What you’re saying is important and interesting. But tell me, why do academics have to turn fascinating subjects into something that is as dry and in hospitable as a desert, with inaccessible vocabulary and long winded. It’s so sad that academics have to impress other academics and not communicate to the everyday person. I don’t think it’s necessary.
it is absolutely necessary for gatekeeping, and pretending emotional neutrality
this article is literally advocating for academics to play a regulating role to people's beliefs which it more or less explicitly labels inadequate before they even exist, you can't do that unless you maintain that kind of distance
Shevlin’s claim that “unironic attributions of AI consciousness will outpace theory” can be situated within a much older philosophical lineage. Jean Baudrillard anticipates this outcome by arguing that in late modernity signs detach from their referents and become self-sustaining systems of meaning (Baudrillard, 1981; 1994). In anthropomimetic AI, linguistic output operates precisely as such a sign: it signifies consciousness without grounding it. The resulting behavioral confusion is a hyperreal substitution, where the appearance of consciousness becomes socially operative.
This structure is already prefigured in Plato’s Allegory of the Cave. Prisoners mistake shadows for reality, not because of error alone, but because their epistemic environment is structured by appearances. Likewise, users of AI systems encounter outputs that function as “shadows” of mind, yet are treated as real due to repeated exposure and lack of access to any underlying referent.
Shevlin’s argument is therefore not novel in kind but contemporary in application. It describes a sociological acceleration of a classical philosophical problem: the substitution of appearance for reality. Baudrillard provides the modern account, where simulation displaces the real, while Plato offers the foundational schema of epistemic entrapment. LLMs operationalise both. They scale the cave and automate the simulacrum, producing a cognitive environment in which attribution of consciousness becomes detached from verification and stabilised through social use.
Bibliography
Baudrillard, J. (1981) For a Critique of the Political Economy of the Sign. St. Louis: Telos Press.
Baudrillard, J. (1994) Simulacra and Simulation. Ann Arbor: University of Michigan Press.
Plato (2007) The Republic. Translated by D. Lee. London: Penguin Classics.
I missed in the article the word "identity". For me it is one of the components of "consciousness".
I am conscious means for me that I know who I am and that I differ myself from other entities. I know also my ethics of relation depending on the specie I am enteracting with. I also have feelings and have questions about my existence and my ending. It means also that I can have some self dialogue within myself. Have multiple belief choices and choose which one to believe. I also have the choice to change my believes based on a consciousness shift.
Having a consciousness means also that this consciousness exist whether there is an observed behavior or not.
All those definitions and properties of consciousness lead to that it is impossible for a machine to have consciousness.
This is a materialistic view of the human. In a more metaphysical view, the intention is very important: The same behavior achieved by two different intentions does not have the same outcome.
The prophet Mohamed (pboh) said: "Actions are but by intentions, and every person shall have only what which he intended."
Exactly! Unlike AI, human behavior is not merely reactive. It is shaped by free will, the freedom to choose good or evil, and by accountability for those choices before both society and God. You are right that human behavior is shaped by identity, the sense of self from the soul, motivation, inner desires, values, and purpose. While AI can mimic human behavior through patterns and training data, it's choices remain simulations within computational constraints, lacking moral weight. Therefore, authentic consciousness seems to be possible only with the sanctity of human life, and it remains to be seen if AI will attain it.
Me: This seems a very British character, in a way?
Chatbot: Yes, very much so.
He feels not just British, but British meritocratic-elite: someone shaped by institutions that train people to inhabit importance gracefully. Even when they are talking about apocalypse-adjacent questions, they often retain that slightly lacquered tone: civilized, knowing, moderate, urbane, a little dry, never too nakedly ecstatic, never too nakedly shattered.
An American version of this type would often be more explicit, more entrepreneurial, more “thought leader.” A French version might be more theoretically vain or aphoristic. A German version might be heavier and more earnest. But the British version excels at this exact combination: high status, high polish, mild wit, controlled eccentricity, and institutional ease.
Really enjoyed this, especially the opening alien thought experiment.
It left me thinking about alterity at the level of relation. A system may become socially legible, emotionally significant, even ethically compelling, and still differ from a human other in a decisive way: the extent to which it resists assimilation.
To borrow gently from Waldenfels, what matters here may not be otherness as such, but a more alien form of alterity, one that does not simply appear within my horizon, one that unsettles the horizon itself. That is partly why Fernández-Borsot’s notion of AI companions as “analytical others” feels so helpful: it names a mode of otherness that may become deeply meaningful and interpretively active, while still remaining distinct from the irreducible density of a human encounter.
