Monday, March 26, 2018

Junk Science: Universe As Simulation Theory

On CBC Radio's "The Current" this morning, there was a long segment about the bogus hypothesis that the universe we live in is really just a giant computer simulation. I've heard this concept before - it's rattled about in computer science circles since I was an undergraduate student in the 1980s.

Superficially, this is little more than the philosophy "Brain in a Vat" thought experiment.  It has some interesting uses for those interested in the subject of ontology, but it makes for awful science.

At its most fundamental level, it's bad science because it is unfalsifiable. A little like the question of "does God exist?", testing it requires the ability to inspect our universe from outside.  So far, we have no means of even describing how that might happen, much less making it actually occur.  This is one of the reasons that any hypothesis that ends with "and therefore God" (or some variation thereof) is fundamentally not science.  The minute you can invoke an unknowable external entity which has no meaningful description, you are simply not engaging in science.



Since I am not a theoretical physicist, I'm not going to go after the idea that we live in a simulation through that lens.  I'm fairly certain that people who are a lot smarter than me can explain just how problematic the idea is in understanding the universe. The physics of the universe are complex and only partially understood today.

From a computational perspective, it's much easier for me to critique.

Limits of Current Computing

At its most fundamental level, current computer architectures are not adequate for modelling the kind of complex, dynamic system that our universe clearly is.  Digital computing is finite and deterministic in its processes - everything ultimately must produce a discrete result.  Our universe is quite different in this regard.  It tends towards equilibrium, which is an analogue state of relative stability.  This is, of course, miles removed from what is readily modelled in a computer. Even modelling an integral in a computer is at best a coarse approximation.

As yet, we have not been terribly successful in modelling even the most fundamental aspects of biology in computing.  Attempts to model intelligence have given us such wondrous sounding technologies as neural networks, but even those cannot be said to accurately model their biological counterparts.

Analog computing might eventually provide us with modelling that extends beyond the current limitations of digital computing, but progress in that domain has been limited, overshadowed by the rapid pace of development in the digital computing domain.

Practically speaking today, we would have great difficulty building a simulation of an analogue universe which would approach the level of complexity needed to describe the complex universe we live in.

A Thought Experiment

However, for a moment, let's engage in a thought experiment.  Assume that we could create a simulation that describes a "digital universe" which would work within the frameworks of our current computational abilities.  Further, let us assume that this universe forms a digital equivalent of Abbott's Flatland - call it DigitLand.  We can "peer in" to this universe and observe the goings on in DigitLand.

Just as the physical nature of our universe makes it impossible for us to step outside of our universe, the virtual nature of DigitLand would render it impossible for the inhabitants of DigitLand to step out of their own universe.  Any program that runs in a computer is only able to reach beyond that computer to the extent that the computer itself has connected inputs that it can access.  So, on that basis alone, it is impossible for the inhabitants of DigitLand to test the idea that there is some metaphysical (metadigital?) world beyond their universe.

Just as the Halting Problem makes it abundantly clear that writing "code to inspect other code" is deeply problematic, so is trying to break out of your current universe using the laws of that universe. Godel's Incompleteness Theorems serve quite nicely to generalize this principle beyond the Halting Problem.  Formal systems have limits, and it has been well demonstrated that you cannot use a formal system to prove the completeness of itself ... bad things happen when you try.

So, even if we are part of a "giant computer simulation", the idea that we "prove" this within the laws of our universe seems highly unlikely.  It would be impossible for the denizens of DigitLand to reach out of whatever computer their universe was hosted in, and we are similarly limited in our universe if it were part of some kind of grandiose simulation.

Counter Arguments

There are a number of counter arguments that could be postulated to the objections I have raised:

1. A Meta Universe Could Have Different Laws

Absolutely possible, and such concepts have been explored within various branches of modern Cosmology and Theoretical Physics.  We don't need the idea that we are part of a simulation to meaningfully explore this idea.  

2. More Advanced Computing Abilities

We can imagine all sorts of things here. However, doing so is more an act of writing science fiction than it is one of science.  There is no meaningful way that such a proposition could be inspected, and as such until there is a reasonable way of inspecting it.  Guessing that certain patterns might occur in the universe or in matter is unlikely to be persuasive unless there were no other possible explanation. 

3. A Mathematics That Supersedes Godel's Incompleteness Theorems

From my point of view, this is a "come and talk to me when you achieve that".  I won't say it's "impossible", because I like to believe that we are always able to come up with new, and innovative constructs. The fact we haven't for a few centuries doesn't mean it can't happen.  However, it hasn't, and supposing such a construct is similarly irrelevant. 

4. Pascal's Wager

Often the last gambit of the faithful in arguing for the reality of God, Pascal's Wager is a "but what if I'm really right" argument.  Pascal's Wager is a desperation argument.  This can ignored because it fails to introduce any new information into the discourse.  

Fundamentally, proponents of Simulation Theory desperately need to introduce some new information into the equation before their idea is much more than a piece of science fiction.  From a computing perspective, the idea seems to ignore some fairly fundamental problems in both mathematics and computational domains, and it presents us with no compelling new evidence which would suggest the concept is valid beyond the thought experiment it began as. 

What Is Simulation Theory Really?

The more I look at it, Simulation Theory falls into the same bin of junk science as so-called "Intelligent Design".  It is an attempt to wrap the recurring quest to "find God" in the cloak of "respectable science".   

Science is fundamentally atheistic because any idea that ends with "and therefore, God" is fundamentally unfalsifiable.  There is no way to inspect, test or challenge the truth of that last clause. The Kitzmiller v. Dover Trial brought that point to the foreground of Intelligent Design (ID), and demonstrated that even though ID's proponents had been very careful to avoid the language of God, fundamentally their concept was little more than an attempt to twist the interpretation of science to derive the existence of God.  

Simulation Theory is Theoretical Physics' parallel to ID.  It is at best a mildly interesting thought experiment, at worst, a Trojan Horse for wedging religion back into science.  



No comments: