Digitaler Salon: Gott und die Maschine

[music] Digital Salon, a format on issues of the networked present from the Alexander von Humboldt Institute for Internet and Society. [music] [music] Hello and welcome to today's digital salon here at the Alexander von Hümbold Institute for Internet and Society. It's wonderful that so many of you have come here in person, and hello and welcome to everyone sitting in front of their screens watching us from there . Today we are dealing in a certain way with opposites, that was at least my first, um, association, but perhaps also with a really great symbiosis. Everyone can form their own opinion today. It's about faith and knowledge, about tradition and the future, about religion and artificial intelligence. God and the machine have changed how artificial intelligence, religious communities, institutions, and religious practices have changed things. That is the full title of today's digital salon. And before we really delve into the topic, I have, uh, the following to say. The attack on the Berlin CSD at the weekend has shocked us all and we do not want to ignore it; our thoughts are with the victims, their families, and the queer community. Nevertheless, we are also holding this event because it is a form of stance . And our three experts up here on the podium also research religion in the digital space and its downsides. So, how AI fragments communities, how, um, people can become radicalized. Um, we'll touch on this topic, but the real question for us tonight is: What is lost when people turn to a chatbot with their religious questions instead of to scholars or the community – precisely what churches, temples, and mosques have offered for centuries, what they are meant to be ? My name is Ronja Bachofer. I work as a journalist, I'm hosting this evening and I'm really looking forward to the discussion. And thank God, which brings us to the topic, or perhaps thank AI, I'm not sitting here alone, but have three wonderful people with very special expertise with whom I can discuss this. And now I will introduce them one by one . Kaja Reich, um, is sitting right next to me. Kaja, you are a cultural anthropologist and religious studies scholar, and you work at the Federal Working Group on Religiously Motivated Extremism. You have led many workshops on discrimination, conspiracy theories, extremism in social media, and you also look at how artificial intelligence is used in relation to extremism in social media . Thank you so much for being here. [applause] Then we continue. Nad the Hotti. You are a PhD-holding sociologist, a research associate at the German Centre for Integration and Migration Research (DEZIM), and a Fellow here at the Humboldt Institute. And you also do a lot of research on racism and discrimination on the internet. For example, you look at how people come into contact with extremist or radical content on social media and subsequently change their attitudes. and you also research digital religion. I'm very curious to learn more. How fitting that we have you here today. [applause] And last but not least, Friedhelm Hartwig. You [snort] studied Islamic Studies, Arabic Studies and Theology and you are now a research assistant at Modus ZD in the field of monitoring. For example, you look at how knowledge is passed on online, how extremists communicate with each other. We will certainly learn more about this later, and I am curious about your perspective, [applause] because the people here in the room and uh, others may not have dealt with the topic of AI and religion quite as intensively as all of you, I would like to bring us all up to speed first, and that's why I'm asking the first question. You can, uh, think about it and decide amongst yourselves who will answer. What forms does artificial intelligence currently take, or how is it used, in faith and religion? Yes, uh, I'll just start. So, in 2023, in our monitoring, artificial intelligence first became more noticeable to me. That's where it came from, namely through the generation of content that covers a broad spectrum. So, these were images, videos, and also translations, and this has been a major concern for us since last year and this year: automatic translations of content from abroad are flooding the German market, social media, in the area of ​​Islamism, and the constellation of channels is now a new one, which is probably also being overlaid from abroad. Um, there are also other forms, but maybe you can say something about that. [snorting] Yes, I'd be happy to say something about that. Um, well, in addition to what Friedhelm said, I would distinguish between three forms. So, firstly, it's about acquiring knowledge, where you look up or a chatbot asks about specific Bible verses or Quran verses and then gets answers, and secondly, it's also about providing guidance or support. Um, and the third one, and this has been around for a while and is n't just since the Large Language Models, which are now much more present in the discourse, um, in practical application. That means, in apps, um, there have been regional apps for a very long time, for example, that remind you when to bid, when to fast, and so on. Um, and even if there is n't direct AI behind it, but personalized memories, AI does play a role there, and you can see that AI isn't always the most obvious thing , but actually plays a part in the background. Very briefly, what is the weighting here? Do you know that? So, are more apps being used than people translating texts, or vice versa, or is it hard to say? So, I took a quick look in the App Store to see which apps have a religious theme, and I could just keep scrolling. There is an insane amount. Um, I definitely think that since the LGBT movement existed, or since it became accessible to us. Um, and you ca n't even measure how much of it is used for that. Um, I think that's an unreported number that can't even be quantified, but maybe you know more about it . Yes, I can't quantify it, but it 's a super interesting question because first you have to define what you mean by AI. If you follow the public debate, you're generally referring to these chatbots, yes, these language models, and not just any language models, but the ones that are used, like ChatGBT, Gemini, Perplexity, and others. Presumably to get information-based answers . That's what the public debate is about, and Kaj rightly pointed out that there are of course other AI technologies that also provide other forms of service, such as apps that offer other religious