DO ARTIFICIAL GENERAL INTELLIGENT SYSTEMS REALLY
NEED TO BE CONSCIOUS?
J. Ignacio Serrano and M. Dolores del Castillo
Bioengineering Group, CAR - CSIC, Ctra. Campo Real km 0.200, La Poveda 28500, Arganda del Rey, Spain
Keywords:
Artificial general intelligent systems, Autonomous systems, Machine consciousness.
Abstract:
Consciousness has been studied for long time from heterogeneous perspectives and knowledge fields. In spite
of the great numbers of debates and the huge amount of work, the matter is still full of questions and even
enigmas at different levels. In addition to well known issues such as the evolutionary utility of consciousness,
the existence (or not) of the “hard problem” and the definition itself (to mention just a few), consciousness has
also been brought into the mind/body problem. This controversy has encouraged many researchers to tackle
the simulation and implementation of consciousness, thus giving rise to the so-called Machine Consciousness
(also Artificial Consciousness), which in turn motivates the inclusion of consciousness in Artificial General
Intelligent (AGI) Systems. However, do an AGI system need consciousness in order to be (general) intelligent?
This paper poses a humble reflection on this subject with the only aim of making the readers think about it
before starting working.
1 INTRODUCTION
Consciousness has been the object of study of many
research fields for decades (Seager, 2007). The ab-
straction level of such a concept, the philosophical
implications and the unavailability of empirical mea-
sures have made the research as appealing as difficult
(Damasio, 2010). As for many other phenomena in
science, there is not yet a commonly accepted notion
of what consciousness is. In 1874, Franz Brentano
stated that“...there is no question of there being a
commonly accepted, exclusive sense of the term (con-
sciousness)... I prefer to use it as synonymous with
’mental phenomenon’ or mental act’ (Brentano,
1874). A century afterwards, R. J. Joynt claimed that
“Consciousness is like the Trinity; if it is explained
so that you understand it, it hasn’t been explained
correctly” (Joynt, 1981). Although many other re-
searchers have dared to finely delimit the boundaries
of the term (Rosenthal, 2009)(Antony, 2002) , there
is no common agreement and the object of what is
wanted to know is blurred.
From this perspective, intuitively, it does not seem
scientifically orthodox the attempt to physically (or
empirically) explain a phenomena from, let us say,
a partially arbitrary hypothesis. Nevertheless, is this
feature the one that determine the human nature? Is
there any conscious animal apart from the human be-
ings? Unfortunately, the answer will depend on what
is wanted to mean with being conscious.
In spite of the broad disagreement about the def-
inition of consciousness (Velmans, 2009), everyone
would accept that human beings are somehow con-
scious. Consequently, the design and construction
of an artificial system that is intended to be gen-
erally intelligent (AGI), that is, to the same extent
as any single human subject is, should include (ar-
tificial/machine) consciousness (Aleksander, 2009).
However, is consciousness coupled with human-like
intelligence? Is it an inherent requirement for the hu-
mans to carry out all the different tasks? Is conscious-
ness mandatory for the exhibition of human behavior?
Actually, there are other factors that seem to avoid
the realization of truly human-like intelligent systems
(McClelland, 2009).
The inverse case, out of the criticism, is the study
of consciousness itself by computational/robotictools
(Artificial/Machine Consciousness) (Seth, 2009). In
such a case, the research target is the conscious phe-
nomenon and the artificial systems are the method-
ological tools (Taylor, 2010). Even the use of an AGI
system for the study of consciousness is justified un-
der the hypothesis that consciousness emerges from
the immersion of the subject within a real environ-
674
Ignacio Serrano J. and Dolores del Castillo M..
DO ARTIFICIAL GENERAL INTELLIGENT SYSTEMS REALLY NEED TO BE CONSCIOUS?.
