But employing a content moderator each 150 people, is basically exactly the same thing as employing a few hundred thousand to do so.

But, we don't even have to go all that far. As he also points out in his speech.
One of the main problems is not only the content itself but the ad models.
a) It should be forbidden to micromanage ads to too specific demographics.
(In this sense there should be a max restriction of "within this state, within this age range")
b) All ads, especially political ads which are paid for, have to be fact checked.

I think doing this would already go a long way.

Eventually, these companies could develop algorithms to detect improper content (what they do already anyway) and then manually review those.
Similarly, reported content should undergo the same thing.
Additionally to that, content which is on the verge of going viral could also undergo similar checks.