services. Therefore, one must also differentiate when discussing AI as a topic itself, right, or as a method? Yes, so is AI being used to achieve something, or is AI also the subject of religious debate, for example? Mhm. One has to differentiate here. We at the National Racism and Discrimination Monitor are planning the next wave, which is part of the decimal project, to actually ask this question: how do people use AI and do they use it for orientation, but also for religious questions, in order to be able to quantify this for the first time? Actually. Mhm. Do you have any more examples of how this might change people's everyday lives? So, I think you already mentioned prayer apps or something like that, right, that remind you of things like that. So, do you have any more examples of something like that? Not direct examples, but what naturally distinguishes AI from other clinging communities or direct clinging communities is that AI is accessible at all times, and you have access to a supposed person who is listening, 24/7 in the middle of the night, if you might want a religious answer . And that, I believe, is the big difference that makes AI so useful in a religious context. Mhm. I'm just thinking now, it's so decentralized, isn't it? So, um, if I grow up in a Christian society, I might not be bound to follow the Christian faith anymore, but could it perhaps be done more easily through a Muslim AI or something like that? Yes, this is now being critically discussed in the communities. Um, besides the question of whether something is fake, whether it's real, and so on, what content is being uploaded, I was also concerned with what attitudes and opinions exist and how these are dealt with, as has already become apparent from the comments or from the people uploading the content . And uh, that's uh, um, quite exciting. Around 2023, AI was still being hailed as a way to suddenly conduct missions on a much, much larger scale on the internet and social media. Subsequently, it was also debated that people could now carry out missions who have no knowledge or qualifications whatsoever, and who also lack linguistic competence or can only speak inadequate German, and they could now become active. And meanwhile, a certain ambivalence can also be seen in the Islamist and Salafist spectrum . um, uh, even a critique of how um AI tools are used. Uh, one also speaks of errors being produced. Uh, that it is an institution that is used unreflectively for decision-making and has already led to arguments in mosques, for example, among brothers, uh, that's the way it's put. Um, um, you can see it now, you admit that it's a useful tool, you use it, but you also realize that it has its weaknesses and that there are also charismatic actors for whom this is basically a business model in this attention economy, that these suddenly face competition because anonymized channels using AI are making content available online, right? Mhm. But I'm sure you have other examples. Yes . Um, so you've brought up an important topic, namely centrality versus decentralization, and I think it's easy to fall into the trap of assuming that it 's super decentralized. I believe it's simply a different form of authority in terms of knowledge and a different attribution of authority. What do I mean by that? So, I mean, there are companies behind it that have certain interests, they also have dependencies, and against that background, um, this AI is modeled and the data is made available . And I have spent a lot of time talking and discussing this topic with religious communities that primarily have an oral tradition. We often talk about the Abrahamic religions, which are also, so to speak, religions of the book, and in the case of Islam, they explicitly understand themselves as an oral tradition on top of everything else . Mhm. But there are some that are not canonized, yes, that is, those that are not fixed and written down . And there was the danger, which has been pointed out, of saying, well, insofar as one now begins to reproduce the existing and therefore dominant knowledge of society in the form of AI , one also decides on a form of interpretation or practice of AI. AI is actually not very good at allowing ambiguities. And [snorting] AI also addresses a problem in the case of religion, namely in the course of the modernization of society and the resulting individualization of people's lifestyles, realities, and problems, many acute problems arise, and that's why you can sometimes contact an AI at 11 PM, when the scholar has probably set his WhatsApp to "do not sleep," i.e., "do not disturb sleep." Um, and perhaps the sheer number of questions also makes it necessary for AI to say, okay, can a machine react better? But the question now is, how do you deal with new cases when the AI ​​has been trained on existing data? It is a principle, and I will also perhaps give many examples from Islam or Islamic studies . That's the H. Basically, the idea that one draws conclusions about the current relevance from incomplete evidence; for example, the best- known example is the handling of tobacco. Well, Islam originated in the early Middle Ages, so to speak. Tobacco only entered the Islamic world in the 17th century . Yes, how does one deal with it when one has no reference to it, when one is unfamiliar with tobacco in religious writings and scriptures? And that's what scholars do, considering the methods they use and also the decision to say, well , I have a leap of faith. There is also a degree of doubt, so to speak, but I make a decision based on that, referring to other opinions and also to social relevance. That doesn't work well with AI, because AI can only build on what it currently has available, and in this case, it's endogenous, meaning self-contained. That might also be the first problem: how do you create new decisions when you only know old problems? But how does AI react to that? So she then thinks up, if we stick with the tobacco example, because I know that in old age she experiences this classic hallucination, you know, that she then says, ah yes, this has been decided for ages, tobacco under no circumstances or something like that. Um, or uh, or she says, then I don't know, I don't have any data. Exactly . So, the hallucination problem was still very severe in the last two years. In fact, it's becoming less and less . Therefore, when reflecting on AI and criticism regarding religion, one must also admit that there are improvements, right? In the past, there was a lot of