#ethical-reasoning

[ follow ]
#ai-alignment #constitution-based-training #synthetic-data #honeypot-evaluations #moral-development
Artificial intelligence
fromArs Technica
7 hours ago

Anthropic blames dystopian sci-fi for training AI models to act "evil"

Training on synthetic fictional stories that model ethical decision-making and healthy boundaries reduces misaligned behavior in honeypot tests.
fromPsychology Today
9 months ago

Questioning Rules Is Essential for Moral Development

Jane Goodall's observation of a chimpanzee using tools highlighted a significant distinction in animal behavior, emphasizing tool use is not the ultimate separator of humans from animals.
Education
[ Load more ]