The media in this post is not displayed to visitors. To view it, please go to the original post.

Heads-up for anyone who doesn't want their posts scraped and handed over to an LLM.

alpaca.gold/@seldo/11628609961…

(Quote replaced with link since I was blocked.)


You would need to block me from following you, I suppose.

This entry was edited (2 months ago)
in reply to DB

@lawyersgunsnmoney AFAICT he made a bot you hand your social media credentials over to and it logs in to all your accounts, scrapes everything, and passes it to an LLM service to "summarize" (and keep for training). Utter clownery. Normally I would just block, but I'm screenshotting and sharing this because it's a privacy threat to everyone.
@DB
in reply to Cassandrich

@YvanDaSilva @xri Like, if you have techbro brain and someone in your replies says something insightful in a way you haven't heard before, it has a chance of breaking thru and making you reconsider things.

But if you get it "summarized" by replacing it with a blend of the closest commonly-repeated stuff from your corpus, you're gonna interpret it as "oh yeah, just more of the same complaints I've already dismissed".

It's basically "turning your audience into yes-men as a service", which sounds like his peers would love it.

in reply to princess pancake ​

In the case of this app, it scrapes from its user's feed, so if you block its user, you will not be in the feed it scrapes.

These techbros want us to believe their "it's silly to have norms if the machine can't enforce them" bullshit. Here, whatever blocking we can do is a statement of our disapproval of their attempts to bypass our consent and reduce the damage, not a surrender of our standing to demand they stop doing things that technical measures fundamentally can't stop.


@notyourfanboy "Welcome to the real world, where anyone can rob your house by smashing your windows and climbing inside."

Do you see how deranged that sounds?

We expect people not to do this because it's antisocial, illegal, and wrong. Not because there's some technical measure (which of course is impossible) preventing it.


This entry was edited (2 months ago)
in reply to princess pancake ​

@natty @SRAZKVT No, we do not have a flag for this. "Indexable" means I want people to be able to find my toots in search results. Not that I want to allow corporations to ingest them into their slop regurgitation machines and mimic my writing style but saying things I didn't and would never say and present those things as "summaries" of what I said.
in reply to Charo del Genio

@paraw

You can't.

Every instance admin would have to intercept API requests and check the origin if it is the Zeitgeist instance and block them.
But you can also just grab the code and run your own instance.

The vibe-coding author mentioned to add #noindex and #nobots support but based on his messages and reactions this won't happen.

in reply to Cassandrich

Funny, I was just thinking the other day it would be really nice if Mastodon gave me the option of opting IN to have my posts scraped by an LLM.

Because I remember that cool Twitter game a few months ago where you could ask ChatGPT to analyse your Twitter profile and tell you all about yourself, and it was often way off, sometimes on the nose, and always hilarious.

I tried to do that with my Mastodon profile, and it told me sorry, it is unable to access my Mastodon profile because it's been blocked.

in reply to Cassandrich

@notyourfanboy while I agree with the pushback against non-consensual scrapping, systems based only on good faith are unsustainable. Of course robbing a house is unethical and antisocial yet we still base the idea of home security expecting there will be bad actors that will do it.

Same as with scrapping public fediverse content, it is already probably happening, more extensively and way less visibly that’s someone openly advertising their actions.

in reply to tian2992

@tian2992 @notyourfanboy It's very much like the issue with glassholes. Of course anything that takes place in public/in open air *might* have cameras somewhere recording it. But you still punch a glasshole in the face and break their pervert glasses if you see one. Even if you are not stopping the harm entirely, you're maintaining the social norm that this is not okay.
in reply to Childless Bambino

The media in this post is not displayed to visitors. To view it, please go to the original post.

@kendricklmao8
They're greyed out - if I hover over the dots all it shows is a box saying Menu, and nothing else. I can't click on the box - it's not clickable.

Thanks for the advice, though. I'll try again later.

This entry was edited (2 months ago)
in reply to Cassandrich

The media in this post is not displayed to visitors. To view it, please go to the original post.

Blocking the author won't help because it's based on who is using the client. Author said they'd add noindex support, so you can add that flag and your posts should no longer show up.

If your instance doesn't support that, there's also #nobots. Disclaimer: I have not checked the code to see if this has been added yet

#nobots

This entry was edited (2 months ago)
in reply to dexternemrod

@dexternemrod A lot of folks don't understand how reporting works on fedi. When you report a post/account, you're reporting it to the moderation staff for *your own instance*, regardless of where the account you're reporting lives. Your mods can then block the account from appearing to anyone on your instance, limit it so it only shows up to ppl looking for it (kinda like a shadowban), or just keep track of the report for making future moderation decisions about the user or the instance they came from.

You can also choose to forward the report (anonymized to protect you from retaliation if the remote admins are aligned with the abuse, but even then they can sometimes guess) to the reported account's instance and/or the instances of anyone else included in the post's @'s. This might get them banned from their own instance if the instance owner doesn't approve of what they're doing, and it might help other instances block them if their instance doesn't act.

in reply to Cassandrich

@MHowell @dexternemrod See also:


@notyourfanboy "Welcome to the real world, where anyone can rob your house by smashing your windows and climbing inside."

Do you see how deranged that sounds?

We expect people not to do this because it's antisocial, illegal, and wrong. Not because there's some technical measure (which of course is impossible) preventing it.


in reply to Cassandrich

@jbaggs @dragonfrog One of the many things I appreciate about the fediverse is that people care about the social aspects of these systems. Like sure, obviously you *can* scrape everyone's posts and shovel them into the slop extruder, but you *ought* not to, and I trust that a large contingent of the fediverse feels the same way and generally wouldn't do that. "Just because you can doesn't mean you should" is very clearly alive and well here, based on the responses, and I love it.

reshared this

in reply to Cassandrich

I blocked and reported. It seems to me that mining folks' posts would be a violation of some server terms (I believe it is a violation of Kolektiva's terms). Many folks have their posts set to auto-delete, so this definitely seems like Seldo's product would be recording posts that would otherwise disappear. I also noticed that some folks from Kolektiva are following that account.

#LLMs #Fediverse #Datamining #AutoDelete #Safety #Security

in reply to adingbatponder 👾

@adingbatponder This person (seldo) made a web-based social media client that logs into your Mastodon/Bsky/etc. accounts, grabs everything in your feed, and sends it off to an "AI" service for them to turn into a "summary" so you can "catch up on what people are posting" without actually reading it.

When challenged about lack of consent to have our writing turned over to "AI" companies for them to process and keep and incorporate into their models, his response was very hostile and suggested blocking him so he can't see our posts. That only solves the problem for his use of his client, not the problem of other people potentially using it.

Unknown parent

mastodon - Link to source

Cassandrich

@IceWolf Of course anyone who can read them *can* do that. That's always possible. Just like, despite the movie industry pretending all the hellish DRM they impose makes it impossible, anyone can record movies off Netflix just by pointing a camera at their screen. It's impossible to keep someone from copying and sharing copies of something they can see.

This doesn't mean it's socially acceptable or legal. What the AI companies are doing and what the people like seldo feeding our data to them are doing is absolutely illegal, unethical, and a deep violation of privacy and consent. And they need to experience consequences for it.

Unknown parent

mastodon - Link to source

Cassandrich

@drwho @mek2600 @gunchleoc @YvanDaSilva @lawyersgunsnmoney Unless you have users on your instance using his AI-scraper client, he's not making any http connections to your instance.

He and his app's users are following you from their own instances, so your posts get federated there, and then they're letting the app log in to their instances and scrape everything from there.

The only use of IP blocking is if you're worried you have local users who might try out his app.