The Current Limits of Modeling

The more I think about it, the more I like the idea. We've known for a long time that scale models aren't perfect surrogates for actual boats, and the smaller the scale, the less useful they are. At the same time digital modeling isn't any better, just different, and not even necessarily cheaper.
However, with the impressive gains in AI and machine learning it should be possible to translate detailed model studies in the towing tank, or bodies of water into increasingly accurate data.
The difference between scale model results and full size reality is exactly the kind of thing machine learning should be good at.
There must be countless examples out there of scale model data and corresponding real world data to pour into the black boxes.