You get that effect when LLMs write whole sentences or paragraphs. Editing human prose for minor consistency improvements doesn’t do that unless you let it run wild and replace >20% of your text.
Slop is a misapplication issue. Just because today’s models can pump out a lot of text doesn’t mean they’re good for it. And just because letting them run wild produces slop, doesn’t mean they don’t work well for appropriately scoped applications.
Slop is a misapplication issue. Just because today’s models can pump out a lot of text doesn’t mean they’re good for it. And just because letting them run wild produces slop, doesn’t mean they don’t work well for appropriately scoped applications.