JBlack comments on Language models seem to be much better than humans at next-token prediction

[JBlack]

The feedback was displayed, but it wasn’t for my sentence. It was for a different sentence, with different word choices.