The experiment improved outputs but the text still felt flat
But the most important thing here is that there is any improvement. I don't know whether these methods will scale to great writing. But they have worked on other things.
On a heroic attempt to undo AI slop: hours of coding, an hour of manual data labelling, $300 in APIs... resulting in text which is flat but serviceable.
@gleech Interesting. Curious, does the rewritten task pass pangram btw?
On a heroic attempt to undo AI slop: hours of coding, an hour of manual data labelling, $300 in APIs... resulting in text which is flat but serviceable.
https://universalprior.substack.com/p/on-slop
But the most important thing here is that there is any improvement. I don't know whether these methods will scale to great writing. But they have worked on other things.
@krishnanrohit low confidence
@gleech Interesting. Curious, does the rewritten task pass pangram btw?
On a heroic attempt to undo AI slop: hours of coding, an hour of manual data labelling, $300 in APIs... resulting in text which is flat but serviceable.
The experiment improved outputs but the text still felt flat
But the most important thing here is that there is any improvement. I don't know whether these methods will scale to great writing. But they have worked on other things.
On a heroic attempt to undo AI slop: hours of coding, an hour of manual data labelling, $300 in APIs... resulting in text which is flat but serviceable.