Summary of the article: This article does three things: gives an operational definition of consciousness, explores the evidence for this operational definition, and, given this definition, answers if machines have consciousness. First, the article explains that consciousness requires two computational processes: C1 or global availability of information, and C2 or self-monitoring, often referred to by psychologists as metacognition. Most processing of the human brain is performed non-consciously, in parallel by various modules. A centralized global workspace functions to make information available and coordinate between these various non-conscious modules, facilitating complex behaviors. This is C1 consciousness. Next, the ability to reflexively self-monitor ones own mental states is C2 consciousness. This includes being aware of what one knows, assessing one's own cognitive processes, and having a sense of confidence about ones particular judgments. C2 consciousness enables self-awareness and introspection, facilitating personal insight, nuanced decision-making and self-regulation. There can be instances where C1 and C2 might come apart. For example, a person might have access to information (C1) without particularly aware about it. This is evident when people can be primed by subliminal information. Conversely, a person can have a sense of confidence about a memory (C2) that never actually occurred, as in the tip-of-the-tongue phenomenon where individuals are aware they know a word but cannot recall it at the moment. - How is "consciousness" defined? - Consciousness is defined as having two essential dimensions: C1 and C2. - C1 is global availability. This is the ability for information to be reflected on and processable by other cognitive systems. - C1 refers to the relationship between a cognitive system and a specific object of thought. For example, the mental representation of "the light" in "The driver is conscious _of_ the light." This information becomes globally available to the organism: we can recall it, act upon it, speak about it, etc. This sense is synonymous with "having the information in mind." - C2 is self-monitoring. This is the ability for a cognitive system to be self-aware. - C2 is self-monitoring, or what psychologists call "meta-cognition," or commonly called introspection: the ability to conceive and make use of internal representation of one's own knowledge and abilities. - What is "unconscious processing (C0)"? - Unconscious processing are those cognitive processes which we are unaware of. This includes processes such as facial recognition, speaker recognition, semantic meaning of sensory input, or decision-making. We tend to underestimate the role of unconscious processes in our mental life. - How do they identify unconscious processes? - Unconscious processing has been identified by tests and brain imaging, such as fMRI. Information displayed subliminally to people has been shown to affect their decision-making and visual recognition. For example, people were flashed a headshot of someone for 50ms (subliminally) and then shown the target headshot immediately after. When the subliminally presented headshot was of the same person as the target headshot, subjects were able to more quickly recognize the target face, despite the subjective experience of the photos being the same whether the subliminal headshot was of a different person or the same person. - What is unconscious learning? - Unconscious processes in the brain whereby knowledge, habits, or skills are acquired without conscious effort. For example, in humans, priming and habituation. Feedforward convolutional neural networks (CNNs) achieve this same kind of learning. - What is "first sense consciousness (C1)"? - First sense consciousness (C1) is the capacity for information to be globally available to other cognitive processes, allowing for this information to be utilized for complex decision-making and planning. There are many competing unconscious cognitive processes, however, a series of informational sieves ensure only the most pertinent information is made globally accessible by various cognitive modules. - C1 consciousness likely evolved out of a need to pool and coordinate information across various cognitive modules, enabling long-term planning and complex decision making and conferring a significant adaptive advantage. This contrasts with simpler cognitive systems, such as flies, which react primarily to sensorimotor stimuli. - C1 consciousness is associated with stable, distributed patterns of brain activity. And is indicated in humans by their ability to verbally report on their internal world, as well as the availability of language itself, a complex representational system that structures and enhances mental processes. - How is information integrated and coordinated? - What is the relation between consciousness and attention? - What is the evidence for the internal global workspace, integration, and broadcasting? - Evidence for human's internal global workspace is indicated by reproducible, stable patterns of brain activity captured by brain imaging. Brain imaging in humans and neuronal recordings in monkeys indicate that the conscious bottleneck is implemented by a network of neurons distributed through the cortex. Broadcasting is indicated by this information being made available to other cognitive modules, as indicated by brain imaging and behavioral expression. - What is the nature of "stability" as a feature of consciousness? - What is "second sense consciousness (C2)"? - How can there be C1 without C2, and C1 without C2? - What do they say about machines and consciousness? How can machines achieve consciousness, if at all? - How are machines implementing unconscious processing? - What can we learn from conscious processing to apply to machine architectures?