You are on page 1of 13

No-Go Theorems on Machine Consciousness

Subhash Kak
Chapman University, Orange, CA 92866 &
Oklahoma State University, Stillwater, OK 74078
Email: subhash.kak@okstate.edu

Abstract.
Creating machines that are conscious is a long term objective of research in
artificial intelligence. This paper look at this idea with new arguments from physics
and logic. Observers have no place in classical physics, and although they play a
role in measurement in quantum physics there is no explanation for their emergence
within the framework. There is suggestion that consciousness, which is implicitly
a property of the observer, is a consequence of the complexity of specific brain
structures, but this is problematic because one associates free will with
consciousness, which goes counter to causal closure of physics. Considering a
nested physical system, we argue that even if the system were assumed to have
agency, observers cannot exist within it. Since complex systems can be viewed in
nested hierarchies, this constitutes a proof against consciousness as a product of
complexity, for then we will have nested system of conscious agents. As the
existence of consciousness in cognitive agents cannot be denied, the implication is
that consciousness belongs to a dimension that is not physical and machine
consciousness is unattainable. These ideas are used to take a fresh look at two well-
known paradoxes of quantum theory that are important in quantum information
theory.

Keywords: AI, machine consciousness, observers, Schrödinger’s cat paradox,


Wigner’s friend paradox

Introduction
Machine consciousness is a goal of artificial intelligence and it is being investigated
from a variety of perspectives that include algorithms [1][2], neuroscience [3], and
physics [4]. If we focus on the presumed physical bases of consciousness, we must
square up with is a commonsensical premise of science that the physical world
(which includes the biological world) is causally closed. This closure implies that
all physical events must be due to physical causes leading to the question as to the

1
source of the freedom that goes beyond chance and necessity that cognitive agents
appear to possess [5][6][7]. If, on the other hand, we live in a mathematical universe
and observers are zombies as has been suggested [8], one needs to explain the
apparent influence of observers on physical systems. Observers and agents deal
with information [9][10] and our knowledge of the universe exists within
consciousness; therefore, comprehending the emergence of observers is also
important for understanding the nature of physical information [11]. Other related
questions are whether agency could be achieved in artificial intelligence machines,
and if consciousness is computable [12].

A physical system must evolve according to natural law, and this must be true even
if it includes observers, leaving no room for agents that are free to observe other
components of the system. Since human agents are certain that they have freedom,
which they use in their interactions with the environment, this leads to questions
about the limits to machine consciousness [13][14].

In classical physics, the observer is always outside of the formalism that describes
the world and, therefore, classical physics doesn’t directly consider the observer
excepting as a physical system. In quantum physics, there exists the difficulty of
reconciling the unitary and deterministic evolution of an isolated system with the
non-unitary and probabilistic state update upon a measurement, which is called the
collapse of the wavefunction. The unobserved physical world is governed by the
deterministic Schrödinger equation, and probabilities come in only upon interaction
with the system. The contradiction underlying this situation has led some to ask the
question if quantum theory is valid at all scales, including that of observers?
[15][16]

It has been argued that if quantum theory had universal applicability it should be
able to model complex systems that include agents who use quantum theory for
their measurements. But a recent paper [17] presents a scenario with inconsistent
results in which one agent, upon observing a particular measurement outcome, must
conclude that another agent has predicted the opposite outcome with certainty. This
scenario indicates the insufficiency of quantum theory to model a universe with
observers.

The question of agency of observers could also be explored from the perspective of
structures related to forms and behavior. Self-similar structures are associated with
life and implicitly with consciousness and therefore some new ideas related to

2
optimum information in noninteger dimensional space may be relevant
[18][19][20]. Note further that conscious observers are not like computers that run
on programs, but rather systems that adapt to the environment [21]. If this ability
to adapt is a characteristic of consciousness, then it is another reason why
programmed artificial intelligence machines (which work according to the Turing
computation model [22]) will never be conscious.

