That is one of the reasons I want it done. We cant tell if AI's are parroting training data without having the whole, training data. Making it old means specific things won't be in it (or will be). We can do more meaningful experiments.