- Contemporary messages sorted: [ by date ] [ by thread ] [ by subject ] [ by author ] [ by messages with attachments ]

From: Marchal <marchal.domain.name.hidden>

Date: Thu Jul 8 10:59:59 1999

I will comment here Chris Maloney, James Higgo and Hans Moravec.

Chris Malloney:

*>[...] yet, about the
*

*>terminology used on this group, in particular, relative
*

*>strong-Self-Sampling Assumption (SSSA)" and "Strong SSSA" (did you
*

*>really mean "strong strong self-sampling assumption, Bruno?).
*

Have I write that ? In any case, I mean only absolute SSSA versus

relative Strong SSA.

SSA = Nick Bostrom's Self Sampling Assumption.

Strong = the domain is the set of observer-instant or kind of

instantaneous state of consciousness (but not on people)

Relative = we only try to define a conditional measure, like

Everett relative states (in fact).

Chris Malloney:

*>[...] we need to establish some kind of
*

*>measure over the set of possible universes, and thus, in theory, over
*

*>the set of all existing observer-moments (I just made the assumption
*

*>that possible == exists, which is a tenent of this group). This
*

*>brings us to the question which was posed in the very first post to
*

*>this group, by Wei Dai: how do we establish this measure?
*

This is indeed the fundamental question. With the computationalist

hypothesis, it is almost the only question. Those who will answer that

question with comp will:

1) solve the flying rabbit problem,

2) complete the reversal (anthropic or better Turing-tropic) between

physical science and theoretical computer science,

3) solve the problem of the origin of the *consciousness of the apparent

physical laws*.

In chapter 1-4 of http://iridia.ulb.ac.be/~marchal, I argue

that comp implies the necessity of solving the reversal problem. In

chapter 5, just to illustrate precisely my point, I make an attempt to

solve the

reversal (alias the rabbit, alias the measure) problem by making

a sort of infinite interview of a self-referentially correct universal

machine.

Chris Malloney:

*>So the "flying rabbit" problem is just a restatement of the "why are
*

*>there physical laws" question, which has also been debated on this
*

*>list myriads of times.
*

Absolutely. It is also a restatement of the *measure problem* as I say

above.

I guess we discussed it a myriad of times because it is so important.

The very formulation of the problem depends on the choice of the type

of SSA, strong or weak (Hal's and Wei Dai's terms), relative or absolute

(the immortality camp favours the "relative", isn'it?).

And so one.

Chris Malloney:

*>I do agree with Jacques M. Mallah that this
*

*>group desperately needs a FAQ, especially to tie different approaches
*

*>to the same questions together.
*

Why not create a SUMMARY thread in which all people could try, to sum up

shortly, from time to time, their evolving mutual understandings and

misunderstandings.

For exemple the group can be easily divided in:

BEGIN

the computationnalist / the non comp. / not yet decided or indifferent;

the immortality positive/the mortality positive/ not yet decided or

indif.;

accept the use of the SSA/doesn't accept SSA/ not yet ...;

Strong SSA/ Weak SSA/ not yet ...;

Relative SSA/Absolute SSA/ not yet ...;

The cells obeys the laws of chemistry/chemistry obeys the laws of cells/

not yet ...;

...;

END

But ...

I don't know. We can think about it. That could perhaps help to agree on

our disagreements. To accelerate a convergence toward a possible

Everything FAQ, indeed.

James Higgo wrote:

(concerning Alastair Malcolm's need for a

primitive universe, or an infinite chain of computer, in case of comp).

*> JH > Why is another universe required?
*

*>
*

*> AM> If the generated universe is somehow specified by the output bit
*

*>string,
*

*> then it cannot be the same as the program and TM (combined if you
*

*>prefer),
*

*> which in turn must operate in some dimension analogous to what we
*

*>think of
*

*> as time (for a step-by-step computation to take place). It seems
*

*>reasonable
*

*> to use the term 'universe' for this realm (as Schmidhuber does), but
*

*>in any
*

*>case it requires an explanation just as ours does.
*

*>
*

*>JH>That's like saying the Mandelbrot set does not exist unless someone runs
*

*>the program on a computer outside that set to generate a graphic
*

*>representation of oit on a computer screen.
*

I agree with James Higgo here. That is why Tegmark can have a

O-information theory explaining how self-aware submathematical structures

will define-discover their (relative) physical realities.

