The cheeky bastards at Open AI presented their new Chat GPT-4 model on Pi Day (14th of March, or 3/14) on youtube. Reading through the comments, I had the impression they were either strongly curated already, or people just did not understand the implications for art.

I suspect most of them simply do not care.

On this side of the discourse pond, instead, I am strongly concerned about the implications it will have for our work as writers and also as readers.

1. Generative text AI will saturate the market further. Genre fiction, especially romance, holds the lion’s share. As genre fiction is the one most prone to the usage of tropes, recursive situations and familiar- repetitive characters, it’s also the area most vulnerable to generative fiction.
   
   
   
2. This will not make human authors extinct – but it will cause us to compete with texts mainly written by AI and edited by humans, which will be much better at understanding the market and thus will receive more favour from the public. In short, we are looking at a sizeable reduction of market, public and income in the next few years.
   
3. The line “but humans want content produced by other humans” is a fallacy in two ways: a) it assigns some kind of intrinsic value to the final product that the reader is able to discern, b) it assumes people will be able to discern AI text from human text. Given the breakneck pace of improvement, it’s likely that AI text will be undistinguishable from human text at least at the micro level (three or four paragraphs). Which brings us to…
   
4. An AI writing an entire novel by itself is still highly unprobable. This both due to intrinsic limitations of deep learning models (as the complexity increases, the computational progress requires increases at least to the fourth power) and the fact that a a wrongly-put line can destroy and entire novel.
   
   ”Then he woke up and it was all a dream”.
   
   See what I mean? (Quantum computing may prove to solve this problem, but I don’t know nearly enough about it to make even a shred of prediction). Further,
   
5. There is a limit to data for Language Models. It has been proposed that we will run out of high-quality words (online books, articles, average Joes writing about his dog on twitter) in a few years. This will likely have another limiting effect, notwithstanding new learning models. And by 2026, I can see an AI teaching itself better. By then, we will be in runaway Singularity territory, and computers taking our books will be a very small problem.
   
   
   
   But that’s not even the worst thing.
   
6. As paragraph-level generation increases in quality, what’s stopping a human actor to string together many paragraphs into a coherent mess? I foresee many ‘writing sweatshops’ publishing hastily-strung-together novels that will inevitably further saturate the market. As always, the fault is not with the tool, is with who uses it.
   
   But for our situation, the difference is just semantic.
   
7. In closing, the one thing that may differentiate human text from AI text might be the process. The so-called human element. I foresee the situation will be quite similar to what’s going on with Chinese-produced porcelain, which goes through a cycle of a) mass-produced goods saturating demand b) artisans carving a niche for themselves out of people fed up with cheap goods c) companies pretending to be artisans saturating that niche with more cheap goods d) consumers wisening up e) artisans creating a further-differentiated niche and so on.
   
   As it becomes increasingly-difficult to distinguish finished product, the value of writing may come to reside in the humans who make it. This would mean focusing on the person who writes – their desires, fears, cravings, faults and limitations. The imperfections would give human-made writing its defining quality.
   
   I find this a disquieting thought.
   
   The Death of the Author has been at the foundation of my narrative and creative choices for years now, starting from my pen name. My idea is that it’s the word that counts, not the person crafting it. If I have a personality, desires, faults, fears and limitations, you could find them on the page.
   
   I am afraid this is not enough anymore – but at the same time I resent the idea of parading myself like a monkey to show other monkey my essence. It would cause focus to shift on the personality and quirks of the person making it, and away from their craft, sensibilities and poetic choices.
   
   I know there are examples of artists who thrived in this (Salvador “Avida Dollars” Dalì come to mind), but I have a feeling that I just cannot adapt to such a development. I have no desire to parade myself, only my stories.
   
   This proposed ‘resurrection of the Author’ might just be a way to cope with the current situation and its future development, but I think we will see it more and more in future.
   
8. One of my favourite artists is the Japanese Hokusai (1760-1849). His most famous image is the impressive The Great Wave off Kanagawa. Ironically enough, the image produced by Hokusai himself has long been lost. Due to the production process, what mattered was the wooden block used for the layered imprinting, and the hundreds of printed copies of that image are those displayed in museums and private collections.
   
   
   
   Art by Robo-san
   
   In short, one of the most-famous and celebrated, influential works of art in human history is itself mass-produced and mechanically-produced.
   
   We still find the time to praise the man who created it in the first place.
   
   Perhaps Hokusai has one more lesson for us all.