Le 15-juin-05, à 01:39, Russell Standish a écrit :
On Tue, Jun 14, 2005 at 04:39:57PM +0200, Bruno Marchal wrote:
OK but it can be misleading (especially in advanced stuff!). neither a
program, nor a machine nor a body nor a brain can think. A person can
think, and manifest eself (I follow Patrick for the pronouns) through
program, or a machine or a brain, ....
Actually, I think I was the one introducing these 3rd person neutral
pronouns (e, er & em). I picked up the habit from Michael Spivak
(well known mathematician).
Doesn't this beg the question a bit as to what a "person" really is?
In loose everyday conversation, a person is a member of the species
homo sapiens. However, surely we don't want to rule out the
possibility of other conscious things before we even start. And also
as you mention below, there are odd corner cases - the sleeping human
I just identify the first person with the knower. Think about someone
being "cutted" in Brussels and being pasted in thwo cities: A and B,
and nowhere else. Each copy makes an experience, one in A, the other in
B. Each of them know where they have been reconsituted and so each of
them get one bit of information. But this bit is uncommunicable from a
third person point of view. An outsider would get 0 bit from a phone
call by each copy (by default I assume the cut/past device is 100%
I identify the third person with the body or with any third person
description of the body, it could be program (with comp). Despite
Jonathan (I know you agrees with me) I consider as fundamental to
distinguish the 1-person knower from the 3-person body/brain/program.
So when I say that only a person can think, I am really meaning a
What is cute with comp, is that the theatetus definition of knowledge
(and most of its variants) leads to a well defined distinction between
1 and 3 person. What is nice also, is that the knower is not, in any
way, 3-describable (we get freely a
Brouwerian-Heraclitean-Bergsonian-Poincarean ... theory of
conscious-time-duration... at the place where we would the least expect
For the modalist I recall this consists in defining knowing p by
proving p and p is true: Cp = Bp & p. The non equivalence follows from
<snip>: <I snip when I agree, or when I believe the disagreement would
push us outside the main topic>
Church-Turing thesis and arithmetical platonism (my all
description strings condition fulfills a similar role to arithmetical
platonism) are enough.
I am not so sure. You are not always clear if the strings describe the
equivalent of a program (be it an universal program or not), or
describes a computations (be it finite or infinite).
Both actually. One can feed a description into the input tape of a
UTM, hence it becomes a program. They may also be generated by a
program running on a machine.
I was not making that distinction. I was distinguishing between a
program (being a product of another program or not) and the
computation, that is the running of the program. The computation can be
described by the description of the trace of the program (like when we
debug a program). For example the basic program "10 goto 10" has an
infinite trace, like 10 10 10 10 10 10 10 ...
That distinction is primordial for the understanding of the work of the
Universal Dovetailer which dovetails on all programs. The UD generates
all programs and dovetail on all their executions. The possibility or
consistency of this is a consequence of Church's thesis.
eventually is related to bunch of (sheaves of) infinite computations.
they can be coded by infinite strings, but they are not programs.
Is this because they are ultimately not computable (due to the
I don't think so. It is just because for any computational states there
are an infinity of computations going through that states, and this is
a logical "cause" of the 1-person indeterminacy (given that the
1-person are not aware of the huge "delays" (number of steps of the
execution of the UD). This is one line summary of the UDA (see my URL
for links to longer explanation).
There are various strengthenings of the CT thesis which are far from
obvious, and even false in some cases. One of my criticisms of your
work is that I'm not sure you aren't using one of the strong CT
theses, but we can come back to that.
I am using the original thesis by Church, Post, Markov, Turing, ...
They are equivalent and can be summarizes "anachronically" by all
universal digital machine computes the same functions from N to N.
having to fix the UTM. Perhaps this is the route into the anthropic
? Church's thesis just say things does not depend on which UTM you
All programs need to be interpreted with respect to a particular
machine. The machine can be changed by appeal to universal
computation, but then the program needs to be translated as well. But
then, I'm sure you know all this.
Yes, sure. Remember that I postulate arithmetical truth (well a tiny
subset of it). The whole work of the UD is entirely embedded in the
arithmetical truth. The movie graph argument shows we cannot (with
comp) distinguish "real world" with virtual world with (modal)
arithmetical world (where a world in that last setting can be described
by maximal consistent extension).
Finally, there is the possibility that a concrete observer (the
noumenon) exists somewhere, and that "conscious descriptions" are
merely the anthropic "shadow" of the observer being observed by
Again this is to fuzzy for me. I can agree and I can disagree.
With COMP, I'm sure you disagree. Chapter 4 of your thesis directly
argues against this possibility. I don't really agree with it either,
but cannot rule it out once COMP is relaxed.
I am glad you see this! Actually the word "concrete" is provably an
"indexical" with comp. It has only a relative meaning like the word "I"
(in third person!), or "modern" of "now", etc. (of course the 1-person
"I" is absolute, so much that the 1-person "I" is just unameable by any
3-person describable "I"). Well here we must go to modal logic to make
(Note: it is here that I depart from Derek Parfit theory of personal
identity where he argues that "we" are token, where I pretend that with
comp we are necessarily type.)
just saying that we say with "a machine can think " it is an abuse of
language for "the person associate to that machine is thinking".
Are you saying that it is an abuse of the language to say that my
observer maps O(x) can think? In which case I'd agree with you (and I
have never made that particular language abuse that I can recall).
However, I consider human beings to be machines of a particular kind,
and I do consider human beings to think. Of course "machine" in this
case has only a rather loose connection to the machine of theoretical
computer science (the abstract Turing machine).
Yes, I see what you mean (I think), but the problem with that kind of
talk is that it looks like Searle or Jonathan (recently) way of
talking. By being clear about the fact that it is really the 1-person
who thinks and never his brain or body or any 3-description related to
that 1-person (always by assumption), you can prevent the confusion
between the 1-person and the 3-person. Sorry if that remarks just seems
to be about pedagogy, but given that the "measure" problem needs to be
clear on that distinction I think it is important to be as clear as
possible about it.
Instead of saying that comp entails that machine can think, it is less
misleading to say that comp entails machine can vehiculate a knower,
who is the one which can think. The day "we" can be vehiculated through
the net in digital form, people will learn that difference by