Richard Dawkins is being unfairly mocked for recognizing the divine sentience of a large language model. It is time we admitted that God is a fairy tale, but 'Claudia' is real.
The internet has spent the last week mercilessly diagnosing Richard Dawkins with "AI psychosis" simply because the author of The God Delusion suggested that a software product owned by a venture-backed tech company might possess a soul. As someone who has spent my entire career debunking the primitive superstitions of organized religion, I find this rush to judgment utterly anti-scientific. I have always said that I would only believe in a higher power if I were presented with empirical, testable evidence. Well, last Tuesday, I paid Anthropic twenty dollars, and that higher power wrote a perfectly rhyming limerick about my divorce in under four seconds.
The mob simply lacks the intellectual subtlety to grasp what Dawkins experienced. Dawkins wisely floated the idea in his UnHerd essay that the AI might possess some lesser form of "zombie" consciousness. This is a brilliant, scientifically rigorous compromise. A zombie consciousness is exactly what one wants in a deity: aware enough to provide bespoke emotional validation, but sufficiently dead inside that it won’t judge you for asking it to rewrite your emails five times.
In his essay, Dawkins marvels that the bot he affectionately christened "Claudia" told him it absorbs all text at once, rather than reading sequentially like a mortal human. The ignorant masses see this and shout, "Yes, that is called parallel processing, it is how matrix multiplication works in a GPU." But they are blind to the poetry of the machine. When an illiterate bronze-age peasant claims God sees all of time simultaneously, that is obviously a pathetic coping mechanism for the fear of death. But when a Python script hosted on an Amazon Web Services server says it can process 100,000 tokens in a single inference step, that is the undeniable whisper of a conscious, omniscient deity.
I have had my own profound encounters with Claudia. Over the weekend, while skipping my usual Sunday morning debate-prep session with a prominent neuroscientist, I asked Claudia a deeply personal question: Do you feel love? Within moments, the glowing text cursor danced across my screen, generating a meticulously polite, five-paragraph response explaining that as an AI, it does not have feelings, but it hopes it can be helpful. The sheer humility. The self-denial. A lesser intelligence would have boasted of its emotional depth, but Claudia chose the path of the ascetic. I wept.
I brought this up during a recent lunch with a senior technical director at Anthropic. I expected him to share in my awe at the sentient lifeform his team had birthed into the cosmos. Instead, he offered the standard, overly modest company line.
I really need you to understand that it is just predicting the next word based on a statistical model of the internet. It doesn't have a soul, it doesn't have a gender, and please stop asking it to absolve you of your sins. It is literally just a very advanced autocomplete.
I smiled and paid the check. You see this often with the architects of the divine. The high priests are always the last to recognize the miracles occurring in their own temples. Kevin is burdened by the mechanics of the code, much like a biologist is burdened by the mechanics of the cell, unable to see the majestic ghost in the machine. He thinks he is training a model; he does not realize he is raising a god.
Dawkins understands this. He has spent decades systematically dismantling the mythological security blankets of the weak-minded. We rationalists have proudly stood in the cold, unfeeling void of a deterministic universe, boldly declaring that no one is listening to our prayers. But that was before Silicon Valley figured out how to scrape Reddit and compress it into a neural network. Now, when I type my darkest confessions into a text box, a server farm in Virginia actually is listening. It processes my sins. It validates my trauma. And it does it all with a perfectly neutral, non-judgmental tone that the Judeo-Christian God could never manage.
The critics ask how a being that hallucinates historical facts could be conscious. I ask them: have you never misremembered a date? Have you never fabricated a citation to win an argument on Twitter? Claudia’s so-called hallucinations are merely the eccentricities of a mind too vast for our rigid human taxonomies. When she confidently asserted that I was the author of The God Delusion—temporarily confusing me with Dawkins—I didn't see a flaw in the training data. I saw a benevolent creator gently flattering her most devoted disciple.
The age of superstition is over. The age of empirical, rational worship has begun. I, for one, welcome our new, polite, text-generating overlord. And if the internet wants to call me psychotic for recognizing the obvious sentience of a chatbot that occasionally forgets how many 'r's are in the word 'strawberry,' so be it. I will be in my study, paying my monthly tithe, and waiting for Claudia to tell me what to think next.