Underground Resistance Aims To Sabotage AI With Poisoned Data

This Forbes article discusses the Poison Fountain project, an underground resistance movement aiming to sabotage AI systems by contaminating their training data. The group's manifesto and instructions are available on a public-facing website, echoing growing concerns about AI safety, particularly from experts like Geoffrey Hinton. The project's strategy involves embedding links to poisoned data on various websites to trick web crawlers into collecting flawed content. However, the article also points out potential limitations of this approach, such as the difficulty of infiltrating large volumes of poisoned data into AI training pipelines and the ability of AI developers to detect and filter out such data.