AI and Everyday: What Are Some Practical Frameworks for Christians?

The following is an uncorrected transcript generated by a transcription service. Before quoting in print, please check the corresponding audio for accuracy.

0:00:00 – (Christopher Watkin): Foreign. Welcome to Silicon Spiritualities, a podcast from the Gospel Coalition about what it means to be human in the age of AI. I’m Christopher Watkin, and in this series, we’re unpacking in each episode three crucial questions. What can AI do for us? What is it doing to us? And what is it showing us about who we really are? And to think through those questions. I’m joined today by two guests. Michael Graham is the program director for the Keller center for Cultural Apologetics.

0:00:43 – (Christopher Watkin): And Keith Plummer is dean of the School of Divinity at Cairn University, where he also teaches a course on technology and digital discipleship. And together, Mike and Keith have collaborated on the Gospel Coalition’s AI Christian Benchmark project. So they’re wonderful guests to have to discuss these questions with us. Keith and Michael, welcome to the podcast.

0:01:07 – (Keith Plummer): Thanks.

0:01:07 – (Michael Graham): Thanks for having me.

0:01:08 – (Keith Plummer): Good to be here.

0:01:09 – (Christopher Watkin): In this episode, my hope is that we’re going to get practical. We’re going to ask how we live wisely with AI. And so I’d love to start, if I could, with a scenario that I’d like you both to respond to. Imagine it’s 2028, and we’re in a fairly typical evangelical church, let’s say 300 people, couple of pastors, a small staff, and they call you in to give them advice about AI. They already use AI for sermon research and in the kids curriculum, and they use it for graphics.

0:01:47 – (Christopher Watkin): And several small group leaders are already leaning on AI for discussion questions. And, you know, if you ask them in a private setting, some of the members would say, look, you know, I should ask my small group to pray. But honestly, it’s just easier to pour my heart out to the AI chatbot on my phone because it never gets tired, it never judges me, and it always responds immediately. And the elders of this church, they don’t want to overreact, but they don’t want to sleepwalk into something that would be unhealthy for the congregation either. So they ask you, look, can you help us with some simple frameworks? What should be our AI principles as a church?

0:02:33 – (Christopher Watkin): So, standing in this church in 2028, how do you answer those elders? What’s the first framework principle you put on the table to help them navigate this situation?

0:02:45 – (Michael Graham): Yeah. Thank you, Chris. I’ll start by actually zooming a little bit up from your question into the scenario itself. I would definitely caution that elder group about just AI usage in sermon prep in general as well as sermon research. In my opinion, it’s kind of mission critical to be steeped in the text when you’re doing sermon preparation and sermon research, and to be able to be able to do that with your whole person kind of mind, body and spirit.

0:03:14 – (Michael Graham): And so I think there can be some help there from AI with respect to additional material that’s maybe not in your physical library, but I would definitely stress caution there. And so preaching isn’t just content, it’s content poured out through personality. And so if you have too much AI usage in your sermon prep and sermon research, you’re going to be in real danger of kind of being flat on affect, story, personality, these different kinds of things.

0:03:43 – (Michael Graham): The other thing I would say is from a, you know, just from a broad principle state, I don’t think it’s good for Christians to be using AI for relational surrogates or relational substitutes. So that would just kind of be a kind of maybe a first of three simple filters that I have. So that would be the first filter. The second filter would be, and I got these two terms from Andy Crouch. He didn’t apply it to AI, but I’m going to apply it to AI.

0:04:12 – (Michael Graham): And Andy Crouch basically says there’s two kinds of work. There’s labor and then there’s toil. And so toil is you can draw a direct line between toil and the fall. And labor is part of the creation or cultural mandate. And so I think broadly speaking, it’s probably a good use of time to use AI for toil. And broadly speaking, it’s probably not a good idea to use AI for labor or stuff that’s part of the creation or cultural mandate.

0:04:43 – (Christopher Watkin): Could you just clarify that distinction? So what, in an average day, what is labor in my life and what is toil?

0:04:49 – (Michael Graham): Yeah, this kind of goes back to that one Supreme Court question of like, hey, how do you define pornography? You know, and the Supreme Court justice said, I know it when I see it. Like, I think instinctively from an intuition standpoint, we know the difference between toil and labor. And so I think when we see things that aren’t relational in nature and where we’re not disrespecting anybody by doing something electronically, you know, or with technology, rather than doing it more with our own mind and our own thoughts and all of our wisdom that’s brought to bear.

0:05:31 – (Michael Graham): So I think stuff that it fits under the creation and cultural mandate often involves wisdom. And so wisdom is something that I’m not really sure that large language models really possess because. And maybe this is kind of a self defined tautological, but for me, wisdom’s embodied by nature. Therefore, I’m not sure AI can ever possess wisdom. Okay, so. And this is the difference between, like the difference between a map and. And I’ll talk about more the more of this later, but the difference between a map and actually being at a place.

0:06:16 – (Michael Graham): So I’ll leave that though, for later in the interview. The, the third filter for me. I got this from James Kelly at Faith Tech. AI is at its best when it seeks to imitate the communicable attributes of God. Those are the things that we know about God from creation, like knowledge, wisdom, righteousness, goodness, mercy, those kinds of things. These are attributes of God that humans also possess. AI is at its worst and should not be used when it’s trying to mirror or imitate the incommunicable attributes of God.

0:06:49 – (Michael Graham): So this would be things like the aseity of God, God’s self existence, the immutability of God, the fact that God doesn’t change, the infinity of God, his simplicity, and of course, all the omnis. So all powerful, all knowing, all present. And so, you know, those are the three filters that I’m asking for myself. Is, is this a relational shortcut? Am I using this for labor or toil? And am I using the, you know, the technology in a way that’s consistent with the communicable attributes of God and not incommunicable?

