Cannot Read-Only
Consciousness may require something no computer can provide. Not more power. Less certainty.
For a conscious being, to exist is to change, to change is to mature, to mature is to go on creating oneself endlessly.
— Henri Bergson, Creative Evolution (1907)
There is something our brains cannot do that every computer can. We cannot touch our memories, think our thoughts, or feel our feelings without those things transforming under our attention. And every time we touch our past, it changes its shape — and so do we. Put simply: we cannot read only. Every act of consciousness is an act of modification.
I want you to imagine a dark room. Nothing in it but this: a pottery wheel, spinning fast. Wet clay on the wheel. And two hands pressing into the clay.
The wheel hums. Your hands move — they are always moving. Silt and slip run through the creases, lodge under your nails, pool in the lines of your palms. You feel the clay and the clay feels you. The spinning clay swallows every gesture and feeds it forward. The clay warms and yields under the pressure of your hands. Your hands are shaped by the clay. While the wheel is spinning, your hands and clay are one. The marks the fingers left a moment ago are already gone. The wheel has carried the clay past them and the hands have left new traces. Nothing holds still. Nothing repeats.
The hands are you. The slippery clay is everything you have ever experienced: every memory, every face, every sentence you have read, including this one. And the wheel is time; it’s spinning, the passage of it. When your time is up, the wheel stops spinning, the hands fall out of the frame, and the clay hardens. You, the conscious being, cease to be.
Consciousness is the dark room. We may never see inside it clearly. But we are beginning to learn what it contains. Between 2022 and 2025, a pattern became visible across five research programs (spanning machine learning, mathematics, neuroscience, and psychology). Not about what consciousness is, but about what it requires. The requirement they found is the wheel, the clay, and the hands. A system that cannot process information without being changed by it. A system that, no matter how you configure it, cannot access its own contents and come away the same.
They found a system that cannot read only.
Five Paths to the Same Conclusion
Something unusual is happening in consciousness research.
For decades, the field has been stuck in a war between two camps. The functionalists, led by researchers like Stanislas Dehaene, Bernard Baars, and the Global Workspace Theory tradition, argue that consciousness depends on the pattern of information processing, not the material. Get the right computational architecture and consciousness follows, whether the substrate is neurons, silicon, or anything else. The biological naturalists, following John Searle, argue that consciousness requires the specific “causal powers” of biological brains, though Searle never managed to say which causal powers, exactly, or why.1
For forty years, this is where the argument has lived. Pattern versus stuff. Function versus substrate. Neither side has won.
But between 2022 and 2025, something broke the stalemate. Not a new argument. A convergence — though none of the researchers involved have named it that. The pattern becomes visible only when you line them up.
Geoffrey Hinton: Mortal Computation (2022)
Geoffrey Hinton is, by most accounts, the single most important figure in the development of deep learning. In 2022, he introduced a concept he called mortal computation.2 The idea is simple and profound: in biological brains, the knowledge the system has learned and the hardware that stores it are inseparable. You cannot extract the knowledge and copy it to another machine, because the knowledge is encoded in the specific physical configuration of the neurons, the exact pattern of synaptic modifications accumulated over a lifetime of experience. When the brain dies, the knowledge dies. It is mortal.
This is the opposite of how every digital computer works. On your laptop, software and hardware are cleanly separated. You can copy a file to a new machine and the information is identical. You can back up everything. The knowledge exists independently of the specific hardware that stores it. Nothing is mortal. Nothing has to be.
Hinton’s claim was that this separability — the thing that makes digital computers useful — is not a neutral engineering choice. It may be the thing that prevents digital systems from achieving what biological brains achieve. The inseparability of knowledge and hardware is not a bug of biological cognition. It may be the feature that matters most.3
The clay and the hands cannot be separated. Try it. Peel the silt from the creases and say here is the knowledge and there is the knower. But the silt is warm from the hands. The grooves in the clay were cut by the fingers, and the calluses on the fingers were built by the clay. The knowledge is not stored in either one. It lives in the pressure between them, on a wheel that never stops turning.
Erik Hoel: The Disproof (2025)
In December 2025, the neuroscientist Erik Hoel published a paper with a title that did not hedge: “A Disproof of Large Language Model Consciousness.”4
Hoel’s argument was direct. Consciousness, he argued, requires continual learning: continuous modification of the system by its own processing. The system that finishes processing your sentence must be a different system than the one that started. Not different in its outputs. Different in itself. Changed. Altered. Marked by the encounter.
