An interesting use for AI right now would be using it as a gatekeeping filter, selecting social media for quality based on customisable definitions of quality.
Using it as a filter instead of a generator would provide information about which content has real social value, which content doesn't, and what the many dimensions of "value" are.
The current maximalist "Use AI to generate as much as possible" trend is the opposite of social intelligence.
I think that's right. Twitter without ads, showing you content you _do_ want to see using some embeddings magic, with decent blocking mechanisms, and not being run as a personal mouthpiece by the world's most unpopular man ... certainly not the worst idea.
It's a nice idea in principle, but would probably immediately become a way by the admins to promote some views and discourage others with the excuse of some opinions being of lower quality.
That's what moderation is and is perfectly fine. Dang does that here on HN and for good reason.
It's not moderation, for one thing it never will be used with moderation.