Salta al contenuto
0
  • Home
  • Piero Bosio
  • Blog
  • Mondo
  • Fediverso
  • News
  • Categorie
  • Old Web Site
  • Recenti
  • Popolare
  • Tag
  • Utenti
  • Home
  • Piero Bosio
  • Blog
  • Mondo
  • Fediverso
  • News
  • Categorie
  • Old Web Site
  • Recenti
  • Popolare
  • Tag
  • Utenti
Skin
  • Chiaro
  • Brite
  • Cerulean
  • Cosmo
  • Flatly
  • Journal
  • Litera
  • Lumen
  • Lux
  • Materia
  • Minty
  • Morph
  • Pulse
  • Sandstone
  • Simplex
  • Sketchy
  • Spacelab
  • United
  • Yeti
  • Zephyr
  • Scuro
  • Cyborg
  • Darkly
  • Quartz
  • Slate
  • Solar
  • Superhero
  • Vapor

  • Predefinito (Nessuna skin)
  • Nessuna skin
Collassa

Piero Bosio Social Web Site Personale Logo Fediverso

Social Forum federato con il resto del mondo. Non contano le istanze, contano le persone
  1. Home
  2. Categorie
  3. Fediverse
  4. AI-assisted moderation in the fediverse is happening. Now what?

AI-assisted moderation in the fediverse is happening. Now what?

Pianificato Fissato Bloccato Spostato Fediverse
fediversemoderation
1 Post multipli 177 Post 88 Autori 28 Visualizzazioni
  • Da Vecchi a Nuovi
  • Da Nuovi a Vecchi
  • Più Voti
Rispondi
  • Risposta alla discussione
Effettua l'accesso per rispondere
Questa discussione è stata eliminata. Solo gli utenti con diritti di gestione possono vederla.
  • minfapper@piefed.social minfapper@piefed.social

    Not from a legal perspective

    magnum@infosec.pub
    magnum@infosec.pub
    magnum@infosec.pub
    scritto su ultima modifica di
    #141

    It is. The federated Data is the same public data available through the website but delivered in a protocol for interoperability. It is not non public accessible data they share with a third party to compute or profit further.

    1 Risposta Ultima Risposta
    0
    • fratermus@piefed.social fratermus@piefed.social

      They do this by running a script/app that sends the user’s comment history to OpenAI with the question “analyze this content for evidence of specific political ideology sentiment.

      To me the problem is what they are looking for not how they are doing it. Thought experiment: in what way would it be qualitatively different if they hired a team of people in Upper Elbonia to do the same thing?

      vagrancyand@sh.itjust.works
      vagrancyand@sh.itjust.works
      vagrancyand@sh.itjust.works
      scritto su ultima modifica di
      #142

      Some political ideologies and opinions are objectively not okay to exist in modern society, and those who hold those opinions should be eliminated. This isn't a controversial opinion outside a tiny bubble in the US who think deep down every nazi is just a misunderstood cash cow waiting to be secreted away by the CIA.

      1 Risposta Ultima Risposta
      0
      • rimu@piefed.social rimu@piefed.social

        I recently discovered that some popular federated instances have been using LLM-assisted moderation tooling that evaluates whether someone has said something bannable. They do this by running a script/app that sends the user’s comment history to OpenAI with the question “analyze this content for evidence of specific political ideology sentiment. Also identify any related political ideology tropes“.

        OpenAI’s LLM (they’re using GPT-5.3-mini) then responds with something like:

        image

        and so on, hundreds of comments.

        I have not named the instances or people involved, to give them time to consider the results of this discussion, make any corrective changes they want and disclose their practices at their own pace and in their own way. I have also redacted the evidence to avoid personal attacks and dogpiling. Let’s focus on the system, not the individuals involved. Today these instances and people are using it and maybe we’re ok with that because it’s being used by groups we agree with but what if people we strongly disagree with used it on their instances tomorrow?

        The use and existence of this tooling raises a lot of other questions too.

        What are the risks? Fedi moderators are often unsupervised, untrained volunteers and these are powerful tools.

        What safeguards do we need?

        Would asking a LLM “please evaluate this person’s political opinions” give different results than “find evidence we can use to ban them” (as used in the cases I’ve seen)?

        What are our transparency expectations?

        Is this acceptable and normal?

        Should this tooling be disclosed? (it was not – should it have been?)

        If you were given a choice, would you have opted out of it?

        Can we opt out?