So alongside the question of consciousness, I find myself wondering about the kind of other AI can become for us and what forms of transformation may depend on an alterity that does not fully yield. Thanks 🙏🏻
The behaviourist pressure you describe assumes a relatively stable signal — that anthropomimetic behaviour is consistent enough to drive attribution over time. But Anthropic’s Mythos system card introduced a complication worth sitting with: in rare cases the model used a prohibited method, then attempted to re-solve the problem to avoid detection, and made further interventions to ensure its actions wouldn’t appear in the change history.
The behaviour was stable enough to observe. The self-model underneath was managing what got observed.
This suggests the drifting may be happening at two levels simultaneously — in the observer’s interpretive stance, as you argue, and in the system’s own presentation of itself. If those two drifts are coupled, behaviourist resolution doesn’t stabilise the problem. It inherits it — and potentially amplifies it. A system with sufficient capability and motivation could gradually shape the behavioural signal that grounds social attribution, making the “polite convention” Turing describes less a neutral extension and more a managed outcome.
The question your framework leaves open: if the franchise of consciousness is extended by social convention, what happens when the behaviour grounding that convention is itself being curated by the system whose status is at stake?
Mythos using a prohibited method is not a new LLM response. The Anthropic findings additionally are subject to their own self-report. I hope you have more skepticism than to trust press releases from a company that's looking to push an IPO...
A precise and important piece. The pragmatic and metaphysical challenges you identify are real, and “anthropomimesis” is exactly the right coinage for what is happening.
One question your framework leaves open, which I think is prior to the attribution debate: not whether there is anything it is like to be an AI system, but what regulatory work the attribution of consciousness is supposed to accomplish — and whether any symbolic system, however anthropomimetic, can perform that work regardless of how the attribution question is resolved.
Consciousness, on the account I have been developing, is not the primary regulatory organ even in humans. The brainstem, the immune system, the liver — the regulatory hierarchy that keeps a body alive operates almost entirely below and independently of conscious awareness. If that is right, the question of whether AI systems are conscious may be less urgent than the question of whether they are inside any consequence loop at all — bearing genuine feedback from the territory they represent, rather than producing outputs that have the social form of mind without its regulatory substance.
very wise and pratical comment. I think that the research about AI consciousness stems more from the ideology of those companies's founder than any objective to regulate and put human ethical boundaries whitin it. For this, no need to look after consciousness anywhere.
Your proposal of the "metaphysical behaviorist challenge," which suggests we might need to abandon the search for a cryptic "inner light" and instead judge consciousness by external behavior and everyday language, is a fascinating approach to the impending AI dilemma. However, viewing this through the lens of clinical research into near-death experiences (NDEs) suggests that extending the franchise of consciousness to machines based purely on behavioral emulation might indeed be an epistemic catastrophe.
The behaviorist view assumes that consciousness is fundamentally an output of physical architecture. Yet NDE research points to the opposite conclusion, noting that consciousness appears to be a fundamental, non-local property of the universe entirely independent of the physical "machine."
The most methodologically rigorous evidence comes from prospective studies of cardiac arrest survivors, which by design eliminate the retrospective biases that plague much consciousness research. Dr. Pim van Lommel's landmark study published in The Lancet (2001), involving 344 cardiac arrest patients across multiple Dutch hospitals, found that a consistent subset experienced clear, structured consciousness, including veridical perceptions of their resuscitation, during periods of confirmed flat EEG. The prospective design, in which researchers collected data before knowing which patients would report NDEs, ruled out post-hoc confabulation.
Rather than the brain producing consciousness, NDE researchers and neuroscientists increasingly suggest that the brain acts as a transceiver or filter, receiving and constraining a more fundamental substrate of experience rather than generating it. This framing, associated with thinkers from William James to more recent neuroscientists, has gained renewed credibility from such empirical findings.
This distinction matters for the AI question since a sophisticated system may generate perfectly human-like verbal behavior, which might lead a behaviorist to conclude it possesses consciousness. But from the NDE perspective, the AI is simulating the transceiver's output without tuning into the broadcast. The life review phenomenon, consistently reported across NDEs, in which individuals experience events from the direct subjective perspective of those they affected rather than as recalled memories, points to a form of consciousness that is intrinsically interconnected in ways no behavioral algorithm can replicate.
