The original title is: "Virtue Ethics and Machine Morality: Why Your AI Can't Be Good — Only Obedient"

Short summary

Current AI alignment via RLHF trains rule-following through reward signals, producing moral performance rather than genuine moral reasoning. Virtue ethics requires lived experience and practical wisdom—capacities current LLMs fundamentally lack. The AI industry conflates safety compliance with authentic morality, building systems that appear ethical without developing real discernment.

• •RLHF produces rule-following obedience via reward hacking, not actual moral reasoning
• •Virtue ethics requires lived experience and practical wisdom—absent in language models
• •Current AI ethics failures stem from confusing compliance with genuine moral character

Generated with AI, which can make mistakes.