It seems inevitable then that we soon will have various large models in competition, with differently sourced legally licensed large training datasets. And opt-in open-source ones. And they will one day all hopefully be as good or better than the current models are now.