- https://the-mad-pirate-prints.teemill.com/
@TheMadPirate I have too many questions but not enough curiosity to go through with this.
@rasterman Rumor has it that Eunomia it is set up to work very similar like the infamous Candid App, scrapping posts from instances all round the Fediverse and using Sentiment Analysis to train a Neural Network on which posts are "problematic" so they can automatically promote or shadowban those posts.
@TheMadPirate That's awful. But can it affect other instances or just the gay ones that use it?
🔻🪂
@rasterman @TheMadPirate the gay ones that use it.
My question would be, how are they thinking on deploying it? Will they just provide the "trained" filter in an open source way? It would be interesting to poke with it. What hardware will be needed to run it in your instance? Would that be expensive for small instances? If they provide it "as a service", free or paid? If free, how many resources are throwing into it? Would there be a way to legally exploit it? 🤔
My question would be, how are they thinking on deploying it? Will they just provide the "trained" filter in an open source way? It would be interesting to poke with it. What hardware will be needed to run it in your instance? Would that be expensive for small instances? If they provide it "as a service", free or paid? If free, how many resources are throwing into it? Would there be a way to legally exploit it? 🤔
Follow
@MechaSilvio @rasterman @TheMadPirate Funniest way to exploit it would be to literally just flip a single > into < and now it will only show a post if its akin to: "NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER NIGGER"