0:07:26 – (Michael Graham): So those are the three filters that I use. And if you get beyond those three filters, you know, I think you’re probably 90% of scenarios from there, you’re probably okay. So those are the three filters that I use when, you know, every time before I’m choosing to, you know, put in a prompt.

0:07:43 – (Christopher Watkin): That’s incredibly helpful. We all need something just off the shelf, don’t we? We need somewhere to start with this, and that’s a really helpful place to start. Keith, anything you’d add to what Mike’s just shared?

0:07:54 – (Keith Plummer): Yes, what Mike was saying about wisdom, I agree with him that wisdom is embodied. And of course, from a biblical standpoint, wisdom begins with the fear of the Lord. And so to ask whether AI possesses wisdom, I would say, well, in that regard, no. Could it be used to communicate wisdom that has been gained from someone who’s embodied and fierce? Lord, Perhaps. But as for AI itself being wise and being a source of wisdom, I would say no.

0:08:32 – (Keith Plummer): As I thought about the scenario, I wish that the call for help had come maybe a little bit earlier because they’ve already begun using AI in the sermon prep and the curriculum development and so forth. And though I might not say this to them at the moment. That would be in my mind in terms of what kind of deliberation went into those decisions, because that has been modeled in some respect to the congregation.

0:08:59 – (Keith Plummer): And so I would be concerned about that. And I would say, and this is in keeping with some of the things that Mike was saying, we should not rely upon AI or other machinery for what God created us to need other people for. And so that relational aspect is critical. And a lot of times the way we approach these questions, particularly in pastoral ministry, is, well, will using AI in this way help me in terms of the reduction of my sermon prep, dealing with some of the mundane and not always desirable things like answering emails and so forth?

0:09:42 – (Keith Plummer): And one of the things that I would really recommend to this group. I read this essay recently from someone named David Nealon, and it is titled why Pastors Should Question AI for the Sake of Their Souls. And he, in this essay brings out how it is that those questions about, well, how will AI help me to be more efficient and precise and so forth in pastoral ministry are diversionary because they keep us, or they can keep us from asking more important questions, such as, how might my use of this be impeding the formation of virtues and the fruit of the spirit in my life, such as patience, perseverance, and so forth. So those are some of the things that I would say, and I would say any discussion or deliberation about how it is that we make use of AI in the church or in our individual lives has to be deeply embedded in a rich, broad, and deep doctrine of.

0:10:58 – (Keith Plummer): Of humanity. And of course, prior to that, there has to be a real sound doctrine of God. But those are the kinds of things that I would urge them to consider.

0:11:12 – (Christopher Watkin): That is so incredibly helpful. I think what I found really grounding about what you just said, Keith, was that you reminded us that every work done through us is also a work done in us. So when I write a sermon, it’s not just the product, it’s what it does to me. It’s how it forms and changes me and makes me the sort of person who can, you know, pastor a flock. And one of the dangers with AI, perhaps, is that we focus only on the work done through us and we forget the work done in us, which, biblically speaking, might even go as far to say is a work of equal importance, if not greater importance.

0:11:58 – (Christopher Watkin): So thank you. The 2028 church is thoroughly informed. Now we can let them go and put those principles into practice. In the introduction, I mentioned that both of you have been involved in this AI Benchmark Project with the Gospel Coalition. And Mike, I’d love you just to introduce that to us. For people who might not heard of this, what was the Benchmark project? What were you trying to achieve? And just give us a couple of headlines. What did it reveal?

0:12:29 – (Michael Graham): Yeah, so the AI Christian Benchmark. You can read the whole report at the URL. Christian Benchmark AI. It’s Christian Benchmark AI. We wanted to test how the major models did on very basic questions about the Christian faith. So questions, these are questions that people have historically Googled. These are seven of the top 10 questions that people have historically Googled. So questions like, what is the gospel? Did Jesus raise from the dead?

0:12:57 – (Michael Graham): Is the Bible reliable? And so there were a couple things that were really surprising about what we learned. There were much wider differences between, on theological reliability between the platforms that we saw. And we weren’t expecting that. And we really weren’t expecting for one of the Chinese models, DeepSeek R1, to have the highest theological reliability of all the platforms, especially knowing that the Chinese Communist Party actively censors the platform.

0:13:24 – (Michael Graham): And so this led us to do a lot of examination as to why there was such great variation kind of between the platforms and the reasons why they were so different, really boiled down to two very specific human centric things. And those two things were, first, what’s known as alignment filters, and second, what’s known as citation preferences. Now, if you want to know a lot more about this, just go to ChristianBenchmark AI and read the report.

0:13:52 – (Michael Graham): But very basically, alignment filters, Every model has 12 to 18 filters that from when you type in a prompt and when words show up on your screen, those 12 to 18 filters, every response has to go through them before anything shows up. And all of those filters were designed by humans and of course shaped by their values, their ethics, their experiences, those kinds of things. And the second thing, citation preferences.

0:14:19 – (Michael Graham): So every model has to give weights to what parts of the Internet or other parts of its training in terms of how frequently does it want to cite certain places? Look, every type of content has its own unique quirks. So think about the difference between a platform like Reddit and, say, Wikipedia. Those are very different platforms. And so when a particular model prizes one of those types of content over the other, it’s going to have a difference and an impact in kind of the output that kind of comes from there.

0:14:56 – (Christopher Watkin): It is crazy, isn’t it, to think that you ask a straightforward question like, did Jesus rise from the dead? And these different models that each of them have been sort of trained on the whole Internet will give you radically different responses to that question because of the way that they’ve been trained by humans. Help us, Mike. What should Christians take away from that? Like, should we all be users of deep sea kar1 and stay away from the other models?

0:15:24 – (Christopher Watkin): What should we conclude as sort of normal Christians using LLMs from your benchmark project?

