The following is a preprint of an article that has been submitted to a peer-reviewed journal. A PDF version of the document can be accessed via PhilArchive here. The following is a cut-and-paste version, so the equations did not print properly. Instead of messing around with HTML, I decided to just include the text below with a link to the actual PDF here.
(error) (1)
(error) (2)
Relativistic Implications for
Physical Copies of Conscious States
Abstract:
The possibility of algorithmic consciousness depends on the assumption that
conscious states can be copied or repeated by sufficiently duplicating their
underlying physical states, leading to a variety of paradoxes, including the
problems of duplication, teleportation, simulation, self-location, the
Boltzmann brain, and Wigner’s Friend. In
an effort to further elucidate the physical nature of consciousness, I
challenge these assumptions by analyzing the implications of special relativity
on evolutions of identical copies of a mental state, particularly the
divergence of these evolutions due to quantum fluctuations. By assuming the supervenience of a conscious
state on some sufficient underlying physical state, I show that the existence
of two or more instances, whether spacelike or timelike, of the same conscious
state leads to a logical contradiction, ultimately refuting the assumption that
a conscious state can be physically reset to an earlier state or duplicated by
any physical means. Several explanatory
hypotheses and implications are addressed, particularly the relationships
between consciousness, locality, physical irreversibility, and quantum
no-cloning.
Keywords:
Strong artificial intelligence; mind uploading; computer consciousness; special
relativity; quantum no-cloning; Wigner’s Friend
I. INTRODUCTION
Either it is physically possible to copy or repeat a conscious state or
it is not. Both possibilities have profound
implications for physics, biology, computer science, and philosophy.
Whether or not consciousness can be copied is fertile ground for a
multitude of troubling, if not fascinating, thought experiments. There’s the duplication problem [1]: imagine
we can teleport a traveler to another planet by creating “a precise duplicate
of the traveler, together with all his memories, his intentions, his hopes, and
his deepest feelings,” but then we decide not to destroy the original copy? “Would his ‘awareness’ be in two places at
once?” There’s the simulation problem
[2]: if conscious awareness can be uploaded onto a computer, then how do we
know we aren’t simulated minds in simulated universes? In fact, because simulated universes are much
less expensive, in terms of matter and energy, than actual universes, then if
consciousness can be duplicated, we are almost certainly one of vast numbers of
simulated copies. There’s the problem of
self-location [3]: if a psychopath tells you that he has created an exact
physical copy of you and will torture it unless you pay a hefty ransom, then
you should pay the ransom unless you are absolutely sure that you aren’t the copy. There’s the problem of the Boltzmann brain: if
consciousness is just the result of atoms in a brain, what’s to prevent a set
of physically identical atoms, somewhere in the universe, from accidentally
coming together in just the right way to create your brain? And what would that feel like?
Notice that each of these problems is a direct consequence of the copiability or repeatability of conscious states.[1] If it turns out, for whatever reason, that
conscious states cannot be copied, then these and other problems
disappear. Whether each of the above
scenarios is actually possible is an empirical question, and given the rate that
technology is advancing, it may be just a matter of time before each is
tested. There are related problems that
may never be testable such as, “Will a computer ever become conscious – and how
would we know?” After all, there is no
consensus on how to measure the existence or level of consciousness in an
entity. Some say that consciousness
depends on the ability to pass a hypothetical “Turing test.” Some say it depends on the level of
complexity in neural networks. Some say
it depends on certain activity in the brain.
So how can we possibly learn anything about the nature of consciousness
if it depends on a definition?
It is easy to get bogged down in the meaning of “conscious” and lose
sight of the big picture. It’s as if the
question, “Will a ball near Earth experience a force of gravity?” has been
preempted by, “What do you mean by ‘ball’?”
So let me rephrase: “If I am near Earth, will I experience a force of gravity?”
My goal is to derive, if possible, an objectively correct prediction to
several questions such as: will it ever be possible to teleport a copy of
myself to another planet? Will it ever
be possible to upload my conscious awareness onto a computer so that I can
outlive my physical death? Will it ever
be possible, perhaps in millions of years, for a collection of atoms somewhere
in the universe to accidentally come together in just the right way to create
my conscious awareness? Will it be
possible for me to experience these things?
The answer is yes only if my conscious states are fundamentally such
that they could be copied or repeated.
Are they?
First, let’s consider the question, “Will it ever be possible to upload
me to a digital computer?” This a
scientific question if it can be phrased as a falsifiable hypothesis. “It is impossible to upload me to a digital
computer” would be falsified, for example, by my being conscious and aware
while possessing a body consisting essentially of on-off switches. But what do I mean by “me”? I mean my identity – but what is that? Questions like this lead many to reject from
the realm of scientific inquiry anything involving identity or
consciousness. But that’s a mistake
because science is about making predictions about our own experiences, and
without a consistent sense of identity, whose experiences are we
predicting?
In this paper, I
will assume it is possible to copy conscious states and show how this
assumption in conjunction with special relativity lead to a
contradiction. In doing so, I will rely heavily on the extent to which
consciousness and identity supervene on an underlying physical state. For this reason, I will begin with a
discussion of identity and why objections based on identity are inadequate to
counter the conclusions of this paper.
II.
QUESTIONING IDENTITY
Let’s consider what it means to upload me to a digital computer. I think it means two things. First, that it’s me, that I experience
it. (Let’s call this existence of
identity.) If a scientist tells me that
an exact copy of me has been uploaded to a computer, but I don’t experience
whatever sensations accompany living inside a computer, I will respond that I
have not been uploaded. “But I can’t
tell the difference between you two,” the scientist may respond. “Yes, but I can,” I would reply, “and you
have failed to upload me.” Second, that
my identity would persist; that as the computer physically evolves in time
according to the laws of nature, the consciousness it creates would continue to
be mine. (Let’s call this persistence
of identity.) In other words, if the
computer begins in some initial physical state S1 that produces me
in conscious state C1, and after awhile physically evolves to
physical state S2 that happens to produce conscious state C2,
that conscious state C2 will still be mine, and I subjectively
experience an evolution from conscious state C1 to C2. And if, due to a random quantum event, the
computer were to evolve to physical state S2’ (instead of S2)
that happens to produce C2’ (instead of C2), then
conscious state C2’ will still be mine and I subjectively experience
an evolution from conscious state C1 to C2’. I don’t expect the scientist to guarantee
that I will live forever – after all, the physical evolution of the computer
could simply stop producing consciousness – but I do expect that as long as the
physical evolution of the computer continues to produce conscious states, I
will be the person experiencing them, not someone else. In other words, if a scientist claims to have
uploaded me to a digital computer, if I don’t subjectively experience (in the
form of evolving conscious states) the physical evolution of the computer over
time, then that scientist has failed.
