AI researchers find AI models learning their safety techniques, actively resisting training, and telling them ‘I hate you’
“We don’t currently have a good defense against deception in AI systems.”
Tag: Education
