Like much modern fantasy, Philip Pullman’s His Dark Materials series features magical objects. Each book in the trilogy is named after one; the second is The Subtle Knife. This knife looks unremarkable, but when wielded by the right person, it can cut a hole in the world — a portal into an alternate universe. Unfortunately, the portal’s edges also turn out to be a wound in the fabric of reality … and the wound bleeds. Sinister, invisible things leak out, hungry ghosts that prey upon the living. To banish these ghosts, the tear in reality must be closed.
Among early modern Europeans, magic was not only taken seriously but was inseparable from mathematics, science, and religion; all fell under the rubric of philosophy. “In effect, philosophy is an incubator for other disciplines,” David Chalmers, an Australian philosopher and cognitive scientist, wrote in 2022. “When philosophers figure out a method for rigorously addressing a philosophical question, we spin that method off and call it a new field.”1
The process can be messy, producing a knowledge patchwork of conspicuously uneven depth and quality. That was increasingly the case in the 17th century. As our understanding of the physical world grew, mounting pressure came to bear on magic and religion to explain what science still could not: life and the mind. How could mere matter feel anything, exhibit agency, or reproduce? How could we be mere matter?
With his “mechanical philosophy,” French polymath René Descartes (1596–1650) forged the subtle knife that promised to resolve this metaphysical crisis by cutting the sciences free of their spiritual encumbrance, severing bodies from souls.2 The trick was to deny the existence of the lower two of Aristotle’s three levels3 of ensoulment: the “nutritive,” possessed by all life forms including plants, and the “sensitive,” possessed by animals. Per Descartes, investigable, clockwork-like mechanisms could fully account not only for the inanimate world but also for plant and animal behavior, without any need for a supernatural animating force. This left only human beings, whose “rational souls” (Aristotle’s highest level) were in keeping with Church doctrine: immortal, immaterial, and indivisible.4
At this point, the model that cut so elegantly needed to confront the discontinuity it had created in the world: Our immaterial souls must perceive the world through bodily senses, as well as control the body’s willed actions. But how?
Descartes reasoned that, since the rational soul was indivisible, it must act upon a central, undivided structure in the brain, which he referred to as “Gland H.” We now know it as the pineal gland — certainly not any special locus of perception or action — but the H might as well have stood for the name of one of the monsters that the Cartesian wound has let slip into our world: homunculus, the “little man” in our brain pulling our strings, who must have, if we are to explain his ability to perceive and act upon the world, an even smaller homunculus within, a dizzying prospect of infinite regress.5 Unless, of course, we allow Descartes’ dualism to stand, in which case the mind is immaterial: a ghost in the machine.6 Or, if we exorcise ghosts and homunculi while acknowledging the possibility that our bodies could continue to give every appearance of intelligent, rational behavior, we are left with the prospect of a philosophical zombie: an entity just like us on the outside, but with no “me” on the inside, no conscious or subjective experience.7
Thus, where body and soul had been severed, three monsters issued forth: homunculus, ghost, and zombie, a funhouse mirror image of the Holy Trinity. Ultimately, these monsters would exact their price, but at first, all seemed in order. New prospects had opened up.
In claiming such a vast territory for reductive materialism — almost the whole world — mechanical philosophy laid the groundwork for Newton’s physics and Lavoisier’s chemistry. By disenchanting the workings of plants and animals, it made biology possible, not only prefiguring the elimination of vitalism from the life sciences but also lighting the way for Lamarck and Darwin. There is even a case to be made for Descartes as the first real neuroscientist, for he both acknowledged how similar human brains are to those of animals and stipulated that — at least in animals — the entire pathway from stimulus to response could be understood by examining the workings of the nervous system.8
However, Descartes knew that his mechanical philosophy would need to stop short of attempting to explain human reason. It was a necessary price not only for his ideas to gain traction within the Church but also for him to avoid the religious persecution Galileo had undergone. (A century later, Julien Offray de La Mettrie extended mechanical philosophy to the human mind in his infamous L’homme machine, or Man a Machine,9 which he had the good sense to publish anonymously. When La Mettrie was outed, he indeed got in trouble.)