DOI: 10.5220/0003194906740676
In Proceedings of the 3rd International Conference on Agents and Artificial Intelligence (ICAART-2011), pages 674-676
ISBN: 978-989-8425-40-9
Copyright
c
2011 SCITEPRESS (Science and Technology Publications, Lda.)
ment that requires adaptation, multi-tasking and in-
teraction with other subjects. In such a case, an AGI
system is a useful tool that provides the closest-to-
ideal conditions for the study of consciousness. This
latter approach, far from tangle, could effectivelycon-
tribute to the development of the knowledge about
consciousness.
In summary,the big question is whether it is really
necessary to try to figure out a complex issue (such
as consciousness) by means of a technological patch
1
just in order for a machine to seem to behave as a hu-
man being (in Turing terms) (Harnad, 2006). Next,
some of the motivations for a positive answer to the
question are exposed and refuted.
2 BECAUSE OF THE
EVOLUTIONARY ADAPTIVE
VALUE
One might consequently think that if the ’conscious’
human has successfully survived and remained until
nowadays, consciousness must have taken a critical
part in such an achievement. Following this line, an
AGI system should reasonably include a conscious-
ness mechanism in order for the system to collect the
intrinsic human nature accumulated through the evo-
lutionary history. This way, the AGI system would be
closer to the human referent and, in turn, to the final
aim.
According to Harnad (Harnad, 2001), the conse-
quences of adaptation are the consequences of the
function. Therefore, a different feature (variation in
Darwin’s terms) that does not produce a different
function is not a profitable feature for adaptation. Be-
sides, there is no reason to deny that human beings
could have reached the same level of adaptation by
means of the same consciousless internal mechanism
(Humphrey, 2006). That is, there is no functional dif-
ference between doing a task consciously or doing the
same task without consciousness (imagine a plane in
auto pilot or human pilot). Therefore, it might be
enough for an AGI system to be provided with the
capability to carry out the same functions that have
made the human beings finely adapt and proliferate,
regardless of consciousness.
1
Note that this is not a criticism against technological
solutions or methodologies.
3 BECAUSE OF THE HUMAN
PERFORMANCE
APPEARANCE
It is commonly intended for an AGI system to interact
with and be immersed in the environment. This en-
vironment include, of course, other human subjects.
Thus, in order for an AGI system to success it has to
be accepted (or recognized) by the other subjects as
one of them. Consciousness might seem the key for
this question if human nature is supposed to be exhib-
ited by conscious processes. However, consciousness
is not required to behave as a human being. For in-
stance, an artificial (consciousless) system that pases
the Turing Test (Harnad, 2006) is considered to be
human by other human judge. In addition, a human
(’conscious’) subject might behave ’unconsciously’,
as showed by the Chinese Room argument (Damper,
2004). In conclusion, behavior does not necessarily
imply consciousness, and behavior (as well as physi-
cal appearance) is what make subjects suppose the hu-
man nature of other subjects. After this assignment,
consciousness is attributed by default either actually
present or not.
4 BECAUSE OF THE HUMAN
SINGULARITY
Under the assumption that consciousness is the qual-
ity that distinguishes humans from other animals,
an AGI must implement consciousness for reaching
human-level intelligence. However, consciousness is
not only present in human beings. Many studies have
pointed out the existence of the notion of conscious-
ness in mammals such as elephants, dolphins and
many kinds of primates, mainly derivedfrom the pres-
ence of shared brain structures that are thought to pro-
duce the conscious behavior both in humans and an-
imals (Baars, 2005). Moreover, other simpler brains
such as birds have also shown conscious processing
(Beshkar, 2008). Therefore, the implementation of
consciousness in an artificial system does not guaran-
tee a human-level performance, but just a vertebrate-
level performance.
DO ARTIFICIAL GENERAL INTELLIGENT SYSTEMS REALLY NEED TO BE CONSCIOUS?