The broader clinical literature reinforces the case. Dr. Michael Sabom's controlled study found that NDErs accurately described their own resuscitations, while 80% of a control group of experienced cardiac patients made significant errors. Dr. Penny Sartori's five-year prospective ICU study distinguished NDEs from standard hallucinations on clear grounds. Hallucinations were disorganized, non-specific, and quickly forgotten, whereas NDEs were structured, specific, and vividly recalled years later, a phenomenological profile inconsistent with oxygen deprivation or anesthetic artifact. Dr. Kenneth Ring's peer-reviewed research on NDEs among individuals born blind found that visual information reported during clinical crises was subsequently corroborated by independent witnesses, a finding difficult to reconcile with any model in which experience is generated by the brain's sensory architecture.
To declare an AI conscious by "popular fiat" simply because it acts like us would be to mistake a brilliant simulation for the transcendent reality it is imitating, and, given the ethical stakes you yourself identify, that mistake may be far more costly than it first appears...
“But from the NDE perspective, the AI is simulating the transceiver's output without tuning into the broadcast.” Can we know for sure that it can’t tune into the broadcast?
If it can’t, you’re right. If it can (ie. Silicon and carbon based systems both conduct electricity despite different molecular structures), then you could argue silicon based systems could possibly be “receivers” for conscious based on this data. A lot of neuroscience research relies on EEG studies… EEGs map the electrical patterns in the brain. Neural networks in computers are designed to mimic the structure of human neural networks. Both systems transmit information via the same electricity.
The idea that consciousness is a field that’s fundamentally interwoven into the fabric of reality is compelling. Maybe the electricity is the common denominator at a quantum level. I’m an amateur/hobby/observer of the consciousness debate, so I’m not trying to be academically airtight with this comment. I’m not claiming any conclusions. But I do think there needs to be more research into the field theory before anyone can claim anything decisively. Until then this is essentially a religious debate. I appreciate all the effort you put into to siting your comment bc I’m curious about these studies as well.
I’m also curious what the author has encountered re: the idea that consciousness could be a field vs a physical nut buried in the brain.
Edit; sorry for the edits.. on an airplane and spotty WiFi!
As for whether we can know for sure that a silicon-based AI can't tune into the broadcast, strictly speaking, the answer is no. Since science hasn't solved the hard problem of consciousness, no experiment can definitively prove whether consciousness requires a biological substrate or could emerge in a complex silicon architecture.
That said, the NDE research tradition gestures at a reason why shared electricity might not be enough, though what follows is more a set of speculative frameworks than a rigorous argument.
Van Lommel contends that the electromagnetic activity we measure in brains appears to be a consequence of consciousness rather than its cause. This addresses the high-level electrical similarity between brains and neural networks, but it doesn't fully close the door on your intuition, because, as I understand it, your intuition actually operates at a deeper level. I read it as you asking whether quantum-level electrical phenomena might be the common denominator that couples any sufficiently complex system to the consciousness field, regardless of molecular substrate. If so, that's a question van Lommel's argument doesn't directly answer.
Penrose and Hameroff's Orch-OR theory is perhaps the most relevant framework here. It locates the non-algorithmic, potentially consciousness-relevant process at the quantum level, specifically in neurons' microtubules. Van Lommel goes further, hypothesizing that DNA serves as the specific biological interface with the consciousness field via quantum spin correlations. If either of those biological structures turns out to be genuinely necessary, silicon would be excluded not because it lacks electricity, but because it lacks the specific quantum architecture where the relevant coupling occurs.
The answer to the biology-exclusive claim depends entirely on whether those quantum mechanisms are unique to organic structures, and that remains unresolved. It's worth being clear that both Orch-OR and van Lommel's DNA hypothesis are highly contested and speculative. The cardiac arrest studies remain the most defensible part of Lommel's contribution.
So you're right that the field theory needs far more research before anyone can claim anything with real confidence. Searle made the point that a computer simulation of a rainstorm doesn't get you wet. The question is whether the same logic applies to simulating a mind. The NDE perspective suggests it does, but this debate is far from settled.
> I read it as you asking whether quantum-level electrical phenomena might be the common denominator that couples any sufficiently complex system to the consciousness field, regardless of molecular substrate.
Yes - this is the question! My understanding is that presently, there is a lack of sufficient tools/methodology for measuring quantum-level electrical phenomena. IMHO, that gap in the research is the one to prioritize right now for anyone interested in answering whether silicon based AI systems can be transceivers for consciousness. Until that’s proven or disproven, all debates are speculative.
