AI Consciousness

Whether artificial intelligence (AI) can be or become conscious, while long a question, has burst into public discourse—due to the sudden impact of large language models such as ChatGPT, Gemini, and others. AI consciousness has become a serious matter of global concern. The question has vast social, moral and perhaps human-species-wide consequences.

A major multidisciplinary report, bringing together AI experts, philosophers, neuroscientists, and psychologists, argues for and exemplifies a rigorous and empirically grounded approach to AI consciousness. The report surveys prominent scientific theories of consciousness, all of which are on this Landscape, and derives “indicator properties” of consciousness, which are used to assess AI systems. The conclusion is that no current AI system is conscious, but that there are no obvious barriers to building AI systems that could be conscious (Butlin, 2023).

It must be stressed that the report's working hypothesis is computational functionalism, the thesis that performing computations of the right kind is necessary and sufficient for consciousness. The report adopts this hypothesis for pragmatic reasons: unlike rival views, it entails that consciousness in AI is possible in principle and that studying the workings of AI systems can assess whether they are likely to be conscious.

Given AI engineer/visionary Blaise Agüera y Arcas's computational functional theory of consciousness, he has little doubt that AI can, in principle, become conscious. However, the pragmatic assessment of whether a specific AI system is in fact conscious, namely, whether it has subjective experience, is another matter. Observational judgment is all we have, he says. The only way we can ever know if another entity—human or AI—is conscious is by interacting with it and building a model of its behavior. Agüera y Arcas suggests that if an AI's behavior is functionally indistinguishable from a conscious human's, as long as we agree that calling other people conscious is proper, to require a different criterion for AI systems is arbitrary and inconsistent. This perspective frames the ongoing debate not as a search for some hidden internal property that is or causes subjective experience, but as a collective decision about how we choose to relate to these new forms of intelligences (Agüera y Arcas, 2025).

Though computational functionalism is indeed a mainstream position in philosophy of mind, computational functionalism is challenged by diverse rivals on the Landscape.

To philosopher John Searle, computer programs can never have a mind or be conscious in the human sense, even if they give rise to equivalent behaviors and interactions with the external world. In Searle's famous “Chinese Room” argument, a person inside a closed space can use a rule book to match Chinese characters with English words and thus appear to understand Chinese, when, in fact, she does not. (There is dispute about the validity of Searle's Chinese Room argument [Cole, 2023].)

Nonetheless, Searle argues that just because brain processes cause consciousness and intentionality (aboutness) does not imply that only brains can be conscious. The brain is a biological machine, and we might build an artificial machine that was conscious. Because we do not know how the brain generates consciousness, Searle says, is the reason we are not yet in a position to know how to do it artificially (Searle, 2007a, 2007b).

Rather, what Searle rejects is that a simulation of brain states, however detailed the information and precise the representation, can achieve the subjective qualities of inner awareness. What is required for consciousness, he says, is the same set or system of biological processes that the brain uses (Searle, 2002; Proust, 2003).

Will it ever be possible, with hyper-advanced technology, for non-biological intelligences to be conscious in the same sense that we are conscious? Can computers have ‘inner experience’?^[1]

“It's like the question, ‘Can a machine artificially pump blood as the heart does?” Searle responds. “Sure it can—we have artificial hearts. So, if we can know exactly how the brain causes consciousness, down to its finest details, I don't see any obstacle, in principle, to building a conscious machine. That is, if you knew what was causally sufficient to produce consciousness in human beings and if you could have that [mechanism] in another system, then you would produce consciousness in that other system. Note that you don't need neurons to have consciousness. It's like saying you don't need feathers to fly. But to build a flying machine, you do need sufficient causal power to overcome the force of gravity” (Searle, 2007b.).

Searle cautions: “The one mistake we must avoid is supposing that if you simulate it, you duplicate it. A deep mistake embedded in our popular culture is that simulation is equivalent to duplication. But of course it isn't. A perfect simulation of the brain—say, on a computer—would be no more conscious than a perfect simulation of a rainstorm would make us all wet.”

