AI Church Toolkit

Part two of the AI Church Toolkit podcast explores how AI affects Christianity

>> Peter: SA.

>> Mercedes: Welcome to episode five, live, part two of the AI Church Toolkit podcast, where we empower church leaders with tools for faithful ministry in a digital age.

>> Peter: I'm Peter here with Mercedes, and together we explore how faith and technology intersect in this new era of generative AI that we're in. Um, guided by the baptismal covenant.

>> Mercedes: Today, we're discussing a big question. How do we relate to AI? This is part two of that discussion, so if you haven't already, please check out part one.

>> Peter: While we can't have a relationship with AI like we do with people, we do interact with it in meaningful ways. So in this part of the episode, we'll continue exploring what it means to engage with AI uh, thoughtfully and faithfully as a tool rather than as a person. Okay, so this is less of an example of how relating to these bots as a person can be harmful and more of how it is just unproductive. Um, and, you know, as we are discussing this and, and sharing with you on this podcast, we want us all to be able to, to use these, these chatbots for the, for our, you know, our work in our ministries in service of the gospel and in service of the people around us. So, um, you know, this is not, uh, this may not be an actively harmful situation, but it's still not as, as helpful as it could be.

ChatGPT is a generative AI that answers questions based on context

So with that framing, I'll explain what, what happened. I, I had this experience where I was talking with a colleague about, about generative AI, asked them if they had ever used it, and they, uh, responded saying they had tried it a couple times, but they didn't find it to be very useful. And they told me about this, uh, a long conversation that they had had with ChatGPT, where they introduced themselves to it, asked it about its theology. Uh, at some point they talked about how they wanted to not, uh, use male pronouns for God. When ChatGPT used male pronouns for God, they were like, well, I, I don't like that. Um, and then so they became confused when. And frustrated with the fact that ChatGPT seemed to continually be changing up its theological standpoint. It said different things that were contradictory based on, um, how they were interacting with the, the chat bot. And, uh, they were super confused and frustrated and basically gave up because they thought it wasn't a good, uh, theological conversation partner. But, you know, we, uh, didn't get into all this. But I think what I would say is, like, it. It's not a bot. Like, the bot does not. Is not a person with opinions of its own. It's uh, what it is made to do is just give you the best or most likely or most preferred response to whatever you prompt it with and with some, you know, inherent randomness, uh, to it. So it's going to be different each time. And so in each interaction it's constantly learning what you want and responding with more tailored responses to whatever you prompt it with. Uh, so in that conversation, as my colleague was Talking with it, ChatGPT was shifting its theology from uh, one moment to the next because it was um, trying to give theology, ah, theological response that would be most helpful. And it started off with the male pronouns for God because that is what is most was common, um, but then, and then changed as it learned more about who it was talking with.

>> Mercedes: Right. And I think this is a good uh, point to remind, we talked about this a couple episodes ago, that uh, LLMs, uh, the, the gen AI text models, uh, and, and even the other ones are just pattern, uh, matching based on the rules and training that it received. So uh, as the colleague continued to ask questions, it was giving little bits of details to the GPT to allow it to come up with a better pattern match. Right. But it, it does not actually have a fixed theology of any sort. In fact, if anything it has uh, as much information, public, publicly available information uh, that it, it, it can get. And so it's just which one it thinks matches the question best.

>> Peter: Right? Yeah, there are a lot of interesting, kind of funny, uh, but also a bit stupid examples of people asking like you know, asking uh, chatgpt, do you believe in God? Do you whatever? And screenshotting it and saying look, ChatGPT believes in God or whatever. And you know, that's just because it is responding the way that it thinks you want it to. Right.

>> Mercedes: And to, to that point that's actually also one of the strengths ultimately that we've talked about when we say that uh, in order to get a good answer from ChatGPT, you have to build context with it. So that is built either in the prompt where you give it the very specific information of the context so that it has a better sense uh, of the pattern that it's supposed to match. And uh. But I was laughing as we were working on this episode because the flip side of that is I have heard myself say many times that my chatgpt knows me or knows my context and uh, it doesn't really uh, it's just that my conversations with it have built its training to include my context. And uh, ironically uh, in another example of how they use Terminology to encourage anthropomorphization. I can't say that ChatGPT calls that memories when we know that computers don't actually have memories. Um, so they're just the puzzle pieces that it uses to help, uh, build the overall picture, right?

