This is more or less a 2D positional encoding where the vertical dimension indicates who and horizontal indicates when. It seems like it would save a lot of headaches. Imagine if everyone’s voice,…

Posters: eshear. eshear

2 likes recorded by the source corpus.