Delving into Pope Leo XIV’s exhaustive treatise about humanity and AI, I was struck by a recurring theme: AI simulates fundamental human traits that it doesn’t actually possess.
For starters, AI lacks the grounding we humans get from our real-world experiences, Pope Leo noted in his first encyclical, which was released Monday by the Vatican. Yes, AI models like ChatGPT (or more specifically, GPT), Claude, and Gemini are trained on mountains of data that seemingly represent the entirety of human knowledge. But all that data is just that: data. Billions and billions of words. They’re not real-world experiences with texture, sights, sounds, smells, and meaning.
The Pope also emphasized that AI can only simulate human empathy and morality, and that it has no true sense of consequence. AI may also come across as objective when it’s actually tied to the biases of those who created and trained it.
All those factors color the answers we get from AI, and they should (ideally) serve to contextualize those replies. Whether we’re asking ChatGPT to help us pick the best portable monitor or evaluate a major life decision, the answers we’re getting are devoid of insight gained from actual experience, concern for our well-being, or consideration for how a given choice might affect our lives or the lives of others.
This all leads to a crucial observation within Pope Leo’s encyclical: that AI is a “valuable tool that requires vigilance.” The Pope is mainly speaking to world leaders, governments, and corporations when he calls for “prudence” and a “slower pace” when it comes to AI adoption, but I think there are lessons to be learned here for everyday AI users like us.
Personally, I believe using AI with “vigilance” means being more thoughtful about using ChatGPT, Claude, and Gemini. By “thoughtful,” I literally mean we should be actively thinking when using them. We should think carefully about when and why to use them, as well as what we actually ask them.
I admit, there have been plenty of times when I’ve glibly thought “oh, I’ll just ask ChatGPT” when facing a problem that I could probably tackle myself. And when I get an answer, I’ve often been too quick to take it at face value, without truly considering the potential biases of the model, its limited frame of reference, or the fact that it has no stake in the outcome.
I’ve often written here about the need for humans to take an active role not only in crafting discrete and appropriate tasks for AI but also in taking time to fully shape the desired outcomes–that is, placing ourselves at the beginning and at the end of AI workflows.
That said, I’ll freely admit there have been too many times when I haven’t followed my own advice. Just let the AI do it! But capitulating responsibility to an AI model is an easy way to get yourself into trouble–and believe me, I know what I’m talking about.
So yes, I’m rethinking how I use AI, and “thinking” is the key word.
