Global Journal of Computer Science and Technology: D

Neural & Artificial Intelligence

Volume 19 Issue 1 Version 1.0 Year 2019
Type: Double Blind Peer Reviewed International Research Journal
Publisher: Global Journals
Online ISSN: 0975-4172 & Print ISSN: 0975-4350

An Integration of Deep Learning and Neuroscience for Machine

Consciousness
By Ali Mallakin
Abstract- Conscious processing is a useful aspect of brain function that can be used as a model to
design artificial-intelligence devices. There are still certain computational features that our conscious
brains possess, and which machines currently fail to perform those. This paper discusses the necessary
elements needed to make the device conscious and suggests if those implemented, the resulting
machine would likely to be considered conscious. Consciousness mainly presented as a computational
tool that evolved to connect the modular organization of the brain. Specialized modules of the brain
process information unconsciously and what we subjectively experience as consciousness is the global
availability of data, which is made possible by a non modular global workspace. During conscious
perception, the global neuronal work space at parieto-frontal part of the brain selectively amplifies relevant
pieces of information. Supported by large neurons with long axons, which makes the long-distance
connectivity possible, the selected portions of information stabilized and transmitted to all other brain
modules. The brain areas that have structuring ability seem to match to a specific computational problem.
The global workspace maintains this information in an active state for as long as it is needed. In this
paper, a broad range of theories and specific problems have been discussed, which need to be solved to
make the machine conscious. Later particular implications of these hypotheses for research approach in
neuroscience and machine learning are debated.
GJCST-D Classification: I.2.6

AnIntegrationofDeepLearningandNeuroscienceforMachineConsciousness
Strictly as per the compliance and regulations of:

© 2019. Ali Mallakin. This is a research/review paper, distributed under the terms of the Creative Commons Attribution-
Noncommercial 3.0 Unported License [Link] permitting all non-commercial use,
distribution, and reproduction in any medium, provided the original work is properly cited.
 An Integration of Deep Learning and
Neuroscience for Machine Consciousness
Ali Mallakin

Abstract- Conscious processing is a useful aspect of brain integrated information and global workspace theories.
function that can be used as a model to design artificial- The first attempts to look at consciousness by learning
intelligence devices. There are still certain computational more about the physical processes that motivate our
features that our conscious brains possess, and which

2 019
conscious experiences[3]. This theory attempts to
machines currently fail to perform those. This paper discusses
create a measure of the integrated information that
the necessary elements needed to make the device conscious

Year
and suggests if those implemented, the resulting machine forms consciousness and reveals that the level of
would likely to be considered conscious. Consciousness integration represents the quality of an organism's
consciousness. It explains the nature and source of 21
mainly presented as a computational tool that evolved to
connect the modular organization of the brain. Specialized consciousness and claims that consciousness is

Global Journal of Computer Science and Technology ( D ) Volume XIX Issue I Version I
modules of the brain process information unconsciously and identical to a particular type of information, which
what we subjectively experience as consciousness is the requires physical realization, not merely functional, and
global availability of data, which is made possible by a non which can be measured mathematically according to
modular global workspace. During conscious perception, the
the Φ metric. The second theory suggests that we have
global neuronal work space at parieto-frontal part of the brain
a memory bank from which the brain draws information
selectively amplifies relevant pieces of information. Supported
by large neurons with long axons, which makes the long- to form the experience of conscious awareness[4].
distance connectivity possible, the selected portions of While integrated information theory focuses more on
information stabilized and transmitted to all other brain identifying whether an organism is aware, the global
modules. The brain areas that have structuring ability seem to workspace theory offers abroader approach to
match to a specific computational problem. The global understanding how consciousness works. My motivation
workspace maintains this information in an active state for as to have useful models in artificial consciousness is
long as it is needed. In this paper, a broad range of theories toward the design of communication technology that is
and specific problems have been discussed, which need to be
convenient in everyday human activities.
solved to make the machine conscious. Later particular
There is a long way to go in our understanding
implications of these hypotheses for research approach in
neuroscience and machine learning are debated. of consciousness as investigators continue to explore
the different bases of consciousness such as the
I. Scientific View and Definition of physical, social, cultural and psychological influences
Consciousness that contribute to our conscious awareness.
Consciousness defined as a state of awareness which

O
ne of the difficulties with the study of features apprehensiveness, subjectivity, experience
consciousness is the lack of a universally through sensory perceptions, the state of wakefulness,
accepted definition. Another struggle is that the sense of ego, and the control of the mind with
empirical scientists must deal with questions, which has knowledge of thought processes [5-7].Other most
no quantitative answer, and responses can touch upon essential attributes of consciousness are curiosity and
unscientific subjects. The very act of thinking creativity, which could be a factor that distinguishes
demonstrates the reality of one’s existence and humans from other earth beings[8,9].Intelligence is also
consciousness, which viewed as an individual’s another fundamental aspect of brain requirement and
awareness of its internal states as well as the events in consciousness, which some beliefs like consciousness
the surrounding environment. Recent research on it is a property of a physical mind, not a metaphysical
consciousness has focused on understanding the phenomenon[10]; however, not a sufficient condition for
neuroscience behind our conscious experiences. Some creativity [11]. Apart from humans, other earth beings
investigators have used brain scanning technologies to also show certain levels of creativity[12]. In defining
seek out specific brain modules or neuronal networks consciousness, the problem emerges from attempts
that may link to different conscious events[1,2]. Two that try to explain it in purely physical terms. That is why
major theories of consciousness, which will be psychology and quantum mechanics have been
discussed further through this paper, consist of integrated to challenge the materialistic view of
consciousness [13].
Author: West Coast Biomedius, Vancouver, British Columbia, Canada. Recent advances in artificial intelligence (AI)
e-mail: WCBIOMED@[Link] have revived the possibility that machines would

