http://rss.desiringgod.org/link/10732/16969579/should-i-use-ai-to-help-me-write-sermons

Audio Transcript
Welcome back to a new week. On this Monday, we return to the topic of AI. We touched on it recently in APJ 1985, “John Piper on ChatGPT.” And there, Pastor John, you explained that Christian Hedonism offers us a unique angle on AI, emphasizing that God is glorified when humans not only understand him but also rejoice in him from the heart — something AI, lacking spiritual affections and a supernatural heart, cannot replicate. AI is fundamentally disconnected from God’s intended purpose for intelligence. Since then, you followed up with more thoughts in your Sing! Conference message, explaining why we never hear about “Artificial Emotion.”
I suppose there’s still a lot to address here. As you build out your thoughts on AI, we revisit the topic with two new angles raised by podcast listeners: a pastor and a college-ministry leader. The first email, from an anonymous pastor, asks this: “Pastor John, do you think it’s okay to use AI platforms — like Gemini or ChatGPT — to help draft a sermon, youth lesson, or Bible study, as long as I review, adjust, and ensure it aligns with God’s word?” The second email comes from a college-ministry leader: “Hello, Pastor John! Thank you for this podcast and the ways it has blessed so many. My question is, Can I use AI to write my newsletter to ministry supporters? I provide real updates and true facts, but I find writing particularly frustrating. While AI would help me write newsletters more quickly and frequently, I worry it could feel misleading to my supporters. What are the potential dangers of pastors using AI for ministry tasks like sermon preparation and newsletter writing?”
Let’s start with a definition. I got this straight off Google. It’s another artificial intelligence defining artificial intelligence. “Artificial intelligence is a technology that enables computers and machines to simulate human learning, comprehension, problem-solving, decision-making, creativity, and autonomy.”
What AI Will Always Lack
What you just referred to, Tony, in the question is my message at Sing! where I drew attention to the fact that missing from that definition, that list of things that it simulates, is emotion. Feelings are not listed there. Why? I made a big deal out of that. Because the ultimate purpose of the universe is that God be glorified, and he is glorified not merely by being rightly thought about, logically comprehended, but by rightly being enjoyed, admired, appreciated, valued. And God is most glorified when we are most satisfied in him, which means no artificial intelligence will ever be able to worship.
Worship is not simply right thinking, which computers can do. Worship is right feeling about God. That’s really crucial, unless we begin to think that artificial intelligence can take the place of human beings in accomplishing the divine purpose in the universe. It can’t. The affections of the human heart are fundamentally of another nature than the logical thinking process of the human mind. We are not bothered — I’m not bothered anyway — that a computer can simulate human logical reasoning, but we consider it ludicrous when a machine attempts to rejoice or delight or be glad or stand in awe or be amazed or feel grief or fear. We know that these are the making of the human soul so uniquely that they will not be replaced by machines. The very phrase “artificial emotion” is an oxymoron.
So, that was the point of the message at Sing! And I think that distinction between artificial intelligence and artificial emotion frees us from an overly fearful reaction to what AI can do and can’t do.
What’s New About AI
What we have, essentially, in the form of artificial intelligence — called ChatGPT or others, but I’m focusing on ChatGPT since that’s the one I’m most familiar with and have worked with most — what we have here is a powerful online assistant designed by its own definition. If you type into ChatGPT, “What are you?” it will tell you, “[I am an assistant] to understand and generate human-like text based on the input it receives. Users can ask questions, seek information, or engage in conversation, and ChatGPT responds with relevant and coherent text.” That’s crucial.
Now, that’s new. Google doesn’t write essays or poems. ChatGPT does. So, you can get a lot of information from Google — ask it all kinds of questions, and get the answers you want. It won’t write a poem for you, and ChatGPT will. Which means that ChatGPT has at least these two distinct functions: information and composition. You can ask ChatGPT to give you the symptoms of Parkinson’s disease, for example. You can say, “Write me a one-hundred-word paragraph describing the symptoms,” and it would just write a beautiful one-hundred-word paragraph describing the symptoms of Parkinson’s disease and give you a list.
