> I and I assumed a lot of people, already thought performance was a function on model size (# of parameters). I guess the disagreement has been in whether thi

> I and I assumed a lot of people, already thought performance was a function on model size (# of parameters). I guess the disagreement has been in whether this function is "continuous" or not. I do not think the title is misleading, considering the article answers to quite specific claims in other articles. I agree it sounds misleading if you do not put it in that context.