2023-01-24 06:09:19
2023-01-11 05:12:26
2023-01-11 05:11:44
1373000
!Friendica Support Do we have an idea...?
Why is my news feed schizophrenic?
My news feed on Mastadon/Fediverse has a split personality. Sometimes the feed is filled with a certain profile of people (teenager/cutesy/twee/animal pictures/lefty) sometimes another (more right/way older/etc). It looks like there's a bot or a series of bots switching between two sets of k means clustered sets of profiles and I have no idea how this is happening or a way to tailor what I want to see. How is the set of posts determined?
If this turns out to be algorithmically determined again like on Facebook I'm abandoning ship.
Spencer
•Given the tone and tenor of that person's other posts, they seem quick to anger and judgment. I hope they find answers to their question, and also, thank you for alerting me to someone I prefer to defederate from.
VegOS
•https://axbom.com/fediverse/
https://fediverse.party
Kristian (inactive/moved)
•@Spencer@Friendica Support
VegOS
•Kristian (inactive/moved)
•(a) Size. One of the frequently-annou... show more
(a) Size. One of the frequently-announced thing is that the Fediverse should grow around small instances, in best cases individual instances. Knowing this is difficult, this might fix your problem seen on instances > 20 people. Get yourself a small crowd of people you trust and roll your own, knowing to block or defederate servers that cause havoc or are annoying to you and yours.
(b) ChatGPT and recognizing ML generated content. Sure, I think it's more than likely that ChatGPT and similar tools could be "weaponized" in such a way. This feels threatening, but I don't see this to be something a protocol such as ActivityPub could ever prevent. Maybe you have some suggestions on how to handle that?
(c) I don't get that last part of your messages ("... it's because this software doesn't work as advertised"). What exactly do you mean by that?
@VegOS
VegOS
•Kristian (inactive/moved)
•Filtered word: nsfw
VegOS
•Kristian (inactive/moved)
•@VegOS
Hypolite Petovan
•…ᘛ⁐̤ᕐᐷ jesuisatire bitPickup
•@Kristian @VegOS @Peter Weyand
I absolutely can confirm Peter's claim, same over here.
Not going to discuss this, much less with some maybe bot that's set to make me loose time or energy.
The issue ends up on the instance level, big instances, no check on the users and abusers .. bad.
From that perspective @ Eugen has or might have some issues because the biggest instances are build by him. I recon that's good for several reasons, like testing and scaling, but if they become polluters, I'm going to do the same with them than I did with some chatBOT that just ran out of code when I called "him" out.
The world we live in is not Eugen's fault.
If this goes on I'll start to publicly call out sites that apparently are the worst representatives as unsave, block them and ask my respective admins and community to block them to.
Those are our tools and we'll have to work it out.
Anyone really believes "they" will allow us to happen just like that?
Kristian (inactive/moved)
•@VegOS
@Peter Weyand
Kristian (inactive/moved)
•This is an interesting take that made me think very much. Maybe that's a dimension of the whole ChatGPT and ML "mess" we didn't even consider so far - does the mere existence of these tool also call for a new line of "defense" in our reasoning and communication, claiming we can't or won't try clarifying our views to each other because there's a legit assumption we're "just" discussing with a bot? That's quite scary.
That aside, well, I agree on that last part: Those are our tools and we'll have to work it out. Following Peters statement, I somehow get the feeling that there's an assumption of possibilities offered by ActivityPub which the protocol simply can't provide, and most likely both because it was never meant... show more
This is an interesting take that made me think very much. Maybe that's a dimension of the whole ChatGPT and ML "mess" we didn't even consider so far - does the mere existence of these tool also call for a new line of "defense" in our reasoning and communication, claiming we can't or won't try clarifying our views to each other because there's a legit assumption we're "just" discussing with a bot? That's quite scary.
That aside, well, I agree on that last part: Those are our tools and we'll have to work it out. Following Peters statement, I somehow get the feeling that there's an assumption of possibilities offered by ActivityPub which the protocol simply can't provide, and most likely both because it was never meant to do so and because these are social not technical issues and need to be addressed on a social not technical level. So far, I have yet to see a valid solution for all this "trust" issue that works without having met a human being "in person" once and exchanged some sort of "credential" that allows for re-recognizing them online as well. But hasn't it been like this ever since? Was that different back then in the age of mailing lists or news groups? It doesn't seem new, just more ... difficult with the arise of ChatGPT et al...?