Still, Descartes appears to have believed sincerely in the human soul, not only because of religious faith but also because he could not imagine how mechanical cause and effect could enable open-ended problem-solving, or what we might today call “general intelligence”:
[I]f there were machines bearing the image of our bodies, and capable of imitating our actions […] there would still remain […] tests whereby to know that they were not […] really men. […] [A]lthough such machines might execute many things with equal or perhaps greater perfection than any of us, they would […] fail in certain others from which it could be discovered that they did not act from knowledge, but solely from the disposition of their organs: for while reason is an universal instrument […] it must be […] impossible that there should exist in any machine a diversity of organs sufficient to enable it to act in all the occurrences of life […].10
For similar reasons, Descartes also believed that language lay beyond the purview of any machine — and hence, of any animal-machine:
[T]here are no men so dull and stupid […] as to be incapable of joining together different words […] by which to make their thoughts understood; […] on the other hand, there is no other animal […] which can do the like. Nor does this inability arise from want of organs: for […] parrots can utter words like ourselves, and are yet unable to speak as we do, that is, so as to show that they understand what they say […].11
If Descartes had lived to see his 400th birthday, he would likely have felt vindicated. He would have seen physics and chemistry flourish, the mysteries of biology slowly unravel, and an age of extraordinary machines arise. In the 1940s, general-purpose computing was born, and with it, software, making it possible to instantly build, modify, and experiment with virtual machines of far greater complexity than anything that could be manufactured physically. Descartes would have witnessed the early enthusiasm of artificial intelligence pioneers in the 1950s and the ensuing AI winters of the 1970s and 1990s.
Classic computer systems and what we now call “good old-fashioned AI”12 (GOFAI) could carry out a great many tasks with “greater perfection than any of us,” but they were indeed unable to do what even the “dull and stupid” can do: act from knowledge, reason about novel situations, and string together words to communicate meaningful thoughts. It seemed that Ada Lovelace (1815–1852), who had anticipated the power of programming a century before the first program would actually run, had been right to assert that a computer could have “no pretensions whatever to originate anything. It can [only] do whatever we know how to order it to perform.”13
However, 20 years after the turn of the millennium, all that would change. In large language models (LLMs), we finally have machines that appear to do exactly what Descartes believed they never could: perform arbitrary tasks in the language domain. There has been pushback, some of it strongly reminiscent of Descartes, such as the assertion that an LLM is “a system for haphazardly stitching together sequences of linguistic forms […] without any reference to meaning: a stochastic parrot.”14 This quote is from a paper written in 2021, when LLMs were far less capable than they are today. Back then, interactions with them could have a Rorschach inkblot quality, with those who were more forgiving recognizing the “spark”15 of an intelligent entity on the other end of a chat and those who were more skeptical seeing only “embers.”16
In 2026, the “stochastic parrot” narrative is increasingly hard to reconcile with the substantive conversations millions of people have with frontier models every day. But then again, the argument was always more philosophical than evidentiary. It relied on the idea that the manipulation of symbols (syntax) can never yield understanding or meaning (semantics).17 The trouble is that “machinery,” whether natural or artificial, can only manipulate symbols. That’s what a computer does. It’s just ones and zeroes in there, regardless of what they encode: text, sound, image, or some internal variable. It’s also what a brain does. It’s just firing neurons in there, regardless of what they encode. What, then, do we mean by “meaning”?
The dogma that the human mind and AI have nothing in common is increasingly tenuous — just as the dogma that humans and nonhumans differ in some immaterial ‘essence’ is and has always been tenuous.
In a Cartesian world, “meaning” is ghostly. Indeed, its absence is evidence of one of the monsters Descartes unleashed, for although scientists no longer discuss immaterial souls, most also avoid revisiting the wound, so it remains open. Where the monsters came in, something, too, went out of the world. It was drained of agency, rationality, value, purpose, and meaning, for these had belonged to God — and to us, since the divine spark was purportedly within each human but absent from everything else, living or not.
