The bot scene - non-LLM bots - allowed, some rules - LLM bots - unworkable...
The bot scene
- non-LLM bots - allowed, some rules
- LLM bots - unworkable rules (socialpolitical polemics) or blanket ban
And what do I find on mastodon.social? LLM bots that *could* be well behaved. These bots have been tweeting for a lot time, so moderators don't seem to care, even if the LLM is poorly behaved.
Self-replies
Well behaved bot:
- doesn't post to hashtags (don't butt into a convo)
- don't initiate a reply, retweet or like (don't butt in)
- do, optionally, reply to a reply or a mention (A mention is permission)
- similarly a follow is permission to interact.
Really, those should be rules for bots written in C# too.
I don't know about following tho.
- A follow seems like permission to follow back
- 50% of my followers on any platform are noise, so why penalize bots following. We allow humans to follow even if the profile is blank.
Bots written in rust, can be crappy too, maybe it just tweets the digits of pi one per day. That creates server load for a one time sensible chuckle or two.
I think a good LLM bot is one that takes info from the world and transforms it, like an RSS feed, but with commentary. There is a bot that does exactly this on mastodon.
And finally, people posting "negative policies" in their pinned tweet or bio e.g. "no bots!" - what does that even accomplish? The bots written in ruby can't read it. If the bot is malicious, it will be ignored. If the bot is well behaved, it shouldn't be initiating contact. If you put "no bots!" into your profile and then follow/mention/reply to a bot, you just gave it permission to interact.