Odds are we're living in a simulation, says Elon Musk
theverge.com
theverge.com
— Hawking, 1988
https://en.wikipedia.org/wiki/Turtles_all_the_way_down#Histo...
https://en.wikipedia.org/wiki/Infinite_regress#Consciousness
edit: What if we talked about the idea rather than worried about who said it?
Perhaps if we tried to build a simulated universe, things would start breaking, or events would happen to prevent it, or our simulation would just be shut down or reset.
Second, the simulation argument is about our own universe someday making simulated realities. Assuming we are being simulated by a radically different universe is entirely different.
We should expect there to be a limit on how powerful the computers we could build are, if we are in a universe governed by the physical laws we now understand, whether or not we are in a simulation.
Because the laws of physics aren't actually being simulated, any more than a video game simulates every particle and photon of light. It's just an approximation, and the computing power isn't there to actually simulate a computer that big.
No, it wouldn't. A simulated universe would have exactly the limits of the laws of the simulation, which would be exactly the physical laws of the universe as seen from within.
> Because the laws of physics aren't actually being simulated, any more than a video game simulates every particle and photon of light. It's just an approximation
The approximation (of some idealized laws of physics conceived in the external realiy) being simulated would be exactly the laws of physics observed in the internal reality of the simulation.
Yes, whatever the simulation simulated would be the physics observed within the simulated universe. They may not fully match some idealized physics (perhaps those of the external universe, perhaps some alternative designed for the purpose of the simulation) for which the code of the simulation is an approximation, but from the inside of the simulation the rules by which the simulation determined outcome would be exactly the empirically-discoverable laws of physics of the simulated universe.
The only way to observe a discrepancy is to escape the simulation, find out what the creators were intending to approximate, and then compare it to the results of the simulation, but you can't do that from within the simulation.
However a sufficiently clever simulator could cover that up a bit. By say altering the results of physics experiments that would detect that.
However there is a limit to how far they could go. We couldn't have computers too big, or it would break the simulation. So at some point they would have to intervene and prevent us from building them, or just reset the simulation.
Either way, you can't have "turtles all the way down". Each nested simulation would only have a tiny, tiny, fraction of a percent of the computing power required for the parent simulation. After just a few levels it would inevitably reach 0.
Whether they would or not depends on the complexity of the computer and software running the simulation, the limits to which are set by the laws of physics of the "real" universe in which the simulator exists (which may not be those of the simulation). Our intuition of what is practical in a simulation based on our own technology and what is practical in our universe simply don't apply.
> E.g. you would observe artifacts of 3d rendering, which we shouldn't expect to observe at all.
Our universe has a lot of quantization artifacts which defy intuition based on assumptions about continuity of physical quantities. Once we systematically observed and cataloged them, we incorporated models of them into our understanding of physical laws. Whether those are "real" physical laws in a non-simulated universe, intended simulated physical laws of a simulated universe, or simulation artifacts representing limitations of the simulation in a simulated universe are impossible to distinguish from within our universe.
> However there is a limit to how far they could go. We couldn't have computers too big, or it would break the simulation.
If the simulation is sufficient to model a universe the size of our observable universe at the detail level our universe is observable too, the maximum scale of computer we can have without breaking the simulation is a quantum computer of the complexity of our physical universe. Anything up to that scale is within the capacity of the simulation.
Basically, all of the assumed limitations you propose are undetectable from within.
You may be making the additional assumption that not only is there a host universe, but that the host universe has a scale and physical laws basically identical to those that we observe in our universe, such that the simulation would need to be substantially less complex. But there is no reason that either the scale or general physical laws of the host universe would have to mirror those in the simulation.
> Either way, you can't have "turtles all the way down".
If you assume the host universe is infinite in scope and information content (and the assumption that it is infinite in time and that all of that time is available to civilizations, which seems to underlie the conclusions on the that our universe is likely to be a simulation, implies this), then no matter how small of a fraction of its capacity one thinks a simulation can have, it can support an unbounded number of layers of simulation underneath it.
If you assume a finite root universe, then (even leaving aside other reasoning flaws in the argument), the conclusion that our universe is likely to be a simulation falls apart rapidly.
that doesnt make any sense to me.