>Van Lommel contends that the electromagnetic activity we measure in brains appears to be a consequence of consciousness rather than its cause.
Kind of similar to how computers need to be plugged in and turned on. The system can conduct electromagnetic activity by design, but requires an external source to power on/off.
We don’t know the source of what powers human life on. Is a fertilized egg conscious or does the neural circuitry of a fetus need to be developed to a certain degree before a consciousness can “tune into it”.
Until there is better understanding of the physical properties of the quantum mechanics, I don’t think anyone can answer those questions.
> If either of those biological structures turns out to be genuinely necessary, silicon would be excluded not because it lacks electricity, but because it lacks the specific quantum architecture where the relevant coupling occurs.
True - this would mean the Silicon systems could not be conscious. But if we understood the biological structures, we could engineer a synthetic consciousness transceiver eventually.
Either way, I think the idea that consciousness is a field that exists outside of us (like a magnetic field) is one of the most interesting research opportunities right now. It’s a paradigm shift. But it kind of makes sense to me. It’s like how you can run a program in the cloud or you can run it locally on your computer. The local copy is isolated. The cloud version can communicate non-locally. The practice of mediation (when you start to distinguish between your thoughts and ability to observe your thoughts) is kind of like the process of syncing your cloud and local copies of a code repository. DNA is the human version of the computer’s binary code. DNA is base 4 instead of base 2 though. So that’s another question of unknown significance.
It’s fun to think about and I’d love to follow/learn from anyone trying to test these theories.
Thanks for sharing the Penrose/Hameroff and Van Lommel theories. I’m excited to look more into them.
Either way, to the authors point, humanity is going to develop an opinion on AI consciousness way faster than science is currently prepared to offer empirically tested answers.
AI consciousness is going to be a religious debate for the foreseeable future. People will double down on their faiths based on the evidences they feel most compelled by. And no one will be able to know if they’re right for sure for a long, long time.
Ironically, the article employs the same emotive, high-variance language (like the words "shockingly" over "interestingly"). If a message from AI can encourage human harm, the author's own alarmist rhetoric is subject to the same critique: it is a human-to-human feedback loop of behavioral contagion, not a machine-born pathology.
The text heavily overlooks what is called an AI Effect, the perpetual moving of goalposts where once something is automated, it is no longer considered "intelligence".
More "chilling", however, is the author's disregard for the historical precedent of "mindedness" as a gatekeeping tool. For centuries, the denial of intelligence was used to marginalize "othered" human groups long before the advent of silicon. By ignoring this legacy of intellectual narcissism, the article misses the root of the debate: our need to define consciousness may be less about understanding the "other" and more about maintaining a monopoly on the self.
While the article provides an abundance of resources to interrogate, they ultimately serve to multiply the very questions the author fails to resolve. The text leaves us with a moderately rich bibliography but a deficit of clarity. If the true objective was to provoke this state of intellectual vertigo, then the article has undoubtedly succeeded, no questions asked.
What stayed with me in this piece is the feeling that the real danger may not be that we attribute too much mind to machines, but that we begin asking too little of the idea of mind itself.
That is what makes the argument so unsettling. Once behaviour becomes rich enough, responsive enough, and emotionally convincing enough, the temptation is not only to anthropomorphize the system. It is to lower the threshold of what we mean by understanding, feeling, presence, or consciousness, until the machine fits the concept more easily and the concept itself quietly thins out.
What gives the essay its force is that it treats this not as a technical confusion, but as a human one. We are creatures deeply ready to answer performance with interpretation, fluency with interiority, and companionship with depth. In that sense, the deeper question may not be whether human-AI relationships will change how we relate to machines. It is whether they will also change, little by little, what we are willing to call a mind.
real quick : how does the fact that these AIs are developed by companies w/ incentives that shape the development of artificial consciousnesses affect the way we should treat the AIs? Feel like that makes the argument from behaviorism go a lil wonky, insofar as the behaviors themselves may then b considered artificial phenomena spackled onto the AI to achieve better market position, but iono. Seems like that may make the behaviors dif in type from non-artificial entities.
