Does Bostrom's simulation argument make sense?

Bostrom has famously argued that we live in a simulation. One of his key assumptions is that other civilizations exist that can simulate us. Why do we make this assumption?

Why assume that other civilizations exist? We obviously would not be able to see them in our universe. If we did, that would imply that we do, in fact, live in "base reality." If that would be the case (that we live in base reality, yet it is simulated by others in this base reality), it would imply an infinite chain of causality, which seems dubious at best.

Why, then, assume that other universes exist besides ours? We simply do not know if 1) other universes exist at all, and 2) life can be supported in that universe, much less intelligent life. It doesn't seem that probability can be applied here, either. How can one possibly, rationally, assign probability to this?

As a bit of an aside, it seems that the argument also assumes reductive physicalism. Is this not contested heavily?

• Where does he make this assumption? That looks more like part of the conclusion. Can you cite the part of the paper that makes this assumption? – Eliran Aug 1 '19 at 21:12
• @Eliran If you're talking about the assumption that other civilizations exist that can simulate us, it is a requirement of the argument. If that's not the case, then we ourselves must simulate us and that's... circular at best. If it's the assumption of reductive physicalism, he assumes that consciousness can arise from the hardware of a computer, which I think implies that it can reduced to it. Is that not correct? – Josh Aug 1 '19 at 21:25
• What is infinite change causality? Also see philosophy.stackexchange.com/a/56834/5759 – alanf Aug 2 '19 at 8:31
• The assumption is just that whatever the base reality, many civilizations arise within it. Nowhere does Bostrom assume that many civilizations must exist in our reality even if our reality is a simulation. – Hypnosifl Aug 2 '19 at 12:51
• @alanf Sorry, that was a typo. I meant infinite chain of causality. – Josh Aug 2 '19 at 17:43

You have almost answered your first question in your reply to Eliran: we make that assumption for the sake of argument, and some of us are interested in the argument because it seems, at least at first sight, to provide an interesting perspective on questions about what is real and about our true nature (others might be interested because they want to definitively put a stop to such nonsense!)

You may not think it at all plausible that such a situation could arise, but there are others who do. I cannot speak for anyone else, but I would start from the position that it is not completely ruled out that we could, in principle, create such a simulation, and then the question of whether I am in one follows naturally.

Assuming that such a thing is possible, it also seems at least possible that a conscious entity in it would be incapable, in principle, of discerning that to be so, but its inability to discern the greater truth of its situation would not be a prima facie argument against the possibility of such simulations.

Putting aside the question of whether the simulation argument does necessarily require reductive physicalism, if it did, then the fact that the latter is contested would hardly be grounds for dismissing the former: just about every nontrivial position in philosophy is contested. Whether reductive physicalism is heavily contested is rather a matter of opinion. As it happens (at least, according to the physicist Sabine Hossenfelder), the simulation hypothesis is not popular among physicists.

• Thank you. It seems that Bostrom's argument is essentially probabilistic (I believe he assigns 1/3 probability to us living in a simulation). How can one possibly assign probability in a case like this? How can you meaningfully assign a probability to other universes existing which support life? We simply do not know. Modern physics would have to tell us nothing, I think (if it's all simulated anyway). As far as I can tell, the only way would be a subjective, Bayesian sort of assignment. But, this obviously does not hold much weight. Is this the case here? – Josh Aug 2 '19 at 17:46
• @Josh If you could estimate the probability P that a civilization would reach a level of technology where it could run such a simulation and chooses to do so (and calculate the expected number it will run), and you assume that every simulation has the same probability of reaching that point and running its own sims, then you can calculate the expected number of total simulated civilizations. Our probability of being a simulation is then >= 1/(1+that number). But we can't estimate those probabilities, so if he actually assigned a probability of 1/3, he probably just made a wild guess. – Ray Aug 2 '19 at 19:53
• @Josh If each civilization creates $n$ simulations capable of creating their own simulations with probability $P(n)$, then the expected number of simulated civilizations is the closed form of $f() = 1 + \sum_{i=0}^∞ i * f() * P(i)$. For some distributions P(n), this will be infinite and require infinite processing power (or infinite time), at which point Bostrom's argument really falls apart. – Ray Aug 2 '19 at 19:59
• @josh Maybe my notion of relative probabilities came from some commentary on Bostrom's thesis. I guess what you recall would explain the 1/3, but I do not know that there is any justification for Bostrom to assume the unknown probabilities are equal... WRT laws: Conifold has identified a common fallacy in these discussions (i.e. assuming 'our' laws are 'their' laws.) In a very tongue-in-cheek sense, I wonder if the growing 'weirdness' of known fundamental physics is because the laws of our simulation are not entirely consistent! – sdenham Aug 3 '19 at 18:30
• @Josh Bostrom's argument assumes that the toplevel civilization is capable of running perfect-fidelity simulations of the history of its own universe, up to and including the point where the universe starts simulations of its own, and that it chooses to do so. If you grant these assumptions, a lower bound on the expected number of simulated universes can be expressed as a simple recurrence relationship. But if you don't accept those assumptions (and in my view, you shouldn't; they're utter nonsense given everything we know of physics), then the probability I stated doesn't apply at all. – Ray Aug 4 '19 at 3:36

