Digitaler Salon: Made by Human

Yes, hello and welcome to the digital salon at the Alexander von Humboldt Institute for Digital Society . I'm incredibly happy that so many people are here today, especially in this heat outside. Absolutely fantastic. I would also be incredibly happy if we could have a conversation later, if you all join the discussion here. And of course I'm delighted to see everyone joining me here in the livestream with Alex Berlin. Welcome. I look forward to an interesting discussion. Um, and we'll start with a picture, [ahem] that some of you might recognize, we'll find out in a moment. For us, two years ago, this haunted the internet and became iconic relatively quickly. Not because it shows Jesus, but because millions of people have shared it, liked it, and interacted with it. It was an image of Jesus in water made of shrimp. I really want to know, does anyone recognize this picture? Maybe you just want to hold up your hand, who can relate? A few, then. Yes, good. Okay, so you're apparently in the same bubble as me, where you get pictures like this [laughter] flooded into your feed. Um, to everyone who did n't get that, congratulations. Uh, I think that's quite an achievement, to have a healthy algorithm without potentially encountering such images. Um, this image has developed an incredible symbolic power because it has simply become one of the most prominent examples of AI Slop . And there, LOP refers to AIs that are generated in large quantities and flooded the internet, and which primarily have one goal: to generate attention. And that's just one example. The NGO AI Forensics already determined in a study in 2025 that one in four of the top results examined on TikTok at that time had been created using AI . So, one can imagine that there has probably been even more since then, and that shows how relevant this topic is. Um, yes, today we want to discuss what AI Slop actually is and what it does to us, and we want to ask ourselves how we can recognize, now that AI Slop is everywhere on the internet, what we still find valuable, what is credible and creative on the internet, and whether we might even rediscover the value of human creativity? My name is Felix Leitmeier. I am a journalist and I am incredibly happy to be able to guide you through the evening here today. And we have three wonderful guests for that. Let's give a big round of applause for the three of us in the group. [applause] You have all brought very different perspectives with you today, and I would like to introduce you, starting with you, Rahne Rateisen. You are a research assistant at the Humboldt IP Institute and are pursuing your doctorate at the chair of Professor Katharina della Durontai in the field of data law. Mhm. Among other things, you'll be dealing with the question of how the law should regulate access to and use of AI models, and we'll be talking a lot about that today, um, what kind of rules there might be . Otherwise, there's also Markus Bch. here in the group. Hi Markus. Um, you research, among other things, the mechanisms of propaganda in the AI ​​age, with a particular focus on synthetic media and TikTok. We will talk about that a lot today. Um, and you just submitted your dissertation at the University of Münster. You teach at HAU Düsseldorf and TH Cologne, and you publish the newsletter Understanding TikTok. We want to hear more from you about TikTok today. Exactly. [snorting] Boris Eldxen is also still in the group . Boris, you are a photo and media artist and you have been working with generative AI for many years. In 2023, you sparked nothing less than a worldwide debate about AI images when you summarily declined the prestigious Sony World Photography Award . Uh, interesting story. The image at the time was AI- generated; you stated that as the reason . You also teach design at the Barcelona School of Arts and produce an AI podcast for the German photography magazine Fotorad. And that is the best prerequisite. Please give a round of applause to our wonderful guests who are here today. [applause] Okay , now a question for the group, for the three of you. When was the last time you saw something on the internet that made you think, my God, that wouldn't have existed without AI? Maybe you'd like to start from the beginning. Yes gladly. Um, today I heard a song that was obviously Kigneriert, which a friend sent me as a gift, which was a parody World Cup song. Um, what I found only worked because it was AI, because it was such an exaggeration of already pushing the limits of pop music, and I thought if you want to exaggerate that even further, then you actually need AI. Therefore, that was the last thing I saw. Exciting. What do you say, Markus? Then I'll stick with sounds. It's not exactly new anymore, but everything that has been created around the election Timmy Hope, from fan chants to visualizations with butterfly wings, seagull wings, and other images of this election soaring into freedom . Um, that certainly had a distinctly new quality. swam out and sang songs to the election . People had them before , but producing them on the internet in this quality is certainly completely new. Yes, I definitely had the Hope content in my feed as well. I believe that applies here to some, or perhaps someone would like to raise their hand. I find that quite interesting to see. So, this whole election story. I know it happened somehow , but it did bother me . Interesting. Woris, what was the last thing you saw that made you think, my God, that really wouldn't be possible on the internet without AI? I see it every day. Therefore, I don't want to pick anything out of it. Um , if you look at what percentage of the internet is already generated by AI, that's now the majority. This is also the new normal on social media platforms. And what's striking are always combinations that wouldn't have been possible before , because we were simply crossing genres or living beings. That wouldn't work. Yes, the new normal. Perhaps this leads to a question about AI Slop, our topic today. Markus, what exactly is AI Slop ? What is your definition of that? Oh God, oh God. There was a really nice article on Substack, I think the publication is called K Neeling Bus, because terms like that always have a career and Slop already has a career behind it. Uh, the article said: "Slop is basically everything we dislike on the internet these days." So, the latest neologism I've heard in this context, um, is Taste Slop. For example, Remov suitcases, as can certainly be confirmed, have a Taste Slop function. Um, that means this term, just like the term vibe, has already gone through its first hype cycle , because the German translation was always garbage or mud. Um, that was the general situation: as cheap as possible, as diverse as possible, yes, with its own aesthetic, uh, content