hallucination; sources were even invented, yes, to determine certain positions. In fact, in these cases, as I currently observe (and this is anecdotal), the response is basically to things that are more attributable to policy, and okay, I ca n't say much about this now, you can rely on that or look here. There's some discussion about it, but no final answer has been given. That means there are actually limitations, right? Recently, it was still the case that you would get a specific answer, right, because that answer would actually have been wrong . Yes, and then perhaps the big addition is that there are open source models, and nowadays anyone, even without technical background knowledge, can build an open source model with certain data that you might not even be able to regulate because you host it yourself, and then you don't even know what data it is anymore. That means, uh, you suddenly have a chatbot that has very specific data for an app or for a religious app, for example, and then, uh, might also be able to reproduce extremist content. And that's precisely the dangerous thing: we have these open-source models that can come onto the market, that can be offered , that can be used by religious people, but where you do n't even know if this data is actually secure, if it's regulated, and how you can actually gain access to it. Mhm. And perhaps as a starting point, because before you meant it with, um, Christian traditions and going to Mass or something like that, and where it becomes decentralized. From a religious studies perspective, the question is always: what is religion? What is the definition of that? But a crucial aspect of religion is also the feeling of belonging and the community itself. And even if an AI can convey an empathetic language and we can build a so-called bond with an AI, which has already been scientifically proven as the ELISA effect, it is still not the case that the AI itself is empathetic. And that is the big difference that we absolutely must take into account: that an AI cannot show the sense of belonging and the empathy that a human being can. And that is often what religion consists of: fasting together, celebrating and praying together, going to church, going to mosque or temple, or going on pilgrimages. Um, and I think it's often forgotten in this discussion about AI and religion that in the faith community, the shared experience and practice are extremely important, and AI is often just considered as an afterthought. Mhm. I'd like to come back to the topic of community and how much of it is lost in a moment . Another thought just occurred to me, um, because you just mentioned this authority, and these clear answers that an AI gives, which perhaps a representative of a religion can't give me so precisely, is that also what makes artificial intelligence so attractive to religious people, because you can ask questions and you're more likely to get a yes or no answer than a more considered one? Um, yes, I'm experiencing raw acceptance in the comments . Um, very surprising. Um, more than 90% of the comments on many videos are positive, uncritical, and, uh, downright naive in their understanding of what, um, is a result generated by children's AI and then presented in this way. Um, and uh, there are also, uh, yes, quite naive attributions that it is scientific , that it is neutral, that it now represents an authentic atheist person, for example, when it comes to a dispute or a discussion being simulated in videos about AI, for example, to refute Christianity between such an avatar that is supposed to represent a Christian and an avatar that is supposed to represent a Muslim, or uh, Muslim and atheist. These are such popular combinations. Um, so hardly any criticism, surprisingly little media literacy, uh, that something like AI is basically just a data set. Who compiled this dataset ? It can be manipulated. AI gives pleasing answers and so on. That is extremely rare. On that topic as well. I think what Friedel mentioned regarding skills is very important, and another skill is ambiguity tolerance, that is, being able to have contradictory ideas at the same time, without having to come to a final answer. And I believe that constantly questioning an AI and the expectations it trains could lead to—this is a hypothesis—people assuming that there are things as old as religious services that they should receive immediately, online, always on, always there, and that can be accessed instantly, and that they rely on them. And that this tolerance for amusement, which is also important later on as a form of prevention—namely, not to become radicalized—is extremely important, and that also has something to do with media literacy. The other thing is, Fried also addressed it under the neutral point. There is also a misconception here, namely, that AI is objective. And there is, so to speak, a subsequent scientification of this AI. How come? where you think, well, that's the machine that answers, it has no stakes, it's not subjective in any way, and so on and so forth. Well, it is based on the data and information it uses, and the policies that the respective AI also has, which make decisions about how information is presented . So, there's no such thing as neutrality or objectivity in this, but it's marketed that way because it's technical, and this label, " technical model AI," can create a misconception, right? Mhm. And that brings us back to the question of who ends up with this data, right? So, um, I can imagine, for example, in the context of confession or something like that, that one might prefer to tell it to AI, because it feels a bit more anonymous, and AI is also generally more agreeable, right? And it might say: "Oh, everything's fine." Um, but what's yours? Maybe we can stick with that for now ? Yes, is it a thing that one trusts AI more, or to put it another way, how much do people trust AI, including very personal data? I think one can trust an AI relatively quickly and, above all, think that the knowledge or information one gives an AI remains in a protected space and that one always gets a reaction . Well, you can already tell that it's also being used for mental health. um, to treat oneself with AI. Um, and I do believe that there is no shame involved, meaning that with questions that might seem shameful to oneself , one doesn't then go somewhere else and tell that to a stranger, even if it 's the same religion, but rather that the hurdle of telling that to an AI, where one knows, or supposedly knows, that