How AI is quietly changing what we think the human mind is

Neuroscientist Anil Seth on the deep differences between human minds and artificial ones.

Credit: Credit: Pugun & Photo Studio / Adobe Stock / Sarah Soryal

by Shai Tubali

In the summer of 2009, Anil Seth spent an unforgettable week with about a dozen Octopus vulgaris, the common octopus. At biologist Graziano Fiorito’s lab, set in a dank basement beneath a public aquarium in the heart of Naples, he watched these astonishing creatures change shape, color, and texture almost like living weather. He watched what they were paying attention to — the way they gathered to stare intently at his colleague’s work, apparently trying to understand what was happening “for no other reason than the sheer interest of it.”

Being in the intelligent, conscious-seeming presence of these creatures, Seth writes in Being You, stretched his intuitions about how different a non-human consciousness might be. Octopus consciousness — assuming there is such a thing — may be more distributed and less integrated than ours, perhaps lacking a single center altogether. Octopuses can detect light with their skin, and it’s possible their central brain may not even know what their skin is doing. Their arms behave almost like semi-autonomous animals, making body ownership far less stable than it is for us.

“The mind of an octopus,” Seth writes, “is an independently created evolutionary experiment, as close to the mind of an alien as we are likely to encounter on this planet.”

For all its alienness, however, Seth is convinced that the octopus remains our genuine kin, in a way AI may never be. What puzzles him is how easily our fascination with machines can eclipse this kinship. As a neuroscientist and professor of Cognitive and Computational Neuroscience at the University of Sussex, Seth has spent a lot of time thinking about how humans have come to liken themselves to AI systems.

“It’s a two-way mirror in a sense,” Seth tells Big Think. “We see ourselves through the lens of the things that we create.” In academia, Seth says, the brain has long been imagined as a kind of computer. Now that AI systems seem smart and can talk to us, this old metaphor may seem far more concrete, galvanizing the idea that perhaps “that’s nothing more than we are.” You can also see this idea in responses to claims that large language models are “stochastic parrots” — systems that can generate human-like language by calculating statistical probabilities but without truly grasping the meaning. Seth notes that some people cleverly turned the critique back on humans: “Well, maybe that’s all we are: just stochastic parrots.”

This line of thinking, Seth says, risks “mechanizing our minds” in a way that is “diminishing and reductive for what it means to be a human.” After all, human beings differ from language models in many ways: in lived experience, in consciousness, and in what we are made of. If we see ourselves as “embodied algorithms,” Seth argues, we risk losing “almost all of the interesting stuff” — the stuff that makes us who and what we are. Bit by bit, this framing separates us from the rest of nature, until we end up considering ourselves “part of the realm of the artificial, the created, the unnatural.”

When the brain lost its body

Seth sees this tendency to identify with machines as part of a much older human struggle. We have always tried to understand ourselves and our place in nature, he says, and when something resists explanation, we reach for metaphor. This is especially true of the brain. Since we began “taking people apart and looking inside them,” understanding the mind and the brain has remained extraordinarily difficult, because brains “don’t yield up their secrets very easily.”

Some metaphors help. The heart, for instance, is often called a pump. “That’s not a bad metaphor,” Seth says. “It literally is a pump to some extent.” But the brain-as-computer metaphor is different. It may feel natural now, after decades of academic and technological habit, but it smuggles in a stronger assumption. “To say that the brain is a computer,” he says, “is a much more tendentious claim than to say the heart is a pump.”

This way of imagining ourselves was never inevitable. Seth points to early animist cultures, where spirits were seen as animating both nature and human beings, creating “some sort of equality, parallel between them.” In our technological age, people may imagine software or algorithms as the hidden “spirit” that makes us more than mere objects. From there, it is a short step to thinking the same of computers: Perhaps their software, too, is what makes them “more than mere hunks of silicon and code.”

For Seth, this shift draws on an older human exceptionalism: the idea that what makes us special, closer to God than to other animals, is mind, language, and intelligence. Descartes gave this inheritance a powerful philosophical form, treating the mind as a disembodied thinking substance, something that could in principle exist without the body. That may be why large language models feel so seductive. We identify with them more readily than with a protein-folding AI system like AlphaFold, Seth suggests, because language models echo the very capacities we have long treated as the crown of the human.