that has generative AI content and can be scaled relatively cheaply. May I jump in here? Absolutely. That's where I would have asked my next question. But we are still affected by this garbage and mud. Perhaps you'd like to say something about this? I would also like to work on the definition , it's a bit of a fetish of mine and has always been defined differently. And the best definition I've heard was also published today. Annika Meer is sitting here, has done an interview with Mat Tryhurst, one of the AI pioneers, and Mat puts it into the right perspective. He says: "Slop is the media that attention networks promptly make humans." That means anything produced solely to boost my social media ranking via the algorithm is already slop. That means Slop actually existed before AI. AI can do many of these things better. That's why I think it's great to have a TikTok expert here, because we ca n't think about Slop and AI without social media. For me, social media is more like the new premium gasoline and the actual engine that lies at the center of social media. The driving force at the center is social media. In order. Um, but now we have AI slope content everywhere, especially on social media, and a large proportion of people are feeling affected by it . How do you explain that? Um, it simply makes no difference how something was produced. Over the past few months, I've taken a closer look at this for a keynote speech, and I 've seen four different levels where we are simply touched by things on the internet, regardless of whether it's art, advertising, or entertainment. Uh, first you need an emotional trigger, a kind of building effect or rage bait that drastically changes our emotional state , but it can also be very brief . And it doesn't matter at all whether it 's generated by an AI agent or by a human. The second stage would then be a resonance. As an example, I always use a song; I listen to a song. Lyrics are about a relationship. I think, hey, that describes my relationship and, uh, how can that be? That really sticks with me somehow . And what's interesting is that this very resonance is being perfected within advertising and entertainment. People want so-called Fractureless Optimization. That means I know exactly how everyone in the audience moves around the internet, and the AI ​​can personalize that. We can all do differently and better what already works. In art, on the other hand, something must be stuck in my throat. It must be something I didn't order . There has to be friction. In the next emotional level, I then try to build a relationship. I'm listening to this song. I think this person who did this knows me, feels me, is in the same situation. Then I start researching what the person did before, what they are doing now, and what they will do in the future. Firstly, authenticity is important here; I need to feel that it was made by a human, but all of it is simulated by AI. And at the final level, it's about how this song touched me so deeply that I see my relationship differently, live my life differently, and uh, for that, it doesn't matter who made the song or how they made it. So, now we come back to the big arc for Slop. That is to say, um, we are conditioned by the internet, by social media, to be emotionally triggered. Slop only has this task and AI can do it very effectively, and then it doesn't matter how it's done, because social media is simply also in certain areas, um, the algorithm doesn't care about quality. Yes, it can be as awful as it possibly is. We just need to stop. And in that respect, AI is the ideal instrument to satisfy precisely this need. Ideal instrument. The only question is, is it something new that is being created here? Um, I'd like to throw that out there as the next question . We've had clickbait, trash TV, and irrelevant content generated by AI for years now. One might say that actually not much has changed. Anne, what would you say to that? This is something new. Um, I'll say something about that in a moment. I would like to briefly say something about the AI ​​slot term, because I believe that when we engage in such a concept formation, which is what is happening right now , it is important to see what we are using the term for, a) how it is used and b) what we need it for. I think what you described is something we actually use the term "content" for. That which is seen, which is played back, uh, and which quickly grabs our attention. I think Slot has a judgmental aesthetic component, which is somehow associated with something derogatory, which is also what Taste Slop is about, and I believe what is new and represents a new quality, now I come to your question, is also this asymmetry of the cost distribution. So, it's very cheap to produce. It's very fast; you can implement an idea very quickly without major problems. Uh, and that results in both quantitative and qualitative differences, I would say, due to the scaling and the asymmetry of the cost distribution. Um, and at the same time, one has this supposed, um, competence. Well, that all looks good enough right away. It can be produced quickly, and sometimes it's relatively intense; it costs more to take it apart than to produce it, or the reception actually costs more than the production. I think what's new is actually just this sheer breed, you say, ultimately, right? And the cheaply produced mass-market stuff, but trash has always existed. Okay, so you're saying that a theoretical phenomenon is primarily something new due to its sheer volume, if I'm translating correctly. And because it's so cheap to produce. I believe that's the essential difference. I'd like to briefly chime in with some content. Um, everything is content. This is content. It will be available on the internet , but we still have a certain claim. Yes, we want to achieve something with this conversation. Eislop does n't have that; he's simply concerned with , uh, gaining reach. In most cases, there is no claim, no content. Uh, it simply sets the emotional trigger points, and yes, you're right, it's fast and it 's cheap. In most cases, you are therefore saying no claim. Um, will you go with that, Markus? Sorry, where are we in the discussion right now? What exactly is the question? The difference between content and slope. Does Slop lack any substantive claim, unlike Content? Well, that's a big question, isn't it? So, Slop can always have no claim. This is a discussion that I would divide into a matter of reception. Um, what's his first name, Jonathan Fransen, in the newsletter? He talked about AI, something that probably everyone here shares. Oh my God, AI is disgusting. Um, that qualifies you as knowledgeable, ah, that's AI, I wo n't touch it, while the lower classes get the slop . And I believe that this discussion, this division, will not get us anywhere . And I would say