No Zombies, Only Feelies?
In this fictitious dialogue Arthur
has read the paper
Why and How We Are Not Zombies and Dean has read
Visions of Mind and Body
The Perspex Machine,
as well as the papers
Perspex Machine III:
Continuity Over the Turing Operations,
Computation of the General Linear Transformations.
Arthur advances the position
that we can never have a scientific theory of what feelings are,
because we can never know what anything feels like to another
person. Dean, on the other hand, defends the point of view set out
in the Book of Paragon that, in the far distant future, we might be
able to do this by building robots that can be transformed into
human beings and other animals, thereby allowing the robots to experience
the feelings of many creatures - and ending up as each one of us so
that we know what it feels like to be another being. In the mean
time, Dean advances a theory of how to build practical robots that
have feelings and describes, in broad outline, how a robot might
shift from one body to another. Dean argues that it is impossible to
build a human zombie that works in every way like a human being,
except that it has no feelings. He tries to show why any close
analogue of a human being must have feelings. In effect, Dean argues
that there are no zombies, only feelies, though he does describe
theoretical conditions in which a robot might be devoid of feeling.
Arthur: Hello, Dean. Have you read
paper? It says, in effect, that each one of us can know if we have feelings,
but we can never know if another being has feelings. This is part of
the other minds problem. We can
know our own mind, but we can never know that another being has a
mind, and we can never know what that mind feels or thinks. Do you
agree with that?
Dean: In my everyday life I do not
experience direct knowledge of other peoples' minds. I do not
experience telepathy, and I doubt if any biological creature does. So,
here and now, I have to say that I do not believe that any of us has
access to another's mind. But I think you go too far when you say we
humans can never know another being's mind.
Arthur: I'll grant that by
observing another being we can be pretty sure that it has a mind,
but we can never be absolutely certain. For example, a robot that
appears to go about its daily life in an intelligent way might
actually be programmed to behave in a pseudo random way, without
anything I would call intelligence. This being the case, we can
never be entirely certain that another being has a mind and, even if
it does, we cannot know what it feels or thinks even if we believe
we have a pretty good idea of what is going on in there. I'll swear
that my wife seems to be able to read my mind even though she says
she is not telepathic. But do you really believe we could have
access to another's mind one day?
Dean: Certainly, but the
argument is a long one. I will sketch it out for you, if you will
allow me to make the materialistic assumption.
Arthur: Ah, yes. The assumption that
everything that exists is physical. Very popular amongst scientists,
I understand. Well, for the purposes of debate, I will grant you the
materialistic assumption. But how does that help us to experience
Dean: For a start it means
that everything has a physical basis. Minds, feelings, and,
mathematics are all expressed in some physical medium, be it a body,
text book, computer, or whatever. There are no Platonic ideals that
exist only in the abstract. Everything is physical. However, I will
admit that a mind might live on after the death of its body, but
only if it is reincarnated in another body, whether biological,
robotic, or spiritual. Because I have adopted the materialistic
assumption, I suppose that the spiritual world exists as a physical
reality, if it exists at all.
Arthur: Let's not stray too
far from the zombie debate. How might this help us feel what another
Dean: The perspex machine is a
theoretical machine, but it is more powerful than any theoretically
possible digital computer. It can describe continuous things, not
just discrete, digital things. I suppose it can describe the
whole universe regardless
of whether the universe is actually continuous or else quantal in
Arthur: Now hang on a minute.
That is a mighty big assumption to make!
Dean: Not really. All
contemporary physical theories are expressed in symbols, in
particular they can be expressed in computers and can even be
simulated in computers. But this means that a perspex machine can
express and simulate all contemporary physical theories because the
perspex machine can do everything that a digital computer can, and
more. So I am certain that any part of the universe that is known to
the literature of the physical sciences can be described by a
perspex machine. I then make the assumption, as scientists do, that
the unknown part of the universe can be described in more or less
the same way as the known part. I might be wrong on this, and other
scientists might be wrong on this when they make their assumptions, but until
there is good evidence to give up the hypothesis, I will continue to
believe that the perspex machine can describe the whole universe.
Arthur: Well, what if it can?
