Elon Musk feels we live in a "simulation" -- if so, should we Occupy Mars?

Topic: 

Elon Musk likes to say pretty controversial things off the cuff, and so do I, but he inspired a number of threads by saying at Re:Code that there's a billion to one chance we're living in base reality. To use his word, this world is almost surely a "simulation."

While a lot of press attributed the idea to him, Musk is actually restating almost exactly the well known thesis of Nick Bostrom on this topic, which has spawned much debate (some of which can be seen at the site linked.) The short precis of the thesis is as follows:

If you accept that the eventual progression of our work in creating digital (or "simulated") worlds is to make ones that match our reality, then you probably accept that once we can do this, we will do it a whole lot, and that eventually there will be very large numbers of created digital worlds, many based on our own. If that's true, then the probability that any particular world (including this one, of course) is the original one is vanishingly small.

Like many, I find the argument interesting, though not quite so compelling, as it contains some logical fallacies. For one, even in the "root" universe, the argument is equally compelling, but also clearly false.

I also oppose the term "simulation." For far too many, "simulated" means "not real" or "less real." This world is clearly "real" even if it is synthetic and based on computation. If you accept the truth of "I think therefore I am," then you are thinking, not engaging in a simulation of thinking. (Just as AlphaGo doesn't simulate playing Go, it plays Go.)

Better terms include "Computational" and "Snythetic" or other synonyms like "digital," "emulated," or "artificial."

Leaving aside the debate over the merits of the argument, let's assume it's true for the moment. The biggest consequence of synthetic is that it means created. As in, "there is a creator/god" in the sense of a being who created this universe and who is in some limited way omnipotent over it and in another limited way omniscient about it. I say a limited way, because this "god" is perhaps a programmer named Martha who has a few hundred digital Earths running in her dorm room. A being perhaps (but not surely) exactly like us in her world, but with the potential ability to observe and change anything about this one.

That is a theistic view, though quite unlike typical theist doctrines. (It bears a small and bizarre similarity to Mormon theology which teaches that our god was once an ordinary being on another world who was rewarded with his own new world to be god of.)

From what we can observe, Martha doesn't interfere overtly with this world. As such, the first conclusion is that even if you believe in this, it should not change very much about how you live your life. If you have no shot at interaction with the "parent" universe, and there is always the chance this whole thesis is false, you should go about being you as though you felt you lived in the root or "first" universe -- what you might incorrectly call the "real" one.

There are some changes that are justified if you believe this, though. They are grand philosophical changes, but some apply to Elon Musk himself.

You see, Elon has made it his prime life goal to get humanity off the Earth. To stop us from being a "one planet species" which would be wiped out if something catastrophic happens here. History shows that bad things have happened naturally (like asteroid strikes) and more bad things could happen due to the works of humanity, like killer diseases or nuclear winters. As such, Elon's goal of getting a self-supporting colony on Mars is a grand one, well worthy of being a prime life-goal for a world-shaker.

But it's taken down a peg if you accept the synthetic world hypothesis. Now, you conclude it's very likely that this is very much not the only cradle of humanity. That there are probably millions or billions of them. That even this one quite probably has backups taken every so often, so that even if we wipe ourselves out, all can be preserved and even restarted, if Martha wants to.

We don't know anything about Martha's motives, other than she appears to not do any noticeable interference. Martha might not even be remotely human, though once again, the probability is (at least from our viewpoint) that beings would create more synthetic worlds like their own than entirely different experiments. But if you believe in Martha than you believe we are not alone and that alters goals about the future of humanity.

If you want to get more extreme, there is also an issue with Mars. While again, we have no information on Martha's goals for this project, it seems likely, unless resources are truly free, that most synthetic worlds will be just the surface of the Earth, just the interesting part in question. Running an entire galaxy or an entire universe is many orders of magnitude more costly. Sure, you might run some of them, but if you can run a trillion Earths for the cost of a couple of galaxies, that's gotta bend things a bit.

As such, the rest of the universe truly is "simulated" in that it's just being computed with barely enough resources to make the few photons which reach us be realistic. (Or it's just a playback of an earlier run.) Many fans of this theory like that it explains Fermi's famous paradox -- no aliens have visited because there are not any -- in this universe.

It's hard to imagine, unless computation is totally free, that there would not be any "optimization" of the computation. Now, at the extreme, this would mean the parts of your house that nobody is looking at would be computed at a lower resolution, and that indeed, if a tree fell in the forest and nobody was there to hear it, it truly would not make a sound in a full way. That's very philosophically spooky, but less spooky is the idea that until we went to it, Mars the planet would not even be "booted up" into our universe. When probes arrived, it might have been fully started, but more likely only where the probes went -- the rest would just be a recorded copy of the original Mars, presuming there is such a thing in Martha's world, as the whole sky would be.

