comp and Maxwell demon

[Evgenii Rudnyi]

I am trying to understand what the hypothesis comp is about. It would be
interesting to understand whether the Maxwell demon could be simulated
or not. Any idea?

Evgenii

[Bruno Marchal]

You can simulate it. But you cannot program it. It uses a huge amount
of information that it is hard to extract from the environment. I
would say.
In a larger sense, in which we are ourself Maxwell demon, you can
filter it from long computation. They will transform selective
information into energy. they will economize.
The Universal Dovetailer does generate all possible Maxwell daemons,
but I guess they are typically distributed sparsely in the set of all
computations.
Quantum Maxwell demon? That might explains the origin of life, but
that might be a too cheap explanation. I should revised quantum
information theory, but I have already hardly the time to do the real
thing, which is to extract the quantum information from classical
information + self-duplication and other self-perturbation (or to
convince others to do the work, or that such a work has to be done to
get both the qualia and the quanta).
Interesting but hard question.

Bruno

> --
> You received this message because you are subscribed to the Google
> Groups "Everything List" group.
> To post to this group, send email to everyth...@googlegroups.com.
> To unsubscribe from this group, send email to everything-li...@googlegroups.com
> .
> For more options, visit this group at http://groups.google.com/group/everything-list?hl=en
> .
>

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

[Evgenii Rudnyi]

Thanks for the answer. My question was due to the fact that comp
presumably can simulate more than we observe. Let me check if I have
understood your answer correctly. I will cite a paragraph from my text

http://blog.rudnyi.ru/2010/12/entropy-and-artificial-life.html

We bring a glass of hot water in the room and leave it there. Eventually
the temperature of the water will be equal to the ambient temperature.
In classical thermodynamics, this process is considered as irreversible,
that is, the Second Law forbids that the temperature in the glass will
be hot again spontaneously. It is in complete agreement with our
experience, so one would expect the same from statistical mechanics.
However there the entropy has some statistical meaning and there is a
nonzero chance that the water will be hot again. Moreover, there is a
theorem (Poincar� recurrence) that states that if we wait long enough
then the temperature of the glass must be hot again. No doubt, the
chances are very small and time to wait is very long, in a way this is
negligible. Some people are happy with such statistical explanation,
some not.

Have you meant that the Universal Dovetailer will act for such a
situation according to Poincar� recurrence?

on 16.01.2011 17:50 Bruno Marchal said the following:

[Stephen Paul King]

Dear Bruno and Evgenii,

 

    Pardon my interjection. How exactly does it make sense to say that we "transform selective information into energy"? If information is just a form of energy, such that that phasing can make sense, does it not follow that there should be a unique relationship between the forms of matter that can implement a given information structure and a given material structure? From what I can tell there exists multiple ways that one and the same information can be implemented in matter. It is also true that multiple different information structures can represent one and the same material system. This implies two distinct and irreducible categories: a category of information structures and a category of material structures. There is also a relationship between these two categories, but that relation is not an identity mapping in the usual sense because it is almost always not a bijection. Because of this line of reasoning I am having a hard time understanding how it is that we can make statements like "transforming information into energy" and think that those kind of statements are sound.

    OTOH, it make sense to me that we can make claims like: "we can use information about material systems to induce changes within them that optimize their energy distributions" and so forth. We can maintain our belief in comp and UD as they apply to information and its structures, such as Boolean and Heyting algebras, but still consider that there has to be some form of material implementation possible at least in principle for it, otherwise we are stuck with the conclusion that the physical world is just a causally ineffective epiphenomena and with no explanation as to why we necessarily experience it. All forms of monism have this problem.

 

    If the distribution of Maxwell daemons is spare in its distribution over the set of possible computations, does this not imply that the chances of finding oneself as a Maxwell Daemon to be identically sparse. If the set of computations is infinite this make the chance of finding oneself as a MD to be vanishingly small, 1/inf. Does this not seem to indicate against the necessity of ever experiencing oneself as a MD in the first place? If we go further and notice that we seem to experience a relatively long chain of experiences that are MD in nature, does this not make it even more miraculous given derivation of the statistics of distributions of MD in the ensemble of possible computations? What if there is a mechanism that allows for the biasing of those statistics such that it is no longer a miracle that we are not just spontaneous Boltzmann Brains. The observer selection ideas of Bostrum seem to give us some good clues about this mechanism. Could it be that a feedback type of relationship might be involved? The choice of what form the material structure takes as it evolves in time could be used to select the information representation that best describes it, but this requires a categorical difference between matter and information such that we cannot think of them as just being transformable into each other.

 

 

Onward!

