ChatGPT and the future of the publishing industry

00:00:04

Christian Kallenberg: Welcome to the second Audience Development Deep Dive 2023 from Sprylab Technologies and We Like Mags. Our guest today is Professor Dr. Thomas Hess from Ludwig Maximilian University in Munich. He is the director of the Institute for Digital Management and New Media there. We're talking, of course, about ChatGPT, the automation and personalization of the publishing industry, and we're clarifying the question of how the industry will actually make money in the future. We, that's me, Christian Kallenberg and of course my co-host, Benjamin Kolb. Hello Benny.

00:00:35

Benjamin Kolb: Good morning Christian.

00:00:37

Christian Kallenberg: And here we go. Hello, Prof. Dr. Hess.

00:00:41

Prof. Dr. Thomas Hess: Hello.

00:00:42

Christian Kallenberg: It's on everyone's lips at the moment, ChatGPT. What do you think about this new technology? How much will it change the publishing industry?

00:00:49

Prof. Dr. Thomas Hess: I think ChatGPT in the version that is now being made available to the general public is first and foremost a marketing tool. A tool that can simply show the general public and interested professional users how texts or even coding can be simplified today with the help of trained AI applications. That, I think, is something that you have to note first. It's not technically new. The current version is 3.5. The 3 version already existed. The next version could be interesting. That is my general assessment. With regard to the publishers, I believe that ChatGPT in the freely available version again points out that it is already gradually going in the direction that content, especially texts can be created by the support of machines. In the specific case, however, I don't think ChatGPT will have much use with publishers because the copyright issue in particular is not really settled. I also suspect that the providers have deliberately left this open to show the general population rather than professionals. For media companies, the next version, which is also pending, ChatGPT 4 could be the interesting thing.

00:02:03

Christian Kallenberg: Have you actually tried this yourself?

00:02:05

Prof. Dr. Thomas Hess: Yes, I have, out of interest for the media industry, but also as a university lecturer. You can already see that interesting texts are coming in, and in the future the university will be asking itself "what about formats like seminar papers? Do we have to change something there?". I took a look at it out of both interests.

00:02:23

Christian Kallenberg: And, do you have to change anything?

00:02:25

Prof. Dr. Thomas Hess: If this works to some extent, it will be the case that in the medium term we will have to think about moving away from papers that simply summarize something that is done at the beginning of the bachelor's degree to papers that go further, not just summarize and prepare. Ultimately, we are moving from texts that students summarize to a dialogue. You can't do that with the machine.

00:02:50

Christian Kallenberg: Are there other types of automatic text generation you are working on at your institute?

00:02:55

Prof. Dr. Thomas Hess: Last year, in 2022, we held a major event with a number of media companies and technology providers, where we generally addressed the topic of AI in the media industry. There are several fields. One is the personalization of content, but one is also, ultimately this is what it's all about, the creation of content. There, too, you've seen that technology use has already come out of an infantile mode. For a long time, the typical example was low-class sporting events and the artificial intelligence writes a nice text about the county league match. Today this is also possible, but it has already gone further. Today, the content creator can already be comprehensively supported. As in many other areas of artificial intelligence, it is foreseeable that some tasks will be enriched, but they will certainly not be replaced by the machine. What the machine cannot do at all, for example, is ultimately generate opinions.

00:03:57

Benjamin Kolb: You've already mentioned that you're active in several fields, including personalization. We recently read through the study on publishing trends with KPMG and MVFP, which focused on personalization. It said that only 16 percent of the publishers surveyed currently use personalization. To what extent were you surprised by this result?

00:04:19

Prof. Dr. Thomas Hess: Yes, that did surprise me. I would have expected a lower number, i.e., a lower percentage. In the study that I helped to conduct, we asked relatively broad questions about personalization. In other words, not just the content that media people focus on, but also personalization of the customer approach, and interpreted very broadly, which we also did deliberately there, I think this 60 percent is understandable. It is not the content area, but as you have also seen, the areas that are in front, it is the customer approach, i.e. interaction on billing issues, etc. and not the actual provision of content.

00:05:00

Benjamin Kolb: Exactly, that was the most common example. Personalization in newsletter distribution, we still see a lot of room for improvement. Where do you see the greatest potential for publishers in the area of personalization?

