I think that high volume validation datasets for predicted/measured behaviors will be important. Maybe automation in experiment is necessary to get to that size

I think that high volume validation datasets for predicted/measured behaviors will be important. Maybe automation in experiment is necessary to get to that size, but if you can't verify it, it's not science.