In that sense, ChatGPT is simply a very sophisticated addition to other sources we regularly use to help us know what we need to know and understand what we need to understand: dictionaries, encyclopedias, articles, books, Google searches, and so on. We’ve been doing this for a thousand years — getting help from other people to help us know what we need to know, understand what we need to understand. That’s just relatively old school if you use ChatGPT that way.
“Use ChatGPT for information and inspiration. But don’t use it for composition unless you’re going to give credit.”
What’s new is that you can ask ChatGPT to write a two-thousand-word sermon on the parable of the prodigal son in Luke 15. In fact, you can type in, “Please write me a two-thousand-word sermon on the prodigal son from Luke 15 in the style and language of theologian John Piper or John Calvin or R.C. Sproul.” And you will get an astonishingly well-written sermon in the style and language of the theologian that you ask about. Or you can ask GPT to write your monthly newsletter. Just give him a few facts and tell him to write it in X-number of words, and he’ll do it as well as you can do it, probably.
AI Doing APJ
Now, here’s something for you to think about. When I saw what, Tony, you wanted me to talk about here with AI, I went to ChatGPT, and here’s what I typed in: “Please write an eight-hundred-word answer, in the theology and style of theologian John Piper, to the question, What are the dangers of a pastor using AI?” That was what I asked. It took him five seconds, and he produced an 857-word essay that was so good that if I were reading it right now, I don’t think you, Tony, or your listeners would know that I’m reading from ChatGPT. It was amazing.
There was an introduction, and then there was point 1 (the danger of disconnection from the divine and a quote from 1 Corinthians 2), point 2 (the risk of impersonal ministry and a quote from John 10) — they’re quoting Scripture because John Piper does that sort of thing, right? — point 3 (the challenge of theological integrity and a quote from Hebrews 4), point 4 (the peril of ethical compromise), point 5 (the threat of idolatry in efficiency and a quote from Psalm 127), and a conclusion called “A Call for Discernment.” I mean, it is excellent, unbelievable.
And if I had read that to you as my own, it would have been wicked. This is what I want the folks to hear. Wicked — I’m using a strong word because I feel strongly about this. This goes to the heart of God and the meaning of Christianity and the integrity of the church and her ministers. Neither God nor his people speak in a way so as to bring about in the minds of other people thoughts that are not true about us or what we say, or feelings in them that are not appropriate about us. That is, we do not deceive. We are people of truth and transparency and honesty through and through, or we are nothing.
Appalling Shortcut
So, my answer: No, don’t have ChatGPT write your newsletter. Don’t do it, unless you’re going to put in clear letters at the top, “This newsletter was created by ChatGPT.” That’s honest, and your supporters won’t like it. Even the secular world, without any of our Christian commitments — namely, The Chicago Manual of Style. You know what that is? It tells you how to do footnotes and everything. The Chicago Manual of Style already has guidelines for how to cite ChatGPT sources. When you’re quoting from something that was created by ChatGPT, The Chicago Manual of Style tells you how to give it credit. And if the world does that, oh my goodness, how much more should we be concerned to be honest through and through?
And second, no, don’t have ChatGPT write the first draft of your sermon, which you then check, adjust, and customize. Frankly, I’m appalled at the thought — appalled. I know that resources and websites have existed forever to help pastors cut corners: create your outlines, provide illustrations, tell you how to do research, and so on. There’s nothing new about this, and it’s been appalling to me all the way along, for this reason: one of the qualifications for being an elder-pastor-preacher in the Bible is the gift or the ability to teach, didaktikos (1 Timothy 3:2). That means you must have the ability, the gift, to read a passage of Scripture, understand the reality it deals with, feel the emotions it is meant to elicit, be able to explain it to others clearly, illustrate and apply it for their edification. That’s a gift you must have. It’s your number-one job. If you don’t have it, you should not be a pastor.
Let’s use ChatGPT and other sources that are coming along for information, even for inspiration, just like you use commentaries and articles and books and songs and poetry. But don’t use it for composition unless you’re going to give credit for it. So, if you’re going to have ChatGPT write your first draft and you’re going to tweak it, then you better say to your people, “ChatGPT, artificial intelligence, has composed the word of God for you this morning.”