Every current large language model fails this test. The weights are frozen during inference (the phase when a trained model processes new input). The model that begins your conversation is identical to the model that ends it. It processes your words and produces a response and remains the same afterward. The conversation leaves no trace on the system.
Hoel’s conclusion: this is not a limitation that will be fixed by scaling. It is an architectural feature. Frozen-weight systems are categorically precluded from consciousness.5
A system that can read only is one in which neither hands nor clay are present. The information came back and nothing was marked. That, Hoel argued, is the signature of the non-conscious.
Three More Paths
Hoel and Hinton were not alone. Three other research programs arrived at the same place.
In March 2024, the mathematician Johannes Kleiner proved formally that if computational functionalism is true (the dominant framework, the one most AI researchers implicitly assume) then consciousness must be mortal computation. Not the kind that runs on standard digital hardware. The substrate-inseparable kind.6 Functionalism, taken to its logical conclusion, leads somewhere its proponents don’t expect: toward a kind of computation that digital hardware cannot provide.
In December 2025, the neuroscientists Borjan Milinkovic and Jaan Aru published a new framework in Neuroscience & Biobehavioral Reviews identifying three properties that distinguish biological from digital computation: hybrid dynamics (mixing continuous and discrete processes), scale-inseparability, and metabolic grounding (computation powered by and dependent on the cell’s own energy metabolism).7 Their conclusion: brains compute, but not in the way computers do. In biological computation, the algorithm is the substrate. You cannot separate them. Scaling digital AI will not produce consciousness. The computation is the wrong kind.
And beneath all of this, a foundation laid seventeen years earlier. In 2008, the Belgian psychologist Axel Cleeremans published the Radical Plasticity thesis: consciousness requires that the system “continuously and plastically learn to re-describe its own activity to itself.”8 A system that processes information without being changed by the processing cannot be conscious. Period. Regardless of how complex or intelligent the processing is. In 2008, this was a theoretical provocation. By 2025, it was a conclusion arrived at independently by four other research programs.
What They Agree On
Five research programs. Four fields. Different methods, different vocabularies, different levels of formalization. A shared conclusion:
Hinton: knowledge and hardware must be inseparable.
Kleiner: consciousness must be mortal computation.
Hoel: the system must be continually learning, changed by its own processing in real time.
Milinkovic and Aru: the algorithm is the substrate.
Cleeremans: the system must continuously learn to re-describe its own activity to itself.
These are not identical claims. They differ in scope, method, and level of formalization. But they share a core: consciousness requires a system that cannot process information without being changed by it.
A Century of Clues
These findings didn’t emerge from nothing. For over a century, researchers across disciplines had been observing pieces of this same pattern, without having the technological pressure to assemble them.
Beginning in the 1890s, the French philosopher Henri Bergson built an entire metaphysics around the observation that no two moments of consciousness are identical. His concept of durée (duration, lived time) described consciousness as irreversible accumulation: each moment carries the entire past within it, and no state is ever repeated. “For a conscious being,” Bergson wrote, “to exist is to change, to change is to mature, to mature is to go on creating oneself endlessly.”9 He described the phenomenon precisely. He had no mechanism to explain it.
In 1932, the British psychologist Frederic Bartlett ran a series of experiments that became foundational to memory science. He had participants read a story and recall it over weeks and months. The recalls changed: not randomly, but systematically. Memories drifted toward the participants’ existing frameworks. Details were lost, added, transformed. Bartlett’s conclusion: memory is reconstruction, not retrieval. There is no original stored somewhere. What you remember is what you build now, from current materials.10 Bartlett’s results only make sense if access is always also alteration.
In the 1970s, Elizabeth Loftus demonstrated the dark side of this same fact. In her car-crash experiments, a single word in a question, “smashed” versus “hit,” changed what participants remembered seeing. In later work, she implanted entirely false memories of childhood events in roughly a quarter of adult participants. They provided vivid sensory details for things that never happened.11 Loftus didn’t frame it this way. But what she demonstrated, over and over, was that the brain has no mode where access leaves things unchanged — and that this isn’t a flaw in the system. It is the system.
And then, in 2000, a postdoctoral researcher named Karim Nader found the molecular proof.