        Are there GDPR implications? Privacy implications? Should these tools be described in a privacy policy?

        Are private messages being scanned and sent to OpenAI?

        How long should these assessments be retained and can we request to see it, or ask for it to be deleted?

        Once the user’s comments are sent to OpenAI, is it used to train their models?

        What will the effect be on our discourse and culture if people know they are being politically profiled?

        Where are the lines between normal moderation assistance tools, political profiling and opaque 3rd-party data processing?

        I hope that by chewing over these questions we can begin to establish some norms and expectations around this technology. The fediverse doesn’t have any centralized enforcement so we need discussions like this to develop an awareness of what people want in terms of disclosure, privacy, consent and acceptable use. Then people can make choices about which instances they join and which ones they interact with remotely.

        And of course there are the other issues with LLMs relating to environmental sustainability, erosion of worker’s rights, increasing the cost of living and on and on. I can’t see PieFed adding any functionality like this anytime soon. But it’s happening out there anyway so now we need to talk about it.

        What do you make of this?

        jankatarch@lemmy.world
        jankatarch@lemmy.world
        jankatarch@lemmy.world
        scritto su ultima modifica di
        #143

        Not comfortable with this. Not at all.

        1 Risposta Ultima Risposta
        0
        • rimu@piefed.social rimu@piefed.social

          I recently discovered that some popular federated instances have been using LLM-assisted moderation tooling that evaluates whether someone has said something bannable. They do this by running a script/app that sends the user’s comment history to OpenAI with the question “analyze this content for evidence of specific political ideology sentiment. Also identify any related political ideology tropes“.

          OpenAI’s LLM (they’re using GPT-5.3-mini) then responds with something like:

          image

          and so on, hundreds of comments.

          I have not named the instances or people involved, to give them time to consider the results of this discussion, make any corrective changes they want and disclose their practices at their own pace and in their own way. I have also redacted the evidence to avoid personal attacks and dogpiling. Let’s focus on the system, not the individuals involved. Today these instances and people are using it and maybe we’re ok with that because it’s being used by groups we agree with but what if people we strongly disagree with used it on their instances tomorrow?

          The use and existence of this tooling raises a lot of other questions too.

          What are the risks? Fedi moderators are often unsupervised, untrained volunteers and these are powerful tools.

          What safeguards do we need?

          Would asking a LLM “please evaluate this person’s political opinions” give different results than “find evidence we can use to ban them” (as used in the cases I’ve seen)?

          What are our transparency expectations?

          Is this acceptable and normal?

          Should this tooling be disclosed? (it was not – should it have been?)

          If you were given a choice, would you have opted out of it?

          Can we opt out?

          Are there GDPR implications? Privacy implications? Should these tools be described in a privacy policy?

          Are private messages being scanned and sent to OpenAI?

          How long should these assessments be retained and can we request to see it, or ask for it to be deleted?

          Once the user’s comments are sent to OpenAI, is it used to train their models?

          What will the effect be on our discourse and culture if people know they are being politically profiled?

          Where are the lines between normal moderation assistance tools, political profiling and opaque 3rd-party data processing?

          I hope that by chewing over these questions we can begin to establish some norms and expectations around this technology. The fediverse doesn’t have any centralized enforcement so we need discussions like this to develop an awareness of what people want in terms of disclosure, privacy, consent and acceptable use. Then people can make choices about which instances they join and which ones they interact with remotely.

          And of course there are the other issues with LLMs relating to environmental sustainability, erosion of worker’s rights, increasing the cost of living and on and on. I can’t see PieFed adding any functionality like this anytime soon. But it’s happening out there anyway so now we need to talk about it.

          What do you make of this?

          arachnidbrilliant@lemmy.dbzer0.com
          arachnidbrilliant@lemmy.dbzer0.com
          arachnidbrilliant@lemmy.dbzer0.com
          scritto su ultima modifica di
          #144

          Well, it was fun while it lasted..

          1 Risposta Ultima Risposta
          0
          • rimu@piefed.social rimu@piefed.social

            I recently discovered that some popular federated instances have been using LLM-assisted moderation tooling that evaluates whether someone has said something bannable. They do this by running a script/app that sends the user’s comment history to OpenAI with the question “analyze this content for evidence of specific political ideology sentiment. Also identify any related political ideology tropes“.

            OpenAI’s LLM (they’re using GPT-5.3-mini) then responds with something like:

            image

            and so on, hundreds of comments.

