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 …

Chemistry is an abstraction of physics, even while fundamental questions remain in the underlying layer...

at the moment we understand both chemistry and physics way way way better than we understand neuroscience.

Huge mysteries remain everywhere but that is irrelevant to the matter of whether ANNs are some kind of abstraction

but relevant to how much confidence we should be expressing

it feels like your arguments shift with every tweet why not just say that you think (A)NNs are a poor or flawed abstraction, perhaps even one that will be eliminated by scientific progress it is most peculiar to demand that they not even be called an abstraction

i never said that couldn’t be called an attempt at an abstraction; the question is whether we can *confidently assume* they are a bona fide abstraction.