Dean: The perspex machine,
like anything that exists, has to be described or instantiated in a
physical medium. Now, the perspex machine describes the position and
motion of perspexes, but everything that exists has some sort of
position and motion, even if they cannot be determined in any
practical way. Perspexes can describe arbitrarily simple or complex
things, and can be degenerate so that they lack position and/or
motion. This means that any physical substance or process can be
used to implement some sort of perspex machine, no matter how large
or small a part of the general perspex machine the substance or
process can support. In fact, I have implemented several simulations
of the perspex machine on digital computers so I know the universe
can support quite complex perspex machines. Of course, a substance
or process might have properties that are not essential to the
implementation of any perspex machine I can describe in words, but I
assume that these properties, along with the whole of the universe,
can be described by some perspex machine. In short, I assume that the
universe can be understood as a perspex machine.
Arthur: For the purposes of
the debate, I'll grant you the big picture. But I thought you said a
perspex machine can do more than a digital computer, so how can you
simulate it on a digital computer? More to the point, how can you
describe it in the words we are using here?
Dean: Sorry, I have been too
technical for you. Simulations are approximate, emulations
are exact. As I said, I can simulate the general perspex
machine on a lap top, but I cannot emulate it on a lap top.
In much the same way, the general Turing machine can be simulated on
a lap top, but it cannot be emulated on one. Similarly, I can
describe the perspex machine approximately in words, but I cannot
describe it exactly in words. Does that make the matter clear?
Arthur: Let's not get into the
theory of computability, Dean. Stick to feelings. If I grant you
that the universe is a perspex machine, how might this allow us to
feel others' feelings?
Dean: You have read the
paper, so you tell me how a robot might have feelings.
Arthur: Well, it is possible
that today's robots have feelings. But I don't believe it for one
minute. I do believe it might be possible to program a robot to have
feelings, but I have no idea how to do it and, even if it could be
done, I cannot see how we could ever know that a robot has
Dean: Let us call robots with
no feelings "Zombies," and those with feelings "Feelies." Now, the
tells us that we can transform one computer program into another so
that a Zombie's program can change over infinitely many,
infinitesimally small, steps into a Feely's program, and vice
versa. If the memories of Zombies
and Feelies are sufficiently similar then a Zombie can remember what
it was like to have feelings as a Feely, and a Feely can remember
what it was like to be devoid of feelings as a Zombie. On the other
hand, if their memories are too dissimilar we can augment the
Zombies and the Feelies with a perspex program that holds the memory
of the other and transforms it into the host's, Zombie or Feely, memory.
Arthur: I suppose it is
possible that you might fail. Zombies' and Feelies' memories might be
so different that you must use the perspex memory translation
program, but, even if you do, the translation might be
incomprehensible. A Zombie might simply be incapable of remembering
feelings, and a Feely might be incapable of remembering the total
absence of feelings.
Dean: Perhaps. But the
tells us that we can make infinitesimal changes to a program. We can
take an infinitesimal step from zombiehood to feeliehood, or vice
versa. So whilst it might be the
case that an extreme Zombie and an extreme Feely cannot understand
each other, it is almost certain that there are infinitely many
robots in between that can understand each other and experience each
other's memories. If this is not the case then it has to be the case
that feelings and
their absence is an exact thing. There is no room for a smooth
transition from feeling to non-feeling. Now, I do not know about
you, but it seems to me that as I grow tired and fall asleep my
sensitivity to the world does decline toward total unconsciousness.
I do experience a diminution in feelings. Conversely, when I wake up
the reverse seems to happen so that I experience an increase in
feelings. The same happens when I am throttled and recover, or when
I am beaten unconscious and recover. Yes, Arthur, I have lived an
adventurous life. But it is one where I have experienced gradations
of feelings. I have also experienced the scholastic life. I know
enough physics and biology to know that animals do not work in an
exact way, yet I am an animal and I have feelings, so I believe that
feelings are not exact, they can be felt by an inexact creature like
me. I suppose you feel the same way and are now ready to admit that
if there is a continuum of robots with different gradations of feeling
then at least some of them can share memories. In
short, they can know each other's minds.
Arthur: I will grant you that
much, for the purposes of discussion, but how does that
help us to
know another's mind?
Dean: Remember the
materialistic assumption. Everything that exists is physical. The
perspex machine is a physical thing. It deals with the position and
motion of objects, but our bodies are made up of molecules, atoms,
and sub-atomic particles that have some sort of position and motion.
We are perspex machines so the
applies as much to our bodies as to robots' bodies. I expect that
one day it will be technologically possible to build a continuum of
robots that can experience the feelings of, at least, their near
neighbours in the continuum and that this continuum will cover all
animals, plants, and inert materials on planet Earth.
And when might this be?
Dean: I expect it will take
thousands of years to build robots that have minds like ours and
millions of years to develop the technology to build the mental
continuum. But it does not matter how long it takes ...
Arthur: ... because I said
never and you said it could be done some time.
