The supposed "knockdown argument" for Simulation Theory -- what did I miss? (IF we can create simulations indistinguishable from reality...)

Question

I have never read any extensive formulation of the argument; I've only listened to people (like Elon) lay it out in Youtube videos; so I don't know what it's called or how to lay it out with rigour; but the popular argument for us existing in a simulation goes something like:

1. 'We can in the future create simulations that are indistinguishable from (base) reality' --> 'Eventually we will inevitably have an extremely high number of such simulations'
2. 'Eventually we will inevitably have an extremely high number of such simulations' --> '(Eventually) it would become highly unlikely that we exist in base reality (one to several billion)'
3. 'We can in the future create simulations that are indistinguishable from (base) reality'

---

4. 'It is right now, already, highly unlikely that we exist in base reality'

What have I missed? The conclusion obviously doesn't follow.

A recent rendition I heard, that one also informal, went something like:

1. 'We can in the future create simulations that are indistinguishable from (base) reality' --> 'Eventually we will inevitably have an extremely high number of such simulations' 2)'Eventually we will inevitably have an extremely high number of such simulations' --> 'It is likely that we already do'
2. 'It is likely that we already do' --> 'It is highly likely that we exist in a simulation'
3. 'We can in the future create simulations that are indistinguishable from (base) reality'

---

5. 'It is highly likely that we exist in a simulation'

I take issue with the second premise of this latter version of the argument. That's what I never got with the argument, with any of the versions I've heard -- even if we grant that we will create these kinds of simulations eventually, and that after a certain point the chance of us existing in a simulation becomes virtually =1, why do we think that we have already gotten there? Either we have or we haven't (I mean, that's of course always true, but you get the point). It feels like a 50-50 thing to me. It feels like the passing of time isn't being thought about here at all.

What have I missed? Or is it the physicists trying their hand at metaphysics that have missed something?

Answer 1

To me, premise one fails because of resolution.

In order to create a simulation that is indistinguishable from base reality, it has to operate at the same resolution. That is, the smallest unit (e.g. Planck Length) in each reality has to be equivalent.

In order to simulate two particles, you would need either:

1. two particles.
2. 1 particle and extra time for it to simulate both particles.

To simulate the earth at perfect resolution, you would need a computer the size of the earth, or a smaller computer and extra time. A computer the size of the earth would almost certainly run into some engineering challenges that might not actually be resolvable. We have not demonstrated that it is even theoretically possible to build such a computer. There is no reason to accept that it is likely that we will.

We make this kind of trade off in simulations all the time. We either try to model very small things that don't take much time as accurately as possible, or we model large things at the lowest resolution that we can while still getting information out of the simulation.

So while it is possible that we are in a simulation, in order to be worth running at the scales we see, that simulation would have to be a lower resolution reality than the base reality in which the simulation hardware exists, and hence distinguishable from that base reality to anyone with access to the base reality.

The one argument in favor of such a theory is that it solves the Fermi Paradox. There are no aliens because they aren't included in the simulation. However, there are also other very clever arguments that demonstrate that humanity is going to go extinct before we reach such a technological level.

So a second counter to p1 is that it is in fact unlikely that our civilization will survive long enough to be able to produce such a simulation, and in fact, it is unlikely that any civilization would survive that long. There are going to be, across an infinite universe, many more civilizations that fail to reach that point than that would, so it's most likely that we're living in one of those.

A third counter to P1 is that is assumes that we're going to continue to improve our ability to simulate reality without limit. This assumption is not only baseless, it's also deeply flawed. To take a trivial example, the simulation argument, applied to cars, implies that eventually cars will travel faster than light using zero energy to do so. Spoiler alert: they won't.

Clearly there are limits to what can be done, and the simulation hypothesis relies on assuming that we know that those limits are beyond a certain point without actually demonstrating that the limits are beyond that point, and counter to demonstrations that the limits are in fact not beyond that point.

These are just a few basic arguments against P1. It's incumbent upon those making the claim to demonstrate that the claim is remotely plausible, and so far, they have failed entirely to do so.

If I may editorialize, the biggest objection to the simulation hypothesis as a whole is that it assumes that we're interesting enough to simulate.

Taken at face value, humanity is nothing more than an almost inconceivably insignificant bunch of semi-evolved apes hell-bent on destroying our habitat. The simulation hypothesis attempts to turn us into something so amazingly important and fascinating that a civilization of god-like beings with near-infinite powers would bother to spend the time and the resources to simulate us.

Given the scale of the universe and the scale of time, and the infinite possibilities available to such a civilization, it is much more likely that they would simulate something other than us than that they would bother to simulate us. So an objection to the argument as a whole is that given that this advanced civilization could simulate almost anything, the idea that they would bother to simulate me is laughable.