that slop is content, but content is not always slop. Okay, subset. Interesting definition. Furthermore, I don't think this discussion about content will get us anywhere . Yes, it is content. Point. So, whether or not this is content doesn't really help us in the discussion about Slop, if everything is content. OK. Whether A Slop is content or not, or new. Islop represents what Bernt Arnold calls the age of new images. Willim Fluss did this in 1985, entering the universe of technical images, and now they are truly here. New images that no longer represent anything in the real world, regardless of its technical nature, but something that no longer needs this visual representation at all. Because Jesus is depicted in a similar way, umbrellas too, but with E Stop there are also things that do n't exist in that way. Fruit Love Island, for example, does not exist. Italian Brain Rod does n't really exist in a remixed version, and I think that explains the Fruit Love Island phenomenon. You might have to tell them that sometime, because I don't think everyone in this room knows about it. So, uh, mine is like a dating show, and now there's also one with fruits that are AI-generated, and some of these fruits act in an extremely racist way, and other fruits eat their children . That, I believe, is the short definition of Fruit Love Island and girlfriends and boyfriends. And above all, the scripts are partly generated with AI, and everything is AI generated, right? So it's really, uh, incredibly successful, I have to say. Yes . And that's new, and I believe it's a paradigm shift, just like we had a paradigm shift from analog to digital photography, where people have no idea about the early history of film. These are paradigm shifts, and we're right in the middle of that now, trying to somehow get a handle on it, trying to somehow define the term, which of course is still evolving. What is considered " slop" now doesn't necessarily have to be "slop" in three weeks. I think that's exciting. So it's developing, it 's spreading incredibly fast . But why at all? That's something I'd still like to know . I'd look in your direction, Boris. Why is AI Slop performing so well? Yes, because that's how media works, and everyone knows that. Yeah, so I need a hook at the beginning, right? And at certain times, certain changes have to happen, and in the end, it's simply about getting a certain reach and making money that way . So, the content doesn't matter at all . So, the platforms are the reason. Yes, the platform dictates how our content should look if we want to be seen. And Slop has internalized this to such an extent that I believe, uh, that what distinguishes Slop is that the creators don't care at all what is actually being produced. I can already see that Markus is n't really on board with this, is he? No, because it does make a difference whether it's propagandistic content . um, AI-generated content that then says, okay, this group of people should be devalued, or whether it's Slop with Jesus AI. Well, one can certainly make substantive distinctions there . I agree that it's about affect and it's about reach, but we can look at both. First, create the visual representation, and then operationalize the whole thing, define what it means in this environment. But in terms of content, there are of course dramatic differences, whether these AI-generated fruits then carry out mass rape, which happens in the context of Love Island, or whether Jesus laughs at a witch who wants to saw him off the tree trunk and puts him in his place with the power of his thoughts . So I would argue that they are different things. Exciting fantasies. So, if these are concrete examples, these are not my fantasies, but these are concrete videos, these are real examples that I posted on Blue Sky. I heard that's how you're supposed to do it here with a platform. Does that change your dreams, Markus? I sleep very well. Thanks. But these examples also show that it's all about getting through the sheer endless masses on the internet, and then you need these strong stimuli. But whether one thing, another, or something else is done is completely irrelevant. Could this be a problem? I would now like to ask: are these platforms doing enough to differentiate what is actually being presented, or is it all just about grabbing attention? Um, well, I think there are significant differences in content, even if it's just entertainment and if it's not a targeted campaign or politically motivated, I believe there are differences in content, and I actually think it's a pretty good example because I think this " Jesus the Umbrella" stands in the tradition of absurd internet humor, a surprising combination of things that didn't exist before, and that's why it has more reach than other things. Therefore, I would say that there are also better and worse AI slopes. I'm not so sure whether the platforms are doing and should be doing that much, especially when it comes to AI-generated content, because I would strongly differentiate whether things are recognizable as AI-generated. And I think, I would define AI as being recognizably AI and also working with this AI aesthetic . I don't see much need for regulation there, because in the end the algorithm does what markets actually do, namely satisfy demand signals. People want to see that there are reinforcement loops; more and more people are seeing them. If people find it entertaining, then that's capitalism for you; if so, then you have a fundamental problem with capitalism, if you have a problem with the fact that people basically have bad taste in the end. I don't get that much of that kind of content delivered to me, but others do . I don't see a strong need for regulation there. In other areas, I see a strong need for regulation, especially when it is not possible to recognize that it is AI-generated. Okay, interesting. We can discuss possible solutions for this platform problem later, or even whether a solution is needed at all, whether regulation is necessary or not. I find that interesting, but you say it's a matter of taste, and one should leave a little leeway there at first. Yes, it also goes a bit in the direction of what Markus said about things, um, about the focus around, um, uh, what was that word "Ick"? You need to explain that again. You mentioned it almost as an aside earlier. You said there are some people who say they have an AI I and then react directly to this AI content with rejection. That gives exactly the I, that's disgusting. And an IG will eventually become an AI paranoia, but Jonathan Fransen explains that much better than I do. But there we are again with this division. I wouldn't intervene in that matter from a regulatory perspective. It 's just super interesting that, for example, the EUAI Act has relied on deepfakes for years now. M and