It seems likely that when a simulation turns out to be especially interesting or unusual due to natural randomness, its simulator will decide to roll out that simulation more times. This wouldn't look like a measurable violation of the laws of physics from the perspective of the simulatees, as long as the simulators don't abuse their resampling power in extremely obvious ways.
Unpredictable simulation rollouts are more scientifically interesting, informative, and fun to observe. So simulatees can expect to live in an unpredictable world, whether their simulation is used for science, prediction, or entertainment.
For example, if someone simulated the entire universe millions of times and finally got a planet with life on it, surely they'd want to roll out that timeline many times, rather than throwing it away and starting from scratch.
Much more speculatively, you could imagine that in a civilization of aliens playing "Earth: The Video Game," one of the aliens happens to play out a timeline where Trump becomes president, they think it's so hilarious they share their checkpoint all over the alien Internet, and it becomes one of the most popular checkpoints to start a game from.
I don't necessarily think worlds where Trump becomes president are the particular kind of world that the simulators would prioritize running. But maybe? In any case, Trump's election hasn't noticeably affected my credence of being simulated (which is pretty dominated by my low starting prior).
In that case, I’d expect to be living in the early 20th century. Would you not say that WW1 & WW2 were especially random, unpredictable, and interesting? With the exception of AI development, I feel this time period was relatively more stable
I don't know, personally I'm much more interested in the ongoing history of AI development than WW1 and WW2, but I guess those wars are also pretty interesting.
If we're indeed living in the hinge of history, and humanity is about to create AIs that spread certain values throughout the universe, it seems very interesting and important to find out what the outcome is. I hear this is especially relevant for acausal trade.
Chains of arguments like these are why the scientific method demands that one makes a claim which is capable of being falsified. If the outcome of a failed test is capable of being the same as the outcome of a successful test, you've engineered a bad test. I believe this is the gist of the argument here?
The argument here is tautological.
You are saying that the state of the world as you see it is what a perfectly normal world would look like and so if a world looks that way even if simulated it must be being simulated to be indistinguishably normal. But it only looks normal because it's your only reference frame.
If an AI agent were 'born' into Minecraft they might make a similar argument that because everything is made of giant blocks — just like 'normal' — that they must be in a sim of a realistic world.
But we can consider that we have discovered over the years a great many things that were very much not considered 'normal' about our world until discovered… and then normalized.
It was a great surprise to many when QM was discovered. One might think that the world's most famous physicist at the time debating whether or not the moon was there if no one was looking at it is exactly the kind of surprising thing that would occur in a simulation. But then the people responsible adopted the phrase "shut up and calculate."
So Bell's paradox about how maybe things aren't real until stateful, the Frauchiger-Renner paradox where maybe relative frames have sync errors, the way the black hole information paradox has led to calculating how closed universes would have a one dimensional Hilbert space without interior relative observer frames — we have several things even in just the last few years (since the formulation of the Simulation Hypothesis even) that have been measured which were quite surprising to people when discovered and have now become normal to our universe, but also happen to be similar to the behavior one might expect of a simulated universe.
Also, I'm skeptical of how heavily you should be weighting your "well obviously the sim should behave as X." This framing seems to adopt the Baysean weakness of hand waving away unknown unknowns. Just because you, at this moment, can't see other likely possible reasons for a realistic sim to be running doesn't mean not to leave room for them in one's reasoning.
For example, an especially surprising outcome to many after the point the Simulation Hypothesis was proposed was finding themselves in a timeline where AI is at its current trending state. My guess is that in 2003 the majority of domain experts would have had an avg of single digit percent estimates for this outcome taking place within 25 years.
So, relative to the widely held expert world models before each were discovered to be true: we find ourselves in an abnormal timeline of AI capabilities in an abnormal universe where attention/interactions apparently turns probabilistic distributions into stateful shared values for overlapping relative frames without which the universe would be empty of information.