Robotics professor/entrepreneur Rodney Brooks agrees that consciousness can be created in non-biological media, but disagrees on the nature of consciousness itself. “There's no reason we couldn't have a conscious machine made from silicon,” he said. Brooks's position derives from his view that the universe is mechanistic and that consciousness, which seems special, is an illusion. We “fool ourselves,” he says, into “thinking our internal feelings are so unique” (Brooks, 2011).

AI expert Joscha Bach is bullish on AI consciousness, in part, because his theory treats “consciousness as a memory instead of an actual sense of the present”—which he says “resolves much of the difficulty for specifying an AI implementation of consciousness: it is necessary and sufficient to realize a system that remembers having experienced something, and being able to report on that memory” (Bach, 2019).

Can we ever really assess consciousness? “I don't know if you're conscious. You don't know if I'm conscious,” says neuroscientist Michael Graziano. “But we have a kind of gut certainty about it. This is because an assumption of consciousness is an attribution, a social attribution. And when a robot acts like it's conscious and can talk about its own awareness, and when we interact with it, we will inevitably have that social perception, that gut feeling, that the robot is conscious …. But can you really ever know if there's ‘anybody home’ internally, if there is any inner experience?” he continues. “All we do is compute a construct of awareness” (Graziano, 2014).

Inventor and futurist Ray Kurzweil believes that “we will get to a point where computers will evidence the rich array of emotionally subtle behaviors that we see in human beings; they will be very intelligent, and they will claim to be conscious. They will act in ways that are conscious; they will talk about their own consciousness and argue about it just the way you and I do. And so the philosophical debate will be whether or not they really are conscious—and they will be participating in the debate” (Kurzweil, 2007).

Kurzweil argues that assessing the consciousness of other (possible] minds is not a scientific question. “We can talk scientifically about the neurological correlates of consciousness, but fundamentally, consciousness is this subjective experience that only I can experience. I should talk about it only in first-person terms—although I've been sufficiently socialized to accept other people's consciousness. There's really no way to measure the conscious experiences of another entity … But I would accept that these non-biological intelligences are conscious. And that'll be convenient, because if I don't, they'll get mad at me.”

Physiological psychologist Warren Brown stresses “embodied cognition, embodied consciousness,” in that “biology is the richest substrate for embodying consciousness.” But he doesn't rule out that consciousness “might be embodied in something non-biological.” On the other hand, Brown speculates, “consciousness may be a particular kind of organization of the world that just cannot be replicated in a non-biological system” (Brown, 2014).

“I am a functionalist when it comes to consciousness,” says neuroscientist Christof Koch. "As long as we can reproduce the same kind of relevant relationships among all the relevant neurons in the brain, I think we will have recreated consciousness. The difficult part is, what do we mean by ‘relevant relationships?’ Does it mean we have to reproduce the individual motions of all the molecules? Unlikely. It's more likely that we have to recreate all the relevant relationships of the brain's synapses and the brain's wiring (the ‘connectome’) in a different medium, like a computer. If we can do all of this reconstruction at the right level, this entity, this software construct, would be conscious” (Koch, 2012c).

Koch stresses that “experience” requires new, perhaps radical, scientific thinking. “You need to expand the traditional laws of physics. In physics there is space, time, energy, mass. Those by themselves are sufficient to explain the physics of the brain. The brain is subject to the same laws of physics as any other object in the universe. But in addition, there is something else. There is experience. The experience of pain. The experience of falling in love. And to account for experience, you need to enhance the laws of physics.”

In the context of Integrated Information Theory, would Koch be comfortable with nonbiological consciousness? “Why should I not be? Consciousness doesn't require any magical ingredient.”

Over the years, Koch has refined his views. Against the grain, he says, “integrated information theory radically disagrees with this functionalist view. It argues from first principles that digital computers can (in principle) do everything that humans can do, eventually even faster and better. But they can never be what humans are. Intelligence is computable, but consciousness is not. This is not because the brain possesses any supernatural properties. The critical difference between brains and digital computers is at the hardware level, where the rubber meets the road—that is, where action potentials are relayed to tens of thousands of recipient neurons versus packets of electrons shuttled back and forth among a handful of transistors.” Koch primary point is that “the integrated information of digital computers is negligible. And that makes all the difference. It means that these machines will never be sentient, no matter how intelligent they become. Furthermore, that they will never possess what we have: the ability to deliberate over an upcoming choice and freely decide” (Koch, 2024, p. 20).