>> Peter: Yeah. Yeah.

>> Mercedes: So what do we want to do with it?

>> Peter: Well, I, Yeah. So, I mean, some ways to shift your approach if you're, um, trying to engage with these tools in a productive manner. I would say don't ask it. What do you think about XYZ topic? Instead, say, you know, brainstorm some different ways of talking about this topic that might be useful for my context. Whether that's a, you know, urban church in Seattle, uh, is, you know, denominationally Episcopal or something else or whatever it is, you know, give it the context and it can, uh, it can brainstorm with brainstorm things that will fit your context and then you're using it as a brainstorming tool rather than a being that thinks for itself self. Um, so, you know, um, we talked, uh, a few episodes ago about brainstorming, uh, for sermon writing. Another example you can do is like helping brainstorm what new ministry youth, uh, that church could undertake. And instead of asking, you know, instead of turning to chat GBD and saying, what do you think my church should do? What new ministry should we undertake? Say, okay, brainstorm 10 different new ministries that my church could undertake to help us be more visible and active in our neighborhood. Give it a bunch, a detailed context about your church and neighborhood and then see what it brainstorms. And you know, the coolest thing about this that ah, a lot of people don't do is you can go back and say, okay, brainstorm 10 more and then 10 more, or make them better and see what it says.

>> Mercedes: Right. Or focus on, uh, people of a certain age or Right. Or other attributes in order to narrow it down.

>> Peter: Yeah, and it's kind of hilarious how the simple response make them better, um, can be like such a magic prompt in a way. Um, it makes ChatGPT go back, look at what it just sent you, and then try to add a bunch more details to make them more detailed, better ideas. Uh, basically it's tricking it into thinking harder, uh, which can be incredibly useful sometimes.

>> Mercedes: Yeah. Another example where I like to think of it as a little factory line tool is, uh, working on something like a stewardship campaign where the initial brainstorming work is where you need to provide the context in order to shape the message to your environment. Uh, so that might involve choosing a scripture, uh, and a uh, project in your church or maybe something that's uh, high on the priority list or an awareness uh, of where you are as a church, uh, to include in the theme. Uh, and so again, uh, you're starting with brainstorming the theme around the element of focus, whatever you choose for your church context, and maybe a uh, scripture passage when it provides results. Sometimes it's not what you're thinking and so you can give it prompts to direct the brainstorming in one direction or another. Say um, no, I want it to be more nature oriented or no, I want to use the model of building for the future, whatever, to help direct that theme. But then ChatGPT is uh, or other AI are pretty good about framing it out with all the bits and pieces of what go into the theme. Uh, but then where the assembly line part comes along is uh, there are multiple templates out there for letters and emails. Um, and usually the stewardship process begins with pulling out last year's templates and starting to look at how we're going to update them. But that part, ChatGPT can do for you that uh, once the theme is defined, then ask uh, it to upload the documents and ask it to revise for a draft. We still have to review the draft to adapt those to the new theme. Uh, and then, you know, I always include in that sections that uh, where our personal reflection is included, uh, whether that's members of the stewardship team or me or my wardens. So uh, that's always also included in there so that it's still a personalized response as well. But that takes a huge chunk of time in terms of getting that first element of the revision done.

Can I give one more funny example, uh, about how using AI Chat can be problematic

>> Peter: Can I give one more funny example, uh, about how using it as a tool, not a person. Um, so I came across this in a, in a Facebook group called Episcopalians on Facebook, which has, you know, something like 70,000, um, people on it. And someone posted in the group a screenshot of how they asked ChatGPT to roast the Episcopal Church, which I think is hilarious. Like so many people have done, like ask ChatGPT to roast different things. And I've seen that, but I had not seen it for the Episcopal Church yet. And I, it was hilarious. It was like spot on. Um, like I can laugh at ourselves. You know, we're uh, we are a very particular group of people. Um, and, but it was so funny to me how many people got mad@chatgpt in the comments. Uh, more than, you know, the person who posted it or whatever. There it was like, you know, it was just doing exactly what the person asked it to. But there are all these people who were saying, oh, like, uh, who are treating it as if that is ChatGPT's opinion about the Episcopal Church, when in fact it could just as easily respond, uh, to a prompt to write a defense of the Episcopal Church and advocate why it is the best of all churches and everyone should attend it. It's just doing whatever you ask it to. Um, but people were like, oh, this is just another bit, uh, of proof that ChatGPT and all AI is completely of the devil or whatever, you know, some ridiculous comments, um, and it'll just do what you ask it to. So it's. You shouldn't, you know, the output you get is what you ask for. Don't blame the, the tool, Blame, you know, the person who did that.

