Terren,
Sorry for sending my answer with some delay. I thought I send it, but
the post was sleeping in my draft box. I appreciate you insist for the
comment.
Best,
Here it is:
On 24 Jun 2011, at 06:37, Terren Suydam wrote:
Hi Bruno, thanks for your comments... see below.
On Tue, Jun 21, 2011 at 11:17 AM, Bruno Marchal <[email protected]>
wrote:
Comp requires only that you can imagine surviving with an
artificial digital
brain. Then a reasoning shows that your consciousness is "more
attached" to
all the possible 'implementations' of that digital brain in the
arithmetical
truth (or just the sigma_1 tiny part, from inside this changes
nothing).
Then, if you allow thought experiences with amnesia, you can
understand that
a non trivial form of consciousness can be attached to the
universal machine
or relatively universal number.
Isn't it reasonable that only certain kinds of 'programs' have a 1st
person consciousness?
I completely agree with this. Löbian programs and universal programs
are certainly very special kind of programs.
That it depends on the details of how the
'program' is constructed? I mean, the UD executes an infinity of
nonsensical algorithms that might correspond metaphorically to "rocks"
and other inanimate phenomena. Again the idea would be that it is a
particular organization (or class of organization) that is realized by
a particular universal number (or class of universal numbers) that
gives rise to the 1st person experience. If this is the case, I'm not
sure you need the (virgin) universal machine to be conscious.
I have few doubt that by assuming mechanism, Löbian programs are
already conscious, and even self-conscious. Indeed they have already
the rich theology with a Plotinian God and a Platonist notion of
matter. Also, they can be shown maximally introspective in some
technical sense. You can think of them as sort of universal baby. I
certainly recognize myself in there. I have just more specific
relative memories.
Now, either by using amnesia based thought experiments, or by using
some drug, I can give sense to consciousness without a notion of self,
that is non-Löbian consciousness, which leads me to the idea that
being universal might be enough for consciousness.
It is not that I need them to be conscious, but that I realize they
might be conscious, even if it is a form of consciousness which is out
of space and time; in fact out of anything third person describable.
It is a platonic form of consciousness. Löbianity just adds a self to
it (A 3-self like in Bp, and a non nameable 1-self, captured by Bp &
p, that is 'Bp and true(p'), but the notion of truth is not
arithmetical and cannot be defined in the language of the machine).
This makes sense with the classical theory of knowledge.
To be sure, I defend the idea that Löbian machine/number are conscious
since more than 40 years. But the idea that a virgin (non Löbian)
universal machine is already conscious has come in an attempt to
explain more precisely the "mystical experience", as described by
mystics (including the rationalist greek and indian mystics) and the
smoker of Salvia divinorum. This is not in my PhD thesis, I cannot ask
the member of the jury to smoke salvia to verify this :). But I begin
to think that enlightenment might be an experience of losing
Löbianity. To be Löbian would already be a sort of delusion. I have no
certainty at all in that matter, to be sure. With salvia we can be led
to a state of consciousness which is no more personal.
To be sure, the salvia experience is even more challenging, I mean for
the computationalist.
Comp predicted that some special thing might happen when we die, or
when we lost our body/brain, but comp predicted that such an
experience was not memorizable, yet, we seem to be able to live it,
when smoking salvia (and so salvia does seem to trigger some mechanism
close to what might happen when dying), but, very paradoxically, we
seem to be able to remember it, or parts of it, and this still does
not make sense with comp, even when considered as an hallucination.
That is why the machine should not be just a virgin universal
machine, but a
Löbian machine.
Both are virtually in all possible environments/computational
histories.
Both are conscious (I think currently), but only the Löbian one has
the
cognitive ability to introspect and to give sense to other
machines/environment.
So, as examples, the Robinson arithmetic theory (basically logic +
laws of
addition and multiplication) is a Turing universal machines, and
thus is
conscious, but not self-conscious. The Peano arithmetic theory,
which is the
same as Robinson + the axioms of inductions (which are very
powerful) is
self-conscious. But, without further programs/instructions, their
first
person indeterminacy bears on all state of consciousness. Our own
consciousness is their consciousness, somehow.
Reasonably, self-consciousness grows a lot and get much more
intricate when
meeting other selves.