            I have not named the instances or people involved, to give them time to consider the results of this discussion, make any corrective changes they want and disclose their practices at their own pace and in their own way. I have also redacted the evidence to avoid personal attacks and dogpiling. Let’s focus on the system, not the individuals involved. Today these instances and people are using it and maybe we’re ok with that because it’s being used by groups we agree with but what if people we strongly disagree with used it on their instances tomorrow?

            The use and existence of this tooling raises a lot of other questions too.

            What are the risks? Fedi moderators are often unsupervised, untrained volunteers and these are powerful tools.

            What safeguards do we need?

            Would asking a LLM “please evaluate this person’s political opinions” give different results than “find evidence we can use to ban them” (as used in the cases I’ve seen)?

            What are our transparency expectations?

            Is this acceptable and normal?

            Should this tooling be disclosed? (it was not – should it have been?)

            If you were given a choice, would you have opted out of it?

            Can we opt out?

            Are there GDPR implications? Privacy implications? Should these tools be described in a privacy policy?

            Are private messages being scanned and sent to OpenAI?

            How long should these assessments be retained and can we request to see it, or ask for it to be deleted?

            Once the user’s comments are sent to OpenAI, is it used to train their models?

            What will the effect be on our discourse and culture if people know they are being politically profiled?

            Where are the lines between normal moderation assistance tools, political profiling and opaque 3rd-party data processing?

            I hope that by chewing over these questions we can begin to establish some norms and expectations around this technology. The fediverse doesn’t have any centralized enforcement so we need discussions like this to develop an awareness of what people want in terms of disclosure, privacy, consent and acceptable use. Then people can make choices about which instances they join and which ones they interact with remotely.

            And of course there are the other issues with LLMs relating to environmental sustainability, erosion of worker’s rights, increasing the cost of living and on and on. I can’t see PieFed adding any functionality like this anytime soon. But it’s happening out there anyway so now we need to talk about it.

            What do you make of this?

            krusty@quokk.au
            krusty@quokk.au
            krusty@quokk.au
            scritto su ultima modifica di
            #145

            Fedislopslopslopslopslop...

            1 Risposta Ultima Risposta
            0
            • rimu@piefed.social rimu@piefed.social

              I recently discovered that some popular federated instances have been using LLM-assisted moderation tooling that evaluates whether someone has said something bannable. They do this by running a script/app that sends the user’s comment history to OpenAI with the question “analyze this content for evidence of specific political ideology sentiment. Also identify any related political ideology tropes“.

              OpenAI’s LLM (they’re using GPT-5.3-mini) then responds with something like:

              image

              and so on, hundreds of comments.

              I have not named the instances or people involved, to give them time to consider the results of this discussion, make any corrective changes they want and disclose their practices at their own pace and in their own way. I have also redacted the evidence to avoid personal attacks and dogpiling. Let’s focus on the system, not the individuals involved. Today these instances and people are using it and maybe we’re ok with that because it’s being used by groups we agree with but what if people we strongly disagree with used it on their instances tomorrow?

              The use and existence of this tooling raises a lot of other questions too.

              What are the risks? Fedi moderators are often unsupervised, untrained volunteers and these are powerful tools.

              What safeguards do we need?

              Would asking a LLM “please evaluate this person’s political opinions” give different results than “find evidence we can use to ban them” (as used in the cases I’ve seen)?

              What are our transparency expectations?

              Is this acceptable and normal?

              Should this tooling be disclosed? (it was not – should it have been?)

              If you were given a choice, would you have opted out of it?

              Can we opt out?

              Are there GDPR implications? Privacy implications? Should these tools be described in a privacy policy?

              Are private messages being scanned and sent to OpenAI?

              How long should these assessments be retained and can we request to see it, or ask for it to be deleted?

              Once the user’s comments are sent to OpenAI, is it used to train their models?

              What will the effect be on our discourse and culture if people know they are being politically profiled?

              Where are the lines between normal moderation assistance tools, political profiling and opaque 3rd-party data processing?

              I hope that by chewing over these questions we can begin to establish some norms and expectations around this technology. The fediverse doesn’t have any centralized enforcement so we need discussions like this to develop an awareness of what people want in terms of disclosure, privacy, consent and acceptable use. Then people can make choices about which instances they join and which ones they interact with remotely.

