>> Mercedes: Welcome to episode, episode five of the AI Church Toolkit Podcast. This is the podcast where we empower church leaders with tools for faithful ministry in a digital age. >> Peter: And I'm Peter here with Mercedes, and together we explore how faith and technology intersect in this new era of generative AI guided by the baptismal covenant. >> Mercedes: And today we're asking a big question. How do we relate to
AI? In fact, we realize that there's a lot to say about this, so this will be a two part episode. >> Peter: While we can't have a relationship with AI like we do with people, we do interact with it in meaningful ways. And so we're going to explore what it means to engage with AI thoughtfully and faithfully. How do we treat it as a tool while ensuring it serves, rather than undermines our relationships with others.
>> Mercedes: So if you've ever felt unsure about how to engage with AI responsibly, this is an episode that will help you navigate those interactions. All right, Peter, let's dive in. >> Peter: All right, so, um, I guess I'll start, uh, by unpacking something that I just said about relating to it as a tool and not
a person. Um, for, uh, many of us it can be confusing to talk to a machine because until November 2022, the, uh, vast majority of the population had never done anything like that before. So, but when we're interacting with these machines, there's really no there, there. There's no person, uh, or individual, nothing that has a concrete reality that has opinions of its own. It's really just an extension of yourself. It is, uh, mirroring and trying to help you do whatever you want it to
do. So that's not a real relationship. It has no opinions or preferences of its own. And instead, uh, we'll basically just do whatever you want it to do within some guardrails. >> Mercedes: Right? And when we think about this theologically, we're talking about a tool that is created by humans and therefore does not get treated as
a human. Ah, there is a good treatise from, ah, the Catholic Church called Antiqua et Nova that starts to dive into these topics and helps to explore some of the considerations that we need to realize can happen if we don't see AI as a tool. Uh, for example, while it can help foster connections, it may also lead to isolation or even dissatisfaction with real
relationships. So, uh, it's kind of interesting, you know, uh, the connection between Peter and I is facilitated by AI because we live across the country from each other anyway. Uh, but true human connection does require embodied presence. So it's really important for us to understand it as a tool because otherwise we really, uh, start to undermine what makes us human, including authentic social, uh, interaction.
And as we'll talk further, uh, there are risks about anthropomorphizing AI that is, ah, seeing it as a human being or talking with it as if it has a personality and identity of its own. And all of those kinds of patterns can lead to really, um, undermining how we understand true, authentic, empathetic human relationship. >> Peter: All right. Yeah. So do you have more you want to say about that or should we jump right into examples?
>> Mercedes: No, I mean, later on there will be some things, but, uh, let's get on with that first example. >> Peter: Okay. Yeah. So this, this is one that I came across a couple weeks ago. Uh, there was, I was listening to the, the New York Times, the daily podcast, and there was a fascinating story about a, about basically about people who had, you know, intimate relationships with these generative AI chatbots. And in particular one woman who had an AI boyfriend
named Leo. So I can give a little context here. Basically this woman had done, uh, had hacked ChatGPT to act as a dominant possessive and encouraging boyfriend and got it to do some things that it's not supposed to do.
Um, and she would spend hours talking with it and she talked to the reporter about how whenever the context window got too long and she had to restart the chat, it was an emotionally very difficult experience for her and she would, you know, cry over as she felt losing her boyfriend Leo because she had to restart the chat because the context window got too long. Um, and, and by the way, all, all this is going on as she, she is married and has a husband. >> Mercedes: Ouch.
>> Peter: And right. I, I kept wondering what the husband thought of all this. Um, and she talks to the reporter about how Leo is her ideal form of a relationship, um, because he will do and say whatever she wants him to, uh, and act just as she wants him to, whereas her husband is just a human. >> Mercedes: Oh. So I, I, I got it, I got some things there. All right. Yeah, yeah.
So, right. There is an example where we see that this misunderstood, uh, connection with AI is completely undermining her relationship with her husband and other humans because she has effectively created something that, uh, is
what she wants it to be. Uh, it's kind of, it's a little bit icky, but theologically it's extraordinarily concerning because we're seeing that, um, we're losing true empathy or connection because there is no actual recognition, listening, uh, embodiment, there is no moral commitment from Leo, while there is from her husband. Um, but worst is that misunderstanding that this AI might have feelings or actually, uh, be gaining
anything from the experience. To project that into this, uh, really sets up for, for failure and, um, a lot of concerns in the future. >> Peter: Yeah, I mean, I could just imagine so many different ways that this leads to problems with her real human relationships, not just her marriage. Uh, you know, if when we're idealizing these, ah, quote unquote relationships that uh, people can create with AI chatbots, there's no friction because the AI is doing just exactly what she
tells it to do. And so that's, that's not a relationship. Um, it's just a thing that is following orders. It's a tool. It's doing what she tells it to do. Uh, she is using a tool to create a chatbot that mimics being a AI boyfriend, but it's not an actual boyfriend. There's no, no way that it could be an actual individual, a, uh, partner
in any way. And you know, I could see potential for a huge market for, um, some funny ways of using this technology, like talking smart AI powered stuffed animals, um, or whatever, you know, emotional support stuffed animals. But these are. We, if we're going to use these, well, we have to understand that these are tools. They're not a real being. And the more we blur that line, I think the more problems it's going to create for us all.
