Wait did they talk about scaling _up_ to datasets of gigabytes in size? Those are the size of toy datasets I use to prototype methods... If this is an accurat

Wait did they talk about scaling _up_ to datasets of gigabytes in size? Those are the size of toy datasets I use to prototype methods... If this is an accurate reflection of how people approach ML problems then I can understand the "growing pains" (heh) I hope it's not.