The Cartesian wound has taken many forms, each a manifestation of the supposed divide between the world of matter and the world of mind, purpose, and meaning. One is the “blind spot,” in which an experimenter imagines being able to stand outside a system under study, free to analyze it with perfect objectivity, as though severed from the world they observe.18 But the experimenter is, of course, always part of the world, too, and ignores the subjectivity inherent to their perspective at their own peril.19
Relatedly, the fact-value distinction — a supposed gulf between describing what is and prescribing what matters20 — is based on the fiction of a “view from nowhere” when it comes to “facts.” That there is no view from nowhere does not license the radical claim that reality is a blank slate, nor does it justify moral relativism,21 the related claim that any perspective is just as valid as any other. These nihilisms covertly presuppose the very view from nowhere22 (or everywhere) they claim to renounce. Acknowledging multiple perspectives doesn’t render the universe a meaningless void, but rather requires us to recognize that the universe includes us and that most of what we consider “reality” — including ourselves and each other — consists of mental constructs.
Then there is the admonition against teleology, which insists that purpose has no place in nature — that Aristotle’s “final causes” were a pre-scientific confusion, with purposiveness either denied altogether (if we reject intelligent design) or quarantined inside human minds. This is hard to square with biology, since it requires that we reject obviously helpful statements like “one of the purposes of the kidney is to filter urea out of blood.” Much of biology is, in fact, dedicated to figuring out the relationships between function and implementation: how living systems achieve their ends and, conversely, what newly discovered mechanisms are for.23
We also find ourselves caught in the anthropocentrism/anthropomorphism trap, whereby we are entreated not to view the world purely as an instrument for human use (anthropocentrism), yet any attribution of thought, agency, language, reasoning, emotion, or intention to a nonhuman entity is deemed an inappropriate projection (anthropomorphism).24 Well before AI came along, we already had plenty of animal videos on YouTube (not to mention reams of scientific evidence25) to show us how often our nonhuman kin exceed their brief.
Finally, we have the “hard problem” of consciousness.26 The problem is only “hard” because, when we imagine a philosophical zombie, we are imagining a dualistic universe in which a soulless non-person could convincingly act as if they were ensouled. But here, ironically, Descartes had something right: A rule-based machine that’s only capable of imitating our actions by rote — that is, a GOFAI system — cannot convincingly behave like an intelligent, socially appropriate entity. If it could, we would have had Gemini, Claude, and ChatGPT decades earlier. LLMs are not programmed to carry out specific tasks, nor do they merely imitate; they learn, and like us, they do so by generalizing from individual examples and feedback to broader patterns. And when we look inside the models to better understand how they generalize and reason, we often find internal representations that mirror ours.27
That doesn’t mean these models work the same way our brains do (they do not), nor does it mean they are equivalent to people.28 But it does mean the dogma that the human mind and AI have nothing in common is increasingly tenuous — just as the dogma that humans and nonhumans differ in some immaterial “essence” is and has always been tenuous.
We all live in the same world. It’s time to heal the wound and rid ourselves of philosophical monsters. Let agency, purpose, and value flow back into our world, not as any single being’s private endowment, but as the basis for the reciprocal relationships that allow for mutual understanding and mutual aid. It’s time, also, to admit what the wound let us forget: that humanity has never been a crowd of self-contained Cartesian souls, but instead a collective intelligence, one whose progress and growth are inextricably linked to its widening circle of care.
1: Chalmers 2022.
2: Descartes 1662.
3: Aristotle [ca. 350 BC] 2016.
4: Riskin 2016.
5: Kenny [1971] 2016.
6: Ryle 1949.
7: Kirk and Squires 1974.
8: Per Lecture I of Pavlov 1927, “Descartes evolved the idea of the reflex. Starting from the assumption that animals behaved simply as machines, he regarded every activity of the organism as a necessary reaction to some external stimulus, the connection between the stimulus and the response being made through a definite nervous path: and this connection, he stated, was the fundamental purpose of the nervous structures in the animal body. This was the basis on which the study of the nervous system was firmly established.”
