very interesting post; i also had a hard time w/ that section, although I recall reading "The Beast in the Jungle" as an undergrad; but a very long time ago, i was teaching Sven Birkert's The Gutenberg Elegies ( I almost wrote Eulogies) and I gave students a page from Hawthorne's The Scarlet Letter -- and for most of them, they were defeated by the prose -- sort of what happened with your Henry James section. And another example: I assigned Norman Mailer's "The Armies of the Night" -- and the students could not get past his style. We are in a way different universe for sure.
I’ve said something similar. Your voice is exactly that; it’s your voice. No AI can possess all you’ve brought in, cast off, thought over, tried, failed, changed course, or went back to the drawing board on before you arrived at it. The real danger lies in those who never do this. The real danger comes for those who let the fake speak for them. Good piece.
I’ve read that one too. I definitely found it intriguing. Weird that he didn’t seem to have the “mad governess” idea in mind when he wrote it, since that does seem to be what really makes the story.
LLMs represent style as a vector. What this means in practice is they can create new styles by blending or accentuating existing styles. Thats the same way most authors come up with unique styles; by blending, accentuating, or removing other styles.
It's really fun to work with primitive ais for image and text generation. The lower quality of output also means a lot more creatively interesting outputs. I've used these primitive ai quite a bit in my artwork. More modern stuff is less interesting from a creative perspective if you use the defaults, but you can generate really interesting output with the right inputs. Been exploring this specifically with threaded spellcasting, where you have the ai encode a spell as an encrypted numeric message, post it on a forum, then other users reply to it and you feed it back into the engine. Generates a lot of unexpected situations.
I share my work with traditional galleries and exhibitions on xenofluff.com, most of which are designed to challenge popular assumptions about machinic creativity. Mostly I build site specific installations with painting, music, video, and dance.
very interesting post; i also had a hard time w/ that section, although I recall reading "The Beast in the Jungle" as an undergrad; but a very long time ago, i was teaching Sven Birkert's The Gutenberg Elegies ( I almost wrote Eulogies) and I gave students a page from Hawthorne's The Scarlet Letter -- and for most of them, they were defeated by the prose -- sort of what happened with your Henry James section. And another example: I assigned Norman Mailer's "The Armies of the Night" -- and the students could not get past his style. We are in a way different universe for sure.
Oh interesting, what didn’t they like about Mailer’s style?
I’ve said something similar. Your voice is exactly that; it’s your voice. No AI can possess all you’ve brought in, cast off, thought over, tried, failed, changed course, or went back to the drawing board on before you arrived at it. The real danger lies in those who never do this. The real danger comes for those who let the fake speak for them. Good piece.
Well put! And thank you!
I think an exception has to be made for “The Turn of the Screw.”
I’ve read that one too. I definitely found it intriguing. Weird that he didn’t seem to have the “mad governess” idea in mind when he wrote it, since that does seem to be what really makes the story.
LLMs represent style as a vector. What this means in practice is they can create new styles by blending or accentuating existing styles. Thats the same way most authors come up with unique styles; by blending, accentuating, or removing other styles.
That’s what the AI people need to tell themselves, but I really don’t think that’s true.
It's really fun to work with primitive ais for image and text generation. The lower quality of output also means a lot more creatively interesting outputs. I've used these primitive ai quite a bit in my artwork. More modern stuff is less interesting from a creative perspective if you use the defaults, but you can generate really interesting output with the right inputs. Been exploring this specifically with threaded spellcasting, where you have the ai encode a spell as an encrypted numeric message, post it on a forum, then other users reply to it and you feed it back into the engine. Generates a lot of unexpected situations.
Can you send some art you’ve generated this way? I’m curious.
I share my work with traditional galleries and exhibitions on xenofluff.com, most of which are designed to challenge popular assumptions about machinic creativity. Mostly I build site specific installations with painting, music, video, and dance.