Deep Fake is a word that has only really made a name for itself on Reddit since 2017 . It's included now, but it's obviously not the problem that's currently on our minds when we talk about Aislop. And things like Cheapes, if you stick to this definition, are cheaply made, um, things; Eislop would actually fall under that category. Um, they're not even mentioned in the EU AI Act, not by name. I find that interesting. And we'll need to talk a little more about the AC later . That is, um, if we have a big topic, as you all expressed a great need in our preliminary discussions. Um, I'd like to talk again about how all of this AI, Slop, is now a problem for our society. And perhaps, uh, Markus, you could start , since you've dealt a lot with propaganda and even gave a lecture on propaganda this morning. In what way is Rlob a societal problem? And then we need to hear from all of you right away , what the biggest problem is for you. Okay, this way of thinking is moving too fast for me again. Oh, there's something new. That's your problem. We need to regulate and prohibit it. Smartphones too, and children as well. Um, the point about propaganda is that throughout human history, propaganda has always used existing technological resources . Poster propaganda, radio propaganda, TV propaganda. Then, in the early days of the internet and social media, there was computational propaganda. And I believe that what we can observe now with AI can be described using terms like synthetic propaganda or post-digital propaganda, and it is being used accordingly. Is that surprising? No. Is that dangerous? Yes. Can we regulate that? Probably not. No. Uh, probably not. But we should at least be aware that we are having an informed discussion about what is happening right now. And now I'll return to this epochal shift, the age of new images. Um, propaganda is being conducted differently here than before, and uh, we still don't understand that well enough. I think that might sum it up . Okay, but the problem existed even before the technology. I'll take this from what you said. In any case, Jackie Lös, a propaganda researcher, said in the 1916s that propaganda is always total, but it only ever uses the available technical possibilities, and of course, AI is currently being used because we have just learned that it is cost-effective, totally scalable, affect- based, and a perfect tool. Boris, what is perhaps the biggest problem for you when we talk about AI Slop? We've just heard, with some caution, that these problems exist, that it might amplify vibrations that already exist, e.g., propaganda. What would you say is the biggest problem? I would like to break ground for creative people, yes, those who produce and create things for other reasons and also hope to reach an interested audience via the internet. And the great lie of AI is, once again, democratization. What is not democratic, of course, are the distribution channels and artists, who then also have to become social media specialists themselves in order to show their work there; they are becoming less and less visible due to AI slope, because it is automated, it is fast, it is a lot. Uh, how can one still be seen there? Ultimately, as an artist, you have no choice. I have to obey the platform's rules, and I will also have to obey the rules of digital marketing in the future , which is dominated by AI, because the AI ​​marketing of the future is real-time, personalized, uh, everyone will be addressed differently. Um, if I don't do that and don't want to, I'm invisible. And if I want to be involved in that, I also need my own agents to promote the whole thing on the platform. This is something that is talked about too little , because the more this slope exists and the bigger these mountains become, the more invisible all those who do not produce a slope become. Okay, so the slop is pushing the arts out of the internet because they simply don't promote engagement as well, for example. The question is, do you actually no longer have a choice about doing certain things on social media to gain reach? Yes, so how many young female painters have to film themselves painting while scantily clad for the algorithm to push them to the top? Is its content then seen because it's about the painting or the figure? This is a very clear platform logic, and it determines what more people see. Doesn't that also apply to all forms of content, including journalism and other forms of content that might now be pushed aside, you say? Right, that's a big problem with social media. What could be done about that? Perhaps force the platform to change its business model. Yes, away from targeting individual people and from advertising models towards a subscription model. Uh, then we as democracies would suddenly have fewer problems. Okay, Anne, I'm going to look at you now too. If that were allowed, the platform could simply say, hey, wait a minute, we need to promote artistic, substantive, journalistic content a little more. I do n't think it's a viable approach to simply ban this business model . So, a recommendation, especially one that was currently absolutely out of the question, because recommendation systems are considered legitimate mechanisms by the European legislator in the DSA and also in the AI ​​regulation, and it only becomes problematic when it involves dark patterns and manipulative technologies . But a well- functioning algorithm that basically shows what I, as a user , want to see is not a dark pattern in that sense. So, from that perspective, it's not realistic to ban social media . Um, and I would say that's just how it is when you consider that this is a market-based organization of cultural production, and of course, if you want to get out of that, you have to find other ways of financing cultural workers. I think these are important considerations to make, that one is not dependent on selling one's content, but that there are other financing options that are independent of the specific attention mechanism to which Berg is so bound, but perhaps other financing opportunities, funding opportunities. Um, I think those are things to consider; banning social media just doesn't seem sensible. Um, I'd like to chime in here. Therefore, the assumption that the algorithm shows what people want to see is naive. I've had a second X account for a year now due to a login error . I already had one and just had this second one lying around . I never followed anyone, I never liked anything, and in recent months I've only received meager content, so much crap and AfD stuff that has no basis in my behavior on the platform . This is extremely manipulative and suggests that the algorithm only shows us what we really want to see. Uh, I don't know where that comes from. And here, democracy is more important than the business model of American platforms. So