Is there a popular philosophical argument that purports to show or corroborate the claim that the reality is a computer simulation? I am wondering if any philosopher tried to tackle this question with some strong philosophical arguments, because most of the arguments I've heard seem to rely on pseudoscience or highly speculative arguments based on certain scientific facts that aren't convincing.
The basic idea promoting the hypothesis is a statistical one (and that includes the Bostrum paper): An extrapolation of the amount of the physical substrate of computing resources available to a civilization as a function of the time it exists paired with an estimate of the amount of said resources needed to run the universe as we perceive it as a simulation.
It is argued that the proportion allows for running an very large number of simulations, much larger than the number of civilizations that could possibly evolve in the universe over its lifetime.
However, imho the reasoning is flawed:
1. It is not addressed why any civilization should want to run the multitude of simulations.
2. The reasoning is of no help to answer the specific question whether precisely our civilization is a simulation.
(of course, 'civilization' here is restricted to civs with a tech level at least comparable of ours; at least, engineered information processing should be commonplace)
Philosopher Nick Bostrom's simulation argument is specifically based on a form of anthropic reasoning known as the self-sampling assumption, which is also used in the doomsday argument that is closely related to the simulation argument. The self-sampling assumption basically says that when you know you're part of a certain group of intelligent beings but don't know certain properties of the group (such as its size, or the size of different subsets), it often makes sense to reason as if your identity was randomly sampled from all the beings in the group. Bostrom had an FAQ on self-sampling and the doomsday argument on his old site which quoted the following thought-experiment from astrophysicist John Leslie:
A firm plan was formed to rear humans in two batches: the first batch to be of three humans of one sex, the second of five thousand of the other sex. The plan called for rearing the first batch in one century. Many centuries later, the five thousand humans of the other sex would be reared. Imagine that you learn you’re one of the humans in question. You don’t know which centuries the plan specified, but you are aware of being female. You very reasonably conclude that the large batch was to be female, almost certainly. If adopted by every human in the experiment, the policy of betting that the large batch was of the same sex as oneself would yield only three failures and five thousand successes. ... [Y]ou mustn’t say: ‘My genes are female, so I have to observe myself to be female, no matter whether the female batch was to be small or large. Hence I can have no special reason for believing it was to be large.
Leslie's example is perhaps a little ambiguous in that he doesn't say what the prior probabilities were that the large batch would be female vs. male, though perhaps he is assuming that in the absence of any other information you should assign a 50/50 subjective probability to each possibility, or perhaps he's implicitly assuming we know that the planners of the experiment made the choice using a random process with 50/50 odds like a coin flip. Either way, if you have some prior probabilities for each outcome, and you then treat yourself as a random sample of all the people who will be created in this experiment, then you can use Bayesian inference to update your prior probabilities, meaning if you are a female then you will upgrade the subjective probability that the large batch was female and downgrade the subjective probability that the large batch was male. The self-sampling assumption would say this is exactly analogous to a problem where an urn is filled with 3 balls of one color and 5000 of a different color, but you don't know which color is the majority one--then if you draw a ball and see it's a given color, say red, that should cause you to significantly upgrade your subjective probability that red is the majority color.
In the simplest case where you had 50/50 prior probability that red would be the majority color, after the draw you would update your probabilities to say there was a 5000/5003 chance that red was the majority color, and only a 3/5003 chance that red was the minority color. Similarly in Leslie's thought experiment, if our prior knowledge of the setup said that there was a 50/50 chance that the larger batch would be all-female (again, we could just assume the planners flipped a coin to decide the majority sex), then you should observe your own sex and update your subjective probabilities to a 5000/5003 that the larger batch was the same sex as you--if every individual in the experiment reasons the same way, 5000 will be right and only 3 will be wrong, so it seems like a good bet in that sense.
The doomsday argument is a very similar application of the self-sampling assumption. We don't know how many humans will ever be born (or human-like intelligences 'descended' from human civilization), but say that a priori you would assign some non-negligible probability to the hypothesis that some kind of doomsday is coming soon that will cause a collapse of civilization and a massive dieoff (say, nuclear war, or rampant global warming) so that the number of humans born in the future is smaller than (or at least not much larger than) the total number that have been born up until now, estimated at around 107 billion. And say that you would also assign a non-neglible a priori probability to the hypothesis that humans or their intelligent descendants will avoid doomsday, and have a long and wonderful future where they colonize many different regions of space and in doing so become very difficult to drive extinct, such that a trillion or more individuals will be exist in the future. Then according to the self-sampling assumption, the observation that you yourself are somewhere around the 100 billionth person to be born must sadly cause you to upgrade the subjective probability of the "doomsday soon" scenario, and downgrade the subjective probability of the long and glorious spacefaring future.
If you accept the self-sampling assumption, this is analogous to a situation where you know a numbered ball will be drawn from one of two urns, each containing a collection of balls labeled sequentially, where one urn has a much larger total than the others--for example, in urn A the balls might be labeled 1-100, in urn B they might be labeled labeled 1-10,000. If you don't know in advance which urn the ball is going to be drawn from, but you assign some subjective probability to each beforehand, then if you see the ball that's picked has the number 80 on it, that should significantly increase the subjective possibility you assign to the hypothesis it was drawn from urn A, and significantly decrease the subjective probability you assign to the hypothesis it was drawn from urn B.
