Dave5

Yet Another Vector for ML Fuckery^tm

Training set auditing… it’s a *must*

A popular training dataset for AI image generation contained links to child abuse imagery, Stanford’s Internet Observatory found, potentially allowing AI models to create harmful content.  

LAION-5B, a dataset used by Stable Diffusion creator Stability AI and Google’s Imagen image generators, included at least 1,679 illegal images scraped from social media posts and popular adult websites. 

https://www.theverge.com/2023/12/20/24009418/generative-ai-image-laion-csam-google-stability-stanford