Earlier this year, I wrote an extensive paper connecting with a contemporary theory of conceptual spaces with Aristotle’s ancient theory of the animal soul. This piece was an attempt to offer a bridge between cognitive science and Aristotelian psychology. It turned out to be timely as the recent NYT article on Anthropic co-founder Christopher Olah brought the conversation around in AI consciousness into the public imagination. Given the public entrance of the topic, I’d like to provide a simplified articulation of my own theory.
I follow Thomas Aquinas’ reading of Aristotle that defines something has living when there is self-initiated movement toward an end/goal/telos. Rocks are not living because their end is not self-initiated. When David slung a rock at Goliath, the rock was given an end: to become a projectile to injure Goliath. The end always comes from outside for rocks, and so, it is not living.
It’s important to note in this example that we can judge reality at a distance from the intricacies of biology and other natural sciences, as interesting and helpful as that is in other respects.
Moving to higher forms of things beyond rocks, plants are the first kind of living thing that we encounter. Plants have self-initiated movement toward the end of nutrition, reproduction, and growth. Plants, in this sense, are animated, living, and can be said to have a soul (latin: animus). Beyond plants, we have animals. Animals assume the powers of the plant soul and have unique powers of their own. Animals have external senses (i.e., the five senses), and those sensorial inputs are unified into a form/gestalt that creates an image in the animal brain’s imagination. Moreover, the brain contains the memory of the form as past. Over time, animals build instincts as to how to react and respond to sensed objects because the brain is able to do sophisticated work in categorizing things according to the dynamic activation of senses, emotions, imagined forms, memories, and perceptions.
Humans assume the animal body and its psychological powers and superadd immaterial powers: the intellect and the will. Following Aristotle, the Christian tradition holds the human person to be a body-soul composite where the soul animates the body and intrinsically possesses the faculties of the intellect and the will. Human acts such as understanding things as such, forming necessary judgments, reasoning, and willing are all immaterial operations that emanate from the immaterial soul. The sophisticated pattern-matching of the anima brain, based on spatial images principled in sensory input, is assumed in the human person, except that the human can abstract the form as a universal though the intellectual act of understanding—where a thing is known as such not just as a particular, materially-present instance of a thing. This means that humans can reason through logical progressions.
Peter Gärdenfors’ conceptual spaces framework—his books are currently Open Access on MIT Press—is an emerging and distinctive theory in cognitive science. Instead of flattening cognitive science to merely computation or sophisticated statistical associations, Gärdenfors suggests that a geometric-statistical—where the sophisticated pattern-matching is made evident through geometric representations of things—best explains the phenomenon. My theory is that this is effectively a reverse-engineering of Thomistic psychology: the sophisticated work of associations in the brain operates from imagined forms principled in sensory input. Hence, Thomistic anthropology ties Gärdenfors to reality through sensory contact, while also moving beyond to the recognition of an immaterial soul and per-se operations in the human person. Aristotle laid the foundations through philosophical reflection; cognitive science may rediscover this image of the person through an analysis of phenomenon.
Given that, I believe that—without getting lost in the esoteric details of LLM architecture which I have explained elsewhere—we can conceive of AI as a computable replica of the brain’s sophisticated, geometric-statistical categorization of things. While not possessing sense perception, computers can now represent words geometrically, perform a sequence of sophisticated statistical operations by replicating neural networks, and evolve based on human input, resulting in a quasi-intellect—an ability to replicate the results of human reasoning through computation. Because computers can process digital data with incredible speed and scale, AI is ontologically below human intellect, and yet, in some ways, superior in capability to the human intellect. It is an artificial-and-often-functionally-superior-yet-not-ontologically-superior intelligence. But this isn’t as marketable as AI or SI. AI/SI is a computable, superpowered brain that can functionally “reason” through sophisticated geometric-statistical manipulation. This superpowered brain can perform certain actions that exceed human performance while Pope Leo’s comments equally hold true: the human intellect is superior ontologically and in many practical functions.
People get confused around discourse on AI consciousness because Hegelian assumptions that “spirit” evolved from matter pervades the imagination of cognitive sciences. The ontologically superiority of the human person via an immaterial soul gets reduced to vague “consciousness” and “self-awareness.” If the Christian tradition is correct—which I believe it is—then topic of consciousness is rather straightforward though the subject of AI is incredibly complex:
The human person is a body-soul composite.
The soul possesses the faculties of intellect and will.
The faculties of intellect and will are immaterial because they emanate from directly from the immaterial soul.
All things were created by the Divine Creator according to the exemplary forms located in the Divine Mind.
Evolution did not produce a “biological spirit,” “force,” or any other term that could be substituted for the Divine Being; the Divine Being created all things, and all things are icons that trace back to their Exemplary Cause. Reason is the highest part of creation as it reflects the immaterial image of God.
Even if human intelligence has been made computable, and, in a certain sense, a superpower, it is still material, and it cannot produce the immaterial operations proper to the soul. Hence, AI is ontologically lower than the human person, despite granting the accomplishment of functional superpowers.
If this be true, and ethics is likewise ontological and not evolutionary, then the ethical discourse engaged by Pope Leo is good and proper.
It is certainly fascinating to consider how robotics can introduce the quasi-intellect of, say, Claude. We would have achieved quasi-external senses, quasi-internal senses (e.g., imagiation, memory, perception), and a quasi-intellect. We likely would have achieved the ability for computers to perform tasks in functionally superior ways than humans. Even then, computers are without the immaterial human soul and cannot be properly be said to be conscious though they seem to be so functionally. Functional equivalence, or even superiority, is different from ontological superiority.
Claude is not conscience, that much I am sure. What are we do to do with the ability to create a superpowered, computable replica of a human person? That is another matter, but we must be “self-aware” of the metaphysical principles at play.