© 2019 Global Journals

 An Integration of Deep Learning and Neuroscience for Machine Consciousness

eventually mimic all the brain’s abilities, including towards developing basic properties that can engineer
consciousness. Deep learning networks, inspired by into machines. It is considered an essential part of the
neuroscience, have led to the creation of artificial neural brain responsible for data processing where conscious
networks (ANN) that can even occasionally surpass is an internal model of attention. Supporters of this
human capacity [14,15]. These types of the network theory claimed that the attention schema theory
without having biophysical properties of real neurons provides a possible answer to the puzzle of subjective
have several neurobiological features, including experience by which the brain computes a simplified
nonlinear input-output functions, layers with converging model of the process and the current state of attention
projections, and modifiable synaptic weights. Advances which is the basis of subjective intelligence[20].
in computer field now allow such networks to operate on Consciousness made by the application of ANN
complex problems with capacity that previously was a may be better explained by exploring the designs that
privilege of a real brain. Computations implemented by allow the human brain to generate consciousness, then
2 019

current deep-learning systems that correspond mostly transferring those understandings into computer
to non conscious processes in the human brain, algorithms. The aim is to study aspects of the cognitive
Year

although it often considered as the highpoint of the neuroscience of consciousness that may be pertinent
22 brain and something impossible to achieve by for learning machines. Consciousness given by brain
machines, in this paper is claimed to be otherwise. This capacity seeks to understand its relationship to our
paper reviews the previous studies and much of what objective world. It states the relationship between a
Global Journal of Computer Science and Technology ( D ) Volume XIX Issue I Version I

we know about how brains generate consciousness, cognitive system and a specific object of thought, such
what are the relationships among conscious, as a mental representation. Conscious information
subconscious and non conscious brains and how these becomes globally available to us, which gives us the
findings can be used in advance artificial intelligence. ability to recollect, act upon it, and speak about it.
This paper reviews specific existing models of Having the data in mind among the immense range of
consciousnesses and presents the approaches that thoughts that can become conscious at a given time,
make a machine more conscious. Besides, it discusses only the ones that are universally available constitute the
the role and relationship between quantum mechanics content of consciousness of an objective world. In
(QM) and remote viewing (RV). This relationship another hand, consciousness scan be reflexive or
between quantum non locality and conscious or spontaneous, in such a way that the cognitive system
subconscious brain functions is essential, as quantum can monitor its processing and obtain information about
processes in biological phenomena are becoming more itself in a self-referential manner. This self-examination
apparent. Further, it discusses non-local consciousness ability of consciousness is commonly known as meta-
or remote viewing (RV) as an application of cognition, which is having self-awareness and the ability
consciousness that permits a viewer to describe details to conceive and make use one's knowledge and
regarding a target that is inaccessible to the ordinary abilities[21].Some suggested that global and reflexive
senses. consciousnesses may constitute independent
dimensions of conscious computations. Both can
II. Current Theories for Machine overlap physical substrates, and in the brain, both
Consciousness depend on the prefrontal cortex[22]. Theoretically, the
two may come apart as there can be global
There are various models of machine consciousness without reflexive consciousness, such as
consciousness, each with certain strengths and when reportable processing is not accompanied by
limitations. Machine consciousness was classified accurate meta-cognition, or reflexive consciousness
earlier into four groups [16], and later to five categories without global one, such as when a self-monitoring
based on recurring matters on the fundamental issues operation unfolds without being consciously reportable
that are most fundamental to consciousness[17]. The [23]. Sepertae computations can be performed before
most recent classification includes a global workspace, we consider their synergy. Furthermore, many estimates
information integration, an internal self-model, higher- involve neither of these two conscious and therefore
level representation, and attention mechanisms[17]. called unconscious. What we pursue is a satisfying
Although various models presented, the creation of an scientific theory of consciousness that predicts under
intelligent machine has not been much less successful which conditions any particular physical system,
due to the computational issues and the inability to whether it is a complex network of neurons or electronic
explain the application of high-level cognitive algorithms circuit board has awareness experience. Moreover, we
in terms of neuro-computational processing[18]. need to investigate what makes the quality of
Mechanistic theory of consciousness hypothesizes that experiences different and if they have a specific
how a brain with an attention schema may have function. Such a theory will allow us to conclude which
subjective awareness[19]. In the attention schema systems will experience anything. Experience has shown
theory, consciousness viewed beyond philosophy, that without an approach that brings testable
© 2019 Global Journals
 An Integration of Deep Learning and Neuroscience for Machine Consciousness

predictions, any speculation based on our intuition approach may proceed with identifying a minimal
would not be reliable. experimental paradigm that allows to contrast visible
and invisible stimuli [27,28]. Then, carefully quantify the
III. The Relationship among Conscious, subject’s introspection, which is recognizing what it
Subconscious and Unconscious known. Self-examination defines the conscious of
Minds subjective perception and must, therefore, be recorded
alongside other objective measures of brain activity. The
Sigmund Freud psychoanalytic theory capacity to report a piece of information to oneself or
presented in the early twentieth century has retain edits other scan be the next phase of the study, in which this
hold on the shaping of views regarding the doctrine of sense of consciousness that is called report ability can
the human mind despite the new advancements. be distinguished from other concepts such as attention,
Neuroscience has identified the physical brain watchfulness, or self-consciousness. The last can be the

