The controversial question of whether machines may ever be conscious must be based on a careful consideration of how consciousness arises in the only physical system that undoubtedly possesses it: the human brain. We suggest that the word “consciousness” conflates two different types of information-processing computations in the brain: the selection of information for global broadcasting, thus making it flexibly available for computation and report (C1, consciousness in the first sense), and the self-monitoring of those computations, leading to a subjective sense of certainty or error (C2, consciousness in the second sense). We argue that despite their recent successes, current machines are still mostly implementing computations that reflect unconscious processing (C0) in the human brain. We review the psychological and neural science of unconscious (C0) and conscious computations (C1 and C2) and outline how they may inspire novel machine architectures.C1: Global availability
This corresponds to the transitive meaning of consciousness (as in “The driver is conscious of the light”)... We can recall it, act upon it, and speak about it. This sense is synonymous with “having the information in mind”; among the vast repertoire of thoughts that can become conscious at a given time, only that which is globally available constitutes the content of C1 consciousness.C2: Self-monitoring
Another meaning of consciousness is reflexive. It refers to a self-referential relationship in which the cognitive system is able to monitor its own processing and obtain information about itself..This sense of consciousness corresponds to what is commonly called introspection, or what psychologists call “meta-cognition”—the ability to conceive and make use of internal representations of one’s own knowledge and abilities.CO: Unconscious processing: Where most of our intelligence lies
...many computations involve neither C1 nor C2 and therefore are properly called “unconscious” ...Cognitive neuroscience confirms that complex computations such as face or speech recognition, chess-game evaluation, sentence parsing, and meaning extraction occur unconsciously in the human brain—under conditions that yield neither global reportability nor self-monitoring. The brain appears to operate, in part, as a juxtaposition of specialized processors or “modules” that operate nonconsciously and, we argue, correspond tightly to the operation of current feedforward deep-learning networks.
The phenomenon of priming illustrates the remarkable depth of unconscious processing...Subliminal digits, words, faces, or objects can be invariantly recognized and influence motor, semantic, and decision levels of processing. Neuroimaging methods reveal that the vast majority of brain areas can be activated nonconsciously...Subliminal priming generalizes across visual-auditory modalities...Even the semantic meaning of sensory input can be processed without awareness by the human brain.
...subliminal primes can influence prefrontal mechanisms of cognitive control involved in the selection of a task...Neural mechanisms of decision-making involve accumulating sensory evidence that affects the probability of the various choices until a threshold is attained. This accumulation of probabilistic knowledge continues to happen even with subliminal stimuli. Bayesian inference and evidence accumulation, which are cornerstone computations for AI, are basic unconscious mechanisms for humans.
Reinforcement learning algorithms, which capture how humans and animals shape their future actions on the basis of history of past rewards, have excelled in attaining supra-human AI performance in several applications, such as playing Go. Remarkably, in humans, such learning appears to proceed even when the cues, reward, or motivation signals are presented below the consciousness threshold.What additional computations are required for conscious processing?
C1: Global availability of relevant information
The need for integration and coordination. Integrating all of the available evidence to converge toward a single decision is a computational requirement that, we contend, must be faced by any animal or autonomous AI system and corresponds to our first functional definition of consciousness: global availability (C1)...Such decision-making requires a sophisticated architecture for (i) efficiently pooling over all available sources of information, including multisensory and memory cues; (ii) considering the available options and selecting the best one on the basis of this large information pool; (iii) sticking to this choice over time; and (iv) coordinating all internal and external processes toward the achievement of that goal.
Consciousness as access to an internal global workspace. We hypothesize that...On top of a deep hierarchy of specialized modules, a “global neuronal workspace,” with limited capacity, evolved to select a piece of information, hold it over time, and share it across modules. We call “conscious” whichever representation, at a given time, wins the competition for access to this mental arena and gets selected for global sharing and decision-making.
Relation between consciousness and attention. William James described attention as “the taking possession by the mind, in clear and vivid form, of one out of what seem several simultaneously possible objects or trains of thought”. This definition is close to what we mean by C1: the selection of a single piece of information for entry into the global workspace. There is, however, a clear-cut distinction between this final step, which corresponds to conscious access, and the previous stages of attentional selection, which can operate unconsciously...What we call attention is a hierarchical system of sieves that operate unconsciously. Such unconscious systems compute with probability distributions, but only a single sample, drawn from this probabilistic distribution, becomes conscious at a given time. We may become aware of several alternative interpretations, but only by sampling their unconscious distributions over time.
Evidence for all-or-none selection in a capacity-limited system. The primate brain comprises a conscious bottleneck and can only consciously access a single item at a time. For instance, rivaling pictures or ambiguous words are perceived in an all-or-none manner; at any given time, we subjectively perceive only a single interpretation out of many possible ones [even though the others continue to be processed unconsciously]...Brain imaging in humans and neuronal recordings in monkeys indicate that the conscious bottleneck is implemented by a network of neurons that is distributed through the cortex, but with a stronger emphasis on high-level associative areas. ... Single-cell recordings indicate that each specific conscious percept, such as a person’s face, is encoded by the all-or-none firing of a subset of neurons in high-level temporal and prefrontal cortices, whereas others remain silent...the stable, reproducible representation of high-quality information by a distributed activity pattern in higher cortical areas is a feature of conscious processing. Such transient “meta-stability” seems to be necessary for the nervous system to integrate information from a variety of modules and then broadcast it back to them, achieving flexible cross-module routing.
