That is what I have repeated so many times in the last 2 years over and over. I consider Yi Tay's response [1] a mere technicality that is actually irrelevant. What is relevant is how predictable "interpolatable" the data are, how predictable we are.