On 12 Mar 2011, at 19:17, Brent Meeker wrote:
On 3/12/2011 1:48 AM, Bruno Marchal wrote:
On 11 Mar 2011, at 19:53, Brent Meeker wrote:
On 3/9/2011 9:14 AM, Bruno Marchal wrote:
On 08 Mar 2011, at 20:11, Brent Meeker wrote:
On 3/8/2011 10:41 AM, Bruno Marchal wrote:
We could start with lambda terms, or combinators instead. A
computation (of phi_4(5) is just a sequence
phi_4^0 (5) phi_4^1 (5) phi_4^2 (5) phi_4^3 (5) phi_4^4
(5) phi_4^5 (5) phi_4^6 (5) ...
_4 is the program. 5 is the input. ^i is the ith step (i = 0,
1, 2, ...), and phi refers implicitly to a universal number.
Bruno, I don't think I understand what a universal number is.
Could you point me to an explication.
The expression "universal numbers" is mine, but the idea is
implicit in any textbook on theoretical computer science, or of
recursion theory (like books by Cutland, or Rogers, or Boolos and
Jeffrey, ...).
Fix any universal system, for example numbers+addition
+multiplication, or LISP programs.
You can enumerate the programs:
P_0, P_1, P_2, ...
So that you can enumerate the corresponding phi_i
phi_0, phi_1, phi_2, ...
Take a computable bijection between NXN and N, so that couples of
numbers <x,y> are code by numbers, and you can mechanically
extract x and y from <x,y>
Then u is a universal number if for all x and y you have that
phi_u(<x,y>) = phi_x(y).
In practice x is called program, and y is called the input.
Now, I use, as fixed initial universal system, a Robinson
Arithmetic prover. I will say that a number u is universal if RA
can prove the (purely arithmetical) relation phi_u(<x,y>) =
phi_x(y).
The notion is not entirely intrinsic (so to be universal is not
like to be prime), but this is not important because from the
machine's point of view, all universal numbers have to be taken
into account.
By "not intrinsic" I assume you refer to the fact that the Godel
numbering scheme that idenitfies a number u with a program is not
something intrinsic to arithmetic; it is a mapping we create.
Right?
Partially right. If you start from Robinson arithmetic, each
universal numbers will have many syntactical variants due to the
provable existence of different coding scheme. This does play a
role in the measure problem a priori. But this is true for your
brain too (clearly so with comp).
But then what does it mean to say "all universal numbers have to
taken into account"? If I understand correctly *every* natural
number is a universal number, in some numbering scheme or
another. So what does "taking into account" mean?
Not every natural number will be universal. Once you have choose
your initial system, only a sparse set of numbers will be
universal, and the computations will rely on the relations between
the numbers. You are not supposed to change the initial system. The
laws of mind and the laws of matter does not depend on the choice
of the theory (initial system), but you cannot change the theory,
once working in the theory. To say that all universal numbers have
to be taken into account means that you have to take into account
all the relations which constitute computations in the initial
theory. The observer cannot distinguish them from its first person
point of view. The measure does not depend on the codings, but does
depend on the existence of infinitely many coding schemes.
OK. That's what I thought, but I wanted to be sure.
Let me be completely specific, and instead of using RA, let us use
a universal system of diophantine polynomials:
The phi_i are the partial computable functions. The W_i are their
domain, and constitutes the recursively enumerable sets.
The Turing universal number relation (X is in W_Nu) can be coded as
a system of diophantine polynomials(*). I will say that a number Nu
is universal if W_Nu is the graph of a universal computable function.
The detail of the polynomials is not important. I want just to show
that such numbers are very specific. Then you can represent
computations by their recursively enumerable sequence of steps,
etc. Löbian machine can likewise be represented by their set of
beliefs, themselves represented by RE set W_i, etc. The same is
true for "interaction" between Löbian machine and environments
(computable or computable with oracle), etc.
For any choice of representation, the number relation organizes
itself into a well defined mathematical structure with the same
internal hypostases.
Here is a universal system of diophantine polynomial, taken in a
paper by James Jones
James P. Jones, Universal Diophantine Equation, The Journal of
Symbolic Logic, Vol. 47, No. 3. (Sept. 1982), pp. 549-571.
To define universal numbers here is very short. To do the same with
RA would be much longer and tedious, but the principle is the same.
Once the initial universal system is chosen, there is no more
arbitrariness in the notion of universal numbers, computations,
etc. Except for different internal codings, but this is the price
for comp. I can change your brain in many ways, but I have to keep
the relations between your constituents intact, making your brain a
non relative notion. A brain is not intrinsic, but is not arbitrary
either.
In the universal system of diophantine polynomials below, there is
an infinity of number Nu coding the primes numbers, or the even
numbers, or the universal numbers, etc. If all numbers code precise
things, it is false that precise things can be coded by arbitrary
numbers.
(*) (Numbers are named by single capital letters, or by single
capital letters + one non capital letter. In the first line Nu, Z,
U, Y, are integers.
They are all integers, aren't they? So am I to understand Z, U, and
Y as parameters which can be chosen and each choice will pick out a
set of numbers, such as the primes, the evens,..., Goldbach pairs?
Yes. Even non negative integers in that particular example.
Then X belongs to W_Nu is equivalent with
So the solutions, X, will be a set of integers which, in our chosen
coding scheme, express some recursively enumerable set of numbers?