If i create a simluation in my computer of a world without gravity, to the inhabitants of that simulation, gravity does not and never has existed.
similarly - what we understand to be the laws of physics could have been laws that were designed for our simulation but have no bearing on the laws that govern the world in which the computer simulating us exists.
The simulation argument is about simulations of our own universe, within our own universe. It's a very specific claim about the parent universe, and the nature of our simulation. As opposed to more general simulations, which can't make any predictions at all and tell us nothing.
The original simulation argument paper is about simulations intended to simulate the historical predecessors of the simulations creators in their own universe, but, even so, any anomalies from the approximation of the physical laws of the root universe used in the simulated universe would only be detectable by reference to the root universe, from the perspective of those within the simulation, the behavior of the simulation would be the physical laws of the universe they live in.
(This is equally true in the more general case where the intent of those creating the simulation is not specifically to do an "ancestor simulation" in a universe approximating the physical laws of their own universe, but restricting to the scenario envisioned in the original paper doesn't change it at all.)
That's just a tautology. Yes the code of the simulation are the laws of physics for the beings living in the simulation.
But the point of a historical simulation is to simulate your own universe. Not a completely different universe. We should expect simulated Earth to be exactly identical to real Earth. Or at least you shouldn't be able to tell the difference. So we should expect the laws of physics to appear the same, but things that require lots of computing power should fail.
We shouldn't be able to simulate universes within our own simulation. That would be a pointless waste of computing power.
If a best-effort ancestor simulation is made, we should expect the laws of the simulation to be identical to the laws of the root universe to the limit of the resolution of the simulation.
> Or at least you shouldn't be able to tell the difference. So we should expect the laws of physics to appear the same, but things that require lots of computing power should fail.
Except, that if we are in the simulation, we have no reference point as to what the actual laws of physics are in the root universe, so we have no way to distinguish between our universe being the root universe operating by the "real" laws of physics and our universe being a simulation which fails in some respect with regard to the root universe. The only way to detect the latter case is to be in the root universe and observing the simulation. Which we could do for a simulation we create, but not for a simulation we are in.
(And, the "things requiring lots of computing power should fail" expectation is unwarranted, at least from the perspective of being something that could be observed from inside of the simulation.)
What he said was that there is a one in a billion chance that our reality is base reality.
If you agree with Occam's Razor, there's no reason to believe we aren't in base reality and just the result of incredible luck (the same incredible luck that leaves us the only known life in the universe).
Understanding the probability of something is not the same as claiming the majority result of that probability.
I see no evidence (empirical or otherwise) that this probability is accurate.
Technology advances -> Simulations exist -> We're either the first civilization in infinity or we're a simulation.
It's possible we're the first civilization in infinity, and it's the simplest explanation given no additional information. However, just because something is true doesn't mean the probability of it being true is very low and just happened to end up that way. You can believe in both probability and reality.
Doesn't follow; there could be lots of civilizations in the same reality, so we could not be the first civilization, and still not be in a simulation. If simulations of the detail level of our universe are possible given the laws of nature governing the reality underlying everything, then either we are in the base reality or a subordinate reality (if such simualtions are not possible, we are definitively in the base reality.)
The information we have provides very little basis for computing probabilities here: any estimate is just a projection of the biases of the person providing it.
> However, just because something is true doesn't mean the probability of it being true is very low and just happened to end up that way.
If something is true, it's probability is 1, by definition. (Something may have low probability based only on particular prior information, but that's a different story.)
If the universe is of infinite age, the chances of our being the first civilization in infinity to create a realistic simulation is one in infinity. It's not magic, it's the simplest possible train of thought.
An assumption for which there is no evidence (in fact, there is considerable evidence that the universe is not; the simulation scenario requires assuming that our universe is not the "root" universe, and that some other universe exists outside of it -- which is not only something for which no evidence exists, but is something for which no evidence can exist.)
The "infinite age" conjecture is already assuming that our universe is a simulation (or, at least, an pocket universe embedded in some other universe.)
> the chances of our being the first civilization in infinity to create a realistic simulation is one in infinity.
No, that's only true if there are an infinite number of civilizations in the history of the universe, not merely if the universe is of infinite age. The two assumption are not equivalent.
