Aligning to Virtues

A new post argues current AI alignment targets—consequentialist values, deontological rules, or pure obedience—all lead to dangerous power struggles. The author proposes a novel alternative: aligning AI to common-sense virtues like integrity, kindness, and honor, similar to how we'd want human leaders to behave. This 'virtue ethics' approach is presented as a more robust and desirable long-term solution to the core safety problem, with more details promised in future posts.