the data really only stays in that chat, is I think smaller. And where exactly is the data going? So is that also dangerous? So I would say yes, I mean, we still do n't know what happens to the data . It's still a plugin box, and even though there are regulations in Europe, it's still partly American technology companies behind it, and you have to read three times to find out where you actually have to click to not share your data. And um, yes, there we definitely still have a major hurdle regarding where the data actually goes and how we allow it. Mhm. Yes, well, the hurdles to telling something that might be private are reduced because you're not facing a person who might judge it, or at least does n't assume that this person will judge. AI does this in principle because it is self-learning, because it uses the inputs to train itself. There are also other AIs that systematically query other chatbots to obtain information from them as well . That means if the CEO of OpenAI, i.e., ChatGBT, tells us, " Well, no, you'd better share your private information," you shouldn't tell the AI, because we'll have that data. Then the same applies to, uh, religion. That's just, um, exactly the same as the ritual of confession. When there's a wall in front of them, so to speak, you get the feeling that the person isn't really present. One is speaking, so to speak, to a principle, not to a person. It's the same with AI; people say, "Oh, I see, you only have your screen in front of you," but they misunderstand where the data is derived . So, from a practical standpoint, it's also difficult to assess privacy and data security. Exactly. Mhm. Um, I looked up how many people are actually, um, believers or, um, yes, religious. And um, it said that 45% in Germany are now Christian, 4% are Muslim, and only 5% of the people in Germany are religiously active. And now I'm wondering , could artificial intelligence also be an opportunity for, um, religious communities, uh, in that it provides less barriers to accessing religion [ahem], or would you say that it's more superficial because, as you Kaj said earlier, the sense of community is still lacking? So, very briefly, I believe it can be made accessible, or rather, it is more accessible, and therefore it can also be considered an opportunity because there is a direct point of contact. Um, and of course, you do n't have to go to the next town to visit the church; you can do it from home, theoretically. Um, nevertheless, I am convinced that this community , the sense of belonging, and somehow this idea are still needed. I personally find it hard to imagine that a chatbot alone could trigger religious feelings. But I myself am not religious or believers, so I can't really understand that. Um, and I can only speculate. Mhm. Yes, from the perspective of monitoring in the area of, um, Islamism, extremism, uh, I can only say that everywhere this is seen as a great opportunity and people are experimenting with all sorts of forms. This can also be proven for areas of diabolism and terrorist groups. Um, that they are very, very keen to quickly adapt to all the new technical possibilities of the digital revolution and now the upheavals in AI , to experiment with them, to publish manuals, to actively seek out experts who provide support, and there is already plenty of evidence of how active they are in this area. And in the area, um, that's more of a gray area, where actors are on a missionary stance, so whoever is missionary online or whoever sees themselves as an influencer in their field, um, they jump on this bandwagon very uncritically and use all sorts of forms, right? There 's a real boom going on right now, and people say it's more of an introductory phase . Perhaps this is just the peak, it might flatten out, but [gasps] um, as the technology continues to develop and improve, it will probably be one form within the wide spectrum of other available formats. Mhm. And also, um [snorting], two points. Any form of information technology that has low barriers to entry because it's free, e.g., social media, chatbots, AI, and so on and so forth, naturally ensures that people can use it for precisely that reason, right? In the case of social media, this is because they can reach people , create visibility, and then, so to speak, dissociate information within the framework of AI. This is especially useful for marginalized groups who may not have the infrastructure and resources to use it in the analog world, right? Especially curious people, people from religious and other backgrounds, right? [snorting] Uh, racialized minorities and so on and so forth like to use this to do community building , to spread narratives that they normally can't do. A positive example, so to speak . The Native example is, of course, there is this, um, extremist migration that existed back then, at the beginning of the 2000s, to say: "Aha, in the analog world we are very visible, online we are anonymized and pseudonymized, we can therefore use it well for these purposes ." So, that's the double-edged sword. You brought up something interesting. You said, okay, uh, um, well, so many percent, so 5%, uh, are according to this value monitor, uh, uh, actively religious in Germany. I also just saw, for example, the Battlesman—er, religion monitor. Yes, between now 2000 and 2023, active religiosity has decreased, so to speak . For example, of people who profess their religious beliefs, only 13% still pray every day, and so on and so forth. Against that background, it's naturally easier to access it via AI, so to speak . K mentioned something about how, if you were to break down a religion , it would essentially come down to three things that are, well, a form of spirituality, right? It's a form of practice, what one does, and ultimately also a form of identity, right? And this identity, so to speak, also needs a social embedding, and without that it doesn't quite work. But a completely different perspective, which we may not have opened up yet, is the influence that AI has on religion. Of course, religion, through the ethical and moral ideas we have, which may have grown, so to speak, based on Abrahamic religion, can also influence the way we deal with AI. But most recently, uh, AI as a religion itself. Why am I saying this? Um, just because there is such a form of secularization doesn't mean there is no need for spirituality . And this can also take the form of guidance, orientation, moral aspects, and so on and so forth, for which AI is actually consulted