@VegOS
@Peter Weyand
Kristian (inactive/moved)
•@Hypolite Petovan@VegOS@Friendica Support@Peter Weyand
Kristian (inactive/moved)
•@Hypolite Petovan
@VegOS
…ᘛ⁐̤ᕐᐷ jesuisatire bitPickup
•Although the description of the problem itself is true, and the level of language quite outstanding, we are just facing same problems, same shit, in a different way.
For example, algos to feed us what we want to see (or the contrary) are normal on the other sites. So that "lady" normally just sell's you some ad's or lure you to sex.com. AI as in the means of understanding and semantics interpretation is a real threat and danger to society, I guess much worse in wallet gardens as out here where we are, where we might uncover them or evade them easier.
btw
friendica has also a positive server list as restriction too, that's most likely one of the features to have in mind also.
Kristian (inactive/moved)
•Geo Rg
•Kristian (inactive/moved)
•... show more
https://jlottosen.wordpress.com/2022/11/08/how-do-you-eat-a-mastodon/
https://jlottosen.files.wordpress.com/2022/11/12b1b5e24edb287e.jpg?w=743
Das sind Algorithmen, die haben Randbedingungen und Eigenschaften, und wenn ich die Eigenschaften kenne, kann ich manipulieren, indem ich mit diesen Eigenschaften arbeite. Beispielskonstrukt: Ich habe eine chronologische Timeline, also werden Posts nach Datum absteigend einsortiert. So eine Timeline kann ich potentiell "fluten", indem ich in loser Folge extrem viel poste - dann sehen Leute, die mich in ihrer Home-Timeline oder meine Instanz in ihrer förderierten Timeline haben, quasi nur das, was ich will, oder müssen extrem viel scrollen (Algorithmen bei Twitter, Instagram, Flickr ... unterbinden das ganz gut). Oder: Ich erzeuge eine Instanz, auf der nur Unsinn gepostet wird (etwa von Bots, die in hoher Frequenz viel Content generieren). Wenn ich dort einen plausiblen Nutzer bekomme, dem jemand anders auf einer anderen Instanz folgt (oder mehrere Nutzer auf anderen Instanzen erzeuge, die meinem Nutzer auf meiner bösen Instanz folgen), dann wird die förderierte Timeline der anderen Instanz auch diesen ganzen Kram anzeigen. Das ist vereinfacht abstrakt formuliert; ich hab keine Idee, ob und wie die gängigen Implementationen (Mastodon, Friendica, ...) das technisch irgendwie unterbinden.
Lösungen wären etwa: Suspekten Nutzern oder Nutzern von suspekten Instanzen nicht folgen. Oder diese Instanz stummschalten / blocken. Oder erkennbare Accounts, die zu viel posten, stummschalten. Oder ...? Herausforderung dabei: Erkennt man, ob der Content, der gepostet wird, "Unfug" ist oder nicht? Ich glaube, Kernprobleme sind hier grundsätzliche Erwartungshaltung an die zugrundeliegende Technologie und Systeme sowie schwierige Vorstellung vom Begriff "Algorithmus". In den Beiträgen bzw Fragen hier scheinen mir sehr viele Dinge durcheinanderzugehen, die irgendwie miteinander interagieren, aber potentiell verschiedene Probleme sind.
🕊️ Das Leben ist schön
•Das Blocken von Instanzen ist ein schlechtes Mittel. Das Blocken und/oder Stummschalten einzelner Accounts wird schnell zur Mammutaufgabe bei steigenden Nutzerzahlen, so handhabe ich das aber derzeit...