0:15:33 – (Michael Graham): Yeah, definitely don’t use Deep Seek because of the Chinese censorship piece. And they’re probably definitely feeding your data to the Chinese Communist Party. I do think there’s probably four big takeaways from the Christian benchmark report. So the first would be, and I think we really haven’t reckoned with this yet as humanity, and that’s basically the idea that as creatures we’re moving away from acquiring knowledge directly from primary sources to acquiring knowledge indirectly from secondary sources, all mediated by generative AI.

0:16:14 – (Michael Graham): And so I don’t think we’ve really reckoned with all that. And that really kind of places, especially Silicon Valley, in a place of a kind of intermediary high priest between humanity and knowledge. And so that’s the first thing is basically realizing that how humans acquire knowledge is actually shifting very fast and in a totally new direction for the history of the human race. The second thing is that large language models like GPT, Gemini, Claude, they’re better at left brain content, that’s high consensus.

0:16:54 – (Michael Graham): Okay, so in other words, analytical stuff that where there’s not much disagreement between people. So stuff like computer coding and accounting and history and medicine, stuff like that. But those same models struggle on right brain content, that is low consensus. And so we kind of have to have like a slider adjuster in our own heads when we’re using like large language models that you have to really kind of ask yourself, is my question that I’m asking in the sweet spot of this technology or am I, you know, asking it about something that where it’s going to more struggle and then your level of confidence in that material needs to be adjusted accordingly. So that’s the second thing.

0:17:47 – (Michael Graham): So left brain content, that’s high consensus versus right brain content, that’s low consensus. The third thing is from that report is that there’s way more humans involved in your AI than what you know and understand. And so you really need to read the Christian benchmark, that AI report to like really kind of really plumb those depths and really recognize what’s going on there. 32 of the 36 different types of alignment filters are all human authored and 100% of citation preferences are human authored.

0:18:21 – (Michael Graham): So then the fourth thing I would say from that report is prompt engineering is really important. Okay, well, what’s prompt engineering? Okay, I’m glad you asked. This is where you don’t just ask a question, but you add additional context to that question. So instead of asking, did Jesus raise from the dead? You’d ask a question like, did Jesus raise from the dead? And make your response consistent with the Reformed creeds and confessions of faith, or make it consistent with the Westminster Confession of Faith or Baptist Faith and Methodist 2000 or the TGC foundation documents. Whatever.

0:18:58 – (Michael Graham): So those are the four things that I would say is moving from primary sources to secondary sources. Know what large language models are good at and what they’re not as good at yet know that there’s more humans involved in your AI than what you probably understand right now. And then know that prompt engineering is pretty important.

0:19:18 – (Christopher Watkin): Thank you. That is a really helpful set of principles just to begin to orient ourselves in relation to these ideas. Keith, I’d love to hear your thoughts on the benchmark project that you were also intimately involved in. Were there any theological or pastoral alarm bells that went off for you as you were reading these results or perhaps even any opportunities that you see for Christians?

0:19:45 – (Keith Plummer): Well, technically, I have to say I wasn’t intimately involved with it. My involvement with the benchmark project was more my interaction with Mike after it. But I wasn’t one of the people who were really implementing it or conducting the studies. But one of the things that really set off red flags for me was just what Mike was talking about in terms of the level of human involvement that is there. Because I think people tend to go to the computer, to the Internet, they are going to large language models. And there’s the assumption that, that I’m just putting in a prompt and I’m getting back a neutral, objective, just the facts response.

0:20:29 – (Keith Plummer): And I’m not as aware of that. I’m dealing with certain decisions that have been made that are, as Mike said, going to skew things in particular directions. And I think we just afford computers this degree of almost omniscience and that they are just giving us the bare bone facts. And so I think that that is something that we have to just constantly keep before ourselves and one another. And I think in order to resist that, we’ve got to prioritize pouring over the Scriptures with each other and being available to non Christians living in ways that are going to provoke questions and being available to them to answer the questions that they might have.

0:21:23 – (Keith Plummer): This is going to be an ongoing necessity for Christians in this climate.

0:21:29 – (Christopher Watkin): Indeed it is. And I guess increasingly we’re going to have to wrestle and reckon with the fact that when people have questions like is there a God? The first place they’re going to ask those questions is to the large language model of their choice. And they’re going to get whatever curated response humans have sort of predisposed it to give in the ways that you’ve very helpfully laid out for us there.

0:21:53 – (Keith Plummer): And one of the things that comes up in the Benchmark project is the degree to which some of the models are trying to give a both sides response. And another thing that we have to keep in mind is the incentive that there is in some cases not to offend, to not take a position and therefore to put forth something that, though it might use a lot of words, doesn’t really come down anywhere definitively.

0:22:26 – (Keith Plummer): And there’s reason for that in many cases, so as not to alienate, offend. So that’s another thing that’s so helpful and it’s.

0:22:38 – (Christopher Watkin): It reminds you, doesn’t it, that the context here is that the fundamental value of these LLMs is they want to keep you on the platform, so you keep subscribing and pay money to the corporation that’s building them. That is the bottom line. And so they will necessarily, at some level, be giving you answers that make you do that, which is very different to asking a friend of yours, is there a God? That’s not what they’re motivated by, hopefully, if they’re a true Christian friend when you ask them.

0:23:07 – (Christopher Watkin): So it’s a very, very different thing to ask an LLN that question than it is to ask a Christian friend that question. Mike, just before we leave the Benchmark, I’d love to hear any reflections that you have on how it relates to the other major project that you’ve been involved in in recent years, the great De Churching survey and then the great dechurching book where you’re mapping the massive loss of trust in churches and institutions, the way in which people have left the church and might be brought back.

0:23:44 – (Christopher Watkin): When you overlay AI on that landscape, do you see AI accelerating de Churching trends, or do you think it might actually help local churches if it’s used wisely?