There is no need for a primitive reality to sustained a concrete dreaming

computer because "primitive realities" could be defined by the possible

computationnal histories.

Actually I believe, like Penrose (on this point!), that it is quite

plausible that the very Mandelbrot Set codes all computational histories.

It would be a sort of projection of UD* (the complete execution of the

universal dovetailer, the complete work of the great programmer, ...).

This can help for the measure problem, but I have not taken this path.

Now, as I said before, I don't think Tegmark will succeed by taking

the whole Mathematics, which is a monster which has swallowed all

mathematicians who have try to give it a (mathematical) name.

The miracle, for the computationnalist, is that Church's thesis, permit

to give a name to the effective equivalent of the whole Mathematics,

namely the Universal Turing Machine (or his extended execution).

This makes effective (see also Wei Dai at the begining of the discussion)

the derivation of the probabilities.

The fact that this concept does not lead to a contradiction is ultimately

linked to the closure of the set of partial recursive functions (with or

without oracle) with respect to the more transcending mathematical

operation ever invented i.e. diagonalisation.

Hans Moravec wrote:

*>Attribution of subjective experience is a festering issue for those
*

*>of us who presume to assemble messes of inanimate components into
*

*>robots,
*

*>and then wish to claim the results have a subjective life on the basis
*

*>of internal models that allow the robots to discuss their experiences,
*

*>goals, likes and dislikes just like conventionally grown persons.
*

*>
*

*>On the one hand, an engineer who completely understands a robot has no
*

*>need to attribute subjectivity to it, since she can fully explain, and
*

*>even predict, everything the robot does from a mechanistic model
*

*>of the designed interaction of its parts.
*

It is possible that genuine intelligent behavior will appear when

engineers

will build complex machines whith lot of independant layers of redondant

and self-repairing parts. By allowing indefinite iteration of

self-modifications by the machines, during a possible long time,

it will be hard for an engineers to say that she understand completely

what happen in the machine. Like surprising patterns appearing from

simple rules (like in cellular automata), surprising behavioral patterns

could emerge.

This is confirmed by the results in the field "Computational Learning

Theory" where all theorems are necessarily highly non constructive.

We could still understand the main psychology of the machine. We can link

consciousness to the automatic ability to infer models (including

unnameable self-models) for exemple. But if the machine get our (or more

than our) intelligence, we will no more really understand it, will we ?

I guess this follows by what you are saying. This makes *third person

attribution of consciousness* relative, it does not make necessarily

*first person consciousness* relative (although the *content* of a first

person consciousness is relative!).

BTW Bravo for Mind Children :-)

Bruno

Bruno MARCHAL Phone : +32 (0)2 6502711

Universite Libre Fax : +32 (0)2 6502715

de Bruxelles Prive : +32 (0)2 3439666

Avenue F.D. Roosevelt, 50 IRIDIA, CP 194/6

B-1050 BRUSSELS Email : marchal.domain.name.hidden

Belgium URL : http://iridia.ulb.ac.be/~marchal

Received on Thu Jul 08 1999 - 10:59:59 PDT

Date: Thu Jul 8 10:59:59 1999

I will comment here Chris Maloney, James Higgo and Hans Moravec.

Chris Malloney:

Have I write that ? In any case, I mean only absolute SSSA versus

relative Strong SSA.

SSA = Nick Bostrom's Self Sampling Assumption.

Strong = the domain is the set of observer-instant or kind of

instantaneous state of consciousness (but not on people)

Relative = we only try to define a conditional measure, like

Everett relative states (in fact).

Chris Malloney:

This is indeed the fundamental question. With the computationalist

hypothesis, it is almost the only question. Those who will answer that

question with comp will:

1) solve the flying rabbit problem,

2) complete the reversal (anthropic or better Turing-tropic) between

physical science and theoretical computer science,

3) solve the problem of the origin of the *consciousness of the apparent

physical laws*.