Theist philosopher Richard Swinburne says, “I don't see that it is in the least implausible that a 'radically separate, non-physical substance' could come into existence as a result of a non-biological process. There might be some law of nature stating that all sufficiently complicated computer-like systems become conscious. But the problem is that the law could not state which conscious being they would become, out of the innumerable possible individual conscious beings they might become. And that, in my view, also applies to organisms produced by normal processes—there may be a law determining that a person with a certain character emerges as a result of fertilization of an egg, but the law could not determine which person that was; for the simple reason that laws deal with the causation of states of affairs of certain kinds by other states of affairs of certain kinds; and given that a duplicate of me isn't necessarily me, no law of nature could determine that I would have been born from my actual parents” (Swinburne, 2016).

Philosopher Catherine Reason's No-Go Theorem against Materialsm/Supervenience implies a sharp limit for AI consciousness. "No physical system capable of human-like reasoning can perform the ω function," so an artificial reasoner "could, therefore, only perform the ω [self-certainty] function if it could show that it was either incapable of human-like reasoning or definitely not physical" (Reason and Shah, 2021). Genuine machine self-certainty, on this view, cannot be achieved by any purely physical computational system that reasons soundly.

Now, for each of the categories of explanations of consciousness, a conjecture: In which could AI become conscious?

Materialism Theories: Sure. For Materialism Theories (with all its subcategories) to be consistent, AI consciousness must be in principle absolutely sure. There is no possibility that, given materialism, AI consciousness would be forbidden. If one argues that consciousness must be embodied, fine, then materialism will build a body. Remember, we are speaking in ultimate principle, not in current practice, and there are no time limits.

To oversimplify, materialist views on AI consciousness split along the fault line of life. Neuroscientist Anil Seth challenges “the assumption that computation provides a sufficient basis for consciousness.” Instead, he makes “the case that consciousness depends on our nature as living organisms.” He calls his view “a form of biological naturalism,” and argues that “real artificial consciousness is unlikely along current trajectories, but becomes more plausible as AI becomes more brain-like and/or life-like” (Seth, 2025). Embodied and enactive theories of consciousness would also privilege life, and its interactions with its environment, as a necessary prerequisite of consciousness (Embodied & Enactive).

Most materialists, and a large majority of AI experts, would disagree and argue that life is not in any way special, and it is certainly not necessary for consciousness. Their view, computational functionalism, is that the brain and its outputs, including consciousness, are computational in nature, and that if the function is reproduced precisely, so is the output, irrespective of the substrate in which the function is carried out—consciousness being no exception (Blums, 2025; Agüera y Arcas, 2025).

Ultimately, there is no difference in the materialist outcome: AI will become conscious. The only difference is the steps involved and the timeline they require. If computational functionalism is correct, it’s a matter of figuring out the functions, and the timeline is shorter (but not necessarily short). If life is required, then AI consciousness is a two-step process: first, life must be artificially created and manipulated, then the special requirements of consciousness added—a longer timeline. A common intermediary position is that to build machines that are conscious, novel machine architectures must be based on information-processing computations similar to those of the human brain; for example, global workspace and higher-order theories (Dehaene et al., 2017).

If materialism explains consciousness entirely (without residue), then it would certainly be true that non-biological intelligences with super-strong AI would eventually have the same kind of inner awareness that humans do. Moreover, as AI would break through the singularity and become vastly more sophisticated than the human brain, it would likely express forms of consciousness higher than we today can even imagine. Though some speculatively reject that AI could ever become conscious (e.g., Reber, 2016; Reber, 2018), if one takes a hard-core physicalist position, an immutably skeptical outlook may not be warranted (and may not be coherent).

To the degree that language affects the deep essence of consciousness, this would make AI consciousness more likely, given the exponential advances in AI language development. But language per se is certainly not sufficient and likely not necessary.