2 019
mechanisms underlying subjective processes. Following use of modern neuro-imaging tools to compare the
those breakthroughs, consciousness could then be behaviors and brain activity patterns evoked by

Year
seen as a computational property associated with a reportable and unreportable stimuli, thus uncovering the
certain level of information processing. Three signatures of consciousness. 23
computational levels may be distinguished based on our
conscious, subconscious or unconscious mind. It is IV. Specific Signatures of a Brain

Global Journal of Computer Science and Technology ( D ) Volume XIX Issue I Version I
estimated that just 10% of the cognitive function is made Consciousness
up of conscious thought, which can direct our focus and
able us to envision what is not real. Our subconscious is Further study of consciousness in the human
the storage point of recent memories, which needed for brain reveals that although subliminal stimuli can induce
quick recall and holds current information that we use considerable activity in many human minds, conscious
every day. The unconscious mind is where our perception is associated with a set of specific
memories and past experiences reside and include signatures. The first sign is amplification and access to
memories that have been repressed and those that have the prefrontal cortex, in which a conscious image is
been merely consciously forgotten. The unconscious amplified and gains access to higher levels of
brain continually communicates with the conscious mind representation, particularly in prefrontal and parietal
via the subconscious, which provides us with the cortices. Next would be tracing the propagation of
meaning of our interactions with the world. It breathes conscious and unconscious images that shows
through feelings, emotions, thoughts, sensations, and unconscious activity can be strong in the cortex, yet
dreams. At unconscious mind algorithms process perish in a brief time within higher cortical areas. This
symbols blindly and, apparently without any awareness. conscious image is amplified in a non-linear manner, in
For instance, our visual system blindly and an event known as global ignition.
unconsciously processes the images. By this mean, Nevertheless, in a short period brain activity
many of the brain’s unconscious computations are becomes more stable when the stimulus is conscious
rational computations, and unexpectedly any machine than when it is not [29].This conscious ignition is
that attempts towards objectivity would be submitted to accompanied by increased in bidirectional exchanges of
similar human-like deceptions. information in the human brain. During a conscious
Above the unconscious processing level, two episode, the cortex activates at greater distances, and
higher levels of information processing are defined that correlations of brain signals demonstrate this. The brain
correspond to primary and secondary consciousness can spontaneously generate its patterns of circulated
[24]. The first level known as conscious access, which is activity with constant changes, even in the absence of
a selective attribution of a piece of information, selected stimuli[30]. This resting state activity can partially predict
for its relevance and became the focus of additional the content of consciousness. Finally, neurons in
processing. Second, is conscious self-representation, in prefrontal and anterior temporal cortex fire to a specific
which the cognitive system shows one or several concept and do so only when the corresponding word
representations of its knowledge, for instance, it may or image is presented consciously. This activity acted as
know what it is currently focusing on, or it made an error. a signature of conscious perception and known as the
Consequently, the system not only commits its all-or-none law that applies to the firing of neurons[31].
resources to a specific piece of information but also The findings of brain-specific signatures and
recognizes that it knows. The assumption is that this explanation of generic consciousness are compatible
self-knowledge is represented in the same format as the with the Global Neuronal Workspace (GNW) hypothesis,
knowledge of other individuals, which allows this which tries to address the serious question of the
information to be shared with others[25,26]. function of consciousness[27,28,32].Consciousness
Those definitions made it entirely possible to appears to be required for specific operations, and it is
study consciousness experimentally. The experimental not just a mere epiphenomenon. Subliminal information

© 2019 Global Journals

 An Integration of Deep Learning and Neuroscience for Machine Consciousness

is temporary, but conscious knowledge is stabilized and A conscious machine should know when it is
available for long-term thinking. Consciousness converts wrong or when it is uncertain about something. In the
the incoming data and reduces it to a form that can be human brain, this corresponds to meta-cognitive
reported or stored, while unconscious processes knowledge that has been linked to the prefrontal cortex.
compute with an entire probability distribution, and In several ways, a computer could be equipped with
consciousness samples from it through the similar functionality. Primarily, it could be equipped with
subconscious mind. Consciousness is also involved in statistical programs that do not just give an answer, but
routing information to other processing phases, which also compute the probability that this answer is correct.
allows us to perform random chains of operations [33]. Subsequently, it can have an error-detection system,
similar to the brain’s error-negativity, which continually
V. Machine Consciousness Should be
compares the current activity with prior expectations and
Self-Organizing instinctively reacts if the current behavior is likely to be
2 019

The phenomenon of consciousness may wrong. Finally, this error detection device could be
require a self-organizing system, similar to the brain’s attached to a corrective mechanism that allows the
Year

physical structure, which current machines lack this system to continually looks for alternative ways to get
24 critical feature. Adaptive self-organizing mechanisms the correct answer.
can be designed to be as sophisticated as the human Another point which may be unique to humans
brain; however, at the moment we require a is the ability to represent self-knowledge and have a
Global Journal of Computer Science and Technology ( D ) Volume XIX Issue I Version I

