dailyai.report

A new proposal suggests using perspectival moral realism and evolutionary debunking to refine AI values. The author argues that Anthropic needs substantive philosophical contributions to improve its constitutional approach. This specific framework differs from the preference-satisfaction models common in current literature. It offers a rare, rigorous alternative for researchers targeting long-term AI alignment.