As such, it would mean going to a place that only "fully exists" (which is to say is being computed at full resolution) because we went there. Somewhat less satisfying.

Still worth going?

People imagining the idea of a synthetic, computed Earth do like to speculate about the motives of its creation. If Martha is just like us, then they probably have rules and ethics about doing this. There are huge ethical questions about all the suffering and evil that comes with creating a universe. One rule I've imagined is that the creator really has some duties to the people inside. Those might include having a heaven of some sorts, or even letting people graduate up to the parent universe and gaining rights there. The most impressive might even get to chat with Martha, though she only has time for a few. Perhaps somebody who does something truly great, like taking humanity off-planet, gets some reward for it. We can suppose this because we might do something like that if we were making these computed places. But we really have no evidence for any of that. Some would argue there is almost nothing ethical about creating a world with so much misery and keeping the inhabitants in the dark about the reality to boot. At least by our standards -- not theirs.

Is there a root?

One popular theme is to suggest that Martha's universe is also synthetic, and there is another creator above her. I describe this by saying, "It's turtles all the way up." Nobody can truly be sure they are in the root of the tree.

This is particularly interesting if you speculate that the rules of our universe, when we finally learn them in depth, will show that computation lies at the bottom of everything. This has often been speculated, and most of the quests for a unified "theory-o'-everything" tend to try to express the rules in simpler and simpler mathematics. People care about that because for now, this theory is based on the idea that computation is being used to simulate the physics of a "real" universe, one made of particles and forces. We are only able to see the particles and forces, and so might conclude we aren't digital. Particularly when emulating the activity of subatomic particles is today very expensive computationally. It makes implementing a synthetic universe at the deep level seem impossible to us. If there are deeper rules that are computational, then you can also postulate that the "root" universe could also be computational. In fact, you sort of need that, because it's hard to figure out how to get the resources needed to have worlds within worlds if you have to implement particles based on computation done with particles which are based on computation and so on. You quickly run out. If, on the other hand, you are in a universe of computation and you create sub-worlds, you can just give those sub-worlds access to the computation substrate of your own, and it scales a lot better.

We like to believe our universe is made of particles which are physical and bounce off one another and follow analog rules. But we don't know that's true. The rules of our universe are a mystery to us. We don't know where they came from, and we can't even declare that whatever they are, any parent or root universe might not run on the same rules or a variant of them.

So should you believe this is a synthetic, computational universe -- or simulation if you insist? Well, you can, but unless you are leading a mission to Mars it is not greatly productive. When the time comes -- as it will -- that we make our own small digital worlds that match our own for reality, doubt will of course increase, but as long as Martha remains hands-off, live your life as you always would have. One of the more spooky ideas in this theory is Last Thursdayism -- the idea that there is no way to tell this world wasn't forked from a backup last Thursday, that all of your memories before then happened to a predecessor. Perhaps that's true, but again it doesn't alter how you should spend your days. Indeed, it is not my goal to convince Elon to abandon his quest for Mars at all; that's worthy even if it doesn't help save humanity.

Unless the creator wants to show up. Martha?

Comments

I'm in agreement with you - I think a lot of people who speculate about our universe being simulated underestimate the efficiency losses of multiple layers of emulation.

However, there's another aspect of this puzzle that I think is interesting, which is, what kinds of things can Martha actually do?

First off, if Martha is simulating an entire universe, she might not even be aware that we exist. We're a very small needle in a very large haystack. Now, one can imagine that there's some sort of "intelligent life detector" built into the operating system that looks for signs of intelligent life or a technological civilization, but this assumes that you know what it is that you are looking for and can distinguish it from signals produced by natural phenomena or other non-intelligent processes. For example, one might make a detector for electromagnetic signals that contain a certain degree of entropy, but this assumes that (a) all or most technological civilizations would produce such signals, and (b) that natural phenomena producing similar signals (false positives) would be less probable than technological civilizations. Given the diversity and strangeness and size of the universe, I think it's parochial to assume that other intelligent species would be recognizable by the same features with which we recognize ourselves.

Note that even if Martha is only simulating one planet, if we assume that intelligent life is vanishingly rare (which appears to be the case), then she still has to simulate a lot of instances of that planet before she gets a 'hit'. So the question then becomes, what does a 'hit' consist of, and how is it recognized?