mathematical theory of computation for these types of theory of mind, which allows us to model other accounts
system. Perhaps it’s true that only biological machines and to use this knowledge to maximize the usefulness of
can be sufficiently creative and flexible. That may information that we can provide to them [34]. Additional
suggest start working on engineering new biological attributes that should be considered include producing
structures that are or can become conscious. a machine with a consciousness that has qualities,
Conscious as a useful aspect of brain function likely to instincts, and creativity. Also, states of consciousness
be helpful to a design of artificial-intelligence devices. that can change experiences such as emotions and
Certain computational features can be considered, expression scan be considered. Current machines often
which if they are implemented, the resulting machine lack such significance and a device that could simulate
would be likely to be considered conscious, or at least its user’s mind would undoubtedly provide more
more aware than what is currently available. The first to relevant information. Algorithms that handle such
be considered is a workspace for global availability, the recursive representations of other accounts are currently
act of selecting and making a piece of information being developed [35,36].On the other hand, certain
accessible for processing and decision-making by the human behaviors such as creativity and a sense of
whole system. Global availability highlights or draws freedom, which do not appear to come from logic or
attention to information that remained unconscious until calculations, cannot be found in machines.
that moment. Even though the brain possesses Another viewing platform on consciousness
specialized modules that operate non-consciously and comes from quantum theory, which is one of the most
are dedicated to specific tasks, it also maintains a in-depth theories of physics. According to the orthodox
global neuronal workspace, where accurate information Copenhagen Interpretation, consciousness and the
is selected and shared across all modules. According to physical world are balancing aspects of the same
the GNW hypothesis, consciousness evolved to break reality. When a person observes some aspect of the
this modularity. The GNW can extract relevant physical world, that person’s conscious interaction
information from virtually any brain module, and make it causes a noticeable change. Since this theory takes
available to the entire organism, in which the prefrontal consciousness as a given and not derived from physics,
cortex appears to act as a central information sharing the Copenhagen Interpretation may be considered as a
device. The information presents in this global phenomenon that exists by itself, while it requires brains
workspace at any given time is known as conscious. to become real. This view was popular with the pioneers
Machines may benefit from a similar architecture for of quantum theory such as Niels Bohr, Werner
flexible information sharing, capable of broadcasting to Heisenberg, and Erwin Schrödinger. A well-known
the entire system a potentially relevant piece of example of this is the paradox of Schrödinger's cat, in
information. It would be interesting to pursue this idea in which a cat is placed in a situation that results in it being
the context of present-day machine-learning algorithms, equally likely to survive or die – and the act of
which can make the best use of the broadcasted observation itself is what makes the outcome inevitable.
information. Another feature for a conscious machine is The opposing view reveals that consciousness
its inherent abilities: a database that contains a list of its emerges from biology, just as biology itself emerges
apps, the kind of knowledge they possess, what goals from chemistry which, in turn, arises from physics. This
they can fulfill, how fast can operate, and how likely they less broad concept of consciousness agrees with the
should be corrected. neuroscientists view that the processes of the mind are
© 2019 Global Journals
 An Integration of Deep Learning and Neuroscience for Machine Consciousness

identical to states and operations of the brain. It also VII. Neuroscience and Consciousness
coincides with a more recent interpretation of quantum
theory motivated by an attempt to rid it of paradoxes, Consciousness is qualitative and coherent, in
the Many Worlds Interpretation, in which observers are a which for each conscious experience some qualitative
part of the mathematics of physics. entities have that experience and each piece of this
conscious experience fits precisely into a global picture
VI. Particle and Consciousness of our world [40]. As conscious experience depends on
Entanglements brain activity; therefore, neuroscience has an extensive
contribution in explaining consciousness. Except for
Since the past few decades, quantum neuroscience, no other area can provide so much
processes in the brain have been given attention to information and advances concerning consciousness,
explain consciousness and its enigmatic features. As which focuses on the neural correlates of

2 019
functional quantum processes in biology are becoming consciousness. Considering that not all neural
more and more apparent, recent evidence suggests that correlates are explanatory, so identifying the correlates

Year
quantum non locality occurs in conscious and is the first step in the neuroscience of consciousness. A
subconscious brain functions. Remote viewing (RV) is a thorough study of consciousness through neuroscience
mental ability that allows a viewer to describe or give 25
requires that we understand the relevant neural
details about a target that is inaccessible to normal properties at the right level of analysis. Understanding of