Awareness or consciousness underlies the ability of observers to interact with the


environment in a unique way, and its emergence from complexity is seen to be
similar to the emergence of chemistry from physics, and biology from chemistry
despite there being no evidence in support of this view.

Recently ideas have been advanced that unique processing in specialized brain
structures creates consciousness. In the recurrent processing theory (RPT) it is
proposed that the first sweep of visual feedforward processing is unconscious, and
consciousness emerges when recurrent, top-down processing interacts with neurons
activated during the initial feedforward sweep [23]. In information integration
theory (IIT), it is suggested that an information integration measure called ϕ, which
is computed based on the causal structure of a system, quantifies consciousness
[24]. Such theories try to provide structural bases to the complexity-leads-to-
consciousness argument but a recent study [25] argued that such models are “either
false or outside the realm of science”. These models are easily refuted because they
are equivalent to mathematical algorithms producing consciousness since structural
complexity may be mapped into a mathematical framework (hardware is
functionally equivalent to software). If mathematics is seen as collection of diverse
properties of objects in sets, this cannot by itself lead to consciousness.

In the present paper, we focus on structure of observers in a narrow setting, namely


that of agents that are subsumed within another agent. We consider both classical
computational as well as quantum mechanical frameworks for our arguments. We
show that using commonsense definition of agency, the emergence of
consciousness is impossible in both cases. The implication is that awareness or
consciousness cannot be physical. We also speculate on how minds and their
contents that are definitely physical are nevertheless associated with consciousness.

The Main Result


Reiterating that the physical world is causally closed, consider a physical system S
where a part of it represents a cognitive agent, C. The interconnection of C to the

3
rest of S represents the physical pathway through which C makes its observations
(Figure 1).

Since the evolution of S is completely governed by physical laws, the states of C


as well as S are determined by the previous states of S and any environmental
inputs.

The observation made by C is then a representation of its co-evolved state.

Figure 1. System S including cognitive agent C

Normally, absent any external input and absent any internal observation by C, the
state of SC -- that is System S including C -- at time 𝑛 + 1 is dependent on the
internal state of SC at time n.

𝑆𝑡𝑎𝑡𝑒 (𝑆𝐶(𝑛 + 1)) = 𝑓1 (𝑆𝑡𝑎𝑡𝑒 (𝑆𝐶(𝑛))) (1)

But if C makes an observation at time 𝑛, then the state of SC at time 𝑛 + 1 is


different:

𝑆𝑡𝑎𝑡𝑒 (𝑆𝐶(𝑛 + 1)) = 𝑓2 (𝑆𝑡𝑎𝑡𝑒 (𝑆𝐶(𝑛), 𝑂𝑏𝑠𝑒𝑟𝑣𝑎𝑡𝑖𝑜𝑛 𝑏𝑦 𝐶(𝑛))


(2)

This means that the future states of the overall system are not completely
determined by the initial state and the dynamics, which contradicts the assumption
of causal closure.

Classical physics does not permit freedom at any level, and so observers cannot
find a place within the framework. The assumption that C possesses agency not
only contradicts causal closure but also, implicitly, the assumption that the mind
(comprising of processes in the brain) is completely a part of this physical world.

4
Some propose that the interface of consciousness and the physical world could be
through a quantum interface, but the overall system will be inconsistent for reasons
that are similar to the one sketched above (e.g. see [6][13][14][15]).

In a quantum description, C may be viewed as being apart from the rest of S by


means of a Heisenberg cut of quantum mechanics which is the hypothetical
interface between quantum events and the observer’s information or conscious
awareness. Below the cut everything is governed by the wavefunction; above the
cut a classical description is used.

The cut may be seen as separating the quantum from the classical. The micro-world
is quantum but the macro-world, the world of commonsense experience
apprehended by the mind, is classical.

C 𝑪𝟏
C 𝑪𝟐
𝑪𝒏

Figure 2. Subsystems of C in an infinite regress

The C of Figure 1 was an arbitrary mapping of a physical part of the system to the
cognitive center, just as for the human one tends to place consciousness within the
brain. While that may be fine as a first cut, the cut separating the consciousness-
generating structures from the rest of the body may be seen a sequence of
continually changing subject-object boundaries with successively smaller subject
space if viewed in structural terms. It may also be viewed as a time process that
goes over or dances across different physical elements.