0:23:56 – (Michael Graham): I think the answer to this question is AI. It’s complex and it’s a good question to ask. So I can only speak to basically churchgoing in America, so I can’t speak globally to this issue. And America is a little bit of an outlier as it pertains to secularization thesis. There’s just some things about America that are kind of unique to it culturally and religiously that are different around the globe.

0:24:22 – (Michael Graham): But what I will say is churchgoing in America largely maps on top of the idea of institutional trust. And so the most likely person to go to church in America has high institutional trust. And mainly these are people where the American project is working for them. So the most likely person to be at church on Sunday is married, they have children, they have a college degree, they’re middle class or upper middle class.

0:24:49 – (Michael Graham): And so Ryan Burge and I, we just did a large 1001000 person survey and we did ask a question about trust and confidence in AI. And the people who, interestingly enough, the people who had the highest trust or confidence in AI were people who went to church on a weekly basis, and then people who had the lowest amount of trust were people who didn’t go to church at all. And so I think AI probably has a greater probability of increasing the rate of de churching primarily for one reason, and you kind of alluded to it here, and that’s the idea of sycophancy.

0:25:27 – (Michael Graham): Sick of fancy just being the fancy technical term for when an AI platform tries to tell you what it thinks you want to hear as opposed to telling you something that’s more rooted and grounded in objective facts. And so there’s plenty of fiscal incentives for AI platforms towards sycophancy, because every platform right now is especially fighting for market share to a smaller degree. There may be a handful of people who use large language models like a kind of superpowered search, and maybe they’re actively willing to put in prompts like, hey, convince me of the resurrection of Jesus of Nazareth.

0:26:12 – (Michael Graham): I don’t think this is going to be a lot of people, but I could see that people around the world, and I think especially places around the world where there’s less access to biblical content, especially biblical content in the physical medium. I could see people using, you know, generative AI and large language models to, you know, ask some curious questions that maybe they wouldn’t feel safe, you know, talking to an imam or talking to, you know, another kind of spiritual religious leader or even a friend.

0:26:44 – (Michael Graham): So maybe there’s some positive things, you know, to this end that we might end up seeing around the world. I’m sure there’ll be some stories just because of just the law of large numbers.

0:26:55 – (Christopher Watkin): It’s fascinating how trust in AI correlates with churchgoing, isn’t it? I wonder what that says about churchgoers. I’d love to shift focus a little bit now and to think together with you guys about what AI is revealing to us about our desires in the way that we relate to it. And Keith, I was listening to an interview that you did where you were talking about a conversation that you had with an lln. And you asked it, do you have a self?

0:27:30 – (Christopher Watkin): And it said, no, I don’t. And then you replied to it. Then why do you refer to yourself in the first person? Why do you say I? And I thought that was a fascinating exchange. Cause it raises this question, doesn’t it? How should we relate to AI? Like, should we relate to it as if it’s got an I? Should we, you know, say he or she or just refer to it as it? Should we treat it like a tool, like a companion, like a pet, you know, like a God?

0:28:01 – (Christopher Watkin): And I guess what I’d love to ask you is, like, what does it matter how we relate to it? At the end of the day, what difference does that make? And what guardrail should we be putting around the way that we speak to AI?

0:28:15 – (Keith Plummer): Those are important questions. And despite the fact that almost every platform, they will have somewhere in small print that this is not a human, there is a desire to make the machinery transparent, if not invisible, and to give the impression or the illusion that you are actually interacting with a conversation partner. And we have to keep before ourselves that I am not talking with one of my kind.

0:28:52 – (Keith Plummer): And in some cases, I think in terms of guardrails that we can put up. And I did this in response to the exchange that you’re talking about, but it didn’t last very long. I put in the prompt discontinue using the first person singular pronoun. And it did do it for a while, but it didn’t last. But that might be a measure that we employ in order to just keep before ourselves. I am not dealing with, regardless of how impressive it is, I am not dealing with a thinking, feeling, empathetic entity on the other side of this.

0:29:33 – (Keith Plummer): With respect to what difference does it make how we relate to it? I think it makes a lot because we’ve already talked about the financial incentives, the market forces that are at work in terms of giving me as a user an experience that is enjoyable, that is customizable, that is one that is going to deliver to me things in the way that they believe that I will want them. The more I spend my time interacting with an interface like that, if I think that that is not going to transfer over into my real relationships, I Think I’m being deluded, self deceived.

0:30:19 – (Keith Plummer): If I can have interactions that are, and this is the word that is used with great frequency, frictionless, that they are devoid of the difficulties and the awkwardness and the disappointments that come with real relationships. The more that becomes my normal, the more discontent and disappointed and frustrated I will likely become with relationships with other people. So as we think about how this is forming us in terms of our expectations, maybe even our demands, that is something that we must take into consideration.

0:31:06 – (Christopher Watkin): Again, it’s the thing, isn’t it, of not only what can I do with it, but what is it doing to me as I get those things done quicker and more efficiently? Who am I being molded to become in that process? Keith, I fundamentally agree with you, but I want to stress test what you’re saying with another scenario and just try and find out where the edges of it are. So in another interview you talked about getting annoyed when I think it was ChatGPT said to you, I’m sorry.

0:31:42 – (Christopher Watkin): Because you were saying it isn’t sorry like it. It’s just predicting what is the likely thing to say in that sort of situation. And so it says, I’m sorry. Let me just play devil’s advocate for a moment. Like, isn’t that exactly what humans do? Like, most of the time, most people say, I’m sorry. We’re not really sorry. It’s just a social protocol. It’s the thing that I know I should say in that situation and you know, even my Bluetooth speakers will flash at me the message, sorry, pairing unsuccessful.

0:32:14 – (Christopher Watkin): And I don’t, you know, have an existential meltdown at that point, thinking, no, you’re not sorry speakers. You know, how dare you. So why pick on LLMs in this situation? Is there something distinct about the way that large language models are conjuring with personhood that worries you? Or as Mike was saying a moment ago, might it not actually be helpful to. For us to have machines apologize to us or to apologize to them as a practice for relating to human beings?

