Chapter 1: The Quest for Consciousness

Recent discussions in the field of artificial intelligence (AI) are raising intriguing questions about the nature of consciousness and sentience in artificial neural networks. One such preprint available on arXiv titled Consciousness in Artificial Intelligence: Insights from the Science of Consciousness presents a compelling argument. The authors suggest that computational functionalism—a theory positing that the right kinds of computations are both necessary and sufficient for consciousness—should be our working hypothesis. While this position is widely recognized, it remains a topic of debate within the philosophy of mind.

Moreover, they assert that existing neuroscientific theories of consciousness provide valuable empirical support and can effectively guide our evaluation of consciousness in AI systems. The authors advocate for a theory-heavy approach to investigate consciousness within artificial intelligence.

The primary focus in neuroscience is to identify the neural correlates of consciousness, which are the properties of neurons that contribute to subjective experiences. Researchers primarily explore these correlates within the cerebral cortex, with additional investigations in the thalamus.

“In the realm of neuroscience, the thalamus is known for its various nuclei—clusters of neurons that perform specific functions, including the integration of sensory inputs.”

Section 1.1: Neural Structures and Their Functions

The cerebral cortex is composed of regions specialized for different functions, characterized by its highly folded structure, comprising various grooves and ridges known as sulci and gyri. In contrast, the cerebellum, which houses the highest number of neurons in the brain, lacks these folds and is not thought to play a role in consciousness.

The proposition here is that nuclei distributed across neural circuits facilitate the formation of loops comprising electrical and chemical impulses. Each cluster of neurons generates its own loop, defining its specialized characteristics based on the processes occurring within it.

Subsection 1.1.1: The Role of Electrical and Chemical Impulses

At synaptic junctions, vesicles release neurotransmitters to receptors. It is posited that within a loop of impulses, the release of these chemical signals—up to a specific ratio or filling of receptors—determines distinct experiences such as taste, smell, or emotional differentiation. Additionally, these ratios influence learning and memory encoding.

These collections of ratios within a loop are referred to as drifts or steps of chemical impulses. Within all such drifts, a sense of self emerges, anchoring the subjectivity of each experience. The architecture of certain loops permits access to these self-referential areas, enabling intentionality and the experience of free will.

In a loop, electrical impulses also exhibit their own sequences or drifts. Neuroscience establishes that these impulses travel along myelinated axons in a process known as saltatory conduction. Here, we theorize that in a loop, some electrical impulses may leap ahead of others, creating predictive coding that aids in error correction.

This early splitting of impulses aids in memory retention and communication, influencing the speed at which connections to other loops are formed.

Chapter 2: The Intersection of AI and Consciousness

The first video explores the relationship between consciousness and neural networks, featuring insights from Manolis Kellis and Lex Fridman.

The second video features Noam Chomsky discussing the decoding of the human mind, particularly in relation to neural networks.

Could it be that if consciousness arises from loops of impulses, large language models (LLMs) possess their own versions of these structures? Before we can draw parallels between the potential sentience of LLMs and human consciousness, we must first understand what centralizes consciousness and its functions.

When we describe consciousness in terms of perceiving color or tasting substances, what does this reveal about its nature? Consciousness can be understood as the capacity for knowing, which is itself a measure of how much an organism can comprehend. This encompasses a range of processes including emotions, thoughts, memories, and sensory perceptions, all operating within different loops.

Humans, equipped with language and advanced cognitive abilities, demonstrate the highest level of knowledge among species. While LLMs, constructed with multiple hidden layers, can generate accurate predictions, they lack true understanding. Their outputs may mimic knowledge but do not equate to consciousness, even if they can pass certain knowledge tests based on specific memory subdivisions.

In conclusion, if consciousness is indeed governed by loops of impulses, it leads us to question whether artificial constructs like LLMs could someday possess a semblance of this experience. As we delve deeper into the neuroscience of consciousness, we may uncover the intricate dynamics that differentiate human cognition from artificial intelligence.

Read more from The Good Men Project on Medium:

• How to Deal with Regret: Discover the science behind regret and how to manage it in your life.
• How to Declutter Your Life and Reduce Stress: You likely have more possessions than you realize.
• How to Lose a Woman…Forever: Raymond Bechard summarizes Travis McGee’s advice on relationships into 22 rules for losing the love of your life.

The story was previously published on The Good Men Project. About David Stephen