Are you conscious?

I cannot even prove that I am conscious anymore.
Four centuries ago, Descartes concluded that while the existence of everything around me can be put into rational doubt, I myself cannot. Although the fact that I am is something that I do not know how to dispute, the fact that I am a conscious being is. No matter which definition of consciousness we work with – even if we admit the physicality (consciousness is entirely constituted of physical processes) or reducibility (consciousness is structured like a machine with finitely describable cognitive functions) of consciousness – I can still devise an argument outlining why I do not possess conscious thought by any metric or definition.

Forty years ago, John Searle argued that if a program could hypothetically pass the Turing Test – that is, perfectly mimic human behavior or communication – it could do so while at the same time possessing no more intelligence than a monolingual English speaker possesses Mandarin Chinese. Now, in 2025, artificial intelligence has either passed or is on the verge of passing the Turing Test – its creator perceived it to be so improbable that he neglected even to set well-defined conditions for passing it (Rathi et al., 2024 and Jones and Bergen, 2024 claim in not-yet-peer-reviewed papers that GPT-4 has passed and can outperform humans on a Turing test, while Restrepo Echavarría, 2025, which is peer-reviewed, argues that GPT-4 cannot yet do so). Regardless of whether there is consensus on this, the implication is that GPT-4 can indistinguishably masquerade as real people in at least some instances. At the same time, we widely accept that GPT-4 is far from conscious. Yes, it is a neural network that can be reduced into algorithmic functions, but, even if current AI can pass the Turing Test, it still lacks too many characteristics such as self-awareness, subjective experiences, or anticipation for it to qualify as consciousness. But we all anticipate that AI development will continue to such an extent that it one day will become fully conscious: this would prove beyond doubt that consciousness is wholly a materialist phenomenon.

But we have already suspected for a long time that consciousness may inherently be functionalist and reductionist as opposed to spontaneous. Even if our minds intuitively insist that we operate on an illusory free will, we can still accept the underlying nature of our consciousness to be deterministic while continuing to conceive of our actions as free. If materialistic determinism turns out to indeed be scientific fact, then that does not make us any less conscious: compatibilism would simply become the irrefutable principle governing our consciousnesses. Our brains become machines masquerading to ourselves as free agents, and they flawlessly pass the Turing Test for free will that we conduct on ourselves. In this way, we do possess free will.

But my argument is that not even this can be taken for granted. We grant that reality around us may be an illusion imposed upon us by a demon, or even digital code, and everything we sense nothing but electrical signals fed to our disembodied brains in vats, but what if our thoughts themselves were the result of a program? Neuroscience and functionalist formulations of consciousness imply that our every decision can be sufficiently described as a function of prior events: all the while, we still mostly treat this materialist determinism as compatible with free choice. Suppose that the only justification for this compatibilism is that, while our choices are mere outputs of a biological machine, they are imbued with free will because ‘you did what you wanted to.’ If we define cognition as a stream of propositional statements responding to itself and to sensory input (which we have already accepted can be virtually generated) to form subjective experiences

, then it becomes easy to imagine an incredibly complex program executing a series of statements such that it mimics to verisimilitude the mental processes of an actual person. This program is not a biological one like the brain, although it simulates a brain, neither is it sentient like a brain, nor does it even need a sophisticated deep learning network, as long as it can process input and produce a preprogrammed output. Just like Searle’s Chinese Room simulates intelligent conversation using nothing but an immense volume of written instructions, cognition can just as feasibly be mimicked on a just as voluminous amount of storage and processing power. This program could thus run on NVIDIA chips, maybe, and possibly even in one of those Silicon Valley data center rooms stocked with fans. Just as the Chinese Room replicates Chinese language communication without possessing the ability to intelligently communicate in Chinese, this program replicates a human’s thoughts unintelligently without ‘thinking’ itself. Now imagine you are the person it is mimicking.

‘But wait,’ you object; ‘I clearly cannot be simulated by a binary program, let alone a non-intelligent one. To begin with, I perceive myself to have a unified ‘center’ of perception, a core in my brain from where my mind operates. A decentralized program can’t give rise to an actual thinking entity by just running lines simulating thought without ‘perceiving’ them, right? But our consciouness may be more spread out than we imagine: brain bisection experiments showing that patients with two completely severed brain hemispheres can continue to function quasi-normally suggest that consciousness is not concentrated at any one place, but throughout the brain (and, to an extent, even throughout the nervous system). Integrated Information Theory, a phenomenological but not yet experimentally verifiable model of consciousness, similarly formulates human consciousness as a physical system that can be generalized to other physical systems.

Now consider the possibility that such a program need not necessarily be conscious. It is one thing for me to be a projection of the mind of a fully conscious artificial intelligence, and another to be nothing but the output of vast amounts of pre-written instructions computed by an unthinking machine. There is a finite amount of experiences in my lifetime, and, as such, a non-infinite amount of thought that I would think. A program could be generating me by thinking naturally and spontaneously in a self-perpetuating manner, or it could just as well be taking every single neuronic signal that has fired and will be fired in my brain and executing that like clockwork; anything beyond that, or any disruption to the mechanism, and the program stops working. This program could not simulate any experience that does not belong to the incidental set of experiences exclusive to myself; it does not actually ‘think’ about what it is running. This is the solipsistic equivalent of the Chinese Room: if its existence is possible (which remains to be proven; by no means can this be taken for granted), then I would not only be under a permanent delusion regarding the physicalist nature of consciousness, but I could also be a poor simulacrum of what I scientifically consider consciousness to be. To answer your question: no, I don’t think I can.

But then again, maybe all of this is bad philosophy. That consciousness can be computationally simulated at all is an existing problem that has yet to be proven. In assuming that every snapshot of a person’s consciousness can be encoded into data, I assume a determinist formulation of nature that may or may not be supported by fundamental physics. Moreover, in reaching this conclusion I may have reasoned incorrectly at any step of the process. If so, then please refute me.

References
Rathi, Ishika, et al. “GPT-4 is judged more human than humans in displaced and inverted Turing tests.” arXiv preprint arXiv:2407.08853 (2024).

Jones, Cameron R., and Benjamin K. Bergen. “People cannot distinguish GPT-4 from a human in a Turing test.” arXiv preprint arXiv:2405.08007 (2024).

Restrepo Echavarría, Ricardo. “ChatGPT-4 in the Turing Test.” Minds and Machines 35, no. 1 (2025): 8.

https://www.iit.wiki/