My Interview with the Creator of that AI-Generated Church Service
300 people flocked to a service written by a machine. Why? What comes next?
A couple of weeks ago, newspapers around the world began reporting on a German church service written almost entirely by AI. As with much reporting on religion and tech the coverage was fairly superficial, so I reached out to the service’s creator, Jonas Simmerlein, who is a professor at the University of Vienna.
Simmerlein was kind enough to grant me an interview, and we had a great conversation about reactions to the service and the potential for religious use of AI more generally.
Below is an edited transcript of our conversation.
David Zvi Kalman: The service that you ran received a lot of press coverage. I’m curious what the response was and how you felt about it.
Jonas Simmerlein: There was this short period of time where everyone was interested in it, but more on a sensational level, people were interested in it like a spectacle. I felt they were really pushing their own narratives, and using it as a vessel. It was interesting that especially the populist and right movement media was interested and giving it more thought than a lot of other media outlets.
People thought it was like an AI cult. So they misread what we were doing, which was using it as a tool in a worshipping context. They thought we were replacing God with the AI, which was just never the case and no one wanted to replace anybody. The intention from my part was to find out if people would have emotional, spiritual experiences.
DZK: I want to get to the experiment in a second, but just to stay on the media coverage for a minute: Yuval Noah Harari has recently talked about how we should be concerned that AIs will develop new religious texts that will cause people to follow them, which to me seems very farfetched. It seems like it's coming out of a place of misunderstanding how religion functions and how religious people operate. How do you understand that mismatch between the media coverage and what you were trying to do?
JS: Well, I think that especially the far right outlets are interested in pushing this narrative of a decay of our society, like we are losing our core values.
I personally got a lot of emails, just people telling me that I'm the Antichrist and Satan and that the church we did the service in is not sacred anymore, it's desecrated or something like that. It's really harsh. I got really hateful responses, as well as positive ones.
I think it all comes from the same place of being afraid, of having this idealistic view of a probably never-existing society, which was homogenous and great, this time where everyone was religious and everyone was faithful and stuff like that. I think a lot of that comes from this place.
DZK: Can you talk about that idea of sacredness, because that's something I picked up on as well. You're now introducing this software that is just as capable of producing a Christian sermon as it is producing a Jewish sermon or producing a sermon about worshipping the devil or a sermon that says that all religions are bad and terrible. How do you think about sacredness in relationship to AI and what it means to use an AI to develop a religious language?
JS: I think we’re going back to the question of, are we generating this AI religion? I think that is really a misunderstanding of what religion is. Religion is not a duty you have; you don't have to be religious, right? It's something people practice because it gives them hopefully something to hold on to and to cling to and that warms their hearts.
So from that stance it is really the question if it's any good at all. The product we had—if I call it a product—the AI church service was kind of mediocre, because it was just an algorithm, a fancy autocorrect module. So it really was important to me to make people understand how banal it is, how trivial it is.
What religion is to a lot of people, to most people, is this connection between humans. The question I had was: is it possible to still have this spiritual connection even though there is not a human pastor involved standing in the front? When we prayed together—for example, when we pray the Our Father together with the AI, there are humans involved.
“People would…be able to have this entity that was a spiritual caregiver to their great-great-grandparents. These are the kinds of application where it really becomes interesting.”
In Protestantism, it’s something that occurs in the congregation itself. It is really not that necessary to have this kind of exalted priest or something that does the religious thing, but we rather believe that God can work through all things. I would leave it to the people that are actually participating to decide if it touches them on an emotional, spiritual, religious level, or if it doesn't. But if it doesn't, I was just curious what was missing.
And I think that there is something missing. I have strong doubts that AI will eventually replace humans, especially in the religious sphere. I much rather would say it could be an interesting addendum, like something we could add to allow us new experiences.
I can give you an example. For example, there is Xian’er, a Buddhist robot monk. In theory—I'm not sure if it works already—it could store the experiences of the community over several hundred years because it's immortal, right? So people would then be able to have this entity that was a spiritual caregiver to their great-great-grandparents. These are the kinds of application where it really becomes interesting.
The misconception is that AI is better because it has more information, whereas spirituality is often the reduction of knowledge—the specific knowledge between, for example, two people, between a spiritual leader and a member of the congregation. This reduction might actually be what is at the core of spirituality, more than the abundance of information.
DZK: Part of what I hear you saying is that anxiety around the use of these models comes from questions around what they mean for religious leaders. But if you’re thinking about these models in terms of what they can do for communities, understanding that spirituality is something that emerges organically within the community—then it doesn't really matter how you're getting there as long as you're helping the community to develop that spirituality.
You already alluded to the fact that this was an experiment. Now that you’ve run this experiment once, what would you imagine doing with it in the future?
JS: I won't do it exactly the same way. I don't see the merit of it because no one was like, “okay, let's do this every Sunday.”
There’s two different ways to approach this project. The first one would be to realize a project where the human element is reduced as much as possible and see how far the AI can take us. That was the experiment I conducted. What I find way more interesting personally and promising is what I alluded to, this synergy between human experience and AI.