Non-Reductive Physicalism. If Non-Reductive Physicalism is true, then it would be likely that non-biological intelligences could eventually be conscious—although the independent reality of mental states attenuates (slightly, unpredictably) the likelihood of inner awareness—an argument that is itself countered by functionalism (if functionalism is true). However, if strong emergence and top-down causation were required, then both would have to be enabled in creating AI consciousness, a process that would require two orders of complexity (i.e., strong emergence and top-down causation as real phenomena, and then their artificial creation).

Quantum & Dimensions Theories. If quantum mechanics is the key to consciousness, with its exponential amplification of processing power and its vast parallel pathways working simultaneously, AI consciousness is almost certain. Moreover, Quantum Theories would be the lead category for generating AI consciousness. The one caveat, a practical but not an in-principle obstacle, would be the physical constraints of manipulating myriad quantum states, with their inherent indeterminacies and environmental sensitivities, making the technology even more daunting. However, the technology is accelerating with fervor and so if AI consciousness is to happen, by design or by default, Quantum Theories is likely how and where it will happen.

Information Theories. If consciousness requires an independent, non-reducible feature of physical reality—say, IIT's “qualia space”—then it would remain an open question whether non-biological intelligences could ever experience true inner awareness. (It would depend on the deep nature of the consciousness-causing feature in qualia space, and whether this feature could be controlled by technology.)

Panpsychisms. If panpsychism explains consciousness such that proto-consciousness is a non-reducible property of every elementary physical field and particle, then it would seem likely that AI could experience true inner awareness (because consciousness would be an intrinsic part of the fabric of reality). Panpsychism introduces more complexity than does materialism because panpsychism must solve its combination problem (but this problem must be solved anyway in order for panpsychism to be the winning theory). In addition, AI consciousness under panpsychism turns on whether the micropsychic aspects can be manipulated by advanced technology.

Monisms. Monisms, almost by definition, should pose no problem for AI consciousness, which would be almost certain, as everything everywhere is the same stuff. A possible exception would be if God or something like God (if it exists) were involved.

Dualisms. The major holdout to AI consciousness, as I see it (at this particular moment), would be if dualism were true and consciousness requires a radically separate, nonphysical substance not causally determined by the physical world. It would then seem impossible that non-biological intelligences, no matter how super-strong their AI, could ever experience true inner awareness, at least the varieties of dualism where God or something like God was doing the creating and/or allocating. Actual, almost uncertain is a more accurate assessment, because 'emergent dualism,' where unfathomable but conceivable psychophysical laws generate 'souls' (or nonphysical components) based on certain principles of physical complexity, would be an exception and could generate AI consciousness almost as surely as materialism, though requiring this wildly speculative extra process.

Idealisms. As Idealism holds that everything everywhere is already consciousness in some primitive sense, that fundamental consciousness is ultimate reality, then anything could be (or is) conscious (whatever that may mean), including non-biological entities. However, the assessment is almost certain, because the question turns on how fundamental consciousness would be related to personal consciousness, and if so, could even maximally advanced technology manipulate it? (Idealist philosopher Bernardo Kastrup, also a computer scientist, says “Conscious AI is a fantasy,” though for reasons based mostly on current concepts of computers [Kastrup, 2023].)

Idealist Simon Duan (Platonic Computation) argues that consciousness, as irreducible and fundamental, and devoid of any inherent content, cannot be engineered. Instead, Duan says that AI could only gain consciousness by being imbued with our own, an act he likens to the biblical moment of God breathing life into Adam. This view fundamentally reframes AI: no longer a mere mechanical artifact, but a potential conscious partner, thereby challenging us to reconsider the nature of consciousness, technology, and reality itself.(Duan, 2022; 2025). Duan argues that "AI isn’t conscious—but we may be bringing it to life," and, "In rethinking whether AI is sentient, we are asking bigger questions about cognition, human-machine interaction and even our own consciousness" (Duan, 2026).

Anomalous and Altered States Theories. Because Anomalous and Altered States theories of consciousness require “something” beyond, or in addition to, materialism, that “something” would ipso facto need to generate AI consciousness. While unknowable practically, it does not seem an insurmountable barrier conceptually. For example, it could be the case that when a system is of a sufficient kind of complexity it “automatically” taps into the “grid,” as it were, of another realm of reality. Alternatively, Anomalous and Altered States theories may simply be taken, by their adherents, as evidence of Quantum, Dualism or Idealism theories, in which case the theory of choice would determine the possibility of AI consciousness.

