Probably there is a lower ceiling in text generation to things that are considered “progress” by everyone. People can verbalize exactly what is wrong with pictures and video, but for text, if it is coherent and gets the point across what is progress beyond passing the Turing test?
I feel like on image/video generation, I can agree that the recent months have shown a lot of progress.
I’m not so much feeling it on the text generation front, it’s felt about the same to me for a while here, notably impressive, but still a bit… off…
Probably there is a lower ceiling in text generation to things that are considered “progress” by everyone. People can verbalize exactly what is wrong with pictures and video, but for text, if it is coherent and gets the point across what is progress beyond passing the Turing test?