Praying for Simulator Intervention
An essay in which I provide disconfirming evidence that we are living in a simulation
Philosopher Nick Bostrom has an extremely interesting and weird argument that we could be living inside of a simulation. This argument has garnered some mainstream attention and seems to have become a bit of a meme. Bostrom’s paper “Are You Living in a Computer Simulation?” proposes the trilemma:
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.
Basically, the idea is that eventually intelligent life is going to achieve technological advancements so profound that it permits the simulation of ancestral populations. For example, perhaps if humans had the computing technology, they would want to explore a bunch of counterfactual questions like “What if there was no American Revolutionary War?” or “What if we made humans way more tribalistic?”. Bostrom says there are a number of reasons that people might want this in his essay “Why Make a Matrix? And Why You Might Be In One”:
Why might they choose to build a Matrix like our reality? One can think of many possible reasons—setting aside the daft idea of using human brains as batteries. But perhaps future historians would create a Matrix that mimicked the history of their own species. They might do this to find out more about their past, or to explore counterfactual historical scenarios. In the world of the Architect(s), Napoleon may have succeeded in conquering Europe, and our world might be a Matrix created to research what would have happened if Napoleon had been defeated. Or perhaps there will be future artists who create Matrices as an art form much like we create movies and operas. Or perhaps the tourist industry will create simulations of interesting historical epochs so that their contemporaries can go on themed holidays to some bygone age by entering into the simulation and interacting with its inhabitants. The possible motives are myriad, and if future people are anything like present people, and if they have the technological might and the legal right to create Matrices, we would expect that many Matrices would be created, including ones that would look like the world that we are experiencing.
If I were in a simulation, I don’t know exactly what the purpose would be, but it seems possible that the simulation is being monitored by the simulators. They could just using the simulation to gather data or train an artificial intelligence, but it could be the case that someone is running it for amusement, like playing a game of Sims or SimCity. If that is the case, then it is possible that the simulator could be inclined to modify the simulation.
If the simulation could be modified and the simulators are monitoring the simulation, then it would be possible to petition for intervention—kind of like intercessory prayer for Christians. If there are billions of simulations, it may be difficult to get attention. If I had to think of a topic that would attract attention from simulators, it would be discussion of the simulation itself. Perhaps, that would raise some flag for a conscious entity to investigate.
I find this all very improbably, but if (1) we are living in a simulation, (2) you can attract simulator attention, (3) simulators are willing to intervene in the simulation, then it is at least possible to provide confirming and disconfirming evidence of the simulation hypothesis. For example, if I did 50 coin flips and asked the simulators for each one to be heads and that occurred (E), I would update my priors toward the simulation hypothesis (H) being more likely because P(H| E) > P(H). For 50 coin flips, this would be a major update because the probability of that occurring is 1/1125899906842624. However, if the experiment fails, it would provide only very weak disconfirmatory evidence.
Using Siri’s flip a coin feature, I got: Heads Heads Tails. I think that this is unimaginably weak disconfirming evidence for the simulation hypothesis and by extension weak evidence for an approaching doomsday. Is this the first test of this kind? I’d be interested to see someone else using a method like this. Let me know in the comments. I didn’t find anything after a brief search.
This is all somewhat out there, but there might be a somewhat relevant moral implication. If the post-human stage will be achieved and they will be interested in many ancestral simulations, but we are not living in a simulation, then it means simulation may not be possible. This is a massive if, but I find it interesting nonetheless. If simulation is impossible, then it may be because Bostrom’s substrate-indifference assumption from his original paper is false.
A common assumption in the philosophy of mind is that of substrate-independence. The idea is that mental states can supervene on any of a broad class of physical substrates. Provided a system implements the right sort of computational structures and processes, it can be associated with conscious experiences. It is not an essential property of consciousness that it is implemented on carbon-based biological neural networks inside a cranium: silicon-based processors inside a computer could in principle do the trick as well.
If that is the case, then we may not have ethical obligations toward simulations. If we do have ethical obligations towards simulations, then I think simulation happiness would eventually outweigh human happiness and become a utility monster. I explain this position in my essay “Artificial Intelligence as the Ultimate Utility Monster.” Personally, I am skeptical of the substrate-indifference argument and so, I don’t think that Bostrom’s simulation argument works. I realize this is pretty outlandish, but it’s all in good fun, right?
Praying for Simulator Intervention
"(2) any posthuman civilization is extremely unlikely to run a significant number of simulations of their evolutionary history (or variations thereof)"
Should this have said "likely" instead of "unlikely"?
There are many non-simulation-hypothesis explanations for a sequence of coin tosses with many heads (or tails) in a row. ("The coin is a trick coin" being one of the simpler ones.)
P(H | S) > P(H) _is not enough_. You must also have P(H | S) > P(H | M) for _all_ other Mundane hypotheses.
For an explicit example.
I have a prior of S=10^-6, and M=10^-4.
(In the below I'm using lowercase X to mean ~X. So e.g. Sm means 'Simulation and not Mundane')
I know the following probabilities:
P(H | SM)=4/5
P(H | sM)=4/5
P(H | Sm) = 3/4
P(H | sm) = 1/2
I flip a thousand heads in a row. What is my resulting posterior?
1000 heads:
SM: 1e-06
Sm: 9.359075874475835e-31
sM: 0.999999
sm: 7.585312510276936e-201
S_: 1e-06
_M: 1.0
...oh. The dominant hypothesis is the mundane explanation.