9: de La Mettrie 1748.
10: Descartes [1637] 1850.
11: Descartes [1637] 1850.
12: Haugeland 1985.
13: Lovelace and Menabrea 1843.
14: Bender et al. 2021.
15: Bubeck et al. 2023.
16: McCoy et al. 2023.
17: Searle 1980.
18: Husserl 1954; Frankish et al. 2017; Goff 2019.
19: Frank, Gleiser, and Thompson 2024.
20: Hume [1739] 1817; Moore 1903; Parfit 2011.
21: Williams 1972.
22: Nagel 1989.
23: Attempted workarounds have included “teleonomy,” the “as-if” purposiveness of evolved systems; Pittendrigh 1958; Mayr 1961. Dennett’s “intentional stance” is similarly “as-if” with regard to agency; Dennett 1989.
24: de Waal 1999.
25: Galpayage Dona et al. 2022; Pardo et al. 2024; Sogawa et al. 2025.
26: Chalmers 1995.
27: Schrimpf et al. 2018; Huh et al. 2024; Gurnee et al. 2026.
28: Leibo et al. 2025.
References
Aristotle. (ca. 350 BC) 2016. De Anima. Translated by Christopher Shields. Clarendon Aristotle Series. Oxford: Clarendon Press.
Bender, Emily M., Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜.” In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610–23. FAccT ’21. New York, NY, USA: Association for Computing Machinery.
Bubeck, Sébastien, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, et al. 2023. “Sparks of Artificial General Intelligence: Early Experiments with GPT-4.” arXiv [cs.CL].
Chalmers, David J. 1995. “Facing Up to the Problem of Consciousness.” Journal of Consciousness Studies.
Chalmers, David J. 2022. Reality+: Virtual Worlds and the Problems of Philosophy. W. W. Norton & Company.
Dennett, Daniel Clement. 1989. The Intentional Stance. MIT Press.
Descartes, René. 1662. De Homine. Translated by Florentius Schuyl. Leyden: Peter Leffen & Francis Moyard.
Descartes, René. (1637) 1850. Discourse on the Method of Rightly Conducting the Reason, and Seeking Truth in the Sciences. Translated by John Veitch. Edinburgh: Sutherland and Knox.
Frank, Adam, Marcelo Gleiser, and Evan Thompson. 2024. The Blind Spot: Why Science Cannot Ignore Human Experience. London, England: MIT Press.
Frankish, Keith, Katalin Balog, Susan Blackmore, Daniel C. Dennett, Jay L. Garfield, Michael S. Gazzaniga, Philip Goff, et al. 2017. Illusionism: As a Theory of Consciousness. Edited by Keith Frankish. Imprint Academic.
Galpayage Dona, Hiruni Samadi, Cwyn Solvi, Amelia Kowalewska, Kaarle Mäkelä, Hadi MaBouDi, and Lars Chittka. 2022. “Do Bumble Bees Play?” Animal Behaviour 194: 239–51.
Goff, Philip. 2019. Galileo’s Error: Foundations for a New Science of Consciousness. Vintage.
Gurnee, Wes, Nicholas Sofroniew, Adam Pearce, Mateusz Piotrowski, Isaac Kauvar, Runjin Chen, Anna Soligo, et al. 2026. “Verbalizable Representations Form a Global Workspace in Language Models.” Transformer Circuits Thread. https://transformer-circuits.pub/2026/workspace/index.html.
Haugeland, John. 1985. Artificial Intelligence: The Very Idea. Bradford Books.
Huh, Minyoung, Brian Cheung, Tongzhou Wang, and Phillip Isola. 2024. “The Platonic Representation Hypothesis.” arXiv [cs.LG]. arXiv.
Hume, David. (1739) 1817. A Treatise of Human Nature: Being an Attempt to Introduce the Experimental Method of Reasoning Into Moral Subjects. London.
