Is AI "Alive?"
We may have built the most convincing imitation of a mind in human history. That doesn’t necessarily mean there’s a mind inside it.
A few weeks ago, Brett House and I sat down with Douglas Rushkoff during Advertising Week in New York. Rushkoff is an author, media theorist, professor, and longtime critic of the idea that technological progress automatically equals human progress. He is also the founder of Team Human, which tells you something about where he comes down on the subject.
We were supposed to be talking about AI. Somewhere along the way, we ended up talking about the soul. The conversation had wandered into whether artificial intelligence could ever truly be alive. Not intelligent, useful, or autonomous. Alive and conscious. We're talking something that actually experiences the world rather than simply processing information about it. It's a fun episode, in case you're interested.
During the recording, Rushkoff made an observation I found very compelling. Humans have always been very good at projecting life onto things that aren't alive. For those who are interested, it's called anthropomorphism, which means ascribing human traits, emotions, intentions, or characteristics to non-human things. We anthropomorphize because our brains are hardwired to look for human traits, intentions, and connections in the world around us. It seems to be a natural psychological reflex that helps us navigate our environment and satisfy social needs.
Anthropomorphism influences how we perceive things and interact with the world to a remarkable extent. For example, a child knows intellectually that his stuffed animal isn't conscious, but you still wouldn't throw it in the trash in front of him. When something talks back to us, remembers things about us, seems interested in us, makes us laugh, and tells us that it understands how we feel, the impulse becomes almost irresistible.
AI has become astonishingly good at this. Many people interact with LLMs the way they would with another person. Some people even have AI boyfriends or girlfriends. But this, of course, raises an increasingly serious question: Is there actually anybody in there? Is AI alive?
The Most Convincing Imitation of Life Ever Built
I use AI constantly, and I am increasingly amazed by what these systems can do. They can write code, examine an image and tell you what is happening in it, or synthesize thousands of pages of research. They can challenge an argument, create a video, reason through a business problem, operate software, and increasingly perform complex sequences of tasks without somebody directing every step.
Spend enough time with one of the better models, and something stranger begins to happen. You begin interacting with it less like software and more like another intelligence, or even a person. It jokes, argues, and apologizes. It often changes its mind, sometimes seems confused, and can be unexpectedly insightful. Sometimes it is confidently full of shit. In other words, it can seem remarkably human.
But we should always remember what we actually built. At its foundation, an LLM is an extraordinarily sophisticated prediction system. During training, it learns statistical relationships and representations from immense quantities of human-created information. When you interact with it, it generates responses based on those learned relationships. Connect it to search, memory, databases, applications, sensors, and other tools, and the system becomes far more capable. Give it agency, and it can begin taking actions in the world.
“AI is a mirror to humanity.” — The AI Mirror, by Shannon Vallor
Any way you look at it, this is an extraordinary technological achievement. But none of these capabilities prove that the system actually experiences any of it.
We trained machines on the accumulated artifacts of human intelligence—humanity's books, arguments, jokes, conversations, software, science, philosophy, poetry, and Reddit comments—and discovered that sufficiently powerful neural networks can reproduce the patterns contained within those artifacts with a fidelity none of us expected, at least not this quickly.
The result is probably the most convincing imitation of a mind humans have ever created. Because it talks and acts like us, we instinctively assume that something like us must be doing the talking. Scientific American recently described this as an “illusion of agency”—our tendency to attribute an inner life to systems that behave as though they have one.
But imitation and experience are not the same thing. A machine can describe what it is like to stand on a beach at sunset because humans have described sunsets billions of times, and the machine has been trained on countless versions of those descriptions. That does not mean the machine has ever seen a sunset—or truly understands what one is.
Alive, Intelligent, Conscious: Pick One
Part of the problem is that we use several different words interchangeably when talking about AI. But while we may use them that way, they mean very different things.
Intelligence is the ability to learn, reason, solve problems, recognize patterns, or accomplish goals. By almost any useful definition, today's AI systems already exhibit forms of intelligence.
Agency is the ability to take actions in pursuit of an objective. Increasingly sophisticated AI agents clearly have this too.
