On 25 Dec 2008, at 08:05, Abram Demski wrote:
>
> Bruno,
>
> I agree with Gunther about the two types of machine. The broader
> machine is any system that can be logically described-- a system that
> is governed by rules and has a definite description.
Then Church thesis entails it is not broader, unless you mean that the
rules are not effective.
> Such machines are
> of course not necessarily computable; oracle machines and so on can be
> logically described (depending of course on the definition of the word
> "logical", since they cannot be described using 1st-order logic with
> its standard semantics).
UDA still works with very big weakening of comp, which I don't mention
usually for pedagogical purpose. The fact that the first person cannot
be aware of delays, together with the fact that the UD generates the
reals extend the comp consequences to machine with all kind of oracles.
The AUDA is even less demanding, and works for highly non effective
notion of "belief". Instead of using the Gödel provability predicate
we can use non effective notion like "truth in all model of ZF", or
even "truth in all transitive models of ZF". In that last case G and
G* can be effectively extended.
To my knowledge the only scientist being explicitly non mechanist is
Penrose. Even Searle who pretends to be a non mechanist appears to
refer to machine, for the brain, which are Turing emulable. Then
Searles make error in its conception of "mind implementation" and
"simulation" like Hofstadter and Dennett have already very well
criticized. The comp reasoning begins to be in trouble with machines
using discrete set of actual infinities. Analog machine based on
notion of interval are mostly Turing emulable. You have to diagonalize
or use other logical tools in some sophisticate way to build
analytical machine which are non turing emulable. Nothing in physics
or in nature points on the existence of those "mathematical
weirdness", with the notable "collapse of the wave packet" (exploited
by Penrose, but also by many dualists).
>
>
> The narrower type of machine is restricted to be computable.
It is logically narrower. But no weakening of comp based on nature is
known to escape the replicability. Even the non cloning theorem in QM
cannot be used to escape the UDA conclusion. You have to introduce
explicit use of actual infinities. This is very akin to a
"substantialisation" of soul. I respect that move, but I have to
criticize unconvincing motivations for it.
Comp entails the existence of uncomputable observable phenomena. It is
normal to be attracted to the idea that non computability could play a
role in the mind. But this consists to build a machine based on the
many sharable computations going through the turing state of the
machine and this gives "quantum machine", which are turing emulable
although not in real time, but then they play their role in the
Universal Deployment.
Of course you can just say "NO" to the doctor. But by invoking a non
turing emulable "machine", you take the risk of being asked which one.
Up to now, as far as I know, this exists in mathematics, but there are
no evidence it exists in nature, except those using the kind of
indeterminacy which can be explain with the comp hypothesis.
>
>
>>
>> All known physical causal system are Turing emulable.
>>
>
> I am no physicist, but I've been trying to look up stuff on that
> issue... Schmidhuber asserts in multiple places that the fact that
> differential equations are used to describe physics does not
> contradict its computability, but he does not explain.
The SWE is linear. It makes the quantum object directly turing
emulable (mostly by dovetailing if you are using a sequential
processor). The solution are linear combination of complex
exponential. Obviously e, PI and i are computable reals.
It is far more difficult, and perhaps false, to say that Newtonian
Physics is Turing emulable. Newton himself was aware of action at a
distance for its gravitational law. But anything so weird has been
usually considered as an evidence that Newtonian Physics could not be
taken literaly.
To reintroduce such bizarre feature in nature just to contradict the
comp hyp is a bit ironical. It is like Bohmian reformulation of
Quantum Mechanics: to make a theory more complex to avoid
interpretation judged as unpleasant.
This subject is made difficult because there are no standard notion of
computability with the real numbers (despite many attempts to find one).
If someone know better ... Non comp theories have to be rather exotic.
Of course this is not an argument for the truth of comp.
> I know that,
> for example, Wolfram is attempting a computable foundation for
> physics, but I don't know about any real progress... so any info would
> be appreciated.
Wolfram like Schmidhuber believes there could be a computable
universe. The "whole" could be computable. But in that case the UDA
shows that the universe is a mathematical one, and indeed can be
described by any universal dovetailer. But the *physical* universe
emerging from inside will have some uncomputable feature. This comes
from the fact that we are necessarily ignorant about which
computations support us, and that, below our substitution level, we
have to take into account of a a non enumerable set of computations.
(More in the UDA).
An interesting book on the computability with the reals containing
some of the construction mentioned above, is:
POUR-EL M. B., RICHARD J. I., 1989, Computability in Analysis and
Physics, Springer-Verlag, Berlin.
If I remember well you can find in this book a computable function
having a non computable derivative!
See also this link, and related links for the analog/digital problem,
or search on "universal analog machine" or "turing analog machine".
http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6V1G-4M6XMHG-3&_user=10&_rdoc=1&_fmt=&_orig=search&_sort=d&view=c&_acct=C000050221&_version=1&_urlVersion=0&_userid=10&md5=165aa939bc8414dbea9554ee03ea1b9e
Bruno
http://iridia.ulb.ac.be/~marchal/
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "Everything List" group.
To post to this group, send email to everything-list.domain.name.hidden
To unsubscribe from this group, send email to everything-list+unsubscribe.domain.name.hidden
For more options, visit this group at
http://groups.google.com/group/everything-list?hl=en
-~----------~----~----~----~------~----~------~--~---
Received on Thu Dec 25 2008 - 13:23:30 PST