Here's another way to look at the spam filtering problem. Consider this hypothetical?

1. Let's say you have a new account on the network (Account A) at they reply 100 times to the same account (Account B) with no engagement back.

2. It doesn't actually matter if Account A is run by a human or a bot with AI. 

3. If you have 1000 accounts like Account B, Account A will just stop using the app. They'll move to another network or a messaging app.

4. Ideally, Account B would reply thoughtfully a few times and Account A engages with them, and then it increases over time as they build a relationship.

5. If you say "well 100 times is too much, but 50 times is fine", then you're admitting humans can be spammy and we're now arguing over the definition.

6. Side note: I don't think anyone is ready for a world where bots powered by AI are as interesting—or even more interesting—than humans.

Not enough focus on incentivizing quality.

When there are so many projects, tokens, systems and mechanisms that incentivize engagement - and only a few small projects that try to promote quality content - it's not surprising that we get so much engagement farming.

You can't solve this with LLMs alone bc you're filtering too many user accounts that mostly use the platform normally but may also want to profit from the incentives (or maybe they're just caught in the LLM filter regardless). But if that's the approach there needs to be greater bottom-up user feedback on spammers.

If the incentives were stronger to create quality content you'd have "quality content farming" - by both users and bots. No one would be complaining about either.

Working on /impact ||


Abundance Economy book: book.abundance.id

I think they're just solving it algorithmically, which is also part of the core of your point: 

They aren't looking at the content, they're only looking at the engagement metrics (how many times they've posted and how much engagement that has generated from the OP).

Not that llm's could actually assess quality content either, but they're not even going that far, just stopping right at engagement as the only qualitative factor.

Could start a game of "bot or not" where users are presented with endless random posts by users with low follow counts and they swipe left or right based on if they think it's quality content or trash 🤣

💯. Big part of the problem.

LLM can potentially get better at detecting real bots/spammers (as well as quality content) but they need to be trained by data from users. Without this bottom-up feedback users will be caught in these filters for normal behavior