              And of course there are the other issues with LLMs relating to environmental sustainability, erosion of worker’s rights, increasing the cost of living and on and on. I can’t see PieFed adding any functionality like this anytime soon. But it’s happening out there anyway so now we need to talk about it.

              What do you make of this?

              j_z@feddit.nu
              j_z@feddit.nu
              j_z@feddit.nu
              scritto su ultima modifica di
              #146

              I guess, given the already open nature of the Fediverse, my takeaway from this thread is that op is using their freedom to say they don’t like this particular style of moderation. Which might be useful, or not, for some moderators

              totallynotjessica@lemmy.blahaj.zone 1 Risposta Ultima Risposta
              0
              • alcoholicorn@mander.xyz alcoholicorn@mander.xyz

                Mods of any instance you're federated with can do this

                sortekanin@feddit.dk
                sortekanin@feddit.dk
                sortekanin@feddit.dk
                scritto su ultima modifica di
                #147

                They don't need to federate even, they're probably already scraping everything.

                1 Risposta Ultima Risposta
                0
                • rimu@piefed.social rimu@piefed.social

                  I recently discovered that some popular federated instances have been using LLM-assisted moderation tooling that evaluates whether someone has said something bannable. They do this by running a script/app that sends the user’s comment history to OpenAI with the question “analyze this content for evidence of specific political ideology sentiment. Also identify any related political ideology tropes“.

                  OpenAI’s LLM (they’re using GPT-5.3-mini) then responds with something like:

                  image

                  and so on, hundreds of comments.

                  I have not named the instances or people involved, to give them time to consider the results of this discussion, make any corrective changes they want and disclose their practices at their own pace and in their own way. I have also redacted the evidence to avoid personal attacks and dogpiling. Let’s focus on the system, not the individuals involved. Today these instances and people are using it and maybe we’re ok with that because it’s being used by groups we agree with but what if people we strongly disagree with used it on their instances tomorrow?

                  The use and existence of this tooling raises a lot of other questions too.

                  What are the risks? Fedi moderators are often unsupervised, untrained volunteers and these are powerful tools.

                  What safeguards do we need?

                  Would asking a LLM “please evaluate this person’s political opinions” give different results than “find evidence we can use to ban them” (as used in the cases I’ve seen)?

                  What are our transparency expectations?

                  Is this acceptable and normal?

                  Should this tooling be disclosed? (it was not – should it have been?)

                  If you were given a choice, would you have opted out of it?

                  Can we opt out?

                  Are there GDPR implications? Privacy implications? Should these tools be described in a privacy policy?

                  Are private messages being scanned and sent to OpenAI?

                  How long should these assessments be retained and can we request to see it, or ask for it to be deleted?

                  Once the user’s comments are sent to OpenAI, is it used to train their models?

                  What will the effect be on our discourse and culture if people know they are being politically profiled?

                  Where are the lines between normal moderation assistance tools, political profiling and opaque 3rd-party data processing?

                  I hope that by chewing over these questions we can begin to establish some norms and expectations around this technology. The fediverse doesn’t have any centralized enforcement so we need discussions like this to develop an awareness of what people want in terms of disclosure, privacy, consent and acceptable use. Then people can make choices about which instances they join and which ones they interact with remotely.

                  And of course there are the other issues with LLMs relating to environmental sustainability, erosion of worker’s rights, increasing the cost of living and on and on. I can’t see PieFed adding any functionality like this anytime soon. But it’s happening out there anyway so now we need to talk about it.

                  What do you make of this?

                  fedegenerate@fedinsfw.app
                  fedegenerate@fedinsfw.app
                  fedegenerate@fedinsfw.app
                  scritto su ultima modifica di
                  #148

                  I think that's an acceptable price of being on the open and free internet.

                  Everything I do is in the open, and you are free to do as you wish with it. Likewise, everything you do is in the open and you accept, by being here, that I am free to do what I wish with it.

                  Defederation is likely what you're looking for, find an instance that defederates GenAI friendly instances.

                  1 Risposta Ultima Risposta
                  0
                  • goferking0@ttrpg.network goferking0@ttrpg.network

                    Or just delete them entirely from piefed.social social 😂

                    blaze@piefed.zip
                    blaze@piefed.zip
                    blaze@piefed.zip
                    scritto su ultima modifica di
                    #149

                    Those removal happened in the context of a mod calling Rimu a zionist (which he's not)

                    It didn't happen out of nowhere.

                    goferking0@ttrpg.network 1 Risposta Ultima Risposta
                    0
                    • azzu@lemmy.dbzer0.com azzu@lemmy.dbzer0.com

                      I'm not sure that is true. As far as I know, GDPR only covers personally identifiable information, if the data is completely anonymous, and can not be traced back to your person, it is fine to use under the GDPR. This is a common misconception about the GDPR.