TL;DR: The argument that clearly the simulation hypothesis isn't worth thinking about because of how normal of a universe we are presented with forces adopting a view of normality that only appears in a post-selected hindsight.
I think he meant something like that:
Reality is quite large and detailed. It’s so large and detailed that it contains a plausible explanation for how the AI itself came to be. It’s so large that it contains radically more computing resources than it takes to run the AI (compared to most games and simulated environments, which are much smaller than the AI). The laws of physics and the evolved primates in the AI’s environment are the kinds of things that could actually yield a real AI at scale, unlike (say) the laws of the computer game Minecraft.
(from https://ifanyonebuildsit.com/5/what-if-we-make-it-think-its-in-a-simulation)
I don't think this affects your central point, but I'd expect a fairly large class of simulations to have exactly one miracle in them, and I don't think your argument weighs against this.
For example, someone with a question like "what if technology X had been discovered 100 years earlier?" might run a simulation with exactly one miraculous intervention to cause technology X to be discovered at the time of the simulator's choosing. Or "what if WW2 had ended differently?" could lead to a simulation with exactly one miraculous intervention to change the outcome of the war, etc.
Similarly, simulations about alternate rules ("what if humans had psychic powers?") might choose to start when things are getting interesting (e.g. when humans evolve, or when civilization kicks off) rather than starting with the big bang, and might fill in the history before that point from some other data source (e.g. basement reality), and then the simulation will have exactly one miraculous discontinuity in its history (at the moment the simulation started). h/t to (major spoilers)
the novel Ra by qntm
(It has occurred to me that "that's when the simulation started" is a possible explanation for (at most one of) various historical discontinuities like the year the singularity was canceled, though I don't give it much credence.)
This doesn't change your take-away that you shouldn't expect miracles in the future.
This is a floor estimate of the abilities of superbeings. There could be such a thing as naturalistic traceless interventions by supersimulators in such a manner that the simulacra would be unable to notice, anticipate or even explain. Thus drawing distributions which rule the Donald Trump election in or out as evidence of a (bored to local heat death) simulator are both a little dubious (ie there is a simulacra frame of reference where it is not possible to tell apart whether "I don't care about Trump's election" is a local projection of "... I don't care about Donald Trump in this universe" or "Dude, I actually don't care about Donald Trump's election" ).
Worse, a caring simulator can easily traceless intervene (for example wagging some rando dinosaur's tail) such that "I don't care about Trump's election" is a local projection of "I care about Trump's election and I have made it happen". That is, we may never even be able to tell usefully that we are in an interventionist or non-interventionist universe.
Do you believe that God intervened in the 2016 election? All these arguments carry over identically to God.
One usually assumes that, for many simulations, various forms of entertainment (possibly including “immersive VR diving”) would be among the goals.
Then, on one hand, one would expect lazy rendering of simulation to save resources, adding those resources only as needed (this is compatible both with the Fermi paradox and with subsequent “discoveries” that other civilizations “actually” exist).
And on the other hand, the level of ridiculousness in having a president with this kind of last name, this kind of behavior, and this kind of ridiculous circumstances of his recurrent coming to power is highly entertaining (at least to those not directly affected by the consequences).
And… this is not the only thing which seems to be rather ridiculous in a somewhat entertaining fashion about our reality. What we should really ask: is what we observe “unnaturally ridiculous” altogether or is it within reason for a “naturalistic reality”?
I am skeptical of how well calibrated your "ridiculous" meter is. Like if you lived in the past most kings were corrupt and ridiculous like Trump. Perhaps the last 30 years or so were anomalously normal seeming. A brief look at history makes it seem that way.
My reference class is US Presidents (and, more generally, elected leaders of democratic countries). One requirement for quite a long time was “being proper”, “demonstrating some kind of externally visible conformity with conventions”, this kind of things.
It reminds me about later post about Why Biological Anchors Never Work where was the moment with "imagine how funny it could be for aliens to choose this exact moment when you say your phone won't hover, to make it hover" and that in fact it won't hover. In a sense, it is dictated by priors. There is some giant number of universes where just one miracle happens. But all of them diverge, all of them are different. And amount of universes where miracles are regular is less than that, but they converge. So all the universes where just one miracle happens generally look like perfectly regular universes.