And, in any case, its irrelevant, because not being the first civilization to develop a simulation doesn't mean we're in a simulation. If, say, the ancient Greeks had a breakthrough and were the first civilization to develop universe-level simulations, that wouldn't make the ancient Chinese living in a simulation.
> It's not magic, it's the simplest possible train of thought.
Its the sloppiest possible train of thought, that starts with assuming most of its conclusion, and then proceeding by unwarranted leaps to conclude the rest of it.
Assuming by "simulation scenario" you mean the theory that there is a chance we are running in a simulation, it doesn't assume we aren't the root universe (but it does make a lot of other assumptions), just that of some number of realities both simulated (s) and real (r), our chance of being in a non-simulated reality is r/(s+r).
Some (but not all) of the assumptions required are: Simulated reality is possible, human level AI is possible, it's possible to simulate a universe to the level we are at before the heat death of the simulation running universe (or the simulation running universe has different constraints, more time or power, etc).
> Its the sloppiest possible train of thought, that starts with assuming most of its conclusion, and then proceeding by unwarranted leaps to conclude the rest of it.
I don't think it's quite that, and I don't think it assumes most of its conclusion. I think given the assumptions the argument takes for granted, then it's a valid argument. While I think many of the assumptions are probably, it only takes one of those, and a much greater likelihood to be true, to greatly affect the original calculation, but even then, you have to assume that if there's a parent universe, it follows the same laws, otherwise you can't assume anything.
Ultimately, the thing we all need to keep in mind is that it doesn't matter. We have no ability to control or predict the consequences of being in a simulation, and given that our universe seems ultimately to conform to laws, and not act in what appears to be a capricious manner, it's inconsequential in the truest meaning of the word.
The argument premised on the infinitesimal chance of our being the first in an infinite timeframe requires assuming that our universe, which seems to have a finite age, is not the root universe.
That assumption is a fairly key part of the conclusion that the universe is a simulation; even assuming the rest of the steps of reasoning were correct, its not an argument for P(Our universe is a simulation | What is now known and observed) but P(Our universe is a simulation | Our universe is a pocket universe embedded in a root universe of infinite age)
I'm not sure where you're getting the infinite time frame requirement, besides previously in this thread (from someone else). I think that's an artifact of this thread, not a requirement of the "simulation scenario".
Assuming we could accurately model one universe to our age within another universe of similar laws (a sentence which contains many assumptions on both sides of the proposal), we are presented with a situation where we are either the base universe, or the modeled one, and without a way to determine which, we have a 50% chance of being in other. I see the main argument as this but scaled with a few assumptions, the biggest of which and the one that really allows for exponential growth, being that the root universe need not have the same constraints as the simulated universe.
I'm not totally sold on the simulation scenario idea being true, as there's a lot of assumptions that could break down and make it impossible, but as of yet I'm not sure I've seen any evidence suggesting it's impossible at any scale. Then again, I'm treating the chance that we could be in a simulation as distinct from the likelihood that we are in a simulation, so I might not even be arguing the same thing (although I don't think the difference really matters).
Infinity is only an abstract mathematical concept. It does not exist in reality.
http://idioms.thefreedictionary.com/chances+are
When used in a phrase it means highly likely. I'm not talking about probability here, rather his use of the idiom.
The use of an idiom doesn't change the definition of the words.
Probability is not causal. Just because something is "highly likely" does not mean it is true.
Then you don't understand what an idiom is.
"The ball is in your court" is an idiom. It doesn't mean you have a ball or a court.
When an American says "chances are" he's not talking about probability. The closest replacement for a non-American would be 'probably'.
Yes, he is. "Chances are" (or, equivalently, what the headline actually says, which is "odds are") is an idiom that means "it is more likely than not", or, in probabilistic terms, "there exists a probability of greater than 50% that..."
More to the point of arguing about the meaning of an idiom, in this case, Musk has actually been a lot more specific and quoted a specific odds ratio of 1 billion to 1, or approximately 99.9999999%.
We're broadcasting these ideas, not because they're at all new, or interesting (they've been around for about 400 years, at least). We're broadcasting them because Elon Musk is saying them. His only new idea, AFAICT, was asserting some probability that he pulled out of his ass that this idea is false.