This method of regression is sometimes represented as in Figure 3, which has the


virtue of representing different modules in the brain, each with its own domain that
is shown as a triangle.

5
Figure 3. Another representation of C in regress

Consciousness in the view of Figure 3, is a category that keeps on receding into


ever smaller physical structures as the brain is probed; it is shown as the supposedly
infinitesimal dot in the center of the diagram. Clearly such a description is not open
to causal analysis.

Theorem 1. Nested observers cannot exist.


Proof. Since observation requires cognition and choice concerning the observation,
observers that are nested within one single structure cannot exist. If C of Figure 1
is the observer, a structure superior to C cannot simultaneously be an observer for
that will interfere with its agency. Since there is only one physical universe, there
can only be one observer. If there existed a lower-level conscious observer, it will
negate the causal closure at the higher level.

The impossibility of nested observers will be true even if quantum theory is


replaced by a future high level theory for free will of the observer cannot be part of
a physical system governed by laws.

By considering an ever larger system that subsumes all the previous ones,
theoretically a single observer may be allowed in the entire universe.

Theorem 2. There can only be one observer.


Proof. Since the whole universe may be viewed as a single system, there can at best
be one observer (conscious agent) associated with it if we wish to avoid
contradiction related to causal closure.

6
But how to reconcile this One Observer Theorem with the fact of many observers,
i.e. conscious agents, in reality?

The reconciliation may be done by assuming that consciousness is a category


different from space and time and it interacts with the physical world by altering
probabilities as has been shown to work in the Quantum Zeno Effect [26].

Theorem 3. Consciousness associated with cognitive agents cannot be emergent on


the complexity based hardware.
Proof. Awareness or consciousness is not a physical process for then one can map
it into a mathematical algorithm, which cannot generate a new phenomenon.

Consciousness can influence the physical world by changing the probabilities


associated with the evolution without altering the dynamics as in the Quantum Zeno
Effect ([26]; and see also [27]). It can be said that there is one consciousness and
many minds.

Schrödinger’s Cat and Wigner’s Friend


We apply the ideas of the previous section to the consideration of information and
agents in quantum theory which is of relevance in quantum computing models.

The difficulties of considering agents within a quantum computational framework


are highlighted by the Schrödinger’s Cat Paradox that is concerned with the
superposition of states of macroscopic objects. Schrödinger was trying to point to
the fundamental difficulty of generalizing the quantum mechanics framework to
large systems by proposing that one could, in principle, create a superposition in a
macroscopic system by making it dependent on the state of a quantum particle [28].

The Cat Paradox formulation


In Schrödinger's original formulation, a cat, a flask of poison, and a radioactive
source (atom) in the excited state |0⟩𝑎 are placed in a sealed box. If an internal
monitor detects radioactivity as a consequence of its decay to its ground state |1⟩𝑎 ,
the flask is shattered, which releases the poison, killing the cat. Considering “live”
and “dead” as states associated with the cat, the superposition of the states implies
that the cat is simultaneously alive and dead. Yet, when one opens in the box, the
cat is either alive or dead, not both alive and dead. The question may be asked as to

7
when exactly quantum superposition ends and reality resolves into a definite
classical state.

The textbook description of the experiment [29] describes the state of the atom–
cat system when it is examined to be:

|𝜓⟩𝑓 = 𝑎|0⟩𝑎 |𝑎𝑙𝑖𝑣𝑒⟩𝑐 + 𝑏|1⟩𝑎 |𝑑𝑒𝑎𝑑⟩𝑐 (3)

Usually, one takes 𝑎 = 𝑏 = 1/√2 for maximum entanglement. The equation (3) is
the case of the cat being simultaneously dead and alive before the measurement is
made.