you're saying that the algorithm does n't necessarily always want to give people what they want. Well, what simply works with the algorithm is rage bait, when people get upset about it. The AfD has mastered it perfectly, the other parties have always been too refined for it, and this is simply reflected in election results. But often I do have the feeling that I get exactly what I might want at that moment. It 's hot outside now and suddenly I'm getting church ice cream recipes. So I can imagine that the algorithm is largely based on that, and I don't know if that's just naive. Yes, how do you explain this example from X? Exactly . Anne, maybe you know Ja, because um, X has a different quality, but because targeted political influence is exerted there by Elon Musk. Um, and what I mean are well- functioning algorithms that depend on their user base not dwindling. Because if I were to see things all the time , if I were to see only poor content all the time, then I would simply stop using Instagram or TikTok . Um, and I, um, and Rage Bate is something that satisfies human needs, and that's why it's displayed. So I do believe that it's about satisfying human needs, which is being optimized. One might find it problematic that some base human needs are being satisfied , but it is also a bit paternalistic to say, yes, that it simply has to be forbidden. Markus, what do you have to say about that? Uh, we've now reached the depths of anecdotal evidence. I would like to take a step back because I found this point with the artist and Slop quite interesting. Firstly, there are already counter-strategies in the context of Dark Forest and, um, some US actors who are trying to build their own smaller structures or a left-wing YouTube media empire like Joshua Zitarella. What I'd like to know is, if this is considered a slop, and then I come across artists, what is the name for AI-generated content that isn't a slop? Does he have a name? I believe that was a question for Boris, if I interpret it correctly. Well, it's just AI-generated image, video, music, or whatever . It's simply AI-generated. But that would be like saying it's a picture painted with a brush. It's a picture, a video, or a sound. And what still distinguishes it now is the need for people to see transparency, how it was done and whether what I see is actually photographed, filmed or generated; I no longer see a problem there . That's transparency. And I find that aspect equally exciting because the German debate has flared up again, because various well-known people have generated texts with or by AI, which at the moment obviously still evokes a defensive reflex, while the vast majority of people in this country now do this as a normal work process every day. I find that really exciting because it shows that we are in a period of upheaval. This is new, and it's a bit like me going online onto the data highway, and this is where we are at the moment, this is the equivalent of the situation we are in right now, that we still have so much to talk about regarding Slop, because we have such a fascination with the age of new images . So you brought up this whole debate about various publicists, politicians, who have published AI-generated texts or supposedly AI-generated texts . Yes, that culminated in this Döpfner idiocy, where they publish an AI-generated text in a medium and say, " But this was generated with an AI." Um, what is trolling for beginners in? So, for anyone who might have heard, Matthias Stöffner, the CEO of Axel Springer, published a text that was really only based on a prompt , and if I remember correctly—you'll have to correct me if I'm wrong—but he wanted to make a statement, to say that this is now a counter-proposal; this text is explicitly AI- generated from top to bottom. Yes, it was essentially a response to the accusations from people who, uh, told these other journalists , "But you did this with AI, or had it done by AI." I think what's interesting is more like a painter with a brush, in that the AI ​​doesn't do anything alone , but that we always have to talk about teams of humans and machines. Because people who input something into an AI are writing their entire world experience into it. This can be seen in propaganda videos, namely when the videos are not made by native speakers, that sometimes the Germans demonstrate with German flags but cannot speak any German at all , but use some cryptic formulations. Um, and I think this aspect is still a bit under-explored in this whole AI topic at the moment. Okay, but I'm already hearing from the debate that at least we can agree that many people see the need for content to have a certain quality , perhaps even be created by humans; that this desire for humanity already exists, that's also a big issue. Berlin coffee, for example, meaning Berlin and coffee; it has to be a manufactory, we are afraid of industrialization. So, people are looking for quality, for things made by people, for things that are perhaps tailored to their needs. Um, this is now a question for all of you. I'm very curious about this and I know that you have different opinions on it. Is there currently a counter-movement to AI Slop that suggests a human comeback is possible? For example, um, could you mention the example of Apple releasing an advertisement for the new MacBook where suddenly many small items float around, a bit like in stop-motion videos or something, and IM showed how they do it. Or Coinbase did an ad where everyone had to dress up as if they were some kind of 3D generated rendering of people from the 90s or early 2000s. Um, and uh, it all had a bit of a DIY look, or a conscious, very natural look. What is your stance on this? Would you like to start? Anne, uh, would you say there's a kind of comeback towards humanity happening here right now, precisely because AI Slop is everywhere? Yes, now we're descending again into the depths of anecdotal [laughter] evidence, but I would say that I experience this in myself, in my environment and [snorting] also in texts about it from people who deal with this, that there is a new appreciation for mistakes, for imperfection. Um, I think that's been around for a while. I also believe that when you think about music, music that is basically totally generic and uniform has really gotten on your nerves, and AI is now putting its finger on the wound, and this uniformity, which becomes totally interchangeable, triggers a new appreciation for imperfections and for irritations, I think. At least that's how I perceive it. I also believe, um, that this is something that distinguishes art, that there are such rule violations, that it is not expected, and that the irritation is actually something that people totally appreciate. And I can imagine that this might create a new awareness of it on a broader scale, but maybe