Feb 13, 2017

6 min read

The Degeneracy of the Simulation Argument

The Fermi Paradox is a real thing, and these days I feel outright annoyance when people fail to make any decent accounting of it within their arguments. While some refutations are rationally plausible (the Doomsday Argument is one) they leave much to be desired due to a feeling of immense unlikeliness of their assumptions.

After having made it through the Superintelligence book by the same author, it’s about time for me to come back around to the paper arguing the Simulation Argument.

After a 2nd or 3rd review of the arguments, my underlying take-away has firmed up — The Simulation Argument doesn’t answer any question I had an interest in. Even if I accept it at face value (which is a big if), then it does nothing to resolve the problem of the Fermi Paradox. If anything, it presents a new philosophical problem that may or may not have any practical implications.

More than anything else, the Simulation Argument is a misquoted academic line of reasoning.

Reconciles the observational fact that we are alone in the universe, because advanced civilizations find the “lonely blue dot” phase of history vastly more interesting than other times relative to the number of individuals living.

Given that we exist in this phase of history, it predicts what type of existence we are. This presents a resolution to why we live in an unlikely phase of history, but only relative to another another phase of history.

Replacing one Fermi Paradox for Another

Instead of using the equations themselves, let me ballpark some numbers. The number of people alive today is 10⁶ and at issue is how we explains why we should happen to be among that small number compared to the number that an advanced civilization could potentially maintain. I will take the latter to be on the order of 10⁶⁰. That means that our probability of being an early human compared to a latter post-singularity individual would be 1 in 10⁵⁴. Very unlikely.

Allowing myself to get more speculative, I’ll assign a number to ancestor-simulations that a civilization like ours will go on to conduct. I’m using the terminology and conventions from the paper, and 1 ancestor-simulation means that the full history of humans is simulated, which I’ll have to ballpark as 10⁶ here, with the acknowledgement that it should really be one power-of-ten higher (bickering over the exact numbers raises just as many questions, if not more, than addressing the points themselves does).

I consider myself generous in assuming that the expected value number of ancestor-civilizations created by that which evolves from Earth civilization is 10¹⁰. This is accounting for failed civilizations, so if our probability of surviving long enough to become durable space-faring people is 1 in 10⁴ then each surviving space-faring civilization must produce 10¹⁴ ancestor-civilizations, for a total of 10²⁰ simulated individuals.

The Simulation Argument accepts our human-like experience as a Bayesian prior, and from that explains that we have only a 1 in 10¹⁴ or 10¹⁰ (depending on your scenario selection) chance of experiencing a genuine human experience versus a simulation one.

But recall the strictly observational flavor to the Fermi Paradox. The classic Paradox merely states that it would be unlikely to be born as a human-like existence if some other advanced civilization may eventually come about which hosts a vastly larger number of individuals.

The Simulation Argument only narrows the improbability from 1 in 10⁵⁴ to 1 in 10⁴⁰. In the world I described, there is an approximately a 1 in 1 chance of being born as a normal citizen of the advanced civilization and only marginal possibilities of being born as a citizen who is lied to in this very specific way such that they believe themselves to be a long-since dead ancestor.

Extremely Impractical Lies

If the “singularity” demarcates the time after which civilization can control the substrate of existence (like transferring a brain from flesh into silicon), then I don’t doubt that many people in this world will be effectively “lied to by their parents”. Society expects some people to contribute a great amount of toil to the progress of particular projects. If the means were available to make the person more effective at their projects by reshaping the apparent reality, then it may follow that they will do so.

This makes no good argument for doing the vast environmental simulation necessary to make the virtual world (the virtual Earth) convincing. There’s nothing readily apparent about the intellectual products we produce that would justify either the ethical hazard of putting us in a simulation or the substantial cost necessary to do so.

Even if you made the completely bonkers assertion that an advanced civilization would devote 50% of its resources to ancestor-simulation, then you would still rationally conclude that the “normal” citizens would vastly outnumber the human-like citizens living in the simulation. Making a convincing reality is much more expensive than simply granting access to the basic facts of the civilization that the individual is living in. So this rules out any very close correspondence between the simulated individual number and the total individual number. I gave a reasonable breakdown of 10⁶, 10²⁰, and 10⁶⁰. I think that giving answers like 10⁶, 10⁵⁰, 10⁶⁰ would be preposterously optimistic about both the efficiency of creating a full conspiracy reality as well as overly-pessimistic about the efficiency of a straightforward internet-connected virtual brain. Keep in mind that I haven’t yet even put in the number for failed or convergent independently evolved civilizations which would boost the first number without budging the 3rd number, further eroding any value that the Simulation Argument added to the resolution of the Fermi Paradox.

No matter how you slice it, it is extraordinarily unlikely that an arbitrary individual is not born as a “normal” member of the advanced civilization. It follows that the Fermi Paradox persists, with little revision to the numbers.

Any Other Options Out There?

I can be conciliatory to saying that the Simulation Argument stands a reasonable chance of making intellectual progress of some time toward genuine answers, but it should be relatively apparent that our knowledge of the universe is just so deeply lacking that we don’t have the answers. But we should be clear that, even if fully correct, the Simulation Argument only barely erodes the vast gulf of improbability that must be explained about our existence.

In the first place, defining the factor of the number of ancestor-simulations that an advanced civilization will run is a matter of prioritization relative to available resources. We might as well go ahead and assume that the fraction of resources spent on this (apparently pointless) endeavor is small. The civilization will be certain to have a great number of priorities, which may be exponentially more numerous and diverse than those of society today as we know it.

It might be good to narrow in on some Elephant-in-the-room priority of an advanced civilization. Depending on what research into fundamental physics yields, I’m tempted to believe that black hole engineering of baby universes is an extremely appealing candidate for this. It might be the case that they can create alternative universes that don’t just drain their own resources, but truly produce something from nothing, and may go on to make their own baby universes. However, I will in-advance volunteer the fact that this does not solve the observation of the Fermi Paradox because recursing the process of spontaneous life generation and evolution into advanced civilizations doesn’t ultimately change the ratio of advanced society individuals to human-like individuals.


There might be some kind of argument rooted in the mathematics of infinite sums. It is entirely plausible that 10⁵⁴ individuals of an advanced society could spawn 10¹⁰⁰ genuine individuals living in a primitive society. The only way this works is because those will go on to turn into 10¹⁵⁴ individuals of the 2nd generation advanced societies. Intuitively I do not think this stands a good chance of resolving the Fermi Paradox observation, but I have not done an actual mathematical proof and it may still be the case that it does.