That reminds me of question "if we live in Tegmark 4, why there aren't irregularities like clothes turning into a crocodile". The answer to me seems that there is no reason to prioritize human concepts like cloth or crocodile. Most of bitstrings of irregularities will look completely random to humans, like quantum noise. Which, we actually observe. So it seems to me like an evidence that we actually do live in Tegmark 4. If we lived in classic deterministic universe where there is no quantum randomness, it would be evidence against us being in Tegmark 4. But it ends up that all possible programs with irregularities which diverge from regular laws of physics sum up to being random in each particular universe. (Which also seems to me as evidence that quantum mechanics isn't some high level physics law, but rather bottom level result of how programs are structured in Tegmark 4.)
What I infer from that is that if we were to expect some miracles we would have already seen them, and since we haven't seen them, then we should expect miracles which are on average not different from randomness.
Or to pose in the god framework, then as I now understand, the main problem of God hypotheses is "which one", all the excuses like "divine ways are mysterious" don't allow you to determine whether it is Yhwh, Allah or Zeus, and unless all of them exist simultaneously, you need to distribute probability among all of them, and so each one of them becomes unlikely already on priors, even if you take only socially established Gods, not just all descriptions of Gods. It doesn't make sense to argue with concrete religion, because first you need to choose why even that religion.
And the same prior argument works for optimization targets. You can't avoid improbability by saying it was being optimized, because you now just need to prove the improbability of some optimization target among all optimization targets.
This rules out interventionist simulations, and inaccurate ones in the sense of miracles. But I think it is important to point out that this is quite different from ruling out simulations that differ from the universe running them altogether, where the rules were never the same rules to begin with.
Plenty of things get simulated with fixed rules for reasons that have nothing to do with modeling the world outside them. An untouched world doesn't tell us much about what the world is for.
It also seems probable that there would be a large class of simulated universes which
are trying to gain information about an object of simulation, and which are therefore
realistic up to the limits of realism. So ... among the versions of you that are
simulated, the vast majority of people exactly like you are in realistic simulations.
Information-gathering simulations only become "the vast majority" if gathering information is far and away the most common reason to run a world like this. More common than fiction or anything else, and I don't see how we can know that. A superintelligence that runs simulations to gather information is a very of our era guess about what would be doing also.
(Reposted from Facebook, on January 17, 2017.)
I am concerned about the number of people I've heard joking about Trump's election being evidence for the Simulation Hypothesis.
Yes, I know it's a joke. I'm still concerned.
Warning: #Essay, #LongEssay
So as not to engage in Logical Fallacy: Appeal to Consequences, before I talk about why this joke is worrying, I shall first discuss why Trump's election does not in fact mean we are living in a simulation. And neither does the Berenstein/Berenstain Bears thing, etcetera.
Because atheism generalizes.
No, I'm not about to commit the Noncentral Fallacy (aka The Worst Argument In The World) by yelling "The Simulation Hypothesis is religious!"
But once upon a decade, there was a time when lots of people believed in God. A time when atheism had to be argued, not just taken for granted. There was a time when believing in atheism made you one of those weird, loud people with arguments that only people with unusually good epistemology could follow, and other people talked about you exactly the way that the anti-LessWrong tumblrsphere now talks about LessWrong.
Today, of course, atheism is just something that everyone in your filter bubble believes. But back then atheism had to be argued, and assessing the validity of those arguments required unusually good epistemology that left a lot of the audience behind, and other people laughed at those weird atheists and their snooty yet flawed epistemology on their primitive equivalents of Tumblr.
A certain key atheistic argument is so obvious as to be boring, which is of course a good sign for its validity:
"We don't see any sign of God's presence. Therefore there is no God."
"Ah," says the theist, retreating to their motte, "that's because God doesn't usually intervene in His creation. Now go away so I can resume telling people to pray for earthquake victims."