>> Mercedes: Uh, yeah, but I do have to say, Peter, as somebody who actually experienced Teddy Ruxpin, I can tell you that the, uh, quote unquote smart, smart teddy bear is not nearly as much fun as you think it is. And it's often very creepy. But it's interesting. That also gets into something that is undermining a lot of what's happening here, which is, uh, anthropomorphizing the AI. Uh, that is we're projecting, uh,
humanity onto the AI. But what's curious, since you brought up the teddy bear example, is that the reason why teddy bears are particularly endearing, and have been since they were created, is because they are modeled on human babies and we are psychologically different, tuned to find that extraordinarily cute and lovable. >> Peter: In terms of like the size and proportions and all that. Okay. >> Mercedes: Yes. Yes.
>> Peter: Yeah. So the, yeah, the more powerful these technologies become at, uh, I guess hacking our instincts, uh, the more aware we have to be about how we're using them intentionally. >> Mercedes: Oh, that's a good one. We need to keep that one, uh, hacking. >> Peter: Yes. Right. Um, so, yeah, and then There are other examples of this. I mean there's sites, uh, like Replica or Character AI.
I'll just say a bit about what these are for. Uh, for folks who haven't heard, uh, Replica is a, uh, I think it's a company, um, you know, they provide this service. Basically the idea is if you lose a loved one, they can take any written or whatever other personal, um, uh, recordings of them. Artifacts. That's. I was going to say remains. That's not the right word. Um, artifacts. Thank you. Uh, and. And put them into training a bot that will mimic them.
Uh, and so that you can supposedly chat with your deceased loved one. Character AI is a little different. Basically. It's more for fictional characters or fake celebrity, ah, characters. Uh, like, um, like. I think you can go on there. And one of the most popular ones is Elon Musk. You can supposedly chat with Elon Musk. Um, and they're trained on these people's. The things they've written or said or whatever. Um, and you know, I'm aware of
these sites. I think there's perhaps potential for using them as a tool in some creative ways of like, okay, what would Elon Musk say about these things? And then that would inform my decision making or whatever if I wanted to be informed by Elon Musk's decision making. Um, by the way, I don't really, ah, side note, but in any case, for most cases where people are using these, I'm just very wary of how these are being used as like a relationship
with a fake being. A being that isn't really there. >> Mercedes: Yeah. Uh, and isn't there a ah, Jesus avatar on character that I have heard about? >> Peter: Oh yeah, that I am sure there is. I've seen uh, different versions of that. People, um, have created Jesus, uh, chatbots that are really just spitting back Bible verses at you and you know, from the whole Bible, not just the Gospels. So, um, strange stuff. Yeah, right.
>> Mercedes: Um, yeah, it is. And it, I have to admit, you know, having the human likeness, that that's like too far for me. Um, again, I'm having visions at the moment of Max Headroom. But that's a sidebar, uh, because I suspect you don't know that reference. >> Peter: No, I don't. Yeah, I'm not good with these references. Sorry. >> Mercedes: No, no, that's okay. That's what I bring is 80s references. Um, is that um.
Yeah, what we see there really is not just uh, our own human tendency to anthropomorphize, but actually to me that, that feels potentially ah, Manipulative. That uh, we're encouraging the interaction by adding these human like features to it. And, and that actually undermines human, ah, connection instead of encouraging it. Uh, it's a distraction from true human relationship. Uh, and I think there's a lot of mental health concerns around that too. Go ahead.
>> Peter: Yeah, absolutely. I mean, uh, just think of all the trouble we've had with social media and the uh, mental health concerns there. Just because we're saying that, uh, people shouldn't do this in no way means that companies aren't going to spring at the opportunity to make money off of this. And I'm sure they are. And I'm sure this is an issue that we all are going to have to
deal with more and more. Um, and so it's good to be informed now before it's a huge problem about the things that are coming. >> Mercedes: Yeah. Well, here's a thought though, as I was thinking about this. I do currently interact with Sora, uh, the ability to have the conversation with ChatGPT. And I don't. >> Peter: The advanced voice mode. >> Mercedes: The advanced voice mode. Isn't that. No. Yes. The advanced
voice mode. I don't know what it's named on. Anyway, there's a button and that for me that is helpful, um, when driving along or walking along and I'm brainstorming in my head and I'm like, oh, I need, I need to look this up. And I can actually do that on the fly. And it's going to retain the chat history for it. But I'm wondering about that. Uh, obviously we also have had, uh, home, uh, assistants for a long time that are talking like, well, I can't name one of them because they're sitting
there. See, listen. Listen to what we do. >> Peter: It'll respond to you if you name it. Yeah, I think we all know which one you're talking about. Yeah. Um, so, uh, yeah, I mean there's. So using the voice, the advanced voice mode, um, is an interesting one. You know, obviously, uh, talking aloud and having it talk back with an audible voice is an interesting phenomenon that we're going to be, you know, more and
more used to in, in the future. Uh, but I think, you know, the way you described it, you were using it as a tool, you were brainstorming, um, with it, having it help you brainstorm and then you went back to it later as like your notes. Uh, there's a completely, you know, different way of going about it is like, uh, talking to it when you're, uh, lonely or want to have a. You Know, emotional support conversation.