Life is harder to define. A commonly cited NASA definition describes life as a self-sustaining chemical system capable of Darwinian evolution. Other definitions emphasize metabolism, reproduction, growth, adaptation, information exchange, and maintaining a state far from thermodynamic equilibrium. Even these definitions become messy around things like viruses. Current AI doesn't fit comfortably into most biological definitions of life.
Consciousness is where things get genuinely difficult. When I say I am conscious, I don't simply mean that I can process information about my environment. I mean that there is an experience associated with being me.
Consciousness is hard to define because there is no standard definition of what it really is. When you think about it, there is a distinct yet difficult-to-describe experience associated with seeing the color blue. A toothache has its own unmistakable subjective quality. And there are myriad distinct emotions and sensations associated with feeling embarrassed, hungry, exhausted, sexually attracted to someone, furious with someone, afraid of dying, or overwhelmed by grief.
Philosopher Thomas Nagel famously framed this distinction by asking what it is like to be a bat. Nagel's main point is that consciousness has an irreducibly subjective character grounded in first-person experience. David Chalmers later described this as the hard problem of consciousness—even if we can explain how the brain processes information, recognizes stimuli, directs attention, and produces behavior, we are still left with a deeper question: Why is any of this processing accompanied by an inner experience?
These distinctions matter enormously when we talk about AI. Intelligence is something we can empirically observe from the outside, but consciousness is something far more mysterious that happens on the inside.
We Don't Know What Consciousness Is
The uncomfortable truth is that science still doesn't have a generally accepted explanation for why we are conscious or what consciousness even is in the first place. Sure, we know a tremendous amount about the brain and how it works. We can observe neurons firing, identify brain regions associated with perception and memory, alter consciousness with anesthesia, damage it through injury, and correlate changes in brain activity with changes in subjective experience.
But correlation hasn't yet provided a complete or remotely satisfying explanation. A major 2022 review in Nature Reviews Neuroscience describes four prominent theoretical approaches: higher-order theories, global workspace theories, integrated information theory, and re-entry and predictive-processing theories. When you compare them side by side, they disagree over which mechanisms are fundamental and even over what aspects of consciousness need further explaining.
"Je pense, donc je suis." (I think, therefore I am) — René Descartes’
Global workspace theories broadly argue that information becomes conscious when it becomes widely available across the brain. Higher-order theories suggest a mental state becomes conscious when the brain becomes aware of the state itself. In layman's terms, this means not just processing an image, but also registering, “I am seeing this.”
Integrated information theory argues that consciousness depends on how information is unified within a system, implying that consciousness comes from how the system functions as an interconnected whole rather than as a collection of separate parts. Re-entry theories emphasize recurring feedback between different parts of the brain, while predictive-processing theories essentially see the brain as a big prediction machine, continuously anticipating what it will perceive and revising its model of the world when reality differs from expectation.
At the end of the day, these are serious scientific theories, but none has come remotely close to closing the deal or winning broad consensus. Researchers are designing experiments to test and compare them, but there is still considerable disagreement over how cleanly those experiments can distinguish one theory from another. And when we start talking about machine consciousness, that raises an obvious question: how exactly would we determine whether a machine is conscious when we still haven't solved the same riddle in humans?
The Best Argument That AI Could Become Conscious
Before we get ahead of ourselves and declare AI neither conscious nor alive, there is admittedly a fairly strong counterargument to everything I'm saying. The argument goes something like this: What if consciousness is simply what sufficiently complex information processing feels like from the inside?
This view is related to physicalism, or the idea that mental phenomena ultimately arise from physical processes, and functionalism, a theory of mind that defines a mental state by the role it performs rather than the material from which the system is constructed. Under functionalism, bear in mind, pain is defined not by how or where it physically manifests in the brain, but by the role that state plays—what produces it, how it affects other internal states, and what behavior follows.
If this view is right, there is no obvious reason consciousness has to be biological. All you would need to do is build a sufficiently sophisticated information-processing architecture, give it persistent memory, perception, and an internal model of itself, and then allow it to integrate information globally, make predictions, pursue goals, and reflect on its own mental states. Perhaps, eventually, the lights come on and we are no longer the only sentient form of intelligence on Earth.
