From:	THEORM::STAPP        23-JUL-1997 17:25:59.66
To:	@KLEINDIS.DIS
CC:	STAPP
Subj:	Games and Minds

Dear Aaron,                                         July 23, 1997

Our recent efforts to define more carefully the issue that divides us 
may be revealing the root of the problem. I have looked at the web site 
reference that you gave [http://www.cs.bham.ac.uk/~axs/misc/supervenience] 
and it appears to me that our differences are closely connected to the 
difficulty that you are having in formulating your position carefully .

Your paper was ostensibly designed (at least in part) to meet my 
request for a clean definition of "implement". I had said that: 

"But as regards our differences everything seems to hang on what your 
definition of  `implementation' is: What is the connection between these 
two levels of causal description? ... Your words are unclear on this point." 

It is significant, I think, that your 26 page effort did not come up with 
an answer: it mentioned some problems, and argued that the question was
a difficult one that needed much more study. But I am inclining now to
the view that our differences stem basically from an inherent incoherence in
what you are striving for: that the mire that you seem to be sinking into,
in even defining what your idea is, is the consequence of your being on a 
basically wrong track that stems from trying to evade discussing physics
in what is basically a physics question. 

I shall use here your words, given in your paper, to support this view.

I start (again) from the point that we do not even have a well defined
issue before us unless you give a well defined meaning for "X is 
implemented in Y". You say/suggest that this is supposed to mean the same 
thing as "X is supervenient on Y".

The basic definition of "X is supervenient on Y" is that "Fixing Y fixes X"

But, as you emphasize, there are more complex versions of "supervenience", 
and this simplest one may not be the one that really gets at the basic issue. 
For example, Chalmers finally settles on a variation "X logically
supervenes globally on Y" as the version that best gets at the core issue.

You have not endorsed Dave's version, and I shall describe below my
my objections to it. I use the concept "X is constructible from Y" 
instead of "X is implementable in Y".

"X is implemented in Y" might be more general. But excess generality 
generates logical flabiness that allows disparate things to be confounded.

A key condition on the towers of virtual machines of interest here is that
they can in principle be built  out of physical materials. I want to 
capture this key condition.

As philosophers and theorists we are interested in logical issues.
Our access to  "physical materials" is via our theories about them.

A question amenable to logical analysis is, for some X: "Is X constructible 
from physical matter that conforms to the principles of classical physics?"

This is an engineering-type of question: Can something of a specified kind
be built out of materials of a specified kind? This is a logical issue 
if the things involved are characterized in logical terms.

The question posed above is the one most relevant to science and technology:
if classically conceived matter cannot do the job, then QT must be tried, 
and the entire complexion of the problem is radically altered, as I shall 
explain below.

First, let me comment on your 26 page paper from the perspective
that the pertinent logical question is whether mind can be constructed
from classically conceived matter.

You say at the beginning, in response to similar statement by me, that:
> I am interested in how actual psychological processes (X), including
>conscious and unconscious parts, can be implemented in physical brains
>of various kinds and also computers and other types of machines (Y).

The apparent thrust of that characterization is retained by changing
"implemented in" to "constructed out of physical matter organized into". 
Then, to give restrictive logical meaning to the phrase
"physical matter", I change it to "classically conceived physical 
matter", and, correspondingly, change "how" to "whether".

You write:

>One thing I see no basis for is the fairly common assumption that the 
>only (em/real) kind of causality is physical causality.

A central theme in your paper, and the basic origin of our differences 
---insofar as the classical framework is being assumed---is your effort 
to deny or evade the fact that the causal properties of the 
higher-levels structures are consequences, ultimately, of the 
causal properties inherited from the matter out of which the tower is 
built.

Thus you say that you will address the question:

>If X is implemented in Y, and Y is causally complete, does that mean 
>that X has no causal powers, i.e., is epiphenomenal?

No one can deny that a tornado has  causal powers! But the question
is whether it has any causal powers that are not completely explainable, in 
principle, in terms of the causal powers of the matter out of which it is
made? Does it have any "independent" or "nonreducible" causal powers?

