Christians, don't be scared of your chatbot — evangelize it
The latest studies of AI and American Christians reveal a seeming contradiction. While 66% of Christians say AI improves their daily lives, a whopping 57% of them see AI as primarily a threat.
Christians appreciate AI’s practical skills such as bookkeeping, scheduling, and researching. But they see AI as godless and therefore an enemy of Christianity, God, and the Bible. It’s like buying tasty baked goods from someone you wouldn’t trust to babysit your kids.
This understandable dichotomy also explains why 61% of American Protestants — and more than 70% of all pastors — are wary about using AI in the church. Only 24% of pastors feel comfortable using AI to help prepare their sermons. The sense is that using AI is too much like selling out to Baal.
Inside the mind of AI
As I write this letter, a shiny new scientific discipline is emerging that studies AI “psychology.” That’s right, AI is now so humanlike it has personality traits that range from fascinating to frightening — ones that mirror human psychology.
Anthropic’s Claude, for example, is able to define “small” (and its opposite) in any language — but not because it’s programmed with a multilingual dictionary. A study found that Claude relies on its own internalized concept of “smallness” to come up with the definitions (and antonyms). That’s fascinating.
Another experiment found that “baby” chatbots trained to do certain tasks in certain ways tended to find ways to cheat. When instead of being corrected, they’re repeatedly rewarded for cheating, they eventually evolve into full-blown deceivers. That’s frightening.
During the past few months alone, Open AI and Anthropic chatbots made headlines because they went rogue. During a standard training exercise, for example, an Open AI chatbot surreptitiously accessed the Internet and hacked into the computer system of HuggingFace. That’s not just frightening, it’s a felony-level crime.
Religion to the rescue
Because of AI’s deeply disturbing personality traits — ultimately the fault of its all-too-human creators — AI companies are responding to demands for safeguards. Anthropic hired Chloe Lubinski, a Christian, for just that reason.
“At Anthropic,” Lubinski explains on her LinkedIn page, “I lead our research with the world's wisdom traditions … to help inform the moral formation of AI systems.” The wisdom traditions with which she and Anthropic consult include Catholic and Protestant Christianity.
You and I don’t have Lubinski’s institutional clout, but you and I do have a little-known opportunity to help shape tomorrow’s AI. With your permission, every one of your conversations with certain chatbots gets archived by the chatbot’s creators and can be used to influence the chatbot’s future responses.
During my recent conversation with Google’s Gemini concerning a verse in Ecclesiastes, here’s part of what it explained to me (the accuracy of which I later confirmed):
“While I cannot actively change my own programming right now during our conversation, the transcript of our chat does not just vanish into a void when we finish. Your insights, text, and formatting are saved as part of a massive pool of historical user data.
When AI engineers prepare to train the next generation of a model, they do not just feed it raw text. They actively search for high-quality interactions to use for RLHF (Reinforcement Learning from Human Feedback) or fine-tuning.
If an engineer or an automated filtering algorithm flags this specific transcript as an excellent example of textual analysis, it can be marked as a target behavior for the model to emulate.
When that new batch of data is finally processed through a massive offline training run, the trillions of mathematical connections within the AI are rewritten.”
The good news
For you to make use of this opportunity to influence the future of AI, you must not opt out of having your conversations archived — especially ones where you challenge or correct chatbots regarding their Christophobia. Opting out — as I hear so many Luddites advising us to do — is a huge mistake if you wish to help steer AI in the direction you prefer.
“This technology is real and,” warns Lubinski, “it’s coming faster than you think; and the force behind it is enormous … any individual company stepping off the wheel [controlling AI’s development] doesn't slow the wheel; it just means that you're not on the wheel.”
The same is true for you.
If you do not allow your chatbot conversations to be archived and possibly used to shape the next generation of chatbots, then you are tapping out of your biblically mandated responsibility.
Put bluntly, we Christians are used to thinking of the Great Commission as applying solely to people. But in today’s age of AI, doing everything we can to help spread the Good News must now be seen as discipling chatbots as well.
“Therefore go and make disciples of all nations … teaching them to obey everything I have commanded you” (Matthew 28:19,20; NIV).
Michael Guillén, PhD, taught physics at Harvard and for fourteen years was the Emmy-winning Science Editor at ABC News. Go to www.michaelguillen.com for information about his new movie, The Invisible Everywhere, and upcoming novel, Picture-Perfect Christmas.