Global Journal of Computer Science and Technology ( D ) Volume XIX Issue I Version I
senses [37,38]. For example, a viewer might be asked consciousness will unquestionably emerge from
to specify a location in a different part of the world, neuroscience but cracking its mysteries will need the
which a reviewer has never visited; or a viewer might be knowledge and discoveries of many other scientific
asked to describe an event that happened long ago disciplines. To understand the connection between
without being said anything about the target. Therefore, brain activity and consciousness, we need to focus on
RV deals directly with target information processed by neurons, study computational and spiritual models, and
human consciousness. QM deal with particles such as conduct a theoretical analysis. Here we will focus on
electrons and photons, in which the concept of particle identifying such principles by considering the activity of
entanglement refers to local measurements on a particle neurons. The effort is to address how neural properties
that will instantaneously interact with an entangled may determine when a state is conscious and how they
particle, without considering the distance and how far may explain what the content is of that conscious. As a
apart the particles are [39].QM and RV both involve phenomenal consciousness is not reducible to either of
entanglement, in which particle entanglement can be these cases, there is a limitation to empirical
described in QM and consciousness entanglement in explanations of phenomenal consciousness as
RV. Entanglement in non-quantum or classical version empirical explanations are fundamentally either
arises in situations where we have partial knowledge of structural or functional.
the state of two systems. In QM entanglement, the Explanation of consciousness remains in the
quantum information is encoded in such a way that neural action that presents conscious contents and the
none of the two qubits (quantum bit: unit of quantum) unknown element that changes a state with that content
carries any clear information on its own as all of the data from unconsciousness to consciousness. A variety of
is encoded in their joint properties. Such entanglement theories can describe generic consciousness that offers
is one of the counter-intuitive features of QM and leads a need for the presence of higher-level neural properties
to its inconsistencies. Consciousness entanglement in necessary or adequate for the consciousness of a given
the RV process refers to the local intentions by a viewer state. Global Neuronal Workspace is one of these
on target information that will instantaneously interact theories, which suggests access into the neural
with entangled target information, no matter how far workspace is essential and sufficient for a state to be
apart the viewer and target are. For the viewer, the consciousness. In contrast, supporters of Recurrent
entangled experience is the target. RV experimental Processing Theory believe that frequent processing in
result shows that nonlocal precognitive connections are, sensory areas is necessary and enough for perceptual
in fact, part of our reality. Our memory fits with the consciousness, and therefore admission to the
simple linear model of time as the only reality based on Workspace is not necessary. In Higher-Order Theories,
our accepted experience. While the non-linear QM seem the presence of a higher-order state connected to
model has not been experimentally verified in physics prefrontal areas of a brain is necessary and sufficient for
laboratories, RV precognition experimental data support a phenomenal experience. Subsequently, recurrent
the notion of backward through time motion. Therefore, processing in sensory areas or access into the
non-local interactions may also involve general aspects workspace is not necessary. At last Information
of space-time described by the theory of relativity. Integration Theories consider a form of integration of
information as essential and adequate parts for a state
to be conscious.

© 2019 Global Journals

 An Integration of Deep Learning and Neuroscience for Machine Consciousness

Most neuroscientists instead of explaining the emphasizes on properties of first-order neural

existence of consciousness in the biotic domain, try to representation as explaining consciousness. It claimed
explain generic consciousness by identifying neural that repetitive processing is necessary and sufficient for
properties that can turn consciousness on and off and consciousness [45-47]. This theory focuses mainly on
by determining the neural representational basis of the visual sensory modality, it argued to apply to all of
conscious contents. Modern neuroscience of the senses, and further, it distinguishes conscious and
consciousness attempts to explain consciousness by unconscious information. These distinctions were made
focusing on neural correlates of consciousness (NCCs) along the lines of different types of processing that
that is an essential initial step in understanding happen within the visual cortex and include the feed
consciousness[41,42]. Talk of correlates is deep-rooted forward, locally recurrent, and globally recurrent types.
in the neuroscience of consciousness, so one must Recurrent processing occurs where sensory systems
remember that the goal is to find the subset of neural are highly interconnected and involve feed forward and
2 019

correlates that are explanatory, in answering particular feedback connections.

questions. Neuroscientific theories that contribute to The well-known Lamme’s theory of recurrent
Year

explaining specific and generic consciousness further processing identifies four stages of normal visual
26 discussed here. We realize that there are limits to neural processing that consist of: I) In superficial feed forward
explanations of consciousness, precisely because of the processing stage visual signals are directed locally
explanatory gap [43].The reference to neural correlates within the optical system; II) In deep feed forward
Global Journal of Computer Science and Technology ( D ) Volume XIX Issue I Version I

here means neural explanatory correlate of processing step visual signals have traveled further
consciousness. forward and they can influence action; III) In superficial
The theories discussed here provide higher recurrent processing phase information has traveled
level neural properties that are necessary or sufficient for back into former visual areas, leading to local, repetitive
generic consciousness of a given state. The first is processing; IV) During widespread intermittent
global neuronal workspace theory that is tied to the processing information activates extensive regions
access of the brain architecture and assumes [45-47]. Based on this theory recurrent processing is
perceptual, aide-mémoire, attentional, valuational and necessary and adequate for consciousness. Therefore,
motoric functions of a cortical structure, which involves the visual state is conscious once a specific intermittent
workspace neurons with long-range connections linking processing state attains the suitable visual circuitry.
arrangements. The global workspace in neural terms Consequently, the global neuronal workspace embraces
defined as long-range workspace neurons within that recurrent processing at Stage 4 is necessary for
different systems that can constitute the workspace, but consciousness while the recurrent processing theory
they should not necessarily identify within the holds that repetitive processing at Stage 3 is sufficient.
workspace. A subset of workspace neurons becomes Base on this difference, recurrent processing theory
the workspace when they exemplify specific neural affirms phenomenal consciousness without access by
properties. What determines which workspace neurons the global neuronal workspace. Each approach has its
constitute the workspace at a given time is the activity of answer, which is different from the other, one requiring
those neurons given the subject's current state. The access but then the other denying it. The organizational
workspace then is not a rigid neural structure but a challenge in testing for the presence of phenomenal
rapidly changing neural network, typically only a proper consciousness independently of access remains a
subset of all workspace neurons. This theory can also hurdle for both theories.
provide an account of phenomenal consciousness in Higher-Order Theory reveals that one is in a
addition to access consciousness. As imaging data can conscious state if and only if one relevantly represents
be obtained that reveal widespread activation when oneself as being in such a state [48,49]. It can be
consciousness is present, it can predict widespread described by a conscious visual state of seeing an
activation of a cortical workspace network as it object if and only if one appropriately represents oneself
correlates with phenomenal conscious experience[44]. being in that visual state. This higher-order state by
As we track phenomenal consciousness by access in representing the first-order state that represents the
the introspective report, therefore, widespread activity world results in having the first order state that is
during reports of conscious experience correlates with conscious. The justification for such theories is that if
both access and phenomenal consciousness. However, one is in a visual state but not aware of that state, then
this correlation cannot state whether the observed the visual state would not be conscious. Consequently,
activity in a report is the basis of phenomenal to be in a conscious state, one must be aware of it and
consciousness or access consciousness. represent it [50]. Higher-order theories tie high-order
Other theories such as Recurrent Processing representations with activity in prefrontal cortex, which is
Theory has a different explanation and ties perceptual assumed to be the area with the required higher-order
consciousness to recurrent activity in sensory areas and representations. On certain higher-order theories, one
processing independent of the workspace, with can be in a conscious visual state even without any
© 2019 Global Journals
 An Integration of Deep Learning and Neuroscience for Machine Consciousness

