The "opposite" version of that quote makes less sense to me than "other way around" or "reverse". The opposite of "man oppresses man" would be "man doesn't oppress man", no...?

Maybe, I just googled the quote because I didn't remember it exactly. But I think it's actually clear that "man doesn't opress man" is not the intended meaning precisely because that wouldn't make sense.

Reminds me of the quote, "Give a man a gun and he can rob a bank, give a man a bank and he can rob the world"

Under post-capitalism, the automated systems serving capital oppress man?

... woman inherits the Earth?

IMHO attribuiting actions to automated systems rather than their owners hides who's actually responsible. If we don't accept abusing power we shouldn't accept abusing power using AI either.

I wouldn't be able to guarantee "capital" (i.e. people with access to large amounts of it) are smart enough to understand the outcomes of sufficiently complex systems.

I think the mortgage derivative swaps crisis taught us that. Powerful fools are still fools, and there's nothing about the system as-is that prevents some fools from becoming powerful.

In that sense, it's entirely possible we have bad outcomes from automated systems created by people who don't understand them. And that's the ambivalent case.

The actively evil case is: "How much would I have to pay you to design an unethical automated system?"

You're right, active evilness is not the only case or necessarily the bigger problem. Not knowing what you're doing doesn't mean you aren't acting unethically though when you risk others' welfare for your own gain.

But actions your automated systems are still your actions, regardless of intentions and outcomes. If we hold people responsible for negligently causing damage, we should also hold them responsible for negligently causing damage using an automated system. And of course, we should positive outcomes a system causes to the people behind it as well.