"Okay," says the atheist, "I'll be more precise: the argument shows that there's no interventionist God. And if we're not playing motte-and-bailey then that kicks hell out of most religions. All the scriptural religions record interventions. If we're not inventing elaborate excuses, then yes a powerful and benevolent God would stop earthquakes that randomly kill lots of innocent people in developing countries."
"Maybe He wouldn't," says the theist. "Who are we to know the mind of God?"
"You can only avoid that conclusion by overcomplicating things," says the atheist. "Ask a six-year-old whether a good person with superpowers would cure malaria. The six-year-old will say yes. The six-year-old is RIGHT. Since malaria is still here, there is no good person with superpowers who can act on our Solar System; and this argument doesn't suddenly go away when somebody invokes the word 'God'." The atheist pauses, thinking. "In fact, this argument generalizes to many other hypothetical entities with desires and superpowers. If there are potatoes in the world, we can rule out the existence of a sufficiently powerful comic-book supervillain who hates potatoes. If to every appearance our world just follows simple mathematical laws, this is strong evidence against every class of superpowered being whose utility function wants anything more than It wants our universe following simple mathematical laws. If there are other places, other parts of Reality, under the dominion of superpowerful beings running around optimizing those places to meet various desires as much as possible, those places really don't look like our part of the universe! The six-year-old knows that, and the six-year-old is right."
"Ah," the theist says wisely, "but what if God wants the world to look exactly like it's following mathematical laws?"
"Well," says the atheist, "as an instrumental strategy, I can imagine that we are figments of the imagination of some superbeing who for whatever reason wants to know a more exact probability distribution about what happens in universes like this one. Say, somebody in a universe like this one has hidden a buried treasure, and the superbeing is mentally simulating many different imaginary versions of our universe to refine its guess as to where the treasure might have been hidden. In this case, to intervene inside the imagined universe would make the imaginary world unrealistic, and thereby invalidate the imagination's information-value for the real-world question the superbeing is trying to answer. This would produce a superbeing that was consistently noninterventionist with respect to its imagination of our dreamed universe; and we must also suppose that the superbeing doesn't care when imagined people die of malaria. This hypothesis is compatible with our observations. In fact it is exactly as compatible with our observations, as our universe not being imaginary. And it is no coincidence that this story consistently implies that we will also not observe any miracles in the future."
"Well, what about the election of Donald Trump?" says the theist. "It's not proof, I know, but isn't that at least some evidence that there's an interventionist God?"
"Why on Earth would it be?" says the atheist.
"Because God could have made Trump be elected," says the theist, "and therefore, worlds in which Trump is elected are more likely given the God Hypothesis."
"A person with superpowers who cares about electoral results also could have made Hillary Clinton be elected," says the atheist. "The superbeing hypothesis class can't make every outcome more probable. The probabilities still have to sum to one."
"Okay," says the theist, "but the prediction markets thought that Trump being elected was unlikely, presumably under the implicit assumption of our universe running under mathematical laws. If the universe doesn't have to follow those strict mathematical laws, if those laws can sometimes be violated, doesn't that move the extreme probability of Trump not being elected back toward the middle? Then Trump's election is more probable given the God Hypothesis than not."
The atheist throws up their hands. "What kind of interventionist superbeing cares enough to intervene to make Trump win an election, and doesn't care about a million other aspects of the world that they could be intervening on?"
"Truly the ways of Heaven are mysterious," the theist says.
"The ways of science are not mysterious," says the atheist, "and they say that we don't just get to throw up our hands and say 'the ways of the experimental apparatus are mysterious' when observation fails to accord with the obvious prediction. It's just a generic excuse to blur out all the failed predictions of the theory, and if you apply it evenhandedly, you shouldn't be using that theory to predict anything else either."
"Anyway," says the theist, "I'm asking if you agree that the God Hypothesis predicted with at least slightly greater probability that Trump would be elected."