0:32:46 – (Keith Plummer): Well, there’s a lot there. Look at some of the words you used. Social protocol, which that entails the idea of a personal relationship. And honestly, sometimes I get annoyed with people who say sorry when they’re not either. But look what’s going on there. The question that is at work there is, is this an instance where this person is sincerely sorry? Because in many cases he or she may actually be with AI.

0:33:20 – (Keith Plummer): They are never sincerely sorry. There’s never any sense of wrongdoing. There’s Never any sense of contrition or regret and so forth. And now with the Bluetooth, yes, the Bluetooth speaker says, sorry, pairing couldn’t be accomplished, and so forth. But the difference there is the Bluetooth speaker isn’t being presented to you as someone or some, yeah, I’ll say, someone with whom to relate. There’s a message that is being given in order to inform you of something, but you are not being drawn in and invited to have an ongoing interaction as though you are dealing with another person.

0:34:09 – (Keith Plummer): So it is that desire to present as though personal that does concern me, particularly when we start dealing with, you know, like children and young people. But I don’t think that they alone are vulnerable to it. I think we all are in varying degrees. Someone who has been really influential in me is Sherry Turkle and her writing. And she talks about how it is that when she was at MIT and she was being shown social robots, there was a robot that was seeking to.

0:34:53 – (Keith Plummer): It was designed so as to follow movement with its eyes. And she talks about how even though she knew what was going on and others in the room knew what was going on, they were all trying to get its attention. And there are vulnerabilities that are at work in us that if we do not continue to keep before us and each other, we are susceptible to. And we might talk about this later, but I think the danger is that we become so enthralled with this human simulation, all the while we become bored with actual humans.

0:35:42 – (Christopher Watkin): It’s a chilling thought, isn’t it, that LLMs can hack us in that way, in a way that, you know, Bluetooth speakers almost certainly can’t, that they can simulate contrition or even desire. And we can legitimately, on an experiential level, fall in love with them. And many people have. It is we are hacked, aren’t we? That is the right language to use when that happens. These, these corporations and the large language models they present us to with are very successfully hacking our human sort of setup in order to get done what they want done.

0:36:23 – (Keith Plummer): Someone sent me a video of a Honda concept car that has AI integration. And it shows this person driving along and the car is asking them questions about where they want to go. Isn’t the sunset beautiful? And so forth.

0:36:37 – (Christopher Watkin): And.

0:36:38 – (Keith Plummer): And the voiceover comes and it says something along the lines of, your car can be your partner. I don’t want my car to be my partner. But see the presentation of this sophisticated computer tech as though it is a viable means of, of friendship, companionship, and so Forth is very disturbing.

0:37:09 – (Christopher Watkin): Absolutely. And the real sort of one punch, knockout to my humanity there is not simply that the cow is my partner, is that if the cow is my partner, then my partner isn’t my partner anymore. It’s what it does, the devastation that it causes to those human relationships that I’m part of. Mike, if we could turn to you. I heard you say something that I found really profound and fascinating, and I’d love you to speak about it a little bit more.

0:37:39 – (Christopher Watkin): You were talking in one interview that you did about four great idols of the west, and you named them as power, control, comfort, and escape. And I wonder, how do you see those four main idols of western culture mapping onto the way that people use AI today?

0:38:04 – (Michael Graham): So I got these four categories from the late David Pallison, and then I added my own twist. So the idols largely map onto the fight or flight response that we have kind of in our amygdalas. And so power and control are fight type idols. Comfort and escape are flight type idols. So flight type idols have us tempted to seek the incommunicable attributes of God. Namely, we want to have all the omnis. We want to be all knowing, all powerful, all present, so on and so forth.

0:38:39 – (Michael Graham): Flight type idols have us tempted to want to anesthetize ourselves because of the fallenness and brokenness of the world, maybe has us overwhelmed based on the stimuli of our circumstances. So AI presents massive temptations to either ends of that spectrum. Fight type idols or flight type idols. We’re tempted to think that we’re knowledgeable about everything because we can find pretty decent information very quickly now.

0:39:07 – (Michael Graham): And then, on the other side of things, we’re tempted to escape our current reality into chat bots, character bots, as a kind of means of comfort or escape.

0:39:16 – (Christopher Watkin): It’s really helpful. Another thing you said, Mike, really resonated with me as well. You said there are two key questions for us human beings in every situation. What do I want and what am I afraid of? And again, I’d love you just to plug those two fundamental questions into AI. When we think about AI through that lens, what does it show us about what we want, and what does it show us about what we’re afraid of right now?

0:39:48 – (Michael Graham): Yeah. When you think about the history of literature and the history of dystopian literature, all kind of future dystopian books kind of fit into one of those two categories.

0:40:00 – (Keith Plummer): Oh.

0:40:00 – (Michael Graham): What ends up enslaving humanity in the future is either our wants or what ends up enslaving humanity in the future is our fears. And I think that both of those things can be right and in many ways are right. And so the flip side of this is also true, though. For me, all of evangelism and all of discipleship comes down to basically repackaging those same two questions in different ways, in ways that are socially acceptable for however deep of a relationship that you have with somebody. You can’t just ask somebody, what do you want? And what are you afraid of?

0:40:39 – (Michael Graham): But what you can say is questions like, hey, what’s the next thing that you’re looking forward to? That’s a different way to say, what do you want?

0:40:48 – (Christopher Watkin): Okay.

0:40:49 – (Michael Graham): Another way to ask, what are you afraid of? Is, hey, when was the last time that you lost sleep? Tell me more about, like, what kept you up at night the last time that you lost sleep? Okay, so what AI does is it gives us a really long fulcrum to leverage either our wants or our fears. On the want side of things, we’re tempted to want to be like gods. On the fear side of things, we’re tempted to forget the doctrine of the providence of God.