visual system activity, as long as one represents oneself once the brain does not present the necessary
as being in that state. attributes, it does not create consciousness.
Empirical tests of the higher-order theory Encouraged by these predictions, it generalizes its
against other accounts show the need to study the claims beyond human consciousness to animal and
prefrontal cortex of a brain [51]. Based on the higher- artificial consciousness. Since it identifies the subjective
order theory, lesions to prefrontal cortex should affect experience of consciousness with objectively
consciousness; therefore, testing the prefrontal cortex measurable dynamics of a system, it can be measurable
for consciousness is necessary [52]. Against higher- in principle, and consequently, proposes the Φ metric to
order theories, recent studies claim that patients with quantify consciousness [58]. Recent new studies
surgically removed prefrontal cortex maintain preserved present more novel and practical methods to estimate Φ
perceptual consciousness [53], which gives support to from high-density EEG that can apply to different states
recurrent processing theories, therefore, deem that of consciousness. The investigation of the EEG

2 019
prefrontal cortical activity is not necessary for properties corresponding to Φ metric enables to gather
consciousness. information on large-scale network correlates of various

Year
At last, Integrated Information Theory states of consciousness [59,60].
emphasizes on the concept of integrated information to
VIII. Conclusion 27
explain generic consciousness, and it assumes
consciousness needs a grouping of elements within a This paper attempts to gather and present the

Global Journal of Computer Science and Technology ( D ) Volume XIX Issue I Version I
system that has physical cause-effect power upon one features necessary to produce a machine
another[54-56].It postulates that any conscious system consciousness with a more human-like experience for
must possess three properties, which include: I) The artificial systems. If the essential computational features
existence of consciousness advocates a system of that conscious brains possess and currently lack in
mechanisms with a specific cause-effect power. II) The machines can be implemented, then artificial systems
organizational nature of consciousness suggests that its that appear natural would be possible. In addition to the
mechanical elements within a system must have the interpretation of consciousness as a metaphysical
ability to combine and that those combinations have phenomenon, this paper presents its concrete definition
cause-effect power. III) Since consciousness is based on the models and biological study of conscious
informative; it must specify or differentiate one brains. Here the author compared the various existing
experience from others. models of consciousness to highlight the essential and
This theory claims that consciousness is distinct features of what makes the machine aware.
identical to a certain kind of information, in which its Even though, the definition of consciousness very much
realization requires physical, not merely functional equals a biological perspective; the proposed machine
integration, which can be measured mathematically consciousness is entirely computational and includes
according to the Φ metric. In principle, it proposes the Φ various phenomenological characteristics of
metric to quantify consciousness and attempts to consciousness. Based on the models and features
explain the nature and source of consciousness by presented in this paper, the organization of machine
incorporating that the system carries integrated consciousness is a functional property and
information if the useful informational content of the progressively approximated by incorporating features
whole is greater than the sum of the informational such as a workspace for global information sharing, up
content of the parts. If there is no partitioning where the to date self-knowledge, statistical programs, error
summed informational content of the parts equals the detection systems, and theory of mind, which the latter
whole; then the system carries integrated information, allows modeling other minds and use this information to
and it has a positive value for Φ. Intuitively, the maximize the usefulness of data. Also, machine
interaction of the components adds more to the system consciousness needs to be an emerging phenomenon
than the parts acting alone. as the development of the human brain indicates that
This theory attempts to bring a balance by functional units such as prefrontal and anterior temporal
preserving the cartesian intuitions that experience is cortex are responsible for the emergence of
immediate, direct and unified, but on the other hand, it consciousness. As the central administration that
takes neuroscientific accounts of the brain for an controls and coordinates all processes, whether
understanding of what the nature of a physical system conscious or sub conscious is necessary for developing
must be to be conscious [57].The methodology of the consciousness, the machine model of consciousness
most developers of this theory that are neuroscientists should mimic the biological systems functionally and
involves characterizing the primarily subjective nature of retains a distinct architecture and basic algorithm for
consciousness and postulating the physical traits implementing consciousness. The conscious machine
necessary for a system to realize it. The theory claims with its proposed features might be neither complete nor
that these predictions agree with observations of the practically feasible at present; however, it should guide
brain's physical realization of consciousness; therefore, building models of conscious machines in the future.