The other question is: what kinds of "miracles" can Martha create? Moving mountains is trivial: just take a bunch of particles in a defined zone of space and offset their X-coordinates by some factor. But there are other things that are algorithmically far more difficult, even (I think) compared to the task of simulating a universe. An example would be something like biological metamorphosis (such as a werewolf), or instantly transforming someone from an Atheist into a Buddhist without changing other aspects of their personality - because the process which determines physical anatomy or mental conviction is a one-way function whose output emerges over time. Simulating a universe takes enormous computational power, but the actual algorithms are quite simple - whereas scanning a network of neurons and trying to extract some sort of meaning out of it requires far more sophistication.

In other words, even if you possess nigh-infinite computational power, you are still limited by your own ability to comprehend complex software.

Well, the ability to run a planet comes a long time before the ability to boot up a whole universe, and even then, doing individual planets is probably far more interesting than experiments with whole universes. And doing alternate versions of your own planet even more interesting. So I predict it's probably there's a lot more of that than whole universes.

If you have the power to do whole universes, I think you have a lot of pretty powerful tools and AI capabilities at your disposal. Certainly enough to survey your universe for interesting activity. And probably, if you've been at this long enough, tools to give you lots of godlike powers in your datasets, though not the true combination of omniscient superintelligence of a classical diety.

Martha may have spun up this planet last Thursday, or at the dawn of humanity, or the dawn of civilization or many other fork points. Depends what she wants to study and play with. While those who start with a blank slate have more options, once somebody's blank slate produces something interesting, I presume everybody will start forking that and experimenting with it. Especially if it takes lots of time to get from a big bang to a full interesting universe. After all, rarely does one write your OS from scratch, you build on interesting prior work.

In a synthetic world there is no need to simulate particles unless someone in the synthetic world is actually trying to see them. It's the same as your argument about a tree falling in the forest or empty rooms in a house. Newtonian mechanics is sufficient for 99.99999% of the computations. Perhaps that's a clue that it is a synthetic world because there is a simple theory that's relatively easy to implement on the computer.

Of course, I don't believe this is a digital world. But that doesn't mean that there isn't a higher order world behind all this. It just may be a different kind and of a different order. The question you need to ask yourself is whether, throughout history, there are any incursions from the higher order world into our world? I think the evidence on that is pretty clear and non-negative. Perhaps Martha has a real name.

You don't need to compute at full resolution that which won't affect any of the other things (ie. people, probably) that you are interested in. But because of the chaos of the universe, that's actually a high bar. Quantum events do have macroscopic consequences, though they are of course hard to see without knowing what they are. However, when you talk about Mars, a recording of the photons from an earlier Mars is going to suffice while nobody is going there.

It is not absolutely necessary to call it "digital" in that we don't actually know by what means a synthetic reality would be made. There are other theories about how to do it. It's just that the only thing we humans today have imagined is virtual worlds inside a computer. So I am preferring "computed" to "digital" but even that presumes too much.

Of course most religions believe our world is an artifact of a deity. The big difference of this concept is that the "deity" is possibly somebody just like us, simply using more advanced tools than we know, which is why I conjure the imaginary Martha. I don't know of any significant religion where the deity is not superhuman.

While there seem to be quantum effects people have gotten by without worrying about them for thousands of years. I don't see any way synthetic computation would be able to include quantum effects of every particle in the synthesized world. But, if not, how can the computer program determine when quantum effects need to be taken into account when they don't, it seems like that means the program would still have to look at a lot of particles to make that determination unless it uses the no one is looking or listening approach. Perhaps this is the crux of the explanation of the role the observer plays in QM. Is the cat in the box dead or alive? You have to look, which means, the program has to do some QM calculation to get the answer.

So perhaps Martha is starting to have high electric bills because more and more scientists here are look into the quantum world so Martha's computer(s) are going to have to deal with more and more particles and quantum effects. If her electric bill gets too high she may have to pull the plug, unless of course she has unlimited resources somehow, but if she is like us, she doesn't.

"For one, even in the “root” universe, the argument is equally compelling, but also clearly false."

This is not really a good argument against the thesis. To take a simple example, I can claim that with 95 per cent confidence I am not #1 when numbers are assigned randomly to 20 individuals. By construction, this will be true for exactly one person, though the argument applies to this person as well. It is unlikely to win the lottery, but someone wins every week.

"there’s a billion to one chance we’re living in base reality."

Should be the other way around!

Note that once you accept the possibility of being in a simulation, you cannot calculate the probability of being in the "root world", since this assumes that the physics of the simulation is the same as the physics in the root world, which is not necessarily the case.

The argument does not calculate the probability as a number, simply that it is very small. As such, it does not depend on any assumptions about the physics of either world.