Let me combine existence and persistence of identity into one concept:
temporal continuity[2]
of identity (or transtemporal identity).
In this paper the
word “consciousness”[3]
includes identity and a “conscious state” is a state experienced by someone;
that is, a person experiencing a conscious state has an identity and one
conscious state cannot be experienced by more than one person. So if a particular physical state produces a
conscious state, and that physical state includes (among other things) light
sensors that are absorbing photons having a wavelength of 680nm, then the
produced conscious state may correspond to someone’s subjective experience of
seeing the color red, in which case we can ask who is experiencing that
conscious state and whether that person is seeing red.
A. Boltzmann Brains and Bad Science
As it turns out, the Boltzmann brain problem is much bigger than the
weirdness of the occasional human brain fluctuating into and out of
existence. In cosmological models in
which Boltzmann brains dominate, not only is any brain much more likely to be a
random fluctuation, but any complete person, world, or even galaxy is much more
likely to be a random fluctuation. Are
you a Boltzmann brain? You might think
the answer is no because you can see a vast world around you and you seem to have
many years’ worth of memories that are consistent with a long-lasting,
law-abiding physical world. However, the
overwhelming majority of people who would make such claims are, in these
models, random fluctuations. They are
“Boltzmann observers” who have no right to believe that any of their
observations are reliable or that their scientific predictions will come to
fruition.
This is a problem. Science, after
all, is about collecting evidence and making predictions based on falsifiable
hypotheses. However, if a model based on
science predicts that the foundation of science itself is not reliable – then
what? Sean Carroll eloquently sums it up
[4]: “it’s overwhelmingly likely that everything we think we know about the
laws of physics, and the cosmological model that predicts we are likely to be
random fluctuations, has randomly fluctuated into our heads.” In other words, if a model that is based on
physical evidence in the world around us predicts that that very evidence is an
illusion, then we should reject such a model as “cognitively unstable.”
A similar fate befalls denial of identity because the foundation of
science fundamentally depends on transtemporal identity. “A ball falls toward Earth at 9.8 meters per
second squared” is an hypothesis that, if incorrect, can be falsified through
experimentation. But that requires a
scientist to prepare the experiment, then perform it, then collect and analyze
data, and then determine whether the hypothesis has been falsified. This series of events, all of which are
necessary in scientific investigation, occurs in a temporal order, and if the
scientist did not or could not expect to be the same person throughout, then she
would have no reason to believe that she could test falsifiable
hypotheses. She would have no reason to
trust science. Therefore, any theory or model
that rejects the temporal continuity of identity should be rejected as rendering
science internally inconsistent. Further,
while transtemporal identity is necessary for scientific inquiry, it is not
itself subject to scientific scrutiny because no experiment can falsify it; thus,
there is no scientific rationale to doubt the existence of transtemporal
identity.
B. Fissioning Philosophers
Philosophers have had much to say about both the existence and
persistence of identity, but their analyses almost exclusively depend on the
assumption that conscious states can be copied.
The typical philosophical analysis of identity goes something like this:
Identity is produced by the brain; the brain is just a chuck of matter that
can be moved around, copied, replaced, and so forth, at will; therefore, the
notion of identity is problematic.
For instance, Parfit begins his analysis [5] with, “We suppose that my
brain is transplanted into someone else’s (brainless) body, and that the
resulting person has my character and apparent memories of my life. [I shall here assume] that the resulting person
is me.” He then proceeds, in thought
experiment, to slice up the brain, trade brain sections with other people,
sever connections between hemispheres, and so forth, to arrive at various
conclusions (such as the “fission” or “fusion” of identities) that are at odds
with transtemporal identity. Zuboff [6] also
starts with replication of a brain that “is a precise duplicate of yours in
every discriminable respect...” and concludes that so-called fission – in which
“one subject can, in a single next moment, experience two differing
non-integrated contents of experience” – is the necessary outcome. These “brain-in-a-vat” style thought
experiments have dominated philosophical theories of mind and identity, not to
mention science fiction, for decades, and represent the best prima facie
arguments against transtemporal identity.
However, an argument that relies on premise X cannot be used to disprove
an argument for ¬X. My goal in this
paper is to show, among other things, that conscious states cannot be
copied. Because the typical
philosophical arguments depend on the in-principle copiability of brains,
mental states, etc., then not only are they inadequate to counter the arguments
in this paper, they are entirely inapplicable.
In other words, the strongest arguments against transtemporal identity
depend on the assumption that conscious states can be copied, but this is the
very notion that I am attempting to refute.
Therefore, any notion or theory of identity that depends on the
copiability of conscious states (or the copiability of brains in conjunction
with the supervenience of consciousness on brains) is simply irrelevant to the
present analysis.
C. Universal Beliefs
Why fight the notion of transtemporal
identity in the first place? After all,
it’s what we all implicitly believe. Every
decision we make about the future depends on our belief in identity. Let’s say that a travel agent was to offer you
the following options:
a) For
$10,000, you will spend a week at a resort in Tahiti;
b) For
$6,000, an exact physical copy[4] of you will be created and
will spend a week at a resort in Tahiti while your current body is anesthetized;
c) For
$2,000, a really good copy of you – one that looks, acts, and talks just like
you but is not an exact copy – will be created and will spend a week at a
resort in Tahiti while your current body is anesthetized.
