Monday, December 17, 2012

Some Ongoing Problems of the Simulation Argument

For Simulation Argument background, go here.  For previous articles on the topic, go here.
Summary:   the definitions of simulation, and whether the origin and purpose of the simulation (if any) matter to the discussion, are sloppy perhaps to the point of meaninglessness.  First, it is often supposed that the simulators can perfectly avoid detection by patching laws of physics, erasing memories if they are discovered, etc.; if this is the case, then in principle the answer to the simulation question is unknowable and is a PEP (pointless epistemological problem).  There also may be no way to distinguish simulators revealing themselves from super-intelligences.  

The distinction between simulator-universe and simulated universe is incoherent and breaks down.  Simulated entities exist as real entities in the simulating universe.  Although the simulated entities may be able to interpret information only from a narrow slice of the universe, e.g. the hard drive they're running on, all consciousness is necessarily a provincial representation of information, each living in its arbitrary simulation of the universe; that is, the simulators might have a broader view than the simulated, but there is only a difference of degree, rather than of kind.  Implied in arguments about simulations is that conscious simulators exist and are intentionally deceiving us, but even if we are only perceiving some narrow slice of spacetime we cannot assume we know anything about simulators' intentions or even that intentional entities are directing the simulation at all, and in any event this information is irrelevant to the argument.  Even if we do show that we are in something usefully called a simulation, this is more likely to (again) expand our idea of the dimensions of spacetime, similar to the way astronomy expanded our knowledge of the universe outside the Milky Way a century ago.

There are several problems with the simulation argument.  Some of these problems involve the definitions of terms in the argument, which as commonly understood seem very sloppy almost to the point of meaninglessness.  The question was originally put as a special case of the self-indication assumption, where we can assume that if humans develop the ability to simulate historical events, they will; and that if they do, any given conscious being is overwhelmingly likely to be a simulation.  This is one way to make the experiment concrete, but it is unnecessarily provincial; it can be re-stated by saying that if conscious entities can be created within simulations, then any given conscious being is likely to be a simulation.  The two questions here:  what is a simulation, and what is its origin and/or purpose?

1) What is meant by "simulation"?  Typically this is conceived as a world of "false" sense experiences created by intentional agents ("real" humans, AIs, aliens, etc.); and their intention is apparently to deceive us.  (Already we're making unwarranted assumptions about the nature of the simulators, including their very existence, but more on this in #2 below). This classical, provincial way of imagining a simulation is basically a Matrix simulation, except where there are not necessarily bodies in the "outside" world corresponding to each consciousness.

In a very real sense, we are definitely living in a simulation - we experience certain sense data (visible light but not microwaves) and knit it together in one certain way but not another, along withour beliefs and emotions and our somatic senses that are all internal and subjective and invented.  We are creating this sensory experience in a not-at-all necessary way; we have built a certain kind of simulation of the world beyond our nervous systems.

The difference between a simulation/not simulation seems to blur into one of degree rather than kind.  If you put on rose-colored glasses, are you now in a simulation?  How about if you become schizophrenic and hear voices and believe the CIA is after you?  For most people adhering to the classic use, these are inadequate to call a "simulation".  What about a Derren Brown-style manipulation where people are seeing the real world, but props and people are being moved around in such a way as to convince them of something not true? Further down this path, how about a DMT trip, where your sense world is completely replaced?  The experience of DMT is not mere hallucination-icing on top of the consensus reality cake as with LSD - you're completely in another world.  Are those experiences simulations?  If a DMT trip is not a simulation but the one in the Matrix is, the definition seems (very strangely!) to hinge not on the content of the sensory experience but on whether there are deliberate controllers intentionally deceiving the simulatees moment-to-moment.  Method of deception and intent of controllers both seem spurious considerations in thinking about such an idea.