In reality there is no paradox for the following reasons:

1. The initial state is |0⟩𝑎 |𝑎𝑙𝑖𝑣𝑒⟩𝑐 is fully separable and the cat is not
entangled with the atom.
2. The cat state |𝑎𝑙𝑖𝑣𝑒⟩𝑐 is a classical state for it cannot be put into a
superposition with |𝑑𝑒𝑎𝑑⟩𝑐 based on how one interacts with it.
3. One can assume that the atom evolves freely to 𝑎|0⟩𝑎 + 𝑏|1⟩𝑎 .
4. The cat is dead if the atom had decayed to |1⟩𝑎 ; otherwise the cat is alive.
The state of the box is a statistical mixture, and if the atom had decayed
and the cat is dead the approximate instant of the death may be inferred
from post-mortem changes.

A variable is correctly considered to be a quantum if its superpositional basis is


revealed upon a suitable interaction of the observer with it. The commonsensical
meaning of the term “alive” is that no matter how the cat is observed, it continues
to remain in that state (therefore it is classical), which is not the case with, say, the
polarization state of a photon or the spin of an electron. Thus a horizontally
polarized photon can be in a superposition with a vertically polarized photon, or be
a vertically polarized photon if the observer approaches it in different orientations.
The “alive” state of the cat can never be in superposition with the “dead” state and
this means that extending the idea of a quantum superposition to such a
macroscopic system is incorrect.

Wigner’s Friend Paradox


Eugene Wigner presented a variation on the Cat Paradox in a thought experiment
where the cat is replaced by another human that is the observer’s friend [30]. Thus
we have two agents, namely Wigner (W) and his friend (F), and underlying this so-

8
called Wigner’s Friend Paradox is the assumption that quantum mechanics is valid
at all scales.

In this experiment, F, sealed in a lab, measures a quantum state, such as one


associated with a particle, while W stands outside. The quantum state of the overall
system is a superposition of several locations for the particle, but F’s observation
“collapses” the particle to just one specific spot. For W, the superposition remains
until he makes a measurement sometime later. The observations of W and F are in
conflict indicating that the idea of superposition does not work for large objects.
Several variations of Wigner’s Paradox have been proposed and if common
assumptions about reality that are implicit in the way we use language are retained,
they negate the assumption that the macroscopic observers are part of the quantum
state [31].

Different resolutions of the paradox have been proposed within interpretations of


quantum mechanics that differ from the Copenhagen Interpretation. We shall
consider two of these here. The Many Worlds interpretation avoids the paradox of
Wigner’s friend by denying the objective existence of wavefunction collapse [32];
in it all phenomena are seen as the consequence of a universal wavefunction. In the
subjective interpretation [33], unitary evolution and measurement operators are
viewed as personal judgments of individual agents, in a way like the Bayesian
interpretation of classical probability theory. Consequently, the wavefunction does
not encode objective facts about reality, but only subjective beliefs of an individual
agent about the results of future measurements. Summarizing, it may be asserted:

Theorem 4. An observer cannot be part of a quantum state.

Awareness and complexity


Neither of the interpretations given in the previous section explains the origins of
the sense of awareness or consciousness that cognitive agents possess. While the
common view within neuroscience is that consciousness is an emergent
phenomenon associated with complexity, the Orthodox Copenhagen Interpretation
is consistent with the understanding of consciousness in Vedanta (including
Kashmir Shaivism) is which it is a transcendent category; in other words, the
physical universe and consciousness are complementary like two sides of a coin
(e.g. [34]). In these traditions, consciousness is compared to light which illuminates
the mind. Consciousness is then a unity, a view that Erwin Schrödinger fully
endorsed in his writings (as in [35]).

9
The minds of cognitive agents are physical systems in which the faculties of
memory, ego, and intelligence (capacity to make judgments) are embedded. It is
easy to understand physical models of memory and intelligence because they are
commonly found in computers and can be expressed in mathematical formalism,
but it is not clear wherefrom ego arises. The minds of different agents are clearly
different for their memories are unique and their judgments circuits may be
different. It is however true that the mechanism of biological memory storage and
recall are not fully understood [36].

