The frustrating thing about all this speculations is, that we don't know what was in the training data, but I think we should know that, to have any meaningful discussion about it.