There is an intuitive argument that more resources would go into universes that are "interesting" and "like the parent" based on the fact that we would probably do that. Of course, we have no idea if the parent universe is people like us. Since you should be able to run a billion Earths for the cost of running a full galaxy -- presuming there is a cost to computation, it makes it more likely you would be in a "just Earth" universe.

Of course, our physics might be much simpler than the parent. The first digital worlds we made were cellular automata like Conway's "Life." It runs on an extremely simple physics, and yet is turing complete and capable of great complexity.

Two comments:
I think your riskiest assumptions are that Martha has limited energy and would spend the most time studying universes similar to hers. One of my favorite "discussions" is to guess what will happen when we have virtually limitless energy. Fusion? Anti-matter?

And my wife's name is Martha, and I know she would not be interested in doing any of this. :-)

Peace,
Randy

Since we know almost nothing about Martha and her universe, all assumptions are risky. But what this argument tries to do is decide what's most probable. Bostrom's argument doesn't say the world is synthetic, it just declares it likely.

So here are some assumptions on probability, all of which could be wrong, but...

  1. We would be likely to create more Earths than places with random physics. We are very narcissistic.
  2. If a world is of type X, there is a higher probability the parent is also of type X

Of course, there will be lot of type X worlds making experimental worlds of type Y. It's just a supposition about what the most likely situation might be. It could be entirely wrong. All we know is what we are like.

You need truly limitless resources to run around doing a lot of full galaxies or universes. Because, virtually limitless is not the same as limitless. With anything other than limitless, the reality is you have to choose between a billion planets or one whole galaxy, so the odds are very high you have many more solo planets than galaxies, and thus the odds that you are in a solo planet universe are high.

Harking back to Arthur C. Clarke's prescient "I Remember Babylon", and the current state of the internet, surely most resources would be used to simulate sexually active individuals. :-)

More seriously, if one is interested in "human" behaviour, why not take the "don't boot up Mars until someone gets there" to the extreme and simulate just brains, i.e. simulate (in the simulation) all the input to the (simulated) brain?

Again, I don't feel simulated is the right word, but indeed, there is much speculation of just how much a computed world designed for humans would optimize, to the point, as I said, of not computing at full resolution the things that nobody is looking at. One can debate if that's true or not, but it's hard to imagine you would fully compute Mars before the first probe lands on it, just to send a few pixels to digital humans on Earth.

In wrestling with this concept over the past few weeks, I've vacillated between belief/disbelief about life as a synthetic reality, and between optimism/pessimism about the resulting outlook for the humanity we know. Presently, I'm trending a bit towards belief & pessimism, and there are two things that have put me there:

1) Does it really seem likely that a sim-creator has a moral obligation to their simulated consciousnesses? That may assume the sim-creator feels an unlikely level of affinity towards our present level of consciousness. There are analogies here to our own existence: Surely, we do feel moral obligations when we do animal testing - on chimpanzees, or even rabbits, because they are relatively similar to us. But how many people would expend the effort to give an afterlife for every ant that dies in their ant farm? My guess is very few, because we perceive ants as so far removed from what we consider an ethically-relevant form of consciousness that their death isn't something we feel overly guilty about. We shy away from needlessly killing them, certainly, but we don't consider ants meaningfully conscious or intelligent, and it might take another lucky billion-ish years for them to evolve to an intelligence that we would consider cruel to terminate without the great deliberation we afford our own. And so back to our sim-creators: even if they are far more ethically nuanced than us, they likely also see themselves as so evolved (& their though processes so much faster) that our meager consciousness seems as distant as the give-it-a-billion-years-and-it-will-be-something-worth-watching ant's consciousness does to us. This gets even more likely if there is truly some form of "Great Filter" that the sim-creator has luckily passed - because they might have the knowledge that even among space-faring consciousnesses, only 1 in 10*XYZ ever even make it past that filter, and thus even advanced civilization simulations fall in the ethically-grey category of "they were doomed to fail before achieving anything, anyway".

2) What if we aren't even remotely associated with the purpose of our simulation? Knowing that my consciousness is a fabricated piece of a "humanity sim" is mildly depressing; slightly more depressing would be knowing that my consciousness is just an equation-balancing setpiece of "elephant extinction sim" or "galactic resource allocation sim" - meaning - not directly the desired outcome of the simulation, but tangentially-related enough to still require computing in full. More concerning than either of those scenarios, though, is one where my consciousness is a completely irrelevant (& maybe unnoticed) emergent side-effect of some huge, complex, abstract calculation being performed by a superintelligent future race. They aren't looking for a simulation of the galaxy at all - they are just trying to crunch numbers on how long it takes a universe-sized chunk of energy to cool down, and our existence is just the unanticipated mirage of order inside the CPU of their big chaotic computer - civilization-scale Karman vortices & Boltzmann brains... ....not even as meaningful as undesired algae growing on the walls of an experimental fish tank, and at best having the significance of that serendipitous coincidence where from a certain vantage point, in specific lighting, and for a few split seconds, one cloud in the sky looked just like the entire observable universe.

