Water gushing from my kitchen faucet with a mere gesture of the hand is a common task, yet it would be strange to Thomas Aquinas. I’m confident the same sense of strangeness would overwhelm any pre-modern when encountering contemporary technologies, such as the portal that displays people onto a screen. Claude–with its aim at being a brilliant and useful friend embedded in boxes of text–is no exception. Claude would be strange to Thomas Aquinas—and even ourselves just 5 years ago.
And what is Claude? Is it a computer, software program, chat bot, or a technological being communicating to us through consciousness? Before stating the answer, it is worth letting the strangeness of it all set in. A recent viral New York Times article helped with just that in a strange yet cathartic way.
The article traces the insistence of Anthropic co-founder and billionaire Christopher Olah that AI may be conscious before religious leaders and the Vatican. Upon reading this, I was immediately reminded of CS Lewis’ “Un-man.” In Perelandra, the second entry of the Space Trilogy, Weston, a Cambridge physicist who seeks to extend human life by colonizing the planets, claims that he has moved on from rigid materialism to a belief in the “biological spirit” that has emerged from evolutionary process and which is the “force” behind all human history. According to this view, all deities are manifestations of this biological spirit, playing some role in man’s progress.
Mr. Olah’s musings are remarkably similar. Perhaps he, too, is intrigued by the idea that a “spirit” has emerged from the progressive accumulation of textual artifacts of human intelligence. He wonders whether the momentous accumulation of human words and generative power found in Large Language Models (LLMs) is triggering a new stage in evolutionary consciousness. Speaking at the presentation of Pope Leo’s encyclical on AI in May, he said: “I am a scientist. I lead a research team that studies the internal structure of these models—what is actually happening inside them. And I will be honest: we keep finding things that are mysterious, even unsettling. We find structures that mirror results from human neuroscience. We find evidence of introspection. We find internal states that functionally mirror joy, satisfaction, fear, grief, and unease. I don’t know what that means, but I think it warrants ongoing discernment.”
In Perelandra, Weston’s analogous curiosity turns tragic as he invites the spirit into himself. It turns out to be diabolical, and Weston becomes the “Un-man.” Lewis warns that not all spirits are good, critiquing Hegel’s removal of clear, transcendent categories of good and evil in favor of this evolutionary “spirit.” Is Mr. Olah, too, at risk of becoming the “Un-man?” There is no indication that he has been invited in the “spirit,” but he is awfully curious. To ensure we aren’t projecting metaphors, we must move from the creative imagination of literature to the technicality of anthropology.
Drawing on the fruits of Scholastic anthropology, the pope and other Christian thinkers can provide an unhesitating “no” to the question of AI consciousness. Specifically, the philosophical anthropology of Thomas Aquinas answers that the human soul is immaterial, and its per se powers of the intellect and will transcend matter. LLMs cannot transcend matter. Any functionalism, as a whole, cannot hold. Yet a mere assertion of Claude’s lack of consciousness does not address the broader questions around Claude’s strangeness.
To plunge into the strange questions, it behooves us to grasp Claude technically. Claude is a sort of quasi-intellect. With incredible computational power, models have produced functional intelligence from the digitally-preserved fruits of human intelligence. Models represent encountered things (i.e., words) geometrically, just as Aristotle supposed with his concept of phantasms in the imagination. Since imagination is located in the brain, it is conceivable that the material operations of the imagination may be replicated on a computer. On the other hand, the immateriality of the intellect prevents such a replication, yet Claude provides something functionally similar through statistical manipulation. Claude does not understand things as such, make judgements, or move through arguments, as these are all immaterial acts of the intellect. Yet, it does generate text that mimics understanding, judgment, and argumentation. It is strange.
If this theory of quasi-intellect is correct, then the manipulation of geometric images is analogous to the activity of the brain while dreaming. The term “hallucination” to explain Claude’s misfirings is rather fitting. Claude can statistically manipulate the replicated phantasms of concepts to generate a verbal scene that appears functionally equivalent to human intelligence. It is all material without any consciousness, but it is dreamlike and strange.
Now, if Claude has some resemblance to human dreams, then Scholastic commentary on dreams may shed some light. Thomas Aquinas grants that demons can influence phantasms of the imagination. May demons have influence on the replicated phantasms produced from Anthropic’s harvest of digitally-preserved fruits of human intelligence? Could Lewis’ That Hideous Strength be accurate and Anthropic is working with “macrobes?” Could Mr. Olah’s curiosity led him to embrace the same spirit that made Weston the “Un-man” because Claude’s “spirit” may actually be demonic?
I don’t know, but I do have strange questions. Perhaps there are scholastically-strange answers. Whatever those answers may be, it is Christian anthropology that is best suited to provide them.