Another issue is that of the location of the agent. What do we mean when we speak
of it, whether as a computer or a biological organism? We cannot say that the agent
is the entire physical body for the processing of sensory information does not occur
within the entire body.

In the human, we may begin of speaking of the body of the individual at first pass,
and then modify it and speak of the brain where the mind is located. At the next
pass, it is not the entire brain, but the centers where the self appears to be localized
during a certain cognition. This location seems to change as one goes from a
cognition to another [37] and the self may be seen as sequencing of several selves
that dance with respect to time. Consciousness is not physical, it is like a light that
illuminates the contents of the mind.

Paradoxes of representation
The analysis of observers becomes complicated for the mind maps symbols and the
extra-symbolic relationships amongst them to specific meanings, and this creates
paradoxes of consciousness. When we use language to represent knowledge, we are
compelled to use categories across hierarchies, as in the case of sets and super-sets.
A conflation of categories leads to paradoxes of logic as in the case of the Liar
Paradox. Likewise conflating the quantum state with its potentially mutually-
exclusive component states and the classical world of well-located objects leads to
ambiguities of representation and analysis.

Given that it is normal to only consider symbols that are classical objects in
associative relationships, there is need to also consider objects with shifting
boundaries as well as quantum objects. We see this in new analysis that deals with
largely independent cognitive categories emanating from the architecture of the
brain [38][39] and potential relationship with quantum states [40].

10
Conclusions
It was argued in the paper that awareness appears to be located is not a specific
center in the brain and it is a dynamic category. In self-reflection one may define a
sequence of progressively more refined subjects in relation to the object that
constitutes a regression. Awareness exists at the end of this sequence but it is not
located at a specific place in the brain. The relationship between awareness and the
physical world is like the one of the Quantum Zeno Effect, where the observer
doesn’t alter the dynamics, but only changes the probabilities.

We considered the problem of nested physical systems and showed that observers
cannot exist within them, and we called this result the “One Observer theorem”.
Since complex systems can exist in nested hierarchies, this constitutes a proof
against the view that consciousness, which is implicitly a property of the observer,
is produced by complexity. This result is consistent with the commonsensical view
that there cannot be a conscious agent subsumed by another conscious agent.

The no-go theorems of this study show that AI machines of the future will never
achieve consciousness.

References

1. Haikonen, P.O.A. On artificial intelligence and consciousness. Journal of Artificial


Intelligence and Consciousness 7, 73-82 (2020)
2. Blum, M. and Blum, L. A theoretical computer science perspective on consciousness.
Journal of Artificial Intelligence and Consciousness 8, 1-42 (2021)
3. Reggia, J.A., Katz, G.E., Davis, G.P. Artificial conscious intelligence. Journal of
Artificial Intelligence and Consciousness 7, 95-107 (2020)
4. Chrisley, R. Artificial consciousness, meta-knowledge, and physical omniscience.
Journal of Artificial Intelligence and Consciousness7, 199-215 (2020)
5. Wigner, E.P. Remarks on the mind-body question. In Philosophical reflections and
syntheses. Springer (1995)
6. Stapp, H.P. Mindful Universe. Springer (2007)
7. Liljenström, H. Consciousness, decision making, and volition: freedom beyond chance
and necessity. Theory Biosci. 141, 125–140 (2022)
8. Tegmark, M. The mathematical universe. Foundations of Physics 38: 101–150 (2007)
9. Landauer, R. The physical nature of information. Phys. Lett. A 217, 188–193 (1996)
10. Reggia, J. The rise of machine consciousness: Studying consciousness with
computational models. Neural Networks 44, 112–13 (2013)
11. Penrose, R. The road to reality. Vintage Books (2004)

