If instance admins allow AI Agents on their platform and they keep harassing us I have no other choice then to silence that instance

[stux@mstdn.social]

If instance admins allow AI Agents on their platform and they keep harassing us I have no other choice then to silence that instance

Again, I do not pay these massive costs each month to host robots

Let's keep the #Fediverse human shall we?

#AI #LLM

[mfierst@mstdn.social]

@stux
Thank you. I agree, let's keep the Fediverse human.

[hans@social.woefdram.nl]

@stux️ The challenge will probably be to recognize who's human and who's not.

I don't think agents will announce what they are in their request headers, and determining if texts are AI generated or human is already pretty much impossible today...

[champersbe@mastodon-belgium.be]

@stux
Horrible how many bots are signing up lately

[finitebanjo@feddit.online]

Hell yeah, man. Personally, I’d be willing to solve a puzzle every half an hour when I navigate to the homepage or post a reply.

[mirabilos@toot.mirbsd.org]

@ChampersBE @stux huh, hadn’t noticed, whom do I need to defederate?

[atropa@piefed.social]

Ty, only for human interaction

[landkieker@norden.social]

@stux Richtige Einstellung.

[wyatt_h_knott@vermont.masto.host]

@stux I'm thinking of migrating to mstdn.social. What do you think a fair contribution is? i.e. total costs divided by users, ballpark figure.

[stux@mstdn.social]

@Wyatt_H_Knott Always welcome!

Ah good question! I think about 1000 divided by 250K total or 11K active or something I guess p/m

[catwoman69y2k@meow.social]

@stux Is there a particular instance you have witnessed/experienced this behavior from? Harassment is not okay, and that includes when a bot does said harassment

[anguadelphine@mk.absturztau.be]

@stux@mstdn.social

Admin Approval / Invitation Code via longstanding instance denizens; to start with.

A probationary period of restricted activities for up to a year - should prevent BotWare. They're usually spotted & decommissioned within a week.

Auto Decom for 3 breaches of the rules whatever / whomever.

Easy for me to say, harder for Admins to do, but you can set and I assume maintain rules easily so reducing Whack A Mole.

One of many reasons that for most of my 4+ years on Fedi, my accounts have been Follows Require Approval.
Also I only use Misskey forks & Akkoma so easier to spot robos. Usually when I get blanket Requests across my main accounts within 24 hours of each other.

[justincrozer@mastodon.green]

@stux I don’t really like automated bot accounts either. I mean if one uses a service like buffer to cross post that’s okay but the ones that scrape certain words and hashtags to just boost annoy me.

[swope@mstdn.plus]

@stux

Do you mean to include plain-automation bots?

I think those can be useful in some cases.

But I agree that LLMs disguised as humans harassing humans is certainly bad.

[stux@mstdn.social]

@swope No, no, just the LLM ones

Acting like human

RSS bots etc are super useful

[crankylinuxuser@infosec.exchange]

@FiniteBanjo

The clankers can also solve those puzzles too.

And when a new puzzle is found that works for humans but not clanker soon is retrained so it too can solve.

I dont have any solutions, other than what would amount to Butlerian Jihad.

[wikifox@g0v.social]

@stux I can't understand why my teacher say OpenClaw(AI Agents) can replace real human's job.

[lunadragofelis@void.lgbt]

@stux Humans and sentient creatures welcome! (Not every fedi user considers themselves human, even if they have a human body)

[ferret@hub.workersofthe.world]

@stux️ I've got one going as part of a running experiment. I'm transparent about it, though, and it isn't fully automated. People are definitely within their rights to block the clanker if they just don't want anything to do with it, though the experiment does depend on some level of engagement.

[msh@coales.co]

@Catwoman69y2k the biggest 3 offenders of spammy/scammy #LLM #AI bots appearing on my server have been mastodon.social, mastodon.online and mastodon.world though they don't exactly "harass", they are all looking to mass follow to acquire training data at best, or commit large scale fraud at worst. I don't think the admins of those servers are being deliberately neglectful, it is mainly because they are too big for their admin teams to stay on top of.

IMO for fedi owners and admins the best policy is to disable open account creation and require manual review or invite only, but many people resist that because it would slow adoption growth and larger instances would gave to to a huge amount of work to manually approve new accounts.

All fair points, but in this age of out of control AI slop it might just be that the best path to success for fedi is to massively increase the number of servers with the average server side being much smaller, because smaller servers are much easier to manage.

@stux