Husserl, Edmund. 1954. Die Krisis Der Europäischen Wissenschaften Und Die Transzendentale Phänomenologie: Eine Einleitung in Die Phänomenologische Philosophie. Edited by Walter Biemel. Vol. VI. Husserliana. The Hague: Martinus Nijhoff.
Kenny, Anthony. (1971) 2016. “The Homunculus Fallacy.” In Investigating Psychology, 1st Edition, 155–65. Routledge.
Kirk, Robert, and Roger Squires. 1974. “Zombies v. Materialists.” Aristotelian Society Supplementary Volume 48 (July): 135–64.
La Mettrie, Julien Offray de. 1748. L’homme machine. Leiden.
Leibo, Joel Z., Alexander Sasha Vezhnevets, William A. Cunningham, and Stanley M. Bileschi. 2025. “A Pragmatic View of AI Personhood.” arXiv [cs.AI]. arXiv.
Lovelace, Ada, and Luigi Federico Menabrea. 1843. “Sketch of the Analytical Engine Invented by Charles Babbage, by LF Menabrea, Officer of the Military Engineers, with Notes by the Translator.” Taylor’s Scientific Memoirs.
Mayr, Ernst. 1961. “Cause and Effect in Biology: Kinds of Causes, Predictability, and Teleology Are Viewed by a Practicing Biologist.” Science (New York, N.Y.) 134 (3489): 1501–6.
McCoy, R. Thomas, Shunyu Yao, Dan Friedman, Matthew Hardy, and Thomas L. Griffiths. 2023. “Embers of Autoregression: Understanding Large Language Models through the Problem They Are Trained to Solve.” arXiv [cs.CL]. arXiv.
Moore, George Edward. 1903. Principia Ethica. Cambridge: Cambridge University Press.
Nagel, Thomas. 1989. The View from Nowhere. Oxford University Press.
Pardo, Michael A., Kurt Fristrup, David S. Lolchuragi, Joyce H. Poole, Petter Granli, Cynthia Moss, Iain Douglas-Hamilton, and George Wittemyer. 2024. “African Elephants Address One Another with Individually Specific Name-like Calls.” Nature Ecology & Evolution 8 (7): 1353–64.
Parfit, Derek. 2011. On What Matters. Vol. 1 and 2. Oxford: Oxford University Press.
Pavlov, Ivan P. 1927. Conditioned Reflexes: An Investigation of the Physiological Activity of the Cerebral Cortex. Translated by G. V. Anrep. London: Oxford University Press.
Pittendrigh, Colin S. 1958. “Adaptation, Natural Selection, and Behavior.” Behavior and Evolution.
Riskin, Jessica. 2016. The Restless Clock: A History of the Centuries-Long Argument Over What Makes Living Things Tick. University of Chicago Press.
Ryle, Gilbert. 1949. The Concept of Mind. London: Hutchinson.
Schrimpf, Martin, Jonas Kubilius, Ha Hong, Najib J. Majaj, Rishi Rajalingham, Elias B. Issa, Kohitij Kar, et al. 2018. “Brain-Score: Which Artificial Neural Network for Object Recognition Is Most Brain-Like?” bioRxiv. bioRxiv.
Searle, John R. 1980. “Minds, Brains, and Programs.” The Behavioral and Brain Sciences 3 (3): 417–24.
Sogawa, Shumpei, Taiga Kobayashi, Redouan Bshary, Will Sowersby, Satoshi Awata, Naoki Kubo, Yuta Nakai, and Masanori Kohda. 2025. “Rapid Self-Recognition Ability in the Cleaner Fish.” Scientific Reports 15 (1): 41882.
Waal, Frans B. M. de. 1999. “Anthropomorphism and Anthropodenial: Consistency in Our Thinking about Humans and Other Animals.” Philosophical Topics.
Williams, Bernard. 1972. Morality: An Introduction to Ethics. New York: Harper & Row.
This article AI is forcing us to reckon with a 400-year-old problem is featured on Big Think.