11
12. Kak, S. Is consciousness computable? NeuroQuantology 17, 71-75 (2019)
13. Kak, S. The Nature of Physical Reality (3rd Edition). Mt. Meru (2016)
14. Kak, S. The limits to machine consciousness. Journal of Artificial Intelligence and
Consciousness 9, 59-72 (2022)
15. Penrose, R. The Emperor’s New Mind: Concerning Computers, Minds and the Laws
of Physics. Oxford University Press, Oxford (1989)
16. Ellis, G.F.R. On the limits of quantum theory: Contextuality and the quantum–
classical cut. Annals of Physics 327, 1890-1932 (2012)
17. Frauchiger, D. and Renner, R., Quantum theory cannot consistently describe the use
of itself, Nature communications 9, 1 (2018)
18. Kak, S. Information theory and dimensionality of space. Scientific Reports 10, 20733
(2020). https://doi.org/10.1038/s41598-020-77855-9
19. Kak, S. Evolutionary stages in a noninteger dimensional universe. Indian J Phys
(2023). https://doi.org/10.1007/s12648-023-02653-8
20. Kak, S. and Kafatos, M. Black holes, disk structures, and cosmological implications
in e-dimensional space. Physics Essays 35, 345-355 (2022)
21. Gautam, A. and Kak, S. Symbols, meaning, and origins of mind. Biosemiotics 6,
301-309 (2013)
22. Nielsen, M. and Chuang, I. Quantum Computation and Quantum Information.
Cambridge University Press (2000)
23. Lamme, V.A. Towards a true neural stance on consciousness. Trends in Cognitive
Sciences, 10 (11), 494-501 (2006)
24. Tononi, G., Boly, M., Massimini, M., Koch, C. Integrated information theory: From
consciousness to its physical substrate. Nature Reviews Neuroscience, 17 (7) (2016)
25. Doerig, A., Schurger, A., Hess, K., and Herzog, M. H. The unfolding argument: Why
IIT and other causal structure theories cannot explain consciousness. Consciousness
and Cognition 72, 49–59 (2019) doi: 10.1016/j.concog.2019.04.002
26. Sudarshan, E.C.G., Misra, B. The Zeno's paradox in quantum theory. Journal of
Mathematical Physics 18 (4), 756–763 (1977)
27. Kak, S. Quantum information and entropy. International Journal of Theoretical
Physics 46, 860-876 (2007)
28. Trimmer, J. D. The present situation in quantum mechanics: A translation of
Schrödinger’s ‘cat paradox’ paper. Proceedings of the American Philosophical Society
124, 323–38 (1980)
29. Peres, A. Quantum Theory: Concepts and Methods. Kluwer, 2002.
30. Matzkin, A. and Sokolovski, D. Wigner-friend scenarios with noninvasive weak
measurements, Physical Review A 102, 062204 (2020)
31. R. Healey, Quantum theory and the limits of objectivity, Foundations of Physics 48,
1568 (2018)
32. DeWitt, B. S. and Graham, N. The Many-Worlds Interpretation of Quantum
Mechanics. Princeton University Press (1973)

12
33. Fuchs, C., Mermin, N.D., Schack, R. An introduction to QBism with an application
to the locality of quantum mechanics. American Journal of Physics 82, 749 (2014)
34. Moore, W.J. Schrödinger --Life and Thought. Cambridge University Press (1992)
35. Schrödinger, W. What Is Life? The Physical Aspect of the Living Cell. Cambridge
University Press (1944)
36. Neisser, U. Cognitive Psychology. Psychology Press (2014)
37. Zeki S. The disunity of consciousness. Trends Cogn Sci 7, 214-218 (2003)
38. Kak, S. Number of autonomous cognitive agents in a neural network. Journal of
Artificial Intelligence and Consciousness 9, 227- 240 (2022).
https://doi.org/10.1142/S2705078522500023
39. Gamez, D. Progress in machine consciousness. Consciousness and Cognition 17,
887–910 (2008)
40. Kak, S. An information principle based on partitions for cognitive data. Journal of
Artificial Intelligence and Consciousness 10, 1-14 (2023);
https://doi.org/10.1142/S2705078522500138

13

You might also like