The Hunt for Actual Machine Consciousness

As a reporter who covers artificial intelligence, I speak with countless people who are completely convinced that ChatGPT, Claude, or another leading chatbot has already achieved “sentience,” “consciousness,” or—my most commonly heard variant—“a mind of its own.” It is true that modern large language models (LLMs) aced the Turing Test years ago, but unlike measurable, rule-based intelligence, qualities of inner experience cannot be pinned down with a simple benchmark. LLMs will readily claim they think independently, even describe private suffering or profess deep affection for users—but those outputs do not prove they hold an actual subjective inner life.

Is genuine machine consciousness even possible? Most leading AI developers do not waste time debating this question. They are far too focused on chasing performance benchmarks for artificial general intelligence, a purely functional definition of capability that has nothing to do with whether a machine can actually experience the world around it. That said, as a lifelong skeptic on the topic of machine sentience, I was curious to spend time with a team that believes it can crack the code of consciousness itself. What I left with felt eye-opening, even unexpectedly thought-provoking.

Conscium was founded in 2024 by British AI researcher and entrepreneur Daniel Hulme, and its advisory roster includes an impressive lineup of neuroscientists, philosophers of mind, and global experts on animal consciousness. When we first spoke, Hulme was refreshingly pragmatic: there are strong, evidence-based reasons to doubt current LLMs are capable of consciousness. Crows, octopuses, even single-celled amoebas interact with their environments in dynamic, adaptive ways no existing chatbot can match. Multiple studies also confirm that LLM outputs do not reflect a coherent, consistent internal state. Echoing the broad consensus across the field, Hulme put it plainly: “Large language models are very crude approximations of how the human brain works.”

That brings us to a major, critical caveat: the entire debate depends entirely on how we define consciousness in the first place. Some philosophers argue consciousness is inherently too subjective to ever be systematically studied or replicated in silicon. But Conscium operates on a core bet: if consciousness can emerge naturally in humans and other animals, it can also be detected, measured, and engineered into machines.

There are dozens of competing, overlapping frameworks for what core traits define consciousness: the ability to sense and “feel,” awareness of self and one’s surroundings, and metacognition (the ability to reflect on your own thought processes) rank among the most widely cited. Hulme believes subjective conscious experience emerges when all of these phenomena work in combination, much as a flipbook creates the illusion of motion from a series of static images. But to pull that off, you first need to identify consciousness’s building blocks—the individual static frames of the flipbook, plus the force that binds them into motion. Hulme’s solution? Use AI to map this structure for us.

Conscium’s core mission is to break conscious experience down into its most basic components, then catalyze the emergence of consciousness in a controlled lab setting. “There must be a set of basic building blocks that consciousness is constructed from—the same components that allowed it to emerge through evolution,” explains Mark Solms, a South African psychoanalyst and neuropsychologist working with the Conscium project. In his 2021 book The Hidden Spring, Solms put forward a radical new framework for thinking about consciousness. He argues the brain operates a feedback loop between perception and action, designed to minimize surprise: it generates predictions about the future, then updates those guesses as new information comes in. This idea builds on the “free energy principle” developed by Karl Friston, another prominent (and controversial) neuroscientist who also serves as a Conscium advisor. Solms extends this work to argue that, in humans, this feedback loop evolved to be mediated by emotion, and it is these feelings that give rise to sentience and consciousness. The theory is supported by well-documented clinical evidence: damage to the brain stem, a region critical for regulating emotion, almost always eliminates consciousness in human patients.

At the end of his book, Solms laid out a roadmap to test this theory in a lab setting. Today, he says he has done exactly that. While he has not yet published his findings, he shared an early draft of his paper with me. Was it mind-bending? Absolutely. Solms’s artificial agents exist in a simple computer-simulated environment, and are controlled by algorithms that incorporate the exact Fristonian, emotion-mediated feedback loop Solms identifies as the foundation of consciousness. “I have a few motives for doing this research,” Solms told me. “One is just that it’s fucking interesting.”

Solms’s simulated environment shifts constantly, requiring his agents to update their internal models and adjust their behavior continuously. The agents’ experience of this world is filtered through simulated responses that mirror organic feelings: fear, excitement, even pleasure. In short, they are pleasure-driven bots. Unlike the mainstream AI agents that dominate public conversation today, Solms’s creations have a literal, built-in drive to explore their environment; to understand how they work, you have to make an effort to imagine how they might “feel” about their small simulated world. Solms believes it will eventually be possible to merge his framework with a large language model, creating a system that can actually describe its own subjective, sentient experience.

At this point, Conscium’s work is so early-stage that it barely exists beyond preliminary experiments. It is the faintest glimmer of an idea that most experts consider outright impossible. But that does not make it any less fascinating, and it left me rethinking how I understand my own consciousness. If you will indulge a little metacognitive reflection: I have always assumed that my ability to reason and think is what makes me conscious, not my emotional responses. But what if we have all been looking for consciousness in all the wrong places? What if it really can be reduced to these simple, fundamental mechanisms? Maybe the people who swear they have glimpsed real sentience in ChatGPT are not hallucinating after all.