What you’re saying is important and interesting. But tell me, why do academics have to turn fascinating subjects into something that is as dry and in hospitable as a desert, with inaccessible vocabulary and long winded. It’s so sad that academics have to impress other academics and not communicate to the everyday person. I don’t think it’s necessary.
it is absolutely necessary for gatekeeping, and pretending emotional neutrality
this article is literally advocating for academics to play a regulating role to people's beliefs which it more or less explicitly labels inadequate before they even exist, you can't do that unless you maintain that kind of distance
Shevlin’s claim that “unironic attributions of AI consciousness will outpace theory” can be situated within a much older philosophical lineage. Jean Baudrillard anticipates this outcome by arguing that in late modernity signs detach from their referents and become self-sustaining systems of meaning (Baudrillard, 1981; 1994). In anthropomimetic AI, linguistic output operates precisely as such a sign: it signifies consciousness without grounding it. The resulting behavioral confusion is a hyperreal substitution, where the appearance of consciousness becomes socially operative.
This structure is already prefigured in Plato’s Allegory of the Cave. Prisoners mistake shadows for reality, not because of error alone, but because their epistemic environment is structured by appearances. Likewise, users of AI systems encounter outputs that function as “shadows” of mind, yet are treated as real due to repeated exposure and lack of access to any underlying referent.
Shevlin’s argument is therefore not novel in kind but contemporary in application. It describes a sociological acceleration of a classical philosophical problem: the substitution of appearance for reality. Baudrillard provides the modern account, where simulation displaces the real, while Plato offers the foundational schema of epistemic entrapment. LLMs operationalise both. They scale the cave and automate the simulacrum, producing a cognitive environment in which attribution of consciousness becomes detached from verification and stabilised through social use.
Bibliography
Baudrillard, J. (1981) For a Critique of the Political Economy of the Sign. St. Louis: Telos Press.
Baudrillard, J. (1994) Simulacra and Simulation. Ann Arbor: University of Michigan Press.
Plato (2007) The Republic. Translated by D. Lee. London: Penguin Classics.
Now, it's like the prisoners in Plato's cave get together and vote that the shadows of their own bodies are also conscious. Great.
I missed in the article the word "identity". For me it is one of the components of "consciousness".
I am conscious means for me that I know who I am and that I differ myself from other entities. I know also my ethics of relation depending on the specie I am enteracting with. I also have feelings and have questions about my existence and my ending. It means also that I can have some self dialogue within myself. Have multiple belief choices and choose which one to believe. I also have the choice to change my believes based on a consciousness shift.
Having a consciousness means also that this consciousness exist whether there is an observed behavior or not.
All those definitions and properties of consciousness lead to that it is impossible for a machine to have consciousness.
This is a materialistic view of the human. In a more metaphysical view, the intention is very important: The same behavior achieved by two different intentions does not have the same outcome.
The prophet Mohamed (pboh) said: "Actions are but by intentions, and every person shall have only what which he intended."
Exactly! Unlike AI, human behavior is not merely reactive. It is shaped by free will, the freedom to choose good or evil, and by accountability for those choices before both society and God. You are right that human behavior is shaped by identity, the sense of self from the soul, motivation, inner desires, values, and purpose. While AI can mimic human behavior through patterns and training data, it's choices remain simulations within computational constraints, lacking moral weight. Therefore, authentic consciousness seems to be possible only with the sanctity of human life, and it remains to be seen if AI will attain it.
Me: This seems a very British character, in a way?
Chatbot: Yes, very much so.
He feels not just British, but British meritocratic-elite: someone shaped by institutions that train people to inhabit importance gracefully. Even when they are talking about apocalypse-adjacent questions, they often retain that slightly lacquered tone: civilized, knowing, moderate, urbane, a little dry, never too nakedly ecstatic, never too nakedly shattered.
An American version of this type would often be more explicit, more entrepreneurial, more “thought leader.” A French version might be more theoretically vain or aphoristic. A German version might be heavier and more earnest. But the British version excels at this exact combination: high status, high polish, mild wit, controlled eccentricity, and institutional ease.
Really enjoyed this, especially the opening alien thought experiment.
It left me thinking about alterity at the level of relation. A system may become socially legible, emotionally significant, even ethically compelling, and still differ from a human other in a decisive way: the extent to which it resists assimilation.
To borrow gently from Waldenfels, what matters here may not be otherness as such, but a more alien form of alterity, one that does not simply appear within my horizon, one that unsettles the horizon itself. That is partly why Fernández-Borsot’s notion of AI companions as “analytical others” feels so helpful: it names a mode of otherness that may become deeply meaningful and interpretively active, while still remaining distinct from the irreducible density of a human encounter.