00:05:10

Prof. Dr. Thomas Hess: I think it was probably the right thing to do, from a commercial point of view, to tackle areas like customer interaction, the quick wins. Where it gets more complicated, but which is certainly the core area of the publishers, is ultimately the compilation of content for the user. I think this has become clear again through the study, there are already two cases there. There is the simpler case, where ultimately a consumer or a user has a specific interest, for example, to be entertained. This is what we see with video or audio streaming platforms. So I get something suggested tonight, I'm tired from the day, I want an entertainment filler, then I get something. That's already integrated with the big technology providers, technology-based media providers. After all, it's not much different than if I'm ultimately looking for a consumer product, even if it's a closet. The second case is the more complicated one, and this is the one that is more interesting for media companies in terms of publishing, you can see it well in the example of news, the periodic provision of content. I look at an offer and ultimately want to be informed periodically. The question is what can and should be personalized, and publishers are already experimenting with this. There is the example of FAZ, which we also presented here in the study. They are quite interesting. They actually offer a normal page, but the personalized area as an add-on, i.e., as a supplement to the normal. The user gets what the editors define, which is also very important, and can add to the topics or view interesting content about individual editors.

00:06:54

Benjamin Kolb: With the FAZ case, it's also the case, you just mentioned it, that the editors do some manual personalization, just like the users themselves. Would you say that this is a transition to fully automated personalization or is this status quo also what you see for the future?

00:07:14

Prof. Dr. Thomas Hess: For the short and medium-term future, I can already see, especially with these periodic media, that the provider consciously sets topics. That is also one of the tasks of media companies, to set topics. That's what's expected of the FAZ, for example, and the Süddeutsche. I believe that the editorial teams will ultimately not allow themselves to be deprived of this sovereignty. They want to set topics based on an orientation or an evaluation of the news, and I believe that in the long term, in the periodical area, there will still be this setting by the editorial staff.

00:07:51

Christian Kallenberg: But what about the end customer? In the example with the FAZ, it was active personalization, i.e. the consumer selected his interests, so to speak. What do you think about automated personalization in this area, i.e., an AI that learns what topics the consumer is interested in and compiles the editorial content accordingly? Is that something that would be superior to active personalization in the medium term?

00:08:17

Prof. Dr. Thomas Hess: Yes, so technically, of course, it's possible, just as you described. Maybe one more step to the personalization. We have a typical case where I have a reader history and then the system or the algorithm considers which classes of messages he or she fits into and then selects the one that fits best. You can do that indirectly based on the data, then the user falls into it, or you can ask the user, but he or she doesn't like to answer. I do believe that this prioritization is there, but I think that two things speak a bit against it. One is the well-known effect that we also know from traditional and print media, when you flip through the newspaper, you come across something that doesn't correspond to your area of interest. So if you normally don't look at the feature section and you flip through and yet you find something, so that surprising thing would go away if you completely automate that. The second thing is, there's a long discussion about this, there could be bubbles. Maybe to the first one, there are, but this is not yet so well-developed, also algorithmic solutions under the keyword CERN-Dipity. They try to bring in this randomness of selection, so you look in your two areas, let's say you're interested in business and sports, then you get news or background reports, content on one area, but every now and then something is interspersed, which comes from the topic of finance, travel, etc. That's not technically advanced yet. That's not technically there yet. I think that's important to bring that surprise in. This is related to the second topic I just mentioned, namely that there has always been a fear that strong personalization will cause users to isolate themselves in individual areas, i.e., to form bubbles and to move around in these bubbles. I'm not a communications scientist, but research says that this is probably not the case and that similar bubbles existed in the past, only in a different way, but there is at least the feeling with personalization that people are reduced to individual - you can also call it something else - echo chambers. I can see both of those things happening. The discussion about this will certainly also shape the topic of personalization in the coming year.