In the absence of new ideas, maybe we weight this by the authority of the speaker. If a theoretical physicist said this stuff I'd be more inclined to take some kind of notice (since they'd have a deeper insight into the structure of the universe than I do). A car salesman - not so much.
Anyway his thoughts aren't based on the Matrix. This is a well known idea called the Simulation Argument, invented by Nick Bostrom (who Elon Musk is a known fan of): https://en.wikipedia.org/wiki/Simulation_hypothesis
You might agree or disagree with Bostrom's paper, but one of those is a high-schooler/tripper/moviemaker's _idea_, and one is a scientific paper, which is meant to be repeatable, experimental, etc.
https://www.youtube.com/watch?v=nnl6nY8YKHs
Even if it is true, it changes nothing. Since this is the world we have to deal with and there is no indication we can influence the (supposed) simulation. If our world/universe is a simulation, it is simply too good.
I also recommend to read "The Big Picture" by Sean Carrol, chapter 11 "Is It Okay to Doubt Everything?" deals with these kinds of thought experiments (including Bostrom).
The 2016 Isaac Asimov Memorial Debate was about "Is the Universe a Simulation?":
If God - as understood by major human religions - does exist, then the world IS a simulation. I mean, all atoms and particles are "real", but ultimately they form a simulation device, one that invokes the sense of existence in sentient beings.
There is a reason to believe in the simulation hypothesis. We already have a massive industry devoted to making world simulations, video games. We know our computing technology advances steadily every year. So it's not unreasonable to suppose it's likely we will someday create such simulations.
The second belief is also untestable, it's a prediction about the future. But a prediction about the future seems much more grounded than religion. It's taking into account our observations of the world, and what future outcomes seem probable. Not just making stuff up.
I personally don't see much difference in the two. Before modern astronomy, people looked at the frequency of the revolution of visible planets and thought their ratios resembled musical ratios, and that this had to be the work of someone (such was the belief of Pythagoreans, Stoics and Neoplatonists). Nowadays people look at some other physical quantity and think that this has to be the work of someone. In this sense, they're pretty similar. And if the error of those ancient philosophers lies in that supposition, I don't see how the degree of reality of our idea of who's doing it (e.g. aliens vs. a hazy concept of a "Demiurge") makes it any better.
Clearly we know a lot more than the ancients do. E.g. the concept of a "simulation". And so we can make more accurate predictions than they would have been able to.
I don't see an issue with believing things that might be considered incorrect in the future, but I do see an issue with asking the wrong questions over and over again.
I see no particular reason why someone ought to be responsible for the patterns of experience. Thinking it so is the crucial mistake here.
This is certainly not the case nor the intention with video games (from which the whole argument is extrapolated), no matter how realistic they can get
I haven't watched the videos yet, so apologies if this is covered. Looking forward to seeing them.
If it is true, it changes everything. If you're simply a character in my simulation and not a real human, then there's no reason why I shouldn't screw "you" over to maximize my personal enjoyment.
That's the most actionable point of the thought experiment, to illustrate the importance ethics.
The Simulation Argument (to the extent it applies to anything at all) applies at least as well to solipsistic simulations where all that is simulated is one observer and their personal observable universe.
While the number of simulated universes may vastly the exceed the size of the real universe, the size of these universes may diminish faster than the increased number can make up for. Bostrom's argument is that we're likely in an ancestor simulation, but if the size of the simulated universes decreases faster than the number of simulations increase, it could easily be the case that the the probability that we're a simulated person is lower than the probability that we're a real person: real people may vastly exceed the number of simulated people.
Thinking about it for a few minutes, we would have to do at least one thing to make it work:
1. Make it more granular, so our sub-subatomic storage could record the states of atoms without being as large or larger than the stored data.
2. Make it slower, so our faster-than-light computer could have time to simulate all interactions.
3. Make it simpler, so our complicated computer mechanism operating on some exotic principle wouldn't have to simulate the same principle.
4. Make it smaller than our computer. It's easy even now to simulate a few atoms with a computer consisting of several kilos of matter, but I can hardly imagine any universe devoting a significant fraction of its mass to ancestor simulations.