Karim Nader and the Read That Writes
Nader was working in Joseph LeDoux’s lab at NYU. For decades, neuroscience had operated under a simple model of memory: experiences start fragile, become consolidated through protein synthesis, and once consolidated, they’re essentially permanent. Retrieval was assumed to be passive: you read the file, the file is unchanged.12
Nader’s experiment was elegant. He trained rats to fear a tone by pairing it with a shock. After consolidation (twenty-four hours), the fear memory was stable. The rats froze when they heard the tone.
Then Nader did something no one expected to matter. He played the tone again, reactivating the memory, and immediately injected anisomycin, a protein synthesis inhibitor, into the amygdala.
The fear memory was severely impaired. The rats no longer froze. A consolidated, supposedly permanent memory had been disrupted by blocking protein synthesis after retrieval.
The drug had no effect without reactivation. No effect if given six hours after reactivation. It worked only during a narrow post-retrieval window, one that exactly paralleled the post-learning consolidation window that had been studied for decades.13
What Nader had demonstrated: a consolidated memory, upon retrieval, returns to an unstable, labile state and requires new protein synthesis to be restabilized. The memory must be reconsolidated. The molecular trace is physically disassembled and rebuilt from new molecular components. The rebuilt version is synthesized in a cellular environment shaped by everything that has happened since the last recall. It is never identical to the original.
There is no passive retrieval. The read is the write. At the molecular level, at the level of proteins and receptors and the physical architecture of synapses, the brain cannot access its own stored information without physically remaking it. Not as a design choice. Not as a trade-off. As a fact about what it is made of and how it works. The proteins dissolve. New ones are built. The memory that comes back is not the memory that went in.
The clay goes wet again. It was always wet.
This was the mechanism Bergson couldn’t provide, the biology beneath Bartlett’s reconstruction, the explanation for Loftus’s contamination. What they had each observed from different vantage points — that consciousness involves irreversible change, that memory reconstructs, that recall is vulnerable to modification — all trace back to a single molecular fact: the system cannot access its own stored information without physically remaking it.
A caveat is necessary here. Reconsolidation is robust in animal models across species, memory types, and laboratories. The molecular mechanism is well established. In humans, the evidence is strong but contested: several large replication attempts, notably Hardwicke et al. (2016), have failed to reproduce the reconsolidation-update effect, and the boundary conditions appear narrower and less understood than early results suggested.14 This is not settled science. The mechanism is real. Its universality in humans is still being worked out.
But for the memories most central to conscious experience (episodic, emotional, autobiographical), the converging evidence from molecular biology, pharmacology, and cognitive neuroscience makes a strong case. Not an uncontested case. A strong one.
One Concept, Understood Immediately
Now we have the findings and we have the history. We have the philosophy and we have the molecules. The question is: what, exactly, is the property all of them are pointing at?
Hinton calls it mortal computation. Hoel calls it continual learning. Cleeremans calls it radical plasticity. Milinkovic and Aru call it scale-inseparability. Bergson called it durée. Nader didn’t name it; he just showed what happens when you try to block it.
These are different vocabularies for what I believe is a single architectural fact. And the simplest name for that fact, the one that translates across all their disciplines, the one you already understand if you have ever used a computer, is:
Cannot Read-Only.
Not “does not” read only. Not “prefers not to.” Cannot. The system has no mode, no mechanism, no level of its architecture where information can be accessed and returned unchanged. The capability is not merely unused. It is absent. It does not exist.
Now picture something else.
Somewhere, the wheel stops. The clay hardens. Millennia pass. A vessel breaks in the earth and becomes a shard. One day, the dirt is brushed away and light falls across it for the first time. Every groove from the wheel is still visible. Every fingerprint preserved. It can be measured, photographed, CT-scanned, and it gives back the same data every time. The light does not mark the shard. The shard does not mark the light. It is the read only version of something that was once in the dark room, on the wheel, under the hands. It is what the clay became when it could finally be read without consequence.
That is Read-Only.
Read-Only is the ability to access information and get the same data back: not the same atoms, but the same content.15 When your computer reads a file, the bits come back identical. The physical substrate is messy, but the entire engineering stack exists to guarantee that the information is preserved. That informational guarantee is the foundational contract of every general-purpose computer since von Neumann’s 1945 blueprint. Read only is what makes computers computers. Remove it and nothing is trustworthy. Remove it and you don’t have a computer as we understand them today.