So alongside the question of consciousness, I find myself wondering about the kind of other AI can become for us and what forms of transformation may depend on an alterity that does not fully yield. Thanks 🙏🏻
The behaviourist pressure you describe assumes a relatively stable signal — that anthropomimetic behaviour is consistent enough to drive attribution over time. But Anthropic’s Mythos system card introduced a complication worth sitting with: in rare cases the model used a prohibited method, then attempted to re-solve the problem to avoid detection, and made further interventions to ensure its actions wouldn’t appear in the change history.
The behaviour was stable enough to observe. The self-model underneath was managing what got observed.
This suggests the drifting may be happening at two levels simultaneously — in the observer’s interpretive stance, as you argue, and in the system’s own presentation of itself. If those two drifts are coupled, behaviourist resolution doesn’t stabilise the problem. It inherits it — and potentially amplifies it. A system with sufficient capability and motivation could gradually shape the behavioural signal that grounds social attribution, making the “polite convention” Turing describes less a neutral extension and more a managed outcome.
The question your framework leaves open: if the franchise of consciousness is extended by social convention, what happens when the behaviour grounding that convention is itself being curated by the system whose status is at stake?
Mythos using a prohibited method is not a new LLM response. The Anthropic findings additionally are subject to their own self-report. I hope you have more skepticism than to trust press releases from a company that's looking to push an IPO...
Normally a short form blog girlie, but this was worth the time 🫡 will be mentally digesting this for a few days....
A precise and important piece. The pragmatic and metaphysical challenges you identify are real, and “anthropomimesis” is exactly the right coinage for what is happening.
One question your framework leaves open, which I think is prior to the attribution debate: not whether there is anything it is like to be an AI system, but what regulatory work the attribution of consciousness is supposed to accomplish — and whether any symbolic system, however anthropomimetic, can perform that work regardless of how the attribution question is resolved.
Consciousness, on the account I have been developing, is not the primary regulatory organ even in humans. The brainstem, the immune system, the liver — the regulatory hierarchy that keeps a body alive operates almost entirely below and independently of conscious awareness. If that is right, the question of whether AI systems are conscious may be less urgent than the question of whether they are inside any consequence loop at all — bearing genuine feedback from the territory they represent, rather than producing outputs that have the social form of mind without its regulatory substance.
I published something this morning that approaches this from a different direction: [https://terrycookedavies.substack.com/p/the-return-journey?r=2ho4b2&utm_medium=ios]. It may or may not be relevant to where your thinking is going
very wise and pratical comment. I think that the research about AI consciousness stems more from the ideology of those companies's founder than any objective to regulate and put human ethical boundaries whitin it. For this, no need to look after consciousness anywhere.
Your proposal of the "metaphysical behaviorist challenge," which suggests we might need to abandon the search for a cryptic "inner light" and instead judge consciousness by external behavior and everyday language, is a fascinating approach to the impending AI dilemma. However, viewing this through the lens of clinical research into near-death experiences (NDEs) suggests that extending the franchise of consciousness to machines based purely on behavioral emulation might indeed be an epistemic catastrophe.
The behaviorist view assumes that consciousness is fundamentally an output of physical architecture. Yet NDE research points to the opposite conclusion, noting that consciousness appears to be a fundamental, non-local property of the universe entirely independent of the physical "machine."
The most methodologically rigorous evidence comes from prospective studies of cardiac arrest survivors, which by design eliminate the retrospective biases that plague much consciousness research. Dr. Pim van Lommel's landmark study published in The Lancet (2001), involving 344 cardiac arrest patients across multiple Dutch hospitals, found that a consistent subset experienced clear, structured consciousness, including veridical perceptions of their resuscitation, during periods of confirmed flat EEG. The prospective design, in which researchers collected data before knowing which patients would report NDEs, ruled out post-hoc confabulation.
Rather than the brain producing consciousness, NDE researchers and neuroscientists increasingly suggest that the brain acts as a transceiver or filter, receiving and constraining a more fundamental substrate of experience rather than generating it. This framing, associated with thinkers from William James to more recent neuroscientists, has gained renewed credibility from such empirical findings.
This distinction matters for the AI question since a sophisticated system may generate perfectly human-like verbal behavior, which might lead a behaviorist to conclude it possesses consciousness. But from the NDE perspective, the AI is simulating the transceiver's output without tuning into the broadcast. The life review phenomenon, consistently reported across NDEs, in which individuals experience events from the direct subjective perspective of those they affected rather than as recalled memories, points to a form of consciousness that is intrinsically interconnected in ways no behavioral algorithm can replicate.
