#dude sounds like the type of guy I used to avoid at SF cons

[spikewriter]

I saw another anti-AI post where the first words out of someone's mouth was "Plagiarism!" That is why it's so difficult to have reasonable discussions about these new tools--and how they be useful as tools--because people start screeching, "You're not a real writer!"

The article at the core of the post, however, is worth discussing because, yup, it is exactly what the antis are yelling about. The post, by the way, did not include a link to the article, just a screenshot of Publisher Weekly's Twitter promo of said article. Which is actually a rewrite of a Newsweek article about a man who was about to release his 97th ChatGPT-written "novel." I'll explain the quotes later on.

I've included a link to the original article because it's worth a read no matter what side of the argument you're on. The headline is absolutely clickbait. It's also full of self-aggrandizing bullshit.

Tim Boucher (the article is written by him, or, rather, 60% written by ChatGPT by his own admission) admits to making $2000 over the course of 7 months. Hardly the thousands of the headline. He's sold 574 books as of the article, which equals out to an average 5-6 copies per book, or an average of just under $21 per book. The books are 2,000 to 5,000 words each, so they're not really novels, but serial chapters. He is also, by the way, not selling on Amazon or any other distributor, possibly because some of the stories are too short for them to accept.

It also means he has an extremely small, niche audience who are interested in "dystopian pulp sci-fi with compelling AI world-building." He writes "majority of my readers being repeat buyers, many having bought more than a dozen titles. In one case, a reader has bought more than thirty titles."

I found this paragraph particularly illuminating:

"It's very difficult, for example, to have longer written pieces that maintain a coherent single storyline or character arc. So instead, I've tended to lean into short "flash" fiction slice-of-life collections, interspersed with fictional encyclopedia entries that deliver world-building and backstory, and point the reader towards other volumes where they can continue down the rabbit holes that appeal to them the most."

Right there is the issue with current LLM programs. You can get a coherent storyline and character arc with ChatGPT or Sudowrite, but it takes manipulation on the author's part. It takes being willing to put in the work to revise and massage the outlines. Dear god, don't use it to write scenes, because the quality of dialogue and description is horrendous.

This guy isn't. He's only willing to put in 6-8 hours to create and publish a book, which may include generating the cover and any brainstorming. What he is doing is the tech boy grift of inflating what the program is capable of and his own accomplishments. He's trying to shout, "I am a disruptor! I am the future!" (And taking a look at his website, he's also a conspiracy theorist about underground cities in Antartica.)

Sadly, this is exactly the type of person other tech bros who might be making decisions are going to listen to. And because he's publicity-hungry, he's making everyone else who is trying to use these tools to assist, not replace, the process look like a grifter as well.

Oh, and I can't help including this article written in response to the Newsweek one.