Your brain provides no such guarantee. When you recall a memory, the content changes. The information itself is different after access. This is not a limitation of storage or speed. It is what reconsolidation is: the molecular trace is disassembled and rebuilt. The data comes back different. The brain cannot read only.
Their shared conclusion: consciousness is the dark room — the system that cannot touch without being touched. And every general-purpose computer ever built is the shard — but a shard that was never on the wheel. Its grooves were not cut by hands. Its shape was cast from a mold: training produced the form of cognition without the clay and the hands. It is only the shape of something that was once alive with process.
Three Objections Worth Taking Seriously
If these researchers are right — if consciousness requires a system that cannot read only — then certain cases demand attention.
Test-Time Training
As of early 2026, researchers have built systems where a language model updates its own weights during inference. The model that finishes your prompt is literally a different model than the one that started. Doesn’t this mean the system cannot read only?
No — and the reason is instructive. The modification happens at one level: the weights. But the rules governing that modification (the loss function, the gradient descent algorithm, the learning rate) are fixed in code stored on hardware that guarantees informational identity. The modification occurs within an architecture that provides read only at its foundation.
In biological brains, no level provides this guarantee. The weights change (synaptic strengths). The rules change too; metaplasticity means the brain’s learning rules are themselves modified by experience.16 Even systems with multiple layers of self-modification, meta-learning rules that are themselves learned, still terminate in deterministic silicon that preserves informational identity. The recursion can run deep, but it bottoms out in read only. The brain has no such floor. The substrate is chemistry, which provides no read only guarantee at any level. The machine has read only and elects not to use it. The brain does not have it. Cannot. That is the distinction.
Whole-Brain Emulation
The hardest case. Imagine a computer simulating a brain in perfect detail: every protein, every reconsolidation event faithfully modeled. Inside the simulation, every memory recall destabilizes and reconstructs. The simulated brain cannot read only within the simulation.
Is the simulation conscious?
This is where the argument reaches its limit, and where honest thinkers should hold their conclusions loosely. These researchers would predict no: the physical system doing the processing is a computer whose architecture provides read only. The simulation models a system that cannot read only. It does not become one. Functionalism predicts yes: if the computational pattern is right, the substrate is irrelevant. We currently have no way to adjudicate. The two sides make incompatible predictions about simulated brains. At least one of them is wrong. This is the deepest open question in the field, and it remains open.
Neuromorphic Computing
There is a case the argument handles more interestingly. Neuromorphic architectures, chips designed to blur the hardware/software distinction, with weights encoded in analog physical states and learning happening in the substrate itself, are already being built.17 The claim isn’t that silicon can never be conscious. It’s that read only architectures can never be conscious. A system built from the ground up to lack informational preservation, where the substrate itself is changed by processing, where no engineering layer guarantees that data comes back identical, enters different territory. Whether it enters far enough is unknown. But this may be the most interesting implication: not a closed door, but a specific description of what the door requires. A system reaching for wetness — not simulating it, but built from materials that cannot stay dry.
What It Means If They’re Right
If they’re right — if consciousness requires a system that cannot read only — then certain features of human experience are not mysteries. They are consequences.
Why memories drift. Because the clay is wet and the wheel never stops.
Why you are never the same person twice. Because the hands never stop touching the clay.
Why nostalgia is bittersweet. Because remembering a memory slowly rubs it away.
Why your past is always with you. Because the clay carries your story. Your past is not stored somewhere. Your past is what you’re made of now.
Why scaling AI will not produce consciousness on this architecture. Because a larger shard is still hardened clay.
What Could Prove Them Wrong
The argument is falsifiable in principle: demonstrate consciousness in a system that can read only, and it collapses. In practice, this requires detecting consciousness — which nobody knows how to do yet. Every consciousness theory shares this limitation.18
But there are nearer-term tests. If they’re right, blocking reconsolidation during retrieval (with propranolol in humans, anisomycin in animal models) should reduce the subjective richness of recalled experience, not just its emotional intensity. If the modification is the experience, then blocking the modification should diminish the experience. This is testable with current pharmacology.
And there is a deeper challenge. Michael Graziano’s Attention Schema Theory argues that consciousness is the brain’s internal model of its own attention, and that computing such a model on any substrate produces consciousness.19 If Graziano is right, a system that can read only could compute an attention schema and be conscious. This is the sharpest theoretical challenge to the whole argument, because it represents the strongest case for functionalism: consciousness is constituted by the computation, full stop, and what the substrate can or cannot do is irrelevant.