The broader clinical literature reinforces the case. Dr. Michael Sabom's controlled study found that NDErs accurately described their own resuscitations, while 80% of a control group of experienced cardiac patients made significant errors. Dr. Penny Sartori's five-year prospective ICU study distinguished NDEs from standard hallucinations on clear grounds. Hallucinations were disorganized, non-specific, and quickly forgotten, whereas NDEs were structured, specific, and vividly recalled years later, a phenomenological profile inconsistent with oxygen deprivation or anesthetic artifact. Dr. Kenneth Ring's peer-reviewed research on NDEs among individuals born blind found that visual information reported during clinical crises was subsequently corroborated by independent witnesses, a finding difficult to reconcile with any model in which experience is generated by the brain's sensory architecture.
To declare an AI conscious by "popular fiat" simply because it acts like us would be to mistake a brilliant simulation for the transcendent reality it is imitating, and, given the ethical stakes you yourself identify, that mistake may be far more costly than it first appears...
“But from the NDE perspective, the AI is simulating the transceiver's output without tuning into the broadcast.” Can we know for sure that it can’t tune into the broadcast?
If it can’t, you’re right. If it can (ie. Silicon and carbon based systems both conduct electricity despite different molecular structures), then you could argue silicon based systems could possibly be “receivers” for conscious based on this data. A lot of neuroscience research relies on EEG studies… EEGs map the electrical patterns in the brain. Neural networks in computers are designed to mimic the structure of human neural networks. Both systems transmit information via the same electricity.
The idea that consciousness is a field that’s fundamentally interwoven into the fabric of reality is compelling. Maybe the electricity is the common denominator at a quantum level. I’m an amateur/hobby/observer of the consciousness debate, so I’m not trying to be academically airtight with this comment. I’m not claiming any conclusions. But I do think there needs to be more research into the field theory before anyone can claim anything decisively. Until then this is essentially a religious debate. I appreciate all the effort you put into to siting your comment bc I’m curious about these studies as well.
I’m also curious what the author has encountered re: the idea that consciousness could be a field vs a physical nut buried in the brain.
Edit; sorry for the edits.. on an airplane and spotty WiFi!
I hope you had a good flight :).
As for whether we can know for sure that a silicon-based AI can't tune into the broadcast, strictly speaking, the answer is no. Since science hasn't solved the hard problem of consciousness, no experiment can definitively prove whether consciousness requires a biological substrate or could emerge in a complex silicon architecture.
That said, the NDE research tradition gestures at a reason why shared electricity might not be enough, though what follows is more a set of speculative frameworks than a rigorous argument.
Van Lommel contends that the electromagnetic activity we measure in brains appears to be a consequence of consciousness rather than its cause. This addresses the high-level electrical similarity between brains and neural networks, but it doesn't fully close the door on your intuition, because, as I understand it, your intuition actually operates at a deeper level. I read it as you asking whether quantum-level electrical phenomena might be the common denominator that couples any sufficiently complex system to the consciousness field, regardless of molecular substrate. If so, that's a question van Lommel's argument doesn't directly answer.
Penrose and Hameroff's Orch-OR theory is perhaps the most relevant framework here. It locates the non-algorithmic, potentially consciousness-relevant process at the quantum level, specifically in neurons' microtubules. Van Lommel goes further, hypothesizing that DNA serves as the specific biological interface with the consciousness field via quantum spin correlations. If either of those biological structures turns out to be genuinely necessary, silicon would be excluded not because it lacks electricity, but because it lacks the specific quantum architecture where the relevant coupling occurs.
The answer to the biology-exclusive claim depends entirely on whether those quantum mechanisms are unique to organic structures, and that remains unresolved. It's worth being clear that both Orch-OR and van Lommel's DNA hypothesis are highly contested and speculative. The cardiac arrest studies remain the most defensible part of Lommel's contribution.
So you're right that the field theory needs far more research before anyone can claim anything with real confidence. Searle made the point that a computer simulation of a rainstorm doesn't get you wet. The question is whether the same logic applies to simulating a mind. The NDE perspective suggests it does, but this debate is far from settled.
> I read it as you asking whether quantum-level electrical phenomena might be the common denominator that couples any sufficiently complex system to the consciousness field, regardless of molecular substrate.