there is also, so maybe I am optimistic about that. You say art is what often has perfection. What are you basing that on? Yes, but I would like to answer your question from a different perspective . Three weeks ago I was invited to the congress of Swiss advertising agencies, and they also proclaimed AI Fatti, and uh, Ogilvi advised all their clients that if they have a high-priced advertising clip , they should make a making-of video for it, preferably shaky, a bit rough, so that everyone can see that it's not AI. This way they sell two films instead of one, which is very clever. Uh, that won't matter eventually. I could just imitate the shaky making-of video . I can also emit the faultiness. Since November of last year, scalers and various image generators have offered the option to restore an image to its real, "no make-up" state , so that it's no longer obvious that it's such a smooth, Journey-style image. And the same will be true for sound and text. And how many people have already said to Chat GPT, please don't use that dash now, right? Take him out. And then there are three different words for this dash. You have to list them all so that it doesn't happen again. So, these are always movements, right? And, uh, we have to accept that, uh, we will no longer be able to detect any differences in the future, starting from year X, simply by watching, listening, or reading . Everything can be faked. And so, coming back to the EU AI , I find their latest proposal not only very charming, but also very impactful. What do you say to that, Markus? So, you're saying, uh, Boris, you're more in that direction, right? This isn't really a comeback of humanity. That's more of an illusion or a trend. I gathered from Anne that you could embrace the idea that humanity truly exists again and that there is a desire for it. At least in the short term. What is your opinion on this, Markus? On a meta level, I would interpret this as us now feeling challenged again as humanity to reflect on, seek, and find our humanity. If we suddenly have cases where people marry their chatbot or use it as a therapist, and we realize that we're not so special after all, if the AI ​​text is better than the text we were ever able to put on paper. Um, we're looking for something, so it's a bit like , if you look back to the intertextuality of the 1960s, the idea that there is no new text at all, because every text is a mosaic of other texts, because they are there and we just put them together again, then something breaks away. So, in Ron Bart's case, the death of the author—well, he doesn't really die, but he's no longer the genius he once was. And I think we're at that point again now. But we want to be something special. But we also want to be better than any kind of entity. That's what I'm reading from it . So I interpret it as a great desire that this somehow awakens overall. We also need to discuss what kind of answers should follow. We've been hearing a lot about the AI Act lately, haven't we? Um, that's something we've all been talking about. The EU wants to stipulate, among other things, through this AI Act from August onwards, that, um, yes, it will be mandatory, at least in large parts of AI content, that there is a seal and that it states that it is now generated by AI, among other things; that is what is, uh, realistic. Perhaps Boris could make a start, since you create a lot of artwork. Could you always add "uh" to it, and do you perhaps even have to? uh, I've been doing it all along. So in the future, I won't need a displayed label anymore. It has to be written underneath. But uh, to briefly bring everyone up to speed regarding the EUI Act, at the beginning all the creative people said, but everything has to be labelled, AI, and then it was said , but the industry and here, and we only make deepfakes. Deepfakes are important. Now comes the irony of it all. Two weeks ago, the EU published a Code of Practice for the manufacturers of the models and suddenly redefined deepfake, opening it up completely. This means that anything that could be mistaken for being authentic, including all images that look photorealistic, videos, and actually also sounds and texts, must all be labelled. There are two labels: AI generated and AI manipulated. Nobody knows what that's supposed to be like . Yes, I'm curious to see if any further compromises will be made. It has n't caught on in Germany yet , nor in America. Big Tech will be furious; there will be a new trade war. Uh, but it's a radical idea. Yes, that means we in Europe will then really learn to differentiate. We will see it every day. We will become more sensitive to this. And I think that's the right way to go. I want to see what happens. I would also like to see Musk's first angry tweet on this subject. So you're also alluding to the fact that, among other things, or generally, voices from the USA keep coming up, even regarding the DSA, the Digital Services Act, which forces large platforms to deregister due to systemic risks, as it's called, that they view all of this critically, largely from the USA, perhaps also in the direction of Ahne. Um, are these regulatory options something that might actually work, and what will it change? So, to be honest, I wouldn't place too much hope in this transparency obligation, which I believe is in Article 50. The Code of Practice, as such, has no real binding force; it is subordinate to legislation. Um, no, I can chime in again later. Did you read the text two weeks ago? The Code of Practice? Yes, but that's not actually in the AI ​​itself; it only contains a presumption that it meets the requirements of Article 50. From what I've read, he clearly states that if you as a company do n't adopt our system, you have to show us that you have an equivalent that works. You also have to pay for a comparison studio between the two, and if you don't pay for that, uh, then the relevant interrogation authorities will come and have everything shown to them. A concrete threat . Yes, but the [ahem] Code of Practice does not have the force of law as such. That is, if you don't follow it, then you are not violating the AI Actomatic. It has no real binding effect as such. However, it lists what happens if you don't follow it. Yes . Yes, but that doesn't automatically mean it's in time, and it's also possible to meet the requirements of Article 50 outside of the Code of Practice . That's possible, that's just how it is, it's like a gentleman's agreement between the AI ​​Office and the AI companies that have also signed the agreement . Um, well, to be honest, we'll have to figure out the technical design later. I believe, therefore, that the AIO is relatively well staffed compared to other departments of the Commission . um, but of course, in terms of resources, um, they are significantly more so and in comparison to