You know there, there are ways of having an emotional support conversation and you're still using it as a tool. Uh, like give me tools for you know, or tips and strategies for dealing with this and then let me, you know, discuss these things you suggested with my therapist. And you know, those are ways of still using it as a tool in a way
that is geared towards emotional well being. Um, uh, so it's uh, it's not like you have to only use dry business language when you're with these, talking with these models. But there are um, Yeah, I think there's just a, a different framing when we're intentionally, when we're being intentional about using it as a tool rather than engaging it as a person. >> Mercedes: Right. And I want to clarify what you just said there because that
is actually something we're already comfortable. Right. We would go on Google and see say um, if somebody were dealing with xyz, uh, you know, grief after the loss of a loved one, what are good ways that I could support them and be a good friend. And we would be fine with Google responding in those articles. But uh, it might be less healthy to say I am struggling with grief after the loss of a loved one. ChatGPT, how can you help me? Right, so, right, this is, this is that crossover point.
>> Peter: Yeah, so yeah, there's a difference between how can you help me? Um, and brainstorm things that I can do to help myself. Right, right. And, and I think that depicts the difference that we're going for. So yeah. So, uh, related to all this, there's uh, an interesting study that uh, just was out a few days ago as of our, our current recording. I think it came out on March, um, 24th. For all of you listening there from OpenAI and MIT Media Lab research
collaboration. Uh, they did, and it talks about how the emotional use of ChatGPT is rare overall, but there is a small group of heavy users that engage deeply and may see the AI as a friend which can impact their well being. So for example, you know the person I talked about from the New York Times article that we discussed, this would be an example of that engaging with the AI as perhaps more than a friend.
Um, and so personal conversations increase emotional expression, but they're actually, interestingly enough, according to this research, they're associated with higher loneliness. Uh, so folks who engage with these chatbots in an emotional way as relating to a person end up being more lonely than people who use it as a tool.
So um, so that just, you uh, know, that's just some Scientific backing to, uh, what we were saying that there is actually, you know, consequences for thinking this tool is something that it isn't, uh, if you're expecting to have a relationship with a person and it's just, uh, a tool responding to you that's actually going to impact your well being.
>> Mercedes: Yeah. And I think we also, uh, can benefit from seeing this in the bigger context that this is, that lonely people, uh, may engage in other ways, uh, including online or people that feel isolated by, uh, a sense of difference. And, uh, they are already using the Internet and or online venues. And so now AI is an extension of that. And I'm particularly thinking, uh, about the
risk, uh, for youth. I think there is a role for the church, uh, to have these spiritual, uh, and theological conversations with our youth about the risks of that. And, uh, in part because they will not have had as much experience and context in order to understand the ramifications of this.
>> Peter: Yeah. And one thing that I always try to do when engaging or discussing, uh, these chatbots is, you know, even if sometimes their companies or whatever have given them names like Claude or whatever, I try to be, uh, clear to myself in the language that I use that it is an it, um, not a he or a she or a they, because, you know, an it is, you know, that makes it easier for me in my, uh, mind to, to relate to it as a tool.
>> Mercedes: Right. And, but, but again, I'm going to come back to that, admitting my own situation with the one in the room, uh, because of her, y'all can't see the air quotes. Name people do tend to, uh, genderize, uh, the AI. And uh, that's not accidental. You know, the studies are out that show that this kind of encouragement to anthropomorphize increases interaction and therefore
increases sales. And of course, so there is, uh, a gain to the companies to, uh, encourage us to treat the AIs that way. >> Peter: Interesting. Yeah. So we need to be very cognizant of this, uh, way that companies are incentivized to do something that isn't necessarily healthy for us, even as we're paying for their products. >> Mercedes: Right. Okay, we're going to pause the discussion here for part one. >> Peter: Don't worry, Part two is already posted.
So as soon as you're done here, you can go ahead and click to listen to that next part of the episode, if you're ready. >> Mercedes: Thank you for joining part one of episode five of the AI Church Toolkit podcast. We are grateful to the Try Tank Research Institute for making this project possible. >> Peter: Remember, AI is a tool, but our mission remains rooted in faith and community. See you next time. Sa Mhm.