Why We Almost Certainly Live in a Simulation — And Why We Probably Don’t

On the appeal of the simulation hypothesis and the problems it ignores

We probably live in a simulation. This is not some conspiracy theory. It is an idea that has several strong arguments behind it. This article is a review of the most popular arguments in favor of the simulation hypothesis, an analysis of how strong those arguments really are, and how they can be refuted.

The main argument in favor of the simulation hypothesis is Bostrom’s Trilemma. The trilemma states that at least one of the following three statements must be true:

1. Almost no civilizations reach technological maturity, also called the posthuman era.
2. Mature posthuman civilizations choose not to run many ancestor simulations, meaning simulations of their history or variations of it.
3. We almost certainly live in a simulation.

First, let’s discuss why all of these statements cannot be false at the same time. Let’s assume the opposite. There are many civilizations that reach posthuman-level technology. Many of those civilizations decide to run simulations of their history or variations of it. One base reality can produce many simulations. In addition, simulated realities can themselves produce even more simulations at deeper levels. This means there are a countless number of realities that are indistinguishable from the base reality, while there is only one base reality. Mathematically speaking, this implies that the probability of any particular reality being the base reality is almost zero.

This reasoning implies that for us to have a reasonably high probability of living in the base reality, at least one of the first two statements must be false. So either most civilizations disappear before reaching advanced technology, or they reach it but decide not to use it to create realistic simulations of their world.

This argument is often called the simulation argument, but on its own it does not claim that we live in a simulation. To reach that conclusion, we need additional reasoning explaining why the first two statements are unlikely to be true.

The claim that almost no civilizations reach technological maturity, or that those which do choose not to create simulations, is both difficult to prove and to disprove. At the present moment, we have no knowledge of other civilizations, so we cannot know whether there is some universal reason that prevents civilizations from advancing beyond a certain technological point. There could be countless reasons for civilizations to disappear, ranging from external factors like asteroid impacts to internal ones like global wars. However, there is no clear reason to believe that there exists a single common factor that stops all civilizations from reaching the required technological level.

Some scholars argue that the trilemma is self-defeating, claiming that the apparent absurdity of us living in a simulation implies that one of the first two statements must be true. This argument makes little sense, since there is no proof that the claim that we live in a simulation is inherently false or absurd.

Others argue that there is no proof that sufficiently advanced simulation technology is even possible. However, given the overall trend of technological development, the many times in history when humans created technologies once thought impossible, and the existence of similar subjective experiences such as dreams, it seems likely that hyper-realistic simulations could be possible at some point.

Another objection is that advanced civilizations might deliberately choose not to simulate realities, even if the technology exists. This claim has little concrete basis. Based on human history, it is reasonable to assume that if such technology were possible, humanity would attempt to simulate reality. There is no clear reason to believe that something would universally prevent civilizations from doing so.

From all of this, it may seem extremely likely that we live in a simulation, since the other two claims in the trilemma appear unlikely. However, this is where the main problem arises.

The premise that this argument is a trilemma is itself misleading.

In reality, all three statements could be false, and other explanations may exist.

My first argument against Bostrom’s Trilemma is that this technology might simply not be possible in the way the argument assumes. If we really are living in a simulation, then it has to be possible to create not just intelligent behavior, but actual conscious experience. And right now we do not even have a clear idea of what consciousness is. Most definitions are circular, or just rename the problem instead of explaining it. The trilemma more or less assumes that consciousness can be reduced to computation or physical simulation, but that is far from obvious. It may turn out that consciousness depends on very specific features of reality that cannot be recreated artificially, or that only exist in a base-level physical world. Even with all the recent progress in AI, what we see are systems that copy patterns of behavior, not systems that we have any reason to believe are actually experiencing anything. So the assumption that conscious beings can be created inside simulations is doing a lot of work here, and at the moment it is still a big open question rather than something we can safely treat as likely.

My second argument is that even if creating consciousness were possible, simulating an entire world might still be out of reach. To get anything close to our reality, a simulation would have to deal with an enormous amount of physical detail and somehow keep it consistent over time. We are talking about interactions between something like 10^80 particles, all following laws that cannot just break at random. Even thinking about this already starts to feel unrealistic. Right now, we cannot properly simulate a single human brain, and even fairly simple physical systems, like fluids, can be difficult to simulate.

Now imagine scaling that up to billions of conscious beings interacting with a complex and constantly changing environment.

At that point it is not just difficult, it starts to look like the kind of problem that might never really be solved in practice.

One might object that a simulation does not need to model everything, only what observers actually experience. But even this kind of simplified approach runs into serious problems. A world that feels real over long periods of time has to remain consistent across memory, history, and interactions between many different observers. Once you have large numbers of agents sharing the same environment, shortcuts start to break down. Objects need persistence, events need causes, and past states need to remain compatible with the present. Over time, maintaining even the illusion of a stable world may require something very close to a full simulation anyway. And even if such a simulation were technically possible in some abstract sense, it could easily be so expensive and resource-hungry that there would be little reason to actually build it. In that case, we end up back at the second option of the trilemma, where advanced civilizations simply choose not to run these simulations, not for moral reasons, but because the cost just does not make sense.

My final conclusion is that while Bostrom’s Trilemma sounds extremely convincing, it is misleading to call it a trilemma. There are many important factors that it does not account for. Aside from Bostrom’s Trilemma, which might be more accurately described as Pentalemma, there are no strong independent arguments in favor of the simulation hypothesis. This makes it much more likely that we live in a real world rather than in a simulation.