00:10:33

Benjamin Kolb: The so-called filter bubble. Now, of course, the algorithms in the personalization of somewhat more complex recommender engines can bring in exactly this randomness again in order to break up exactly this pattern, both this surprise effect for the user, that he discovers something that interests him, and also not to let this filter bubble become too narrow. We see quite clearly that this is also possible with relatively simple means. Now, with our product, there are also two different approaches to compiling these topics. One is, of course, topic-based, in that the AI learns, so to speak, topic areas and the interests of the user and brings them together, or it observes the behavior of other users and makes available what we know from Amazon, such as "the users who bought this cabinet also looked at this chest of drawers", via certain cohort formation. How do you evaluate the two approaches in news? Is that something you can roughly compare and where you can see a clear favorite, or can you actually not compare that at all?

00:11:33

Prof. Dr. Thomas Hess: To mention the second point first, what's behind it, you know that too, that's collaborative filtering, i.e. the algorithm has the logic, you look at two users, you look at what they have in common and what the complementary area is and you think, "if they have a lot in common", you can determine that via statistical measures, "then I also recommend the other area". I think so. You can do that, but I don't think it's in the periodic area, so it's not in the news. I can imagine that when it comes to music, that's not periodic or also with videos, for entertainment or also for information, but when I look at the news area, my impression is that you have to go more to the higher categories. You can say there is such a subject area, you mentioned it or classic categories in the news area, I think it's good that you sort by that, also like to look at what this one has done, but that you couldn't break that down further. Example: If someone from the business sector is interested in a specific industry, for example aviation, then if you did that a couple of times, the system will say "this is someone who is interested in aviation", but that was maybe only a temporary interest from an investment reason or from the professional context. My impression is that the media companies would be better advised to be a little more general in the news area, in the periodical media area, to focus on general subject areas and not break them down so precisely, because you just get such different news in the case that it's hard to classify. I would like to touch on two other aspects that we haven't had yet. One, what you lose through personalization, of course, is what used to be called a bit of follow-up communication. Means that the next day in the context of friends' work and so on, you talk about the television event or the content event of last night. That's disappearing a little bit. A second aspect, which we can perhaps discuss in more detail later, is that public service providers, i.e. not private providers, have a somewhat different target function. They don't want to maximize anything, they want to create relevance. That makes personalization all the more difficult.

00:13:49

Benjamin Kolb: You just meant the campfire effect, which was also frequently cited in the course of the World Cup or now with the jungle camp just on RTL, so as a social bonding agent, so to speak, and you're afraid that this will now fall away? Did I understand that correctly?

00:14:05

Prof. Dr. Thomas Hess: Exactly, that is the campfire effect. That's the right image for it. That people get something that first corresponds to their interests, that they simply break down into these smaller segments.

00:14:18

Christian Kallenberg: We often have people on as interview guests who are actively shaping the publishing industry. You are, if I may say so, above all an observer, so I can ask you this, who is doing particularly well at the moment in terms of automated personalization, but also content generation, etc.?

00:14:36

Prof. Dr. Thomas Hess: I would be happy to name someone, but I don't see the perfect solution yet. I think the topic is starting, that's why we took it up last year, we didn't have the case of the FAZ in it by chance, not because it's so sophisticated what they do, but because it reflects quite well where the editorial office has a strong significance in the news area, where it's a bit personalized, where the user and also the editorial office were strongly introduced. I therefore find this approach interesting. The second thing that has to be said is that this is of course a problem for smaller providers in terms of competencies. You have to prepare the data, you have to implement it, even if it's a simple process. You also have to get the people who know how to do it, even if you pay them, you have to get them first. That's something that larger providers have to practice a bit first. That's why I always distinguished between periodic and non-periodic. Especially with periodical goods, the specifics of the media industry have to be taken into account to a much greater extent. Maybe one more point that just comes to mind. In personalization, there are other areas of life than content in the media, and there, in addition to the user's profile, which is shaped by his or her history, the context of use is also becoming very important. Transferred to the media industry, this means that I receive personalization, for example in the music sector, which is based on my history, what I have otherwise enjoyed listening to or which genres. It also includes whether I do it when I'm on the road or when I'm at home, for example, and also what time of day it is, for example. More recent developments are moving towards placing even greater emphasis on context, which is the umbrella term for all these factors. Ultimately, all of this is a matter for the future. I remember that many years ago we tried out a project like this with a large telecommunications company, where we also tried to recognize the user's emotion, i.e., whether he or she is in a positive or negative mood, and to control the choice of music accordingly. At least at that time, the customer didn't want to have it.

