"Semantically equivalent decompression" is a very interesting idea, but I take issue with the idea that different words can share semantic equivalence. But I've

"Semantically equivalent decompression" is a very interesting idea, but I take issue with the idea that different words can share semantic equivalence. But I've also been reading a lot of poetry today, so subtle word choice feels important, even though I can recognize it may not be that important...a LLM could easily produce "better" or "worse" text than a human (while being lossy in either case), but then I have questions about the reason for using words in the first place.