Later you say:
 
>8. It may be assumed that for X to be implemented in Y the causal laws
>inherent in X should be derivable from those of Y. However, that cannot be 
>so in general. If X is a chess playing machine then its laws will be those 
>of chess. However, it is trivial to change the laws in X, e.g. so as to 
>allow the pawns to move backwards under some conditions, without changing 
>the laws of physics. Therefore the laws of physics do not determine the 
>laws of virtual machines implemented in physical systems.
>
>9. What about requiring the laws of X to be derivable from the laws of Y
>together with facts about Y, e.g. facts about the actual structure of the 
>machine Y? That is also too strong, for the concepts of chess (e.g. 
>"pawns") do not stand in any definable relationship to those of physics,
>so no set of sentences describing laws of physics together with the
> physical structure of a physical machine Y can entail a statement such
>as
>   "pawns can never move backwards"
>which would be a law-like true description of any chess-playing virtual
>machine. This notion of ``definitional disconnection'' is very 
>important: in general different virtual machines have ontologies that 
>cannot be defined in terms of those of other virtual machines, including
>those in which they are implemented.
>
>However, this suggestion that somehow the combination of the structure
>of the machine Y, plus the laws of the domaine Y (i.e., ``boundary 
>conditions'' plus laws) should determine whether an implementation of X
>exists or not is not along the right lines, and I'll return to it later.
>However, the determination cannot be logical, or mathematical, at least 
>not when the ontologies of X and Y are definitionally disconnected.
>That's what the notion of "emergence" is all about.

Insofar as the principles of classical mechanics hold, all "emergent"
properties of physical structures are just complex manifestations of
the basic physical laws in particular physical situations or configurations.
If the classical principles hold, nothing fundamentally mysterious, or 
in principle nonreducible to the basic-physics level, can emerge when 
physical materials are put together in complex ways.

The thrust of your remarks is to try to slide around the simple fact that 
if---within the classical physics framework---a machine that is constructible 
out of physical materials functions/behaves as a chess-playing machine then 
that functioning/behaviour must be simply a consequence of the physical 
structure/configuration of the machine and the laws of classical physics.  
That's what classical physics is all about!

There is nothing remotely mysterious---fundamentally---about the behaviour 
of classically conceived tornadoes and chess-playing machines. There can be
high-level rules/laws governing high-level functioning/behaviour. But the
notion of epiphenomenalism is not appropriate: it is a trivial thing that
the high-level laws and causal powers are completely reducible to the
laws and causal powers of the physical materials out of which they are made,
because these things themselves, namely the tornadoes and chess-playing 
machines, insofar as they can be classically conceived, are just 
configurations of those physical materials. [Of course, these names, 
"tornadoes" and "chess-playing machines", are human inventions that are 
dependent on human beings, but we can and do  abstract objective 
meanings, and imagine those physical structures as existing independently 
of us.] 

So there is no problem about the ontological/nomological/causal
properties of things constructible from classically conceived matter.

However, when we consider conscious experiences, within the 
classical framework, two opposite tacks are possible. 

The first tack is to try to maintain that conscious experience is 
a high-level structure that is *fundamentally* just like a tornado or a 
chess-playing machine (though with some more complex functional/behavioural 
properties): fundamentally it is nothing but physical matter arranged in 
a complex way. Then the question of epiphenomalism  never arises: 
high-levels structures can certainly have their own laws of behaviour, 
but these are just manifestations of the laws of the underlying 
classically conceived matter. Those manifestations depend, of course, 
on the configuration of the matter out of which the tower is constructed.

I sometimes believed that you took this first tack. But your 26 page paper 
seems to go off on a completely different tack: you appear to be denying
that the high-level laws are simply complex manifestations of the low-level
(i.e. physical) laws.

You seem to be hewing to a second tack, which claims that the high-level
ontological/nomological/causal structures can be  "emergent" in a sense 
that makes them not fully reducible to the physical-level properties.

The only reason I can see for rejecting the first tack is a judgement,
or prejudice, that conscious is not reducible in the required way.

I do not know how or why  you came to such a conclusion: 
you seem to differ on this score from many other CS people.

But let me descibe *my* reasons for rejecting the first tack.

The assumption, here, is that the principles of classical physics are
adequate. So to get a clear idea of what can by constructed out
of classically conceived physical matter one should have a clear
idea of the classical-physics ontology. Occam's razor tells us to
take an austere ontology: to not pile on needless realities.

The austere natural ontology for classical physics says 
that the only realities are the "local variables". To make this more 
concrete, for computer scientists, imagine a battery of computers, with 
each one assigned to one of a collection of overlapping spacetime regions, 
and programmed to grind out, at an appropriate point in the operation of 
this battery machines, the evolution of the physical system through this 
spacetime region. Each of these computers has in its memory banks only
the physical parameters associated with its own tiny spacetime region. 
[For all regions lying later than some intial time the associated registers 
start with null values, and the grinding out of the computations eventually 
sets the values in these registers to the values associated with that
spacetime region.] There is no additional physical reality: e.g., the sum
of values in registers associated with two distant regions is not part of 
the reality, because this sum is not contained in any register. Similarly, 
what might be imagined to be "in God's mind" is not part of the classically 
described reality: it plays no role in the determination of the evolution 
of the physical reality, and is not part of the classical-physics reality.