 

Stephen

[Bruno Marchal]

On 16 Jan 2011, at 22:27, Evgenii Rudnyi wrote:

> Thanks for the answer. My question was due to the fact that comp
> presumably can simulate more than we observe. Let me check if I have
> understood your answer correctly. I will cite a paragraph from my text
>
> http://blog.rudnyi.ru/2010/12/entropy-and-artificial-life.html
>
> We bring a glass of hot water in the room and leave it there.
> Eventually the temperature of the water will be equal to the ambient
> temperature. In classical thermodynamics, this process is considered
> as irreversible, that is, the Second Law forbids that the
> temperature in the glass will be hot again spontaneously. It is in
> complete agreement with our experience, so one would expect the same
> from statistical mechanics. However there the entropy has some
> statistical meaning and there is a nonzero chance that the water

> will be hot again. Moreover, there is a theorem (Poincaré

> recurrence) that states that if we wait long enough then the
> temperature of the glass must be hot again. No doubt, the chances
> are very small and time to wait is very long, in a way this is
> negligible. Some people are happy with such statistical explanation,
> some not.
>
> Have you meant that the Universal Dovetailer will act for such a

> situation according to Poincaré recurrence?

The UD will do that an infinity of times, given that the Poincaré
recurrence is a computable process. But the physical laws are sum on
first person views, based on a continuum of histories, so to relate
thermodynamic to the UD is certainly not obvious at all. The UD is
just a way to provide the minimal third person ontology (the
'everything') needed when we assume mechanism, and its role is to
build a mathematical formulation of the mind-body problem (if only to
illustrate that science has not yet choose between Plato and Aristotle).

Bruno

[Bruno Marchal]

I agree. But it is a usual Maxwell demon phrasing. It can be used to explain why deleting a bit of information needs some minial amount amount of energy, and why computing does not need any energy at all. The quantum vacuum gives an example of universal dovetailing which does not use any energy.

    OTOH, it make sense to me that we can make claims like: "we can use information about material systems to induce changes within them that optimize their energy distributions" and so forth. We can maintain our belief in comp and UD as they apply to information and its structures, such as Boolean and Heyting algebras, but still consider that there has to be some form of material implementation possible at least in principle for it,

Not at all. That would contradict the movie graph argument. Look at MGA in the archive.

otherwise we are stuck with the conclusion that the physical world is just a causally ineffective epiphenomena and with no explanation as to why we necessarily experience it. All forms of monism have this problem.

I disagree. A *primitive* physical world just does not exist, and the illusion of why universal numbers believes in a physical worlds is explained by the logic of self-reference, and this in a precise enough way to already compare the physics dreamed by universal numbers with our observation. 

 

    If the distribution of Maxwell daemons is spare in its distribution over the set of possible computations, does this not imply that the chances of finding oneself as a Maxwell Daemon to be identically sparse. If the set of computations is infinite this make the chance of finding oneself as a MD to be vanishingly small, 1/inf. Does this not seem to indicate against the necessity of ever experiencing oneself as a MD in the first place?

That is a form of Bayesian reasoning that we have already often criticized in this list. It assumes an absolute measure which makes no sense, imo.

If we go further and notice that we seem to experience a relatively long chain of experiences that are MD in nature, does this not make it even more miraculous given derivation of the statistics of distributions of MD in the ensemble of possible computations?

Only the relative probabilities makes sense to extract fits person plural views. 

What if there is a mechanism that allows for the biasing of those statistics such that it is no longer a miracle that we are not just spontaneous Boltzmann Brains. The observer selection ideas of Bostrum

See also my older papers. Such selection is the base of the first person indeterminacy, and of everything which follows from that.

UDA step 7 and 8 already explain that we cannot be 'spontaneous Boltzmann Brains'. If we are machine, we are provably the sum on all 'number theoretical 'Boltzmann brains' or universal numbers/computations. 

seem to give us some good clues about this mechanism. Could it be that a feedback type of relationship might be involved? The choice of what form the material structure takes as it evolves in time could be used to select the information representation that best describes it, but this requires a categorical difference between matter and information such that we cannot think of them as just being transformable into each other.

But that is exactly what is provided by the 'material hypostases". Matter is not information. It is information stabilized by a continuum of computations as seen by numbers. I follow you against easy links between energy and information. Both term are very fuzzy and admit quite different sort of formal specifications. Better to use a top down approach which prevents the evacuation of the subject, consciousness and/or the first person views.

Bruno

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

[Evgenii Rudnyi]

on 17.01.2011 14:00 Bruno Marchal said the following:

>
> On 16 Jan 2011, at 22:27, Evgenii Rudnyi wrote:
>

...

