Hal Finney wrote:
>Bruno Marchal writes:
>> Methodologically your ON theory suffers (at first sight)the same
>> problem as Wolfram, or Schmidhuber's approaches. The problem consists in
>> failing to realise the fact that if we are turing-emulable, then
>> the association between mind-dynamics and matter-dynamics cannot be
>> one-one. You can still attach a mind to the appearance of a
>> machine, but you cannot attach a machine to the appearance of a
>> mind, you can only attach an infinity of machines, and histories,
>> to the appearance of a mind.
>
>I think what you are saying is that if a mind can be implemented by more
>than one machine, there is first-person indeterminacy about which
>machine is immplementing it.
Yes.
>However, wouldn't it still be the case that to the extent that the mind
>can look out and see the machine, learn about the machine and its rules,
>that it will still find only a unique answer? There would be a subjective
>"split" similar to the MWI splits. For all possible observations in a
>given experiment to learn the natural laws of the universe/machine that
>was running the mind, the mind will split into subsets that observe each
>possible result.
Yes.
>So it is still possible to make progress on the question of the nature of
>the machine that is the universe, just as you can make progress on any
>other observational question, right?
Almost right. We can make progress on the question of the nature of
the average machine that is the average "universe" (computational history)
which defined our most probable neighborhood.
>Also, isn't it possible that, once enough observations have been made,
>there is essentially only one answer to the question about what this
>machine is like? Just as there will often be only one answer to any
>other factual question?
Only if you observe yourself above your level of substitution. Below
that level, repeated observations should give you trace of the comp
indeterminacy. Like in QM. For example, you will discover that precise
position of some of your particles are undefined. Below the level
of substitution the statistics will be non classical for they must take
into account our inability to distinguish the computational histories.
>Of course, it's always possible that the machine is itself being emulated
>by another machine, since one computer can emulate another. But we could
>still at least say that the observed laws of physics correspond to a
>particular computer program which could be most naturally implemented on a
>particular architecture.
I don't think that that could be the case. It could only be an
approximation.
Below the level of substitution we must find a sort of vagueness
related to our incapacity to distinguish one computation from the many others
which are possible. With comp the laws of physics must emerge from that
average. You are coherent because this follows from the UDA part which
you admittedly have still some problem with.
cf:
http://www.escribe.com/science/theory/m3817.html
A little "TOE-program" is still possible, but then it must be extracted
from that average---in fact it must run the definition of that average,
in the case such a computational definition exists, and that is doubtful.
But even if that was the case, that definition must be derived from
that comp average. That's why I suspect a quantum universal dovetailer
is still a possible candidate of our uni/multiverse.
>We can never be sure that the universe machine
>isn't sitting in someone's basement in a super-universe with totally
>different laws of physics, but we can at least define the laws of physics
>of our own universe, in terms of a computer program or mathematical model.
I don't think so. We belong to an infinity of computational histories
from which the (beliefs of the) laws of physics emerge, from which the
appearance of "a universe" emerges too. "our universe" is a not
well defined expression (provably so with the comp hyp).
Bruno
Received on Tue Nov 26 2002 - 06:29:24 PST