Gary Marcus

astonished that anyone knowledgeable could claim that neural nets are (obviously) an “abstraction of neural processing” when we don’t yet know how brains work. if you don’t know how Y works you can’t really speak with certainty about whether X is an abstraction of Y. Period.https://twitter.com/tyrell_turing/status/1200072223299657728 …

Does _any_ model used in science abstract a process that is completely understood?

Yes, one useful case is when it is computationally inefficient to use a more complete model than a simplified/abstract one. This is why abstraction is useful, must decide when to trade off completeness for speed.

I understand that abstract point, noting that it was not actually accompanied by an example of a single natural process that it describes (which I could not then turn around and say all the ways that we don not understand that natural process, as gmarcus does)

If you consider computer science a real science, abstraction is used in this sense for hierarchical reinforcement learning (https://h2r.cs.brown.edu/wp-content/uploads/2016/05/gopalan16.pdf …). Also, many simulation use simplified physics models (lambertian reflectance https://en.m.wikipedia.org/wiki/Lambertian_reflectance …)