New research: There's a 50-50 chance we're living in a simulation and here's how to find out

… which in my view is a worthwhile exercise that everyone should engage in at least once in their life. If nothing else, it should help one bear in mind that empirical science is by definition a process of best approximations of whatever ‘reality’ is and how it works.


With that I have no problem. And obviously, I enjoy thinking about it even though I don’t take it seriously. I also enjoy thinking about theism and other unscientific questions. The simulation hypothesis just isn’t getting passed of as empirical science to me, because it ain’t.


I have just watched a nice presentation of Conways ‘Game of Life’, and how it is Turing-complete. We can simulate a computer within ‘Life’, and simulate ‘Life’ within that computer, and so on.

Here’s a slow zoom out on a Life pattern that is executing Life and also displaying it.

We can see that this takes perhaps a thousand clock cycles to one simulation clock cycle. We can all see that each super cell is about a thousand pixels square. However, the actual calculation is happening at the edges, and all the square is doing is showing ‘Life’ to us super-beings that are out of the plane of the game. So let us assume 1% of the cell is actually doing the calculation. We then have 1000x1000x1000/100 pixel-clock events in ‘Life’ to 1 pixel-clock event in the simulation. This is an approximation, but the implementation looks pretty efficient to me. So we have an efficiency ratio of 10 million to one. A similar calculation based on an 8-bit processor came up with a similar number. The exact number is not important - just note that the simulation is very inefficient. So, if all ‘Life’ was simulations of itself (which makes no sense, but it is an upper limit), the chance that any given clock-event is a simulation rather than real is 10 million to one.

Conway’s Life works on a regular grid. It is uniquely best placed to simulate itself. It is likely that simulating our universe with its extra dimension and no known grid is going to be hugely less efficient.

The simulation may be greatly sped up by omitting the details. We can render realistic scenes using general physics engines, rather that calculating every particle. That could give a scene in a head-up display that an observer cannot accurately distinguish from a real scene. While this is not ‘living in a simulation’, it is an experiment that we can do.


Sadly, it is usually the more ignorant, hateful, and profitable views that people choose to get caught up in.

I say it every time simulation comes up. Elon Musk is living in a simulation. It’s a simulation created by human beings in this universe to shield Musk from having to suffer emotional reactions to the consequences of the system that allows Musk wealth and power. Musk reaches for a fantastical idea of a simulated world to deal with the nagging suspicion that something isn’t right.

Also, for god’s sakes, all these “proofs” are just rehashed versions of the watchmaker proof for God’s existence, gussied up with new technology.


I get that, but at the same time this particular mess can have more drastically negative effects.

I referenced QAnon in my previous post. The ability of humans to go completely off the rails based on so very little that makes sense is incredible, and this particular farce is basically just thousands of years old beliefs with a new user interface.


Any sufficiently advanced simulation theory is indistinguishable from theology.


The whole idea of calculating the probability the universe is a simulation is complete nonsense. In order to calculate the probability of something you need to make certain underlying assumptions. If you selected a die at random from all dice in the world (hooray for thought experiments) and rolled it six times and got a one every time, you’d know the odds of that were one in 47,000 unless the die was loaded, so you might think the die is loaded. But that’s a terrible conclusion. First of all, you’d have to know how many dice in the world are loaded. If only one in 100,000 are loaded then it’s still more likely the die isn’t loaded. But there’s more. how many of those dice are loaded to roll ones? How much more likely are those loaded dice to roll ones than to roll other numbers (it can’t happen every time)?

So even if you could conduct an experiment that gives evidence this universe is a simulation (which I think is impossible to begin with) to judge the quality of that evidence and come up with an actual probability you’d need an estimate on the probability that any given universe is a simulation. If we count every single instance of Conway’s game of life as a simulation of a universe then it sure seems like by far most are. But we probably don’t. There is probably something about our universe that we want to stipulate a simulation must live up to before we count it in the pool of universes that we are talking about (much like only talking about dice that are loaded to roll ones). And we run into a problem there because I think I’m on very firm footing saying that we just don’t know what traits out universe has that would make another universe “similar” to it enough to count. Does it have to have to have quantum phenomena like we ones we observe? Does the gravitational constant or the speed of light have to be right? Does it have to contain conscious things (and if so, now we have to define what that meants).

