@emollick
Neat paper suggesting that human augmentation and task automation are not necessarily related. Models that are really good at doing work are not always good at helping humans do work better. Given the pressure to make models good agents, this may undermine human-AI cowork.