up our sim?
- Would there be infinite simulations? How far _up_ does the rabbit hole go?
- Another idea, and this is reaching, would be to consider the motivation behind running simulations like these. Are the "controllers" running simulations to try and find solutions to problems they are facing? Of course, this is the stuff right out of Scifi movies.
What is really being researched is whether the universe is computable, because that's something you can prove.
I see what you mean though, it would be impossible (with our current understanding of ... everything) to actually determine is we're in a sim or not.
If you prove the universe is computable, that still doesn't give you any answer whether it's a simulation or not.
As it stands, the theory is a tautology, because that's just proving the universe is something that it is.
In 1969, Zuse suggested the concept of a computation-based universe in his book Rechnender Raum (Calculating Space).
http://en.wikipedia.org/wiki/Konrad_Zuse
The problem is that this is human bias. You can describe any theory by information manipulation, and it is as good a description as any other.
That sounds distinctly like Anselm's ontological proof for the existence of God, and thus likely suffers from the same logical fallacies.
That said, I do often feel that at the literal edges of our universe (in all the senses of that term), our uncertainties simply dominate everything and all of our guesses about the true nature of reality are dominated by that uncertainty. Something (or, depending, multiple somethings) must be true for some value of true, but I'm skeptical if we can discover it. If it popped up and simply flat-out told us what the truth was somehow (the Great Simulator writing the message in the stars [1]), we probably couldn't confirm it.
[1]: http://lesswrong.com/lw/qk/that_alien_message/
My favorite theory was that all the humans were passengers on a generational seed ship traveling to the next star, kept in near stasis to keep them alive and mostly sane during the millennial transit.
That would have played nicely with The Architect's explanation that in the perfect simulated world, humans went insane. So The Architect devised worlds with struggles. But the iteration we observed got out of hand, due to The Oracle's meddling.
Such a wasted opportunity.
Yeah, but there were a lot of explosions.
Pretty sure under our current understanding of physics, it's impossible to simulate a universe like our own within our own.
Impossible to perfectly model a universe, but not impossible to simulate something that looks like a tiny part of a universe.
http://en.wikipedia.org/wiki/Ontological_argument
(no comment, just providing context)
The simulation argument says: if simulating minds is possible, and technologies eventually advance to the point that they can do it, a civilization would make lots of them. Thus, most minds are virtual, so your mind has a high base rate of being one.
Anselm's argument is: the best thing in the world would have to exist, since existence is better than non-existence. I define God as the best thing, so God must exist.
I don't quite see the parallel, but if you see something specific, I'd like to hear it.