Dean: [Nods in ascent.]
Arthur: Dean, I had always
supposed that the lowest level of a computer is a word of some kind,
machine code, micro code, or something like that. But you deny this?
You claim that the lowest level of anything is something
geometrical. So, presumably, the lowest level of a computer is the
geometrical layout of electronic tracks in silicon, or something
Dean: Of course. If you
believe that the lowest level of a computer is a word then I invite
you to edit the words in my lap top and turn it from base metal and
plastic into gold. Better still, I would like a couple of kilograms
of platinum, please.
Arthur: Ha, ha. You've got me
there. Lucky for me your lap top isn't here!
Dean: No problem. I will fax
it to you. [Dean winks at Arthur.]
Arthur: I will grant that you
have hypothesised a solution to the other minds problem, but you
have said nothing about what feelings are. Have you anything to say
Dean: Certainly. Many feelings
serve a functional purpose. If I bang my toe on a stone I move out
of the way so as not to fall over it or, perhaps, I suppose that my
kicking a stone proves some philosophical point to my companion.
These kinds of feeling have a purpose. It is less clear that other
kinds of feeling have a purpose. Many amputees say they can
still feel their amputated limbs. Some even take analgesics to
reduce the pain they feel in the amputated limb. I am not sure what
kind of purpose the feeling of phantom limbs has, but it is reported
as a feeling.
Arthur: I agree that some
feelings have a purpose, but that does not tell us why they feel
like anything. For example, why cannot a robot have abstract
mathematical feelings implemented in functions that do not feel
Dean: Because you granted me
the materialistic assumption. There are no abstract feelings, no
functions devoid of a physical basis. There is a physical
environment, or content, to every implemented function. This
physical content is part of what we call a feeling. The functional
relationships are the other part. For example, the book
Visions of Mind and Body
argues that today's computers feel the timeness of time both as a
functional ticking in their clocks and as the impact on their
performance of the elapsed time since their
last re-boot and original manufacture.
Arthur: Too abstract! Tell me
how a robot can feel. How can it feel a sunrise, or the cold of the
Dean: If a robot is too cold
it cannot function at all. As it warms up toward operating
temperature it goes through various error states that cause it to
operate briefly before re-booting. It is disoriented. At operating
temperature it operates correctly. It senses all manner of things
with its cameras, temperature sensors on the surface of its body,
microphones in its head, and so on. All of these sensations are
correlated. The sun supplies light for the cameras, heat for the
temperature sensors, and changes the density and currents in the air
so that the robot hears different things. As the temperature rises
beyond operating temperature the robot suffers hallucinations as
memory errors cause it to mix up the data between its senses. It
makes erroneous judgements and, eventually, it stops work and melts
into a puddle. In much the same way, when I suffered hypothermia I
was disoriented, but when I returned to normal body temperature I
operated correctly. At normal temperature I could see the sun, feel
the warmth of its rays on my face, and hear the gulls squawking in
the distance. I over did it a bit in thermal clothing. I began to
hallucinate and fell unconscious. Fortunately my buddies sorted me
out long before I melted into a puddle! Now, all of these things
that happened to me, or could happen to a robot, have a common
physical cause. The effects of heat on a physical body. The
functional relationships and the physical content could be similar,
differing only in the way that silicon chips differ from biological
neurons - or whatever. Apart from function and physical content
there is no other possible difference so, if the physical basis of a
robot is sufficiently close to my physical basis, it feels. I do not
claim that all robots can feel, only that some can. And I claim that
there is no more to feeling than functions and physical content.
Arthur: You claim that some
robots might not have feelings. How can that be if they have
functions with physical content?
The Perspex Machine
explains this. Feelings are not
atomic things. One can be conscious of a feeling, remember a
feeling, be mistaken about a feeling, and so on. There are many
facets to feeling. The glossary of the book hypothesises functions
that give rise to consciousness, feeling, intelligence, morality,
and so on. It is then a simple technical matter to block a robot's
sensation of the world, or break one of the two relationships needed
for consciousness. Such a robot might be devoid of feeling, but it
would be practically useless. It would not sense the world, it would
not be able to form any relationship between its ideas and the
world. In fact, its ideas would be disjoint from each other. It
might exist, but it would be bloody useless. A biological creature
that worked that way would not survive long in a competitive
environment. In fact, devoid of any form of homeostasis, it would
scarcely survive at all. So I am pretty sure that all animals, and plants, have
feelings. Moreover, I believe that my lap top feels the timeness of time
and, perhaps, a few other things.
