Re: COMP refutation paper - finally out

[Bruno Marchal]

On 14 Jun 2011, at 21:19, Terren Suydam wrote:
Thanks for the reply Bruno, comments below...

On Tue, Jun 14, 2011 at 9:53 AM, Bruno Marchal <marc...@ulb.ac.be>  
wrote:
doesn't that imply the possibility
of an artificial intelligence?

In a weak sense of Artificial Intelligence, yes. In a strong sense,  
no.

If you are duplicated at the right substitution level, few would  
say that
"you" have become an "artificial intelligence". It would be a case  
of the
good old natural intelligence, but with new clothes.

Sure, but the distinction between artificial and natural intelligence
is not that important assuming comp.

I agree with you. The difference between artificial and natural is ...  
artificial (and thus it is natural indexical done by any entity having  
some big "ego").




The point is simply that if I can
be simulated (which I agree requires some faith), that implies that
intelligence does not require biology (or any other particular
"physical" substrate), that strong artificial intelligence is possible
in principle, ignoring for the moment the question of whether we can
provably construct it.

I agree. My point was similar to the recent post of Russell Standish,  
that somehow Colin get a result which is a consequence of comp, and  
can't be use against comp. Only against a misunderstood view of comp.




In fact, if we are machine, we cannot know which machine we are,  
and that is
why you need some luck when saying "yes" to a doctor who will build  
a copy
of you/your-body, at some level of description of your body.

This is an old result. Already in 1922,  Emil Post, who discovered  
"Church
thesis" ten years before Church and Turing (and others) realized  
that the
"Gödelian argument" against Mechanism (that Post discovered and  
refuted 30
years before Lucas, and 60 years before Penrose), when corrected,  
shows only
that a machine cannot build a machine with equivalent qualification  
to its
own qualification (for example with equivalent provability power in
arithmetic)  *in a provable way*. I have refered to this, in this  
list,
under the name of "Benacerraf principle", who rediscovered this  
later.

We just cannot do artificial intelligence in a provable manner. We  
need
chance, or luck. Even if we get some intelligent machine, we will not
know-it-for sure (perhaps just believe it correctly).

Doesn't this objection only apply to attempts to construct an AI with
human-equivalent intelligence?  As a counter example I'm thinking here
of Ben Goertzel's OpenCog, an attempt at artificial general
intelligence (AGI), whose design is informed by a theory of
intelligence that does not attempt to mirror or model human
intelligence. In light of the "Benacerraf principle", isn't it
possible in principle to provably construct AIs so long as we're not
trying to emulate or model human intelligence?

I think that comp might imply that simple virgin (non programmed)  
universal (and immaterial) machine are already conscious. Perhaps even  
maximally conscious. Then adding induction gives them Löbianity, and  
this makes them self-conscious (which might already be a delusion of  
some sort). Unfortunately the hard task is to interface such (self)- 
consciousness with our probable realities (computational histories).  
This is what we can hardly be sure about.
I still don't know if the brain is just a filter of consciousness, in  
which case losing neurons might enhance consciousness (and some data  
in neurophysiology might confirm this). I think Goertzel is more  
creating a competent machine than an intelligent one, from what I have  
read about it. I oppose intelligence/consciousness and competence/ 
ingenuity. The first is needed to develop the later, but the later has  
a negative feedback on the first.

Bruno


On Thu, Jun 9, 2011 at 4:53 PM, Bruno Marchal <marc...@ulb.ac.be>  
wrote:

Hi Colin,

On 07 Jun 2011, at 09:42, Colin Hales wrote:

Hi,

Hales, C. G. 'On the Status of Computationalism as a Law of  
Nature',
International Journal of Machine Consciousness vol. 3, no. 1,  
2011.
1-35.

http://dx.doi.org/10.1142/S1793843011000613


The paper has finally been published. Phew what an epic!


Congratulation Colin.