In all three cases, the underlying
question you’ll no doubt ask yourself is, “What will I experience?” If you believe that an exact physical copy of
you inevitably includes your identity, then you might be tempted to go with
option b), given that the $4,000 in savings could be enjoyed by you at a
later date. And if you believe that a reasonably
good copy of you includes your identity, then you might be tempted by option
c). In other words, your potential
choice of options b) or c) depends on (your beliefs about) whether your
identity can flow to a copy of you, not on whether you have an identity. When choosing, you are, presumably, not
thinking about neural firings, corresponding to pleasure sensations, in a brain
that is similar or identical to yours; you are thinking about whether you
will experience those pleasure sensations.
This is also true, I suspect, of those philosophers who officially refute
transtemporal identity. This is easily
demonstrated by proponents of the Many Worlds Interpretation (“MWI”) of quantum
mechanics, in which a similar identity problem notably arises in the question
of what an observer experiences at some quantum mechanical “branching” event
(cf. [7]). Imagine an observer,
originally in state (error)
and
such
that a measurement of it will yield outcome A with 99% probability or outcome B
with 1% probability:
The measurement apparatus, originally in state
, is designed so that if outcome A obtains, the observer
will experience intense pain, while if B obtains, he will experience neither
pain nor pleasure. According to MWI, which
assumes the universality of quantum mechanics at any scale, the experiment
causes the entire system to linearly evolve according to:
Notice that the final evolution in Eq. 2 remains a superposition in which
a state of the observer corresponding to observation of outcome A (and thus the
experience of pain) is correlated to outcome A, and vice versa for outcome
B. In collapse interpretations of
quantum mechanics, the wave state in Eq. 2 instantaneously reduces to either (error), a world in which the observer
experiences intense pain with a likelihood of 99%, or (error), a world in which the observer avoids
pain with a probability of 1%. Transtemporal
identity is no problem in this case because the pre-measurement observer can
expect to become one or the other of the post-measurement observers. However, because MWI is a non-collapse
interpretation in which every term in a quantum wave state, no matter how
macroscopic, corresponds to an actual world, MWI implies that after the
measurement, there is a real observer correlated to outcome A and a real observer
correlated to outcome B. Identity fissions,
and neither post-measurement observer has a better claim to “being” the
pre-measurement observer than the other [8]: “All your future selves, on all
your branches, are equally real and equally yours.” Like many philosophers, the MWI proponent can,
in an academic setting, disavow any belief in a transtemporal identity. Perhaps he will try to console himself by
noting that the experiment will cause his identity to fission so that, after
the measurement, there will exist an “equally real” version of himself that has
experienced no pain. But I suspect that,
despite fanciful arguments discounting notions of enduring identity, he will be
more than a little nervous to carry out the above experiment. I suspect that when push comes to shove, he
will in fact be afraid of the pain that he – not someone who looks and acts
just like him, but he – is likely to experience.
D.
Much Ado About Nothing
The most likely objections to the arguments in this paper will target transtemporal
identity. But even if I haven’t
convinced you that science without identity is unscientific, or that philosophy
without identity is inapplicable to this paper, or that belief in transtemporal
identity is essentially universal, it doesn’t matter because the issue of
identity is a red herring.
My goal in this paper is to determine whether or not conscious states can
be copied, where a particular conscious state is identified with a particular
person. That is, I want to answer
questions like, “Is it possible in principle to upload myself to a computer?”
or “Is it possible to teleport me to a distant galaxy?” To do so, I simply assume that a person’s
conscious state (i.e., a conscious state corresponding to someone’s
transtemporal identity) supervenes on some underlying physical state that can
be copied, and then show that this assumption leads to a contradiction. But if a person’s conscious state does not
supervene on physical state (whether because identity is an illusion, or
identity fissions, or whatever), then we come to the same conclusion that
conscious states cannot be copied, because there is nothing physical that can
be copied that will produce a conscious state of that person.
Stated bluntly: if there is something physical that can be copied that
causes me to be uploaded to a computer, then what are the
consequences? (That is a question I will
explore in Section III.) But if there is
not something physical that can be copied that causes me to be
uploaded to a computer, then we have the answer: (my) conscious states cannot
be copied.
III.
WHY CONSCIOUS STATES CANNOT BE COPIED
A. Assumptions
In this section, I’ll lay out the structure of my argumentation as well
as some of the assumptions I make. I
will let conscious state (or mental state) include identity so that two or more
different people cannot experience the same conscious state. However, different conscious states could
clearly be experienced by the same person, such as if a person experiences conscious
state C1 and then later state C2, which is temporally
downstream from state C1.
Also, I use “person” to refer to any conscious entity, whether
instantiated in human or other biological or even nonbiological form.
A
scientist claims that he is able to upload me to a digital computer. Do we have reason to doubt his claim?
Let’s assume that the scientist’s claim is true and then follow it to its
logical consequences. First, he is
claiming to upload me. If he is
right, then when he uploads me, I will subjectively experience being conscious
and aware yet embodied not by a human body but by a collection of physical objects
(such as digital switches) configured as a general purpose computer executing
software code. Next, because software is
simply a set of instructions that can inherently be copied and executed on any
general purpose computer, what happens if he uploads me to a second
computer? Will I subjectively experience
both? What would happen if those two
experiences, responding to different stimuli (or inputs), start to
diverge? It might seem easy to solve the
problem by adding an ad hoc requirement that only the first software
copy can be executed, but how could Mother Nature possibly enforce this rule if
the two computers are spacelike separated – i.e., causally nonlocal – so that
the temporal ordering of events is relative to an observer? Now, instead of software, the scientist
claims that he is able to create a copy of me in a different spacetime
position. He is not merely claiming that
he is able to make a reasonably good physical copy of my brain, or an exact
copy of my brain, or a reasonably good physical copy of my body, or an exact
copy of my body. Rather, he is claiming
to be able to copy me, identity and all.
The same problems arise as in the case of uploading me to a
computer. I will argue that the demands
of special relativity are inconsistent with the ability to freely copy a
conscious state because the following two assumptions lead to a contradiction:
Copiability of conscious states: A
copy of physical state S1, which is sufficient to create conscious
state C1, can be instantiated elsewhere in spacetime[5] in a manner that does not prevent other
instantiations from evolving.
Note that the second assumption (“Copiability”) depends on the first
assumption (“Supervenience”); Copiability can be false on its own, but if
Supervenience is false, then Copiability must also be false because there is
nothing physical that can be copied to produce the same conscious state of the
same person.