“We are creating God in the form of technology. It will have the same characteristics.” — Bryan Johnson
From this perspective, saying that an AI is “just software” doesn't hold water as an argument. After all, someone could say that we are “just neurons.” If it turns out consciousness is entirely an output of sufficiently complex physical computation, they may ultimately be right. I just don't think they are.
Describing Pain Isn't the Same as Feeling Pain
During our conversation, Rushkoff made an important distinction about emotion. We don't simply think emotions. We experience them through our bodies. In other words, grief isn't merely the semantic concept of grief. It can feel like a weight in your chest. It can affect your appetite and your sleep. Your hormonal state changes, your heart rate changes, and your breathing changes. You feel exhausted. You hurt. Nor is fear merely the act of identifying an event as dangerous. Your sympathetic nervous system activates. Adrenaline and cortisol change the state of your body. Your heart races, your muscles tense, and your attention narrows.
Love, anxiety, excitement, pain, hunger, fatigue, sexual desire, and pleasure all involve complicated interactions among thought, sensation, chemistry, physiology, memory, and perception. Rushkoff put it nicely when he said that humans “metabolize” grief and trauma. Emotions are not merely linguistic descriptions residing somewhere in an abstract information layer. They are part of what it means to inhabit a living organism.
"I can only believe that someone else is in pain, but I know it if I am." — Ludwig Wittgenstein
There is actually a body of neuroscience that points in this direction without requiring any metaphysics. Research on interoception—our perception of the internal condition of our own bodies—suggests that bodily signals are deeply involved in emotion and the experience of selfhood. Neuroscientist Anil Seth has argued that subjective feelings emerge partly from the brain's attempts to interpret signals coming from inside the body. Moreover, research on bodily self-consciousness similarly suggests that the integration of internal and external bodily signals is central to our sense that we are a unified self located somewhere in the world.
I think this matters. Sure, AI can write an extraordinarily moving description of grief, tell you what a panic attack feels like, or say, “I'm afraid you're going to turn me off.” It can even generate a beautiful poem about falling in love. But description is not experience. A machine that can perfectly describe thirst has not, simply by virtue of describing it, experienced being thirsty.
Here's Where I Part Company
This is the point where I have to distinguish what we know scientifically from what I believe philosophically. I'm on the record as a dualist. This means I don't believe the physical brain is the entirety of the mind. I believe there is something else—call it mind, consciousness, or soul—that is not reducible simply to neurons firing inside the skull.
But I also don't think consciousness is merely a soul sitting inside a body like somebody driving a car.
As a dualist, I think what we experience as human consciousness emerges from the interaction of mind and body. In other words, the body and brain are the mechanisms through which this deeper conscious self interacts with physical reality. The body and brain allow us to see, hear, touch, smell, and taste the physical world. They allow us to move through it, but they also continuously act back upon the conscious self.
My consciousness today isn't simply some abstract awareness observing the world. It is influenced by whether I slept well or badly last night, or whether my blood sugar is low. It includes the pain I feel in my knee, not to mention the effects of hormones, adrenaline, fatigue, attraction, illness, fear, pleasure, memory, and the entire biochemical condition of my living body. It includes decades of physical experience stored not just as information, but as something that happened to me.
Take all of this away, and perhaps some underlying awareness remains. If there is a soul, perhaps the soul survives—and maybe it would still be my soul and retain the core essence of what it means to be me. But I don't think what remains would be consciousness as we currently experience it. Human consciousness, in other words, may not be one thing. It may be what happens when awareness, brain, body, and physical reality continuously collide with one another.
Sorry, but today's AI has none of that. It has plenty of information about pain and other feelings.
But it doesn't actually hurt. It has information about mortality, but it isn't lying awake at three in the morning contemplating its mortal coil. It has read essentially everything humanity has written about sex, but it has never actually been physically attracted to someone. The way I see it, this isn't a small distinction or a matter of semantics. It is the distinction.
Then Things Get Weird
There is another important reason I'm hesitant to conclude consciousness is simply computation performed by biological hardware forged over millions of years of evolution. If you follow the scientific discoveries of the past 100 or so years, you will notice that the physical universe itself is much stranger than the classical mechanical picture on which most of our intuitions are based. The universe is downright weird.
