Ever since artificial intelligence became the talk of the town with ChatGPT, AI has been finding its way into countless applications and areas of life \u2014 publishing included, of course. A new development is, however, quite shocking. Not necessarily in its present form, but in terms of the perspective it opens up for the book market. I also keep getting the impression that many people in the literary world have very little idea what\u2019s heading their way \u2014 both with artificial intelligence and with demographic changes in the German-speaking world. In this post, I want to jot down a few thoughts on that.<\/p>\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
The company Media Control has announced that it will make its latest AI-based solution, DemandSens<\/a>, available to paying publishers starting next year. For any book listed with an ISBN number, the application is supposed to predict how successful that book will be. The hit rate is put at 85% \u2014 and up to 99% in certain segments. Five billion data records are said to be analyzed in roughly one second and used to generate a forecast. So, before a book even hits the market, a prediction can be made of how many copies it will sell. And Media Control can draw on very high-quality data: in addition to daily book sales figures, they also use data from TikTok and other metadata, for example.<\/p>\r\n\r\n\r\n\r\n If you look around a bit, surprisingly little is being written about this software. The S\u00fcddeutsche Zeitung calls it an \u201catomic bomb\u201d<\/a>, the B\u00f6rsenblatt has reported on it<\/a>, and the NZZ has written about it<\/a> as well. As so often, the Swiss approach the topic critically and with realism. Some statements in the NZZ piece shocked me. For one thing, they confirm a development I already discussed in an earlier post<\/a>: fewer people are reading books \u2014 though those who do seem to be reading more. The readership is shrinking, and with it diversity; this is reflected in the total number of titles published and in a 10% drop in new translations.<\/p>\r\n\r\n\r\n\r\n \u201c[In 2021] 273 million books were sold in Germany \u2014 but they were spread across only one million titles.\u201d<\/p>\r\nnzz.ch<\/a><\/cite><\/blockquote>\r\n\r\n\r\n\r\n So there\u2019s a very clear focus on a small number of titles (if you can call a million \u201csmall,\u201d though the ratio is interesting). And when I look at the Spiegel bestseller list, I get a bad feeling. Now, enter an AI that helps publishers preselect the most lucrative books. If I consider the figures for the German book market<\/a>, it\u2019s obvious that publishers will be forced to pull out all the stops. A publishing house is a business that must be profitable. In the current economic climate \u2014 with the massive scaling back of the German economy, significantly higher inflation, and a wage\u2013price spiral \u2014 only those publishers that can secure their profits will survive. Right now, TikTok is being treated as the magic bullet. Ten years ago it was book blogs, then Instagram, then BookTubers, and now BookTokers. But let\u2019s be honest: the readership isn\u2019t growing \u2014 it\u2019s shrinking, by 33% in Germany over the past ten years. An AI tool that predicts a book\u2019s chances of success can massively increase a publisher\u2019s efficiency.<\/p>\r\n\r\n\r\n\r\n If you read more closely about what DemandSens actually offers, the functionality is limited. Ulrike Altig, managing director at Media Control, downplays it in an interview with Deutschlandfunk<\/a>. Books have to be listed already \u2014 i.e., have an ISBN. So it\u2019s merely a helper tool for estimating print runs, not an automated scan of a manuscript or proposal that offers a prediction before publication. No big deal, in other words. Ulrike von Stenglin, managing director at Gutkind Verlag, takes a similar view in an interview with SWR<\/a>. And that reaction is exactly what I\u2019ve come to expect from the literary sector \u2014 it often reminds me of civil servants just ambling along unperturbed.