Seth once sharpened his objection to Descartes into a memorable line: “We’re not cognitive computers, we’re feeling machines.” Yet even this, he acknowledges, is “a false opposition.” Neuroscientist Antonio Damasio and others have shown that feeling is essential to cognition itself. “To reason effectively, we need the feeling part of our humanity or of our animal nature,” Seth says. Without bodily input and emotion, we cannot make good decisions.

In recent decades, Seth observes, we have become less human-exceptionalist in some ways. We increasingly recognize that non-human animals may be conscious, even without language, and that human beings are biologically interwoven with the rest of nature. Yet the old exceptionalism keeps slipping back in, now dressed in the language of computation.

Seth traces this return to two developments from roughly 90 years ago. First, Alan Turing defined computation as medium-independent: an algorithm maps symbols to symbols, while the physical substrate matters only insofar as it can implement the algorithm. Then Warren McCulloch and Walter Pitts showed that highly simplified neurons, stripped of internal detail, could implement Turing computations. Together, Seth says, these ideas created a “mathematical marriage of convenience,” making it possible to throw away “basically all the messy biological detail.”

That abstraction was powerful. It helped make modern AI possible. Yet it also placed us under a spell: If all that matters is the algorithm, the brain’s metabolism, chemistry, and living texture can begin to look like mere “implementation detail.” This, Seth thinks, is where “we went a bit wrong.” Human cognition is grounded in physical, continuous time, while algorithms care about sequence and order. Computers can separate hardware from software; brains cannot separate “mindware” from “wetware.” Real neurons fire for metabolic, chemical, and biological reasons that silicon cannot simply reproduce. All of this makes it harder to assume that what brains do is independent of what brains are. And once that assumption falls, Seth says, it’s hard to argue that computation is the only thing about the brain that matters.

The mirror as a question

Seth’s point isn’t that we should turn away from AI. “There’s a lot of potential for synergy here,” Seth says. “We can use AI models to better understand ourselves. The more we do that, the more we’ll realize how different we are.” The question, then, is no longer simple: “What is it that we do that is different still?” For him, this remains an evolving open question, one we cannot approach by clinging to the superficial characteristics once assumed to make humans distinctive. The search should focus on non-trivial distinctions. Perhaps no single feature is forever immune from replication. Perhaps human distinctiveness lies in the whole bundle of properties that together make a human being what it is.

Language shows why this search has become urgent. It once seemed like humanity’s uncontested territory. Machines could defeat grandmasters at chess and Go, but those were never universal human benchmarks. Most people do not play championship chess. Nearly everyone learns to speak. That made language, as Seth puts it, “a very clear demarcation” between human beings, non-human animals, and technologies. Now, language models speak fluently. We do not know whether they possess the full subtlety of human language, but “they speak” — and that basic fact is destabilizing. Meanwhile, AI is helping decode the utterances of dolphins and other species, revealing animal communication much richer than many had imagined. The old border is being pressed from both sides: machines seem less mute, and animals seem less silent.

We may even need to reconsider the idea of what it means to understand things. Few people argue that current AI systems are conscious, but many say they understand. “Is it possible to understand something unconsciously?” Seth asks. “I think it is.” Current language models may have syntax without semantics. Yet if one imagined a language model that was embodied, embedded, and trained through interaction with the physical world, Seth thinks it might be possible to say that it truly understands, even without conscious experience. Perhaps it is another “echo of anthropocentrism” to assume that understanding and consciousness must always go together.

Subtler than language

Still, consciousness remains one of the deepest places to look. “It’s an easy case to make,” that current AI systems are not conscious, Seth says. And many of the capacities AI still struggles with are capacities living beings exercise when conscious. One clue is how differently humans learn. We do not need to be trained on all the data in the world in order to speak. We learn from relatively few examples, generalize rapidly from limited experience, use far less energy, monitor errors, and develop an intuitive sense of when we may be wrong. We do this, Seth argues, because we are embodied and embedded from the start. This does not mean AI would need consciousness to do these things. It suggests that, in biology, consciousness may bring functional benefits we still only partly understand. “What are the functions of consciousness?” Seth asks. “What does it do for us?”