>> Mercedes: Right.

>> Peter: Um, although I still think that person had some good sense of humor posting it in the Episcopalians on Facebook group. But, yes, absolutely.

>> Mercedes: Right. Yeah, yeah. Don't, don't think there's somebody sitting behind it actually doing the roast. There's nobody trained it to do that. It just knows, uh, how to follow that pattern. Right, right. All right. Well, okay. So, uh, what then, as we think about these various examples, uh, what are some guidelines that we can think about going forward to, uh, help manage, uh, our tendency, ah, to anthropomorphize this device?

>> Peter: Yeah. So, I mean, um, for me, I think, uh, the thing I've repeated ad nauseam already is just use it as a tool, not a person. Make sure you're not using it for emotional support as a relationship. Um, and I've said before, I find it useful to refer to it with it pronouns. That helps me, uh, know that I'm talking about a tool and not a person. One other thing I haven't said yet related to that, um, is that there's this whole school of thought that you should be polite to your AI Chat boss because that helps you get a better response. I, uh, am actually finding myself more and more of the opinion of the opposite camp, or where you don't be polite. Um, don't say please and thank you. You don't have to be rude, but, um, but just tell it what to do and expect it to respond because it's a tool, it's not a person.

>> Mercedes: Yeah, I think those of us from the south are really struggling with that one because it's just like, ingrained. Ah, and I, you know, again, back, uh, to the bot sitting over on the shelf, um, There, there is a tendency in this household to say please and thank you because we know our children are listening. So we're like modeling our manners just like this. So, uh. But yeah, you're right, it's misleading us. But I think uh, we also have to come back to not just how we are choosing to handle the situation, but also having a responsibility, uh, to our members and particularly the youth, to be clear educationally about this and uh, to recognize the risks, uh, to mental health and theological misunderstandings or, uh, uh, poor. You know, uh, if somebody's going to go out and ask what ChatGPT believes and uh, not understand that that's not necessarily what we as a church teach and believe. That's really important that we specifically, uh, offer education around those risks. And as we mentioned earlier, you know, the, the using the human, like features, like a face, uh, really kind of gray area relative to the fact that it is actually incentivizing the use. Even though some people might see it fun. I think there's also like an important understanding of recognizing that it is human created behind the scenes through the training models, et cetera. It's something that uh, is not independent. Uh, sometimes we get, uh, perspectives from movies, uh, classic science fiction books, but especially in movies where we see it, uh, look real can uh, lead us to think that that's what's happening, uh, in the systems that we're interacting with.

>> Peter: Sure, yeah. Lots of examples of that. Foreign. So great.

Mercedes: AI has no place in sacramental rites

Well, shall we wrap it up by talking about the baptismal covenant?

>> Mercedes: Absolutely.

>> Peter: So, uh, yeah, the first one. Mercedes, what do you think about this? Will you continue in the apostles teaching and fellowship in the breaking of the bread and in the prayers. How does that relate to our conversation today?

>> Mercedes: Right. Well, I, uh, still remember, uh, the bishop sitting at the table who said, I don't want to see a robot making Eucharist. So I have that image stuck in my head, uh, although I haven't looked at the details of it. But there is a European church that tried, ah, AI led church. Um, but specifically, uh, AI has no place in the uh, sacramental rites in terms of officiating that which has been, uh, ordained to humanity, if that makes sense.

>> Peter: Right. Since it's not a person itself, it can't perform any, uh, rites, it can't pray. It certainly is not ordained and can't perform sacraments. Um, and. But you know, thinking about how it could be used as a tool, uh, you know, you could have a robot, uh, stand at the altar over in front of, you know, some bread and wine and do all the motions. No prayer would happen. Uh, there would be no Eucharist or whatever. Maybe that could be used as a training tool. I don't know. I'm curious.