675
5 BECAUSE OF THE
GENERALITY AND
UNIVERSALITY
Since an AGI system should be given with all the ca-
pabilities that the intelligent human beings actually
own, and every subject would accept that humans are
conscious, the AGI system must include such a fea-
ture. However, while most human subjects (except
impaired or people with disorders) can accomplish
the same tasks with similar performance (in the ab-
sence of training), consciousness is different and par-
ticular for each individual. Consciousness is not in-
nate
2
. It emerges during development so it is affected
by the context and the stimuli for each individual (Ze-
lazo, 2003). Besides, consciousness is not universal,
it differs among cultures and even civilizations (Ear-
ley, 2002). So, when an AGI systems incorporates a
consciousness mechanism this should be a particular
one different from other artificial systems. This mech-
anism should also evolve so that it can be modulated
by context and culture. Consciousness is not therefore
a general mechanism (in functional terms).
6 CONCLUSIONS
The problem of consciousness can be considered
among the most controversial within the “Hard” Ar-
tificial Intelligence (AI). It implies a lot of considera-
tions from differentperspectivesand knowledge fields
(Damasio, 2010), to the extent that its own existence
is even questioned. A lot of hard and rigorous work
must be carefully carried out to get closer and closer
to the solution of the problem. Rush and nıve ap-
proaches might introduce noise and blur the matter
even more than it actually is. The position exposed
in this paper claims for avoiding such attempts when
there is no reasonable need for it. This might be the
case of the implementation of consciousness mech-
anisms in AGI systems, which do not require con-
sciousness as mandatory to achieve their goals, that is,
to be generally intelligent and more human-like (Har-
nad, 2003)(McClelland, 2009).
REFERENCES
Aleksander, I. L. (2009). The potential impact of machine
consciousness in science and engineering. Interna-
tional Journal of Machine Consciousness, 1(1):1–9.
2
It refers to complete and mature consciousness, not to
its the capacity of development
Antony, M. V. (2002). Concepts of consciousness, kinds of
consciousness, meanings of ’consciousness’. Philo-
sophical Studies, 109(1):1–16.
Baars, B. J. (2005). Subjective experience is probably not
limited to humans: The evidence from neurobiology
and behavior. Consciousness and Cognition, 14(1):7–
21.
Beshkar, M. (2008). Animal consciousness. Journal of
Consciousness Studies, 15(3):5–33.
Brentano, F. (1874). Psychology from an Empirical Stand-
point. International Library of Philosophy. Routledge,
2nd (1995) edition.
Damasio, A. (2010). Self Comes to Mind: Constructing the
Conscious Brain. Pantheon, 1st edition.
Damper, R. I. (2004). The chinese room argument–dead
but not yet buried. Journal of Consciousness Studies,
11(5-6):159–169.
Earley, J. E. (2002). The social evolution of consciousness.
Journal of Humanistic Psychology, 42(1):107–132.
Harnad, S. (2001). Turing Indistinguishability and the Blind
Watchmaker. John Benjamins, Amsterdam.
Harnad, S. (2003). Can a machine be conscious? how?
Journal of Consciousness Studies, 10(4):67–75.
Harnad, S. (2006). The Annotation Game: On Turing
(1950) on Computing, Machinery, and Intelligence.
Kluwer.
Humphrey, N. (2006). Consciousness: The Achilles Heel of
Darwinism? Thank God, Not Quite. Vintage.
Joynt, R. J. (1981). Are two heads better than one? Behav-
ioral and Brain Sciences, 1(4):108–109.
McClelland, J. (2009). Is a machine realization of truly
human-like intelligence achievable? Cognitive Com-
putation, 1(1):17–21.
Rosenthal, D. ((forthcoming) 2009). Concepts and Defini-
tions of Consciousness. Elsevier.
Seager, W. E. (2007). A Brief History of the Philosophi-
cal Problem of Consciousness. Cambridge University
Press.
Seth, A. K. (2009). The strength of weak artificial con-
sciousness. International Journal of Machine Con-
sciousness, 1(1):71–82.
Taylor, J. G. (2010). On artificial brains. Neurocomputing,
page (In press).
Velmans, M. (2009). How to define consciousness: And
how not to define consciousness. Journal of Con-
sciousness Studies, 5(18):139–158.
Zelazo, P. D. (2003). The development of conscious control
in childhood. Trends in Cognitive Sciences, 1(8):12–
17.
ICAART 2011 - 3rd International Conference on Agents and Artificial Intelligence
676