Since ChatGPT’s release in November 2022, headlines about artificial intelligence (AI) have been sensational. AI is being touted as a panacea for everything from educating our youth to curing cancer.
As pressure mounts to fundamentally change the way we do everything, so does anxiety. While some are bullishly enthusiastic about AI and others find the hype nauseating, AI is affecting an ever-increasing segment of the population, including Catholics.
According to Catholic theologian Henry Karlson, the Church has long engaged the sciences even as it has cautioned against their misapplication “in theories which do not have enough evidence or in fads which die out.” With that in mind, it’s worth taking a sober look at artificial intelligence — what it is, what it is not, what it can do, and what it cannot do.
There is no consensus on a standard definition of artificial intelligence. Nevertheless, AI is generally understood as the field of science and engineering concerned with building a semblance of intelligence into computer systems (e.g., the ability to reason, learn and act). One definition of AI that I like is from Elaine Rich and Kevin Knight’s Artificial Intelligence: “The study of how to make computers do things which, at the moment, people are better at.”
For instance, Visa’s processing of nearly 4 million transactions per minute is not AI, because a computer can do it far more efficiently than a person can. On the other hand, driving a car is considered AI because, at present, a human can drive better than a computer can. The blurred boundaries between what does and does not qualify as AI are evident: While the processing of those millions of transactions is not AI, detecting fraud within them is.
The practical exposure of AI to the public over the past three years has animated discussions about which aspects of our work will be replaced by it.
During his first address to the College of Cardinals, Pope Leo XIV said that the Church would address the risks AI poses to “human dignity, justice and labor.” At the core of these discussions is the fundamental question of what it means to be human.
In January 2025, the Dicastery for the Doctrine of the Faith and the Dicastery for Culture and Education co-issued a doctrinal note titled Antiqua et Nova (Ancient and New). The document is subtitled “Note on the Relationship Between Artificial Intelligence and Human Intelligence.”
There is often an inverse relationship between human and artificial intelligence. Actions that people perform without thought are challenging for computers, while complex cognitive tasks requiring significant human intelligence are often easy for computers.
For instance, playing an outstanding game of chess requires significant human intelligence. Yet computers have been regularly beating grandmasters for almost 30 years. By contrast, an infant can almost instantaneously recognize his or her mother when she enters a room. But programming computers to recognize specific objects in images and video — an application of AI called “computer vision” — remains a difficult problem.
Despite the frenetic pace at which AI technologies have advanced in the past decade, AI is neither new nor novel.
The term “artificial intelligence” was coined in 1956 at a summer research workshop at Dartmouth College, where a small group of researchers from a variety of disciplines gathered to explore the idea of writing programs that could teach computers to learn and reason.
Even today’s popular applications of AI — self-driving cars like Teslas and chatbots like ChatGPT — are not novel. In the mid-1960s, Joseph Weizenbaum, a professor at MIT, developed ELIZA, which was a program that implemented rudimentary communication between a human and a computer using natural language.
Weizenbaum also questioned the wisdom of aspiring to create “intelligence” in computers — a philosophical question as relevant today as it was 60 years ago. By contrast, questions about whether computers can think are “just as meaningful as the question whether submarines can swim,” as Edsger Dijkstra famously said during a discussion titled “Computers and Society.”
We are often duped by the appearance of human behaviors in computer systems. Today’s chatbots, despite their use of large language models (LLMs), do not have a model of language. Rather, they possess sophisticated statistical models of correlations between words.
An LLM generates a response to a user prompt by taking the sequence of words in the prompt and predicting the next word, then using that extended sequence to predict the following word, and so on. This iterative prediction process — akin to the autocomplete feature in email applications like Outlook — is called “generative AI” and creates the illusion that we are having a natural dialogue with ChatGPT.
This deception is reinforced by the words used to describe these computer processes — words such as training, learning and understanding, which refer to human activities. The “deep” in deep learning does not refer to depth of understanding — the system has no understanding of anything. Rather, it is a technical term referring to the structure of the neural network, which contains many (or “deep”) layers.
“Artificial intelligence,” however, is an apt term to describe the field because AI is not natural intelligence; it is a simulation of natural intelligence. As Joseph MacRae Mellichamp, emeritus professor of management at the University of Alabama, has said, “the ‘artificial’ in artificial intelligence is real,” but the intelligence is not. (The word artificial comes from Latin roots meaning “skill” and “to make.”)
AI is also not magical software that will one day spring to life, as in the fictional creation of Mary Shelley’s Victor Frankenstein. Nor will AI replace human beings in some recurring doomsday Kafkaesque nightmare. Mathematician Hannah Fry has observed that worrying about such catastrophes is like worrying about overpopulation on Mars.
The scientific term for humans, homo sapiens (“wise man”), highlights the truth that intelligence is fundamental to man and inseparably woven into what it means to be human. Human beings are created in the image and likeness of God, and no advances in science or technology are ever going to change that metaphysical reality. Human intelligence involves consciousness and free will. AI systems are not sentient; they are created and controlled by humans.
What the field of AI has achieved — and this is no small scientific feat — are simulations of isolated facets of natural intelligence, targeted to a specific set of tasks within particular application domains. This is known as narrow AI. Narrow AI excels at straightforward, tedious and highly repeatable tasks, especially where the consequences of failure are low.
In stark contrast to narrow AI stands the amorphous concept of artificial general intelligence (AGI), which refers to “a single system capable of operating across all cognitive domains and performing any task within the scope of human intelligence.” At present, AGI remains a fantasy.
AI is part of a long continuum of technological advances that shape human history and, like those before it, will likely bring short-term disruptions. At this point, the magnitude of any long-term effect of AI on culture and society remains unclear. As in all matters, we should look to Holy Mother Church for eternal truth and wisdom.
As we continue to grapple with the complex issues surrounding AI, I pray: “Lord, the Earth is filled with the fruit of Thy works. May we cling to hope and joy as we face the unknown. Everything is Yours — do with it what You will.”
For more information on the artificial intelligence concepts introduced in this article, Perugini has a short course titled Demystifying Artificial Intelligence from a Catholic Perspective, which is freely available in the Ave Maria University Pursuit of Wisdom short course series at thepursuitofwisdom.org.
link