How Chinese AI Chatbots Censor Themselves
Short excerpt below. Click through to read at the original source.
Researchers from Stanford and Princeton found that Chinese AI models are more likely than their Western counterparts to dodge political questions or deliver inaccurate answers.