I really like this vision and idea, it would be amazing if we would be able to achieve this. However I think that we are quite far off for now. Especially when he described how few materials we actually know the properties of. I would think that we would always need a combination of both the practical and computational. However stepping away from trail and error experiments will be important, because, in my opinion, we must utilise the limited possibilities to do these kinds of experiments to validate and improve on the computational models.