I don't know of Bostrom has ever said this specifically, but I would bet that the doomsday argument was part of the inspiration for his simulation argument, since the simulation argument uses self-sampling in a similar way but exploits a kind of loophole that allows us to avoid the conclusion that a near-future doomsday is far more likely than long-term survival of technological civilization. The loophole here is that if a very long-lived advanced civilization acquires vast computing resources and performs "ancestor simulations" containing conscious minds that aren't aware they're part of a historical civilization, then these minds will be deceived about their "birth order". So even if you seem to be only the 100 millionth human (or humanlike mind) that ever arose, you might actually be the 100 trillionth, the 100 quintillionth, etc. And Bostrom is a transhumanist who seems hopeful that our civilization might develop into a "posthuman" one that spawns a vastly larger number of minds--see his paper The Future of Humanity where he discusses the possibility of posthuman civilizations and on p. 20 gives one of the qualifying features as a population greater than 1 trillion--so it seems natural that he'd be interested in any loophole that allows us to avoid the conclusion that our own birth order makes it far less likely that our world will give birth to such a super-civilization. On p. 25 of this paper he also comments that there are plausible (to him) arguments that "the current century, or the next few centuries, will be a critical phase for humanity, such that if we make it through this period then the life expectancy of human civilization could become extremely high", noting the possibility of superintelligent AI giving our civilization much improved "foresight and planning", along with the point that "once a human or posthuman civilization becomes dispersed over multiple planets and solar systems, the risk of extinction declines."
Note though that Bostrom tries to be somewhat "conservative" in the conclusions he actually draws from the self-sampling assumption and the possibility of ancestor simulations. He does not actually say the possibility that we live in a simulation is far more likely than any of the alternative, his conclusion as stated in his simulation argument FAQ is that we should believe in one of the three options in a trilemma:
The argument shows that at least one of the following propositions is true: (1) the human species is very likely to go extinct before reaching a “posthuman” stage; (2) any posthuman civilization is extremely unlikely to run a significant number of simulations of their evolutionary history (or variations thereof); (3) we are almost certainly living in a computer simulation. It follows that the belief that there is a significant chance that we will one day become posthumans who run ancestor-simulations is false, unless we are currently living in a simulation.
So our living in a computer simulation is just considered to be one of the three viable possibilities. If you apply the self-sampling assumption to human-like intelligences similar to ourselves, not just specifically to humans, then for (1) to be true, it would have to be true that the vast majority of civilizations at a level of development similar to ours fail to reach the technological breakout point discussed in The Future of Humanity, i.e. there would have to be a Great Filter lying in the near future of all civilizations at a point similar to ours, something close to an ironclad historical law that civilizations like ours almost never survive. And for (2) to be true, there would have to be a different sort of ironclad historical law that civilizations which do reach that breakout point almost never perform significant numbers of ancestor simulations, given the assumption that civilizations that do make it past this breakout point produce many orders of magnitude more intelligent beings than the 100 billion or so that have existed in our own (apparent) history. It also seems to me that whatever prior probability one assigns to (2), the self-sampling assumption should cause you to significantly downgrade the subjective probability of (2), since only an extreme minority of individuals will have an observed birth order so low in that scenario; I'm not sure why he includes this one with the others.
So if one accepts the arguments for Bostrom's trilemma (or for a dilemma just consisting of the scenarios (1) and (3)), but one finds historical laws required by (1) and (2) to be implausible compared to (3), that would be a reason to assign a larger subjective probability to the hypothesis that we're living in a simulation. Note that the only element of this argument that might be considered a properly philosophical issue is the use of the self-sampling assumption, the rest is just a non-philosophical argument about different possible plausible outcomes for future civilization.
Pro: The hypothesis provides an elegant answer to the open question "Why is mathematics a language to express the laws of nature?"
Contra: We do not observe typical failures of the natural processes which can be explained as rounding errors due to the mathematical computation which governs the simulation.
"Brian Greene: The Hidden Reality. Penguin Books (2012). See Chap. 10 Universes, Computers, and Mathematical Reality. The Simulated and Ultimate Multiverses.
The popular argument is simple: Civilisations capable of building conscious simulations will build many more than one. Typically they will be simulating their own civilisation. Furthermore, the conscious simulants they create will in turn build simulations of their world, in an ever-deeper nest of levels. Therefore there will be far more conscious beings living in simulations than in the top-level real world. Therefore, the chances are overwhelming that we are living in a simulation.
That's it in a nutshell. But I think it a mistake to compound "popular" and "philosophical" in the same sentence. They say that you can use statistics to prove anything, and I have seldom seen a better example of that in action.
The whole argument is based on the premise that any appropriately complex information system will become sentient. This assumption, exemplified by Integrated Information Theory, is based on the observation that wet hydrocarbons are nothing special, but that what is special about the sentient brain is the information it processes. This is treated not as a philosophical or metaphysical proposition but as scientifically demonstrable. Otherwise, creating simulants would not be possible.