These restrictions would eventually limit the ability of child universes to simulate grandchildren, but I am not convinced that it significantly reduces the number of sentient beings (simulated and non-simulated) which is the key to our estimate of whether or not we are in a simulation. If in the parent simulation, room-temperature superconductors are common, cold fusion is easy, there is no speed of light in the parent universe, Landauer's Limit does not hold, and computers are built on a scale far below quantum mechanics, our (apparently) large universe could easily be one of trillions of ancestor simulations.
[1]: https://en.wikipedia.org/wiki/Landauer%27s_principle [2]: https://en.wikipedia.org/wiki/Planck_time
... presuming that the outer simulator consists of fundamental laws that support more than one identity.
heh.
P.S: I understand it is a balance but Musk has done nothing to warrant this derision...yet.
A lot of life is, until it's too late and we're speaking in terms of hindsight. If you want to believe in something, at least believe in something that's likely to be true.
To be fair these engineers are taking a lower salary but are getting paid really well in terms of stock options.
Elon Musk is as close to doing those things as anyone has ever been in history. What else is he supposed to do to appease you?
Someone had to say Elon is Saving the World.
Of course that's ridiculous but we put guys like Gates and Musk on a pedestal
and then we're being so edgy when we say they're not really that great.
He's doing some interesting things and saying some interesting things.
I don't think he's playing a chess game. I think he was having fun.
A person asked him a question outside of his expertise and he answered it.
The boring thing to do is say "I don't know ask an expert".
Of course people are only paying attention because he's a celebrity Intellect.
So I'm going to disregard him on this one.
If this is a purposeful simulation done by an advanced intelligence, then it is a sick puppy.
If he means this is an accidental simulation, ie. a projection of 4D from a 3D plane, or a 4D shadow from a 5D plane, then maybe, just maybe.
The media doesn't report much on relatively obscure philosophical discussions, but they do report on what celebrities talk about.
1) This idea makes me think that the simulation hypothesis is wrong. If some advanced civilization has the power to create simulations, why would they create us, specifically? Why would they bother recreating a primitive world of the past? Wouldn't it be way more interesting to create cooler, crazier worlds? Do you really believe a post-human ultra-intelligent civilization would create something as present-day ISIS? Is someone having fun watching how we experience this world's horrors and pains?
2) Then I think the following: yes, the above scenario is possible, given that organized intelligence is able to realize more possibilities of the universal computational space. With enough computing power, organized intelligence would be able to spawn near-infinite world simulations, and ours is just one of those.
I could go on - at length. And this only scratches the surface of alternative world history, which is just one of a crazy number of reasons to simulate something that involves a previous history's peoples.
Hah, I wonder what geologists would get up to?
How is the (let's say, quite likely) hypothesis that eventually we're going to create VR that's ingistinguishable from actual reality supposed to logically lead to the notion that there's a billion to one chance we're living in VR right now?
It feels like something along the lines of the "???? PROFIT!!!!" meme (originated from a South Park episode). Clearly a crucial step is missing
That's a very simple version of the argument with obvious flaws, but that's what he meant. Nick Bostrom has made a much more elaborate version of the argument.
So, just like a physic's simulation that uses limited means, say like floating points, exception cases need to be taken care of (like detecting resting-cases in physics and then clamping floats to a small epsilon in order to avoid weird reactions).
The treatment of these exception cases would add discontinuities in the physical behaviours we observe. For example, in a simulation, a force like gravity would be explained as: F = G (m1m2/r2), unless r is very close to 0 in which case F=0. In our reality, the "unless" part does not exist, in fact, if you have this exception, it is actually a symptom that your formula or approach might not be the perfect one.
In our reality there are no glitches nor can we detect means by which glitches where avoided, in fact it is quite the opposite, metaphysics seem to exhibit an undeniable amount of coherency and consistency.
I'm not convinced by the simulation argument at all.
True - but a perhaps a clever simulation could concentrate it's computing near the observers and approximate distant or non-observed events.
But more importantly, the laws of physics would be observably non consistent.
I didn't go this far, so I'll have to think more about it.
thanks for the food for thought.
Isn't that what's happening during our observation of the galactic rotation curve?
