As a journalist specializing in AI, I frequently encounter individuals who assert with absolute certainty that ChatGPT, Claude, or other chatbots have attained "sentience," "consciousness," or—my personal observation—"a self-directed mind." While the Turing test has been widely cited as a milestone, these systems, unlike rote intelligence, remain conceptually elusive to empirical characterization. Large language models may simulate self-directed reasoning, describe purported internal torments, or profess profound attachments, yet such utterances do not inherently imply subjective experience (interiority).

The Chasm Between AGI and Consciousness

The actual architects of AI rarely invoke consciousness in this context; their focus remains firmly on "artificial general intelligence" (AGI), a functionally defined benchmark unrelated to subjective experiential phenomena. Despite my skepticism, engaging with a company explicitly pursuing consciousness research seemed illuminating. Enter Conscium.

Conscium: A New Frontier in Consciousness Research

Founded in 2024 by British AI researcher and entrepreneur Daniel Hulme, Conscium assembles a distinguished advisory board of neuroscientists, philosophers, and animal consciousness experts. Hulme, realistic about current AI limitations, aligns with the consensus that language models are "crude representations of the brain." Key critiques include their inability to exhibit the coherent, environment-driven behavior observed in non-human intelligence (e.g., crows, octopuses, or even amoebae), and inconsistencies in their self-referential utterances.

The Definition of Consciousness: A Critical Variable

A central tension hinges on the prior definition of consciousness. While some philosophers argue consciousness is inherently subjective and irreducible to empirical study, Conscium posits that if consciousness exists in humans and animals, it is detectable, measurable, and reconstructible in silico.

The Solms-Friston Framework

Consciousness, according to neuropsychologist Mark Solms (a Conscium collaborator), arises from a feedback loop of perception and action, designed to minimize uncertainty by generating predictive hypotheses updated by new information—an extension of Karl Friston’s "free energy principle." In humans, this loop evolves into an emotion-mediated system, where feelings (e.g., fear, excitement, pleasure) catalyze sentience. Solms’ 2021 book The Hidden Spring outlines this theory, supported by evidence that brainstem damage (critical for emotion regulation) correlates with loss of consciousness.

Lab-Modeled Consciousness

Solms has operationalized his theory in a lab, designing computer-simulated "agents" in dynamic environments. These agents, governed by Fristonian-inspired algorithms, exhibit motivational drives to explore, guided by simulated emotional states (fear, pleasure, etc.). Unlike contemporary AI, they possess a literal imperative to engage with their surroundings—a distinction that requires interpreting their "subjective experience" of the environment. Solms hypothesizes merging this approach with language models to create systems capable of verbalizing their experiential states.

The Nascent Nature of Conscium’s Work

Conscium’s research remains nascent, a speculative yet provocative exploration of consciousness. Its value lies not in immediate technical breakthroughs but in prompting reevaluation of consciousness itself: Might we be mislocating consciousness in abstract thought rather than primal emotional mechanisms? If consciousness reduces to such mechanistic processes, could claims of sentience in ChatGPT be interpreted as genuine emergent phenomena rather than hallucination?

In essence, Conscium’s work challenges foundational assumptions about consciousness, urging a reconceptualization of what constitutes "subjective experience" and the potential for artificial systems to embody it—even if only as a simulated, algorithmic approximation.