Yes.
The set of numbers X will constitute such a W_Nu, and is a recursively
enumerable set of all solutions of the system of diophantine
polynomials for some choice of the parameters Z, U, Y. And all W_i can
be retrieved by chosing well Z, U, Y.
A number Nu can be said to be a universal numbers if Nu = ((ZUY)^2 +
U)^2 + Y, + the other polynoms, and W_nu is a universal recursively
enumerable set (a creative set, a m-complete set or a sigma_1 complete
set, those are equivalent notion).
There will be an infinity of them, but few numbers, in proportion,
will be universal. Like there is an infinity of codes for a LISP
interpreter, but few programs are LISP interpreter.
It is hard to believe that a super growing function like y =
x^(x^(x^(x^ ...x) with x occurrence of x (which grows quicker than all
exponential functions!) can be represented by a polynomial. But this
has been proved and is a direct consequence, of course, of the
existence of universal diophantine polynomial. There is also parameter
Z, U, Y coding the current state of our galaxy, or coding the quantum
vacuum ... Those are (sophisticate) example of universal numbers.
Note that the degree of the universal polynomial below is 5^60. We can
reduce the degree to four, but then we will need 58 variables.
Note that if the variables are real numbers, we get a decidable, even
finite, and thus non universal, set of solutions. The universality
comes from the digitality, or from the natural numbers. The natural
numbers behave in an infinitely more complex way than the real
numbers, and reminds us of the non triviality of what is a
computation. That can perhaps be used to mitigate the idea that a rock
can think, because I can imagine that point particles (in quantum non
relativistic physics for example) can simulate the real-number version
of that polynomial, but I am less sure for the diophantine (integers)
version.
Note also that, although the first order theory of rational numbers is
Turing universal, it is an open problem if there is a universal
polynomial on the rational numbers. Another open problem is the
existence of a cubic universal diophantine polynomial. We know that
there is no quadratic universal polynomial, though.
Bruno
Brent
Nu = ((ZUY)^2 + U)^2 + Y
ELG^2 + Al = (B - XY)Q^2
Qu = B^(5^60)
La + Qu^4 = 1 + LaB^5
Th + 2Z = B^5
L = U + TTh
E = Y + MTh
N = Q^16
R = [G + EQ^3 + LQ^5 + (2(E - ZLa)(1 + XB^5 + G)^4 + LaB^5 + +
LaB^5Q^4)Q^4](N^2 -N)
+ [Q^3 -BL + L + ThLaQ^3 + (B^5 - 2)Q^5] (N^2 - 1)
P = 2W(S^2)(R^2)N^2
(P^2)K^2 - K^2 + 1 = Ta^2
4(c - KSN^2)^2 + Et = K^2
K = R + 1 + HP - H
A = (WN^2 + 1)RSN^2
C = 2R + 1 Ph
D = BW + CA -2C + 4AGa -5Ga
D^2 = (A^2 - 1)C^2 + 1
F^2 = (A^2 - 1)(I^2)C^4 + 1
(D + OF)^2 = ((A + F^2(D^2 - A^2))^2 - 1)(2R + 1 + JC)^2 + 1
Z, U, Y codes just the parameter Nu. The other "letters" are the
variables.
We could transform this into one giant universal polynomial, and we
could manage its degree to be equal to 4. But that would take a lot
of place and variables. The solutions of that equations provide a
universal dovetailing, notably because computations can be seen as
RE relations. For further reference, I will call that system JJ
(for James Jones). JJ, like RA, is an acceptable theory of
everything, at the ontological level.
Only the left (terrestrial) hypostases can be named in that system,
the right hypostases, the divine one, are, in the general case
purely epistemological and cannot be named in the system, like the
universal soul itself S4Grz. Machines can point on it only
indirectly, but can study, nevertheless, the complete theory for
simpler Löbian machines than themselves. So JJ provides the
ontology, but the internal epistemology necessarily escapes the
ontological TOE (whatever the initial theory choice). This explain
that you can *see* consciousness there, no more than you can see
consciousness by dissecting a brain.
I recall the 8 hypostases (three of them splits according to the G/
G* splitting)
p
Bp
Bp & p
Bp & Dp
Bp a Dp & p
The "left one" are the one ruled by G, and the right one are the
one ruled by G*. Which corresponds the terrestrial and the divine
respectively in the arithmetical interpretation of Plotinus. Those
with (& p) are not solution of the JJ equations, nor any of them
when extended in first order modal logic. Philosophy of mind, with
comp, is intrinsically very complex mathematics. The contrary would
have been astonishing, to be sure.
To sum up: don't confuse the flexibility of the codings with the
non arbitrariness of the relations we obtain between those numbers
in any initial theory chosen.
Note that JJ is simpler than RA. It has fewer assumption. It does
not need logics, but only equality axioms, and a the existential
quantifier (or the distinction variable/parameter). It is the
equivalent of the other "everthing ontological theory" based on the
combinators:
((K x) y) = x
(((S x) y) z) = ((x z) (y z))
Which is quite much readable and shorter! In that theory you might
understand that there will be an infinity of universal combinators
(which will be very long!), but to be universal will not be an
arbitrary notion. Non-intrinsic does not mean arbitrary.
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
.
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.
