IT security student. Programmer of digital duck tape, learning to write digital plumbing. Aspiring cryptographer. Personal research interests include zero-knowledge credentials (to counter online antisocial conduct while protecting privacy) and AV/image non-repudiation (to counter deepfakes). Anti-exceptionalist egalitarian. Empirical pragmatist: evidence over ideology, results over rationalization. Theory is as good as the least of its falsifiability, its internal consistency, and its withstanding of attempted disproof. It is as useful as one is willing to revise or replace it when it fails.
David
@deFractal@infosec.exchange
infosec.exchange
From #GamersNexus: “It’s Time to Poison AI.”
Steve reports on the necessity and means of #dataPoisoning to impose negative value in defence against nonconsensual LLM data scraping.
Thanks, Steve.
https://www.youtube.com/watch?v=0WP6kEO3i8g
#noAI
Dr. Or M. Bialik
@ombialik@mastodon.world
Geology, climate change researcher, and too many other things | Posting about the writing experience, science, SFF and academia. Opinions are my own or of the characters in my head.
mastodon.world
LLMs have been with us for a few years now, and in that time have generated both behavioral changes and a lot of... stuff, including what this study calls "LLM Pollution", which presents a possible issue in trying to understand humans as the data source contains more non-human inputs and people themselves change behaviors.
How to deal with this calls for some new thinking.
Link: https://www.nature.com/articles/s41467-026-74621-9
#DataPoisoning #LLM
𝕂𝚞𝚋𝚒𝚔ℙ𝚒𝚡𝚎𝚕™
@kubikpixel@chaos.social
~/# I'm a ASD coffee loving WebDev and FLOSS geek with some 3D and cryptography experiences. Anarcho with Punk attitude and I had my own radio show. I post here mostly in German and in English and the are deleted after one month. This is a personal account and work in progress… searchable quality propaganda. -- Copyleft 🄯 CC BY-SA 4.0
chaos.social
«KI-Suche erklärt Trump für tot — Warum #DataPoisoning #FakeNews so gefährlich macht:
Künstliche Intelligenz soll das Auffinden von Informationen vereinfachen. Doch was passiert, wenn die Datenbasis gezielt mit absurden Details manipuliert wird, um Verwirrung zu stiften? Ein aktueller Vorfall demonstriert die Schwächen von #Suchmaschine'n eindrucksvoll.»
Fiele User glauben dass die #KI neutral sei, doch so geht #Manipulation und das will kaum jemand wahrhaben wollen.
🤨 https://t3n.de/news/ki-suche-trump-tot-data-poisoning-fake-news-1749884/
You've seen all posts