C1 consciousness in human and nonhuman animals. C1 consciousness is an elementary property that is present in human infants as well as in animals. Nonhuman primates exhibit similar visual illusions, attentional blink, and central capacity limits as human subjects.C2: Self-monitoring
Whereas C1 consciousness reflects the capacity to access external information, consciousness in the second sense (C2) is characterized by the ability to reflexively represent oneself ("metacognition")
A probabilistic sense of confidence. Confidence can be assessed nonverbally, either retrospectively, by measuring whether humans persist in their initial choice, or prospectively, by allowing them to opt out from a task without even attempting it. Both measures have been used in nonhuman animals to show that they too possess metacognitive abilities. By contrast, most current neural networks lack them: Although they can learn, they generally lack meta-knowledge of the reliability and limits of what has been learned...Magnetic resonance imaging (MRI) studies in humans and physiological recordings in primates and even in rats have specifically linked such confidence processing to the prefrontal cortex. Inactivation of the prefrontal cortex can induce a specific deficit in second-order (metacognitive) judgements while sparing performance on the first-order task. Thus, circuits in the prefrontal cortex may have evolved to monitor the performance of other brain processes.
Error detection: Reflecting on one’s own mistakes ...just after responding, we sometimes realize that we made an error and change our mind. Error detection is reflected by two components of electroencephalography (EEG) activity: the error-relativity negativity (ERN) and the positivity upon error (Pe), which emerge in cingulate and medial prefrontal cortex just after a wrong response but before any feedback is received...A possibility compatible with the remarkable speed of error detection is that two parallel circuits, a low-level sensory-motor circuit and a higher-level intention circuit, operate on the same sensory data and signal an error whenever their conclusions diverge. Self-monitoring is such a basic ability that it is already present during infancy. The ERN, indicating error monitoring, is observed when 1-year-old infants make a wrong choice in a perceptual decision task.
Meta-memory The term “meta-memory” was coined to capture the fact that humans report feelings of knowing, confidence, and doubts on their memories. ...Meta-memory is associated with prefrontal structures whose pharmacological inactivation leads to a metacognitive impairment while sparing memory performance itself. Metamemory is crucial to human learning and education by allowing learners to develop strategies such as increasing the amount of study or adapting the time allocated to memory encoding and rehearsal.
Reality monitoring. In addition to monitoring the quality of sensory and memory representations, the human brain must also distinguish self-generated versus externally driven representations - we can perceive things, but we also conjure them from imagination or memory...Neuroimaging studies have linked this kind of reality monitoring to the anterior prefrontal cortexDissociations between C1 and C2
According to our analysis, C1 and C2 are largely orthogonal and complementary dimensions of what we call consciousness. On one side of this double dissociation, self-monitoring can exist for unreportable stimuli (C2 without C1). Automatic typing provides a good example: Subjects slow down after a typing mistake, even when they fail to consciously notice the error. Similarly, at the neural level, an ERN can occur for subjectively undetected errors. On the other side of this dissociation, consciously reportable contents sometimes fail to be accompanied with an adequate sense of confidence (C1 without C2). For instance, when we retrieve a memory, it pops into consciousness (C1) but sometimes without any accurate evaluation of its confidence (C2), leading to false memories.Synergies between C1 and C2 consciousness
Because C1 and C2 are orthogonal, their joint possession may have synergistic benefits to organisms. In one direction, bringing probabilistic metacognitive information (C2) into the global workspace (C1) allows it to be held over time, integrated into explicit long-term reflection, and shared with others...In the converse direction, the possession of an explicit repertoire of one’s own abilities (C2) improves the efficiency with which C1 information is processed. During mental arithmetic, children can perform a C2-level evaluation of their available competences (for example, counting, adding, multiplying, or memory retrieval) and use this information to evaluate how to best face a given arithmetic problem.Endowing machines with C1 and C2
[Note: I am not abstracting this section as I did the above descriptions of consciousness. It describes numerous approaches rising above the level of most present day machines to making machines able to perform C1 and C2 operations.]
Most present-day machine-learning systems are devoid of any self-monitoring; they compute (C0) without representing the extent and limits of their knowledge or the fact that others may have a different viewpoint than their own. There are a few exceptions: Bayesian networks or programs compute with probability distributions and therefore keep track of how likely they are to be correct. Even when the primary computation is performed by a classical CNN, and is therefore opaque to introspection, it is possible to train a second, hierarchically higher neural network to predict the first one’s performance.Concluding remarks
Our stance is based on a simple hypothesis: What we call “consciousness” results from specific types of information-processing computations, physically realized by the hardware of the brain. It differs from other theories in being resolutely computational; we surmise that mere information-theoretic quantities do not suffice to define consciousness unless one also considers the nature and depth of the information being processed.
We contend that a machine endowed with C1 and C2 would behave as though it were conscious; for instance, it would know that it is seeing something, would express confidence in it, would report it to others, could suffer hallucinations when its monitoring mechanisms break down, and may even experience the same perceptual illusions as humans. Still, such a purely functional definition of consciousness may leave some readers unsatisfied. Are we “over-intellectualizing” consciousness, by assuming that some high-level cognitive functions are necessarily tied to consciousness? Are we leaving aside the experiential component (“what it is like” to be conscious)? Does subjective experience escape a computational definition?
Although those philosophical questions lie beyond the scope of the present paper, we close by noting that empirically, in humans the loss of C1 and C2 computations covaries with a loss of subjective experience. For example, in humans, damage to the primary visual cortex may lead to a neurological condition called “blindsight,” in which the patients report being blind in the affected visual field. Remarkably, those patients can localize visual stimuli in their blind field but cannot report them (C1), nor can they effectively assess their likelihood of success (C2)—they believe that they are merely “guessing.” In this example, at least, subjective experience appears to cohere with possession of C1 and C2. Although centuries of philosophical dualism have led us to consider consciousness as unreducible to physical interactions, the empirical evidence is compatible with the possibility that consciousness arises from nothing more than specific computations.