Page 34 - AIH-2-3

Artificial Intelligence in Health Machine consciousness

system could literally possess a mind – in other words, that aside the notoriously difficult task of pinning down an
executing the right algorithms might generate genuine exact definition of consciousness and instead agreeing
understanding and cognitive states indistinguishable on practical operational criteria. Levy argues that
from those of humans. This perspective implies that, insisting on a rigid definition may be counterproductive;
76
at some level, the functional organization of a machine rather, if the community can settle on a shared intuitive
could support conscious in the same sense a brain does. In understanding of what consciousness functionally entails,
contrast, the weak AI position holds that machines, at best, researchers could “simply use the word and get on with it”
simulate consciousness without any real inner experience in developing systems that meet those criteria. 78(p210) This
or awareness. From this viewpoint, even the most approach reflects a practical mindset: Even if we lack a
77
advanced AI today (for example, sophisticated language perfect definition of consciousness, we might still engineer
models or robotic assistants) lack subjective sentience or systems that everyone agrees exhibit key properties of
genuine understanding; they merely manipulate symbols consciousness (such as complex adaptivity, learning, and
and exhibit behaviors that mimic consciousness without self-report), thereby moving the field forward without
actually experiencing the world. becoming mired in semantics.
The clash between these perspectives highlights a Other researchers emphasize specific features thought
core conceptual challenge: Explaining how subjective to be indispensable for consciousness. Chatila et al.
79
experience (the essence of consciousness) might emerge focus on self-awareness as the cornerstone of machine
from purely physical or computational processes. This is consciousness, proposing a framework for self-aware
essentially the classic “hard problem of consciousness” robots grounded in several cognitive abilities. They outline
applied to machines: The difficulty of explaining how and fundamental principles by which a robot could be designed
why a physical system could produce the felt quality of to understand its environment and its own role within it,
experience. Even in humans, consciousness defies simple to be cognizant of its actions, and to respond appropriately
explanation; present scientific understanding of brain in real time to changes. Crucially, a self-aware robot
function has yet to fully bridge the gap between neural should also be able to learn from its own experiences and
circuitry and subjective feeling. mistakes and to explicitly demonstrate that it has learned
When considering artificial agents, we are further – for instance, by documenting or communicating its
constrained by our human-centric intuitions: Our acquired knowledge. These capabilities mirror aspects of
understanding of consciousness is largely shaped by human consciousness: Humans continuously monitor
the first-person experience of our own mind, making it their surroundings and their own internal states, adjust
challenging to objectively evaluate whether a machine – behavior on the fly, learn from feedback, and can report on
accessible only from an external, third-person perspective what they have learned. Chatila’s framework thus attempts
– could possess anything akin to a conscious mind. to imbue machines with a form of reflective cognition
analogous to that of humans, on the premise that such
In summary, the strong AI versus weak AI dichotomy
sets the stage for discussing machine consciousness by reflection (knowing what one knows, and knowing what
one does) is a pre-requisite for any genuine consciousness.
asking whether replicating intelligent behavior is sufficient
for authentic consciousness (strong AI) or whether A complementary perspective is offered by Kinouchi
subjective awareness is a qualitatively distinct property and Mackin, who propose that consciousness serves a
80
that machines inherently lack (weak AI). This foundational functional role as an integrative system-level adaptation
debate provides a context for interpreting the progress in mechanism in complex agents. In their architecture, a
neuroscience-inspired AI frameworks and guides our multitude of lower-level processing units (analogous to
skepticism or optimism regarding artificial consciousness. distributed modules in the brain or in a large AI system)
operate in parallel, each handling specific tasks or
4.2. Neuroscience-inspired functional frameworks sensory inputs. Machine consciousness, in this view, is
for artificial consciousness the higher-level function that coordinates and organizes
Amid these philosophical debates, researchers have the outputs of these parallel processes, synthesizing them
proposed various frameworks for building or recognizing into a coherent state that can guide the agent’s overall
consciousness-like properties in machines. Often drawing behavior adaptively. This coordinating role is likened to
inspiration from neuroscience and cognitive science, these the way human consciousness creates a unified experience
frameworks focus on replicating functional attributes and decision-making process out of the brain’s many
of human consciousness in an artificial medium. One simultaneous unconscious computations. Kinouchi and
pragmatic stance, advocated by Levy, suggests setting Mackin and Hildt explicitly draw an analogy to the
81
78
80
Volume 2 Issue 3 (2025) 28 doi: 10.36922/aih.5690