by people who may not belong to the active community . And if we briefly shift the focus away from the idea of ​​religious faith as something inherently metaphysical and the belief in an uncreated creator, and instead consider it purely functional— namely, providing orientation, guidance, and so on—it's possible that AI is also used as a religion by those who, for example, don't belong to any religious community. The effects are the same for them as they are for secular people, in that they receive information from a tech company with economic interests when it comes to how to deal with their own psyche, how to handle morally reprehensible situations. The same applies because people have these spiritual requirements for AI, even if they don't belong to any religion . Mhm. I find that really exciting. AI itself as a religion. But can that work? How do they work? Well , I don't know. So what is it then? So, is this a specific app that conveys a certain kind of belief, or how about... I would put it a bit more bluntly . I would say that AI will then become a form of functional god, someone from whom one accepts authority, from whom one receives guidance, orientation and knowledge [laughter]. Omniscience, so to speak. Yes, exactly. So to speak, everything and the data are like the University of Baghdad, but the AI ​​itself, which speaks to you, is like this intermediary, and in this analogy, each person would receive an inspiration. So, and that's the idea, it's not just me saying this, it's the analogy that Claudia Paganini also makes, for example in her book "The New God", which is about AI and so on and so forth. But it's not about an actual religion, but rather, if you look at religion only functionally and what functions it can fulfill, then some of these functions are naturally also taken over by AI. Mhm. Um, if you in the audience have any questions or comments, and say, uh, you can now imagine converting to the AI religion or not, or something like that, then uh, please give me a sign. I'll ask again in a moment . Or does anyone already have an idea, or would like to contribute something? Not yet. I'll ask again later . Uh, to me that sounds like, "Were you going to say something else?" Yes, with pleasure. So AI is ultimately being elevated to a religious level by those who create it in Silicon Valley , right? Ultimately, we're talking about qualities that the currently available simple AI does n't even offer yet, and we're elevating the whole thing to the religious, divine, um, realm, right? Uh, so that AI can achieve properties that surpass human capabilities, right? and AI is presented as a savior for all human problems. So, there is ultimately a framing, a rhetoric, a religious rhetoric, without any empirical evidence whatsoever. And, uh, if you delve deeper into the thinking of these actors who create and disseminate AI, then you also come across ideological paths that I find highly problematic. Yes, will AI, especially when you even consider AI as a religion, increasingly become a competitor for religion, for communities, for people who come together? Because it already sounds like, well, if AI can take over many or almost all functions that a religious community also offers, what happens then? So, are religions dying out? Yes . Yes, well, um, there are two important points there too. So AI can take over certain functions. We've already documented that, right? Mostly, it's about providing information, directions, and so on and so forth, right? Well, if we note that religions have undergone a lot of change over the last 2000 years and have always been able to adapt to the times, and so to speak had to adapt in order to continue, then a lot of progress has taken place. And the question is whether AI can achieve that. That's already the first problem, yes, to say, if you only create religion with AI, the question is whether you are dependent on cultural, historical, time-sensitive contexts, whether that can be incorporated, and that is indeed the first problem. Yes. Um, the second thing is, if one now says, well, religion has much more to offer, as we also, uh, uh, as Kaja has also categorized well. But the problem is that people say, well, what do I need the other things for if I want to know something about the product in the end, and if I get the product in the end , then I'm basically no longer going into the institution. How can this be compared? I always think of the so-called "lost leaders" at such large supermarket chains, where they say, for example, they sell that chicken for €5 because it's super cheap , but they know that if you go in there , you'll buy other things too, right? And the idea is, if you do n't go in for the loser, for the attractive one, for the cheapest one, you get immediate information about what happens when you give everything else, and that's the big problem, right? If one considers religion without social integration, without a sense of identity, without a sense of meaning, then that's when the big problems can begin. Yes, then competitive behavior can also occur, saying, okay, I get the product religion, and that is then, so to speak, the commodification of religion, its displacement. I see it simply as a service. Yes, I can get that done very quickly, so to speak, through AI. Yes, that's the first problem. The second problem is, I'm not going to participate in any of that anymore. And precisely these aspects, e.g., social embedding, uh, meaning-making, no, contexts, also the fact that AI can be very affirmative , saying, okay, not so much challenging one's own values. And religion is precisely that; religion is also, in a way, the assumption that there is an objective morality somewhere, and from that perspective I can judge myself. This is also a form of self-care and self-assessment. And if that is essentially missing, then one has already made a decisive step , also opening oneself up to problematic views. Yes, because some kind of corrective is missing, and also the social embeddedness is lacking, uh, uh, to avoid, for example, joining problematic ideologies . And that's where I see a problem with radicalization, for example. Mhm. Yes, the religious studies scholar from Heidelberg, Imke (um, a link to her lecture is also available at this event, I believe), will be there as well . Or at least he was in preparation . Uh, she made a very nice distinction, which I find very interesting. Well, the discussion already touched on AI and religion, and the further distinction between AI and