AI should write books? We've had all that before and, above all, better

In addition to all the bad news from all the current crises, there's now this: As recently reported, British publisher Faber & Faber will be labeling the new book by author Sarah Hall with the "Human Written" sticker. The author is using it to protest the illegal use of copyrighted works by tech giants.

But the sticker has its pitfalls. In this paradoxical self-parody, it actually admits its futility: Even an organic label can no longer help. What regulatory authority wants and can inspect books or authors and certify them as AI-free organic businesses? And which authority oversees the oversight of an industry that has made fiction, or rather lies, its business model?

There's no way out of this AI-generated mess now. "But where there is danger," as Hölderlin would have said at this point, "there also grows / That which saves." It's approaching us in the form of AI itself. For there are a few good reasons why artificial intelligence shouldn't write novels or poems; and there are also a few reasons why it can do so without harm.

Today, those who should know better naively assure us that AI will be a long way from being able to write entire novels. And certainly not ones that are worth anything. All we can say is: Dear people, the world has been much more advanced in the past.

We've known powerful automated writing systems for centuries. By far the most beautiful was conceived by the British poet Laurence Sterne, who introduced it in 1765 in the eighth volume of his "Tristram Shandy": Of all the ways to begin a book, Sterne wrote, he considered his the best and most pious. "For I begin by writing the first sentence—and I trust in God Almighty for the second."

It is the original model of the automatic writing machine. The writer is the tool of a higher authority. This model was refined over time. Theodor Fontane dreamed of a pantograph so that his writing utensils would be moved by the world spirit, or at least by an external power. Others have resorted to alcohol or drugs to free writing from the shackles of the burdensome consciousness. The Surrealists invented "automatic writing," which led to highly dubious results. And the Frenchman Raymond Queneau came up with the glorious idea—as the reincarnation of Jesus at the miraculous multiplication of the loaves and fishes on the Sea of Galilee—of transforming ten sonnets into "Cent mille milliards de poèmes." He cut each sonnet into its individual verses, which could then be recombined into an almost infinite number of new poems.

But let's be honest: Except for Laurence Sterne's God Almighty, none of this really worked. The writing hand and consciousness remain chained together, and sometimes the former slows things down, sometimes the latter, but art emerges—and even then rarely enough—only from this friction. But how can one teach AI to wrestle with itself?

It won't be long before AI writes the perfect Martin Suter novel. And probably even before that, it will spit out the better remake in the never-ending series of Annie Ernaux's autofictional books . But who wants to read something like that? Another book by Annie Ernaux, another glimpse behind the scenes of a middle-class life? And who wants a flawless Martin Suter after we've made our peace with the fact that his books always have a little snag, a design flaw, or an unintentional comic touch? And now it's supposed to be slick?

A flawless Martin Suter would be a bad Martin Suter, and improving his books would be a violation of common decency. Here, Laurence Sterne's "Tristram Shandy" comes to our aid. Given the choice of exactly reproducing a picture and thus making it less impressive, he always chooses the lesser evil: "In that it seems even more forgivable to sin against truth than against beauty." Martin Suter would sign that immediately, only AI wouldn't know what he was talking about.

Because that's the first law of the realistic novel. It must invent the truth so that it is also beautiful. Casablanca, for example, shouldn't look like the AI depicts the city using Wikipedia knowledge in a novel. Casablanca has to look, smell, and be noisy exactly as we know it from the film of the same name. The average person only appreciates truth and beauty as clichés. A diligent AI would only spoil the fun here.

Heinrich von Kleist had also designed an automatic writing machine as early as 1811, a sort of godless version of Laurence Sterne's. Kleist's method is called "gradual formation of thoughts while speaking."

What Sterne considered God Almighty, Kleist's language itself is embedded in language. Language is the flywheel of thought and produces the unheard-of, the unprecedented. The machine capable of such a thing is yet to be built, as are the power plants and server farms capable of providing the necessary computing power.

Kleist's confidence, however, is boundless: "I believe that many a great orator, the moment he opened his mouth, didn't yet know what he would say. But the conviction that he would draw the necessary wealth of thought from the circumstances and the resulting emotional excitement made him bold enough to make a start, hoping for luck." In all honesty, one must admit: Man's advantage over machine is marginal here. It's an open secret that even great orators open their mouths and, all too often, don't know what they want to say when they finally close them again.

And unfortunately, God forbid, this also applies to books bearing the organic label "Humanly Written": The production of ideas while writing doesn't come naturally. Many an author is "bold enough to start with a good start, hoping for the best." But luck deserts them within the first few pages, and even by the end, they still don't know where the beginning could or should have led.

Books without an organic label have the not inconsiderable advantage of not having to read them. And since the market seems to be flooded with such books—Amazon can barely contain the onslaught on its self-publishing platform—there are more and more books you don't have to read. This also applies to many organic-label books, by the way. This frees up an enormous amount of time.

One could then, for example, reread Laurence Sterne and ponder how such books were incredibly successful, while nowadays one hardly has the patience, let alone the knowledge, to enjoy such storytelling with such delight. Thus, the flood of AI books would have the fantastic effect of driving us straight into the arms of 18th-century poets.

In the 1930s, Thomas Mann coined the term "novel industry," which didn't win him any friends and angered his brother Heinrich for a long time. What would he say today if he saw the books of Stephen King, Isabelle Allende, John Grisham, and the like? Doesn't anyone really feel sorry for these heroic authors who write their fingers to the bone in the service of their readers (and their bank accounts)? AI should take pity on them. It wouldn't hurt their bank balances. And finally, they'd have time to throw away the money they've accumulated with both hands.