There is no trouble fitting tornados and chess-playing computers into
this this ontology (although, as mentioned above, the words "tornado")
and "chess-playing machine", and the concepts they represent are not
parts of this austere classical-physics reality). But our conscious
experiences cannot be part of this austere ontology: how, for example,
can my concept of a prime number be contained in a *disjunction* of
the numbers in the registers associated with tiny space-time regions.
If the regions are taken to be very small then only a small amount of
information is contained in any region, but my idea of the 29th prime
number is quite complex. I find the possiblity that complex human thoughts
could by fully represented in this way so unlikely that I think a search
for an alternative possibility is fully warranted, particularly since 
the implausible possibility is based on a conception of physical reality 
that is known to be grossly wrong at the fundamental level.

In the quantum case the analogous austere reality is far richer: for each 
temporal slice there is a register for EACH OF THE ALTERNATIVE POSSIBLE
CONFIGURATIONS of the values of the registers in the classical model that 
could exist on that temporal slice. That is, there is a "value" (actually
a complex number) for each of the ALTERNATIVE POSSIBLE *entire classical
worlds, at each instant of time,* and the values at earlier times feed 
into the determination of the values at later times. [I am talking about
the state vector PSI[PHI(x),t], the phase factor PSI that is a functional
of the set of fields PHI(x), each of which is a function of x]. So in this
austere quantum ontology there is an element of reality associated with
each alternative possible entire classical world. This includes realities
associated with each alternative possible state of my entire brain.
So we now have some realities that can reasonably be imagined to 
correspond to the complex structure of human thoughts.

But this complex reality, rich as it is, is only part of the quantum
reality: something that lies outside and beyond that physical ontology
is needed to complete the full dynamical description. A fair description
of this extra thing would probably be to say that it is something like 
"the mind of God", because it must do such high-powered things.

In each of the various quantum ontologies that has been proposed the 
other realities must do important things. I will stick here to a 
Bohr/Heisenberg/vonNeumnn/Wigner interpretation.

The extra reality must do three things:
1. select a question (choose a basis)  
2. answer that particular question (actualize one of the basis states)
3. bring the physical state of universe into concordance with 
   that answer. 

Most importantly, insofar as we know today, each question must be a 
question about what some human experience will be, and the answer must
bring the state of the brain of that human being into correspondence
with that experience. This is the way the theory is set up, in order
to account for *all* phenomena, including phenomena adequately described
by using the classical approximation. 

Although only events associated with human experiences are required by
the theory, the theory  allows similar events not associated with human 
beings. But no evidence for their existence has yet been uncovered.

This completes my explanation of why I reject the classical-physics 
ontology, and how our conscious experiences fit into a quantum-physics 
ontology.

Let  me explain why I think Dave's selected version of the supervenience
condition is objectionable. Basically, it is because it tacitly assumes 
a classical-physics ontology.

The basic step in Dave's argument is when he argues that Fixing the entire
course of physical-world history does not logically fix our conscious 
experiences.

But without any specification of what physical reality is like
someone could argue that "Brain action A invokes experience E"
is scientifically equivalent to "Brain action A is experience E",
and that the "Brain action A" is merely some outside observer's 
experience/interpretation of what is "actually" experience E,
which is merely one of the host of experiences that constitute
physical reality. Indeed, the QT conception of "physical reality" 
seems to be moving in the direction of a conception of the world
as a collection of `experiential' realities that somehow bind each other
together into a structure of great mathematical regularity.
The whole intuition that drives Dave's argument is that physical
reality is basically something like our classical conception of it:
that there are microphysical facts, and other physics facts that
are not basically made up out of phenomenal facts.

So this assumption, which may be grossly incorrect (as QT may be 
suggesting), ought to be made up front. (Perhaps Dave does cover himself 
on this point, somewhere in his book.)

If Dave's argument is applied to the sort of world that QT suggests it
to be then the argument fails.

First, there is the problem with the idea that we can think of the 
whole course of history as all laid out: according to the orthodox 
interpretations of quantum theory one should not even imagine
the future as being well defined: it only becomes well defined
when experiences occur. But in that case the logical disjunction
between the factual world and the phenomenal world that Dave tries to
defend cannot be justified: the only true facts are phenomenal facts.

My message here is again that the basic issue is a physics question, and 
that approaches that blithely ignore the central physics questions, 
by tacitly assuming the validity of the classical-physics ontology, are 
blinding themselves to *the* essential question of how our experiences 
enter into, and influence, the flow of natural events.

Henry