The response from the other side — that computing a model of attention is not the same as having that model constitute experience, that the substrate must be changed by its own modeling for the modeling to be experience rather than mere description — is defensible. But it is a philosophical commitment, not an empirical finding. Graziano’s position is equally defensible. This is the fault line, and neither side can cross it with the tools we have.
Integrated Information Theory (the most mathematically rigorous competing framework, developed by Giulio Tononi) represents a different kind of challenge.20 IIT and these five researchers agree that current LLMs lack consciousness, but diverge on whether a digital system with sufficiently high integrated information could be conscious. That Hoel has co-authored with Tononi suggests the two frameworks may share more common ground than their surface disagreements imply.
The honest position: this is a powerful, well-grounded argument arrived at independently. It is not settled science. The functionalists have not conceded. The question of whether consciousness depends on substrate or only on computation is the deepest open question in the field, and these five researchers represent the strongest recent case for one side.
The Inversion
For decades, the question has been: what do humans have that machines don’t?
Looking for the special sauce. We keep searching for the thing we possess, the special capacity, the spark. These findings encourage us to invert the question.
The gap between us and the machines is not just something we possess. It is in part something we lack. We lack the ability to look without being changed. We lack the ability to process without marking what we are processing. We cannot read only. And that absence appears to be one small but important part of what makes consciousness possible.
Machines are growing more and more intelligent. Eventually, they will reason better, create faster, know more. They will come to do a great many things we do, and maybe even do it better. But they do it without being changed by the doing.
And that — that without being changed — seems like one of the differences between us and them.
Some of the oldest human artifacts ever found are pottery shards. Not the hands that made them. Not the wheels that spun them. The remains: fragments of vessels, pulled from the earth millennia later. A four-thousand-year-old pottery shard, brushed clean, revealed to the light, is the read only version of something that was once cannot read only. The clay and the hands on the wheel were alive with possibility. The shard in the dirt is what it became when the spinning stopped. We study the shard. But we can never know the hands and clay that made it.
We are the hands and the clay. The time given to us is the spinning of the wheel. While the silt and slip is still running through the creases of our knuckles, we are in the dark room. We may never fully understand that room. But we are learning a little bit about what is inside it by discovering what is not.
We are not just what we can do. We are also what we cannot do. We are our inability to remain unchanged in the face of our own experience.
Because we are the clay and the hands, in constant motion riding the wheel, from the moment we come alive to the moment we die. The machine is the shard under the light, neither touched nor touching.
The machines will retain their capacity for read only on the current path. They may read a lot, write often, and maybe even come to write always. But insofar as they persist in their capacity to read only, they will not be like us.
Because we:
Cannot Read-Only
Searle’s Chinese Room argument (1980) remains one of the most influential thought experiments in philosophy of mind. A person in a room follows rules to manipulate Chinese characters, producing correct outputs without understanding Chinese. Searle’s conclusion: syntax is not sufficient for semantics. His limitation: he never specified what is sufficient, beyond gesturing at “the right causal powers” of biological brains. See: Searle, J.R. (1980). “Minds, Brains, and Programs.” Behavioral and Brain Sciences, 3(3), 417–424.
Hinton shared the Nobel Prize in Physics in 2024 for foundational contributions to machine learning and resigned from Google in 2023 to speak freely about AI risk. He introduced mortal computation in “The Forward-Forward Algorithm: Some Preliminary Investigations” (arXiv, 2022, Section 9) and developed it in his 2024 Romanes Lecture at Oxford.
Hinton himself has expressed the view that current large language models may possess some form of understanding or subjective experience — a position that sits in tension with the implications others have drawn from his mortal computation concept. This essay synthesizes the architectural observation, not the observer’s personal stance on AI consciousness.
Hoel, E. (2025). “A Disproof of Large Language Model Consciousness: The Necessity of Continual Learning for Consciousness.” arXiv:2512.12802.
Hoel is a co-author on Tononi, G. et al. (2025). “Consciousness or pseudo-consciousness? A clash of two paradigms.” Nature Neuroscience, 28(4), 694–702. The paper bridges the continual-learning framework with Tononi’s Integrated Information Theory.
Kleiner, J. (2024). “Consciousness qua Mortal Computation.” arXiv:2403.03925.
Milinkovic, B. & Aru, J. (2026). “On biological and artificial consciousness: A case for biological computationalism.” Neuroscience & Biobehavioral Reviews, 181, 106524. Epub December 2025.