There's a further problem with the hierarchical conception of simulations.  In one sense, there's a very clear hierarchy - a baseball bat in the world of the simulators would end the experience of the simulated.  At the same time, the simulated entities are every bit as real as the simulators.  The simulators can use instruments to show how magnetic fields on certain areas of the electronic medium (hard disk, volatile memory, whatever) are coherent entities with prolonged, distinct existence.  Look!  There goes Jake, that pattern of 0's and 1's right there!  Of course, that pattern of zeroes and ones is having the subjective experience that he's playing frisbee in a park, only because his information-processing system is knitting together the events of magnetic fields on a hard drive in a certain way, just like you're knitting together the events of electromagnetic radiation and temperature and pressure waves in a certain way.  I'm intentionally avoiding questisons of whether information equals consciousness, but Jake certainly exists in the simulators' world, just in ways he doesn't understand.  Again this makes the simulator-simulated distinction collapse, since it is certainly the case that there are aspects of ourselves we don't understand that are obscured by the way our nervous systems work.

2) Who or what are the simulators, if any exist?  First, if we are simulated, and there are simulators (two different things!), then do the simulators' intentions matter to this argument, i.e. whether they are trying to deceive us?  Furthermore, assuming we're "running" on some computer in a wider metaverse, how can we say for sure that the physics of that universe demand active entities to build a computer?  Maybe there's a metal-rich moon somewhere on which there was some kind of natural selection for coherent spin-flipped-domain entities - software.  This is akin to the idea of a Boltzmann brain.  Either way, assuming we're in a historical simulation built by future humans hellbent on continuing our deception, and that we have a "real" body waiting for us to wake up, is a very narrow conception of possible ways our perception of reality could be systematically narrower than would otherwise be possible.  (For an exercise in throwing out unwarranted assumptions when you're asking a question that cuts so deep into reality, read Nozick's Why Is There Something Rather Than Nothing in Philosophical Explanations.)

3) Pointless Epistemological Problems (PEPs).  There are a number of ideas (the technological singularity is another, as is the god of several religions) where a concept is argued to be fundamentally closed to human reason - that we cannot, even in principle, ever understand it.  This results in unfalsifiable arguments.  "We can't know if the singularity occurred because we're not bright enough to recognize AI behavior" is the same as saying "It might already have occurred and we can't know."  Similarly, the simulation argument is vulnerable to lots of smart-ass answers - if humans ever figure too much out about our simulation, the simulators will just hit pause and fix it, or they'll alter the software detailing that person's mental state and that person will forget, and we'll be back to square one.  So if we can't know, why even talk about it?  If we think the rules of figuring out the truth make it worth our time to entertain such ideas, then we must certainly also discuss at once the theory that not just the moon, but the whole universe, is made of green cheese.

Previously I wrote about an experiment that physicists recently proposed to test the simulation argument.  Although experiments around Bell's Theorem have repeatedly not supported local hidden variables (by some interpretations, discrediting simulation arguments), suppose that the new experiment shows unambiguously that we are in a simulation.  What then?   How does this new knowledge affect our future actions?

This experiment can show us nothing about the nature or purposes of the simulators, or indeed, that they exist at all - and maybe future experiments can.  For now, all we'll know is that the slice of spacetime we perceive is a smaller part of a bigger whole.  The justified change in our worldview will not be to suddenly resign ourselves to being  meaningless play things in an alien god's video game (which we must have been the whole time).  It will be to realize, again, that the universe is bigger and stranger than we knew before, more akin to astronomy's discovery of a universe outside the Milky Way, or the standard model's intuitively incomprehensible higher dimensions.  That is to say, the physicists doing the experiment, if their result is positive, will be remembered more like Copernicus than Morpheus.  Then we can begin to explore the physics of the universe outside our narrow slice of it.  If you're familiar with Conway's game of life, imagine you're the simulator, and you leave it running it overnight to find they've built a glider gun that has deduced the real law of gravitation when you get up in the morning.  That's what our job becomes.

No comments: