Making sense of EPW

Jeff Bone jbone at
Tue Apr 29 01:02:20 PDT 2003

On Monday, Apr 28, 2003, at 23:28 US/Central, Russell Turpin wrote:

> There are some obvious analogies, but I think the
> Every Possible World ontology really is much more
> radical. Cosmologists envision bubbling universes
> where some fundamental constants may vary but
> physical law is the same, and MWI posits every
> possible universe that obeys the laws of QM and
> is a quantum mechanical branching from a quantum
> mechanically possible past. EPW goes far beyond
> either of these...

As you'll see if you dig into this, EPW doesn't require throwing the 
baby out with the bathwater.  Even if all possible worlds exist that 
doesn't imply that they're all equally likely.

> In essence, EPW denies the distinction between
> 'real' and 'pretend,' except in some sort of
> complexity ordering, i.e., some universes are
> simple enough that we can wholy imagine them...
> It's just all too Rudy Ruckerish to me.
> I'm surprised, though, that there's not a reference
> to this viewpoint.

This view's got some similarities to (and roots, for me, in) Egan (Perm 
City [1] and others) as mentioned, and I'm sure a host of other 
writers, both sci-fi and scientists.  There's been a lot of stuff about 
this over the last few years.  In particular, you're going to want to 
check out Jurgen Schmidhuber's _A Computer Scientist's View of Life, 
the Universe, and Everything_ [2] and other stuff [3,4].

Now let's rescue reality.  There are a number of measures by which 
different universes could be regarded as, in some sense, more "real" 
than others.  (More probable = more real, etc.)  Note that 
Champerknowne's number in its infinite expansion contains all possible 
finite bitstrings within it.  (Bonus points:  prove that all finite 
bitstrings occur as substrings an infinite number of times in 
Champerknowne's infinite expansion.)  But the distance between 
instances of particular bitstrings, or classes of bitstrings, etc. 
might be greater or less vs. other bitstrings / classes.  You could 
then regard those recurring bitstrings as in some sense "more probable" 
than other bitstrings.  (if it's not clear, we're interpreting these 
bitstrings as snapshots or slices through an uber-phase space;  
increased frequency would then mean that certain states would more 
likely occur in any random sampling.)  The distributions of such across 
the entire expansion might have some kind of relationship with certain 
hypotheses about the distribution of primes, etc. i.e. a kind of 

Anyway, there's no obvious reason to assume that all sequences of all 
configurations of phase space are equally likely;  and though there 
aren't any necessary a priori constraints on what we could interpret as 
phase transitions in this iterated phase space, it's still the case 
that transitions between any such more-likely similar states would be 
themselves more likely.  And the similarities between those states then 
give us clustered, consensus realities in which we have things like c, 
G, and so on.  And thus perhaps there's a higher-order set of rules to 
be discovered.



More information about the FoRK mailing list