Answer 2

There are a few other premises that you haven't included, and may be omitted from some people's presentation of the argument as they consider them too obvious to have to state. One of them is that we should based our probability estimates on what worlds are most common. There's a thought experiment called the Sleeping Beauty problem:

Sleeping Beauty volunteers to undergo the following experiment and is told all of the following details: On Sunday she will be put to sleep. Once or twice, during the experiment, Sleeping Beauty will be awakened, interviewed, and put back to sleep with an amnesia-inducing drug that makes her forget that awakening. A fair coin will be tossed to determine which experimental procedure to undertake:

If the coin comes up heads, Sleeping Beauty will be awakened and interviewed on Monday only. If the coin comes up tails, she will be awakened and interviewed on Monday and Tuesday. In either case, she will be awakened on Wednesday without interview and the experiment ends.

Any time Sleeping Beauty is awakened and interviewed she will not be able to tell which day it is or whether she has been awakened before. During the interview Sleeping Beauty is asked: "What is your credence now for the proposition that the coin landed heads?"

If you say that the probability is 1/2, then you're probably not going to be convinced by the simulation argument. But if you say it's 1/3, then the simulation argument is probably more convincing.

That's what I never got with the argument, with any of the versions I've heard -- even if we grant that we will create these kinds of simulations eventually, and that after a certain point the chance of us existing in a simulation becomes virtually =1, why do we think that we have already gotten there?

Because, according to the argument, the number of people who exist in a universe in which we have gotten there is massively larger than the number of people who live in a universe where we haven't, therefore the overwhelming probability is that we live in a universe where we have gotten there. In vitro fertilization is involved in a tiny minority of births. If you have no other information as to whether in vitro fertilization was involved in your birth, would you assume that the probability is significantly lower that 50%?

Answer 3

UPDATE

A comment by @Hypnosifl reminded me of the original of this argument, and I now realize that my answer below is not correct. I would remove it, but it has generated some discussion, so I'm leaving it for now in case anyone wants to comment further.

WRONG ANSWER

I believe the argument goes something like this:

Premise 1: Eventually it will be possible to create simulations that are indistinguishable from reality. (This strikes me as likely but not certain).

Premise 2: If humans could create simulations indistinguishable from reality, then some humans would live their entire lives in such a simulation and never know that they do not live in a simulation. (This strikes me as almost certainly false).

Conclusion 3 (from 1 and 2): Eventually some humans will live in a simulation indistinguishable from reality and not know that they are living in a simulation.

Premise 4: One of the simulations that a significant number of people will live in will be a simulation of the current era (This one strikes me as absurd. People have the option of living like gods, and they choose instead to live in this world of scarcity and woe, subject to injury, disease, famines, murder, and war? Only someone who is relatively young and lives an unusually coddled life would think so).

Conclusion 5 (from 3 and 4): Eventually a significant number of people will live in a simulation of the current era and not know that they are living in a simulation of the current era.

Premise 6: The time during which a significant number of people live in a simulation of the current era and don't know that they are living in a simulation will be much longer than the current era.

Conclusion 7 (from 6): Eventually, the number of people who live in a simulation of the current era and don't know it is a simulation will be much larger than the number of humans who live in the current era.

Argument: Suppose you are a human being who is apparently living in the current era and you want to decide whether you really do live in the current era or whether you live in a simulation of the current era. You have no evidence to distinguish one from other, so the only thing you have to go on is probability. Since many more people will live in simulations of the current era than live in the current era, you are probably one of the ones who lives in a simulation.

Answer 4

The “knockdown argument” for simulations theory was a paper by Nick Bostrom in 2003: https://www.simulation-argument.com/simulation.pdf

Bostrom did not spell out all the steps of his argument clearly. As most people are wont to do, he missed some key assumptions that he simply takes as givens. Filling those in, his argument was as follows:

1. Assume that consciousness and selfhood is computational/algorithmic, such that eventually software could be alive and conscious
2. Software consciousness could exist in real time in a real world, or in simulations
3. Technology has shown a trajectory of significant advances, and projecting into the future, we can presume the potential of nearly limitless computational power for advanced civilizations. Bostrum uses the term post-human for such civilizations.
4. The “carrying cost” of simulated beings is far less than those of physical ones, and if a post human civilization creates simulated lives (in a simulated environment), then the number of such simulated lives would exceed the number of non-simulated beings over all of history by orders of magnitude.
5. As we are still apparently prior to “post human”, Bostrom limits his argument to “history” simulations, where the past of a race is being simulated.
6. Simulated lives could not tell if they are in a simulation
7. We can perform a probability calculation as to whether we are simulated beings, and that probability is basically: Fraction of civilizations that become post human X number of history simulations they run X the number of lives in an actual history before a civilization becomes post human – divided by this product + the number of lives in an actual history.
8. This probability becomes massively large, unless basically no civilizations become post human, or no post human civilizations ever run ancestor simulations.