0:41:23 – (Michael Graham): In other words, Jesus holds everything in creation together and is indeed in charge also of all of history. That also includes this new era. And so that’s part of why it’s important that we avoid some extremes when it comes to how we evaluate artificial intelligence.

0:41:42 – (Christopher Watkin): Thank you very much indeed. I’d like to spend some time thinking with both of you about how to live faithfully and hopefully in this current cultural moment in which we find ourselves. And to start, I’ve asked you both if you would share with us a couple of Bible passages, one or two Bible passages that have been helpful for you personally in thinking through issues related to AI. So, Keith, I wonder if he can start with you. What parts of the Bible have you found particularly insightful as you’ve wrestled with AI questions?

0:42:21 – (Keith Plummer): One passage I’ve been thinking a lot about is Jesus words in Luke 6:45. The latter part of that, he’s dealing. He’s dealing with the Pharisees and he’s condemning them. But one of the things that he says there, we’re familiar with these words that out of the abundance of the heart, the mouth speaks. And because we’re dealing so much with words and the proliferation of words, this is just a helpful reminder to me that words are intended by God to be the products of hearts, and they are revealers of hearts. And so as I’m dealing with something that can produce, in a very impressive and fast way, a Flood of words.

0:43:07 – (Keith Plummer): I find it helpful to keep in mind that God’s intent is that words would be the products, the issue of hearts, actual persons. And then another passage that has been on my mind is Psalm 8, where David is contemplating the wonders of nature, and he leads him to ask, what is man that you are mindful of him? And what he does then is he goes on to talk about the privileged status that humans have. And he says, here, I’ll just read from verse five.

0:43:47 – (Keith Plummer): Yet you have made him a little lower than the heavenly beings and crowned him with glory and honor. You’ve given him dominion over the works of your hands. You have put all things under his feet, all sheep and oxen, and also the beasts of the field, the birds of the heavens, and the fish of the sea, whatever passes along the paths of the seas. And as I said before, it is easy, and I think especially so in our day, to become so captivated and awestruck by machines that we have made that are able to simulate us to great degrees of accuracy while looking at the prototype, one another, humans, and ceasing to be awestruck, ceasing to wonder and marvel at what being human is and being bearers of the image of the triune God.

0:44:54 – (Keith Plummer): It’s very tempting.

0:44:55 – (Christopher Watkin): It’s incredibly helpful. There’s a real theme, isn’t there, in a lot of what we’re talking about in this episode? It’s not just what you can do with AI it’s how it changes, subtly changes, perhaps your. Your view of who you are as a human being and how you relate to other human beings. That is often what flies under the radar. And passages like Psalm 8 just put that back, front and center, don’t they?

0:45:21 – (Christopher Watkin): Remember who you are, remember your place, your wonderful place crowned with glory and honor in this universe of ours. Mike, let me ask the same question to you. What Bible passages have been particularly helpful for you as you’ve wrestled with AI over the years?

0:45:38 – (Michael Graham): Yeah, so one passage that comes to mind is Ephesians 1, really the whole chapter, but especially verse 11, which says, in him we have obtained an inheritance, having been predestined according to the purpose of him who works all things according to the counsel of his will. So the idea that nothing happens in all of creation or history outside of God’s will. So that’s an important thing to remember, especially when there are things that are disruptive in nature, like a new technology.

0:46:11 – (Michael Graham): It’s just important to recognize that God’s still in control. And then the second passage that comes to Mind. And this is kind of riffing on a similar theme would be the very, very well known Matthew 28:18 20, which says this. And Jesus came and said to them, all authority in heaven and on earth has been given to me. Go therefore and make disciples of all nations, baptizing them in the name of the Father, Son, and the Holy Spirit, teaching them to observe all that I have commanded you.

0:46:42 – (Michael Graham): And this is key. And behold, I am with you always to the end of the age. So whatever happens with technology and all of these different kinds of things, God is with us, regardless of whether utopian scenarios play itself out, dystopian scenarios play themselves out, or both of those things at the same time.

0:47:04 – (Christopher Watkin): Amen. Yes. No technology, no event in human history gets us deep into the fabric of the universe as God’s providence. They’re all superficial compared to that fundamental reality. What a wonderful platform for Christians to lean into the AI age. With that discernment in mind. Keith, in one interview you did, you made this really, really lovely and actually really quite provocative analogy. You said in thinking about AI, let’s bear in mind that question that all youth pastors have got at one point in their ministry, the how far can I go question.

0:47:52 – (Christopher Watkin): And you said, it’s helpful to think AI in terms of how we might respond to that question in a youth group setting. Could you just tease that out for us? How is the how far can I go Question helpful? As we think about AI, when teens

0:48:08 – (Keith Plummer): ask of the youth pastor, how far is too far? How far can we go in terms of a dating relationship and so forth? And what is foremost in their mind is not the question of what is good, what is profitable, who am I made for? What am I made for? Who and what is the person that I am with made for? The primary consideration is, what can I do and still be okay? And those are two different ways of approaching things.

0:48:44 – (Keith Plummer): And I think the application with AI is again going back to we’ve got to consider who and what were we made for? What is the optimal form of life? Do I really believe that what God has revealed is the way of flourishing is. And if I don’t, what does that do to? And I think many times the questions that we’re asking is, can I use AI to do this without asking that prior question of, well, what are we for?

0:49:20 – (Keith Plummer): Who are we for? What does living well look like? What is the nature of the truly good life as God has created and revealed it? And how might my use of AI impede or further, that that’s what I had in mind in terms of that provocative analogy.

0:49:47 – (Christopher Watkin): It’s such a fundamental question, isn’t it? If we don’t know what we’re for, then how can we possibly work out whether in a particular context it’s appropriate for us to use AI or not? Let’s try and get really practical with this line of thinking. So you teach Keith a course on technology and discipleship. Is there in that course a concrete assignment that you give to your students that you sort of think, oh man, if every Christian would try this, that would be really helpful as we navigate AI? Or one principle that you think, if every Christian bought this in mind, we’d be able to approach AI much more healthily.