                      So the real question is if your comments can be used to identify you personally. This might be the case, or it might not.

                      general_effort@lemmy.world
                      general_effort@lemmy.world
                      general_effort@lemmy.world
                      scritto su ultima modifica di
                      #150

                      No. You're thinking of California law. PII is not a thing in the GDPR.

                      1 Risposta Ultima Risposta
                      0
                      • rimu@piefed.social rimu@piefed.social

                        I recently discovered that some popular federated instances have been using LLM-assisted moderation tooling that evaluates whether someone has said something bannable. They do this by running a script/app that sends the user’s comment history to OpenAI with the question “analyze this content for evidence of specific political ideology sentiment. Also identify any related political ideology tropes“.

                        OpenAI’s LLM (they’re using GPT-5.3-mini) then responds with something like:

                        image

                        and so on, hundreds of comments.

                        I have not named the instances or people involved, to give them time to consider the results of this discussion, make any corrective changes they want and disclose their practices at their own pace and in their own way. I have also redacted the evidence to avoid personal attacks and dogpiling. Let’s focus on the system, not the individuals involved. Today these instances and people are using it and maybe we’re ok with that because it’s being used by groups we agree with but what if people we strongly disagree with used it on their instances tomorrow?

                        The use and existence of this tooling raises a lot of other questions too.

                        What are the risks? Fedi moderators are often unsupervised, untrained volunteers and these are powerful tools.

                        What safeguards do we need?

                        Would asking a LLM “please evaluate this person’s political opinions” give different results than “find evidence we can use to ban them” (as used in the cases I’ve seen)?

                        What are our transparency expectations?

                        Is this acceptable and normal?

                        Should this tooling be disclosed? (it was not – should it have been?)

                        If you were given a choice, would you have opted out of it?

                        Can we opt out?

                        Are there GDPR implications? Privacy implications? Should these tools be described in a privacy policy?

                        Are private messages being scanned and sent to OpenAI?

                        How long should these assessments be retained and can we request to see it, or ask for it to be deleted?

                        Once the user’s comments are sent to OpenAI, is it used to train their models?

                        What will the effect be on our discourse and culture if people know they are being politically profiled?

                        Where are the lines between normal moderation assistance tools, political profiling and opaque 3rd-party data processing?

                        I hope that by chewing over these questions we can begin to establish some norms and expectations around this technology. The fediverse doesn’t have any centralized enforcement so we need discussions like this to develop an awareness of what people want in terms of disclosure, privacy, consent and acceptable use. Then people can make choices about which instances they join and which ones they interact with remotely.

                        And of course there are the other issues with LLMs relating to environmental sustainability, erosion of worker’s rights, increasing the cost of living and on and on. I can’t see PieFed adding any functionality like this anytime soon. But it’s happening out there anyway so now we need to talk about it.

                        What do you make of this?

                        general_effort@lemmy.world
                        general_effort@lemmy.world
                        general_effort@lemmy.world
                        scritto su ultima modifica di
                        #151

                        Are there GDPR implications?

                        Hahaha. Oh boy...

                        This is amazingly illegal. The emotional damages might be high enough to make it worthwhile to sue. Neither Meta nor almost any of the other companies that users here love to hate, have ever done anything even remotely as bad.

                        1 Risposta Ultima Risposta
                        0
                        • blaze@piefed.zip blaze@piefed.zip

                          Those removal happened in the context of a mod calling Rimu a zionist (which he's not)

                          It didn't happen out of nowhere.

                          goferking0@ttrpg.network
                          goferking0@ttrpg.network
                          goferking0@ttrpg.network
                          scritto su ultima modifica di
                          #152

                          He's been doing that for a long time.

                          Those removal happened in the context of a mod calling Rimu a zionist.

                          Is that supposed to make it better?

                          blaze@piefed.zip 1 Risposta Ultima Risposta
                          0
                          • rimu@piefed.social rimu@piefed.social

                            I recently discovered that some popular federated instances have been using LLM-assisted moderation tooling that evaluates whether someone has said something bannable. They do this by running a script/app that sends the user’s comment history to OpenAI with the question “analyze this content for evidence of specific political ideology sentiment. Also identify any related political ideology tropes“.