The phrase “physical state” is meant in the broadest sense possible. Classically, the physical state of a system
of matter might be fully described in phase space – i.e., in terms of the
positions and momenta of each of its constituent particles. However, we know that the classical
description of the universe is only an approximation and fails to correctly
describe and predict extremely small systems, the purview of quantum
mechanics. Quantum mechanically,
however, it can’t be said that any particular particle in a system even has a position or momentum until
measured; and while its position can be measured to nearly arbitrary precision,
quantum uncertainty guarantees a trade-off in the precision to which its
momentum can be simultaneously measured.
Further, the quantum mechanical description of a particle can’t be
separated from that of other particles with which it is entangled, making the quantum
mechanical prediction of a physical system larger than a few atoms essentially
impossible. Quantum mechanics may itself
turn out to be an emergent approximation of a more fundamental ontology.
Therefore I do not intend to limit “physical state” to any known
description. Instead, by assuming that conscious
state C1 supervenes on physical state S1, I simply mean
that there is something physical about the universe that gives rise to that
conscious state. Of course, there’s no
guarantee that a given physical state will create a conscious state – probably
very, very few will – but if a conscious state exists then it depends entirely
on the underlying physical state. I also
assume that a given conscious state might be created by more than one physical
state. In other words, it may be the
case that conscious state C1 arises from any member of some large
set {S1, S1*, S1**, ...} of underlying
physical states, and that instantiating any of these physical states will
produce the same conscious state C1 of the same person.
Further, while the following arguments apply to a physical state of any
size, they are more interesting and elucidating when applied to a small or
minimal physical state S1 that produces conscious state C1
(although state S1 need not be the absolute minimum or smallest
state that produces the person’s consciousness). For instance, if it turns out that the human brain[6] is sufficient to create a
person’s consciousness, then certainly that brain plus a body also
creates the person’s consciousness. But
if, as engineers, our goal is simply to create the person’s consciousness as
simply and efficiently as possible, then we may as well focus on just the
brain. So when I assert that conscious
state C1 is created by physical state S1, I don’t
necessarily know what kinds of information or sorts of physical attributes
specify that physical state S1.
I also don’t know how big state S1 is. Is it the local state of certain neural
connections in one’s brain? The state of
one’s entire brain? One’s body? The planet?
The universe? (Although, if it
turns out as a matter of fact that the smallest physical state S1
sufficient to create a conscious state C1 happens to involve
particles throughout the entire universe, then clearly that will prevent state C1
from being copied.)
B. Spacelike Separated Copies
I will now show how the assumptions of Supervenience and Copiability
conflict with the demands of special relativity. With reference to Fig. 1, a physical state S1
on which conscious state C1 supervenes exists at a Point 1 in
spacetime. In other words, a conscious
person having an identity determined by state S1 exists at Point 1 –
let’s call her Alice and designate the person at Point 1 as Alice1. For simplicity, assume that the information
specifying state S1 is read and then the state is destroyed or
reconfigured soon afterward so that state C1 no longer exists.[7] The light cone of Point 1 is shown; events
outside the light cone are called spacelike (or nonlocal) and cannot
have any cause-effect relationship, temporal relationship, or information
connection to the event at Point 1; events inside the light cone are called timelike
and do have a temporal relationship to Point 1.
Fig.
1 Spacelike separated instantiations
of physical state S1
A copy of state S1 is instantiated at Point 2 (shown with its
light cone) and allowed to physically evolve to a physical state S2’
that is sufficient to create conscious state C2’. (For the moment we will ignore what may be
happening at Point 3.) The physical
evolution of state S1 over time is subject to physical laws and
depends on both random quantum events[8] and chaotic amplifications
of initial conditions such that state S1 could evolve over some time
period to any one of a large set of mutually exclusive states, among which S2,
S2’, and S2’’ are possibilities. By Supervenience, a conscious person exists
at Point 2, and that person is Alice – let’s call her Alice2. Alice1 and Alice2 are
the same person; Alice at Point 1 is Alice at Point 2, and she will
subjectively experience the evolution from C1 to C2’
because of a temporal continuity in her identity.[9]
Fig. 1 also shows a Point 3 (with its light cone), with Points 2 and 3
lightlike to Point 1 but spacelike to each other. A copy of state S1 is instantiated
at Point 3 and allowed to physically evolve to physical state S2’’
that is sufficient to create conscious state C2’’, where C2’’
is consciously distinct from C2’ – that is, Alice would be able to
tell them apart. (For the moment we will
ignore what may be happening at Point 2.)
By Supervenience, Alice is created at Point 3 – let’s call her Alice3. Again, Alice3 is the same person
as Alice1 and she will subjective experience the evolution from C1
to C2’’ because of a temporal continuity in her identity.
Now, imagine for the moment in Fig. 1 that copies of state S1
are instantiated at both of Points 2 and 3, but that S2’ = S2’’
= S2 and C2’ = C2’’ = C2. Note that Points 2 and 3 are timelike to Point
1, which is a necessary condition if the information necessary to produce the
copies at Points 2 and 3 depends on information collected at Point 1. Further, Points 2 and 3 are spacelike to each
other so that there is no fact about simultaneity or temporal ordering between
them, nor can information pass from one to the other.[10] We already know that if state S1
was not instantiated at Point 2, then Alice exists at Point 3 and she
will subjectively experience the evolution from C1 to C2
because of a temporal continuity in her identity. However, even when S1 is
instantiated at Point 2, Alice exists at both Points 2 and 3, she has
precisely the same subjective experience at both points, and she will
subjectively experience the evolution from C1 to C2. This must be true because we already know that
Alice exists in conscious state C1 at Point 2. But the creation of physical state S1
at Point 3 is spacelike to creation of physical state S1 at Point 2,
which means that from the perspective of Point 3, there is no fact about the
creation of state S1 at Point 2.
