Reading

General, no-spoilers overview:

Some of the changes are obvious “doing it this way worked better on-screen” things. Scenes that were just-MB in the book become group efforts, giving the PresAux actors more to do. Plot points that were just inner-monologue realizations in the book are delivered in conversations instead.

I mostly like them! Even with the characters, even a few dramatic personality shifts — look, I’ll be mad if some of them start bleeding into book!fandom, and fans stop writing the original versions of the characters. But as a standalone AU, most of them work really well.

The few changes I actively don’t like are all “why did you even add this, what was the point?” kind of things. No huge dealbreakers. Just some low-key annoyances.

There are a few particular exchanges from the book that you really have to get right to make a satisfying adaptation. They’ve all landed. And a bunch of the comedy moments have been had-to-stop-the-episode-while-I-cracked-up funny.

The biggest advantage of doing Murderbot on TV is, The Rise And Fall Of Sanctuary Moon is also TV. Which means the showrunners can film Actual 100% Authentic Sanctuary Moon Footage, and cut to it while MB is watching. It’s ridiculous and amazing.

Edited version of an IBM rule: A Computer Can Never Be Spiteful Or Horny, Therefore A Computer Must Never Make Art

Saw my first “did using an LLM screw up your business? We can help you find someone to fix it” ad in the wild today. (It was a Fiverr commercial on Youtube.) Wonder how many more of those are coming.

—

From 2023: Novelist Alexander Wales blogged a bit about trying to get an LLM to generate a publishable novel. He made a good-faith effort, took a lot of different thoughtful approaches, and documented enough of it to be a good read. Part 1: “I’ve been trying my hand at writing with the assistance of ChatGPT and occasionally other tools. Mostly, it sucks…” Part 2: “I’m still trying to get an LLM to write me a novel, and experiencing the first major setbacks while working on chapter 2.” (There is no post 3.)

And from this January: “A dad just can’t seem to figure out why his six-year-old daughter wasn’t impressed by the AI toy he gave her for Christmas. […] He writes that he cannot understand why his daughter disabled the dinosaur plushie’s built-in AI voice — opting, instead, to play with it like a regular toy, and dressing it with clothes she made.“

LLMs are an interesting novelty the first time you play with them, but for people with actual creativity — whether it’s writers, artists, or Literally Any Child — you overrun their limits and get bored with them so fast.

(What really gets to me about the dinosaur one is the dad saying he “wasn’t able to really understand where’s the resistance.” Instead of approaching the problem as “let me analyze this toy to figure out why it hasn’t earned my kid’s interest,” he’s gone with “of course the toy is entitled to my kid’s interest, let me analyze her to figure out why she’s ‘resisting’.”)

—

From this week, a writer trying to get ChatGPT to quote/summarize some linked essays: “The lines you quote are not lines I wrote. They are not in the piece. What is going on here?”

Ending on a golden note from FFA: “Hi my name is Loquacious Techbro Midjourney ChatGPT Claude AI and I have long, beige, run on sentences (that’s how I got my name) with purple prose streaks and red flag tips that reach into the stratosphere and icy blue prompts that like using limpid tears and a lot of people tell me I look like 300 Ghibli characters smashed together(If you don’t know what that is get da hell out of here!).”

Erin Watches: Murderbot (TV), season 1

General, no-spoilers overview:

Detailed reaction, with spoilers:

and a lot of people tell me I look like 300 Ghibli characters smashed together