                            OpenAI’s LLM (they’re using GPT-5.3-mini) then responds with something like:

                            image

                            and so on, hundreds of comments.

                            I have not named the instances or people involved, to give them time to consider the results of this discussion, make any corrective changes they want and disclose their practices at their own pace and in their own way. I have also redacted the evidence to avoid personal attacks and dogpiling. Let’s focus on the system, not the individuals involved. Today these instances and people are using it and maybe we’re ok with that because it’s being used by groups we agree with but what if people we strongly disagree with used it on their instances tomorrow?

                            The use and existence of this tooling raises a lot of other questions too.

                            What are the risks? Fedi moderators are often unsupervised, untrained volunteers and these are powerful tools.

                            What safeguards do we need?

                            Would asking a LLM “please evaluate this person’s political opinions” give different results than “find evidence we can use to ban them” (as used in the cases I’ve seen)?

                            What are our transparency expectations?

                            Is this acceptable and normal?

                            Should this tooling be disclosed? (it was not – should it have been?)

                            If you were given a choice, would you have opted out of it?

                            Can we opt out?

                            Are there GDPR implications? Privacy implications? Should these tools be described in a privacy policy?

                            Are private messages being scanned and sent to OpenAI?

                            How long should these assessments be retained and can we request to see it, or ask for it to be deleted?

                            Once the user’s comments are sent to OpenAI, is it used to train their models?

                            What will the effect be on our discourse and culture if people know they are being politically profiled?

                            Where are the lines between normal moderation assistance tools, political profiling and opaque 3rd-party data processing?

                            I hope that by chewing over these questions we can begin to establish some norms and expectations around this technology. The fediverse doesn’t have any centralized enforcement so we need discussions like this to develop an awareness of what people want in terms of disclosure, privacy, consent and acceptable use. Then people can make choices about which instances they join and which ones they interact with remotely.

                            And of course there are the other issues with LLMs relating to environmental sustainability, erosion of worker’s rights, increasing the cost of living and on and on. I can’t see PieFed adding any functionality like this anytime soon. But it’s happening out there anyway so now we need to talk about it.

                            What do you make of this?

                            emmy67@lemmy.world
                            emmy67@lemmy.world
                            emmy67@lemmy.world
                            scritto su ultima modifica di
                            #153

                            I hate the use of LLMs. But I hate Nazis more. If this targets them or associated ideology. I can get behind that

                            1 Risposta Ultima Risposta
                            0
                            • goferking0@ttrpg.network goferking0@ttrpg.network

                              He's been doing that for a long time.

                              Those removal happened in the context of a mod calling Rimu a zionist.

                              Is that supposed to make it better?

                              blaze@piefed.zip
                              blaze@piefed.zip
                              blaze@piefed.zip
                              scritto su ultima modifica di
                              #154

                              Is that supposed to make it better?

                              Rimu is not a Zionist. It's this wrong accusation that escalated the tensions. I'll clarify my comment.

                              goferking0@ttrpg.network 1 Risposta Ultima Risposta
                              0
                              • db0@lemmy.dbzer0.com db0@lemmy.dbzer0.com

                                You talk about instances utilizing this tooling, but in your comments you admit it's just some mods. This is misleading, as talking about instances doing it assumes admin access and relevant instance policy, something which invites calls for defederation (as can be clearly be seen from the comments in your post).

                                A random mod doing something is not the same as an instance doing it. Literally anyone can be a mod and they don't get any more access than an anonymous account by doing so.

                                This is the second time in one week I see you throwing careless statements like chum in the water. I can't help but notice a pattern emerging.

                                rimu@piefed.social
                                rimu@piefed.social
                                rimu@piefed.social
                                scritto su ultima modifica di
                                #155

                                Ahh sorry - I just checked more closely. It's an admin doing it.

                                1 Risposta Ultima Risposta
                                0
                                • blaze@piefed.zip blaze@piefed.zip

                                  Is that supposed to make it better?

                                  Rimu is not a Zionist. It's this wrong accusation that escalated the tensions. I'll clarify my comment.

                                  goferking0@ttrpg.network
                                  goferking0@ttrpg.network
                                  goferking0@ttrpg.network
                                  scritto su ultima modifica di
                                  #156

                                  That didn't address anything even with the clarification as it's his go to response. He's been doing it since the instance was stood up

                                  blaze@piefed.zip 1 Risposta Ultima Risposta
                                  0
                                  • rimu@piefed.social rimu@piefed.social

                                    It can be traced objectively.