"I'm frankly not sure of that in this case," says the atheist. "When I actually try to envision that tiny fraction of weird possibilities in which some superbeing who never intervened on any definitely visible occasion, cares enough to intervene in the US Presidential election, it's not clear to me that they favor Trump over Clinton."
"I'm not saying that God, as I understand Him, would favor Trump over Clinton," says the theist. "It's just that I, in my mortal limitation, don't have reason to believe that the ineffable God is 4 times as likely to favor Clinton over Trump. If in my own imagination God would be twice as likely to favor Clinton over Trump in his ineffable plan--that is, given that God intervened, I would have guessed 2:1 odds for Him intervening for Clinton--then to the extent one believes God might intervene in the US election, and that prediction markets have not already taken this into account, one ought to shift the market odds of 4 : 1 for Clinton : Trump towards the less extreme 2 : 1 odds of God favoring Clinton over Trump. Maybe just a little, but even so, it would be some little bit of evidence."
The atheist crosses their arms, looking skeptical. "I'm a little suspicious that you would maybe have given God stronger odds than 2 : 1 of favoring Clinton over Trump, if asked in advance. But I suppose, when I envision weird beings with superpowers intervening there and nowhere else, I don't imagine them swinging strongly for Clinton. So the point carries for me, if not for you."
"So you agree, then," the theist says, "that Trump's election was to some small degree evidence for the existence of God."
The atheist thinks. "I'm not sure that's valid reasoning," the atheist says slowly. "To get a higher predicted probability that Trump wins the election, it's not enough to hypothesize a superbeing. We have to hypothesize a superbeing that cares about Trump winning the election, but doesn't care about malaria, and doesn't care about making planets move in perfectly circular orbits instead of elliptical orbits, and so on. I'm worried that you're pulling some equivalent of seeing a coin come up heads-heads-tails and saying, 'Ah, that's evidence that a superbeing made the coin come up heads-heads-tails, because the coin would be more likely to come up HHT if a superbeing wanted it to.'"
"But we're not talking about a random-looking bitstring," says the theist. "Let's say there's some event which has as simple description, but is incredibly unlikely assuming the universe runs on physics as we know it. Like, you flip a coin 1024 times and the resulting bitstring has a SHA256 hash of exactly zero. Would you not agree that this result is much more probable given the God Hypothesis?"
"Okay," the atheist says after a moment, "I think I accept that, and in that sense, I accept that the Exists Some Rarely Interventionist Superbeing Hypothesis nudges extreme probability assignments over any simply describable outcome back toward 50-50. Now conversely, by the law of Conservation of Expected Evidence, we must say the Superbeing Hypothesis is disconfirmed to some small degree every time we see a rock fall down. After all, a superbeing could have made the rock fall upward, or hover in the air; and if that did happen you would no doubt shout 'Ah, that is more likely in worlds where there is a God!' So when it does not happen, that is a tiny little bit more evidence against an interventionist God."
"So you do concede that Trump's election is evidence for God's existence!" the theist says, smiling.
The atheist throws up their hands. "There ought to be a name for the fallacy of taking a coin that comes up heads-heads-tails, looking at just the tails coinflip, and saying, 'Aha, that's evidence for the coin being biased towards tails!' Yes! Yes if you consider just that single coinflip as an isolated event, it is from a strict Bayesian standpoint a small unit of evidence for the coin being biased tails! But when a prediction market says something has a probability of 20%, it ought to happen around one time in 5! If we're all inside the strictly realistic imagination of a superbeing who is imagining many different versions of universes like this one in order to gain more information about some real-world variable, then most of those worlds, just like the real world, look back and see a big class of events in which 20%-probable events are happening about 1 time in 5. And 4 times out of 5 they don't happen. Trump's election was not so unlikely that it outweighs every other dog not barking, every other observation of the world following strict physics. Trump's election was not the sort of thing that only ever happens in the unfettered imaginations of superbeings who don't care about strict realism. It doesn't happen only in worlds with active superpowered villains."