To answer your main question: Yes, it does make sense.

"One of his key assumptions is that other civilizations exist that can simulate us. Why do we make this assumption?"

That assumption is not derived from a physical evidence, but is an assumption in itself for the sake of argument. So, Bostrom's argument is that (as I understand him from this interview: https://www.youtube.com/watch?v=nnl6nY8YKHs):

1. If, just by their nature, civilisations don't destroy themselves (in a nuclear war, for instance) before they reach the levels to be capable of creating simulations as advanced as our universe is...

2. If, once reached the required levels, they don't lose interest in creating such simulations (e.g. due to ethical issues)...

3. Then, the conclusion is that we probably live in a simulation.

He also stated that the chances of the third one is less than 50 percent because the first two assumptions are pretty strong ones. Well, his actual structure of Simulation Argument is a little bit different, but I restated it the above way for the sake of simplicity. I think I didn't lose the main idea.

I, personally, tend to believe that we actually live in a simulation and here is my speculation:

With our technological advances it is not difficult to imagine our own civilisation if continued uninterrupted for thousands or tens of thousands of years would reach levels being capable of creating pretty realistic simulations. In 60s several megabytes of random access memory was almost a scifi, while now we have gigabytes of RAM in our cell phones, let alone nowadays supercomputers, and that's just in 50 years. Extrapolate that into future for hundreds or thousands of years and it is not difficult to see that realistic simulations are not so unrealistic. With this and Nick's arguments put together we are left to believe in either of these two:

• We are in a top-level universe (base reality, if you will). We are not a simulation but an actual reality.
• We are just another simulation among many other simulations.

Last I checked when humans thought of themselves being something special:

1. That the earth is the centre of the universe, the sun revolves around the earth, etc. The earth isn't even the center in our Solar system, let alone galaxy or the universe.
2. That there are animals and there is a human being, created in the image of God. Which turned out that we are just another species among thousands of other species (if not hundreds of thousands).

So, believing that we are in the top-level universe would, similarly, put us in a special position (odds were against us previously).

My another argument is that we live in a deterministic world. If we look at the cosmos, every state of the universe is determined by its previous state. We might think that us, humans, have a free will and we can influence the future. But our decisions are made by our brains, by atoms and electrons bouncing around in our brains. Position of atoms and electrons in our brains is determined by the previous positions of atoms and electrons in our brain. So, it's one chain. There is no branching of future, just one possible future, which awaits us. This is consistent with simulation hypothesis. The creators just had to input the initial conditions (laws of physics and constants, if you will) and push the button. Or the simulator automatically tries different inputs on each iteration (also a possibility).

So, "We are not special" + "Deterministic universe" arguments.

There is also an argument than if this is a simulation than its creators must be cruel, because of so much suffering in this world. I don't believe anyone is watching us real time. Let me give you a rough example. Researchers at OpenAI run simulations to train artificial neural networks to do meaningful things. For instance, in this simulation researchers try to teach agents to play hide and seek: https://www.youtube.com/watch?v=Lu56xVlZ40M

Millions of simulations are run and no one is watching them play around real time. The researchers are interested in the end results. If someday we discover that those red and blue agents in a simulation actually feel pain (just imagine for the sake of argument), it would definitely raise the ethical questions. There would be people who would advocate for stopping those simulations. But, the thing is millions of simulations will have already been run.

Millions of runs of the simulations of this universe have been conducted already and millions will be run after our universe ends. While I am typing this, our creators are in the process of blinking their eyes (assuming they have eyelids), just an ordinary day at their office. They will be still in the process of that same blinking in many universes to come. At the end of their work week they will look at the end results and raise the question, maybe, they should stop these experiments.

• Welcome to SE Philosophy! Thanks for your contribution. Please take a quick moment to take the tour or find help. You can perform searches here or seek additional clarification at the meta site. – J D Mar 11 at 15:57