But let's look at how that helps in some cases.

So if we already know the object is a cup, and we know how it's positioned, then its shadow is an actual guide to its particular geometry.

So in cases where we have enough a priori scientific information, we can rig datasets (shadows) to be informative of the target domain.

Here the target is discrete: say the peaks and tips of a mountain line. Now can we rig a photo of a mountain to have in its ink an informative structure?

Sure. Now if we didn't know it was a mounting apparent peaks aren't even 'peaks' at all, they're just patterns of ink.

A priori explanatory models are needed to rig data for statistical modelling.

No such rigging can take place absent them

And with them, we aren't really discovering any new science -- rather we're gaining highly particular knowledge typically useful in engineering

Biologists in these areas describe this research as quite trivial low-hangimg fruit. It's not of much research interest just to automate these kinds of investigations

My masters was on a very similar project applied to quantum metrology -- it's always 'useful' but it's always also just a kind of engineering utility. We couldn't even do it if we hadn't already done the science

They just fed protein sequences. They did not alter the architecture in any way. To the transformer, it may as well have been any random assemblage of letters and numbers.

Functions like secondary structure, contacts, and biological activity were found because those things are implicit in the creation of the data, not because the model was "rigged" in any way.

that is where the science occurs -- the data analysis is just an administrative task after science has taken place

The only "experiments" performed here were done by biology and evolution.