[ExI] Some new angle about AI

scerir scerir at libero.it
Fri Jan 1 03:57:05 UTC 2010

> There is a way to produce algorithmic randomness with a Turing
> machine, requiring something of a trick. The Turing machine runs a
> program which generates a virtual world containing an observer, and
> the observer has a piece of paper with a bitstring written on it. At
> regular intervals the program duplicates the entire virtual world
> including the observer, but in one copy appends 1 to the bitstring and
> in the other appends 0. From the point of view of the observer,
> whether the next bit will be a 1 or a 0 is indeterminate, and the
> bitstring he has so far truly random. This is despite the fact that
> the program is completely deterministic from the perspective of an
> outside observer. It will be obvious that this is a model of quantum
> randomness under the MWI of QM: God does not play dice, but his
> creatures do.
> Stathis Papaioannou

A Turing machine can compute many things. It cannot compute other things, 
like (in general) real numbers (because of their incompressibility). I can 
agree that, like in your example, a perspective from within is different from 
the perspective of an outsider. A God who does not play dice is well possible 
(even the late Dirac had that opinion) but the God who plays the ManyWorlds or 
the Great Programmer who computes all evolutions of all universes, and not  the 
specific evolution of the specific universe, are lazy, IMO.

More information about the extropy-chat mailing list