Bostrom then goes on in the paper to discuss the reasons for thinking which of his three options in the paper is more likely.

In a 2011 follow on, Bostrom admits to a flaw in the assumptions that could change the calculation in 7) significantly, and suggests some corrections. https://philpapers.org/rec/BOSAPF-3 Most of Bostrom's critics focus on prior assumptions, not on the calculation, so this “correction” did not engage with that criticism.

Following are a sample of the critics of Bostrom:

This paper attacks the presumption that premise 6 is basically free – by noting that the massive amount of coding needed to achieve an undetectable simulation environment would be massively difficult/expensive to create, even if the “processing” is basically free. From that and a variety of other estimates, the critic argues that product of the two main terms in 7 does not become extremely large, as Bostrom assumed. https://philpapers.org/archive/ROBCEI-2.pdf

This paper critiques multiple aspects of Bostrom's argument Microsoft Word - http://philsci-archive.pitt.edu/11537/1/Computer_Simulation_paper_revised.pdf It cites Penrose arguing that Godel’s incompleteness theorem shows that not just consciousness, but basically nothing is reality is computable, and Feynman arguing that our universe is analog, and therefore not computable. It notes that One True Logic is needed for the thesis, which is not compatible with logical pluralism: https://www.cambridge.org/core/journals/think/article/abs/guide-to-logical-pluralism-for-nonlogicians/EDFDFA1C9EB65DB71848DABD6B12D877. And it wraps up with a claim that Bostrom's assertion is not testable and therefore not science, unless one accepts that the tests are in a real world, hence refuting the simulation premise being tested.

This blog critiques computationalism as a theory of consciousness https://thethink.institute/articles/simulation-theory-debunked. A key assumption is makes is that unlike things cannot spontaneously emerge (denies any emergence principles are possible). The importance of emergence in contemporary philosophy of science undercuts this critique. A second critique is that beliefs must cohere rationally, and the simulation hypothesis cannot do so (logic can change in a simulation). This second argument depends on One True Logic, which as noted earlier, is widely recognized now to be refuted by logical pluralism. This paper https://www.researchgate.net/publication/257560619_On_the_'Simulation_Argument'_and_Selective_Scepticism argues that Bostrom engages in selective skepticism -- where he does not apply extreme skepticism to the limits and nature of computation, but does apply extreme skepticism to the physical world -- a selective approach to skepticism which if made more consistent, changes Bostrom's conclusions.

These critics primarily attack assumption 1, premise 6, and the validity of the calculation of 8, plus a wholistic critique of the hypothesis and reasoning structure Bostrom uses

To give Bostrom credit, he admits that 1) is speculative, and disputed. As all theories of consciousness are speculative, and more widely considered refuted than proven, that there are major concerns relative to 1) just comes with the territory in Philosophy of Mind. It is reasonable to credit Bostrom with developing inferences that apply to only one of a disputed set of worldviews, and run with the one he chose, and see where it goes.

The critiques of premise 6 seem to be more substantial. 6 can't be implemented to infinite precision, by definition. And between the logic problems of imperfect simulation, Godel, and physics being indeterministic, the computationalism assumption about the universe that is intrinsic to the hypothesis is one that appears to be intrinsically flawed. And Bostrom does not deserve a pass on that aspect of computationalism, because he can’t point to a disputed set of mutually rejected models.

The attacks on logical coherence and testability of the thesis in general – appear to be raising the bar fallacies. No views logically cohere – as all our worldviews have pluralistic bases, which leads to logic contradictions. And the listing of problems discussed for the Simulation Hypothesis serve very effectively to refute the claim it is not testable. Both philosophy and science are pragmatic and tentative processes, and that Bostrom's assumptions are not bombproof is just SOP for pragmatic reasoning. I at least found those critiques to be less convincing than the authors of the papers found them.

Bostrom's thinking is therefore not really a "knockdown argument" and there are a variety of critical approaches taken to critique it, with varying validity to these sorts of critiques.

Answer 5

If Carl Jung's principle of Synchronicity is true, then there is a phenomenon that could only be created in some kind of simulation (or hyper-world). Furthermore, that simulation could only be a replica of 'reality' if reality also included synchronicity, in which case — as stated — reality would be some kind of simulation. In either case, the base level of the simulation would need to be unlike the simulated reality in order to be able to incorporate the extraordinary phenomenon, and there is a potential knockdown argument against "simulations indistinguishable from [base] reality."

https://en.wikipedia.org/wiki/Synchronicity#Philosophy_of_science