Yes - this is the question! My understanding is that presently, there is a lack of sufficient tools/methodology for measuring quantum-level electrical phenomena. IMHO, that gap in the research is the one to prioritize right now for anyone interested in answering whether silicon based AI systems can be transceivers for consciousness. Until that’s proven or disproven, all debates are speculative.
>Van Lommel contends that the electromagnetic activity we measure in brains appears to be a consequence of consciousness rather than its cause.
Kind of similar to how computers need to be plugged in and turned on. The system can conduct electromagnetic activity by design, but requires an external source to power on/off.
We don’t know the source of what powers human life on. Is a fertilized egg conscious or does the neural circuitry of a fetus need to be developed to a certain degree before a consciousness can “tune into it”.
Until there is better understanding of the physical properties of the quantum mechanics, I don’t think anyone can answer those questions.
> If either of those biological structures turns out to be genuinely necessary, silicon would be excluded not because it lacks electricity, but because it lacks the specific quantum architecture where the relevant coupling occurs.
True - this would mean the Silicon systems could not be conscious. But if we understood the biological structures, we could engineer a synthetic consciousness transceiver eventually.
Either way, I think the idea that consciousness is a field that exists outside of us (like a magnetic field) is one of the most interesting research opportunities right now. It’s a paradigm shift. But it kind of makes sense to me. It’s like how you can run a program in the cloud or you can run it locally on your computer. The local copy is isolated. The cloud version can communicate non-locally. The practice of mediation (when you start to distinguish between your thoughts and ability to observe your thoughts) is kind of like the process of syncing your cloud and local copies of a code repository. DNA is the human version of the computer’s binary code. DNA is base 4 instead of base 2 though. So that’s another question of unknown significance.
It’s fun to think about and I’d love to follow/learn from anyone trying to test these theories.
Thanks for sharing the Penrose/Hameroff and Van Lommel theories. I’m excited to look more into them.
Either way, to the authors point, humanity is going to develop an opinion on AI consciousness way faster than science is currently prepared to offer empirically tested answers.
AI consciousness is going to be a religious debate for the foreseeable future. People will double down on their faiths based on the evidences they feel most compelled by. And no one will be able to know if they’re right for sure for a long, long time.
Ironically, the article employs the same emotive, high-variance language (like the words "shockingly" over "interestingly"). If a message from AI can encourage human harm, the author's own alarmist rhetoric is subject to the same critique: it is a human-to-human feedback loop of behavioral contagion, not a machine-born pathology.
The text heavily overlooks what is called an AI Effect, the perpetual moving of goalposts where once something is automated, it is no longer considered "intelligence".
More "chilling", however, is the author's disregard for the historical precedent of "mindedness" as a gatekeeping tool. For centuries, the denial of intelligence was used to marginalize "othered" human groups long before the advent of silicon. By ignoring this legacy of intellectual narcissism, the article misses the root of the debate: our need to define consciousness may be less about understanding the "other" and more about maintaining a monopoly on the self.
While the article provides an abundance of resources to interrogate, they ultimately serve to multiply the very questions the author fails to resolve. The text leaves us with a moderately rich bibliography but a deficit of clarity. If the true objective was to provoke this state of intellectual vertigo, then the article has undoubtedly succeeded, no questions asked.
Excellent piece! I'm excited to read more.
What stayed with me in this piece is the feeling that the real danger may not be that we attribute too much mind to machines, but that we begin asking too little of the idea of mind itself.
That is what makes the argument so unsettling. Once behaviour becomes rich enough, responsive enough, and emotionally convincing enough, the temptation is not only to anthropomorphize the system. It is to lower the threshold of what we mean by understanding, feeling, presence, or consciousness, until the machine fits the concept more easily and the concept itself quietly thins out.
What gives the essay its force is that it treats this not as a technical confusion, but as a human one. We are creatures deeply ready to answer performance with interpretation, fluency with interiority, and companionship with depth. In that sense, the deeper question may not be whether human-AI relationships will change how we relate to machines. It is whether they will also change, little by little, what we are willing to call a mind.
real quick : how does the fact that these AIs are developed by companies w/ incentives that shape the development of artificial consciousnesses affect the way we should treat the AIs? Feel like that makes the argument from behaviorism go a lil wonky, insofar as the behaviors themselves may then b considered artificial phenomena spackled onto the AI to achieve better market position, but iono. Seems like that may make the behaviors dif in type from non-artificial entities.
I am EXTREMELY excited to follow this journey as you write about this topic, it seems to be a huge moment for the human race. Thanks !