large KIU companies. Um, not really, so in the DSA we have n't yet seen that the regulatory power of the Commission, um, the enforcement authority of the DSA, has developed such a strong impact. Um, nobody really knows what systemic [ahem] risks are supposed to mean . There is also a transparency obligation in Article 35. Um, I'm skeptical about how much impact that will have. So I'm curious, um, but also curious about how strongly companies feel bound by this Code of Practice , especially if they do n't know exactly what it's supposed to mean . The commission doesn't know, and isn't even called, exactly what it's supposed to mean. Um, I know I'm here. I would say that's, well, I don't see that as such a big deal. Okay, so the deregulation ideas that exist now aren't such a hard blow, you're saying Boris? I got that impression from you too, didn't I? You do n't believe that this will be implemented immediately, do you? Um, there are a lot of things that are still unclear, right? And August 2nd is just around the corner. Uh, everyone is kind of curious to see how this is going to be implemented. Me too. However, I now consider the term " deepfake" to be extending this path to such an extent. Very exciting. That totally surprised me and, uh, I'm also curious, uh, yes, to what extent Europe will react if nothing happens from America. And I wish that, the Americans have no interest in Europe. They want to destroy the EU . This is competition. We are caught between China and America on one side. Russia is still there, too. We have very few friends outside of Europe, maybe in Canada. Yes. Um, I wish the EU would become a bit more consistent and combative. That, too, is of course a matter of interpretation. I would now like to open the floor for questions . We've already talked about so many things, and we've also noticed that there are already big differences in the rules and laws that are currently being drafted and are about to come into effect, depending on how you want to interpret them and perhaps also when you look at the effects. Um, are there any points in this room that directly relate to this? Yes, Mr. [Name] is here in front, a microphone is coming right here, so we can hear you too, namely in the live stream. I find the discussion, when it came to the question of taste and paternalism, whether the state should interfere, very interesting. Uh, one could probably also bring critical theory and, uh, the culture industry into it. The new thing, however, would be that through Demokratis, the AI ​​Slop, if we now understand the AI ​​Slop to mean poor quality and bad taste, is produced. And that 's something new for the culture industry, where it was industrially produced as an entertainment medium . Now, when we produce texts, like the one that Thuringian Minister-President Döpfner defended, because he wanted to write his own speech or have it written by AI and not by a speechwriter, not by a specialist, then we produce the slop that we have there, which we can then read in the redo or on the website. Ask the start, and we will continue to do this because we do n't have this sensitivity, whether it's with photography, music, or texts. And that's an interesting point that one might have to consider: that we all participate, without intending to, in producing this AI slop, and all of us practically extend the range of what can be said, seen, or heard downwards in terms of taste. So we all participate by the fact that our training data might be used, that we consume it. No, no, that's not what I mean at all . I mean that everyone who ca n't actually do it should either leave writing or composing alone, or at least take on the training until they can. Everyone who uses AI, with the app that exists for it , yes, that they then produce junk with it, without being able to recognize it themselves . In this experimentation, which is also what one does as an artist, so that's what it sounds like, well, Vog didn't just rubber-stamp it because he thought he'd get into trouble, but with a Holocaust memorial speech, yes, he thought, that's okay, sounds okay, and that will be the future, that we rubber- stamp the things that are offered to us, and that sounds okay to us, okay to those who don't have that sensitivity, but not to others. And so we will be able to differentiate – by "we" I don't mean those in the room, but I mean us as a society, or those who don't have this expertise. Bus, do you see that too? You do a lot with, uh, I mean, you do a lot of AI art, and you experiment a lot, and in the process, you might create material that you don't need. Um, so before this quote from Mat Tryhe, my definition of slope was the statistical mean and what is below it. And I believe that those who are now using AI, but who have no talent, no experience, no training, will not produce the garbage down here , but rather the AI ​​will raise everyone to the statistical average. You have to keep that in mind somehow. And now it gets interesting for the artistic field, because suddenly the dregs, the really, really bad stuff, are gone. And I'm waiting for the next generation of female art students to say, "We're going to occupy this now." And what might that look like? That's an interesting question for me, but I believe that those who don't possess that talent all end up at the statistical average. The AI ​​does that automatically. So Boris, you're already waiting for, um, deliberately created sloppy clothes to ultimately become a trend. Well, for me, if it's done consciously now and it now has, let's say, an expression of the human condition and it's not just about click rates and satisfying the algorithm, then it has a different quality. I would n't call it a slop then. It then plays with an aesthetic that has previously been associated with untalented professionals, and that can be redefined. Okay, back to the question: what is Loppen? What exactly isn't there? Were there any further questions or comments up front? Thank you very much. Yes, thank you very much. I wanted to ask, um, if you actually think that Slop is trash and trash is thrown away, or what exactly is the idea with these seals and such? It's actually very irrational to store all that trash and just put a sticker on it saying " trash". Um, and do you think it would change platforms if you could sort it out like an email spam filter ? Would that be possible? So, would that be technically possible, would it be legally possible, and how would it change the platform? Would such a filter be possible, perhaps Markus? Would that even be a good idea? I believe that "slop" is a transitional term. I don't think we'll be talking about Slop anymore in the near future . This is a transitional term we need because we are trying to somehow grasp what is being produced. But who defines when a slope is still a slope, or when it becomes