<\/p>\r\n\r\n\r\n\r\n If I were a publisher or an editor, I\u2019d be moderately panicked. Why? Media Control isn\u2019t one of the big players, and I suspect getting the tool to its current state was already a heavy lift. A company like Amazon is operating at an entirely different scale. Amazon has digitized copies of essentially all books, it has sales data, and it knows the interconnections between books \u2014 i.e., it knows what else consumers read. With Kindle and Audible, it even knows what gets read within books: where readers drop off, how fast they read certain passages \u2014 perfect training material for an AI that evaluates consumer-friendly content. With Goodreads, they have additional stats from heavy readers. Amazon has AWS \u2014 massive infrastructure and affordable compute. And Amazon already applies AI as daily bread. To me, it\u2019s only a matter of time before this knowledge is fused and made accessible via AI. Feeding a manuscript into an AI that has ingested nearly all previously published books, knows the sales figures, and contains vast amounts of generalized knowledge \u2014 and that can judge within seconds whether a book will sell \u2014 isn\u2019t far-fetched. And that\u2019s what it\u2019s about: will a book sell or not. Nothing else. Publishers and companies like Amazon don\u2019t have a cultural or educational mandate. It\u2019s about profit, full stop.<\/p>\r\n\r\n\r\n\r\n People then point to classics that were innovative and became surprise successes \u2014 and cultural enrichments \u2014 with Ulysses<\/em> often the first example named. Literature is art; you can\u2019t just squeeze it into an algorithm. It\u2019s about people, zeitgeist, the processing of political, emotional, cultural \u2014 indeed, all spheres of life. That\u2019s true, but at the end of the day, economic success is what counts. A bankrupt publisher can\u2019t publish new titles. And let\u2019s not kid ourselves: most books sold aren\u2019t \u201cgreat literature.\u201d The bread-and-butter business \u2014 alongside professional and school books \u2014 is entertainment. I already trust an AI to write the hundredth Sarah Lark pastiche. Or the thousandth young-adult novel. Or the umpteenth melodrama where a young woman finds letters or diaries in her grandmother\u2019s attic, uncovers Granny\u2019s love story, and, naturally, falls in love herself. This theme has been kneaded a thousand times in countless settings. Even if an AI doesn\u2019t write such a book itself, I readily believe it can provide a realistic assessment of a manuscript\u2019s chances.<\/p>\r\n\r\n\r\n\r\n \u201cAnother snapshot from the industry that may require more analysis than anything else: more and more young people don\u2019t get into books at all \u2018because they can\u2019t read with comprehension,\u2019 as B\u00f6rsenverein head Karin Schmidt-Friderichs puts it. In the face of this educational catastrophe, debates about the dangers of artificial intelligence seem petty.\u201d<\/p>\r\nnzz.ch<\/a><\/cite><\/blockquote>\r\n\r\n\r\n\r\n The bulk of consumed books simply aren\u2019t deep novels or classics. And taking that NZZ quote at face value, it\u2019s hard to deny. The readership is changing. Books compete with the entertainment value of the latest Netflix series and must therefore be pleasant and easy to consume. Standards will continue to drop \u2014 if only because children\u2019s reading competence is declining sharply. Whereas in 2001 every sixth fourth-grader had reading difficulties, by 2021 it was already every fourth child<\/a>. Add to that the failed integration of migrants in Germany. Many children who have migrated, of course, do not receive sufficient reading support<\/a>. For publishers, the market will thus become even more difficult, because fewer customers are moving up the pipeline. Further price explosions due to inflation, a shortage of skilled workers, and, on top of that, a shrinking market \u2014 we\u2019re nowhere near the bottom yet.<\/p>\r\n\r\n\r\n\r\n So leaning back, as Ulrike von Stenglin does, and playing down the current solution strikes me as a big mistake. And you can already see publishers running into trouble. Suhrkamp, for example, faced serious financial difficulties before it got a new owner<\/a>. Whether that will solve the publisher\u2019s structural problems is questionable. I have to be honest: as a heavy reader, Suhrkamp is completely uninteresting to me. Sure, they have many classics \u2014 I have a few Hesse titles on my shelf. But let\u2019s be honest: innovative, attractive, or exciting is something else. The design is simply atrocious; the books have looked the same for a hundred years; there are no noteworthy new releases. The Insel Verlag, also under the umbrella, is among