                                    It's mods.

                                    They're using software that they made to do this quickly and easily, it's not a manual copy and paste situation.

                                    flatworm7591@lemmy.dbzer0.com
                                    flatworm7591@lemmy.dbzer0.com
                                    flatworm7591@lemmy.dbzer0.com
                                    scritto su ultima modifica di
                                    #157

                                    Sure bro, keep on making shit up, I guess? By the way I have removed all the community bans for yourself and Skavau. You only had to ask btw. Might have saved a lot of pointless drama.

                                    1 Risposta Ultima Risposta
                                    0
                                    • rimu@piefed.social rimu@piefed.social

                                      I recently discovered that some popular federated instances have been using LLM-assisted moderation tooling that evaluates whether someone has said something bannable. They do this by running a script/app that sends the user’s comment history to OpenAI with the question “analyze this content for evidence of specific political ideology sentiment. Also identify any related political ideology tropes“.

                                      OpenAI’s LLM (they’re using GPT-5.3-mini) then responds with something like:

                                      image

                                      and so on, hundreds of comments.

                                      I have not named the instances or people involved, to give them time to consider the results of this discussion, make any corrective changes they want and disclose their practices at their own pace and in their own way. I have also redacted the evidence to avoid personal attacks and dogpiling. Let’s focus on the system, not the individuals involved. Today these instances and people are using it and maybe we’re ok with that because it’s being used by groups we agree with but what if people we strongly disagree with used it on their instances tomorrow?

                                      The use and existence of this tooling raises a lot of other questions too.

                                      What are the risks? Fedi moderators are often unsupervised, untrained volunteers and these are powerful tools.

                                      What safeguards do we need?

                                      Would asking a LLM “please evaluate this person’s political opinions” give different results than “find evidence we can use to ban them” (as used in the cases I’ve seen)?

                                      What are our transparency expectations?

                                      Is this acceptable and normal?

                                      Should this tooling be disclosed? (it was not – should it have been?)

                                      If you were given a choice, would you have opted out of it?

                                      Can we opt out?

                                      Are there GDPR implications? Privacy implications? Should these tools be described in a privacy policy?

                                      Are private messages being scanned and sent to OpenAI?

                                      How long should these assessments be retained and can we request to see it, or ask for it to be deleted?

                                      Once the user’s comments are sent to OpenAI, is it used to train their models?

                                      What will the effect be on our discourse and culture if people know they are being politically profiled?

                                      Where are the lines between normal moderation assistance tools, political profiling and opaque 3rd-party data processing?

                                      I hope that by chewing over these questions we can begin to establish some norms and expectations around this technology. The fediverse doesn’t have any centralized enforcement so we need discussions like this to develop an awareness of what people want in terms of disclosure, privacy, consent and acceptable use. Then people can make choices about which instances they join and which ones they interact with remotely.

                                      And of course there are the other issues with LLMs relating to environmental sustainability, erosion of worker’s rights, increasing the cost of living and on and on. I can’t see PieFed adding any functionality like this anytime soon. But it’s happening out there anyway so now we need to talk about it.

                                      What do you make of this?

                                      freefall@lemmy.world
                                      freefall@lemmy.world
                                      freefall@lemmy.world
                                      scritto su ultima modifica di
                                      #158

                                      Ai use should be against the rules for mods. That shit is poison.

                                      1 Risposta Ultima Risposta
                                      0
                                      • alvaro@lemmy.blahaj.zone alvaro@lemmy.blahaj.zone

                                        Wthout going into the issue itself, it is such a ridiculous waste to use an llm for something that a far simpler model could do like 100x faster and locally for essentially free...

                                        Just search for "machine learning text moderation" and you will find all kinds of options. Not to talk about the fact that a simple 4B LLM could do this as well.

                                        One thing I really hate is how LLMs have completely overshadowed the entire ML/AI field and people just use them for everything.

                                        Using a trillion parameter LLM model for basic text moderation is like using a gaming rig to play candy crush.

                                        sl00k@programming.dev
                                        sl00k@programming.dev
                                        sl00k@programming.dev
                                        scritto su ultima modifica di
                                        #159

                                        I've been doing a lot of work around this area and the issue tends to pop up around context. There are instances a haiku model will catch something far more accurately than gpt-oss-safeguard. Obviously you pay in costs, realistically, you need a full system here for a proper implementation so some flows to tiny LLM some flows exclusively ML, some flows to a higher intelligence.