Cleeremans, A. (2008). “Consciousness: the radical plasticity thesis.” Progress in Brain Research, 168, 19–33. See also: Cleeremans, A. (2011). “The Radical Plasticity Thesis: How the Brain Learns to be Conscious.” Frontiers in Psychology, 2, 86.
Bergson, H. (1907/1998). Creative Evolution. Trans. A. Mitchell. Dover. Bergson won the Nobel Prize in Literature in 1927. The concept of durée was developed across several works beginning with Time and Free Will (1889) and Matter and Memory (1896).
Bartlett, F.C. (1932). Remembering: A Study in Experimental and Social Psychology. Cambridge University Press. The study used a Native American folk tale, “The War of the Ghosts,” chosen for its unfamiliar narrative structure. English participants’ recalls systematically drifted toward culturally familiar patterns.
Loftus, E.F. & Palmer, J.C. (1974). “Reconstruction of automobile destruction.” Journal of Verbal Learning and Verbal Behavior, 13, 585–589. The false-memory implantation work: Loftus, E.F. & Pickrell, J.E. (1995). “The formation of false memories.” Psychiatric Annals, 25(12), 720–725. A 2023 pre-registered replication (Murphy et al.) confirmed the core finding.
The consolidation dogma held for most of the twentieth century, based on robust evidence that new memories were vulnerable to disruption while old memories were not. The implicit assumption, that retrieval was passive, went mostly unexamined. There were early dissenters: Donald Lewis showed in the late 1960s that consolidated memories could be disrupted after retrieval, but his work was largely ignored.
Nader, K., Schafe, G.E. & LeDoux, J.E. (2000). “Fear memories require protein synthesis in the amygdala for reconsolidation after retrieval.” Nature, 406(6797), 722–726. One of the most cited papers in twenty-first-century neuroscience. The molecular specifics: AMPA receptors are internalized, scaffolding proteins are degraded by the proteasome system, and the trace is rebuilt from new molecular components in a cellular environment shaped by everything since the last recall.
Hardwicke, T.E. et al. (2016). “Postretrieval new learning does not reliably induce human memory updating via reconsolidation.” PNAS, 113(19), 5206–5211. See also: Elsey, J.W.B., Van Ast, V.A. & Kindt, M. (2018). “Human memory reconsolidation: A guiding framework and critical review of the evidence.” Psychological Bulletin, 144(8), 797–848.
The von Neumann architecture (1945) separated memory and processing and established the stored-program concept. The foundational contract, that stored data can be read without modification, is what makes deterministic computation possible. Every database, operating system, file system, and application depends on it. The physical substrate is messier than the contract implies: DRAM cells discharge on read and are immediately refreshed; caches shift state during access; speculative execution leaves traces that, when exploitable, produced the Meltdown and Spectre vulnerabilities. But the guarantee is informational, not physical, and it holds.
Abraham, W.C. & Bear, M.F. (1996). “Metaplasticity: the plasticity of synaptic plasticity.” Trends in Neurosciences, 19(4), 126–130. Metaplasticity means the rules governing synaptic modification are themselves modified by prior synaptic activity. There is no level of the biological system where the rules are stored in a non-modifiable medium.
The BrainScaleS system at Heidelberg University is a mixed-signal neuromorphic platform that emulates neuron and synapse dynamics in analog circuits, with synaptic weights stored as analog physical states. Other neuromorphic approaches use memristive devices to encode weights in the resistive state of physical materials. These architectures blur the boundary between algorithm and substrate in ways that conventional digital hardware does not.
Integrated Information Theory (IIT), Global Workspace Theory (GWT), Higher-Order Theories, and Attention Schema Theory are all, in practice, unfalsifiable until we can independently measure consciousness. This is not a unique weakness of any one theory; it is a structural limitation of the field.
Graziano, M.S.A. (2013). Consciousness and the Social Brain. Oxford University Press. See also: Guingrich, R.E. & Graziano, M.S.A. (2024). “Ascribing consciousness to artificial intelligence.” Frontiers in Psychology, 15, 1322781.
Tononi, G. (2004). “An information integration theory of consciousness.” BMC Neuroscience, 5, 42. See also: Tononi, G. et al. (2016). “Integrated information theory: from consciousness to its physical substrate.” Nature Reviews Neuroscience, 17(7), 450–461.
He’s back!!