To me, the abstraction implied by "without further
programs/instructions" renders the notion of self-consciousness
obsolete. What I can accept is that the Löbian machine represents the
minimum logical framework to *support* self-consciousness as
"embodied" by the relations of *particular* universal numbers...
otherwise we dilute the meaning of the term "self-conscious", which at
a bare minimum requires some kind of distinction between an embodied
self and the 'other' in which it is situated. What would that 'other'
be? How would it interact with it?
The Löbian machine have the notion of 'other' (indeed they can see
themselves as other, they can do the whole UDA thought experiments,
etc. They are not human, but they have a very rich theology, including
a physics like ours. The further program/instruction/experience will
just differentiate them, like in duplication experiments.
I illustrate how we can interview them on fundamental questions, and
their discourse is very sophisticate and quite close to platonist
theology.
A rock is not a person. In fact a rock or any piece of matter is a
pattern
*we* make from a infinite sum of computational histories. That
exists only
as a stable appearances. It might eventually "contains" universal
dovetailing, and thus, trivially, all consciousness of all persons.
But the
rock is none of those person, so it makes no sense to say that a
rock is
conscious. The same for the whole physical universe: it is a
projection that
*we*, or all Löbian machines are making. Thus, comp is quite the
opposite to
panpsychism. Only person, incarnated by relations among natural
numbers (or
combinators, java program, etc.) can be conscious or self conscious.
But couldn't you make the same argument to say that the 'virgin'
universal machine is not conscious, because it is none of those
persons in particular?
I think the virgin universal machine *is* all those persons in
particular.
At that level we are the same person, but in different relative
contexts.
That is why I thought (before trying salvia) that the mystical
experience was a return to our virgin, yet still Löbian, state. But
the experiments, and the reading of reports of other experiment leads
me to think that consciousness is even more primitive than I thought.
In your case, we are left wondering how the
consciousness of the virgin universal machine "interfaces" with
specific universal numbers, and what would explain the
differences in
consciousness among them.
The difference will come from their different experiences
relatively to the
different computational histories which supports them. This will
entail
different memories, personalities, characters, etc.
Sure, but I was talking less about the content of individual
consciousnesses, and more about the quality of such... e.g. what it's
like to be a bat. How would you distinguish between a creature that
(most of us believe) is conscious, like a cat, and a creature most of
us believe is not, like a bacterium? It seems to me that if you have
an answer to that question, you have the makings of a theory of
consciousness that does not depend on the attribution of some "source
consciousness" of the virgin universal machine.
The bacterium might be already conscious, but is not Löbian. Bacteria
are already Turing universal (I can prove that!).
The cat is Löbian. I think.
The difference is that the bacterium has no notion of time, space, and
others. But he has already a notion of good and bad, and it will try
to maximize the good and avoid the bad. The cat has empathy and some
notion of other beings, and in principle a notion of death.
Humans have powerful hands, so they are more gifted in calculus
(digits), languages, tools, and eventually modifying more and more
their environment. They have develop many competence, but I am not
sure they are more *intelligent* than cats and dogs. In a sense,
humans are more stupid because they feel superior. They believe that
God prefers them to other creatures. That is a necessarily open
question, assuming comp. If we want approach "God", we have to be a
bit more modest.
That's why I favor the idea that consciousness arises from certain
kinds of cybernetic (autopoeitic) organization (which is consistent
with comp).
Sure. Given that everything is defined through self-reference, comp
should
have friendly relationship with autopoiesis. Self-reference and
self-organization is crucial for the development of consciousness and
self-consciousness. I talked to Varela and he was aware and
interested by
the work of Judson Webb on mechanism, and very open to comp and
comp's
consequences.
Cool!
In fact I think it is still consistent with much of what
you're saying... but it is your assertion that comp denies strong
AI
that implies you would find fault with that idea.
The only fault is related to the idea that we can build an AI ,
*AND* give
some proof that it is an AI. The same for an artificial brain. You
need to
do some act of faith. Most pausibly, we and nature do instinctively
or
automatically such act of faith, for example in believing in other
people.
The real question is not "can a machine think", the real question
is "are
you OK if your son or daughter decides to marry a machine?".
haha, well said... so far as that goes. But the real issue here is
your original assertion - the one I responded to initially - where you
said "Actually, comp prevents
"artificial intelligence".