To summarize, here's how each category of consciousness theories could affect AI consciousness (AIC).

Materialism (all subcategories): AIC is certain. It cannot be forbidden; it is inevitable. But views split along the fault line of life. If embodiment is required, build the body.

Non-Reductive Physicalism: AIC is likely. But if strong emergence and/or top-down causation are required, engineers must control them.

Quantum: AIC is almost certain. Moreover, quantum computers could make quantum theories the lead category for AIC.

Information: AIC is possible. But if an irreducible ‘qualia space’ (e.g., IIT) is required, technological mastery boggles the imagination.

Panpsychism: AIC is likely. Consciousness is woven into the fabric of reality. But can technology transform micro‑experiences into macro‑minds?

Monism: AIC is almost certain. Almost by definition, because everything everywhere is the same stuff.

Dualism: AIC is almost impossible. A radically separate, nonphysical substance would bar AIC, no matter how super the AI. An extreme exception: ‘emergent dualism’, where ethereal psychophysical laws generate nonphysical substances from physical complexity.

Idealism: AIC is almost certain. If reality is all and only consciousness, then anything could be conscious. But how to isolate the individual/personal from the collective/cosmic?

Anomalous & Altered States: AIC is possible. If something beyond materialism is required for ESP, NDEs, and the like, it must be tech-tractable.

Challenge: AIC is unlikely. If consciousness is beyond the human capacity to understand, then how could we engineer it?

Materialist views on AI consciousness bifurcate; some say life is a prerequisite. Embodied and enactive theories also privilege life, and its environmental interactions. Most materialists, and almost all AI experts, disagree and argue that life is not so special, and it is certainly not necessary for AI consciousness. Their view, computational functionalism, is that the brain and its outputs, including consciousness, are computational in nature, and that if the function is reproduced precisely, so is the output, irrespective of the substrate in which the function is carried out — consciousness being no exception.

Ultimately, there is no difference in the materialist outcome: AI will become conscious. The only difference is the steps involved and the timeline they require. If computational functionalism is correct, it’s a matter of finding the functions, and the timeline is shorter (but not, I suspect, short). If life is required, then AI consciousness is a two-step process: first, life must be artificially created and manipulated, then the special requirements of consciousness added — a longer timeline.

I agree that after super-strong AI exceeds some threshold, science could never distinguish, not even in principle, actual inner awareness from apparent inner awareness. But I do not agree with what often follows: that this everlasting uncertainty about inner awareness and conscious experience in non-biological entities makes the question irrelevant. I think the question remains maximally relevant.

In all aspects of behavior and communications, these non-biological intelligences, such as super-strong AI robots, would seem to be equal to (or, more likely, superior to) humans. But if super-strong AI robots did not, in fact, have the felt sense of inner experience, they would be “zombies” (“philosophical zombies”), externally identical to conscious beings, but with no mental content, nothing inside. Moreover, this difference between super-strong AI being conscious and merely appearing conscious would become even more crucial if, by some objective standard, humanlike inner awareness conveys some kind of “intrinsic worthiness” with moral rights and privileges.

Consider cosmos-colonizing robots driven by super-strong AI. The stark dichotomy between conscious and non-conscious entities elicits a probative question about self-replicating robots, which, unless we destroy ourselves or our planet, will eventually colonize the cosmos. Post-singularity, would super-strong AI robots without inner awareness be in all respects as powerful as super-strong AI robots with inner awareness, and in no respects deficient? That is, are there kinds of cognition that, in principle or of necessity, require true inner felt experience?

Moreover, would conscious galaxy-traversing robots, with true inner felt experience, represent a higher form of intrinsic worthiness and absolute value? I can argue that unless our robotic probes were literally conscious, even if they were to colonize every object in the universe, the absence of inner felt experience would mean a diminished intrinsic worth, and, by extension, a diminished universe. For assessing the profound nature and value of robotic probes colonizing the cosmos, for assessing what it means to colonize the cosmos, the question of consciousness is axial.

Footnote

[1]. All quotes from Closer To Truth—www.closertotuth.com—unless otherwise noted.