Just poking fun at the way these things are presented in pop.sci. media. Anyway, talking over my head, I'm not shure formal systems are a good way of modelling the mind itself. Formal systems are (usually) considered consistent. The mind is not. Neither are neural nets. And it's interesting how neural nets, and other statistical models/algorithms that drop the requirement of always beeing right, seems more much more capable in certain practical matters.

>see the mind as computation over a formal system

Wow, is that what people actually think a "computational theory of mind" is? Look, just because every program can be trivially rewritten as some kind of formal proof system, doesn't mean that any given program meaningfully has the semantics of a formal proof system, let alone the mind.