Sure, I agree.

The issue is the word "computer" means not "device we have made" but "universal turing machine".

Ie., a computer is any system which realises a function from the Naturals to the Naturals.

Physics barely, if at all, has any use whatsoever for those functions. It is a very important point.

Computer scientists (ie., discrete mathematicians) are not the people who are even able to describe, engineer and build whatever is needed for an AI -- if, as I claim, continuous dynamical properties are needed.

(As, for example, they are needed by pretty much every system.)

Computer scientists do study computers that use real numbers, for example it's known that such computers can solve all problems in #P in polynomial time. Many other areas of computer science also use plenty of real analysis.