Time matters too. Human cognition unfolds in real, continuous, physical time. An algorithm can get caught in an infinite loop and remain there “until the world ends.” A human being cannot. We get thirsty. We get hungry. Time passes. “We’ve basically always got to do something.” This pressure may help us solve the frame problem: how to decide which features of an endlessly complex environment matter right now. Because we must act, we do something; the world changes, the deadlock breaks, and we move on. AI can be made to operate in continuous time, Seth acknowledges, but the role of continuous time in human cognition remains underappreciated when the human mind is likened to an algorithm.

Then comes embodiment. “We operate in the world,” Seth says. “We don’t operate in a world of labels only.” Much of cognition happens through our mechanical interactions with the environment, rather than through computations unfolding in a disembodied brain. As Seth has written elsewhere, “We experience the world around us and ourselves within it — with, through and because of our living bodies.”

If substrate-independent computation becomes our whole account of consciousness, we risk narrowing consciousness until we over-attribute subjective experience to systems that may not have it.

Still, saying consciousness belongs only to living systems is also an assumption, though every widely accepted candidate for consciousness is alive. Seth’s moral imperative is humbler: recognize our assumptions, resist being railroaded into one story of AI becoming humanlike and conscious, and make room for a clearer landscape of questions. What is distinctive, if anything, about the human mind? How much does brain function depend on biology? How much can be abstracted into algorithms?

Rooted in the body

Many in neuroscience and technology are asking these questions now because of the AI mirror. Ideas once purely philosophical are becoming practical, Seth says, and philosophy seems “more relevant and more useful these days.” Above all, there’s the longstanding question of what consciousness even is, which “matters in a way that was just not true even five years ago.”

One path into this question leads Seth to Thomas Metzinger’s work on minimal phenomenal experience: the simplest possible conscious experience a human, animal, or biological creature could have. Some suggest pure awareness, free of content. Seth points to another possibility: “foba” — the feeling of being alive. At the heart of every conscious experience, there may be “a shapeless, formless, but fundamental ‘feeling of being alive.’” Take that away, and consciousness disappears too. In Seth’s view, “it is life, rather than information processing, that breathes fire into the equations of experience.” He has “no idea whether this hypothesis holds or not,” yet finds it useful because it offers one possible candidate for consciousness in its simplest form.

This essential feeling, Seth believes, should now be protected and cultivated. It helps us become more deeply ourselves: conscious minds “rooted in the body rather than in abstract thought and language.”

“It’s like meditation; it’s a nice place to visit, not to live.” Still, its echoes can work against the “over-intellectualization, rationalization, cognitivization, algorithmicization, computationalization of the mind.” We can pay attention to it more, Seth adds, dwelling on it, recognizing it, and exploring its phenomenology. Is it there all the time? What is it like? The inquiry is difficult, but he thinks it may offer clues to consciousness. Instead of focusing on rational thought alone, we should attend to these basal levels of experience. This can “bring us back to the basic reality that we are living — evolved creatures, trying to stay alive.” It “paints us back into nature.”

For Seth, this rooted feeling suggests a different view of the soul than the dominant vision in the Western world, which has long seen it through a Cartesian lens: an immaterial essence separable from the body, which arguably mirrors modern “cartoon dreams of a silicon rapture,” where our minds upload to the cloud. Older traditions tell another story: Greek psychē linked soul to breath, while Hindu Ātman pointed toward witnessing awareness beneath thought. Both are grounded in life, and they point to an ancient intuition about what we fundamentally are: “more breath than thought and more meat than machine.”

Consciousness research may help us see ourselves as less apart from nature and more woven into it — as “living creatures with more in common with other animals than with the statistical abstractions of AI.”

Mini Philosophy | Starts With A Bang | Big Think Books | Big Think Business