I've been reviewing a lot of LLM-generated documents lately and while the ability to express ideas in unexpected term struck me as impressive at first, it's now just consistently annoying.
It's still pretty evident that the models lack in understanding the underlying concepts...
This is perhaps an unexpected emergent property, but heavily speaking in metaphors to reason about things I would have presumed to only be possible to express in a latent space wasn’t on my 2026 bingo card.
Model output will sometimes strike you quite deeply as in “oh what a strange permutation of words, but i know what it means by that”.
@sun_girlxo I think a lot of it has to do with being able to drop excessive pride/ego. Saying this as someone in a particular episode of struggling with forgiveness
using ChatGPT Images v2 feels like you get access to see everything that could ever happen. you can be the first to discover something no one has ever seen before. have not had this much fun generating ever
@gabeeegoooh and team are absolutely incredible
You don't watch your phone in new york, no one's stealing phones from ppl's hands. Extreme contrast with e.g. London where it happens all the time. Still puzzling to me why