© 2019 Global Journals

 An Integration of Deep Learning and Neuroscience for Machine Consciousness

We may not need to wait and see if computers 11. Karwowski, M., Dul, J., Gralewski, J., Jauk, E.,
become conscious as the design of more complicated Jankowska, D.M., Gajda, A., Chruszczewski, M.H.,
computers has not yielded any satisfactory result. Benedek, M. (2016).Is creativity without intelligence
Instead, we may need to be more foresighted and possible? a necessary condition analysis.
proactively integrate machine consciousness in the new Intelligence. 57:105–117.
design processes to have immediate benefit from its 12. Kaufman, A.B., Butt, A.E., Kaufman, J.C., Colbert-
[Link] philosophical and empirical studies White, E.N. (2011). Towards a neurobiology of
can be done, which involve other fields in addition to creativity in nonhuman animals. Journal of
computer sciences such as humanities, divinity, natural Comparative Psychology.125(3): 255-272.
sciences, and psychology as an empirical science. It 13. Brabant, O. (2016). More than meets the eye:
may be more useful to design a machine in a way that it Toward a post-materialist model of consciousness.
concludes it has consciousness and can prove this EXPLORE: The Journal of Science and Healing.
2 019

conclusion. This intelligent machine then can use its 12:347–354.

self-model to regulate its data flow and understand the 14. Silver, D., Schrittwieser, J., Simonyan, K.,
Year

actions of others. Antonoglou, I., Huang, A., Guez, A., Hubert, T.,
Baker, L., Lai, M., Bolton, A., Chen, Y., Lillicrap, T.,
28 References Références Referencias
Hui, F., Sifre, L., Van den Driessche, G., Graepel, T.,
1. Carhart-Harris, R.L., Leech, R., Hellyer, P.J., Hassabis, D. (2017).Mastering the game of Go
Global Journal of Computer Science and Technology ( D ) Volume XIX Issue I Version I

Shanahan, M., Feilding, A., Tagliazucchi, E., without human knowledge. Nature. 550:354-359.
Chialvo, D.R., Nutt, D. (2014). The entropic brain: a 15. Lake, B.M., Ullman, T.D. (2017). Tenenbaum JB,
theory of conscious states informed by neuro Gershman SJ. Building machines that learn and
imaging research with psychedelic drugs. Front think like people. Behav Brain Sci. 40:e253.
Hum Neuro sci. 3(8): 20. 16. Gamez, D. (2008). Progress in machine
2. Fins, J.J. (2008). Neuroethics and neuroimaging: consciousness. Consciousness and Cognition.
Moving toward transparency. American Journal of 17:887–910 (2008).
Bioethics. 8:46–52. 17. Reggia, J.A. (2013). The rise of machine
3. Tononi G1.(2005). Consciousness, information consciousness: Studying consciousness with
integration, and the brain. Prog Brain Res.150: computational models. Neural Networks. 44:
109-26. 112–131.
4. Baars, B.J., Franklin, S. (2009).Consciousness is 18. Reggia, J.A., Monner, D., Sylvester, J. (2014). The
computational: The LIDA model of global computational explanatory gap,” Journal of
workspace theory. International Journal of Machine Consciousness Studies. 21:153–178
Consciousness. 1: 23-32. 19. Graziano, M.S.A.,Webb, T.W. (2014). A mechanistic
5. Baars, B.J. (1997). In the theatre of consciousness. theory of consciousness. International Journal of
global workspace theory, a rigorous scientific theory Machine Consciousness. 6:163–176.
of consciousness. Journal of Consciousness 20. Graziano, M.S.A. (2017). The Attention Schema
Studies.4:292–309. Theory: A Foundation for Engineering Artificial
6. Revonsuo, A. (2006). Inner presence: Consciousness. Front. Robot. AI.
Consciousness as a biological phenomenon. MIT 21. Peters, F. (2013).Theories of Consciousness as
Press. Reflexivity. The Philosophical Forum. 44:341-372.
7. Haladjian, H.H., Montemayor, C. (2016). Artificial 22. De Sousa, A. (2013).Towards an integrative theory
consciousness and the consciousness-attention of consciousness: Part 1 (neurobiological and
dissociation. Consciousness and Cognition. cognitive models). Mens Sana Monogr. 11:100–150.
45:210–225. 23. Peebles, G. (2013).Reflexive theories of
8. Hunter, J.A., Abraham, E.H., Hunter, A.G.: Goldberg consciousness and unconscious perception.
LC, and Eastwood JD. (2016). Personality and Philosophical Psychology. 31:25-43.
boredom proneness in the prediction of creativity 24. Edelman, G. (1989). The remembered present.
and curiosity. Thinking Skills and Creativity. 22: Basic Books: New York.
48–57. 25. Bahrami, B., Olsen, K., Latham, P.E., Roepstorff, A.,
9. Lin, W-L., Shih, Y-L. (2016).The developmental Rees, G., Frith, C.D. (2010). Optimally interacting
trends of different creative potentials in relation to minds. Science. 329: 1081-5.
children’s reasoning abilities: From a cognitive 26. Kerr, N.L., Tindale, RS. (2004). Group performance
theoretical perspective. Thinking Skills and and decision making. Annu Rev Psychol. 55:623-55.
Creativity. 22:36–47. 27. Baars, B.A. (1989).Cognitive theory of
10. Starzyk, J.A., Prasad, D.K. (2011). A computational consciousness. Cambridge, Mass.: Cambridge
model for machine consciousness. Int. J. Mach. University Press.
Conscious. 3: 255-281.