the better ones \u2014 but those books have always been plain ugly too. Compared with Manesse \u2014 likewise a traditional house with many classics \u2014 the difference is stark. In their list and in the redesigns of famous world classics, you can see at least an attempt to adapt to the market. Klett-Cotta, too, is trying to reach new audiences with colored edges and attractive book design (hello TikTok). They publish old books in new clothes \u2014 sometimes even as paperbacks \u2014 which shows how undemanding the new readership is: cheap production with a pretty colored edge. Let\u2019s not pretend this is a newfound love of bibliophilia. Still, I\u2019m glad publishers are trying new approaches and meeting these readers where they are. Or take the B\u00fcchergilde<\/a>. Aside from a handful of beauties, most of their designs are utterly subpar; I genuinely wonder who they aim to reach. If they carry on like this, they\u2019ll lose their customer base simply because the older generation is dying out; their program, with its often stodgy, greasy look, certainly won\u2019t appeal to a young audience.<\/p>\r\n\r\n\r\n\r\n In such a strained environment, I\u2019d be very interested in any new solution as a publisher. AI in particular offers massive optimization potential. With increased efficiency and the financial leeway it creates, publishers can again cross-subsidize the so-called \u201chigh literature.\u201d<\/p>\r\n\r\n\r\n\r\n At the same time, there\u2019s a major risk that quality will go down the drain if production is focused solely on the mass market. If AI makes it possible to churn out high-selling books with little effort, the market will take over \u2014 and the publishers that know how to exploit that will be the ones that survive.<\/p>\r\n\r\n\r\n\r\n A useful comparison is the video game industry. There are lavishly crafted, beautiful games \u2014 I\u2019m thinking of titles like Horizon Zero Dawn or Cyberpunk 2077 \u2014 with strong storytelling, wonderful graphics, and extremely high-quality music. And then there are games that lean on in-game purchases \u2014 cheap mobile time-killers for the undemanding, mini-games to kill time. \u201cIn 2023, in-game purchases (microtransactions, virtual add-ons, etc.) accounted for around 71 percent of total revenue in the German video game market (excluding hardware).\u201d (source: statista.com<\/a>). From a business perspective, it\u2019s obviously much easier to push out the thousandth rehash that brings in quick money with little effort than to invest five years and a nine-figure budget to create cultural goods. Risk management. A publisher will be more inclined to release a completely undemanding mass-market book if they know there\u2019s an extremely high probability of profit than to publish a truly novel work with great risk and likely losses. If an AI offers a decisive pointer in more than 80% of cases, publishers will trust it \u2014 they must, because those are the rules of the market; otherwise, they\u2019re at a competitive disadvantage. The SZ headline \u2014 calling it an \u201catomic bomb\u201d \u2014 seems spot on to me. DemandSens isn\u2019t that yet, but the trajectory is. Am I exaggerating? Maybe.<\/p>\r\n\r\n\r\n\r\n As always, the truth lies somewhere in the middle. The world won\u2019t end (it never does), but it also won\u2019t develop optimally (it never has \u2014 that much is evident). That\u2019s my take on AI and the publishing landscape, and I think the sluggish, academic literary sector needs to start moving if it doesn\u2019t want to end up out on the street tomorrow, scraping putty from the window frames. But hey, what do I know? I\u2019m just a voracious reader \u2014 and the inexhaustible well of wonderful books is already mine. I\u2019m perfectly relaxed.<\/p>\r\n\r\n\r\n\r\n What do you think? Will AI shake up the book market? Or will nothing change? Do you work in publishing? Are you panicking? Or am I exaggerating again and everything\u2019s fine? What\u2019s your take?<\/strong><\/p>\r\n\r\n","protected":false},"excerpt":{"rendered":" Ever since artificial intelligence became the talk of the town with ChatGPT, AI has been finding its way into countless applications and areas of life \u2014 publishing included, of course. A new development is, however, quite shocking. Not necessarily in its present form, but in terms of the perspective it opens up for the book … <\/p>\n\r\n
\r\n