I think we could combine these to generate new experiences. So for example, AI is now pretty decent at creating imagery. We could imagine, for example, having a sermon which is accompanied by this interesting visual conceptualization of what is said. Just a pretty basic example.
There are also some robots, for example, that people use for their prayer practice. For example, there's CelesTE, which several people use in their daily evening prayer routine. I personally am not really impressed by CelesTE on a technical level, but I have to acknowledge that these people say that they are finding some value in praying with their robot basically every evening before going to bed.
What also we saw happened when Replika was used broadly by people as something like a friendship AI, or for some even a form of partnership between AI and human. We can see that people are actually capable of forming some bond and integrating it into their emotional life.
“For some people listening to Bach is some form of religious experience, but no one would ever consider worshipping the organ.”
And for that reason, I think we should also think about what it does to religious practices without, like, throwing off the guardrails, right? I'm writing a paper on this specific question: What is specific to religious practices that need this special care from an ethical standpoint? Because many tech enthusiasts are not religious.
DZK: I wanted to go back in time a little bit. In the 16th century, and also earlier, there were prayer robots or prayer automata, physical devices that looked like supplicants praying. They seem to have been designed especially within Catholic circles, as a kind of devotional object, and creating this praying automaton was a way of expressing one's devotion to God.
There was some amount of controversy, especially from Protestants, arguing that there is something kind of inappropriate or inauthentic about this. I'm wondering if you want to draw a connection between that debate and the contemporary debate.
JS: What makes it hard for me from a research standpoint is knowing how it was actually used. We know these artifacts exist but it's hard to tell if they were more like artistry objects, something to boast about when you had visitors, or if people were actually having religious experiences with it. That's just something we can't go back to.
Insofar as there is a connection, I think it’s that there were always a lot of voices being critical about technical innovation in religion and churches. When the organ was introduced into the church—that's a comparison I use a lot because for some people listening to Bach is some form of religious experience, but no one would ever consider worshipping the organ.
There is always this fear of introducing something that is inauthentic, separating the believer from God because it introduces this tool, which then kind of stands between the believer and God. That was the reason why people would tear down pictures from church walls during the reformation. Not the only, but one of the reasons.
I think this is always an interesting and valuable conversation because it takes us back to what actually matters about religion, right? I think it should be led because people might rightly criticize introducing something that then becomes not a tool, but [an end] in itself. And people might at some point actually have a stronger connection to the AI algorithm than they do to humans, and then it could actually be problematic insofar as religion is originally intended to create community. It's hard to tell; people might actually gather around AI and form new bonds and communities and congregations.
This might not be a one size fits all model. It was interesting: I let the people who participate take a survey about it. And although the data is not analyzed completely yet, it is already clear that there is an age differentiation. Not surprisingly, people from 50 to 60 upwards were more defensive, or even disgusted or bored by it, while younger people were way more open to it and were like, yeah, it's not there yet, but if it's technically more advanced, I would actually welcome having these agents in the broadest sense in our religious community or practices.
DZK: I have one more question for you. In a lot of religious contexts people are encouraged to treat the things they don't understand with a sense of reverence, that there are things in the world that are beyond the comprehension of humans, that is one of the reasons to have reverence for God. Now, there are lots of things within AI that we don't understand. There are all kinds of AI hallucinations, AI making ideas up that we don't fully comprehend and may never comprehend because of the complicated ways in which these systems are developed.
Do you think it would be appropriate to read that same sense of wonder or awe into AI hallucinations? Is it legitimate to treat the parts of AI that we don't understand as having some kind of holiness in them or something godly in them, just because they are incomprehensible?
Maybe I’ll phrase this differently. There are some people who try to communicate with God by doing something random, say like opening the Bible to a random page and letting it fall open. Do you think there's a way in which the hallucination aspect of AI could also be seen in that same way?
JS: I would trace it back to the experience of the believers again. Right now I'm rather skeptical because as the things stand there is not an overwhelming reaction of people having these awe moments with AI.
I think we have this really technical view on AI still. We are interested in: How does it do? Why does it do these things? We are looking at AI from a really scientific point of view. In my experience, we are curious about it in a different way than we are curious about the Grand Canyon, which is kind of like this prototypical natural and awe-inspiring thing. We are just too used to viewing AI or digital media as a tool and are just not used to viewing it as something religious.
I think that this is also a matter of culture. In the Shinto tradition, which imbues objects with spirits, for example, they have a far easier time to include robots into religious practices, whereas European Christians are not coming from a tradition that is really accustomed to this. We are also just not yet experienced enough to form these bonds. If we are used to AI all around us, we might actually have an easier time connecting, also in a spiritual or religious way.
“We are curious about [AI] in a different way than we are curious about the Grand Canyon…We are just too used to viewing AI or digital media as a tool and are just not used to viewing it as something religious.”
Many scholars—for example, John M. Jordan—pointed out that the problem with robots is that we have this pretty advanced scientific view of what they could be. We have these great examples, like R2D2 and the Terminator. So it makes it even harder not to be disillusioned by them, because our hope, our expectations, are way higher. We already have this sci-fi idea of what they could be—and then the reality is so far back that we actually find it hard to connect to them because we have already formed this ideal version of what they could be.