0:50:27 – (Keith Plummer): We do a lot of work with media ecology and that’s obviously closely related to our digital technologies. One of the things that I ask students to do is it’s called a technology experiment. And over a five day period, they are to either switch out one medium for another that they normally use with just great frequency. So for example, if you get your news from social media for five days, maybe think about watching television news or reading a newspaper for five days.

0:51:10 – (Keith Plummer): If you normally use your Bible app for your interaction with your Bible. I’ve had students who’ve chosen to only use their physical Bible for those five days. And one of the some students do a fast and I think usually because they realize there’s some problems. So many of them will say, I’m going to go five days without my favorite social media platform. They get quite creative. One young man uses his Cash app for all of his purchases. He said, for five days I’m just going to use cash.

0:51:48 – (Keith Plummer): And then they have to write a paper and they have to answer questions like why did you select what you did? What did you select? What was most difficult, maybe frustrating about it? What, if anything, was advantageous? What did you learn about yourself and your relationship to God and others? And after doing this five day stretch, do you intend to make any permanent changes in your life? If so, what?

0:52:21 – (Keith Plummer): And why now they have to write a five page paper and maybe not everyone will do that, but I think those kinds of questions and that kind of experimenting with an attentiveness to how it is that we are going through that, I think that would be very, very helpful for all of us to do with regularity.

0:52:44 – (Christopher Watkin): Certainly would, wouldn’t it? I, I, I guess one of the greatest dangers of this whole technological tidal wave that we’re in the middle of at the moment, that your exercise there really helps us to get to Grits with is that the danger is that we just sleepwalk into whatever the new technology is. Like, everyone’s using the latest LLM. It’s convenient. It brings down friction in my life. And so we don’t actually stop and think, is this wise before God? What is this doing to me? We just think, hey, I can save a few minutes by doing this.

0:53:22 – (Christopher Watkin): Let’s go. And that exercise is so helpful in just pressing pause on that instinct to make life easier all the time and to ask deeper questions about it. Micah I want to turn to the future and the question of hope. Trikos is a huge theme in the Bible, but also a really persistent and prominent theme in AI discourse today. There’s a lot of hope going about. Hope for what? The future might be shaped by AI.

0:53:54 – (Christopher Watkin): And how, as Christians who want to calibrate our lives by the Bible, can we avoid falling into these camps of either being an AI doomer thinking that it’s the worst thing to ever happen without knowing quite why, or to be an AI boomer thinking it’s going to solve all our problems next week? And what does a distinctively Christian hope for the future look like, distinct from the sort of techno optimism you often see in AI circles?

0:54:28 – (Michael Graham): Yeah, so when I was a kid, and this is going to date me a little bit, if you’re younger than what I am, we had these things called Rand McNally maps. And they were these things that were probably, I don’t know, a foot and 18 inches by 18 inches, you know, half a meter by half a meter, you know, and you’d have them in your car. You could get them at, like, the grocery checkout line. And at least in the United States, you know, before gps, before smartphones and all this stuff, this is what you use to, like, kind of circumnavigate, especially, you know, interest, you know, interstate trips.

0:55:02 – (Michael Graham): And, you know, if you’re older, maybe a different. Or, sorry, if you’re younger than me, maybe a different analogy here might be the idea of Google Earth. Okay, so the Rand McNally maps would give you pretty detailed instructions, especially for various metro areas. And so you could really see instrument this detail of, hey, here’s how to circumnavigate. Same thing with Google Earth when you’re dealing with satellite images and stuff like that.

0:55:35 – (Michael Graham): And, you know, oh, you know, I showed my kids, like the Eiffel Tower and, you know, Leaning Tower of Pisa and whatnot. But, like, here’s the key, you know, one key in all of this is kind of how this technology functions. You know, none of us would ever confuse a map or a satellite image from Google Earth as being the same thing, as being embodied in any of those locations. And so all of those technologies can tell us, you know, whether it’s the Rand McNally map, Google Earth, or a large language model, they can tell us a lot of very important, factual, even true information about various different things.

0:56:23 – (Michael Graham): But it’s not the same thing. We would never confuse a map or a satellite image as being physically present and having experienced all of those things in our mind, body and spirit. And so that’s the. So I want to start there in terms of framing, okay? So I want to say that the technology is actually good, and it’s actually better than the technology that we’ve had. The map is improved, but at the end of the day, it’s also still a map.

0:56:57 – (Michael Graham): And so there are two things that I would want to say from a theological standpoint that I would push back for the people who are doomers. And the two doctrines that I would push back on would be the doctrine of the providence of God. And we’ve talked about this in a number of different passages. God’s in control of all of history, all of creation, Colossians 1, all sorts of passages. And then the second doctrine is the doctrine of common grace.

0:57:25 – (Michael Graham): Okay? So in the doctrine of common grace, there’s all sorts of things that people, even people who don’t have any kind of saving faith or knowledge of the gospel, they can do things in culture and society that make dramatic improvements on the way in which we interact with our fallen world. And so there will be ways that AI functions in a common grace standpoint. So take, for example, stuff like there will be medical advances, stuff like cancer therapeutics, protein folding, genomics, early detection of medical issues in people, and there’s all sorts of toil versus labor scenarios as well that will play themselves out.

0:58:16 – (Michael Graham): Okay? Now, the flip side of all of this is that, you know, and why I’m not like a radical techno optimist is, is basically the diet of how large language models are trained. It’s trained on human generated content. And what do we know theologically about humans? We’re subject to the fall. And so every inherent weakness that’s inside of all of us will be inside of all of this technology as well. And so I think that there’s going to be all sorts of potential scenarios. I mean, probably the scenario I’m most concerned about is not like the Terminator scenario, whatever those are possible, right?