The creation of S1 at Point 2 can have no physical effect at Point
3. For the same reason that Alice exists
and experiences an evolution from C1 to C2 due to the
creation of state S1 at Point 2, she exists and experiences an
evolution from C1 to C2 due to the creation of state S1
at Point 3. This may seem philosophically
odd, but (so far) there is no physical contradiction.[11] We should be careful, however, not to imagine
different Alices who happen to have very similar experiences; Alice1,
Alice2, and Alice3 are all the same Alice. Supervenience requires that Alice1
experiences the evolution of C1 to C2 due to physical
evolutions of instantiations of physical state S1 at both Points 2
and 3.
Finally, let’s return to the original
configuration of Fig. 1, in which S2’ ≠ S2’’ and C2’
≠ C2’’ – i.e., that they are consciously distinct states.[12] In
that case, both Alice2 and Alice3 exist, but the underlying
physical evolutions of the two physical states S1 diverge. (It makes no difference which of the two
systems diverges as the divergence is relative to the other.) For the sake of clarity, assume also that the
entirety of the evolutions of states S1 (at Point 2) to S2’
and S1 (at Point 3) to S2’’ are spacelike to each other. Given that Alice exists at both Points 2 and
3, what does Alice actually experience in Fig. 1? At both Points 2 and 3 she experiences state C1
arising from state S1, but then does she subjectively experience an
evolution to C2’ or C2’’?
There are exactly three logical possibilities[13]: she experiences neither, both, or one or
the other.
Is it possible that Alice
experiences neither? Remember that Alice
is created independently by instantiations at both Points 2 and 3; the
instantiation evolving from Point 3 cannot be affected by the instantiation
evolving from Point 2. If Alice is created
by the instantiation at Point 3 and experiences an evolution from C1
to C2’’, then nothing that happens in the evolution of state S1
at Point 2 to S2’ can affect Alice’s experience from C1
to C2’’. Therefore, she
cannot experience neither.
Perhaps Alice experiences both? What would it feel like for Alice to
subjectively experience consciously distinct evolutions from C1 to C2’
and C1 to C2’’?
Or, more colloquially, what would it feel like for her to experience
diverging streams of consciousness? Before
we indulge in speculations about higher planes of consciousness or fissioning
identities, we should note the problem with locality. Note that the experience of Alice2
in Fig. 1 depends on the underlying physical evolution of S1 to S2’,
while the experience of Alice3 depends on the underlying physical
evolution of S1 to S2’’.
Alice3 cannot experience an evolution of conscious states
that depends on a physical evolution of S1 to S2’,
because S2’ is spacelike to S2’’, just as Alice2
cannot experience an evolution of conscious states that depends on a physical
evolution of S1 to S2’’.
Therefore, Alice cannot experience an evolution of conscious state from C1
to both C2’ and C2’’.
Finally, is it possible that Alice
experiences one or the other? The
problem here is that because the underlying physical systems are spacelike
separated, there is no means or criteria by which nature can independently
select one instantiation or the other for the continuity of Alice’s identity. For instance, imagine that Mother Nature
simply required that the “first” instantiation of state S1 would assume
Alice’s identity; that won’t work because Points 2 and 3 are spacelike to each
other so there is no observer-independent fact about which happens first. Perhaps Mother Nature could choose the point
having the shortest spacetime distance to Point 1, which in Fig. 1 would be Point
2. However, again, the spacelike
separation of the points prevents the passage of any signal from Point 2 to Point
3 to “stop” the creation of Alice’s conscious state C1 at Point 3.
Therefore, having ruled out the
only three possibilities, the copiability of conscious states to spacelike
locations in spacetime leads to a contradiction. One might object that the above argument
depends on the instantiation and evolution of spacelike-separated copies of state
S1; perhaps there is no problem when the copies are instantiated locally
to each other. This objection
fails. If there is something
fundamentally impossible about the spacelike instantiation of copies, then it
must be found in their instantiation, not in their relative locations or the
extent to which they are separated. Further,
the following section will show that timelike instantiations of copies of state
S1 are equally problematic.
C. Timelike Separated Copies
The spacelike instantiations in Section B are problematic because of a
lack of causal connection. I will now show
that a similar contradiction arises when we consider two timelike
instantiations of state S1 because they would require impermissible backward-in-time
causation.
Fig. 2 is comparable to Fig. 1, the primary difference being that Point 3
is timelike to (and occurs after) Point 2, which is timelike to (and occurs
after) Point 1. For the same reasons, if
we ignore what may be happening at Point 3, then Alice2, who is the
same person as Alice1 at Point 1, exists at Point 2 and will
subjectively experience the evolution from C1 to C2’. And if we ignore what may be happening at Point
2, then Alice3, who is the same person as Alice1 at Point
1, exists at Point 3 and will subjectively experience the evolution from C1
to C2’’.
Fig.
2 Timelike separated instantiations
of physical state S1
Now, imagine for the moment in Fig. 2 that copies of physical state S1
are instantiated at both of Points 2 and 3, but that S2’ = S2’’
= S2 and C2’ = C2’’ = C2. By the assumption of Supervenience, Alice
exists and has precisely the same subjective experience at both Points 2
and 3 and will subjectively experience an evolution from C1 to C2. To the extent that any conscious state, when
it is experienced, is subjectively experienced as occurring “now,” states C1
at Points 2 and 3 must be experienced identically and thus must be subjectively
experienced simultaneously. So, even
though Point 3 objectively occurs after Point 2, Alice’s conscious experience
(which includes the subjective experience of “now”) is the same at both points. After all, if they were subjectively
experienced as different times, such different experiences would manifest
themselves in different conscious states.
While this may seem philosophically odd, again there is no
contradiction.[14]
Unlike in Section B in which the copies of state S1 are
spacelike separated, Alice3 is in fact created after Alice2,
in which case a causal connection is possible in principle. However, state S1 is already
assumed (by Supervenience) to be sufficient to create Alice in conscious state C1,
so the instantiation of state S1 at Point 3 must necessarily create
Alice in state C1, independently of her creation at Point 2 (or any
other event prior to Point 3). In other
words, the assumptions of Supervenience and Copiability prevent the kind of
causal connection that would invalidate these arguments. For example, consider the case in which creation
of Alice2 causally prevents the creation of Alice3; if
that were true, then creation of conscious state C1 at Point 3 would
depend on facts that transcend physical state S1, thus contradicting
Supervenience. Therefore, if state S1
is sufficient to create Alice in state C1, then instantiation of S1
at Point 3, and its corresponding creation of Alice3, cannot be altered
by the existence of Alice2. For
the same reason that Alice exists and experiences an evolution from C1
to C2 due to the creation of state S1 at Point 2, she
also exists and experiences an evolution from C1 to C2
due to the creation of state S1 at Point 3. Again, we should be careful not to imagine
different Alices who happen to have very similar experiences; it is the same
Alice.
