Notes on technical alignment via human-like social drives

Short summary

A lengthy research post explores using human innate social and moral drives as a foundation for technical alignment of brain-like AGI. The author examines specific human instincts, three potential failure modes (wrong moral circle, balance-of-power issues, consequentialist desires overriding virtue ethics), and implementation details for transplanting social drives into AGI source code. The work is exploratory and seeks community feedback for deconfusion.

• •Proposes human social instincts as a starting point for AGI alignment
• •Identifies three failure modes: moral circle errors, power imbalance, and consequentialist override of virtues
• •Discusses person-first vs desire-first pathways for virtue development in AGI

Generated with AI, which can make mistakes.