>> Have you meant that the Universal Dovetailer will act for such a

>> situation according to Poincar� recurrence?
>
> The UD will do that an infinity of times, given that the Poincar�

> recurrence is a computable process. But the physical laws are sum on
> first person views, based on a continuum of histories, so to relate
> thermodynamic to the UD is certainly not obvious at all. The UD is
> just a way to provide the minimal third person ontology (the
> 'everything') needed when we assume mechanism, and its role is to
> build a mathematical formulation of the mind-body problem (if only to
> illustrate that science has not yet choose between Plato and
> Aristotle).

Let me write down how I understand this. The 3rd person view gives us a
complete Universe of numbers and the 1st person view perceives just a
part of it. Is this correct?

[Bruno Marchal]

On 17 Jan 2011, at 22:12, Evgenii Rudnyi wrote:

> on 17.01.2011 14:00 Bruno Marchal said the following:
>>
>> On 16 Jan 2011, at 22:27, Evgenii Rudnyi wrote:
>>
> ...
>
>>> Have you meant that the Universal Dovetailer will act for such a

>>> situation according to Poincaré recurrence?
>>
>> The UD will do that an infinity of times, given that the Poincaré

>> recurrence is a computable process. But the physical laws are sum on
>> first person views, based on a continuum of histories, so to relate
>> thermodynamic to the UD is certainly not obvious at all. The UD is
>> just a way to provide the minimal third person ontology (the
>> 'everything') needed when we assume mechanism, and its role is to
>> build a mathematical formulation of the mind-body problem (if only to
>> illustrate that science has not yet choose between Plato and
>> Aristotle).
>
> Let me write down how I understand this. The 3rd person view gives
> us a complete Universe of numbers and the 1st person view perceives
> just a part of it. Is this correct?

Unfortunately it is a bit more complex. Let me try to explain, even if
I introduce simplification, which eventually are wrong. They can be
handled only by the math, which are counter-intuitive on this.

If you want you have the "ultimate" third person point of view, which
is, assuming comp, just arithmetical truth. That is all the truth that
you can write in the language of arithmetic, that is the true formula
build from classical logic (with the symbols "&", "v", "~", "->", but
also "E" (it exists) and "A" (for all), together with the arithmetical
symbols "+", "*", s (successor) and "0".

The semantics of arithmetic formula is rather simple, because we have
a good intuition of the natural numbers. The symbol "0" is interpreted
by the number zero. s(x) represent the x + 1, etc.
For example the semantics of AxAy(x + y = y + x) is given by its truth
condition in the usual structure (N, + *). The formula
"AxAy(x + y = y + x)" is true if it is the case that for all numbers n
and m it is the case that n + m = m + n. OK?
Arithmetical truth is the collection of all those true formula. It is
a highly undecidable set. It contains Fermat theorem, but this has
taken centuries of complex math to prove. We don't know if it contains
Goldbach's conjecture, nor Riemann hypothesis, etc.

This can play the role of your "complete Universe of numbers". It
plays the role of GOD, or the ONE, in the (toy?) theology of the
Löbian machine. It is a highly non effective and non constructive
system, beyond the ability of any machine, and even any machine + a
hierarchy of strong non effective oracle. Actually, such an all
encompassing notion of truth cannot even be defined in the
arithmetical language of any machine (by a theorem due to Tarski).

Now enter the (digital) Löbian machine. A Löbian machine is a
universal machine/number. Universal means that it can mimic any
computable process if you give it enough time and (memory)-space. It
does not mean that it can PROVE all true statement of arithmetic, nor
even define its own conception of truth. By incompleteness it proves
only a tiny part of arithmetical truth. But "Löbian" means that it can
prove its own universality, and so it can prove its own Gödelian
limitations. It is a machine which "knows" that it is ignorant, or
more exactly that it has to be ignorant if it is consistent.

The beliefs of that machine are still third person view, a priori. Its
beliefs can be modeled by its (Gödel) provability predicate, and this
is an arithmetical predicate, and it belongs to the language of that
machine. We have, for p and q arithmetical formula, that
1) if the machine proves p, then the machines proves Bp
2) the machine proves B(p -> q) -> (Bp -> Bq)
3) the machines proves Bp -> BBp
And 4) the Löb formula, the machine proves B(Bp -> p) -> Bp.

Bp is the third person view of the machine. Think about the WM self-
duplication. The machine can prove, given the protocol that she will
be at W and at M. She looks at herself or itself in a third person way
(trusting the doctor about the description of her body).