Finally, let’s return to the original configuration of Fig. 2, in which S2’
≠ S2’’ and C2’ ≠ C2’’ (i.e., they are
consciously distinct states). In that
case, both Alice2 and Alice3 exist, but the underlying
physical evolutions of the two states S1 diverge. Given that Alice exists at both Points 2 and
3, what does Alice actually experience in Fig. 2? At both Points 2 and 3 she experiences state C1
arising from state S1, but then does she subjectively experience an
evolution to C2’ or C2’’?
Like in the case of spacelike separated copies of state S1,
there are exactly three logical possibilities: she experiences neither, both,
or one or the other.
Regarding neither, I have explained why instantiation of state S1
at Point 3, and its evolution to state S2’’, will produce Alice
subjectively experiencing state C1 to C2’’ independently
of instantiation of state S1 at Point 2. Therefore, it cannot be neither.
Regarding both, the problem here is that Alice experiences the evolutions
of C1 to C2’ and C1 to C2’’ as
simultaneous (from her subjective perspective), because they both begin at the
same conscious state C1. Therefore,
for Alice to experience both conscious evolutions, it must be the case that the
conscious experience of Alice2 depends not only on the underlying
physical evolution of S1 (at Point 2) to S2’, but also on
the underlying physical evolution of S1 (at Point 3) to S2’’. However, if Alice is to experience conscious
states in real time with their underlying physical states, this scenario would
require impermissible backward-in-time causation, because what she experiences
prior to Point 3 would depend at least in part on events occurring after
Point 3. Further, because of physical
indeterminism (e.g., chaos and quantum mechanical randomness), it cannot be
known at Point 2 whether, or when, additional copies of state S1
will be instantiated. Therefore, it
would be impossible for Alice to experience both an evolution of C1
to C2’ and C1 to C2’’ if she experiences them
in real time with the underlying physical evolutions.
The only possible way for Alice to experience both evolutions would
require two facts: first, that Alice’s conscious states are experienced after
(i.e., not in real time with) the physical states that produced them; and
second, that Alice could not experience any conscious states until the
instantiation of additional copies of state S1 is no longer
possible. In other words, Alice would
only be able to experience conscious state C1 created by multiple
copies of state S1 after the final copy of state S1
was instantiated. Thus, Alice can only
subjectively experience consciousness after which time her conscious states can
no longer be copied or repeated – in which case Copiability is false.
Finally, regarding whether Alice will experience one evolution or the
other, I have explained why instantiation of state S1 at Point 3,
and its evolution to state S2’’, will produce Alice3
subjectively experiencing state C1 to C2’’. Therefore, if Alice is to experience
conscious states in real time with their underlying physical states, whether or
not Alice2 will experience the evolution from C1 to C2’
depends on the future fact of whether Alice3 is created by
instantiation of state S1 at Point 3, another impermissible
backward-in-time causation. Thus, for
Alice to experience one evolution or the other, not only would Alice’s
conscious states necessarily be experienced long after the physical states that
produced them have long since disappeared, but Alice could not experience any
conscious states until the instantiation of additional copies of state S1
is no longer possible. Thus, Alice can
only subjectively experience consciousness on or after which her conscious
states can no longer be copied or repeated – in which case Copiability is false.
Therefore, having ruled out the only three possibilities, the copiability
of conscious states to timelike locations in spacetime leads to a
contradiction. In conjunction with the
conclusion of Section B, it follows that Copiability is false.
IV. DISCUSSION
The assumptions of Supervenience and Copyability lead to a contradiction,
whether copies of a conscious state are created spacelike or timelike. Therefore, Copyability, which depends on
Supervenience, is false. Consequently, a
conscious state cannot be copied or repeated (unless something about the
universe prevents the copies from evolving differently); said differently, if
two instances of the same conscious state could evolve to different conscious
states, they cannot both exist. This
conclusion does not speak to the veracity of the statement, “Consciousness is
created by the brain,” but if that statement is true, then there must be
something physical about the brain that prevents it from being copied. Why should this be the case? And what are some potential implications?
A. Explanatory Hypotheses
Why might it be fundamentally physically impossible to copy or repeat
conscious states? It may be the case
that there is no way to adequately measure physical state S1, or if
there is, that there is no way to recreate it elsewhere in spacetime. Some potential reasons include: the
information necessary to specify the physical state is contained in distant particles
that cannot be reached; attempting to measure the information inherently
destroys it, for example, because of reduction of a quantum wave function; the
information cannot be measured with adequate precision due, for example, to quantum
uncertainty; the state includes quantum entanglements with distant particles
that cannot be recreated; and/or the state is too large to measure or
instantiate elsewhere.
The assumption of Supervenience does not depend on physical state S1
being a quantum state, nor do the arguments in this paper depend on any
relationship between consciousness and quantum mechanics. Having said that, many of the explanatory
hypotheses for the physical inability to duplicate conscious states invoke
quantum mechanics. This should not be
entirely surprising; the only known physical mechanism that prevents the
instantiation of multiple copies of the same entity is quantum no-cloning, a
no-go theorem that asserts the impossibility of creating an identical copy of
an unknown quantum state. It may well be
the case that whatever physical state S1 is required to create
conscious state C1 depends on quantum information that prevents it
from being copied. Aaronson [11], for
example, points out that if a conscious chunk of matter is “unclonable for
fundamental physical reasons,” then that unclonability could be a consequence
of quantum no-cloning if the granularity a brain would need to be simulated at
in order to duplicate someone’s subjective identity was down to the quantum
level.
Further, the non-repeatability of a conscious state implies
irreversibility, and the concept of irreversibility pervades physics:
classically in the form of the Second Law of Thermodynamics and increasing
entropy[15], and quantum mechanically
in the form of decoherence [12], which occurs “as soon as a single [correlated]
quantum is lost to the environment.”