Like others,  I don't succeed in getting it, neither at home nor  
at the
university.

From the abstract I am afraid you might not have taken into  
account our
(many) conversations. Most of what you say about the  
impossibility of
building an artificial scientist is provably correct in the  
(weak) comp
theory.  It is unfortunate that you derive this from comp 
+materialism,
which
is inconsistent. Actually, comp prevents "artificial  
intelligence". This
does not prevent the existence, and even the apparition, of  
intelligent
machines. But this might happen *despite* humans, instead of  
'thanks to
the
humans'. This is related with the fact that we cannot know which  
machine
we
are ourselves. Yet, we can make copy at some level (in which case  
we
don't
know what we are really creating or recreating, and then, also,
descendent
of bugs in regular programs can evolve. Or we can get them
serendipitously.
 It is also relate to the fact that we don't *want* intelligent  
machine,
which is really a computer who will choose its user, if ... he  
want one.
We
prefer them to be slaves. It will take time before we recognize  
them
(apparently).
Of course the 'naturalist comp' theory is inconsistent. Not sure  
you take
that into account too.

Artificial intelligence will always be more mike fishing or  
exploring
spaces, and we might *discover* strange creatures. Arithmetical  
truth is
a
universal zoo. Well, no, it is really a jungle. We don't know  
what is in
there. We can only scratch a tiny bit of it.

Now, let us distinguish two things, which are very different:

1) intelligence-consciousness-free-will-emotion

and

2) cleverness-competence-ingenuity-gifted-learning-ability

"1)" is necessary for the developpment of "2)", but "2)" has a  
negative
feedback on "1)".

I have already given on this list what I call the smallest theory  
of
intelligence.

By definition a machine is intelligent if it is not stupid. And a  
machine
can be stupid for two reason:
she believes that she is intelligent, or
she believes that she is stupid.

Of course, this is arithmetized immediately in a weakening of G,  
the
theory
C having as axioms the modal normal axioms and rules + Dp ->  
~BDp. So Dt
(arithmetical consistency) can play the role of intelligence, and  
Bf
(inconsistance) plays the role of stupidity. G* and G proves BDt - 
> Bf
and
G* proves BBf -> Bf (but not G!).

This illustrates that "1)" above might come from Löbianity, and  
"2)"
above
(the scientist) is governed by theoretical artificial  
intelligence (Case
and
Smith, Oherson, Stob, Weinstein). Here the results are not just
NON-constructive, but are *necessarily* so. Cleverness is just  
something
that we cannot program. But we can prove, non constructively, the
existence
of powerful learning machine. We just cannot recognize them, or  
build
them.
It is like with the algorithmically random strings, we cannot  
generate
them
by a short algorithm, but we can generate all of them by a very  
short
algorithm.

So, concerning intelligence/consciousness (as opposed to  
cleverness), I
think we have passed the "singularity". Nothing is more
intelligent/conscious than a virgin universal machine. By  
programming it,
we
can only make his "soul" fell, and, in the worst case, we might get
something as stupid as human, capable of feeling itself superior,  
for
example.

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@googlegroups.com 
.
To unsubscribe from this group, send email to
everything-list+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/everything-list?hl=en.



--
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@googlegroups.com 
.
To unsubscribe from this group, send email to
everything-list+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/everything-list?hl=en.


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- 
l...@googlegroups.com.
To unsubscribe from this group, send email to
everything-list+unsubscr...@googlegroups.com.
For more options, visit this group at
http://groups.google.com/group/everything-list?hl=en.



--
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@googlegroups.com.
To unsubscribe from this group, send email to everything-list+unsubscr...@googlegroups.com 
.
For more options, visit this group at http://groups.google.com/group/everything-list?hl=en 
.


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@googlegroups.com.
To unsubscribe from this group, send email to 
everything-list+unsubscr...@googlegroups.com.
For more options, visit this group at 
http://groups.google.com/group/everything-list?hl=en.