The first person view is a notion far more subtle. It includes
consciousness or knowledge, which contains an implicit reference to
truth, and this can be used to show that neither consciousness nor
knowledge, nor the very notion of first person can be defined by the
machine. Yet, assuming mechanism, we can meta-define it very
precisely, by introducing an operator linking the belief of a
proposition p (Bp) with a clause saying that it is the case that the
proposition p (is true): Bp & p.
Socrate was already aware of the difficulty to define knowledge, and
in the "Theaetetus" Plato defines knowledge by the true justified
opinion, that is here Bp & p. The machine cannot define it, but could
perhaps reason on the meta-definition, like you and me.

The machine does not prove Bp -> p, nor p -> Bp, so Bp and Bp & p will
obey different logics, having different semantics.

What I call the first person view correspond to the logic of Bp & p.
It is indeed a logic of knowledge, even of evolving, self-developing,
knowledge. It is richer than the logic of Bp in many respect, and it
corresponds to a branch of a splitting and fusing tree, with a
topology akin to the topology of the real or complex numbers (although
a lot of works remains to makes those statements more precise).

And then there is that Skolem-like phenomenon which makes that first
person view richer than any third person view available to the
machine. This is due to the relation between proof (Bp) and the non
definable truth (p). If we describe in a third person view the
(intensional) content of that knowledge, it is much greater than what
the machine can prove, or even that GOD can prove! This is why I say
that arithmetical truth, seen from inside, is greater than
arithmetical truth (seen as a set of third person descriptions). The
thought experiment shows that such truth includes the many contingent
facts, like being in W or being in M, that you, but no one else, can
"know" by self-localization, in time and space.

For example, you are conscious here and now, because some arithmetical
proposition, involving your "body" at the right level of comp-
substitution, happens to be true and provable. But only *you*, (the
non definable knower) can know this, and this introduces a whole
different perspective on arithmetical truth, as seen by you.

Then the observability is defined in a similar way by Bp & Dp (Dp =
~B ~p, Dp is the consistency of p, the fact that there is a world
where p is true). This will explain why you are not only "here and
now", but will most probably stay near "here and now". Again this
makes sense because, by incompleteness, Bp does not imply Dp, for the
machine. Bp -> Dp is true, but not provable, and this changes again
the machine's perspective, and the logic and the semantics available
to the machine. Bp & p leads to intuitionist logic (with topological
interpretations), and Bp & Dp leads to a quantum logic, when "p" is
restricted to the UD-accessible arithmetical propositions, with
(hopefully) an Hilbert/von Neumann algebraic sort of semantics (linear
algebra, quantum mechanics).

Then sensibility is given by Bp & Dp & p. (Reapplication of the
Theaetetus trick). This leads to a qualia logic, extending the quanta.
To be sure, the quanta appears only there, making them qualia, and
making the whole physical reality a first person plural construction.
There is no *primitive* physical universe, only a secondary physical
phenomenon.

To sum up, mechanism leads to a neutral monism, where the basic truth
is arithmetic, and the "views from inside" are given by modal variant
of the self-reference logic (G, the logic of Bp, and their intensional
variants Bp & p, Bp & Dp, etc.). The person are abstraction, and in a
sense "Bp & p" is not a machine. The relation with the truth makes it
non definable in machine or number language. And indeed, the thought
experiment will link the knowledge of the machine with infinities of
consistent extensions which play a role in the semantics of Bp & p.

The ontology is very simple: arithmetical truth
But the theology and epistemology of the numbers will be terribly
complex. If *we* are machine, the theology is beyond the whole of
math, not just beyond arithmetic.

This is a bit like a painting, think of Mona Lisa. The paper on which
it is painted is very simple (a rectangle), but the drawing shape of
it is quite complex. A subset a of a set b, can be much more complex
than the set b. Likewise, the set of total computable function is much
more complex than the set of partial computable function which
contains it, and that is why the arithmetical intuitionistic logic of
machine is more complex than the whole truth, even if none of them are
definable by the machine. You can think also to the Mandelbrot set: http://www.youtube.com/watch?v=9G6uO7ZHtK8

So, to make my point clearer, it is not correct to say that the first
person point of view is a part of the third person point of view. It
is an entirely different *perspective* and a deep enrichment of the
truth, obeying different kinds of logic. The difference is almost as
big as the difference between a book on the history of the humans, and
the fact of being that particular human, or the difference between an
iterated self-duplication resulting in 2^1000 persons, and the fact of
being one of them.

I hope this can help a bit. It is because the notion of first person
is intrinsically complex that I am using the modal logic of self-
reference to handle them. Also, with mechanism, we have no choice: we
have to take into account the self-limitation results.

Best,

Bruno

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