Re: on simply being an SAS

From: Russell Standish <>
Date: Sun, 9 Jan 100 17:11:08 +1100 (EST)

> Russell Standish wrote:
> >> BM: Are you saying that you can survive with artificial neurons, but
> >> not with artificial *digital* neurons?
> >
> >RS: You are probably being a little sloppy with the word digital
> >here. However, if we are talking about my substitution process I
> >mentioned above, then the neurons would most definitely have to be
> >analogue.
> Yes but there are two kinds of analogue devices. Those who are still
> turing emulable and those who are not (at the relevant level).

By Turing emulable, presumably you mean a digital simulation of the
neuron carried out at some sufficiently high level of resolution,
using pseudo random numbers where necessary. However, assuming the
human brain exhibits chaotic behaviour, then no two runs of the human
brain within the same environment will yield identical output, which
is manifestly not the case of the simulation. Whether this is
significant for conciousness or not is part of the current debate.

> It seems you agree with Penrose that there must be uncomputable
> feature in our brain. These features would be not emulable at any
> level.

This accusation has been leveled at me before. I have not read
Penrose, so I don't actually know whether I agree with him. However,
from some of the discussions of his ideas in New Scientist re (eg
quantum effects being amplified by the microtubules), I suspect I part
company from him rather early.

> >> To believe in comp, it is not necessary that neurons are Turing
> >> emulable, only that there is a level such that the relevant working
> >> of the neurons are turing emulable.
> >
> >And I'm saying that randomness is an essential component to the
> >operation of the brain. Randomness is not Turing emulable at any
> >level.
> But Everett just showed us how to emulate randomness with UTMs.
> It is enough to emulate solutions of schroedinger equation applied
> to the observer (emulable itself by comp). This does not create
> randomnes relatively to us, but it creates randomness relatively
> to the simulated observers. You get the same with the computationalist
> self-multiplication kind of indeterminacy.
> You cannot get 3-randomness, but you can easily get 1-randomness, with
> turing machine. My interpretation of Everett is that he makes
> disappearing the need of 3-randomness in QM.

I'm not entirely sure how to interpret this. If what you're saying is
that QM is computable (of course we can solve the Schroedinger
equation and evolve it deterministically), therefore the individual
observers in the multiverse are computed by this simulation, then you
are wrong. The computation does not contain within it the actual
observations by the observers - all it contains is probability
distributions for the attributes the observers see. The actual
observations are simply resolved by chance.

This is easier to discuss using the Schmidhuber Plenitude, which i
ascribe to as a great idea. Unfortunately, the Schmidhuber Plentiude
does not imply COMP (If it does, I'd be interested to see the proof!)

> >Nevertheless, it may be possible for a Turing machine to "pass the
> >Turing test" without genuine randomness. Its not clear to me whether
> >this might lead to "zombies" as you call it.
> I'm afraid it will.
> Here your "non-computationalism" is a little worst than Penrose's one :-(
> At least Penrose believes that not only consciousness need uncomputability
> but even conscious behavior need it. So there is no Zombie in Penrose.
> If you are correct, it would mean that from a 3-person point of view I can
> survive teletransportation, but that from "your" point of view my digital
> copy is a zombie, and then you can kill me!

If your "digital" copy passes the Turing test, I will not know whether
it is a zombie or not.

> By "not genuine randomness" do you mean the randomnes produced by
> pseudo-random generator, or by the randomness due to self-multiplication
> (i.e. the one of the MWI or comp) ?
> Bruno
> PS There is a beautiful paper in theoretical computer science which
> shows indeed that some problem are soluble with UTM + Random Oracle, and
> not soluble with any UTM + any Pseudo Random Oracle. (Kurtz, S. A., On
> the Random Oracle Hypothesis, Information and control, 57:40-47).
> But that does not change the fact that self-multiplication and the
> 1/3-distinction gives, in the comp realm, a phenomenology of "genuine"
> randomnes.

Dr. Russell Standish Director
High Performance Computing Support Unit,
University of NSW Phone 9385 6967
Sydney 2052 Fax 9385 6965
Room 2075, Red Centre
Received on Sat Jan 08 2000 - 22:07:35 PST

This archive was generated by hypermail 2.3.0 : Fri Feb 16 2018 - 13:20:06 PST