But it sounds like what you really meant to say is "Actually, comp
prevents us from proving AI" which is a very different statement.
OK. Sorry if I have been unclear. I said this in the context of Colin
Hales paper.
I think I understand your point here with regard to consciousness -
given that you're saying it's a property of the platonic 'virgin'
universal machine. But if you assert that about intelligence,
aren't
you saying that intelligence isn't computable (i.e. comp denies
strong
ai)?
Comp implies strong AI (but not vice versa: machine can think does
not
entail that only machine can think).
Comp => STRONG AI: If I am a machine, then some machine can think
(assuming
that I can think).
But comp denies that "we can prove that a machine can think". Of
course we
can prove that some machine has this or that competence. But for
intelligence/consciousness, this is not possible. (Unless we are not
machine. Some non-machine can prove that some machine are
intelligent, but
this is purely academical until we find something which is both a
person and
a non-machine).
With you here...
I use "intelligence" is the large sense (it is close to being
conscious). So
it is related to the first person indeterminacy, which is infinite.
You need
this to stabilize consciousness, and attach it to a notion of normal
computational history. You don't need this for one instant of
intelligence,
but you need it for two instants, so to speak.
but lost me here.
This assumes the understanding of the UD argument. Consciousness
(which is a purely first person notion) is not attached to any
particular instantiation of a working program, but on all such
instantiation. Like in Everett QM, we (first person) do have an
infinity of "bodies", which are parts of the infinitely many
computations going through our state, and executed (platonically) by
the UD. This is counterintuitive and even shocking, for most, but is a
consequence of taking the mechanist assumption seriously enough. Don't
hesitate to ask more questions. I will explain the whole UDA on a
salvia forum soon.
That creativity is sourced in subjective indeterminacy?
I don't think so. The universal machine is already creative, but its
creativity needs some histories to bring stable results.
Note that the machine can lose its creativity in some histories,
like bad
education can discourage students. But at the start, both
consciousness and
creativity are "maximal" in some way. The more we are aware of our
universality (like Löbian machines/numbers already are), the more
we can use
our initial creativity (if society and contingencies allow it).
Creativity
might be encouraged, and some heuristics can be taught (like with
de Bono),
but creativity per se is at the heart of universality. I think that
the
Mandelbrot set is creative, and that Emil Post "creative sets" are
too. That
is why he called those set creative, and it has been proved that
creativity
in the sense of Post is just a set theoretical mathematical
characterization
of (Turing) universality or sigma_1 completeness.
To the extent I buy into your mathematical formulations of such heavy
concepts like consciousness, intelligence, and creativity, then that
makes sense to me. But I am left wondering if your logics-based
definitions are the best way to make sense of those concepts, assuming
comp of course.
To be sure, neither truth, nor consciousness, nor any first person
notions have a full third person description. What is amazing, is that
such non-definability can be proved once we assume both comp and some
theory of knowledge (like the classical one where knowledge obeys
axiomatically to the axioms of the S4 modal logic:
K(p -> q) -> (Kp -> Kq) if I know that p implies q, then if I know p,
I will know q.
Kp -> p if I know p, then p is true.
Kp -> KKp if I know p then I will know that I know p.
+ the rules of modus ponens, and necessitation (if I prove p, then I
can deduce Kp).
So we can get formal systems describing, at some meta level, non-
formalizable notions!. This is a subtle point, and it mainly comes
from Tarski theorem of the non-definability of any notion of truth
sufficiently encompassing to allow self-reference.
But I don't want to give you the wrong impression here
either, because I am deeply impressed by your thoughts on this
forum... thanks for taking the time to articulate them and to respond
to folks like myself.
The pleasure is mine. It is always a pleasure to discuss with open
minded person. Well, that is probably the reason why I like to discuss
with the abstract Löbian machine to start with: not only she has none
of the human prejudices, but she lacks also the deepest prejudices
that all animals have develop through millions of years of life
struggle (to eat or to be eaten).
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 [email protected].
To unsubscribe from this group, send email to
[email protected].
For more options, visit this group at
http://groups.google.com/group/everything-list?hl=en.