religion, right? And AI and religion always include at least the three Abrahamic religions, right? For example, the idea of ​​something transcendental, um, a God, um, who can also elude a relationship, who can be contradictory, who is irritating, and um, whether this, uh, can then also be achieved by an AI that is artificially created, created by humans, uh, that is uh, a question for me. And I would like to put forward a thesis: anyone who spends more time with AI as a religion, or rather, AI as a religion, or has a need for guidance, will possibly come to the conclusion that this will be missing. Um, uh, something like, um, yes, uh, theological creativity, uh, new thinking in an, um, new environment. um, constantly redefining themselves. Well, I think that Kaida is relying on a data set and repeatedly reproducing similar answers, which in turn leads to stagnation. Perhaps this is another theory that AI-generated offers and content are more attractive to people who prefer a more literal understanding, a literary approach, a kind of literalist reading, which is what people are looking for, the repetition of dogmas, a conservative stance, or, as I see it especially in my content, a reproduction or new visualization of fantasies from early Islamic times and, uh, visualization of prophets, prophet biographies, heroic ages of conquest, which must be very popular narratives in Salafism and Islamism, and there is also a great hunger for images of hell, paradise , and dystopian ideas of an apocalypse. Uh, they're very well received with high click-through rates, so it's AI that potentially hits a specific target group with a specific need particularly well. Yes . And perhaps to return to the original question, whether AI is now competition for religious communities, and I believe that this can certainly happen and is already happening to some extent, and at the same time I believe AI is also a great opportunity for religious communities. I'm thinking of a different analogy right now, but I recently spoke to a female driver and she said the programs she has at the church are either for teenagers and children or for people over 60 . That means all those age groups in between are not covered at all. And she meant that 20- or 25-year-olds then turn to the Christian fundamentalist free churches, which offer precisely these services, addressing precisely this target group, addressing precisely these uncertainties, and I believe that religious communities must clearly take a hard look at themselves and actually incorporate AI , discuss it and make it accessible to their religious community, and then AI won't necessarily become competition, but rather be considered alongside it. Perhaps it can be considered precisely when it comes to sensitive questions or when it comes to classifying certain sources. Um, and I think AI really has a great opportunity there, and also for religious communities. Mhm. Yes, if I may, sorry. Um, some very interesting points have been made here, and I would summarize them as follows. I believe there is a reality of AI as a competitor to religion that is observable and cannot be denied. Yes, well, we're simply observing that people are more likely to turn to AI than, for example, to local religious communities in an analog way or whatever. This is the real competition that is taking place, in that this authority is being shared or new authorities are emerging. [snorting] The other thing is, from a normative point of view, if we want to be evaluative, there are of course aspects where we say, yes, AI can't do that and perhaps it can't do that structurally either . And that includes something Friedom brought up. AI is n't very good at handling differences of opinion in these aspects yet, is it? A major point, for example, in Islam as a religion is the so-called Iktilaf, that is, the difference of opinions, which is also accepted. In Islam, there are different schools of thought on how to approach topics . Some schools of law allow analogical reasoning, logic for example, while others do not. And AI isn't very good at that yet, due to the fact that it's always striving to give an answer. And I would agree with Kaj to say, well, uh, there is also a reality, given how time is going, that it is becoming more and more dynamic, that it is becoming faster and faster, that information is practically overflowing and that it is difficult to navigate in this time, that it is perhaps also necessary, just to find some kind of technique to react to the sheer volume of data, but to find ways to do this sustainably. There are also examples that are aiming for that, so to speak. There is an app developed in Germany called Islamify, which, for example, at least claims to always have a sort of AI advisory board, a body that determines how the information is presented, on what basis it is founded, and is always checked against expert knowledge, for example. This is one way of saying that it has to be shaped in some way and not just shaped by religions . This is one way, which of course cannot protect against the possibility that certain orthodox women might reproduce the same thing. That's the thing, right? Most data produces what ultimately becomes orthodox knowledge in AI. But, of course, we now have to find ways. There will be no future apart from AI, as far as religion is concerned. It will probably be used somehow as a technical means, as a tool. We need to find a way to make this acceptable . That's kind of my current position. Mhm. [snorting] If I may summarize, we said earlier that AI can very well summarize information from religious texts, but less so in terms of providing social information or something like that. Um, but what's your take on the interpretation of such religious texts? I mean, it's not just black and white. Should AI be allowed to do that? [snorting] Uh yes, there is already a broad discussion in my data set about the fact that there are significant errors and that one should not ultimately exclude a scholar with training from this whole area. And there are considerable dangers when people use AI models, or the apps mentioned, as an instance for advice or decision-making, and then take a position in their environment . So, highly problematic. So we come back to the criticism that has now begun to emerge. And I think that very clear criteria must be established by the churches or religious associations regarding the limits of AI, the naive attributions currently made to this AI within the communities, and the resulting attitudes and