A story about ChatGPT, reading, and Nick Cave
February 22, 2023
A story about ChatGPT, reading, and Nick Cave

I've always liked to read science-fiction. More particularly anticipation.I think it really started when my dad gave me two plastic tots full of sci-fi books. It was nothing fancy. Mainly from publishers Fleuve noir and J'ai lu (Flammarion) and where from the 60s to the 90s. It had classics (which I didn't know were classics at the time) like Foundation, Dune and Ubik and a lot of Robert Silverberg novels. The French translations where not alwats good but it didn't matter to me. It also had a lot of less known books from authors who seemed to have disapeared without leaving any traces on the internet.

Although I might have a more critical eye now, it all seemed wonderful back then. I loved the eerie, the new questions it opened, the world-building. I quickly became fascinated by anticipation and it's dystopias, obsessed even, and begun my quest to absorb as much of it as I could.

Unfortunately, like a lot of active adults, I have much less time to read as I would like. Especially a genre that is as taxing as anticipation, that requires me to completely disconnect from the outside world. I buy books, yes, but I barely make a dent in them. So my books purchases where merely an attempt to reassure me that I am STILL an avid reader.

That changed when my wife gave me Dangerous visions[1] for Christmas. It's an anthology book by Harlan Ellison. Instead of simply making an anthology of different texts he selected, Harlan Ellison commissioned them from famous authors. Philip K. Dick, Robert Silverberg, Poul Anderson, Carol Emshwiller, Damon Knight, Philip Jose Farmer (who would win a Hugo Award with his submission) are some of the authors who wrote in Dangerous visions. It was incredibly exciting to be able to write short stories written by those heavy-weight of the new-wave sci-fi. Finally I was able to read a story for 15 minutes and move to another one later, and still be able to read a book from start to finish ! I felt my craving for reading grow again.

After that, I started scouting the internet for anthologies. I found cool website like Strange Horizons[2] but the one I liked the most was, and by a large margin, Clarkesworld[3]. It was like a monthly delivery of sci-fi and sci-fi related non-fiction. The quality would usually range from great (some award winners in there !) to good, with, of course, the occasional story that I felt was lacking but that's the good thing with anthologies isnt't it ? I could order a physical copy for the price of magazine and that was that.

Well, last night my wife who had assessed that the news couldn't wait (and she was right) woke me up to tell me that Clarkesworld wouldn't take submissions anymore. Neil Clarke, the editor of Clarkesworld, published in a blog post[4] that they would stop taking submissions for the time being. The cause ? Language model AIs, more precisely ChatGPT. You see, after chatGPT opened to the public, Clarkesworld became flooded by submissions. AI generated submissions. If you think I'm exagerating take a look at the graph posted by Neil Clarke :

Graph starts in June 2019 and displays monthly data through February. Minor bars start showing up in April 2020. Mid-21 through Sept 22 are a bit higher, but it starts growing sharply from there out. Where months were typically below 20, it hits 25 in November, 50 in December, over 100 in January, and nearly 350 so far in February 2023.

The reasons are simple :

  • Clarkesworld pays the published authors
  • ChatGPT generates a story in seconds
  • It's free

In other words, writing a short story from scratch and sending it to a publisher just to make money is not worse it, but having an AI generate stories and send them to a publisher is quick and doesn't cost anything. It's a free-entry lotery. Clarkesworkd is not the only publisher who is facing this issue. In the same post, Neil Clarke mentions :

I’ve reached out to several editors and the situation I’m experiencing is by no means unique. It does appear to be hitting higher-profile “always open” markets much harder than those with limited submission windows or lower pay rates. This isn’t terribly surprising since the websites and channels that promote “write for money” schemes tend to focus more attention on “always open” markets with higher per-word rates.

Now, the issue, or at least to me, is not to know wether or not the story has been written by a human (this is another can of worm I'd rather not open in this, already too long, post) but the fact that :

  • The submission rate went through the roof
  • AI generated stories are terrible

If the art that AI could create was equal to what a human can achieve, the question would be more of a moral one. Should we allow those stories with no authors (or with all the authors perhaps, knowing how those models are created ?) to be published ? Are we even allowed to enjoy them ? The media has picked up on the story and hasn't missed the irony that the whole premise could be a Clarkesworld's published story itself. I find the idea of a competent, art-creating AI more interesting. What if art, music, litterature where infinite ? Every book you read and love, unique and tailored for you ? No matter the life-altering masterpiece you've just finished, your attempt to share it with the world would be met with disdain as everybody would be too busy with the ones made for them, and it would disappear as fast as it was created.

But this idea is indeed science-fiction. For now at least. Because the issue at hand here is not that we can't make the difference between AI generated novels and human ones. We very much can, do the test yourself, chatGPT is pretty terrible at every creative task. So clarkesworld is flooded by submissions, but the ratio of rejected and approved submissions changed drastically. There are still the same number of good stories sent to them, but they have to commit an immensely bigger amount of work in order to filter out the garbage. It's one thing to read a terrible novel from a passionate writer and I'm sure it gets dull sometimes. I can't imagine how draining it must be to have to read through hundreds of AI generated nonsense to weed it out.

Nick Cave recently talked about it when someone sent him a ChatGPT-generated song "in the style of Nick Cave"[5] (I'm stll baffled that anyone could think that it was a good idea). Suffice to say, he didn't like it :

This song sucks

We could talk in length about why machine learning is excellent at figuring out strategies that have a measurable outcome and why it sucks at creating more abstract things (without a human observer, there is an easy way to tell who the winner of a chess game is, not so much to tell if a song is good) or creating new things. But frankly I don't have the credentials for that and the media is already spewing enough nonsense about all of this for me to add more to it.

I would add, that AI generated graphic art is pretty impressive, at least for the uninitiated like myself. Midjourney is giving impressive results and as won some local contests, but then again, never innovates.

Nick Cave seems to believe that their is an inherent ability humans have in the way we feel and create. Something that could never be replicated by machines or programs :

This is part of the authentic creative struggle that precedes the invention of a unique lyric of actual value; it is the breathless confrontation with one’s vulnerability, one’s perilousness, one’s smallness, pitted against a sense of sudden shocking discovery; it is the redemptive artistic act that stirs the heart of the listener, where the listener recognizes in the inner workings of the song their own blood, their own struggle, their own suffering. This is what we humble humans can offer, that AI can only mimic, the transcendent journey of the artist that forever grapples with his or her own shortcomings.

I must admit, as a musician, and songwriter myself (although this side of me has suffered the same fate that my eager-reader side as of late) I admire Mr. Cave's optimism and I somehow hope that he is right.Then again I am sure that he hope that he is right too. A computer can already do most things better that we can, what are we to be if it feels better than us as well ? The good news if that if we never succeed in making good, disruptive and innovative AI generated art , then maybe we are special and Nick Cave is right. The bad news is that it might also be the way to transcend human intelligence and consciousness (another recurring sci-fi theme) if that was ever something we had in mind to achieve as a specy

But this at least won't be an issue for the short future. Right now the issue is that we're using new technologies to mass produce mediocrity in order to make financial gain. It's starting to look like a trend and it's not about to stop any time soon. As a result, my favorite magazine, is not taking submissions anymore.

If there is one thing that I understand in all of that, is that reading about dystopias is more fun than experiencing them.


[1] https://en.wikipedia.org/wiki/Dangerous_Visions

[2] http://strangehorizons.com/

[3] https://clarkesworldmagazine.com/

[4] http://neil-clarke.com/a-concerning-trend/

[5] https://www.theredhandfiles.com/chat-gpt-what-do-you-think/

Add a comment.
250 characters left
mastodon github linkedin