(This circumstance being true would further reduce the ethical burden of a sim-creator as discussed in #1 above, in much the same way that a commercial fisherman most certainly feels less guilt when they unintentionally kill some bycatch in their nets than when an aquarium-keeper unintentionally kills a similar fish in their own tank. Ethical burdens routinely show as easier to ignore when they only present themselves as unrelated side-effects of some grander endeavor.)

Religions have pondered the idea of the world having been created by some sort of being/deity for a very long time. And more recently, with the rise of atheism, the idea of the universe "just existing" and just happening to have evolved us. These are the two most common scenarios in philosophy today, and are they any more bleak or pleasant than the scenarios you describe? The "just a side effect" scenario is not different from the "universe has no design or purpose" that most atheists (including myself) hold to, at least in terms of how you live your life.

One key difference from synthetic world theory and deism is that in deism the creator is some special supernatural being, usually of immense power and wisdom, able to have a plan for this universe which we might not understand but imagine is "higher" in some way. In synthetic world theory, the creator need not be any more than we are. (After all, the theory arises from the idea that we are on the cusp of being able to do this.) Of course it doesn't demand such a creator -- the creator could also be superhuman (though not supernatural.) The creator may or may not have a plan, but for now that plan does not manifest itself in what we observe.

Now if the creator is Martha -- a human not too different from us -- then she probably does care about other humans created and her society may have an ethics about these experiments. I think we would have such an ethics when we get around to doing this.

Definitely - guessing the moral system of a creator puts us solidly in the domain of religion. The rabbit hole of creator ethics is so deep and strange that virtually all possibilities can be found there, and it may be overly optimistic and anthropocentric to suppose that their value system is a higher-but-still-recognizable form of ours. Even if they recently evolved from us, it may be that becoming a post-biological species will change our morality rapidly and in ways that are foreign to us now. Most religions pegged their creator ethics to whatever the ruling class of their century envisioned as conducive to their idea of utopia; us now doing the same thing in our current era seems... unsatisfying. In that light, the most defensible position we are able to take may indeed be the admission that the moral system of any creator (should there be one) is to us inscrutable and indistinguishable from a state of true chaos.

In the more specific comparison between "this universe has no design or purpose" & "this synthetic universe is just a side effect", an existential optimist would believe a sufficiently advanced civilization could transcend the chaos (or the simulation) to create meaning in some way... the only distinction the latter presents is that there is an additional "kill-switch in the control room", so to speak, where need to not only transcend our meaningless existence, but also do so before the whole thing gets shut down for reasons that might be completely orthogonal to anything we'd ever be able to anticipate. Obviously this doesn't change the overall nature of our struggle, but it does make it one level more difficult.

Remember, the Bostrum argument is based on what is the most probable situation. The argument is that the the child universe being similar to the parent is the most probable, since it seems that's what we would probably do. But of course it's entirely the case that this will be entirely wrong in many universes.

However, it is also true that theology hasn't spent a lot of time on the idea of the creator being another human, not "higher" in any way. There are religions which have limited and flawed gods, most notably the pantheistic religions, but the idea of a single creator (sort of) who is no more than human is a bit novel to theology.

I say "sort of" because of course the hypothetical Martha did not create the Earth on her own. She pushed the create button, but using vast libraries developed by her civilization.

Would you believe the simulation story if it were told by Warren Buffet? Bill Gates? Charles Koch? What makes Elon Musk, or any other person, more credible on this topic?

Really, this is why we have Occam's Razor. Once you've thrown that out, you're free to debate whether Martha's simulation also includes a large number of angels dancing on the head of every pinhead. After all, in a simulation with infinite compute power, angels don't come at any cost, and the universe being infinite, anything that is imaginable must be true, right?

Neurophysiological experiments have shown that the brain can be made to trigger an "experience of the presence of God" in some people. If your reason can't allow you to believe in traditional religion, yet your brain is forcing you to believe in some higher power, then a simulated universe provides an escape route to resolve that conflict.

The idea is not Elon's -- it's been around for 12 years or more. He just gave it some more attention.

And yes, Occam's razor suggests not to believe it. But Occam's razor is just a guideline, the idea that the simpler is more likely. It doesn't prove or disprove anything any more than Bostrom's argument does

Add new comment