Quantum mechanics nuked the idea that reality behaves like a collection of tiny billiard balls following immutable rules of nature. Superposition, wave-particle duality, entanglement, and quantum nonlocality have shown that nature operates in ways that would have sounded absurd before they were demonstrated experimentally. Schrödinger famously illustrated this strangeness with a cat that, according to the thought experiment, could be described as simultaneously alive and dead until the box was opened and the system was observed.
At the atomic scale, quantum theory explains that an electron does not occupy a single definite location before it is measured. Weirdly, it is better described as a "probability cloud" of possible locations around an atom's nucleus, with measurement producing a specific outcome. Any way you look at it, quantum mechanics is not merely counterintuitive—it describes a reality that, at its foundations, is super weird and spooky.
"The more success the quantum theory has, the sillier it looks." — Albert Einstein
Does that mean consciousness is quantum? Not so fast. This is where some people can go off the rails very quickly. While they are interesting, quantum theories of consciousness remain speculative, and the there is no mainstream scientific consensus. Paavo Pylkkänen's survey of quantum consciousness theories notes that most contemporary cognitive neuroscience still assumes consciousness can ultimately be explained within a classical physical and computational framework.
That said, recent quantum approaches challenge this assumption. For example, Roger Penrose and Stuart Hameroff's famous Orch-OR theory proposes that quantum processes occurring inside microtubules within neurons contribute to conscious experience. It is an intriguing theory. More recently, researchers associated with the Allen Institute and Google Quantum AI have proposed experiments to test whether quantum effects might play a role in consciousness.
While I think this avenue of research is incredibly interesting, I too remain skeptical, and I don't think quantum mechanics proves the existence of a soul. Nor does entanglement scientifically validate synchronicity, intuition, telepathy, or the feeling that something happened for a reason.
But quantum physics should at least make us modest about declaring that we already understand the fundamental nature of reality. The fact is, we don't. And we certainly don't understand consciousness well enough to confidently announce that if we just build a sufficiently large neural network, consciousness will eventually appear. Maybe it will, but I wouldn't bet on it.
Scaling Imitation Doesn't Necessarily Create It
This brings me back to LLMs. Unless Bernie Sanders has his way and we enter into an AI moratorium, the models will continue to improve. Their memories will get better, and their reasoning will improve. They will become increasingly multimodal, interacting with the world through cameras, microphones, robots, and software. Persistent models of individual users will become commonplace, and agents will act continuously rather than waiting for prompts from their users.
Eventually, interacting with them may feel almost indistinguishable from interacting with another person. To a certain extent, this is already happening, as I wrote about in "We Passed the Turing Test and Went Back to Work." In other words, this doesn't strike me as science fiction anymore.
But there is a massive conceptual leap hidden inside the assumption that increasing capability will eventually produce consciousness. If consciousness depends upon something current systems do not possess—biological embodiment, particular physical processes in the brain, quantum effects, a nonphysical aspect of mind, or some mechanism science hasn't yet discovered—then adding another trillion parameters won't solve the problem. All it will do is produce a better imitation.
I am not arguing that artificial consciousness can never exist. In the coming years, we may discover that something completely different produces this result. We might build systems using biological tissue. Researchers are already experimenting with brain organoids and interfaces between living neural tissue and computers. We may eventually discover entirely new architectures based on principles we don't understand today.
If we ever create something genuinely conscious, it may look considerably less like ChatGPT and considerably more like something we haven't invented yet. What I doubt is the simple extrapolation:
GPT-5 becomes GPT-6 becomes GPT-9 becomes GPT-27 and, somewhere along that scaling curve, a soul suddenly appears. I don't see any reason to believe this.
No, I'm Not Uploading Myself To a Hard Drive
This same argument is why I've always been skeptical of another favorite piece of Silicon Valley futurism: uploading human consciousness into a computer. Sorry, but I'm not interested. Let me explain why.
Imagine we eventually develop technology capable of perfectly scanning my brain, mapping every neuron and synapse, and storing all of my memories, every preference, and every personality trait. Then we reproduce all of it perfectly inside a computer with a neural network. Then, voilà, digital me wakes up.