>> Mercedes: I'm not going down that one.

>> Peter: Yeah. Yeah. Okay. That's ah. That was just a wondering, not a, not an actual opinion. Um, but I do think that, um, I would say that AI, as we've been talking, uh, about it, can be useful in writing prayers, in writing other things and brainstorming as we're going, uh, back and forth and engaging with it as a thought partner. But it's not able to perform a prayer itself, um, because it's a tool and not a person.

>> Mercedes: Yeah.

We are called to love God and love our neighbors

Next up, uh, we have. I'm actually, if it's okay, I'm gonna take two of them together at least for my response. Uh, okay. Uh, persevere in resisting evil, and whenever you fall into sin, repent and return to the Lord. And will you seek and serve Christ in all persons, loving your neighbors as yourself. Uh, so I'm gonna, um. I'm, um, being in priest role, like it might be a little preachy in this moment, but yes, we are called to love God and love our neighbors. And uh, the relational aspect of this is central, uh, to our beliefs. Turning that, uh, core expectation to find Christ in each other and trying to get, uh, that relationship from a tool, a robot, a computer, is, in our terminology, sinful because we are turning away from God and it's drawing us away from, uh, our relationship with God and our neighbor.

>> Peter: Right. I mean, I can see a world in which becoming too emotionally absorbed in a machine rather than in the people around you is evil because it draws you away from, from the relationships that we are, uh, that we are made for. Uh, as you're harming yourself and perhaps harming others by, uh, refusing real relationship, I can see how that can lead towards sin.

>> Mercedes: Absolutely.

>> Peter: So what about, uh, will you proclaim by word and example the good news of God in Christ?

>> Mercedes: Okay, so again, I'm going to get kind of theologically specific here. The good news is the Gospel or, you know, the story of Jesus. And I would say that the very incarnation and the theology of incarnation of Jesus emphasizes the importance of, uh, relationships, uh, with God and each other and in that connection that is gained and how we mature and grow spiritually through that connection. Uh, I'd also say that the Trinity, uh, is another place that we can lean into understanding as Christians how we are supposed to be involved in, uh, relationships that way.

Peter: Since AI is not a person, it can't be harmed

Which brings us to, uh, our final one. Peter, what do you got for. Will you strive for justice and peace among all people and respect the dignity of every human being?

>> Peter: Yeah, I mean, I think the. The simple thing to reiterate is that since AI Is not a person, it can't be harmed. Um, uh, you know, that doesn't mean abuse it verbally. Uh, you know, if we're thinking about it as a tool, then, uh, perhaps we can think about how just as a gardener, for example, might take good care of their tools, we can do likewise. Um, you know, and taking care of our tools, well, also m. Is a way of taking care of ourselves. Uh, but I wouldn't worry about harming the chatbot, because you can't actually harm it.

>> Mercedes: Right, but on the other side of that, I would argue that, uh, actually treating AI As a human instead of a tool diminishes the sacred, uh, uniqueness of our identity as humans and, uh, children of God made in the image of God.

>> Peter: So there you go.

>> Mercedes: Okay, so I, um, hope you've enjoyed this conversation. Um, here's the short version. The takeaway. AI, ah, can support ministry, uh, as a tool, but it's our relationships with God and others that matter most. Uh, and going forward, I hope you will consider subscribing to the AI Church Toolkit podcast and leave us a review.

As always, we have some listener homework. Uh, reflect on how you currently relate to AI

>> Peter: So, as always, we have some listener homework. Uh, reflect on the way you currently relate to AI as you're using it. Uh, assuming you've been using it as you've been listening along, ask yourself, do the ways that I'm using it enhance relationships or replace them? And are there certain boundaries I need to set in my own use of A.I.

>> Mercedes: Uh, and now I'm wondering about that at home, too. Uh, so, uh, there's, uh, several articles that we will actually link in the show. Notes for this, uh, that we didn't mention specifically, but, uh, were sources for our discussion today. So, friends, uh, thank you for joining us for episode five. Uh, and we want to let you know that the AI Church Toolkit podcast is made possible by the Try Tank Research Institute.

>> Peter: Yes, thank you very much to them. And so remember, AI Is a tool, but our mission remains rooted in faith and community. See you next time.