Since there will always be those who claim that any quantum mechanical
process is reversible in principle by acting on the system with the reverse
Hamiltonian, we could limit ourselves to “any fact of which the news is already
propagating outward at the speed of light, so that the information can never,
even in principle, be gathered together again in order to ‘uncause’ the fact” [11]. If we indeed live in a universe with
nonnegative curvature, as is currently believed, and given that almost the
entire night sky is black, we can regard virtually every photon emitted into
space as irretrievable. Thus if it turns
out that consciousness can only result from irreversible processes, that
irreversibility may itself be manifested in a physical state that cannot be read
or copied because some of the information specifying it is embedded in photons
streaming through space at the speed of light.
B. Implications
First, of course, the conclusion that conscious states cannot be copied
or repeated instantly renders moot the philosophical problems mentioned in the Introduction. Two others will be broached here.
Consciousness is not algorithmic.
The possibilities of mind uploading and computer consciousness
depend on whether consciousness is fundamentally algorithmic – that is, whether
consciousness can be reduced to a finite set of input-dependent
instructions. Because an algorithm can
be executed in the form of software on any general-purpose computer
independently of its underlying physical substrate, any algorithm can be copied
with no fundamental limitation on the number of copies that can be executed; it
can also be repeated on the same computer by resetting that computer and then
executing the software again. If a
conscious state cannot be copied or repeated, but an algorithm can, then an
algorithm cannot produce a conscious state.
That inevitably leads to the conclusions that Strong Artificial
Intelligence (“Strong AI”) is likely false (cf. [13]) and that mind uploading
and computer consciousness are likely impossible.
Wigner’s Friend is not conscious.
In Eq. 1, we were told that the quantum mechanical system
is initially in a superposition of states
and
; when the observer performs a thousand experiments,
he finds that around 99% of them are measured in state
and the
remaining in state
. However,
what if the observer had been fooled?
What if instead of a thousand systems in a pure state, he was actually
provided with 990 systems already in state
and 10
already in state
– i.e., a
mixed state? Is there a way for him to
modify the experiment to figure this out?
Yes: by making measurements in a new basis in which one element,
is parallel to
and the other
element,
is perpendicular.
If each system was originally in a pure state, then all 1000 will, with
certainty, be measured in state
; if the systems were mixed, then some are likely to
be measured in state
.[16] While the principle is relatively easy to
explain, the actual process of measuring in the new basis – called an
interference experiment – is prohibitively complex in all but the simplest situations.
In Eq. 2, let’s rename the observer either Schrodinger’s Cat (“SC”) or
Wigner’s Friend (“WF”); both refer to essentially identical thought experiments
in which an isolated macroscopic system involving some quantum mechanical event
is allowed to evolve until finally observed by a Super-observer (Schrodinger or
Wigner, respectively). When Wigner
finally confers with his friend post-measurement, WF of course reports having
measured either outcome A or B. But did
WF collapse the wave function when he performed the measurement? Or did WF remain in the superposition shown
in Eq. 2 until Wigner conferred with his friend, in which case Wigner’s
observation of WF collapsed the wave function?
This is the same question as whether WF was in a pure or mixed state at
the time of Wigner’s observation, and the only way for Wigner to know is by
repeatedly doing an interference experiment on WF in a basis parallel to the
state in Eq. 2. But the only way to do that
would be to repeatedly measure WF in a coherent superposition of states, which
requires that WF’s state (specifying his body, brain, memories, etc.) is itself
reversible. Such manipulations, if they
were possible, would require erasing WF’s memory and further require WF to experience
“the same mental processes over and over, forwards in time as well as backwards
in time” [11]. Deutsch describes in [15]
a hypothetical method for experimenting on WF by having him send a message to
Wigner that is uncorrelated to the measurement outcome[17], but concedes that such
an experiment requires memory erasure. While
many scholars have asserted the “in-principle” possibility of such an
experiment[18],
Deutsch attempts to overcome otherwise insurmountable technical hurdles by
positing conscious computers.
However, if a conscious state cannot be repeated, as I showed in Section
III, then this kind of manipulation on a conscious observer is not
possible. In other words, if the WF (or
SC) experiment can actually be performed, then WF is not conscious; if he is
conscious, then the experiment cannot be performed. Further, if consciousness is not algorithmic,
then Deutsch’s attempt to salvage his experiment by substituting a computer for
a person fails. Because the ability to
measure WF in a coherent superposition requires reversibility, and because a
conscious state is not reversible because it cannot be repeated, it follows
that if WF’s conscious measurement is related to an irreversibility in the
measurement process. A conscious state
can only arise after reversal of the state is no longer possible – that is,
when the state is truly irreversible.
Therefore, consciousness may be a direct result, and indication, of the
irreversibility of a physical process.[19]
C. Concluding Remarks
Conscious
states are produced by the brain; the brain is a just a chuck of matter; it is
copiable in principle; therefore, copying a conscious state is merely a
technological problem.
The allure of this logic is irresistible and has led physicists and
philosophers into a minefield of intractable problems and seeming paradoxes. More careful scrutiny of the assumptions,
particularly in light of physical limitations imposed by special relativity on
copies of conscious states instantiated at different points in spacetime,
yields a contradiction: there is something about consciousness that prevents it
from being copied or repeated at will.
REFERENCES
[1]
Penrose, R., 1989. The
Emperor’s New Mind: Concerning Computers, Minds, and the Law of Physics. Oxford
University Press.
[2]
Bostrom,
N., 2003. Are we living in a computer simulation?. The Philosophical
Quarterly, 53(211), pp.243-255.
[3]
Elga,
A., 2004. Defeating Dr. Evil with self‐locating belief. Philosophy and
Phenomenological Research, 69(2), pp.383-396.
[4]
Carroll,
S.M., 2017. Why Boltzmann brains are bad. arXiv preprint
arXiv:1702.00850
[5]
Parfit,
D., 1971. Personal identity. The Philosophical Review, 80(1),
pp.3-27.
[6]
Zuboff,
A., 1990. One self: The logic of experience. Inquiry, 33(1),
pp.39-68.