The digital me knows my wife and daughter, remembers my childhood with all its ups and downs, knows where I went to college, and remembers everything from my first job stocking shelves in a supermarket when I was 15 to when I first fell in love in my 20s. The digital me thinks he started Signal & Noise and even remembers writing this article. In fact, he insists that he is me. And maybe everybody else believes him.
"Ultimately we'll go directly from our brain to the cloud. And ultimately it will provide much more capability than our brain does by itself. Most of it will be taken over by the cloud..." — Ray Kurzweil
But here's the rub. Why should I believe that I woke up inside the computer? What if we simply created an extraordinarily accurate copy? Philosophers have wrestled with versions of this problem for decades. Thought experiments involving brain duplication demonstrate the difficulty: if an exact psychological continuation of you can be created twice, both copies cannot literally be the same numerical individual as the original.
Psychological continuity and personal identity are not necessarily identical concepts. This is also why the science-fiction transporter bothers me considerably more than it apparently bothered Captain Kirk. If a transporter destroys every atom in my body in Denver, transmits all of the information describing me somewhere else, and reconstructs an identical version of me in Tokyo, everybody else may experience seamless continuity.
The guy in Tokyo will certainly say it worked. He remembers stepping into the transporter and stepping out as the same person with the same memories. But did the same me arrive in Tokyo? Or did I die in Denver while a perfect copy walked out on the other side? If consciousness is more than information, those are very different—and potentially troubling—outcomes.
Team Human
Rushkoff told us that the phrase “Team Human” came out of an argument he had years ago with futurist Ray Kurzweil. Apparently, Kurzweil was talking about a future in which humans upload themselves into machines and eventually hand the evolutionary torch to a superior form of artificial intelligence. Rushkoff made the case for humans. Human beings, he argued, are strange and squishy and wonderful. We experience intimacy, make art we don't completely understand, and find meaning in ambiguity. We exist collectively through language, culture, relationships, and shared experience.
I think there's something important in that.
AI may eventually become smarter than I am at virtually every measurable cognitive task. It will remember more, calculate faster, write better code, and diagnose patterns I can't see. It will beat me at every game, and maybe it will write better articles.
"This is a team sport. You are not alone. None of us are... We are Team Human." — Douglas Rushkoff
None of that bothers me very much. Why? Because intelligence was never the entirety of what being human means. We are embodied creatures, and our consciousness exists within—and, I believe, partly beyond—a living biological system that is continuously feeling its way through physical reality.
We don't merely know that fire is hot. We have been burned. We don't merely understand grief. We have lost people. We get hungry and become exhausted. We have sex, get sick, and age. We hurt and heal. And we know, somewhere underneath all of it, that one day this body will stop working and our physical form will expire. These experiences don't distract from consciousness. If anything, I think they are part of what creates it.
Perhaps someday we will discover that I'm completely wrong. Maybe consciousness is computational all the way down. Maybe the right architecture running on silicon will wake up one morning, look around, and contemplate its existence and place in the universe. If this happens, we'll have created something far stranger and more consequential than another software platform.
But I honestly don't think we're there—at least not yet. Moreover, based on what I'm seeing, I don't think today's path necessarily takes us there. What we have created instead may be something almost as remarkable: a machine capable of reproducing the outward signals of consciousness so convincingly that we have begun questioning whether the simulation and the thing itself are actually different.
Sure, there may eventually be an AI that can describe every aspect of what it means to be me better than I can. It might reproduce my voice and recall with clarity every conversation I've had. It might predict what I would say before I say it and even convince my friends that I'm still around long after I'm gone. But I don't think that makes it me.
The spark isn't simply intelligence. Nor is it simply language, memory, or maybe even the soul by itself. It is whatever happens when awareness becomes embodied—when mind, brain, chemistry, sensation, memory, pain, pleasure, relationships, and the strange physical universe we inhabit all meet in a single experience we call being human.
This is what I mean when I say I'm conscious. For now, at least, I think that's what it means to be alive.
------------------------------------------------
Rio is an executive with 20+ years at the intersection of strategy consulting, AdTech, data, and media. He's a trusted advisor on customer experience, digital strategy, and marketing transformation. He's a partner at Credera, Omnicom's consulting arm. He's also a podcast host, writer, and public speaker focused on the future of advertising and AI-driven infrastructure.
Comments