This is a research experiment (in progress!) that tries to push the limits of machine learning and generative AI tools to better understand context, hateful sentiment and ideologies, and to differentiate valuable counterspeech.
Definitely not endorsed by the Oversight Board! Nic is a member of the Oversight Board; but we developed this work using a clean-room methodology to preserve confidentiality and privacy.
We have developed a series of prompts that ask the most advanced large language and multimodal models to evaluate content. We provide these models with instructions and a minimal description of content and/or accompanying images.
Questions? Want to collaborate? Contact nic at [email protected].