© 2019 Global Journals

 An Integration of Deep Learning and Neuroscience for Machine Consciousness

28. Dehaene, S. (2014). Consciousness and the Brain: 45. Lamme, V.A.F, Roelfsema P.R. (2000). The distinct
Deciphering How the Brain Codes Our Thoughts modes of vision offered by feed forward and
(Reprint edition). Penguin Books. recurrent processing. Trends in neurosciences. 23:
29. Schurger, A., Sarigiannidis, I., Naccache, L., Sitt, 571-579
J.D., Dehaene, S. (2015). Cortical activity is more 46. Lamme, V.A.F. (2006). Towards a True Neural
stable when sensory stimuli are consciously Stance on Consciousness. Trends in Cognitive
perceived. Proceedings of the National Academy of Sciences, 10: 494–501
Sciences of theUnited States of America. 112, 47. Lamme, V.A.F. (2010). How Neuroscience Will
E2083-2092 Change Our View on Consciousness. Cognitive
30. Barttfeld, P., Uhrig, L., Sitt, J.D., Sigman, M., Neuroscience. 1: 204–220
Jarraya, B., Dehaene, S. (2015). Signature of 48. Rosenthal, D.M.(1986). Two concepts of
consciousness in the dynamics of resting-state consciousness. Philosophical Studies. 49: 329–359

2 019
brain activity. Proceedings of the National Academy 49. Rosenthal, D.M.(2005). Consciousness and Mind.
of Sciences of the United States of Oxford: Oxford University Press

Year
America.112:887-892 50. Rosenthal, D.M.(2002). Explaining Consciousness”,
31. Quiroga, R.Q., Mukamel, R., Isham, E.A., Malach, In Philosophy of Mind: Classical and Contemporary 29
R., Fried, I. (2008). Human single-neuron responses Readings, David J. Chalmers (ed.), Oxford: Oxford
at the threshold of conscious recognition. Proc Natl University Press, 109-131

Global Journal of Computer Science and Technology ( D ) Volume XIX Issue I Version I
Acad Sci USA. 105:3599-604 51. Hakwan, L. and Rosenthal, D.M.(2011). Empirical
32. Dehaene, S., Changeux, J.P. (2011).Experimental Support for Higher-Order Theories of Conscious
and theoretical approaches to conscious Awareness. Trends in Cognitive Sciences. 15:
processing. Neuron. 70:200–27 365–373
33. Sackur, J., Dehaene, S.(2009). The cognitive 52. Kozuch, B. (2014). Prefrontal Lesion Evidence
architecture for chaining of two mental operations. against Higher-Order Theories of Consciousness.
Cognition. 111:187-211 Philosophical Studies. 167: 721–746
34. Sperber, D., Wilson, D. (1988). Communication and 53. Boly, Melanie, Marcello Massimini, Naotsugu
cognition. Behavioral and Brain. Sciences. 10: Tsuchiya, Bradley R. Postle, Christof Koch, and
697-789 Giulio Tononi. (2017). Are the Neural Correlates of
35. Baker, C.L., Saxe, R., Tenenbaum, J.B. (2009). Consciousness in the Front or in the Back of the
Action understanding as inverse planning. Cerebral Cortex? Clinical and Neuro imaging
Cognition.113:329-49 Evidence. Journal of Neuroscience, 37: 9603–9613
36. Daunizeau, J., den Ouden, H.E., Pessiglione, M., 54. Tononi, G. (2004). An Information Integration Theory
Kiebel, S.J., Stephan, K.E., & Friston, K.J. (2010). of Consciousness. BMC. 5: 42
Observing the observer (I): meta-bayesian models 55. Tononi, G. (2008). Consciousness as Integrated
of learning and decision-making. PLoS Information: A Provisional Manifesto. The Biological
One.5(12):e15554 Bulletin. 215(3): 216–242
37. Brown, C. (2005). Remote Viewing: The Science and 56. Tononi, G. (2015). Integrated Information Theory.
Theory of Nonphysical Perception. Farsight Press. Scholarpedia. 10 (1): 4164
38. Marks, D. (2000). The Psychology of the Psychic 57. Edelman, G. M., & Tononi, G. (2000). A universe of
(2nd Edition). Prometheus Books. consciousness: How matter becomes imagination.
39. Einstein, A. Podolsky, B., and Rosen, N. (1935). A universe of consciousness: How matter becomes
Can Quantum-Mechanical Description of Physical imagination. New York, NY, US: Basic Books.
Reality Be Considered Complete? Physical Review. 58. Tononi G., Boly M., Massimini M., Koch C.(2016).
47: 777-780 Integrated information theory: from consciousness
40. Nagel, T. (1974). What is it like to be a bat? to its physical substrate. Nat. Rev. Neurosci. 17:
Philosophical Review. 83: 435-50 450–461
41. Crick F. and C. Koch.(1998). Consciousness and 59. Eagleman, Sarah L.; Drover, David R. (2018).
neuroscience. Cerebral Cortex. 8: 297–107 Calculations of consciousness: electroence
42. Crick F. and C. Koch. (2003). A framework for phalography analyses to determine anesthetic
consciousness. Nature Neuroscience. 6: 119–126 depth. Current Opinion in Anesthesiology. 31:
43. Levine J. (1983). Materialism and qualia: The 431–438
explanatory gap. Pacific Philosophical Quarterly. 60. Hyoungkyu Kim., Anthony G. Hudetz, Joseph Lee,1
64:354-61 George A. Mashour, UnCheol L. and the ReCCognition
44. Dehaene S. and Changeux JP. (2011). Experimental Study Group. (2018). Estimating the Integrated
and theoretical approaches to conscious Information Measure Phi from High-Density
Electroencephalography during States of Consciousness
processing. Neuron. 70:200-27
in Humans. Front Hum Neurosci. 12: 42

© 2019 Global Journals