In chapter 1-4 of http://iridia.ulb.ac.be/~marchal, I argue

that comp implies the necessity of solving the reversal problem. In

chapter 5, just to illustrate precisely my point, I make an attempt to

solve the

reversal (alias the rabbit, alias the measure) problem by making

a sort of infinite interview of a self-referentially correct universal

machine.

Chris Malloney:

Absolutely. It is also a restatement of the *measure problem* as I say

above.

I guess we discussed it a myriad of times because it is so important.

The very formulation of the problem depends on the choice of the type

of SSA, strong or weak (Hal's and Wei Dai's terms), relative or absolute

(the immortality camp favours the "relative", isn'it?).

And so one.

Chris Malloney:

Why not create a SUMMARY thread in which all people could try, to sum up

shortly, from time to time, their evolving mutual understandings and

misunderstandings.

For exemple the group can be easily divided in:

BEGIN

the computationnalist / the non comp. / not yet decided or indifferent;

the immortality positive/the mortality positive/ not yet decided or

indif.;

accept the use of the SSA/doesn't accept SSA/ not yet ...;

Strong SSA/ Weak SSA/ not yet ...;

Relative SSA/Absolute SSA/ not yet ...;

The cells obeys the laws of chemistry/chemistry obeys the laws of cells/

not yet ...;

...;

END

But ...

I don't know. We can think about it. That could perhaps help to agree on

our disagreements. To accelerate a convergence toward a possible

Everything FAQ, indeed.

James Higgo wrote:

(concerning Alastair Malcolm's need for a

primitive universe, or an infinite chain of computer, in case of comp).

I agree with James Higgo here. That is why Tegmark can have a

O-information theory explaining how self-aware submathematical structures

will define-discover their (relative) physical realities.

There is no need for a primitive reality to sustained a concrete dreaming

computer because "primitive realities" could be defined by the possible

computationnal histories.

Actually I believe, like Penrose (on this point!), that it is quite

plausible that the very Mandelbrot Set codes all computational histories.

It would be a sort of projection of UD* (the complete execution of the

universal dovetailer, the complete work of the great programmer, ...).

This can help for the measure problem, but I have not taken this path.

Now, as I said before, I don't think Tegmark will succeed by taking

the whole Mathematics, which is a monster which has swallowed all

mathematicians who have try to give it a (mathematical) name.

The miracle, for the computationnalist, is that Church's thesis, permit

to give a name to the effective equivalent of the whole Mathematics,

namely the Universal Turing Machine (or his extended execution).

This makes effective (see also Wei Dai at the begining of the discussion)

the derivation of the probabilities.

The fact that this concept does not lead to a contradiction is ultimately

linked to the closure of the set of partial recursive functions (with or

without oracle) with respect to the more transcending mathematical

operation ever invented i.e. diagonalisation.

Hans Moravec wrote:

It is possible that genuine intelligent behavior will appear when

engineers

will build complex machines whith lot of independant layers of redondant

and self-repairing parts. By allowing indefinite iteration of

self-modifications by the machines, during a possible long time,

it will be hard for an engineers to say that she understand completely

what happen in the machine. Like surprising patterns appearing from

simple rules (like in cellular automata), surprising behavioral patterns

could emerge.

This is confirmed by the results in the field "Computational Learning

Theory" where all theorems are necessarily highly non constructive.

We could still understand the main psychology of the machine. We can link

consciousness to the automatic ability to infer models (including

unnameable self-models) for exemple. But if the machine get our (or more

than our) intelligence, we will no more really understand it, will we ?

I guess this follows by what you are saying. This makes *third person

attribution of consciousness* relative, it does not make necessarily

*first person consciousness* relative (although the *content* of a first

person consciousness is relative!).

BTW Bravo for Mind Children :-)

Bruno

Bruno MARCHAL Phone : +32 (0)2 6502711

Universite Libre Fax : +32 (0)2 6502715

de Bruxelles Prive : +32 (0)2 3439666

Avenue F.D. Roosevelt, 50 IRIDIA, CP 194/6

B-1050 BRUSSELS Email : marchal.domain.name.hidden

Belgium URL : http://iridia.ulb.ac.be/~marchal

Received on Thu Jul 08 1999 - 10:59:59 PDT

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