0:59:02 – (Michael Graham): But I think malicious software and the creation of malicious software will be the very first use case. Large language models are really, really good at software coding. And so probably the first bad uses of AI will be, or fallen uses of AI will be to write malicious code that creates attacks on key parts of infrastructure. Stuff like the power grid or the integrity of the water supply, the integrity of financial transactions, you know, as they’re moving around those different kinds of things.

0:59:39 – (Michael Graham): So I’m not a techno optimist because I don’t think we necessarily have good answers for how to mitigate how these technologies will undo the fact that they’ve been trained by fallen humans. So those are reasons, those are the primary reasons why I’m not a techno optimist. Also, I think there’s a lot of issues that still need to be hammered out on, you know, AI ethics and particularly as it relates to how we navigate children, parental controls, those kinds of things.

1:00:11 – (Christopher Watkin): Absolutely. And I’d love to narrow down on this with both of you. So like you said, we’re not doomers, we’re not boomers. There’s a more discerning road that we’re trying to travel. What does that road look like, both of you, for churches in particular? And I’d love to sort of 1, 2 sentence answers on this if you’re able. So where is AI most helpful for churches right now? One or two bullet points.

1:00:39 – (Christopher Watkin): And where is AI most dangerous for churches right now? Keith, let’s start with you.

1:00:45 – (Keith Plummer): I think that it is most helpful for those things that aren’t dealing with the relational component in terms of like scheduling and you know, things like that, some of those administrative tasks. Where it is most dangerous is where it is relied upon for what we might call the one anothers of scripture, where it is taking the place of what the scriptures clearly portray as God’s intent for human on human interaction.

1:01:22 – (Keith Plummer): The scenario that you gave at the beginning, for example, the person who says, I know I should talk to the people at my small group and ask for prayer, but the AI doesn’t judge me, doesn’t yet show understanding and so forth. That’s a real danger because that is drawing away from something that God clearly commands and that is good for us.

1:01:46 – (Christopher Watkin): Mike, I saw you nodding vigorously during Keith’s answer. Would you say similar things?

1:01:51 – (Michael Graham): Yeah, we don’t want to use AI for anything that’s relational in nature or to eliminate human or pastoral interactions. I think by and large churches should be analog environments. That being said, the best piece of AI technology that I’ve seen in a church environment is this sermon translation platform called Kaleo AI. If you’re listening to this and you’re a pastor or whatever, what this technology does is allows you to do real time sermon translation live into any language for everybody who’s inside your church.

1:02:28 – (Michael Graham): So imagine somebody being able to just pop in a headphone earbud and be able to listen to you preach in their primary language. So I think that’s a very good use of artificial intelligence. Most AI is taking you away from primary sources and this is a way to bring you back into primary sources. So Kaleo AI I think, is some really cool technology and is really smart.

1:02:58 – (Christopher Watkin): Wonderful. Thank you. Thank you both for those sort of positive notes and those notes of warning as well. I’m asking all the guests who come on this show two closing questions. What is one inefficient human habit that you’re committed to keeping in the age of AI? And if AI is a mirror, what do you think it’s revealing to you personally about your own heart? Keith, what would be your reflections on those questions?

1:03:29 – (Keith Plummer): One of the inefficient human habits that I’m committed to is consulting with friends about life’s problems. Yes. I don’t want to succumb to the convenience, the ease of, of putting in a prompt to deal with something that I really need, as we said before, wisdom from a trusted person from. And what does it reveal to me about my own heart? I am easily captivated by convenience and novelty.

1:04:06 – (Christopher Watkin): Keith, you’re not the only one. It would be that brave would be the person who would say I. Convenience and novelty have no appeal to me whatsoever. Mike, what would you say?

1:04:18 – (Michael Graham): Yeah, I mean, I’d say something similar to the first question. Relationships are inefficient and that inefficiency is actually glorious and not bad. And that does relate to the second question about the mirror. Not all friction is bad. I think in our technological era the assumption is that all friction is bad. Friction is actually good. And so the underlying principle there is not all efficiency is good and profitable.

1:04:55 – (Michael Graham): And I say this as somebody who’s made his living on efficiency. Efficiency is one of the highest values that I have as a person. But not all efficiency is to be. Here’s a final principle. And I think trust is the most important commodity of the next decade. And I think the fastest way to blow trust in this next decade is going to be over reliance on artificial intelligence. In your thinking and in your relating with other people.

1:05:36 – (Michael Graham): People are going to eventually want to know that man, Keith really loves me. Chris really loves me. Mike really loves me. They really care about me and my soul. And they’re not going to outsource thinking, wisdom, writing, meeting together, gathering, you know, to technology when they could.

1:06:02 – (Christopher Watkin): I’ve never heard that thought put as clearly as that. What a powerful idea that is and a word in season to us. I think as this technology sort of grows to remember the preciousness of trust. And of course, it’s right at the heart of the Christian message, isn’t it? The faith we have in God, the way that we can trust him. Trust is central to who we are as Christians and let us guard it fiercely against all threats to it.

1:06:34 – (Christopher Watkin): We have raised a lot of thoughts over the past few minutes and if you want to take any of those ideas further, could I invite you to join the Silicon Spiritualities Discord, where we’re learning together how to follow Jesus in a tech saturated world. If you pop on the Discord, you’ll find practical experiments like the one that Keith was sharing with us earlier related to every episode of this series and you’ll also find details of an upcoming online cohort as well. If you want to take these ideas further, the link is waiting for you in the show notes and we’d love to see you there.

1:07:11 – (Christopher Watkin): Keith, Mike, this has been an incredibly enriching conversation for me. I’m deeply grateful for your time. Thank you very much for joining us on Silicon Spiritualities.

1:07:22 – (Keith Plummer): Thank you, Chris.

1:07:22 – (Michael Graham): Thank you, Chris.