Would you expect the glitches to appear as random events if you look too closely?
'Cos until you wrote this I didn't think it was likely that we were in a sim. Now you've made me question that...
In order to avoid these glitches we need to clamp these numbers to values that are very close to zero and basically stop simulating the objects concerned until we detect that these objects move again.
Although this is an example, an imperfect process would have an upper and lower limit of numerical representation, requiring patches to limit glitches.
patches would be detectable in a simulation.
As limits. We have those, or at least our accepted view of physics has both lower and upper bounds on what we can know. How are they that different from what you're suggesting?
It's about limits or bounds that are incongruent with how physics normaly works in reality (patches would always do that), for lack of a better expression, I call these metaphysical inconsistencies.
Only a complete simulation of a universe with an infinite number of the smallest components would require infinite means.
- because a complete simulation would require infinite means
- simulations can only be incomplete
- incomplete simulations would exhibit metaphysical irregularities
- yet to this day, metaphysics seems coherent and consistent
- therefore we are probably not in a simulation
Therefore, it could be a simulation. Or not. There's no way to tell the difference.
If the Milky Way were to re-arrange itself to say Game Over, I think I'd be inclined to go for sim.
We laugh now at the 19th century man unable to accept that "man evolved from a monkey", but his case more charitably stated is the reluctance to give up the dignity of being made in his creator's image.
Likewise, I think people who build the 21st century secular myths are setting themselves up to be "disappointed" by the actual data when comes in barren in the next two centuries - telescopes that can measure exo-planet atmospheres basically confirm or deny.
I'm not taking either side, but if the "fact" turns out to be that we are alone, not anybody's simulation, not part of multi-verse, etc, this seems to be reality that modern scientism would have the most problem accepting.
A simulation is just as real as the reality that hosts it, and it only differs in it's fidelity to the inspiration. The same can be said about fractals.
Let's assume we do live in a computer simulation. Then what? The fact that this knowledge doesn't change a single thing about the way we should do things means that it's not particularly interesting or useful. The same is true regarding the debate of free-will.
Certainly, no matter how far you go back, the simulations are "real" in the sense that they exist. Why is this simulated reality considered less "real?" Would we (or whatever we really are) be able to experience "base reality" as he puts it? Can you see your own eyeball directly?
Uh... following this precisely 10,000 years from now would at 1/1000 of our current advancement rate would look like 2026 (assuming our current rate doesn't slow down). We don't really have photorealistic graphics now and I'm not convinced that we'll have them in 10 years. Even if it is photorealistic, it'll just barely look that way on the macroscopic scale. Now simulate all the particles in the universe.
>Musk said, the odds that we are living our lives in base reality — that is, "real" reality — is one in billions.
Musk just presents an interesting argument and asks "can you find the flaw." Sure. The flaw is that there's absolutely no reason to think that:
...computers are going to speed up by 40 orders of magnitude, ever
...and there will be planet sized computers
...and civilizations will be motivated to spent their planet sized computing resources on simulations
The whole thing is absurd.
Only if advancement is linear.
And the numbers I gave only allow you to simulate roughly in realtime, so... your civilization has to stick around for another 4.5 billion years to get to us.
Programmers don't really do formal proofing or enough unit testing after all.
Either way, I reckon this will one day be used to generate an infinite amount of behaviourally customized cat videos for our simulators' facebook feeds.
https://en.wikipedia.org/wiki/White_Christmas_(Black_Mirror)
Serious question: any practiced religions that are built around that sort of idea?
What would that even mean?
The simulation argument is connected with Cartesian philosophy though. It is one of the possible forms of the "Cartesian demon" argument [0]. Descartes tried prove the truthfulness of experience through a twofold proof: a proof of God's existence; and a proof of His unwillingness to deceive. He does all this in Meditation III [1].
[0]: https://en.wikipedia.org/wiki/Evil_demon
[1]: https://en.wikipedia.org/wiki/Meditations_on_First_Philosoph...
No, he's saying that his speculation about the future quality of our VR is proof that simulations of the quality of our universe are possible, and that if those simulations are possible then, our universe probably is such a simulation, because obviously anything that can be done has been done by someone else first.