"But it is more likely if the God Hypothesis is true," says the theist.
The atheist sighs. "By every rule of the ordinary meanings of ordinary words, and irrespective of some technical interpretation you could make of the Bayesian import of isolated events, we do not say that Trump's election shows there is a God--at least, no more than a flipped coin coming up heads three times in a row shows there is a God."
===
No, the Simulation Hypothesis is not just the revenge of religious reasoning.
But the reasoning against interventionist supernatural deities goes through unchanged for interventionist naturalistic simulators, and I'm a little bit worried about this not being obvious.
You know how it's a bad sign when scientists leave the laboratory and suddenly forget that beliefs require evidence?
I'm worried if the anti-interventionist-superbeing argument suddenly isn't obvious, once the argument is no longer about that God thingy that nobody inside your filter bubble believes in anyway, and is instead about that cool-sounding simulation thingy.
Epicurus (or whoever) was right to deduce the absence of an interventionist benevolent God by observing suffering. Later atheists were right to further to deduce the absence of any kind of interventionist God with desires such that the world could obviously be improved further with respect to those desires. Atheists were right to look around them and see a world where, so far as they could tell, no verifiable miracle had ever occurred. Naturalistic philosophers were right to see a world where, so far as they could tell, improbable things happened around as often as they should, and no more often than that. And it was valid reasoning to conclude by induction that we shouldn't expect miracles, or disproportionately frequent improbabilities, in the future.
There are arguments against noninterventionist Gods that don't go through for noninterventionist simulators, such as "God sure sounds like an ontologically basic mental entity, and minds are made of parts", or "This does not sound like a hypothesis we would naturally trip over by starting from only the observable world and extending our generalizations from there."
But all the arguments against interventionist Gods--every way the universe does not look like there is a God running around inside it fiddling with things--that all goes through without a hitch for any other kind of interventionist being with superpowers. If there are perfectly normal interstellar aliens watching us from Jupiter, they're noninterventionist aliens; they didn't intervene on Earth and they also didn't enclose all the stars in Dyson shells. Not all of the arguments against God are arguments against the existence of naturalistic aliens; but the arguments against interventionist Gods are arguments against interventionist naturalistic aliens being here in this Solar System.
So if this is a simulation, it's a realistic simulation. And the obvious reason why that some superintelligence would be rolling a realistic simulation, is that this is a simulation meant to gain information about a lower-level universe that is very similar to this one.
In which case there will be no miracles now or ever, nor is it a coincidence that this is what the obvious hypothesis ends up saying.
As we once looked around ourselves and saw a universe that runs only on math, we should, by the same and unchanged eyes, look around and see a simulation that is perfectly faithful to the kind of universe that runs only on math.
Regardless of the absolute proportion of people in simulated universes, and people in basement universes, I think we are licensed to expect that the vast majority of simulated universes where the simulators are messing around with the simulation or filtering its conditions in weird and unrealistic ways, do not look like this seemingly untouched universe. It also seems probable that there would be a large class of simulated universes which are trying to gain information about an object of simulation, and which are therefore realistic up to the limits of realism. So whatever proportion of people like you are in simulations--whether that proportion is large or small compared to the proportion in basement universes--among the versions of you that are simulated, the vast majority of people exactly like you are in realistic simulations. These proportions are not nudged more than infinitesimally by conditioning on Donald Trump having been elected President.
Similarly, the hypothesis of a realistic simulation does not even help explain anything that still seems weird and mysterious. The Fermi Paradox is that we look around and see an empty universe with seemingly uninhabited stars and frequent planets, even though the stars are old enough for life to have evolved long before now. Well, the Realistic Simulation Hypothesis doesn't help explain the Fermi Paradox any more than it explains Trump's election. If this is a sim, it's a realistic one meant to gain information about a similar basement universe, so why would that universe have a Fermi Paradox?
It must all add up to normality in the end. And the normality that it must all add up to hasn't changed since someone thought up the Simulation Hypothesis.