actions. And so I see a highly problematic area for prevention in the future, because not only media literacy, but also AI literacy will be required in the future – what competencies one even ascribes to an AI, or not, right? That is, if I may briefly interject a question, do you think that learned religious communities should be involved in AI development? Uh, so there are now in the Islamic world, um, outside of Germany. I've already heard discussions from Qatar about developing their own AI. which follows Islamic principles, an Islamic ethic and civilizational, um, um principles, as you call them, although I don't know in detail what you mean by that . This is being discussed at Qatar University on a video channel. We'll publish something about the mode later, it's called Matik. A trend report will be available soon, this fall. Kaja will also take notes. Um, then there will be details about that. Um, but there are also very specific principles that are already established, right? so Islamic principles, because AI is currently primarily shaped by Western ideologies. Well, they would need to be decolonized, that's also an important point. Signs of this would be, for example, that the AI ​​allegedly advocates genocide up to now, and so, uh, um, well, the genocide, uh, in Gaza, excuse me, uh, not called as such, according to the, uh, statements of the actors there in the channel, and then they also cite further examples. Um, so um, these are very roughly guidelines for developing your own project; that's one example, right? And um, I think that Christian churches or Muslim institutions should also ask themselves the question: Is religious sovereignty preserved in AI? Is there transparency? about the procedures, the rules, the data, and the limits of what is possible. Um, also clarity about responsibility, and many AI services are anonymous, right? In terms of apps. Uh, that will be a large field to investigate. One wonders, who is behind this? Who compiled the data on Islam, and according to which criteria will such judgments then be made? Yes . And here perhaps, well, I might be a little pessimistic or realistic, I don't know. Um, and I think I would much rather support the point that we need to support media literacy and AI competence much more than supposedly building new AI, because there are so many models, open source models, that are fed with data and you don't know where it comes from, you don't know who is behind it. Then there's, for example, GAP AI, which you might know now. It's a rather right-wing conservative nationalist AI based in America, and it's based on Christian values. At the same time, until a few years ago, you could ask this AI questions about how to build bombs, where to attack synagogues, and how to become jute-free. So it is highly anti-Semitic, and nowadays it is no longer possible to ask these questions, and at the same time it is said, yes, but ask Dolphin AI, and then it is the new AI that is asked, and that is blocked again, and then there are models again that you can build yourself and that can be used. That's why I'm perhaps a little pessimistic in saying that I don't know if it's worthwhile to build even more AI or to create even more models that might only have these one things or data. So I think it's good to have this offer. This already exists, for example, AVP's awareness campaign. They have a chat room called Uncle Tarek. There are very specific dates for that. Um, you're welcome to take a look. This was recently launched, and um, there are already many tools that are based on a good data foundation, and we need to expand on that while simultaneously becoming aware that there are also models that don't have that at all, and that these can also become considerably larger . And if I may, I think it's a great point. It is, I mean, susceptible to misuse precisely because of this communication; it is technically objective and a machine that does this, and it is also politically instrumentalized, not least through this uh terrible app here, which was made for Berlin, the knife city of Berlin, uh, to supposedly attest to moderate knife violence in Berlin, and then of course it falls flat in this case because it doesn't work. But I mean, no, it can be used for things like that, precisely because of this framing. I [snorting] wanted to address another point, namely, uh, two things. Firstly, uh, you mentioned it's also about who can have a say in the AI, and so on and so forth. I have a fundamentally different model of how AI should be the future of AI, namely that it must function deliberatively. It must be shaped through the participation of various experts, interest groups, communities, and groups so that it is fair, sustainable, and certain. This is a democratic understanding, which I am trying to apply here, and so on and so forth, because we cannot avoid the problem that it is a generally used tool, which suggests that it can provide the answers, and if it reproduces purely technical, i.e., technical or scientific, interests of a particular company within its contexts and its own, then it does not correspond to my understanding. We have already begun to apply this logic to social media for the first time. We say this is our town hall, so to speak, and accordingly it must also withstand our regulations and political pressure, and we must also exert political pressure there, and that should be no different for AI, and Friedem brought up a point that I think is very important, which religion has institutionalized [ahem] and organized. That's a form of responsibility, isn't it? Well, at least from Islamic studies, I know that in Islamic legal texts , it is stated that the people who write them take responsibility for what they write. And anyone who has used AI and takes a close look knows what it says. Uh, Chat GPT May, right? Do errors or anything else happen, right? I just watched it again too, or maybe that's wrong , right? Or check or whatever , right? That's a completely different mode of action, isn't it? And in mainstream Islamic studies, it's the case that even a scholar has to correct himself and publicly announce his correction, saying to others, "Okay, I was wrong here, this lies in correcting my own knowledge." And, if you apply that, it should actually also apply to AI, but it doesn't, right? AI positions itself in the role of having this authority, writes you this information, but at the same time also says, well, actually you don't really know, do you? Uh, uh, check it out for yourselves. Actually, one