Don't get me wrong, I'm not a real supporter of LLM's but it's fair to say that not all text which is generated by an AI is spam. Depending on the effort put into prompting and preparation it is quite possible that it is valuable , useful and original content but that the author has used an AI to polish the language.

The problem is of course how to distinguish the one from the other. Just using the "AI detection" tools is however not going to get the answer we need. Most probably the usual upvoting schemes will continue to be the best way.

I'm going to be quite happy if the quality of the grammar, and in some cases the language, improves. Better quality text makes it easier to concentrate on the real thing I'm interested in: the content and ideas!

but let's say I post something and there's 10,000 comments in the first 10 minutes of the post. your going to need something to filter it out.

the only other options is pay to post. but even then it's way more lucrative to post spam then real comments.

I consider reddit comments quality to be like a age- and exposure- ladder and not everyone needs to read everything. By that I mean all age groups and expert levels (and now bots) are at one place making it hard to know what is relevant for my context. LLMs will help us navigate the reddit comments on client side. LLMs will prepare a summary report of the reddit discussion and highlight those which need our attention.

I see what you mean. But that's not a new problem caused by AI's, which I think was the original point.

I'm sure that all (useful) social media sites have a way to flag, or down vote such massive influxes.

Bot recognition of some sort of rate-limit transgression is another issue entirely, and not what I was trying to comment about (content based spam labelling based on "AI" authorship).

Perhaps I missed the point of the OP, in which case please ignore me ;-)

there are other forms of filtering/ rate limiting like by reputation, HN does this already to some extent for new users < 500 .

Recency of reputation changes can be factors so historical accounts don't accrue a lot of value for bots to buy.

We will downvote spam or low quality content anyway . Accounts (users or bots) with poor rep can be more harshly restricted.

Ultimately does it matter if a human is actually posting if the post is insightful / useful ?

An incentive structure for original human-made content, maybe. AI detection seems like a futile fight, thought. Even if you employed the best AI detection algorithms on the market, you'd still have to deal with false positives, missed spam and rapidly-evolving models that can be fine-tuned to adversarial data.