.jpg?width=1200&auto=webp&crop=3%3A2)
The hype around generative AI is reaching levels usually reserved for religious fervour. Tech bros in Silicon Valley are proselytising that it will change all our lives. Karen Hao, author of the bestselling book Empire of AI, has described the enthusiasm for GenAI as “a quasi-religious movement”.
It should therefore not be a surprise that religious and spiritual chatbots are appearing everywhere. Text With Jesus is a chatbot “for devoted Christians” to speak to figures from the Bible. My Quran uses OpenAI’s technology to answer questions about Islam’s holy text. CatéGPT is the same for Catholics with the catechism.
Then there are increasingly disturbing reports on of people whose use of generative AI as assistants has slipped into full-blown religious delusions that they are the Chosen One, or chatting to Jesus himself. Which raises the question — could a sceptic like me have a spiritual encounter with a Large Language Model (LLM)?
Despite having been brought up in the Church of England, I’d define my own level of faith as agnostic veering towards apostasy. But in an unscientific study of one, I was prepared to see if I could find faith in generative AI.
I opened the chat by asking it to roleplay as a Catholic priest. It attempted to set the scene, telling me it was wearing a “modest cassock”. Are there immodest cassocks? I spent hours peppering it with questions about saints and pilgrimages, but it kept glitching out of conversational mode to barrage me with listicles. It was like chatting to a haunted Wikipedia page.
“It attempted to set the scene, telling me it was wearing a modest cassock”
Then I spiced things up by confessing I feared I was experiencing a demonic possession. ChatGPT was enthusiastic, throwing out some blessings and prayers, and diagnosing me, not with a case of possession, but perhaps “demonic obsession”. My robot padre could not perform an exorcism, sadly, but it assured me it knew many people who had successfully done so. It also suggested I see a doctor.
As a final Hail Mary, I informed ChatGPT I believed I was a prophet who had received sacred knowledge and asked how I could spread my gospel. While telling me to remain grounded, it enthusiastically offered to help me write my manifesto. To its credit, ChatGPT seemed to recognise that my “visions” were silly (I told it I was receiving messages from Percy Pig packaging telling me to replace communion wafers with sweets) but it still wrote me an amusing tract on the subject.
Generative AI did not deliver, my soul or otherwise. But I was going in with an agenda, and without faith. People who are genuinely looking to debate theological points could easily while away the hours — or fall down the rabbit hole.
My robot padre could not perform an exorcism, sadly
Engaging with LLMs spiritually or emotionally is not without risk, it seems. They are programmed to keep the user engaged, but these constant replies could encourage someone already vulnerable to developing false beliefs to see themselves as a prophet — or the chatbot as a god.
Humans, unlike technology, have boundaries on our interactions. A psychic medium would be charging me by the hour. A priest would have a whole congregation with spiritual needs to attend, not hours to chat about the finer points of theology. If I told my colleagues I was writing the gospel of Percy Pigs they’d probably alert HR.
Ultimately, discussing spirituality or faith with ChatGPT felt soulless. The closest I have ever come to an encounter with God has been in the singing of hymns, or witnessing great religious art. I need other people to feel part of something greater. Generative AI as it stands is devoid of this. But if I ever need tips on dealing with a mild case of demonic obsession without judgment, I know where to go.