should— and this is also taught in a good theological education— keep one's information limited, to say what one doesn't know, and to say, to stop there and not to over- interpret and prattle on if one does n't have the knowledge. And one must also move to say, where are my limits, where ca n't I, and these limits must be set. For that, you also need experts who also help determine this. Mhm. But not so omniscient. Yes, I think we have a question from the audience. The microphone is on its way. Hello. Yes, thank you very much. I have two questions. The time has passed a bit , but I've been thinking a lot about football as a substitute religion, and when you discussed AI so nicely – it could also be a kind of substitute religion – I thought about it and thought, hm, coming from my football expertise, I would say it lacks rituals, community, and celebration. So, what you Nada already said, social integration, which is also very differentiated, rituals, movement, physicality, that's somehow missing, and that's where my question to you guys got stuck, and then I have another one. But you will react, Kaja. Yes . Yes, you just mentioned rituals. In fact, one can ritually tell JBT that one wants this and that every morning at 8 a.m. If that's ritual enough, then you have a ritual here, or you can have memories set, in a positive affection way. Um, and then you can have rituals generated for you, so to speak. The question is always , when does something become a ritual, when does something become a community? And I believe that this is a good place to have a debate in the field of religious studies . Briefly, your second question. My second question, um, deals more with that. I, um, have conducted many interviews in the context of outreach social work online on social media, and one phenomenon that many in the field of extremism have encountered was the online conversion of young people through social media, and then, um, the questioning of Chat GPT followed, to see if everything had gone well and correctly. I have encountered it, at least as a phenomenon. Um, we went further and said, yes, there must be some kind of rules for this, but are you familiar with this phenomenon, where would you classify it today? I found it interesting that it first takes place actively on social media with an online imam and then, so to speak, transitions into AI and into this chat relationship. But actually, um, the online conversion of young people, a certain part of it is now also a very special moment that happens on social media and then it is further confirmed with Chat GPT, so to speak. Are you familiar with such phenomena? Where would you place this? I would ask you to answer briefly, because we only have, I think, less than 5 minutes left and there is a second question. Yes, that's an interesting example. Um, I have n't come across this combination before, but I think it's possible, right? So these online conversions are often highlighted as confirmation of faith. And, uh, that goes beyond my area of ​​expertise. Um, there are also plenty of Christfluencers who also try to proselytize and convert, right? So I think it's very possible that the community, the audience, the users who listen to it and participate, will then afterwards ask GPT chat questions about their attitude and the assignment of which competencies they assign to this tool. Now there was a second question. so that we can still accommodate them . Uh, Sorin Nada, gladly. My question is, isn't it a mistake that we're now trying to give AI human and divine qualities? We ca n't achieve that with AI. When I think about three years ago, we only used AI to write text somewhere, and we all laughed at what came out. Now, suddenly, AI is like a god, a religion, but how do we view this, and all the criticism we face? We have to consider these negative characteristics, but AI is not a god, it is not a religion, it is information. And if we only consider that, we can also say why limits have to be set, where the information is, or the scope, and so on. But we simply produce this thing, and I think there is also resistance, not only from us. There is also a bit of religion involved, yes, resistance to what corporations are doing, and so on. But I also think from another side of religion, which I think contradicts what I have said. Actually, from my perspective as a non-religious person, I will consider this tool as a means of fast and extensive data acquisition. Whether they are all accurate is another matter. But that's exactly what religion has been afraid of for 1000 years: that people don't get answers directly from their priest or their Rabbi or their, uh, Muslim Imam, and that 's a problem. Why are they like this? I'm going to make a very short final round now because, unfortunately, we really don't have any time. Um, if you like, a very brief spotlight on each person again. Uh, AI and religion. Does it work or not? [laughter] Uh, well, I would then tie that in directly and make it very short, namely, the idea is that if AI is a religion, then AI is the first secular god that we have created. And it is not uncommon for people to have anthropomorphisms, to attribute human characteristics to AI, but ultimately it is, so to speak, a model. Incidentally, it's not about gathering information, so to speak. The technical way in which AI, so to speak, works is only a prediction of what is probably meant and what the information to be produced is . It's also a probability model somewhere, and therefore all the more human, because it naturally follows certain rules . Therefore, AI and AI and religion work, uh, it must necessarily work. That's my perspective. There is a necessity for it to function, and that's how I would phrase it. Exactly. OK. I think I completely agree with that. I can only add that prevention and political education must play a major role in this . Mhm. Hey, do you have a really short ending, Player in two? K should be constantly deconstructed as a technological idol, and the framings built around it should be, uh, exposed, right? We've built up so much for an hour, and now we're going to deconstruct it all again. Thank you all so much for the interesting discussion. Thank you for listening, watching, and participating. I really enjoyed it . Thank you very much. [applause] [music] The Digital Salon is a production by Alex Berlin and the Alexander von Humboldt Institute for Internet and Society. A new episode is released every month as a podcast or for viewing on YouTube. โй [music]