00:16:42

Christian Kallenberg: And for what reason did he not want it? What was the realization?

00:16:44

Prof. Dr. Thomas Hess: People don't like this derivation from the face. The system recognizes, sounds a bit like Big Brother, what mood you're in. So they said you could also do it via heart rate. They didn't want that at all. Rationally speaking, the user weighs it up and says "what do I get out of personalization?" and "what's wrong with it?". The perceived negative aspect of a machine interpreting me was greater than a bit of customization.

00:17:18

Benjamin Kolb: We have another topic that we wanted to discuss with you, and that is that the publishing industry is under pressure from declining print revenues and is fighting against that with increasing digital revenues, with whatever business model. In the beginning, a lot of emphasis was placed on advertising. Advertising then came under a bit of pressure. The new model is paywalls. Almost every publisher, at least in Germany, is trying out paywalls, some more and some less successfully. What is the state of your research in this area? Are there any special findings that can be derived for paywalls, let's say in a more dedicated way, there are news publishers, there are magazine publishers that have different binding strengths, are there any derivations that can already be seen?

00:18:06

Prof. Dr. Thomas Hess: That is one of the basic problems of information content providers. The background is also clear: in the past, in the context of print, people were more willing to pay. There was the free wave, and now the publishers have been trying for 15 years rather than ten to get away from this free logic and to put a price on what is offered digitally. The background is also clear, the idea that you realize everything through advertising revenues doesn't work. I see two things. The first, today positioning content behind the paywall is still a niche topic and the revenues that are realized there are homeopathic. They're growing, but they're really small and homeopathic. Now I wouldn't expect that to go up in the average case, now not the New York Times, of a news provider. It's a trial. The second comment, you lose a little bit of journalistic relevance. When I put something behind the paywall, only a small portion of users ever see that first. Many companies in the news sector, less so in the entertainment sector, define themselves by the fact that they shape opinion. As a media entrepreneur, you have to consider that if I put it behind the paywall, I might motivate my user, who otherwise pays nothing, to ultimately pay, for example, to take out a monthly subscription, but I lose a bit of relevance. Of course, what I also lose if I put a lot behind the paywall, I lose advertising revenue. It's already a more complex optimization problem, what content you put behind it. I don't think it's the salvific solution that you might find where you look at the systems, which are quite convincingly done, but it could be relevant in the medium to long term, but it's already a longer way. I had just given an example of a major American news provider for a reason. They gain something, of course, because they are visible worldwide through the language and the brand. Whether now, let's take the counter-example, a regional news provider at the end of the day, when he adds up everything and sees the investments, whether he will be better off economically and journalistically in the end, I am at least cautious about that.

00:20:31

Christian Kallenberg: Where do you see the future of the publishing industry if it's not the paywall and if advertising revenues, as we just said, are also declining? How will the industry earn money in the future, or will the industry no longer exist in the long term? Will the public broadcasters do all that?

00:20:47

Prof. Dr. Thomas Hess: I believe the latter less. There will certainly be a concentration, as you can already see. Larger publishing groups will emerge, especially in view of the technological investments and the platforms that have to be built. That would be my first comment on that.

00:21:03

Christian Kallenberg: Benjamin Kolb is very pleased about that (laughs).

00:21:08

Prof. Dr. Thomas Hess: The second thing is, of course, that publishers have always tried to diversify in a business sense, i.e. to get away from the classic "I provide content" business. That can now be traditional, I organize trips for my readers. It can also go a bit further. The Burda Group, for example, has often been criticized for diversifying further and saying "you can transport everything with this medium" and you can also diversify information and products and be strong in e-commerce. I believe that ultimately publishers will have to build something on this path in addition to their classic content business. There are also really interesting examples beyond the daily newspaper of who has done this very well. Now in the broader media sense, the Ravensburger Group, which is also known for children's games, etc., they have ultimately also built a hardware business, so for small children, when they buy a pen and they click on it, then the cow moos. That's something that is of course a wonderful diversification. I believe that, in addition to concentration, the development of additional business areas is a strategic option that you simply have to keep in mind, even if you might not like it that way.

