Sunday, April 20, 2008

Are we living in a simulation?

Today I came across one of Dr. Nick Bostrom's existential philosophy papers regarding life vs. sim-life (aka are we living in the Matrix?).

To me, the really interesting question is the assumption of substrate-independence (because I don't believe we're living in a simulation, more on that in a bit) -- that sentience, sapience, and self-awareness can arise from any appropriately complex material, including computer processors. Is there some minimal complexity bound for intelligence? (First, tell me what you mean by intelligence.) On one hand, we already know that a virus is just a particular aggregation of molecules, and that any assemblage of those particular atoms will exhibit the same viral behavior. On the other, does that extend to a connection between viruses and the rest of the living world, and by analogy, to bottom-up intelligence?

Will computers be able to exhibit sentience, sapience, or self-awareness?

As an aside, although most people seem to know Seth Lloyd's paper on the ultimate limits of computing, I tend to prefer Warren D. Smith's Fundamental Physical Limits on Computation and Fundamental physical limits on information storage as being more useful equation-wise (and he has very interesting papers on election systems and voting, which doesn't surprise me when it concludes that our current voting system is nearly the worst mathematically possible, and that Range Voting is a much better algorithm).

(Note that some the papers are in PS format, so you will need a PostScript reader such as GSview along with GPL Ghostscript to read.)

Before I get too sidetracked, let me outline my reasoning for why I don't believe we're living in a simulation:

  • A simple quantum computer with 500 entangled pairs generates more information than could be simulated by any non-quantum computer in this Universe (2^500 >> 10E123)
  • If the Universe is not simulated to a quantum degree of accuracy, the simulation can be immediately exposed via Bell's inequality
  • Thus, in order to create a virtual universe sufficient to withstand experimental quantum physics tests, you need 10E123 qubits (e.g., the Universe)

Anyways, this is a very interesting topic, but I should continue my sidetracking avoidance and get back to my research.



Warren Smith doesn't say that plurality voting is the worst system possible; only that it has a lot of the properties you'd expect if you were trying to develop a bad voting method. There are much worse voting methods.

Also, the argument against our being in a simulation because of the amount of information in n entangled particles is irrelevant, since if this is a simulation we're living in, we have no idea whether the properties of the "real" world are anything like those of this fake one.

Also we don't have a way of even knowing whether we exist at all, thought that's a little more esoteric.

Clay Shentrup
Chief Evangelist
The Center for Range Voting

Adam Getchell said...


I've adjusted my comments about Range Voting (though it really doesn't change the overall conclusion).

However, I don't think you're understanding the argument with respect to Bell's theorem, so let me lay it out.

Quantum mechanics (and Quantum Field Theory) describe, not a single theory, but a series of mathematical rules for constructing theories. We choose those rules that reflect our observations.

If our current reality were virtualized, that means that the currently indeterminate state vector of a system, say |a>, is actually determined by the physics of the underlying reality running our simulation.

This is equivalent to a hidden variable theory. Bell's inequality provides a mathematical restriction on the limits of hidden variable theories, and experimentation has shown that our universe exceeds these limits.

Thus,there cannot be an underlying simulation universe that provides hidden inputs to our quantum mechanical state vectors, and that still complies with what we see today.

I'll have more to say about this later.