                                        People on the fediverse are generally pretty anti ai but it's basically impossible to scale a platform without AI moderation. I would fully welcome any instance trying an implementation of [Osprey](https://discord.com/blog/osprey-open-sourcing-our-rule-engine with LLMs.

                                        1 Risposta Ultima Risposta
                                        0
                                        • loco_mex@sh.itjust.works loco_mex@sh.itjust.works

                                          I’m fine with a general overview. Leaving the whole desicion to a bot is not a good design choice, but AI could help on summarising things.

                                          I suspect that is the case and what we will find once Rimu stops drama farming and says who it is.

                                          flatworm7591@lemmy.dbzer0.com
                                          flatworm7591@lemmy.dbzer0.com
                                          flatworm7591@lemmy.dbzer0.com
                                          scritto su ultima modifica di
                                          #160

                                          That is exactly the case. I really can't be bothered reading this whole fucking post, so I'm gonna just reply to you, since you sound reasonably sane.

                                          First, it should be obvious to everyone that their post and comment histories are completely public, and accessible to anyone on the fediverse. There is zero privacy on the fediverse. Your comment histories have already been scraped a thousand times over by every big model out there.

                                          All we have at the moment is a simple script, developed by one of our mods (who will be publishing it on codeberg soon), that any user can run that logs into lemmy using your own account, and downloads a set number (or time period) of comments into a text file. There is no abuse of admin powers going on, it's just the stock lemmy/piefed api. This is massively faster than manually paging through comment histories on lemmy, and can make mod decisions more robust and more informed.

                                          Using the text file, mods or admins can quickly search for keywords or whatever, using a simple text editor, or simply skim read it. Another option is to ingest it into an LLM to provide a summary. I tried doing that just a handful of times, for testing, but honestly I found it a bit cumbersome and who knows if the summary is actually accurate given the tendency for hallucinations? A couple of tests seemed consistent with my own assessment, and a couple were way off base.

                                          That told me everything I need to know about how much to trust the summaries.... very little. I honestly don't think they are much of a value add, because you just can't reliably trust the results. In any case, we have absolutely no plans to use llms for that on a regular basis, I just wanted to see what it came up with, and how well it matched a manual assessment.

                                          And to also clarify, there is/was absolutely no automated scanning of users. The process is the same as always. We get a report, we investigate the report, and make a human mod decision. The only difference in this case is that the investigation can be done more efficiently, because we don't have to go slowly paging through comments and searching through the Lemmy UI for the relevant data.

                                          Obviously as well, no mod or admin is gonna download the entire comments history of a user unless it is a complicated report that is difficult to get to the bottom of from a quick look at the report. 90%+ of mod actions would never need that much detail.

                                          The way OPs post was written was obviously designed from the ground up to stir up drama about AI use. Honestly, I don't know why he's still malding, but Rimu seems to be engaging in a lot of very bad faith hit pieces at the moment, designed solely to stir up drama, and directed at our instance. It's really shitty behaviour.

                                          rimu@piefed.social 1 Risposta Ultima Risposta
                                          0

                                          Ciao! Sembra che tu sia interessato a questa conversazione, ma non hai ancora un account.

                                          Stanco di dover scorrere gli stessi post a ogni visita? Quando registri un account, tornerai sempre esattamente dove eri rimasto e potrai scegliere di essere avvisato delle nuove risposte (tramite email o notifica push). Potrai anche salvare segnalibri e votare i post per mostrare il tuo apprezzamento agli altri membri della comunità.

                                          Con il tuo contributo, questo post potrebbe essere ancora migliore 💗

                                          Registrati Accedi
                                          Rispondi
                                          • Risposta alla discussione
                                          Effettua l'accesso per rispondere
                                          • Da Vecchi a Nuovi
                                          • Da Nuovi a Vecchi
                                          • Più Voti


                                          • 1
                                          • 2
                                          • 5
                                          • 6
                                          • 7
                                          • 8
                                          • 9
                                          Feed RSS
                                          AI-assisted moderation in the fediverse is happening. Now what?
                                          @pierobosio@soc.bosio.info
                                          NodeBB Contributors
                                          • Accedi

                                          • Accedi o registrati per effettuare la ricerca.
                                          • Primo post
                                            Ultimo post