art? Uh, we're talking about AI-generated images, and maybe we can all agree that most of us in this room have created something with AI before. Whether this is sloppy or not, others may judge, and this is the new normal. Whether I like it or not, or whether it gives people the idea to do something completely different, whether that's even possible, because there are people like Rod Meer who say it's impossible, because the whole industrial tool, by its very nature, leads directly to fascism. So, we cannot overcome this system from within the system itself. That 's a point for discussion. I don't think I see it that way, but um, would that be a good idea, given that there's now a democratized form of content creation that admittedly consumes a lot of resources? But why should we sort them out and throw them away, only to then celebrate the artists who paint something with a brush at night, and then we don't even like it afterwards? Any other points of connection to this? The person at the very front left is just coming in with the microphone. One moment. Uh, yeah, thanks. This is a really exciting discussion, but um, one aspect that hasn't been explored enough so far, I think it was hinted at a bit, um, is that the production conditions of, let's call it AI Slop, um, are only really possible in this form because, um, the training data largely consists of, for example, images or texts, um, where the people who wrote, painted, or photographed them weren't asked for their consent and weren't paid for their use . Um, and if I create an image or graphic with AI, I don't have to pay a graphic designer or photographer for it. So yes, this materialistic aspect has been somewhat neglected, and there is, for example, an article by Gareth Watkins and the New Socialist, which argues that this aspect is also a big part of the aesthetics of AI, and I would be interested to know what you think about that. Yes, Anne, maybe you were just addressed. Yes, I can answer that, because it's basically the biggest copyright debate on the whole topic in several years . And [snorting] one can of course take the position that this is the biggest theft of intellectual property that has taken place. And, uh, that's fundamentally unfair, that you basically take something, extract it, and build something from it that replaces what you built it on the basis of. From a copyright perspective, um, that's not so easy. Most people would say that what AI training has done or is doing is not covered by copyright protection, because just as I can read a book and then say , yes, I'll write a similar book that might be better, which I'll sell better, and then the old book won't be bought anymore. That's basically how the technology works, too: if you have a very cheap production method that can replace everything, you run the risk of human training data no longer being produced . Um, and then you run into a wall . Um, and I think that's why it's also important to think about how to establish other financing options for artists outside of traditional copyright law, which works in a more object-related way, so that they do n't lose their means of earning a living. Um, there's a discussion about Levi, for example, which is based on AI revenue, so to speak, that they skim off a bit of it and redistribute it. I believe this is the most politically sensible way forward. Then, of course, there are distributional questions. So who is supposed to get it then? What criteria are used to distribute it? Probably not based on what ended up in it as training data, but rather on a value-laden, normative distribution key. Um, who manages that [ahem] is another question, but I believe that's the political way to solve precisely this problem , that you basically have indirect substitution effects at the end that copyright as such cannot control . Did that answer the question ? Something like that . Okay, satisfied face. Thank you for the great question, and I've already seen some agreement here on Markus's page. There are definitely more questions. I think we still have time for one. Not currently, or any further input in any form. If not, I'd like to do one last quick lap to finish up . Let's ask ourselves, if we could somehow rebuild the internet, and um, we have some ideas about what we could do to make it run on the platforms according to our wishes, right? Um, what do you think could be improved quickly? The very simple questions at the end. I think that also provides a very simple answer. So, much like Groundhog Day, we simply relive the year 1994 over and over again with all this new beginning, er, the euphoria of new beginnings, and dark spaces on the internet where you feel understood and where nobody knows you're a dog, and then the next year is 1994 again, and it just stays that way. How do you look? Um, I can only repeat myself, I wish that social media platforms would be forced to move away from their advertising business plan to a subscription model. I see no alternative and I couldn't care less about the income of this platform . Democracy is more important to me. One point where An doesn't agree. You have heard before. Oh, yes, democracy is important to me too , but I think, um, I think I would like an internet with other users. I think that's my biggest problem. Yes, so we have n't found the one answer yet, have we? Various approaches attempt to address how to deal with AI Slop. One more thing, you can build these as agents, right? They'll talk to you there . However, it also runs out of power, so the half-life is not that long. But sometimes AI is also the answer to how to deal with AI, perhaps. So, platforms were discussed about legal issues, about, um, human creativity and its importance, and about how AI might sometimes remind us what humanity actually is. Interesting points. Um, first of all, thank you to my guests here on the panel and thank you for the great participation and for all of you being here. Like I said, despite the heat, it's gotten quite warm here too. Um, the official part is now drawing to a close, but I would like to ask you to stay here a long time longer and, uh, talk to our guests and, um, generally answer any questions about AI. Um, there's still plenty of room for that, and I've already heard that they'd like to stay and, um, be here for discussion. Um, exactly, that's why thank you so much. Um, the next digital salon is on the last Wednesday in July . Um, we would be happy if you came back again. And um, exactly, now we can move on to discussion. Um, we were n't really able to answer all the questions, but maybe one more quick answer to wrap things up. Did you still leave the discussion feeling satisfied? You've gotten closer. [laughter] That's good to hear. Thank you very much, and um, yes, have a safe journey home and enjoy the discussion. [applause] [applause] อ