00:22:23

Christian Kallenberg: So the challenges in the industry remain great. What about you? Did you anticipate this development back in 1997, or why did you turn your back on the industry again after your time as assistant to the Bertelsmann Executive Board?

00:22:37

Prof. Dr. Thomas Hess: I would say it the other way around, I was at Bertelsmann for two years, which was exactly when the digital business was being built up. At the same time, I still had in the back of my mind that maybe I should go into the scientific field and I took that with me. I took that to the chair in Munich until now and tried to look at it from a different perspective. That's why the media industry has always been interesting, because it's creatively and socially important, but also because the media industry is ultimately at the spearhead of digitization. After all, the biggest first example was Napster, the file-sharing platform for music, which has always fascinated me. I think that if you look at the media industry and what it's doing alongside other industries, you can see a lot of issues that many other industries will have to wait until later.

00:23:24

Benjamin Kolb: Christian said earlier that you're just an observer of the media industry, but that's not quite true, because you're also co-editor of the quarterly magazine "Medienwirtschaft. What exactly is your job there?

00:23:36

Prof. Dr. Thomas Hess: "Medienwirtschaft" is a periodical that is made available in print and online, where we ultimately want to reach the practitioner who wants to read scientifically solid, but still relevant results. We summarize studies there that have been done in classical science, for example, also on the topic of personalization, and prepare that every quarter and do a kind of quality control, a kind of gatekeeper function, and that's how it came into being back then as a pure print medium 10-15 years ago. We are a group of five editors, and due to my disciplinary background, I have the focus on companies and on technology development. In addition, we have colleagues who come, for example, from journalism, from the legal field or from the field of communications science, and so we want to continue to accompany the development of media entrepreneurs, in particular digitization, from different perspectives and provide practical input. The journal can also be accessed online via the Beck offering. We also organize, for example, the Munich Media Day, where we initiate discussions on current topics, together with Bayrischer Rundfunk. We also try to bring a bit of diversification, although of course we don't intend to make a profit, ultimately there.

00:25:06

Benjamin Kolb: But even without a profit motive, what is your digital strategy for the magazine?

00:25:11

Prof. Dr. Thomas Hess: Ultimately, the strategy means that we are glad - we are a small medium, after all - that we ultimately have the opportunity to make our content available via a third party, via the Beck Group, through our publishing house in Hamburg. We have a classic hybrid strategy. There is still the periodical, which continues to be read, but we also offer the content online and the specific in the academic area, which is also aimed at students in the internship, that such content is fed into larger packages. So University X, University Y does not subscribe to individual journals or content from journals, but packages from publishers. We are part of a package and therefore we are glad that we can reach the second target group, the students. Those are our two main stakes, i.e., to address both channels, the second via packages. We don't do other things that you could also imagine with magazines. There is, for example, an American magazine, a practitioner's magazine, which then makes a film from the article. We would perhaps also like to do that with statements, but we simply don't have the resources.

00:26:26

Benjamin Kolb: What do you actually read in your private life? What media offerings do you use?

00:26:29

Prof. Dr. Thomas Hess: For the information sector in Munich, I make extensive use of BR24. That's a public service that's available on the radio, but also via the app and on TV. I don't use television as much, but when I'm driving or otherwise have time, I use it for up-to-date information. Then one or two digital offerings from the FAZ and the süddeutsche Zeitung. What I still hold in my hand, in the information area, is a business periodical that comes every week, Wirtschaftswoche. In the entertainment area, I'm probably relatively typical, so I use the two big streaming providers and a few supplementary media libraries of the public broadcasters.

00:27:12

Christian Kallenberg: You continue to support the industry privately, I think that's great. Thank you very much for your time today, and I'm curious to hear how our listeners respond.

00:27:19

Prof. Dr. Thomas Hess: Thank you very much for your time.

00:27:21

Benjamin Kolb: Thank you very much.

00:27:23

Prof. Dr. Thomas Hess: Thank you for the interview.

00:27:24

Christian Kallenberg: And if you, dear listeners, enjoyed this episode, I would be delighted if you subscribe to us on the podcast platform of your choice.

‍