People who actually think they can estimate a probability we are living in a simulation run on a thing we would recognize as a computer if we joined the simulators in their “reality” really haven’t given much thought about what they mean by that. And I swear to god what some of them mean is that one day they’ll “break out” of the simulation and wake up under a glowing sign reading, “Congratulations! You made it to the real world! (This is really, really, real)”


I think everything you wrote after your first paragraph is spot on. But no, I don’t need to know how many dice there are or how many are loaded to update my beliefs based on what I see. You’re treating the assumptions as strict numbers, when what you need is to assign a probability distribution over all possibilities. I have plenty of data to conclude “most dice I encounter are not intentionally loaded,” and data on whether I know where this die came from and why I’m rolling it, all of which I can update based on. If you are going for strict Bayesian probability calculations, everything is evidence about everything else, no matter how quickly that becomes computationally intractable in practice. We all make these kinds of assumptions all the time without even noticing, because there’s no other possible way to function in the world. At least in your example I can push the die partly off the edge of a table in different orientations to see where its center of mass is to quickly and roughly test my guess.

Basically, Kipping is arguing that yes, with regard to the simulation hypothesis, we are in a state of maximum uncertainty, and then he goes on to claim that the maxentropy prior we should adopt is exactly 50-50. I think you and he are in agreement about the first part of that claim, but not the second?

Definitely. My thought experiment began with the rather impossible selection of a random die from the world and went on to try to calculate the precise probability it was loaded. In real life we constantly balance all kinds of information we have.

So the fact that we don’t know the precise number of dice that are loaded in the world or whatever doesn’t invalid our suspicion that the die in front of us rolls way too many sixes (I had three dice I used to roll D&D character stats, I am 100% sure they rolled 4,5,6 too often). You skip the precise data collection part and go with accumulated knowledge.

The problem is when you try to apply the same thing to simulated universe it doesn’t work. I have literally zero experience with or information about the different possible kinds of universes out there. The intuitive systems that help us get along in day to day life point us in the total wrong direction in this case. It may be the case that this is really the only way that a universe can be. It could easily be that running a simulation of the type of universe we are in just isn’t a thing, that’s it’s pure fantasy based on a misunderstanding of what the universe is. So I don’t think I really agree with any part of what Kipping says. To me, Kipping might as well being say that we are the dreams of magic dragons until they can address in an I’m-pretty-sure-impossible way what is being simulated and what it means to be simulated.


During the Enlightenment, belief in a clockwork universe was popular, because clockwork was the most sophisticated technology people could conceive of at the time. With the advent of mainframes it was popular to think of our minds as being programmed, because how a mainframe worked was the next most mysterious thing after our own minds.

Basically, when trying to describe the weird, unknowable nature of reality, we use metaphors. A compelling metaphor in any age is to compare the unknowable to the most cutting edge technology. It works because most people have no idea how that technology works either, so the comparison feels compelling. In retrospect it turns out to have missed the point in important ways, though.


In this case, I think it may be simulators all the way up.


Well, certainly a portion. But that’s a pretty wide net you’ve cast – the majority of the world’s populace hold spiritual or religious views, and I don’t believe all of them to be more awful because of it. Hasn’t been my experience, that it’s nearly so universal. A lot of people with these views don’t wear them on their chest quite so loudly in day-to-day circumstances – by definition you’re seeing more of the noisy ones.

Although unlike “pure” theology, it could potentially be scientifically investigated.

Fair enough. So then, you’re both claiming complete ignorance on the matter, which is reasonable, with the difference that he is saying that that still means he has some maxentropy prior for it being right or wrong, but chooses a ridiculously wrong one, whereas you’re claiming to not have any such prior, on the basis of the hypothesis not being sufficiently specified in any way we could possibly evaluate.

I think I take a bit more expansive of a view of what counts as evidence for this kind of thing, but also think that it really doesn’t matter as much as some people seem to think. The world is real whether it is a simulation or not. It’s the same as my response to thought experiments like if you made an exact perfect copy of me: there wouldn’t be an original or a copy, there would be two originals.

1 Like

I think at this point we’re just a backward-compatibility layer to keep the earth running long enough for it to kill us off and get back to normal. The primate process seemed like a good idea at first, but it’s gotten worse with every update.


It is. The problem is that it isn’t funny any more. :confused:


ps -ef | grep primate
kill -9

It always seems to me that the non-continuous bits of reality, like planck constants, give off the feeling of seeing the dithering effects of running on a digital platform.


I see you omitted the bit of my post that literally covered this entire point. Which is a bit of a cheap shot, don’t you think?

1 Like