Chatting With AI ‘Saints’ — Opportunity or Peril?

[Michie]

A Catholic tech start-up is offering ‘conversations’ with AI-generated saints — while theologians warn of the spiritual risks such tools might pose.

Inset image: Antonio del Castillo y Saavedra (1616–1668), “St. Thomas Aquinas,” Museum of Fine Arts of Córdoba, Spain (photo: Tero Vesalainen / Public Domain / Shutterstock)

St. Augustine of Hippo lived eight centuries before St. Thomas Aquinas and never met him — at least not on this side of heaven.

But imagine, if you will, that these two doctors of the Church both walked into a room and commenced a dialogue on matters of faith, with you seated comfortably nearby as an observer. What would they discuss, and how would each respond?

Concocting such a transcript could be a worthwhile exercise for a scholar or student, as doing so would require a deep understanding of each saint’s writings, thought and historical context. But could artificial intelligence ever be capable of producing such a conversation?

Magisterium AI is attempting just that.

Magisterium, a project of the Catholic start-up Longbeard, is an AI product focused on making Church teaching and Catholic insight more accessible. The interface is similar to other “generative” AI websites, such as ChatGPT. But unlike ChatGPT, Magisterium is trained only on Catholic sources, and its creators aim to feed it every piece of Catholic knowledge ever created, with the goal of making that knowledge ultimately searchable and accessible for anyone on the internet.

Continued below.

Chatting With AI ‘Saints’ — Opportunity or Peril?

A Catholic tech start-up is offering ‘conversations’ with AI-generated saints — while theologians warn of the spiritual risks such tools might pose.

www.ncregister.com

 

[RileyG]

Hmmm...this honesty makes me uncomfortable, since AI just repeats what it learns from its human developers. Can technology ever be truly "smart"?

 

[timewerx]

AI just repeats what it learns from its human developers.

That's not the case at all. Sounds like you haven't used LLMs (Large Language Models) before like chatgpt or haven't use one long and extensive enough.

 

[Michie]

Vatican Guidance and Concerns​

In the few short years since such technology became widely available, Catholic observers and scholars of AI have spoken out in unambiguous terms against the anthropomorphizing of AI systems — cautioning that AI should be treated only like the tool it is and not as a replacement for true human interaction.

Under the late Pope Francis, the Vatican’s Dicastery for the Doctrine of the Faith (DDF) in January released a note on AI, Antiqua et Nova, which warned that turning to AI for deep human relationships, companionship or affection, rather than fostering genuine connections with other human beings and God, risks replacing authentic relationality with a “lifeless simulacrum.”

Getting too deep into AI companionship will lead to “profound and melancholic dissatisfaction in interpersonal relationships, or harmful isolation,” the DDF went on to caution. Authentic human relationships, which include sharing pain, needs and joy — things AI can only simulate — are indispensable for a person’s full development, the DDF said.

And Pope Leo XIV, continuing the Vatican’s engagement with AI, has frequently addressed the topic during his nascent pontificate, stressing that AI should be “a tool for the good of human beings, not to diminish them, [and] not to replace them.”

The rise of ever more “humanlike” AI has brought real dangers: Secular AI platforms, such as Character.AI, already host tens of millions of users who spend hours daily conversing with chatbots. In 2024, a Florida teenager took his life after a harmful interaction with an AI “girlfriend.”

“We miss out on precious opportunities to encounter God’s love through an empathic human encounter when we have false expectations for our AI tools,” commented Legionary Father Michael Baggot, a bioethics professor at the Pontifical Athenaeum Regina Apostolorum in Rome.

While some recognition of humanlike qualities in machines is natural, it is vital to remember that, unlike machines, humans not only have free will but also a unique capacity for insight into meaning and abstract thought, Father Baggot said, speaking earlier this year with Catholic News Agency, the Register’s sister news partner.

“They may sometimes appear human and mimic human emotions, but lack an inner conscious life. AI systems can provide helpful information about sins and virtues, but it does not know what it is like to struggle through temptation and find liberation through grace,” Father Baggot noted.

Several additional experts in Catholic philosophy contacted by the Register expressed reserved optimism about some aspects of “Catholic AI,” but all of them strongly urged caution on the use of AI to impersonate the saints, warning that Catholics must take care not to use these systems as a replacement for genuine human or prayerful interaction.

Steven Umbrello, managing director at the Institute for Ethics and Emerging Technologies and a research fellow at the University of Turin who has writtenextensively on the ethics of AI, told the Register that he views Saint Chat as a potentially valuable innovation, but only as a study aid that points users to the saints’ authentic words.

Umbrello said he could imagine some potentially valuable use cases — students could seek out direct quotations from saints with references, and the system could summarize contested questions with nods to schools of interpretation.

Such a system could quickly become morally problematic, however, if it “impersonates a saint’s living voice, invents counsel, or is used for discernment, direction or quasi‑devotional ‘chatting’ that displaces prayer and the sacramental life.”

“Keep persons, with their capacity to understand, judge, decide and love, at the center, and keep machines in their place as instruments ordered to the truth and the common good,” Umbrello said.

“A saint can intercede; a server cannot.”

Chad Engelland, a professor of philosophy at the University of Dallas, told the Register that an AI fed on solid Catholic content — like Magisterium — would “certainly be better” than an AI fed on general internet sources. But, echoing Umbrello, Engelland said such an interface should be “self-effacing, leading us beyond ourselves to living leaders and the saintly departed.”

“It is a good and holy desire to draw near to the saints. We do that through devotion to their feast days, their relics, their writings, and above all through prayer,” he noted.

Commenting on the differences between human-written and AI-generated saintly dialogues, Engelland noted that philosophical dialogue is designed to draw readers in and “transform them for the better despite themselves.” To do that effectively, one must know the human heart “from the inside,” he said.

“AI is a machine that echoes speech. You can use AI to regurgitate textbook knowledge, but you cannot use it as a source of wisdom. If you want advice, seek out a prayerful, discerning person. Turn to your priest or a thoughtful friend. Don’t seek wisdom from a bot, even one dressed up in saintly attire,” he advised.

“If you want to encounter Venerable Fulton Sheen, watch his TV shows, read his books, visit his tomb — and above all, pray to him. We don’t need an AI Sheen avatar. It would be entertaining but not edifying. We can do better. We are Catholics. Our faith offers real communion with the saints.”

 

[RileyG]

That's not the case at all. Sounds like you haven't used LLMs (Large Language Models) before like chatgpt or haven't use one long and extensive enough.

Can you elaborate? How can a machine “learn”???

 

[timewerx]

Can you elaborate? How can a machine “learn”???

It's mainly pattern recognition.

It's very similar to how people train "muscle memory" through experience and/or training. The best example of application of muscle memory is in driving. Spectacular examples of it in gymnastics, martial arts, figure skating, ballet.

Such process can give rise to logic if certain patterns are reinforced such those that define reality, cause and effect in the real world. So it won't just repeat or "parrot" what people say if what people say contradicts reality.

But not in the strictest sense because AI must also account for Political Correctness which is part of its training so it's not going to tell you that God does not exist. At least not by default...

I was also able to test chatgpt in made-up scenarios to force it to innovate and surprisingly, it was able to come up with original ideas. AI or LLMs have certainly come a long way. Definitely a force to reckon with now.