Arthur: [Putting down
The Perspex Machine.]
Wow. That glossary is dense. I'll grant that you have hypothesised a
solution to the mind-body problem. And you have hypothesised that it
could be put to the test in a few million years' time. But how does
that help us here and now?
Dean: That depends on how
plausible you think my hypothesised solutions to the other minds
problem and the mind-body problem are. That is a matter entirely for
you. I do not care a fig for these problems. I am building a perspex
robot and want to see what it can do using the technology available
to me today, including the technology I have invented, and might go
on to invent during the remainder of my working life.
Arthur: Why did you invent the
number nullity explained in
Computation of the General Linear Transformations?
Dean: Because of the
Arthur: Let me see. It was
once thought that the eyes focus the world on the pineal body in the
brain. So it was thought that the pineal body sees the world. But
the real question is how the pineal body could do that without
involving a little person, or homunculus, sitting in the pineal body
that does the seeing. In fact, what is wanted is a
neurophysiological explanation of the visual pathways such as, or
better than, we have today. Whenever explaining a human faculty we
want a mechanistic explanation, not one that pushes words around and
leaves the problem right back in the human. But what has that to do
Dean: Integer arithmetic is
fine, but rational arithmetic, and more advanced arithmetics all
involve division by zero. Division by zero is not defined so
whenever it arises a human mathematician has to get involved and try
to sort it out. Division by zero turns up in an awful lot of
mathematics and, so too, do related geometrical properties. Points
that are co-punctal or co-linear or co-planar are banned from all
sorts of geometrical operations. When they turn up a human
mathematician has to sort them out on a case by case basis. And so
it goes on. Almost all of mathematics is infected by the homunculus
problem - corner cases are not defined so a human mathematician has
to get involved to try to sort them out. This is one of the things
that makes computer algebra so hard. All of the corner cases have to
be defined so that a computer can solve algebraic problems without
assistance from a human.
Arthur: So how does the number
Dean: I defined a canonical
form for numbers divided by zero and let the rules of arithmetic
hold regardless of division by zero. This produced a new arithmetic
that contains the arithmetics that people commonly use, but one in
which division by zero is well defined. Thus, I removed the
homunculus problem from this part of mathematics. That was all that
I needed to do in order to define the perspex machine in a way that
is guaranteed to be able to operate without human intervention. The
number nullity, together with the number infinity, makes the perspex
machine a suitable physical substrate for a robot's mind and body.
That is what I wanted to achieve. Back then I did not know how to
implement a mind using conventional mathematics - so I changed
mathematics. Now I could do it in conventional mathematics, but I do
not want to. Nullity makes all sorts of calculations easier.
The Perspex Machine
defines causality. Can you shed any
light on that?
Dean: [Smiling] Dear, Arthur.
It is a technical matter. The definition of causality in the that
book is the definition of causality in the theoretical perspex
machine, not in the universe. It is hypothesised that the
theoretical perspex machine can describe the universe, but it has
already been proved that it cannot do it in a direct way. The
explanation of causality in the universe would be a very complex
thing in the theoretical perspex machine and that explanation would
operate, picking out objects in the world and giving explanations,
in the causality explained in the book. It is simple really. The
causality in the book explains the theoretical perspex machine and
the theoretical perspex machine can, it is supposed, explain
causality in the universe.
Arthur: But how can the
perspex machine explain random things that occur in the universe, if
they do occur, that is!
Dean: Easily. A genuinely
random number can be generated by an infinitely long Turing program,
but the perspex machine can execute all Turing programs, even those
the Turing machine cannot complete, so it can complete the infinite
Turing program, and can generate any number of genuinely random
numbers. Now, no existing physical theory employs the number
nullity, so one can put the infinite machinery of random number
generation in a nullity subspace and have it operate at a distance
on a non-nullity subspace that is identical to the geometrical
spaces used in contemporary physical theories. Thus, the perspex
machine is a superset of existing geometrical theories of physics
and can have, or fail to have, genuinely random numbers as is wanted
in a theory.
Arthur: How does the geometry
of space affect the perspex machine?
Dean: One has some freedom to
choose the geometry and the instruction the perspex machine
executes, but not total freedom. Some geometries limit the
instructions that can be embedded in them, and some instructions
limit the geometries they can be embedded in. When one designs a
perspex machine the choice of geometry and instruction affects the
bodies and minds that can arise in the machine. The same holds, of
course, if one is in a position to design a physical universe.
Arthur: You said, "one has
freedom." Does the perspex machine explain free will?
Arthur: Oh, bugger!