Kristian (inactive/moved)
•@Georg aus Bakum
Geo Rg
•Geo Rg
•Kristian (inactive/moved)
•Interessant etwa auch: https://www.jwz.org/blog/2022/11/mastodon-stampede/ : Wenn ich einen Link auf eine Website ins Fediverse poste, werden die Instanzen, die diesen Link sehen, den Server, auf den verlinkt wird, kontaktieren. Konsequenz: Habe ich einen Link, der in großem Umfang förderiert wird, greifen plötzlich potentiell ein paar tausend Instanzen völlig unkoordiniert auf den Server zu, auf den der Link zeigt. Wenn der Link auf einen Webserver auf einem RaspPi hinter dünner Leitung oder ein System wie https://solar.lowtechmagazine.com/ zeigt, kann das schon echt Stress bedeuten - oder umgekehrt die Anforderungen an Server- und Netzwerk-Ressourcen erhöhen, was wir aus verschiedenen Gründen (Stichwort mindestens Energiesparen) eigentlich nicht wollen.
Oder aber (hab den Link grad nicht mehr zur Hand) die Bitte eines Mastodon-Nutzers, große Videos bitte nicht über Mastodon zu teilen, weil das zur Konsequenz hat, dass die Videos (wie auch Bilder) vielfach zwischen Instanzen übertragen, lokal gespeichert werden und damit die Systeme voll-laufen lassen.
Das beides sind schon Dinge, die weder überraschen noch erschrecken, die aus meiner Sicht auch nicht grundsätzlich schlimm oder unbeherrschbar sind - aber eben Themen, an denen "zentrale" Systeme gewisse Vorteile haben können und mit denen sich die Fediverse-Community vermutlich irgendwann irgendwie beschäftigen will. Und die Frage, inwieweit Botnetze irgendwelchen Stuss tun können, indem sie die Systeme mit Content fluten, den man nicht will und maschinell schlecht filtern kann, wird möglicherweise auch dazugehören - siehe etwa auch die Qualität, die manche Spam- und Phishing-E-Mails dieser Tage haben und die so "gut" sind, dass diese Nachrichten selbst durch gute automatische Filter rutschen und Menschen Geld an Unbekannte überweisen.
@Das Leben ist schön@Peter Weyand
Geo Rg
•Geo Rg
•Geo Rg
•Geo Rg
•Geo Rg
•Kristian (inactive/moved)
•Kristian (inactive/moved)
•Geo Rg
•Geo Rg
•Geo Rg
•Kristian (inactive/moved)
•@Georg aus Bakum
Kristian (inactive/moved)
•As for that other part, that idea of anonymously viewing content, that somehow feels funny because it seems to completely oppose that other idea - if I want to have a system very keen on accessing content anonymously, isn't it the very idea to leave as little clues as possible about who a particular actor is? For that example you described, this is one of the arguments quite some of the fediverse propoments come up with when talking about different instances dedicated to particular subtopics, social communities, ... : Of course you don't want to have just one account in there everything is tied to. You want several accounts for things you want to keep separated, much alike most people I know (at least in tech) do have at least two Twitter accounts - one for business and professional stuff, one for personal stuff, and they are _very_ careful not to have any links between them. And I think the fediverse actually is rather good here - it's pretty easy to have accounts that are as anonymous as they probably could be. So most likely you won't block people just to make sure they don't see which porn you're looking at - why even bother if you can be reasonably sure no one else out there knows who "you" in this particular case are? If you do that, however, you end up with your initial problem again - you have no idea who hides behind a particular account, whether it's the person it claims to be, or whether it's a person at all. That's a bit unsatifying, but then again.... hasn't it always been this way? In usenet? In IRC? On mailing lists? Maybe the "solution" here is to, for situation in which it matters, choose tools that do the job as good as they can - and in others, know as much as possible about the shortcomings of a particular tool to still act in a responsible and safe manner? I don't have any ideas that improve that without making it much worse at the same time...
@Georg aus Bakum
Der Marek
•https://en.wikipedia.org/wiki/Turing_test
…ᘛ⁐̤ᕐᐷ jesuisatire bitPickup
•> The Activity Pub algorithm enables this by its' design.
As much as anything else that creates something out of nothing.
Sry but you end up talking nonsense, truth as such never has existed except of in terms of cultural interpretation demanding something to be absolutely true and imposing that supposed truth upon others.
A decentralized setup in any case gives you the chance to discover "the brick wall at the back of the theater".
Centralized setups even deny you that certainty or chance.
---
and yes, I'm a bot too.