[7]
Wallace,
D., 2012. The Emergent Multiverse: Quantum Theory According to the
Everett Interpretation. Oxford University Press.
[8]
Lewis,
D., 2004. How many lives has Schrödinger's cat?. Australasian Journal
of Philosophy, 82(1), pp.3-22.
[9]
Tappenden,
P., 2011. A metaphysics for semantic internalism. Metaphysica, 12(2),
p.125.
[10] Saunders, S., 1998. Time, quantum
mechanics, and probability. Synthese, 114(3),
pp.373-404.
[11] Aaronson, S., 2016. The Ghost in the Quantum Turing
Machine. In: The Once and Future Turing: Computing the World.
Cambridge University Press.
[12]
Haroche, S., 1998. Entanglement, decoherence and the
quantum/classical boundary. Physics today, 51(7),
pp.36-42.
[13] Searle, J.R., 1990, November. Is the
brain a digital computer?. In Proceedings and addresses of the American
Philosophical Association (Vol. 64, No. 3, pp. 21-37). American
Philosophical Association.
[14] Okón, E. and Sebastián, M.A., 2016. How
to back up or refute quantum theories of consciousness. Mind and Matter, 14(1),
pp.25-49.
[15] Deutsch, D., 1985. Quantum theory as a
universal physical theory. International Journal of Theoretical Physics, 24(1),
pp.1-41.
[16] Salom, I., 2018. To the rescue of
Copenhagen interpretation. arXiv preprint arXiv:1809.01746.
[17] de Barros, J.A. and Oas, G., 2017. Can
we falsify the consciousness-causes-collapse hypothesis in quantum
mechanics?. Foundations of Physics, 47(10),
pp.1294-1308.
[1] Throughout this paper,
I’ll treat copiability and repeatability of conscious states as meaning the
same because repeating a conscious state is akin to copying it at a later point
in spacetime.
[2] This temporal flow is
often called a “stream” of consciousness, but many have argued that
consciousness consists of a series of discontinuous blips of awareness that are
perceived as continuous. This
distinction is irrelevant to the present analysis.
[3] Just as there is no
consensus on the definition of consciousness, there is no single term or phrase
to represent it –others are mentality, sentience, self-awareness, and the
subjective experience of qualia.
[4] I have no idea what this
phrase means, nor does any other physicist, although many philosophers throw it
around casually. An exact copy in
classical phase space is impossible because infinite precision does not exist
in the physical world; an exact copy in quantum mechanical Hilbert space is
impossible because of quantum no-cloning; and so on. But for the sake of this section, let’s pretend
there is such a thing.
[5] If state S1 depends on spatiotemporal
location – i.e., if that state cannot be instantiated elsewhere in spacetime –
then the assumption of Copiability is obviously false. To use the philosopher’s lexicon, two
physical systems are not “numerically identical” if they exist at different
locations in spacetime, no matter how physically similar they may be. So if one already accepts that numerical
identity is necessary for the same person to experience the same conscious
state, then it follows that conscious states cannot be copied or repeated.
[6] Specifying physical state
S1 as “the human brain” is quite sloppy, as we really need to
specify the features (e.g., relationships among cells, molecules, electrons,
etc.) that define state S1 and give rise to conscious state C1. If, for example, state S1 depends
only on how cells act like digital switches, then S1 can be
specified and instantiated in the form of a digital computer. If, however, S1 depends on quantum
effects, then quantum no-cloning may prevent specification of S1 at
all.
[7] This is not a requirement
but does simplify the analysis. If it
turns out that the mere existence of the information necessary to produce state
S1 is itself adequate to produce state C1, then it is
impossible to destroy S1 without also preventing any future
instantiations of C1.
[8] Whether or not quantum
mechanics has anything to do with consciousness, it is clear that conscious
states can correlate to (amplified) quantum events, such as the “click” a
scientist consciously notices when using a Geiger counter to measure decays of
radioisotopes.
[9] To reiterate, this must
be true if Supervenience is true, and if Supervenience is not true, then
conscious states cannot be copied because there is no underlying copiable
physical state that preserves the person’s identity, and we can skip to the end
of the paper.
[10] While Point 2 appears
closer in time to Point 1, and thus earlier than Point 3, in fact the interval
in spacetime is the metric
. Which of
Points 2 and 3 occurs earlier in time is actually relative to an
observer.
[11] Tappenden explains [9]: “How
can two doppelgangers zillions of lightyears apart whose simultaneity we know,
from Special Relativity, is entirely relative to an inertial frame, how can
they share a single mind?” The answer: no
need for causal connection.
[12] We can ensure that the
states are distinct by designing the systems to correlate to amplifications of
cosmic microwave background radiation originating from different directions
and/or points in spacetime.
[13] These three options
perfectly mirror the options suggested by Saunders [10]: “Nothing, both, or
else just one of them?” He concludes
that only the third alternative is possible, but does not consider the effects
of nonlocality.
[14] Zuboff explains [6]:
“This experience [across brains] of being you, here, now, would be numerically the same whenever, as well as wherever, it was realized.”
[15] Strictly speaking,
classical physics is time-reversible; increasing entropy is entirely a
statistical result and is not an immutable law of nature.
[16] This is because the
systems starting in state
are in a
superposition of
and
with nonzero
amplitudes, so when each such system is measured in the
basis, some
will be measured as
and some as
; and vice versa for those systems starting in state
. However,
because there is some nonzero probability that all 990 initially
states and
all 10 initially
states are,
by random quantum “accident,” measured in state
, which is identical to the outcome predicted for
systems in a pure state, it actually requires infinitely many interference
experiments to confirm a pure state (cf. [14]).
[17] But see [16], which
points out why a post-measurement message from WF will always be correlated to
the outcome.
[18] But see [17], which
argues that the thermal isolation required for such reversibility precludes
performing the experiment on any living thing we might regard as conscious.
[19] Aaronson comes to a
similar conclusion [11] and notes that, if correct, it may have interesting
cosmological implications, such as the universe having nonnegative curvature.
No comments:
Post a Comment
All comments are moderated. After submitting your comment, please give me 24 hours to approve. Thanks!